Paul Goren - On Voter Competence

On Voter Competence
Series in Political Psychology
Series Editor
John T. Jost
Editorial Board
Mahzarin Banaji, Gian Vittorio Caprara, Christopher Federico, Don Green,
John Hibbing, Jon Krosnick, Arie Kruglanski, Kathleen McGraw, David Sears,
Jim Sidanius, Phil Tetlock, Tom Tyler
Image Bite Politics: News and the Visual Framing of Elections

Maria E. Grabe and Erik P. Bucy
Social and Psychological Bases of Ideology and System Justification

John T. Jost, Aaron C. Kay, and Hulda Thorisdottir
The Political Psychology of Democratic Citizenship

Eugene Borgida, Christopher Federico, and John Sullivan
On Behalf of Others: The Psychology of Care in a Global World

Sarah Scuzzarello, Catarina Kinnvall, and Kristen Renwick Monroe
The Obamas and a (Post) Racial America?

Gregory S. Parks and Matthew W. Hughey
Ideology, Psychology, and Law

Jon Hanson
The Impacts of Lasting Occupation: Lessons from Israeli Society

Daniel Bar-Tal and Izhak Schnell
Representing Red and Blue: How the Culture Wars Change the Way Citizens
Speak and Politicians Listen
David C. Barker and Christopher Jan Carman
On Voter Competence
Paul Goren
On Voter Competence
Paul Goren
1
3
Oxford University Press is a department of the University of Oxford. It furthers the
University’s objective of excellence in research, scholarship, and education by publishing
worldwide.
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trade mark of Oxford University Press in the UK and
certain other countries.
Published in the United States of America by
Oxford University Press
198 Madison Avenue, New York, NY 10016
© Oxford University Press 2013
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, without the prior permission in writing
of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed
with the appropriate reproduction rights organization. Inquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department, Oxford University
Press, at the address above.
You must not circulate this work in any other form
and you must impose this same condition on any acquirer.
Library of Congress Cataloging-in-Publication Data

Goren, Paul. On voter competence / Paul Goren.
p. cm. – (Series in political psychology)
Includes bibliographical references and index.
ISBN 978–0–19–539614–0 (hardback : alk. paper) 1. Voting—United States.
2. Voting—United States—History. I. Title.
JK1967.G58 2012
324.973—dc23
2012011459
9 8 7 6 5 4 3 2 1
Printed in the United States of America on acid-free paper
Contents
Acknowledgments vii
1 The Indictment of the American Voter 1
2 Policy Cleavages in Historical Context 15
3 Policy Attitudes and Political Sophistication 43
4 The Three Faces of Policy Voting 63
5 The Availability of Policy Principles 89
6 The Centrality of Policy Principles 123
7 The Origins of Policy Principles 159
8 The Electoral Consequences of Policy Principles 195
9 The Exoneration of the American Voter? 233
Appendix: Measurement of Key Variables 245
References 249
Index 263
v
This page intentionally left blank
Acknowledgments
On Voter Competence is my first book. I must confess that the experience of

writing it was often miserable. What, exactly, am I trying to say, and why
can’t I say it clearly? Well, this chapter doesn’t fit; I wish I had seen this before.
Ugh . . . my SEM program won’t run on the new OS. Taking the stairs down
to my basement office to write—and rewrite and rewrite and rewrite—left
me feeling as if I was trapped in a course-correcting time loop in which the
proper course was difficult to chart. At other times I felt like a character in a
Beckett novel, compelled to move forward, but not sure how: “I can’t go on,
I’ll go on.” Progress came, of course, in fits and starts, fits mostly, and often
left me feeling irritable and exhausted at day’s end. But there were moments
of joy as well once everything began to come together. The book is done at
last, which brings me to this especially gratifying task. Over the past 4 years
I have leaned heavily on the good graces and patience of innumerable col-
leagues, friends, and, most importantly, my family. I can finally acknowledge
these debts and thank those whose support has meant very much to me.
I begin with the first rate editorial team at Oxford University Press and
the Series in Political Psychology. Lori Handelman contacted me about my
idea long before I had put pen to paper, waited patiently for the prospectus
to arrive, and provided much good cheer along the way. Abby Gross proved
remarkably patient and helpful as I brought the project to a close and worked
tirelessly to ensure a key deadline was met. Aaron van Dorn and Joanna Ng
handled all queries I had quickly and professionally. John Jost provided criti-
cal feedback on the proposal and draft manuscript and, thankfully, saved me
from a critical error late in the game. Finally, comments from the anonymous
reviewers, much of which has been incorporated into these pages, improved
the manuscript immeasurably.
Friends and colleagues at the University of Minnesota and elsewhere delivered
timely and invaluable feedback on various theoretical, substantive, and method-
ological issues. I’m not sure how the book would have turned out without their
vii
viii ACKNOWLEDGMENTS
generous help. I do know it would have been much worse. Howie Lavine
read every chapter, gave excellent commentary across the board, and supplied
plenty of help along the way. Chris Federico read most of the book, made
many astute comments about psychological theory and American politics, and
offered sage statistical advice. I thank him too for scoring tickets for the Boris ~
Kurihara and GBV shows. Chris Chapp somehow managed to read my man-
uscript while completing his own book. His commentary was always expert
and his suggestions spot on (and his spots in the rack greatly appreciated). Ed
Schiappa read the entire manuscript with remarkable dispatch, offered cogent
advice, and pushed me to keep my eye on the big picture, all the while serving
as chair of the Department of Communication Studies at Minnesota. Where he
found the time I will never know. Tim Johnson and David Kimball read the
complete manuscript as well and made a number of very useful suggestions.
They too pushed me to emphasize broad themes and clarify my prose. David
Samuels, Markus Prior, and Aaron Hoffman deserve credit on this score as
well. Joanne Miller offered much needed criticism on early drafts of the con-
ceptual/theoretical chapters, forcing me to simplify matters in my own mind.
Two historians, James Patterson at Brown University and Robert Collins at
the University of Missouri, answered an out-of-the-blue cry for help from
a political scientist desperately trying to condense a great deal of historical
material into a coherent narrative. Both graciously read drafts of Chapter 2
with a careful and critical eye, saved me from several embarrassing factual
errors, and taught me a thing or two about the historian’s craft. Finally, I thank
Paul Sniderman. Though I did not realize it at the time, the idea for the book
originated in a conference paper I presented at the American Political Science
Association’s annual meeting in 2001. After the presentation, Paul introduced
himself, shared some kind words about my efforts, and encouraged me to
think big. Later on in another venue, he furnished critical advice about the
project, all of this despite the fact that some of what I was doing challenged
his work.
Many other people provided helpful comments or assistance along the way,
including Scott Abernathy, Liz Beaumont, John Bullock, Logan Dancey, Mat-
thew DeBell, John Freeman, Kim Fridkin, Sarah Allen Gershon, Jim Gimpel,
Jessica Goren, Rick Herrera, Jon Hurwitz, Bill Jacoby, Stephen Jessee, Andy
Karch, Pat Kenney, Miki Caul Kittilson, Pay Luevano, Makoto Kawabata, Ron
Krebs, Jay McCann, Bill McCready, Takashi Mizutani, Jeff Mondak, Christopher
Muste, Kathryn Pearson, Brian Rathbun, Mark Ramirez, Shalom Schwartz, Phil
Shively, Dara Strolovitch, John Sullivan, Shawn Treier, and Joan Tronto. Finan-
cial support for the three surveys described in Chapter 7 was provided by the
Graduate School and the Center for the Study of Political Psychology at the
Acknowledgments ix
University of Minnesota and the International Studies Association. I thank Bill

Chittick and Jason Reifler for bringing me onboard the ISA project. Despite all
this generous help errors surely remain, for which I accept full responsibility.
My greatest debts are to my family. My mother, Elaine Goren, taught me a
love of books at an early age and my father, Nurullah Goren, instilled in me the
discipline it would take to one day write one. My in-laws, Brian and Rosemary
Dingle, offered much encouragement over the years, as did my sisters Jessica
Goren and Leslie Doyle. I thank them all here. My children, Nurry, Elsie, and
Indigo, though not always sure why I had to work so much, reminded me
daily about what is most important in life. Finally, my wife Lisa has borne
many heavy burdens the past several years. Through all the late nights and
parental shirking and lost weekends, she remained cheerful, supportive, and
understanding. For this and so much more, I will always be grateful. I dedi-
cate this book to her.
On Voter Competence
CHAPTER 1
The Indictment of the American Voter
U.S. presidential campaigns revolve around competing claims about the

good and just society. During election season, presidential hopefuls invoke
the terms “liberal,” “moderate,” and “conservative” to describe these visions.
They do so in their stump speeches and political advertisements, in staged
media events, when they debate one another before tens of millions of tele-
vision viewers, and when responding to questions posed by journalists and
citizens in various forums. Their messages are covered by national and local
newspapers and magazines such as Time and Newsweek. Network and cable
news broadcasts provide daily coverage. Hard news programs such as Meet
the Press and The News Hour assess the ideological proclivities of the candi-
dates. Jon Stewart and Stephen Colbert mock conservatives on The Daily Show
and The Colbert Report, while Rush Limbaugh and Laura Ingraham lambaste
liberals on talk radio. On the World Wide Web political bloggers at Daily Kos
and Redstate dissect the candidates’ proposals in real time. In short, liberal
and conservative rhetoric is everywhere during the campaign season.
Consider some memorable lines delivered at the national party conventions
over the past 24 years. In 1988, Ronald Reagan declared “It’s time to talk issues;
to use the dreaded ‘L’ word; to say the policies of our opposition and the congres-
sional leadership of his party are liberal, liberal, liberal” (cited in von Bothmer,
2010, p. 97). Four years later, Bill Clinton pivoted to the center: “We offer our
people a new choice based on old values. We offer opportunity. We demand
responsibility. We will build an American community again. The choice we offer
is not conservative or liberal . . . . It is different. It is new. And it will work.” At the
2000 GOP convention, George W. Bush posed as the center right candidate: “Big
government is not the answer. But the alternative to bureaucracy is not indif-
ference. It is to put conservative values and conservative ideas into the thick of
the fight for justice and opportunity. This is what I mean by compassionate con-
servatism. And on this ground we will govern our nation.” Finally, recall Barack
Obama’s postpartisan yearnings at the 2004 Democratic convention: “there are
1
2 ON VOTER COMPETENCE
those who are preparing to divide us, the spin masters and the negative ad ped-
dlers who embrace the politics of anything goes. Well, I say to them tonight,
there’s not a liberal America and a conservative America; there’s the United
States of America” (Washington Post, 2004).1
These examples illustrate how political leaders invoke liberal and conser-
vative labels to communicate with voters. They do so because these labels
summarize information about candidate and party positions on the proper
economic, social, and political order; the role of government; and positions
on diverse issues. Those comfortable with ideological frames of reference can
follow the dialogue with ease. When a politician is described as conservative,
exceptionally sophisticated citizens—those who are deeply informed about
government and politics—can infer the candidate favors lower taxes, smaller
government, the right to life, higher military spending, and so on. And when
hearing the opponent is liberal, they recognize this candidate holds oppos-
ing positions. Knowing where the candidates fall on the liberal–conservative
continuum, highly sophisticated voters utilize a simple and effective decision
rule: choose the candidate who lies closer to them on the liberal–conservative
spectrum. Policy voting is thereby assured.
But what of those who quickly scan the front page of the local paper before
turning to the sports pages? What of those who pause briefly, if at all, on the
evening news before switching to reruns of their favorite sitcom? How about
people who surf the web for everything but political news? What about those
who have never listened to Keith Olbermann or Bill O’Reilly, who have never
read Paul Krugman or David Brooks, who know more about the celebrity
scandal du jour than the federal budget? Do these citizens know what liber-
alism and conservatism mean? Can they use ideological labels to summarize
positions on dozens of issues? Will they ground their votes in abstract liberal
or conservative principles?
Scholars’ answers to each of these questions is “no.” There is no doubt
that locations along the liberal–conservative continuum represent the ultimate
political shorthand, the political heuristic par excellence, in American political
discourse (Downs, 1957). But since most members of the public do not under-
stand what liberalism and conservatism mean, widespread ideological voting
does not occur. Citizens perform no better when it comes to concrete issues.
Notwithstanding the occasional controversy that captures the public’s atten-
tion, large swatches of the electorate lack meaningful attitudes on the issues of
the day. Hence, the prospects for pervasive issue voting appear equally dim.
1
Unless otherwise noted, all convention and state of the union quotes are from
Woolley and Peters (2011).
The Indictment of the American Voter 3
This pessimistic view has dominated the study of voting behavior for over
half a century. In 1960, Campbell, Converse, Miller, and Stokes concluded their
landmark work The American Voter on this dour note.
Some individuals . . . know what they want their government to do and

they use their vote in a very purposive manner to achieve within their
power the policy alternatives that they prefer. Such people do not make
up a very large proportion of the electorate. (p. 542)
Nearly half a century later, Kinder (2006) concurred: “when it comes to poli-
tics, most citizens are ideologically innocent: indifferent to standard ideolog-
ical concepts, lacking a consistent outlook on public policy, in possession of
genuine opinions on only a few issues, and knowing damn little” (p. 199).
The normative corollaries attending this empirical portrait are bleak. In
some recent reflections, Converse (2006) opined:
I cannot say that voters have to be users of the ideology heuristic to vote
‘sensibly.’ On the other hand, since it offers unusual powers of economy in
both understanding arriving flows of political information and in retain-
ing it, and since the world of politics is objectively complex and the voter
is a cognitive miser, its use is to be recommended, and the fact that it is not
widely understood is, to me, a central indicator of the problem surround-
ing voter competence. (p. 310)
On Voter Competence takes a far more optimistic view of the American voter.
In this book, I argue everyday citizens choose based on core policy principles,
but that professional students of electoral behavior have missed this because
they have searched for evidence of policy voting in the wrong places. Once we
turn away from liberal–conservative predispositions and preferences on dis-
crete issues, we discover that nearly everyone in the mass public holds policy
principles and uses these to guide candidate selection in U.S. presidential elec-
tions. Contrary to the indictment leveled by much of the scholarly commu-
nity, citizens who are not deeply informed about public affairs prove as adept
as their more sophisticated counterparts at grounding presidential votes in
abstract views about public policy.
Three principles corresponding directly to the major policy cleavages that
have divided the Democratic and Republican parties for the past several
decades are paramount: limited government, traditional morality, and military
strength. My most important claims are that attitudes toward these principles
are (1) available in the minds of nearly all citizens, (2) function as central heu-
ristics in the belief systems of the politically sophisticated and unsophisticated
alike, (3) are rooted deeply in basic human values, and (4) guide presidential
choice to a comparable degree for voters across the sophistication spectrum.
Put simply, the critical point to take away from my book is that most citizens
have genuine policy principles and rely heavily on these when casting pres-
idential ballots. Insofar as we can equate the development and use of policy
principles with citizen competence, the American voter performs far better
than has been recognized.
THE IMPORTANCE OF POLICY PRINCIPLES

How can those too unsophisticated to ground their votes in ideological or
issue preferences come to use policy principles? This is the key puzzle ani-
mating my book. This is an intriguing question because most citizens seem-
ingly know too little about government and politics to policy vote. To take
some particularly egregious examples of public ignorance: a 1964 survey
found that only 38% of the public realized the Soviet Union was not in the
anti-Soviet defense alliance NATO (Page & Shapiro, 1992, p. 9); in another
survey approximately 40% of the American public believed Israel was an
Arab nation (Sniderman, Brody, & Tetlock, 1991, p. 15); a 1992 study found
that 86% of likely voters knew the name of President Bush’s dog but only 15%
were aware that Bush and Clinton favored the death penalty (Delli Carpini
& Keeter, 1996, p. 63); a July 2010 Pew Research Center poll found that 28%
of voting age adults knew John Roberts was the Chief Justice of the Supreme
Court, whereas 85% could identify Twitter (Kohut & Keeter, 2010); and a
November 2010 Pew survey taken a week after a historical midterm election
in which the GOP picked up 63 seats to reclaim the House of Representatives
found that only 46% of the public knew Republicans would be in the majority
(Kohut, Doherty, Dimock, & Keeter, 2010). Given that so many know so little
about public affairs, our inclination is to scoff at the notion of widespread
policy voting.
To address the puzzle of how unsophisticated citizens policy vote, I begin
by defining policy principles and explaining how they differ from liberal–
conservative attitudes and issue preferences, the mainstays of research on
voting behavior. I sketch the conceptual distinction here before elaborating it in
Chapter 3. Policy principles represent bottom line judgments about the proper
course of action to take in the key issue areas that comprise American politics.
They are crowning postures that guide political judgment and shape prefer-
ence formation in a specific policy domain. Because of this, they are sometimes
called domain-specific principles (Peffley & Hurwitz, 1985). Throughout the
book, I use “policy principles,” “domain-specific principles,” and the related
phrase “core principles” interchangeably. Limited government, traditional

morality, and military strength represent the focal principles in this study. As
I detail in Chapter 2, all three warrant attention because they represent highly
visible cleavages over which the presidential candidates and national parties
have battled for a very long time.
What do these terms mean? Attitudes toward limited government reflect
judgments about the role the federal government should play in underwriting
the economic security of the populace. On the one hand, those who believe the
government in Washington has a responsibility to help individuals cope with
the hazards of the market economy support activist government. On the other
hand, those who believe individuals are entirely responsible for their own eco-
nomic well-being favor limited government (Markus, 2001). Next, traditional
morality centers on the degree to which conservative or orthodox moral stan-
dards should guide the public and private life of the nation. Traditionalists
view long-established standards of moral rectitude as absolute. They hold that
conventional views of good and bad and right and wrong are sacrosanct and
inviolable. As such, society should defend and promote traditional values.
Against the claim that orthodox standards apply universally, progressives
take a more tolerant and pluralistic stance, and thus, resist collective efforts to
promote and enforce traditional morality (Hunter, 1992). By military strength,
I mean the extent to which U.S. foreign policy and national security objectives
are best served by emphasizing American power. Hawks favor the develop-
ment and maintenance of a strong military and believe it should be used when
necessary. Doves hold that U.S. interests are better served through the applica-
tion of softer forms of power, such as diplomacy, negotiation, and compromise
(Mueller, 1973).
These principles are distinguishable from other types of preferences linked
to policy voting, namely liberal and conservative attitudes and issue attitudes.
To begin with the former, it may strike some that the principles articulated
above are simply manifestations of ideological predispositions rather than
independent psychological entities. If so, the conceptual distinction between
liberal–conservative attitudes and domain-specific principles vanishes. Now,
if would be electors consistently took left- or right-wing positions across all
three dimensions, we could conclude that locations on the liberal–conserva-
tive continuum summarize these positions in the manner described by Downs
(1957). To illustrate, someone deeply committed to weak government, tra-
ditional morals, and military power could be placed near the conservative
end of the ideological continuum. Conversely, someone enthusiastic about
reformist government, moral progressivism, and diplomacy would reside at
the opposite end of the spectrum. For such individuals, liberal–conservative
placements would capture general political principles as well as preferences

on many separate issues.
Such reasoning sounds nice in theory, but as my discussion in Chapter 3
will show most citizens do not think about politics like this. Instead, liberal–
conservative orientations are better viewed in terms of evaluations of various
social groups and political symbols associated with these labels (Conover &
Feldman, 1981; Levitin & Miller, 1979; Sears, 2001). For some, the labels repre-
sent “feminists” or “gays” or “big business” or “Bible thumpers.” For others,
the terms connote images such as “taxes,” “permissive,” “religious,” and so
on. Thus, liberal–conservative attitudes are symbolic predispositions largely
devoid of policy content. In contrast, policy principles, as defined here, are
all about policy, about what should be done in a given issue area. Next, issue
preferences can be defined as evaluations of proposals about how to solve
some specific problem or address a particular controversy. Examples include
abortion, health care reform, tax proposals, Social Security privatization, fight-
ing terrorism, and innumerable other concerns. Given that an issue references
a single policy whereas core principles transcend all such policies in a given
domain, the conceptual distinction between them should be clear.
To reiterate, policy principles, liberal–conservative orientations, and issue
preferences embody distinct types of political attitudes. Policy principles are
abstract ideas about the proper course of action to pursue in a single policy
domain. Liberal–conservative attitudes are evaluations of a few social groups
or political symbols tied to these labels. Issue attitudes reflect judgments about
particular policy proposals on offer in a given campaign. Because the manifest
content of each policy entity differs, the attitudes can be distinguished from
one another.2
PREVIEW OF THE THEORETICAL ARGUMENT

To date, the study of policy voting has been dominated by liberal and conser-
vative orientations and issue preferences. Multiple studies demonstrate that
knowledgeable citizens alone base their votes on liberal–conservative atti-
tudes (Goren, 1997; Knight, 1985; Lau & Redlawsk, 2006; Neuman, 1986). For
2
Policy principles also stand apart from basic human or personal values, a con-
cept with which they are sometimes confused. Policy principles center on what
people think should be done in a given issue area in the political sphere, whereas
personal values revolve around the importance individuals attach to abstract
goals that transcend politics (Rokeach, 1973; Schwartz, 1992). I expand on these
differences in Chapter 7.
most voters these attitudes are imbued with little policy content, revolving
instead around gut-level feelings about disparate social groups and symbols.
Even among the most sophisticated third or so of the public, ideological vot-
ing is probably best viewed as an expression of symbolic affinity for a given
candidate rather than an attempt to send policy signals about what the can-
didate should do if elected. In a similar vein, accumulated research finds that
those who are knowledgeable about politics in general or some particular con-
troversy issue vote more than those lacking knowledge (Anand & Krosnick,
2003; Delli Carpini & Keeter, 1996). Other research suggests a salient issue may
weigh heavily on voter choice in a given election, but fail to matter in subse-
quent contests (Abramowitz, 1995).
On Voter Competence accepts that most Americans are innocent of ideology
and fuzzy on most issues, but it rejects the corollary that their presidential
votes are untouched by policy considerations. In the pages to follow, I argue a
set of bedrock principles shapes voter choice to a comparable degree for politi-
cally aware and unaware citizens. It matters little whether people know a
great deal about government and politics, whether they prefer The News Hour
or Smackdown, whether they read Thomas Sowell and Frank Rich or Stephen
King and Nora Roberts. Nearly everyone acquires and subsequently uses pol-
icy principles to guide candidate choice.
At this point, it should prove helpful to preview the conceptual and theoret-
ical arguments I make throughout the book. First, I draw upon the Eagly and
Chaiken definition of the attitude construct to argue that liberal–conservative
orientations, issue preferences, and policy principles are best conceptualized
as different types of policy attitudes. Eagly and Chaiken (1993) define an atti-
tude as “a psychological tendency that is expressed by evaluating a particular
entity with some degree of favor or disfavor” (p. 1). This inclusive definition
runs the gamut from dispositional psychological states that endure over time
to temporary evaluations constructed on the spot as demanded by the situa-
tion. It applies readily to the aforementioned policy attitudes.
Next, I argue that most individuals develop and maintain genuine atti-
tudes toward limited government, traditional morality, and military strength.
By develop and maintain, I mean that domain-specific principles are available
and accessible in mass belief systems. To understand how ordinary men and
women acquire these, it must be recognized that the idea embodied by each
principle is sufficiently clear that it can be evaluated without difficulty (cf.
Carmines & Stimson, 1980). In contrast to ideological labels, whose meaning
eludes the uninitiated, there is no need to learn what “government help for
average people” or “traditional morals” or “force versus diplomacy” signify.
Large stores of political knowledge are not required for deciding how we feel
about these ideas. Because the meaning of each is self-evident, evaluation fol-
lows automatically upon exposure.
The question remains as to whether citizens receive sufficient exposure to
these ideas to actually evaluate them. There are compelling grounds for believ-
ing that they do. Political debate in this country has revolved around the role
the federal government should play in the economic welfare domain since
Franklin Roosevelt established the New Deal in the 1930s, around what moral
vision should prevail in American society since the 1960s, and about the proper
use of military force since the onset of the Cold War in the late 1940s. Because
these divisions have been so prominent for so long members of the voting
public need not monitor political debate in real time to learn what the major
policy dimensions are. Political and campaign discourse are rich in overlap-
ping cues that tell the same story about political conflict election after election
and year after year (Feldman, 1988; Rahn, Aldrich, Borgida, & Sullivan, 1990).
Although most people do not know what liberalism and conservatism mean
and remain unclear about, if not oblivious to, the latest controversy preoccu-
pying the chattering classes in Washington, DC and the blogsphere, they see
where the major fault lines lie in the political system. As such, they can form
durable impressions about the core conflicts that drive American politics. In
this manner, attitudes toward limited government, moral traditionalism, and
American military power become lodged in mass belief systems.
I then integrate research on political psychology and public opinion to
argue these orientations operate as central political heuristics in the minds of
most citizens. Political attitudes do not exist in isolation, but rather are linked
together in broader associative networks (Hurwitz & Peffley, 1987). The key
distinction in these attitude structures is between their central and periph-
eral elements. Central attitudes systematically affect peripheral evaluations
without, in turn, being shaped by them. Policy principles, I argue, influence
short-term political evaluations along these lines. Because almost all voters
have some experience using principles to guide the construction of issue pref-
erences and other transitory evaluations, they can be applied heuristically to
the problem of candidate choice. Doing so allows voters to decide quickly
and effectively without taxing their finite cognitive resources. The question
of origins is then taken up. I posit that basic human values and party identi-
fication are candidates to shape policy principles. Moreover, personal values
should guide the positions people take on policy principles both directly and
indirectly via their influence over partisanship.
Lastly, I posit that policy principles drive the presidential vote to an anal-
ogous degree for sophisticated and unsophisticated voters. This should hold
when factors such as partisanship, liberal and conservative orientations,
retrospective judgments, and particular issues are taken into account. Because
views on government power, moral conventions, and military might are
widely held and readily invoked to inform short-term preferences, voters
should encounter little difficulty linking principles to candidate evaluations as
long as they recognize where the candidates stand on each dimension. Because
presidential contenders and the national parties have a long history of taking
clear and contrasting stands on these principles, reminders of which appear
regularly in ongoing political debate and over the course of every presidential
campaign, most voters develop a sense that Democrats stand to the left of the
GOP on each cleavage. In this way, cue consistency and redundancy facili-
tate principle-based choice. In contrast to ideological and issue voting, both of
which are conditional on prior knowledge and interest, principle-based vot-
ing is not limited to a thin slice of the electorate. Instead, policy principles
shape candidate choice for nearly everyone.
To sum up, we must broaden our theoretical understanding of policy vot-
ing beyond liberal–conservative orientations and issue preferences. Once we
do we find that regardless of how much or little Americans know about poli-
tics, abstract policy principles are available in their political psyches, serve as
central heuristics in their belief systems, and guide their presidential votes.
The combination of political ignorance and ideological naiveté bemoaned so
often and so vociferously by so many does not preclude widespread policy
voting in the American electorate.
CONTRIBUTIONS
Numerous variants of democratic theory hold that if the demos are to rule in any
meaningful sense of the term, they must play a leading role in the determina-
tion of public policy (Pennock, 1979; Thompson, 1970). For this to happen, indi-
viduals must hold policy attitudes and use these to guide their choices come
Election Day. The question of whether citizens meet these conditions has gen-
erated a tremendous amount of empirical research over the years. Despite the
occasional dissent (Aldrich, Sullivan, & Borgida, 1989; Ansolabehere, Rodden,
& Snyder, 2008), the dominant view holds that many if not most voting age
adults lack genuine ideological and issue attitudes, implying that policy vot-
ing lies beyond the reach of the typical American voter (Berelson, Lazarsfeld,
& McPhee, 1954; Campbell et al., 1960; Converse, 1964; Delli Carpini & Keeter,
1996; Lau & Redlawsk, 2006; Lewis-Beck, Jacoby, Norpoth, & Weisberg, 2008;
Luskin, 2002; Neuman, 1986). The normative corollary is clear: the demos fail
to meet a key criterion of political competence. This is perhaps the most dam-
aging charge in the indictment against the American voter.
By refuting this perspective, On Voter Competence contributes to the study

of electoral behavior, political psychology, and citizen competence on several
fronts. First, scores of studies have raised serious questions about the quality
of policy attitudes in the mass public. The most pessimistic accounts main-
tain that everyday people lack the motivation and ability to develop, hold,
and draw on genuine policy attitudes. My claim is that although this gen-
erally holds true with respect to evaluations of liberalism and conservatism
and numerous issue controversies, it does not apply to all classes of policy
attitudes. By demonstrating that limited government, traditional morality,
and military strength are available, durable, grounded in personal values, and
consequential, my study shows that abstract policy orientations occupy cen-
tral positions in citizens’ minds. In essence, policy principles lie in the sweet
spot of mass belief systems, neither too abstract such as liberal-conservative
worldviews nor too concrete such as particular issues to elude all but the most
diligent citizens.
Second, my work speaks directly to how much policy voting occurs in U.S.
presidential elections. As noted, many studies find ideological voting and
issue voting are concentrated among highly engaged segments of the public.
Other work establishes that some issues matter in a given election but fall off
the agenda thereafter. Although agreeing with these points, I reject the impli-
cation that policy voting necessarily lies beyond the pale for the less informed.
Instead, attitudes toward limited government, enduring moral standards, and
military power drive candidate choice to a comparable degree for all voters.
Policy voting is broad and deep in the U.S. population.
Third, my work challenges some key tenets of the sophistication inter-
action model that dominates the study of electoral behavior, public opin-
ion, and political psychology. It has long been clear that political awareness
is a prerequisite for ideological reasoning and judgment (Converse, 1964;
Sniderman et al., 1991). I have no quarrel here. However, ardent defenders
of the model go beyond this by claiming the ability to use abstract predis-
positions is always conditional on sophistication. Delli Carpini and Keeter
(1996) aver that “the greater the store of information, the more often citizens
will be able to connect their values with concrete matters of politics” (p. 229).
In a similar vein, Luskin (2002) holds that “There are many reasons to think
sophistication important, but perhaps its greatest importance lies in its con-
ditioning of the relationship between values and policy and candidate pref-
erences, which can be expected to be tighter among the more sophisticated”
(p. 220). By demonstrating that individuals across the sophistication con-
tinuum hold policy principles and apply them to their electoral choices, my
findings suggest the conventional wisdom is too facile. Sophistication may
indeed facilitate some desirable political behaviors, but its absence need not
preclude them.
This is not to say that sophistication is irrelevant. As I show in Chapter 5,
sophistication promotes attitude stability for both policy principles and party
identification. Similarly, evidence in Chapter 6 suggests the sophisticated rely
a bit more heavily on general principles than the unsophisticated when fig-
uring out issue positions. Findings presented in Chapter 8 reveal that sophis-
tication enhances ideological voting. The sophistication interaction model
possesses some validity. But, as I will show, these are marginal rather than
fundamental differences. Said otherwise, the sophistication gradient is mild
rather than steep. My book affirms that sophistication matters in some ways,
but less so than typically imagined.
Fourth, On Voter Competence has positive implications regarding the polit-
ical intelligence of the American voter. Normative and empirical theories of
democracy insist that policy views must impact voter choice (Fiorina, 1981;
Key, 1966; Pennock, 1979). Policy views signal what people want the national
government to do and allow them to hold elected officials accountable for
their actions in office. Indeed, these signals provide a means through which
nonelites exercise some measure of popular control over government. Because
my work shows that most citizens possess genuine attitudes about the funda-
mental divisions that define policy space in the United States and call upon
these to reward or punish the candidates and parties, as well as signal the
direction government should take in a given policy area, fears about voter
incompetence are overblown. Put simply, the typical American voter performs
reasonably well as judged by these tenets of democratic theory.
PLAN OF THE BOOK

My book proceeds along the following lines. The next three chapters set the
stage for the empirical analyses to come later on. Chapter 2 presents a his-
torical overview of the major developments, as I see them, in the economic
welfare, cultural issues, and foreign policy domains. My review suggests
that public discourse in the respective issue areas can be encapsulated by the
broader dimensions of limited government, moral traditionalism, and mili-
tarism, and that presidential incumbents, presidential candidates, and their
parties have taken distinctive positions on each over the past several decades.
The clarity of these cues has implications for the development of policy prin-
ciples. In Chapter 3, I apply the Eagly and Chaiken definition of the attitude
construct to liberal and conservative attitudes, issue preferences, and policy
principles. The sophistication construct is also defined. Note that the terms
“political sophistication,” “political knowledge,” and “political awareness”

are treated as synonyms throughout the book.
Chapter 4 then integrates the historical and conceptual analyses into a the-
ory of voter choice. I adapt the requirements of issue voting laid out by the
authors of The American Voter to argue that citizens must satisfy three condi-
tions to policy vote. These are availability, centrality, and position matching.
When policy attitudes reside in long-term memory, perform as central heuris-
tics, and are matched to the positions held by the respective candidates, policy
voting should occur. I also delineate the role political sophistication plays in
moderating reliance on each class of attitudes. Theoretical reasoning and prior
empirical work lead me to propose that ideological voting and issue voting
bypass uninformed voters. In contrast, the politically sophisticated and unso-
phisticated should root their votes in policy principles.
My theoretical framework holds that limited government, traditional
morality, and military strength are widely available in political belief systems.
Empirically, this implies that responses to opinion items designed to measure
latent principles will be structured coherently and equivalently for individu-
als at different levels of knowledge. Moreover, these principles should prove
reasonably stable over time. Chapter 5 tests these predictions. Exploiting data
from multiple National Election Study (NES) cross-sectional and panel sur-
veys, this is precisely what I find. Chapter 6 takes up the centrality condi-
tion. According to my theory, policy principles are highly diagnostic and thus
should weigh heavily on short-term issue preferences. To test this proposition, I
use NES panel data to estimate a series of dynamic constraint models whereby
current positions on general principles and issue preferences are modeled as a
function of one another and lagged positions on both variables. My estimates
demonstrate that limited government and moral standards shape health care
and abortion preferences, respectively. Comparable data to test for the central-
ity of military strength are unavailable, but Peffley and Hurwitz (1993) have
reported results similar to what I uncover here. I then turn to cross-sectional
data to estimate the effects principles have on issue preferences for citizens
across the sophistication continuum. I find that the less sophisticated almost
always use principles to construct issue preferences. Moreover, some evidence
suggests awareness strengthens these connections. But the bottom line is that
although sophistication may enhance reliance on these heuristics, its absence
does not foreclose such usage. Finally, my results show that policy principles
often outperform other heuristics––including party identification––as predic-
tors of issue preferences.
Having examined the properties and some of the effects of policy principles,
I turn to their building blocks in Chapter 7. I argue that party identification
and basic human values shape evaluations of domain-specific principles. I test

these expectations using data from three new national surveys.3 My analysis
shows that although partisanship and other factors shape policy principles,
the impact of basic human values is larger––perhaps much larger. To put it
another way, personal beliefs about what is important in life constrain atti-
tudes toward limited government, traditional morality, and military power
more so than partisan affinities do. This holds for both sophisticated and
unsophisticated citizens. These findings speak directly to the question of citi-
zen competence. So far as citizens are expected to evaluate public policy using
abstract normative criteria over simplistic group cues, my results suggest citi-
zens perform reasonably well by this standard.
Chapter 8 takes up the business of the policy principle–vote choice rela-
tionship. I begin by summarizing leading theories of electoral behavior: the
partisan voter model (Campbell et al., 1960), the retrospective voter model
(Fiorina, 1981), and the ideological voter model (Sniderman et al., 1991). Next,
I use data from the 1988–2008 NES surveys to model the presidential vote
as a function of predictors drawn from each perspective alongside the lim-
ited government, moral tradition, and military strength variables. My analysis
demonstrates that policy principles systematically affect candidate choice in
every election and, most importantly, that the unsophisticated rely as much
as the sophisticated on them. These results suggest that when it comes to the
presidential vote core principles behave like long-term factors in the minds of
voters. This does not mean sophistication plays no role in the electoral calcu-
lus. Consistent with prior work, I find that sophistication promotes liberal and
conservative voting.
Chapter 9 brings the book to a close. It begins with a restatement of the
theoretical argument, reviews the lessons learned, and highlights the limita-
tions of what I have done and can claim. The bulk of the chapter considers
the broader theoretical, practical, and normative ramifications my findings
have for evaluating voter decision making, citizen competence, and American
politics.
3
I designed two of these surveys and commissioned Knowledge Networks to field
them in 2007 and 2008. The third survey was designed by a team of researchers
led by William Chittick and Jason Reifler and administered in 2011 by YouGov/
Polimetrix.
CHAPTER 2
Policy Cleavages in Historical Context
My examination of the American voter focuses on three policy principles. First,

the limited government dimension arrays those who favor federal efforts to
underwrite economic security against those who believe individuals must
take on that responsibility entirely by themselves. Second, traditional morality
reflects a collective commitment to the preservation of orthodox conceptions
of right and wrong. Traditionalists endorse this goal whereas progressives,
tolerant of alternative values and lifestyles, do not. Third, military strength
denotes the extent to which military power is preferable to softer forms of
power as a means of advancing U.S. foreign policy and national security
objectives. Hawks favor military muscle whereas doves are more comfortable
with the tools of statecraft.
Attitudes toward these principles do not arise spontaneously in the minds
of citizens. Instead, they develop when individuals react to the fundamental
policy cleavages that dominate public affairs and electoral politics. This pro-
cess begins when some economic, social, or political shock—or set of shocks—
bursts onto the national agenda. Strategic politicians in search of votes attempt
to mobilize those harmed by the crisis. Successful mobilization typically gen-
erates backlash from those with contrasting views, who can then be mobilized
by other politicians. Some disputes generate political friction for relatively
brief periods before vanishing. Others persist long enough to become deeply
enmeshed within the party system and intertwined with presidential politics.
As controversies morph from short-term disagreements into long-term cleav-
ages, citizens can develop attitudes toward them, assimilate these into their
belief systems, and call upon them to guide political judgment (Carmines &
Stimson, 1989; Claggett & Shafer, 2010; Layman, 2001).
This chapter reviews twentieth century American history to chart the
development of the elemental policy cleavages that have structured national
politics for the past several decades in the economic welfare, cultural issues,
and national security domains. In the pages to come, I describe the emergence
15
of each cleavage, trace its evolution over the long haul, and examine its con-
nections to presidential politics and the national parties. Doing so helps illu-
minate the psychological processes I elaborate in Chapters 3 and 4 by which
ordinary citizens acquire and subsequently apply policy principles to their
electoral choices. Readers should note that I delimit my historical analysis in
three ways. First, discussion centers on what I see as the major developments
within the aforementioned policy domains. A detailed account of every issue
controversy lies beyond the scope of my inquiry. Second, given my interest in
policy voting in U.S. presidential elections, I highlight the politics of presiden-
tial campaigning and governance. Far less attention is devoted to the congres-
sional side of the ledger. Third, issues in other domains that have experienced
short lifespans command little attention in what follows.
THE LIMITED GOVERNMENT CLEAVAGE

I begin with the politics of economic welfare.1 The question of whether the
federal government should provide some measure of material security for
average Americans moved onto the political agenda in earnest during the
Great Depression. Between 1929 and 1932, gross national product plum-
meted 50%, while the unemployment rate exploded from 3% in 1929 to 25%
in 1932. Wages stagnated or declined for many of those who held onto their
jobs and the specter of poverty haunted the middle class as never before.
President Herbert Hoover did not stand by idly and simply hope for recovery.
Instead, he adopted several novel policies to arrest the decline. For example,
the president created the innovative Reconstruction Finance Corporation in
1932 to provide emergency loans to banks and other sectors of the econ-
omy. Through bank relief Hoover hoped to stabilize the banking system and
stave off economic collapse. Yet when it came to providing unemployment
assistance and relief for working class and poor citizens, Hoover balked. In
earlier downturns state and local governments, along with private charities,
undertook relief efforts, and the president saw virtue in continuing along
these lines. Unfortunately, the crisis overwhelmed these institutions. When
Hoover sought reelection, the Great Depression was a millstone around his
neck. The image of the Republican Party was badly damaged as well.
1
My discussion of each issue area has been informed by many outstanding works
of history and political science. For the economic welfare domain, I draw on Blum
(1991), Boyer (2001), Brewer and Stonecash (2007, 2009), Collins (2000, 2007),
Dallek (2003), Edsall and Edsall (1991), Hamby (1992), Kennedy (1999), Page
(1978), Patterson (1994, 1996, 2005), Sundquist (1968, 1983), and Wilentz (2008).
Policy Cleavages in Historical Context 17
Into the breach stepped Franklin Delano Roosevelt, a skilled politician and
inveterate reformer who believed the national government was obligated to
use its power to restrain private interests on behalf of the public good. He
planned to use the levers of federal power to alleviate economic suffering and
stabilize the devastated economy. At the Democratic Party convention that
July, Roosevelt previewed his New Deal philosophy.
[T]he federal government will assume bold leadership in distress relief.

For years Washington has alternated between putting its head in the sand
and saying there is no large number of destitute people in our midst who
need food and clothing, and then saying the States should take care of
them, if there are. . . . I say that while primary responsibility for relief rests
with localities now, as ever, yet the federal government has always had
and still has a continuing responsibility for the broader public welfare. It
will soon fulfill that responsibility.2
That message resonated with the electorate as Roosevelt trounced Hoover

by a 57–40% margin on Election Day. During the first 100 days in office, his
administration secured passage of measures designed to prevent the depres-
sion from worsening. These included the Emergency Banking Act and a
4-day banking “holiday,” which stemmed the banking panic; the Civilian
Conservation Corps (CCC), which enabled the federal government to put
hundreds of thousands of young men to work on conservation projects; and
the Federal Emergency Relief Administration, which provided direct assis-
tance to the unemployed and the poor. This frenetic pace was unsustainable
over the long haul, but Roosevelt never wavered in his commitment to using
the machinery of government to help ordinary people. Between 1933 and 1938
Roosevelt and his Democratic allies in Congress built the foundation of the
American welfare state. Landmark New Deal programs included the Works
Project Administration (WPA) program (1935), which placed roughly one-
third of the unemployed on the federal payroll at some point, and the Social
Security Act (1935), which established the federal pension program for retir-
ees, an unemployment compensation system, disability insurance, and aid to
single mothers with dependent children.
Full recovery remained elusive until the start of World War II, in part because
of problems with the global economy and administration missteps such as the
premature bid to restore fiscal austerity in 1937. Despite this, Roosevelt’s read-
iness to marshal governmental resources on behalf of the dispossessed, his
2
Recall that all convention quotes are from Woolley and Peters (2011).
comforting fireside chats, and the distribution of tangible benefits, whether

in the form of mortgage relief, a CCC or WPA job, unemployment insurance,
direct cash payments, or the promise of old age pensions, left no doubt where
he and his party stood on the role of government. The contrast with Hoover
and the GOP could not have been any starker.
Since the Roosevelt era, Democratic presidents have usually sought to
enlarge the capacities of government to insulate Americans from market
risk. To begin with Harry Truman, his top domestic priorities were maintain-
ing prosperity and institutionalizing the New Deal. The latter goal featured
heavily in his 1948 “Give ‘em Hell” campaign. Echoing Roosevelt’s attacks
on “economic royalists,” Truman sought to mobilize working class and lower
status voters by savaging “Wall Street reactionaries” and “bloodsuckers” and
portraying Republicans as stalwart defenders of a discredited governing phi-
losophy. Additionally, he proposed wider Social Security coverage and higher
benefits, national health insurance, and a rise in unemployment compensa-
tion, all the while heaping scorn on the “do nothing” Republican Congress for
resisting such proposals.
In office, Truman invoked the memory of Roosevelt by labeling his reform
package the Fair Deal. His pursuit of reform legislation was uneven, in part
because foreign policy crises occupied much of his time. Nevertheless, there
were a few noteworthy accomplishments, including the Social Security Act
of 1950 that expanded coverage and benefits, along with a housing program
that helped some middle and working class families find affordable housing.
There were failures as well. Foremost among these, his goal of establishing a
national health insurance program was stymied. Likewise, his call for a full
employment bill mandating public expenditures to fight joblessness failed
to become law. But even in defeat Truman reaffirmed the Democratic Party’s
commitment to the “little guy.”
Next, as the Eisenhower presidency drew to a close in 1960, Democratic
candidate John F. Kennedy positioned himself in the ideological center of
American politics. This did not reflect centrism per se as a governing philos-
ophy, but rather, a mix of Cold War toughness in foreign policy and liberal
inclinations on the domestic front. On social welfare, Kennedy sounded like
a New Deal Democrat minus the biting attacks on “economic royalists” and
Wall Street “bloodsuckers.” From the Democratic national convention to the
fall campaign, Kennedy chastised Republicans for ignoring the unemployed,
senior citizens, working families, and the poor. During his presidency, Kennedy
focused heavily on restoring and maintaining economic growth at home; deal-
ing with overseas crises in Laos, Cuba, Berlin, and Vietnam; and managing
U.S.–Soviet tensions. Nevertheless, economic welfare issues received attention
as well. Congress backed the administration’s plan to ease chronic unemploy-

ment in West Virginia and similarly afflicted states and heeded its calls for
expanded Social Security and unemployment benefits. Kennedy also signed
legislation expanding Aid to Families with Dependent Children. Finally, the
president set to work on an antipoverty program he wanted to push in 1964,
but these plans fell into abeyance when an assassin’s bullet struck him down
on November 22, 1963.
After succeeding the slain president, Lyndon Johnson seized the reform
mantle by promising to build a “Great Society.” Johnson believed ever rising
economic growth would provide the revenue needed to fund an increasingly
generous, opportunity-enhancing welfare state. During his 1964 state of the
union address, Johnson declared “unconditional war on poverty” and assured
the public that “the richest nation on earth can afford to win it.” Over the
next couple of years Johnson worked tirelessly on behalf of his ambitious plat-
form. Highlights included the Civil Rights Act (1964) and the Voting Rights
Act (1965) that guaranteed full legal equality and the franchise to African-
Americans; the establishment of Medicare and Medicaid as part of the Social
Security Act of 1965; establishment of the Office of Economic Opportunity
that administered remedial education and job training programs; the expan-
sion of existing antipoverty programs such as food stamps; the develop-
ment of new programs such as child nutrition and Head Start; and urban
revitalization efforts launched under the auspices of the new Housing and
Urban Development cabinet agency. The flood of Great Society legislation
in 1964–1965 reinforced the connection between strong government and the
Democratic Party to a degree unseen since the New Deal era. Yet by 1966 the
moment had passed. A combination of economic problems caused by funding
the Great Society and war in Vietnam along with domestic unrest (detailed
below) precluded further reform.
Post-Great Society, Democrats continued to champion domestic policy lib-
eralism, albeit in a more tempered form. Universal health care remained a top
priority. On the hustings, candidates such as Walter Mondale and Michael
Dukakis called for expanded health care access, but the party had no real
prospects for success until 1992 when Bill Clinton unseated George H. W.
Bush. Candidate Clinton pledged federal action to attack rising health care
costs and provide coverage for those who lacked it. He also staked out centrist
ground by pledging to “end welfare as we know it,” a standard Republican
refrain. But after running as a “new kind of Democrat,” it was telling that
president Clinton took up health care first. After a promising start in 1993, the
reform effort fell apart the following year when Congress began to consider
the plan. Resistance from mid-sized insurance companies and small business,
lockstep Republican opposition, and public ambivalence combined to sink the

proposal. But even in defeat Clinton, like Truman some four decades earlier,
underscored his party’s commitment to government activism to help vulner-
able subsets of the population.
On some economic welfare issues, Clinton took more centrist positions
than those adopted by his Democratic predecessors. Most memorably, he
signed welfare reform legislation in August 1996 that transformed cash assis-
tance programs for the poor in ways that troubled many liberals. Clinton’s
deviation from party orthodoxy on this matter did not mean he had aban-
doned Democrats’ commitment to the middle, working, and lower classes. He
initially vetoed two Republican welfare bills as too draconian and promised
to fix the one he signed after the election. Moreover, in battles with the GOP
Congress his spirited defense of social insurance and safety net programs such
as Medicare, Medicaid, and other programs (e.g., school lunches) reaffirmed
his dedication to helping economically vulnerable constituencies.
During the 2008 election season, Barack Obama took up the reform cause
once again by vowing to fix the health care mess. Like Clinton, Obama backed
federal efforts to ensure universal coverage and control escalating health care
costs. After the election, the Democratic Congress spent much of 2009 and
early 2010 marking up several bills. The political environment was compli-
cated by unified Republican opposition and rising anger among “Tea Party”
sympathizers. The fate of health care legislation was often in doubt, but the
Democrats prevailed over unanimous Republican opposition when president
Obama signed the Patient Protection and Affordable Care Act in March 2010.
This arduous struggle demonstrated anew where the parties stood on strong
versus limited government. The same point holds true for the nearly 800 bil-
lion dollar stimulus package passed by the Democratic Congress, with almost
no GOP support, and signed into law by the president in February 2009.
Having surveyed the Democratic Party’s stance toward government
from Roosevelt to Obama, I now address the Republican side of the story.
In the aftermath of the 1932 election debacle, the GOP continued to defend
small government and laissez-faire economics. During Roosevelt’s first term
Herbert Hoover frequently denounced the New Deal. Alf Landon followed
suit on the campaign trail in 1936, but after his crushing defeat GOP opposi-
tion to the New Deal softened. Recognizing that the country had sided with
Roosevelt, Republican candidates such as Wendell Willkie in the 1940s, Dwight
Eisenhower in the 1950s, and Richard Nixon in the 1960s no longer called for
its eradication. Notwithstanding this moderation, Republicans remained com-
mitted to smaller government and fiscal prudence. Moreover, partisan images
forged during the Great Depression lingered for decades. Voter recollections
of mass joblessness, soup lines, and “Hoovervilles” did lasting damage to the
Republican brand (Campbell et al., 1960; Lubell, 1956).
In 1964, Barry Goldwater offered America “a choice, not an echo” by
roundly condemning the welfare state. The Senator from Arizona had long
railed against big government and bureaucratic centralization and praised
individual initiative and the free market. Although he backed away from
some earlier controversial comments, such as making Social Security a vol-
untary program, he left no doubt about what he believed. After Johnson
routed Goldwater, neither Nixon nor his successor Gerald Ford challenged the
legitimacy of core New Deal functions. Indeed, although Nixon derided the
Great Society and often clashed with congressional Democrats over specific
programs, his administration oversaw a sizable expansion of social services
(e.g., expanded Social Security benefits) and federal regulatory capacities (e.g.,
the Environmental Protection Agency and Occupational Safety and Health
Administration were launched during his administration). Of course, the driv-
ing forces here were strong Democratic majorities in the House and Senate.
Nixon cared more about foreign policy and was content to sign Democratic
sponsored bills backed by the public.
By the late 1970s a faltering economy, declining trust in government, and a
resurgent business community helped pave the way for a renewed assault on
the welfare state. Campaigning in 1980, Ronald Reagan proclaimed “govern-
ment is not the solution to our problems, government is the problem.” Reagan
argued that cutting taxes and slashing government spending would revital-
ize an economy ravaged by inflation, unemployment, and murderous inter-
est rates. After his victory, Reagan moved quickly to pass his program. He
delivered in the summer of 1981 by signing the Economic Recovery Tax Act
that phased in substantial personal income and business tax cuts over 3 years,
reduced a number of means tested programs for the poor, terminated public
employment programs, and removed hundreds of thousands of beneficiaries
from the Social Security disability program.
Although the administration had some success in scaling back safety net
programs, wholesale restructuring of the welfare state remained elusive.
Nevertheless, Reagan transformed the political landscape (Collins, 2007;
Wilentz, 2008). Before Reagan, the Republican Party had for the most part
accommodated itself to the existence of the welfare state. Republicans might
not have liked it and frequently sought to resist its expansion, but they knew
where public sentiment lay. After the triumph of “Reaganomics,” limited gov-
ernment and lower taxes became defining features of the Republican Party
brand. Efforts by GOP congresses in the mid-1990s to reduce funding for var-
ious safety net programs, the Social Security partial privatization plan offered
by George W. Bush in 2000 and 2004–2005, unified Republican opposition to

the Clinton and Obama health care initiatives, and the 2001 and 2003 Bush tax
cuts underscore this point. Despite their commitment to smaller government,
Republican presidents have deviated occasionally from party orthodoxy
when economic or political imperatives compelled strategic moderation. For
example, both Reagan and the first president Bush raised taxes to cut mush-
rooming federal deficits. Without denying the importance of such exceptions,
the broader point to keep in mind is that the GOP has long been philosophi-
cally opposed to Washington led efforts to help citizens do what, Republicans
believe, citizens should do for themselves.
To sum up, my historical sketch establishes several points about the poli-
tics of economic welfare. First, the question of whether the federal government
should help those struggling to help themselves has been a central feature—
arguably the central feature—of the domestic policy agenda since 1932.3 Key
(1966) provides a lucid summation of the changes wrought by the New Deal.
Before 1932 the federal government had been a remote authority with
limited range. It operated the postal system, improved rivers and har-
bors, maintained armed forces on a scale fearsome only to banana repub-
lics, and performed other functions of which the average citizen was
hardly aware. Within a brief time it became an institution that affected
intimately the lives and fortunes of most, if not all, citizens. Measures of
recovery and of reform . . . contributed to this fundamental alteration of
federal activities. (p. 31)
Since the New Deal era, philosophical battles over what the federal gov-
ernment should do have profoundly impacted American politics and soci-
ety. Debates over the Fair Deal, the Great Society, Reaganomics, health care
3
Related questions about the role of government have infused public debate at
different periods in American history. For instance, questions about federal sup-
port for agriculture were paramount during the New Deal era. Likewise, ques-
tions about government regulation of business have arisen at various moments
in time since the 1970s. My point is not that these are trivial controversies. They
clearly matter, especially to those most directly affected by the action in ques-
tion. Instead, my claim is that ordinary citizens are less likely to pay attention to
matters such as agricultural subsidies or deregulation than to the broader ques-
tion of whether the government in Washington has a responsibility to help those
struggling in the market economy. For evidence that the public thinks about gov-
ernment activism along these lines, see Campbell et al. (1960), Jacoby (1994), and
Lewis-Beck et al. (2008).
reform, tax versus spending tradeoffs, and entitlement reforms have domi-
nated domestic policy like no other cluster of issues. Other conflicts have mat-
tered during various intervals, including issues such as civil rights for blacks
in the 1960s and busing and affirmative action in the 1970s (I have more to say
about race later in the chapter); however, in terms of salience and persistence
as a fundamental policy cleavage none can approach, let alone match, conflict
over the size and scope of government.
Second, presidential incumbents, those seeking the office, and the national
parties have, generally speaking, taken distinct positions on this dimension.
Sometimes the space between the candidates has been immense, such as 1936,
1964, 1980, and 2008. At other times, the distance has narrowed, such as 1952,
1956, 1960, and 1976. Moreover, some candidates have adopted positions on
discrete issues associated with the opposition party, such as Clinton’s 1992
welfare reform pledge or Bush’s vow in 2000 to seek a Medicare prescrip-
tion drug benefit. Although deviations like these suggest Clinton was less
committed to activist government than Johnson and Bush less conservative
than Goldwater, historical continuities surely matter less to voters than where
the current candidates stand. The bottom line is that although the personali-
ties and politics have changed a good deal from the 1930s to the present, the
Democratic Party’s commitment to marshalling federal power to help those at
the economic margins of society has held steady. And although the GOP has
had its share of centrists in the past, the party has always stood to the right of
Democrats on this continuum (cf. Gerring, 1998).
THE TRADITIONAL MORALITY CLEAVAGE

Divisions over which set of moral standards should guide the cultural and
social life of the nation moved onto the political agenda in the 1960s.4 In this
section I consider the social upheavals that have rocked American society
these past 50 years and explain how presidents and the parties responded to
these accumulated shocks. To begin, at the dawn of the 1960s traditional (i.e.,
Christian and bourgeois) standards of moral behavior governed American
society. By decade’s end this was no longer the case. This break from the past
had far reaching implications for American politics. This is not to say moral
4
The following works have guided my analysis of cultural issues: Boyer (2001),
Blum (1991), Brewer and Stonecash (2007, 2009), Carter (1995), Collins (2007),
Halberstam (1993), Hunter (1992), Layman (2001), Liebman, Wuthnow, and Guth
(1983), Patterson (1996, 2005), Perlstein (2008), Scammon and Wattenberg (1970),
Sundquist (1983), Wilentz (2008), and Wuthnow (1988).
issues were inconsequential prior to the 1960s, but rather that their relevance
ascended to a degree hitherto unseen. Moreover, previous moral disputes had
usually been tied to single issues such as abolition, evolution, or temperance,
to name a few, rather than reflecting a broader cultural syndrome.
The Supreme Court fired some early salvos in the culture wars with two
controversial decisions passed down early in the decade. In Engel v. Vitale
(1962) the Court ruled voluntary and nondenominational prayers in public
schools violated the first amendment establishment clause and, hence, were
unconstitutional. The following year the Court barred schools from mandat-
ing Bible study (Abington School District v. Schempp, 1963). Ignoring nearly two
centuries of practice, within a year the Supreme Court had banished God and
religion from the nation’s classrooms. Although hailed by some, these deci-
sions angered the much larger share of the public that believed religion had a
role to play in public schools (Page & Shapiro, 1992).
Conflict in the schools spilled over to the curriculum. One set of skirmishes
centered on the Biblical view of creation. The Supreme Court undercut tra-
ditional viewpoints again by invalidating state laws prohibiting the teaching of
evolution (Epperson v. Arkansas, 1968). Since that time the Supreme Court and
other lower courts have repeatedly barred creationism from public schools.
Battles over sex education represented another front in the curriculum wars.
The topic had been taught in some areas of the country for decades, but the
emphasis had been on biology, chastity, and personal hygiene. In the 1960s
coverage expanded to more controversial topics and new parts of the country.
Previously taboo subjects, such as masturbation, birth control, and premari-
tal sex, now supplemented standard fare. Needless to say these changes pro-
voked resistance in many areas, dissension that continues to this day.
Next, we have the sexual revolution. The development and dissemination
of cheap and reliable oral contraceptives helped transform sexual practice by
giving women more control over their sexual choices and reducing the risks
of unplanned pregnancies. The 1965 Supreme Court decision in Griswold v.
Connecticut, which decreed married couples were entitled to contraceptives,
and the 1972 Eisenstadt v. Baird case, which extended this right to everyone
else, facilitated dissemination. Additionally, discussion of human sexuality
moved into public discourse in ways previously unimaginable. The pioneer-
ing studies of Alfred Kinsey in the late 1940s and early 1950s on male and
female sexual behavior, along with William Masters and Virginia Johnson’s
1963 study on the physiology of sex, catalyzed this process. Mass circulation
publications reflected evolving mores as well. To take a couple of examples,
Playboy magazine, with its photographs of nude women, debuted in 1953 and
grew in popularity over the next two decades. In 1962, Helen Gurley Brown’s
book Sex and the Single Girl, which encouraged women to find sexual fulfill-
ment before—or in place of—marriage, was published and sold well. As a
result of these and other factors, public attitudes toward sex liberalized, par-
ticularly in matters of premarital relations and cohabitation. These attitudes,
in turn, reinforced changing practices, especially among the young, many of
whom openly flaunted the conventions of their parents and elders. This is
not to say that most joined the revolution or that traditional courtship rituals
such as “going steady” were abandoned. The difference was that once settled
norms were now competing with newer, more lenient standards.
Sex spread to popular culture. From everyday speech to receding hemlines
to advertising, sex was out in the open like never before. On television, 1950s
family sitcoms such as Leave it to Beaver and Ozzie and Harriet that featured
wholesome middle class families gave way to 1970s shows such as Charlie’s
Angels and Three’s Company that featured attractive women prancing about
and adopting provocative poses on the sets. By the 1990s, programs such as
Sex and the City celebrated casual sex while pornography was available widely
on cable television and the Internet. Crassness permeated other areas of popu-
lar culture as well: music became louder and more offensive, Hollywood fare
increasingly lurid, and standup comedy more profane.
Another departure from convention concerned the changing role of women
in American society. At the beginning of the 1960s, women were expected to
marry relatively young, raise children, and leave careers to their husbands,
in short, to place devotion to family above all else. “The two big steps that
women must take are to help their husbands decide where they are going and
use their pretty heads to help them get there” Mrs. Dale Carnegie intoned in
Better Homes and Gardens in 1955 (Halberstam, 1993, p. 592). Cultural mores
began to change rapidly over the next decade. Betty Friedan’s 1963 book The
Feminine Mystique attacked gender norms for condemning (educated, middle
class) women to lives of domestic tedium and unfulfilled potential. Millions
of women proved receptive to this message. Changing mindsets were accom-
panied by interest group mobilization. In 1966, activists founded the National
Organization for Women (NOW) to advocate for federal action across a range
of women’s issues. Although few women thought of themselves as feminists
or joined such organizations, many sympathized with movement aspirations
and goals. Of course, millions more women, and not a few men, viewed fem-
inism with utter disdain.
Abortion came next. Building on its earlier Griswold decision, which estab-
lished an inherent right to privacy, the Supreme Court in Roe v. Wade (1973)
invalidated state laws criminalizing abortion. This tremendous victory for the
women’s movement evoked disbelief and horror in the minds of those who
revered the “sanctity of life.” The decision had real consequences as abortion
rates accelerated thereafter before leveling off in the 1980s (Brewer & Stonecash,
2007, p. 97). Since then pro-life and pro-choice forces have clashed at abortion
clinics, at the ballot box, in statehouses, in Congress, and in state and fed-
eral courts. Two notable Supreme Court cases (Webster v. Reproductive Health
Services, 1989 and Planned Parenthood v. Casey, 1992) placed some restrictions
on abortion but preserved a woman’s right to choose. However, in Gonzales
v. Carhart (2007) the court upheld a ban on partial birth abortions, a decision
widely viewed as a setback for abortion rights.
Social change has affected the American family as well. Female labor partic-
ipation, including that of mothers with young children, climbed steadily after
1960 (Brewer & Stonecash, 2007, pp. 99, 102). Although many were untroubled
by this, others despaired at the falling number of stay at home moms and the
growing number of “latchkey” children home alone. The increasing economic
and sexual independence of women along with a rise in no fault divorce laws
contributed to a jump in the divorce rate between 1960 and the mid-1970s
(Brewer & Stonecash, 2007, p. 104). Lastly, society witnessed an ever mount-
ing rise in the number of out-of-wedlock births from the early 1960s onward
(Brewer & Stonecash, 2007, p. 105). These changes left many children in single
parent (almost always female headed) households. For those committed to
traditional family values, such developments were deeply unsettling.
Another threat to old fashioned ways emerged from an unanticipated
direction—that of the homosexual community. Throughout American history,
gays and lesbians had been ridiculed, marginalized, and persecuted. To take a
few examples, same sex relations were criminalized, the American Psychiatric
Association (APA) classified homosexuality as a mental illness, and gays were
barred from military service. A few homophile organizations founded in the
1950s sought to improve matters, but there existed no mass movement work-
ing on behalf of gays and lesbians. Things began to change in the summer of
1969. On June 27 police raided the Stonewall Inn in Greenwich Village, New
York City, a bar with many gay patrons that was subject to constant police
harassment. On this night, gays fought back for the first time and battled cops
for several days. The “Stonewall Riot” awakened group consciousness and
the gay liberation movement was born. Although the movement experienced
some early successes (e.g., gay pride parades were organized in many large
cities and the APA declassified homosexuality as a mental disease), struggles
persisted over antisodomy laws, AIDS, gays in the military, same sex marriage,
and other issues. The marriage issue in particular has been tremendously divi-
sive, especially at the state level where multiple ballot initiatives banning gay
marriage have appeared in recent years.
I conclude with the “law and order” issue. On August 6, 1965, President
Johnson signed the Voting Rights Act into law. Many hoped this landmark
achievement, along with the 1964 Civil Rights Act, would mark the begin-
ning of the end of America’s race problem. Five days later, the predominantly
black Watts neighborhood in Los Angeles was engulfed in flames after white
policemen pulled over a black motorist for speeding. A routine moving viola-
tion quickly spiraled out of control, causing a riot that lasted 6 days, claimed
34 lives, and caused 35 million dollars worth of damage (Blum, 1991, p. 253).
Riots erupted in many other cities over the next three summers. Many blacks,
long frustrated by discrimination and segregation, a lack of opportunity, and
police harassment, felt trapped and hopeless. For many, formal legal equality
and the franchise did little to improve their daily lives.
Worries about law and order were not confined to central cities. Unrest
on college campuses fed the belief among some people that the social order
was collapsing. New Left organizations, such as Students for a Democratic
Society, began protesting Johnson’s Vietnam policies in 1964. There were other
grievances. For instance, the 1964 Free Speech Movement at the University of
California, Berkeley sought to expand the rights of political speech on campus.
It garnered much attention as thousands of student protesters clashed openly
with university police. Another notable rebellion occurred in 1968 when stu-
dent radicals at Columbia University occupied several campus buildings for
6 days to decry various university policies before being routed by police.
Although many young people were sympathetic to causes espoused by cam-
pus activists, others in the broader public were mortified as they watched
events unfold on their television screens.
Fear of crime has long outlived the 1960s. Though media coverage often
inflamed public anxiety, such fears were tied to real world events as rates of
murder, violent crime, juvenile delinquency, and drug usage climbed inex-
orably throughout the 1960s and 1970s (Brewer & Stonecash, 2007, p. 107;
Miringoff & Miringoff, 1999, p. 45). In the 1980s, new concerns arose about the
underclass, crack cocaine, gangs, and teenage drug use. Public outrage helped
pave the way for the war on drugs, “three strikes and you’re out” sentencing
guidelines, and an exploding prison population.
To sum up, the far reaching changes unleashed during the 1960s struck at
the heart of traditional morality. The fact that such changes touched on the
most basic and personal notions of right and wrong ensured cultural con-
flict would attract extraordinary attention. Though many who came of age
during these turbulent times embraced novelty or took it in stride, the large
share of the public wedded to conventional moral standards was appalled
by what they saw. Among the most distressed were evangelical Christians
and committed Roman Catholics. Many members of the middle and working
classes, though not especially pious, were also deeply troubled by the ravages
of modernity. For culturally conservative groups such as these, the disparate
issues were symptomatic of a broader problem—the collapse of moral author-
ity. In their view, America was becoming too liberal, too secular, too permissive.
As the sociologist Jerome Himmelstein (1983, p. 16) describes the syndrome:
“Abortion, the Equal Rights Amendment, busing, affirmative action, sexual
promiscuity, drugs, prohibitions on school prayer, the secular curriculum in
public schools, and many similar things are opposed on the grounds that they
contribute to this process of social breakdown and moral decay.”
Several developments would help pave the way for the eventual politiciza-
tion of these grievances. To begin, the post-World War II era witnessed a sharp
rise in the size of the Protestant evangelical community along with a corre-
sponding decline in mainline Protestant denominations, providing future
leaders with a wellspring of religious conservatives to back demands for
moral restoration. Moreover, differences in the educational and social bases of
Protestant and Catholic denominations waned, thereby enhancing the pros-
pects of interfaith cooperation among conservative Christians on the right or
religious liberals on the left. A 1978 Internal Revenue Service (IRS) proposal to
strip the tax exempt status of private Christian schools that failed to meet its
standard of racial integration ignited a firestorm of protest within the evangel-
ical community. Finally, skilled leadership channeled moral indignation into
a potent political force.
To build on this last point, charismatic preachers used local pulpits, tel-
evision ministries, and radio programs to rail against the heresy of liberal
Christian denominations and secular humanism more broadly. Jerry Falwell,
Pat Robertson, James Dobson, and others lamented America’s cultural decline
and began organizing in the mid-1970s. Political operatives played a key role
in the rise of the New Christian or Religious Right. Conservative activists such
as Paul Weyrich and Richard Viguerie, who had been involved in secular pol-
itics, wanted a means to push the Republican Party to the right. Recognizing
that previously apolitical evangelicals and fundamentalists could help serve
this goal, they persuaded religious leaders such as Falwell to get involved in
politics. By the end of the 1970s, the Religious Right had coalesced into a loose
collection of organizations capable of rallying evangelicals. Organizations such
as the Moral Majority, the Religious Roundtable, the Christian Voice, Focus on
the Family, the Family Research Council, and then, in the 1980s and 1990s,
the Christian Coalition, sought to harness devotees’ anger—and dollars—
to support public officials who shared their views. Other groups mobilized
around a particular cause, such as the Concerned Women for America, Phyllis
Schlafly’s Stop ERA and Eagle Forum, Randall Terry’s Operation Rescue, and
various antigay and antipornography organizations. Though airing different
complaints, they were united by a transcendent concern with moral decline.
This burgeoning discontent provided strategic politicians with an oppor-
tunity to reap success at the ballot box by capitalizing on collective anxieties
about social change. George Wallace was the first to campaign on the “social
issue” during his long shot bid for the 1964 Democratic Party nomination.
During the general election campaign that fall Barry Goldwater also lamented
moral decline. For both men this strategy proved futile against the Johnson
juggernaut. Two years later, however, Ronald Reagan’s decisive victory over
incumbent Pat Brown in the 1966 California gubernatorial race suggested
cultural backlash could yield electoral success. Taking notice, Richard Nixon
aped Reagan in 1968 by campaigning on law and order, time honored values,
and moral decency in order to court support from what he would later call the
“silent majority.”
Throughout his term in office and during the 1972 campaign, Nixon posi-
tioned himself as the champion of “Middle America” and the traditional
values it subscribed to. That year Democratic challenger George McGovern
played the foil by taking left of center positions on new cultural issues such as
busing, amnesty for Vietnam draft evaders, and penalties for marijuana use.
The Democrats’ commitment to cultural liberalism was on full display at their
national convention that July. As presidential campaign chronicler Theodore
White described it (1973):
One could . . . watch the parade of women across the podium . . . forc-
ing into politics matters never before publicly discussed at a national con-
vention—for example, the laws of sex . . . Then came homosexuals to the
microphone and camera, men openly demanding before the nation that
the coupling of males be accepted not furtively, but as a natural and legal
right. (p. 180)
With Nixon defending establishment values and McGovern ensconced in

the modernist camp, partisan images began to take shape before crystalliz-
ing some years later when the national parties polarized on hot button issues
such as abortion, gay rights, gun control, and related concerns (Adams, 1997;
Layman, 2001; Lindaman & Haider-Markel, 2002).
By 1980, GOP activists recognized that the social issue presented a golden
opportunity to fracture the Democratic coalition. Many whites loyal to the
party of Roosevelt because of bread and butter economic issues were more con-
servative than the party on moral issues. Therefore, a culturally conservative
candidate atop the ticket might prove attractive to some Democrats (e.g.,
working class Roman Catholics, southern evangelicals). That candidate was
Ronald Reagan. Given the Religious Right’s organizational prowess and poten-
tial to mobilize votes, Reagan eagerly courted their support. At the Religious
Roundtable’s annual meeting in Dallas that summer, Reagan delighted his
audience by saying “You may not endorse me, but I endorse you” (Wilentz,
2008, p. 123). New Christian Right leaders let the faithful know Reagan was
their man in 1980 and again in 1984. Although he delivered little in the way of
tangible results, President Reagan offered symbolic and rhetorical support to
those who revered traditional and religious values. Conservative Christians
and other Middle Americans moved steadily into the GOP fold throughout
the 1980s (Greeley & Hout, 2006; Layman, 2001).
Cultural divisions have intensified since then. Recall Patrick Buchanan’s
clarion call for a relgious war at the 1992 Republican Party convention. Other
examples of partisan cultural conflict spring to mind, such as Republican
attacks on Clinton’s moral character that culminated in his 1998 impeachment;
the divisive battles over gay rights during the Clinton and Bush II years at
the national and state level; renewed struggles over the teaching of creation-
ism, rebranded as intelligent design, in public schools; and the intractable and
sometimes deadly fights over abortion.
Having reviewed social and political developments in the cultural issues
domain over the past 40 years, I now summarize the takeaway points. Moral
conflict moved into the public sphere in the 1960s and has remained deeply
entrenched ever since. Although the battles have often focused on concrete
issues such as school prayer, feminism, abortion, gay rights, and gun control,
the broader battle has always been about what set of moral standards should
govern the public and private life of the nation (Hunter, 1992). As the historian
Robert Collins (2007, p. 173) has put it, this “clash between the older ortho-
doxy of Christian values and the beleaguered but still dominant bourgeois
order on the one hand and the new radical cultural forces on the other consti-
tuted the central battle of the culture war.”
As the tides of change swept across American society and advocates of
cultural liberalism mobilized to press a disparate set of demands, backlash
emerged from the right as cultural and religious conservatives entered poli-
tics to protect traditional rules of moral behavior. GOP officials, on the lookout
for new issues to blunt the Democratic advantage in the economic welfare
domain, moved quickly to exploit the opportunities afforded by this backlash.
Throughout his first term Nixon spoke for those who feared or resented social
change by paying tribute to traditional family values, while the Democratic
Party moved left on moral standards, especially during the 1972 campaign.
Although the parties did not polarize immediately on these matters, the seeds
of future conflict were planted by the mid-1970s. With the election of Reagan
in 1980 polarization began before accelerating once Clinton moved into the
Oval Office 12 years later. The process continued during the Bush presidency
of 2001–2009. In this manner, apparently disconnected battles over a plethora
of moral issues have evolved into a deep-seated cleavage dividing the parties
and their presidential standard bearers.
THE MILITARY STRENGTH CLEAVAGE

