SEEING BLUE: A POLICE-CENTERED EXPLANATION OF

PROTEST POLICING*

Jennifer Earl and Sarah A. Soule†

Existing explanations of repression and the policing of protest focus on the interests of
political elites, with research indicating that a chief predictor of state repression is the level of
threat protesters pose to elite interests. However, prior research has only paid sporadic
attention to how the institutional and organizational characteristics of local law enforcement
agencies shape the character of protest policing. This article addresses this significant theo-
retical gap by developing a police-centered, or “blue,” approach to protest policing. Using
data on the policing of public protest events in New York State between 1968 and 1973, this
article finds support for the blue approach. Specifically, the situational threats posed by
protesters to those agents who actually perform repression—local police—are critical predic-
tors of police presence and action. Results also show some residual support for the role of
elite threats in structuring repression.

Violent authoritarian regimes and explosive confrontations between authorities and demon-
strators create a public caricature of state terror and repression that is as harrowing as it is
familiar. From political murders committed by the Guatemalan government (Ball 2005), to
Tiananmen Square where millions around the world watched as Chinese military tanks
crushed student demonstrators (Francisco 2005), to the streets of South Africa where apar-
theid was contested in pitched battles between white police officers and black subjects
(Francisco 2005), instances of repression remind citizens that states are, at times, willing to
wield their formidable power against their own citizens to injurious and deadly ends.
Of course, such violent interactions with insurgents are not the exclusive—or even the
dominant—reaction of the state to protest (della Porta and Reiter 1998; Earl, Soule and
McCarthy 2003). Nonetheless, one legacy of these eruptions of state power is the academic
attempt to explain why authorities react as they do to protest. Scholars have sought answers
by addressing two major concerns. First, some scholars have attempted to explain how and
why authorities’ dominant responses to protest change over time (Jaime-Jiménez and Reinares
1998; McCarthy, McPhail, and Crist 1999; McPhail, Schweingrubber, and McCarthy 1998;
Reiner 1998; Reiter 1998; Winter 1998). This research overlooks differences between protest
events, protesters, and police agencies at a given historical moment, focusing instead on more
general changes in levels or strategies of repression. Second, some researchers have taken the
converse position by focusing on the “allocation of repression” (Cunningham 2003a) across
protest events and social movements during a particular moment in history (Earl et al. 2003).
*
This research was supported by grants from the following: the National Science Foundation (SBR-9709337, SBR-
9709356, SES-9874000, a Dissertation Improvement Grant, and a Graduate Research Fellowship Grant), the
University of Arizona’s Social and Behavioral Sciences Research Institute, and the University of Arizona Vice-
President for Research Small Grants Program. We thank each of those granting agencies for their support, as well as
Doug McAdam, Susan Olzak, and John McCarthy for their critical role in the collection of the protest event data used
in this paper, Calvin Morrill, Elisabeth Clemens, Ronald Breiger, Mayer Zald, and David Meyer for their comments
on this project, and Katrina Kimport for her research assistance.
†
Jennifer Earl is Assistant Professor of Sociology at the University of California, Santa Barbara, CA 93106. Sarah
Soule is Professor of Sociology at University of Arizona, Tucson, AZ 85721. Please direct correspondence to Jennifer
Earl at jearl @soc.ucsb.edu.

© Mobilization: An International Journal 11(2): 145-164

145
146 Mobilization

Research in this area uses variation in repression within a given historical moment to answer
questions about why repressive agents direct their attention and action toward specific
movements or protest events, out of a wider range of movements and protests within a
historical period.
In this article, we focus on the second area of inquiry by building on existing case studies
and ethnographic work to outline a novel approach to explaining one form of repression—the
policing of public protest events—at the height of the 1960s protest wave in the U.S.
(Cunningham 2004; della Porta 1995, 1998; della Porta and Reiter 1998; Waddington 1994,
1998). Specifically, we develop and test an innovative approach that focuses on how the
institutional and organizational characteristics of policing may affect repression.1 Our com-
parison of this police-centered, or “blue,” approach to other major approaches to protest pol-
icing uses data on the public policing of protest events in New York State from 1968 to 1973.
Results confirm the importance of attending to policing as an institution and, to a lesser ex-
tent, police agencies as organizations.

PRIOR RESEARCH ON THE ALLOCATION OF PROTEST POLICING

A large amount of scholarship has attempted to explain which social movements and/or
protest events will be targeted for police monitoring and action.2 By far, the dominant
explanation for repression generally, and protest policing specifically, is the “threat approach”
(Davenport 2000a; Earl 2003; Earl et al. 2003). In this model of repression, larger threats to
political elites predict greater repression in terms of both frequency and severity. Threats to
political elites are argued to include such things as groups that use noninstitutional and
confrontational tactics (McAdam 1982), groups pursuing revolutionary or radical goals
(Bromley and Shupe 1983), countercultural groups (Wisler and Giugni 1999), groups with
multiple targets or goals (McAdam 1982), and large protest sizes or high levels of movement
mobilization (Davenport 2000a). Less threatening groups tend to be “accepted groups” with
small goals (Tilly 1978). Davenport (2000a) and others (Earl 2003; Earl et al. 2003) conclude
that threat is the most widely accepted and empirically supported explanation of repression
developed thus far.
Alternatively, some argue for a “weakness” approach, first advanced by Gamson (1975).
This approach argues that because states risk public embarrassment if they fail in repressive
attempts, states should have an interest in publicly repressing only insurgents who they think
they can defeat. In other words, groups may be threatening enough to catch elites’ eyes, but
they must be weak enough to attract overt state action against them. Others suggest parallel
arguments: Duvall and Stohl (1983) claim that weaker groups are vulnerable targets for op-
portunistic state power.3
Earl et al. (2003) examine this approach, focusing on two types of weakness: weakness-
from-within and weakness-from-without. Weakness-from-within suggests that protest events
composed of participants that lack substantial access to government and government officials
are “weaker” than those composed of participants who have more institutional and regular
access to government and government officials for two reasons: (1) the lack of access and
representation are expected to lower the costs of repression; and (2) it is expected that “weak”
protesters will have fewer routes for redress in the wake of repressive action. For instance,
protests that include or are primarily composed of marginalized groups, such as racial and
ethnic minorities, religious minorities, and the poor (Piven and Cloward 1977) may be less
able to resist repression and/or less able to retaliate politically against repressors (Stockdill
1996). Other work suggests that the political legitimacy of protest claims may affect the
image of protesters and their policing (de Biasi 1998; della Porta 1998). Consistent with the
resource mobilization approach, resources and organizational capacity could contribute to the
relative ability of protesters to discourage and/or react to repression such that protests with
Seeing Blue 147

fewer or no social movement organizations (SMOs) present would be relatively “weaker.”

