Shipan MechanismsPolicyDiffusion 2008

The Mechanisms of Policy Diffusion
Author(s): Charles R. Shipan and Craig Volden

Source: American Journal of Political Science , Oct., 2008, Vol. 52, No. 4 (Oct., 2008), pp.
840-857
Published by: Midwest Political Science Association
Stable URL: https://www.jstor.org/stable/25193853
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Midwest Political Science Association is collaborating with JSTOR to digitize, preserve and
extend access to American Journal of Political Science
This content downloaded from

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
All use subject to https://about.jstor.org/terms
The Mechanisms of Policy Diffusion
Charles R. Shipan University of Michigan
Craig Volden The Ohio State University
Local policy adoptions provide an excellent opportunity to test among potential mechanisms of policy diffusion. By examining
three types of antismoking policy choices by the 675 largest U.S. cities between 1975 and 2000, we uncover robust patterns
of policy diffusion, yielding three key findings. First, we distinguish among and find evidence for four mechanisms of
policy diffusion: learning from earlier adopters, economic competition among proximate cities, imitation of larger cities,
and coercion by state governments. Second, we find a temporal component to these effects, with imitation being a more
short-lived diffusion process than the others. Third, we show that these mechanisms are conditional, with larger cities being
better able to learn from others, less fearful of economic spillovers, and less likely to rely on imitation.
The promise of state and local policymaking in a states and localities is desirable. Policy adoption based on
federal system is that these subnational govern learning about effective policies elsewhere leads to good
ments may serve as laboratories of democracy, outcomes, whereas the negative externalities arising from
where they experiment with different policies and learn competition can produce bad outcomes. Imitating other
from one another. The peril is that each government may governments by simply copying their policies may result
advance its own interests at the expense of others, leading in inappropriate policy choices. And policy choices based
to the possibility of destructive competition and coercion. on coercion by other governments are unlikely to be op
Scholars of federalism and intergovernmental relations timal. Thus, exploring the conditions under which each
often recognize these competing pressures, but rarely sep of these mechanisms drives policy diffusion is norma
arate one from the next. For example, in the sizable liter tively important, and, from a social scientific perspective,
ature on policy diffusion, authors frequently take note of is essential for a better understanding of the political in
the multiple mechanisms that lurk behind the spread of centives behind policy decisions.
policies across governments. However, because learning, In this article we demonstrate how scholars can dis
competition, imitation, coercion, and other mechanisms entangle these four mechanisms of policy diffusion. In
typically all point to an increased likelihood of policy particular, we establish that each of these diffusion mech
adoption when neighbors adopt the policy, scholars often anisms affected the spread of antismoking laws across
simply assess whether such a neighbor effect exists, with the 675 largest U.S. cities between 1975 and 2000. We
out concern for which mechanism is driving the result. then demonstrate the temporal and conditional nature of
Yet, from a normative perspective, uncovering the these diffusion mechanisms, illustrating the fleeting na
various mechanisms of policy diffusion is crucial to un ture of imitation and the varying susceptibility of cities of
derstanding when the devolution of policy control to different sizes to these various mechanisms. Ultimately,
Charles R. Shipan is professor of political science, University of Michigan, 505 S. State Street, 7764 Haven Hall, Ann Arbor, MI 48109-1045
([email protected]). Craig Volden is professor of political science, The Ohio State University, 2147 Derby Hall, 154 N. Oval Mall,
Columbus, OH 43210-1373 ([email protected]).
The authors would like to thank Jacob Nelson, Ken Moffett, Tracy Finlayson, and Chad Diefenderfer for valuable research assistance;
Ted Brader, Fred Boehmke, Rob Franzese, and Kurt Weyland for useful discussions and suggestions; and seminar participants at the
University of Arizona, Florida State University, Keio University, University of Illinois, University of Michigan, New York University, The
Ohio State University, Pennsylvania State University, Stanford University, and the Midwest Political Science Association meetings for
helpful comments. We also thank the Robert Wood Johnson Foundation for financial support and Jamie Chriqui for providing us with the
updated version of the National Cancer Institute's State Cancer Legislative Database. Local tobacco control ordinance data were provided
by the American Nonsmokers' Rights Foundation Local Tobacco Control Ordinance Database? and data on city-level demographics were
obtained from the Taubman Center for State and Local Government.
American Journal of Political Science, Vol. 52, No. 4, October 2008, Pp. 840-857
?2008, Midwest Political Science Association ISSN 0092-5853
840
189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
MECHANISMS OF POLICY DIFFUSION 841
we find that smaller governments face a relative disadvan Although these works have uncovered a great deal of
tage in policy formulation in the American federal system. evidence that policies do diffuse, much less is understood
They appear to be less capable of learning from the policy about the specific mechanisms that cause a policy to spread
choices of others, more susceptible to economic com from one government to another. That is, if a second gov
petition, more likely to engage in simple imitation, and ernment adopts a policy because a first government has
strongly at risk of coercion from their state government. already done so, what explains that second government's
To establish these findings, we proceed as follows. action? Here we focus on four mechanisms of diffusion:
First, we briefly survey the literature on policy diffusion, learning, economic competition, imitation, and coercion.
in order to place this study in context and to motivate While these mechanisms are also relevant to diffusion
our main theoretical arguments and testable hypotheses. across states and countries (e.g., Simmons, Dobbin, and
We then discuss the literature on the adoption of anti Garrett 2006), our focus on city-to-city diffusion allows
smoking laws, introduce our data, and detail how we test us to examine each mechanism individually as well as
our hypotheses. Finally, we highlight our results and their in conjunction with one another. Previous scholarship
importance. has often referred to multiple mechanisms of diffusion,
but with few exceptions (e.g., Berry and Baybeck 2005;
Boehmke and Witmer 2004; Weyland 2005, 2007) these
studies have not tested one explanation against another.
Local Policy Adoption and the Throughout this discussion, for the purpose of sim
Mechanisms of Diffusion plicity, we often write of the "city" taking action?
learning or competing, for example. In reality, individual
Policy innovation occurs whenever a government?a na decision makers?mayors, managers, council members,
tional legislature, a state agency, a city?adopts a new bureaucrats, and others?are the critical actors in these
policy (Mintrom 1997a; Walker 1969). The impetus for cities. Because of this large number of individuals and
this policy innovation can come from within the polity, forms of government, we rely on the shorthand of refer
such as when interest groups within a state push for the ring to cities as actors. As is common in the diffusion lit
adoption of a new policy, or when electoral and institu erature, we believe the individual decision makers within
tional forces within a legislature affect the likelihood of these cities are interested in adopting beneficial policies,
adoption. Pressure for policy innovation also can come either as a means to reelection or reappointment or as an
from outside the polity, with the spread of innovations end in themselves. Such motivations are at work across
from one government to another, a process known as all four mechanisms explored here.
policy diffusion. The first mechanism of diffusion that we explore?
The literature on policy diffusion is vast and expand learning?is the process that leads states to be called lab
ing rapidly. Building on a series of classic early studies oratories of democracy (Brandeis 1932). By observing the
(e.g., Crain 1966; Gray 1973; Walker 1969), as well as politics of policy adoption and the impact of those poli
more recent significant theoretical and methodological cies, policymakers can learn from the experiences of other
advances (e.g., Berry and Berry 1990; Berry and Baybeck governments. We follow most previous studies in adopt
2005), scholars have conducted a number of studies of dif ing a general definition of learning: as Berry and Baybeck
fusion during the past decade. These studies have focused note, for example, "[w]hen confronted with a problem,
on the diffusion of a range of policies, including same-sex decision makers simplify the task of finding a solution
marriage bans (Haider-Markel 2001), education reform by choosing an alternative that has proven successful
(Mintrom 1997a), abortion (Mooney and Lee 1995), the elsewhere" (2005, 505). Most generally, then, learning
death penalty (Mooney and Lee 1999), and HMO reforms involves a determination of whether a policy adopted
(Baila 2001), among many others. In addition, these and
other studies have shed light on the processes by which study of fluoridation, Knoke's (1982) analysis of the adoption of
diffusion takes place, focusing on factors that enable or municipal reforms, Godwin and Schroedel's (2000) investigation
hinder diffusion, including the policy's success (Volden of local gun control ordinances in California, and Martin's (2001)
examination of living wage laws. These studies, however, face a
2006), policy entrepreneurs (Baila 2001; Mintrom 1997a,
number of limitations, such as conducting tests only within one
1997b), and the initiative process (Boehmke 2005).l state (e.g., Godwin and Schroedel 2000), relying on bivariate rather
than multivariate analysis (e.g., Crain 1966), focusing solely on
internal determinants (e.g., Martin 2001), or looking at structural
1 Previous studies of diffusion in American politics have focused innovations rather than public policies (e.g., Knoke 1982). Ours is
overwhelmingly on state-to-state diffusion. Evidence of city-to-city thus the first large-N, multistate, quantitative study of the diffusion
diffusion does exist, of course, as demonstrated by Crain's (1966) of city-level policy adoptions.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
842 CHARLES R. SHIPANAND CRAIG VOLDEN
elsewhere has been successful. If the policy is deemed to Baybeck (2005) argue that learning can take place across
be successful, then a city is more likely to adopt it. states generally, while economic competition is typically
Ideally, political and policy success would be read confined to individuals living near state borders. Using
ily observable to decision makers and researchers alike. geographic information systems (GIS) technology, they
When success is difficult to measure (as it is at the city isolate the effects of learning and of competition to ex
level for antismoking policies), various shortcuts that are plain lottery adoptions and welfare benefit levels.
consistent with learning are taken. For instance, poli Like these studies and others, we contend that eco
cymakers may interpret the broad adoption of a policy nomic competition can lead to the diffusion of policies
without subsequent abandonment over time as evidence with economic spillovers across jurisdictions. State wel
of the success of the policy, or at least as evidence of main fare policy is a classic example. Fearful of becoming "wel
tained political support. Researchers, in turn, may explore fare magnets" (Peterson and Rom 1990), states may face
the effect of the "opportunity to learn" on policy choice incentives to engage in a "race to the bottom" in welfare
as a substitute for direct evidence of learning. Put sim benefits due to competitive federalism (e.g., Bailey and
ply, policymakers cannot learn about policies that have Rom 2004; Volden 2002). Such competition may also take
not yet been tried. They can learn more when multiple place at national or local levels of government, in policy
governments try the policy, and even more when such areas ranging from education and the environment to in
policies affect larger segments of society. The reliance of frastructure, minimum wages, and antismoking policies.
researchers on opportunity to learn is more appropriate In each instance, policymakers consider the economic
for policies that are eventually clearly identified as suc effects of adoption (or lack of adoption) by other govern
cesses, both politically (as evidenced by lack of repeals) ments. If there are negative economic spillovers, where
and on policy grounds (as evidenced by studies of ef the government will be hurt if it adopts a policy that its
fectiveness in general). The antismoking policies studied neighbors lack, then it will be less likely to adopt the policy
here meet both of these criteria.2 If they did not?for ex itself. On the other hand, if there are positive spillovers,
ample, if evidence of success were limited or not found? such as are found by establishing uniformity in infra
then it would be more difficult to discern exactly what structure, then governments will be more likely to adopt
cities learn from the experiments of others. For our case, the policy of others. Consistent with both theoretical and
this "opportunity to learn" idea is expressed as follows. empirical approaches to economic competition, we offer
the following hypothesis.
Learning Hypothesis: The likelihood of a city
adopting a policy increases when the same policy Economic Competition Hypothesis: The likeli
is adopted broadly by other cities throughout the hood of a city adopting a policy decreases when
state. there are negative economic spillovers from that
adoption to nearby cities and increases with pos
A second mechanism?economic competition?is itive spillovers from nearby cities.
often raised in conjunction with learning, and these two
mechanisms are viewed, at least implicitly, as the most
A third diffusion mechanism?imitation?has re
common processes explaining policy diffusion. Two re ceived much less attention in the state politics literature,
cent state politics studies have sought to disentangle these but arises more frequently in comparative politics (e.g.,
two mechanisms. Boehmke and Witmer (2004) explore Meseguer 2006; Simmons, Dobbin, and Garrett 2006)
state adoption of Indian gaming compacts, arguing that and has roots in social psychology and in studies of the
diffusion of innovations across a multitude of fields of
learning and economic competition are both important
in explaining initial adoptions, whereas only economic study (Rogers 1995). Sometimes also referred to as emu
competition explains subsequent compacts because pre lation, imitation involves copying the actions of another
in order to look like that other. The nature of imita
vious experience with one's own compacts removes the
need to learn from the experience of others. Berry and tion can be understood in contrast to learning. In learn
ing, policymakers focus on the policy itself?how was it
adopted, was it effective, what were its political conse
2 This assertion is well supported by the scarcity of antismoking law
repeals and by the public health literature, with numerous studies quences? In contrast, imitation involves a focus on the
showing both that antismoking measures can have positive effects, other government?what did that government do and
such as reductions in the rate of smoking (e.g., Evans, Farrelly, how can we appear to be the same? The crucial distinc
and Montgomery 1999; Ross and Chaloupka 2004), and can avoid
negative effects, such as the loss of business and restaurant income tion is that learning focuses on the action (i.e., the policy
(e.g., Glantz and Charlesworth 1999). being adopted by another government), while imitation