The hawk–dove division represents the third great policy cleavage that has
structured American politics these past several decades.5 Prior to World War II
the main foreign policy cleavage pitted isolationists, who believed American
interests were best served by avoiding entangling alliances in the old world,
against internationalists, who felt engagement advanced U.S. objectives. The
stay-at-home sentiment, prevalent throughout much of the early twentieth
century, surged during the 1930s in response to economic calamity at home
and war clouds gathering abroad, until the carnage of World War II eroded
mass and elite support for isolationism (Berinsky, 2009; Herring, 2008; Page
& Shapiro, 1992). The catalyzing event was the surprise Japanese assault on
Pearl Harbor on December 7, 1941, which shattered the myth that America was
safe from attack. More broadly, the war underscored the need for American
involvement in the international arena to help ensure collective security and,
thus, avoid future conflagrations that could draw the United States in once
again.
A critical problem confronting policymakers in the war’s immediate after-
math centered on what combination of hard and soft power would advance
U.S. foreign policy and national security objectives. Officials across the par-
tisan spectrum generally agreed that military strength was essential to this
task. In the early Cold War years, consensus was reflected in bipartisan back-
ing for Truman’s containment policy, which sought to check Soviet expan-
sion through a combination of military deterrence, economic aid, and political
measures; the decision to support West Berlin against communist pressure;
and Truman’s decision to fight in Korea. There were limits to bipartisanship,
5
These works guided my review of foreign policy history: Addington (2000),
Aitken (1993), Berinsky (2009), Blum (1991), Collins (2007), Dallek (2003), Gaddis
(2005), Halberstam (2001), Herring (2002, 2008), Oberdofer (1998), Packer (2005),
Patterson (1996, 2005), Perlstein (2008), Ricks (2006), Sundquist (1983), Suskind
(2006), and Wilentz (2008).
of course. Disputes predictably arose over which party could better protect
the nation. For instance, after Chinese forces entered the Korean War in 1950,
some congressional Republicans backed Douglas MacArthur’s call to expand
the war to the Chinese mainland. Likewise, during the 1952 campaign the
GOP tarred Democrats as soft on communism. Eight years later John Kennedy
returned fire by promising to fix a “missile gap” that allegedly developed on
Eisenhower’s watch. Without gainsaying these differences, the important
point to remember is that politicians from both sides of the aisle were commit-
ted to maintaining American power and prestige via military strength.
The first strains in the consensus surfaced during the Vietnam War era. In
1964, U.S. intelligence indicated that absent sustained U.S. support the gov-
ernment of South Vietnam would fall to communist insurgents, known as the
National Liberation Front (NLF), who were backed by Ho Chi Minh’s commu-
nist government in North Vietnam. Recalling what “losing China” had cost
Truman, Johnson feared disengaging in Vietnam would render him vulner-
able to blistering criticism from the right. But he knew that an intensifying
military conflict during an election year would be politically toxic as well. So
he campaigned as a man of peace, promising the American public “we are not
about to send American boys nine or ten thousand miles away from home
to do what Asian boys ought to be doing for themselves” (Goldberg, 1995,
p. 231). Taking no chances, he further implied GOP opponent Barry Goldwater
was an erratic warmonger not to be trusted with his finger on the nuclear
button.
After the election the choice confronting the administration was whether
to escalate or negotiate a settlement with the communists, which would
effectively doom the government of South Vietnam. Johnson chose escala-
tion, in large part because he feared letting South Vietnam fall would under-
cut domestic support at home and thereby cripple his ability to construct his
Great Society. From early 1965 to the end of 1967, the U.S. troop presence
shot up from 23,000 to 485,000 (Herring, 2002, p. 182). Despite this rise and
a sustained bombing campaign against enemy positions, the NLF and North
Vietnam matched U.S. escalation at every turn. Seeking to allay the concerns
of an increasingly anxious public, the Johnson administration launched a
public relations offensive in November 1967. Vietnam Commander William
Westmoreland returned home to shore up public support for the war, telling
reporters “We are making real progress” and “We have reached an important
point where the end begins to come into view” (Herring, 2002, p. 221). Vice
president Hubert Humphrey toed the administration line a few days later:
“there has been progress on every front in Vietnam. . . . There is no military
stalemate” (DeGroot, 2008, p. 282).
This sense of optimism was shattered 3 months later on January 31, 1968
when the communists launched the Tet Offensive in scores of urban areas
throughout South Vietnam thought to be immune from danger, the most
notable example of which was the assault on the U.S. embassy in Saigon.
Overcoming early setbacks in the field, U.S. and South Vietnamese regulars
regrouped to inflict punishing losses on enemy personnel. Although suf-
fering a bruising military defeat, communist forces scored a major political
victory by helping tip U.S. mass and elite opinion against the war. During
the February 27 broadcast of the CBS evening news, venerable anchorman
Walter Cronkite concluded “we are mired in a stalemate.” Public support for
Johnson’s handling of the war dropped precipitously (Page & Shapiro, 1992,
pp. 57, 232–233). Several weeks later in a nationally televised speech on March
31, Johnson announced a bombing halt, reaffirmed his commitment to peace,
and stunned the nation by withdrawing from the presidential race. On April
9 the administration announced a troop ceiling and the gradual transfer of
responsibility for waging war to South Vietnamese forces.
The war fomented deep divisions in American society. Antiwar dissent
was confined initially to college campuses but eventually spread to other
sectors of society. A number of clergy spoke out against Vietnam, Martin
Luther King, Jr. most notably. Celebrities criticized the war as well, such as
the pediatrician Dr. Benjamin Spock, the actress Jane Fonda, and heavyweight
boxing champion Muhammad Ali. Mass demonstrations, including a 1967
march on the Pentagon, gave vent to public frustration with the war. Johnson
probably could have weathered such criticism, but when dissent spread to
the Democratic Party establishment his position became increasingly ten-
uous. In 1966 Senator William Fulbright, in his capacity as chairman of the
U.S. Senate Foreign Relations Committee, held televised hearings critical of
the war. Adding to Johnson’s woes, Senator Robert Kennedy, a potential rival
for the 1968 nomination, later called for a negotiated settlement with Ho Chi
Minh’s government. Finally, during the 1968 New Hampshire primary, peace
candidate Eugene McCarthy shocked the country by capturing 42% of the
vote (compared to the 49% write in vote for President Johnson). At its chaotic
national convention in Chicago that August, the Democrats’ hawk and dove
factions clashed openly on the convention floor. The postwar consensus in
favor of military power was collapsing within the Democratic Party.
After the election, Vietnam became Nixon’s war. Adopting his predeces-
sor’s “Vietnamization” policy as his own, the new commander in chief sought
to transfer the combat burden from American to South Vietnamese troops
while increasing U.S. aid to its ally. Given the likelihood that U.S. withdrawal
might occur before South Vietnamese forces were ready to take up the fight,
this strategy carried considerable risk. Consequently, Nixon deployed a vari-

ety of aggressive military tactics to degrade enemy capabilities. These included
several massive bombing campaigns against North Vietnamese targets at
levels far above those permitted by Johnson, authorizing a separate (initially
secret) bombing campaign against enemy sanctuaries in Cambodia in 1969,
and the 1970 invasion of that country. Yet Nixon also fulfilled his promise to
draw down U.S. troops, which declined from a peak of 543,000 in April 1969
to 24,000 by the end of 1972 (Herring, 2002, p. 182). With the signing of the
Paris Peace Accords in January 1973, the U.S. presence in Southeast Asia effec-
tively ended. The war lasted another 2 years until communist forces overran
Saigon on April 30, 1975. Vietnam left deep scars on the national psyche. In the
post-Vietnam era, the public grew more reluctant to use military force abroad
(Page & Shapiro, 1992). In addition to this, the disaster challenged prevailing
elite assumptions about the propriety and efficacy of using military force to
advance U.S. objectives. The “ghosts of Vietnam” would haunt foreign policy
debates for many years.
The second major foreign policy issue confronting Nixon was how to man-
age relations with the communist superpowers. Nixon, along with National
Security Advisor Henry Kissinger, believed that conciliation with our adver-
saries would lock the superpowers into a mutually beneficial “structure of
peace.” Recognizing the relative decline in U.S. power, Nixon and Kissinger
believed exploiting the Sino-Soviet split could enhance U.S. security, promote
international stability, and elevate Nixon’s stature at home in advance of his
reelection bid. Nixon’s initial gambit was the opening to China. His week-
long trip to that country in February 1972 had enormous symbolic value and
yielded some tangible benefits, most immediately, leverage to use in subse-
quent negotiations with the Soviets and some oblique assurances the Chinese
would not intervene in Vietnam. Three months later, the president met with
Soviet Premiere Leonid Brezhnev in Moscow. They signed the Strategic Arms
Limitations Talks (SALT I) and Anti-Ballistic Missile (ABM) treaties, which
placed some limits on the arms race. Both nations also stood to benefit from
increased trade.
When the presidential campaign kicked off that fall, the combination of
the nearly complete withdrawal of U.S. ground forces from South Vietnam,
the prospect of a peace accord with the communist North, and the diplomatic
coups with China and the U.S.S.R. put Nixon in a commanding position. He
could claim to be a man of peace. And although McGovern had long taken
positions well to the left of Nixon on Vietnam and the U.S. defense posture,
it is hard to see how conditions in Southeast Asia would have deviated much
under a new McGovern administration.
Questions about the utility of military power remained after Nixon resigned
in disgrace in 1974, but party positions would not crystallize fully until the
early 1980s. Gerald Ford initially supported the Nixon and Kissinger policy
of détente. However, hard liners in both parties attacked the policy, while
relations with the Soviet Union deteriorated throughout the 1970s. Despite
the increasing tensions, arms control talks continued and ultimately culmi-
nated in the 1979 SALT II treaty, which Jimmy Carter and Brezhnev signed in
June. But by the end of the year, NATO had consented to the deployment of
new missiles in West Europe and the Soviet Union had invaded Afghanistan.
Carter withdrew SALT II from Senate consideration, imposed a grain embargo
against the Soviets, and called for rapid defense spending increases. Détente
was dead. In addition to inflamed Cold War tensions, the Iranian hostage cri-
sis fed public perceptions of U.S. vulnerability in an increasingly dangerous
and unstable world.
During the 1980 presidential campaign, Ronald Reagan attacked Carter for
being weak on national security and promised to rebuild American military
strength and restore U.S. prestige in the world. Throughout much of his first
term, Reagan damned the Soviets. For example, in his first press conference he
condemned Soviet leaders for reserving “unto themselves the right to commit
any crime: to lie, to cheat” and in a 1983 speech called the U.S.S.R. an “evil
empire” (Collins, 2007, p. 195; Patterson, 2005, p. 194). Reagan sought “peace
through strength” by overseeing the largest peacetime military buildup in
U.S. history. NATO’s 1983 deployment of the next generation of cruise mis-
siles on West European soil aroused intense opposition from the Soviets and
revitalized the peace movement. The administration supported a number of
anticommunist movements throughout the world, including the contras in
Central America and the mujahedeen in Afghanistan. U.S. military force was
deployed in the 1983 invasion of Grenada and the 1986 bombing of Libya.
Finally, Reagan’s support for the controversial Strategic Defense Initiative
(SDI), a space-based defensive shield against intercontinental missiles, accen-
tuated the president’s commitment to military strength (defensive in this
case). Of course, Reagan saw SDI as a means to transcend nuclear conflict, but
that position was not always clear to the public at the time. All these signs left
little doubt where Reagan, and by extension the GOP, fell on the hawk–dove
continuum. Consistent Democratic opposition to the Reagan defense posture
helped clarify party positions on this dimension.
Moving beyond the military buildup and saber rattling of his first term,
Reagan pursued a different tact after reelection. Reagan and Soviet leader
Mikhail Gorbachev held summits in Geneva in 1985, Reykjavik in 1986,
Washington in 1987, and Moscow in 1988. Reagan, who was genuinely
dedicated to the elimination of nuclear weapons, and Gorbachev, who saw

economic, political, and strategic benefits in a reduction of Cold War ten-
sions, developed a good rapport during the first summit. Two years later,
they signed the landmark Intermediate-range Nuclear Forces (INF) Treaty in
Washington, DC that for the first time eliminated entire classes of nuclear mis-
siles. Reagan’s deft use of the tools of diplomacy enhanced U.S. foreign policy
and security goals.
Important though such developments were, their ability to revise how
people viewed the parties was soon overtaken by the 1988 presidential cam-
paign in which GOP nominee George H. W. Bush promised to maintain
U.S. military supremacy and portrayed his democratic opponent Michael
Dukakis as weak on defense. Throughout election season, the candidates car-
ried on as if no major thaw in Cold War tensions had occurred. As president,
Bush called upon the armed forces in two major military operations. First,
after a series of provocations by Manuel Noriega, the president dispatched
28,000 U.S. troops to Panama in December 1989 where they overpowered
local defense forces and took custody of the dictator within a week. The sec-
ond occasion was the 1991 Persian Gulf War. In August 1990, Iraqi forces
invaded and brutally occupied Kuwait. Appalled by this brazen and vicious
attack, Bush vowed to liberate the country. By the end of August, the admin-
istration had secured passage of U.N. resolutions condemning the invasion,
imposing a trade embargo, and prohibiting Iraqi oil exports. A subsequent
U.N. resolution set January 15, 1991 as a deadline for Iraqi leader Saddam
Hussein to withdraw his army from Kuwait or face war. After Hussein failed
to comply, coalition forces attacked on January 17. A 5-week bombing cam-
paign decimated enemy targets in Kuwait and Iraq. The ground phase began
on February 24 and lasted a mere 100 hours before Hussein surrendered.
Panama and Iraq showed that Bush was willing to use military force in pur-
suit of U.S. foreign policy objectives, thereby reinforcing the GOP’s image as
the party of military strength.
The other major national security development during Bush’s term was the
collapse of Communism in East Europe and the Soviet Union. The president
handled the transition to a post-Cold War world with great skill, but since
U.S. national security was never in obvious jeopardy, these developments had
little impact on the 1992 presidential campaign. Bush believed voters would
reward him for his foreign policy stewardship, but he found them far more
concerned about deteriorating economic conditions at home than the fall of
the Berlin Wall, the demise of communism, and victory in the Gulf War.
Throughout the 1992 campaign, Bill Clinton repeatedly charged that Bush
was out of touch with economic problems at home. As such, foreign policy
issues played little role in his campaign message. In office Clinton sought
to capitalize on a “peace dividend” by calling for defense spending cuts to
free up resources for domestic priorities. But, as always, foreign policy cri-
ses demanded presidential attention. Although U.S. national security was not
seriously threatened at any point during the decade, Clinton used military
force on a number of occasions, foremost in humanitarian and peacekeeping
efforts in Somalia and the Balkans and air and missile strikes against targets in
the Balkans, Iraq, Afghanistan, and the Sudan. In any case, because the Cold
War was over and no ground wars occurred on Clinton’s watch, national secu-
rity took a back seat to domestic policy throughout the 1990s.
Things changed dramatically after the terrorist attacks on September 11,
2001 killed nearly 3000 innocents on American soil. Osama bin Laden, the
leader of the al Qaeda terrorist organization based in Afghanistan, claimed
responsibility for the attacks. President George W. Bush promised swift ret-
ribution against al Qaeda and those who harbored its members. When the
Taliban movement refused to surrender the terrorist organization’s leaders,
and meet several additional demands, war quickly ensued. Bombing began
in early October. Six weeks later the Taliban were driven from power and
al Qaeda’s forces scattered along Afghanistan’s mountainous eastern border
with Pakistan.
The war in Afghanistan was the first battle in a broader “global war on
terror” in which the Bush administration promised a new approach to defend-
ing U.S. security. The full meaning of this became clear during the president’s
commencement address at West Point in June 2002 where he unveiled what
came to be known as the Bush Doctrine. During the speech the president
asserted the right of the United States to initiate preventive war: “We must
take the battle to the enemy, disrupt his plans, and confront the worst threats
before they emerge” (Ricks, 2006, p. 38). Discarding the decades old practice
of deterrence, Bush proposed the most aggressively militant foreign policy
in American history. Throughout the summer and fall top administration
officials conflated al Qaeda, the 9–11 terrorist attacks, and Saddam Hussein,
implying the Iraqi dictator posed an immediate threat to the United States.
The most frightening charge was that Hussein might furnish terrorists with
weapons of mass destruction to attack targets on American soil. In a speech
that October the president laid out the case for taking action: “America must
not ignore the threat gathering against us. Facing clear evidence of peril, we
cannot wait for the final proof, the smoking gun that could come in the form
of a mushroom cloud” (Ricks, 2006, p. 61). Although some diplomatic efforts
were undertaken, there seemed little doubt the administration was preparing
for war.
After efforts to peacefully resolve the crisis failed, combat began on March
20, 2003. A combination of highly effective precision bombing and a lighten-
ing quick ground invasion decimated Iraqi resistance in short order. In a May
1 televised address, Bush triumphantly—and prematurely—declared that
“Major combat operations in Iraq have ended. In the battle of Iraq, the United
States and our allies have prevailed” (Ricks, 2006, p. 145). As it turned out, the
real war was just beginning. A series of U.S. mistakes, including the failure to
develop an adequate postwar plan to secure the country; the de-Baathification
order in which some 35,000 Sunnis, many of whom had committed no crimes,
lost their jobs; the dissolution of the Iraqi army, which left hundreds of thou-
sands of armed Iraqi men unemployed and angry; and the decision to indefi-
nitely table the creation of an interim Iraqi government, ignited the insurgency.
From the summer of 2003 to the spring of 2007 the war went poorly. By the
time the president made the gutsy decision to back a “surge” strategy in early
2007 much blood and treasure had been squandered. During the 2008 election
Barack Obama attacked the Bush administration for its “misguided war” and
promised to bring it to an end. GOP rival John McCain defended his support
for the war, if not the initial war plan, and insisted the United States had to
finish the job. From 2003 to 2008, then, there was little ambiguity about which
party “owned” Iraq.
This completes my historical review. To reiterate, my purpose in this sec-
tion has been to provide a synopsis of major developments in the national
security domain in the post-World War II era. The fundamental issue driving
debate in this domain has centered on the role of military power. From the late
1940s to the mid-1960s bipartisan consensus backed the idea, codified initially
in Truman’s containment policy, that a strong military and the will to use it
were essential for protecting U.S. national security and geopolitical interests.
But when victory in Vietnam could not be achieved at an acceptable cost, elite
consensus began to unravel in the Democratic Party. For the first time, estab-
lishment politicians questioned whether hard power advanced or undermined
America’s foreign policy goals. After 1968, Vietnam became Nixon’s war and
the parties began to polarize on the conflict with Democrats becoming more
dovish and Republicans increasingly hawkish (Berinsky, 2009). Yet American
involvement in Vietnam ended in 1973 and normalization of relations with
China and détente with the Soviets belied claims that Nixon was an ardent
hawk.
If the Nixon–Kissinger–Ford commitment to détente temporarily clouded
the picture of where the parties stood on the hawk–dove dimension, Reagan’s
election clarified matters. Reagan’s first term mixed the largest peacetime
expansion of military power in American history, bellicose anti-Soviet rhetoric,
varying degrees of support for anticommunist forces across the globe, and the
commitment to expand nuclear weapons into space, leaving little doubt about
where the president and his party stood. And though Reagan used diplomacy
to reduce U.S. and Soviet tensions during his second term, the parties quickly
reverted to form when the 1988 election rolled around as Bush hit Dukakis
for being too soft on national security. The first and second Bush presidencies
included a major military operation in Panama, two wars with Iraq, a war in
Afghanistan, and a global war on terror. Coupled with consistent reluctance
by Democrats to deploy military force—especially boots on the ground—
throughout much of this period, party positions on this cleavage were plain to
see. In short, the military power question has been salient for much of the past
60 years and the parties have, generally speaking, taken opposing positions on
this dimension since the early 1980s.6
A DURABLE RACIAL CLEAVAGE?

Race represents another critical issue area in American politics, so readers may
wonder why this cleavage does not feature in my discussion. My answer is that
for the years covered by my subsequent empirical analyses (1988–2008) racial
issues have been less significant than divisions over government power, moral
standards, and military strength. To explain, with passage of the Civil Rights
Act in 1964 and the Voting Rights Act in 1965, Lyndon Johnson effectively took
the racial equality cleavage off the political agenda. Although civil rights fea-
tured prominently in some presidential elections from the 1940s through the
1960s (e.g., 1948, 1960, and 1964), going forward no major party contender
would seek to deny blacks equal rights. Although candidates such as Nixon or
Reagan might be perceived as indifferent or perhaps even hostile to the inter-
ests of blacks, neither they nor anyone else campaigned openly to turn back the
clock on race relations. Indeed, there is no longer any serious difference among
political elites (or citizens) as to whether races should be treated the same.
This is not to deny the extraordinary importance of race in the histori-
cal period covered above. In the 1930s, Roosevelt was very careful not to
6
The Vietnam War led some to question not only the efficacy of force, but the
purpose of U.S. foreign policy more broadly, especially in the post-Cold War era
when the issues of humanitarian intervention and nation building arose dur-
ing the 1990s (Halberstam, 2001). Although such questions were of paramount
importance to U.S. policymakers and the foreign policy elite, the discussion
probably eluded most in the mass public who continued, as always, to care pri-
marily about U.S. security (Page & Bouton, 2006).
offend southern sensibilities and did little to advance the cause of civil rights.
Thereafter, the civil rights movement pressured the national government to
move toward full racial equality. Truman, Kennedy, albeit reluctantly, and
Johnson, with great courage and conviction, employed a combination of
moral persuasion, executive orders, and legislative pressure to ultimately dis-
mantle the Jim Crow system of discrimination and segregation and secure the
franchise for blacks. Democrats paid a steep political price for doing the right
thing. Commenting on passage of the Voting Rights Act, Johnson despaired to
an aid: “I think we’ve just handed the South over to the Republican Party for
the rest of our lives” (Halberstam, 2001, p. 61). He was right. The GOP became
the party of racial conservatism in order to enhance its appeal among whites
in the south and elsewhere. Although it would be going too far to claim such
appeal was based primarily on racial considerations, inasmuch as the party’s
stance on social welfare, cultural, and national security issues attracted con-
servative support, race was an important factor (Abramowitz, 1994; Carmines
& Stimson, 1989). As the party grew in strength in the Sunbelt region and its
ranks of moderates thinned, Republican support rose among whites and all
but disappeared among blacks. Hence, with respect to party coalitions the
centrality of race is beyond dispute. Indeed, if we define cleavages in terms
of social groups rather than policy dimensions, as I do here, race is probably
the preeminent division in American politics (Abramson, Aldrich, & Rohde,
2009).
In terms of issues, racial controversies have sometimes generated divisive
public argument since the mid-1960s. For instance, conflict flared over forced
busing in the 1970s and affirmative action in the late 1970s and mid-1990s.
Periodic battles over the reauthorization of the Voting Rights Act and federal
efforts to ensure equal opportunity and fair treatment have arisen at various
intervals. Immigration has proven especially nettlesome in recent years. These
significant problems affect the lives of millions of people. But given the rela-
tively short lifespan of these issues, it seems fair to conclude that no durable
racial policy cleavage exists along the lines of strong versus limited govern-
ment, traditional standards of moral conduct, and the utility of military force.
Put otherwise, these three cleavages have impacted presidential politics and
the party system to a much greater extent than battles over affirmative action,
minority set asides, or, after 1965, voter registration.
Although overt racial conflict has not dominated the American political
scene since the 1960s, implicit racial issues have mattered a great deal. Racially
conservative politicians rarely make explicit antiblack appeals because doing
so violates the norm of racial egalitarianism endorsed by most whites, and
thus risks provoking backlash at the ballot box and, more speculatively,
damage to a party’s brand. However, public opinion on a number of osten-

sibly race neutral issues is race coded in practice, in part because media cov-
erage of these issues reinforces stereotypical images of blacks (Gilens, 1999;
Mendelberg, 2001). When politicians decry “welfare queens” or promise to
get tough on “inner city” criminals or lambaste programs such as “midnight
basketball,” racial antipathies are activated in the minds of whites, which in
turn undermine support for liberal policies. Insofar as politicians successfully
play the “race card,” racial resentment remains a potent and insidious force in
American politics.
Yet this is not the same thing as saying the question of what government
should do to help blacks constitutes a long-lasting and deep-seated policy
cleavage on par with limited government, traditional morality, and military
strength. Said otherwise, although race shapes party coalitions and affects
white opinion on some issues in rather subtle ways, concerns over the role
of government, the nature of moral standards, and the application of military
force have been more salient and persistent cleavages since the civil rights rev-
olution in the mid-1960s (cf. Sniderman & Piazza, 1993).
CONCLUSIONS
In this chapter I provided a historical sketch of the origin and evolution of the
bedrock cleavages in the economic welfare, cultural issues, and foreign policy
domains. For each, I suggested that particular controversies at given points
in time should be viewed as manifestations of a deeper underlying conflict.
Economic welfare issues, such as poor relief, unemployment insurance, Social
Security and Medicare, food stamps, and health care reform, reduce to phil-
osophical differences over what the government in Washington should do to
help those struggling in the market economy. In the cultural domain, issues
such as school prayer, abortion, gay marriage, pornography, and stem cell
research are linked to broader concerns about the preservation of traditional
norms of moral behavior. Finally, national security issues, such as U.S.–Soviet
relations, rapprochement with China, shooting wars in Korea, Vietnam, Iraq,
and Afghanistan, smaller military operations in places such as Grenada and
Panama, and missile strikes against terrorist camps and state sponsors of ter-
rorism, reflect the more basic issue of the role military strength should play in
advancing U.S. foreign policy objectives.
This chapter’s second takeaway point is that each policy cleavage has occu-
pied a central position on the national political agenda for a very long time.
Battles over the size and scope of the welfare state date back to the 1930s.
Concerns about moral decline and restoration have shaken American society
since the 1960s. Finally, the hard versus soft power tradeoff has structured for-
eign policy discourse since the Cold War began in the late 1940s. This is not to
say all other issues are irrelevant. Some disputes have generated passionate
debate at various moments in time, such as civil rights for African Americans
in the 1950s and 1960s or clashes on immigration over the past several years.
In my judgment, no other policy cleavage can match the visibility and dura-
bility of the three identified here.
The final point I stress is that each policy cleavage has become deeply
entwined with presidential politics and the party system. The major parties
and their standard bearers have taken contrasting positions on government
activism since the New Deal, on traditional morality since the rise of the coun-
terculture, and on the hawk–dove continuum since the early years of the
Reagan revolution. My reading of the historical record is that Democrats are
the party of activist government whereas Republicans stand for moral tradi-
tion and military strength. Public opinion research shows that citizens stereo-
type the parties and their presidential candidates precisely along these lines
(Campbell et al., 1960; Green, Palmquist, and Schickler, 2002; Hayes, 2005;
Lewis-Beck et al., 2008; Petrocik, 1996).
Of course, the historical record is more complicated than this so far as
a party’s presidential candidates do not always fall on the same point on a
given policy dimension. The economic welfare reform programs advanced by
Roosevelt and Johnson were far more comprehensive than anything proposed
by Clinton and Obama. Likewise, Eisenhower and Nixon did not contest the
legitimacy of the New Deal like Goldwater and Reagan. Be that as it may, it
seems safe to presume that when marking ballots voters rarely dwell on his-
torical comparisons within parties. When the time to decide arrives, the com-
parison that matters is where the nominees on the ballot before them fall on
a given cleavage. In 2008, the fact that Obama’s domestic policy agenda was
not as ambitious as Roosevelt’s surely meant little to most people. What mat-
tered is that a vote for Obama was a vote for more government than the coun-
try would get under McCain.
To conclude, I have argued that conflict in the primary issue areas on the
American political agenda reduces to more basic dimensions, that each dimen-
sion has been plain to see for decades, and that the parties and their lead-
ers have taken clear and consistent positions opposing one another on each
dimension. The genesis and evolution of policy cleavages have implications
for how citizens come to understand the political world and render electoral
choices. In the next two chapters, I define policy principles as they exist in the
belief systems of everyday people and explain how citizens acquire and sub-
sequently use them to inform their presidential votes.
CHAPTER 3
Policy Attitudes and Political Sophistication
The first step in developing my theory of policy voting is explaining what I

mean by policy attitudes and political sophistication—the key concepts that
animate my work. Scholars have identified three broad classes of policy atti-
tudes that people may hold: liberal–conservative attitudes, issue preferences,
and policy principles. Drawing on the Eagly and Chaiken framework (1993),
I define each of these as an evaluation of some political object that endures
to a greater or lesser degree. The type of object in question is what distin-
guishes them. After defining these concepts, I take up political sophistication
and assess its distribution in the mass public.
THE ATTITUDE CONCEPT

What is an attitude? Social psychologists have wrestled with this for a very
long time. Early in the field’s history, some saw attitudes as the dominant force
in the human mind. For instance, Allport (1935) defined an attitude as “A men-
tal and neural state of readiness, organized through experience, exerting a
directive or dynamic influence upon the individual’s response to all objects
and situations with which it is related” (p. 810). In this perspective, attitudes
are mental predispositions that systematically affect perceptions, judgments,
and behaviors related to the object in question. This conceptualization has been
called the “dispositional” model of attitudes (Campbell, 1963).
The dispositional view was accepted widely for many years, but when
systematic reviews of the interplay between attitudes and behavior uncov-
ered surprisingly little overlap, psychologists began rethinking it (Wicker,
1969). Some posited that attitudes are the epiphenomenal byproducts of the
behaviors they are supposed to explain (Bem, 1972). Others sought to identify
characteristics of attitudes or situations that promote tighter attitude behavior
congruence (Petty & Krosnick, 1995). More recently, another challenge has
emerged from those who see attitudes as temporary evaluations constructed
43
on the spot when situational demands compel perceivers to generate evalua-

tive responses to a given object (Schwarz & Bohner, 2001; Wilson & Hodges,
1992). This “constructionist” view holds that attitudes are not enduring dispo-
sitions, but rather they are temporary creations that arise as needed only to be
dispensed with when the need has passed (Gawronski, 2007).
Because the dispositional and constructionist views of attitudes have merit
under different circumstances, assertions that one approach is preferable to
the other seem misguided. Instead, a broader definition that can accommo-
date these ostensible rivals under a single conceptual rubric would be helpful.
Eagly and Chaiken (1993) provide such a definition. Building on comparable
approaches that center on object evaluation associations (Fazio, 1989; Zanna
& Rempel, 1988), they define an attitude as “a psychological tendency that is
expressed by evaluating a particular entity with some degree of favor or dis-
favor” (p. 1). In this light, an attitude can be seen as lying somewhere along a
disposition–construction continuum bounded by extremely stable evaluative
tendencies at one end and temporary evaluative constructions at the other,
what Converse (1970) has called the attitude–nonattitude continuum. When
a person thinks a lot about a given object, the result is a durable evaluative
tendency, a crystallized attitude, residing at the dispositional end of the con-
tinuum. In contrast, when someone evaluates an object for the first time, the
attitude must be constructed then and there. Temporarily constructed evalua-
tions are called uncrystallized attitudes or nonattitudes.
This definition has three attributes requiring elaboration: entity, evalua-
tion, and tendency. For starters, “entity” refers to discrete objects someone
can discern. In the language of social psychology, these are known as atti-
tude objects. These objects of attention are remarkably diverse. To take some
examples, they can range from the concrete (e.g., $23.42) to the abstract (e.g.,
inflation), from the individual (e.g., Barack Obama) to the collective (e.g., the
American public), from the familiar (e.g., Frank Sinatra) to the obscure (e.g.,
Acid Mothers Temple), and so on. Beyond people and things, attitude objects
can denote ideas, principles, and systems of thought. For instance, cultural
values (e.g., self-reliance), scientific theories (e.g., evolution), religious ideas
(e.g., the state of grace), and political philosophies (e.g., liberalism and conser-
vatism) can serve as perceptual targets.
Next, we have the “evaluation” component of the definition. An evaluation
represents a bottom line judgment about the desirability of an entity. These
summaries can be simple dichotomous judgments, such as the object is good
or bad, or more nuanced placements along a latent continuum in which the
entity can be seen in varying degrees of goodness or badness or relative to
some other object (Zanna & Rempel, 1988). Returning to the aforementioned
Policy Attitudes and Political Sophistication 45
examples, someone may dislike Barack Obama a bit, feel very positively
toward the ideas of grace and self-reliance, and find conservatism more appeal-
ing than liberalism. The key point, again, is that evaluations of the object are
expressed in summary terms.
Finally, we have the “tendency” attribute of the definition. To understand
what this means, it helps to think about how attitudes form. The process often
begins with conscious perception of some entity (implicit attitudes matter too,
but I do not consider these here). Upon detection, people decide how they feel
about it. The first time this happens, a mental residue is stored in long-term
memory. This “evaluative knowledge” predisposes the individual to react to
the entity in a similar fashion when it is encountered again. The more often
this process repeats, the stronger the object evaluation association becomes
(Fazio, 2007). For an attitude object that has been considered many times, the
net result is an enduring disposition to evaluate the entity in a consistent man-
ner. For other entities, typically obscure topics, no prior evaluative tendencies
exist. Instead, when the person perceives the object he or she will need to con-
struct a temporary judgment. If the object is beheld no more, this evaluative
knowledge fades quickly from memory. It is as if the attitude never existed.
And if per chance the object is encountered again after the attitude has disap-
peared, the object evaluation association must be constructed anew. In short,
the tendency attribute in the Eagly and Chaiken model accommodates both
attitudes as dispositions (or crystallized attitudes) and attitudes as construc-
tions (or uncrystallized attitudes).
To recap, an attitude represents a psychological tendency to evaluate a
given entity with some degree of favor or disfavor. The entity can refer to any-
thing perceived, global evaluations can be based on some absolute standard
or relative to another object, evaluations can range from extremely positive to
extremely negative, and the tendency can vary from immovable to nonexis-
tent. The Eagly and Chaiken definition has much value because it incorporates
attitudes as dispositions, attitudes as temporary constructions, and attitudes
lying between these extremes. As we shall see in the next section, this inclu-
sive definition applies readily to the study of policy attitudes.
POLICY ATTITUDES
I now use the Eagly and Chaiken framework to define liberal–conservative
attitudes, issue attitudes, and attitudes toward policy principles. Each con-
cept represents a psychological tendency to evaluate a given political entity
with some measure of liking or disliking. What distinguishes them is the
entity being evaluated. Standard terminology in political science uses the
terms “issue preferences” and “policy attitudes” interchangeably, but readers

should note I do not follow that practice here. Because my definition of lib-
eral–conservative attitudes, issue preferences, and policy principles ties back
to the umbrella definition of attitudes, I use the phrase “policy attitudes” as
a shorthand to denote all three. I begin with liberal–conservative orientations
before moving onto issue attitudes and policy principles.
Liberal–Conservative Ideology
The study of ideology is complicated for two reasons. First, scholars often con-
flate liberal–conservative attitudes and liberal–conservative ideology. The two
are not the same. To avoid confusion, I elaborate what liberal–conservative
ideology means before turning to liberal–conservative attitudes. The second
factor muddying the analysis of liberal–conservative attitudes lies in the great
variety of meanings people attach to ideological labels. Hence, care must be
taken in specifying what these terms signify to different perceivers. I do that
after ideology has been defined.
At the broadest level, ideologies are best described as cognitive networks
of attitudes and beliefs organized by superordinate principles in an area
of human activity such as religion, economics, or politics (Kerlinger, 1984).
In the political domain, ideology can be defined as “a particularly elabo-
rate, close-woven, and far-ranging structure of attitudes” organized by the
“liberal-conservative continuum—the ‘right’ and the ‘left’ of a political spec-
trum” (Campbell et al., 1960, pp. 192–193). Although other capstone principles
might bring some coherence to political attitudes, the liberal–conservative
continuum has monopolized scholarly attention because it represents the key
dimension dividing political elites (McCarty, Poole, & Rosenthal, 2006). This
continuum centers on the role the national government should play in the
economic, social, and political life of the country. Per Campbell et al. (1960):
“Differences between liberal and conservative tend to focus upon the degree
to which the government should assume interest, responsibility, and control
over these sectors of endeavor” (p. 194). By this account, citizens’ views on
various political controversies, their preferences on multiple issues, and their
evaluations of elected officials, political candidates, and the parties are encap-
sulated by points on the liberal–conservative continuum.
To provide context for assessing what ordinary citizens know about the
continuum, I elaborate what liberalism and conservatism mean to those versed
in the argot of ideology. To delineate liberal and conservative doctrines, I draw
on Kerlinger (1984) and McClosky and Zaller (1984). To begin, contemporary
liberal philosophy maintains that public officials should harness the power
of the federal government to improve conditions for people in general, and
the disadvantaged in particular, by reforming institutions that stifle human

potential and social progress. Operationally, this implies vigorous government
regulation of the free market and oversight of business behavior; a generous
welfare state; various forms of aid to oppressed, stigmatized, or otherwise
marginalized groups; and support for personal autonomy in matters of pri-
vate behavior. In the foreign policy arena, liberals believe that engagement
with the world community, international cooperation, and flexibility best
serve the national interest. Military force may be necessary in some situations,
but only if other mechanisms of conflict resolution have failed.
To turn to the right, conservative doctrine holds that a strong federal gov-
ernment threatens individual liberty, private property, and social stability.
Rather than looking to government for help, citizens should adhere to tried
and true values such as self-reliance, religious faith, the traditional nuclear
family, and respect for authority. In domestic politics, conservatism calls for
minimal federal intervention into the market, a limited welfare state, the
privatization of risk and reward, and commitment to orthodox conceptions
of morality. In international affairs, conservatives believe the United States
should pursue its national interest unilaterally, that it should be willing to
use military force if necessary to further these goals, and that cooperation and
negotiation are often ineffective tools when dealing with other nations.
For those immersed in the world of public affairs, such as politicians, polit-
ical activists, journalists, and scholars, liberalism and conservatism serve as
philosophical postures that organize attitudes, beliefs, and perceptions and
facilitate information processing and decision making. Writing in 1984, Enelow
and Hinich presented the idea this way:
Information on policy issues is frequently scant . . . with candidates being

described in general terms, such as “fiscal conservative” or “New Deal
liberal.” However, such terms do carry substantial information for the vot-
ers on the policy positions of the candidates. A New Deal liberal may be
expected to favor increases in spending on social welfare programs, oppose
tax cuts for the rich, favor cuts in defense spending, favor increased gov-
ernment regulation of business, favor measures to aid labor unions, and a
host of similar positions. Thus, political labels are convenient devices for
simplifying discussion of policy issues. (p. 38)
Political elites use ideological labels along these lines. But what about non-
elites whose busy lives leave little time for politics? Do they fit this profile?
Scholars have utilized multiple techniques to see if survey respondents
treat liberal and conservative labels as political shorthand. One common
approach examines whether citizens employ liberal–conservative frames of

reference when describing presidential candidates and political parties. Using
data from the 1956 National Election Study (NES) survey, Campbell and his
coauthors (1960) analyzed respondent comments to a series of open ended
questions about what they liked and disliked about the candidates and par-
ties in order to classify them into different “levels of conceptualization.” The
levels sort citizens based on the organizing devices they bring to bear on their
political evaluations. Respondents who explicitly used liberal or conserva-
tive language were placed in the “ideologue” category. These individuals
demonstrated at least a passing familiarity with the liberal–conservative con-
tinuum and an ability to place the candidates and parties along this dimen-
sion. Ideologues, in other words, use liberal or conservative cues with some
effectiveness.
To take a typical example, a respondent who noted “the Democratic Party
tends to favor socialized medicine” and said Republicans are “more middle-
of-the-road—more conservative. . . . They are not so subject to radical change”
was classified as an ideologue. Another respondent liked the Democratic
Party for “Nothing except it being a more liberal party, and I think of the
Republicans as being more conservative and interested in big business.” This
respondent offered an additional reason for disliking the Democrats: “I think
extravagance, primarily” (Campbell et al., 1960, pp. 228–229).
As these comments reveal, some in the top category were not deeply
conversant with ideological concepts. They did not necessarily employ lib-
eral–conservative terminology in rigorous fashion, but instead, knew a few
buzzwords associated with the labels. Reading quotes such as these, we sus-
pect that if Campbell and his colleagues had not used lax standards, almost no
one would have made it into the ideologue group. Overall, the authors placed
2.5% of the sample in the highest level of ideological thinking. Unimpressed,
they concluded that “Our failure to locate more than a trace of ‘ideological’
thinking in the protocols of our surveys emphasizes the general impoverish-
ment of political thought in a large proportion of the electorate” (Campbell
et al., 1960, p. 543).
Many political scientists disputed this unsettling conclusion. One line of
criticism held that The American Voter revealed more about the quiescent pol-
itics of the 1950s than about the “failings” of ordinary citizens. According to
Pomper (1972, p. 416) “the findings . . . show a low degree of ideology among
voters in 1956, but only because the 1956 election did not stimulate ideo-
logical feelings” (cf. Nie, Verba, & Petrocik, 1979). In an era when moderate
candidates such as Dwight Eisenhower and Adlai Stevenson stood atop the
party tickets; when the congressional wing of the Democratic Party contained
many southern conservatives and the GOP had a fair share of moderates and
a sprinkling of liberals; and when Lyndon Johnson and Sam Rayburn ran the
Senate and House, respectively, from the political center, this point rings true.
In a more ideologically charged climate, citizens might become more adept at
liberal–conservative thinking.
Following this intuition, Hans-Dieter Klingemann used 1968 NES data
to calculate the proportion of ideologues in the public. Given the Johnson–
Goldwater election in 1964, civil rights and Great Society legislation, escala-
tion and quagmire in Vietnam, and cultural change and social unrest at home,
the possibility that the number of ideologues rose seemed plausible. A rise
did occur, but a paltry one at that. The percentage of ideologues increased
from 2.5% in 1956 to 6% in 1968 (as cited in Converse, 1975, p. 102). Although
deployment of ideological labels was over twice as high by 1968, it remained
confined to a sliver of the electorate.
Given the time it takes for information in elite political discourse to diffuse
throughout the general public, it may have been asking too much to expect
ideological conceptualization to have increased substantially between 1956
and 1968. A much longer timeframe may be needed before major advances
in ideological thinking can be detected. Two long-term changes seem espe-
cially likely to have facilitated a rise in the share of ideologues. First, levels
of education have exploded over the past several decades. According to U.S.
census data, 7.7% of the public aged 25 and older held a 4-year college degree
in 1960 compared to 25.5% in 2000 (U.S. Census Bureau, 2011). Since education
promotes ideological thinking, we would expect the ideologues’ share of the
electorate to have risen over time.
In addition, the Democratic and Republican parties have polarized sharply
since the 1970s. The rightward shift of the GOP, the decimation of the Boll
Weevil Democrats, and the virtual disappearance of moderates from both
sides of the aisle have pushed the national parties further apart on the lib-
eral–conservative continuum than at any time since the Gilded Age (McCarty
et al., 2006). Given ideological polarization at the elite level, perhaps many
citizens followed suit by adopting liberal–conservative mindsets. In short, the
American public is more educated than ever before and the ideological cues
it receives from public officials have rarely been clearer. In conjunction, these
factors should have augmented ideological thinking well beyond the levels
observed during the Eisenhower administration.
Lewis-Beck and his colleagues (2008) used NES data to estimate the propor-
tion of ideologues in the public during the 2000 election season. As in previous
efforts, lenient criteria were used: “Generally speaking, we gave respondents
the benefit of the doubt when placing them in the levels of conceptualization.
The result is that many of our ideologues definitely do not possess a fully
articulated personal political philosophy” (p. 264). The authors report that
only 10.5% of the electorate made it into the ideologue category. Now, com-
pared to the 2.5% figure in 1956, 10.5% is demonstrably better. But although
this fourfold increase is notable in relative terms, it fails to impress in absolute
terms. What we see is a modest uptick rather than a fundamental change in
the character of the electorate. Dramatically higher levels of education cou-
pled with elite polarization over the past 40 plus years have coincided with
incremental growth in the portion of the electorate that actively thinks about
politics in liberal–conservative terms. To sum up, my first cut at the evidence
suggests few citizens use liberal–conservative principles to structure their
views about the candidates and parties. Even those in the most ideological
category lack a sophisticated understanding of what these terms mean. Recall
that the comment “the Democratic Party tends to favor socialized medicine”
earns classification as an ideologue.1
Though used by many scholars, the levels of conceptualization index
has not escaped criticism (Smith, 1989). Hence, it would be foolish to rely
1
The next level in the conceptualization index contains “near ideologues.”
Respondents in this category did not seem to explicitly recognize liberal–
conservative terminology but evaluated the candidates and parties using other
abstractions that hinted, however remotely, at the possibility of ideological
thought. To take one example from The American Voter, an Ohio respondent liked
Republicans because “they play up to individual rights, which is good. That’s
good—it makes a person feel more independent.” The respondent disliked the
GOP because “They believe in big industry, utilities, etc. They’ve passed a lot of
labor bills I don’t approve of.” Next, a New York respondent disliked Democrats
because “The Communists linked to Roosevelt and Truman. Corruption. Tax
scandals. I don’t like any of those things” and approved of Republicans because “I
also like the conservative element in the Republican Party” (Campbell et al., 1960,
pp. 230–232). Campbell and his coauthors found that 9% of the sample qualified as
near ideologues in 1956. Lewis-Beck and colleagues (2008) report a comparable fig-
ure for the 2000 electorate. Scholars who use the index often lump ideologues and
near ideologues into a single category. Reports put this figure in the 20–25% range
(Pierce & Hagner, 1982). This coding procedure is helpful for some purposes, but it
has had the unintended consequence of conveying the misleading impression that
20% or more of the public thinks about politics in liberal–conservative terms. This
is not so. Instead, 20% or so of the electorate has a few clues regarding the symbols
associated with the labels and applies these to the candidates and parties in very
simple ways. This is something, but it is definitely not a “particularly elaborate,
close-woven and far-ranging structure of attitudes.”
exclusively on this measure to estimate the prevalence of ideological thought

in the mass public. Because using ideological terminology to evaluate polit-
ical actors is more demanding than recognizing and understanding what
the terms mean, the public may know more than implied by the index. Said
otherwise, those who do not use liberal–conservative labels to describe the
candidates and parties may still understand what the terms mean. Perhaps
many of those who judge the candidates and parties based on more immedi-
ate concerns (e.g., the state of the economy) know enough about liberalism and
conservatism to structure some judgments in ideologically coherent ways.
This possibility was not addressed in The American Voter, but Converse
(1964) took it up in his belief systems essay by examining whether NES
respondents recognized and understood what ideological labels imply.
Using 1960 data, he found that 17% of the sample had a sense of what these
terms meant (e.g., conservatives are for the free market, liberals are for
change), 46% had a more superficial understanding (e.g., conservatives are
careful with money, liberals waste money), and 37% had no idea. As with
the levels of conceptualization index, charity was shown in granting credit
for correct responses. As Converse admits in a footnote, “it should probably
be mentioned that a teacher grading papers would be unlikely to give pass-
ing marks to . . . 10 percent of the total sample. We made an effort, however,
to be as generous as possible in our assignments” (p. 257). Thus, whether
the more generous 17% or stricter 10% figure is counted, the test shows
few people had a deep understanding of ideological labels. But perhaps,
the critic might object, the 1960 election, which pitted the centrist Nixon
against the centrist Kennedy, was the wrong place to search for ideological
comprehension.
Fair enough, but work by Bennett (1995) covering the 1980–1992 period
paints a similarly dim picture. Bennett used NES data to assess how respon-
dents performed when asked to define “liberal” and “conservative.” Survey
participants were told that “people have different things in mind when they
say that someone’s political views are liberal or conservative” and then asked
“what sorts of things do you have in mind when you say that someone’s polit-
ical views are liberal” and “conservative.” Respondents could provide up to
three answers per label for six total responses. Bennett reported 45% of the
samples provided zero or one passable answer whereas only 12% could fur-
nish five or six definitions. The mean number of correct responses across the
surveys was 1.93. As in 1960, the American public turned in a remarkably
lackluster performance.
A final line of research bears on ideological thinking. Both the levels of
conceptualization and the recognition and understanding tests demand
articulation of ideological terminology, and thus they may overlook the

propensity for people to think about politics in ideologically implicit ways.
Perhaps large strata in the electorate see politics in left–right terms but lack
the ability to say so in response to open ended questions. Though unable to
verbalize what liberalism and conservatism mean, they might rely on latent
liberal–conservative dispositions to organize political views. If so, we would
anticipate that their responses to questions on particular issues would be cor-
related (Converse, 1964).
Scores of analysts have examined this possibility over the years, but
found little evidence that issue preferences are structured by any such out-
look. Research shows that policy preferences in a given issue area do covary
to some extent, but correlations across areas are much weaker (Knight, 1985;
Layman & Carsey, 2002; Peffley & Hurwitz, 1985). To explain, opinion on
issues within the economic welfare domain (e.g., health care and government
spending) or the cultural issues domain (e.g., abortion and gay rights) corre-
late more strongly within than across domains (e.g., health care and abortion).
Results such as these imply multiple predispositions are at work rather than
a single liberal–conservative orientation (I say more about the dimensionality
of political opinion in Chapter 5). These findings imply that most people do
not organize political attitudes in left–right terms. Articulation difficulties are
not the problem.
It is time to bring my discussion on ideology to a close. Recall that lib-
eral–conservative ideologies are cognitive networks of political attitudes and
beliefs organized by the liberal–conservative continuum. My review shows
that few citizens think about politics this way. Most people know little about
what these terms mean. They do not develop abstract liberal or conservative
ideas to guide their thinking about the candidates, the parties, or politics more
generally. Nor do they implicitly utilize the left–right dimension to inform
their preferences on specific issues. Some 50 years after Converse, ideological
innocence remains the norm (Kinder, 2006).
Liberal–Conservative Attitudes
Although few people possess political ideologies, many more could hold
liberal–conservative attitudes. Surveys routinely show that when asked to
describe their political views on a standard liberal–conservative self-place-
ment scale, two-thirds or more of the public obliges. This begs the question of
what the labels mean to these respondents. Because the labels reflect neither
abstract political beliefs nor policy preference summaries, self-placements
must reflect something else. But what exactly?
I define liberal–conservative attitudes as psychological tendencies to eval-

uate liberal and/or conservative labels with some degree of favor or disfa-
vor.2 The key to the definition centers on the meaning people attribute to the
labels. In essence, these function as political shorthand for a plethora of objects
as perceived by different individuals. Generally speaking, these evaluations
reflect how individuals feel about visible social groups or political symbols
tied to the labels. For some, ideological evaluations are grounded in feelings
toward blacks. For others, evaluations center on various “special interests”
associated with each label, such as feminists, gays and lesbians, and environ-
mentalists on the left and big business and born again Christians on the right
(Conover & Feldman, 1981). In a similar vein, some people link the labels to
symbols such as “equality” or “social reform” or a narrowly defined referent
such as “taxes” or “affirmative action” (Edsall & Edsall, 1991; Jost, 2006; Sears
& Citrin, 1985).
For those who equate liberalism with helping minorities, their global
judgments about liberal and conservative labels are based on their feelings
toward minorities. For others, liberalism might connote atheism and conserv-
atism could symbolize religion. When evaluating political labels, these images
would spring to mind. And so on for different perceivers. Levitin and Miller
(1979) do a nice job summarizing this view: “many Americans use ideological
labels in ways that suggest only a partial understanding of the terms ‘liberal’
and ‘conservative’ and of their implications . . . those labels nevertheless have
political significance for these citizens” (p. 752).
This conceptualization is on the mark. It fits with the evidence reviewed
earlier and other work indicating that liberal–conservative self-placements are
rooted in feelings toward social groups and political symbols. When it comes
to liberalism and conservatism, what people have in mind is nothing like a
wide ranging set of political preferences held together by abstract liberal–
conservative principles. Yet that diminishes neither the motivational potency
nor behavioral consequences that liberal–conservative attitudes have for mil-
lions upon millions of ordinary citizens. Liberal–conservative attitudes have
real meaning, but it is tied to symbolic concerns and group identities rather
than explicit policy content.
2
Liberal–conservative attitudes are embedded within liberal–conservative ide-
ologies, but they are not the same thing. Hence, the application of the Eagly and
Chaiken framework to liberal–conservative attitudes dispenses with the need to
think about these in terms of organized preferences.
Issue Attitudes
Scholars have focused a lot of attention on standalone issue preferences,
which represent the second face of policy attitudes animating electoral behav-
ior research. I define an issue attitude as a psychological tendency to evaluate
a specific policy proposal or rival proposals with some degree of positive or
negative affect. A policy proposal denotes an idea about an explicit course of
action the federal government could take to address some identifiable prob-
lem. The issue may center on something unique to a given campaign, such
as the death penalty in 1988, health care reform in 1992 or 2008, or the par-
tial privatization of Social Security in 2004. Other issues recur over a series
of elections, such as abortion and tax cuts. In contrast to liberal–conservative
attitudes, which are based on perceptions of diverse entities that vary across
respondents, issue preferences are tied to the same stimuli.
During election season, presidential candidates bombard the voting public
with multiple policy proposals. The fluidity of the campaign further complicates
things for voters as new controversies emerge and old ones fall off the agenda.
Over the past half dozen or so elections candidate proposals have addressed
many topics, including abortion, business regulation, business taxes, climate
change, deficit reduction, energy policy, economic stimulus, estate taxes, fam-
ily leave, federal bailouts, gun control, health care reform, homeland security,
income taxes, investment taxes, Medicaid and Medicare, missile defense, pri-
vate school vouchers, education reform, Social Security, stem cell research, ter-
rorism, welfare reform, the war in Afghanistan, the war in Iraq, and so on.
Clearly, this is a lot to keep track of. The next chapter addresses questions about
the breadth and depth of issue attitudes that citizens hold.
POLICY PRINCIPLES
What They Are
For more than 50 years, liberal–conservative predispositions and issue prefer-
ences have dominated the study of voter choice. More recently, a third class
of policy attitudes has emerged as a focal point in the fields of political psy-
chology and public opinion, if not electoral behavior. These orientations have
gone by various names, including core principles, domain-specific principles,
and policy-related predispositions (Feldman, 1988; Hurwitz & Peffley, 1987;
Miller & Shanks, 1996). Despite some subtle differences, these terms describe
the same basic construct. I refer to this class of attitudes as policy principles.
A policy principle is a psychological tendency to evaluate with some
degree of favor or disfavor a general claim about the proper course of action
to follow in a given issue domain.3 Attitudes toward policy principles differ

from liberal–conservative and issue attitudes in the following ways. First,
policy principles are less diffuse than liberal–conservative evaluations. The
former revolve around overarching claims about what should be done in a
given policy area and the latter are tied to a range of discrete entities that var-
ies across individuals. Second, policy principles reference more abstract ideas
than preferences about a lone issue. Put otherwise, core principles stand above
the issues that comprise a given policy cluster.
I see limited government, traditional morality, and military strength as the
capstone principles operating within the respective areas. Two concerns arise
immediately. First, what, precisely, do these principles mean? Second, why
focus on these as opposed to other orientations? I define each principle first
before turning to the selection issue. To start with limited government, atti-
tudes toward this principle reflect evaluations of the claim that the national
government should use its power to provide some minimal level of security
from market risks, especially for those at the economic margins of society or
who otherwise have trouble fending for themselves and/or their loved ones
(Kinder & Sanders, 1996; Markus, 2001). Limited government represents an
abstract principle because it centers on a general claim about whether or not
federal power should be used in the economic welfare domain rather than on
behalf of any single group (e.g., the working poor) or through any identified
policy tool (e.g., the earned income tax credit). Those who favor activism back
strong government and those who oppose such efforts prefer limited govern-
ment. Of course, individuals take positions between these extremes as well.
Next, attitudes toward traditional morality reflect judgments about the
preservation of traditional moral standards in American society. The idea is
abstract because it focuses on basic standards of right and wrong rather than
on narrowly defined issues such as abortion, school prayer, and so on (Hunter,
1992). The claim that America must preserve long established ideas about moral
decency, through collective action if necessary, anchors the right wing of this
dimension. Traditionalists revere absolutist codes of moral conduct, believ-
ing them applicable in all circumstances and thus immutable; therefore, they
have little tolerance for alternative views. Progressives reject black and white
views of morality. For them, ideas about right and wrong should respond to
changing times and be updated when necessary. Progressives take a relativ-
istic stance toward moral standards and reject the claim that traditional ways
3
My thinking has been influenced by Peffley and Hurwitz (1985; Hurwitz &
Peffley, 1987), Feldman (1988), Feldman and Zaller (1992), and Miller and Shanks
(1996).
require government sanction. And, of course, other people hold positions that
strike a balance between these competing views. Lastly, note that whereas tra-
ditional views are influenced by religious backgrounds, they are not reducible
to religious beliefs. Many who are not especially religious endorse traditional
values.
Military strength represents the final posture under consideration. Attitudes
toward this principle reflect the degree to which someone feels American
military power and the will to use it best serve American foreign policy and
national security objectives. Hawks believe military power is vital for defend-
ing and advancing U.S. geopolitical interests in the international arena. In
contrast, doves are cooler toward military strength because they see it as
less effective and more dangerous than the tools of statecraft (Mueller, 1973).
Between these extremes lie foreign policy moderates who believe a mix of
hard and soft power serves U.S. goals. Hawk and dove principles are broader
than views of hot and cold wars, conflicts with particular foreign adversaries,
specific weapons systems, defense spending, and other policy controversies
(e.g., United Nations dues).
Why These Three