Earl et al. (2003) also suggest a weakness-from-without conceptualization. Protesters are
seen as relying on external monitoring of the state to limit repression (della Porta 1998; della
Porta and Reiter 1998), with weaker protesters hoping that influential elites will monitor
repression and contest overzealous acts of coercion on their behalf. Wisler and Giugni’s
(1999) work, which finds an inverse relationship between the level of media coverage and the
level of repression, could be interpreted in this light.
Noticeably absent from these major approaches is a discussion of police agencies as
organizations or policing as an institution, despite the fact that police, at least in the U.S., are
responsible for most protest policing. The focus of both the threat and weakness approaches is
on the will of political elites and the meaning of various characteristics of protesters and
protest events to political elites. Research suggests, though, that the institutional and organi-
zational characteristics of police may matter, but this work has not resulted in the creation of a
coherent explanatory approach.
For instance, Earl et al. (2003) find that police force capacity, but not prior reports of
protest-related brutality, affected police presence at protest events, but that neither variable
affected police action. Earl et al. acknowledge that these two variables do not comprise a
comprehensive account of institutional and organizational police influences, viewing work in
this area as nascent.
Waddington (1994, 1998) differentiates between “on the job” and “in the job” trouble,
arguing that police attempt to react to protest in ways that minimize both sources of trouble.
“On the job” trouble, where protests are concerned, “arises from the potential for disorder and
violence that might result in damage to property and injury to participants, including the
police” (Waddington 1998: 119). “In the job trouble” occurs when external bodies review
police decisions because of controversy, such as official inquiries following a violent
protester-police confrontation. Waddington’s ethnographic studies of British policing suggest
the importance of these two sources of trouble, but these sources of trouble do not frame a
dominant explanatory approach to repression.
Della Porta and Reiter (1998) outline a model of protest policing in their edited volume
on the policing of mass demonstrations in Western democracies. According to their approach,
stable political opportunity structures are comprised of the structural relationship of the police
to political elites, statutory and judicial restrictions on police, political culture, and the vintage
of the democracy, among other factors. These stable political opportunities combine with the
“current configurations of power” (Kriesi 1995; or “volatile political opportunities” in della
Porta and Reiter’s 1998 language) and with the history of police organizations’ interactions
with specific social movements to produce “police knowledge.” Police knowledge, defined as
“the police’s perception of their role and of the external reality” (1998: 22), then directly
structures protest policing.
Even in this more elaborated approach, institutional and organizational police
characteristics are limited to the history of interactions between specific police organizations
and protesters, as well as the relationship of the police to the state and political elites. In other
words, although this work mentions organizational and institutional police characteristics, the
approach centers on how police fit into stable and volatile political opportunities. Further,
della Porta and Reiter’s approach is designed to explain aggregate styles of policing that differ
crossnationally and longitudinally, not to explain the allocation of repression in a given time
period and place.
Waddington, Jones and Critcher (1989) develop a model of “flashpoints” in which they
discuss how characteristics of the police, the polity, and interactions between police and
civilians may escalate conflicts. However, because their analytic focus is on explaining
flashpoints, they do not offer a more general theory of how police characteristics and patterns
of interaction shape a broad range of police-protester encounters.
Finally, Cunningham’s research on the FBI’s COINTELPROs against the New Left and
148 Mobilization

white hate groups highlights the importance of organizational structure in the allocation of re-
pression (2003a, 2003b, 2003c, 2004). He argues that because the FBI Director’s Office was
the only one with access to information from other FBI field offices and the ability to sanction
field offices and agents who were not engaged enough in repressive activities, the Director’s
Office strongly structured the allocation of repression. Even with Cunningham’s concern for
organization structure, he does not suggest applications of his approach to local police
departments, which are engaged in the lion’s share of public protest policing in the U.S.
Further, institutional concerns of law enforcement generally are downplayed as analytic
attention is directed toward the peculiar role of the FBI Director’s Office.
The lack of a coherent explanatory approach to repression that focuses on the police is an
important oversight when one considers both that local law enforcement agencies are
responsible for the lion’s share of U.S. protest policing and that sociolegal scholars and
criminologists have long recognized that police agencies have distinctive institutional char-
acteristics that shape how agencies and officers perform their duties (Lundman 1980;
Skolnick 1966). In the U.S., these distinctive characteristics include high levels of discretion
for line-officers in the performance of routine duties, historically varied attachments (and
distance) from political elites, an insular police culture cultivated formally and informally by
police agencies, and particularly salient organizational concerns over officer safety. Police
agencies also have organizational characteristics that vary across agencies (e.g., personnel and
resources) and might impact protest control.
To address this gap in the literature, we suggest a new approach that offers a more com-
prehensive account of the role of police characteristics in structuring the allocation of repress-
sion and more deeply embraces the particular institutional and organizational features of
police agencies. After outlining our “blue” approach, we empirically compare that approach
to the threat and weakness approaches using data on protest events that occurred in New York
State from 1968 to 1973.4

A “BLUE” APPROACH TO EXPLAINING POLICE ACTION AT PROTESTS

The police-centered, or “blue,” approach outlined here focuses on two main elements. First,
our approach focuses on key institutional features of policing. These institutional features are
characteristics that should be common among U.S. police agencies. This focus on institutional
features is in keeping with other advances in the social sciences. For instance, new institu-
tionalist research suggests that core elements of institutional fields should directly affect
organizational activities (DiMaggio and Powell 1991). In particular, we focus on the insti-
tutional imperative for control in police encounters that characterizes policing generally, and
American police in particular. Second, the blue approach acknowledges that not all police
agencies are cut of the same cloth, despite important institutional similarities. Thus, the second
prong of the approach examines organizational characteristics that vary across police agen-
cies, suggesting what kinds of organizational variation may advance explanations of protest
policing.

Institutional Concerns: Rethinking Threat

Maintaining public order and enforcing laws are the primary aims of the police (Bittner
1967a). The emphasis on public order is reflected in institutional concerns about maintaining
control in interactions and over communities (Balch 1972; Rubinstein 1980; Skolnick 1966;
Tauber 1967; Wilson 1968). Many within policing see public disorder as a loss of control by
police and a mark of failure. The importance of control in police agencies is pervasive and
manifest. Agencies evaluate themselves on their ability to maintain order and control over
situations and subjects. These institutional imperatives filter down to line-level officers who
Seeing Blue 149

grade their effectiveness and job performance according to their ability to maintain control
over situations and the persons they encounter (Hartjen 1972).
Further, American police cement the importance of control by postulating a relationship
between the maintenance of control and officer safety. Prior research shows that although
police officers objectively face risk in the line of duty, this risk is transformed through social
and organizational processes into a sense of danger and into a set of routines for identifying
and responding to danger (Tauber 1967).5 That is, policing involves a host of risks to officers
and to agencies, but officers must learn about these risks and learn how to identify them,
which are fundamentally social processes. The social construction of danger and cues of
danger means that not all objective risks are attended to by officers and agencies and that some
objective risks are exaggerated (Tauber 1967).
This institutional structuring of danger affects how police understand and relate to key
institutional imperatives such as maintaining public order, controlling communities, and
controlling interactions with the public (Balch 1972; Rubinstein 1980; Skolnick 1966; Tauber
1967; Wilson 1968). Rubinstein (1980) argues that the importance of control and the threat
represented by losing control encourage vigilant monitoring:

[The police officer] must learn to control his fears and anxiety by looking for signs of danger
in the places and people he approaches; he must learn to examine people for signs of
resistance, flight and threat, to limit their chances of hurting him or creating situations he can-
not control or can control only with the use of force, which is inappropriate to the circum-
stances. . . . He must accept and welcome the fact that, as a policeman, he must be in control
of the situation lest it be in control of him (75).

Thus, although political scientists and sociologists have hotly debated exactly what
political elites perceive to be threatening,6 sociolegal scholars have been able to pin-point one
major perceived threat by police officers: the loss of control. By focusing on common
institutional concerns within policing, we reconfigure the theoretical impulse of (elite) threat
theorists. Instead of studying threats to political elites, the blue approach examines what
police agencies and officers are likely to find threatening. We argue that this peculiar insti-
tutional characteristic of the police structures protest control.
More specifically, the blue approach argues that while elites may be concerned about
more diffuse threats, such as the articulation of revolutionary goals by a protest group or
movement, the police are more concerned with situational threats that indicate that they may
lose control of a community or a crowd (or have already begun to do so).7 This insight reson-
ates with other recent work examining the effects of police-protester interactions on police
action at protests (McPhail and McCarthy 1998; McCarthy, Martin, and McPhail 2005).
In order to create a viable analytic approach to studying repression, however, situational
indicators of actual or possible loss of control must be identified. We argue that several
indicators are available from the existing literature. First, ethnographic studies of protest
policing have shown that police officers find the presence of counterdemonstrators sub-
stantially increases the potential for conflict at protest events (Waddington 1994). The logic of
this insight is intuitively clear: it is not just that counterdemonstrators passively disagree with
the political claims of protesters. Rather, counterdemonstrators often vehemently disagree
with the political claims of protesters and actively seek to thwart opposing mobilizations
(Meyer and Staggenborg 1996). Protests that bring together groups diametrically opposed to
one another would seem, on average, much more difficult to control because of the potential
for explosive interactions.
We argue, therefore, that the presence of counterdemonstrators should (1) increase the
likelihood of police presence, as police seek to ensure control, and (2) increase the likelihood
of police action, as police who are present on the scene seek to maintain control over the
demonstrators and counterdemonstrators. This is clearly a story about police threat, not pol-
itical elite threat, since political elites should prefer counterdemonstrator presence. For elites,
150 Mobilization