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
focuses on the actor (i.e., the other government that is minimize trade barriers and other coercive mechanisms
adopting the policy). Outside of the policy adoption con across the states.
text, a classic example of learning is avoiding touching Although horizontal coercion across states or local
the hot burner after observing someone doing so with ities in the American federal system is therefore limited,
bad effects, whereas imitation is jumping off the garage vertical (or top-down) coercion is still quite possible. This
roof after observing your older brother doing so, without should be of no surprise to scholars of policy diffusion,
regard for the consequences. In the former case, it is the who have long noted that grants from the federal gov
action that matters; in the latter, the actor. In the former, ernment to states and localities often stimulate policy
you learn about consequences; in the latter you simply adoptions (e.g., Allen, Pettus, and Haider-Markel 2004;
aspire to be like the other actor. Karch 2006; Shipan and Volden 2006; Walker 1973; Welch
Although imitation sometimes has been ignored or and Thompson 1980). For instance, the threat of lost
even mislabeled in the policy diffusion literature, it is highway funds coerced states into adopting lower speed
wholly consistent with early studies of local and state limits and higher drinking ages. In addition to their in
policy adoptions. This literature focused on which states fluence through intergovernmental grants, higher levels
and cities were "leaders" or "laggards" (e.g., Crain 1966, of government can coercively influence the actions of
Grupp and Richards 1975; Walker 1969). Innovative lead lower levels by taking the lead in that policy area, setting
ers were found to be larger, wealthier, and more cos their own minimum wage or antismoking restrictions, for
mopolitan. Smaller communities aspire to be like these example.3
leaders, and therefore adopt the same policies as these Even more coercive are preemptive policies. Because
leaders without necessarily thinking about the conse cities are creatures of the state, with no constitutionally
quences of such adoptions. Clearly, policymakers in these specified sovereignty, state governments can pass laws that
smaller cities also may learn from the policy experiences disallow any city action contrary to state law. In the area
of those in larger cities. And they also may worry about of antismoking policy, for example, such state-level pre
competition, leading them to adopt policies in an attempt emptive policies were commonplace and were an explicit
to stem the flight of citizens and businesses to these larger strategy of the tobacco industry to fight back increasingly
cities. But above and beyond learning and competition, stringent local laws (e.g., Givel and Glantz 2001). A local
decision makers in smaller cities also may adopt poli ity still might pass weaker laws that ensure the continu
cies simply because they want their communities to be ance of the policy if the state were to reverse its stance; or
as favorably viewed as the cities that are seen as lead it could enact alternative laws in order to provoke a court
ers. They hope that such imitation will raise their profile challenge; but, in either case, the usefulness and hence the
and make them more attractive places to live, like their likelihood of passage of such laws are greatly diminished.4
larger, wealthier, and more cosmopolitan neighbors. In
our context, therefore, imitation may appear as smaller Coercion Hypothesis: The likelihood of a city
cities copying the policies of their larger neighbors. adopting a policy decreases when the state adopts
a similar policy that covers the city. This decrease
Imitation Hypothesis: The likelihood of a city is even more substantial when the state law pre
adopting a policy increases when its nearest big empts either future local laws on the same policy
ger neighbor adopts the same policy. or future stronger laws.
The fourth mechanism of diffusion?coercion? We acknowledge that the theoretical distinctions

