Describing and Accounting for the Trends in US Protest Policing,1960−1995

Abstract

Numerous scholars have observed a decline in more coercive police tactics used to control demonstrations since the 1960s in North America and Western Europe. Such claims, however, are largely based on rather unsystematic observation, and almost no research directly examines the evolution of protest policing during this entire period. To address this gap, the authors use semiparametric logistic regression to examine reported police presence, the use of arrests, and the use of force at 15,965 US protests occurring between 1960 and 1995. The results confirm that while there has been an absolute decline in more repressive policing behavior, the transitional process was not a monotonic, linear process. The authors also investigate the different evolutionary patterns of each type of protest policing. The authors further demonstrate that African American initiated events, government targets, social movement organization presence, protest forms, the use of force, and arrests have variable impacts on police responses over time.

Keywords

protest policing state repression social movements

Most observers agree that protest policing in the United States has changed fundamentally since the protest wave of the 1960s. Indeed, among scholarly observers there had developed, until recently, a near consensus that the high level of police repression of the 1960s was followed by a more conciliatory form of protest control not only in the United States but in Western democracies more generally. In a widely noted study, McPhail, Schweingruber, and McCarthy (1998) argue this shift could be conceptualized as a transition away from a more coercive escalated force approach to more amicable negotiated management policing strategies. This distinction, and other similar ones (e.g., della Porta, Peterson, and Reiter 2006) has been widely adopted by those who examine protest policing in western democracies after the 1960s. While more current research (e.g., Vitale 2005, 2007) has raised the possibility that several elements of negotiated management practices have been on the wane after the turn of the century, the escalated force/negotiated management transition argument remains well entrenched among protest policing researchers (Soule and Davenport 2009).

Following the early work by della Porta (1996) and Waddington (1994), scholarship on how police officers and agencies plan for demonstrations and how they handle interactions with protesters during those gatherings has become much more common (e.g., della Porta and Reiter 1998; della Porta, Peterson, and Reiter 2006). Nonetheless, a considerable portion of subsequent work has continued to rely mostly on anecdotal evidence or case studies (e.g., McCarthy and McPhail 1998; McPhail and McCarthy 2005), though exceptions do exist (see Fillieule 1997). A recent study by Soule and Davenport (2009) provides one of the most attentive analyses of the transitional period in the United States. They examine police responses to protest events occurring between 1960 and 1990, using part of the same data source we employ. The central thrust of their analysis is to assess whether a trend toward less coercive police responses to protests was evident. They do so by differentiating between before and after 1969, making a plausible case that the transition toward negotiated management should have been well along by 1969. Soule and Davenport’s (2009) results do confirm an overall decline in coercive protest control, but they emphasize the continuity of protest policing styles since more threatening events are policed more aggressively regardless of when they take place.

Soule and Davenport’s (2009) analysis provides an important first step toward understanding the ebb and flow of American protest policing, but there are limitations to their analysis primarily due to their treatment of the transition as a discrete construct. This inhibits a more dynamic understanding of protest policing over time, which may help explain the limited impact of the broader institutional changes or differences in police strategies. More systematic and comprehensive attention to the transitional period is clearly necessary to be able to generate a more adequate assessment of these widely held expectations. Our study extends the existing literature by more concretely engaging with the ebb and flow of protest control. We pose three main questions: (1) Is there evidence of a decline in coercive protest policing strategies between 1960 and 1995? (2) If temporal trends do exist, are they consistent across different elements of police responses to protest? and (3) How might we account for the patterns we uncover, and what are the implications of such ideas and assessments for the processes hypothesized to have motivated the policing transition? While our first question is theoretically important, we expect that given the literature cited earlier it is largely confirmatory. In contrast, there has been little direct examination of our second and third questions, leading to a paucity of research identifying not only whether widely held expectations about different police behaviors are consistent but also the factors that can be used to explain the emergent patterns.

To address these questions, we review past work that specifies the key elements of the contrasting police strategies hypothesized to exist before and after the completion of the transition. We seek theoretically grounded changes that can be used to understand the forces responsible for generating the transition between stages. We then describe our empirical evidence drawn from over 15,000 protest events covered in the New York Times (NYT) between 1960 and 1995. We use semiparametric logistic regression models that control for theoretically relevant factors to examine the full temporal trends in reports of police presence at protest events, the police use of arrests, and the police use of force. We conclude by discussing the implications of our findings for protest policing theory.

The Temporal Evolution of American Protest Policing

While social movements scholars have long recognized the impact of state repression on social movement dynamism (Gamson 1975; Tilly 1978), a more recent stream of research has focused on the policing of protest events (e.g., della Porta 1995; della Porta and Reiter 1998; Earl, Soule, and McCarthy 2003; Rafail 2010). One important aim of these scholars has been to provide idiographic accounts of several cases involving highly coercive policing, such as the events that transpired during the 1968 Democratic National Convention in Chicago (Skolnick 1969) or the 1999 World Trade Organization protests in Seattle (Gillham and Noakes 2007). Another key goal among protest policing scholars has been to develop a parsimonious theoretical account of more routinized, day-to-day protest control. Such ideal typical models have largely been created on the basis of a small number of cases but generalized to the level of nation-state over the period from the peak of the protest cycles of the 1960s and 1970s into the late twentieth century, with an emphasis on western democratic countries. One of the most widely recognized accounts of trends in US protest policing over the last forty years is that of McPhail and colleagues (1998), who make the distinction between the escalated force and negotiated management periods. Table 1 summarizes the central differences between the escalated force and negotiated management approaches to protest control.

Table 1.

Comparison of the Escalated Force and Negotiated Management Protest Policing Strategies

	Escalated	Negotiated
Major dimensions of protest policing	Force	Management
Concern for First Amendment rights	Low	High
Tolerance for community disruption	Low	High
Communication between police and protestors	Rare	Frequent
Use of arrests to control participants	Frequent	Rare
Use of force to control participants	Frequent	Rare

Source: Adapted from McPhail, Schweingruber, and McCarthy (1998).