counterdemonstrators show wider audiences that multiple points of view exist and that some
“average citizens” support political elite policies.
Second, particular types of provocative actions on the part of a crowd signal to police
dispatchers that a crowd requires police supervision, or, in the presence of police, signal that
the crowd is becoming unmanageable.8 Specifically, a dispatcher who receives a report of
property damage or violence should realize that a police deployment is necessary. Likewise,
when police are already present, police will recognize that the crowd is no longer under con-
trol when protesters damage property or use violence.
Critical readers may object to this as a police threat, arguing that police in such a situ-
ation would be responding to the illegality of violence and property damage rather than to the
threat of losing control. This alternative perspective, however, does not fit with research on
policing generally, or protest policing specifically. A substantial amount of prior research has
demonstrated that police do not react automatically to all instances of law breaking (Black
and Reiss 1970; Bittner 1967a; Clark and Sykes 1974; Lundman 1974; Peterson 1971;
Piliavin and Briar 1964; Wilson 1968), and Waddington’s (1994, 1998) and della Porta’s
work (1998; della Porta and Reiter 1998) support this claim where protests are concerned.
Instead, enforcement usually relates to other goals or concerns of the officer (Bittner 1967a,
1967b; Hartjen 1972; Lundman 1974, 1980; Pastor 1978; Wilson 1968). As Bittner notes:

It is the rare exception that the law is invoked merely because the specifications of the law are
met. . . . It could be said that patrolmen do not really enforce the law, even when they do in-
voke it, but merely use it as a resource to solve certain pressing practical problems in keeping
the peace. (1967a: 710)

Legal violations such as property damage and physical violence may indeed signal both law
breaking and the loss of control, but, based on the extensive policing research on enforcement,
we argue that it is the loss of control signal that actually cues police action at protest events.
Finally, since police cues signaling losses of control are socially produced, some his-
torical variability may exist. In the period examined here, we argue that “missile” throwing
(e.g., rock, bottle, brick, etc.) was a particularly salient cue. The Kerner Commission (1968)
argued that the use of missiles by crowds was an important common factor among crowds that
escalated into rioting mobs. Police were taught to watch for and react to this signal.9 We ex-
pect that missile throwing will significantly affect protest policing net of property damage and
violence.

Organizational Effects

Just as police agencies have theoretically meaningful similarities, there are also
differences between police departments, such as varying departmental concern for and levels
of professionalism, organizational structure, and size. In addition to taking core institutional
characteristics of policing seriously, a police-centered approach should address differences
between law enforcement agencies.
We seek to incorporate these sources of organizational variation into our police-centered
approach to protest policing. Central among the organizational characteristics we are con-
cerned with are staffing levels. That is, we expect that departments with larger staffs will be
able to devote more police-hours to the control of protest without compromising other police
functions (i.e., patrol and emergency response), and that these departments will have more
tactical options available for protest policing because of higher staffing levels. We also expect
that departments with more resources will have more access to training and equipment.
We expect that varying levels of departmental professionalism should affect protest
policing. According to Earl’s (2002) study of the development of protest-policing protocols
from 1960 to 1980 in the U.S., the growth of police professionalism, particularly in the late
1960s and thereafter, emphasized the need to reduce uses of force in protest situations.10
Seeing Blue 151

Given the emphasis on training in professional models of policing, we also expect that more
professional departments will be better trained to handle protest.
Taken together, we argue that institutional features of policing—the emphasis on control
and institutional constructions of how control can be achieved and maintained—and organi-
zational variations between police agencies—including staffing levels and professionalism—
structure protest policing. In the next section we introduce the data and methods that will be
used to compare the political elite threat, weakness, and blue approaches.

DATA AND METHODS

This article analyzes data from events that occurred in New York State from 1968 to 1973. In
order to test the organizational aspects of the approach, we chose a location that allowed
enough variation along organizational dimensions while being limited enough geographically
that organizational data could be collected. New York State fulfills both requirements since
police agencies in the state vary widely in organizational characteristics but it was feasible to
collect county-level data about police organizational characteristics. Further, as Earl et al. (2003)
note, New York during this period was “a microcosm of police-protester and police-public
interactions” (587) and a microcosm of developments within policing (Alex 1976; Astor
1971; Lardner and Reppetto 2000; Niederhoffer 1969; Silverman 1999).

Gathering Data on Protest Events and Protest Policing

Data on collective action events were collected from daily editions of the New York
Times as part of a larger research project. Our analysis includes protest events that occurred
within New York State between 1968 and 1973 and which were reported upon in the New
York Times.11 In order to be included in our analysis, protest events had to meet several other
criteria: (1) more than one person had to participate in an event since our concern is with
collective action; (2) participants must have articulated some claim, whether it be a grievance
or an expression of support; and (3) the event must have happened in the public sphere. Of the
events that meet those criteria, we exclude several types of events from analysis: (1) those that
would not, under normal circumstances, be likely to experience police presence, such as
letter-writing campaigns, lawsuits, and press conferences; (2) those that took place in total
institutions (Goffman 1984)—such as prisons, jails, and mental institutions—since the dy-
namics of protest policing in such institutions are likely to differ significantly from protest
policing in less controlled environments; and (3) labor-related events since the dynamics of
labor-police interactions and the length of many labor actions (e.g., strikes) may substantially
differ from nonlabor protest events. In the end, our data set included 1,905 events.12

Data Collection and Coding

In choosing newspaper reporting as a data source, we follow a rich history in the field of
social movement research. While the merits of this method far outweigh the problems
associated with it, the research design of this project attempted to deal with some of the fre-
quently noted problems associated with newspaper data (see Earl, Martin, Soule, and
McCarthy 2004 for a summary). For instance, two of the largest problems that the project
team avoided are that, unlike many prior studies using newspapers as a source of data on
protest events, the project team did not use an index of the New York Times to identify events
nor did they sample days of the newspaper. Instead, researchers skimmed daily editions of the
newspaper between 1968 and 1973 and identified all collective action events that were
reported. Research assistants then content coded these events, achieving intercoder reliability
rates that were consistently at or above 90% agreement.13
152 Mobilization

Modeling Police Presence and Action

Using this rich data source, we empirically compare the above approaches using a series
of logistic and multinomial logistic regressions, following recent developments in protest
policing research.14 As Earl et al. (2003) argue, this modeling strategy better conceptualizes
protest policing, casting it as a two-step process in which police first decide whether or not to
attend a protest and then decide how to respond to the event only after arriving at it (Tilly
1978). The first stage of this process—police attendance (or lack thereof) at a protest event—
is modeled using logistic regressions where the dependent variable is the log odds of police
presence at a protest event (Long 1997). The logistic regression models use the entire sample
discussed above.
When examining the second stage of this process—police action at protest events—the
sample is limited to events at which police were present. In operationalizing the dependent
variable, we follow Earl et al. (2003) and use a five-category dependent variable constructed
to capture combinations of police tactics used at protest events. This modeling strategy better
represents the combinatorial logic of police action. Since police empirically tend to make
interrelated decisions about how to respond to protest—not unrelated, discrete, yes-no
decisions about certain actions as implied by the use of logistic regressions—modeling needs
to be sensitive to combinations of actions.
The first approach (and first category of the dependent variable) that we examine is the
“Do Nothing” approach. In this approach, officers show up at a protest event but take no
further action. Second, there is the “Nothing to See Here” approach in which officers show up
and take only limited action (which excludes making arrests or using force). Third, in the
“Ounce of Prevention/Legal Eagles,” officers attempt to prevent disorder by erecting bar-
ricades, reacting to protest by making arrests, or combining the use of barricades with arrests.
The fourth response, the so-called “Dirty Harry” strategy, exclusively uses force (including
hand-to-hand conflict and/or the use of weapons).15 In the final category, the “Calling All
Cars” approach, officers combine both force and arrests/prevention. We use multinomial
logistic regression to examine this five-category dependent variable.16