I defined limited government, traditional morality, and military strength as
summary judgments about the proper course of action to take in the economic
welfare, cultural affairs, and foreign policy domains, respectively. My next task
is to explain why these serve as first principles in my model of voter choice.
That is, on what grounds should these three take precedence over alternative
principles such as economic individualism, political trust, authoritarianism,
or isolationism?
Two criteria have guided my selections. First, these principles merit close
attention because, as my historical overview in Chapter 2 revealed, they
encapsulate the major policy cleavages that have animated public discourse
for decades on end. Since the advent of the New Deal, the question of whether
the federal government is doing too little or too much to help citizens cope
with economic distress has manifested itself in multiple controversies, rang-
ing from work and poor relief in the 1930s to Truman’s proposed national
health insurance program in the late 1940s to the war on poverty in the 1960s
to the ambitious health care reform efforts in 1965, 1993–1994, and 2009–2010.
Next, battles over whether society should adhere to traditional moral stan-
dards have cleaved the body politic since the 1960s. Above culture war issues
such as abortion, pornography, school prayer, and the like stand overarching
concerns about cultural change and, depending on one’s view, moral progress
or decline. Lastly, in the foreign policy and security arena, the problem of how
much weight should be attached to military strength has structured public
discourse since the late 1940s. Here, too, issues tied to various adversaries,
whether nuclear armed superpowers such as the Soviet Union and China or
other threats such as the National Liberation Front (NLF) in South Vietnam
and Saddam Hussein in Iraq, often come down to the relative merits of hard
versus soft power. Put simply, divisions over the role of government in the
economic welfare domain, what set of moral standards should guide behavior
in American society, and the projection of American military power abroad
have been pervasive in American politics.
Importantly, these cleavages have taken on partisan hues for nearly as
long. As we have seen, presidential candidates and the national parties have
adopted clear positions on these dimensions for many years. From the 1930s to
the present, Democrats have backed strong government and the Republicans
have not. Since the 1970s, Republicans have proven more conservative than
Democrats on moral standards. Partisan divisions on the hawk–dove dimen-
sion started emerging after Nixon assumed responsibility for the Vietnam War
in 1969, but the 1973 Paris Peace Accords, rapprochement with China, and
détente with the Soviets clouded the picture for a time. When Ronald Reagan
won in 1980, party positions on this divide crystallized. At the presidential
level, then, the Democratic Party has generally fallen to the left of the GOP on
limited government, moral standards, and military power.
The second reason I analyze one posture per issue area centers on the psy-
chological principle of cognitive efficiency. Although other general attitudes
and beliefs might guide political judgment in the three domains, my emphasis
on the aforementioned principles comports well with abundant evidence that
citizens are cognitive misers whose innate ability and motivation to process
political information are greatly constrained (Lavine, Johnston, & Steenbergen,
2012; Sniderman et al., 1991). Given this, it seems reasonable to focus on the
principles especially likely to matter in the various policy domains (cf. Peffley
& Hurwitz, 1985). Of course, this presumes a strong case can be made for
identifying the germane principles. To the extent my historical synopsis is
on the mark, limited government, moral traditionalism, and military power
are eminently reasonable focal points. Moreover, empirical research suggests
these principles exert more consistent and powerful effects on some political
judgments than other plausible claimants, such as economic individualism or
free market capitalism in the social welfare realm, authoritarianism in the cul-
tural issues domain, and isolationism in foreign policy (Feldman, 1988; Goren,
2004; Hetherington & Weiler, 2009; Hurwitz & Peffley, 1987). To sum up, the
enduring visibility of the three great policy conflicts, the manner in which they
have been incorporated into presidential and party images, and the inherent
cognitive limitations that confront all human beings underlie my claim that
limited government, traditional moralism, and military strength function as
central heuristics in mass belief systems. I adduce a good deal of empirical
support for this claim in Chapters 5 through 8.
POLITICAL SOPHISTICATION
Having defined policy principles and distinguished them from liberal–
conservative and issue attitudes, I turn now to political sophistication. This
variable represents a critical moderator in the study of political choice and
it features heavily throughout the book. The importance of sophistication
was established in Converse’s seminal 1964 essay. Despite the theoretical and
empirical force of his argument, investigations as to how and why sophistica-
tion affects political judgment were largely neglected until the publication of
three important books in the 1990s.4 These were Reasoning and Choice by Paul
Sniderman, Richard Brody, and Philip Tetlock in 1991, The Nature and Origins
of Mass Opinion by John Zaller in 1992, and What Americans Know about Politics
and Why It Matters by Michael Delli Carpini and Scott Keeter in 1996. Because
of these works, the “sophistication interaction” model of political judgment
is now standard fare in the fields of political psychology, public opinion, and
electoral behavior.
Following prior work, I see political sophistication as a combination of
factual and associational political knowledge stored in long-term memory
(Converse, 1964; Luskin, 1987; Neuman, 1986). Factual knowledge refers to
correct information about national politics, including leading public offi-
cials, the parties, political institutions, government processes, and domestic
and foreign policy issues. Politically sophisticated individuals have many
discrete political facts at their disposal whereas the unsophisticated do not.
Associational knowledge denotes the organization of these bits of data into
broader knowledge structures.5 These structures help people manipulate new
4
Stimson (1975), Chong, McClosky, and Zaller (1983), and Neuman (1986) repre-
sent notable exceptions. These efforts aside, most work that engaged Converse
sought to refute his claims by arguing that the public as a whole was more
sophisticated than he allowed, thus sidestepping questions about the effects of
sophistication variance.
5
Note that associational knowledge structures as described here are distinct from
liberal–conservative ideologies as described earlier in the chapter. Knowledge
structures refer to connections between political facts whereas ideological struc-
tures denote connections between political attitudes and beliefs.
political information in an efficient and effective manner. That is, among the
politically sophisticated incoming information can be understood, stored,
integrated, and retrieved without much trouble. In contrast, the unsophisti-
cated experience great difficulty processing whatever information they stum-
ble across. Moreover, they cannot readily store nor recall the information to
which they have been exposed (Fiske, Lau, & Smith, 1990; Lau & Redlawsk,
2006). A final point. Political sophistication is best seen as varying along an
underlying continuum. People are politically knowledgeable or ignorant to
a greater or lesser degree. To put it another way, sophistication differences
reflect differences in degree rather than differences in kind. This point should
be kept in mind when I discuss the behavior of sophisticated and unsophisti-
cated groups—and experts and novices—later on in the book.
For many analysts, political sophistication is the lynchpin of democratic
citizenship. In light of this, the obvious question is just how sophisticated is
the American public. Regardless of how the concept has been measured, the
answer is always the same. Most people are not terribly informed about gov-
ernment and politics. Just as importantly, there exists considerable variation in
what they know. Some are deeply informed about politics, others know little
or nothing, and others fall at points between (Delli Carpini & Keeter, 1996).
This point can be illustrated using public opinion data from the 2004 NES
survey that contains factual items covering a range of political topics. To take
some examples, one question asked citizens to place the Republican Party on
a seven-point scale ranging from extremely liberal to extremely conservative;
66% put the GOP in the conservative end of the scale. Although very easy by
the standards of political elites, it stumped a third of the public during the
heat of an ideologically charged election. And this is as good as it gets. When
asked which party had the most members in the Senate before the election,
51% of the sample correctly said the Republicans. Of course, if everyone were
guessing randomly we would arrive at the same figure. To take one last exam-
ple, a question asked whether George W. Bush felt it was more important to
6
To many readers of this book, these are laughably easy questions. However,
the point to recognize is that the items do a fine job in discriminating between
knowledge-rich and knowledge-poor citizens. If respondents are given harder
questions most will fail to answer correctly, thereby leaving analysts with too lit-
tle variance to permit systematic study. My point is not to criticize the public, but
rather to emphasize the fact that when we talk about knowledge, respondents
who score high are not very well informed by the standards of those who prac-
tice, work, teach, or study politics for a living.
10
6
Percent
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Number of Correct Responses
Mean = 8.22, Stdev = 4.28
Source: 2004 NES
Figure 3.1. Distribution of Political Knowledge in the U.S. Public, 2004.
protect the environment or preserve jobs. Only 36% of the sample saw Bush as
conservative on the issue.6
To see what the distribution of knowledge in the American public looks
like, I use information items from the 2004 NES to create a knowledge scale
and plot the results. Each respondent’s score represents the number of correct
responses he or she gave to 15 questions. The histogram in Figure 3.1 shows
the distribution. Two points stand out. First, the public as a whole performs
rather poorly, as indicated by a mean score of 8.22 correct answers out of 15
(or 55%, a failing grade). Second, we see a lot of variation across the scale (the
standard deviation = 4.28 right answers). A sizable share of the public got
most questions right, many others turned in a middling performance, and
a nontrivial segment performed abysmally. To be specific, about 17% of the
sample provided three or fewer correct answers while another 21% got 13 to
15 answers right.
These are typical results, a point affirmed by looking at data from the 1992
NES. This instrument contains items asking about the job or political office
held by Dan Quayle and Tom Foley, which branch of government is respon-
sible for determining the constitutionality of a law, and additional items. The
sample performed best on the Quayle item with 88% identifying him as Vice
President. Beyond this, the public fared much worse. For instance, 58% knew
the responsibility for determining the constitutionality of a law rests with the
Supreme Court. Only 26% knew Foley was House Speaker at the time. The
mean number of correct responses on the 11-item scale was 5.67 (about 52%).
14
12
10
Percent
0
0 1 2 3 4 5 6 7 8 9 10 11
Number of Correct Responses
Mean = 5.67, Stdev = 3.07
Source: 1992 NES
Figure 3.2. Distribution of Political Knowledge in the U.S. Public, 1992.
Figure 3.2 shows the 1992 knowledge distribution. Once again, there is exten-
sive variation across the scale (standard deviation = 3.07). About 20% of the
sample did poorly, scoring two or less, while another 21% got nine to eleven
answers right. When it comes to political knowledge in the American public,
the mean is low and the variance is high (Converse, 1990). The conventional
view is that such variation has profound implications for political judgment
and voter competence.
CONCLUSIONS
This chapter has defined the key concepts upon which my theoretical edifice
rests. To review, I applied the Eagly and Chaiken definition of attitudes to
three classes of policy attitudes. Liberal–conservative attitudes represent eval-
uations of liberal–conservative labels, which are tied to politically relevant
symbols or social groups that vary across perceivers. Issue preferences cor-
respond to evaluations of specific government plans to address some readily
identifiable problem. Policy principles are evaluations of claims about what
should be done in the primary issue areas that define American politics. Policy
principles are similar to liberal–conservative attitudes and issues attitudes in
that they represent psychological tendencies to evaluate some aspect of public
policy favorably or unfavorably, but they differ with respect to the attitude
object at hand. Policy principles are more focused than liberal–conservative
attitudes and more abstract than concrete issues.
Limited government, traditional morality, and military strength represent

the foundational principles in the minds of ordinary citizens. These principles
reflect the overriding policy cleavages that have animated U.S. political debate
for the past several decades. Limited government refers to global judgments
about federal efforts to provide material security for the American people.
Traditional morality summarizes feelings about the standards of right and
wrong that should guide the public and private life of the country. Military
strength represents judgments about the extent to which hard or soft forms
of power should govern U.S. foreign and security policy. Finally, I defined
political sophistication as a combination of factual and associational political
knowledge stored in long-term memory. As I shall argue in the next chap-
ter, there are compelling theoretical reasons for positing that these principles
function as the most important type of policy attitude in mass belief systems,
and thus should guide voter choice for individuals across the sophistication
spectrum.
CHAPTER 4
The Three Faces of Policy Voting
Ordinary citizens vary widely in terms of what they know about public affairs.
The customary view among scholars is that such differences have profound
implications for understanding the quality of policy attitudes and how these
shape electoral choice. According to the sophistication interaction model,
sophisticated citizens hold a wider range of more fully crystallized, tightly
connected policy attitudes than the unsophisticated. This approach posits fur-
ther that ideological predispositions and issue preferences shape candidate
choice to a much greater degree among the politically aware. In conjunction,
these propositions sustain the claim that the sophisticated better approximate
the standards of political competence than the unsophisticated. The findings
from this perspective thus pose a puzzle: how can citizens who are too unso-
phisticated to vote based on liberal–conservative and issue attitudes come to
rely on policy principles?
My task in this chapter is to explain how and why policy principles guide
candidate choice for practically all voters. I do so by specifying a set of condi-
tions individuals must satisfy to policy vote and then applying this frame-
work to liberal–conservative attitudes, issue attitudes, and policy principles.
The three conditions are availability, meaning the policy attitude is stored in
voters’ memories; centrality, which means it functions as a critical heuristic
in political judgment; and position matching, which implies citizens figure
out which candidate lies closer to them on a given policy dimension and vote
for that candidate. To preview the argument, I posit that political sophistica-
tion enhances the likelihood that citizens meet these conditions for liberal–
conservative and issue attitudes. In contrast, both the politically aware and
unaware should prove capable of satisfying these conditions for the use of
limited government, traditional moralism, and military strength. In short, the
first two faces of policy voting are conditional on sophistication, but the third
face is not.
63
Before proceeding, readers should note that policy voting, as I use the
phrase, stands for choosing based on liberal–conservative attitudes, issue atti-
tudes, and policy principles. That is, policy voting covers the three classes of
policy attitudes rather than applying to discrete issues alone.
THE CONDITIONS OF POLICY VOTING

The Classic Perspective
When it comes to electoral behavior, the obvious place to begin is The American
Voter. Campbell, Converse, Miller, and Stokes (1960, pp. 170–171) delineated
three conditions for issue voting, but these requirements apply readily to all
types of policy attitudes. The first condition entails that voters have a genuine
attitude on a given issue, that it “be cognized in some form.” In other words, a
voter must know that controversy exists on some topic and must have given it
enough thought to form an evaluation. This evaluative knowledge can then be
stored in long-term memory, ready to be recalled in the future. When this con-
dition is satisfied, we say the policy attitude is available in memory (Aldrich
et al., 1989; Fazio, 2007).
The second condition for policy voting requires that the issue “arouse some
minimal intensity of feeling” in the minds of voters. Familiarity with the con-
troversy is not enough to stimulate policy-based choice; instead, feelings about
the issue must be sufficiently powerful to move a voter to action. This prop-
osition appears reasonable on its face, but further reflection raises questions
about what, exactly, it means in psychological terms. The authors characterize
it variously as personal importance,1 attitude extremity,2 and value expres-
sion.3 Because these represent distinct properties of attitudes, the ambigui-
ties inherent in the second condition are transparent. Because this condition
conflates strength-related attributes, it lacks sufficient theoretical precision to
serve as a general condition for policy voting. Partly for this reason, I replace
the intensity condition with centrality as elaborated below.
Third, a given policy evaluation “must be accompanied by some percep-
tion that one party represents the person’s own position better than do the
1
“The second condition requires that there be some sense of the importance of an
issue” (Campbell et al., 1960, p. 170).
2
“[A]mong persons who are familiar with an issue, what variation can be
observed in a quantitative measure of intensity, extremity, or strength of
opinion?”(Campbell et al., 1960, p. 177).
3
“For a given problem of government, intensity of opinion will depend on the
importance of the values involved” (Campbell et al., 1960, p. 177).
The Three Faces of Policy Voting 65
other parties.” If voters hold a real attitude on some policy and correctly per-
ceive where the Democratic and Republican nominees stand, they can deter-
mine who falls nearer to them on this dimension. Armed with this knowledge,
voters can be ready to back that candidate. Note my emphasis on “real” atti-
tude. Policy voting occurs only when the person holds a durable evaluative
disposition about the subject, recognizes where the candidates stand, and
chooses accordingly. People who take a position on an issue may perceive
that their favored candidate is closer to them than the candidate actually is, or
may adjust their position to fit their perception of the favored candidate’s pos-
ition. In neither case can we conclude that position matching has transpired.
Instead, the “match” results from voter projection or candidate persuasion.
To reiterate, Campbell et al. described three conditions citizens must sat-
isfy for policy voting to take place. People must hold real attitudes on some
controversy, have strong feelings about it, and see which candidate is in closer
proximity to their own views. If these requirements are met, policy voting
should happen. In what follows, I elaborate the first and third conditions,
which I denote as availability and position matching, respectively, and substi-
tute a second condition which, per Converse (1964, 1970), I call centrality.
Attitude Availability
The first policy voting requirement is that a policy evaluation be available in
the person’s belief system. Availability means the attitude is encoded some-
where in long-term memory—“cognized in some form”—and hence that it
has the potential, once activated, to shape electoral choice (Higgins & King,
1981). The obvious question at this juncture is how do policy evaluations wind
up in memory in the first place?
Social scientists have developed an array of complementary models
accounting for the acquisition and maintenance of political attitudes. The
essential points from these models can be condensed into a four-step process.4
Policy attitudes, again, defined broadly to include liberal–conservative orienta-
tions, issue attitudes, and policy principles, come into being when people are
exposed to information about some policy entity, comprehend what they have
been exposed to, render a bottom line evaluation about the object, and store the
impression in their memory banks. Let us consider each step in the process.
Exposure occurs when someone spends time, however limited, contem-
plating some entity. Exposure cannot be reduced to mere perception. It is not
4
My perspective draws heavily on Converse (1964), Chong et al. (1983), and
Feldman and Zaller (1992; Feldman 1988; Zaller 1992).
enough for people to notice something in passing and then immediately forget
about it, nor to perceive it at a subconscious level. Rather, the individual must
be aware of the stimulus and mull it over a bit. The degree of engagement with
the entity need not be extensive; indeed, it can be quite brief. But there must be
a conscious recognition of the object. Of course, the more salient a policy entity
is in public debate, the greater the likelihood an individual will come across it
at some point. Ongoing controversies are especially likely to attract notice.
Next, comprehension means the information received is readily under-
stood. For a stimulus requiring neither elaboration nor explanation, com-
prehension is instantaneous. Examples of readily understood topics include
a declaration of war, a race riot, or the death penalty. Because people know
what these phrases signify in a literal sense, they can be evaluated directly. For
obscure topics comprehension will be limited to those who possess the back-
ground knowledge needed to decipher the message. Some examples include
the Central America Free Trade Agreement (CAFTA), capital gains tax cut pro-
posals, or cap and trade pollution credits.
Exposure and comprehension represent the critical stages for developing
political attitudes, because evaluation and internalization usually happen once
the first two requirements have been met. To illustrate, consider the principle
of limited government. The question of how much the national government
should do for those in need has been a recurring theme in political discourse
since the 1930s. People are likely to come across this idea at some point early
in the political life cycle. Exposure may come from stories on the front page of
the local paper, hearing a political discussion between their parents or taking
part in one at school, or watching campaign advertisements during commer-
cial breaks from their favorite television shows. Because the meaning of this
idea is transparent, a person can form an impression and encode it in memory
without difficulty. With additional exposure to the idea, the attitude is evoked
anew, leading the person to think about government power some more. As
this is done, the principle evaluation association grows stronger. Over time, a
summary judgment about limited government becomes embedded within the
person’s belief system.
To wrap up, when a person perceives some policy entity, knows what it
means, renders a global evaluation of it, and stores the judgment in memory,
an attitude forms. Subsequent exposure to the entity reinforces the object eval-
uation association, pushing the attitude a bit closer to the disposition end of
the attitude–nonattitude continuum. The attitude now resides in the voter’s
mind, ready to guide information processing and decision making. I leave
until Chapter 7 the issue of what factors predispose citizens to render positive
or negative evaluations of the object.
Attitude Centrality
The second requirement for policy voting is that the attitude functions as a cen-
tral heuristic in someone’s mind. A number of available attitudes might con-
ceivably influence the vote, but it seems improbable all will do so. Presumably,
some attitudes are more consequential than others, so they should manifest
stronger effects on political judgment. The question on hand is what charac-
teristics of different classes of attitudes render them powerful with respect to
political choice. This is where centrality comes into play.
Early work in the field of psychology presumed that attitudes guide
behavior related to the object across many situations. To invoke Allport
again (1935), “Attitudes determine for each individual what he will see
and hear, what he will think, and what he will do” (p. 806). Today, psy-
chologists recognize that the power attitudes wield over behavior is con-
tingent and have identified factors that moderate this relationship. These
include attitude accessibility, which refers to how quickly an object evalua-
tion association can be moved from long-term to working memory; attitude
importance, which denotes how much personal significance an individual
attaches to the matter; attitude certainty or the amount of confidence a per-
son has in his or her attitude; and attitude ambivalence or the degree to
which the object evokes positive and negative reactions (Visser, Bizer, &
Krosnick, 2006, pp. 3–4).
Strength-related properties tell us a lot about how, why, and under what
conditions a given attitude influences behavior. For instance, accessible atti-
tudes shape behavior more than inaccessible attitudes, personally important
concerns have larger effects than unimportant concerns, and so on (see Petty
& Krosnick, 1995, for an excellent treatment of this topic). However, my pri-
mary theoretical interest lies in examining how different types of policy evalu-
ations guide electoral choice vis-à-vis one another. Put otherwise, I am more
interested in the relative influence of liberal–conservative attitudes, issue pref-
erences, and policy principles on the vote than in how variation in strength-
related attributes for a single attitude conditions choice. Attitude centrality
represents the critical attribute on this score.
To see what centrality is and why it matters, I must first explain how polit-
ical attitudes are organized in the minds of citizens. Recall that an attitude,
an association between some discernible entity and an evaluation, is a dis-
crete mental construct. For example, evaluations of Barack Obama are psy-
chologically distinct from evaluations of African-Americans. So, too, are
feelings toward government health care reform and the minimum wage. But
attitudes do not exist in isolation. Instead, they are linked to one another in
broader associative networks known as attitude structures (Campbell et al.,

1960; Judd, Drake, Downing, & Krosnick, 1991).5 Attitudes connected in this
fashion are described as “constrained.” Converse (1964) defined constraint as
“the success we would have in predicting, given initial knowledge that an
individual holds a specified attitude, that he holds certain further ideas and
attitudes” (p. 207).
Constraint takes two forms. First, when two (or more) attitudes at the same
level of abstraction (e.g., two issue attitudes) are linked in a predictable way,
they are constrained horizontally. To illustrate, imagine someone favors a gov-
ernment plan to guarantee universal health care coverage. Knowing this, we
predict that they also favor raising the minimum wage. If our prediction is
confirmed, we conclude their attitudes are horizontally constrained. Next,
vertical constraint denotes a hierarchical link between a general attitude and a
specific attitude whereby the former shapes the latter. When someone evalu-
ates a particular entity their reaction is shaped by predispositions germane
to the topic at hand. Returning to our hypothetical subject, we might predict
their attitudes toward health care reform and the minimum wage derive from
a broader stance toward government activism. If such connections exist, we
say that their issue positions are vertically constrained by their disposition
toward government power.
Constraining agents can be described as central attitudes. I define centrality
as the extent to which an attitude vertically constrains other attitudes, beliefs,
and perceptions. The greater the number of attitudes constrained by a given
disposition, the greater its centrality in the structure. Although Converse con-
ceptualized centrality in terms of the total number of interattitudinal connec-
tions rather than vertical connections alone, my definition follows directly
from his. In contrast, attitudes that are shaped by broader evaluative disposi-
tions without, in turn, constraining other attitudes can be described as periph-
eral. Peripheral attitudes exercise comparatively little force over other idea
elements in the cognitive network.
So far, I have focused on links between pairs of policy attitudes. Of course,
attitude structures incorporate a much wider range of evaluations. These struc-
tures are usually conceptualized in hierarchical or top-down terms whereby
a small number of central attitudes vertically constrains a much larger set
of peripheral attitudes (Hurwitz & Peffley, 1987; Peffley & Hurwitz, 1985).
5
I am referring to interattitudinal structures, which are different from intraattitu-
dinal structures. The latter refer to the affective and cognitive bases that underlie
a single attitude.
Limited Trad Military

Govt morality strength
Health
Minimum School War in Defense
care Tax Cuts Abortion
wage prayer Iraq spending
reform
Figure 4.1. A Hierarchical Political Attitude Structure.
Said another way, a few long-term postures give rise to an array of short-term
evaluations such as new issues, presidential job performance assessments, and
so on. Figure 4.1 illustrates the logic. In this diagram, the three abstract prin-
ciples function as central attitudes that vertically constrain issue preferences
in each domain. For example, limited government shapes attitudes toward
health care reform, the minimum wage, and tax cuts without being recipro-
cally shaped by them. Likewise, moral traditionalism guides the positions citi-
zens take on abortion and school prayer, but these issues do not influence the
broader principle. This is a strong assumption. I test it in Chapter 6.
The next theoretical step requires explaining how centrality facilitates can-
didate choice. This can be accomplished by considering how people make
political judgments in general. I begin by noting, as many have before, that
social life is bewilderingly complex. When thinking about some matter or
weighing alternatives, cognitive misers utilize decision-making strategies to
reduce this complexity to a manageable level. Moreover, they do so in the
face of innate limitations on the cognitive and motivational resources they can
bring to bear on the task (Fiske & Taylor, 1991).
Heuristics help us make sensible choices under these constraints. Heuristics
are rules of thumb used to simplify judgmental tasks in ways that often yield
acceptable, though not necessarily optimal, decisions (Chaiken, 1987). Rather
than attending to all relevant bits of information that might conceivably matter
for the decision, people search out diagnostic cues that can act as effective sub-
stitutes for complete information. Heuristics thus serve as computational aids.
According to Sniderman et al. (1991):
Heuristics are judgmental shortcuts, efficient ways to organize and sim-

plify political choices, efficient in the double sense of requiring relatively
little information to execute, yet yielding dependable answers even to
complex problems of choice. . . . Insofar as they can be brought into play,
people can be knowledgeable in their reasoning about political choices
without necessarily possessing a large body of knowledge about politics.
(p. 19)
Central attitudes can be seen as a type of political heuristic. The rule of

thumb is simply to deduce preferences on a specific matter from the general
attitude. In lieu of considering all aspects of the problem, people take a posi-
tion consistent with broader predispositions relevant to the choice at hand.
There are two reasons why central attitudes can be expected to function as
heuristics. First, it is much easier to acquire and maintain a few basic attitudes
than to develop and preserve a collection of attitudes covering an extensive
set of political objects. Indeed, a few postures can go a long way toward guid-
ing the construction of a wide range of more concrete short-term evaluations.
Second, reliance on central attitudes can lead to the same decision that would
result from more effortful processing. Inasmuch as abstract predispositions
serve this purpose, ordinary citizens can make sensible choices without being
deeply informed about public affairs. Nevertheless, it is important to recog-
nize that heuristics sometimes lead decision makers astray. Better choices
might obtain through systematic processing. To conclude, by relying on cen-
tral policy attitudes citizens can make reasonably good decisions in a timely
manner, but the possibility of error is real.
Now to see why centrality is so important for electoral choice, consider the
volume of policy information a prospective voter confronts during a presiden-
tial campaign. To begin, the candidates take positions on many discrete issues
within and across policy domains. Sometimes they offer detailed blueprints for
what they plan to do. At other times their stances are more nuanced and harder
to pin down. The fact that these proposals may evolve over the campaign adds
to the burden of staying abreast of candidate positions. Next, there is ample
discussion about each nominee’s prior experience and record of public service.
Candidates selectively highlight past stances and attack their opponents for
being on the wrong side of some affair. Here, too, past and current positions
may be at odds. As voters contemplate their options, the prospect of information
overload is ever present. They cannot evaluate every policy datum they encoun-
ter, encode each one, and recall them all when comparing candidates. They need
to simplify the choice by employing cognitive strategies that facilitate decision
making efficiency. Moreover, to be handy these strategies must have a reason-
able likelihood of yielding a choice consistent with fully informed preferences.
The challenge, again, is to make a good choice with minimal effort.
This is where central policy attitudes come into play. Because people
already use these to construct other judgments, they can employ these trust-
worthy guides in the area of candidate choice. Instead of attending to all pol-
icy proposals raised during the campaign or studying the candidates’ prior
records, voters use central attitudes as rough proxies to guide their selection.
For instance, if attitudes toward traditional standards of right and wrong
are more central than attitudes toward abortion, sex education, and school
prayer, the general principle should have a stronger effect on candidate eval-
uation than these issue attitudes. Indeed, the existence of general principles
dispenses with the need to think about most specific issues.
To sum up, the use of central policy attitudes is economical and likely to
produce the same choice that would result if a more comprehensive and delib-
erative approach were taken. In essence, central policy attitudes are akin to
partisan identities, transcending elections and candidates, ready to be acti-
vated during the campaign. They are probably not as powerful as partisan
identities in shaping electoral choice, but that does not diminish their impor-
tance. To the degree that central attitudes guide the construction of short-term
political judgments, they should prove consequential for voter decision mak-
ing. But for this to happen, one final hurdle must be overcome.
Position Matching
Per The American Voter, citizens must recognize which candidate will better
represent their position on a given policy dimension. This occurs when vot-
ers compare their position to those of the rival candidates and go for the one
whose position is more similar. This is called position matching. The key mat-
ter here concerns the mechanisms by which a voter learns candidate positions.
There are several channels through which a voter can acquire the requisite
information. First, the voter may know where the presidential hopefuls stand
at the outset of the campaign. For the most part, early learning is a function
of extant sophistication. Those who are politically knowledgeable in general
are likely to know (or quickly find out) where the nominees stand when elec-
tion season kicks off. Second, voters who lack this information at the start of
the campaign have numerous opportunities to acquire it from an information
environment flush with multiple overlapping cues about candidate positions
(Rahn et al., 1990). Key campaign events, hard and soft news in all sorts of
media, and ubiquitous campaign commercials all supply a steady stream of
redundant cues.
Likewise, party labels can fill in missing information about candidate posi-
tions. A voter can leverage prior knowledge about the parties to infer where the
candidates lie on the central dimensions. This knowledge may take the form
of correct perceptions of party positions, stereotypical beliefs about the party
and its coalition, or party issue handling reputations (Conover & Feldman,
1989; Green et al., 2002; Petrocik, 1996). Because party images and positions
hold steady over time, a voter can make some reasonable assumptions about
party stances and use these to fill in the blanks about candidate positions.
Finally, citizens may learn candidate views through informal channels.
Individuals are plugged into social networks that furnish some political infor-
mation during election season. Whether through political discussion or inci-
dental learning that occurs as a byproduct of social interactions at work, with
neighbors, with friends, or in religious or social organizations, policy cues can
reach those who plan to show up on Election Day (Huckfeldt & Sprague, 1995;
Popkin, 1994).
I use some examples to illustrate position matching. First, consider those
for whom liberal–conservative placements are an important heuristic. Upon
learning of the ideological leanings of the nominees, it is a simple matter for
political sophisticates to see which one lies closer to them and, other things
equal, throw their support to that candidate. Second, assume someone
favors traditional morality but does not know what the candidates believe.
If a voter knows the GOP is more committed to traditional moral values than
Democrats, he or she can deduce the same holds true for the respective nomi-
nees. Or perhaps this person stereotypes Republicans as the party of religious
orthodoxy. Here, too, this cue facilitates inferences about candidate positions
on moral standards. Finally, the voter may overhear a conversation at work
whereby one colleague is complaining to another that Republicans are too
narrow minded and intolerant on moral issues. Whether via direct knowl-
edge, campaign learning, party-based inference, social contact, or incidental
learning, voters can usually meet the position matching condition. Insofar as
they do, voting based on central policy attitudes should occur.
Tying It All Together

My discussion of policy voting has covered a lot of ground. Adopting a
condition-based approach, I postulated that policy voting occurs when
three requirements are met. First, a given policy attitude must be present in
long-term memory (the “availability” condition). By definition, nonexistent
attitudes cannot impact candidate choice. Second, discrete policy attitudes are
organized into hierarchical structures. Within these structures, central policy
attitudes vertically constrain peripheral attitudes. As these central heuristics
enable the construction of short-term evaluations, a voter can call upon them
when casting a presidential ballot. Therefore, given a set of available policy
attitudes, those occupying the central nodes within political belief systems
are better positioned to guide citizen choice than those at the periphery (the
“centrality” condition). Third, the voter must have a good idea of where the
candidates fall on a given policy continuum. They may learn this during the
campaign, use party schema to deduce candidate positions, or acquire it as
a byproduct of interactions within various social networks. An accurate per-
ception of candidate stances engenders policy-informed voting (the “position
matching” condition).
In the rest of this chapter, I examine the degree to which citizens can be
expected to satisfy these conditions for liberal–conservative attitudes, issue
preferences, and policy principles, paying especially close attention to the
role sophistication plays in the process. Because liberal–conservative voting
and issue voting have been studied extensively, my discussion of the first
two faces of policy voting is grounded in theoretical reasoning and empirical
evidence. In contrast, the lack of research on domain-specific principles and
electoral behavior means my discussion is guided by theoretical logic alone.
Miller and Shanks (1996) present a rare exception, but they do not analyze the
role sophistication plays.
LIBERAL–CONSERVATIVE ATTITUDES AND

VOTER CHOICE
The ideological thoughts and feelings of the American voter have preoc-
cupied electoral behavior analysts since the field’s inception. I now review
this body of work in light of my adapted model of policy voting, begin-
ning with the availability condition. Recall liberal–conservative attitudes are
conceptualized here as global evaluations of the social groups or political
symbols associated with ideological labels. They are not, as some presume,
worldview or policy position summations. Liberal and conservative atti-
tudes develop when citizens perceive ideological labels, attribute meaning
to them, assess their goodness or badness, and store the judgments in mem-
ory. Those who find liberal groups or symbols pleasing or dislike conserva-
tive imagery internalize proliberal sentiments. Conversely, those for whom
conservative images hold more appeal develop a positive stance toward the
conservative label.
This conceptualization bears a close resemblance to theories of motivated

social cognition advanced in the field of social psychology, most prominently
in the research program of John Jost and his colleagues (Jost, 2006; Jost, Glaser,
Kruglanski, & Sulloway, 2003). This approach postulates that liberal and
conservative attitudes reflect views about equality, hierarchy, tradition, and
social change as well as cognitive and motivational needs to reduce threat
and uncertainty. Citizens are drawn to ideological labels in part based on their
feelings about the symbols associated with equality and tradition and how
these resonate with liberal and conservative labels, as well as on how these
labels mesh with their basic needs and underlying psychological proclivities.
This conceptualization shares a close affinity with the one with which I am
operating, although it places greater emphasis on existential needs to cope
with danger in an uncertain environment and personality factors. Most crit-
ically, the motivated social cognition approach posits that citizens who are
more engaged, informed, and involved with public affairs are more likely to
find the label that better reflects their feelings about egalitarian and tradition
symbols and suits their psychological needs (Jost, Federico, & Napier, 2009).
Said otherwise, sophistication counts in both perspectives.
To elaborate, ideological evaluation occurs for those who pay enough
attention to politics to develop a sense, however limited, of what the labels
mean. Because the politically sophisticated track news about government and
public affairs, they are exposed regularly to the ideological frames of reference
in elite debate. When people encounter ideological labels for the first time the
political connotations are not obvious, but with subsequent contact they start
to attach some meaning to the labels. In contrast, the lexicon of liberalism and
conservatism remains elusive to those who typically shun elite debate. Given
this, liberal and conservative attitudes should be more prevalent in the minds
of the politically aware, a central tenet in the sophistication interaction model
of political judgment (Converse, 1964; Delli Carpini & Keeter, 1996; Sniderman
et al., 1991).
While the sophistication interaction model has many devotees, its les-
sons sometimes go unheeded. Let us see what happens when the sophistica-
tion variable is not taken into account in tests of attitude availability, vertical
constraint, and electoral choice. First, in opinion surveys many participants
describe their political views in liberal–conservative terms, suggesting the
labels mean something to those who adopt them. Since 1972, National Election
Study (NES) respondents have been asked the following:
We hear a lot of talk these days about liberals and conservatives. Here is a
seven-point scale on which the political views that people might hold are
arranged from extremely liberal to extremely conservative. Where would

you place yourself on this scale, or haven’t you thought much about this?
The response options are “extremely liberal,” “liberal,” “slightly liberal,”

“moderate, middle of the road,” “slightly conservative,” “conservative,” and
“extremely conservative.” A follow-up question probes whether those who
failed to take a position or chose the moderate option feel closer to liberals or
conservatives. Usually two-thirds to three-fourths of respondents locate them-
selves on the scale. The figure sometimes reaches or even exceeds 90% when
respondents who answered the probe are included. Some interpret these
results as prima facie evidence that liberal–conservative attitudes are availa-
ble in the minds of these respondents.
Studies on response stability also speak to availability. In his essay
Converse (1964) reasoned that if respondents hold real attitudes they should
give the same responses at different points in time to opinion items designed
to measure them. Applying this technique to the study of liberal–conservative
identities, researchers have found that responses to the self-placement item
are quite stable, suggesting many hold genuine liberal–conservative attitudes.
Studies report uncorrected continuity correlations around .60 over varying
time frames, less than estimates for party identification, which typically lie in
the neighborhood of .80, but greater than those for issues and candidate evalu-
ations (Converse & Markus, 1979; Jennings, 1992). Moreover, when structural
equation modeling techniques are used to disentangle attitude stability from
random measurement error, presumably induced by faulty item wording,
liberal–conservative identifications prove remarkably persistent (Alwin &
Krosnick, 1991).
Regarding centrality, some adduce evidence that most people ground issue
preferences in liberal–conservative attitudes. Using data from national and
state samples, this work shows that ideological self-placements predict opin-
ion on government spending, taxes, health insurance, busing, crime, and other
issues (Sears & Citrin, 1985; Sears, Hensler, & Speer, 1979; Sears, Lau, Tyler, &
Allen, 1980). Such results imply that liberal–conservative identities operate as
central heuristics for the lion’s share of the public.
Tallied up, works that neglect the sophistication variable find that liberal–
conservative predispositions are dispersed widely throughout the general
public and function as central heuristics in mass belief systems. But once
sophistication and related individual difference variables are incorporated
into the statistical analyses as moderators, problems emerge. This can be seen
through careful review of research that looks for sophistication effects. To
begin, the work on the levels of conceptualization and ideological knowledge
summarized in the last chapter implies that exposure to and comprehension

of liberal–conservative terminology are limited to a small subset of educated,
politically informed citizens. The sophistication interaction perspective fur-
ther suggests liberal and conservative views will be more stable in the minds
of political experts. Here the evidence also affirms the sophistication model.
For example, Delli Carpini and Keeter (1996) used NES panel data to show
that 70% or so of respondents in the top knowledge quartile gave the same
response on the seven-point liberal–conservative scale over a 2-year interval
compared to about 33% in the bottom quartile. So, although many respondents
take a position on the scale, a nontrivial portion of the least sophisticated evi-
dently offer top of the head responses signifying nothing. Given these results,
scholars should be wary of presuming that fairly high continuity correlations
from whole sample analyses indicate that most respondents have genuine
ideological attitudes.
Finally, we have the proposition that sophistication strengthens the con-
nections between ideological identities and issue preferences. The empirical
record again buttresses the sophistication interaction model. When sophis-
tication and related variables are ignored the self-placement term usually
manifests significant effects on citizen preferences, implying everybody
uses liberal or conservative cues. But in sophistication-stratified samples,
the evidence indicates that these orientations shape opinion only among the
politically aware (Zaller, 1992) and well educated (Sniderman et al., 1991).6
The record demonstrates that the sophistication interaction model provides
a highly persuasive account of ideological availability and centrality. The
methodological lesson should be clear. Research on the development and use
of policy attitudes that fails to consider the role sophistication plays can yield
misleading results.
The third policy voting condition demands voters match their ideologi-
cal proclivities to those espoused by the candidates. If potential voters know
where the nominees lie on the liberal–conservative spectrum, ideological
voting will likely follow. Some know this at the start of the campaign. Those
unsure of where to place the contenders can learn this during election season
or infer it from party schema. And though there may be some slippage in the
inferential process, the key is that the Democratic nominee should be seen to
the left of his or her GOP rival.
6
Authors sometimes employ education as a proxy for political sophistication,
an admittedly crude, though empirically defensible approach as the correlation
between education and knowledge scales typically lies in the neighborhood of .50.
The sophistication interaction model posits that political knowledge

facilitates ideological voting because the politically aware alone hold genu-
ine liberal–conservative attitudes, use these to guide opinion and judgment,
know where the candidates stand, and vote accordingly. Studies that do not
bring sophistication into the model often present evidence that the liberal–
conservative self-placement scale manifests statistically significant and sub-
stantively powerful effects on candidate evaluation and voter choice. Works in
this vein find that liberals vote Democratic and conservatives go Republican
(e.g., Levitin & Miller, 1979; Miller & Shanks 1996; Sears et al., 1980).
As the following two studies attest, the picture changes once sophistication
is taken into account. First, we have Knight’s (1985) thoughtful analysis that
used 1980 NES data to classify respondents into different levels of ideological
conceptualization to see if such variation moderated the association between
liberal–conservative identities and candidate evaluation. Of the sample 22%
made it into the highly sophisticated group, while the remaining 78% were dis-
tributed across less sophisticated categories. Next, she modeled the Reagan–
Carter feeling thermometer difference as a function of these identifications
and control variables. In the whole sample she found that stronger conserva-
tive identification predicts increasingly pro-Reagan evaluations (standardized
beta = .12). Efforts that ignore sophistication would stop here and conclude
erroneously that liberal–conservative attitudes matter for nearly everyone.
Instead of stopping, Knight estimated separate candidate evaluation models
across sophistication-stratified samples and found liberal–conservative iden-
tities significantly impact candidate evaluations only in the most sophisticated
group (standardized beta = .32 in the high sophistication group and about .09
in the low sophistication groups). Clearly, sophistication matters a great deal.
To take an example from my own work (Goren, 1997), I used data from the
1984 and 1988 NES surveys to model the presidential vote as a function of ideo-
logical proximity to the candidates, political knowledge, the corresponding
interaction term, and various controls. Like Knight, I found the more knowl-
edgeable the respondent, the greater the impact of liberal–conservative prox-
imity on the vote. In 1984, liberal–conservative voting was nonexistent among
political novices (those at the 25th percentile of knowledge). Among sophisti-
cates (those at the 75th percentile) movement from the most liberal to the most
conservative score increased the probability of a Reagan vote by .25. Other
studies, employing a variety of research designs, sophistication measures,
representative and convenience samples, and statistical models report similar
conclusions (e.g., Delli Carpini & Keeter, 1996; Lau & Redlawsk, 2006).
Overall, the evidence yields an unequivocal verdict. The politically sophis-
ticated alone meet the availability, centrality, and position-matching conditions
for liberal–conservative voting. By my count, perhaps 30–40% of the public

relies on these attitudes to inform Election Day choices. The flip side is that
liberal–conservative voting eludes the remaining 60–70%. As others have
noted, when it comes to the formation and use of liberal–conservative atti-
tudes, sophistication cannot be ignored. Doing so runs the risk of presenting
a deceptive picture about the ideological proclivities of the American public,
a point widely recognized by political and social psychologists (e.g., Lau &
Redlawsk, 2006; Jost et al., 2009; Sniderman et al., 1991).
ISSUE PREFERENCES AND VOTER CHOICE

I conceptualize issue attitudes as the degree to which proposals that address
specific problems, either in isolation or competition, are positively or neg-
atively evaluated. Scholars have searched long and hard for evidence of
issue-based choice. Because presidential candidates tender multiple policy
proposals during the campaign, it is natural to wonder whether such issues
matter to voters. For instance, in 2000 Al Gore and George W. Bush took much
different positions on what to do with federal budget surpluses. Gore argued
these funds should be used to shore up Social Security and Medicare whereas
Bush vowed to return some of the surplus to the American people via tax cuts.
During their first debate Gore promised to “put Medicare and Social Security
in a lockbox and protect them.” Bush countered “I want to send one-quarter
of the surplus back to the people who pay the bills. I want everybody who
pays taxes to have their tax rates cut” (Woolley & Peters, 2011). Perhaps voters
weigh proposals such as these.
If issue preferences are available in memory, operate as central heuristics,
and line up clearly with those of one candidate, issue voting can transpire. Do
voters meet these conditions? Let us consider availability first. Given the large
number of issues on offer during a presidential campaign, combined with the
cognitive and motivational constraints confronting voters, it seems implausible
that most people will have crystallized attitudes on most issues. The empirical
record backs this claim. Studies on issue availability have focused heavily on
response instability. The idea, again, is that issue dispositions should endure
over time in the minds of individuals, leading them to take consistent posi-
tions on an opinion scale at different dates. Converse (1964) relied on survey
data from the 1956–1958–1960 NES panel to assess attitude stability for eight
separate issues. For the 1958–1960 interval, seven of the eight continuity corre-
lations resided in the .30–.40 range—embarrassingly low figures. In contrast,
the continuity correlation for party identification exceeded .70, indicating that
some orientations are quite durable.
These results suggest that when it comes to issues, instability rules, and
hence that issue preferences fall near the constructionist end of the attitude–
nonattitude continuum. However, correlations between opinions in successive
panel waves cannot reveal whether change is systematic or random. If such
movement is systematic, the presumption that people hold genuine issue atti-
tudes may be defensible. If this is so, the correlations between adjacent panel
waves should exceed the correlations between the first and third waves. This
is not what Converse found. Instead, correlations between the 1956–1958 and
1958–1960 waves were similar to those for 1956–1960. These results debunked
the claim that observed opinion change reflects levelheaded responses to
evolving events. Instead, issue preferences fluctuate without much rhyme or
reason.
Converse probed further to determine what mix of respondents might
account for these curious patterns. The best fitting model partitioned the
sample into two groups. The first consisted of respondents whose attitudes
never changed, whereas the second, much larger group was composed of
subjects whose opinions varied randomly over time. This “black and white”
model provided a nearly perfect fit to the data on one of the eight issues—the
“power and housing” item. He found that less than 20% of the public held
“real and stable attitudes” on this issue; the remaining 80% either lacked such
preferences or held preferences that moved randomly over time (Converse,
1970, p. 176). Of the remaining seven items, the black and white model fit the
data reasonably well, though not quite as good as the power and housing
item. Evidently, many respondents constructed opinions on the spot because
no evaluative knowledge was available for recall from long-term memory.
Converse (1964) concluded famously “that large portions of an electorate do
not have meaningful beliefs, even on issues that have formed the basis for
intense political controversy among elites for substantial periods of time” (p.
245). This position is known as the “nonattitudes” thesis. Subsequent work on
response instability confirmed these were not anomalous findings.7
Given the dispiriting implications these findings have for citizen compe-
tence, many have challenged the nonattitudes thesis. Achen (1975) has articu-
lated the most prominent critique. He speculated that response instability may
be a function of inadequate questions rather than inadequate respondents.
7
Converse and Markus (1979) conducted similar analyses using data from the
1972–1974–1976 NES panel and found that classic guns and butter issue prefer-
ences were as wobbly in the 1970s as they were in the 1950s. The authors also
found party identification to be exceptionally stable and that new cultural issues
were moderately stable as well.
This interpretation, known as the “measurement error” critique, maintains

that citizens hold genuine issue attitudes but respond inconsistently to opin-
ion items at different times due to question flaws. For example, a respond-
ent whose latent evaluation of federal aid to education remains marginally
positive over time may offer inconsistent answers to an item whose response
options are “strongly agree,” “agree,” “neither agree nor disagree,” “dis-
agree,” and “strongly disagree” because of wording ambiguities. The “agree”
option connotes too much enthusiasm and the “neither agree nor disagree”
option too little. Which position should the respondent choose at a given point
in time? The answer is not obvious. The person might choose “agree” at time
one and “neither agree nor disagree” at time two. By switching from liberal to
centrist opinion, it appears that the person lacks a true attitude even though
his or her latent evaluation has not changed.
How is it possible to determine whether faulty respondents or questions
drive response instability? Achen reasoned that if respondents were to blame,
there should be less error variance in the opinions proffered by politically
engaged respondents. Using the same data as Converse, Achen modeled error
variance as a function of political interest, education, and several control vari-
ables. He found the responses of the engaged fluctuated as much as those
of the unengaged, implying that flawed questions drive response instability.
Achen then used statistical techniques to calculate corrected continuity cor-
relations, which in turn suggested that issue opinions are almost perfectly
stable. For example, the correlation for the power and housing item, the exem-
plar of Converse’s black and white model, averaged .96 across all waves of
the panel versus .37 for the uncorrected estimates. Moreover, the policy pref-
erence correlations were on par with the party identification estimates. Again,
these results implied issue preferences hew much closer to the dispositional
end of the attitude–nonattitude continuum, but deficient instruments convey
the erroneous impression that such evaluations are doorstep opinions.
More recent works provide evidence consistent with both views of response
instability. Achen sought to determine whether the amount of error variance
in issue opinions varied systematically across levels of education and interest
in politics, variables that at the time seemed like good proxies for political
sophistication. However, subsequent work has established that sophistica-
tion is best measured using factual political knowledge scales (see my dis-
cussion on the measurement of sophistication in Chapter 5). When Norpoth
and Lodge (1985) and Feldman (1989) modeled error variance as a function
of political knowledge, they discovered the variance was smaller among the
more informed. Although modest, these differences support the nonattitudes
thesis. Consistent with the measurement error critique, Krosnick and Berent
(1993) showed that response instability is partly attributable to flawed ques-

tions. These researchers identified a pair of problems with the standard seven-
point issue scales and found that a fully labeled branching format yields more
reliable responses than the partially labeled nonbranching format favored by
the NES. These results provided the first direct evidence that previous esti-
mates of response stability in issue opinion reports were downwardly biased
because of imperfectly worded items.
What can we make of these conflicting findings? How should we appor-
tion response instability between respondents and questions? Here is my
judgment. First, response instability arises in part because many respondents
worry little about politics. When asked to report issue opinions, the unsophis-
ticated bulk of the public often provides top of the head responses that flut-
ter somewhat unpredictably over time. Sophisticated individuals, in contrast,
offer more thoughtful responses that hold steadier. These findings lend some
support to the nonattitudes thesis. Second, faulty questions deserve a share of
the blame as well. When items are written using best practices, measurement
unreliability declines significantly. This evidence is consistent with the mea-
surement error interpretation. Hence, response instability arises because of
faulty respondents and faulty questions. Although this is an obvious answer,
the implications it has for assessing attitude stability have not been acted
upon. I say more about this in the next chapter.
To return, at last, to the attitude availability and centrality conditions, these
findings have obvious implications. Even when latent attitudes are measured
using best practices, the level of random error in opinion responses remains
quite high, suggesting that issue attitudes are not terribly durable. Likewise,
party identification proves more persistent than issue attitudes, irrespective of
how the constructs are measured and often by a wide margin, which further
underscores the flimsiness of issue attitudes. In light of this, I conclude that
most potential voters do not hold enduring evaluations on most issues. They
may hold strong preferences on one or two concerns but, generally speaking,
issue preferences are not terribly crystallized. Now, those who think regularly
about public affairs evaluate a wider range of issues than those who pay little
attention, but even here there are simply too many issues and the cognitive
constraints are too daunting for the sophisticated to hold crystallized attitudes
across the board. Given that issue preferences are often unavailable and those
that exist are at least partially uncrystallized, it follows they will not occupy
central positions in attitude hierarchies.
The third condition of policy voting calls for citizens to compare their
issue positions to those adopted by the candidates and vote for the better
match. Although many citizens fail to satisfy the availability and centrality
requirements, some make it past these hurdles. For whom should issues mat-
ter? Scholars have advanced two complementary theories. To begin with,
the sophistication interaction model applies once more. Because the politi-
cally aware think longer and harder about public affairs, they are more apt to
develop firm attitudes on at least some issues and compare their positions to
those espoused by the candidates (Delli Carpini & Keeter, 1996). Another the-
ory focuses on single issue voters (Krosnick, 1990). People who care intensely
about an issue attach a great degree of personal importance to their evalu-
ations. The personal importance of the issue motivates these voters to learn
what the candidates believe so the right one can be supported. Finally, some
work shows single issue voting is especially pronounced among knowledge-
able citizens (Anand & Krosnick, 2003).
This is not to say that issues never matter. Occasionally, an issue gener-
ates enough controversy to shape candidate choice throughout the electorate.
For instance, Abramowitz (1995) has shown that attitudes toward abortion
weighed heavily on voter decision making in 1992, a plausible finding in light
of Patrick Buchanan’s “culture war” speech at the GOP convention and Bill
Clinton’s vigorous defense of abortion rights. Nevertheless, the broader point
to take away is that issues are unlikely to manifest consistently strong effects
on electoral choice for most voters in most elections. Given the large number
of issues raised during presidential campaigns, the innate cognitive limita-
tions that burden voters’ processing capabilities, the peripheral status of issue
preferences in mass belief systems, and the difficulty in learning candidate
positions, issue voting is perforce limited.8
POLICY PRINCIPLES AND VOTER CHOICE

To date, most work on policy voting has examined liberal–conservative atti-
tudes and issue preferences. Since neither class of attitudes is widely availa-
ble or highly central in political belief systems, it comes as little surprise that
policy voting, so defined, is limited to politically knowledgeable and engaged
pockets within the electorate. From this vantage point, many conclude that
policy voting lies beyond the reach of everyday people. Some even go so far
as to belittle citizens: “The picture of uninformed voters in the election booth
8
Recent work by Ansolabehere et al. (2008) suggests that issue preferences are
stable for nearly everybody and issue voting is not conditional on sophistication.
Their critique of prior work is grounded in methodological rather than theoreti-
cal considerations. I address it in Chapter 5, which takes up measurement.
staring vainly at their shoes in search of cues to help in their vote decision is in
all likelihood not a hyperbole” (Neuman, 1986, p. 173).
Such pessimism is unwarranted. The possibility that American voters
incorporate an alternative class of policy evaluations into their electoral calcu-
lus must be explored before we conclude that policy voting lies beyond their
reach. This brings us to the third face of policy voting, selecting based on pol-
icy principles, which I defined as global evaluations about the proper course
of action to take in the major issue areas in the public sphere. As always, policy
attitudes should guide candidate choice when the standard conditions apply.
First, attitudes toward limited government, traditional morality, and military
strength must be available in memory. Second, each must function as a central
heuristic in political judgment. Third, voters must identify the candidate who
lies closer to them on each dimension. If these conditions are met, principle-
guided choice should ensue.
To begin with availability, citizens may hold real attitudes toward each
principle if they are exposed to information about the idea, comprehend and
evaluate it, and assimilate these feelings. Because the manifest content of each
principle is transparent, citizens do not need ongoing contact with elite debate
to learn what the claims entail. All the same, they need some exposure, and
because many care little about public affairs even minimal levels of exposure
cannot be assumed (Zaller, 1992).
As my earlier discussion has shown, the bulk of the electorate knows little or
nothing about liberalism and conservatism. Likewise, large slices of the elector-
ate remain ignorant about particular issues. This holds true for moderate to low
salience political battles, which is not terribly surprising. It also holds for high
profile issues such as the budget deficit, the Bush tax cuts, and the war in Iraq
(Bartels, 2008; Delli Carpini & Keeter, 1996; Kull, Ramsay, & Lewis, 2003/2004).
But as we have seen in the historical chapter, public discourse in the main issue
areas usually reduces to a more fundamental division. People do not need to
stay abreast of day-to-day developments in a given domain to acquire policy
principles. All that is necessary are occasional encounters with some aspect of the
major policy cleavages that animate public debate over the long haul (Feldman
& Zaller, 1992). This exposure accrues through a combination of secondary and
postsecondary education; the hullabaloo accompanying presidential campaigns;
hard news coverage of economic and political crises; coverage that spills over
into soft news outlets; viral sensations on the Internet; social interactions with
family, friends, colleagues, and other acquaintances; and incidental learning.
Through multiple channels, voters can encounter policy principles.
Given exposure, can we expect individuals to understand what these
principles mean? The critical point here is that the idea underlying each one
is easy to understand. Should the government in Washington do more for

those in need or does it do too much already? Is there an absolute standard
of right and wrong that holds for all circumstances or should moral views be
updated as times change? Should the United States rely on coercive military
force to ensure national security or will softer forms of power do? Men and
women can answer these questions directly. They need not be privy to the lat-
est health care reform proposals emerging from the House Ways and Means
Committee to appreciate the difference between limited and activist govern-
ment. Immersion in debates over neoconservative doctrine is unnecessary to
understand what force versus diplomacy means. Because the policy content
underlying each principle is self-evident, background knowledge is rendered
unnecessary.
Upon exposure and comprehension, evaluation and encoding follow eas-
ily. The politically sophisticated and unsophisticated alike can make sense of
each principle, formulate bottom line judgments, and store their impressions
away for future use. And when they encounter the idea later on, the associa-
tion between principle and evaluation grows tighter. In this manner, individu-
als come to hold real attitudes toward government activism, moral standards,
and American power.
Although maintaining that all citizens hold policy principles, I recognize
that sophistication probably affects their dispositional quality. Those pas-
sionate about public affairs contemplate principles more frequently than the
less interested; therefore, the former should hold more crystallized attitudes
toward each posture. Sophistication matters in some respects. But it does not
follow that because the sophisticated hold more durable attitudes, the unso-
phisticated lack meaningful attitudes. With the exception of those utterly
divorced from public life, nearly everyone should encounter these ideas at
different moments in time. And when they do, they evaluate the ideas and
store their feelings. The key point is that although core principles probably lie
a bit closer to the dispositional end of the attitude–nonattitude continuum for
the politically aware, the attitudes are meaningful for citizens at all levels of
sophistication.
Turning to centrality, I posit that policy principles serve as central heu-
ristics in the minds of nearly everyone. These principles help people render
sensible judgments while minimizing information-processing costs. Rather
than deliberating at length about the question at hand, the individual judges
the controversy in a manner indicative of broader principles. When someone
thinks about what government should do in a given instance, how it should
handle some crisis, or how an incumbent officeholder is handling the job, pol-
icy principles guide their reactions. To take a pair of examples, Peffley and
Hurwitz (1993) showed that hawk–dove predispositions shape evaluations of

issues such as defense spending and aid to anticommunist guerillas in Central
America. Likewise, McCann (1997) discovered that attitudes toward tra-
ditional moralism systematically affect abortion preferences. In both studies,
the reciprocal effects of policy preferences on broader predispositions were
negligible. Such evidence implies that domain-specific principles operate as
central heuristics in political attitude structures.
This is not to say that such influence entirely determines preferences, for
other factors matter as well. Nevertheless, the leverage of policy principles
is sizable for many such judgments. Because people have some experience
using principles, the application to electoral choice is straightforward. Core
principles are in place long before a presidential campaign begins, and thus
are well positioned to steer the vote. Rather than learning what it means to say
someone is liberal or conservative or assessing the candidates’ positions on a
dozen specific issues, the voter concentrates on first principles. In short, policy
principles should behave like central heuristics for expert and novice voters.
My final policy voting condition demands the individual recognize where
each candidate stands on a given dimension. With this information in hand, it
is possible to figure out which ballot to mark. The key is whether most voters
ascertain this information. There are convincing reasons for positing that they
do. To begin, the campaign generates a tremendous amount of useful data.
The candidates and their partisan allies frequently signal beliefs about what
should be done in the major issue clusters. There may be confusion over the
details of specific policies, but the main points come through clearly enough.
Democratic nominees favor a more active role for government in the eco-
nomic welfare domain whereas their GOP counterparts are more committed
to conventional morality and a tough foreign policy. Candidates broadcast
these postures during the party conventions, in campaign speeches, during
the presidential and vice presidential debates, in staged media events, and
in countless television advertisements. Daily coverage appears on cable tel-
evision, local television, radio, in print, and on the Internet, so much so that
even an intermittent glimpse at the morning headlines or the evening news
can prove informative. Beyond the campaign and traditional media, poten-
tial voters learn from soft news, such as morning and daytime talk shows.
Indeed, such outlets ramp up political coverage during presidential cam-
paigns, thereby reaching those who might otherwise miss the news (Barker,
2002; Baum, 2002).
In addition, individuals can acquire the requisite information by drawing
inferences from party schema (Conover & Feldman, 1989; Feldman & Conover,
1983). As we have seen in Chapter 2, policy principles are inextricably linked
to enduring party images. Given that the principle party linkages are few
in number and have long been visible, they are easy to learn and remember
(Carmines & Stimson, 1980). Very large shares of the electorate know that
Democrats “want to help the middle class and the poor” while Republicans
stand for “traditional morals” and “strong defense” (Green et al., 2002; Lewis-
Beck et al., 2008; Petrocik, 1996). As such, people can manipulate party labels to
make reasonable guesses about where the nominees stand on first principles.
Lastly, information about policy principles flows through multiple social
networks. Some might encounter cues as a byproduct of interacting with fam-
ily members, friends, neighbors, co-workers, fellow parishioners, and so on
(Huckfeldt & Sprague, 1995; Popkin, 1994). They may learn during a brief
exchange with a co-worker at the water cooler or on the shop floor, from a fel-
low power lifter or runner at the local gym, from another parent at their son’s
gymnastics practice or their daughter’s debate tournament, or from a friend
at the weekly pool hall gathering or bingo session at church. Even those who
pay relatively little attention to the campaign and have limited knowledge
stores likely receive some exposure to information about candidate positions
on general principles.
Overall, multiple sources saturate the information environment with rein-
forcing signals about where the candidates and parties stand on the fundamen-
tal policy cleavages that define American politics. Although no one receives
all campaign messages, most probably encounter a few from time to time from
different sources. And because much of this information is repeated over and
over, those who miss it at one point have plenty of chances to hear it later on
(Just et al., 1996; Rahn et al., 1990).
To reiterate, my argument maintains that nearly everyone holds genu-
ine attitudes toward limited government, moral traditionalism, and military
strength, thereby satisfying the availability condition. This is because both
political culture and campaign debate repeatedly highlight the fundamen-
tal policy cleavages, making it easy to absorb and evaluate these ideas. Next,
these principles function as central elements in political attitude hierarchies
where they vertically constrain short-term political evaluations, thus satisfy-
ing the centrality condition. Insofar as voters have experience calling upon
these heuristics to guide other short-term judgments, the ability to apply them
to electoral choice follows naturally. Finally, given an information-rich cam-
paign environment flush with repetitive cues, most of those who go to the
polls should have a decent idea of where the candidates stand on basic dimen-
sions, thereby satisfying the position matching condition. Since abstract pol-
icy principles are widely available, central, and linked to candidate positions,
citizens should rely on these heuristics when casting their presidential ballots.
In contrast to the first two faces of policy voting, restricted as they are to the
politically sophisticated, the third face of policy voting should elude relatively
few citizens.
CONCLUSIONS
For over half a century two research questions have dominated the study of
policy voting. First, to what extent do liberal–conservative orientations shape
electoral choice? Second, how much do issue preferences guide the vote? The
main finding from this body of work is that for the most part policy attitudes,
so defined, matter little. If we stop here, it would seem that policy voting
lies beyond the grasp of the unsophisticated bulk of the American elector-
ate. Before accepting this troubling conclusion, a systematic examination of
the interplay between policy principles, political sophistication, and candi-
date choice must be undertaken. Insofar as core principles are available, cen-
tral, and matched to candidate positions for citizens across the sophistication
divide, a third face of policy voting may exist. To date, no such examination
has been performed.
I hypothesize that practically all citizens hold genuine attitudes toward lim-
ited government, traditional morality, and military strength; that each oper-
ates as a central heuristic in their belief systems; and that the sophisticated
and unsophisticated alike are adept at matching positions to those held by the
candidates and grounding their votes in abstract principles. The remainder of
this book tests these claims. I begin in the next chapter by analyzing the degree
to which policy principles are present in the minds of political sophisticates
and novices. Chapter 6 examines attitude centrality and heuristic reasoning. I
then take up the origins of general principles in Chapter 7. Finally, Chapter 8
presents the vote choice analyses.
CHAPTER 5
The Availability of Policy Principles
Attitude availability represents the first condition of policy voting. A policy

attitude stored in long-term memory, once activated, has the potential to guide
candidate selection. The costs of acquiring and maintaining liberal–conserva-
tive orientations and issue preferences are sufficiently high to deter large por-
tions of the electorate from developing them. As the review in Chapter 4 shows,
a great deal of research supports this claim. In contrast, the costs of doing so
are low enough for policy principles for most people to acquire them. This is
because cues about policy principles are diffused widely throughout political
culture and campaign discourse, so much so that the bulk of the electorate
likely receives sufficient exposure to form summary judgments about these
ideas. Because the meaning of each principle is transparent, the evaluation,
storage, and retrieval processes can proceed without difficulty. Sophistication
confers no significant advantages in this regard.
This chapter examines whether attitudes toward limited government, tra-
ditional morality, and military strength are present in the belief systems of
cognitively heterogeneous populations. I begin by summarizing prior work
on attitude availability and then lay out the approach I take here. Next, I esti-
mate a series of measurement models, continuity correlations, and opinion
holding rates across sophistication stratified samples to see if everyone holds
core principles. Collectively, my analyses suggest that policy principles are
bedrock dispositions in the minds of all citizens, but are somewhat more crys-
tallized for the highly aware.
ASSESSING ATTITUDE AVAILABILITY

The Opinion Holding Method
How do we determine whether a political attitude is stored in memory?
This problem is hard to solve because attitudes are not directly observable.
Because an attitude cannot be seen, its existence must be inferred from some
89
observable indicator designed to reflect it. The issue then becomes what meas-
urement procedures should be used to establish that the attitude exists. A
number of studies have examined this matter with great care (Aldrich et al.,
1989; Campbell et al., 1960; Delli Carpini & Keeter, 1996; Lewis-Beck et al.,
2008). These efforts gauge availability via opinion-holding measures. When a
survey respondent answers a policy query, we call the response an opinion.
To illustrate, consider the following item that appears regularly on National
Election Study (NES) surveys.
There is much concern about the rapid rise in medical and hospital costs.
Some people feel there should be a government insurance plan which
would cover all medical and hospital expenses for everyone. Others feel
that all medical expenses should be paid by individuals, and through pri-
vate insurance plans like Blue Cross or other company paid plans. Where
would you place yourself on this scale, or haven’t you thought much
about this?
The response options lie along a seven-point scale with the endpoints labeled
“government insurance plan” or “private insurance plan.” When someone
chooses a point on the scale, the presumption is that the response reflects a
meaningful attitude. In contrast, when someone admits that he or she “hasn’t
thought much about this” or says “I don’t know,” analysts conclude that no
evaluation resides in memory, that the attitude is unavailable.
Although opinion holding seems like a reasonable way to explore attitude
availability, it has serious flaws. For starters, social desirability pressures com-
promise the technique’s utility. Respondents sometimes report opinions on
issues to which they have given little or no thought. They do so to convey
the impression that they care about the issue, that they are neither apathetic
nor uninformed. Although “pseudo-opinions” are prevalent on many obscure
and hard issues, they may come into play on standard political issues as well
(Bishop, Oldendick, Tuchfarber, & Bennett, 1980). Inasmuch as survey respon-
dents dissemble in this fashion, opinion-holding measures overstate the prev-
alence of policy attitudes.
To discourage pseudo-opinions, researchers and pollsters rely on “no opin-
ion” (NO) filters, such as the one above in the health care question. Common
sense suggests NO filters provide a legitimate out for respondents to admit
that they have no preference. Unfortunately, the filters do not work as adver-
tised. In a comprehensive review of their use, Krosnick (2002) has shown
that NO filters dissuade some who hold genuine attitudes from expressing
them. This may occur because respondents are ambivalent about the issue,
The Availability of Policy Principles 91
they aim to complete the interview as quickly as possible, or question word-

ing is somehow ambiguous. Moreover, experimental work shows that data
quality (i.e., reliability and construct validity) does not suffer when NO fil-
ters are removed from survey items, which raises additional concerns about
their use. It appears, then, that NO responses cannot be taken as prima facie
evidence that a given policy attitude does not exist. Because of this, opinion-
holding measures equipped with NO filters may understate levels of attitude
availability.
What can we conclude about these procedures? The method overstates
availability when subjects who hold no attitudes respond as if they do. NO
filters are used to mitigate this problem, but they overcompensate by screen-
ing out some who hold real attitudes, thereby underestimating availability. It
is impossible to disentangle the incidence of these two processes in a single
sample, let alone in sophistication-stratified samples in which overreporting
and underreporting surely vary as a function of respondents’ interest, knowl-
edge, and motivation. In light of this, opinion-holding measures should be
used cautiously, if at all, to assess attitude availability. Clearly, a better method
is needed.
The Measurement Model Method

To determine whether cognitively distinct groups hold authentic policy atti-
tudes, three requirements must be satisfied. First, multiple measures of each
latent attitude should be employed whenever possible. Given the fallibili-
ties of single items, more are preferable to less. Second, NO filters should be
avoided. This ensures that respondents answer almost every question, which
lets the analyst examine response quality empirically in lieu of assuming NO
responses reflect nonattitudes. Third, a statistical method that distinguishes
between systematic and on the fly answers is needed. Given these criteria,
confirmatory factor analysis (CFA) techniques fit the bill quite nicely.
With these techniques, it is posited that responses to a set of opinion
items are guided by a latent evaluative predisposition. If the attitude exists,
it should govern how subjects respond to survey queries designed to meas-
ure it. Statistically, this implies that answers to questions will covary signifi-
cantly, the observed covariances will depend on a latent factor, and a specified
measurement model linking latent and observed variables will accurately
reproduce the sample covariance matrix (assuming there are enough indica-
tors to estimate an overidentified model). In contrast, if the attitude does not
exist, opinions will reflect little more than temporary evaluations constructed
wily-nilly. Statistically, there would be little item covariation, no evidence of a
latent factor, and poor model fit. This example pertains to a single attitude that
presumably exists for everyone, but the approach generalizes to multiple atti-
tudes and groups. If a set of policy attitudes is present in the minds of groups
at various levels of sophistication, comparable factor structures and loadings
should emerge across the groups. When the evidence follows this pattern,
measurement invariance holds (Bollen, 1989). However, if policy principles
exist only in the minds of the politically aware, evidence of structure will be
found there and nowhere else.
To test for measurement invariance or equivalence, I trichotomize my
samples by political sophistication, specify a CFA model that links the latent
variables to multiple indicators, and determine whether the indicators tap
the latent factors to a similar degree across the groups. If so, the data should
reveal good model fit in the samples, which would imply analogous factor
structures. Likewise, similar factor loadings would suggest the unsophisti-
cated do as well as the sophisticated in translating unobservable attitudes into
observable opinions. I focus on standardized factor loadings that reflect the
correlation between latent factors and observed indicators because they have
an intuitively clear interpretation. Insofar as the model fit and item factor cor-
relations are similar across samples, I can infer measurement invariance holds.
If I find discrepant model fit or widely varying loadings, invariance is not pre-
sent. Before carrying out these tests, I describe my data and measures.
DATA AND MEASURES