McPhail et al. (1998) emphasize that the distinctions between each of the five major dimensions of protest policing are meant to be understood relative to one another. Thus, compared to the escalated force era, there is a much lower likelihood of heavy-handed policing in the negotiated management period, particularly in regards to the extent and frequency of arrests and the use of force, which are two of the elements of policing we examine below. In line with McPhail et al. (1998) and following Soule and Davenport (2009), we expect that there has been an absolute decline in more aggressive policing of protest between 1960 and 1995, all other things being equal. While an absolute decline in the likelihood of police aggressiveness during protest events is generally accepted by many researchers, less attention has been paid to the transitional period between escalated force and negotiated management itself. The next section provides a more detailed account of the key changes and sources of stability in policing and social movement dynamics that are suggested to have led to the decline of the coercive protest policing style of the 1960s.

Accounting for the Shift in Protest Policing

To develop an accounting of the evolution of protest policing, we begin by distinguishing between (1) factors that we believe have remained constant over time and (2) factors that we expect to vary over the transitional period. Following Soule and Davenport (2009), we argue that more militant or contentious demonstrations can be expected to have consistently elicited more repressive police responses, regardless of the temporal period in which a demonstration occurred. Next, we discuss several institutional and structural forces that are thought to have altered protest management strategies, such as the institutionalization of protest and social movement organizations (SMOs) and the development of public order management systems (POMS), which provide a theoretical grounding for both temporal variations as well as the observed absolute decline in more forceful policing.

Factors Remaining Constant during the Transition

The relationship between a given protest’s level of threat and coercive policing is perhaps the most universally supported finding in the literature. A wide variety of factors can be considered threatening. Research by Schweingruber (2000), for example, has suggested that police are suspicious of large gatherings, thus may be more willing to use force or arrests when faced with large demonstrations. The most consistently important factors in shaping police responses have been the behavior of the participants during a protest event. Soule and Davenport (2009) convincingly argue that there is a great deal of temporal stability in the predictors of harsh police responses to protest, such as the effects of violent or extremely disruptive protest on police response.

Other research confirms that more contentious protestor behaviors, such as civil disobedience, property damage, riots, and mob violence are likely to be met with more aggressive policing compared to tactics such as candlelight vigils or peaceful rallies, regardless of the time period when the event itself occurs (McAdam 1983, 1999; Earl, Soule, and McCarthy 2003; Earl and Soule 2006). Indeed, McCarthy, Martin, and McPhail (2007) have argued that protest events may be privileged as the result of respect for the First Amendment right to assemble compared to largely apolitical convivial college student gatherings. This privileging only applies to peaceful protest however, since their findings suggest that once the protest participants engage in violent or illegal behavior, the police tend to respond in kind. Though there is wide agreement about the importance of threat, it is important to stress that the threat–repression relationship is variable, and several scholars note that the police are occasionally willing to tolerate illegal or threatening behavior, perhaps in hopes of avoiding heightened conflict (e.g., della Porta 1995; McPhail and McCarthy 2005).

Factors that Varied during the Transition

In contrast to the largely constant impact of situational threats, scholars of social movements and policing point toward aggregate changes that have taken place in the United States since the 1960s wave of protest. Although several broad social, demographic, and cultural changes have clearly taken place since then, we focus our discussion on how the historical evolution of both protest groups and police agencies are the primary causal factors in the transformation of protest policing.

Shifts in the Organization of Protest and Protest Groups

Since the wave of protest in the 1960s, there have been widespread changes in how social movements organize themselves, gain members, accumulate capital, collect other forms of resources, and become integrated into the political process. Meyer and Tarrow (1998) synthesize these changes into what they refer to as the “social movement society” hypothesis. A social movement society combines three central features: first, protest is a perpetual feature of political life and monotonically increases over time; second, protest has diffused to groups across the political spectrum; third, there has been a trend toward professionalization and institutionalization within SMOs. Since Meyer and Tarrow (1998) proposed the social movement society hypothesis, certain elements of their model have been questioned. For example, Soule and Earl (2005) and Koopmans (1993) have analyzed temporal trends in protest frequency, and their results suggest that the ebb and flow of protest activity is far from a linear, monotonic process. Despite this, there is relatively widespread agreement among researchers that protest has become institutionalized in the United States (McCarthy and McPhail 1998). This transformation has considerable implications for the policing of protest, which we now discuss more extensively.

According to McCarthy and McPhail (1998), the institutionalization of protest and SMOs refers to the trend that groups sponsoring and organizing protests are increasingly likely to be highly formalized and professionalized, making them more amenable to the enhanced predictability of protests scenarios that result from cooperative advanced planning. Such behaviors facilitate the adoption of negotiated management strategies since such groups are more likely to work with law enforcement prior to and during protest events. This trend of professionalization has been well documented in the literature (e.g., Meyer and Tarrow 1998; Meyer 2007) and is consistent with McCarthy and Zald’s (1977) argument that the social movement sector will become increasingly professionalized over time in the United States and in Western nations more generally.¹

A consequence of the institutionalization of social movements is that demonstrations and rallies became less contentious over time. As Tilly (2004, 146) points out, “the street demonstration looks like an all-purpose political tool . . . [it has become] a significant alternative to elections, public opinion polls, and letter writing as a way of voicing public positions.” As a result, we should expect that over time, the police are less likely to use coercive tactics at rallies or demonstrations, all other thing being equal. Another implication of the institutionalization argument is that there should be less aggressive policing when an SMO is present. Since specific SMOs can develop and sustain cooperative relationships with the police (McPhail, Schweingruber, and McCarthy 1998), over time we should expect to see a declining level of police aggressiveness during demonstrations.