Independent Variables

Variables operationalizing elite threat, weakness-from-within and weakness-from-

without, as well as control variables, are summarized in table 1.17 Control variables are inclu-
ded in all models (except where explicitly indicated), although results for these variables are
not reported. Our measures follow existing research in measuring key concepts, and data are
directly drawn from the protest event data set described above.
Because we are testing the blue approach for the first time here, it is important to discuss
in detail the operationalization of these measures. In order to operationalize the police threat
concepts laid out above, we employ several dichotomous variables. We capture the potential
for conflict through a dummy variable turned on by the presence of counterdemonstrators. We
expect that police will be more likely to attend events in order to monitor them when
counterdemonstrators are present, but we also expect that the police will be more likely to
take action at such events. We also include four dummy variables to capture cues that a
situation is quickly getting out of control: (1) protester-committed property damage, (2)
protester-committed violence, (3) the co-commission of property damage and violence by
protesters,18 and (4) rock, brick, and bottle throwing.
To operationalize organizational characteristics that vary between police agencies, we
employ several geographically specific measures. We include a measure of police staffing
capacity equal to per capita police spending at the county level.19 We also include several
measures for police professionalism.20 Because higher education and technical training have
been critical to police professionalism, we include a time-sensitive count of the number of
Seeing Blue 153

Table 1. Nonpolicing Independent Variables and Control Variables Summarized

Concept Measure Notes or Examples

Political Elite Threat
Protest Size Logged number of participants Logged to address heteroskedasticity
Confrontational Dummy turned on for confrontational For example, sit-ins, office takeovers, and
Tactics tactic use meeting disruptions. See the appendix for
a full listing of confrontational tactics
Radical Goals Dummy turned on for advocating radi- For example, racial and ethnic-power
cal goals claims, communist claims, and pro-gay
rights claims. See the appendix for a full
listing of radical goals
Number of Goals Count of the number of protester goals/
claims
Weakness from Within
Minority Presence Dummy variable turned on when pro- For example, black protesters; see the
testers from subordinated groups are appendix for a full listing of participant
present groups that turn on this variable
College Student Dummy variable turned on when Depending on their age and the year of the
Presence college students protest protest, many college students either could
not vote or did not vote in large numbers
SMO Presence Dummy variable turned on when SMO
was present
Number of SMOs Count of the number of SMOs that were
present
Weakness from Without
Front Page News Count of protest events discussed in Front page stories are much higher profile
Coverage front page NYT stories in the prior and therefore much more important to
month external monitoring
Control Variables
Political Climate % vote for Democratic nominee in the Data taken from ICPSR (1995)
most recent presidential election
Antipathy towards Dummy variable turned on when event Included to control for potentially harsh
Civil Rights was a civil rights movement protest policing towards the African-American
civil rights movement
Duration of Risk Logged duration of the protest event Longer events allow more time for police
to arrive and to take action. Variable is
logged to address heteroskedasticity
Detail in NYT Logged count of paragraphs in news- Controls for the level of detail that might
Reporting paper report on protest event result from different lengths of newspaper
reports on events. Variable is logged to
address heteroskedasticity
Location Fixed effects for New York City, Long Included to address autocorrelation that
Island, Westchester County, and Albany could arise from repeated encounters with
(the state capital) protesters in major protest centers
Location in Public Dummy variable turned on when protest Public forum law, which addresses police
Forum occurs on private property powers based on protest location (McPhail
et al. 1998), allows police the most leeway
when protests occur on private property

police studies programs in each county (Earl 2002). In terms of professional associations, we
include two dummy indicators of police professionalism at the leadership level of organi-
zations. One dummy variable is set to 1 when a chief of police or sheriff in the county where
the protest occurs is named president of the International Association of Chiefs of Police
(IACP).21 The other dummy variable is turned on when a chief of police or sheriff in the
154 Mobilization

county where the protest occurs is named president of the New York State Chiefs of Police
Association. Both measures are derived from reports in the IACP’s monthly magazine Police
Chief. We also measure professionalism at the line-level by counting the number of IACP
police members who reside in a county. This measure was compiled from annual membership
lists published in Police Chief. Although this measure may have some error because some
police departments allow officers to reside outside of the locale they serve, we expect that this
was unusual, particularly given that we measure residence at the county, instead of municipal,
level.

EXPLAINING POLICE PRESENCE

Table 2 presents results from logistic regression analysis modeling police presence at public
protest events. The full model includes all theoretically relevant and control variables while
the reduced model includes only theoretically relevant variables that significantly affect the
likelihood of police presence at a protest event at the .10 level and all control variables.
The results indicate support for most political elite threat variables. As expected, police
are more likely to attend and monitor events with larger numbers of participants, events where
confrontational protest tactics are deployed, and events where protesters support radical
claims. Contrary to expectations, the number of claims that protesters make, which is a proxy
for the breadth of change desired by protesters, does not affect the likelihood of police
presence. It is interesting to note that of the three political elite threat variables that are
significant and in the expected direction, two tap more situational concerns: the size of the
protest and the kinds of tactics used by protesters.
However, table 2 suggests that political elites are not able to translate opportunities to
repress weak targets into even police monitoring of those groups. In fact, only two measures
of weakness are significant and both are in the opposite direction than expected by weakness
theorists. Despite the suspected weakness of college students, they were protected from police
monitoring. According to the weakness approach, protests with SMOs present should be less
likely to be policed because of the protective strength provided by SMOs, but the opposite is
true. Since this finding is net of protest size, it is not the case that SMO-sponsored events
attract more protesters, and hence police presence. Other research also found these contrary
effects (Earl et al. 2003).
As predicted by our blue approach, institutional concerns over control independently
structure protest policing. All police threat measures are significant and in the expected
directions. Counterdemonstrator presence, protester initiated violence and/or property dam-
age, and missile throwing all increase the likelihood of police presence. The counter-
demonstrator finding is particularly impressive because it is net of protest size, the tactics
used, and the presence of SMOs. In other words, it is not simply the case that events with
counterdemonstrators are larger, use confrontational tactics, and are likely to be organized by
SMOs. This is a strong indication that police do construct counterdemonstrator presence as a
threat to order and control and thus seek to monitor protest events where counterdemon-
strators are present. The findings regarding property damage and/or violence demonstrate that
the police attend protests that involve clear shows of disorder. Finally, the independent effect
of missile throwing, which is net of property damage and violence occurring at events, de-
monstrates that police are cued by certain socially constructed signals of escalating disorder.
The blue approach also suggests that variation within institutional fields exists such that
organizations differ on certain characteristics such as resources and professionalism and that
this kind of interorganizational variation affects the likelihood of police presence at protest
events. Our findings are weaker on this point. Although we anticipated that personnel re-
sources would affect police presence at protest events, per capita spending on policing does
not affect the likelihood of police presence. One way to understand this unexpected finding is
 Seeing Blue 155