Measuring Political Sophistication
In this chapter, I rely on data from the 1988–2008 NES presidential election year
surveys to measure the constructs and test for attitude availability.1 I begin with
the measurement of political sophistication, which represents a combination
of factual and associational knowledge about government and politics stored
in long-term memory. I measure this concept using political knowledge scales.
To elaborate, each respondent’s sophistication score equals the total number
of correct answers given to a series of factual questions (incorrect, don’t know,
and refusals are counted as wrong) plus the NES interviewer rating of each
respondent’s apparent level of information about politics (on a five-point scale
ranging from “very low” to “very high”). The latter item has been recoded to
1
The combination of a multistage area probability design, the use of the face-to-
face interviewing mode, and high response rates make NES samples the gold
standard in data quality for electoral behavior research. There are at least two
downsides to using these data. First, the questions often diverge from researcher
preferences. Second, necessary measures are sometimes unavailable.
lie on a 0–1 scale. The final knowledge index is standardized this way as well,
with 0 indicating no questions answered correctly and the lowest interviewer
rating and 1 signifying all correct answers plus the highest interviewer rating.
There are 11 to 17 items in these scales, depending on the year.
I use a mix of easy, middling, and hard items. Some ask interviewees
to place the presidential candidates and parties on the correct side of vari-
ous scales, such as liberal–conservative or left–right continua, government
spending, defense spending, and so on. To get credit, respondents must place
Democrats on the liberal side and Republicans in the conservative end. I also
employ items that ask about the offices or positions held by various public
figures.2 Lastly, questions on political institutions and procedures, such as
which party controls the House and Senate, are utilized when available (see
the Appendix for the items).
This strategy can be justified as follows. To begin, these scales tap the fac-
tual knowledge attribute of the sophistication concept in a highly discriminat-
ing manner. The ability to answer 9 or 10 questions correctly suggests far more
information holding than implied by these 10 questions. Similarly, someone
who cannot answer more than a couple would be hard pressed to do better
if given more opportunities. Hence, although the number of items used to
construct each scale is limited, they are diagnostic of vast differences in latent
knowledge (Converse, 2000). Next, studies show that items such as these
reflect a single latent trait, that additive scales based on such items are highly
reliable (the Cronbach α reliability coefficient exceeds .80 for every scale in
this study), and that knowledge scales are construct valid (Delli Carpini &
Keeter, 1993; Luskin, 1987; Neuman, 1986).
With respect to the NES interviewer rating of each respondent’s informa-
tion level, these judgments are made by professional interviewers with con-
siderable experience evaluating respondents in lengthy face-to-face sessions.
Although the interviewers cannot determine objectively how informed their
companions are, their ratings are shaped by respondents’ interest in and
engagement with the survey. Additionally, reliability and validity tests carried
out here and elsewhere demonstrate that this item correlates robustly with
2
I employ the office holder items for 1996 and before, but not thereafter. This is
because “don’t know” responses were not handled consistently in later surveys.
Some respondents were encouraged to take a guess after saying “don’t know”
and others were not. To be clear, I am referring to both the split ballot “don’t
know” wording experiments used in the 2000 NES and the 2004 and 2008 sur-
veys in which no such experiments were administered. In any case, alternative
coding schemes for knowledge do not alter the results in any significant way.
objective knowledge measures, thereby justifying its incorporation into the

index.
What about associational political knowledge? Can information scales
double as measures of political schema that facilitate information processing?
In my judgment, absolutely. People do not amass large stores of incidental
political facts and then fail to integrate them into broader knowledge struc-
tures. To put it another way, discrete bits of political knowledge do not exist in
isolation, but instead are linked in broader cognitive networks. Experimental
and observational studies show that those scoring high on political knowl-
edge scales are more skilled at classifying, storing, and retrieving new polit-
ical information compared to those who score low (Fiske, Kinder, & Larter,
1983; Price & Zaller, 1993). This is precisely what we would expect if these
scales tap knowledge schema. To sum up, political knowledge scales pro-
vide for valid and reliable measurement of the sophistication construct. As
Delli Carpini and Keeter (1993) put it: “A common conclusion in an increas-
ing number of studies is that factual knowledge is the best single indicator of
sophistication” (p. 1180).
Measuring Policy Principles

Multiple indicators are used for each principle. I start off with limited govern-
ment, which reflects the degree to which someone favors or opposes federal
efforts to ensure a basic level of material security for vulnerable groups such
as the elderly, the working class, the poor, the homeless, and so on. Ideally, the
items would inquire whether the federal government should do more or less
for such groups. Unfortunately, no such items are present. Instead, I tap limited
government using responses to questions about federal spending on various
social groups and programs. The “welfare,” “food stamps,” “poor people,”
“homeless,” “Social Security,” and “child care” items appear regularly across
the surveys and combinations of these, depending on what was asked each
year, are used to gauge the broader principle. The number of indicators varies
from four to six depending on the survey (see the Appendix).
For all years except 2008, the measures had three response options: increase
spending, keep it about the same, or decrease spending. Respondents who said
spending should be eliminated are recoded into the decreased category. In 2008,
those who selected increased or decreased were asked to report the intensity
of their preferences, yielding a seven-point scale for each item (i.e., spending
should be increased/decreased a great deal, a moderate amount, a little, or kept
the same). Reponses are coded so that higher scores denote antispending opin-
ion, and thus reflect a preference for limited government. The spending items
do not include NO filters, so “don’t know” responses are minimized. Remember
this is desirable for two reasons. First, NO filters screen out some who hold true
attitudes. Second, because most respondents answer all questions I can deter-
mine whether responses reflect true attitudes or random noise.
Although no single item can effectively tap the general principle, the key
point is that when multiple indicators are combined the cumulative nature of
the measure captures the principle rather nicely. Those who want government
to do more should favor spending increases across the board; those reject-
ing the claim should favor cuts (Jacoby, 1994). Is this assumption reasonable?
The items may seem deficient because they fail to ascertain whether subjects
believe the national government should do more to improve conditions for
these groups, asking instead only whether it should spend more. Moreover,
references to federal spending seem to come uncomfortably close to the types
of preferences limited government is supposed to explain. Although such con-
cerns cannot be mitigated entirely, the spending items possess a number of
attractive features that should alleviate these worries.
First, although the ideas of government “doing more” and “spending
more” differ semantically, respondents will probably react to them in a similar
way. When asked to evaluate federal spending on the poor, the homeless, wel-
fare, and such, the answers given likely depend in large part on feelings about
government assistance more generally. Next, because various subsets of the
population may benefit from government aid, items that evoke group centric
thinking are needed. The spending items satisfy this requirement handsomely.
Some do so explicitly (e.g., poor people, the homeless) and others implicitly.
In the latter category, the welfare items evoke images of welfare recipients, the
Social Security item calls to mind the elderly, and the child care item implies
help for working parents and single mothers. Third, because group deserv-
ingness varies systematically across these measures, the items can differen-
tiate between welfare state supporters and detractors. Citizens committed to
the idea that the federal government has a responsibility to do something for
the “have nots” in society should favor spending more on both sympathetic
(e.g., the elderly) and unsympathetic (e.g., people on welfare) groups. Those
who agree up to a point should favor aid for “deserving” groups alone. And
those who reject the idea should oppose spending across the board. Fourth,
evidence presented in Chapter 6 shows that the latent variable tapped by the
spending items dynamically constrains short-term preferences on a concrete
issue (i.e., health care) without being reciprocally shaped by these, which is
what we would expect if the spending items tap a general policy orientation
rather than a specific issue. Given these virtues, the spending items can serve
as valid and reliable (albeit imperfect) measures of attitudes toward govern-
ment activism.3
Next, I elaborate the measurement of attitudes toward traditional morality,

which reflect how people react to the proposition that traditional moral stan-
dards should be preserved and deviations from these norms discouraged if
not prohibited. This principle is tapped via four items. Respondents are asked
whether they agree strongly, agree somewhat, neither agree nor disagree, dis-
agree somewhat, or disagree strongly with the following statements. The first
reads “The world is always changing and we should adjust our view of moral
behavior to those changes.” The second asserts “We should be more tolerant
of people who choose to live according to their own moral standards even if
they are very different from our own.” Third, we have “This country would
have many fewer problems if there were more emphasis on traditional family
ties.” Finally, “The newer lifestyles are contributing to the breakdown of our
society.” The items lack a no opinion filter, so missing data are rare.
These items let me distinguish between moral traditionalists and pro-
gressives. First, the statements reference the tradition–progressive conflict
in various ways. Respondents who take an uncompromising view of moral
truth should disagree with the tolerance statements in the first two items
and endorse the traditionalist positions in the latter pair. Progressives should
evince opposing responses across the questions. Second, wording is abstract:
the items are tied to no specific groups or issues. Third, in Chapter 6 I demon-
strate that the latent variable captured by these items dynamically constrains
abortion evaluations, suggesting it behaves like a central heuristic. One draw-
back is that no statement asks directly about a single, indisputable moral truth.
Another shortcoming is that the items suffer from a response set that can com-
plicate interpretations of inter-item correlations (Green & Citrin, 1994). Lastly,
although the items allude to a collective response, they make no explicit refer-
ence to government involvement. Despite these problems, the indicators seem
more than serviceable for the job.
Military strength, the final principle under consideration, means bottom
line judgments about the use of military force versus softer forms of power in
3
There are problems with items that appear on NES surveys that seem to be face
valid measures of limited government. First, the seven-point scales provide
respondents with a NO filter, which, as discussed earlier, is unappealing. As
I show later in the chapter, missing data on these questions are high. Second,
some of the questions lack clear group referents. Indeed, no other set of items
can rival the spending battery in terms of breadth and depth of group coverage.
Third, there are fewer items available relative to the spending measures. In short,
although the standard seven-point scales are useful for some purposes, I believe
they are inferior to the spending battery.
the international arena. Hawks are comfortable with using military muscle to
advance foreign policy and national security objectives whereas doves prefer
noncoercive forms of engagement. I use whatever items appear on a given
survey to construct this variable (see the Appendix). Here are some examples.
First, a 1988 item: “The U.S. should maintain its position as the world’s most
powerful nation even if it means going to the brink of war.” Five response
options are given, ranging from “agree strongly” to “disagree strongly.”
Second, a 1992 query: “In the future, how willing should the United States
be to use military force to solve international problems . . . extremely willing,
very willing, somewhat willing, not very willing, or never willing?” Finally,
this question appears on the 2004 NES.
Some people believe the United States should solve international prob-
lems by using diplomacy and other forms of international pressure and
use military force only if absolutely necessary. Suppose we put such peo-
ple at 1 on this scale. Others believe diplomacy and pressure often fail
and the U.S. must be ready to use military force. Suppose we put them at
number 7. And of course others fall in positions in-between, at points 2, 3,
4, 5, and 6. Where would you place yourself on the scale, or haven’t you
thought much about this?
Although the items vary from year to year, the key point is that each elicits
overarching views on military power. The language is generally pitched at a
broad level with no allusion to specific military conflicts, foreign countries,
or regions in the world. Lastly, note that, except for the 2004 item, NO filters
are absent. Hence, missing data are, minus this item, very low. An important
drawback with these data is that only one indicator is available in the 1996
NES and no items appear on the 2000 and 2008 NES.
THE MEASUREMENT MODEL TESTS

Multidimensional Models
My theoretical argument maintains that the ideas underlying limited govern-
ment, moral traditionalism, and militarism are easy to understand, evaluate,
and assimilate. This should hold true for those who follow politics closely as
well as for those who do not. Operationally, this implies (1) a multidimensional
measurement model will fit the data within each sample and (2) robust item
factor correlations should be present in each sample. If such results emerge, I
can conclude that policy principles are available in the minds of the politically
aware and the unaware.
ξ1: ξ2: ξ3:

Limited Trad Military
Govt Morality strength
λ1 λ2 λ3 λ4 λ5 λ6 λ7
x1 x2 x3 x4 x5 x6 x7
δ1 δ2 δ3 δ4 δ5 δ6 δ7
Figure 5.1. Three-Factor Measurement Model.
There are, of course, alternative ways the data might be structured. Two
possibilities are especially pertinent. First, if the nonattitudes thesis is on the
mark, the opinions of the highly and moderately informed should be guided
by latent principles, but there should be little evidence of this among low
knowledge respondents. This pattern of results would suggest the latter do
not hold real principles. Second, for the most knowledgeable, responses to the
policy principle items might derive from a single left–right orientation, which
would undercut my argument that opinion is organized in a domain-specific
manner. For the sophisticated, perhaps everything reduces to a single liberal–
conservative dimension.
The available measures permit specification of three-factor models in 1988,
1992, and 2004 and two-factor models in 1996, 2000, and 2008.4 To illustrate my
approach, Figure 5.1 shows the 1992 measurement model. In this specification
4
The careful reader will recall that with the exception of the 2008 NES the lim-
ited government items have three response options. In light of this, I created
miniscales by adding responses from various pairs of items together to create
five-point measures that are used in the CFA analyses. This procedure ensures
the items better approximate interval level measurement as needed for accu-
rate estimation, although slippage obviously remains. Note also that for the
moral tradition items the agree–disagree response options are compromised
ξ1, ξ2, and ξ3 represent latent limited government, traditional morality, and
military strength, respectively. Each xi corresponds to the ith variable used to
measure a specified factor. The λi are standardized factor loadings that indi-
cate the correlation between the latent variable and the given indicator. Each
δi represents errors of measurement and unique sources of variation for xi.
Two-way arrows between the factors represent zero-order correlations. I esti-
mate the model at different levels of sophistication by dividing each sample
into comparably sized low, medium, and high knowledge groups. If the global
model fit statistics described below imply good fit in each sample, the con-
clusion of equivalent factor structures is justified. I then examine whether the
factor loadings are similar across the groups.
The first set of estimates appears in Table 5.1.5 The top and middle sec-
tions of the table list the latent principles and corresponding indicators in the
first column. The standardized factor loadings for the low, medium, and high
sophistication groups appear in the second through fourth columns, respec-
tively. The bottom part of the table lists the global fit statistics in column one
followed by the sample estimates in the next three columns. The robust χ2 is the
test statistic for the null hypothesis that the population covariance matrix for
the observed variables equals the covariance matrix reproduced by the meas-
urement model. A statistically insignificant result means the null hypothesis
by a response set that decreases the magnitude of the negative correlation

between oppositely worded items (i.e., between items in which “agree” reflects
traditional responses and items in which “agree” reflects tolerant responses),
implying that the attitudes are independent rather than bipolar. The simplest
way to combat this problem is to build scales whereby a “tolerance” item is
subtracted from a “family values” item (Green & Citrin, 1994). I follow that
approach here.
5
The EQS program (version 6.1) generates model estimates using direct maximum
likelihood procedures applied to the raw data matrix. The use of the technique
presumes the observed variables are multivariate normally distributed. When
this assumption is violated, the χ2 test and standard errors are biased. To account
for nonnormality in the data I report a robust χ2 statistic that is more accurate
than the standard maximum likelihood estimates (Curran, West, & Finch, 1996).
The technique further assumes the observed variables are continuous. Most
of the items I use have five to nine response options that approximate, albeit
crudely, interval level measurement. Simulation evidence suggests the applica-
tion of the method to data such as these is not too problematic until the number
of points on the scale drops below four (Bollen, 1989; Finney & Distefano, 2006;
Zumbo, Gadermann, & Zeisser, 2007).
TA B L E 5 . 1 . POLICY PRINCIPLE MEASUREMENT MODELS BY LEVEL OF

SOPHISTICATION, 1988
Standardized Factor Loadings Low Medium High
ξ1 Limited government:
λ1 Spending scale 1 .60 .56 .56
ξ2 Traditional morality:
λ4 Traditional morality scale 1 .68 .79 .76
ξ3 Military strength:
λ6 Military strength item 1 .39 .52 .55
Mean factor loading .58 .62 .68
Model fit:
Yuan–Bentler scaled χ2 34.54 38.28 24.44
Degrees of freedom 17 17 17
p-value <.01 <.01 .11
CFI .95 .96 .99
RMSEA .05 .05 .03
SRMR .04 .04 .03
Number of observations 567 599 571
Notes: Direct maximum likelihood estimates are based on raw data. Standardized factor loadings
represent the correlation between latent and observed variables. All loadings are significant at
p < .05. Latent and observed variables are coded so that higher scores reflect more support for a given
principle; therefore, each λ should be positive. CFI, comparative fit index; RMSEA, root mean square
error of approximation; SRMR, standardized root mean square residual.
Source: 1988 NES.
cannot be rejected and thus supports the inference that the specified model
holds in the population. Hence, in contrast to the typical practice, a statis-
tically insignificant result is desired. The χ2 performs well as a significance
test when used with large samples, when the distributional assumptions hold,
and when the model is specified correctly. Because one or more of these condi-
tions are typically violated in a given application, the χ2 test loses some value
as a significance test (Bollen, 1989). Because of this, I also report three descrip-
tive fit measures that assess global fit according to various statistical criteria.
These include the robust comparative fit index (CFI) where values close to or
in excess of .95 indicate good fit, the root mean square error of approximation
(RMSEA) in which values close to or below .06 suggest reasonable fit, and the
standardized root mean square residual (SRMR) in which values close to or

less than .08 are preferred (Hu & Bentler, 1999).6
Table 5.1 indicates whether policy principles are available in the minds of
cognitively heterogeneous groups in 1988. As shown by the global fit statis-
tics, the hypothesized three-factor model does a good job in reproducing the
covariance matrix. In each group the CFI, RMSEA, and SRMR meet or exceed
the desired cut points. In addition, the χ2 value proves insignificant in the
high sample, though it attains significance in the other two. Next, consider
the magnitude of the factor loadings across the samples. Recall that these val-
ues represent correlations between latent and observed variables. If the item
factor correlations are reasonably robust and of comparable size, this would
suggest respondents at various degrees of sophistication differ little in their
ability to turn latent attitudes into opinion statements. Rather than discussing
the individual items sequentially, I report the mean factor loading within each
group in the middle of the table. The mean λ is .58 for the least informed, .62
for the moderately informed, and .68 for the most informed. Two points bear
highlighting. First, the correlation means are sizable in each group. Second,
as we move up the sophistication ladder the loadings grow larger. I conclude
that citizens in each knowledge tercile do a good job mapping their positions
on latent principles onto opinion scales and the more sophisticated do a bit
better in this regard. To sum up for 1988, abstract principles are structured
coherently and, for the most part, equivalently in the belief systems of the
politically aware and unaware.
Although these results are not surprising for the top two categories, my
finding that the opinions of the least informed are structured by underlying
predispositions deserves some emphasis. Clearly, their reported positions
represent far more than random noise. Said another way, the responses of
the least informed reflect latent attitudes, and, hence, cannot be dismissed as
doorstep opinions. This suggests that all citizens get some exposure to the
major policy cleavages that drive American politics, understand what they
have been exposed to, evaluate the information, and internalize their reac-
tions. To see if the pattern carries forward, I move onto the other years.
Table 5.2 lists the 1992 estimates. To start, the global fit statistics suggest a
three-factor model works in each group. Whereas the χ2 statistic is significant in
6
The CFI is an incremental fit index that assesses the proportionate improvement
in model fit relative to a baseline where all observed covariances are zero. The
RMSEA and SRMR are absolute fit measures that assess how well the posited
model reproduces the sample data (the former includes a parsimony correction).

Standardized factor loadings Low Medium High
Model fit:
p-value <.01 .23 <.01
CFI .95 1.00 .98
RMSEA .05 .02 .05
SRMR .04 .03 .03
represent the correlation between latent and observed variables. All loadings are significant at p < .05.
Latent and observed variables are coded so that higher scores reflect more support for a given principle;
therefore, each λ should be positive. NES sampling weight has been used. CFI, comparative fit index;
RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual.
Source: 1992 NES.
the low and high samples (remember that an insignificant result is preferred),
the descriptive fit measures meet or surpass the fit thresholds in every group.
The next concern is whether the factor loadings are strong within and compa-
rable across the samples. The mean loading equals .57 for the less informed, .68
for the moderately informed, and .70 for the most informed. Like the 1988 esti-
mates, the item factor correlations are substantial in all three groups, albeit more
so among the better informed. Evidently, most folks receive sufficient exposure
to these ideas in the broader political environment to form genuine evaluations.
The 1996, 2000, 2004, and 2008 estimates appear in Tables 5.3–5.6, respec-
tively. The results from all four surveys reinforce the conclusions reported
above. To summarize, model fit ranges from good to excellent in each sample
in each year, the mean item factor correlations are all substantial (varying

Model fit:
p-value .66 .01 .04
CFI 1.00 .98 .99
RMSEA .00 .07 .06
SRMR .01 .03 .03
Source: 1996 NES.
from .56 to .79), and the item factor correlations are always larger in the high
group relative to the low group (on average by .20). Taken in total, the CFA
results confirm my theoretical expectations. The evidence reveals the speci-
fied multidimensional model always fits the data, whereas the factor loading
comparisons imply political sophisticates and novices do well in reporting
their positions (but again, the aware do better). My findings suggest that
although individuals vary greatly in what they know about politics, every-
one possesses policy principles, and, thus, satisfies the first condition of pol-
icy voting.
A Unidimensional Alternative?
Evidence presented above indicates that three distinct policy predispositions
reside in the political attitude structures of everyday people. This allows me
to discount the nonattitudes thesis, which maintains that such attitudes are
absent from the minds of political novices. However, as discussed above, a
λ2 Poor people item .74 .74 .76
Model fit:
p-value .12 <.01 .03
CFI .98 .95 .99
RMSEA .04 .09 .06
SRMR .03 .04 .03
Source: 2000 NES.

Model fit:
104
TA B L E 5 . 5 . (Continued)
Low Medium High
p-value .67 .11 <.01
CFI 1.00 .98 .98
RMSEA .00 .04 .08
SRMR .03 .03 .05
Source: 2004 NES.

λ1 Poor people item .77 .82 .79
λ2 Welfare item .52 .54 .73
λ3 Child care item .57 .62 .71
λ4 Social security item .40 .47 .50
Model fit:
p-value .01 .03 <.01
CFI .97 .99 .99
RMSEA .05 .04 .05
SRMR .03 .03 .03
Source: 2008 NES.
105
competing hypothesis avers that a single left–right predisposition structures

policy opinions among the aware. To test this idea, researchers often calcu-
late correlations between multiple preference variables. Scholars have learned
that left–right consistency between preferences rises with sophistication and
related variables. Although this is sometimes interpreted as evidence of ideo-
logical structuring, it should be recognized that correlations provide no direct
evidence regarding dimensionality. To see if the opinions of the most aware
are structured by a liberal–conservative predisposition, I examine attitude
consistency between the policy principle items and then turn to CFA tech-
niques. The variables are keyed such that higher scores denote conservative
positions; therefore, the correlations should be positive. Moreover, these cor-
relations, reported in Table 5.7, should grow stronger moving from lower to
higher sophistication.
To begin with 1988, the mean correlation between the limited government,
traditional morality, and military strength items equals .27 among the highly
sophisticated, .14 in the medium group, and .12 in the low group. The high–low
difference works out to .15. The same pattern of results holds for all remaining
years. To look at the 2008 data, the correlation means are .37 in the top sample,
.22 in the middle sample, and .17 in the low sample. As predicted, the more
informed respondents are, the greater their degree of attitude consistency.
These results establish that sophistication promotes left–right consistency,
but they do not tell us whether the opinions of the most knowledgeable are
generated by a single left–right orientation. To test this I specify a measurement
model whereby all items are constrained to load on a single factor. If the posi-
tions taken by the politically aware are driven by an ideological disposition,
the one-dimensional model should fit the data as well as the multidimensional
TA B L E 5 . 7 . MEAN PEARSON R FOR POLICY PRINCIPLE ITEMS BY LEVEL

OF SOPHISTICATION, 1988–2008
Low Medium High High–Low
1988 .12 .14 .27 .15

1992 .11 .22 .29 .18
1996 .17 .28 .41 .24
2000 .15 .23 .33 .18
2004 .12 .23 .45 .33
2008 .17 .22 .37 .20
Notes: Cell entries represent mean Pearson product-moment correlation between all observed
variables.
Source: 1988–2008 NES surveys.

model. The one-factor solution will then be preferable on grounds of parsi-

mony. But if the fit of the unidimensional model lags that of the multidimen-
sional alternative, support for the rival hypothesis will evaporate.
Table 5.8 enumerates the results. For the sake of completeness, one-
factor solutions are reported for all samples. I find no evidence that the
one-dimensional model adequately fits the data anywhere. To begin with
TA B L E 5 . 8 . UNIDIMENSIONAL MEASUREMENT MODELS BY LEVEL OF

SOPHISTICATION, 1988–2008
Low Medium High
1988:
p-value <.01 <.01 <.01
CFI .47 .61 .29
RMSEA .14 .14 .18
SRMR .11 .13 .12
Mean standardized factor loading .30 .34 .35
1992:
p-value <.01 <.01 <.01
CFI .41 .56 .77
RMSEA .15 .19 .17
SRMR .10 .13 .10
1996:
p-value <.01 <.01 <.01
CFI .76 .60 .79
RMSEA .12 .26 .25
SRMR .07 .13 .10
2000:
p-value <.01 <.01 <.01
CFI .75 .48 .52
RMSEA .13 .26 .35
SRMR .08 .13 .18
2004:
p-value <.01 <.01 <.01
CFI .39 .69 .80
(Continued)
Low Medium High
RMSEA .13 .15 .21

SRMR .10 .09 .09
2008:
p-value <.01 <.01 <.01
CFI .82 .71 .69
RMSEA .11 .17 .24
SRMR .07 .11 .12
Notes: Direct maximum likelihood estimates are based on raw data. Sample sizes are reported in
Tables 5.1–5.6. NES sample weights have been used. CFI, comparative fit index; RMSEA, root mean
square error of approximation; SRMR, standardized root mean square residual.
Source: 1988–2008 NES.
high knowledge respondents, the χ2 values are all extremely large and
highly significant (p < .00001), every CFI lies below the .95 threshold, all
RMSEA values exceed the .06 cutoff, and the SRMR returns mediocre val-
ues. When the unidimensional estimates are compared to their multidi-
mensional counterparts, the latter’s superiority is self-evident. To take an
example, the 2004 CFI for the high group is .98 in the three-dimensional
model versus .80 in the unidimensional model. Likewise, the RMSEA
equals .08 for the three-factor model compared to .21 for the one-factor
model. The mean factor loadings are also lower in the unidimensional
model compared to the multidimensional model (.66 < .79). Things are
no different in the medium and low knowledge samples. Here too the
multidimensional model bests the unidimensional specification. This, of
course, comes as no surprise given the ideological innocence of much of
the mass public.
To sum up the measurement model results, citizens with various degrees
of political knowledge hold real attitudes toward federal support for strug-
gling Americans, standards of moral rectitude, and the use of military power.
Moreover, my evidence calls into question two rival perspectives. Contrary
to the nonattitudes thesis, the policy opinions of the least informed are struc-
tured coherently and sensibly. Against the ideological sophistication thesis,
the supposition that a global left–right orientation drives opinion for the most
informed receives no support.
THE ATTITUDE STABILITY TESTS

Using Panel Data to Examine Attitude Availability
The CFA results imply that policy principles are available in the minds of
almost everyone. The evidence backing this conclusion rests entirely on the
statistical analysis of cross-sectional data. To strengthen the case, I use panel
data to carry out another test. As my Chapter 4 discussion on response sta-
bility showed, researchers have calculated continuity correlations to exam-
ine whether members of the general public hold real attitudes. When a
person has thought a great deal about something, the psychological result is
an enduring tendency to evaluate the entity with some consistency. If some-
one holds a crystallized attitude, his or her responses to survey questions
measuring it should be very similar when asked on separate occasions. In
contrast, when someone thinks about a political matter for the first time, or
the first time in a long time, the evaluation must be constructed on the spot.
In the case of uncrystallized attitudes, opinion reports should vary unpre-
dictably because little or no evaluative knowledge is stored in memory to
guide the answers.
Of course, a high level of stability need not mean latent principles are crys-
tallized to the same degree for everybody. In fact, there are good reasons for
positing sophistication-based differences on this point. The politically aware
follow public debate and spend more time thinking about politics than the
unaware. More thought leads to more frequent attitude expression, which in
turn pushes evaluations closer to the dispositional end of the attitude–non-
attitude continuum (Converse, 1970; Fazio, 1989). So, even though all citi-
zens acquire general principles, these should be somewhat more crystallized
among the politically engaged. Operationally, this implies higher continuity
correlations in more knowledgeable strata.
But what type of correlation?7 Uncorrected correlations are the preferred
method for some in light of evidence that random measurement error is
attributable to poorly informed respondents (Feldman, 1989; Norpoth &
Lodge, 1985). Others argue that the locus of error lies in faulty measures
7
One important point to note is that the standard deviations of the variables are
usually larger in sophisticated groups, which contributes to higher continu-
ity correlations. These differences reflect the tendency of the less informed to
cluster near or at the middle of the scale whereas more informed respondents
report more extreme positions. The assumption is that these distributions arise
because the less informed hold weak or nonexistent attitudes and choose middle
responses to hide this fact (Converse, 1964, 1970; Jennings, 1992).
rather than know nothing respondents (Achen, 1975; Krosnick & Berent,
1993). Inasmuch as question-induced error is misattributed to survey par-
ticipants, raw correlations represent negatively biased estimates of attitude
stability. The solution is to apply statistical methods that purge the opin-
ion items of random error to generate corrected continuity correlations that
reflect “true” stability.
I argued that random error is a joint product of flawed respondents and
flawed questions. Given this, it seems unwise to insist uncorrected or cor-
rected correlations are superior. The former method underestimates attitude
stability and the latter overstates it. I take a balanced approach here by report-
ing uncorrected and corrected correlations and then taking the mean of these
estimates, which in effect partitions the error equally between respondents
and questions. To the extent that response instability is disproportionately
affected by respondents or questions, my “split the difference” approach is
potentially misleading. Nevertheless, this strategy remains preferable because
it takes more information into account than standard approaches. In brief, the
continuity correlation averages should yield more accurate estimates of atti-
tude stability than either uncorrected or corrected measures.
I rely on data from the 1992–1994–1996 NES panel to estimate the corre-
lations. Each wave contains a sufficient number of limited government and
morality indicators to conduct the tests. The welfare state measure is com-
posed of several of the items used earlier. I tap moral attitudes with the same
items described above. Correlations are calculated for the 1992–1994 and
1994–1996 intervals.
The NES panel does not contain repeated measures of abstract hawk–dove
orientations. Fortunately, Peffley and Hurwitz (1993) have undertaken such
an analysis. They administrated a two-wave panel survey to residents of the
Lexington, Kentucky metropolitan area over a 13-month period and reported
a corrected correlation of .97. An observed correlation was not reported. The
authors did not stratify the sample by political knowledge, educational attain-
ment, or some comparable variable, so it remains unclear whether stability
varies by level of sophistication. However, given the high value of the correla-
tion, any expert novice discrepancies would be marginal.
Party Identification as a Baseline

The examination of attitude stability can be deepened by adding party identifi-
cation to the mix. Party identification provides a useful benchmark for evaluat-
ing policy principles. Given the importance of partisanship to my efforts here,
as well as the role it plays throughout the book, I elaborate its definition and
measurement at this juncture. By party identification, I mean a psychological
attachment to a political party (Campbell et al., 1960). Individuals identify

with a party because they think of themselves in partisan terms and hold feel-
ings toward party labels, not because they are formally registered with a party
or vote consistently for its nominees. Its key conceptual attributes are partisan
self-categorization and partisan affect.
A type of social classification underlies the cognitive component of identi-
fication. Building on the work of The American Voter, Green et al. (2002) argued
that individuals see the parties as representing disparate social groups.
Members of these groups identify with the party they believe is the better
fit for their social group(s). To illustrate, the Democratic Party is often seen
as sympathetic to labor whereas the GOP is viewed as the political home of
business. Thus, a union member is more likely to develop an attachment to the
Democratic Party than a business entrepreneur, who in turn should gravitate
toward the Republican side. Affect plays a central role in most treatments of
partisanship. According to Campbell and his colleagues (1960) “the individ-
ual’s affective orientation to an important group-object in his environment”
(i.e., a political party) is rooted in “the attracting or repelling quality of the
group as the generalized dimension most critical in defining the individual-
group relationship” (p. 121). Someone who feels warmly toward Republicans
and abhors Democrats should develop a psychological affinity for the GOP
label (cf. Sears, 2001).
For the stability tests, I rely on a trio of measures to tap partisan affinities,
which facilitates estimation of corrected continuity correlations. The standard
seven-point scale captures self-categorization. Respondents answer a two-part
item about the direction and strength of their partisanship on a scale rang-
ing from strong Democrat to strong Republican. Second, to capture partisan
affect, feeling thermometer scores are utilized. Respondents rate how warmly
or coolly they feel toward the Democratic Party and Republican Party on 101-
point scales. Higher values indicate warmer feelings. I recalibrate the ther-
mometers as a simple difference measure. Both the self-placement and feeling
thermometer difference variables are normed to a 0–1 range with higher scores
denoting stronger GOP attachments.8
8
The use of the seven-point scale is widely accepted, but the same cannot be
said for the feeling thermometer items. Some maintain that feeling thermom-
eter scores reflect short-term evaluations rather than attitudinal dispositions.
However, some work suggests the feeling thermometers represent valid and
reliable indicators of deeper partisan attachments. CFA results imply the seven-
point scale and feeling thermometers tap a single underlying factor (Goren, 2005).
Other efforts show that responses to the feeling thermometers nearly match the
Regarding stability, many have found that party identification is remark-

ably persistent over time. That is, the relative positions between Democrats
and Republicans in the partisan distribution change little from year to year
(Converse & Markus, 1979; Green et al., 2002). More recent evidence shows
that some occasionally update party loyalties under various conditions, so
these identities are not completely fixed (Carsey & Layman, 2006; Dancey &
Goren, 2010). Yet even when party ties are updated, the simultaneous influ-
ence of party on these attitudes is usually stronger. Finally, when party stabil-
ity is compared to that of other political attitudes and beliefs, party usually
wins by a wide margin. In short, party identification is the most durable polit-
ical predisposition, and thus provides a useful point of reference for assessing
the stability of policy principles.
I make two predictions about partisan stability. First, given the founda-
tional role party identification plays in mass belief systems, its continuity
correlations should surpass those of policy principles. Second, because the
politically sophisticated think more often and deeply about public affairs than
the unsophisticated, these correlations should increase in magnitude at higher
levels of sophistication.
Continuity Correlation Results

I focus initially on the raw correlations in Table 5.9. Remember, the assump-
tion here is that random measurement error is due primarily to befuddled
citizens. Respondents are again stratified into low, medium, and high sophis-
tication categories. To begin with party identification9294 the continuity cor-
relation (Pearson r) is .68 in the low knowledge group, .85 in the medium
group, and .88 in the high group. For 1994–1996, the correlations are .78, .86,
and .93 in the selfsame groups. By convention, these are impressive values.
In addition, partisan identities are, as anticipated, more consistent over time
for sophisticated respondents. The 1992–1994 difference between the high and
low sample is .20, a sizable but not overwhelming gap. The 1994–1996 differ-
ence shrinks a bit to .15. These values imply that although partisan identities
persist for everyone, they are more durable among the better informed.
This is a critical finding because it suggests sophistication-based differ-
ences in continuity correlations should not be viewed as evidence that the
less informed lack true attitudes. Although some interpret differences such
temporal stability of the seven-point scale (Green et al., 2002). Collectively, these
studies suggest that the feeling thermometers serve as good proxies of latent
partisanship.
TA B L E 5 . 9 . UNCORRECTED CONTINUITY CORRELATIONS BY LEVEL OF

SOPHISTICATION, 1992–1994–1996
Party identification
1992–1994 .68 .85 .88 .20
1994–1996 .78 .86 .93 .15
Limited government
1992–1994 .56 .67 .68 .12
1994–1996 .52 .65 .67 .15
Traditional morality
1992–1994 .55 .67 .73 .18
1994–1996 .55 .66 .80 .25
Notes: Cell entries represent observed variable correlations (Pearson r).
Source: 1992–1994–1996 NES panel.
as these as supporting the nonattitudes thesis, no one believes the less aware
lack genuine partisan orientations. Unless we are willing to argue a gap of
.20 means partisan identities are doorstep opinions for the unsophisticated
even though their continuity correlation equals .68, the conclusion that gen-
uine attitudes vary in durability across levels of sophistication seems hard to
dispute. Said differently, the fact that partisan identities are sturdier among
the politically sophisticated does not imply that the unsophisticated lack such
ties. Instead, it means partisan orientations are somewhat less crystallized
among the less knowledgeable. We should keep this in mind when evaluating
the remaining correlations.
How does limited government stack up? For 1992–1994, Table 5.9 reveals
correlations of .56, .67, and .68 in the low, medium, and high groups, respec-
tively. The comparable figures for 1994–1996 are .52, .65, and .67. These cor-
relations are moderately strong in the low group and somewhat better in the
higher groups. The high–low discrepancy is .12 in the first interval and .15 in
the second. As expected, then, sophistication promotes stability. But the key
point to take away is that attitudes toward limited government, measured via
a lower bound estimate, appear reasonably solid for most persons. Note lastly
that attitudes toward government activism fluctuate more than partisan iden-
tities, a finding that should surprise no one.
The traditional morality estimates appear in the bottom row of the table.
For 1992–1994, the correlation equals .55 in the low sample, .67 in the middle
sample, and .73 in the high sample. For 1994–1996, the respective marks are
.55, .66, and .80. These results reflect moderate stability in the bottom group
TA B L E 5 . 1 0 . CORRECTED CONTINUITY CORRELATIONS BY LEVEL OF

1992–1994 .71 .91 .95 .24
1994–1996 .87 .91 .97 .10
Limited government
1992–1994 .90 .91 .89 –.01
1994–1996 .64 .86 .89 .25
1992–1994 .70 .83 .85 .15
1994–1996 .86 .83 .93 .07
Notes: Cell entries represent latent variable correlations.
and considerable stability in the higher two. Next, we can see that the high–
low difference equals .18 in 1992–1994 and .25 in 1994–1996. These gaps imply
that knowledge enhances the stability of moral attitudes, just as it did for the
partisan and economic welfare predispositions. Finally, the moral attitude cor-
relations also lag party identification. Overall, the key points to take away
from Table 5.9 are that partisan identities represent the most durable predis-
position, policy principles hold steady as well, and sophistication promotes
stability. These results reinforce the claim that policy principles are widely
available in the public mind.9
Table 5.10 reports the corrected estimates, which reflect the correla-
tion between latent factors free of measurement error. Here, the operating
assumption is that error arises from problematic questions. Because respon-
dents are absolved of blame, disattenuated correlations can be viewed as
upper bound estimates of attitude crystallization. For starters, latent par-
tisanship proves remarkably enduring. For 1992–1994, the corrected cor-
relation equals .71 in the low group .91 in the middle group, and .95 in the
top group. Over the latter waves, the respective correlations are .87, .91,
and .97. Next, the high–low stability difference is .24 in 1992–1994 and .10
in 1994–1996. These differences suggest knowledge does indeed promote
partisan stability.
9
Part of the difference is because there is more variance in the opinions of aware
respondents.
Moving on, I find impressive stability for limited government in all sam-
ples. For 1992–1994, the corrected correlation is .90 among the least informed
portion of the electorate, .91 among the moderately informed, and .89 for the
most informed. In an unanticipated finding, there is no high–low difference
to speak of. For the second period, the correlations are .64, .86, and .89 for the
respective groups. As predicted, the correlation grows larger in more aware
samples. But the most noteworthy discovery is that the correlations are gen-
erally high, which suggests limited government behaves more like an evalu-
ative disposition than a temporary construction. The final point to stress is that
limited government proves less stable than latent partisanship in the second
panel wave, but not the first. For traditional morality9294 the low, medium, and
high continuity correlations are .70, .83, and .85, respectively. For 1994–1996,
the figures are .86, .83, and .93. I draw three conclusions from these results.
First, evaluations of moral behavior are stable for all. Second, political aware-
ness promotes crystallization. Third, the stability of moral views trails that of
party.
Overall, the corrected correlations demonstrate that latent partisanship,
government power, and traditional moralism are sturdy at each level of
sophistication. This, in turn, reaffirms my claim that these are real attitudes
for most of the public. Party identification may be the most enduring orien-
tation, but limited government and moral traditionalism prove durable as
well. Before proceeding, I offer some commentary to put these results in per-
spective. Critics of these techniques sometimes argue that because the unso-
phisticated provide less reliable responses than the sophisticated, corrected
correlations purge more random error from the responses of the less aware
and thus artifactually minimize stability differences. A superficial comparison
of some results in Tables 5.9 and 5.10 lends credence to this view. For instance,
the high–low difference for limited government9294 is .12 for the uncorrected
case and –.01 for the corrected case. But other examples show the high–low
gap widens after correcting for measurement error. This happens for limited
government9496 (uncorrected difference = .15 vs. corrected difference = .25).
When all high–low differences in each table are taken into account, the mean
difference equals .18 for the uncorrected correlations versus .13 for the cor-
rected ones. Although the gap narrows to some degree, the method of purging
correlations of random error does not obliterate these differences. As such, the
corrected correlations should not be dismissed.
As argued above, the attenuated and disattenuated correlations are biased
measures of attitude stability. The uncorrected estimates are biased down-
ward because responsibility for measurement error is placed squarely on
respondents even though the questions deserve a share of the blame. The cor-
rected estimates are biased upward because the locus of error is presumed
to lie with faulty questions even though individuals are partly responsible.
Given the limits of both methods, my solution is to calculate the mean con-
tinuity correlation. In essence, this move divides the error equally between
respondents and measures. The average correlations in Table 5.11 stand as my
best estimates of attitude stability.
I begin with party identification. For 1992–1994, the party mean assumes
a value of .70 in the low sample, .88 in the medium sample, and .92 in the
high sample. Turning to 1994–1996, the respective correlations are .83, .89, and
.95. In terms of magnitude, five of the six estimates are extremely robust: the
sixth (low9294 = .70) is also impressive. The next point to note is that the high–
low difference equals .22 in period one and .12 in period two. Hence, partisan
identities endure for citizens across the board and, most especially, for the
most informed.
For limited government9294 the mean correlation is .73 in the low group and
.79 in the two higher groups. The corresponding 1994–1996 estimates are .58,
.76, and .78. These are, for the most part, very solid correlations. The high–low
difference is .06 in the first period and .20 in the second, showing anew that
crystallization rises with sophistication. Overall, it seems that for most people
attitudes toward government aid to those with varying degrees of need are
stable belief system elements. I finish up with moral traditionalism. The 1992–
1994 correlation is .63 in the low sample, rises to .75 in the medium group, and
TA B L E 5 . 11 . MEAN CONTINUITY CORRELATIONS BY LEVEL OF

1992–1994 .70 .88 .92 .22
1994–1996 .83 .89 .95 .12
Limited government
1992–1994 .73 .79 .79 .06
1994–1996 .58 .76 .78 .20
1992–1994 .63 .75 .79 .16
1994–1996 .71 .75 .87 .16
Notes: Cell entries represent the mean of the attenuated and disattenuated continuity correlations
from Tables 5.9–5.10.

peaks at .79 in the high group. In 1994–1996 these values are .71, .75, and .87,
respectively. The high–low gap is .16 in both periods, showing that moral pre-
dispositions are firmer in the highest strata. Broadly speaking, moral attitudes
are reasonably to highly stable for the least sophisticated and very stable for
the most sophisticated.
The stability tests have covered a lot of ground, so now is a good time
to tie everything together. First, by the standards used to judge continuity
correlations, my estimates suggest that limited government and traditional
morality represent long-term evaluative dispositions in mass belief systems.
These results imply policy principles are psychologically available regardless
of how much or how little political information citizens have. Second, stabil-
ity is higher for knowledgeable groups. Therefore, although policy principles
reside in attitude hierarchies, their degree of crystallization varies predict-
ably across cognitively heterogeneous populations. Third, party identification
emerges, as always, as the most stable political predisposition. A final caveat
is in order. Readers should note that impressive stability does not mean per-
fect stability. Hence, I make no claim that policy principles are fixed and thus
invulnerable to change. They are persistent, not immutable. They may evolve
over time, but such change occurs at a deliberate pace.
A CHECK ON OPINION HOLDING

I presented two bodies of evidence suggesting that most voting age adults
hold genuine principles, but have avoided a potential problem lurking in
the background. A reasonable criticism of some public opinion work is that
researchers ignore the implications item nonresponse has for their substantive
conclusions (Converse, 1964, 1970). Item nonresponse occurs when respon-
dents say something such as “I’m not sure” or “I haven’t really thought about
it” instead of selecting a valid response option from the menu of choices.
On any given issue, the proportion of the sample reporting no opinion may
range from a point or two up to a third or more. When item nonresponse
is aggregated across multiple questions, opinionation declines rapidly, espe-
cially among uninformed respondents (Delli Carpini & Keeter, 1996).
Nonresponse is a big problem in the NES seven-point issue scales. These
measures provide respondents with the option of saying they “haven’t thought
much” about the issue at hand. Many choose this option. Others volunteer
that they “don’t know” or have “no opinion”. To illustrate with 1992 NES
data, I calculate nonresponse rates for the low, medium, and high groups for
the following issue scales: (1) government services, (2) defense spending, (3)
health care, (4) government guaranteed job, (5) aid to blacks, (6) urban unrest,
and (7) women’s role. In the top knowledge category 79% answered all seven
questions versus 67% in the medium tercile and 40% in the low knowledge
sample. Comparable patterns exist in other NES surveys. Clearly, item nonre-
sponse is rampant among less informed respondents.
Analysts sometimes address this problem using statistical algorithms
to impute opinions for the policy questions respondents failed to answer.
Ansolabehere et al. (2008) rely on this strategy in their study on attitude sta-
bility and issue voting. Using issue scales constructed from up to 24 items,
they found stability levels are much higher than previously reported and that
sophistication-based differences are marginal. They took respondents who
answered at least 75% of the issue items and imputed preferences for their
unanswered questions. In light of the missing data patterns just described,
it seems likely that preference imputation occurs most frequently among the
least aware. A critic might charge that their stability estimates are at least in
part an artifact of the imputation process. To the extent that such respondents
do not belong on the scale at all, the imputation of preferences may be contro-
versial (Converse 1964, 1970).
Rather than taking a position on the merits of this particular imputation
strategy, the points I wish to stress are that, with one exception, levels of non-
response are low for the items used in my analyses and, most critically, are
largely unaffected by sophistication. Table 5.12 reports the percentage of each
sample in each year taking positions on all measures. Some examples illu-
minate the point. First, in 1992 84% in the low knowledge group answered
all 12 items compared to 91% in the middle and high groups. Now, there is
a seven-point difference between the high and low groups, but it is a small one
at that. Moreover, it is tiny compared to the 39 point difference for the seven
issue scales described above. 1992 is not unique. In 2008, 92% of the low and
medium knowledge respondents answered the eight policy principle ques-
tions versus 94% in the high sample. The only exception to this pattern is in
2004 where 93% of the high group answered all 11 items versus 65% for the
low group, a 28-point difference. The problem here is that a hawk–dove item
contains a “don’t know” filter that many of the unsophisticated utilized. In
any case, Table 5.12 shows that subjects answer most of the questions and, as
seen earlier in the chapter, when they do the responses of the uninformed are
as meaningful as those of the informed.
To sum up, my supposition that political novices hold genuine principles is
not compromised by excessive missing data. In conjunction, the factor analy-
sis estimates, the continuity correlation results, and the opinion holding rates
point to the same conclusion. Core principles are very widely available in
political belief systems.
TABLE 5.12. OPINION HOLDING BY LEVEL OF POLITICAL SOPHISTICATION

1988: 13 items 80 86 91 11
1992: 12 items 84 91 91 7
1996: 10 items 93 95 96 3
2000: 9 items 92 90 91 –1
2004: 11 items 65 86 93 28
2008: 8 items 92 92 94 2
Notes: Cell entries represent percentage of respondents giving valid responses to all policy principle
items.
CONCLUSIONS
If policy attitudes are to shape electoral choice, they must reside in long-
term memory. In Chapters 2–4 I argued that policy principles are embed-
ded deeply in political debate and the party system, are invoked regularly
during presidential campaigns, and are readily comprehensible. Because of
this, the American public receives sufficient exposure to evaluate the ideas,
encode their impressions in memory, and retrieve them later on. This chap-
ter examined whether groups at different levels of sophistication hold such
principles. My analysis of opinion data from multiple NES surveys covering
a 20-year period yields an unequivocal verdict. The unsophisticated, mod-
erately sophisticated, and highly sophisticated hold real attitudes toward
limited government, conventional morality, and military power. In the Eagly
and Chaiken framework, policy principles are better characterized as dura-
ble evaluative dispositions rather than temporary evaluations constructed
on the spot. In the language of Converse, core principles fall much closer
to the attitude end of the attitude–nonattitude continuum. Because most
folks hold policy principles, they satisfy the availability condition for policy
voting.
This does not mean sophistication has no impact on the quality of domain-
specific principles. Because the politically sophisticated think about govern-
ment and politics more often than the unsophisticated, their policy attitudes
should be more crystallized. This is precisely what I have found. The opinion
reports of the informed correlate a bit more robustly with the latent attitudes
they are designed to measure. Likewise, although these attitudes are stable for
almost everyone, stability is higher among the more knowledgeable. Without
Limited Government
Density
Strong Govt Limited Govt

Note: Range [0–1], Mean = .31, Stdev = .23
Traditional Morality
Density
Progressive Traditional
Military Strength
Density
Dove Hawk
Source: 2004 NES
Figure 5.2. Public Opinion on Policy Principles, 2004.
gainsaying these two findings, the takeaway point is that policy principles are
long-term forces in the minds of all citizens.
Before proceeding, it should prove helpful to view the distribution of
public opinion on each dimension. Figure 5.2 stacks smoothed histograms
(kernel densities) for collective opinion on limited government, traditional
morality, and military strength in 2004. All variables lie on a 0–1 range and
Limited Government
Density
Strong Govt Limited Govt

Density
Progressive Traditional
Military Strength
Density
Dove Hawk
Source: 1992 NES
Figure 5.3. Public Opinion on Policy Principles, 1992.
are scored so that higher values reflect right-wing opinion. For limited gov-
ernment, Figure 5.2 reveals a left leaning public, one that on average prefers
more government to less. This can be seen in the shape of the distribution,
the bulk of which falls in the liberal end of the scale, and the limited govern-
ment mean, which equals .31. Next, note the American public leans to the
right on traditional morality, as indicated by the distribution’s shape and .56

mean. Americans also tend toward the hawkish side of the military strength
continuum: the distribution lies a bit to the right with a mean of .60. The final
point to note is the considerable variation across the dimensions. Although
many cases are concentrated around the aforementioned central tendencies,
a sizable minority of the electorate falls in the conservative side of limited
government and there are many people in the liberal halves of the morality
and military dimensions. Moreover, the distributions extend to the extreme
values of the scale for all three variables.
These results are typical for the 20-year period covered by my data. To
show this, Figure 5.3 stacks smoothed histograms for each dimension using
the 1992 NES data. Once again we confront a public that favored strong gov-
ernment over limited government (mean = .36), traditional moral standards
over progressive alternatives (mean = .59), and hard power over soft power
(mean = .56). And, once again, there is considerable variation with cases fall-
ing across the range and extending out to the left and right tails of the distri-
butions. More broadly, given that the public leans left on federal activism and
right on the cultural and military dimensions, it is clear why the GOP has long
sought to fracture the Democratic Party coalition by drawing attention away
from the economic welfare cleavage to other policy terrain on which it can
better compete.
To conclude, the evidence marshaled throughout this chapter raises the
possibility that policy principles function as central heuristics in the minds of
voters. The case for such will be strengthened if I can show these principles
guide the construction of short-term preferences for citizens across the aware-
ness divide. I examine this next.
CHAPTER 6
The Centrality of Policy Principles
My theory of voter choice posits that a set of policy principles functions as

central heuristics in the belief systems of politically aware and unaware citi-
zens. Rather than deliberating at length about all aspects of a given problem,
cognitive misers simply take positions consistent with the broader principle
relevant to the choice at hand. Doing so helps them make reasonable politi-
cal judgments in a timely fashion. Insofar as individuals get into the habit of
using policy principles to evaluate issues, this decision-making strategy can
potentially be applied to the problem of candidate choice.
The purpose of this chapter is to test the proposition that domain-
specific principles are central heuristics that guide the positions people
take on the issues of the day. My analysis proceeds as follows. First, if pol-
icy principles act as heuristics, they should constrain issue preferences in
a top-down manner without being constrained by them. Adopting the meth-
odology laid out by Peffley and Hurwitz (1993) and McCann (1997), I use
1992–1994–1996 National Election Study (NES) panel data to unpack the
temporal dynamics between core principles and issue preferences. Second,
if sophistication is not a prerequisite for calling upon principles, all citizens
should invoke these when evaluating political controversies. I employ data
from the 1988–2008 NES cross-sectional surveys to verify this claim. Third,
if policy principles are central idea elements in mass belief systems, their
power to shape issue judgments should eclipse that of cues such as party
identification. The cross-sectional data are used to check this claim as well.
Overall, my analyses demonstrate that limited government, traditional
morality, and military strength serve as powerful heuristics for the mass
public. Those who are knowledgeable about public affairs sometimes prove
more adept at exploiting abstract principles, but even less informed citizens
put them to good use. In brief, all citizens satisfy the centrality condition of
policy voting.
123
ASSESSING ATTITUDE CENTRALITY

Models of Constraint
In political attitude structures, central principles occupy key nodes from which
they vertically constrain peripheral attitudes. Such principles help people cope
with innate cognitive limitations that impact the decision-making process.
Guided by this approach, I argued that attitudes toward limited government,
traditional morality, and military strength, which encapsulate the primary cleav-
ages that have animated political discourse in this country for many years, oper-
ate as central heuristics that structure attitudes about short-term controversies
in the economic welfare, cultural, and national security domains, respectively.
To illustrate the logic of heuristic reasoning, imagine that people are asked
whether they favor a government insurance plan that covers all medical and
hospital expenses for everyone or private insurance plans in which individuals
and employers pay the bills. Assume they have never given the question much
thought and, thus, have no stored attitude to retrieve when a judgment is solic-
ited. Instead, they call upon their views of limited government to construct an
attitude. To the extent they favor small government they choose private insur-
ance plans. Conversely, a preference for activist government translates into sup-
port for the government option. The efficiency and utility of this strategy are
self-evident. Our citizens could think systematically about the relative merits
of these dueling proposals, but doing so would consume too much time and
tax their information-processing capacities. Far easier instead to infer a position
from a broader outlook on government power. Moreover, this simple rule of
thumb will likely yield the same decision they would have reached had they
deliberated at length about the matter.
The obvious question to pose at this juncture is how do we know reasoning
proceeds this way? Typically, analysts presume general principles are central
idea elements, issue preferences are peripheral, and political judgment fol-
lows a top-down path. These assumptions underlie the hierarchical model of
attitude constraint illustrated in Figure 4.1. Of course, an alternative causal
sequence whereby reasoning proceeds in a bottom-up fashion, or general and
specific attitudes shape one another simultaneously, might apply (Luskin,
1987). If attitude structures are organized along these lines, my claim that pol-
icy principles operate as decision-making heuristics cannot stand insofar as
principles should not be shaped by the types of judgments they are supposed
to explain. Rather than trying to sort out the competing claims via assump-
tion, I test rival models of attitude structure using opinion data.
Cross-sectional data are ill-suited to adjudicating between top-down and
bottom-up models of constraint. Because of this, scholars turn to panel data
The Centrality of Policy Principles 125
to unpack temporal connections between variables. If policy principles move

first in a causal sequence, they should facilitate change in issue positions over
time whereas issues induce no concomitant shift in principles. Results such as
these strongly imply (but do not prove) that principles occupy more central
nodes in political belief systems. Using this approach, Peffley and Hurwitz
(1993) discovered that beliefs about military power shape defense spending
attitudes without being shaped by them. Likewise, McCann (1997) found
moral traditionalism constrains abortion preferences but not the reverse.
These works render credible the proposition that domain-specific prin-
ciples vertically constrain preferences within a given issue area. To build the
case that the principles featured in this book are central heuristics, I under-
take a comparable set of analyses by applying structural equation modeling
(SEM) techniques to data from the 1992–1994–1996 NES panel survey. SEM
techniques allow the analyst to estimate dynamic relationships between
latent variables by separating out the stability of the variables from the
unreliability of the instruments. These techniques have been utilized in prior
research on dynamic constraint. I follow suit here by investigating whether
(1) latent policy principles dynamically constrain latent issue preferences
at time t, controlling for issues at time t – 1, and (2) latent issue preferences
constrain principles at time t, holding lagged principles constant. If prin-
ciples sway issue preferences while lying beyond their reach, my claim that
principles behave like central heuristics passes a crucial test. Conversely,
evidence that issues shape principles would nullify the assumption about
the centrality of principles.
Political Sophistication and Centrality

My claim is that nearly everyone relies on domain-specific principles when
formulating issue opinions. If politically aware and unaware citizens rely on
limited government, traditional morality, and military strength when think-
ing about related issues, my argument that these attitudes function as critical
heuristics for everyone will be strengthened. But if the ability to deploy these
principles is confined to especially sophisticated slices of the electorate, the
argument no longer holds.
What does extant research say about this? As noted throughout the book,
many scholars believe that the ability to deduce issues from principles depends
on prior knowledge. Sniderman et al. (1991) have remarked that “one should
obviously expect an interaction between political sophistication and cognition-
driven reasoning, such that the more politically sophisticated citizens are, the
more weight they are likely to attach to abstract cognitive considerations in
making up their minds about political choices” (p. 25). Zaller (1992) concurs:
“the impact of people’s value predispositions always depends on whether
citizens possess the contextual information needed to translate their values
into support for particular policies or candidates” (p. 25). Whereas Sniderman
and colleagues emphasize liberal-conservative orientations, Zaller concep-
tualizes value predispositions more broadly to include ideological and core
principles.
What does Zaller find? For the most part he lacks direct measures of
domain-specific principles on the NES surveys he uses to study opinion
change. To get around this problem he relies on liberal–conservative scales
as proxy measures for value predispositions under the assumption that
ideological self-placements correlate with values (Zaller, 1992, pp. 26–28,
344–345). The potential difficulty with this approach is that, as seen in
Chapter 4, the unsophisticated have little idea what ideological labels mean,
so it seems doubtful that liberal–conservative scales reflect value disposi-
tions for these respondents. This is precisely what I found in prior work
(Goren, 2001, pp. 162–163).
To underscore the point using the data at hand, Table 6.1 reports Pearson r
correlations between respondents’ self-placements on the seven-point liberal-
conservative scale and my three policy principle scales for low, medium,
and high sophistication respondents in the 1988–2008 NES surveys. Higher
values reflect increasingly conservative opinion on all measures, so positive
associations should be the rule. If the Zaller measurement strategy is defen-
sible, the correlations should be about the same across samples. The data
show instead that the associations grow stronger moving from less to more
informed samples. For instance, in 1988 the mean correlation between the lib-
eral–conservative and policy variables increases from .14 in the low group to
.21 in the middle group to .49 in the high group for a high–low difference of
.35. Likewise, in 2004 the respective correlation means equal .07, .36, and .66,
yielding a high–low gap of .59. Given the feeble values in the low knowledge
group, critics might question whether the lack of principle-based reasoning
among the unaware in Zaller’s study results from their lack of knowledge
or the lack of viable measurement instruments. In short, although this work
demonstrates that awareness motivates reliance on liberal–conservative orien-
tations, it cannot speak as directly to the question of whether the same holds
true for other abstract principles.
What happens when the sophistication interaction model is tested
using valid measures of core principles? Studies in this vein, which rely
on cross-sectional data, suggest political novices ground preferences
in crowning postures (Goren, 2004; Pollock, Lilie, & Vittes, 1993). This
TA B L E 6 . 1 . CORRELATIONS BETWEEN THE

LIBERAL–CONSERVATIVE SCALE AND POLICY
PRINCIPLES BY LEVEL OF SOPHISTICATION
1988:
Limited government .14 .19 .53 .39
Traditional morality .18 .30 .46 .28
Military strength .10 .14 .48 .38
Mean .14 .21 .49 .35
1992:
Mean .05 .27 .45 .40
1996:
Mean .09 .29 .46 .37
2000:
Mean .09 .40 .53 .44
2004:
Mean .07 .36 .66 .59
2008:
Limited government –.04 .19 .49 .53
Mean .01 .25 .55 .54
Notes: Cell entries represent Pearson product-moment correlations between the

seven-point liberal–conservative scale and policy principle scales. Scales are
coded to be positively related.
does not mean sophistication is irrelevant, as it tends to boost reliance on

some principles under some circumstances. However, even when condi-
tional effects exist, the unsophisticated usually invoke these predisposi-
tions (Goren, 2001). Such research substantiates my argument that policy
principles facilitate political choice for citizens along the sophistication
continuum. To assemble a more persuasive case here, I use NES data to
analyze whether limited government, traditional moral standards, and

military strength constrain issues for political sophisticates and novices.
Although panel data yield credible evidence on temporal dynamics, they
are less suited for sophistication interaction tests given the small number
of cases available across the panel waves and the lack of national secu-
rity variables. Fortunately, as we shall see below, the panel study results
justify specification of statistical models that treat general principles as
independent variables. As such, large N, cross-sectional data sets that
contain all necessary measures can be exploited.
THE DYNAMIC CONSTRAINT TESTS

The Limited Government and Health Care Model
My first task requires using 1992–1994–1996 panel data to determine what
causes what in the economic welfare domain. To provide a fair test of the com-
peting hypotheses, a high profile issue should be identified. Moreover, meas-
ures of latent principles and issue preferences must appear in multiple panel
waves to estimate the model. Given these constraints, I use the health care
issue because it featured prominently during the 1992 presidential campaign
and was debated vigorously during the early years of the Clinton presidency.
Put simply, health care reform was a high visibility issue.
I measure limited government using two additive scales per panel
wave and tap health care opinion with a seven-point scale. The latter asks
whether respondents favor a government insurance or private insurance
plan (see p. 90 in Chapter 5 for question wording). Because higher scores
reflect more conservative responses, the variables should move in tandem
as elaborated below.
Limited government94 = β1Limited government92 + β2Health care atttiudes94 + ζ1 (1)
Limited government96 = β3Limited government94 + β4Health care atttiudes96 + ζ2 (2)
Health care atttiudes94 = β5Health care atttiudes92 + β6Limited government94 + ζ3 (3)
Health care atttiudes96 = β7Health care atttiudes94 + β8Limited government96 + ζ4 (4)
In these equations the β1 and β3 estimates capture the effect of limited

government at time t – 1 on limited government at time t, controlling for
health care attitudes at time t. Likewise, β5 and β7 record the impact of
lagged health care attitudes on current attitudes, holding current limited
government constant. These coefficients reflect the stability of each latent
variable, controlling for the simultaneous influence of the other. Next, the
structural coefficients β2, β4, β6, and β8 denote the effect of one latent fac-
tor on another at a given point in time, ceteris paribus. If policy principles
drive issue preferences, limited governmentt will predict health caret, con-
trolling for health caret – 1 whereas health caret will not affect limited gov-
ernmentt, controlling for limited governmentt – 1. Conversely, if issues shape
principles, health caret will constrain limited governmentt, controlling for
limited governmentt – 1 and limited governmentt will not sway health caret,
holding issue prefrencest – 1 constant. All variables vary from zero to one,
which means the coefficients can be interpreted as the percentage change in
the dependent variable given movement across the full range of the inde-
pendent variable, holding the other factor constant. Finally, note that equa-
tions (1)–(4) reflect a synchronous specification. Comparable results emerge
when lagged values of principles (issues) are used to predict current values
of issues (principles).
Table 6.2 reports the model fit statistics and unstandardized parame-
ter estimates for the structural equations. To avoid clutter, measurement
model coefficients are omitted. The first column identifies the stability
coefficients, the structural coefficients, and fit statistics. The 1992–1994
estimates appear in column two, followed by the 1994–1996 estimates in
column three. To begin with global fit, the specified model reproduces the
data exceptionally well. The χ2 statistic is insignificant and the compara-
tive fit index (CFI), root mean square error of approximation (RMSEA), and
standardized root mean square residual (SRMR) values surpass the desired
cut points.
Moving onto the parameter estimates, I find that limited government is
quite stable over the first wave (β1 = 0.74) and remarkably stable over the
second (β3 = 0.93). In contrast, health care preferences fluctuate a lot between
1992 and 1994 (β5 = 0.48) before stabilizing thereafter (β7 = 0.69). Formal testing
reveals limited government is probably more durable than health care over
the first panel wave (p = .06) if not necessarily the second (p = .12). In terms
of the structural coefficients, limited government affects health care attitudes
in 1994 but not 1996.1 In 1994, the strongest advocates for small government
1
The health care reform issue received tremendous media coverage from 1992 to
1994, only to be ignored afterward. As a colleague and I report elsewhere, 561
health care reform stories appeared on major television networks during the first
wave of the NES panel versus 10 during the second (Dancey & Goren, 2010).
Given these coverage patterns, the finding that people updated only during
a time of intense political debate makes sense.
TA B L E 6 . 2 . LIMITED GOVERNMENT–HEALTH CARE

OPINION MODELS, 1992–1994–1996
1992–1994 1994–1996
Stability coefficients:
Limited government 0.74* 0.93*
(.12) (.20)
Health care preference 0.48* 0.69*
(.12) (.13)
Structural coefficients:
Health care → Limited government 0.06 0.05
(.08) (.13)
Limited government → Health care 0.72* 0.24
(.21) (.25)
Model fit:
Yuan–Bentler scaled χ2 25.18
Degrees of freedom 18
p-value .12
CFI .99
RMSEA .03
SRMR .03
Number of observations 412
*p < .05 (one-tailed).
Notes: Maximum likelihood estimates are based on raw data. Unstandardized

coefficients are reported. Robust standard errors are in parentheses. All
latent variables are on a 0–1 scale. Structural coefficients should be positively
related. NES sample weights have been used. Measurement model estimates
are not shown. CFI, comparative fit index; RMSEA, root mean square error of
approximation; SRMR, standardized root mean square residual.
score far more conservative on health care than government defenders, hold-
ing lagged health care preferences constant (p < .01). The limited government
coefficient remains sizable in 1996, but falls short of statistical significance. In
contrast, health care shapes limited government in neither interval. The struc-
tural coefficients cannot be reliably distinguished from zero and are substan-
tively trivial. Put simply, limited government dynamically constrains health
care preferences without being constrained by them. The broader implica-
tions should be clear. Attitudes toward limited government lie closer to the
belief system core whereas health care preferences probably reside further
out in the periphery.
The Traditional Morality and Abortion Model