Organizational Change in Policing

Since the 1960s, the organization of American police departments has increasingly favored the adoption of more reflexive, professionalized standards of conduct. One important change concerns the widespread diffusion of POMS. McCarthy and McPhail (1998, 91) define POMS as the “standard policing practices that are designed by authorities for supervising protestors’ access to public space and managing them in that space.” As a whole, POMS have increased the consistency of protest policing across US police departments and have diffused because of widespread federal involvement in training local police in the policing of civil disorders. Such training is designed to inculcate the principles of negotiated management in police departments around the country and was buttressed by enhanced by federal funding. At an organizational level, training programs such as the “Miami Field Force” model of policing and the extensive use of police paramilitary units for day-to-day operations became widely implemented across the United States (Kraska and Kappeler 1997; Kraska and Paulsen 1997). Such training regimens focus on squad-based rather than individual- or partner-based tactical maneuvering, ultimately giving police officers the tools to more efficiently contain large gatherings (McPhail and McCarthy 2005).

Such changes have major implications for protest policing. As Earl and Soule (2006) note, the professionalization of the police should result in a standardization of state responses to social movement mobilization over time. Accordingly, we expect that over time, police reactions to protest should be increasingly decoupled from the content of a challenger’s claims, and the challenger’s characteristics (e.g., race/ethnicity, gender, religion, or sexual orientation). While authorities have a long history of repressing radical groups (Gamson 1975; Goldstein 1978), African American initiated protest (McAdam 1999; Davenport, Soule, and Armstrong 2011), or other minority groups, if the professionalization argument is correct, we should expect to see a marked movement away from discriminatory protest policing practices over time.

To summarize briefly, we have outlined escalated force and negotiated management approaches to protest policing and examined the broad social and institutional changes that may have altered protest policing in the United States. Despite the relatively high level of scholarly agreement on the impact of these factors, we note that most studies discussing the transition have focused on its beginning and ending, with little in the way of direct engagement with the ebb and flow of the transitional period for the main dimensions of police response to protest. The hypothesized causal dynamics of the transition specified by McPhail and McCarthy for the US imply a relatively linear evolution toward negotiated management (McPhail et al. 1998; McCarthy and McPhail 1998). This is linked to changes in several social, legal, and political processes underlying US public order management (POM) strategies that diffused across police departments. McPhail, Schweingruber, and McCarthy (1998) particularly link an accumulation of constitutional guarantees regarding the rights of protesters and the diffusion and legitimation of a POM policing model emphasizing extensive negotiations between police and protest groups. The processes generating the transition, then, suggest that such institutional changes ought to be resistant to regression to earlier practices.

Other work based upon analyses of numerous European cases (della Porta 1996; della Porta and Fillieule 2004) that focuses more theoretical attention on the importance of shifts in political opportunity structures associated with state regimes on policing responses, on the other hand, can be read to imply a less smooth and linear transition toward less coercive policing. This difference is largely due to the response of particular governments to particular protest movements. These accounts focus some attention upon the changing ideological stances of governments, noting that the likelihood and extent of repression is typically greater under conservative governments (della Porta 1996).

Most notably, however, the extant literature is largely silent in generating precise explanations for the transitional period. As a result, we build on these previous studies and suggest that to establish a more thorough accounting of the transformation in protest policing, it is important to examine not only the endpoints but also the shape, scale, and comparability of the transition itself across different police actions. At this point, existing scholarship has not developed any systematic expectations of whether the transition in protest policing was a relatively smooth and monotonic process, or in contrast, if it was a nonlinear and nonmonotonic process. Further, since the available evidence suggests, for example, that the use of arrests may have become preferred over the use of force during unruly protests through the transitional period, it is also important to ask whether there are comparable evolutionary logics across different police responses. Finally, it is important to identify which key causal factors we can use to help understand the shifts in protest management strategies. The remainder of this study addresses these open questions using 15,965 protest events that took place between 1960 and 1995.

Data and Methods

Data Source: The Dynamics of Collective Action, 1960–1995

The unit of analysis in this article is the protest event, which we define as any type of collective activity that is carried out with the goal of articulating a claim against (or in support of) some target. Data on these protest events are drawn from the daily editions of the NYT between 1960 and 1995 and come from the Dynamics of Collective Action Project directed by Doug McAdam, John McCarthy, Susan Olzak, and Sarah Soule.² For a more detailed discussion of the data collection procedures, see Soule and Earl (2005), Soule and King (2008), or Walker, Martin, and McCarthy (2008). The original project coded information on several kinds of protest tactics used at protest events, but herein we analyze only events where the tactics used might plausibly draw police presence and/or action: rallies, demonstrations, marches, vigils, picketing, civil disobedience, ceremonial events, motorcades, dramaturgical demonstrations, symbolic displays, riots, mob violence, and attacks.³

To be included in the data set, a protest event must have met three basic criteria. First, since the interest is in collective action, there must have been more than one participant at the event. Second, participants at an event must have articulated some claim, whether this is a grievance against some target or an expression of support of some target. As well, all events occurred in the public sphere or were open to the public. Finally, these events occurred all over the United States and were not limited to a single locale.⁴

The data were collected in two distinct stages. First, researchers quickly read or “skimmed” each page of all daily issues of the NYT searching for any mention of protest events. This was done to avoid using the index to the NYT because such a procedure risks missing events that were embedded in articles on other (often related) topics. After finding articles on protest events, these were content coded for a variety of different facets of protest events. Intercoder reliability estimates for most items on the codesheet were consistently at or above 90 percent agreement. In all, there are 15,965 protest events reported to have occurred in the United States between 1960 and 1995 fulfilling the preceding criteria.

Newspaper Data

One of the most widely used sources of data on protest events, newspapers, and news sources are not without possible bias (see review in Soule forthcoming).

In general, there here are two chief sources of potential bias (Earl et al. 2004; Ortiz et al. 2005; Davenport 2010; Soule forthcoming): description bias and selection bias. Description bias refers to how well (or poorly) the journalist describes what actually happened at a given event. Most attempts to assess this source of bias conclude that the “hard facts” of the event are generally accurately reported by newspapers, if they are reported at all. Because we draw on “hard facts” of the events (as described in detail in the following) rather than on “soft facts” (such as journalistic opinions on the issue), we are comfortable with the accuracy of our measures for this article.