Table 2. Modeling Police Presence, New York State 1968-1973

Full Reduced
Political Elite Threat
Logged Number of Participants 0.164** 0.168**
(0.035) (0.035)
Confrontational Tactics 1.665** 1.656**
(0.156) (0.154)
Radical Goals 0.556** 0.579**
(0.143) (0.140)
Number of Goals 0.054
(0.076)
Weakness
Minority Presence 0.168
(0.143)
College Student Presence -0.462** -0.479**
(0.168) (0.165)
SMO Presence 0.311* 0.392**
(0.134) (0.123)
Number of SMOs 0.027
(0.032)
Front Page News Coverage -0.012
(0.009)
Police Threat
Counterdemonstrator Presence 1.255** 1.205**
(0.283) (0.279)
Protester Initiated Violence 0.539** 0.556**
(0.184) (0.182)
Property Damage 1.336** 1.347**
(0.331) (0.331)
Protester Initiated Violence and Property Damage 0.782** 0.737**
(0.249) (0.245)
Bottle and Brick Throwing 2.099** 2.133**
(0.395) (0.391)
Organizational Characteristics
Per Capita Police Force Spending -4.528
(6.097)
Number of Police Studies Programs in County -0.180
(0.203)
IACP President 0.026
(0.336)
NYACP President 0.834+ 0.808+
(0.485) (0.484)
IACP Members in County 0.009* 0.005*
(0.004) (0.002)
Constant -5.663** -6.376**
(1.164) (1.089)
Log Likelihood -869.58 -873.72
Observations 1859 1859
Note: Standard errors in parentheses, + significant at 10%; * significant at 5%; ** significant at 1%
156 Mobilization

that police respond to protests that require monitoring whether or not that stretches their
agency thin. It is reasonable to expect during a period in which protest is intense, as is the
case in the period we study, that police might seek to monitor all protests that meet certain
characteristics—such as the characteristics we model through measures like protest size and
counterdemonstrator presence—and they do so despite the hardships it causes their depart-
ments in terms of staffing other competing concerns.
The results for the agency-level professionalism measures are mixed. The number of
police studies programs in a county does not affect the likelihood of police presence. Whether
the police chief or sheriff of a department has been a recent president of the International
Association of Chiefs of Police (IACP) does not affect the likelihood of police presence
either. However, if the police chief or sheriff of a department had been a recent president of
the New York Association for Chiefs of Police, the likelihood of police presence increases at
the .10 significance level. This is a weaker significance level than all other coefficients in the
reduced model and so must be interpreted cautiously. Higher numbers of IACP members in a
county, which also taps the level of police professionalism, increases the likelihood of police
presence as predicted. Thus, although resources did not affect police presence, findings for
some of the professionalism measures suggest that more professionally integrated depart-
ments are more likely to send officers to public protest events.22

EXPLAINING POLICE ACTION AT PROTEST EVENTS

The results for the police action models are similar, but not identical, to the patterns just
discussed. To analyze police action, we limit the sample to events at which police were
present and then use a multinomial logistic regression to model the likelihood of deploying
each of five possible police responses to protest.23 Table 3 contains predicted probabilities for
the five police approaches to protests.24 The first row of the table contains a base probability
for each police response to protest, which is computed with all dummy variables set to zero
and all continuous variables at their means. For example, at an event where all dummy
variables are “off” and continuous variables are at their mean, there is a 41% chance that the
police would deploy a Do Nothing approach if present, but only a 5% chance of deploying a
Calling All Cars approach. Subsequent rows in the table contain the predicted probability for
each police response given either a 0/1 change for dummy variables or a designated change in
percentile value for continuous variables. For instance, while the hypothetical event just
mentioned has a 41% chance that the police would deploy a Do Nothing approach, the
probability of a Do Nothing approach drops to 23% if protesters use confrontational tactics.
Similarly, while there is a 32% probability of using a legal approach at an event in the 10th
percentile of protest size, that probability drops by 14 percentage points to 18% if the protest
size increases to the 90th percentile. Finally, police responses whose probabilities are
significantly altered by the row variable are bolded. Readers should recall that multinomial
models actually produce coefficient estimates for all dependent variable categories (i.e., a 5-
category dependent variable will generate four coefficients and associated statistics for each
variable). Thus, although the likelihood of deploying one police strategy may not be affected
by a variable, the likelihoods of other strategies may be significantly affected. The bolded
probabilities indicate significant changes in likelihood.
As table 3 demonstrates, only two political elite threat variables significantly affect police
action at protest events: increasing protest sizes and the use of confrontational tactics. The use
Seeing Blue 157

Table 3. Predicted Probabilities for Police Action, New York State 1968-1973a

Do Nothing to Legal Calling

Nothing See Here Eagle Dirty Harry All Cars
c b b b b b
Base Probability 41 29 25 approaches 0 5
Political Elite Threat
Protest Size – 10th percentile 32 33 32 approaches 0 4
th
Protest Size – 25 percentile 34 32 30 approaches 0 4
Protest Size – 75th percentile 45 27 23 approaches 0 5
Protest Size – 90th percentile 54 22 18 approaches 0 6
Confrontational Tactics 23 36 33 approaches 0 7
Weakness
College Student Presence 54 24 20 approaches 0 2
Front Page News Coverage – 10th
percentile 50 21 24 approaches 0 4
th
Front Page News Coverage – 25
percentile 48 23 25 approaches 0 4
Front Page News Coverage – 75th
percentile 39 30 26 approaches 0 5
Front Page News Coverage – 90th
percentile 29 41 25 approaches 0 5
Police Threat
Counterdemonstrator Presence 31 36 26 1 6
Protester Initiated Violence 23 34 20 1 20
Protester Initiated Violence and
Property Damage 27 32 22 approaches 0 18
Bottle and Brick Throwing 23 25 38 1 13
Organizational Characteristics
Per Capita Police Force Spending –
10th percentile 39 26 30 approaches 0 5
Per Capita Police Force Spending –
25th percentile 41 27 27 approaches 0 5
Per Capita Police Force Spending –
75th percentile 42 30 23 approaches 0 4
Per Capita Police Force Spending –
90th percentile 43 33 21 approaches 0 4
a
Notes: Probabilities may not sum to 100% due to rounding error.
b
Bolded values indicate a significant change in this probability occurs when the row variable is manipulated.
c
The base probability reflects the probability of each outcome with dummy variables set to 0 and continuous
variables at their means.

of confrontational tactics significantly decreases the probability of using the Do Nothing

approach from a 41% to 23% chance (and consequently increases the probability of all other
police responses to protest). By contrast, larger protest sizes significantly increase the prob-
ability of either Do Nothing or taking action that involves force (i.e., deploying either a Dirty
Harry or a Calling All Cars approach) at a protest event by decreasing the likelihoods of
Nothing to See Here and Legal Eagle approaches. That is, larger protest sizes bifurcate police
responses into either simple presence with no action or serious action involving force. The
increase in the probability of using force is understandable from a political elite threat van-
tage. However, increasing probabilities for Do Nothing is counter to political elite threat
158 Mobilization

claims: one would expect that as protest size increases, so too would the threat posed to
political elites. If police forces directly translate that sense of elite threat into police action
then the probability of the police taking at least some action should increase (i.e., do
something other than Do Nothing), which is not the case here.
This finding is explicable, though, with reference to police threat concepts. As crowd
sizes increase, maintaining control over a crowd can become more difficult. As well, inter-
acting with large crowds can actually ignite contention and make conflict between police and
protesters much more likely. This may lead the police to resist interacting with crowds until it
is vitally necessary. Thus, police would simply monitor protest (i.e., Do Nothing) until a
crowd was beginning to get seriously out of control. At that point, limited action or purely
nonviolent forms of protest control (e.g., Nothing to See Here or Legal Eagle deployments)
may no longer be viable, leading to an increase in the likelihood of using force. This scenario
can explain why police action would be bifurcated into Do Nothing and force-based stra-
tegies, but only by examining the protest situation from the police perspective as opposed to
an elite perspective.
The weakness approach performs poorly in the police action models. Only two weakness
variables are significant—college student presence and total front-page news coverage from
the prior month—although each variable is in the opposite direction than predicted by weak-
ness theorists. The presence of college students decreases the probability of police using a
Calling All Cars response, reducing the probability by over half from 1 in 20 protests to 1 in
50 protests. Thus, just as is the case in police presence models, college student presence un-
expectedly dampens police action instead of opportunistically buttressing it. Total front-page
news coverage is also significant and serves to decrease the probability of a Do Nothing
approach by police. That is, increased front-page news coverage increases the probability that
the police will take at least some action at protest events, contrary to expectations of weakness
theorists.
The police threat variables perform well in the police action models. Counterdemon-
strator presence increases the probability of deploying a Dirty Harry approach. Although the
substantive size of this effect is not large—the probability only rises to 1 in 100 protests—it is
nonetheless a fascinating result. Without a police threat explanation in mind, the presence of
counterdemonstrators would not appear to be an important predictor of protest policing. But,
by using a police-centered threat approach, we are able to partially explain the deployment of
the police approach to protest that most closely approximates Stark’s (1972) police riots.
Although property damage alone does not significantly impact police action, violence
does affect police action. Whether coupled with property damage or not, the use of violence
by protesters increases the probability of using a force-based police strategy such as the Dirty
Harry or Calling All Cars approaches. The use of missiles, net of violence, property damage,
and other protest characteristics, also increases the probability of using a force-based police
strategy. These effects are as predicted.
In terms of organizational measures, only police capacity significantly affects police
action at protest events. Specifically, wealthier departments are less likely to use a Dirty
Harry approach. This finding makes a great deal of intuitive sense because wealthier depart-
ments would be better able to provide their officers with adequate training. Adequate levels of
well-trained manpower may increase the chances that police are able to react in legal ways
and without engaging in a police riot (Stark 1972).