Moving onto the cultural domain, I assess the ties between traditional moral-
ity and abortion preferences. Abortion is arguably the most explosive culture
war issue and thus merits close attention. I measure moral traditionalism
as described in Chapter 5. To tap abortion attitudes, I employ a four-point
scale with response options ranging from extremely pro-life (“By law, abor-
tion should never be permitted”) to extremely pro-choice positions (“By law,
a woman should always be able to obtain an abortion as a matter of personal
choice”). Higher values reflect traditional and pro-choice sympathies on
the respective variables; therefore, they should be inversely related. I estimate
the models specified in equations (1)–(4) with the cultural variables taking the
place of the economic welfare variables.
Table 6.3 contains the estimates. First, model fit leaves almost nothing to
be desired. The χ2 statistic is insignificant and the goodness of fit values are
impressive. Second, both types of policy attitudes hold very steady over time.
The stability coefficients for traditionalism are 0.78 from 1992 to 1994 and
0.94 from 1994 to 1996. The abortion analogues equal 0.98 and 0.88. Abortion
proves more stable over the first interval (p < .05). Third, moral standards con-
strain abortion preferences in 1996, controlling for 1994 preferences (p < .05).
The –0.16 coefficient means the most orthodox respondents in 1996 score 16%
lower on abortion rights than the most progressive respondents. No compa-
rable effect appears over the first wave. Fourth, abortion preferences fail to
shape attitudes toward moral standards. The abortion coefficient approaches
significance in 1994 (p < .10), but there is little practical effect to speak of as
pro-choice respondents score 5% lower on latent traditionalism than pro-life
respondents. In 1996 the variables do not appear to be related. To sum up, the
SEM estimates largely corroborate the hypothesis that principles shape issues
while yielding little evidence consistent with the reverse causal chain.
Overall, the statistical evidence reported in Tables 6.2 and 6.3 affirms my
prediction that policy principles shape issue attitudes in a hierarchical fash-
ion. In contrast, the rival hypothesis that issue preferences shape broader
principles fares poorly. These results, along with those reported by Peffley
and Hurwitz (1993) and McCann (1997), support the inference that policy
principles vertically constrain issue preferences and, presumably, related
political judgments. Attitudes toward limited government, moral standards,
and force versus diplomacy, in other words, function like political heuristics.
But can everyone deploy these or is usage restricted to those who know
a lot about politics? The above estimates ignore this complication and there-
fore provide no insight into the matter. The following section addresses this
question head on.
TA B L E 6 . 3 . TRADITIONAL MORALITY–ABORTION
OPINION MODELS, 1992–1994–1996
1992–1994 1994–1996
Stability coefficients:
Traditional morality 0.78* 0.94*
(.06) (.07)
Abortion 0.98* 0.88*
(.06) (.06)
Structural coefficients:
Abortion → Traditional morality –0.05+ 0.03
(.04) (.04)
Traditional morality → Abortion –0.02 –0.16*
(.08) (.08)
Model fit:
Yuan–Bentler scaled χ2 18.79
Degrees of freedom 20
p-value .54
CFI 1.00
RMSEA .00
SRMR .01
Number of observations 490
+
p < .10; *p < .05 (one-tailed).
Notes: Maximum likelihood estimates are based on raw data. Unstandardized

coefficients are reported. Robust standard errors are in parentheses. All latent
variables are on a 0–1 scale. Structural coefficients should be negatively
related. NES sample weights have been used. Measurement model estimates
are not shown. CFI, comparative fit index; RMSEA, root mean square error of
approximation; SRMR, standardized root mean square residual.
THE SOPHISTICATION INTERACTION TESTS

Limited Government and Health Care
I argued that the propensity to derive issue attitudes from general principles
escapes few people. The sophistication interaction model, of which there are
two variants, maintains that prior knowledge conditions reliance on such
cues. The first variant of the model avers that the sophisticated alone take
advantage of these heuristics. Lacking the contextual knowledge needed to
understand the implications their principles have for specific issues, the unin-
formed make it up as they go along when asked to report an issue opinion
(Zaller, 1992). The less extreme version of the hypothesis agrees that sophisti-
cation sharpens principle-opinion ties, but leaves open the possibility that the
unaware manipulate core principles, albeit with less felicity than their more
sophisticated counterparts.
One seemingly straightforward way to test these ideas would be to rees-
timate the dynamic constraint models for respondents at different levels of
sophistication. Despite its intuitive appeal, this is not a viable strategy for
a few reasons. First, the samples contain about 400–500 observations and
thus are of marginal utility for generating subgroup estimates via struc-
tural equation modeling techniques. Were I to trichotomize my data as
before, the samples would contain about 150 or so cases per group, too few
to generate stable and efficient parameter estimates. Second, the absence
of hawk–dove items precludes testing in the foreign policy and national
security domain. Fortunately, because the results reported in Tables 6.2
and 6.3 point to the temporal priority of abstract principles, I can use the
1988–2008 NES cross-sectional surveys to estimate the statistical models
without having to worry about simultaneity bias.
My analyses revolve around relationships between principles and issue
preferences for which the temporal ordering has been established above or
elsewhere (Peffley & Hurwitz, 1993). Although I would anticipate observing
comparable results for other issues, space limitations dictate a more focused
approach in these pages. The first test features the limited government to
health care link. To begin with the dependent variable, I tap health care opin-
ion using the seven-point scale described above with higher scores denoting
preferences for private insurance plans. Next, limited government is measured
using whatever combination of spending items appears on a given survey.
After identifying the items, I calculate a “person mean” score for respondents
who answered over half the questions used to tap a given attitude (minimum
of three questions asked). To illustrate, five applicable items appear on the
1988 NES. For individuals who answered three of the five, their limited gov-
ernment score equals the mean response for the three items. Similarly, par-
ticipants who answered four questions receive a score based on the average
of those four responses, whereas those who answered all five questions have
their positions calculated accordingly. Respondents who answered less than
half the items are treated as missing data. This procedure allows me to preserve
a small number of cases that would otherwise be dropped from the analysis.
The person mean approach works well when the items tap a single latent con-
struct, scale reliability is high, and over half the available items are employed
(Bernaards & Sijtsma, 1999; Schafer & Graham, 2002). All three conditions
are satisfied here.2 Because higher values on limited government and health
care reflect increasingly conservative sentiments, the variables should be pos-
itively related in the statistical models, which are estimated using ordinary
least squares (OLS) regression.
A standard set of control variables is incorporated into the equations. First,
I code “65 or older” 1 for individuals who meet this criterion and 0 otherwise.
Because most such respondents have Medicare coverage, they should react
coolly to private market schemes. The regression coefficient should thus be
negative. Next, “unemployed” is coded 1 for unemployed respondents and 0
otherwise. Because the unemployed typically lack health insurance, they too
might express qualms about private insurance. A negative coefficient is antici-
pated here as well. Third, party identification is gauged using the seven-point
self-categorization scale. Higher values denote stronger GOP attachments, so
the regression coefficient should be positive. Next, liberal–conservative atti-
tudes are tapped with the seven-point liberal–conservative self-placement
item. As I argued in Chapter 4, sophistication should drive the use of liberal–
conservative attitudes. To test this, the knowledge indicator and its multiplica-
tive product with the left–right scale are added to the model. The coefficient for
the constitutive liberal–conservative term summarizes this variable’s impact
on health care opinion when the sophistication variable equals zero—the
minimum knowledge score.3 Because the least informed usually do not hold
meaningful liberal–conservative attitudes (see Chapter 4) this term should be
insignificant. In contrast, the multiplicative term should be positive and signif-
icant, indicating that the bond between conservative affinities and preferences
for private insurance strengthens with sophistication. Fifth, I tap political trust
using the standard multi-item scale. Following Hetherington (2005), I expect
2
I apply this coding rule to all other multiple indicator attitude scales composed
of three or more items. In each case, the items tap a single concept, the scale is
reliable, and over half the available items are employed in the construction of the
scale. Recall from Chapter 5, Table 5.12, that rates of item nonresponse are, for
the most part, comparable across levels of sophistication. Therefore, the person
mean approach does not routinely drop less sophisticated respondents from the
estimation samples. Note finally that the health care preference variable, along
with the defense spending item used later on, include no opinion filters that
screen out some less sophisticated respondents.
3
Constitutive terms are often interpreted as the main or average effect of the
predictor in question. This is a mistake. These coefficients, again, estimate the
predictor’s effect on the dependent variable when its interaction partner equals
zero, ceteris paribus (Friedrich, 1982).
that increasing trust in government will undercut support for private insur-
ance, so the trust coefficient should be negative.
The key question, to reiterate, is whether the sophisticated alone call
upon policy principles to guide their issue evaluations. To test this, a lim-
ited government × sophistication multiplicative term is added to the base-
line to create a second model. If the stronger version of the sophistication
interaction hypothesis applies, the coefficient for the multiplicative term
will be positive and significant whereas that of the constitutive limited gov-
ernment variable will be insignificant. The latter result would suggest that
the least aware fail to ground health care preferences in abstract postures
about government. If the weaker sophistication prediction holds, the con-
stitutive and multiplicative terms will jointly achieve significance. In that
case, the unaware would make use of limited government, but not to the
same degree as the aware.
Table 6.4 reports the 1988–1996 model estimates. Table 6.5 does the same
for 2000–2008.4 Model 1 includes all variables except the limited government ×
sophistication term, which appears in model 2. Because all variables are scored
on a 0–1 range, the coefficient for each additive term can be interpreted as the
percentage difference on health care opinion between those with the mini-
mum and maximum value on the predictor in question, holding all else con-
stant. The constitutive terms should be interpreted as described above. The
multiplicative terms are elaborated below.
To start with the control variables, age and unemployed perform poorly,
whereas party has the predicted effect in five of the six equations. The trust
variable is usually significant as well, though signed incorrectly, the posi-
tive coefficients implying political trust motivates support for private plans.
Next, as predicted, sophistication stimulates ideological evaluation. The
politically engaged usually rely more on ideological leanings than the unen-
gaged. When conditional slope estimates are calculated for respondents one
4
The NES uses a complex, multistage area probability design to sample respon-
dents from the population. Because such designs affect the accuracy of parameter
estimates and standard errors and the conventional formulas for estimators and
standard errors assume the data are collected using simple random sampling
techniques, adjustments should be made to get the point estimates and standard
errors right (Heeringa, West, & Berglund, 2010). I do so here by incorporating
postelection weight, cluster, and strata variables into the estimation. There is no
weight variable in the 1988 NES, but clustering and strata information are taken
into account.
TA B L E 6 . 4 . LIMITED GOVERNMENT AND HEALTH CARE OPINION, 1988–1996
1988 1988 1992 1992 1996 1996
Model 1 Model 2 Model 1 Model 2 Model 1 Model 2
Constant 0.27* 0.30* 0.20* 0.25* 0.28* 0.31*

(.08) (.08) (.06) (.06) (.10) (.10)
65 or older –0.04 –0.04 0.01 0.01 –0.00 –0.00
(.03) (.03) (.02) (.02) (.02) (.02)
Unemployed –0.02 –0.03 –0.07* –0.07* 0.01 0.01
(.05) (.05) (.04) (.04) (.07) (.07)
Party identification 0.12* 0.12* 0.14* 0.14* 0.14* 0.14*
(.03) (.03) (.03) (.03) (.03) (.03)
Liberal–conservative identification 0.00 0.06 –0.01 0.02 –0.04 –0.00
(.13) (.14) (.11) (.11) (.14) (.15)
Political trust 0.06+ 0.06+ 0.08* 0.08* 0.03 0.03
(.04) (.04) (.04) (.04) (.05) (.05)
Limited government 0.41* 0.17+ 0.35* 0.16* 0.30* 0.16
(.04) (.12) (.04) (.07) (.05) (.14)
Political sophistication –0.05 –0.10 –0.16* –0.24* –0.13 –0.18
(.12) (.12) (.08) (.09) (.12) (.12)
Liberal–conservative identification × sophistication 0.19 0.06 0.31* 0.22+ 0.36* 0.29
(.20) (.22) (.16) (.16) (.22) (.24)
Limited government × sophistication 0.43* 0.35* 0.22
(.17) (.13) (.21)
F statistic 35.56* 37.17* 57.42* 51.61* 48.71* 44.33*
R2 .15 .16 .19 .19 .22 .22
Number of observations 1321 1321 1790 1790 1277 1277
+
p < .10; *p < .05 (one-tailed).
Notes: Ordinary least squares (OLS) regression estimates. Standard errors are in parentheses (adjusted for complex sample design). All variables are on a 0–1
range. Higher values on health care reflect conservative opinion.

TA B L E 6 .5 . LIMITED GOVERNMENT AND HEALTH CARE OPINION, 2000–2008
2000 2000 2004 2004 2008 2008
Constant 0.12 +
0.12 +
0.24* 0.29* 0.17* 0.18*
(.09) (.09) (.08) (.08) (.07) (.08)
65 or older 0.03 0.03 0.06* 0.07* 0.04+ 0.05+
(.03) (.03) (.03) (.03) (.03) (.03)
Unemployed –0.03 –0.03 –0.12* –0.12* –0.02 –0.03
(.06) (.06) (.04) (.04) (.04) (.04)
Party identification 0.14* 0.14* 0.04 0.04 0.17* 0.17*
(.04) (.04) (.05) (.05) (.05) (.05)
Liberal–conservative identification 0.03 0.03 –0.04 0.01 0.05 0.05
(.15) (.16) (.14) (.13) (.13) (.13)
Political trust 0.10* 0.10* 0.14* 0.14* 0.09* 0.09*
(.05) (.05) (.04) (.04) (.03) (.04)
Limited government 0.39* 0.38* 0.22* –0.08 0.37* 0.31*
(.05) (.12) (.06) (.15) (.07) (.14)
Political sophistication 0.04 0.04 –0.17* –0.25* –0.18* –0.20*
(.13) (.12) (.10) (.10) (.10) (.11)
Liberal conservative id × sophistication 0.24 0.23 0.59* 0.46* 0.37* 0.35*
(.22) (.23) (.19) (.19) (.18) (.18)
Limited government × sophistication 0.01 0.50* 0.09
(.17) (.20) (.22)
F statistic 27.82* 24.59* 49.18* 48.72* 37.32* 32.90*
R2 .20 .20 .24 .24 .27 .27
p < .10; *p < .05 (one-tailed).

+
Notes: OLS regression estimates. Standard errors are in parentheses (adjusted for complex sample design). All variables are on a 0–1 range. Higher
values on health care reflect conservative opinion.

standard deviation above the knowledge mean, the liberal–conservative

effect is always statistically significant (p < .05) and substantively large.
Moreover, the constitutive liberal–conservative term attains significance in
no model. The same holds true for respondents a standard deviation below
the knowledge mean. Consistent with prior work, the politically informed
alone make extensive use of the liberal–conservative heuristic (Sniderman
et al., 1991; Zaller, 1992).
What about limited government? Here I find a varied pattern of results.
First, limited government affects health care opinion equally for all respon-
dents in 1996, 2000, and 2008. For these years the additive coefficient in
model 1 achieves significance (at p < .01 or better) whereas the limited gov-
ernment × sophistication terms fall short in model 2. In practical terms, in
1996 those with the maximum limited government score are 30% more con-
servative on health care opinion than those with the minimum score, cete-
ris paribus. Comparable results emerge in 2000 (39% difference) and 2008
(37% difference). The magnitude of these relationships can be illustrated in
graphic form. Figure 6.1 displays the impact movement from the minimum
to maximum score on the limited government variable has on predicted
health care opinion in 1996 and 2008. As indicated in the figure, respon-
dents most enthusiastic about limited government take significantly more
conservative positions on health care reform than those who favor activ-
ist government. Again, these differences lie in the 30–37% range. Note that
relatively few cases fall in the extreme conservative end of the scale (see
Figures 5.2 and 5.3). As such, the simulations are best viewed as illustrative
Support for Private Health Insurance
1.00
0.75
0.50
0.25
1996 2008
0.00
Oppose Favor
Limited Government
Figure 6.1. Limited Government and Health Care Opinion, Additive Effects.
Source: Table 6.4-6.5 estimates; other predictors held at mean/mode. Effects significant.
of the hypothetical maximum effect. More broadly, limited government can

serve as a dependable heuristic for most people.
But it does not always serve as such to the same degree. In 1988, 1992, and
2004, the significant multiplicative term in model 2 reveals that sophistication
facilitates vertical constraint. This can be illustrated by comparing the lim-
ited government slope for respondents scoring one standard deviation over
and under the knowledge mean. In 1988, going from the progovernment to
antigovernment extreme on limited government increases support for private
insurance by 27% for the less aware and 47% for the more aware (both p < .01).
Similar results hold for 1992 (24% and 43% for the respective groups, both p <
.01). Figure 6.2 illustrates the 1992 effect. Moving from left to right on limited
government is associated with a somewhat sharper rise in support for private
market insurance among those with above average knowledge scores relative
to those with below average knowledge scores. For 2004, complete movement
across limited government leads to a 5% rise in support for private health care
for novices (p ns) and a 33% rise for experts (p < .01). The end result is that
with the exception of 2004 limited government shapes health care opinion for
almost everyone, though sophistication tightens connections in 3 of 6 years.
Limited government clearly matters, but it is not the only heuristic that
might be invoked. If this principle really is a central belief system element, it
should hold more sway over health care preferences than other predisposi-
tions. To test this possibility, I compare the limited government effect to that
of party identification, which is seen by many as the ultimate political heuris-
tic. To illustrate with the 2008 estimates (model 1), the coefficient for limited
Support for Private Health Insurance
1.00
0.75
0.50
0.25
Low Sophistication High Sophistication
0.00
Oppose Favor
Limited Government
Figure 6.2. Limited Government and Health Care Opinion, 1992 Interactive Effects.
Source: Table 6.4 estimates; high/low soph = +/– 1 sd from mean; other predictors held at
mean/mode. Effects significant.
government is over twice that of party (0.37 > 0.17). Of course, visual inspection
cannot determine whether such a difference exceeds what might be expected
by chance. To see if the difference is statistically significant, I constrain the two
coefficients to be equal and assess the test result.5 The test statistic indicates
the null hypothesis of equal effects can be rejected with a high degree of con-
fidence (p < .05). Put otherwise, the 0.37 coefficient for limited government is
demonstrably larger than the 0.17 coefficient for party. The same holds true for
the other 2 years in which a significant main effect arises. Limited government
has a greater impact on health care opinion than party in 1996 (0.30 > 0.14, p <
.01) and 2000 (0.39 > 0.14, p < .001). The standardized coefficient for limited
government also surpasses that of party identification for all three of the years
(.22 > .16 in 1996; .24 > .13 in 2000; .22 > .18 in 2008).
Turning next to the equations with a significant limited government ×
sophistication term, I calculate conditional slope estimates for limited
government for those plus or minus one standard deviation from the
knowledge mean and compare these to the party coefficient. Among the
less informed, the limited government coefficient exceeds that of parti-
sanship in 1988 (0.27 > 0.12) and 1992 (0.24 > 0.14), but these differences
are not quite significant (p88 = .15 and p92 = .08). In 2004 neither variable
systematically affects health care opinion among novices. The standard-
ized betas tell a similar story as limited government performs marginally
better than party (.16 > .13 in 1988, .16 > .15 in 1992, .04 < .05 in 2004). For
those with above average knowledge, the regression coefficient for lim-
ited government tops party at p < .01 in 1988 (0.47 > 0.12), 1992 (0.43 >
0.14), and 2004 (0.33 > 0.04). The standardized coefficients reinforce the
point (.29 > .13 in 1988, .29 > .15 in 1992, .24 > .05 in 2004). Collectively,
these results show that for the less sophisticated limited government
either equals or, more often, surpasses the party cue, and it always bests
party among the sophisticated. Party identification may indeed be the best
5
The test result is not influenced by the fact that movement from 0 to 1 spans
the full range of the variables. If limited government and party identification
are scaled another way (e.g., each set to a seven-point scale), the test evalu-
ates whether a one unit change on the two predictors is statistically equivalent.
Although the magnitude of the unstandardized regression coefficient will vary
depending on the variables’ ranges, for either scaling the equality constraint test
result will be identical. Some may prefer to assess the relative importance of dif-
ferent predictors using standardized regression coefficients. Although there are
problems with this approach (see Achen, 1982), I comment on the standardized
coefficients for interested readers.
predictor of many things, but it is not the best predictor of health care
opinion.6 These results underscore limited government’s centrality in mass
belief systems.
Some might wonder how limited government stacks up against the
trust heuristic. As seen above, political trust significantly affects health
care opinion but in the wrong direction. Those who trust government
evaluate private insurance more favorably than the cynics. When I con-
strain the limited government and trust coefficients to be equal across the
various equations, trust always lags limited government in sophisticated
and unsophisticated samples (with one exception) at conventional levels of
significance (the same holds true when standardized betas are compared).
What about ideological labels? Here too, limited government matches or
surpasses the force of liberal–conservative attitudes among politically
informed respondents. To be precise, the limited government coefficient
is significantly larger than the liberal–conservative analogue in two equa-
tions, approaches a significant difference in a third, and is not significantly
different in the remaining three. With respect to the standardized coeffi-
cients, limited government exceeds the liberal–conservative variable, at
times substantially, from 1988 to 2000, and lags it a bit in 2004 and 2008.
In sum, limited government often exerts more influence over health care
attitudes than other notable cues.
On the whole, the evidence buttresses my argument that limited govern-
ment serves as an effective heuristic for politically sophisticated and unsophis-
ticated individuals. By implication, the strong version of the sophistication
interaction model, which posits that the unaware cannot deduce issue attitudes
from broader principles, commands little support. However, the results lend
some credence to the weaker statement of this model. Although the absence of
sophistication does not preclude the use of core principles, its presence often
helps. Said otherwise, limited government shapes health care opinion most of
the time, but sometimes carries more weight for the better informed.
Traditional Morality and Abortion

Next up we have the relationship between moral standards, political sophis-
tication, and abortion preferences. Traditional morality is measured as before
6
My account focuses on direct effects alone and thus ignores the potential indirect
influence party may wield over issue preferences through its effects on limited
government. I take up the effects of party ties on policy principles in Chapter 7,
but I note here that basic human values weigh more heavily on policy principles
than partisanship.
with higher values denoting orthodox views. I use an assortment of questions

to tap abortion attitudes including the four-point item employed earlier and
whatever related items appear on a given survey (e.g., spousal notification,
parental notification for minors, and government funded abortion for poor
women). The four-point item serves as the dependent variable in the 1988 and
1996 equations; multiple indicator scales perform the duty in the other years.
Higher scores reflect pro-choice commitments, so the principle and preference
variables should move in the opposite direction.
As for the control variables, I fall back on age, measured in years, and
dummy variables for female and born again Christian (the latter is unavailable
in 2004). The age and born again predictors should depress abortion approval
whereas female should increase support. Next, party identification should
be inversely related to abortion opinion (recall higher values signify stronger
Republican attachments). In addition, authoritarianism should undercut abor-
tion support. I tap this concept using the standard four-item child rearing scale
(see Hetherington & Weiler, 2009, pp. 47–48), which is scored so higher values
denote increasing authoritarianism (this variable does not appear on the 1988 or
1996 surveys). Next, liberal–conservative identities are interacted with political
knowledge. The multiplicative coefficient should be negative and significant,
meaning the highly engaged make greater use of ideological predispositions
than the less engaged. The constitutive liberal–conservative term should not
differ significantly from zero, which would imply that among the unaware
abortion opinions are not constrained by liberal–conservative views. Finally,
I insert a traditional morality × sophistication multiplicative term into the base-
line equation. If the stronger variant of the sophistication interaction model is
valid, the coefficient for the product term should be negative and statistically
significant, with the constitutive term close to zero. Conversely, if the weaker
hypothesis holds, both coefficients should be negative and significant.
The 1988–1996 estimates are listed in Table 6.6, followed by the 2000–2008
estimates in Table 6.7. Because the 1988 and 1996 dependent variables are four-
point measures, ordered logistic regression is employed these years; otherwise
OLS estimation is used. To begin, the control variables perform unevenly. On
the positive side, born again Christians are always more critical of abortion
than those from other groups. Also, authoritarians consistently evaluate abor-
tion more negatively than libertarians. Next, the liberal–conservative identifi-
cation × sophistication variable is negative and significant across the models,
suggesting ideological sentiments increasingly shape abortion opinion at
higher levels of awareness. The ideological variable behaves as predicted for
respondents one standard deviation above the knowledge mean and remains
insignificant for those a standard deviation below it. The other controls do not
TA B L E 6 . 6 . TRADITIONAL MORALITY AND ABORTION OPINION, 1988–1996
1988 1988 1992 1992 1996 1996
Constant 0.60* 0.50*

(.05) (.06)
Age –0.004+ –0.004* 0.00 0.00 –0.01* –0.01*
(.003) (.003) (.00) (.00) (.00) (.00)
Female 0.19* 0.18+ 0.06* 0.05* 0.20* 0.18*
(.11) (.11) (.01) (.01) (.11) (.11)
Born again Christian –0.80* –0.79* –0.07* –0.07* –0.74* –0.72*
(.11) (.12) (.01) (.01) (.12) (.12)
Party identification 0.30* 0.31* –0.05* –0.05* 0.15 0.17
(.16) (.17) (.02) (.02) (.23) (.23)
Liberal–conservative identification 0.36 0.13 0.08 0.04 –0.32 –0.58*
(.49) (.53) (.08) (.08) (.84) (.85)
Authoritarianism –0.13* –0.12*
(.03) (.03)
Traditional morality –2.73* –1.61* –0.39* –0.18* –2.46* –0.28
(.24) (.63) (.03) (.07) (.39) (1.09)
Political sophistication 2.40* 3.36* 0.47* 0.62* 3.08* 4.82*
(.54) (.60) (.08) (.09) (.80) (1.25)
Liberal–conservative identification × sophistication –2.22* –1.71* –0.47* –0.37* –2.72* –2.12*
(.88) (.99) (.15) (.16) (1.29) (1.30)
(continued)
1988 1988 1992 1992 1996 1996
Traditional morality × sophistication –2.08* –0.36* –3.50*

(1.03) (.13) (1.53)
μ1 –3.50 –2.97 –3.53 –2.41
μ2 –1.40 –0.88 –1.58 –0.45
μ3 –0.49 0.03 –0.78 0.35
F statistic 27.59* 28.17* 131.46* 114.92* 63.54* 52.17*
R2 – – .33 .33 – –
+
p < .10; *p < 0.05 (one-tailed).
Notes: Ordered logistic regression estimates for 1988 and 1996. OLS regression estimates for 1992. Standard errors are in parentheses (adjusted for complex
sample design). All variables are on a 0–1 range. Higher values on abortion reflect pro-choice opinion.

TA B L E 6 . 7 . TRADITIONAL MORALITY AND ABORTION OPINION, 2000–2008
2000 2000 2004 2004 2008 2008
Constant 0.66* 0.65* 0.70* 0.61* 0.63* 0.58*

(.06) (.07) (.07) (.08) (.05) (.07)
Age –0.00 –0.00 0.00 0.00 0.001* 0.002*
(.00) (.00) (.00) (.00) (.00) (.00)
Female –0.03* –0.03* –0.04* –0.04* –0.04* –0.04*
(.01) (.01) (.02) (.02) (.01) (.02)
Born again Christian –0.08* –0.08* –0.13* –0.13*
(.02) (.02) (.02) (.02)
Party identification 0.02 0.02 –0.06+ –0.06+ –0.03 –0.02
(.03) (.03) (.04) (.04) (.03) (.03)
Liberal–conservative identification 0.08 0.07 0.29* 0.20* 0.01 –0.02
(.07) (.08) (.11) (.11) (.08) (.08)
Authoritarianism –0.10* –0.10* –0.19* –0.19* –0.08* –0.07*
(.02) (.02) (.05) (.05) (.03) (.03)
Traditional morality –0.36* –0.34* –0.50* –0.25* –0.27* –0.16+
(.04) (.10) (.06) (.15) (.06) (.11)
(.07) (.08) (.08) (.11) (.07) (.09)
(continued)
2000 2000 2004 2004 2008 2008
Liberal–conservative × sophistication –0.54* –0.52* –0.71* –0.51* –0.36* –0.30*

(.11) (.12) (.16) (.17) (.12) (.12)
Traditional morality × sophistication –0.04 –0.43* –0.17+
(.14) (.21) (.13)
F statistic 57.85* 52.01* 66.17* 72.86* 54.09* 55.45*
R2 .30 .30 .26 .27 .35 .35
+
p < .10; *p < .05 (one-tailed).
Notes: OLS regression estimates. Standard errors are in parentheses (adjusted for complex sample design). All variables are on a 0–1 range.
Higher values on abortion reflect pro-choice opinion.

conform to expectations. Age is associated with abortion opposition in 3 of

6 years. Women prove more supportive of abortion rights than men from 1988
to 1996, but reverse course thereafter by expressing mildly greater opposition
than men. Party identification behaves erratically as well, as Republicans take
a harder line against abortion than Democrats only in 1992 and 2004.
Next, I find that traditional moralism structures abortion opinion for cit-
izens in general and the more sophisticated in particular. To elaborate, the
traditionalism × sophistication term in model 2 approaches or achieves sig-
nificance every year but one (2000). Just as importantly, the coefficient for the
constitutive term does so as well in four of these five models (it misses in
1996). And in the one equation in which the interaction term fails, the additive
term proves highly significant (see model 1 in 2000). To reiterate, although tra-
ditional morality vertically constrains abortion attitudes for nearly everyone,
sophistication usually fortifies these bonds.
To illustrate effect sizes, I manipulate the constitutive and multiplicative
coefficients in some of the models to calculate predicted abortion support scores
for those plus or minus one standard deviation relative to the knowledge mean.
For 1992, I find that movement up the traditionalism scale depresses abortion
support by 27% for individuals a standard deviation below the mean (p < .001)
versus 46% for those lying a standard deviation above it (p < .001). For 2004, the
respective differences equal 37% for the less aware and 61% for the more aware
(both p < .001). Figure 6.3 illustrates the effect graphically for 2004. As we can
see, attitudes toward moral traditionalism shape abortion opinion for the polit-
ically aware and unaware, albeit more so for the former. Figure 6.4 shows the
same pattern for 2008. Among those scoring one standard deviation below the
knowledge mean, movement from the most progressive to the most orthodox
1.00
Support for Abortion Rights
0.75
0.50
0.25

0.00
Oppose Favor
Figure 6.3. Traditional Morality and Abortion Opinion, 2004 Interactive Effects.
mean/mode. Effects significant
Support for Abortion Rights 1.00
0.75
0.50
0.25

0.00
Oppose Favor
Figure 6.4. Traditional Morality and Abortion Opinion, 2008 Interactive Effects.
mean/mode. Effects significant.
stance on traditional morality depresses abortion support by 19%. For those

a standard deviation over the mean, the corresponding effect is 30%.
Moral attitudes perform well in predicting abortion opinion. The case for
its centrality in cultural attitude hierarchies can be strengthened by assessing
its impact relative to other judgmental yardsticks. I do so by constraining the
tradition and party coefficients to be equal across the equations and evaluat-
ing the test results. For the five equations containing a significant tradition ×
sophistication interaction, I employ the same protocol used above. In every
year, traditionalism shapes abortion opinion more so than party identification
for cases plus or minus one standard deviation from the knowledge mean
(all p < .05 except one case for the unsophisticated where p < .10). To take an
example, in 2004 the impact of traditionalism on abortion opinion outpaces
that of party for the sophisticated (p < .01) and unsophisticated (p < .01). Note
also that in 2000 the main effect of traditionalism is greater than partisanship
(p < .01). Given the inconsistent performance turned in by party as a predic-
tor of abortion opinion, these results may not be terribly surprising, though
they are worth underscoring given that the difference between significant and
insignificant coefficients is itself often insignificant (Gelman & Stern, 2006).
The same pattern emerges when I compare traditionalism to other psycho-
logical variables. Among the less knowledgeable, traditionalism either meets
or eclipses the effect of authoritarianism. Among the more informed, tradi-
tionalism nearly always outpaces authoritarianism. For politically informed
respondents traditionalism equals the ideological cue in half the equations
and exceeds it in the other half. For the standardized coefficients the tradi-
tion variable always outperforms party and authoritarianism for the sophisti-
cated and unsophisticated alike, usually by a wide margin. For sophisticated
respondents, the tradition variable matches or surpasses the ideology coef-

ficient. Once again, the germane policy principle usually promotes vertical
constraint more so than other heuristics.
The conclusion I draw from these results is that regardless of how much or
little everyday Americans know about public affairs, they do a pretty good job
in deducing abortion opinions from abstract views about moral standards. This
reaffirms my primary claim that core principles drive issue opinions through-
out the mass public. But consistent with the milder form of the sophistication
interaction model, politically engaged citizens demonstrate more skill in this
case. Like limited government, then, traditional morality serves as a central
heuristic for citizens in general and the politically sophisticated in particular.
Military Strength and Defense Spending

My analysis concludes with the military strength and defense spending
connection. I measure the hawk–dove dimension using all available indi-
cators on each NES survey and tap defense spending with the standard
seven-point query: “Some people believe that we should spend much
less money for defense. Others feel that defense spending should be
greatly increased. Where would you place yourself on this scale, or have-
n’t you thought much about this?” Because higher values reflect hawkish
responses, the regression coefficient for military strength should be posi-
tive. Many of the variables described above reappear as statistical controls
in the equations. I start with the female and born again Christian variables.
The former should be negatively related to defense spending and the latter
should be positively related. Party identification and authoritarianism are
used as well. Because higher values denote more conservative responses
on each predictor, they should generate pro-defense sentiment. The liberal–
conservative identification × sophistication variable makes another
appearance. The constitutive term should be insignificant and the multi-
plicative term positive and significant. This pattern would establish that
the least informed shun the liberal–conservative heuristic favored by the
more informed. To finish, I add a military strength × sophistication vari-
able to the baseline model to learn if sophistication boosts reliance on the
principle. If the constitutive and multiplicative terms attain significance,
the evidence will favor the weaker variant of the sophistication interaction
model. If only the multiplicative term does so, the stronger form of the
model garners support.
The OLS regression estimates appear in Table 6.8. Recall that data are avail-
able in 4 of 6 years (2000 and 2004 are missing). The results can be summarized
quickly. First, among the additive variables partisanship alone performs as
expected with Republicans rendering more positive evaluations of defense
TA B L E 6 . 8 . MILITARY STRENGTH AND DEFENSE SPENDING OPINION, 1988–2004
1988 1988 1992 1992 1996 1996 2004 2004
Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2
Constant 0.33* 0.35* 0.21* 0.22* 0.41* 0.43* 0.37* 0.40*

(.06) (.07) (.05) (.07) (.07) (.09) (.06) (.09)
Female –0.00 –0.00 –0.01 –0.01 –0.03* –0.03* –0.03* –0.03*
(.01) (.01) (.02) (.01) (.01) (.01) (.02) (.02)
Born again Christian 0.00 0.00 0.02 0.02 0.04* 0.04*
(.01) (.01) (.02) (.02) (.02) (.02)
Party identification 0.09* 0.09* 0.09* 0.09* 0.04+ 0.04+ 0.07* 0.07*
(.03) (.03) (.02) (.02) (.03) (.03) (.03) (.03)
Liberal–conservative identification –0.08 –0.07 0.02 0.02 –0.05 –0.04 –0.16 –0.14
(.08) (.08) (.08) (.08) (.11) (.11) (.12) (.12)
Authoritarianism 0.07* 0.07* 0.01 0.01
(.01) (.01) (.03) (.03)
Military strength 0.36* 0.32* 0.30* 0.28* 0.28* 0.22* 0.46* 0.41*
(.03) (.07) (.03) (.08) (.04) (.12) (.03) (.10)
Political sophistication –0.28* –0.30* –0.14* –0.15+ –0.27* –0.32* –0.18* –0.21*
(.08) (.09) (.06) (.09) (.09) (.12) (.07) (.11)
Liberal–conservative identification × sophistication 0.33* 0.30* 0.12 0.12 0.39* 0.38* 0.37* 0.34*
(.14) (.14) (.11) (.10) (.15) (.16) (.15) (.14)
Military strength × sophistication 0.07 0.03 0.10 0.09
(.09) (.12) (.18) (.16)
F statistic 44.09* 38.76* 34.35* 31.15* 39.31* 35.61* 115.39* 106.75*
R2 .22 .23 .18 .18 .17 .17 .30 .30
Number of observations 1378 1378 1641 1641 1252 1252 1011 1011
+
p < .10; *p < .05 (one-tailed).
Notes: OLS regression estimates. Standard errors are in parentheses (adjusted for complex sample design). All variables are on a 0–1 range. Higher values on
defense spending opinion reflect prospending sentiment.

spending than Democrats. The female, born again Christian, and authoritari-
anism variables behave as predicted roughly 50% of the time. Second, sophis-
tication conditions reliance on liberal–conservative predispositions in three
of four models. When conditional slope coefficients are calculated, liberal–
conservative attitudes are insignificant for respondents one standard devia-
tion under the sophistication mean and highly significant for those a standard
deviation above it. As always, relatively sophisticated people take advantage
of liberal–conservative cues.
What about the hawk–dove heuristic? Here I find the sophisticated and
unsophisticated rely equally on views of American power to guide their
defense spending evaluations. Across the equations the military strength coef-
ficient is correctly signed and statistically significant in model 1 whereas the
military strength × sophistication term falls short in model 2. The estimates
lead to the conclusion that the majority of the American electorate puts eval-
uations of military power to good use. Substantively, the military strength
effect proves quite potent. To take a pair of examples, extreme hawks evaluate
defense spending 36% more favorably than extreme doves in 1988 and 46%
more positively in 2004, all else being constant. The effect sizes for 1988 and
2004 are illustrated graphically in Figure 6.5. Next, citizens rely more heavily
on principle than partisanship. In all 4 years the hawk–dove coefficient for-
mally eclipses that of party identification. Military strength also outperforms
authoritarianism and, among the sophisticated, liberal–conservative identi-
fications. Finally, the standardized regression coefficient for the hawk–dove
variable is always significantly larger than that of party, authoritarianism,
and, among respondents one standard deviation above the knowledge mean,
liberal–conservative identities (except for 1996). Put simply, military strength
represents the key ingredient in the derivation of defense spending opinion.
1.00
Support for Defense Spending
0.75
0.50
0.25
1988 2004
0.00
Oppose Favor
Military Strength
Figure 6.5. Military Strength and Defense Spending Opinion, Additive Effects.
Source: Table 6.8 estimates; other predictors held at mean/mode. Effects significant.
Having completed the statistical analyses across the three fundamental

policy domains that define issue space in the United States, I now summarize
the big picture. First, most citizens use policy principles to guide the posi-
tions they take on salient political issues. In the economic welfare domain,
people base health care attitudes on broader postures toward government
power. When attention shifts to social and cultural affairs, the mass public
demonstrates anew its reasoning capacities as moral standards consistently
shape abortion preferences for the sophisticated and unsophisticated. Finally,
in the national security arena practically everyone relies on military power to
structure attitudes toward defense spending. To highlight the essential point
from the analyses, the less sophisticated have shown themselves to be quite
adept at taking advantage of core principles. Across 16 different tests, polit-
ical novices invoked core principles 14 times. Rarely did the less informed
rely more heavily on party or other cues when deducing preferences. Indeed,
the effects of policy principles often exceeded those of alternative factors. Put
simply, policy principles operate as central heuristics for most people—the
unsophisticated included.
Second, some evidence backs the weaker version of the sophistication
interaction model. As seen above, political knowledge enhances verti-
cal constraint between limited government and health care opinion in 3 of
6 years and between traditionalism and abortion opinion in 5 of 6 years. No
interaction effect surfaced in the case of defense spending. Here, the final
tally yields eight significant knowledge interactions in 16 tests, more than
would be expected by chance, but far from universal. Evidently, sophistica-
tion strengthens some ties between policy principles and issue attitudes. But
again, that the sophisticated sometimes do better should not obscure the
fact that the unsophisticated manage reasonably well. Finally, among the
informed the influence of core principles generally outweighs that of other
heuristics, most notably, party identification and, often, liberal–conservative
orientations.
Third, because the least aware habitually utilize domain-specific prin-
ciples, the proposition that high knowledge levels are necessary to deduce
issue preferences from such principles comes up short. But when predispo-
sitions are defined in terms of liberal–conservative attitudes, this perspec-
tive has much to recommend it. The liberal–conservative × sophistication
term was significant in 13 of 16 equations whereas the constitutive term
hardly ever reached significance in the anticipated direction. A similar pat-
tern held when comparisons were restricted to those a standard deviation
above or below the sophistication mean. In brief, the use of the liberal–
conservative heuristic is confined to more informed subsets in the
electorate.
I offer some final comments on the policy principle × sophistication effects.

The varied results raise some intriguing questions. Why does sophistication
promote reliance on limited government in 1988, 1992, and 2004 but not in
1996, 2000, and 2008? And why do the sophisticated weigh moral standards
more than the unsophisticated? In the abortion case, one plausible explanation
holds that a moral frame of reference dominates public discourse on this issue,
leading those who pay the closest attention to politics to conceptualize abor-
tion primarily in moral terms (Brewer, 2003). With respect to health care it may
be that the 1992 and 2004 presidential campaigns, which featured extensive
debate over the size and scope of the welfare state, primed limited government
more readily in the minds of the sophisticated. This may also hold for 1988 as
Vice President Bush kept Michael Dukakis on the defensive throughout the
campaign over taxes and welfare state liberalism, among other things.
These are important puzzles, but their resolution lies beyond the scope
of what I am seeking to accomplish in this book, and thus are best left for
future inquiry. For my purposes, the takeaway points are that sophisticated
and unsophisticated citizens regularly deduce issue preferences from abstract
policy principles, that these effects are substantively meaningful, and that the
sway of principles typically surpasses that of other cues. Broadly speaking,
the politically sophisticated and unsophisticated satisfy the centrality condi-
tion of policy voting.
CONCLUSIONS
The first condition of policy voting holds that if policy attitudes are to shape
candidate choice they must be present in mass belief systems. As demonstrated
in Chapter 5, attitudes toward core principles are available in the minds of
nearly everyone. The second condition holds that among available attitudes
those that play the part of central heuristics are more likely to guide electoral
choice than peripheral attitudes. So far as individuals grow accustomed to
using policy principles to craft evaluations of short-term controversies, these
attitudes may be applicable to candidate selection. In this chapter, my exami-
nation of panel data attests to the fact that policy principles shape issue pref-
erences in a top-down manner without being simultaneously influenced by
them, whereas my analysis of cross-sectional data indicates that experts and
novices rely heavily on these principles to infer issue positions. And although
sophistication sometimes promotes heuristic reasoning, a lack of sophistication
rarely forecloses it. What is more, core principles seem to wield more power
than the usual suspects in citizens’ reasoning chains. As such, my argument
that policy principles function as central heuristics for individuals across the
awareness spectrum rests on a solid empirical foundation.
Having found that the demos satisfy the first two conditions of policy vot-
ing, I am now ready to model the relationship between core principles and
the presidential vote. But before doing so, I take a detour in the next chapter
to study the origins of limited government, traditional morality, and military
strength. Although such an analysis is both interesting and informative in its
own right, it serves a broader normative purpose as well. Though perhaps
willing to concede, based on evidence marshaled so far, that most people hold
real policy principles, the skeptic might charge that said principles reflect little
more than latent partisan biases. If limited government, traditional moralism,
and military strength are shaped primarily by simple-minded cues such as
these, the fact that such principles guide political judgment offers little reas-
surance about the democratic capacities of ordinary citizens. But if policy prin-
ciples reflect first and foremost personal goals, the esteem in which citizens are
held appreciates. The next chapter seeks to sort this all out.
CHAPTER 7
The Origins of Policy Principles
Chapters 5 and 6 have established that policy principles are widely available
and function as central heuristics in the minds of voters. Where do these prin-
ciples come from? Why do some people hold positive views of government
activism and others hold negative views? Why do some individuals favor the
preservation of traditional family ties whereas others denounce such efforts?
What impels different people to become national security hawks or doves?
In Chapter 4, I sketched a model delineating the process by which citizens
acquire policy attitudes. Such attitudes develop when citizens are exposed
to salient ideas in American political discourse, gather what the ideas mean,
evaluate them, and store their impressions in memory. This model reveals
why core principles develop, but it elides the question of what drives evalua-
tion. This chapter aims to shed some light on these matters.
I argue that two factors are paramount in this process: party identification
and personal values. I invoke work on partisan perception (Campbell et al.,
1960) and basic human values (Schwartz, 1992) to explain how and why these
forces shape policy principles. Data from three new national surveys are used
to test the hypotheses. The results suggest that party and values systemati-
cally affect policy principles for the sophisticated and unsophisticated alike,
with the influence of values outweighing that of party. After this, I pursue
an exploratory analysis of the relationship between human values and party
identification. I posit that values drive partisanship and present some evidence
consistent with this claim. The chapter concludes with a brief discussion of the
broader implications these results have for assessing citizen competence.1
1
Readers may wonder why I do not include liberal–conservative orientations
alongside party and basic values as antecedents of policy principles. There
are two reasons. First, as made plain in Chapter 3, the supposition that ideo-
logical principles structure political evaluations may hold for only 5–10% of
the electorate. A considerably larger share of the public holds attitudes toward
159
THE INFLUENCE OF PARTY IDENTIFICATION

Party identification has been defined as a psychological bond to the Democratic
or Republican Party. A rich line of research has demonstrated that partisan
affinities subtly affect political perception and judgment. For instance, parti-
sanship biases evaluations of public officials in a manner consistent with peo-
ple’s underlying loyalties. In prior work I showed that the partisan opponents
of a sitting president rely more than his copartisans on perceptions of charac-
ter weakness when rendering bottom line evaluations about him (Goren, 2002,
2007). Thus, Republicans saw dishonesty as Clinton’s chief flaw and made
greater use of these perceptions than Democrats when evaluating him in 1992
and 1996. Democrats returned the favor in 2004 by relying more heavily than
Republicans on perceptions of Bush’s character weakness, a lack of compas-
sion, when judging him. Partisan ties even color perceptions of objective facts.
Bartels (2002) documented an especially egregious example of this. Although
the inflation rate fell from 13.5% to 4.1% between 1980 and 1988, when asked
whether inflation had improved or worsened during this period 50% of strong
Democrats said it had gotten worse and only 8% said it had gotten better.
In contrast, 13% of strong Republicans said worse and 47% said better. The
fact that Reagan was president evidently outweighed economic reality in the
minds of Democratic respondents.
What underlies partisan perception? Although the possibility that such an
influence arises spontaneously cannot be discounted, there are good reasons
for believing biased processing typically occurs after partisan identities have
been activated by political source cues. Students of attitude change have long
recognized that source cues affect how message recipients react to persua-
sive appeals. Although many personal decisions (e.g., whether to tender or
accept a marriage proposal) require careful deliberation, political decisions
are inconsequential for most people. For judgments such as these individuals
usually rely on simple heuristics to guide their choices. One widely used rule
liberal–conservative labels, but these tend to center on a few symbols or social

groups rather than functioning as philosophical worldviews or policy preference
summaries. And even here, only a fraction of the public holds such orientations.
Many people, in short, lack ideological orientations. Second, policy principles
may develop before liberal–conservative attitudes. For those who lack ideologi-
cal attitudes, policy principles necessarily occupy more central nodes in attitude
structures. Among those who hold ideological attitudes, this developmental
sequence suggests policy principles are as likely to shape liberal–conservative
attitudes as to be shaped by them.
The Origins of Policy Principles 161
of thumb dictates that people decide based on how they feel about the source
behind the message (McGuire, 1969).
The most salient cue in politics is the party label. Sometimes cues are embed-
ded in political communications, such as when Democratic- or Republican-
sponsored campaign commercials air on television. At other times the partisan
leanings of the source are identified by a news organization, such as when
a reporter announces the positions of a Democratic or Republican Senator.
Cues also reach citizens from copartisans in their social networks. Whatever
the impetus, partisan predispositions, once activated, motivate recipients to
accept or reject messages in a manner indicative of party leanings. If the mes-
sage source and recipient share the same affiliation, the latter accepts the claim
at face value without sweating the details. But if the source and target stand
on opposite sides of the political divide, the recipient will probably reject the
message (Goren, Federico, & Kittilson, 2009).
Given the centrality of partisan identities in the minds of citizens and the
prevalence of party cues in a political environment organized as a two-party
system, Democratic and Republican politicians have considerable latitude to
shape public opinion. Party leaders regularly disseminate simple cues about
party positions that reach many people. These cues may be statements about
party principles or calls for specific government actions that signal broader
party stances. So long as party representatives furnish the electorate with a
steady stream of cues, citizens can simply toe the party line when thinking
about policy cleavages. As we have seen in Chapter 2, the Democratic Party
has long stood for government activism in the economic welfare domain and
the GOP has resisted such demands. Republicans have decried social change
and moral decline whereas Democrats have stressed, with varying degrees of
enthusiasm, the virtues of tolerance. In the national security arena, the GOP
has for the most part favored force over diplomacy; Democrats have proven
more tentative about military engagements. Given this, I hypothesize that
Republicans will evaluate limited government, orthodox morality, and mili-
tary strength more favorably than Democrats.
THE INFLUENCE OF PERSONAL VALUES

What Are Personal Values?
Personal values represent another potential antecedent of political attitudes.
The values construct has received a lot of attention over the years from social
psychologists. In his landmark work, Rokeach (1973, p. 5) defined a value as
“an enduring belief that a specific mode of conduct or end-state of existence is
personally or socially preferable to an opposite or converse mode of conduct

or end-state of existence.” Discrete values are organized into broader hierar-
chies; again Rokeach (1973, p. 5) states that “A value system is an enduring
organization of beliefs concerning preferable modes of conduct or end-states
of existence along a continuum of relative importance.” Schwartz (1992) has
built upon Rokeach to develop a universal theory of the content, structure,
and consequences of human values. For Schwartz (1994, p. 20) personal val-
ues are abstract beliefs about desirable end states or behaviors that transcend
specific situations, guide evaluation and behavior, and can be rank ordered in
terms of relative importance. Every value reflects an abstract goal that serves
either the needs of individuals as biological organisms, group welfare needs,
or the need for coordinated social interaction. Satisfaction of these needs helps
ensure survival for individuals, groups, and the broader society.
In Schwartz’s theory, some 50 or so diverse values are encapsulated by
10 broader value types that share compatible motivations. The value types
are benevolence, universalism, self-direction, stimulation, hedonism, achieve-
ment, power, security, conformity, and tradition. Table 7.1 defines the 10 value
types in terms of the goals they express. To offer some examples, tradition val-
ues call for deference to enduring cultural and religious norms. They fit nat-
urally with conformity values that emphasize self-restraint and submission
TA B L E 7 . 1 . DEFINITION OF VALUE TYPES

1. Benevolence values—preserving and enhancing the welfare of those with whom we
are in frequent personal contact
2. Universalism values—understanding, appreciation, tolerance of and protection for
the well being of everyone as well as nature
3. Self-direction values—independent thought and action, choosing, creating, exploring
4. Stimulation values—excitement, novelty, and challenges in life
5. Hedonism values—pleasure or sensuous gratification for yourself
6. Achievement values—personal success acquired by demonstrating competence
according to social standards
7. Power values—social status and prestige, control of or dominance over people and
resources
8. Security values—safety, harmony, and stability of the self, personal relationships, and
the broader society
9. Conformity values—restraint of actions, inclinations, and impulses that are likely to
upset or harm others and violate social expectations or norms
10. Tradition values—respect for, commitment to, and acceptance of the customs and
ideas embodied by your culture or religion
Source: Schwartz (1994).

to social expectations. In contrast, self-direction values stress free choice and

action for individuals whereas stimulation values affirm excitement, nov-
elty, and challenge. These examples show that the attainment of one value
(e.g., self-direction) can obstruct realization of another (e.g., conformity). The
content and structure of values can be represented visually using what psy-
chologists call a circumplex. Figure 7.1 arrays the value types along a circular
motivational continuum in which adjacent values have more in common with
one another than with values further away on the structured circle.
When the 10 value types are considered en masse, they too can be reduced
to a smaller number of higher order values based on motivational congru-
ency. Figure 7.1 shows a common formulation based on a quartet of superor-
dinate values: (1) self-transcendence values (benevolence and universalism)
that denote concern for and commitment to the well being of all individuals,
groups, and society; (2) self-enhancement values (hedonism, achievement,
and power) that elevate the pursuit of self-interest, success, and dominance
over others; (3) openness to change values (self-direction, stimulation, and
hedonism), which express a desire for independent thought and action, nov-
elty, exploration, and self-gratification; and (4) conservation values (security,
conformity, and tradition) that emphasize self-restraint, the preservation of
traditional cultural and religious practices, social stability, and individual
OPENNESS TO Self-direction Universalism SELF

CHANGE TRANSCENDENCE
Stimulation Benevolence
Hedonism
Conformity Tradition
Achievement
SELF CONSERVATION
ENHANCEMENT Power Security
Figure 7.1 The Schwartz Model of Value Relations among the 10 Motivational
Domains.
and collective security (Schwartz, 1994, p. 25).2 Note that self-transcendence

and self-enhancement oppose one another to form a bipolar value dimension.
Openness to change and conservation form a second, independent dimension.
As with the 10 value types, the pursuit or fulfillment of one overarching value
implies downgrading the other.
These are strong claims, so it is natural to wonder how they have fared
empirically. Using data from approximately 350 independent representa-
tive and convenience samples drawn from roughly 75 different countries,3
Schwartz and others have found the posited model turns up most of the time
(e.g., Davidov, Schmidt, & Schwartz, 2008; Schwartz, 1992; Spini, 2003). Recent
work using representative national samples in the United States finds the same
(Goren, 2009). Given consistent results across diverse geographic regions, lin-
guistic communities, and cultures, a reasonable inference to draw is that this
model of value content and structure holds widely, albeit with some local
variations.
Personal values are construed as standards that guide perception, judg-
ment, and behavior in all walks of life. To take a handful of examples, values
shape lifestyle choices, consumer purchases, altruistic behavior, social interac-
tions, college majors, what people worry about, and so on (Bardi & Schwartz,
2003; Maio & Olson, 1995; Schwartz, Sagiv, & Boehnke, 2000; Verplanken &
Holland, 2002). Because values are used so readily in daily life, they may
inform policy principles as well. There is not much prior literature to guide
my thinking here, so I rely on admittedly exploratory logic about congruence
between the abstract goals expressed in values and the ideas embedded in
policy principles. First, I posit universalism vertically constrains attitudes
toward limited government. This value stresses understanding, appreciation,
and concern for the interests and well being of people in general. Insofar as the
denial of federal assistance may harm those on the economic margins of soci-
ety, broader goals are frustrated. Universalism should thus be inversely related
to limited government. Second, self-direction values elevating the importance
of independent thought and action might also impact limited government.
In American political culture, strong government is often seen as a threat to
values such as individual freedom and independence that express these goals
(Feldman & Zaller, 1992). For this reason, self-direction priorities and sup-
port for small government should go hand in hand. Third, self-enhancement
2
Hedonism values are subsumed under both openness to change and self-
enhancement.
3
Personal communication with Shalom Schwartz, December 29, 2008.
values emphasize striving, personal success, and the accumulation of power

at the expense of other elements in society. If government activism on behalf
of the needy is perceived as hindering these pursuits, individuals who pri-
oritize such values should back smaller government. Hence I propose that
achievement and power values will be positively related to limited govern-
ment. My final prediction holds that belief in the desirability of conservation
goals evokes favorable reactions to limited government. Because idleness and
dependency are seen as violations of American cultural norms, shaped as
they are by the Protestant work ethic, conformity and tradition values should
engender support for limited government.
For the policy principle of traditional morality, several values are likely con-
sequential. On the one hand, universalism and self-direction may drive oppo-
sition to claims that traditional family ties must be preserved and alternative
lifestyles prohibited. Those who value personal autonomy for themselves and
protection for others should resist the imposition of one set of moral precepts
for everyone. On the other hand, belief in conservation values should augment
support for moral regulation. To the degree that self-denial, deference to social
expectations and norms, and respect for cultural and religious traditions serve
as guiding principles in peoples’ daily lives, these beliefs should translate into
political preferences that seek the same for the larger community.
For the hawk versus dove cleavage, universalism and benevolence should
be negatively related to militarism. In professing care for the interests of oth-
ers, these values embody a willingness to coexist with, rather than fight, those
who are different. Additionally, because military power furnishes America
with the means to influence if not dominate foreign adversaries, it might gain
assent from citizens who value the accumulation and application of power
in their personal lives. Hence, those who endorse power values should favor
military muscle. Lastly, military strength promises safety from attack and
a bulwark against change; as such, it should elicit approval from those hold-
ing conservation values in high regard. My expectation is that the more people
value security, conformity, and tradition, the greater the appeal of American
power and prestige. Table 7.2 summarizes the disparate hypotheses. A “+” sign
means a given value and principle should be positively related. Conversely, a
“–” sign calls for an inverse relationship between the variables. Finally, a blank
space denotes relationships that will go untested in the present study.4
4
Resource constraints precluded inclusion of the full set of value items across
the surveys as described below. I concentrated on what seem likely to be the key
values affecting principles.
TA B L E 7 . 2 . PREDICTIONS FOR THE IMPACT OF PERSONAL VALUES ON

POLICY PRINCIPLES
Traditional Military
Limited Government Morality Strength
Benevolence –
Universalism – – –
Self-direction + –
Achievement +
Power + +
Security +
Conformity/tradition + + +
Note: + means the value should promote support for the given principle; – means the value should
depress support for the principle. Blank spaces denote untested relationships.
As always, the possibility that political sophistication moderates verti-

cal constraint must be entertained. Proponents of the sophistication interac-
tion model hold that prior knowledge usually strengthens the connections
between abstract beliefs and attitudes. Although this may hold for some pre-
dispositions (e.g., liberal and conservative views) and under some circum-
stances (e.g., technical issues), sophistication seems a poor bet to moderate the
linkages described above. As seen previously, policy principles occupy central
nodes in the belief systems of political sophisticates and novices, suggesting
knowledge is no prerequisite for acquiring and maintaining them. And given
that most folks have a great deal of experience deploying values as judgmen-
tal yardsticks in their personal affairs, they should experience little trouble
applying them to central political predispositions. Put simply, most citizens
should be adept at grounding attitudes toward limited government, tradi-
tional morality, and military power in personal values (Schwartz, Caprara, &
Vecchione, 2010). Before proceeding to the empirical tests, I need to differenti-
ate personal values from two other concepts with which they might be con-
fused. I do so in the next two sections.
Personal Values and Core Political Values Are Different

Like social psychologists, political scientists have spilled much ink on the val-
ues construct over the years, thereby begging the question of how the two
approaches compare and whether one is preferable to the other. There are
important differences between the two. First, psychologists have defined per-
sonal values at a higher level of abstraction. In their conceptualization, values
represent hierarchically organized beliefs about desirable, transsituational
goals that guide evaluations of persons, events, and behaviors in everyday

life. Political scientists have focused on the more narrowly defined construct
of core political values and offered a variety of definitions over the years.
Kinder (1998, p. 808) described values as “relatively abstract and durable
claims about virtue and the good society.” For McCann (1997, p. 565) “A citi-
zen’s set of core political values consists of overarching normative principles
and belief assumptions about government, citizenship, and American soci-
ety.” To Kuklinski (2001, p. 364), “values are people’s normative views of life.”
These definitions set core political values apart from concepts such as liberal–
conservative orientations, issue preferences, and candidate evaluations, but
they seem to function more like a catchall category for general political beliefs
that operate in the domain of politics as opposed to a fundamental class of
personal beliefs that applies across life domains. Said otherwise, core political
values are a political construct centered on beliefs about “government, citizen-
ship, and American society.” Personal values are prepolitical beliefs that have
much wider applicability.
Measurement differences exist between the perspectives as well. The
social–psychological measurement of personal values is taken up in the next
section. At this point I outline some potential problems in how core political
values have been tapped. Consider the following statement that has appeared
on National Election Study (NES) surveys since 1984: “If people were treated
more equally in this country we would have many fewer problems.” The
item does not cue respondents to rank political values in terms of importance;
instead, subjects report on how strongly they agree or disagree with the state-
ment. This is troubling because values, according to most accounts, are pri-
oritized in terms of relative importance. Indeed, the “agree–disagree” format
raises the possibility that NES items are more akin to attitudes than values.
Another worry is that many NES items hint at a collective response through
political action. The aforementioned item blends abstract beliefs about equal-
ity (“people should be treated equally”) with an implicit call for government
action (“our country will be better off if this problem is fixed”).5 Many NES
measures conflate abstract, context-free values with the policy attitudes they
are ultimately supposed to explain. Similar problems are evident on value
items in other surveys. For instance, in the 1991 National Race and Politics
Survey respondents are asked to rate the importance of social justice using
this item: “taking care of the homeless.” Or consider the equality item on the
1994 Multi-Investigator Study: “By EQUALITY we mean narrowing the gap
5
I thank Stanley Feldman for bringing this point to my attention.
in wealth and power between rich and poor.” Again, these measures blend
abstract ideas with identifiable political issues.
This brief discussion indicates that personal values and core political val-
ues are not interchangeable (Schwartz et al., 2010, make the same argument).
Whereas personal values apply broadly to all situations and contexts, core
political values are confined to a single area—politics. Moreover, personal val-
ues are rank ordered in regard to personal priority but political values are
not. Schwartz’s theory of value content and structure specifies a universe of
values and how these relate to one another; political science offers no ana-
logue. Finally, as I will show before long, personal values are measured at a
higher level of abstraction and, thus, unlike political value measures, are not
confounded with public policy. On grounds of theoretical breadth, concep-
tual clarity, and measurement validity, the social–psychological approach has
much to recommend it. I adopt it accordingly.6
Personal Values and Personality Traits Are Different

Some readers may wonder about the distinction between personality traits
and basic human values, so some discussion here may be helpful. At the
broadest level, personality traits can be viewed as fundamental disposi-
tions that influence what people think and believe and how they act across
a variety of contexts and moments in time (McAdams & Pals, 2006). Work
over the past two decades has identified five key trait dimensions—the “Big
Five” in the vernacular of personality researchers. These include openness to
experience, conscientiousness, extraversion, agreeableness, and emotional
stability. The “Big Five” term is a taxonomy that encompasses the majority
of discrete traits that are used to describe personality. This phrase should
not be seen as implying that this quintet reflects all meaningful individual
differences. Instead, the Big Five define personality at the highest level of
abstraction, meaning that each dimension covers more specific facets of per-
sonality, which in turn denote the much larger array of discrete personality
characteristics.
The Big Five can be defined as follows. Openness to experience reflects a
blend of intellect, imagination, a willingness to pursue novel activities, and a
6
Before proceeding, I want to emphasize the key insight emerging from work
on core political values. The consistent finding that those who are “innocent of
ideology” instead use political values to constrain their issue preferences repre-
sents a signal contribution to the study of political psychology and public opin-
ion. Ideological innocence does not imply political incompetence, an incredibly
important and heartening finding.
desire for variety, what John, Naumann, and Soto (2008, p. 120) have described
as “the breadth, depth, originality, and complexity of an individual’s men-
tal and experiential life.” Next, conscientiousness encompasses dependabil-
ity and reliability, effort, a drive to achieve, and self-control, which add up
to “socially prescribed impulse control that facilitates task- and goal-directed
behavior” (John, Naumann, & Soto, 2008, p. 120). Extraversion denotes a gre-
garious, outgoing, and active disposition, “a preference for companionship
and social stimulation” (McCrae & Costa, 1999, p. 143), and “an energetic
approach toward the social and material world” (John et al., 2008, p. 120).
Agreeableness denotes a tendency to strive for positive interactions and rela-
tions with other people, a willingness to defer to them to minimize interper-
sonal conflict and disagreement, and a “prosocial and communal orientation
toward others” (John et al., 2008, p. 120). Lastly, emotion stability has been
defined as a calm and steady disposition versus being tense, anxious, and
prone to negative emotional states.
So, how do personality traits and basic human values differ? First, traits
are innate dispositions to act in comparable ways across a variety of dif-
ferent contexts and over time, whereas personal values are enduring goals
people learn from the environment. Second, traits vary across people in
their intensity and frequency of occurrence, whereas values reflect differ-
ences in the priorities people attach to desirable end-states or modes of
conduct. Third, traits can be judged positively or negatively (e.g., conscien-
tiousness is typically viewed more favorably than neuroticism), but values
are generally seen as desirable to varying degrees. Fourth, personality traits
appear to develop before personal values in a temporal or causal sequence
(McCrae & Costa, 1999; Roccas et al., 2002). Perhaps the simplest way to
summarize the difference is this: “Traits describe what people are like; val-
ues refer to what people consider important” (Caprara et al., 2006, p. 3). Just
as personal values differ from core political values, so too do they differ
from personality traits.
My point here is not to dismiss the theoretical utility of personality traits for
the study of political judgment. On the contrary, a burgeoning line of research
demonstrates that the Big Five have far reaching implications for understand-
ing public opinion, political engagement, political identities, the strength of
those attachments, and other important variables (Carney, Jost, Gosling, &
Potter, 2008; Gerber, Huber, Doherty, Dowling, & Ha, 2010; Mondak, 2010).
Instead, my point is that personality traits and basic human values are differ-
ent constructs and that they should not be viewed as one in the same. Indeed,
had I adequate measures of trait dispositions, they would be included as inde-
pendent variables in the models of core principles.
MODELING POLICY PRINCIPLES

Given the problems with existing value instruments, data from NES and other
publically available surveys cannot be used to explore the correlates of policy
principles. In light of this, I designed two surveys and commissioned Knowledge
Networks (KN) to administer them to nationally representative samples. The first
centered on personal values and economic welfare opinion and was conducted
in July 2007 (n = 1506). The second focused on values and cultural issues and was
fielded in October 2008 (n = 1017).7 To model the origins of military strength, I
turn to data from a third national survey designed by a team of researchers and
fielded by YouGov/Polimetrix in February 2011 (n = 1200).8
The Building Blocks of Limited Government

I begin with the 2007 economic welfare survey. Limited government is measured
with five questions about federal spending on “assistance to the unemployed,”
7
KN has assembled a web-based panel in which random digit dialing techniques
are used to recruit representative samples of the U.S. voting age public to partic-
ipate in a few monthly online surveys on various topics (e.g., consumer choices,
entertainment preferences, and health care usage) in exchange for free Internet
access. Because participants are selected using probability sampling techniques
and current population data are used to create poststratification weights that
reduce the effects of nonresponse and noncoverage biases, KN samples are rep-
resentative of the broader U.S. population.
8
I thank William Chittick and Jason Reifler for inviting me to be part of this project.
YouGov uses sample matching techniques to draw “representative” samples from
nonrandomly selected pools of respondents in online access panels, which consist
of internet users who were recruited via banner ads, purchased email lists, and other
devices. The sample matching technique begins by drawing a stratified national
sample from a target population (in this case, the 2006 American Community
Survey). Rather than contacting these individuals directly, which would be pro-
hibitively expensive, YouGov utilizes matching techniques to construct a compa-
rable sample from its existing internet panel. Members of the matched sample are
then contacted and invited to participate in the survey. After selection, the sample
is weighted to match the target population on a series of demographic factors.
Although the matched sample has been drawn from a nonrandomly selected pool
of opt-in respondents, it can in some respects be treated as if it were a random sam-
ple (Vavreck & Rivers, 2008). These matched samples resemble the broader public
on a number of sociodemographic variables; however, because respondents self-
select into the original panel they may differ from the broader public on unmeas-
ured variables such as political interest and awareness.
“welfare,” “food stamps,” “aid to the homeless,” and “child care.” Respondents
indicated whether spending should be “increased a lot,” “increased some,”
“kept the same,” “decreased some,” or “decreased a lot.” I calculate an average
or person mean score for each respondent who answered more than half the
available items (see Chapter 6 for a review). The limited government variable is
constructed so that higher scores denote more conservative responses.
I include statistical controls for female (1 = female, 0 = male), African-
American (1 = African American, 0 = other), Hispanic (1 = Hispanic, 0 =
other), dummy variables for the middle and top income terciles (roughly),
and dummy variables for high school graduates (1 = high school graduates
and those with some college, 0 = other) and college graduates (1 = 4-year col-
lege graduate, 0 = other). I hypothesize that women, African-Americans, and
Hispanics will evaluate limited government more critically than men and
non-Hispanic whites, respectively. Moreover, respondents in the middle and
high income groups should hold stronger antigovernment views than those
in the bottom third of the income distribution. Similarly, those in the medium
and high education categories should prove more receptive to smaller govern-
ment than the least educated.
Gilens (1999) and Kinder and Kam (2008) have shown that whites who
endorse negative stereotypes about the work ethic of blacks and Hispanics are
more likely to oppose means-tested welfare programs than those who repudi-
ate stereotypes. The same may hold true with respect to the broader principle
of government activism. To test this claim, I include a stereotype variable in
the model, which is measured using the person mean score for respondents
who answered at least three of four seven-point stereotype scales. The first
two items asked respondents to place “blacks in general” and “Hispanics in
general” on a “hardworking” versus “lazy” scale; a second pair of items did
the same for a “responsible” versus “irresponsible” question (Cronbach α =
.79). Because higher scores indicate antiminority sentiments, the stereotype
and limited government variables should be positively related.
Turning to the key independent variables I tap party identification using
the standard seven-point scale, coded so that higher scores reflect GOP self-
categorization. Compared to Democrats, Republicans should evince greater
hostility toward the welfare state. To capture personal values I employ a vari-
ant of the measurement strategy developed and validated in early research by
Schwartz (1992, 1994). The following paragraph introduced the values battery
to respondents:
You are going to see some statements about things some people consider
important in life. Using a number from 0 to 10, please rate how important
each one is to you. If it’s one of the absolutely most important things to
you, give it a 10. If it’s one of the least important things, give it a 0. You’re
free to use any number between 0 and 10, but remember, the more impor-
tant something is to you, the higher the number you should give it.
Responses to multiple statements are used to measure each germane value.