differs from the previous three. Like imitation, it is more among these four categories are starker than are the real
commonly raised in the comparative politics literature world empirical classifications of these diffusion mech
(e.g., Simmons, Dobbin, and Garrett 2006) than in Amer anisms. For example, when a neighboring city restricts
ican politics. In the international setting, for example, smoking in its restaurants, this provides an opportunity to
countries can coerce one another through trade prac learn, raises some economic spillover considerations, and
tices and economic sanctions. They can attempt to coerce
others directly, or can do so through international institu 3 See Volden (2005, 2007) for formal models of these processes and
tions like the United Nations and the International Mon their likely policy effects.
etary Fund, which encourage or pressure governments 4 Similarly, cities could pass weaker laws if the state preemptive
to take actions that meet common expectations. Coer clause allows them to do so (e.g., the state may restrict smoking
only under certain conditions, but allow cities to loosen the set of
cion was such a major concern to the founders of the U.S. restrictions). Even in such a situation, however, far fewer cities will
Constitution that they established the commerce clause to have the incentive to pass such laws.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
844 CHARLES R. SHIPAN AND CRAIG VOLDEN
may induce imitation. Separating these effects from one navigate through the public policy process in order to
another is difficult. Nevertheless, we believe that raising bring about a policy adoption, then all the relevant infor
these theoretical and archetypal mechanisms as distinct mation is revealed at the time of adoption. But if policy
from one another provides guidance for scholars to be makers are interested in knowing the political and policy
gin to disentangle these diffusion processes. Moreover, we consequences of an adoption, then it may take months or
argue that our focus on cities provides the variance neces years to evaluate the effectiveness of a particular policy.
sary to explore these distinct mechanisms. For instance, Regardless of what is being learned, the learning effect is
whereas learning can take place over quite a distance, unlikely to fade quickly?indeed, evidence of the effects
economic spillovers are geographically limited. Whereas of policies, once known, is likely to remain relevant to
one can learn from the experiences of smaller or larger policymakers for a considerable period of time. Second,
cities, imitation is focused only on the larger leader cities. economic competition also should exhibit long-term ef
And whereas larger cities may be imitated, both smaller fects. If governments are worried about the economic
and larger cities alike may present economic competitive spillovers from another government's policies, that com
concerns if sufficiently proximate. Thus, despite overlap petitive pressure will remain for as long as the policy is in
across diffusion mechanisms, we can distinguish among place.
them conceptually. Such conceptual distinctions guide Overall, then, we should expect to find different tem
the variable operationalizations we use to empirically test poral effects for imitation than for learning and competi
our hypotheses. tion. For imitation, we would expect a strong initial effect
that then should fade over time. For the other horizontal
mechanisms, we should expect both an initial effect and
an effect into the future (e.g., a city will continue to learn
The Temporal and Conditional Nature
from other cities two or three years after those other cities
of Policy Diffusion Mechanisms
have adopted a policy). Finally, for the coercion involved
Although a major goal of this article is to differentiate, in vertical diffusion, the temporal effects are less clear.
both theoretically and empirically, among multiple mech Preemptive laws may immediately influence local adop
anisms of diffusion, it is equally important to explore tions and those effects may persevere. Yet, localities may
when each of these mechanisms takes place and why one over time adopt laws testing whether the state restrictions
mechanism may affect some cities more than would an still have teeth. Exploratory work below examines these
other mechanism. Therefore we advance two additional alternatives, while our main temporal effects hypothe
hypotheses. sis spells out our predictions for the horizontal diffusion
First, we consider the temporal nature of each mecha mechanisms.
nism, which allows us to further distinguish among mech
anisms and also to gain additional perspective on whether Temporal Effects Hypothesis: The effects of imi
the operationalizations that we use are appropriate. Our tation are likely to be short-lived. Learning and
starting point here is the realization that some mech economic competition, on the other hand, are
anisms of diffusion should be short-lived, while others likely to exhibit longer-term effects.
should have longer lasting effects. In the former category,
we would expect to find imitation. When one city imitates Second, we consider the conditional nature of each
another, it does so fairly quickly, as policymakers in that mechanism. Some cities are better equipped to learn from
city imitate the actions of cities that are leaders and do so others; some are more susceptible to economic competi
in order to look like those leaders. Because imitation in tion than are others. Some are more likely to follow lead
volves no concern about the effects of policies, but rather ers; and some are more likely to resist coercive actions. Al
only a desire to do whatever a leader city has done, the though there maybe many criteria that divide cities along
response to a policy adoption should be almost immedi these lines, one straightforward and broadly relevant city
ate. If the response does not come quickly, it becomes less characteristic is simply the size of its population, which is
likely over time, as policymakers will decide whether to likely to matter for each of the diffusion mechanisms we
imitate an action or not and then will move on to other are exploring.5 Because larger cities tend to have bigger
ways to imitate the leader.
In contrast to imitation, the other two mechanisms 5 Other city characteristics, like wealth or government structure,
might likewise affect different diffusion mechanisms. However, our
of horizontal diffusion?learning and competition?
goal here is to demonstrate that these diffusion processes can be
should exhibit longer-term effects. First, consider learn conditional, rather than to fully document all possible conditional
ing. If policymakers are concerned only about how to relationships.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
and more professional governments, they are more capa law, the specific action taken in the law, and the adoption
ble of learning from others.6 Larger cities are less likely date. In addition, the SCLD dataset also identifies which
to be influenced by local economic spillovers, partly be state laws preempted future city-level laws. Thus, anti
cause of their economic diversity and partly because their smoking policies provide a useful and appropriate forum
smaller neighbors are less economically threatening. Be for testing our hypotheses.7
cause they are already the leaders that others look up to We focus on three types of antismoking policies in
(e.g., Crain 1966), larger cities also are less likely to en this article: restrictions on smoking in government build
gage in imitative behavior than are smaller cities. Finally, ings, restrictions on smoking in restaurants, and youth
larger cities are more likely than smaller cities to confront access restrictions. Our choice of these three policies re
the coercive power of the state. We present these ideas in flects several considerations. At one extreme, we could
the following Conditional Diffusion Hypotheses. simply look to see whether a city (or state) has adopted
any antismoking law, regardless of type. This, however,
Conditional Learning Hypothesis: Larger cities are would be a level of analysis that is far too aggregated
more likely to learn from other cities. and that ignores variation among types of policies. At the
other extreme, we could assign each law to a category
Conditional Competition Hypothesis: Larger cities based on detailed criteria contained within the law. The
are less susceptible to economic competition. problem with this approach is that the data would then
be far too disaggregated. Laws that place restrictions on
Conditional Imitation Hypothesis: Larger cities smoking in government buildings, for example, can con
are less likely to engage in imitation. tain outright prohibitions on smoking; they can restrict
smoking in common areas, or provide for specific areas
Conditional Coercion Hypothesis: Larger cities are where smoking is allowed, such as individual offices; they
less likely to be coerced effectively. can set limits on smoking near doorways, perhaps man
dating a nonsmoking perimeter near the entrance; and
so on. Because the list of specific modifications is nearly
endless, we have chosen to strike a middle ground, neither
Antismoking Policies at the State aggregating all laws together, nor disaggregating them by
and Local Levels their components, but rather grouping them into three
fairly broad, yet distinct, categories.
Testing these diffusion hypotheses requires, first, a pol Grouping laws into these three categories also has
icy area in which states and cities share jurisdiction, and the beneficial effect of capturing variations across mul
second, data on policy adoptions at both levels of gov tiple types of laws. More specifically, two of our policy
ernment. Antismoking policies meet both requirements. areas?government buildings and restaurants?are gen
States and cities both are active policymakers in this pol erally classified by public health scholars as clean indoor
icy area, passing a large volume of laws that regulate a air laws, since they tend to be spurred by concerns over the
wide range of activities (e.g., Schroeder 2004; Shipan and health effects suffered by nonsmokers who are exposed to
Volden 2006). In addition, we have been able to obtain secondhand smoke. The other policy area?youth access
comparable data on adoptions at both the city and state laws?is instead designed to make it more difficult for
levels of government. Specifically, data on state-level laws young people, especially teenagers, to obtain cigarettes.
come from the National Cancer Institute's State Cancer Policies in this area include regulations regarding the lo
Legislative Database (SCLD), maintained by the Maya cation of vending machines, fines for selling cigarettes
Tech Corporation. For city-level laws we use the American to minors, and restrictions on the sale of cigarettes out
Nonsmokers' Rights (ANR) Foundations Local Tobacco of their original packaging. To the extent that diffusion
Control Ordinance Database. Each of these comprehen mechanisms vary across these three areas, including all
sive databases contains extensive information about each three provides the greatest opportunity to uncover these
law passed by the state or city, including the topic of the distinct mechanisms.
Before turning to the data analysis, it is helpful to con
6Shipan and Volden (2006) similarly find that more professional sider the face validity of the hypotheses suggested above
state governments are better able to learn from local adoptions, in the context of antismoking policies. The public health
thereby facilitating local-to-state diffusion. More generally, Huber,
Shipan, and Pfahler (2001) and Huber and Shipan (2002) find that
more professional governments often are better able to engage in 7 For additional discussions of these dataseis, see Shipan and Volden
policy control. (2006) and Chriqui (2000).