Selection bias refers to the fact that not all protest events are covered by a given newspaper and the possibility that what is covered is likely to be a biased representation of the universe of protests that actually took place. The literature on this kind of bias suggests that more intense events (e.g., larger, more violent, injurious, and conflictual), those with “significant actors” (e.g., celebrities or those defined as powerful and/or culturally legitimate), and those that took place in closer proximity to the newspaper, are more likely to be selected for coverage (Ortiz et al. 2005; Davenport 2010).

To deal with these potential sources of bias, we have implemented several procedures. First, as we note above, unlike many other studies using newspapers as a source of data on protest events, the project team did not use an index of the NYT to identify events nor did it sample days of the newspaper. Instead, researchers reviewed daily editions of the newspaper and identified all collective action events that were reported. This strategy helped to reduce the selection bias that may be introduced by indexing methodology and day-of-the-week rhythms in coverage (Earl et al. 2004; Ortiz et al. 2005). Second, we include in our analysis controls for other common sources of selection bias (e.g., disruptive tactics, violence, event size, and proximity to the news source). Finally, as a robustness test, we randomly selected 10 percent, 20 percent, 30 percent, 40 percent, and 50 percent of all intense events—defined as those where there was protester violence, that were larger than average, or involved property damage—and dropped them from the analysis (see also Soule and Davenport 2009). The logic of this strategy is that if newspapers overreport intense events and we randomly remove some of these intense events, we ought to be able to discern whether any overreporting is affecting our results. The results of these assessments indicate that with the removal of smaller proportions of events, and even when we remove 40 percent of intense events randomly, the same general patterns we report below hold. Our final analytic sample contains 15,965 protests, which are drawn from the larger sample of over 23,000 events reported in the NYT.

Variables

As Earl, Soule, and McCarthy (2003) note, much of the research on the policing of protest has examined policing in a dichotomous fashion—that is, police either show up at a protest event or they do not. We concur with Earl et al. (2003) and several more recent studies (e.g., Earl and Soule 2006; Gillham and Noakes 2007) that emphasize the wide array of options the police have once present at a protest event. Given these considerations, in the analysis that follows we are interested in accounting for the probability of reported police presence, their use of arrests, and their use of force during the 1960–1995 period.⁵ This is far from an exhaustive enumeration of the tactics available to the police (Earl 2003), but these dimensions do represent central theoretical components of the escalated force and negotiated management transition, and can also be relatively accurately measured using newspaper data. Since multiple police tactics can often occur in close succession, we alternate the variables for arrests and the use of force as regressands or regressors depending on the model specifications.⁶ Before proceeding, we note that these variables provide only a partial test of escalated force and negotiated management policing styles. This is particularly the case for negotiated management because it is difficult, if not impossible, to reliably code information on negotiations between protestors and the police from newspaper records. Nonetheless, the police tactics that we focus on do capture theoretically salient dimensions of police reactions to protest events and can also be accurately coded from newspaper coverage. Overall, we suggest that these variables provide useful proxies that allow us to reliably examine trends in protest policing over time.

We use three additional substantive blocks of predictor variables in our regression models of protest policing. First, we use six variables to measure situational threats. Taken together, these variables encompass the actions of protest participants and also several other elements of protest dynamics that can have strong implications for public order considerations. These include measures of whether the protestors engaged in any civil disobedience, whether they damaged property, or committed any acts of violence. These variables are all coded as dichotomies and are given values of 1 if they occur during the event and 0 otherwise. We also use a dichotomous indicator coded as 1 if counterdemonstrators were present and 0 otherwise, since such events often have high levels of conflict between the opposing groups (Waddington 1994). Since several studies (e.g., Earl, Soule, and McCarthy 2003; Earl and Soule 2006) have pointed to the important link between protest size and police responses, we include the natural log of the total number of participants reported to be present.⁷ The final variable in the situational threat block is a measure of the year a particular event took place, which ranges between 1960 and 1995.

The second block of variables gauges transitional threats, which are intended to capture factors that we expect will vary during the transitional period. First, sensitive to the heightened protest activity during the 1960s that was marked by high levels of state repression directed at African Americans (Goldstein 1978; Davenport and Eads 2001; Davenport et al. 2011), we use a dichotomous variable that is coded as 1 if the sole initiating group is African American and 0 otherwise.⁸ Second, even though the targets of collective action are multifaceted (Van Dyke, Soule, and Taylor 2004), several studies have suggested that the police have lower tolerance for protests that target the government (Earl, Soule, and McCarthy 2003; Earl and Soule 2006; Soule and Davenport 2009). This variability in police response could be due to the fact that the tactical repertoire utilized during a protest is influenced by its target (Walker, Martin, and McCarthy 2008). Demonstrations targeting elements of the government may be seen as more likely to disrupt public order compared to events targeting universities and private firms (Soule 2009). This variable is coded 1 if the government is the target of the event and 0 otherwise. Third and finally, following McCarthy and McPhail’s (1998) argument that SMOs became increasingly professionalized over the transition period, we differentiate between events with at least one SMO present (coded as 1) and those without an SMO (coded as 0). Since professionalized SMOs are more likely to develop cooperative relationships with the police, this variable also serves as a weak proxy of the willingness of the protestors to negotiate with the authorities prior to or during an event.

Finally, we use a three-category variable to capture the main form of each event. As we mentioned, in McCarthy, Martin, and McPhail (2007) comparison of convivial campus gatherings and disorderly campus protests, police employ different tactical repertoires when faced with different types of collective action events, and rallies or demonstrations appear to be privileged. Not only that, but if the police have become more professionalized over time, we should not expect peaceful protest forms (such as rallies and demonstrations) to be more heavily policed than other types of gatherings. We therefore differentiate between rallies and demonstrations, riots/mob violence/ethnic violence, and all other protest forms.⁹ We use “all other protest forms” as the reference category in our regression analyses. If the police professionalization argument is supported, we expect that the police became more tolerant of rallies and demonstrations over time.