Universalism is gauged by the following six statements: “Caring for the
weak,” “Tolerating different beliefs and ideas,” “Making sure everyone has
an equal opportunity to succeed,” “Caring for the needy,” “Tolerating peo-
ple who think and act differently,” and “Making sure everyone has an equal
chance to get ahead.” Self-direction is tapped using “Depending on yourself
to make it in life,” “Being free to do whatever you want,” “Relying on your-
self to get ahead,” and “Being free to think whatever you want.” For achieve-
ment values, I rely on a quartet of items: “Working very hard at what you
do,” “Improving your position in life,” “Setting ambitious goals and achiev-
ing them,” and “Being successful in reaching you goals.” Four items assess
power values: “Making a lot of money,” “Having control over other people,”
“Having a lot of material possessions,” and “Getting others to do what you
want them to.” Two statements address conformity values: “Being self-disci-
plined” and “Resisting temptation.” For each scale, I take the person mean
score for respondents who answered more than half the items. The scales are
keyed so that higher scores denote greater personal importance attached to
a given value. As indicated above, universalism should be inversely related
to limited government whereas self-direction, achievement, power, and con-
formity should enhance support for small government.
A few points should be noted about the items. First, question wording cor-
responds closely to the definition of the personal values construct provided
above and references key values used in Schwartz’s (1992) framework. The
introduction to the battery and subsequent statements are designed to elicit
transsituational beliefs about desirable goals that can be rank ordered in terms
of personal importance. Moreover, the language references personal beliefs.
Implicit or explicit policy ideas are absent. Second, in an earlier report on these
measures I found that each set of items captures the respective value dimen-
sion it is supposed to capture and forms a reliable scale (the Cronbach α coef-
ficients vary from .75 to .81; see Goren, 2009).
I also test whether value reasoning is conditional on a proxy measure of
sophistication. Specifically, I estimate whether the relationship between values
and limited government is stronger at higher levels of education. Although
education is not a direct measure of factual and associational political knowl-
edge, it can serve as a useful if rough instrument when knowledge measures
are unavailable. Education correlates reasonably well with political knowl-

edge (typically around .50) and frequently yields results comparable to those
that are obtained when knowledge scales are employed (e.g., Lau & Redlawsk,
2006). I test for sophistication effects with value × education terms. My expec-
tation is that these terms will add little.
The statistical analysis proceeds as follows. I report a baseline linear regres-
sion model predicting limited government from the sociodemographic, party,
and stereotype controls. The universalism, self-direction, achievement, power,
and conformity variables are then added to the baseline to see if the regression
coefficients simultaneously equal zero (assessed using a joint F test), boost
model fit (assessed using the R2 statistic), and, if so, how each variable bears
upon limited government. A final model incorporates the value × education
interaction terms. All variables are normed onto a 0–1 range; therefore, the
additive regression coefficients can be interpreted as the percentage difference
in support for limited government between those who score highest and low-
est on a given predictor, ceteris paribus.
Table 7.3 reports the statistical estimates. In the baseline model, whites,
middle and upper incomes groups, and the college educated judge limited
government more positively than blacks, the bottom income tercile, and the
least educated, respectively. Furthermore, those who think ill of minorities
score 13% higher on limited government than those holding the most positive
assessments (p < .001). Finally, party identification plays its prescribed role
with strong Republicans 17% more antigovernment than strong Democrats
(p < .001).
The second model introduces universalism, self-direction, achievement,
power, and conformity values to the mix. The null hypothesis that the coeffi-
cients for these variables simultaneously equal 0 can be rejected (F test = 15.47,
p < .001) and the R2 increases from .23 to .29. The coefficient estimates support
some but not all of my hypotheses. On the plus side, universalism is strongly
associated with opposition to small government. The more importance peo-
ple attach to the well being of others, the cooler their reactions to economic
conservatism (b = –.38, p < .001). A modest self-direction effect turns up as
well (b = 0.12, p < .05), meaning that a heightened commitment to values such
as freedom and independence arouses antigovernment suspicion. Conformity
values also produce the expected effect. Those who value social norms such
as the work ethic and self-discipline view limited government more positively
than those who reject conformity (b = 0.20, p < .01). On the minus side, the self-
enhancement variables perform erratically. Contrary to my prediction, those
who rate personal achievement highly are somewhat less critical of the federal
government than those who attach little importance to this goal (b = –0.12,
TA B L E 7 . 3 . THE ORIGINS OF LIMITED GOVERNMENT, 2007
Model l Model 2 Model 3
Constant 0.21* 0.38* 0.38*

(.03) (.05) (.11)
Female –0.01+ –0.01 –0.01
(.01) (.01) (.01)
Black –0.08* –0.08* –0.08*
(.02) (.02) (.02)
Hispanic –0.01 –0.03 –0.02
(.02) (.02) (.02)
Second income tercile 0.06* 0.06* 0.06*
(.01) (.01) (.01)
Third income tercile 0.08* 0.08* 0.08*
(.02) (.02) (.02)
High School graduate or some college 0.03+ 0.03+ 0.03
(.02) (.02) (.13)
4-year college graduate 0.04* 0.04* 0.03
(.02) (.02) (.14)
Party identification 0.17* 0.14* 0.13*
(.02) (.02) (.02)
Antiminority stereotypes 0.13* 0.07* 0.07*
(.04) (.04) (.04)
Universalism –0.38* –0.30*
(.05) (.14)
Self-direction 0.12* –0.05
(.06) (.21)
Achievement –0.12* –0.24+
(.06) (.17)
Power 0.01 0.04
(.03) (.08)
Conformity 0.20* 0.40*
(.04) (.12)
Universalism × High School graduate/ 0.02
some college (.15)
Universalism × college graduate –0.31*
(.16)
Self-direction × High School graduate/ 0.16
some college (.22)
Self-direction × college graduate 0.21
(.22)
Achievement × High School graduate/ 0.06
some college (.18)
Achievement × college graduate 0.34*
(.20)
Power × High School graduate/some college –0.07
(.09)
174
TA B L E 7 . 3 . (CONTINUED)
Power × college graduate 0.02

(.10)
Conformity × High School /some college –0.21*
(.13)
Conformity × college graduate –0.26*
(.15)
R2 .23 .29 .31
F statistic 29.75* 30.05* 20.38*
Joint F test for additional block – 15.47* 2.37*
+
p < .10; *p < .05 (one-tailed).
Notes: Ordinary least squares (OLS) regression estimates. Unstandardized coefficients are reported.
HC3 standard errors are in parentheses. All variables are on a 0–1 range. Higher values on limited
government reflect a conservative opinion. Number of observations = 1452. Knowledge Networks
sampling weights are used.
Source: 2007 Knowledge Networks survey.
p < .05). Power values do not appear to constrain attitudes either way (b =
0.01, ns). Overall, the regression estimates suggest that basic human values,
universalism and conformity especially, play a central role in shaping bottom
line judgments about how much government should try to spare citizens eco-
nomic hardship.
To convey effect size, Figure 7.2 plots the predicted limited government
position across the range of scores for universalism and conformity, the dom-
inant values acting on opinion. Bear in mind that because relatively few cases
fall in the bottom end of the value distributions this plot essentially illustrates
the hypothetical maximum effect of a given value on the principle, ceteris pari-
bus (the same holds true for most plots in the chapter). We can see that the
more importance citizens attach to universalism the less they care for limited
government. Going from the lowest to highest importance rating on univer-
salism reveals that support for limited government falls some 38%. The figure
further demonstrates that those most wedded to conformity values score 20%
higher on limited government versus the least committed.
The third regression model tests whether the value principle relation-
ship grows stronger among the better educated. Here, the evidence is mixed.
The null hypothesis that the value × education terms simultaneously equal
zero can be rejected (F = 2.37, p < .001), whereas the R2 rises, albeit modestly,
from .29 to .31. Yet only two of the 10 interactions are correctly signed and
1.00
Support for Limited Government
Universalism
0.75
0.50
0.25
Conformity
0.00
Low High
Importance of Value
Figure 7.2. Personal Values and Limited Government.

Source: Table 7.3 estimates; other predictors held at mean/mode. All effects significant.
statistically significant in a manner suggesting education promotes reliance on

values (universalism × college graduate and achievement × college graduate).
Both of the conformity × education coefficients are negatively signed and sig-
nificant, meaning the value to principle link is strongest in the least educated
group, the opposite of what sophistication theory indicates should happen.
And the remaining six interactions are insignificant. These results do not pro-
vide much evidence that sophistication, or at least the education variable I use
in its stead, promotes reliance on basic human values.
Overall, my statistical examination of the correlates of attitudes toward
limited government yields several conclusions. First, the usual sociode-
mographic suspects, GOP partisanship, and stereotypical beliefs about the
work habits of blacks and Hispanics are associated with antigovernment
sentiment. Second, global evaluations about the role of government are
grounded deeply in personal values. Those concerned about the interests
of others take a much dimmer view of limited government than individuals
who manifest less compassion. Additionally, conformity values induce hos-
tility toward “big” government. Self-direction, as expected, and achieve-
ment, in an unexpected way, affect opinion as well. Finally, power exerts
little sway over this principle. Clearly, universalism and conformity goals
are more consequential than the others. But the broader point stands: per-
sonal values play a central role in guiding the positions citizens take on
limited government.
The Building Blocks of Traditional Morality

The 2008 KN survey includes the standard four-item battery to measure
support for prevailing moral customs, which I assess via the person mean
approach and code so that higher scores denote cultural conservatism.

Controls for female, African-American, Hispanic, and high school and college
graduates are included in the statistical models. I posit that women, blacks,
and Hispanics will favor conventional morality more than men and whites,
respectively. Party identification, measured as before (i.e., high score = strong
Republican), should also encourage cultural orthodoxy.
I gauge values using the Schwartz personal values portrait (PVP) question
format, which has been validated after extensive testing (e.g., Davidov et al.,
2008). For these measures, each subject read a statement about an abstract goal
described as important to a hypothetical person and then indicated how similar
he or she was to that person. The introduction to the battery reads this way:
I will describe some people. Please tell me how much each person is or is
not like you. Is this person very much like you, like you, somewhat like
you, a little like you, not like you, or not like you at all?
Four gender-balanced items are used to evaluate universalism: “She thinks

it is important to accept people who think and act differently,” “Equality is
important to her. She believes everyone should have equal opportunities in
life,” “Recognizing and respecting the rights, beliefs, or customs of others
is important to her,” and “She believes it is important that every person in
the world be treated equally.” The following two statements capture self-
direction values: “It is important that every person is free to act as they think
best” and “It is important to her that everyone is free to do whatever they
want as long as they don’t harm others.” These four items tap conformity
and tradition values: “She thinks it is important to stick to her religious faith
and belief,” “Tradition is important to her. She tries to respect tradition,”
“Being true to God is important to her,” and “It is important to preserve
traditional ideas of right and wrong.” Because conformity and tradition are
close neighbors on the motivational circumplex (see Figure 7.1) and their
empirical correlations are robust, I combine the items into a single scale (cf.
Davidov et al., 2008).
Some commentary is needed on the PVP format. Each portrait describes
another person’s goals in a way that implicitly evokes a given value rather
than asking directly about it. Doing so helps minimize self-presentation
biases, reflects the types of interpersonal comparisons individuals make in
their daily lives, and spares them the task of thinking about what is impor-
tant to them, something they may have less experience with than person-to-
person comparisons (Schwartz, 2003). Question wording centers on personal
beliefs and priorities, making no reference, implicit or otherwise, to govern-

ment action or political response. Finally, in an earlier report on these data
I found the items reflect the value dimensions they are supposed to reflect
and form reliable scales (the Cronbach α coefficients range from .76 to .81;
see Goren, 2009).
For scales based on three or more items, I calculate person mean scores for
participants who answered over half the questions. Each scale is coded so that
higher scores reflect greater dedication to a given value. Recall that univer-
salism and self-direction should motivate resistance to conventional moral-
ity whereas conformity/tradition values should elicit positive reactions. The
variables are standardized to a 0–1 scale so all additive regression coefficients
can be interpreted as the percentage difference on the dependent variable
between those with the minimum and maximum scores on a given predictor,
all else being constant. The final model includes all value × education terms to
test whether education moderates these ties.
The baseline model estimates in Table 7.4 show that blacks may hold more
traditional attitudes than nonblacks (p < .10) and respondents who have earned
a high school or college degree hold more progressive attitudes than those
who did not graduate high school (p < .05). In a surprise finding, Hispanics
are less orthodox than whites. Moreover, the coefficient for the female variable
cannot be distinguished from zero.
With respect to the variables of prime theoretical interest, strong Republicans
are predicted to be 27% more culturally conservative than strong Democrats
(p < .001). However, the effect declines noticeably once personal values enter
the picture. Comparing the baseline and values models, we see that the magni-
tude of the party coefficient diminishes some 60% from 0.27 to 0.11 (it remains
significant at p < .001). Next, values perform well in the second model. The
values block is jointly significant (F = 154.56, p < .001), the R2 jumps from .19
to .54, and the coefficient for each value is correctly signed, statistically sig-
nificant, and substantively meaningful. The effects are illustrated in Figure
7.3, which plots predicted moral opinion across the full range of the univer-
salism, self-direction, and conformity/tradition variables. These plots present
the hypothetical maximum effects values have on opinion. For universalism,
those who prioritize understanding, appreciation, and tolerance for others
score 27% lower on traditional morality than individuals untroubled by the
fate of others. Likewise, those who believe independent thought and action
are extremely important score 15% lower on this policy principle than those
who deem the goals to be unimportant. For the conformity/tradition variable,
individuals who believe most fervently in obeying social norms score 54%
TA B L E 7 . 4 . THE ORIGINS OF TRADITIONAL MORALITY, 2008
Constant 0.52* 0.49* 0.51*

(.03) (.04) (.07)
Female 0.00 –0.01 –0.01
(.02) (.01) (.01)
Black 0.05+ 0.01 0.01
(.04) (.02) (.02)
Hispanic –0.05* –0.03+ –0.03+
(.03) (.02) (.02)
High School graduate or some college –0.05* 0.00 –0.01
(.03) (.02) (.08)
4-year college graduate –0.11* –0.03 –0.09
(.03) (.02) (.11)
(.02) (.02) (.02)
Universalism –0.27* –0.16
(.04) (.12)
Self-direction –0.15* –0.19*
(.03) (.08)
Conformity/tradition 0.54* 0.43*
(.03) (.10)
Universalism × High School graduate/ –0.14
some college (.13)
Universalism × college graduate –0.09

(.15)
Self-direction × High School graduate/ 0.04
some college (.09)
Self-direction × college graduate 0.06

(.10)
Conform/trad × High School graduate/ 0.13
some college (.10)
Conform/tradition × college graduate 0.13
(.11)
R2 .19 .54 .54
F statistic 27.90* 96.17* 59.69*
Joint F test for additional block – 154.56* 0.40
+
p < .10; *p < .05 (one-tailed).
Notes: OLS regression estimates. Unstandardized coefficients are reported. HC3 standard errors are in
parentheses. All variables are on a 0–1 range. Higher values on traditional morality reflect a conservative
opinion. Number of observations = 1000. Knowledge Networks sampling weights are used.
Source: 2008 Knowledge Networks survey.
179
Support for Traditional Morality 1.00
Universalism
0.75
Self-direction
0.50
0.25
Conformity/tradition
0.00
Low High
Importance of Value
Figure 7.3. Personal Values and Traditional Morality.

more conservative on traditional morality relative to those who reject conser-

vation goals.9
The last hypothesis under consideration is whether education strengthens
these connections. Probably not, according to the interaction model estimates
in the last column of Table 7.4. The formal test reveals that the interaction
coefficients cannot be distinguished from zero with any confidence (F = 0.40,
p = .88), the R2 holds steady across the models at .54, and the coefficient for
9
I remind readers that one of the items used to construct my dependent variable
reads as follows: “This country would have many fewer problems if there were
more emphasis on traditional family ties.” Two items in the conformity/tradition
scale mention tradition as well (“Tradition is important to her. She tries to respect
tradition” and “She thinks it is important to preserve traditional ideas of right
and wrong”). The value and opinion items differ in that the value measures ask
respondents to rate how much a hypothetical other for whom a given normative
belief serves as a guiding principle in his or her life is like them, thus tapping
personal priorities. The “family ties” question asks respondents how much they
agree or disagree with the sentiment expressed above and thereby more readily
evaluates attitudes. Because the word “tradition” appears in both the dependent
and independent variables, some might wonder if the observed value–principle
relationship is an artifact of semantic overlap. I tested this possibility by dropping
the “traditional family ties” item from the dependent variable scale and reesti-
mated the model. The coefficient for the conformity/tradition variable declines
from 0.54 to 0.49 and remains highly significant. I then dropped the two personal
value items mentioning tradition and modeled the four-item opinion scale as a
function of this pared down conformity scale and the other predictors. The value
coefficient remains powerful statistically and substantively (b = 0.34, p < .001).
each value × education term falls short of standard or even marginal levels of
significance. If we can place some confidence in education as a sophistication
proxy, then these results infirm the hypothesis that sophistication enhances
value-driven constraint. Instead, the evidence suggests once again that funda-
mental human values—especially universalism and conformity/tradition—
shape policy attitudes throughout the general public.
The Building Blocks of Military Strength

The 2011 YouGov/Polimetrix survey contains the instrumentation necessary
to model attitudes toward American military power. To measure the depen-
dent variable I combine responses from a pair of seven-point items into a sim-
ple additive scale. The first reads:
Some people think that in dealing with other nations our government
should be strong and tough. Suppose these people are at one end of this
scale, at point number 1. Others think that our government should be
understanding and flexible. Suppose these people are at the other end, at
point 7. And, of course, other people have opinions somewhere in between
at points 2, 3, 4, 5, or 6. Where would you place yourself on this scale?
The second item is:
Some people believe the United States should solve international problems
by using diplomacy and other forms of international pressure and use mil-
itary force only if absolutely necessary. Suppose we put such people at 1
on this scale. Others believe diplomacy and pressure often fail and the U.S.
must be ready to use military force. Suppose we put them at number 7.
Where would you place yourself on this scale?
The additive scale is coded so that the minimum score (0) represents the most
dovish position and the maximum score (1) represents the most hawkish
(Cronbach α = .48). Controls for sex, African-American, Hispanic, education,
and party identification, all measured as above, are included in the model.
I expect the force option attracts less support from women than men, from
minorities than whites, from the more educated relative to the less educated,
and from Democrats compared to Republicans.
I use the Schwartz PVP format to measure values. Each variable is com-
posed of two or more gender-matched items respondents viewed after seeing
this introduction: “Here we briefly describe some people. Please read each
description and tick the box on each line that shows how much each person
is or is not like you.” Six response options were offered: “Very much like me,”
“like me,” “somewhat like me,” “a little like me,” “not like me,” and “not like
me at all.” Item wording is as follows. For the benevolence value: “It is impor-
tant to her to be loyal to her friends. She wants to devote herself to people
close to her” and “It is important to her to respond to the needs of others. She
tries to support those she knows.” The universalism items are as follows: “She
thinks it is important that every person in the world should be treated equally.
She believes everyone should have equal opportunities in life” and “She wants
everyone to be treated justly, even people she doesn’t know. It is important to
her to protect the weak in society.” My power measure is constructed from the
following: “It is important to her to get respect from others. She wants people
to do what she says” and “It is important to her to be in charge and tell oth-
ers what to do. She likes to be the leader.” Next, I gauge security using this
pair of statements: “It is important to her to live in secure surroundings. She
avoids anything that might endanger her safety” and “Having a stable soci-
ety is important to her. She is concerned that the social order be protected.”
Tradition is tapped using the following: “It is important to her to be humble
and modest. She tries not to draw attention to herself” and “Tradition is impor-
tant to her. She tries to follow the customs handed down by her religion or her
family.” Conformity is measured this way: “She believes that people should
do what they’re told. She thinks people should follow the rules at all times,
even when no one is watching” and “It is important to her to be obedient. She
believes she should always show respect to her parents and to older people.”
As before, the conformity and tradition responses are joined together.
These items follow the recommendations made by Schwartz after exten-
sive pretesting (Schwartz, 2003). Analysis conducted elsewhere has shown
that the items tap the designated value dimensions and combine to form rea-
sonably reliable scales (Cronbach α varies from .55 to .75; see Goren, Reifler, &
Scotto, 2011). Recall from Table 7.2 that benevolence and universalism ought
to be negatively associated with military strength, whereas power, security,
and conformity/tradition values are expected to relate positively.
The YouGov survey contains a four-item measure of political knowledge.
The questions ask about the number of terms a president can serve, which
branch of government has responsibility for determining the constitutionality
of a law, the office held by Hillary Clinton, and the primary responsibility of
the United Nations (Cronbach α = .45). The mean for this scale is 3.13 correct
answers out of 4. Compared to past research on political knowledge (e.g., Delli
Carpini & Keeter, 1996) this estimate seems high. As I indicated in note 8,
the fact that respondents opted into the online panel may render them differ-
ent from the general public on political knowledge and similar unmeasured
variables. Unfortunately, no data are on hand to examine this possibility. In

any case, I use value × knowledge variables to test the sophistication interac-
tion hypothesis.
The linear regression estimates appear in Table 7.5. The baseline model
reveals that high school and college graduates adopt more dovish positions
on military strength than those lacking a high school degree. In contrast, my
supposition that gender and race impact this principle remains unsubstanti-
ated. Also, military power evokes stronger backing from GOP loyalists than
from Democrats. This result was expected, but it is worth noting that the party
coefficient declines some 30% in magnitude (from 0.27 to 0.19) after personal
values enter the equation.
Turning to column three, I find personal values efficacious one last time.
The possibility that the five value coefficients simultaneously equal zero can be
rejected (F = 31.17, p < .001) and the R2 rises from .20 in the baseline to .31 in the
second equation. The coefficients for universalism, power, and conformity/
tradition lie in the anticipated direction and are statistically and substantively
significant. The benevolence coefficient is incorrectly signed and marginally
significant (p < .10), but is substantively small. Also contrary to expectations,
security values appear, oddly enough, unrelated to hawk–dove orientations.
Figure 7.4 models the predicted hawk–dove score as a function of the three crit-
ical values. Movement from the low to high universalism score depresses sup-
port for military strength some 30%. Not surprisingly, those concerned about
the well being of other people look askance at the use of military power. Next,
respondents who esteem tradition and conformity are 27% more hawkish on
national security than subjects who reject status quo values. Lastly, movement
from the low to high score on power boosts support for military strength by
11%. Those who want to dominate others in their personal lives evidently seek
something similar for their country in the international arena.
The final model in Table 7.5 yields scant evidence that sophistication pro-
motes value-based reasoning. The joint F test is insignificant (F = 1.52, p = .18),
the R2 remains unchanged across the values and interaction models, and none
of the five multiplicative coefficients attain significance at conventional levels
(although three come close at p < .10). The result is that when it comes to judg-
ing military strength, most citizens can ground opinions in values (but recall
that the YouGov sample may be somewhat better informed than the public at
large).
So far, I have shown that personal values systematically affect all three
policy principles, controlling for a range of alternative explanations, party
identification first and foremost. The next question to pose is how the rela-
tive effects of party and values compare. Put otherwise, do values wield more
TA B L E 7 . 5 . THE ORIGINS OF MILITARY STRENGTH, 2011
Constant 0.49* 0.50* 0.43*

(.04) (.05) (.09)
Female –0.01 –0.01 –0.01
(.01) (.01) (.01)
Black –0.00 –0.02 –0.03
(.02) (.02) (.02)
Hispanic –0.02 –0.02 –0.02
(.02) (.02) (.02)
High School graduate or some college –0.07* –0.08* –0.07*
(.04) (.04) (.04)
4-year college graduate –0.12* –0.11* –0.11*
(.04) (.04) (.04)
(.02) (.02) (.02)
Benevolence 0.06+ 0.31+
(.04) (.21)
Universalism –0.30* –0.13
(.03) (.15)
Power 0.11* –0.06
(.03) (.11)
Security –0.03 0.03
(.03) (.13)
Conformity/tradition 0.27* –0.01
(.04) (.21)
Political sophistication 0.09
(.11)
Benevolence × sophistication –0.30+
(.23)
Universalism × sophistication –0.21
(.18)
Power × sophistication 0.22+
(.14)
Security × sophistication –0.07
(.16)
Conformity/tradition × sophistication 0.32+
(.24)
R2 .20 .31 .31
F statistic 44.43* 43.86* 32.23+
Joint F test for additional block – 31.17* 1.52
+
p <.10; *p < .05 (one-tailed).
Notes: OLS regression estimates. Unstandardized coefficients are reported. HC3 standard errors are
in parentheses. All variables are on a 0–1 range. Higher values on military strength reflect a hawkish
opinion. Number of observations = 1019. YouGov sampling weights are used.
Source: 2011 YouGov survey.
184
Support for Military Strength 1.00
Universalism
0.75
0.50 Power
0.25 Conformity/tradition
0.00
Low High
Importance of Value
Figure 7.4. Personal Values and Military Strength.

influence over core principles than party or vice versa? To address this, I test
whether the unstandardized regression coefficient for a given value is signif-
icantly larger than that of party identification. To the degree that values out-
pace party, the evidence will suggest central policy attitudes in mass belief
systems reflect far more than the residue of partisan bias.
Limited government comes first.10 My attention centers on the value dimen-
sions found to be most significant in the regression models. In Table 7.3, the
magnitude of the universalism coefficient (b = –0.38) exceeds that of the party
variable (b = 0.14) at p < .001. And although the conformity coefficient is some
40% larger than that of party (0.20 > 0.14), the null hypothesis of equal weights
cannot be rejected (p = .22). For moral traditionalism (see Table 7.4), the party
coefficient (0.11) lags universalism (–0.27), self-direction (–0.15), and conform-
ity/tradition (0.54) in absolute magnitude. The universalism and conformity
coefficients prove significantly larger than that of party (both p < .002) whereas
the self-direction and party coefficients are statistically indistinguishable. The
coefficients from the military strength equation are as follows: party = 0.19,
universalism = –0.30, conformity/tradition = 0.19, and power = 0.11. Formally,
the party effect trails that of universalism (p < .05), equals the conformity/tra-
dition effect (p = ns), and eclipses power (p < .05).11
10
Party identification is positively related to each policy principle. Some of the
personal values are positively related to a policy variable, but others carry a
negative sign. When party and value are positively related to the dependent
variable, I test the null hypothesis: bPID – bValue = 0. When party and value have
opposing signs, I test this null hypothesis: bPID + bValue = 0.
11
For limited government the standardized beta equals –.29 for universalism, .24
for party identification, and .16 for conformity. For moral traditionalism the
Collectively, my appraisal suggests that when citizens formulate their posi-

tions on limited government, traditional morality, and military strength, basic
personal values usually match and more often surpass the influence of party
identification. Of the eight key tests, the value coefficient is significantly larger
than party four times and of equivalent magnitude in three other tests. Party
triumphs in a single case. Moreover, when we recognize that multiple values
simultaneously impinge upon each policy principle, the influence of values
relative to partisanship is further emphasized.12 Broadly speaking, then, basic
human values, universalism and conformity/tradition primarily, seem to
carry more weight than party identification in shaping domain-specific prin-
ciples. This conclusion parallels what we saw in Chapter 6 where these prin-
ciples were shown to be stronger predictors of issue preferences than party
identification.
The Relationship between Personal Values and Party Identification

To this point I have examined the impact of personal values and partisan loy-
alties on policy attitudes while sidestepping the values–party relationship. Do
personal values shape partisan identities or is it the other way round? A good
deal of political science research, my own included, suggests that the causal
arrow runs from party to values (Goren, 2005; Goren et al., 2009). However, this
work focuses on core political values that, for the reasons delineated above,
differ in key respects from the personal values construct. Personal values
serve as evaluative standards that have implications for daily living whereas
core political values structure judgment in the political arena alone. Moreover,
measures of core political values do not tap personal importance and may
be conflated with the types of policy attitudes they purport to explain. The
personal value measures suffer no such flaws. Hence, results showing parti-
san identities shape political values cannot speak directly to the relationship
between prepolitical, personal values and party ties.
On this score, I suggest personal values come before party ties in a tempo-
ral sequence. Because personal values function as transsituational guides that
standardized coefficients are .58 for conformity/tradition, –.22 for universal-

ism, .18 for party, and –.17 for self-direction. For military strength the coeffi-
cients equal .30 for the party variable, –.29 for universalism, .24 for conformity/
tradition, and .12 for power.
12
I restricted my comparisons to the largest predictors and ignored those with
marginal or insignificant effects. A focus on the key values seems defensible
so long as it is acknowledged that party outperforms the less important value
dimensions.
shape all kinds of preferences and behaviors outside of politics, it seems plau-
sible that they would exercise similar influence on those occasions when atten-
tion turns to politics. Personal values shape what we do for a living, where we
live, who we spend time with, what we spend our money on, and so on. Such
leverage could extend to the political parties with which we identify (Rokeach,
1973). To make the same point another way, values have greater motivational
efficacy than partisanship. We can easily imagine people who go into social
work, organize local food drives, and volunteer willingly and repeatedly to
watch their neighbors’ children because they are motivated by benevolence
and universalism values, by a keen desire to help those near and far. The pos-
sibility that they go into social work or lead food drives because they are a
Democrat seems farfetched. Put simply, the role values play in human judg-
ment renders them more likely to shape partisan loyalties than the reverse.
Other considerations reinforce this argument. A great deal of evidence con-
firms that everyone holds personal values, but the same does not hold true
for partisan attachments. Two bodies of work speak to this point. As reviewed
above, researchers have found that the posited value structure holds in hun-
dreds of samples across scores of countries. The structure has been discovered
in student samples, adult convenience samples, and representative national
samples on six continents over the past three decades. Discrepancies emerge
at times, but for the most part this value hierarchy seems to be a common fea-
ture of the human psyche (much like the five factor model of personality). This
holds true in countries such as the United States that have long-established
party systems, but the value structure also turns up in places where party
organizations and ties are weak or nonexistent. Because values take hold in the
minds of everyone and partisan identities do not, the former can be viewed as
more widely available in, and central elements of, belief systems. The second
body of research that speaks to the temporal priority of values deals with the
distribution of partisanship in the U.S. electorate. Studies always find a large
fraction of the public feels little or no sense of affinity with either party. In the
2008 NES, for instance, 29% of the sample identified weakly with the parties
and another 11% professed complete independence. In terms of raw numbers,
this means that approximately 92,000,000 voting age adults lacked strong
party ties.13 No one supposes that 92,000,000 Americans lack basic values.
To sum up, personal values impinge on a much wider range of judgments
and behaviors than partisan identities. The reach of values extends to almost
13
To calculate this estimate I multiplied a voting age population figure of
230,782,870 (see http://elections.gmu.edu/Turnout_2008G.html) by .40.
all aspects of life, but partisanship functions only in the political domain.
Moreover, values exist on a nearly universal basis, but partisan identities do
not. The implications are clear: personal values likely hold temporal priority
over party ties and thus are more likely to shape such ties than to be shaped
by them. Ideally, I would test this proposition using panel data containing
repeated measures of Schwartz values and party identification that would
allow me to estimate dynamic constraint models or experimental work in
which the consequences of value and partisan identity manipulations might
be explored. No such data are currently available; therefore, my assumption
that values take precedence over party identification is necessarily tentative.
It is not unreasonable, perhaps, but is tentative all the same. Readers should
keep this caution in mind as we proceed.
I now take a look at the relationship between personal values and parti-
san loyalties using data from the three surveys employed earlier. The value–
party expectations are straightforward because the groups that comprise the
party coalitions and the policies pursued by each party have clear implica-
tions for values. The Democratic Party has long represented disadvantaged
and marginalized groups in American society and sought to redress these
imbalances through political reform. The GOP has represented groups whose
positions in the economic, social, and political spheres are established more
securely; therefore, the party has frequently defended status quo arrange-
ments favored by their constituents (Gerring, 1998). Given these profiles, the
party labels should attract support from people with the following value hier-
archies. Those dedicated to self-transcendence values (i.e., benevolence and
universalism) should gravitate toward the Democratic Party more readily
than those who display little affection for these values. Moreover, those who
pledge allegiance to conservation values (i.e., security and conformity/tradi-
tion) should find the GOP brand more appealing compared to those who give
conservation goals less priority. I have no clear predictions regarding self-
direction, achievement, and power values. Both parties favor self-direction
and achievement, though they differ on how to realize these goals. Finally,
neither party is crass enough to come out and say it stands for controlling or
dominating others.
Table 7.6 arrays the regression estimates from the 2007, 2008, and 2011 surveys
in the second through fourth columns, respectively. Demographic controls are
included for sex, African-American, Hispanic, income, and education, all coded
as above. The dependent variable is coded so that higher scores denote GOP lean-
ings; hence the regression coefficients for universalism and benevolence should
be negative and those for security and conformity/tradition should be positive.
TA B L E 7 . 6 . THE ORIGINS OF PARTY IDENTIFICATION, 2007–2011
2007 2008 2011
Constant 0.56* 0.36* 0.63*

(.08) (.09) (.08)
Female –0.06* –0.02 –0.05*
(.02) (.02) (.02)
Black –0.28* –0.28* –0.31*
(.03) (.03) (.03)
Hispanic –0.14* –0.12* –0.12*
(.03) (.03) (.04)
Second income tercile 0.04* 0.04+ 0.02
(.02) (.03) (.03)
Third income tercile 0.11* 0.08* 0.03
(.03) (.03) (.03)
HS graduate or some college 0.04+ 0.12* 0.05
(.03) (.03) (.06)
4-year college graduate –0.02 0.13* 0.03
(.04) (.04) (.06)
Benevolence 0.06
(.06)
Universalism –0.56* –0.44* –0.55*
(.07) (.08) (.05)
Self-direction –0.06 –0.00
(.09) (.05)
Achievement 0.16*
(.09)
Power –0.03 –0.01
(.06) (.04)
Security –0.08+
(.06)
Tradition/conformity 0.35* 0.49* 0.39*
(.07) (.05) (.06)
R2 .17 .26 .23
F statistic 23.10* 24.71* 36.35*
Joint F test for values block 15.08* 40.66* 38.48*
+
p < .10; *p < .05 (one-tailed).
Notes: OLS regression estimates. Unstandardized coefficients are reported. HC3 standard
errors are in parentheses. All variables are on a 0–1 range. Higher values on party
identification reflect Republican attachments.
Sources: 2007 Knowledge Networks survey, 2008 Knowledge Networks survey, 2011
YouGov survey.
189
The estimates show the Democratic brand usually garners more support from
women than from men, from blacks and Hispanics compared to whites, and
from the less educated (marginally) and poor relative to their more educated
and affluent counterparts. None of this is surprising.
More importantly for my purposes, basic human values covary with party
identification as predicted (all joint F tests are significant at p < .001). A univer-
salism effect emerges in all three models; its coefficient is highly significant (p <
.001) and substantively powerful. To illustrate, Figure 7.5 plots predicted parti-
san self-placements across the range of universalism for each year (remember
these illustrate hypothetical maximum effects). The downwardly sloping lines
reveal that the more emphasis respondents place on the well being of others
the weaker their ties to the GOP. Across the three surveys, those who attach
the most importance to universalism score about 44–56% more Democratic
than those who attach the least importance to this value, holding demograph-
ics and other values constant. Conformity/tradition values also behave as
predicted. Each coefficient lies in the posited direction and proves to be highly
significant (p < .001). Figure 7.6 plots partisanship as a function of these goals.
Movement from the minimum to maximum score on conformity/tradition
leads to increases of 35–49% in the direction of GOP identification, all else
being constant. Overall, the regression estimates and plots suggest that uni-
versalism and conformity/tradition values powerfully shape party ties. Three
points warrant emphasis. The first is that the effect sizes are always large; the
second is that consistent effects are obtained even though different question
formats and items are used to measure values; and the third is that the results
are robust over time.
1.00
0.83
Party ID (High = Rep)
0.66
0.50
0.33
0.16
2007 2008 2011
0.00
Low High
Importance of Value
Figure 7.5. Universalism and Party Identification.

1.00
0.83
Party ID (High = Rep)
0.66
0.50
0.33
0.16
2007 2008 2011
0.00
Low High
Importance of Value
Figure 7.6. Conformity/Tradition and Party Identification.

Among the remaining values only two, achievement in 2007 and security in
2011, impact party. The 0.16 achievement coefficient (p < .05) implies that the
more highly we rank personal success, the more likely we are to identify with
the GOP. Security may be negatively related to Republican partisanship in
2011, but the effect is substantively small and marginally significant (b = –0.08,
p < .10). Neither power in 2007 or 2011 nor self-direction in 2007 or 2008 nor
benevolence in 2011 matter. With the exception of benevolence, and security,
I had no firm priors about how these beliefs would affect party. Finally, I find
no evidence that the values–party link is stronger among more educated or
informed respondents (data not shown).
My bottom line conclusion is that universalism and conformity/tradition
hold considerable sway over partisan attachments. So far as my assumption
about the direction of causality is on the mark, the results imply that these two
value domains powerfully shape the positions people take on limited govern-
ment, traditional morality, and military strength. Basic human values exert
strong direct effects on general policy principles that match or, more typically,
surpass those of partisanship. Additionally, values can shape principles indi-
rectly through their influence on party identification, the effects of which also
prove quite substantial (again, experimental and panel studies are needed to
properly evaluate this hypothesis).
CONCLUSIONS
This chapter has examined two factors that correlate strongly with limited
government, traditional morality, and military strength. My evidence sug-
gests that personal values wield a good deal of influence over the positions
people take on these principles. Universalism and conformity/tradition pri-

marily, and self-direction and achievement secondarily, shape judgments
about what government is supposed to do in the economic welfare domain.
Next, universalism, self-direction, and conformity/tradition values con-
strain attitudes toward the preservation of moral customs. And universalism,
power, and conformity/tradition values exert some leverage over the military
strength dimension. Critically, the ability to root policy principles in basic val-
ues does not appear to be conditional on education or political knowledge.
These results suggest everyday people do a pretty good job in deducing policy
principles from broader beliefs. My lack of a direct knowledge measure in the
Knowledge Networks samples and the possibility that the YouGov sample
is more informed then the public at large temper these conclusions to some
degree.
The direct effects of personal values either match or more often eclipse those
of party identification. These results echo the finding in Chapter 6 that policy
principles outweigh partisanship in shaping issue preferences. In light of the
presumed dominance of party identification in models of political preference
formation, the performance turned in by personal values is impressive. And
if we take into account the indirect impact values may have on core principles
through party identification, the motivational potency of values becomes even
more apparent. There is no denying that party identification, long viewed as
the dominant political heuristic in the fields of political psychology, public
opinion, and political behavior, matters a good deal. Personal values seem to
matter at least as much and perhaps more.
Regarding the implications for voter competence, my results suggest that
citizens render thoughtful political choices. To the extent that policy principles
depend on little more than gut-level reactions to party labels, the reasoning
capacities of ordinary citizens seem overly reflexive and simplistic. But once
personal values enter the picture, the limitations of the conventional view
are plain to see. Insofar as policy principles are grounded heavily in abstract
beliefs about desirable end-states and modes of conduct that citizens apply
liberally as evaluative standards in their daily lives, the case for their polit-
ical competence grows stronger. Citizens do not deliberate at length about
the political choices confronting them nor engage in complicated cost–benefit
analyses, but they have sound normative rationales for the positions they take
on the key policy cleavages that define American politics. Calls for collective
or government action that are consistent with their values command support;
demands that contravene their values arouse opposition. That seems emi-
nently reasonable.
To this point in the book, my results establish that policy principles are
available in the minds of citizens, function as central decision-making heuris-
tics, and are infused with normative beliefs about what is important in life. All
of this seems to hold to a similar degree for politically sophisticated and unso-
phisticated individuals. My final task requires examining systematically how
policy principles shape voter choice in U.S. presidential elections. Chapter 8
takes up that charge.
CHAPTER 8
The Electoral Consequences of Policy

Principles
The key question posed by my book is whether policy principles shape the
presidential vote to a comparable degree for politically sophisticated and
unsophisticated individuals. I have argued that policy principles will guide
voter choice if three conditions are satisfied. The first is that each principle
must be present in their minds. Insofar as principles are available, they
have the potential to impact candidate choice. The measurement model
and attitude stability estimates reported in Chapter 5 establish that citizens
meet the availability condition. Second, people must use these principles to
guide their reactions to short-term developments such as issue controver-
sies. To the extent that they do so, policy principles are well placed to guide
electoral choice. The dynamic constraint and regression results summarized
in Chapter 6 suggest that individuals meet the centrality condition. Third,
voters must determine which candidate lies closer to them on a given con-
tinuum and then select that candidate. If people fulfill this position match-
ing condition, the case for the third face of policy voting will be complete.
In this chapter, I model the relationship between policy principles and voter
choice using data from the 1988–2008 National Election Study (NES) cross-sec-
tional surveys. My efforts begin with an overview of the leading theories of
electoral behavior: the partisan voter model, the retrospective voter model, and
the ideological voter model. These theories suggest a set of variables that must
be included in the candidate preference equations. Moreover, each perspective
has implications regarding voter competence. After describing the data and
measures, I estimate the statistical models and summarize the results. To fore-
shadow my findings: (1) policy principles consistently and powerfully affect
candidate preferences for all voters; (2) liberal–conservative attitudes affect
the vote only among the politically sophisticated; and (3) the effects of issues
prove inconsistent, though there is a tendency for sophistication to enhance
195
issue voting. In short, the first and second faces of policy voting elude most
people, but the third face encompasses all electors.
THEORIES OF VOTER CHOICE

The Partisan Voter
The study of electoral choice begins with party identification (Campbell et al.,
1960). Scores of studies have established that latent partisan attachments are
available, crystallized, and central elements in mass belief systems, so their hold
on political judgment is understandable. Nowhere is this more evident than in
the realm of voting behavior. When it comes to presidential elections, those who
feel a sense of attachment to one of the two major parties usually vote for their
side’s nominees. From 1952–2008, 90% or more of strong identifiers have cast a
partisan ballot and roughly 75% of weak identifiers have followed suit. Even
those who claim to be independent but feel a bit closer to one party have sup-
ported its nominees about 75% of the time (Abramson et al., 2009).
What underlies the influence of partisanship? Two mechanisms are para-
mount. The first centers on group solidarity. Those who identify with a party
pull for their “team” to win elections. They do so in part because the par-
ties are seen as representing the social groups with which people identify
(Green et al., 2002). To illustrate, an upper income business owner may have
an affinity for the GOP because he or she sees it as the party for people like
them. In contrast, a public school teacher and union member may be drawn
to the Democratic Party by virtue of its pro-teacher and pro-union stances.
When Election Day arrives, the businessman votes Republican and the school
teacher goes for the Democrat.
The second mechanism underlying party voting is partisan bias. To revisit an
idea raised earlier in the book, partisanship leads people to take positive views
of most things associated with their party while simultaneously denigrating
the opposition. Democrats and Republicans perceive, interpret, evaluate, and
integrate new information in a manner that yields the desired partisan conclu-
sions. The power of the partisan lens is subtle enough that its victims usually
remain ignorant of its influence.1 This influence is especially potent in the realm
of candidate evaluation. As Stokes (1966, p. 127) has put it: “For the millions of
Americans identified with a particular party, the candidate bearing the party
symbol tends to become more of a lion, his opponent more of a wolf.”
1
But not always. Some partisans cast a more critical eye toward their party when
conditions deteriorate under its stewardship or evaluate the opposition more
favorably when times are good (Lavine et al., 2012).
The Electoral Consequences of Policy Principles 197
To reiterate, group solidarity and biased perception motivate partisan vot-

ing. Note that these mechanisms are mutually reinforcing. Those who identify
with the GOP because it represents people like them take a pro-Republican
view of the political world most of the time. The same holds true for Democratic
identifiers. Although the impact these mechanisms have on electoral choice
varies across individuals and circumstances, there is no doubt that party ties
weigh heavily in the electoral calculus of the American voter.
I make two final points about the partisan voter model. First, these identi-
ties are for the most part imbued with minimal policy content (Bartels, 2002;
Green & Palmquist, 1990). Even among the narrow subsets of the public
for whom policy attitudes shape party ties, the reciprocal effects of party
on policy are typically larger (Carsey & Layman, 2006). To the extent that
policy attitudes shape the vote, such influence is rarely transmitted through
partisan feelings. Next, some of the motives underlying party voting cast it
in a normatively unappealing light. Although party labels can serve as an
effective heuristic for voters under some circumstances, the fact that such
usage often amounts to little more than vapid partisan cheerleading, distorts
political views, and reflects little policy content is unsettling. Of course, the
fact that party identification is grounded in social group identities reflects
positively on the status of partisanship. Insofar as people align with parties
that represent their group interests, that is to be applauded. Moreover, if it
is accepted that basic human values inform partisan identities, as I argued
previously, an even stronger case can be made for the normative appeal of
partisan voting.
The Retrospective Voter

A second leading theory of electoral choice centers on retrospective judg-
ments (Fiorina, 1981; Key, 1966). The logic of this perspective is captured well
by remarks Ronald Reagan made during the final 1980 presidential debate.
Reagan asked the audience:
Is there more or less unemployment in the country than there was four
years ago? Is America as respected throughout the world as it was? Do
you feel that our security is as safe, that we’re as strong as we were four
years ago?
People reflect on questions such as these, formulate bottom line judgments

about conditions under the president, and use these to inform their votes. If
things seem better, they support the incumbent. If conditions are deteriorat-
ing, they opt to replace him.
Accumulated research suggests that people reward or punish the president

in two key areas. When voters believe the economy is surging and jobs are
plentiful, they back the incumbent and his party on Election Day. But if they
feel the economy is slumping and jobs are scarce, incumbent support withers.
A similar process plays out in the domain of foreign policy. If peace and stabil-
ity reign in the international arena, voters stick with the commander in chief.
But when the United States is bogged down in a long and costly war or, short
of this, the national interest is somehow imperiled, the president fares poorly.
What does retrospective voting imply about voter competence? This model
has been deployed to rebut the charge that the American voter is too inept
about ideology and confused about the issues to vote sensibly. The strongest
proponents of this view see retrospective voting as quite rational. It is not
hard to see why. Democratic elections are supposed to hold representatives
accountable for their performance in office. To the extent that a president’s
term is marked by peace and prosperity, a reelection vote appears reasonable.
Conversely, if the economy falters badly or American blood and treasure are
squandered in foreign lands, voters should demand change. Knowing this,
politicians work hard to deliver these desiderata.
Although there is much to be said for this position, scholars have raised
questions about the quality of retrospective judgments. Three criticisms are
especially pertinent. First, the connection between perception and reality is
often tenuous because many individuals lack accurate information about con-
ditions at home and abroad (Conover, Feldman, & Knight 1986; Kull et al.,
2003/2004). Second, retrospective judgments are susceptible to the influence
of very recent events and are often detached from what happened before, as
if the president is judged only on the last year of his term, a shortsighted and
incomplete timeframe (Bartels, 2008). Third, retrospective judgments are dis-
torted by partisan perception. The president’s copartisans in the electorate
almost always see conditions better than his partisan opponents, regardless
of the actual state of affairs (Bartels, 2002). To sum up, although the empirical
connection between retrospective judgments and the presidential vote is well
established, it is not clear that this redounds to the credit of voters.
The Ideological Voter

Liberal and conservative attitudes are conceptualized here as psychologi-
cal tendencies to evaluate liberal–conservative labels with some degree of
favor or disfavor. The question then becomes what meaning people attach
to these labels. In Chapter 3, we saw that many people have little or no idea
what liberalism and conservatism mean as political terms. Recall Lewis-Beck
et al. (2008) discovered that about 11% of the public actively used liberal–
conservative frames of reference to evaluate the presidential candidates and
major parties in 2000. Likewise, Bennett’s (1995) careful analysis of 1980–1992
NES data showed that when asked to provide up to three definitions of what
liberal and conservative mean (for a total of six), nearly half the public could
do no better than a single answer. At the high end, only 12% of the samples
furnished five or six passable definitions. These results, which differ margin-
ally from reports in seminal works, suggest that ideological labels are opaque
if not impenetrable to much of the electorate.
But this does not apply to everyone. Those who know a good deal about
public affairs associate some basic political ideas and symbols with ideo-
logical labels and thus form liberal–conservative attitudes (Sears, 2001;
Sniderman et al., 1991). For the politically aware, these labels may connote
a social group or two such as “feminists” or “Christian fundamentalists.”
For others, the labels connote a symbol such as “big government” or “free
enterprise.” Although these associations are understandable, their lack of
breadth and depth cannot be ignored. Among those who hold genuine lib-
eral–conservative attitudes, such evaluations are rarely based on overarch-
ing political philosophies or summaries of multiple issue positions. Instead,
these attitudes are simple reactions to a limited number of attitude objects.
By my reading of the evidence, perhaps 10% of the electorate acquires
a decent understanding of ideological labels. For another 25–30% or so, atti-
tudes toward liberal–conservative labels reflect little more than impressions of
a couple of groups or symbols. And for the rest of the public, there is little or
nothing. Converse is right on this score.
When it comes to evaluating presidential candidates, those who know
something about ideological labels can satisfy the position-matching condition
of policy voting. The process can be illustrated using a pair of examples from
the 1992 campaign. At the GOP convention that summer, Patrick Buchanan
described the gathering of Democratic delegates several weeks earlier as “that
giant masquerade ball . . . where 20,000 liberals and radicals came dressed up
as moderates and centrists in the greatest single exhibition of cross-dressing
in American political history.”2 Three nights later President Bush charged that
“the liberal McGovern wing of the other party, including my opponent, con-
sistently made the wrong choices.”
To those versed in the argot of ideology, references to “radicals and liber-
als . . . dressed up as moderates and centrists” and “the liberal McGovern
2
http://www.americanrhetoric.com/speeches/patrickbuchanan1992rnc.htm.
wing of the other party” had a familiar ring. Some interpreted Buchanan’s
diatribe as an accusation of ideological hypocrisy. To other viewers, Bush’s
allusion to liberal McGovernites evoked stereotypes of Democratic politi-
cians as naive peaceniks who could not be trusted to defend the nation.
Whatever meaning they ascribed to these labels, the politically aware could
see which contender was closer to them on the continuum and vote for the
better match. Nothing comparable would have occurred in the minds of
politically unsophisticated viewers. Lacking prior knowledge about what
the labels meant, they would have experienced difficulty extracting useful
information from these cues. Apart perhaps from a vague sense of unease,
allusions to “radicals and liberals . . . dressed up as moderates and centrists”
and the “liberal McGovern wing” of the Democratic Party would have
left many viewers baffled. In other words, the first face of policy voting is
a function of political sophistication.
The normative implications of the ideological voter model can be sum-
marized as follows. On the one hand, some researchers take comfort in
findings that liberal or conservative orientations guide voter choice for
a portion of the electorate (Delli Carpini & Keeter, 1996; Lau & Redlawsk,
2006; Neuman, 1986). This optimistic reading is based on the assumption
that liberal–conservative orientations are infused with a healthy dose of
policy content. In this view, the politically sophisticated use their votes to
communicate about policies they want the federal government to pursue.
But as pointed out repeatedly, for most people liberal and conservative dis-
positions function neither as abstract worldviews nor as issue preference
markers. Because liberal–conservative attitudes reflect simplistic global
judgments about a few social groups or evocative symbols, the fact that the
politically knowledgeable use them to guide their votes is not necessarily
encouraging from a normative point of view.
The Principled Voter

Partisan identities, retrospective judgments, and liberal–conservative attitudes
(among the politically aware) represent key determinants of the presidential
vote. In addition to the usual suspects, I contend that limited government, tra-
ditional morality, and military strength exert considerable pull over candidate
choice. In contrast to the partisan voter, retrospective voter, and ideological
voter models, there exists little evidence corroborating this claim. Miller and
Shanks (1996) represent a rare exception, but they do not focus on sophistica-
tion interaction effects in their study.
If policy principles are to guide candidate choice, they must be available

in long-term memory, central elements in political attitude structures, and
matched to the candidate positions. All citizens should develop and maintain
attitudes toward these dimensions because they reflect foundational politi-
cal divisions in American society that antedate elections and, consequently,
are visible enough for all to see and evaluate. Furthermore, citizens use these
judgmental yardsticks to guide the positions they take on sundry political
controversies. In so doing, they gain the requisite experience applying core
principles to other short-term judgments, the most important of which is, of
course, the presidential vote. The empirical tests in the Chapters 5 and 6 vali-
dated these propositions.
The final condition asks whether voters compare their positions to those
held by the candidates and select the one lying closer. I argued in Chapter 4
that most people are capable of satisfying this requirement. To review, citi-
zens can discern where the candidates stand on policy principles via one of
three routes. (Of course, some people know this at the start of the campaign.)
First, they can acquire information directly from the campaign. The campaign
and attendant media coverage in hard and soft news outlets supply voters
with an ongoing stream of redundant cues about candidate stances (Just et al.,
1996; Rahn et al., 1990). Whether from the local television news preceding the
weather report, campaign commercials airing during Monday Night Football
or Grey’s Anatomy, time spent watching convention speeches or a presidential
debate, or a clip from the The Daily Show or The O’Reilly Factor downloaded
from the internet, those planning to vote have multiple opportunities to learn
candidate positions on the main policy cleavages.
The second means for accomplishing this is through party-based infer-
ence. Voters can exploit party schema to deduce candidate positions (Conover
& Feldman, 1989; Feldman & Conover, 1983). Diagnostic categories associ-
ated with party labels facilitate this process. For instance, stereotypes link
party brands to familiar social groups, such as “Democrats are for the poor,
Republicans favor the rich.” The same holds true for party issue handling rep-
utations. Democrats are deemed more competent on social welfare whereas
the GOP receives higher marks for defending traditional values and national
security (Petrocik, 1996). Knowing this, voters can deduce where the respec-
tive nominees stand on each principle. Those who see the Democratic Party as
the champion of the “have nots” and the GOP as the party of the “haves” can
presume that the Democratic candidate is more committed to activist govern-
ment than his rival. Similarly, those who see the GOP as the religious party
may surmise its nominee will embrace traditional morality with greater fervor
than the Democratic opposition.
Most voters probably learn candidate positions from the campaign or
through party-based inference. The intention to vote motivates them to do
just this. But even if information from these sources escapes some people,
communication flows through social networks help them fill in the blanks.
People who pay little attention to hard news media and have impoverished
party schemas may learn about candidate positions from a soft news pro-
gram such as Today, from a passing conversation at the lunch counter, from
a chat with neighbors or friends at the local diner, from colleagues at work,
in houses of worship, and so on (Huckfeldt & Sprague, 1995; Popkin, 1994).
Through social channels such as these, cues about candidate positions on gov-
ernment aid, moral standards, and force versus diplomacy can reach voters.
Fiorina (1990, p. 338) sums up this view quite nicely: “Citizens often receive
information in the course of doing other things. There is no question of infor-
mation costs, no question of deciding to gather information as opposed to
doing something else.”
To illustrate the possibilities of campaign learning, I return to the GOP con-
vention speeches invoked above. To begin with Buchanan, although his refer-
ence to cross-dressing radicals and liberals may have mystified millions of
viewers, there was no ambiguity later on in his speech.
Friends, this election is about more than who gets what. It is about who we
are. It is about what we believe and what we stand for as Americans. There
is a religious war going on in this country. It is a cultural war, as critical to
the kind of nation we shall be as the Cold War itself. For this war is for the
soul of America. And in that struggle for the soul of America, Clinton and
Clinton are on the other side, and George Bush is on our side.
Those who watched it live, or saw a snippet rebroadcast on the news at a later
date, or heard about it at work or at church or softball practice, could quickly
realize that President Bush stood to the right of Bill Clinton on traditional
morality. Knowing this, citizens could assess, albeit roughly, where they stood
relative to each contender, thereby increasing the odds of voting for the appro-
priate nominee.
Three nights later, President Bush chastised the Democratic Party for its
alleged spinelessness during the Cold War:
In the seventies, they wanted a hollow army. We wanted a strong fighting

force. In the eighties . . . they wanted a nuclear freeze, and we insisted on
peace through strength. From Angola to Central America, they said, ‘Let’s
negotiate, deliberate, procrastinate.’ We said, ‘Just stand up for freedom.’
A bit later in the speech Bush returned to this theme by declaring that “We
must be a military superpower.” Those who knew nothing about Cold War
proxy battles in Angola and Central America could not mistake what the
president meant by “they wanted a hollow army” and “a nuclear freeze” and
“military superpower.” When encountering this information, whether dur-
ing the speech or in subsequent debates or campaign commercials or con-
versations with friends and co-workers, the implication could not be missed.
Republicans such as Bush were tougher on national security than Democrats
such as Clinton.
I offer a final comment on the normative appeal of the principled voter
model. Policy principles summarize what people want the federal govern-
ment to do in key issue areas and thereby provide a means for substantive
policy representation. Furthermore, when presidents and their allies pur-
sue actions at odds with the policy desires of voters, the latter can support
the challenger when the time comes. Hence, core principles also provide
citizens with a mechanism for holding elected officials accountable for
what they do in office. And because these policy attitudes are grounded
deeply in personal beliefs about what is important in life (see Chapter 7),
they cannot be dismissed as the byproduct of partisan biases or out-group
antipathies. So far as citizens ground candidate preferences in limited gov-
ernment, moral traditionalism, and military strength, the charge that their
votes are bereft of policy concerns, along with the corollary that they are
politically incompetent, will necessitate revision.
Expectations
The study of voting behavior has been animated by three theories. The par-
tisan voter model emphasizes the impact enduring partisan loyalties have
on electoral choice. The retrospective voter model stresses the role played
by judgments about national and international conditions. The ideological
voter model contends that liberal–conservative attitudes shape the choices of
the politically informed. Without denying the utility of these theories, I have
developed an alternative model that centers on basic policy principles.3 The
key theoretical supposition of the principled voter model is that attitudes
toward limited government, traditional moral standards, and military power
3
I take up issue voting later in the chapter.
shape candidate choice to a comparable degree for citizens across the sophis-
tication spectrum. The model’s key normative claim is that reliance on policy
principles speaks well of the political competence of regular people.
From the above frameworks, I derive the following working hypotheses.
First, the more strongly citizens identity with a political party the higher the
likelihood they will vote for their side’s nominee. Second, positive retrospec-
tive judgments will enhance the probability of a pro-incumbent vote. Third, the
more positively citizens react to an ideological label the greater their propen-
sity to support the ideologically congruent candidate, conditional on political
sophistication. Fourth, stronger left-wing positions on each policy principle
should lead to firmer support for the Democratic candidate.
DATA AND MEASURES

I rely on NES cross-sectional data from the 1988–2008 presidential elections to
measure the concepts and test my hypotheses. My dependent variable is the
self-reported presidential vote, coded 1 for a Democratic preference and 0 for
a GOP preference. For 1992, I code the vote this way: 2 = Perot vote, 1 = Clinton
vote, and 0 = Bush vote.
Turning to the explanatory variables, I measure party identification with
the seven-point self-categorization item, rescaled onto a 0–1 range with higher
scores reflecting stronger GOP ties. The party coefficient should be negatively
related to the likelihood of a Democratic vote. Next, I employ several items
to tap retrospective economic and foreign policy judgments. Economic judg-
ments are measured by combining respondent ratings of how much better or
worse “the nation’s economy” and “unemployment in the country” are into
a simple additive scale.4 This variable also has a 0–1 range with higher values
reflecting more favorable evaluations. I anticipate positive coefficients in 1996
and 2000, reflecting support for the incumbent Democratic administration,
and negative coefficients in 1988, 1992, 2004, and 2008, signifying a lack of
support for Democratic challengers. Foreign policy judgments are assessed
with a question about whether over the past year “the United States’ position
in the world has grown weaker, stayed about the same, or grown stronger.”
I create dummy variables for the “stronger” and “same” categories, leaving
4
The latter item is not available in 1996 and not used in 2004 due to a program-
ming error in which the item was not administered to 125 respondents. Rather
than losing these cases, which comprise nearly 12% of the sample, I drop the
unemployment item from the analysis.
respondents who said “weaker” to serve as the reference group. Those who
believe the U.S. position has held steady or improved should favor the incum-
bent, leading me to expect positive coefficients in 1996 and 2000 and negative
coefficients in the remaining years.
I measure liberal–conservative attitudes with the self-placement item
that has respondents describe their “political views” using a seven-point
scale ranging from “extremely liberal” to “extremely conservative.” In each
survey about a quarter to a third of the sample says they “haven’t thought
much about this” or “don’t know.” These respondents receive a follow-up
question asking “If you had to choose, would you consider yourself a lib-
eral or a conservative?” Respondents who selected the “moderate, middle
of the road” option on the seven-point scale get the same probe. Given
the problems with no opinion filters (see Chapter 2), I use the probe to
locate respondents on the seven-point scale. This procedure helps ensure
that the estimation samples are not upwardly biased on the knowledge
variable since those who fail to answer the initial question are typically less
informed than those who comply. The liberal–conservative variable lies on
a 0–1 scale and is keyed so that higher values denote more conservative
leanings, which should depress Democratic voting. To test whether polit-
ical sophistication moderates this relationship, I take the product of the
liberal–conservative and sophistication variables (tapped using the multi-
item political knowledge scales employed in earlier chapters). The coeffi-
cient for the multiplicative term should be negative, which would indicate
that the ideological effect becomes increasingly pronounced at higher lev-
els of sophistication.
Policy principles are measured the usual way. To remind readers, limited
government is a multiple indicator scale composed of four to six federal spend-
ing items depending on the survey (the Cronbach α coefficient varies from .65
to .76). The scale is coded such that higher scores reflect more support for
limited government, so it should be negatively related to voting Democratic.
For moral traditionalism I use the standard quartet of items (the α coefficient
varies from .55 to .67). Because higher scores reflect increasingly conventional
outlooks, this variable should also be negatively related to a Democratic vote.
Lastly, I measure hawk–dove orientations using one to three items depending
on the survey (the α coefficient varies from .51 to .68). Unfortunately, mea-
sures of abstract ideas about military force and diplomacy do not appear in
the 2000 and 2008 surveys. Higher scores denote more hawkish sentiments.
As such, this variable should be negatively related to a Democratic choice. All
principle variables span a 0–1 range. Finally, note that person mean scores are
employed for these scales.
I create a series of policy principle × sophistication interaction terms to

determine if political knowledge amplifies core principle effects. Again, two
conceivable pathways of influence exist. Sophistication may be a necessary
condition for policy voting, implying that those who score low on knowledge
cannot policy vote (the stronger version of the sophistication hypothesis).
Alternatively, most voters may rely on policy principles, with sophistica-
tion tightening attitude-behavior correspondence (the weaker version of the
sophistication hypothesis). My expectation, of course, is that sophistication
will not alter the policy principle weights.
THE VOTE CHOICE TESTS

The 1988 Election
In the analysis that follows, I begin with a baseline logistic regression model
that includes party identification, retrospective judgments, liberal–conserva-
tive orientations, political sophistication, and the ideology × sophistication
term. Policy principle variables are then added to the baseline model to assess
their joint and partial effects controlling for the standard predictors. Lastly, I
add the policy principle × sophistication terms to the equation to test whether
sophistication moderates the impact of principles. I assume, but do not test
due to a lack of suitable measures, that where the policy principle variables
systematically affect the vote choice the position matching condition has been
satisfied. Some work indicates that the results of policy voting analyses change
little when policy variables are measured on their own or in conjunction with
candidate proximity measures (Knight, 1985).
The 1988 models are arrayed in Table 8.1. To begin with the baseline
(model 1), the estimates strongly support the partisan voter, retrospective
voter, and ideological voter models. The coefficient for each variable car-
ries the expected sign and usually achieves statistical significance (p < .05
or better). The estimates reveal that stronger Republican attachments are
inversely related to the probability of a Dukakis vote, that positive views of
the economy and the U.S. position abroad decrease the odds of backing the
challenger, and that the negative relationship between conservative identi-
fication and Dukakis support grows stronger at higher levels of sophistica-
tion. Again, all of this is as expected. The results make clear that if policy
principles are to shape candidate preferences, such influence must operate
in the context of a stringent baseline model.
Model 2 in Table 8.1 shows what happens when policy principles enter the
equation. The first point to note is the null hypothesis that all principle coeffi-
cients simultaneously equal zero can be rejected (partial Wald statistic = 11.98,
TA B L E 8 . 1 . THE PRESIDENTIAL VOTE, 1988
Model 1 Model 2 Model 3
Constant 2.17* 4.36* 4.48*

(.87) (1.15) (1.51)
Party identification –4.61* –4.52* –4.55*
(.35) (.37) (.37)
Retrospective economic judgments –2.30* –2.35* –2.34*
(.60) (.68) (.69)
U.S. world position same –0.36+ –0.43* –0.43*
(.22) (.23) (.23)
U.S. world position stronger –0.76* –0.83* –0.80*
(.24) (.28) (.28)
Liberal–conservative identification 1.92+ 1.91+ 1.85+
(1.29) (1.39) (1.37)
Political sophistication 5.81* 5.06* 4.89*
(1.40) (1.54) (2.16)
Liberal–conservative identification × sophistication –9.34* –7.46* –7.35*
(2.10) (2.44) (2.46)
Limited government –2.94* –2.13*
(.56) (1.26)
Traditional morality –1.21* –2.32+
(.51) (1.51)
Military strength –1.35* –0.68
(.57) (1.25)
Limited government × sophistication –1.50
(2.24)
Traditional morality × sophistication 2.14
(2.34)
Military strength × sophistication –1.37
(1.88)
Wald statistic 40.57* 28.04* 22.26*
Partial Wald statistic – 11.98* 0.46
Notes: Logit estimates. Standard errors are in parentheses (adjusted for complex sample design).
All variables are on a 0–1 range. The partial Wald statistic tests the null hypothesis that the logit
coefficients for the additional variables simultaneously equal 0. Number of observations = 1052.
Source: 1988 NES survey.
p < .001). For the individual variables each coefficient is correctly signed and
statistically significant. The negative coefficient for limited government means
that the more someone opposes federal aid for the economically dispossessed,
the lower the likelihood of a Dukakis vote. The negative coefficients for tra-
ditional morality and military strength imply that right-wing views on each
dimension lower the odds of a Dukakis ballot. These effects hold controlling
for party identification, retrospective judgments, and liberal–conservative
orientations. In short, the model 2 results affirm the proposition that citizens
ground presidential selections in general policy orientations.
How large are the effects? To address this, I use the logit coefficients to sim-
ulate the probability of a Dukakis vote for hypothetical respondents with the
minimum and maximum scores on each principle, holding the other variables
constant at their central tendencies (means for interval level variables and modes
for nominal variables) and plot the results in Figure 8.1. Moving left to right
along the horizontal axis reflects increasingly right leaning positions on each
principle. The vertical axis plots the probability of backing Dukakis. For limited
government, the probability welfare state opponents cast a Dukakis vote is .08
versus .61 for welfare state supporters. For the traditionalism variable, the prob-
ability of Dukakis support equals .57 for those comfortable with cultural change
and .28 for those who reject it. Finally, the probability that national security
doves cast a Democratic vote equals .57 compared to .26 for hawks. These simu-
lations document the power of domain-specific principles to change votes.
The final question is whether the model 2 specification obscures condi-
tionality in the policy principle–vote choice relationship. My argument holds
that principles drive voter choice for everyone. The sophistication interac-
tion model maintains that policy voting is a function of political awareness.
To adjudicate between these perspectives, I estimate a final equation that
assesses how the limited government × sophistication, traditional morality
× sophistication, and military strength × sophistication variables perform
when added to model 2. If my expectations are on the mark, the parameter
1.00
Probability of Dukakis Vote
0.75
0.50
0.25
0.00
Oppose Favor
Limited Govt Traditional Morality Military Strength
Figure 8.1. Policy Principles and the Presidential Vote, 1988.

estimates for the interaction terms should be indistinguishable from zero.