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
literature has provided such an assessment through a va Shipan and Volden (2006), who in turn rely on a slight
riety of case study and single-state analyses, which sug modification of the modeling approach of Wei, Lin, and
gest initial support for our main hypotheses. Skeer et al. Weissfeld (1989), is appropriate because any of the three
(2004), for example, note that neighboring towns in Mas policies could be adopted at any time in any order. The
sachusetts, especially those with high levels of income and results are robust to either pooled or separate analyses.
education, were more likely to adopt similar restaurant re In our statistical tests, we first separately examine
strictions. Jacobson and Wasserman's (1997) case studies the hypotheses regarding the individual diffusion mecha
in seven states show a decrease in local adoption activi nisms. Our second set of results includes all four diffusion
ties following the enactment of state laws.8 And numerous mechanisms in the same model, illustrating the strength
studies point to the tobacco industry's state-level preemp of each effect upon controlling for the other three. The
tion strategy as a way to win battles that otherwise would third part of our analysis presents lagged versions of the
be lost in localities (e.g., Givel and Glantz 2001; Siegel et diffusion variables, to test the Temporal Effects Hypothe
al. 1997). Thus, public health studies provide initial evi sis. Finally, we test the Conditional Diffusion Hypotheses
dence consistent with our expectations. We turn now to by interacting the diffusion variables with city size to il
a systematic analysis of our hypotheses. lustrate which types of cities are more responsive to which
diffusion pressures.
Empirical Approach Mechanism Variables
Our dependent variables of interest capture whether a The Learning Hypothesis holds that a city will be more
city adopts a law in each of the three types of antismoking likely to adopt a policy if other cities in the state have
restrictions we examine. For each of our three categories already done so. To test this hypothesis, we constructed a
of antismoking laws, we construct a dependent variable variable, Proportion of State Population with Local Re
that is initially set equal to 0. In the year the city passes a striction, which is calculated by identifying the cities that
law, this variable is set equal to 1 ; and in following years the have each type of antismoking law at the beginning of
city's observations are removed from the dataset, as the the calendar year, summing up the populations of those
city is no longer at risk of a policy adoption. This allows us cities, and dividing by the overall population within the
to use a standard event history analysis (EHA) to predict state. Our expectation is that as this proportion increases,
the probability that an event will occur given that it has so will the likelihood that the city will adopt the same type
not already occurred. Our analyses include all 675 cities of law. Certainly there may be other ways to formulate a
in the United States with populations of 50,000 or greater variable to capture the possibility of learning.9 And our
as of the year 2000. We do not include smaller cities, for variable may partially capture other mechanisms of dif
which both independent and dependent variables are less fusion beyond learning, as it includes both nearby cities
available and may be less reliable. We examine the period (a possible source of competition) and larger cities (pos
between 1975 and 2000. We focus on this 25-year period sible targets of imitation). Nevertheless, because learn
due to data availability and because very few city-level ing can take place from a broad array of other cities, we
antismoking laws were passed prior to this time. rely on this broad, inclusive measure. Controlling for this
Because we are examining three policy choices by learning effect will demonstrate the residual part of dif
the same cities in the same years, these adoptions may be fusion still captured by our competition and imitation
considered as a type of repeated event (Box-Steffensmeier variables, detailed below. Likewise, controlling for these
and Zorn 2002). In our analyses, we therefore pool the other mechanisms helps uncover the residual diffusion
data together, yielding one observation per city per year effect due to learning.
To test for the second diffusion mechanism, as pre
per policy. This pooling, which follows the approach of
sented in the Economic Competition Hypothesis, we cre
8 On the other hand, Andersen, Begay, and Lawson (2003) illus ated a variable called Outflow. This variable is designed
trate the more classical pattern of positive vertical policy diffusion to capture the city's concern, central to this hypothesis,
by highlighting how state-level funding of local tobacco control
initiatives helped explain their adoption in Massachusetts. Conlisk
et al. (1995) present an interesting counterpoint to this wave of 9Looking instead at the number of cities with the given law yields
research. When North Carolina attached a three-month window largely similar results. We believe that weighting by city population,
to their preemption law, during which time localities were allowed however, better characterizes the opportunity to learn about policy
to adopt and grandfather their own laws, 89 new local regulations impacts, as more buildings, restaurants, and youths are affected by
were passed, compared to a total of 16 prior to that time. policy adoptions in larger cities.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
that it may lose out economically to surrounding cities adopted an antismoking law in the category of interest
if it adopts an antismoking measure?that is, benefits by the start of the year, then this variable is set equal to
may flow out from the city to other cities.10 Such fear is, 1. We expect this variable to have a positive influence on
of course, dependent both on whether the surrounding the dependent variable. This focus on larger neighbors
communities have the same kinds of antismoking laws is consistent with the concept of imitation, but it creates
themselves and on the relative size of those neighbors a coding problem for the largest city in each state. Since
compared to the city being examined. Therefore, to con there is no "nearest bigger city" for the largest city in a
struct this variable, we began by identifying all cities in state, for these cities we look to see whether the second
our dataset that were contiguous to, or within 10 miles of, largest city has adopted a policy. If, alternatively, we simply
the city in question.11 We then summed the population of drop these largest cities from our analysis, our results
those surrounding cities that did not have the antismoking remain substantively unchanged. It is also important to
policy?in other words, those cities to which economic note that we are not arguing that cities fail to learn from, or
benefits might flow if the city in question adopted a law? to compete with, their nearest bigger neighbors. Rather,
and created as our measure the ratio between the sum of here we are assessing whether an imitation effect exists
these populations and the population of the city in ques above and beyond the learning and competition effects
tion. For example, if a small town is next to a city that has controlled for by our other variables.
five times as many people as the town, and neither has The Coercion Hypothesis suggests that state adoption
restaurant restrictions, Outflow takes a value of 5.0 for the of a policy will decrease the likelihood that a city within
town and 0.2 for the city, as the town is more vulnerable that state will adopt a similar law and that preemptive
to the outflow of an average restaurant patron than is clauses in state laws will be an even greater deterrent.
the big city. If there are no cities within 10 miles, or if the To test this hypothesis, we created two variables. First,
surrounding cities all have passed laws, Outflow equals State Law is set equal to 1 in every year after the state
zero. Because fears of economic vulnerability should di has adopted a law and is otherwise set to 0. For example,
minish the likelihood of an antismoking policy adoption, in 1977 Iowa adopted a law restricting smoking in gov
we anticipate a negative coefficient on this variable. ernment buildings, so for the observations dealing with
To test the Imitation Hypothesis, we constructed a government buildings in cities in Iowa, State Law is equal
Nearest Bigger City variable by looking within the state toOin 1975,1976, and 1977, and takes a value of 1 starting
to see whether the nearest city with a larger population in 1978 and continuing through 2000. Second, State Pre
than the city in question has previously adopted the policy emption is constructed analogously to State Law, based
(i.e., restrictions for government buildings, restaurants, on whether the state law explicitly preempted local action
and youth access, respectively). If it has, the variable takes in a relevant category of antismoking restrictions.13 We
on a value of 1; and if it has not, this variable is set equal expect a negative coefficient on these two variables.
to 0.12 For example, the nearest city to Oakland that has
a larger population is San Francisco; if San Francisco has
10Economic competition is likely to affect clean indoor air adop

City, State, and Temporal Controls
tions more substantially than youth access restrictions, as policy
Although we are mainly interested in horizontal and ver
makers may heed warnings about lost restaurant patrons due to
restaurant restrictions and about lost business revenue due to gen tical diffusion patterns, we agree with Peterson's counsel
eral restrictions on public and private workplaces. Rerunning all that "[t]o ignore internal factors altogether would be as
models without including the youth access policy area affirms this
misleading as to treat urban politics and policymaking
expectation.
solely in terms of them" (1981, 4). Therefore, it is essen
11A 10-mile radius seems appropriate as a limit on the distance a
tial to control for city-level factors that may influence the
typical resident will travel to engage in smoking-related activities.
This is consistent with numerous studies finding that the maximum adoption of antismoking policies. In our particular case,
distance restaurant customers are willing to travel to dine is between without such controls we would be unable to discern
five and ten miles (e.g., Purlee 1995; Tanyeri 2007).
12To find the nearest city, we look only within state borders. Thus,
dicate that inclusion of such variables does not substantively change
the nearest bigger city for Toledo, Ohio, is Cleveland, rather than our results.
Detroit, Michigan, even though Detroit is closer. As this exam
ple indicates, our diffusion variables look within state boundaries. 13Offen these preemptive laws are more general than are the sub
Future work to determine whether city-level diffusion stops at state stantive state laws. For example, a state may pass a restriction on
lines would be most welcome, although it is beyond the scope of the smoking in public places coupled with the preemption of local laws
present analysis. Initial exploratory analyses of the effects of local for all clean indoor air policies, meaning that the localities are be
laws in neighboring states (e.g., by considering the proportion of ing coerced to not act on restaurant restrictions even though the
the neighboring state's population that is covered by local laws) in specific state law did not address restaurants at that point in time.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
whether neighboring cities are adopting similar policies the non-minority presence in the city.18 Relative to most
due to their similar political and economic conditions minorities, whites are less likely to smoke, leading us to
or due to policy diffusion (Franzese and Hays 2008). We expect a positive coefficient on this variable. Per Capita
therefore incorporate seven city-level control variables Government Spending is expressed in thousands of dol
that are designed to capture separate internal influences lars per resident and is an indicator of the liberalism or
on city-level policy adoptions. To begin with, City Popu activism of the local government. We therefore anticipate
lation is simply the city's population (scaled in 100,000s a positive coefficient on this variable.19
of city residents). Larger cities have greater capabilities We also need to account for several factors at the state
to pass laws and are anticipated to be early leaders, so level, in part to control for their overall effects within the
we expect this variable to have a positive effect on the state, and in part because good measures of these features
likelihood that the city will adopt a policy. do not exist at the urban level. Percent Smokers captures
The form of government also may affect the likeli the statewide percent of adults who smoke. Ideally, this
hood of a policy adoption or innovation. Mayor-Council measure would be available at the city level. However,
is a dummy variable that captures whether the city has since such a measure does not exist for many cities in many
a mayor-council city governance structure. As Knoke years, it is necessary to control for statewide smoking rates
(1982) notes, many cities have adopted forms of gov as a proxy for local rates. Tobacco Production is a dummy
ernment other than a combination of a mayor and a variable, taking a value of 1 for cities in tobacco-producing
city council, expecting these other forms of government states. We anticipate negative coefficients on both of these
(e.g., commissions, council-managers) to possess more variables.
expertise and to be more effective and efficient at passing Other state-level measures control for governmental
and implementing legislation. Thus, we anticipate Mayor effects. First, we include two measures of interest group
Councilto have a negative influence on the probability that influence, Tobacco Lobbyists and Health Organization
a city will adopt an antismoking law.14 Lobbyists, where the former is the ratio of tobacco lob
Percent Health Employees captures the percentage byists registered at the state level to the total number of
of employed residents in the city who work in health lobbyists registered at the state level, and the latter is a
service professions.15 To the extent that this variable indi similar ratio that uses health organization lobbyists in the
cates the presence of health advocates or a predisposition numerator instead of tobacco lobbyists.20 Our expecta
toward more healthful policies, we would expect a posi tions here are conflicted. On the one hand, a city in a
tive coefficient. Percent High School Graduates captures strong tobacco state will be less likely to adopt an anti
the percent of the city population over age 25 with high smoking law, while a city in a strong pro-public health
school diplomas or equivalencies.16 More educated popu state will be more likely to do so, due to interest group
lations are less likely to smoke, more likely to be concerned pressures extending to the local level. On the other hand,
with health risks, and more liberal; thus we expect them the inability of state officials to act may spur local action.
to be more likely to favor the adoption of antismoking Second, we also include a variable that measures ideology
policies. Per Capita Income is the average income per at the state level, relying on updated versions of Berry and
resident in thousands of dollars.17 We anticipate a posi Colleagues (1998) measure of State Government Ideol
tive coefficient on this variable. Percent White captures ogy, in which higher values represent more liberal views.
More conservative states will be less likely to adopt anti
smoking laws, but this hesitancy may result in local, rather
14 Such findings would be consistent with Moon (2002), who uncov than state, adoptions.
ers positive effects of city size and council-manager governments in
facilitating municipal website adoptions. Interestingly, Frederick
Finally, we include dummy variables for the years
son, lohnson, and Wood (2004) show how these city government in our analysis. Inclusion of these dummies allows us to
types have emulated one another's features, becoming more similar
and homogeneous over time.
18 Further breakdowns into other racial and ethnic categories
15 Including instead the proportion of city spending dedicated to showed some additional variation across groups, but did not change
health did not affect the diffusion relationships discussed below. the support for the hypotheses of interest.
16 Including instead other education variables, such as the current 19The main findings are robust to other city-level controls, such
high school dropout percent or current college enrollment, showed as the population density, percent female, percent of population
weaker effects of education, but did not change the uncovered employed by the government, or average number of vehicles per
diffusion relationships. household, most of which were not themselves significantly related
to antismoking policy adoptions.
17 Similar results followed for the inclusion of household income,
family income, or a variety of poverty and unemployment mea 20These measures are based on the 1994 snapshot for each state,
sures, none of which affected the uncovered diffusion relationships. constructed by Goldstein and Bearman (1996).