Analytic Strategy

To examine reported police presence, arrests, and the use of force, we use semiparametric logistic regression, which allows for the introduction of nonlinear effects into the standard regression framework.¹⁰ Semiparametric models provide a balance between the straightforward interpretability of parametric models and the flexibility of nonparametric analysis. Semiparametric analysis provides a considerable advantage over the inclusion of polynomials to account for nonlinear associations in a regression framework—a practice common in the social sciences—since polynomials often utilize a global function to explain what is often local variation (Beck and Jackman 1998; Keele 2008). Nonparametric estimates are generated for both of our continuous variables (natural log of event size and year of event) using smoothing splines. We use parametric estimates for the categorical components of our models, and these can be interpreted identically to a fully parametric logistic regression model.

As Keele (2008) notes, a central objection to the use of semiparametric methods is that they facilitate overfitting and may direct attention to mostly inconsequential stochastic variation in a dataset. While such concerns have some merit, they are not unique to semiparametric analysis, and the estimation procedure that we use minimizes these concerns. Semiparametric models extend the iteratively reweighted least squares method that is commonly used for generalized linear models (GLMs). Indeed, only a few minor alterations are required for semiparametric regression. Specifically, for the simplest bivariate regression case, semiparametric models minimize:

| | Y_{i} - f_{j} (x_{i j}) | |^{2} + λ \int_{x_{1}}^{x_{n}} {f_{j}}^{''} (x_{i j})^{2} d x .

The leftmost term in equation (1) is the usual vector norm that is minimized for GLMs; however, the rightmost term provides a roughness penalty that is obtained by integrating over the squared second derivative in the span of event i. This procedure can be generalized to include both parametric and nonparametric estimates, as appropriate. An advantage to this specification is if the correspondence between x and y is linear, this will be reflected in the model. If the correspondence is more complex, this will be captured as well. More simply, this approach provides a penalty for smooths that are very “wiggly” (i.e., highly nonlinear), and this “wiggliness” is built into the estimation procedure. As Wood (2006, 130–131) points out, this setup leads to the problem of finding an optimal value for a smoothing parameter (denoted by λ in equation (1). We use cross validation to determine the appropriate amount of smoothness (see Keele 2008, 85–108), which automates the estimation λ in equation (1) based on the data.

Since our response variables are each operationalized as dichotomies, we assume that Y follows a Bernoulli distribution and use a logit link function; hence, $Pr (Y_{i} = 1) = [1 + exp (- η_{i})]^{- 1}$ .¹¹ This yields the final structural form of our model as follows:

\begin{array}{c} η_{i} = {\hat{β}}_{0} + x_{{S i t u a t i o n a l T h r e a t s}_{i}}^{T} \hat{β} + f_{S i z e} ({S i z e}_{i}) + f_{y e a r} ({Y e a r}_{i}) \\ - 12 p t + x_{{T r a n s i t i o n a l T h r e a t s}_{i}}^{T} \hat{β} + x_{P r o t e s t {F o r m}_{i}}^{T} \hat{β} . \end{array}

Consistent with the literature on semiparametric modeling, we use betas to denote parametric effects and $f (\cdot)$ for nonparametric smoothing. All of our models were estimated using Wood’s (2006) multiple generalized cross validation (mgcv; version 1.6-1) package in R (version 2.10.1; R Development Core Team 2009).

Results

Figure 1 displays the temporal trends for each of our police response variables. There is evidence of a general decline in the proportion of protests where police are reported to be present, use arrests, and use other forms of force or violence. The curve for each of these three police responses is quite jagged, however, and there is considerable fluctuation year to year. While this absolute decline is roughly consistent with the escalated force-negotiated management transition, it is clear that if a major transition in protest policing occurred, it was far from a linear, monotonic process. Furthermore, there is an increase in the proportion of protest events where the police are reported to be present or made arrests during the 1980s, counter to arguments about the transition from escalated force to negotiated management. In contrast, the proportion of events where the police use force follows a different pattern that is more consistent with this argument. The police are most likely to use force during the 1960s, and there is a rapid decrease by end of the decade. By 1970 until the end of our analytic period, the police use force at less than 10 percent of all protests. Taken together, these results do not suggest there is any generalizable temporal pattern across the three dimensions of protest policing, and therefore, any conclusions about the evolution of protest control by police are contingent on the specific police tactic that is examined.

The regression results for the parametric terms are presented in Table 2, and the nonparametric components are displayed in Figure 2. Though these estimates are based on the full model specification for each of our response variables, we also used a nested modeling strategy to determine whether including a particular block of variables appreciably improved model fit. For each police tactic we examine, there was a substantive and significant improvement based on a chi-square difference test and comparison of the Akaike information criterion. As a robustness check, we also compared the overall fit for all of our semiparametric analyses to models that contained linear terms for event size and year, and an additional specification with a categorical variable for decade and a linear term for event size. Our semiparametric models considerably increased model fit compared to the alternative specifications. The nested model tests are appended as Tables A1, A2, and A3.

Table 2.