If the sophistication interaction hypothesis provides the better account, the
multiplicative terms should be negative and significant. The estimates in
the last column of Table 8.1 yield no evidence that sophistication matters.
The joint test of whether the principle × sophistication coefficients simul-
taneously equal zero cannot be rejected (p < .72). The t-tests for the multi-
plicative coefficients tell the same story. Hence, in contrast to the powerful
influence sophistication has in promoting liberal–conservative voting, noth-
ing comparable is obtained when it comes to core principles.
Overall, these results imply that most people satisfy the third condition
of policy voting. When Bush criticized Dukakis for planning to cut specific
weapons systems or being dangerously naive about international affairs,
many people got the message that Bush was tougher on national security
than Dukakis. Likewise, those who saw the Dukakis commercial about doing
more to help families living from paycheck to paycheck or heard his call for
expanded health care coverage could have deduced that he was more com-
mitted to strong government. Other voters surely called upon party stereo-
types or issue handling reputations to infer where the candidates fell along
the various continua. And some may have learned from a trusted contact in
their social network. Regardless of how the information was acquired, most
voters evidently matched their policy principles to the positions taken by the
candidates on each dimension and voted for the one in closer proximity. When
it comes to principle-based voting, those deeply informed about politics do
not outshine their less informed counterparts—at least for 1988.
The 1992 Election

The 1992 campaign saw a three candidate battle between Republican President
George H. W. Bush, Democrat challenger Bill Clinton, and billionaire busi-
nessman H. Ross Perot. The statistical models are estimated using multino-
mial logistic regression. Table 8.2 contains the results. I report estimates for
the Clinton versus Bush and Perot versus Bush contrasts for a baseline model
(model 1), a policy principle model (model 2), and a policy principle interac-
tion model (model 3).
In the baseline model, the standard variables conform to expectations. For
starters, stronger GOP leanings diminish the likelihood of a Clinton or Perot
vote relative to Bush. Next, those pleased with conditions at home or abroad
stood by the president more than the displeased. Finally, the impact of lib-
eral–conservative orientations varies across levels of sophistication. Among
knowledgeable strata in the electorate, stronger conservative attachments are
TA B L E 8 . 2 THE PRESIDENTIAL VOTE, 1992
Clinton Perot Clinton Perot versus Clinton Perot

versus Bush versus Bush versus Bush Bush versus Bush versus Bush
Constant 2.82* 1.10 6.04* 3.49* 4.36* 2.36+

(.77) (.87) (.89) (1.04) (1.27) (1.49)
Party identification –5.51* –2.21* –5.40* –2.14* –5.41* –2.14*
(.50) (.40) (.51) (.41) (.51) (.41)
Retrospective economic judgments –2.19* –1.44* –2.14* –1.40* –2.14* –1.37*
(.52) (.39) (.56) (.41) (.58) (.41)
U.S. world position same –0.19 –0.20 –0.30 –0.28 –0.29 –0.27
(.26) (.22) (.26) (.23) (.26) (.23)
U.S. world position stronger –0.72* –0.57* –0.68* –0.54* –0.69* –0.54*
(.27) (.23) (.29) (.25) (.30) (.26)
Liberal–conservative identification 1.70+ 0.43 1.82+ 0.59 1.62 0.59
(1.23) (1.25) (1.21) (1.21) (1.32) (1.29)
(1.15) (1.32) (1.14) (1.38) (1.92) (2.15)
Liberal–conservative identification × sophistication –10.01* –5.00* –7.80* –3.80* –7.39* –3.75+
(1.95) (2.02) (1.97) (2.07) (2.20) (2.28)
Limited government –1.88* –0.21 –1.99 0.32
(.56) (.50) (1.54) (1.46)
Traditional morality –3.09* –2.43* –2.56* –2.54*
(.57) (.45) (1.23) (1.25)
Military strength –1.59* –1.60* 1.11 0.01
(.50) (.47) (1.26) (1.14)
Limited government × sophistication 0.26 –0.86
(2.54) (2.19)
Traditional morality × sophistication –0.87 0.18
(1.94) (1.99)
Military strength × sophistication –4.87* –2.81+
(2.14) (1.80)
Wald statistic 41.97* 21.54* 12.74*
+
p < .10; *p < .05 (one-tailed).
Notes: Multinomial logit estimates. The Bush vote is the base outcome. Standard errors are in parentheses (adjusted for complex sample design). All variables
are on a 0–1 range. The partial Wald statistic tests the null hypothesis that the logit coefficients for the additional variables simultaneously equal 0. Number of
observations = 1522.

associated with a higher likelihood of voting for Bush relative to Clinton (p <
.001) and Perot (p < .05). Liberal–conservative attitudes do little to guide the
preferences of political novices.
I turn next to policy principles. To begin, the possibility that the principles
jointly have zero effect can be rejected without hesitation (p < .001). Next, five
of the six individual coefficients are statistically significant across the candi-
date pairings, suggesting that voters differentiate between the three candi-
dates on policy grounds. To elaborate, those who favor strong government
are more likely to support Clinton over Bush compared to those who dislike
federal activism (p < .001). The odds of choosing Clinton over Perot are also
higher among backers of the welfare state (p < .001). However, these attitudes
are probably unrelated to the Perot–Bush choice (p ns). Put another way, lim-
ited government systematically affects the Clinton–Bush and Clinton–Perot
pairings but not Bush–Perot. Consider what these results tell us about voter
competence. The GOP and president Bush had a well-established record of
opposing government activism. For Perot, controlling federal spending was
a centerpiece of his campaign. This, of course, sounded like Republican boil-
erplate. In contrast, the Democratic Party had long favored federal efforts to
alleviate economic suffering. Although Clinton took moderate positions on
some issues, he promised to help the “forgotten middle class” and “work-
ing families” through new federal initiatives such as health care cost reform,
expanded health care access, family leave, and so on. In other words, Bush
and Perot were much closer to one another on the limited government con-
tinuum than to Clinton. Hence, the finding that limited government shapes
preferences for Clinton over Bush and Clinton versus Perot but fails to do so
for the Bush–Perot contrast underscores voters’ powers of discernment.
Turning to moral traditionalism, we see analogous results. For the Clinton–
Bush contrast, the negative coefficient means that those who view traditional
codes of conduct positively are less inclined to vote for Clinton than moral
progressives (p < .001). Likewise, cultural traditionalists are less likely to favor
Perot than Bush (p < .001). For Clinton versus Perot, moral attitudes exert little
pull either way (p ns). Given the issue handling reputation of the GOP and
the fact that Bush took more socially conservative positions than Clinton and
Perot, the finding that moral attitudes shape Clinton–Bush and Perot–Bush
voting but fail to emerge for the Clinton–Perot match-up suggests information
about core principles captures voter attention. Like the findings about govern-
ment power, then, these results reveal a thoughtful public, one sensitive to
differences between multiple candidates on multiple policy dimensions.
How does military strength affect candidate choice? Here, the odds of pre-
ferring Bush to either Clinton or Perot are significantly higher among hawks
than doves (both p < .001). In contrast, hawk–dove orientations manifest little
influence over the Clinton–Perot choice (p ns). Given the GOP’s reputation as
the party of military strength, which had been reinforced by the decisive 1991
U.S. victory over Iraq in the first Gulf War, these findings again speak to the
policy voting capacities of everyday people. Insofar as the contrast between
Bush and his opponents was much cleaner than that between Clinton and
Perot, the finding that the hawk–dove continuum mattered only for the Bush
comparisons implies that voters are cognizant of where the candidates and
parties stand on this fundamental cleavage.
The statistical estimates validate expectations, but they fail to convey how
strongly policy principles guide candidate preference. Therefore, I simulate
the probability that voters cast a Clinton ballot across the range of each prin-
ciple, holding the other variables at their means or modes. Figure 8.2 presents
the curves. Respondents who endorse limited government are .40 less likely to
fancy Clinton (.22) than those who favor activist government (.62). For moral
standards, the likelihood of a Clinton vote equals .65 among the most progres-
sive versus .28 for the most traditional. Lastly, the probability that extreme
hawks back Clinton is .37 compared to .53 for extreme doves. These simula-
tions demonstrate that abstract principles carry a lot of weight in the minds of
voters. Indeed, movement from very negative to very positive evaluations on
each one changes how citizens decide.
But perhaps these results are wrong. Perhaps policy voting is moder-
ated by sophistication. As the interaction model estimates reveal (model 3,
Table 8.2), evidence that the politically aware make up their minds differ-
ently is thin. The likelihood that all three interaction terms have no effect
cannot be rejected (p < .50). The individual coefficients for the government
1.00
Probability of Clinton Vote
0.75
0.50
0.25
0.00
Oppose Favor

power and moral standards interactions are statistically insignificant as

well. However, the militarism × sophistication variable reaches signifi-
cance in the Clinton–Bush pairing (p < .05) and approaches significance for
Perot versus Bush (p < .07). The coefficients suggest politically knowledge-
able hawks are more likely than knowledgeable doves to favor President
Bush over his rivals. But because the sophistication approach presumes
that the use of abstract predispositions is generally conditional on sophis-
tication and all other sophistication interactions thus far cannot be distin-
guished from zero, it is premature to claim that one positive test confirms
the theory.
To sum up, the 1992 data show that in a competitive three candidate race
limited government, traditional morality, and military strength affect voter
preferences, whereas political sophistication proves largely unnecessary to
help citizens navigate a complex choice set. Given the wide range of disparate
channels through which campaign information flows, it may not be terribly
surprising that policy principles function as critical heuristics in the decision-
making calculus of the American voter.
The 1996 Election

Table 8.3 displays the parameter estimates for the 1996 Clinton–Dole contest.5
As usual, the baseline model does an excellent job in predicting the vote.
Republicans are less likely to vote for Clinton than Democrats, those pleased
with economic and foreign affairs are more likely to support the incum-
bent, and sophisticated conservatives back Clinton at much lower rates than
sophisticated liberals.
The second model shows that policy principles matter to some degree,
although not as much as in 1988 and 1992. The joint test that all three coef-
ficients equal zero can be rejected (p < .05). However, assessed individually,
only limited government reaches significance (p < .001). The negative coef-
ficient means that those approving of small government are more anti-Clinton
than those who favor strong government. In contrast, neither conventional
morality nor military power systematically affects candidate choice. Next,
Figure 8.3 reveals that defenders of the welfare state have a .94 probability
of voting for Clinton versus .26 for welfare state opponents, holding all else
constant. This sizable difference emphasizes the power of limited government
to move votes.
5
I also ran a multinomial logit model with Perot. The conclusions in the text do
not change with Perot included. Given the greatly diminished appeal and sali-
ence of his candidacy, this is not surprising.
Constant 0.32 1.09 1.44

(1.02) (1.36) (2.31)
(.42) (.42) (.43)
Retrospective economic judgments 2.35* 2.63* 2.68*
(.83) (.82) (.80)
U.S. world position same 0.78* 0.81* 0.89*
(.28) (.31) (.32)
U.S. world position stronger 1.29* 1.11* 1.15*
(.33) (.36) (.36)
Liberal–conservative identification 1.94 1.72 2.32
(2.00) (2.08) (2.23)
Political sophistication 4.85* 5.30* 4.36
(2.03) (2.33) (3.66)
(3.41) (3.69) (3.97)
(1.10) (2.57)
Traditional morality –0.17 1.79
(.80) (2.69)
Military strength 0.71 –0.07
(.72) (1.65)
Limited government × sophistication 5.86*
(3.33)
Traditional morality × sophistication –3.14
(4.09)
Military strength × sophistication 1.43
(2.42)
Wald statistic 22.81* 24.74* 19.33*
As always, the possibility that core principles matter more at higher

levels of sophistication must be investigated. My findings suggest the
sophistication variable does not perform as predicted by the sophistication
hypothesis. As shown in the final column of Table 8.3, neither the partial
Wald statistic (p < .30) nor the t-test results for the individual coefficients
Probability of Clinton Vote 1.00
0.75
0.50
0.25
0.00
Oppose Favor
Limited Government

support this claim. There is some indication of a limited government ×

sophistication effect (p < .05), but since the positive coefficient means the
less informed make greater use of this principle, the evidence contradicts
the sophistication model.
My summation for 1996 is that limited government represents the lone
principle to guide voter decision making. Whether from a recollection that
Clinton stood up for average Americans during the government shutdowns
or a campaign commercial attacking Dole for his opposition to the minimum
wage or a party informed inference, voters readily surmised that Clinton
was the progovernment candidate. And although the impact of policy prin-
ciples is confined to a single predisposition, the substantive impact is striking.
Movement across the full range of the variable leads to a bigger change in the
predicted probability of voting Democratic in 1996 (.68) than in 1992 (.40) and
1988 (.53). Yet the fact the other two principles did not behave as predicted
raises the possibility that policy principles are less consequential than I have
claimed.
The 2000 Election

The vote choice results for the 2000 election are shown in Table 8.4. As
expected, partisan identities hold sway as Democrats sustain Gore and
Republicans do the same for Bush. For retrospective assessments, the ver-
dict is mixed. On the positive side, those who saw international conditions
holding steady or getting better were more likely to vote Democratic than
those who saw American decline. On the negative side, national economic
judgments fail to predict candidate choice. People who thought the economy
Constant –0.69 0.28 0.23

(.87) (1.01) (1.32)
(.45) (.43) (.44)
Retrospective economic judgments 0.46 0.51 0.56
(.51) (.52) (.52)
U.S. world position same 0.78* 0.77* 0.78*
(.25) (.26) (.26)
U.S. world position stronger 1.40* 1.30* 1.31*
(.29) (.30) (.31)
Liberal–conservative identification 3.74* 3.95* 4.06*
(1.49) (1.53) (1.63)
(1.57) (1.57) (2.07)
(2.98) (2.95) (3.18)
(.66) (1.30)
Traditional morality –1.09* –0.62
(.61) (1.69)
Limited government × sophistication 1.84
(1.96)
(2.75)
Wald statistic 35.99* 31.99* 25.95*
was better off were, oddly enough, no more likely to prefer Gore than those
who saw a faltering economy. Lastly, liberal–conservative evaluation contin-
ues to emerge conditionally. Sophisticated liberals line up behind the vice
president whereas sophisticated conservatives go for the governor of Texas.
Among the less informed swatches of the public, liberal–conservative atti-
tudes are inconsequential.
The second column contains the policy principle estimates. The 2000 NES
lacks measures of abstract hawk–dove attitudes, so their effects on the vote
cannot be assessed. For the other two principles, Table 8.4 yields unequivocal
support for my key hypothesis. As indicated by the adjusted Wald statistic
(p < .01), it is extremely unlikely that both variables have no impact. The
individual coefficients tell the same story as both are statistically significant
(p < .05). Welfare state supporters are more likely to prefer Gore than wel-
fare state critics, whereas moral traditionalists support Bush to a much greater
degree than progressives. To assess the magnitude of these effects, I generate
predicted probability curves. Figure 8.4 indicates that for limited government,
movement from the most left-wing to the most right-wing position drops the
probability of a Gore vote from .67 to .32, all else being equal. Otherwise typ-
ical voters switch sides based on their evaluations of government. The moral
traditionalism curve mirrors this pattern. For zealous traditionalists, the esti-
mated probability of supporting Gore equals .42. For uncompromising pro-
gressives, the probability of a Gore vote is .69.
Do the politically sophisticated rely more on policy principles to guide their
ballots? The interaction model estimates in the last column of Table 8.4 fail to cor-
roborate this hypothesis, as indicated by the insignificant joint test and individual
t-tests. Although sophistication promotes reliance on liberal–conservative atti-
tudes, its influence does not extend to the use of core principles.
Taken as a whole, the results imply abstract policy principles guide voter
decision making. Information about candidate positions on general principles
was dispersed throughout the information environment, thereby allowing
very large portions of the electorate to satisfy the position matching condition
1.00
Probability of Gore Vote
0.75
0.50
0.25
0.00
Oppose Favor
Limited Government Traditional Morality

of policy voting. Regardless of how much or little they knew about public
affairs in general, most voters made good use of limited government and
moral traditionalism when choosing between Gore and Bush. Note finally that
the 2000 results bear a closer resemblance to those of 1988 and 1992, in which
all three principles mattered, than to 1996, in which limited government alone
was significant.
The 2004 Election

Table 8.5 contains the 2004 parameter estimates. Note first that all base-
line model coefficients are correctly signed and statistically significant.
Democrats are more enthusiastic about Kerry than Republicans, those who
believe national and international conditions improved or held steady are
less inclined to favor Kerry compared to those who saw conditions worsen-
ing, and informed conservatives back Kerry to a much lesser extent than
informed liberals. As always, the partial effects of domain-specific princi-
ples must emerge above and beyond the influence of powerful vote choice
predictors.
The model 2 estimates reveal the desired effects. First, the joint test is highly
significant, suggesting it is very unlikely that the principle coefficients jointly
equal zero (p < .001). Second, each logit coefficient surpasses significance at
p < .01 or better. I illustrate the effect sizes through predicted probability simu-
lations. The curves in Figure 8.5 demonstrate that each principle profoundly
affects candidate choice. For limited government, movement from the least to
most supportive position decreases the probability of a Kerry vote from .81
to .12. Next, the likelihood of a Democratic ballot is .83 for the most progres-
sive and .34 for the most traditional respondents. Finally, the probability that
a strong dove goes for Kerry is .93 compared to .22 for the most committed
hawk. These simulations imply that each principle wields clout sufficient to
change the vote.
In the sophistication interaction equation, the model 3 estimates fail to con-
firm the claim that awareness magnifies principle-based voting. The partial
Wald test shows that the coefficients for the multiplicative terms are jointly
indistinguishable from zero. The same holds true for the individual coeffi-
cients. After five elections, I have yet to uncover any systematic evidence that
political knowledge promotes reliance on domain-specific principles. These
results accrue to the accumulating body of evidence that abstract policy pos-
tures shape candidate choice for all voters. The remarkably competitive 2004
election provided citizens with ample opportunities to learn—or be reminded
TA B L E 8 . 5 THE PRESIDENTIAL VOTE, 2004
Constant 2.64* 6.73* 6.30*

(.78) (1.28) (1.59)
(.60) (.64) (.65)
Retrospective economic judgments –2.62* –2.22* –2.23*
(.85) (.91) (.90)
U.S. world position same –0.89* –0.68* –0.67*
(.27) (.26) (.26)
U.S. world position stronger –1.51* –1.16* –1.14*
(.34) (.40) (.40)
Liberal–conservative identification 2.69* 2.95* 2.70+
(1.40) (1.58) (1.69)
(1.30) (1.67) (2.63)
(2.36) (2.93) (3.11)
Limited government –3.51* –3.03+
(.66) (2.26)
(.81) (2.60)
Military strength –3.78* –4.70*
(.85) (1.92)
Limited government × sophistication –1.10
(4.14)
(4.43)
Military strength × sophistication 1.72
(3.83)
Wald statistic 25.88* 19.31* 28.25*
p < .10; *p < .05 (one-tailed).

+
of—candidate positions on the fundamental policy cleavages that animate

American political discourse. Voters responded by taking all of this in and
throwing their support to whichever candidate seemed the better bet to serve
their general policy interests.
Probability of Kerry Vote 1.00
0.75
0.50
0.25
0.00
Oppose Favor

The 2008 Election

The 2008 estimates are in Table 8.6. Here, the baseline model turns in a mixed
performance. On the plus side, both partisanship and liberal–conservative
orientations behave as they should. Stronger Republican identification
is associated with a sharp drop in the odds of an Obama vote. Likewise,
knowledgeable conservatives are far cooler toward Obama’s candidacy
than knowledgeable liberals. On the minus side, the retrospective voter
model garners minimal support. The coefficients for the national economy
and one of the U.S. position dummy variables fall well below conventional
levels of significance. People who believed that the U.S. economy and posi-
tion overseas were diminished were no more likely to vote for Obama than
the few who saw developments more hopefully. This curious nonfinding
may be a statistical artifact. Because most respondents believed that the
economy was in terrible shape and the U.S. position in the world had been
badly damaged, there may be insufficient variation on these variables to
influence the vote.
I turn now to policy principles, for which measures of limited govern-
ment and traditional morality are available. In model 2, the joint test result
is highly significant (p < .001), the limited government variable approaches
significance (p < .06), and morality is highly significant (p < .001). The more
favorably people evaluate limited government and traditional moral stan-
dards the lower the odds of an Obama vote. However, the model 3 esti-
mates suggest that these relationships may be moderated by sophistication.
The partial Wald statistic approaches significance (p < .07), as does the lim-
ited government × sophistication variable (p < .06), with the positive sign
Constant 2.45* 4.41* 3.82*

(.71) (.89) (1.07)
(.43) (.46) (.46)
Retrospective economic judgments –0.28 –0.42 –0.44
(.77) (.76) (.77)
U.S. world position same –0.54* –0.68* –0.64*
(.29) (.31) (.31)
U.S. world position stronger –0.44 –0.51 –0.51
(.54) (.54) (.56)
Liberal–conservative identification 2.38* 2.38* 2.31*
(1.06) (1.07) (1.10)
(1.07) (1.16) (1.62)
(1.74) (1.86) (1.88)
Limited government –1.22+ –3.11*
(.77) (1.49)
(.55) (1.09)
Limited government × sophistication 3.42+
(2.15)
Traditional morality × sophistication –3.54*
(1.79)
Wald statistic 36.16* 27.90* 27.81*
Partial Wald statistic – 12.45* 2.77+
p < .10; *p < .05 (one-tailed).

+
suggesting that less sophisticated voters rely more heavily on limited gov-
ernment. This, of course, undercuts the sophistication interaction model. As
seen in Figure 8.6, for voters one standard deviation below the knowledge
mean movement from the minimum to maximum score on limited govern-
ment drops the probability of an Obama vote from .85 to .44, a large sub-
stantive effect. For voters a standard deviation above the mean, movement
across the limited government range does not affect the probability of an
Obama vote (the probability ticks up from .75 to .76).
Probability of Obama Vote 1.00
0.75
0.50
0.25

0.00
Oppose Favor
Limited Government
Figure 8.6. Limited Government and the Presidential Vote, 2008.

Source: Table 8.6 estimates; other predictors held at mean/mode. Effect ns for high sample.
The traditional morality interaction is negative and significant, implying

that more informed citizens make greater use of this dimension to guide can-
didate choice. Figure 8.7 illustrates the size of the effect for respondents plus
or minus one standard deviation from the knowledge mean. Among novices,
strong traditionalists are significantly less likely to cast an Obama ballot than
strong progressives (.60 < .88). For the aware, movement across the tradition-
alism range decreases the probability of an Obama vote from .96 to .37. Here,
the sophistication effect is remarkably large.
Tallying up the results from the 1988–2008 elections, we see that policy
principles systematically affect voter choice for citizens at different levels of
sophistication. When it comes to grounding candidate selections in general
1.00
Probability of Obama Vote
0.75
0.50
0.25

0.00
Oppose Favor
Figure 8.7. Traditional Morality and the Presidential Vote, 2008.

Source: Table 8.6 estimates; other predictors held at mean/mode. Effect significant.
policy predispositions, those with deep stores of political knowledge perform

the same as those whose knowledge base is shallower. To the degree that
policy voting represents an indispensable criterion of political competence,
these findings suggest that ordinary citizens perform far better than has been
acknowledged by electoral behavior specialists and political psychologists.
What about Specific Issues?

To this point, the effects of policy principles on the presidential vote hold
controlling for party identification, retrospective judgments, and liberal–
conservative orientations. A natural question to ask is how policy principles
fare when issues are added to the mix. Do the effects of limited government,
traditional morality, and military strength persist when diverse issues from
each domain are entered into the equations, or are their effects washed out? If
the latter happens, the claim that policy principles matter more than specific
issues will no longer stand. However, if core principles continue to hold while
issues prove less potent, the case for the third face of policy voting will be
strengthened.
As a final check, in every case in which a policy principle variable mani-
fested a statistically significant effect on the vote, I include an issue preference
variable from the same domain to assess how the coefficient for the principle
behaves. For example, the 1988 estimates suggest that citizens ground their
votes in limited government, traditional morality, and military strength. I add
tax cut preferences, death penalty opinion, and anticommunism to the model
2 specification to see if the three principles continue to matter. Likewise,
all three principles guide electoral choice in 2004. Here, opinions on Social
Security privatization, gay marriage, and the war in Iraq are included in the
newly specified model. And so on for the other elections.
I used the following selection criteria to pinpoint germane issues. Drawing
on prior empirical, journalistic, and historical accounts of each election,
I identified a discrete positional issue in the respective domains that attracted
attention and then located measures in the NES surveys to tap evaluations.
Table 8.7 lists the issues along with the number of items used to construct each
measure. Although all possible controversies are not included in the model,
those chosen did feature prominently in the respective election seasons.
I estimated a series of additive, interactive, and mixed models to examine
the effects of principles and issues. Table 8.8 reports estimates for the final
model, each of which contains the significant additive and/or interactive
terms that survived the testing protocol. To conserve space, I do not report the
parameter estimates for the other predictors in the model (e.g., party identifi-
cation and national economy). To begin, the logit coefficients for the principle
TA B L E 8 . 7 . DISCRETE ISSUES INCLUDED FOR THE

SPECIFICATION TESTS
Economic Welfare Issues Cultural Issues National Security Issues
1988 Taxes (2) Death penalty (1) Anticommunism (4)

1992 Health care (1) Abortion (4) Iraq (2)
1996 Taxes (1) – –
2000 Budget surplus (2) Abortion (3) NA
2004 Social Security privatization (1) Gay rights (2) Iraq (2)
2008 Health care (1) Abortion (7) NA
Notes: “–” indicates a nonsignificant effect for the policy principle variable in the 1996 model, so
no issue variable was used to assess the principle’s robustness. NA indicates the military strength
variable was not available in the survey, so no issue variable was used to assess the principle’s
robustness. Number of items used to measure each opinion are in parentheses.
variables remain strong predictors of candidate choice. Before issue prefer-

ences are added to the equations, policy principles are significant at p < .05 or
better in 13 of 16 tests and at p < .06 in one other case (see Tables 8.1–8.6 and
note that I focus on the significant principle × sophistication terms in model 3,
Table 8.6). Once issue preferences are incorporated, 11 of these 14 coefficients
remain significant at p < .05 or better and two slip to marginal significance
(military strength88, p < .06, and traditional morality00, p < .10). Moreover,
a comparison of the policy principle coefficients across the two sets of equa-
tions reveals essentially unchanged coefficients in some models (e.g., tra-
ditional morality b = –1.21 in Table 8.1 compared to –1.17 in Table 8.8) and
slightly to somewhat attenuated coefficients in others (e.g., limited govern-
ment b = –1.46 in Table 8.4 versus –1.32 in Table 8.8). And because the issue
variables mediate some of the influence of principles, the Table 8.8 estimates
understate the total effects principles have on the vote.
The issue preference variables perform more erratically. Economic wel-
fare issues manifest a significant main effect in 2008 (health care), a marginal
main effect in 1988 (taxes), and nonsignificant effects in 1996 (Dole 15% tax
cut proposal) and 2004 (Social Security privatization). Political sophistication
promotes reliance on economic welfare issues in 1992 (health care) and 2000
(budget surplus plan). In the cultural issues domain, a significant main effect
emerges in 1988 (death penalty) whereas sophistication-based effects occur in
2000 (abortion), 2004 (gay rights), and 2008 (abortion). The 1992 abortion effect
cannot be reliably distinguished from zero. In the national security arena,
I find a significant main effect in 2004 (Iraq), a sophistication interaction effect
in 1992 (Iraq), and no effect for 1988 (anticommunism). Summing up, the issue
TA B L E 8 . 8 . THE PRESIDENTIAL VOTE, ISSUE SPECIFICATION TESTS
1992 BC 1992 RP
1988 versus GB versus GB 1996 2000 2004 2008
Limited government –2.43* –2.05* 0.08 –5.33* –1.32* –3.38* –2.11

(.59) (.58) (.51) (1.90) (.73) (.70) (2.44)
Traditional morality –1.17* –2.91* –2.29* –1.28 –0.94+ –1.75* 2.12
(.52) (.69) (.50) (1.12) (.71) (.77) (2.14)
Military strength –0.90+ –0.98+ –1.53* 1.27 –3.32*
(.57) (.60) (.46) (1.19) (.87)
Limited government × sophistication 2.54
(3.31)
Traditional morality × sophistication –6.58*
(3.34)
Economic welfare issue –0.51+ 1.58* 1.22 –0.51 –0.11 –0.45 –1.23*
(.35) (.69) (.97) (.75) (1.09) (.39) (.49)
Moral issue –1.18* –0.52 –0.33 1.27 1.31* 0.94
(.30) (.38) (.37) (1.21) (.58) (1.58)
National security issue –0.72 –0.30 0.85 –2.82*
(.66) (.75) (.71) (.54)
Economic welfare issue × sophistication –2.86* –3.15* –2.46+
(1.21) (1.37) (1.84)
Moral issue × sophistication –4.16* –2.91* –3.28+
(2.12) (1.09) (2.42)
National security issue × sophistication –1.23 –2.08*
(1.29) (1.21)
Number of observations 1007 1375 1375 536 993 738 705
p < .10; *p < .05 (one-tailed).

+
Notes: Logit estimates. Standard errors are in parentheses (adjusted for complex sample design). All variables are on a 0–1 range. The issue variables are listed in
Table 8.7. The issue variables are included only when the policy principle was significant in Tables 8.1–8.6. Issues are coded to be negatively related to vote. The
other variables are included in the model, but are not presented in the table to preserve space.

variables had significant main effects in four cases (one at p < .10), nonsignifi-
cant effects in another four, and sophistication interaction effects in six cases
(two at p < .10). This varied pattern of results implies that sophistication pro-
motes issue voting to some extent and, more importantly, that discrete issues
do not have nearly as much influence over candidate choice as abstract policy
principles.
To sum up, limited government, traditional morality, and military strength
have robust effects on the presidential vote for citizens across the sophisti-
cation continuum. The effects emerge in the presence of well-known predic-
tors of the vote such as party identification and retrospective judgments, they
obtain controlling for liberal–conservative orientations, and they hold when
campaign-specific issues are accounted for.6
When Sophistication Matters

Sophistication may be largely irrelevant when it comes to the develop-
ment and use of policy principles, but that does not mean that it plays no
role in the decision-making strategies of some voters. As we have just seen,
the issue voting results in Table 8.8 provide some evidence of a sophistica-
tion interaction effect. Additionally, as shown in Tables 8.1–8.6, the liberal–
conservative identification × sophistication coefficient is correctly signed
and statistically significant in every model. To illustrate how sophistication
1.00
Probability of Dukakis Vote
0.75
0.50
0.25

0.00
Ext Lib Ext Con
Liberal-Conservative ID
Figure 8.8. Liberal–Conservative Views and the Presidential Vote, 1988.

Source: Table 8.1 estimates; other predictors held at mean/mode. Effect ns for low sample.
6
In terms of relative influence, the party variable eclipses the isolated policy
principle effects most of the time. Of course, when principles are considered en
masse, their relative impact rises accordingly.
1.00
Probability of Obama Vote
0.75
0.50
0.25

0.00
Ext Lib Ext Con
Liberal-Conservative ID
Figure 8.9. Liberal–Conservative Views and the Presidential Vote, 2008.

Source: Table 8.6 estimates; other predictors held at mean/mode. Effect ns for low sample.
promotes liberal–conservative voting, I offer some probability simulations for

1988 and 2008 (the same patterns hold for the other years). Each summarizes
how movement from the extreme liberal to the extreme conservative point
on the ideology scale alters the probability of voting Democratic for low and
high sophistication individuals (the other predictors are held at scale means
or modes). The low sophistication curve applies to respondents one standard
deviation below the knowledge mean. Conversely, the high curve is for those
scoring a standard deviation above it.
To begin with 1988, Figure 8.8 shows that sophisticated conservatives are
much less willing to support Dukakis than sophisticated liberals (.10 < .87,
p < .001). The same does not hold true among novices as conservatives are
marginally less likely to cast a Dukakis ballot than liberals (.31 < .49, p = .12).
Next, Figure 8.9 shows that in 2008 the probability of an Obama vote is much
lower for sophisticated conservatives than for sophisticated liberals (.26 < .97,
p < .001). The likelihood of an Obama vote differs little between uninformed
conservatives (.74) and uninformed liberals (.76, p ns). Broadly speaking, these
plots furnish more evidence that sophistication matters in some respects.
When it comes to constructing preferences on specific issues (see Chapter 6)
or marking ballots on Election Day, politically aware citizens make greater use
of liberal–conservative predispositions than the unaware.
CONCLUSIONS
This chapter brings my empirical study of the American voter to a close. To
recap, there exist multiple pathways through which information about can-
didate positions on policy principles can reach prospective voters. They may
acquire the information from the campaign, through party-based inferences,

from contacts within their social networks, or from other by-products of daily
living. Because the same information flows repeatedly through these channels
over the course of the campaign, citizens need not be attuned to every elec-
toral twist and turn to discover what they need to know to vote on grounds
of principle. All that is necessary is a bit of motivation to pay attention to the
campaign from time to time. Given a little information, they can extrapolate
where the presidential hopefuls stand on the main policy cleavages and select
the appropriate one to represent them (cf. Popkin, 1994).
The statistical analyses in this chapter buttress these claims. Year after year
global judgments about government activism, moral standards, and military
power systematically affect the choices citizens make in U.S. presidential elec-
tions, controlling for a number of established predictors of the vote. The joint
test that all policy principle coefficients equal zero was always rejected. The
individual coefficients surpassed statistical significance in most tests across all
six elections. Limited government mattered in every election considered (in
2008 the less knowledgeable made greater use of it than the more informed),
moral traditionalism mattered in five of six battles (in 2008 sophistication pro-
motes reliance on this dimension, but even the less aware made use of it),
and military strength weighed on candidate choice in three of the four years
for which adequate measures were available. The probability simulations
revealed impressive effect sizes. In most cases movement along the range of
the principle changed votes. Finally, and most importantly, little evidence sug-
gests that political sophistication promotes the use of abstract principles. The
highly sophisticated, the moderately sophisticated, and the unsophisticated
are equally adept grounding votes in core policy principles.
As postulated by the sophistication interaction model, the first two faces
of policy voting depend on sophistication to varying degrees. With respect to
ideological attitudes, the liberal–conservative identification × sophistication
coefficient was statistically and substantively significant in every equation.
Said otherwise, the choices rendered by politically aware, but not unaware,
individuals are grounded in liberal–conservative attitudes. For specific issues
such as health care, abortion, gay rights, and Iraq, sophistication enhanced
reliance on some preferences, though not nearly to the degree it did for ideo-
logical leanings. And when sophistication failed to promote issue voting,
little evidence suggests that issues routinely affect the vote in the broader
electorate.
In conclusion, the influence of policy principles on voter choice cannot
be disputed. There is indeed a third face of policy voting, one that extends
broadly and deeply throughout the mass public. This simple conclusion
breaks sharply with the dominant strains in the electoral behavior, political
psychology, and public opinion fields, which have insisted for far too long that
the ability to develop, maintain, and use policy attitudes depends on political
sophistication. This holds true when policy attitudes are defined in terms of
liberal–conservative orientations and issue preferences, but once the concep-
tual rubric of policy attitudes is broadened to encompass core policy prin-
ciples, the limits of the conventional wisdom become clear. The concluding
chapter takes up these and other broader matters.
CHAPTER 9
The Exoneration of the American Voter?
Voters are frequently mistaken for fools. Some critics offer telling anecdotes to
underscore the political stupidity of the masses. To take one example, a survey
found that 22% of American adults could name all five members of Homer
Simpson’s fictional cartoon family, whereas only 1 in 1000 identified all five
first amendment freedoms (Shenkman, 2008). Others have built a case by mar-
shalling an impressive array of evidence showing that most voters are innocent
of ideology and unsure about the issues, the net result of which is that policy
voting eludes them (Campbell et al., 1960; Converse, 1964; Neuman, 1986).
But from findings such as these it does not automatically follow that voters
are incompetent, that they lack genuine positions on the policy cleavages that
define American politics, or that they cannot judge candidates on grounds of
what really matters—the direction the national government should take in the
major policy domains. In this final chapter, I summarize my argument and
findings and situate them in broader literatures, delineate qualifications that
bear on my claims, and discuss the broader implications my results have for
evaluating the American voter.
REVIEW OF THE MODEL AND PRINCIPAL FINDINGS

The seminal works on electoral behavior took a cold hard look at American
voters and found them wanting (Berelson et al., 1954; Campbell et al., 1960).
Half a century later, the landscape seems largely unchanged. Political knowl-
edge has risen little, if at all, over the past 50 years, and although there is
some evidence of heightened ideological and issue awareness in the elector-
ate, such increases have been marginal and confined to politically engaged
subsets of the public (Delli Carpini & Keeter, 1996; Lewis-Beck et al., 2008).
Ideological ignorance remains the norm, few hold meaningful attitudes on
the issues, and citizens remain stunningly uninformed (Kinder, 2006). But not
233
everyone. Differences in the level of sophistication presumably have wide-

ranging implications for electoral choice. To most scholars, “it is unreasonable
to insist that a person who is extremely well informed about politics will make
his voting choice in the same way as a person who habitually pays only mini-
mal attention to public affairs” (Sniderman et al., 1991, p. 178).
Consider how the prevailing view applies to the 2008 presidential cam-
paign. Like most recent elections, candidates offered the people a genuine
choice on policy grounds. On the hustings, McCain adopted conventional
GOP positions on economic welfare, cultural issues, and national security.
First, the Arizona senator committed to lower taxes, smaller government, and
less federal spending. Obama countered by promising to overhaul and greatly
expand the federal role in the health care system, to raise income taxes for
individuals earning over $200,000 and families earning in excess of $250,000
annually, and to fight for a large stimulus package to help reverse the economic
slide. On cultural issues, McCain stood well to the right of Obama on such
matters. To take the most obvious example McCain was pro-life and Obama
was pro-choice. Finally, McCain touted his support for the surge in Iraq and
argued that the United States had to stay the course until the job was finished.
Obama’s position on national security was more nuanced. He charged that
the invasion of Iraq had been a mistake and said he would draw down U.S.
forces in that country, but also proposed increasing the U.S. military presence
in Afghanistan. The candidates also differed on the use of diplomacy. Obama
signaled his willingness to meet with Iranian and North Korean leaders with-
out preconditions, moves McCain dismissed as naive and dangerous.
As voters pondered their choices, a number of factors weighed on their
minds. For tens of millions of people, underlying partisan loyalties were cru-
cially important. For millions more, assessments of conditions at home and
abroad likely carried some weight. The possibility that a black man could
occupy the White House for the first time in American history unquestionably
mattered to large slices of the electorate. Some were troubled by this because
they disliked African-Americans; others more magnanimous welcomed the
prospect. Some fence sitters, perhaps wary of Obama’s inexperience, were no
doubt reassured by his inspiring oratory and cool performance in the debates.
McCain’s status as a war hero and the sacrifices he bore for the nation moved
many people. His self-conscious status as a maverick appealed to some,
although its repeated invocation annoyed others. There were, in short, many
reasons why voters might have opted for Obama or McCain.
But what of policy concerns? Despite the attention the candidates devoted
to their competing visions of American society and politics, and the time they
spent explaining these fundamentally different views to the people, a chorus
The Exoneration of the American Voter? 235
of scholars believe policy means little to most voters. The dominant sophisti-
cation interaction model maintains that when it comes to choosing between
presidential candidates, only those who know a good deal about public affairs
rely on the liberal–conservative heuristic and issue preferences to guide their
choices. Insofar as the first two types of policy voting escape most people, the
general public falls short on this criterion of political competence. Instead,
the much larger share, oblivious to liberalism and conservatism and political
issues, defaults to simple cues such as party identification, party infused judg-
ments about conditions at home or abroad, and so on.
The sophistication interaction account is at once plausible and backed by
much evidence. Instead of showing once again that the first two faces of pol-
icy voting are limited to politically aware citizens, I have explored the pos-
sibility that a third face of policy voting exists, one centered on the role of
core principles. These principles reflect abstract beliefs about what should be
done in one of the three major issue areas that comprise the American political
agenda. First, in the economic welfare domain, limited government centers
on the extent to which the government in Washington should provide some
measure of economic security to those in need or otherwise vulnerable to the
vicissitudes of the market economy. Second, traditional morality, defined in
terms of the moral standards that should guide the public and private life of
the nation, operates in the cultural and social issues domain. Third, military
strength centers on the role of force versus diplomacy in the conduct of U.S.
foreign policy.
These principles were chosen for theoretical and historical reasons. On the
theoretical side, voters are cognitive misers whose natural inclinations are to
make accurate decisions without expending scarce cognitive resources; hence,
my predilection for one bedrock principle per issue area makes sense. But why
these three as opposed to other viable principles? On the historical side of
the ledger, my review of the record suggests limited government, traditional
morality, and military strength have long served as organizing principles in
their respective domains. Divisions over whether the national government is
doing too much or too little to help average Americans and those at the eco-
nomic margins of society extend back to the New Deal era. The 1960s ushered
in a series of disruptive social changes, what came to be known as the “culture
wars,” battles that continue to this day. Questions about striking the proper
balance between military muscle and the tools of statecraft have animated the
foreign policy and national security domain since the aftermath of World War
II. Although the relative visibility of these disparate conflicts has waxed and
waned over time, these divisions have defined political conflict in this country
for decades. Moreover, they have mapped onto the party system for nearly
as long. Democrats and Republicans have taken distinct positions on limited

government since the 1930s, on moral traditionalism starting in the late 1960s,
and on military power by the early 1980s.
Policy voting can occur when three conditions are satisfied. Policy attitudes
must be available in political attitude structures, must function as central
heuristics in mass belief systems, and must be matched to candidate posi-
tions. To test for availability, in Chapter 5 I estimated a series of measurement
models and continuity correlations for survey respondents stratified into low,
medium, and high knowledge terciles (roughly). The analysis suggested that
nearly everyone holds attitudes on each principle, that these are stable over
time, and that the attitudes of the aware are somewhat more crystallized than
those of the less aware. Such results confirm that policy principles are widely
available in the minds of would-be voters.
To evaluate the centrality condition, I conducted tests in Chapter 6 utiliz-
ing panel data to get at temporal priority between principles and issues and
cross-sectional data to see whether most people call upon principles to verti-
cally constrain issue positions. The estimates showed that limited government
and traditional morality dynamically constrain issue preferences without
being reciprocally shaped by them. These findings echo Peffley and Hurwitz’s
report (1993) that military power does the same with respect to national secu-
rity preferences. Additionally, citizens across the sophistication spectrum use
limited government, traditional morality, and military strength to guide the
positions they take on various issue controversies. Sophistication promotes
reliance on limited government on occasion and traditional morality espe-
cially, but even so, political novices make heavy use of general principles. The
effects of principles typically bested those of party identification and other
widely used heuristics.
Chapter 7 examined the building blocks of limited government, traditional
morality, and military strength. Basic human values—abstract normative
beliefs about what is important in life—correlate strongly with opinions about
each political principle. And in something of a surprise, the effects of values,
self-transcendence and conservation foremost, more often than not exceeded
those of partisanship. Insofar as policy attitudes express deeply held values,
this evidence further underscores the centrality of core principles in political
attitude structures (Katz, 1960). Simply put, policy principles function as pow-
erful heuristics in the belief systems of most people.
In Chapter 8, the presidential vote was modeled as a function of policy prin-
ciples and predictors drawn from leading theories of electoral choice. Multiple
tests across six elections revealed that policy principles systematically guide
candidate selection for citizens across the sophistication continuum. Voters
who know relatively little about politics rely on limited government, tra-
ditional morality, and military strength as much as those whose knowledge
stores run far deeper. Core principles manifested strong substantive effects
within and across the elections, controlling for retrospective judgments, party
identification, liberal–conservative orientations, and issue preferences. Lastly,
the estimates revealed that sophistication habitually amplifies ideological
voting and promotes issue voting to a degree as well. At the end of the day,
core principles represent the only class of policy attitudes that consistently
and powerfully shapes voter choice. The first two faces of policy voting may
be confined to more sophisticated strata of the electorate, but the third face
extends largely to all. The American voter is a principled voter.
QUALIFICATIONS
These conclusions rest on a number of simplifying assumptions made through-
out the book. As such, the limits of what I have done and what I can claim
must be emphasized. To begin, the argument rests heavily on my judgment
that the policy space of American politics revolves around three fundamen-
tal cleavages. My coverage of American history from the 1930s to the present
maintains that divisions over the role the federal government should play in
insulating citizens from market risk constitutes the first great policy cleavage;
conflict over social change and the implications it has for moral standards con-
stitutes a second overriding cleavage; and that disputes over whether force,
diplomacy, or some combination thereof best serve U.S. foreign policy and
national security goals expresses a third fundamental cleavage. These strike
me as leading candidates for America’s political fault lines, but they are not
the only ones the analyst might select.
Similar to quantitative work, historical analysis relies on simplification. In
trying to reduce so much of this complexity to the essentials, we court the
risk that something important has been missed. Some readers may be trou-
bled by my neglect of the racial cleavage. But as I explained in Chapter 2,
my argument is not that race is irrelevant. There is no denying the profound
implications race as a social cleavage has had for structuring the party system.
Moreover, explicit racial issues, such as busing and affirmative action, have
moved onto the national agenda on occasion. My claim is simply that after the
1960s the issue of race qua race has diminished in relative importance. Once
politicians embraced the principle of racial equality, whether by choice or
necessity, overt racial issues receded into the background. Conservative poli-
ticians could no longer campaign openly on segregationist or other explicitly
antiblack platforms. This does not mean racial issues disappeared—far from
it. Instead, they became intertwined with other, ostensibly race-neutral issues
such as law and order and welfare (Mendelberg, 2001). Again, race matters,
but not, in my judgment, to the extent that battles over limited government,
social change, and military power do.
A second potential shortcoming lies in my assessment (or lack thereof) of
position matching. Recall this requires prospective voters to compare their
own positions on each dimension to those espoused by the candidates and
then go for the one lying in closer proximity. Although voters struggle when
it comes to placing presidential candidates on the liberal–conservative scale
and various issue continua, I have argued that the barriers to doing so are
much lower for the limited government, traditional morality, and hawk ver-
sus dove dimensions. Lacking direct measures of voter perceptions of can-
didate positions, I presumed the salience of these affairs, the persistence of
party stereotypes and issue handling reputations on first principles, and the
multiple opportunities for campaign learning help most people see where the
candidates stand on each dimension. This is a reasonable assumption, per-
haps, but an assumption all the same.
Third, I studied the relative influence of different classes of policy atti-
tudes on the vote, conditional on a single moderator—political sophistica-
tion. Other approaches suggest variation along intraattitudinal dimensions of
strength matter too. Prior work chronicles that attitude accessibility, ambiva-
lence, importance, and uncertainty influence political judgment and behavior
(Krosnick, 1990; Lavine et al., 2012). Perhaps some people attach much greater
importance to limited government than others. In this case, attitude impor-
tance could alter the magnitude of the relationship between principles and
candidate preferences. Or perhaps those sure of their positions make greater
use of principles than those bedeviled by some uncertainty. Having noted this,
the attitude stability evidence in Chapter 5 implies policy principles are strong
attitudes for most people, hence they should continue to matter as a general
rule.
A fourth issue concerns the generalizability of these results over time. The
1988–2008 period studied here occurred against a backdrop of intense elite
polarization, thereby raising questions about what happened before and what
is likely to transpire going forward [the paucity of adequate measures of pol-
icy principles on National Election Study (NES) surveys before 1988 dictated
my temporal focus]. What occurred earlier depends on when the cleavages
mapped onto the party system. Thus, the prospects of voting on the moral
dimension seem dim prior to the 1960s when widespread social conflict
emerged and the GOP sought to exploit cultural backlash for electoral gain
later that decade. Similarly, widespread hawk–dove voting seems more likely
to have occurred after the parties began to separate on this dimension in the
early 1980s than during the era of Cold War consensus. Truman and Kennedy
were Cold War warriors after all. But when attention turns to limited govern-
ment, this cleavage has probably shaped candidate choice since the New Deal
(cf. Campbell et al., 1960). Indeed, in terms of its durability over time and the
role it has played in structuring electoral choice throughout the culture wars
and various cold and hot wars, we might reasonably posit that limited gov-
ernment is the dominant cleavage in recent American history (cf. Claggett &
Shafer, 2010).
What of the future? Should we expect these cleavages to endure going
forward? Because elite polarization shows no sign of abating anytime soon,
party signals will remain clear for some time on these dimensions. The ques-
tion then becomes how salient these conflicts are likely to be. Battles over the
size and scope of government have continued to rage since Obama’s elec-
tion. Obama and congressional Democrats have clashed with the congressio-
nal wing of the GOP over the stimulus bill, health care reform, government
spending, and taxes, all manifestations of the broader conflict about the role
the federal government should play in American economic life. On the moral
dimension, the trajectory seems more difficult to chart. American society has
become more progressive on some matters over the past decade or so (e.g.,
gay rights, gay marriage, and marijuana usage). Cohort replacement, whereby
older, less tolerant citizens are replaced by younger, more tolerant counter-
parts, would seem to ensure that these liberalizing trends continue apace. But
in the short term, it is easy to imagine that traditional values will continue to
influence candidate preferences so long as a sizable share of the public presses
demands for moral restoration. Lastly, the hawk–dove dimension seems most
dependent on prevailing developments in the world. As I found in Chapter 8,
attitudes toward this dimension did not significantly affect the vote in 1996,
an understandable result given the demise of the Cold War a few years ear-
lier and the lack of a shooting war that year. If America becomes preoccupied
with domestic problems and turns inward or international problems revolve
around economic rather than security issues, the hawk–dove cleavage will
diminish in relevance. Of course, as all presidents have learned the hard way,
foreign policy crises that require difficult choices about the application of mil-
itary power often explode without warning. Another terrorist attack on U.S.
soil, a civil war in Iraq that draws in neighboring countries, or conflict with
Iran or North Korea that spirals out of control could quickly elevate the sali-
ence of this dimension.
The next qualification to stress is that sophistication conditions some
elements of political judgment and electoral choice. Throughout the book I
have made a strong case that sophistication is less important than we have
been led to believe. This should not be read as saying that sophistication is
irrelevant or that it has no bearing whatsoever. To revisit findings attesting
to its significance, the partisan identities and policy principles adopted by
the highly informed are more durable over time than those held by the less
informed (Chapter 5); the sophisticated rely more heavily than the unsophis-
ticated on limited government to some extent, and moral traditionalism to an
even greater extent, to vertically constrain their positions on concrete issues
(Chapter 6); the aware make greater use of liberal–conservative attitudes than
the unaware to construct preferences and guide votes (Chapters 6 and 8); and
sophistication enhances issue voting (Chapter 8). Put simply, sophistication
affects how people decide, but again, its absence does not preclude principled
political thought.
Finally, I must acknowledge my argument’s debt to revisionist work on
public opinion, voting behavior, and citizen competence. I drew heavily on the
domain-specific model of public opinion developed by Peffley and Hurwitz
(1985; Hurwitz & Peffley, 1987), Feldman (1988), Feldman and Zaller (1992),
and Miller and Shanks (1996). In The Reasoning Voter, Popkin (1994) argues
that citizens are adept at heuristic reasoning and learning from campaigns, a
perspective that guides the approach taken in these pages. Aldrich et al. (1989)
make many similar points. Most recently, Ansolabehere et al. (2008) come to
the defense of the American voter by showing that issue attitudes are more
stable over time and policy voting is more prevalent than suggested by the
sophistication interaction model.1
Though similar in spirit, two differences between On Voter Competence
and these efforts deserve emphasis. First, some revisionists remain wedded
to liberalism and conservatism and issue preferences. My view is that much
is gained by elaborating the conceptual and empirical distinction between
the three classes of policy attitudes. Insofar as ideological worldviews are
too abstract for most people to comprehend and there is too much informa-
tion about scores of policy controversies within and across multiple policy
domains to contend with, it is not hard to see why everyday people, beset
as they are by the demands of daily living and innate cognitive constraints,
fail to develop strong attitudes toward ideological labels and political issues.
Policy principles are different. They are neither too abstract to grasp intui-
tively nor too concrete and innumerable to keep up with. Citizens develop
1
Nie et al. (1979) offered an especially comprehensive challenge to the Michigan
model of political behavior, but subsequent work has undermined some of the
key evidence (see Smith, 1989, for a review).
policy principles without much difficulty and deploy them as the primary
policy lenses through which the political world can be evaluated. Indeed, this
approach represents an optimal use of policy heuristics in the realm of polit-
ical choice.
The second critical difference lies in the broader scope of my book. My work
alone centers on the historical, psychological, and behavioral importance of
core policy principles. And whereas most revisionist studies restrict analysis
to part of the big picture, I investigated the development, crystallization, ori-
gins, and consequences of policy principles for understanding public opinion
and the presidential vote. These differences aside, all revisionist work is ani-
mated by the contrarian spirit of V.O. Key (1966), who declared in the pages of
his final work that “The voter emerges as a person who appraises the actions
of government, who has policy preferences, and who relates his vote to those
appraisals and preferences” (pp. 58–59). Voters, in short, are no fools.
BROADER IMPLICATIONS
Doubts about the political intelligence of ordinary people go back a very long
time. Plato warned against letting the demos take hold of the ship of state and
regarded democracy as superior only to tyranny. Citizens could not be trusted
to rule; instead, those responsibilities must be reserved unto wise and benev-
olent philosopher kings. Next, though he championed representative govern-
ment, John Stuart Mill (1861/1991) was so apprehensive about its prospects
that he proposed a plurality voting system in which “the institutions of the
country should stamp the opinions of persons of a more educated class as
entitled to greater weight than those of the less educated” (p. 241). Fears that
the “passions” of ordinary people would undermine the political order and
destroy individual liberty pepper The Federalists Papers. To take a single exam-
ple from Federalist Paper 63, “the people, stimulated by some irregular pas-
sion, or some illicit advantage, or misled by the artful misrepresentations of
interested men, may call for measures which they themselves will afterwards
be the most ready to lament and condemn” (Hamilton, Madison, & Jay, 1961,
p. 384). Joseph Schumpeter, writing in 1942, scoffed at the notion of demo-
cratic governance: “the typical citizen drops down to a lower level of mental
performance as soon as he enters the political field. He argues and analyzes in
a way which he would readily recognize as infantile within the sphere of his
real interests. He becomes a primitive again” (p. 262).
The emergence of scientific opinion polling in the 1940s and 1950s seemed
to confirm this gloomy view. As I have noted throughout my book, schol-
ars have been complaining about voter shortfalls ever since. Carping about
the limitations of ordinary citizens extends far beyond political science.

Journalists ask Just How Stupid Are We and answer, pretty stupid (Shenkman,
2008). Historians complain that we live in The Age of American Unreason (Jacoby,
2009). Economists explode The Myth of the Rational Voter, asserting that “voters
are worse than ignorant; they are, in a word, irrational—and vote accordingly”
(Caplan, 2007, p. 2).
My analysis of voter competence has been confined to the single charge
that policy voting escapes most people. Evidence presented within these pages
shows that when it comes to the major policy cleavages that structure political
conflict in the United States, the sophisticated and unsophisticated alike (1) hold
genuine and durable attitudes toward limited government, traditional moral
standards, and military strength; (2) rely heavily on these principles to guide
the positions they take on short-term controversies within the respective issue
clusters; (3) ground policy principles in abstract beliefs about what is important
in life; and (4) call upon each principle to guide the choices they make in U.S.
presidential elections. The weight of the evidence makes it plain that the accusa-
tion should be dismissed. The American voter is exonerated on this charge.
But does it follow that the American voter is politically competent? Not
necessarily. Normative and empirical theorists have developed a set of criteria
by which citizens can be judged. Policy voting is part of that repertoire, but it
is not the only element of political competence, and perhaps not even the most
important. Other facets include participation, discussion, deliberation, open-
mindedness, support for democratic norms, political tolerance, and holding
political leaders accountable for what they do in office, to name a few (e.g.,
Delli Carpini & Keeter, 1996; Thompson, 1970). Sorely needed is an assess-
ment of how citizens perform in toto across all these dimensions. Complicating
this task is the fact that the various criteria might be weighted differently by
different analysts. Finally, we must be sensitive to the effects political sophis-
tication may have on each performance criterion.
A skeptic of the possibility of citizen competence might concede that policy
voting remains an exception, but counter that political sophistication is surely
a prerequisite for model citizenship. The skeptic has a wealth of evidence to
back this claim. To illustrate, sophistication enhances nearly all forms of politi-
cal participation. Other things being equal, those who know a lot about politics
turn out, discuss public affairs, contact public officials, volunteer, and donate
their labor to campaigns at higher rates than those who know little. Similarly,
knowledgeable individuals back democratic norms and embrace political
tolerance more enthusiastically than the unknowledgeable (Delli Carpini &
Keeter, 1996; McClosky & Zaller, 1984; Verba, Schlozman, & Brady, 1995). On
these desiderata, then, sophistication does what it is supposed to do.
Yet sophistication also facilitates less noble behaviors, a point not always
acknowledged by those who sing its praises. To take some examples, the politi-
cally aware are more likely to process new political information in ways that
reinforce their prior beliefs and ignore evidence that contradicts their expecta-
tions (Lodge & Hamill, 1986; Taber & Lodge, 2006); the pernicious influence of
partisan bias on subjective political judgments and objective facts grows stron-
ger at higher levels of sophistication (Shani, 2006); educated and sophisticated
individuals are especially susceptible to thinking about welfare in race-coded
terms and applying a racial double standard to claims for economic assistance
made on behalf of blacks (Federico, 2006; Goren, 2003); the sophisticated are
more likely to ape the positions taken by political leaders and even political
entertainers such as Rush Limbaugh than their less sophisticated counterparts
(Barker, 2002; Zaller, 1992).
Although political sophistication is ritualistically hailed for promoting the
hallmarks of good citizenship, the dark side of sophistication has been less
widely noted (cf. Lavine et al., 2012). To the extent that sophistication promotes
selective perception, motivated reasoning, dogmatism, blind partisan loyalty,
ideological rigidity, and so on, its status as normatively sacrosanct should be
challenged. This is not to say that ignorance is preferable to knowledge, but
rather that sophistication has both positive and negative ramifications for dem-
ocratic life. The case for sophistication must be assessed civic virtue by civic
virtue and normative judgments should be withheld unless backed by careful
theoretical reasoning and rigorous empirical scrutiny. Ultimately, I would not
be surprised if sophistication’s benefits outweigh its liabilities. At the same
time, it would not surprise me if the differences between the sophisticated and
the unsophisticated turn out to be as modest for other dimensions of political
competence as I have shown them to be in the case of policy voting.
To conclude, Delli Carpini and Keeter (1996) may well be right that
“informed citizens are demonstrably better citizens, as judged by the stan-
dards of democratic theory and practice underpinning the American system”
(p. 272). At the end of the day, I have rebutted a single charge in the indictment
against the American voter. When it comes to developing and holding policy
principles, using these to guide the positions they take on short-term political
judgments, and grounding the presidential vote in such postures, the unso-
phisticated do about as well as the sophisticated. Given all that we have been
told about the cognitive, motivational, and normative deficiencies of ordinary
citizens, these are no mean achievements. But other charges have been leveled
against the people, ones that I have not addressed here. The question of which
way the scale ultimately tips remains to be settled.
Appendix: Measurement of Key Variables
Political sophistication items: I use a combination of factual political knowledge

questions, seven-point issue placement scales (respondents get credit by plac-
ing Democrats in the liberal end and Republicans in the conservative end of
the scale), and the National Election Study (NES) interviewer rating to tap
sophistication.
1988 NES: ideology for Dukakis (v231) and Bush (v232), spending and
services for the Republican (v307) and Democratic Party (v308), defense
spending for the Republican (v315) and Democratic Party (v316), interviewer
rating (v555), party control of House (v878) and Senate (v879), which party is
more conservative (v811), Kennedy position (v871), Schultz position (v872),
Rehnquist position (v873), Gorbachev position (v874), Thatcher position
(v875), Arafat position (v876), Wright position (v877).
1992 NES: ideology for Bush (v3514) and Clinton (v3515), interviewer rat-
ing (v4205), which party is more conservative (v5915), who judges constitu-
tionality of laws (v5920), who nominates federal judges (v5921), party control
of House (v5951) and Senate (v5952), Quayle position (v5916), Rehnquist posi-
tion (v5917), Yeltsin position (v5918), Foley position (v5919).
1996 NES: interviewer rating (v70), ideology for Clinton (v369) and Dole
(v371), ideology for the Democratic (v379) and Republican Party (v380),
change in budget deficit (v392), party control of House (v1072) and Senate
(v1073), Gore position (v1189), Rehnquist position (v1190), Yeltsin position
(v1191), Gingrich position (v1192).
2000 NES: interviewer rating (v1033), party control of House (v1356) and
Senate (v1357), ideology for Clinton (v1371), Gore (v1372), Bush (v1374), and
Buchanan (v1376), ideology for the Democratic (v1382) and Republican Party
(v1383), Cheney home state (v1466), Lieberman home state (v1467).
2004 NES: ideology for Bush (v3087), Kerry (v3088), and Nader (v3089),
ideology for the Democratic (v3090) and Republican Party (v3091), spending
and services for the Republican (v3141) and Democratic Party (v3140), Bush
245
246 APPENDIX: MEASUREMENT OF KEY VARIABLES
environment v. jobs (v3184), interviewer rating (v3403), party control of House

(v5089) and Senate (v5090), left–right scale for the Democratic (v5260) and
Republican Party (v5261), left–right scale for Bush (v5263), Kerry (v5264), and
Nader (v5265).
2008 NES: ideology for Obama (v3070a), McCain (v3070b), ideology for
the Democratic (v3071a) and Republican Party (v3071b), interviewer rating
(v3303), which party is more conservative (v5119a), party control of House
(v5066) and Senate (v5067), left–right scale for the Democratic (v5189a) and
Republican Party (v5189b), left–right scale for McCain (v5190a), and Obama
(v5190b).
2007 and 2008 Knowledge Networks: education dummy variables for 4-year
college graduates and beyond and high school graduates and beyond.
2011 YouGov: number of presidential terms, constitutionality of law, Clinton
position, U.N. chief responsibility.
Limited government items: I use various combinations of items depending on
what is available in a given survey.
1988 NES: Federal spending on food stamps (v349), the unemployed (v379),
child care (v382), the homeless (v385), the elderly (v384), Social Security
(v348).
1992 NES: Federal spending on food stamps (v3725), welfare (v3726), the
homeless (v3730), the poor (v3817), Social Security (v3811), child care (v3813).
1996 NES: Federal spending on food stamps (v496), welfare (v497), the
homeless (v501), the poor (v565), Social Security (v560), child care (v564).
2000 NES: Federal spending on food stamps (v679), welfare (v676), Social
Security (v681), child care (v685), the poor (v680).
2004 NES: Federal spending on welfare (v3169), the poor (v3172), Social
Security (v3170), child care (v3165).
2008 NES: Federal spending on welfare (v3145x), the poor (v3148x), Social
Security (v3141x), child care (v3146x).
2007 Knowledge Networks: Federal spending on welfare, food stamps, the
unemployed, the homeless, child care.
Traditional morality items: I use four items to tap this concept. The items are
the same in every survey: new lifestyles, adjust views of moral behavior, tra-
ditional family ties, tolerate different moral standards.
1988 NES: v951, v952, v953, v954.
1992 NES: v6115, v6116, v6117, v6118.
1996 NES: v1247, v1248, v1249, v1250.
2000 NES: v1530, v1531, v1532, v1533.
2004 NES: v5189, v5190, v5191, v5192.
2008 NES: v5139, v5140, v5141, v5142.
Appendix: Measurement of Key Variables 247
2008 Knowledge Networks: same items.