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
look for patterns over time and to control for temporal State are covered by a similar law. Based on Model 1 re
dependence (Beck, Katz, and Tucker 1998). To save space sults, a 10% increase in the state population covered by a
in the tables, we do not report the coefficients for these similar antismoking restriction is associated with a 49%
year dummies, but note that they follow a general pattern: boost in the odds of such an adoption in the city we are
they tend to be negative in the early years of our series, focused on.
positive in the middle, and negative toward the end, and The results of Model 2 show that the second diffusion
are often (although not always) significant. Such a pattern mechanism, outlined in the Economic Competition Hy
is typical for S-shaped policy diffusions (Gray 1973), with pothesis, is also at work here. As indicated by the negative
a few leaders, a few laggards, and many adopters in the coefficient on the Outflow variable, cities are hesitant to
middle. Our results are not substantively altered if we use adopt antismoking laws when neighboring cities within
year and year-squared variables, rather than yearly dum 10 miles do not yet have such laws.23 Because this variable
mies. All variables and their descriptions are summarized is constructed based on city proximity and relative popu
in the appendix. lation sizes, its effect is somewhat difficult to characterize
in a general sense, but examples can indicate the magni
Results tude of the effect. For instance, compared to a city with
no near neighbors, if a city has three equally sized prox
As described above, we pool our observations across the imate neighbors that do not have the given antismoking
three types of antismoking policies. We test our hypothe policy, its odds of adopting the smoking restriction on
ses using logit, although the results are robust to other its own are 14% lower. Put another way, a one-standard
functional forms, such as probit or the complementary deviation increase in the Outflow (economic competi
log-log function (Buckley and Westerland 2004). Other tion) variable?which indicates that a city is surrounded
distributions of the hazard rates yield very similar results, by more people, relative to its own population, who are
whether based on a Weibull distribution or a Cox propor not covered by the policy?is associated with a 33% drop
tional hazards model. To account for heteroskedasticity in the odds of adoption in any given year. Cities thus are
and correlation across observations, we cluster by city revealed to be hesitant to adopt antismoking policies until
year using the cluster procedure in Stata 9.2, which allows their neighbors act.
the possibility of dependence in the three policy choices As shown in Model 3, Nearest Bigger City demon
within each city in a given year and relies on Huber/White strates strong support for the Imitation Hypothesis, indi
robust standard errors.21 Year dummies help account for cating that the likelihood of a city adopting antismoking
potential patterns of temporal dependence. The number laws increases when the nearest city that is bigger has
of observations is determined by the number of cities at already adopted such a law. Substantively, the effect of
risk for each policy's adoption in each year.22 neighboring city diffusion is quite large. Compared to a
Our initial results focus on the four main diffusion city without a previous adoption by its larger neighbor,
hypotheses examined separately, as shown in Table 1. a city whose nearest bigger neighboring city already has
Although each of these models therefore faces poten the same antismoking restriction in place is three times as
tial omitted variable problems, this preliminary step is likely to adopt its own restriction in any given year.
instructive in setting baselines against which the fully Finally, the results of Model 4 provide clear evidence
specified model can be compared. Because most previ in favor of the Coercion Hypothesis. The negative and sig
ous studies have included only one mechanism or one nificant coefficient on State Law indicates that the adop
diffusion variable, this comparison also underscores the tion of a state law restricting smoking decreases the odds
degree to which those earlier results may be inaccurate. that a city within that state will adopt a similar law. More
Support for the Learning Hypothesis comes from our specifically, the effect of a given state antismoking law is
first horizontal diffusion variable, Proportion of State Pop a 26% decline in the odds of a local adoption. Even more
ulation with Local Restriction. The coefficient on this vari striking is the negative and significant coefficient for the
able in Model 1 is positive and significant, indicating that State Preemption variable, indicating that the adoption of
a city is more likely to adopt antismoking laws when a a preemptive state-level law decreases the odds of a local
greater proportion of people in other cities within the antismoking restriction by 94%.
21 Similar results follow from clustering by city. 23Interestingly, we found no support for an "Inflow" variable that
we also constructed, a variable that sought to determine whether
22 Somewhat fewer observations are found in regressions including a city would be less likely to adopt laws if surrounding cities had
Nearest Bigger City, as this variable is undefined in states with only already adopted such laws (i.e., where the city could hope to draw
one city over 50,000. people from those other cities).

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
CHARLES R. SHIPAN AND CRAIG
Table 1 City-Level Adoption of Antismoking Policies

Model 1 Model 2 Model 3 Model 4
Learning Competition Imitation Coercion
Horizontal Diffusion
Learning (Proportion of State 3.98***
Population with Local Restriction) (0.365)
Competition (Outflow) -0.0426***
(0.0148)
Imitation (Nearest Bigger City)
(0.125)
Vertical Diffusion
Coercion: State Law -0.300***
(0.0980)
Coercion: State Preemption -2.80***
(0.420)
City-Level Controls
City Population (in 100,000s) 0.0453*** 0.0353*** 0.0286*** 0.0371***
(0.0079) (0.0078) (0.0083) (0.0081)
Mayor-Council -0.118 -0.345*** -0.204* -0.334***
(0.132) (0.127) (0.128) (0.126)
Percent Health Employees 0.0453*** 0.0288* 0.0340** 0.0247
(0.0188) (0.0207) (0.0202) (0.0197)
Percent High School Graduates 0.0432*** 0.0253*** 0.0284*** 0.0282***
(0.0079) (0.0084) (0.0082) (0.0081)
Per Capita Income -0.0189 0.0436*** 0.0213* 0.0208*
(0.0164) (0.0143) (0.0149) (0.0142)
Percent White 0.00006 -0.0107t -0.00641 -0.00762
(0.00381) (0.00362) (0.00355) (0.00348)
Per Capita Government 0.241*** 0.0623 0.124 0.130
Spending (0.0990) (0.107) (0.104) (0.106)
State-Level Controls
Percent Smokers -0.0405** -0.0693*** -0.0663*** -0.0795***
(0.0209) (0.0185) (0.0196) (0.0200)
Tobacco Production 0.145 -0.310** -0.216* -0.137
(0.149) (0.136) (0.136) (0.138)
Tobacco Lobbyists 12.6* 10.6* 10.2 6.71
(8.51) (7.92) (8.31) (8.11)
Health Organization Lobbyists -1.43* -2.68*** -2.34*** -1.90**
(0.964) (0.818) (0.893) (0.830)
State Government Ideology -0.00453** -0.00183 -0.00482** -0.00205
(0.00273) (0.00256) (0.00262) (0.00262)
Waldx2 528.2*** 365.8*** 404.1*** 375.3***
N 34,415 34,415 32,810 32,903
Robust standard errors in parentheses, clustered by city-year. All models include yearly dummy variables and a constant, not
shown here due to space considerations.
***p < 0.01, **p < 0.05, *p < 0.1 (one-tailed tests).
tp < 0.01 (two-tailed test with coefficient taking unexpected sign).