Semiparametric Logistic Regression Estimates of Reported Police Presence, Use of Arrests, and Use of Force at US Protest Events, 1960–1995 (n = 15,965)

	Police Presence	Use of Arrests	Use of Force
Police tactic
Use of arrests			2.69*** (0.07)
Use of force		2.73*** (0.07)
Situational threats
Civil disobedience	2.01*** (0.05)	1.88*** (0.06)	0.74*** (0.08)
Property damage	0.30*** (0.08)	0.12 (0.09)	0.03 (0.11)
Use of violence	1.50*** (0.08)	0.46*** (0.09)	1.48*** (0.11)
Counterdemonstrators	1.13*** (0.08)	0.35*** (0.09)	0.46*** (0.11)
Event size (ln)	***see Figure 2A	***see Figure 2C	***see Figure 2E
Year of event	***see Figure 2B	***see Figure 2D	***see Figure 2F
Transitional threats
African American initiated	0.36*** (0.06)	0.26*** (0.07)	0.34*** (0.08)
Government target	0.35*** (0.04)	3.0E-3 (0.05)	0.55*** (0.07)
SMO present	0.08 (0.04)	0.05 (0.05)	0.01 (0.07)
Protest form
Rally/demonstration	0.93*** (0.05)	0.84*** (0.06)	0.14 (0.08)
Riot/clash	1.20*** (0.09)	0.72*** (0.10)	0.75*** (0.12)
Intercept	−2.38*** (0.05)	−3.08*** (0.06)	−4.70*** (0.10)
Model χ²	4,374.10***	4,441.80***	4,349.90***
df	23.30	22.91	26.49

Note: Reference category is all other types of event forms. SMO = social movement organization. *p < .05. **p < .01. ***p < .001 (two-tailed tests).

Reported Police Presence

The results for our model of reported police presence are presented in Table 2 and Figure 2. Beginning with the parametric estimates, as expected, the regression coefficients for civil disobedience, property damage, and protestor violence are all positively related to reported police presence. We also see that events where counterdemonstrators are present have a higher likelihood of reported police presence. Turning to the impact of transitional threats, we see that there is an increased likelihood of reported police presence at African American initiated protests. As well, there is higher likelihood of reported police presence at events where the government is targeted, consistent with previous findings (Soule and Davenport 2009; Soule 2009). Interestingly, the presence of one or more SMOs has a negligible impact on reported police presence. Finally, our results diverge slightly from the previous research, in that relative to all other types of event forms, the police are more likely to be reported present at riots/violent clashes, but contrary to our expectations, they are also more likely to be present at rallies and demonstrations.

The smooth suggests that while there is a positive relationship between event size (number of participants) and reported police presence at the event, the results point toward a threshold where the likelihood of reported police presence increases quite rapidly for very large events. This finding is intuitive since the prospects for public order disruptions can be seen as exponentially larger if a large demonstration takes on a violent or disorderly character. The nonparametric estimate for event year points to an overall decline in reported police presence since 1960; however, there is a strong departure from a monotonic relationship. While there is, in fact, a decline between 1960 and 1980, there is a sudden upsurge in the likelihood of reported police presence in the early 1980s followed by another decline in the late 1980s that lasts for the remainder of our analytic period. Since the period beginning around the 1980s displays important patterns for each of our reported police response variables, we provide more extended synthetic discussion of this upward shift in our Discussion section below.

Arrests

The parametric results for arrests point to a largely similar pattern compared to our models of reported police presence, though the estimate for property damage is not significant at conventional levels. Generally, arrests are more likely at events where the protestors engage in threatening behavior and when counterdemonstrators are present. There is less overall support for our variables measuring transitional threats: only African American initiated protests are statistically significant predictors of arrests, while the impact of government targets and SMO presence have a limited impact. Finally, similar to our model of reported police presence, there is a higher likelihood of arrests at riots or violent clashes and rallies or demonstrations.

The nonparametric effects also suggest an entirely different evolutionary logic for arrests relative to the other police tactics we examine. The nonparametric estimate for arrests points to a more complicated relationship between crowd size and arrests than we expected. Our models suggest that arrests are actually less likely to occur as the size of the event increases, but after a certain threshold (around 8,000 in our data [=exp(9)]), there is a rapid upsurge in the likelihood of arrests. This suggests that arrests are more likely to be used as a tactic when the police are faced with very large crowds but not necessarily when dealing with relatively smaller events. Importantly, since large events are generally rare there is a high level of uncertainty about the effect size for large demonstrations, as can be seen in the width of the confidence interval for these events. The smooth for event year suggests that the likelihood of arrests is largely flat during the 1960s, then slightly declines during the 1970s, and reaches its highest level in the mid-1980s, which is well after the end of the hypothesized escalated force period. This increase is not stable over time, however, as the likelihood of arrests declines considerably between the late 1980s until 1995. This pattern suggests that the changes in arrest patterns proposed by McPhail, Schweingruber, and McCarthy (1998) are not wholly supported in our analysis.

Police Force

Our final model examines the predictors of the police use of force. Overall, the directionality and impact of situational threats is mostly uniform, regardless of whether one looks at reported police presence, arrests, or the use of force. The main exception to this pattern is for the measure of property damage, which is not significant at conventional levels. For our measures of transitional threats, events initiated by African Americans remain a positive and robust predictor of the use of force, as do protests that target the government, while SMO presence does not appear have a significant impact. Unlike our models for reported police presence and arrests, there is not a higher likelihood that the police will use force during demonstrations or rallies. This latter finding suggests that the police are more selective when deciding to use force when faced with more standard forms of protest.

The temporal trends in the use of force by police point toward a different evolutionary logic relative to police presence and arrests. The nonparametric effect for event size also points toward a more complex relationship between protest size and the use of force than we expected. The graph shows a gradual increase in the use of police force as event size grows, however, once approximately 8,000 [=exp(9)] participants are present, there is a rapid decrease in the likelihood that the police use force (though again, there is a high level of uncertainty in the confidence interval). If the implications for the estimates for arrests and the use of force on event size are combined, it appears that the police use force to control protests up until a certain size threshold, then arresting protesters becomes the preferred police tactic. Given that at very large events, the number of protestors can outnumber the police by a considerable margin, the police may be hesitant to physically engage with a crowd that can easily overpower them. Finally, as the escalated force/negotiated management transition argument leads us to expect, there is an increase in the use of force beginning in the 1960s followed by a rapid decline until approximately 1975. For the next twenty years, there is a wavelike pattern in the use of force, which remains at a lower overall level than in the 1960s. Thus, this result supports the notion that there has been an overall decline in the use of force but does not support the idea that the transition was at all consistent.