DISCUSSION AND CONCLUSION

In response to the current lack of coherent theoretical explanation of protest policing that
focuses on the independent role of law enforcement in structuring protest policing, we de-
velop and test a police-centered, or blue, approach to protest policing. This approach argues
Seeing Blue 159

that the peculiar institutional characteristics of the police and interorganizational variation
between police agencies on key dimensions such as staffing and professionalism should
independently structure the dynamics of protest control.
In order to empirically compare the explanatory power of our blue approach to the
political elite threat and weakness approaches, we examine police presence and police action
at protest events that occurred in New York State from 1968 to 1973. We find overwhelming
support for a blue, or police-centered, approach to protest policing in the U.S. Our findings
are particularly strong where the blue approach focuses on key institutional features of
policing, such as the way in which threat is constructed within police agencies and perceived
by officers involved in actual policing operations. More mixed support was found for
organizational variation, although professionalism did somewhat affect police presence and
the resources available to police did affect police action.
Critical readers may be tempted to offer another reading of our results, suggesting that
the measures we label as police threat variables are actually capturing other, new types of
threats to political elites. For example, perhaps elites are likely to be concerned with violence
and property damage. Critical readers may argue, then, that police are acting on the behalf of
political elites when they react to violence and property damage.
We take issue with this alternative interpretation. First, our analyses go beyond modeling
shared concerns. For instance, the statistical significance and substantive importance of
counterdemonstrator presence to explanations of both police presence and police action
directly contradicts an exclusively elite threat interpretation of the model results. As we note
above, political elites should actually prefer counterdemonstrator presence at protests;
counterdemonstrators often support the state’s position on a matter, serve to suggest that pro-
testers do not necessarily represent a majority viewpoint, and often attempt to delegitimate the
movements they oppose. Earl (2004) suggests that, in some instances, counterdemonstrators
or countermovements more generally may even function as protest control agents. However,
qualitative studies of protest policing show that police do not view counterdemonstrators in
such a favorable light (Waddington 1994). Police view counterdemonstrators as adding to the
likelihood that the police could lose control of a situation. As our analyses indicate, this
increasing threat to police control increases the likelihood that police will attend a protest,
thus allowing police to monitor the protest, and increases the probability that police will use a
Dirty Harry approach.
Second, our findings with respect to missile throwing suggest that this kind of aggressive
behavior was particularly salient in explaining protest policing. This finding was net of the
property damage and violence involved at such protests events and thus cannot be explained
by a particular concern for illegality. Further, it is unclear why political elites would have a
particular interest in controlling missile throwing. However, it is clear that missile throwing
directly threatens officer safety and was, during the period we examine, widely argued to be
an indicator that police control was either lost or nearly lost over a situation (Kerner 1968).
Third, with respect to elite threat versus police threat, one could easily argue that the
importance of confrontational tactics in the analysis actually owes to the problems these
tactics pose to police, not the threat they represent to political elite power. It is certainly the
case that police face more immediate trouble in dealing with confrontational protests than
political elites do. Further, extremely radical and confrontational tactics may actually aid
political elites in the long run since these tactics can at times estrange movement groups from
a mainstream base of support. In this way, perhaps it is the case that some supportive elite
threat findings actually owe to the threat posed to police.
This does not suggest that prior research focusing on political elite concerns is without
any merit. Our analyses find support for political elite interests. For instance, we find that
radical goals, which should not matter much to police, affect police presence. Nonetheless, we
do not find support for opportunistic action against “weaker” protests.
Future research should address how generalizable the blue approach is to other types of
160 Mobilization

repression, law enforcement agencies, or even military regimes. For instance, Cunningham’s
(2004) work suggests that the organizational design of a federal law enforcement agency
affected how that agency handled repression. Davenport (1995) and Stanley (1996) find insti-
tutional and organizational effects in nondemocratic states and militarized repressive agen-
cies, although their work departs from other research. Future research should also attempt to
understand how the threats police attend to are organizationally constructed within agencies
and policing more generally and how these organizational productions change over time.

APPENDIX

Confrontational Tactics. We consider the following forms of protest to be confrontational:

Civil disobedience, physical attacks, riots, mob violence, strikes, slow-downs, sick-ins
and other conflicts. We consider the following activities at protest events to be con-
frontational: general civil disobedience, sit-ins and derivatives of sit-ins (e.g. shop-ins,
penny-ins, etc.) physical attacks, verbal attacks or threats, blockades by protesters, build-
ing take-overs, looting, damaging property, kidnapping, and meeting disruptions.

Radical Goals. We coded the following goals articulated at protest events as radical:
For: comparable worth, ERA, socialism, communism, fascism, welfare, freedom of
speech, Affirmative Action, minority political power, minority culture or pride, Black
separatism, minority extremism, gay rights, Native American rights, and farm worker
rights.
Against: discrimination in employment for any minority group, the Vietnam war,
imperialism, capitalism, the U.S. government, ROTC and campus military recruitment,
government surveillance of protesters, the current status of minorities in America, socio-
cultural ghettos, police brutality (particularly against minorities), and the under-
representation of minorities in the media.

Minority Groups
The following groups are considered minority groups in our analysis: the unemployed,
the poor, gays and lesbians, the homeless, Muslims, and racial and ethnic groups except
for undifferentiated whites, Italians, other European nationalities, and the Jewish.
Seeing Blue 161