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
The controls for city-level characteristics behave these mechanisms follow the systematic temporal and
largely as expected. The positive and highly significant conditional patterns predicted by our hypotheses.
coefficient on City Population indicates that, as expected,
larger cities took the lead in adopting antismoking restric Multiple Mechanisms at Work
tions.24 Additionally, wealthier and more highly educated
cities, and those with a greater presence of health profes As noted earlier, although our main independent vari
sionals, were more likely to adopt antismoking policies. ables are designed to isolate key aspects of different diffu
Consistent with a lower level of efficiency, the Mayor sion mechanisms, they also may capture aspects of other
Council variable is negative and statistically significant. mechanisms as well. For example, rather than just imitat
Surprisingly, despite lower smoking rates among whites, ing the nearest larger neighbor and learning from cities
predominantly white cities were actually somewhat less throughout the state, cities may learn from their nearest
likely to pass antismoking measures. larger neighbors or imitate other cities in the state. Com
At the state level, the negative coefficients for Per paring the results of Model 5 in Table 2 to the models from
cent Smokers and Tobacco Production are consistent with Table 1 demonstrates that, although all of the mechanism
self-interested behavior. Meanwhile, the governmental variables remain significant in the multivariate model, the
sizes of the coefficients in Model 5 differ from those in
state-level controls provide some intriguing results. The
coefficient for Tobacco Lobbyists tends to be positive and Table 1. For instance, controlling for the other diffusion
statistically significant; and the negative (although not mechanisms, the coefficient on our learning variable (Pro
always significant) coefficients on Health Organization portion of State Population with Local Restriction) declines
Lobbyists and State Government Ideology are suggestive. As by about 20%, from 3.98 to 3.20. Additional analyses in
we noted earlier, the expectations for these variables are dicate that most of this drop is due to the inclusion of the
unclear; and the results might seem to imply, perversely, imitation variable, indicating that about one-fifth of the
that a strong pro-tobacco presence produces more anti learning effect detected in Model 1 was due to the policy
smoking restrictions, while a strong pro-health presence choice of the nearest bigger city.25
and a liberal state government lead to fewer antismoking Second, the coefficient on the competition variable
restrictions. Another interpretation of these results, how (Outflow) likewise diminishes by one-quarter between
ever, is also possible. Our findings, while not constituting Models 2 and 5. Again, this drop comes mainly from the
hard proof, are nonetheless consistent with the idea that inclusion of Nearest Bigger City, indicating that the effect
policy advocates engage in venue shopping (e.g., Baum of competition shown in Model 2 had been somewhat
gartner and Jones 1993; Boehmke, Gailmard, and Patty overstated by failing to control for imitation of the near
2006; Pralle 2003). When the tobacco lobby is strong at est bigger city. Third, and most dramatically, the coeffi
the state level, for example, antismoking advocates realize cient on the imitation mechanism variable itself (Nearest
that they will have little chance to succeed at the state Bigger City) is cut to less than half of its value between
level and thus will turn their attention to cities. Although Models 3 and 5. This drop is mainly due to inclusion of
we are hesitant to make too strong of a claim in favor the learning variable, indicating that a substantial portion
of venue shopping?we would want to know, for exam of what previously had appeared to be imitation was due
ple, that state-level strength for these advocates does not to learning from other cities, including from the nearest
carry over to local-level strength?these results suggest larger neighbor. Finally, the coercion variables maintain
that future exploration may be fruitful. their sizes upon controlling for the other diffusion mech
In sum, the results in Table 1 show strong patterns of anisms, suggesting that vertical diffusion in the form of
policy diffusion through all four mechanisms of diffusion state-to-local coercion is a process wholly separate from
discussed above. We now expand on these results in two the horizontal spread of policies from city to city.
ways. First, we examine whether these mechanisms collec
tively influence diffusion?that is, do we continue to find
evidence of each of these types of effects even when we Temporal Diffusion Patterns
control for all of the others? Second, we explore whether
The Temporal Effects Hypothesis predicted that imitation
should be a relatively short-lived effect, while the effects
24The results in this table hold when we limit our analysis to cities
with populations under 1,000,000; thus, the results are not being 25 Specifically, including all of the variables in Model 5 except Nearest
driven by a few large cities. In addition, similar results follow from Bigger G'fy produces a coefficient on Proportion of State Population
including population squared rather than population as a control with Local Restriction (as well as on Outflow) that is very near its
variable. Table 1 value.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
CHARLES R. SHIPAN AND CRAIG VOLDEN
Table 2 All Diffusion Mechanisms, Temporal Lags, and Conditional Effects

Model 5 Model? Model 7
All Mechanisms Lagged Effects Conditional E
Learning (Proportion of State 3.20*** 2.47*** 3.02***
Population with Local Restriction) (0.442) (0.474) (0.487)
Learning x Population ? ? 0.316**
(0.190)
Competition (Outflow) -0.0320** -0.0249** -0.0644***
(0.0150) (0.0125) (0.0260) ***
Competition x Population ? ? 0.0438
(0.0171)
Imitation (Nearest Bigger City) 0.456*** 0.243* 0.486***
(0.155) (0.175) (0.161)
Imitation x Population ? ? -0.0392
***
(0.0123)
Coercion: State Law -0.345*** -0.422*** -0.309***
(0.0924) (0.0946) (0.105)
Coercion (State Law) x Population ? ? ?0.0275
(0.0451)
Coercion: State Preemption -2.75*** -2.74*** -2.76***
(0.428) (0.508) (0.470)
Coercion (Preemption) x Population ? ? 0.0376
(0.156)
Waldx2 562.8*** 438.9*** 589.6***
N 32,810 29,813 32,810
Robust standard errors in parentheses, clustered by city-year. All models include city-leve
controls, yearly dummy variables, and a constant, not shown here due to space consideration
***p < 0.01, **p < 0.05, *p < 0.1 (one-tailed tests).
of learning and competition should be moredrop by less than one-quarter and retain t
enduring.
To test this hypothesis, we reran Modeltial 5, replacing
statistical the
significance.27 Taken together,
our
diffusion variables with the same variables findings
lagged by two support the Temporal Effect
years. The results are displayed in Model but 6.26 If the
they also ef
lend additional support to ou
fect of imitation takes place immediately ization
upon theof action
the key variables of learning, com
imitation.
of the Nearest Bigger City and then fades as theAlthough
imi there may be overlap
tating policymakers move on to other variables,
issues, then
theythe
exhibit temporal effects that
with theoretical
coefficient on the lagged imitation variable should be expectations.
substantially diminished for the lagged variable as com
pared to the more immediate short-term effect shown
The
in Model 5. Conversely, if the effects of the Contingent
other hori Nature of Diff
zontal mechanisms are more enduring, their coefficients
The models examined so far have treated
should remain roughly the same size as in Model 5. This
process on
is indeed what we find. While the coefficient as being
Near fundamentally the same f
Each city, for example, was found to be m
est Bigger City drops nearly 50% to 0.243 and becomes
adopt policies found in other cities and less li
statistically significant only with a very lenient p < 0.10
standard and a one-tailed test, the coefficients on Propor
27 Interestingly, despite our lack of clear theoretic
tion of State Population with Local Restriction and Outflow
the coercion variables show temporal effects sim
the competition and learning variables. In addition,
learning, competition, and imitation are unchang
'Other lags showed similar results. use the unlagged versions of the coercion variables.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
policies where the state government had already acted shrinks to zero when city size reaches about 1.2 million
(especially with preemptive laws). There is reason to be people.
lieve, however, that diffusion does not work the same way Fourth, the interactions between Population and the
for every government. Dealing with state-level antismok coercion variables of State Law and State Preemption
ing adoptions, for example, Shipan and Volden (2006) are statistically indistinguishable from zero. Thus, con
find different patterns of diffusion for states with more trary to the Conditional Coercion Hypothesis, both large
professional legislatures than for those with less profes and small cities alike are coerced by the governments
sional legislatures. This article similarly posits that a more of the states in which they are situated.28 Vertical diffu
complete understanding of diffusion must recognize that sion thus appears to be less conditional than horizontal
diffusion effects?more specifically, the effects of various diffusion.
mechanisms?can be conditional. Taken together, these interactions provide strong ev
The Conditional Diffusion Hypotheses suggest that idence in support of the Conditional Diffusion Hypothe
larger cities are better able to learn from others, less sus ses for the three horizontal diffusion mechanisms.29 Put
ceptible to economic competition, less likely to engage simply, the smallest cities in our dataset (with 50,000 res
in imitation, and less vulnerable to coercion. To test idents) are responsive to many diffusion pressures. They
these four hypotheses, we interact City Population (in imitate larger cities, learn somewhat from experiments of
100,000s) with measures for each of the four mechanisms others, and worry about the economic consequences of
of policy diffusion. Model 7 in Table 2 mimics Model their laws. The average city in our analysis (population
5, but with the inclusion of these interactive variables, 150,000) still imitates the biggest cities and is better able to
and produces the following results. First, consistent with learn from experiments elsewhere in the state, but shows
the Conditional Learning Hypothesis, the positive coef little concern for economic spillovers to the (typically
ficient on the interaction between Population and Pro smaller) surrounding communities. Finally, the largest
portion of State Population with Local Restriction shows cities show no evidence of imitation and spillover con
that learning is enhanced in larger cities. Because their cerns, seeming rather to act based on the experiences of
governments are larger and more capable, these bigger earlier adopters throughout the state.
cities are better able to build on the experiments of oth
ers. For example, a city of 50,000 increases its odds of
adoption by 3.2% for each additional percent of the state
population covered by other cities' laws. The compara
Discussion and Conclusion
ble learning-based boost for a larger city of 500,000 is
4.7%. The scholarship to date on diffusion has shown robust
Second, the positive coefficient on the interaction patterns of policies and institutions spreading from coun
between Outflow and Population indicates that smaller try to country and from state to state. This study finds that
cities are more concerned with economic competition localities are also susceptible to horizontal and vertical
than are larger cities. For example, consider two contigu diffusion pressures. More importantly, this work reveals
ous cities of the same size. If the cities are of popula some of the benefits of studying diffusion at the local
tion 50,000, the lack of a policy in the neighboring city level. In particular, we demonstrate one way to disentan
diminishes the other city's odds of adoption by about gle the multiple diffusion mechanisms of learning, eco
4.2%. Yet, if the cities are each of population 147,000, nomic competition, imitation, and coercion. We show
this economic competition effect falls to zero. This find not only that these mechanisms exist, but also that the
ing shows support for the Conditional Competition effect of imitation fades over time while the other mech
Hypothesis. anisms' effects are more persistent. Moreover, diffusion