Visualizing the Transition Period

Having established these baseline effects for reported police presence, arrests, and the use of force, we now turn to a series of interactions between specific explanatory variables and event year. While legal decisions and institutional shifts that have at least in part driven the transition, it is unclear and perhaps unlikely that they would diffuse evenly across American police departments as noted by McPhail, Schweingruber, and McCarthy (1998) and McCarthy and McPhail (1998). By estimating and comparing the results across the different police tactics we examine, whether and how such institutional changes have influenced protest on the ground. This strategy allows us to more concretely assess the impact of theoretically salient predictor variables which we expect to have influenced protest policing during the transition period. We estimate interactions between either the use of arrests or the use of force (depending on the response variable), whether an event was initiated by African Americans, whether the government was the primary target, whether an SMO was present, and if the main event form was a rally or demonstration.¹² These models allow us to relax the assumption that the police tactics we examine are substantively equivalent across our analytic period because we let the effects vary nonparametrically. We rely on a nested modeling strategy to compare different model specifications, and estimate each interaction individually, which is compared to the performance of the based models. A more conservative minimal threshold for p values of .01 is used since the chi-square distribution for these comparisons is approximate.¹³ The results from the interaction models are provided in Table 3.

Table 3.

Chi-Square Comparison of Model Fit for Regression Models with the Specified Interactions Predicting Reported Police Presence, Use of Arrests, and Use of Force at US Protest Events, 1960–1995 (n = 15,965)

	Police Presence	Arrests	Use of Force
Baseline Model	4374.10***	4441.80***	4225.10***
	Change in Chi Square
Interaction with event year and
Use of arrests			116.11***
Police use of force		78.71***
African American initiated	47.85***	7.02**	12.17
Government target	36.26***	16.00	13.94
SMO present	12.72	9.90	12.74
Rally/demonstration	21.18***	21.86***	13.73

Note: SMO = social movement organization.

**p < .01. ***p < .001 (two-tailed tests).

The overall pattern from the interaction results in Table 3 points toward considerable differences between each of the police tactics we examine. The model fit for reported police presence is significantly improved when interaction terms between event year and African American initiated events, government targets, and rallies or demonstrations are included. In contrast, when looking at the models predicting arrests, the interactions between event year and the police use of force, African American initiated protests, and rallies or demonstrations all significantly improve model fit. The fewest number of interaction terms are significant for the police use of force, where only the specification of event year and arrests is significant. Overall, these results point toward quite different evolutionary logics for each police tactic. Since omnibus tests of model fit do not provide an image of what the transitional period looked like, we provide figures of the statistically significant interaction terms in Figures 3 through 5.

The interactions predicting reported police presence are largely compatible with arguments favoring the adoption of negotiated management practices. As can be seen in Figure 3, the likelihood that the police were reported to be present for events with specific claims, targets, or initiating groups declined over time. Though there is an increased likelihood of reported police presence at African American initiated protests during the 1960s, it is likely a by-product of the heightened mobilization during the civil rights movement. However, by 1970, this trend had largely stabilized and remained roughly consistent for the rest of our analytic period. The interactions between event year and both government targeted events and rallies have similar trajectories: there is a slow and steady decrease until the 1980s, followed by a minor upswing and then a gradual decline by the end of the decade.

When looking at arrests, there is an absolute decline in the police use of force and during African American initiated protests over the analytic period, but these findings must be qualified. There is a considerable degree of variability in the use of force while making arrests over time, and the mid- to late 1970s and late 1980s display effects that are comparable in size to the escalated force period. Further, for African American initiated protests, we see a linear but slight decline over our analytic period, suggesting that the police became less likely to heavily police such protests, though this is at least partially offset by the increase in other forms of state repression directed at the black community (Oliver 2008). Finally, there does not appear to be strong support for the expectation that rallies or demonstrations became privileged activities during the transition, at least in terms of the likelihood of arrests. The smooth reaches its highest point in the 1980s, then begins to decline by 1986. This is followed by a more pronounced decline until 1995. The overall pattern of these interactions points to a series of processes that are not fully compatible with the expectations of the escalated force/negotiated management transition hypothesis. When looking at arrests, the evolution of protest policing is mostly nonlinear, and two of the three interactions we examined show at least a partial upsurge in arrests well after the conclusion of hypothesized escalated force period.

The final interaction we discuss focuses on the relationship between arrests and event year in predicting the use of force. The overall trajectory of the curve is highly nonlinear and largely indistinguishable from the main effects in Figure 2. Consistent with our expectations, there is an absolute decoupling of arrests from the use of force over time, but there is a large increase during the late 1970s and late 1980s. Thus, simply focusing on the beginning and ending of the transition period misses a high degree of instability in protest policing for nearly all of the period we examine.

Discussion and Conclusion

We set out to examine three central questions: first, we were interested in whether there had been an absolute decline in coercive protest policing since the 1960s' wave of protest. Second, we wanted to know whether the temporal trends in protest control demonstrated a similar pattern of change over time. Third, we asked which explanations help account for the patterns we uncovered over time. Our findings raise several implications for ideas about the transition of protest policing. Most fundamentally, we have demonstrated that the temporal evolution of protest control is not as straightforward as many scholars have anticipated, even after controlling for several robust predictors of protest policing. Consistent with the expectations proposed by McPhail, Schweingruber, and McCarthy (1998) and others, our results confirm that there has been an absolute decline in more coercive policing, at least in terms of arrests and the use of force (see also Soule and Davenport 2009). However, we demonstrate that there has also been a decrease in the proportion of protests where the police are reported present between 1960 and 1995. This raises the possibility that the apparent decline in harsh police responses is partially due to lower levels of police presence. Overall then, while coercive protest control appears to be on the wane, our results make it clear that the observed changes in police tactics are not consistent with one another. Examining the rightmost column in Figure 2, as well as the interaction effects, we see that there are strikingly different evolutionary patterns for police presence, the use of arrests, and the use of force in our data. Overall, these findings raise several implications for future research.