NOTES
1
We use the term “institutional” to refer to characteristics shared by policing, while we use “organizational” to refer
to characteristics that vary between police agencies. This distinction is elaborated below.
2
A substantial literature on crossnational variation in repression exists (see Davenport 2000b and Wisler and Kriesi
1998). Because we are interested in the allocation of repression within a given state and time period, not in cross-
national variation in aggregate repressiveness, we do not address this work further.
3
Stockdill (1996) suggests a “threat-weakness” interaction, but recent tests of this claim find no support (Earl et al.
2003) and thus we do not consider this interaction further.
4
We use the term “blue” because in American culture this color has a cultural association with policing, perhaps
because it is a common uniform color for American police.
5
This research also implies that it would be inappropriate to view concerns over danger as simple “survival instincts”
since the dangers that police focus on and the routines they use to identify these dangers are socially constructed
through institutional and organizational processes.
6
In addition to the scholarship reviewed on “objective” threats to elites, research has attempted to complicate the de-
finition of a threat, arguing that subjective understanding of threats to elites (or “threat perception” in that literature),
not objective threats, are important (e.g., see Boudreau 2005).
7
We acknowledge that police and political elites have some shared interests in public order, but police also have
concerns that political elites do not share and viceversa. Since we discuss this later in the paper, it sufficient to note
now that one could imagine police and political elite concerns as a Venn diagram in which the police and political
elites have some overlapping concerns but many distinct concerns.
8
We discuss dispatchers because police can be dispatched to a scene in reaction to a “call for service,” as when a
caller dials an emergency operator who dispatches police.
9
Police dispatchers would also be expected to react to this cue, sending police officers to situations they feared could
escalate to riots based on reports of missile throwing.
10
It is also possible that more professional departments are also more authoritarian (Brown 1981). Our analysis can
handle this alternative: coefficients would suggest that professionalism significantly increased the probability of
substantial police action.
11
In addition to NYS being a theoretically informed location, focusing on this state has one additional, and sig-
nificant, empirical advantage. It limits one of the strongest sources of potential bias in newspaper data coverage of
protest events: proximity of the protest event to the news source (Earl et al. 2004; McCarthy et al. 1996).
12
As indicated the tables of results, our logistic regressions use slightly fewer cases—1,859 protest events instead of
1,905—due to missing data on key variables.
13
For other work using these data, see Soule and Earl (2005); Earl and Soule (2006); McAdam and Su (2002); Van
Dyke, Soule and Taylor (2004); and Earl et al. (2003).
14
Interested readers should consult Earl et al. (2003) for additional advantages of this modeling approach, particularly
in comparison to other common alternatives.
15
We recognize alternative characterizations of “Dirty Harry” (e.g., Klocklars 1980). However, we use this
terminology to reflect the extent to which police were willing to use violence alone to control protest. Since violence
should never be used outside of the arrest of a subject, this category is similar to Stark’s (1972) “police riot.”
16
Given recent research comparing arrests and police violence (Earl 2005), we understand these categories to be
distinct but not ordered. However, even if the categories of police action were strictly ordinal, the Parallel Regression
Assumption required for ordinal logistic regression models would be violated, making multinomial logistic regression
the preferred technique (Long 1997).
17
We included one variable to capture the broad political climate of the county in which protest occurred, but we did
not include other measures of political opportunities because research using the same data has shown no effect for
such variables (Earl 2002; Earl and Soule 2006).
18
Critical readers may suggest that the police, at times, provoke violent actions or property damage by protesters.
This would not affect our findings because police are unlikely to see themselves as provoking those actions and are
also unlikely to change their reaction to violence or property damage even if they believe that their actions
encouraged protester violence and property damage. Instead, police agencies might discourage action that would
escalate the intensity of protests, but then allow significant police action once protests became violent or damaged
property (Earl 2002).
19
We use county level data because city level data for most police organizational variables is unavailable despite
extensive efforts to collect such data. Although this limits our analysis in some ways, the prevalence of mutual
assistance agreements that allowed intracounty cooperation for large protests, riots, and other emergencies renders the
county a theoretically viable level for measurement (Earl 2002).
20
Future research should test alternative specifications of police professionalism since this period of policing history
was characterized by multiple visions of professionalism.
21
We use IACP Presidential status as an indicator because it was the leading organization touting police
professionalism in this period (Earl 2002). Francis B. Looney from New York City was IACP President from 1973 to
1974 and George A. Murphy from Oneida, NY was IACP President from 1971 to 1972.
22
It is possible that some of these organizational characteristics did not significantly affect police presence because
interagency differences are “hidden” by the fixed effects for major centers of protest (New York City, Long Island,
Westchester, and Albany) that are included in each model as control variables. That is, since the fixed effects and
162 Mobilization

important police agency characteristics were colinear, many of the interagency effects could have been muted by in-
cluding the fixed effects control variables.
23
The final model has a negative log-likelihood score of -718.948, an N=574, and passes a Hausman test (satisfying
the assumption of the independence of irrelevant alternatives).
24
The probabilities are generated in Stata using the SPost Suite (Long and Freese 2001), as are all model fit scores.
Predicted probabilities are presented in lieu of full model results because of the lengthy and cumbersome presentation
format of actual model coefficients. The probabilities presented are based on a reduced model that excludes non-
significant coefficients (except the fixed effects for location, which correct for spatial autocorrelation). The reduced
model is used so that nonsignificant coefficients cannot affect predicted probabilities. Variables are excluded from
the full model if no coefficient associated with the variable was significant at the .05 level (recalling that multinomial
logistic regressions produce coefficient estimates for N-1 dependent variable categories).

REFERENCES

Alex, Nicolas. 1976. New York Cops Talk Back. New York: John Wiley and Sons.
Astor, Gerald. 1971. The New York Cops. New York: Charles Scribner's Sons.
Balch, Robert W. 1972. “The Police Personality.” Journal of Criminal Law, Criminology, and Police
Science 63: 106-119.
Ball, Patrick. 2005. “On the Quantification of Horror.” Pp. 189-208 in Repression and Mobilization,
Christian Davenport, Hank Johnston and Carol Mueller, eds. Minneapolis: University of Minnesota
Press.
Bittner, Egon. 1967a. “The Police on Skid-Row.” American Sociological Review 32: 699-715.
———. 1967b. “Police Discretion in the Emergency Apprehension of Mentally Ill Persons.” Social
Problems 14: 278-292.
Black, Donald, and Albert J. Reiss. 1970. “Police Control of Juveniles.” American Sociological Review
35: 63-77.
Boudreau, Vince. 2005. “Precarious Regimes and Matchup Problems in the Explanation of Repressive
Policy.” Pp. 33-57 in Repression and Mobilization, Christian Davenport, Hank Johnston and Carol
Mueller, eds. Minneapolis: University of Minnesota Press.
Bromley, David G., and Anson D. Shupe, Jr. 1983. “Repression and the Decline of Social Movements.”
Pp. 335-347 in Social Movements of the Sixties and Seventies, Jo Freeman, ed. New York:
Longman.
Brown, Michael K. 1981. Working the Street. New York: Russell Sage Foundation.
Clark, John P. and Richard Sykes. 1974. “Some Determinants of Police Organization and Practice in a
Modern Industrial Democracy.” Pp. 455-494 in Handbook of Criminology, Daniel Glaser, ed.
Chicago: Rand McNally.
Cunningham, David. 2003a. “State versus Social Movement.” Pp. 45-77 in States, Parties, and Social
Movements, Jack A. Goldstone, ed. New York: Cambridge University Press.
———. 2003b. “Understanding State Responses to Left- versus Right-Wing Threats.” Social Science
History 27: 327-370.
———. 2003c. “The Patterning of Repression.” Social Forces 82: 209-240.
———. 2004. There's Something Happening Here. Berkeley: University of California Press.
Davenport, Christian. 1995. “Assessing the Military's Influence on Political Repression.” Journal of
Political and Military Sociology 23: 119-144.
———. 2000a. “Introduction.” Pp. 1-24 in Paths to State Repression, Christian Davenport, ed. Lanham,
MD: Rowman & Littlefield Publishers.
———, ed. 2000b. Paths to State Repression. Lanham, MD: Rowman & Littlefield.
De Biasi, Rocco. 1998. “The Policing of Hooliganism in Italy.” Pp. 213-227 in Policing Protest,
Donatella della Porta and Herbert Reiter, eds. Minneapolis: University of Minnesota Press.
Della Porta, Donatella. 1995. Social Movements, Political Violence, and the State. New York:
Cambridge University Press.
———. 1998. “Police Knowledge and Protest Policing: Some Reflections on the Italian Case.” Pp. 228-
252 in Policing Protest, Donatella della Porta and Herbert Reiter, eds. Minneapolis: University of
Minnesota Press.
Della Porta, Donatella, and Herbert Reiter. 1998. “The Policing of Protest in Western Democracies.” Pp.
1-32 in Policing Protest, Donatella della Porta and Herbert Reiter, eds. Minneapolis: University of
Minnesota Press.
Seeing Blue 163