Third, consistent with the Conditional Imitation Hy mechanisms play different roles in large cities than in
pothesis, the negative coefficient on the interaction be small cities. Compared to small cities, large cities are
tween Nearest Bigger City and Population indicates that
larger cities are indeed less likely to rely on imitation than 28 Additional analyses indicate that very small cities are somewhat
are smaller cities?that is, the effect of Nearest Bigger City more likely to be deterred by state action than are larger cities.
diminishes for larger cities. For example, a city of 50,000 Perhaps larger cities limit coercion by states earlier in the policy
process, when city officials lobby state legislators against preemptive
has 59% greater odds of an antismoking adoption if its clauses.
nearest bigger city already has the policy. Yet that boost
29 Interactions of the horizontal diffusion variables with Per Capita
in the likelihood of adoption due to imitation drops to Income and with Mayor-Council suggest further conditional diffu
34% for a city of 500,000. Overall, the imitation effect sion effects, but are excluded here due to space considerations.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
equally susceptible to coercion from the state government, governments will experiment and learn from one another.
but are more likely to learn from others' experiments, Here we find such learning, even among the smallest of
and are less likely to simply imitate the policies of others cities in our dataset. However, the other side of devo
or to be deterred in their actions by potential economic lution is that the lower-level governments may compete
competition. with one another in ways that are not mutually beneficial,
Such findings have important normative implica or may adopt policies with negative externalities felt by
tions. First, it is altogether possible that the most appro others. We find that economic competition is relevant to
priate policy for one government may be different from antismoking policy choices and has caused some (par
that for another government serving a different popula ticularly small) cities to hesitate to adopt policies until
tion. Evidence that small cities simply copy their larger their neighbors do the same. When, instead of devolu
neighbors suggests that their policies may not be as well tion, policy centralization takes place, such as with state
suited for their populations as would be ideal. Second, laws including preemptive clauses, it is unsurprising that
when the control of policies is devolved to lower levels such experimentation and competition comes to a nearly
of government, one of the potential benefits is that these complete halt.
Appendix Variable Descriptions, Summary Statistics, Sources

Variable Description Mean St. Dev.
Antismoking Policy Adoption* Dependent variable = 1 if city adopts its first law 0.015 0.123
in this area in this year. Set = 0 if no adoption to
date. Observation removed if already adopted.
Proportion of Population with Proportion of state population living in localities 0.112 0.163
Local Restrictions** with restrictions in this area at start of the year.
Outflow** Ratio of sum of populations of surrounding cities 3.769 9.250
without policy to home city's population.
Nearest Bigger City** Dummy = 1 if the nearest city that is larger 0.211 0.408
than the observation city adopts its law in this area
prior to the observation year.
State Lawc Dummy = 1 if state adopted restriction in 0.364 0.481
this area prior to this year.
State Preemption* Dummy = 1 if state adopted law prior to this 0.144 0.351
year that prohibits or limits city-level
government laws in this area.
City Population City population (in 100,000s) at the time 1.482 3.834
of the nearest census.
Mayor-Council0 Dummy = 1 if the city has a mayor-council 0.326 0.469

form of government.
Percent Health Percent of employed residents working in 9.18 2.77
Employees* health services professions.
Percent High School Percent of adults over age 25 with high school 72.8 11.5
Graduates* diplomas or equivalencies.
Per Capita Income* Average income per resident ($ 1000s). 11.7 4.78
Percent White* Percent of residents self-identified as white. 76.8 17.5
(continued)

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
MECHANISMS OF POLICY DIFFUSION
Appendix Continued
Variable Description Mean St. Dev.
Per Capita Government Government spending in thousands of 0.801 0.538

Spending^ dollars per resident.
Percent Smokers{ Percent of adults in state who smoke. 24.5 3.51
Tobacco Production* Dummy = 1 if tobacco produced in state. 0.329 0.470
Tobacco Lobbyist^1 Proportion of lobbyists in the state working for 0.014 0.007
tobacco industry, based on 1994 snapshot.
Health Organization Proportion of lobbyists in the state working for 0.089 0.062
Lobbyist^1 health organizations, based on 1994 snapshot.
State Government Ideology? Ideology score for state government. 52.5 21.6
Data sources:
Constructed by authors based on American Nonsmokers' Rights Foundation Local Tobacco Control Ordinance Database?.
bConstructed by authors based on U.S. Census data.
cConstructed based on National Cancer Institute, State Cancer Legislative Database Program, Bethesda, MD: SCLD.
dCity and County Databook, various years.
Constructed by authors based on U.S. Census data provided by the Taubman Center, Harvard University.
fCenters for Disease Control and Prevention website (http://apps.nccd.cdc.gov/statesystem/, choose "Detailed Report").
Constructed by authors based on U.S. Department of Agriculture data.
Constructed by authors based on Goldstein and Bearman (1996).
'Berry, Ringquist, Fording, and Hanson (1998) data on ICPSR website.
References ment Ideology in the American States, 1960-93." American