Military strength items: I use various combinations of items depending on
what is available in a given survey.
1988 NES: strong military v. bargaining (v966), strong military important
(v967), United States remain powerful (v972).
1992 NES: maintain military power (v3603), willingness to use force
(v3605).
1996 NES: willingness to use force (v411).
2000 NES: No items available.
2004 NES: preelection military–diplomacy scale (v3107), postelection mili-
tary–diplomacy scale (v5124), strong military important (v5108).
2008 NES: No items available.
2011 YouGov: United States strong and tough versus understanding and
flexible; diplomacy or military force.
References
Abramowitz, A. I. (1994). Issue evolution reconsidered: Racial attitudes and parti-

sanship in the U.S. electorate. American Journal of Political Science, 38, 1–24.
Abramowitz, A. I. (1995). It’s abortion, stupid: Policy voting in the 1992 presiden-
tial election. Journal of Politics, 57, 176–186.
Abramson, P. R., Aldrich, J. H., & Rohde, D. W. (2009). Change and continuity in the
2008 elections. Washington, DC: CQ Press.
Achen, C. H. (1975). Mass political attitudes and the survey response. American
Political Science Review, 69, 1218–1231.
Achen, C. H. (1982). Interpreting and using regression. Beverly Hills, CA: Sage.
Adams, G. D. (1997). Abortion: Evidence of an issue evolution. American Journal of
Political Science, 41, 718–737.
Addington, L. H. (2000). America’s war in Vietnam: A short narrative history.
Bloomington, IN: University of Indiana Press.
Aitken, J. (1993). Nixon: A life. Washington, DC: Regnery.
Aldrich, J. H., Sullivan, J. L., & Borgida, E. (1989). Foreign affairs and issue voting:
Do presidential candidates “waltz before a blind audience?” American Political
Science Review, 83, 123–141.
Allport, G. W. (1935). Attitudes. In C. Murchison (Ed.), A handbook of social psychol-
ogy (pp. 798–844). Worcester, MA: Clark University Press.
Alwin, D. F., & Krosnick, J. A. (1991). Aging, cohorts, and the stability of sociopolit-
ical orientations over the life span. American Journal of Sociology, 97, 169–195.
Anand, S., & Krosnick, J. A. (2003). The impact of attitudes toward foreign policy
goals on public preferences among presidential candidates: A study of issue
publics and the attentive public in the 2000 U.S. presidential election. Presidential
Studies Quarterly, 33, 31–71.
Ansolabehere, S., Rodden, J., & Snyder, J. M. (2008). The strength of issues: Using
multiple measures to gauge preference stability, ideological constraint, and
issue voting. American Political Science Review, 102, 215–232.
Bardi, A., & Schwartz, S. H. (2003). Values and behavior: Strength and structure of
relations. Personality and Social Psychology Bulletin, 29, 1207–1220.
249
250 REFERENCES
Barker, D. C. (2002). Rushed to judgment: Talk radio, persuasion, and American political
behavior. New York, NY: Columbia University Press.
Bartels, L. M. (2002). Beyond the running tally: Partisan bias in political percep-
tions. Political Behavior, 24, 117–150.
Bartels, L. M. (2008). Unequal democracy: The political economy of the new gilded age.
Princeton, NJ: Princeton University Press.
Baum, M. A. (2002). Sex, lies, and war: How soft news brings foreign policy to the
inattentive public. American Political Science Review, 96, 91–109.
Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Advances in experi-
mental social psychology (Vol. 6, pp. 1–62). New York, NY: Academic Press.
Bennett, S. E. (1995). Americans’ knowledge of ideology, 1980–1992. American
Politics Quarterly, 23, 259–278.
Berelson, B. R., Lazarsfeld, P. F., & McPhee, W N. (1954). Voting: A study of opinion
formation in a presidential campaign. Chicago, IL: University of Chicago Press.
Berinsky, A. J. (2009). In time of war: Understanding American public opinion from
World War II to Iraq. Chicago, IL: University of Chicago Press.
Bernaards, C. A., & Sijtsma, K. (1999). Factor analysis of multidimensional poly-
tomous item response data suffering from ignorable item nonresponse.
Multivariate Behavioral Research, 34, 277–313.
Bishop, G. F., Oldendick, R. W., Tuchfarber, A. J., & Bennett, S. E. (1980). Pseudo-
opinions on public affairs. Public Opinion Quarterly, 44, 198–209.
Blum, J. M. (1991). Years of discord: American politics and society, 1961–1974. New York,
NY: W.W. Norton.
Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley.
Boyer, P. S. (Ed.). (2001). The Oxford companion to United States history. New York,
NY: Oxford University Press.
Brewer, M. D., & Stonecash, J. M. (2007). Split: Class and cultural divides in American
politics. Washington, DC: CQ Press.
Brewer, M. D., & Stonecash, J. M. (2009). The dynamics of American political parties.
New York, NY: Cambridge University Press.
Brewer, P. R. (2003). Values, political knowledge, and public opinion about gay
rights: A framing-based account. Public Opinion Quarterly, 67, 173–201.
Campbell, A., Converse, P. E., Miller, W. E., & Stokes, D. E. (1960). The American
voter. New York, NY: Wiley.
Campbell, D. T. (1963). Social attitudes and other acquired behavioral dispositions.
In S. Koch (Ed.), Psychology: A study of a science (Vol. 6, pp. 94–172). New York,
NY: McGraw-Hill.
Caplan, B. (2007). The myth of the rational voter: Why democracies choose bad policies.
Princeton, NJ: Princeton University Press.
Caprara, G. V., Schwartz, S., Capanna, C., Vecchione, M., & Barbaranelli, C. (2006).
Personality and politics: Values, traits, and political choice. Political Psychology,
27, 1–28.
Carmines, E. G., & Stimson, J. A. (1980). The two faces of issue voting. American
Political Science Review, 74, 78–91.
References 251
Carmines, E. G., & Stimson, J. A. (1989). Issue evolution: Race and the transformation
of American politics. Princeton, NJ: Princeton University Press.
Carney, D. R., Jost, J. T., Gosling, S. D., & Potter, J. (2008). The secret lives of liberals
and conservatives: Personality profiles, interaction styles, and the things they
leave behind. Political Psychology, 29, 807–840.
Carsey, T. M., & Layman, G. C. (2006). Changing sides or changing minds? Party
identification and policy preferences in the American electorate. American
Journal of Political Science, 50, 464–477.
Carter, D. T. (1995). The politics of rage: George Wallace, the origins of the new con-
servatism, and the transformation of American politics. New York, NY: Simon &
Schuster.
Chaiken, S. (1987). The heuristic model of persuasion. In M. P. Zanna, J. M. Olson,
& C. P. Herman (Eds.), Social influence: The Ontario Symposium (Volume 5,
pp. 3–39). Hillsdale, NJ: Lawrence Erlbaum.
Chong, D., McClosky, H., & Zaller, J. (1983). Patterns of support for democratic
and capitalist values in the United States. British Journal of Political Science, 13,
401–440.
Claggett, W. J. M., & Shafer, B. E. (2010). The American public mind: The issues struc-
ture of mass politics in the postwar United States. New York, NY: Cambridge
University Press.
Collins, R. M. (2000). More: The politics of economic growth in postwar America.
New York, NY: Oxford University Press.
Collins, R. M. (2007). Transforming America: Politics and culture during the Reagan
years. New York, NY: Columbia University Press.
Conover, P. J., & Feldman, S. (1981). The origins and meaning of liberal/conserva-
tive self identifications. American Journal of Political Science, 25, 617–645.
Conover, P. J., & Feldman, S. (1989). Candidate perception in an ambiguous world:
Campaigns, cues, and inference processes. American Journal of Political Science,
33, 912–940.
Conover, P. J., Feldman, S., & Knight, K. (1986). Judging inflation and unemploy-
ment: The origins of retrospective evaluations. Journal of Politics, 48, 565–588.
Converse, P. E. (1964). The nature of belief systems in mass publics. In D. Apter
(Ed.), Ideology and Discontent (pp. 206–261). New York, NY: Free Press.
Converse, P. E. (1970). Attitudes and non-attitudes: Continuation of a dialogue. In
E. R. Tufte (Ed.), The quantitative analysis of social problems (pp. 168–189). Reading,
MA: Addison-Wesley.
Converse, P. E. (1975). Public opinion and voting behavior. In F. I. Greenstein &
N. W. Polsby (Eds.), Handbook of political science, volume 4: Nongovernmental poli-
tics (pp. 75–169). Reading, MA: Addison-Wesley.
Converse, P. E. (1990). Popular representation and the distribution of information.
In J. A. Ferejohn & J. H. Kuklinski (Eds.), Information and democratic processes
(pp. 369–388). Urbana, IL: University of Illinois Press.
Converse, P. E. (2000). Assessing the capacity of mass electorates. Annual Review of
252 REFERENCES
Converse, P. E. (2006). Democratic theory and electoral reality. Critical Review, 18,
297–329.
Converse, P. E., & Markus, G. B. (1979). Plus ca change . . . : The new CPS election
study panel. American Political Science Review, 73, 32–49.
Curran, P. J., West. S. G., & Finch, J. F. (1996). The robustness of test statistics to non-
normality and specification error in confirmatory factor analysis. Psychological
Methods, 1, 16–29.
Dallek, R. 2003. An unfinished life: John F. Kennedy, 1917–63. Boston, MA: Little
Brown and Co.
Dancey, L., & Goren, P. (2010). Party identification, issue attitudes, and the dynam-
ics of political debate. American Journal of Political Science, 54, 686–699.
Davidov, E., Schmidt, P., & Schwartz, S. H. (2008). Bringing values back in: The
adequacy of the European Social Survey to measure values in 20 countries.
Public Opinion Quarterly, 72, 420–445.
DeGroot, G. J. (2008). The sixties unplugged: A kaleidoscopic history of a disorderly
decade. Cambridge, MA: Harvard University Press.
Delli Carpini, M. X., & Keeter, S. (1993). Measuring political knowledge: Putting
first things first. American Journal of Political Science, 37, 1179–1206.
Delli Carpini, M. X., & Keeter, S. (1996). What Americans know about politics and why
it matters. New Haven, CT: Yale University Press.
Downs, A. (1957). An economic theory of democracy. New York, NY: Harper & Row.
Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX:
Harcourt, Brace, Jovanovich.
Edsall, T. B., & Edsall, M. D. (1991). Chain reaction: The impact of race, rights, and taxes
on American politics. New York, NY: W.W. Norton.
Enelow, J. M., & Hinich, M. J. (1984). The spatial theory of voting: An introduction.
Fazio, R. H. (1989). On the power and functionality of attitudes: The role of atti-
tude accessibility. In A. R. Pratkanis, S. J. Breckler, & A. G. Greenwald (Eds.),
Attitude structure and function (pp. 153–179). Hillsdale, NJ: Lawrence Erlbaum.
Fazio, R. H. (2007). Attitudes as object-evaluation associations of varying strength.
Social Cognition, 25, 603–637.
Federico, C. M. (2006). Race, education, and individualism revisited. Journal of
Politics, 68, 600–610.
Feldman, S. (1988). Structure and consistency in public opinion: The role of core
beliefs and values. American Journal of Political Science, 32, 416–440.
Feldman, S. (1989). Measuring issue preferences: The problem of response instabil-
ity. In J. A. Stimson (Ed.), Political analysis: Volume 1 (pp. 25–60). Ann Arbor, MI:
University of Michigan Press.
Feldman, S, & Conover, P. J. (1983). Candidates, issues, and voters: The role of
inference in political perception. Journal of politics, 45, 810–839.
Feldman, S., & Zaller, J. (1992). The political culture of ambivalence: Ideological
responses to the welfare state. American Journal of Political Science, 36, 268–307.
References 253
Finney, S. J., & DiStefano, C. (2006). Non-normal and categorical data in structural
equation modeling. In G. R. Hancock & R. O. Mueller (Eds.), Structural equa-
tion modeling: A second course (pp. 269–314). Greenwich, CT: Information Age
Publishing.
Fiorina, M. P. (1981). Retrospective voting in American national elections. New Haven,
CT: Yale University Press.
Fiorina, M. P. (1990). Information and rationality in elections. In J. A. Ferejohn &
J. H. Kuklinski (Eds.), Information and democratic processes (pp. 329–342). Urbana,
IL: University of Illinois Press.
Fiske, S. T., Kinder, D. R., & Larter, W. M. (1983). The novice and the expert:
Knowledge-based strategies in political cognition. Journal of Experimental Social
Psychology, 19, 381–400.
Fiske, S. T., Lau, R. L., & Smith, R. A. (1990). On the variety and utility of political
knowledge structures. Social Cognition, 8, 31–48.
Fiske, S. T., & Taylor, S. E. (1991). Social cognition. New York, NY: McGraw-Hill.
Friedrich, R. J. (1982). In defense of multiplicative terms in multiple regression
equations. American Journal of Political Science, 26, 797–833.
Gaddis, J. L. (2005). The cold war: A new history. New York, NY: Penguin Press.
Gawronski, B. (2007). Editorial: Attitudes can be measured! But what is an atti-
tude? Social Cognition, 25, 573–581.
Gelman, A., & Stern, H. (2006). The difference between “significant” and “not
significant” is not itself statistically significant. American Statistician, 60,
328–331.
Gerber, A. S., Huber, G. D., Doherty, D., Dowling, C. M., & Ha, S. E. (2010).
Personality and political attitudes: Relationships across issue domains and
political contexts. American Political Science Review, 104, 111–133.
Gerring, J. (1998). Party ideologies in America, 1828–1996. New York, NY: Cambridge
University Press.
Gilens, M. (1999). Why Americans hate welfare: Race, media, and the politics of antipov-
erty policy. Chicago, IL: University of Chicago Press.
Goldberg, R. A. (1995). Barry Goldwater. New Haven, CT: Yale University Press.
Goren, P. (1997). Political expertise and issue voting in presidential elections.
Political Research Quarterly, 50, 387–412.
Goren, P. (2001). Core principles and policy reasoning in mass publics. A test of
two theories. British Journal of Political Science, 31, 159–177.
Goren, P. (2002). Character weakness, partisan bias, and presidential evaluation.
American Journal of Political Science, 46, 627–641.
Goren, P. (2003). Race, sophistication, and white opinion on government spending.
Political Behavior, 25, 201–220.
Goren, P. (2004). Political sophistication and policy reasoning: A reconsideration.
Goren, P. (2005). Party identification and core political values. American Journal of
254 REFERENCES
Goren, P. (2007). Character weakness, partisan bias, and presidential evaluation:

Modifications and extensions. Political Behavior, 29, 305 –325.
Goren, P. (2009). Personal value systems and public opinion: A preliminary report
on some new value items. Unpublished manuscript.
Goren, P., Federico, C. M., & Kittilson, M. C. (2009). Source cues, partisan identities,
and political value expression. American Journal of Political Science, 53, 805–820.
Goren, P., Reifler, J., & Scotto, T. (2011). Personality, values, and public opinion.
Unpublished manuscript.
Greeley, A. M., & Hout, M. (2006). The truth about conservative Christians: What they
think and what they believe. Chicago, IL: University of Chicago Press.
Green, D., Palmquist, B., & Schickler, E. (2002). Partisan hearts and minds: Political
parties and the social identities of voters. New Haven, CT: Yale University Press.
Green, D. P., & Citrin, J. (1994). Measurement error and the structure of attitudes:
Are positive and negative judgments opposites? American Journal of Political
Science, 38, 256–281.
Green, D. P., & Palmquist, B. (1990). Of artifacts and partisan instability. American
Halberstam, D. (1993). The fifties. New York, NY: Villard Books.
Halberstam, D. (2001). War in a time of peace: Bush, Clinton, and the generals. New York,
NY: Scribner.
Hamby, A. L. (1992). Liberalism and its challengers: From F. D. R. to Bush (2nd ed.).
Hamilton, A., Madison, J., & Jay, J. (1961). The federalist papers. New York, NY:
Penguin Books.
Hayes, D. (2005). Candidate qualities through a partisan lens: A theory of trait
ownership. American Journal of Political Science, 49, 908–923.
Heeringa, S. G., West, B. T., & Berglund, P. A. (2010). Applied survey data analysis.
Boca Raton, FL: CRC Press.
Herring, G. C. (2002). America’s longest war: The United States and Vietnam, 1950–
1975 (4th ed.). New York, NY: McGraw Hill.
Herring, G. C. (2008). From colony to superpower: U.S. foreign relations since 1776.
Hetherington, M. J. (2005). Why trust matters: Declining political trust and the demise
of American liberalism. Princeton, NJ: Princeton University Press.
Hetherington, M. J., & Weiler, J. D. (2009). Authoritarianism and polarization in
American politics. New York, NY: Cambridge University Press.
Higgins, E. T., & King, G. (1981). Accessibility of social constructs: Information
processing consequences of individual and contextual variability. In N. Cantor
& J. Kihlstrom (Eds.), Personality, Cognition, and Social Interaction (pp. 69–121).
Hillsdale, NJ: Lawrence Erlbaum.
Himmelstein, J. (1983). The new right. In R. Liebman & R. Wuthnow (Eds.), The
new Christian right: Mobilization and legitimation (pp. 13–30). Hawthorne, NY:
Aldine Publishing Co.
References 255
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance struc-
ture analysis: Conventional criteria versus new alternatives. Structural Equation
Modeling, 6, 1–55.
Huckfeldt, R. R., & Sprague, J. D. (1995). Citizens, politics, and social communica-
tion: Information and influence in an election campaign. New York, NY: Cambridge
University Press
Hunter, J. D. (1992). Culture wars: The struggle to define America. New York, NY:
Basic Books.
Hurwitz, J., & Peffley, M. (1987). How are foreign policy attitudes structured? A
hierarchical model. American Political Science Review, 81, 1099–1120.
Jacoby, S. (2009). The age of American unreason. New York, NY: Vintage.
Jacoby, W. G. (1994). Public attitudes toward government spending. American
Jennings, M. K. (1992). Ideological thinking among mass publics and political
elites. Public Opinion Quarterly, 56, 419–441.
John, O. P., Naumann L. P., & Soto, C. J. (2008). Paradigm shift to the integrative
big five trait taxonomy: History, measurement, and conceptual Issues.” In O.
P. John, R. W. Robins, & L. A. Pervin, (Eds.), Handbook of Personality: Theory and
research, 3rd ed. (pp. 114–158). New York, NY: Guilford Press.
Jost, J. T. (2006). The end of the end of ideology. American Psychologist, 61,
651–670.
Jost, J. T., Federico, C. M., & Napier, J. L. (2009). Political ideology: Its structure,
functions, and elective affinities. Annual Review of Psychology, 60, 307–337.
Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003). Political conserva-
tism as motivated social cognition. Psychological Bulletin, 129, 339–375.
Judd, C. M., Drake, R. A., Downing, J. M., & Krosnick, J. A. (1991). Some dynamic
properties of attitude structures: Context-induced response facilitation and
polarization. Journal of Personality and Social Psychology, 60, 193–202.
Just, M. R., Crigler, A. N., Alger, D. E., Cook, T. E., Kern, M., & West, D. M. (1996).
Crosstalk: Citizens, candidates, and the media in a presidential campaign. Chicago, IL:
University of Chicago Press.
Katz, D. (1960). The functional approach to the study of attitudes. Public Opinion
Quarterly, 24, 163–204.
Kennedy, D. M. (1999). Freedom from fear: The America people in war and depression,
1929–1945. New York, NY: Oxford University Press.
Kerlinger, F. N. (1984). Liberalism and conservatism: The nature and structure of social
attitudes. Hillsdale, NJ: Lawrence Erlbaum.
Key, V. O. (1966). The responsible electorate: Rationality and presidential voting, 1936–60.
Cambridge, MA: Belknap Press.
Kinder, D. R. (1998). Opinion and action in the realm of politics. In A. R. Pratkanis,
S. J. Breckler, & A. G. Greenwald (Eds.), Handbook of social psychology, 4th ed.
(pp. 778–867). Hillsdale, NJ: Lawrence Erlbaum Press.
Kinder, D. R. (2006). Belief systems today. Critical Review, 18, 197–216.
256 REFERENCES
Kinder, D. R., & Kam, C. D. (2008). Us against them: Ethnocentric foundations of

American opinion. Chicago, IL: University of Chicago Press.
Kinder, D. R., & Sanders, L. M. (1996). Divided by color: Racial politics and democratic
ideals. Chicago, IL: University of Chicago Press.
Knight, K. (1985). Ideology in the 1980 election: Ideological sophistication does
matter. Journal of Politics, 47, 828–853.
Kohut, A., Doherty, C., Dimock, M., & Keeter, S. (2010, November 18). Public knows
basic facts about politics, economics, but struggles with specifics. Retrieved
from http://people-press.org/files/legacy-pdf/677.pdf.
Kohut, A., & Keeter, S. (2010, July 15). Well known: Twitter; little known: John
Roberts. Retrieved from http://people-press.org/files/legacy-pdf/635.pdf.
Krosnick, J. A. (1990). Government policy and citizen passion: A study of issue
publics in contemporary America. Political Behavior, 12, 59–92.
Krosnick, J. A. (2002). The causes of no-opinion responses to attitude measures in
surveys: They rarely are what they appear to be. In R. M. Groves, D. A. Dillman,
J. L. Eltinge, & R. J. A. Little (Eds.), Survey nonresponse (pp. 88–100). New York,
NY: Wiley.
Krosnick, J. A., & Berent, M. K. (1993). Comparisons of party identification and
policy preferences: The impact of survey question format. American Journal of
Kuklinski, J. H. (2001). Introduction. In J. H. Kuklinski (Ed.), Citizens and poli-
tics: Perspectives from political psychology (pp. 355–365). New York: Cambridge
University Press.
Kull, S., Ramsay, C., & Lewis, E. (2003/2004). Misperceptions, the media, and the
Iraq war. Political Science Quarterly, 118, 569–598.
Lau, R. R., & Redlawsk, D. P. (2006). How voters decide: Information processing during
election campaigns. New York, NY: Cambridge University Press.
Lavine, H., Johnston, C., & Steenbergen, M. (2012). The ambivalent partisan: How
critical loyalty promotes democracy. New York, NY: Oxford University Press.
Layman, G. (2001). The great divide: Religious and cultural conflict in American party
politics. New York, NY: Columbia University Press.
Layman, G. C., & Carsey, T. M. (2002). Party polarization and “conflict extension”
in the American electorate. American Journal of Political Science, 46, 786–802.
Levitin, T. E., & Miller, W. E. (1979). Ideological interpretations of presidential elec-
tions. American Political Science Review, 73, 751–771.
Lewis-Beck, M. S., Jacoby, W. G., Norpoth, H., & Weisberg, H. F. (2008). The American
voter revisited. Ann Arbor, MI: University of Michigan Press.
Liebman, R. C., Wuthnow, R., & Guth, J. L. (Eds.). (1983). The new Christian right:
mobilization and legitimation. Hawthorne, NY: Aldine Publishing Co.
Lindaman, K., & Haider-Markel, D. R. (2002). Issue evolution, political parties, and
the culture wars. Political Research Quarterly, 55, 91–110.
Lodge, M., & Hamill, R. (1986). A partisan schema for political information pro-
cessing. American Political Science Review, 80, 505–520.
References 257
Lubell, S. (1956). Revolt of the moderates. New York, NY: Harper.

Luskin, R. C. (1987). Measuring political sophistication. American Journal of Political
Science, 31, 856–899.
Luskin, R. C. (2002). Political psychology, political behavior, and politics:
Questions of aggregation, causal distance, and taste. In J. H. Kuklinski (Ed.),
Thinking about political psychology (pp. 217–252). New York, NY: Cambridge
University Press.
Maio, G. R., & Olson, J. M. (1995). Relations between values, attitudes, and behav-
ioral intentions: The moderating role of attitude function. Journal of Experimental
Social Psychology, 31, 265–285.
Markus, G. B. (2001). American individualism reconsidered. In J. H. Kuklinski
(Ed.), Citizens and politics: Perspectives from political psychology (pp. 401–432).
McAdams, D. P., & Pals, J. L. (2006). A new big five: Fundamental principles for an
integrative science of personality. American Psychologist, 61, 204–217.
McCann, J. A. (1997). Electoral choices and core value change: The 1992 presiden-
tial campaign. American Journal of Political Science, 41, 564–583.
McCarty, N. M., Poole, K. T., & Rosenthal, H. (2006). Polarized America: The dance of
ideology and unequal riches. Cambridge, MA: MIT Press.
McClosky, H., & Zaller, J. (1984). The American ethos: Public attitudes toward capital-
ism and democracy. Cambridge, MA: Harvard University Press.
McCrae, R. R., & Costa, P. T. (1999). A five-factor theory of personality. In L. A.
Pervin & O. P. John (Eds.), Handbook of personality: Theory and research, 2nd ed.
(pp. 139–153). New York, NY: Guilford Press.
McGuire, W. J. (1969). The nature of attitudes and attitude change. In G. Lindzey &
E. Aronson (Eds.), Handbook of social psychology, 2nd ed. (pp. 136–314). Reading,
MA: Addison-Wesley.
Mendelberg, T. (2001). The race card: Campaign strategy, implicit messages, and the
norm of equality. Princeton, NJ: Princeton University Press.
Mill, J. S. (1861/1991). Considerations on representative government. Buffalo, NY:
Prometheus Books.
Miller, W. E., & Shanks, J. M. (1996). The new American voter. Cambridge, MA:
Harvard University Press.
Miringoff, M., & Miringoff, M. L. (1999). The social health of the nation: How America
is really doing. New York, NY: Oxford University Press.
Mondak, J. J. (2010). Personality and the foundations of political behavior. New York,
NY: Cambridge University Press.
Mueller, J. M. (1973). Wars, presidents, and public opinion. New York, NY: Wiley.
Neuman, W. R. (1986). The paradox of mass politics: Knowledge and opinion in the
American electorate. Cambridge, MA: Harvard University Press.
Nie, N. H., Verba, S., & Petrocik, J. R. (1979). The changing American voter (Enlarged
ed.). Cambridge, MA: Harvard University Press.
258 REFERENCES
Norpoth, H., & Lodge, M. (1985). The difference between attitudes and nonatti-
tudes in the mass public: Just measurement? American Journal of Political Science,
29, 291–307.
Oberdofer, D. (1998). From the cold war to a new era: The United States and the Soviet
Union, 1983–1991 (Updated ed.). Baltimore, MD: John Hopkins University
Press.
Packer, G. (2005). The assassin’s gate: America in Iraq. New York, NY: Farrar, Straus,
and Giroux.
Page, B. I. (1978). Choices and echoes in presidential elections: Rational man and electoral
democracy. Chicago, IL: University of Chicago Press.
Page, B. I., & Bouton, M. M. (2006). The foreign policy disconnect: What Americans
want from our leaders but don’t get. Chicago, IL: University of Chicago Press.
Page, B. I., & Shapiro, R. Y. (1992). The rational public: Fifty years of trends in Americans’
policy preferences. Chicago, IL: University of Chicago Press.
Patterson, J. T. (1994). America’s struggle against poverty, 1900–1994. Cambridge,
MA: Harvard University Press.
Patterson, J. T. (1996). Grand expectations: The United States, 1945–74. New York, NY:
Oxford University Press.
Patterson, J. T. (2005). Restless giant: The United States from Watergate to Bush v. Gore.
Peffley, M., & Hurwitz, J. (1985). A hierarchical model of attitude constraint.
Peffley, M., & Hurwitz, J. (1993). Models of attitude constraint in foreign affairs.
Political Behavior, 15, 61–90.
Pennock, J. R. (1979). Democratic political theory. Princeton, NJ: Princeton University
Press.
Perlstein, R. (2008). Nixonland: The rise of a president and the fracturing of America.
New York, NY: Scribner.
Petrocik, J. R. (1996). Issue ownership in presidential elections, with a 1980 case
study. American Journal of Political Science, 40, 825–850.
Petty, R. E., & Krosnick, J. A. (Eds.) (1995). Attitude strength: Antecedents and conse-
quences. Hillsdale, NJ: Lawrence Erlbaum.
Pierce, J. C., & Hagner, P. R. (1982). Conceptualization and party identification,
1956–1976. American Journal of Political Science, 26, 377–387.
Pollock, P. H., Lilie, S. A., & Vittes, E. (1993). Hard issues, core values, and vertical
constraint: The case of nuclear power. British Journal of Political Science, 23, 29–50.
Pomper, G. M. (1972). From confusion to clarity: Issues and American voters, 1956–
1968. American Political Science Review, 66, 415–428.
Popkin, S. L. (1994). The reasoning voter: Communication and persuasion in presidential
campaigns (2nd. ed.). Chicago, IL: University of Chicago Press.
Price, V., & Zaller, J. (1993). Who gets the news? Alternative measures of news
reception and their implications for research. Public Opinion Quarterly, 57,
133–164.
References 259
Rahn, W. M., Aldrich, J. H., Borgida, E., and Sullivan, J. L. (1990). A social cogni-
tive model of candidate appraisal. In J. A. Ferejohn & J. H. Kuklinski (Eds.),
Information and democratic processes (pp. 136–159). Urbana, IL: University of
Illinois Press.
Ricks, T. E. (2006). Fiasco: The American military adventure in Iraq. New York, NY:
Penguin Press.
Roccas, S., Sagiv, L., Schwartz, S. H., & Knafo, A. (2002). The big five personal-
ity factors and personal values. Personality and Social Psychology Bulletin, 28,
789–801.
Rokeach, M. (1973). The nature of human values. New York, NY: Free Press.
Scammon, R. M., & Wattenberg, B. J. (1970). The real majority: An extraordinary exam-
ination of the American electorate. New York, NY: Coward McCann.
Schafer, J. L., & Graham, J. W. (2002). Missing data. Our view of the state of the art.
Psychological Methods, 7, 147–177.
Schumpeter, J. A. (1950). Capitalism, socialism, and democracy, 3rd ed. New York, NY:
Harper and Brothers.
Schwartz, S. H. (1992). Universals in the content and structure of values: Theoretical
advances and empirical tests in 20 countries. In M. Zanna (Ed.), Advances in
experimental social psychology (pp. 1–65). Orlando, FL: Academic Press.
Schwartz, S. H. (1994). Are there universal aspects in the structure and contents of
human values? Journal of Social Issues, 50, 19–45.
Schwartz, S. H. (2003). A proposal for measuring value orientations across nations
(chapter 7). Retrieved from http://www.europeansocialsurvey.org/index.php?
Schwartz, S. H., Caprara, G. V., & Vecchione, M. (2010). Basic personal values, core
political values, and voting: A longitudinal analysis. Political Psychology, 31,
421–452.
Schwartz, S. H., Sagiv, L., & Boehnke, K. (2000). Worries and values. Journal of
Personality, 68, 309–346.
Schwarz, N., & Bohner, G. (2001). The construction of attitudes. In A. Tesser &
N. Schwarz (Eds.), Blackwell handbook of social psychology: Intraindividual processes
(pp. 436–457). Malden, MA: Blackwell.
Sears, D. O. (2001). The role of affect in symbolic politics. In J. H. Kuklinski (Ed.),
Citizens and politics: Perspectives from political psychology (pp. 14–40). New York,
NY: Cambridge University Press.
Sears, D. O., and Citrin, J. (1985). Tax revolt: Something for nothing in California
(Enlarged ed.). Cambridge, MA: Harvard University Press.
Sears, D. O., Hensler, C. P., & Speer, L. K. (1979). Whites opposition to “busing”:
Self-interest or symbolic politics. American Political Science Review, 73, 369–384.
Sears, D. O., Lau, R. R., Tyler, T. R., & Allen, H. M. (1980). Self-interest vs. symbolic
politics in policy attitudes and presidential voting. American Political Science
Review, 74, 670–684.
Shani, D. (2006). Knowing your colors: Can knowledge correct for partisan bias in
political perceptions? Unpublished manuscript.
260 REFERENCES
Shenkman, R. (2008). Just how stupid are we? Facing the truth about the American voter.
New York, NY: Basic Books.
Smith, E. R. A. N. (1989). The unchanging American voter. Berkeley, CA: University
of California Press.
Sniderman, P. M., Brody, R., & Tetlock, P. E. (1991). Reasoning and choice: Explorations
in political psychology. New York, NY: Cambridge University Press.
Sniderman, P. M., & Piazza, T. (1993). The scar of race. Cambridge, MA: Harvard
University Press.
Spini, D. (2003). Measurement equivalence of 10 value types from the Schwartz
value survey across 21 countries. Journal of Cross-Cultural Psychology, 34, 3–23.
Stimson, J. A. (1975). Belief systems: Constraint, complexity, and the 1972 election.
Stokes, D. E. (1966). Party loyalty and deviating elections. In A. Campbell, P. E.
Converse, W. E. Miller, & D. E. Stokes (Eds.), Elections and the political order (pp.
125–135). New York, NY: Wiley.
Sundquist, J. L. (1968). Politics and policy: The Eisenhower, Kennedy, and Johnson years.
Washington, DC: Brookings Institution.
Sundquist, J. L. (1983). Dynamics of the party system: Alignment and realignment of
political parties in the United States (Revised ed.). Washington, DC: Brookings
Institution.
Suskind, R. (2006). The one percent doctrine: Deep inside America’s pursuit of its enemies
since 9/11. New York, NY: Simon & Schuster.
Taber, C. S., & Lodge, M. (2006). Motivated skepticism in the evaluation of political
beliefs. American Journal of Political Science, 50, 755–769.
Thompson, D. F. (1970). The democratic citizen: Social science and democratic theory in
the twentieth century. New York, NY: Cambridge University Press.
United States Census Bureau. (2011). Table A-1. Years of school completed by people 25
years and over, by age and sex: Selected years 1940 to 2010 [XLS Data file]. Retrieved
from http://www.census.gov/hhes/socdemo/education/data/cps/historical/
index.html.
Vavreck, L., & Rivers, D. (2008). The 2008 cooperative congressional election study.
Journal of Elections, Public Opinion and Parties, 18, 355–366.
Verba, S., Schlozman, K. L., & Brady, H. E. (1995). Voice and equality: Civic volunta-
rism in American politics. Cambridge, MA: Harvard University Press.
Verplanken, B., & Holland, R. W. (2002). Motivated decision making: Effects of
activation and self-centrality of values on choices and behavior. Journal of
Personality and Social Psychology, 82, 434–447.
Visser, P. S., Bizer, G. Y., & Krosnick, J. A. (2006). Exploring the latent structure of
strength-related attitude attributes. In M. Zanna (Ed.), Advances in experimental
social psychology (Vol. 38, pp. 1–67). San Diego, CA: Academic Press.
von Bothmer, B. (2010). Framing the sixties: The use and abuse of a decade from Ronald
Reagan to George W. Bush. Amherst, MA: University of Massachusetts Press.
References 261
Washington Post. (2012, February 12). Transcript: Illinois senate candidate Barack
Obama. Retrieved from http://www.washingtonpost.com/wp-dyn/articles/
A19751-2004Jul27.html.
White, T. H. (1973). The making of the president, 1972. New York, NY: Atheneum
Publishers.
Wicker, A. W. (1969). Attitudes versus actions: The relationship of verbal and overt
behavioral responses to attitude objects. Journal of Social Issues, 25, 41–78.
Wilentz, S. (2008). The age of Reagan: A history, 1974–2008. New York, NY: Harper.
Wilson, T. D., & Hodges, S. D. (1992). Attitudes as temporary constructions. In
L. L. Martin & A. Tesser (Eds.), The construction of social judgments (pp. 37–65).
Hillsdale, NJ: Lawrence Erlbaum.
Woolley, J., & Peters, G. (2011, July 25). The American presidency project. Retrieved
from http://www.presidency.ucsb.edu/.
Wuthnow, R. (1988). The restructuring of American religion: Society and faith since
World War II. Princeton, NJ: Princeton University Press.
Zaller, J. R. (1992). The nature and origins of mass opinion. New York, NY: Cambridge
University Press.
Zanna, M. P., & Rempel, J. K. (1988). Attitudes: A new look at an old concept. In
D. Bar-Tal & A. Kruglanski (Eds.), The social psychology of knowledge (pp. 315–334).
Zumbo, B., Gadermann, A. M., & Zeisser, C. (2007). Ordinal versions of coefficients
alpha and theta for Likert rating scales. Journal of Modern Applied Statistical
Methods, 6, 21–29.
INDEX
ABM. See Anti-Ballistic Missile treaties attitudes. See also liberal-conservative

abortion, 25–26 attitudes; policy attitudes; political
culture wars and, 131 attitudes
rights, 82 availability, 65–66, 89–92, 109–10
Supreme Court’s restrictions on, 26 central, 8, 70–71
traditional morality and, 131–32, centrality, 67–71, 124–28
143–51 class of, 54
traditional morality influencing defining, 7, 43–44, 61
preferences in, 85, 131–32 dispositional model of, 43
abortion opinion issue, 6, 54, 131
party identification inversely related on limited government, 55, 113
to, 144 on military strength, 56
traditional morality and models of, as object evaluation associations, 44
131–32, 145t–148t, 149–50 peripheral, 68
traditional morality predicting, 150 policy principles and, 11, 45–46,
traditional morality with interactive 86–87
effects and, 149f–150t stability of, 11
Abramowitz, A. I., 82 structures of, 8
Achen, C. H., 79–80 on traditional morality, 55–56
achievement, 165, 172, 173 traditional morality measurement
Aid to Families with Dependent of, 96
Children, 19 voting behavior influenced by, 67
Aldrich, J. H., 240 attitude stability, 110
Ali, Muhammad, 33 corrected and uncorrected estimates
Allport, G. W., 43, 67 of, 115–16
American Psychiatric Association examining, 110–11
(APA), 26 sophistication-based differences and,
The American Voter, 111 118
criticism of, 48–49 authoritarianism, 150, 154
voting behavior noted in, 3, 64
Ansolabehere, S., 82, 118, 240 belief systems, 236
Anti-Ballistic Missile (ABM) treaties, 34 Bennett, S. E., 51, 199
APA. See American Psychiatric Berent, M. K., 80
Association Big Five personal traits, 168–69
263
264 INDEX
bin Laden, Osama, 37 CFA. See confirmatory factor analysis

black and white model, 79 CFI. See comparative fit index
born again Christian, 144–45, 151, 154 Chaiken, S., 7, 11, 44–45, 61, 119
Brezhnev, Leonid, 34 Chittick, William, 170n8
Brody, Richard, 58 Christian, born again, 144–45, 151, 154
Brooks, David, 2 Christian Coalition, 28
Brown, Helen Gurley, 24 citizens
Brown, Pat, 29 core policy principles choices of, 3
Buchanan, Patrick, 30, 82, 199, 202 democratic, 59
Bush, George H. W., 19, 36 federal government and economic
Clinton, B., charges of economic stress of, 56–57
problems of, 36–37 as ideologically indifferent, 3
Democratic Party chastised by, 202–3 as ideologues, 49–50
presidential campaign with, 209 with liberal-conservative attitudes,
Bush, George W., 1, 22 159n1
federal budget surplus position of, liberal-conservative attitudes of, 75
78 along liberal-conservative
traditional moralists supporting, 218 continuum, 2–3
war on terror of, 37–38 liberal-conservative continuum and
Bush doctrine, 37 thinking of, 52
liberal-conservative voting of, 6–7
CAFTA. See Central America Free liberalism and conservatism
Trade Agreement meaning and, 8
Campbell, A., 46, 48, 50n1, 64–65, 111 party schema inferences of, 85–86
candidate choice policy information exposure of,
central attitudes influencing, 71 83–84
centrality facilitating, 69 policy principle attitudes possessed
policy attitudes shaping, 119 by, 11, 86–87
policy principles guiding, 7, 9, 13, policy principles exposure of, 83–84
127–28 policy principles guiding, 155
position matching and, 71–72 policy principles influencing belief
social networks providing systems of, 236
information for, 202 policy principles long-term forces
sophistication interaction model of, 120
and, 63 political attitudes of, 67–68
candidate evaluation models, 77 political choices of, 192
Carnegie, Dale, 25 political evaluations of, 47–48
Carter, Jimmy, 35 political knowledge of, 59n6, 108,
CCC. See Civilian Conservation Corps 241
Central America Free Trade Agreement politically aware, 109
(CAFTA), 66 political parties and, 204
central attitudes, 8, 70–71 presidential selections of, 208, 230
central heuristics, 75 racial equality arguments of, 40–41
centrality voting behavior and ignorance of, 4
attitudes, 67–71, 124–28 Civilian Conservation Corps (CCC),
evaluating, 236 17–18
political sophistication and, 125–28 civil rights, 40
Index 265
Civil Rights Act, 19, 27, 39 core principles, 118, 156–57, 159
Clinton, Bill, 1, 19, 212 in policy voting, 235
abortion rights defense of, 82 sophistication interaction model
Bush, G. H. W., economic problems measuring, 126–27
charges from, 36–37 voter choice shaped by, 237
centrist position of, 20 correlation
Dole running against, 214 continuity, 112–17
military force used by, 37 corrected continuity, 114t
presidential campaign with, 209 knowledge, 76n6
Clinton, Hillary, 182 for limited government, 116–17
cognition policy principles, 127t
distinct groups with, 91 uncorrected, 109–10
-driven reasoning, 125–26 crime, fear of, 27
efficiency in, 57 Cronkite, Walter, 33
motivated social, 74 cultural liberalism, 29–30
networks with, 94 culture wars, 235–36
Colbert, Stephen, 1 abortion and, 131
Cold War, 31 over traditional morality, 27–28
Collins, Robert, 30 Supreme Court and, 24
comparative fit index (CFI), 100, 129
conditional slope estimates, 142–43 decision-making
confirmatory factor analysis (CFA), 91, heuristics, 124
98n4 limited government guiding, 216
conformity, 172, 173, 182, 191f strategies, 228–29
conscientiousness, 169 defense spending
conservation values, 165 military strength and, 151–56,
conservatism, 8, 51, 240 152t–154t
conservative doctrine, 47 Republican Party favoring, 151–54
conservative sophistication, 217 Delli Carpini, M. X., 10, 58, 76, 94, 243
constraint, 68 democratic citizenship, 59
Converse defining, 68 Democratic Party
horizontal, 68 Bush, G. H. W., chastising, 202–3
models of, 124–25 domestic policy liberalism
resource, 165n4 championed by, 19–20
vertical, 68 federal governments power and, 23
containment policy, 31–32 government capacities enlarged by,
continuity correlation, 112–17 18
conventional morality, 178 marginalized groups represented
Converse, P. E., 64, 78–79 by, 188
attitude-nonattitude continuum military power consensus collapsing
from, 44 in, 33
constraint defined by, 68 moral standards position of, 30–31
ideological labels examination of, 51 direct maximum likelihood, 105
political sophistication established dispositional model of attitudes, 43
by, 58 divorce rates, 26
voter competence reflections of, 3 Dobson, James, 28
core political values, 166–68 Dole, Bob, 214
266 INDEX
domain-specific principles, 7–8, 12–13 liberal philosophy harnessing power

domestic policy liberalism, 19–20 of, 46–47
don’t know responses, 93n2, 118 limited government and spending
Downs, A., 5–6 of, 94–95, 170–76
Dukakis, Michael, 19, 156, 206–8, 229 personal achievement and, 173–74
dynamic constraint private interests restrained by, 17
models, 133 public debate on role of, 22n3
tests, 128–32 The Federalists Papers, 241
feeling thermometer, 111n8
Eagly, A. H., 7, 11, 44–45, 61 Feldman, S., 80, 240
Economic Recovery Act, 21 The Feminine Mystique (Brown), 25
economic security, 5, 15 Fiorina, M. P., 202
economic stress, 56–57 fireside chats, 18
economic welfare, 192 fiscal austerity, 17–18
attention given to, 18–19 fiscal conservative, 47
federal government and, 16 flawed questions, 80–81, 110, 114
political sophistication reliance on, Foley, Tom, 60
225–26 Fonda, Jane, 33
politics of, 22 Ford, Gerald, 35
reform, 42 foreign policy, 204
economy, global, 17–18 nation building in, 39n6
education Nixon’s issue of, 34
knowledge correlation with, 76n6 peace and stability from, 198
political knowledge measuring, U.S. crises in, 37
172–73 free enterprise, 199
Eisenhower, Dwight, 20, 48 Friedan, Betty, 25
Eisenstadt v. Baird, 24 Fulbright, William, 33
electoral behavior research, 92n1
elite polarization, 238 Gilens, M., 171
Emergency Banking Act, 17 global economy, 17–18
Enelow, J. M., 47 Goldwater, Barry, 21, 29, 32
Engel v. Vitale, 24 Gonzales v. Carhart, 26
error variance, 80 Gorbachev, Mikhail, 35
evolution, 24 Gore, Al, 78, 216–19
Goren, P., 6, 57, 77, 111n8, 112, 126, 127,
factual knowledge, 58, 93 129n1, 160, 161, 164, 172, 178, 182,
Fair Deal, 18, 22 198, 243
Falwell, Jerry, 28 government. See also federal
federal budget surplus, 78 government; limited government
Federal Emergency Relief battles over size of, 239
Administration, 17 Democrats enlarging capacities of,
federal government 18
citizens coping with economic stress Reagan stating problem is, 21
and, 56–57 Great Depression, 16
Democratic Party marshaling power Great Society, 19, 32
of, 23 Green, D., 111
economic welfare and, 16 Griswold v. Connecticut, 24
Index 267
Halberstam, D., 23n4, 25, 31n5, 39n6, innocence of, 168n6

40 political knowledge and awareness
hawk-dove continuum, 42, 154, 183, of, 233–34
205 political parties polarization by, 49
hawk-dove dimensions predispositions to, 63
Hurwitz and Peffley findings on, 85, sophistication enhancing voting
110 with, 11
of political parties, 57 sophistication stimulating
presidential candidates and, 213 evaluation of, 135–36
health care, 19 thinking about, 48
limited government and, 128–30, voting behavior and, 10
130t voting with, 7
media coverage of, 129n1 INF. See Intermediate-range Nuclear
Obama’s reform of, 20 Forces Treaty
private insurance and, 143 Ingraham, Laura, 1
health care opinion Intermediate-range Nuclear Forces
limited government and, 136t, (INF) Treaty, 36
138t–139t, 140–42 international conditions, 216–17
seven-point scale tapping, 128–29 Iraq war, 37–38
Hetherington, M. J., 134 issue preferences
heuristics, 69–70 issue voting and, 82n8
hierarchical structures, 73 liberalism and conservatism in, 240
Himmelstein, Jerome, 28 policy principles predicting, 12,
Hinich, M. J., 47 156–57
Hoover, Herbert, 16, 20 principles relationship with, 133–34
horizontal constraint, 68 variables in, 224–25
humanitarian intervention, 39n6 voter choice and, 78–82
Humphrey, Hubert, 32 issues
Hurwitz, J., 12, 123, 236, 240 attitude, 6, 54, 131
hawk-dove findings of, 85, 110 issue preferences and voting on,
military power beliefs findings of, 82n8
125 policy preferences according to, 52
policy principle findings of, 131 racial, 237
Hussein, Saddam, 36, 37, 57 specifications test, 226t–227t
voter choice factoring of, 234
ideological labels voting behavior and, 10
Converse examining, 51
liberal-conservative attitudes Johnson, Lyndon, 19, 39, 49
developed from, 73–74 Johnson, Virginia, 24
presidential candidates evaluated Jost, John, 74
by, 199
voter choice and, 199 Kam, C. D., 171
ideological voter model, 13, 198–200, Keeter, S., 10, 58, 76, 94, 243
203 Kennedy, John, F., 18, 32
ideologues, 49–50 Kennedy, Robert, 33
ideology Kerlinger, F. N., 46
indifferent to, 3 Kerry, John, 219
268 INDEX
Key, V. O., 241 Downs describing, 5–6

Kinder, D. R., 3, 167, 171 policy principles correlation with,
King, Martin Luther, Jr., 33 127t
Kinsey, Alfred, 24 political elites division
Kissinger, Henry, 34 representation of, 46
Klingemann, Hans-Dieter, 49 political parties polarized on, 49
KN. See Knowledge Networks presidential candidates placed
Knight, K., 77 along, 48
knowledge. See also political sophistication and, 127t, 134
knowledge value dispositions in, 126
associational, 58 liberalism, 8, 51, 53, 240
education correlation with, 76n6 liberal philosophy, 46–47
factual, 58, 93 Limbaugh, Rush, 1
liberal-conservative attitudes voting limited government
of, 6–7 achievement and power values
structures, 58n5 related to, 165
Knowledge Networks (KN), 170, 170n7 additive scales measuring, 128
Krosnick, J. A., 80, 90 attitudes on, 55, 113
Krugman, Paul, 2 economic security and, 5, 15
Kuklinski, J. H., 167 federal spending and, 94–95, 170–76
health care model and, 128–30, 130t
Landon, Alf, 20 health care opinion and, 136t,
latchkey children, 26 138t–139t, 140–42
law and order issue, 27 mass belief systems centrality of, 143
Levitin, T. E., 53 origins of, 175t–176t
Lewis-Beck, M. S., 49, 50n1, 198–99 personal values and, 176, 176f
liberal-conservative attitudes political belief system with, 12
central heuristics of, 75 Republican Party featuring, 20–22
citizens with, 159n1 stability result for, 115, 129–30
defining, 45–46, 53 in 2008 presidential election, 223f
ideological labels developing, 73–74 universalism related to, 175
knowledgeable citizen’s voting with, value dimensions of, 185
6–7 voter choice influenced by, 214–16,
meanings attached to, 198–99 239
measuring, 205 voter decision-making guided by,
in 1988 presidential election, 228f 216
organized preferences of, 53n2 linear regression, 173, 183
political principles captured in, 5–6 Lodge, M., 80
political sophistication and voting long-term memory, 64, 72
from, 77–78 Luskin, R. C., 10
in sophistication interaction model,
77 MacArthur, Douglas, 32
sophistication levels and, 155 Markus, G. B., 79n7
liberal-conservative continuum mass belief systems
citizens falling along, 2–3 domain-specific principles in, 7–8
citizens not thinking about politics limited government’s centrality in, 143
in terms of, 52 policy principles in, 8, 10
Index 269
Masters, William, 24 nonresponse rates in, 117–18

McCain, John, 38, 234 policy principle measures in, 238
McCann, J. A., 85, 123, 125, 131, 167 National Liberation Front (NLF),
McCarthy, Eugene, 33 32, 57
McClosky, H., 46 National Organization for Women
McGovern, George, 29, 199–200 (NOW), 25
measurement model method, 91–92 National Race and Politics Survey, 167
media coverage, 129n1, 201 national security, 183
military from military strength, 15, 35, 38,
buildup, 35 96–97, 155
force, 37 Republican Party for, 203
power, 33, 125 nation building, 39n6
military strength, 33, 181–86 The Nature and Origins of Mass Opinion
attitudes on, 56 (Zaller), 58
defense spending and, 151–56, near ideologues, 50n1
152t–154t negative stereotypes, 171
defining, 5 NES. See National Election Study
national security from, 15, 35, 38, New Deal liberal, 47
96–97, 155 New Deal philosophy, 17, 18
origins of, 184t Nixon, Richard, 20
personal values and, 185f aggressive military tactics from, 34
political belief system with, 12 foreign policy issue confronting, 34
political divisions on, 31–32 social services expansion of, 21
politicians committed to NLF. See National Liberation Front
maintaining, 32 NO. See no opinion filters
presidential candidates on, 212–13 nonattitudes thesis, 79
Republican Party backing, 36, 183 nonresponse, 117–18
universalism depressing support no opinion (NO) filters, 90–91
for, 183 Norpoth, H., 80
military tactics, 34 NOW. See National Organization for
Mill, John Stuart, 241 Women
Miller, W. E., 53, 64, 73, 200, 240
models of constraint, 124–25 Obama, Barack, 1, 20, 221, 229
Mondale, Walter, 19 Olbermann, Keith, 2
Moral Majority, 28 opinion holder, 119t
moral standards, 30–31, 155 opinion holding method, 89–91
motivated social cognition, 74 O’Reilly, Bill, 2
motivational domains, 163–64, 163f organized preferences, 53n2
multidimensional models, 97–103
multistage area probability design, panel data, 109–10
135n4 Paris Peace Accords, 34
partisanship, 196
National Election Study (NES), 12, identities, 187–88
74–75 perceptions, 160–61
cross-sectional surveys of, 195–96 personal values relationship with,
electoral behavior research of, 92n1 188
interviewers experience of, 93–94 self-placements, 190
270 INDEX
partisanship (cont.) personality traits different than,

universalism influencing 168–69
attachments of, 191 policy principles impacted by, 166t,
voter choice bias and, 196 191–92
voters with, 196–97 policy principles informed by, 164
partisan voter model, 13 political values compared to, 167–68
party identification traditional morality and, 180f
abortion opinion inversely related personal values portrait (PVP), 177,
to, 144 181
conformity/tradition and, 191f Plato, 241
continuity correlation in, 112 policy attitudes, 45–54
correlations for, 116 candidate choice shaped by, 119
domain-specific principles evaluated classes of, 238, 240–41
by, 12–13 hierarchical structures of, 73
influence of, 160–61 sophistication’s role in, 76, 119–20
measuring, 204–5 policy cleavages, 3–4, 15–16, 42
origins of, 189t policy principles
personal values relationship with, actions to be pursued of, 6
159, 186–91 American voter and, 15
policy principles assessed with, 112, attitude class as, 54
185n10 attitudes on, 11, 45–46, 86–87
political party attachments in, candidate choice guided by, 7, 9, 13,
110–12 127–28
social group identities in, 197 citizen’s exposure to, 83–84
universalism and, 190f citizens learning about, 86–87
party schema, 85–86 as decision-making heuristics, 124
party stereotypes, 209 defining, 4–5
Patient Protection and Affordable Care Hurwitz and Peffley findings on, 131
Act, 20 issue preferences predicted by, 12,
Pearl Harbor, 31 156–57
Peffley, M., 12, 123, 236, 240 liberal-conservative continuum
hawk-dove findings of, 85, 110 correlations with, 127t
military power findings of, 125 in mass belief systems, 8, 10
policy principle findings of, 131 measuring, 94–97, 205
peripheral attitudes, 68 modeling of, 170–91
Perot, H. Ross, 209, 212 NES measures of, 238
Persian Gulf War, 36 in 1988 presidential election, 206–8
personality traits, 168–69 in 1996 presidential election, 216f
personal values personal beliefs influencing, 13
core political values different from, personal values impacting, 164, 166t,
166–68 191–92
limited government and, 176, 176f presidential candidates and, 212–13
military strength and, 185f presidential election and, 213f, 228,
partisan loyalties relationship with, 236–37
188 presidential vote and, 213f
party identification relationship presidential vote driven by, 8–9
with, 159, 186–91 public opinions on, 120f–121f
Index 271
in 2000 presidential election, 218f cognition-driven reasoning

in 2004 presidential election, 221t interaction with, 125–26
values and, 6n2 democratic citizenship and, 59
voter choice and, 82–87, 208f, 218–19, economic welfare reliance of, 225–26
223–24, 230–31 liberal-conservative voting from,
voting behavior and, 15 77–78
policy voting measuring, 92–94
attitude centrality in, 67–71, 124–28 vertical constraint moderated by, 166
conditions required for, 12, 64–65, 236 Pomper, G. M., 48
core principles in, 235 Popkin, S. L., 240
party stereotypes and, 209 position matching
position matching and, 76, 81–82 assessment of, 238
requirements for, 72–73 candidate choice and, 71–72
sophistication moderating, 213–14 policy voting and, 76, 81–82
voter competence and, 242–43 poverty, 19
voting behavior and, 9, 83 power values, 165, 172, 173
political attitudes presidential campaign, 201, 209
of citizens, 67–68 presidential candidates. See also
four-step process of, 65 candidate choice; voter choice
hierarchical structure of, 69f hawk-dove dimensions and, 213
measurement procedures of, 89–91 ideological labels evaluating, 199
personal values antecedent to, liberal-conservative continuum and,
161–62 48
political heuristics, 70 policy principles and, 212–13
political knowledge sophistication interaction model
of citizens, 59n6, 108, 241 choices of, 235
cognitive networks of, 94 on traditional morality, 212
distribution of, 60f–61f presidential elections
education measure of, 172–73 citizen’s choices in, 208, 230
ideological awareness and, 233–34 policy principles and, 213f, 228, 236–37
scales measuring, 92 policy principles driving, 8–9
U.S. distribution of, 60 presidential election, 1988
vertical constraint enhanced by, 155 liberal-conservative attitudes in, 228f
political parties. See also Democratic policy principles in, 206–8
Party; Republican Party voting in, 207t
citizens identifying with, 204 presidential election, 1992, 209–14,
hawk-dove dimensions of, 57 210t–211t
ideological polarization of, 49 presidential election, 1996, 214–16,
military strength divisions of, 31–32 215t, 216f
party identification attachment to, presidential election, 2000, 49–50,
110–12 216–19, 217t, 218f
policy cleavages dividing, 3–4, 42 presidential election, 2004, 219–20,
public opinion shaped by, 161 220t, 221t
traditional morality dividing, 31 presidential election, 2008, 221–24, 234
voting behavior influenced by, 201–2 liberal-conservative attitudes in, 229f
political sophistication limited government and, 223f
centrality and, 125–28 voting of, 222t
272 INDEX
principle-based voting, 209, 219–20 response instability, 80–81

principled voter, 200–204 retrospective voter model, 13, 197–98,
principles, 133–34 203
private insurance, 143 RMSEA. See root mean square error of
private interests, 17 approximation
progressives, 55–56 Robertson, Pat, 28
Protestant evangelical community, 28 Roe v. Wade, 25
pseudo-opinions, 90 Rokeach, M., 161–62
public affairs, 82, 84 Roosevelt, Franklin, 8, 17, 39
public debate, 22n3 root mean square error of
public opinion, 240 approximation (RMSEA), 100, 129
data of, 59
nonresponse in work on, 117 SALT I. See Strategic Arms Limitations
on policy principles, 120f–121f Talks
political parties shaping, 161 sample matching technique, 170n8
PVP. See personal values portrait Schlafly, Phyllis, 28–29
Schumpeter, Joseph, 241
Quayle, Dan, 60 Schwartz, S. H., 162, 172, 177, 181
Schwarz model, 163f
race scientific opinion polling, 241–42
conservatism on, 40 SDI. See Strategic Defense Initiative
equality of, 39–41 secular humanism, 28
issues on, 237 self-categorization, 111
Rayburn, Sam, 49 self-direction values, 164–65, 172, 173
Reagan, Ronald, 1, 29, 197 self-enhancement values, 164–65
government is problem statement self-transcendence values, 188
of, 21 SEM. See structural equation modeling
peacetime military buildup of, 35 September 11, 2001, 37
Religious Right support sought by, seven-point scales, 96n3
30 dependent variable of, 181
Reaganomics, 21 health care opinion tapped with,
Reasoning and Choice (Sniderman, 128–29
Brody and Tetlock), 58 nonresponse in, 117–18
The Reasoning Voter (Popkin), 240 Sex and the Single Girl (Brown), 25
Reconstruction Finance Corporation, 16 sex education, 24
Reifler, Jason, 170n8 sexual liberation, 25
religion, 24 Shanks, J. M., 73, 200, 240
Religious Right, 28, 30 Sniderman, Paul, 58, 70, 125–26
Republican Party social change, 29
as culturally conservative, 178 social groups, 6, 197
defense spending favored by, 151–54 social networks, 86, 202
limited government and lower taxes Social Security, 95
from, 20–22 Social Security Act, 18, 19
military strength backed by, 36, 183 social services, 21
national security from, 203 society
racial conservatism of, 40 progressiveness of, 239
welfare state hostility of, 171–72 traditional values promoted by, 5
Index 273
Vietnam war fomenting divisions political views in, 74

in, 33 presidential candidates choices in,
women’s role in, 25 235
sophistication, 58n4, 143. See also principle-based voting in, 219–20
political sophistication Soviet Union, 35
attitude stability and differences in, spending battery, 96n3
118 Spock, Benjamin, 33
candidate evaluation models and, 77 SRMR. See standardized root mean
corrected continuity correlations of, square residual
114t stability tests, 111, 117
dynamic constraint models and, 133 standard deviations, 109n7
factual knowledge attribute of, 93 standardized factor loading, 99
ideological evaluation stimulated by, standardized root mean square
135–36 residual (SRMR), 101, 129
ideological voting enhanced by, 11 Stevenson, Adlai, 48
liberal-conservative attitudes and Stewart, Jon, 1
levels of, 155 stimulus package, 234
liberal-conservative continuum and, Stokes, D. E., 64, 196
127t, 134 Stonewall Riot, 26
liberals and conservatives with, 217 Strategic Arms Limitations Talks
limited government and, 143 (SALT I), 34
opinion holder and, 119t Strategic Defense Initiative (SDI), 35
policy attitudes and role of, 76, structural equation modeling (SEM),
119–20 125
policy principles and levels of, 156 Supreme Court
policy principles measurement abortion restrictions from, 26
models of, 100t, 102t–105t cases of, 24
policy voting moderated by, 213–14 culture wars and, 24
political judgment and, 239–40 surveys
unidimensional measurement model cross-sectional, 195–96
of, 107t–108t don’t know responses in, 93n2
value-based reasoning promoted by, of economic welfare, 170–71
183 flawed questions in, 80–81
variable standard deviations and, KN administering, 170, 170n7
109n7
vertical constraint facilitated by, 141 taxes, 20–22
voter decision-making strategies Tea Party, 20
and, 228–29 television, 25
sophistication interaction model, 11, Tetlock, Philip, 58
132–56 Tet Offensive, 33
candidate choice and, 63 three-factor measurement model, 98f
core principle measures using, traditional family ties, 180n9
126–27 traditionalism, 150
liberal-conservative attitudes in, 77 traditional morality, 176–81
limited government and, 222 abortion and, 131–32, 143–51
policy voting in, 230 abortion opinion models and,
political awareness in, 208–9 131–32, 145t–148t, 149–50
274 INDEX
traditional morality (cont.) dimensions, 185

abortion opinion predicted by, 150 policy principles and, 6n2
abortion opinion with interactive political, 167
effects and, 149f–150t power, 165, 172, 173
abortion preferences influenced by, reasoning of, 172–73
85, 131–32 relations, 163f
attitude measurement of, 96 self-direction, 164–65, 172, 173
attitudes on, 55–56 self-enhancement, 164–65
Bush, G. W., supported by, 218 self-transcendence, 188
cultural conflict over, 27–28 structure with, 187
origins of, 179t system of, 162
personal values and, 180f traditional morality influenced by,
political belief system with, 12 165
political divisions over, 23–24 types of, 162–64, 162t
political parties divided on, 31 vertical constraint, 68
presidential candidates on, 212 political knowledge enhancing,
stability results for, 115 155
in 2008 presidential election, 223f political sophistication moderating,
values influencing, 165 166
traditional-progressive conflict, 96 sophistication facilitating, 141
Truman, Harry, 18, 31–32 Vietnam war, 33
Viguerie, Richard, 28
unidimensional alternative, voter choice
103–8 core principles shaping, 237
unidimensional measurement model, ideological labels and, 199
107t–108t information for, 229–30
United States (U.S.) issue preferences and, 78–82
foreign policy crises of, 37 issues factoring in, 234
military supremacy maintained in, limited government influencing,
36 214–16, 239
political conflict in, 242 partisan bias in, 196
political knowledge distribution policy principles and, 82–87, 208f,
in, 60 218–19, 223–24, 230–31
universalism, 164–65, 173 tests on, 206–29
limited government related to, 175 voter competence
measuring, 182 Converse reflecting on, 3
partisan attachments influenced by, policy voting and, 242–43
191 political choices and, 192
party identification and, 190f voting behavior
U.S. See United States The American Voter noting, 3, 64
attitudes influencing, 67
values. See also personal values campaign and media supplying
-based reasoning, 183 information influencing, 201
conformity, 172, 173 ideology and issues in, 10
conservation, 165 in presidential election, 219–20, 220t,
core political, 166–68 221t
defining, 161–62 in 1988 presidential election, 207t
Index 275
in 1992 presidential election, 209–14, Wallace, George, 29

210t–211t war on terror, 37–38
in 1996 presidential election, 214–16, welfare programs, 171
215t, 216f welfare state, 21, 171–72
in 2000 presidential election, 49–50, Westmoreland, William, 32
216–19, 217t, 218f Weyrich, Paul, 28
in 2004 presidential election, 219–20, What Americans Know about Politics
220t, 221t and Why It Matters (Carpini and
in 2008 presidential election, 221–24, Keeter), 58
222t, 223f, 229f, 234 White, Theodore, 29
policy principles and, 15 Willkie, Wendell, 20
policy voting and, 9, 83 Works Project Administration (WPA),
political parties influencing, 201–2 17–18
position matching and, 71–72 World War II, 31
principle-based, 209 WPA. See Works Project
public ignorance in, 4 Administration
sophistication and decision-making
strategies if, 228–29 YouGov/Polimetrix, 170, 170n8, 181
theories of, 203–4
Voting Rights Act, 19, 27, 39–40 Zaller, J., 46, 126, 240

Paul Goren - On Voter Competence

Uploaded by

Copyright:

Available Formats

Paul Goren - On Voter Competence

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Paul Goren - On Voter Competence

Uploaded by

Copyright:

Available Formats

On Voter Competence

Series in Political Psychology

Image Bite Politics: News and the Visual Framing of Elections

Social and Psychological Bases of Ideology and System Justiﬁcation

The Political Psychology of Democratic Citizenship

On Behalf of Others: The Psychology of Care in a Global World

The Obamas and a (Post) Racial America?

Ideology, Psychology, and Law

The Impacts of Lasting Occupation: Lessons from Israeli Society

Library of Congress Cataloging-in-Publication Data

1 The Indictment of the American Voter 1

2 Policy Cleavages in Historical Context 15

3 Policy Attitudes and Political Sophistication 43

4 The Three Faces of Policy Voting 63

5 The Availability of Policy Principles 89

6 The Centrality of Policy Principles 123

7 The Origins of Policy Principles 159

8 The Electoral Consequences of Policy Principles 195

9 The Exoneration of the American Voter? 233

Appendix: Measurement of Key Variables 245

On Voter Competence is my ﬁrst book. I must confess that the experience of

University of Minnesota and the International Studies Association. I thank Bill

The Indictment of the American Voter

U.S. presidential campaigns revolve around competing claims about the

Some individuals . . . know what they want their government to do and

THE IMPORTANCE OF POLICY PRINCIPLES

phrase “core principles” interchangeably. Limited government, traditional

placements would capture general political principles as well as preferences

PREVIEW OF THE THEORETICAL ARGUMENT

By refuting this perspective, On Voter Competence contributes to the study

PLAN OF THE BOOK

“political sophistication,” “political knowledge,” and “political awareness”

and basic human values shape evaluations of domain-speciﬁc principles. I test

Policy Cleavages in Historical Context

My examination of the American voter focuses on three policy principles. First,

THE LIMITED GOVERNMENT CLEAVAGE

[T]he federal government will assume bold leadership in distress relief.

That message resonated with the electorate as Roosevelt trounced Hoover

comforting ﬁreside chats, and the distribution of tangible beneﬁts, whether

as well. Congress backed the administration’s plan to ease chronic unemploy-

lockstep Republican opposition, and public ambivalence combined to sink the

by George W. Bush in 2000 and 2004–2005, uniﬁed Republican opposition to

THE TRADITIONAL MORALITY CLEAVAGE

With Nixon defending establishment values and McGovern ensconced in

THE MILITARY STRENGTH CLEAVAGE

this strategy carried considerable risk. Consequently, Nixon deployed a vari-

dedicated to the elimination of nuclear weapons, and Gorbachev, who saw

A DURABLE RACIAL CLEAVAGE?

damage to a party’s brand. However, public opinion on a number of osten-

Policy Attitudes and Political Sophistication

The ﬁrst step in developing my theory of policy voting is explaining what I

THE ATTITUDE CONCEPT

on the spot when situational demands compel perceivers to generate evalua-

terms “issue preferences” and “policy attitudes” interchangeably, but readers

the disadvantaged in particular, by reforming institutions that stiﬂe human

Information on policy issues is frequently scant . . . with candidates being

approach examines whether citizens employ liberal–conservative frames of

exclusively on this measure to estimate the prevalence of ideological thought

articulation of ideological terminology, and thus they may overlook the

I deﬁne liberal–conservative attitudes as psychological tendencies to eval-