DiMaggio, Paul J., and Walter W. Powell. 1991. “Introduction.” Pp. 1-38 in The New Institutionalism in
Organizational Analysis, Walter W. Powell and Paul J. DiMaggio, eds. Chicago: University of Chi-
cago Press.
Duvall, Raymond D. and Michael Stohl. 1983. “Governance by Terror.” Pp. 179-219 in The Politics of
Terrorism, Michael Stohl, ed. New York: Marcel Dekker.
Earl, Jennifer. 2002. “The Banner vs. the Baton.” Ph.D. thesis, University of Arizona, Tucson, AZ.
———. 2003. “Tanks, Tear Gas, and Taxes.” Sociological Theory 21: 44-68.
———. 2004. “Controlling Protest: New Directions for Research on the Social Control of Protest.”
Research in Social Movements, Conflicts and Change, Special Issue on Authority in Contention 25:
55-83.
———. 2005. “You Can Beat the Rap, But You Can't Beat the Ride.” Research in Social Move-ments,
Conflicts and Change 26: 101-139.
Earl, Jennifer, and Sarah A. Soule. 2006. “Political Opportunities and State Repression: The Limits of
Political Elite Influence.” Unpublished manuscript.
Earl, Jennifer, Sarah A. Soule, and John D. McCarthy. 2003. “Protests Under Fire? Explaining Protest
Policing.” American Sociological Review 69: 581-606.
Earl, Jennifer, Andrew Martin, John D. McCarthy, and Sarah A. Soule. 2004. “Newspapers and Protest
Event Analysis.” Annual Review of Sociology 30: 65-80.
Francisco, Ronald A. 2005. “The Dictator's Dilemma.” Pp. 58-81 in Repression and Mobilization,
Christian Davenport, Hank Johnston and Carol Mueller, eds. Minneapolis: University of Minnesota
Press.
Gamson, William A. 1975. The Strategy of Social Protest. Homewood, IL: Dorsey Press.
Goffman, Erving. 1984. “Characteristics of Total Institutions.” Pp. 464-477 in Deviant Behavior, Delos
Kelly, ed. New York: St. Martin's Press.
Hartjen, Clayton. 1972. “Police-Citizen Encounters.” Criminology 10: 61-84.
ICPSR. 1995. “General Election Data for the United States, 1950-1990 [Computer file].” Ann Arbor,
MI: Interuniversity Consortium for Political and Social Research.
Jaime-Jiménez, Oscar, and Fernando Reinares. 1998. “The Policing of Social Protest in Spain: From
Dictatorship to Democracy.” Pp. 166-187 in Policing Protest, Donatella della Porta and Herbert
Reiter, eds. Minneapolis: University of Minnesota Press.
Kerner Commission. 1968. Kerner Commission Report. Washington, D.C: National Advisory Commis-
sion on Civil Disorders.
Klocklars, Carl B. 1980. “The Dirty Harry Problem.” The Annals of the American Academy of Political
and Social Science 452: 33-47.
Kriesi, Hanspeter. 1995. “The Political Opportunity Structure of New Social Movements.” Pp. 167-198
in The Politics of Social Protest, J. Craig Jenkins and Bert Klandermans, eds. Minneapolis:
University of Minnesota Press.
Lardner, James, and James Reppetto. 2000. NYPD. New York: Henry Holt and Company.
Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Thousand
Oaks, CA: Sage.
Long, J. Scott, and Jeremy Freese. 2001. Regression Models for Categorical Dependent Variables using
Stata. College Station, TX: Stata Press.
Lundman, Richard J. 1974. “Routine Police Arrest Practices.” Social Problems 22: 127-141.
———. 1980. Police Behavior. New York: Oxford University Press.
McAdam, Doug. 1982. Political Process and the Development of Black Insurgency, 1930-1970.
Chicago: University of Chicago Press.
McAdam, Doug, and Yang Su. 2002. “The War at Home: Antiwar Protests and Congressional Voting,
1965-1973.” American Sociological Review 67: 696-721.
McCarthy, John D., Andrew Martin, and Clark McPhail. 2005. “Constraints on the Freedom of Public
Assembly: Police Behavior and the Demeanor of Citizens in Disorderly Campus Gatherings.”
Paper presented at the Annual Meeting of the American Sociological Association, Philadelphia,
PA. August 2005.
McCarthy, John D., Clark McPhail, and John Crist. 1999. “The Diffusion and Adoption of Public Order
Management Systems.” Pp. 71-93 in Social Movements in a Globalizing World, Hanspeter Kriesi,
Donatella della Porta, and Dieter Rucht, eds. New York: St. Martin's Press.
McCarthy, John D., Clark McPhail, and Jackie Smith. 1996. “Images of Protest.” American Sociological
Review 61: 478-499.
McPhail, Clark, and John McCarthy. 1998. “Interaction Among and Collective Action by Police and
164 Mobilization

Protestors.” Paper presented at Protest, The Public Sphere and Public Order. University of Geneva,
Switzerland.
McPhail, Clark, David Schweingrubber, and John McCarthy. 1998. “Policing Protest in the United
States: 1960-1995.” Pp. 49-69 in Policing Protest, Donatella della Porta and Herbert Reiter, eds.
Minneapolis: University of Minnesota Press.
Meyer, David, and Suzanne Staggenborg. 1996. “Movements, Countermovements, and the Structure of
Political Opportunity.” American Journal of Sociology 101: 1628-1660.
Niederhoffer, Arthur. 1969. Behind the Shield. Garden City, NY: Doubleday.
Pastor, Paul. 1978. “Mobilization in Public Drunkenness Control.” Social Problems 25: 373-384.
Peterson, David. 1971. “Informal Norms and Police Practice.” Sociology and Social Research 55: 354-
362.
Piliavin, Irving, and Scott Briar. 1964. “Police Encounters with Juveniles.” American Journal of
Sociology 70: 206-214.
Piven, Frances Fox, and Richard A. Cloward. 1977. Poor People's Movements. New York: Vintage Books.
Reiner, Robert. 1998. “Policing, Protest, and Disorder in Britain.” Pp. 35-48 in Policing Protest,
Donatella della Porta and Herbert Reiter, eds. Minneapolis: University of Minnesota Press.
Reiter, Herbert. 1998. “Police and Public Order in Italy, 1944-1948.” Pp. 143-165 in Policing Protest,
Donatella della Porta and Herbert Reiter, eds. Minneapolis: University of Minnesota Press.
Rubinstein, Jonathan. 1980. “Cop's Rules.” Pp. 68-78 in Police Behavior, Richard J. Lundman, ed. New
York: Oxford University Press.
Silverman, Eli B. 1999. NYPD Battles Crime. Boston: Northeastern University Press.
Soule, Sarah A., and Jennifer Earl. 2005. “A Movement Society Evaluated: Collective Protest in the
United States, 1960-1986.” Mobilization 10(3): 345-364.
Skolnick, Jerome. 1966. Justice Without Trial. New York: Wiley.
Stanley, William. 1996. The Protection Racket State. Philadelphia: Temple University Press.
Stark, Rodney. 1972. Police Riots. Belmont, CA: Wadsworth Publishing.
Stockdill, Brett C. 1996. “Multiple Oppressions and Their Influence on Collective Action.” Ph.D.
Thesis, Northwestern University, Evanston, IL.
Tauber, Ronald K. 1967. “Danger and the Police.” Issues in Criminology 3: 69-81.
Tilly, Charles. 1978. From Mobilization to Revolution. Reading, MA: Addison-Wesley.
Van Dyke, Nella, Sarah A. Soule, and Verta Taylor. 2004. “The Targets of Social Movements: Beyond a
Focus on the State.” Research in Social Movements, Conflicts and Change. 25: 27-51.
Waddington, P.A.J. 1994. Liberty and Order. London: University College London Press.
———. 1998. “Controlling Protest in Contemporary Historical and Comparative Perspective.” Pp. 117-
140 in Policing Protest, Donatella della Porta and Herbert Reiter, eds. Minneapolis: University of
Minnesota Press.
Waddington, David, Karen Jones, and Chas Critcher. 1989. Flashpoints: Studies of Public Disorder.
London: Routledge.
Wilson, James Q. 1968. Varieties of Police Behavior. Cambridge, MA: Harvard University Press.
Winter, Martin. 1998. “Police Philosophy and Protest Policing in the Federal Republic of Germany,
1960-1990.” Pp. 188-212 in Policing Protest, Donatella della Porta and Herbert Reiter, eds.
Minneapolis: University of Minnesota Press.
Wisler, Dominique, and Marco Giugni. 1999. “Under the Media Spotlight.” Mobilization 4: 171-187.
Wisler, Dominique, and Hanspeter Kriesi. 1998. “Public Order, Protest Cycles, and Political Process.”
Pp. 91-116 in Policing Protest, Donatella della Porta and Herbert Reiter, eds. Minneapolis:
University of Minnesota Press.