Journal of Political Science 42(1): 327-48.
Allen, Mahalley D, Carrie Pettus, and Donald P. Haider-Markel. Boehmke, Frederick J. 2005. The Indirect Effect of Direct Legisla
2004. "Making the National Local: Specifying the Conditions tion: How Institutions Shape Enterest Group Systems. Colum
for National Government Influence on State Policymaking." bus: Ohio State University Press.
State Politics and Policy Quarterly 4(3): 318-44. Boehmke, Frederick J., Sean Gailmard, and John Wiggs Patty.
2006. "Whose Ear to Bend? Information Sources and Venue
Andersen, B. S., M. E. Begay, and C. B. Lawson. 2003. "Break
ing the Alliance: Defeating the Tobacco Industry's Allies and Choice in Policy Making." Quarterly Journal of Political Sci
Enacting Youth Access Restrictions in Massachusetts." Amer ence 1(2): 139-69.
ican Journal of Public Health 93(11): 1922-28. Boehmke, Frederick J., and Richard Witmer. 2004. "Disentan
Bailey, Michael A., and Mark Carl Rom. 2004. "A Wider gling Diffusion: The Effects of Social Learning and Economic
Race? Interstate Competition across Health and Welfare Pro Competition on State Policy Innovation and Expansion."
grams." Journal of Politics 66(2): 326-47. Political Research Quarterly 57(I): 39-51.
Baila, Steven J. 2001. "Interstate Professional Associations and Box-Steffensmeier, Janet M., and Christopher Zorn. 2002. "Du
the Diffusion of Policy Innovations." American Politics Re ration Models for Repeated Events." Journal of Politics 64(4):
search 29(3): 221-45. 1069-94.
Baumgartner, Frank R., and Bryan D. Jones. 1993. Agendas Brandeis, Louis Dembitz. 1932. "Dissenting Opinion." New
and Instability in American Politics. Chicago: University of State Ice Co. v. Liehmann, 285 U.S. 262, 311.
Chicago Press. Buckley, Jack, and Chad Westerland. 2004. "Duration Depen
Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. dence, Functional Form, and Correct Standard Errors: Im
"Taking Time Seriously: Time-Series-Cross-Section Analy proving EHA Models of State Policy Diffusion." State Politics
sis with a Binary Dependent Variable." American Journal of and Policy Quarterly 4( 1 ): 94-113.
Political Science 42(4): 1260-88. Chriqui, Jamie F. 2000. Restricting Minors Access to To
Berry, Frances Stokes, and William D. Berry. 1990. "State Lot bacco Products: .An Examination of State Legislation
tery Adoptions as Policy Innovations: An Event History and Policy Innovation. Ph.D. dissertation. University of
Analysis." American Political Science Review 84(2): 395? Maryland.
415.
Conlisk, E., et al. 1995. "The Status of Local Smoking Regula
Berry, William D, and Brady Baybeck. 2005. "Using Geo tions in North Carolina Following a State Preemption Bill."
graphic Information Systems to Study Interstate Compe JAMA: Journal of the American Medical Association 273 ( 10):
tition." American Political Science Review 99(4): 505-19. 805-7.
Berry, William D, Evan J. Ringquist, Richard C. Fording, and Crain, Robert L. 1966. "Fluoridation?Diffusion of an Innova
Russell L. Hanson. 1998. "Measuring Citizen and Govern tion among Cities." Social Forces 44(4): 467-76.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
Evans, William N., Matthew C. Farrelly, and Edward Mont Mintrom, Michael. 1997b. "The State-Local Nexus in Policy
gomery. 1999. "Do Workplace Smoking Bans Reduce Smok Innovation Diffusion: The Case of School Choice." Publias:
ing?" American Economic Review 89(4): 728-47. The Journal of Federalism 27(3): 41-59.
Franzese, Robert J., Jr., and Jude C. Hays. 2008. "Interdepen Moon, M. J. 2002. "The Evolution of E-Government among
dence in Comparative Politics: Substance, Theory, Empirics, Municipalities: Rhetoric or Reality?" Public Administration
Substance." Comparative Political Studies 41 (4yf5): 742-80. Review 62(4): 424-33.
Frederickson, H. G., G. A. Johnson, and C. Wood. 2004. "The Mooney, Christopher Z., and Mei-Hsien Lee. 1995. "Legisla
Changing Structure of American Cities: A Study of the Dif tive Morality in the American States: The Case of Pre-Roe
fusion of Innovation." Public Administration Review 64(3): Abortion Regulation Reform." American Journal of Political
320-30. Science 39(3): 599-627.
Givel, Michael S., and Stanton A. Glantz. 2001. "Tobacco Lobby Mooney, Christopher Z., and Mei-Hsien Lee. 1999. "The Tem
Political Influence on U.S. State Legislatures in the 1990s." poral Diffusion of Morality Policy: The Case of Death Penalty
Tobacco Control 10: 124-34. Legislation in the American States." Policy Studies Journal
Glantz, Stanton A., and Annemarie Charlesworth. 1999. 27(4): 766-80.
"Tourism and Hotel Revenues Before and After Passage of Peterson, Paul E. 1981. City Limits. Chicago: University of
Smoke-Free Restaurant Ordinances." Journal of the Ameri Chicago Press.
can Medical Association 281(20; May 26): 1911-18. Peterson, Paul E., and Mark C. Rom. 1990. Welfare Magnets:
Godwin, Marcia L., and Jean R. Schroedel. 2000. "Policy Dif A New Case for a National Standard. Washington, DC: The
fusion and Strategies for Promoting Policy Change: Evi Brookings Institution.
dence from California Local Gun Control Ordinances." Pol
Pralle, Sarah B. 2003. "Venue Shopping, Political Strategy, and
icy Studies Journal 28(4): 760-76. Policy Change: The Internationalization of Canadian Forest
Goldstein, Adam O., and Nathan S. Bearman. 1996. "State To Advocacy." Journal of Public Policy 23(3): 233-60.
bacco Lobbyists and Organizations in the United States: Purlee, Scott. 1995. "Driving Influences: Operators Now Use
Crossed Lines." American Journal of Public Health 86(8): Travel Time to Tally Potential Customers." Restaurants and
1137-42. Institutions 105(10): 33.
Gray, Virginia. 1973. "Innovation in the States: A Diffusion Rogers, Everett M. 1995. The Diffusion of Innovations. 4th ed.
Study." American Political Science Review 67: 1174-85. New York: The Free Press.
Grupp, Fred W, Jr., and Alan R. Richards. 1975. "Variations in Ross, Hana, and Frank J. Chaloupka. 2004. "The Effect of Public
Elite Perceptions of American States as Referents for Pub Policies and Prices on Youth Smoking." Southern Economic
lic Policy Making." American Political Science Review 69(3): Journal 70(4): 796-815.
850-58.
Schroeder, Steven A. 2004. "Tobacco Control in the Wake of the
Haider-Markel, Donald P. 2001. "Policy Diffusion as a Geo 1998 Master Settlement Agreement." New England Journal
graphical Expansion of the Scope of Political Conflict: Same of Medicine 350(3): 293-301.
Sex Marriage Bans in the 1990s." State Politics and Policy
Shipan, Charles R., and Craig Volden. 2006. "Bottom-Up Fed
Quarterly I: 5-26.
eralism: The Diffusion of Antismoking Policies from U.S.
Huber, John D, and Charles R. Shipan. 2002. Deliberate Discre Cities to States." American Journal of Political Science 50(4):
tion? New York: Cambridge University Press. 825-43.
Huber, John D, Charles R. Shipan, and Madelaine Pfahler. Siegel, Michael, et al. 1997. "Preemption in Tobacco Con
2001. "Legislatures and Statutory Control of Bureaucracy." trol: Review of an Emerging Public Health Problem." JAMA
American Journal of Political Science 45(2): 330-45. 278(10): 858-63.
Jacobson, Peter D., and Jeffrey Wasserman. 1997. Tobacco Con Simmons, Beth A., Frank Dobbin, and Geoffrey Gar
trol Laws: Implementation and Enforcement. Santa Monica, rett. 2006. "Introduction: The International Diffusion
CA: RAND. of Liberalism." International Organization 60(4): 781?
Karch, Andrew. 2006. "National Intervention and the Diffusion 810.
of Policy Innovations." American Politics Quarterly 34(4): Skeer, M., et al. 2004. "Town-Level Characteristics and Smok
403-26. ing Policy Adoption in Massachusetts: Are Local Restau
Knoke, David. 1982. "The Spread of Municipal Reform: Tem rant Smoking Regulations Fostering Disparities in Health
poral, Spatial, and Social Dynamics." American Journal of Protection?" American Journal of Public Health 94(2): 286
Sociology 87(6): 1314-39. 92.
Martin, I. 2001. "Dawn of the Living Wage: The Diffusion Tanyeri, Dana. 2007. "So Who Is Your Customer Anyway?"
of a Redistributive Municipal Policy." Urban Affairs Review Restaurant Business 106(8): 26.
36(4): 470-96. Volden, Craig. 2002. "The Politics of Competitive Federalism: A
Meseguer, Covadonga. 2006. "Learning and Economic Policy Race to the Bottom in Welfare Benefits?" American Journal
Choices." European Journal of Political Economy 22: 156-78. of Political Science 46(2): 352-63.
Mintrom, Michael. 1997a. "Policy Entrepreneurs and the Dif Volden, Craig. 2005. "Intergovernmental Political Competition
fusion of Innovation." American Journal of Political Science in American Federalism." American Journal of Political Sci
41(3): 738-70. ence 49(2): 327-42.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00
Volden, Craig. 2006. "States as Policy Laboratories: Emulat Wei, L. J., D. Y. Lin, and L. Weissfeld. 1989. "Regression Analysis
ing Successes in the Children's Health Insurance Program." of Multivariate Incomplete Failure Time Data by Modeling
American Journal of Political Science 50(2): 294-312. Marginal Distributions." Journal of the American Statistical
Volden, Craig. 2007. "Intergovernmental Grants: A Formal Association 84(408): 1065-73.
Model of Interrelated National and Subnational Political Welch, Susan, and Kay Thompson. 1980. "The Impact of Federal
Decisions." Publius: The Journal of Federalism 37(2): 209-43. Incentives on State Policy Innovation." American Journal of
Walker, Jack L. 1969. "The Diffusion of Innovations among the Political Science 24(4): 715-26.
American States." American Political Science Review 63(3): Weyland, Kurt. 2005. "Theories of Policy Diffusion: Lessons
880-99. from Latin American Pension Reform." World Politics 57:
Walker, Jack L. 1973. "Problems in Research on Diffusion of 262-95.
Policy Innovations." American Political Science Review67(A): Weyland, Kurt. 2007. Bounded Rationality and Policy Diffusion.
1186-91. Princeton, NJ: Princeton University Press.

189.192.220.70 on Fri, 16 Feb 2024 18:55:26 +00:00

Shipan MechanismsPolicyDiffusion 2008

Uploaded by

Copyright:

Available Formats

Shipan MechanismsPolicyDiffusion 2008

Uploaded by

Document Information

Original Description:

Original Title

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

Shipan MechanismsPolicyDiffusion 2008

Uploaded by

Copyright:

Available Formats

The Mechanisms of Policy Diffusion

Author(s): Charles R. Shipan and Craig Volden

Stable URL: https://www.jstor.org/stable/25193853

This content downloaded from

This content downloaded from

This content downloaded from

The fourth mechanism of diffusion?coercion? We acknowledge that the theoretical distinctions

This content downloaded from

This content downloaded from

This content downloaded from

Empirical Approach Mechanism Variables

This content downloaded from

10Economic competition is likely to affect clean indoor air adop

This content downloaded from

This content downloaded from

This content downloaded from

Table 1 City-Level Adoption of Antismoking Policies

This content downloaded from

This content downloaded from

Table 2 All Diffusion Mechanisms, Temporal Lags, and Conditional Effects

This content downloaded from

Hypothesis. anisms' effects are more persistent. Moreover, diffusion

This content downloaded from

Appendix Variable Descriptions, Summary Statistics, Sources

Mayor-Council0 Dummy = 1 if the city has a mayor-council 0.326 0.469

This content downloaded from

Per Capita Government Government spending in thousands of 0.801 0.538

References ment Ideology in the American States, 1960-93." American

This content downloaded from

This content downloaded from

This content downloaded from

You might also like