First, in our models predicting arrests, we see that even after controlling for situational threats, transitional threats, and protest form, the police use of force remains a very strong predictor. Though there is a relative separation of arrests from force over our analytic period, it is far from linear or monotonic. This trend is not easily reconciled with the escalated force/negotiated management transition, which argues that the use of force should become increasingly rare over time and that arrests should become increasingly orderly and routinized. If arrests are primarily a function of the level of threat an event poses, and, over time arrests become a crowd control strategy of last resort, we would not expect to see this tight coupling between police violence and arrests after the proper controls are added.

A second major implication is that the 1980s was also a period of transition in protest policing that has been largely unexamined both theoretically and empirically by scholars. While the implications of the escalated force/negotiated management transition proposed by McPhail, Schweingruber, and McCarthy (1998) is largely linear and monotonic, our regression analyses suggest that there was a significant departure from that smooth progression during the 1980s. This may be a result of a combination of several factors. First, if we assume that protest policing serves as a barometer of the broader political opportunity structure, as della Porta (1996) strongly argues, the upsurge in coercive protest policing may be a result of the rightward shift in American politics with the election of Ronald Reagan and the onset of the Reagan revolution. Given the decentralized nature of US police agencies, however, an explanation along these lines would need to locate mechanism of diffusion of new approaches similar to those identified by McPhail and colleagues for the post-1960s period. Another major shift during the 1980s was the large increase in the number of events with protestor initiated violence, which peaked at nearly 35 percent of events reported in the NYT in 1983. A further examination of the data (available upon request) suggests this is a result of a higher than average number of bias crimes during the early 1980s, coupled with an increase in radical environmental and antinuclear activism (see, e.g., Downey’s [1986] discussion of the Clamshell Alliance).

Third and finally, we focus on protest policing at a national level. To some degree, this decision may oversimplify a more complex picture relative to when a finer spatial analytic lens is used, such as regional, state-level, or municipal examinations of protest. It is well established that American police departments are highly autonomous (Bailey 1992) raising the possibility that certain police departments may have abandoned escalated force policies in favor of negotiated management, while others did not. Indeed, in the original statement on the escalated force/negotiated management paradigm, McPhail, Schweingruber, and McCarthy (1998, 68) caution that the diffusion and adoption of their model is far from ubiquitous. At the same time, our temporal analytic period was host to a geographic transition of protest mobilization away from Southern states after the decline of the civil rights movement. In our data, 36.5 percent of protests during the 1960s took place in the South compared to only 19.8 percent by the 1990s, using the four main census regions for comparison. As a result, it is difficult to untangle whether the absolute decline in heavy-handed protest control we observe is a function of a general reduction in police repression nationwide, fewer protests occurring in the Southern states, the decline of the civil rights movement, selection bias, or, more likely, the result of a combination of these changes. If anything then, our results serve to challenge the generality of the escalated force/negotiated management transition at the national level, but it is entirely possible that individual police departments—particularly, the Park Service Police in Washington, DC—did evolve as McPhail and colleagues suggest, and continued to pursue the negotiated management approach. Future research should pay particular attention to these meso-level dynamics to generate a more complete understanding of these factors and how they might influence protest control.

Footnotes

Appendix

Table A3.

Semiparametric Logistic Regression Estimates of the Police Use of Force at US Protest Events, 1960–1995 (n = 15,965)

	Model 1	Model 2	Model 3	Model 4
Police tactic
Use of arrests	3.04*** (0.06)	2.73*** (0.07)	2.71*** (0.07)	2.69*** (0.07)
Situational threats
Civil disobedience		0.54*** (0.07)	0.59*** (0.07)	0.74*** (0.08)
Property damage		0.13 (0.10)	0.09 (0.11)	0.03 (0.11)
Use of violence		1.79*** (0.08)	1.87*** (0.09)	1.48*** (0.11)
Counterdemonstrators		0.40*** (0.11)	0.44*** (0.11)	0.46*** (0.11)
Event Size (ln)		***	***	*** see Figure 2E
Year of event		***	***	*** see Figure 2F
Transitional Threats
African American initiated			0.36*** (0.08)	0.34*** (0.08)
Government target			0.52*** (0.07)	0.55*** (0.07)
SMO present			−0.04 (0.07)	0.01 (0.07)
Protest form
Rally/demonstration				0.14 (0.08)
Riot/clash				0.75*** (0.12)
Intercept	−3.39*** (0.05)	−4.20*** (0.07)	−4.54*** (0.09)	−4.70*** (0.10)
Model χ²	2,947.10***	4,231.00***	4,310.30***	4,349.90***
df	1.00	21.37	24.6	26.49
Change in χ²		1,283.90***	79.37***	39.57***
Change in df		20.37	3.23	1.89

Note: SMO = social movement organization. Reference categories are all other types of event forms and all other targets.

***p < .001 (two-tailed tests).

Authors’ Note

Katherine Johnson, Claudia Nau, Christian Davenport, and the anonymous reviewers provided several helpful comments. Replication data for this article are available at the journal’s website: http://www.jcr.sagepub.com.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by grants from the National Science Foundation (SES-0549930, SBR-9709337, SBR-9709356, SBR-9874000, SES-9911296, and SES-9911431).

Notes

References

Andersen

Robert

. 2009. “Nonparametric Methods for Modeling Nonlinearity in Regression Analysis.” Annual Review of Sociology 35: 67–85.

Bailey

David H

1992. “Comparative Organization of the Police in English-Speaking Countries.” Crime and Justice 15:509–45.

Beck

Nathaniel

Jackman

Simon

. 1998. “Beyond Linearity by Default: Generalized Additive Models.” American Journal of Political Science 42 (2): 596–627.

Davenport

Christian

. 2010. Media Bias, Perspective and State Repression: The Black Panther Party. New York: Cambridge University Press.

Davenport

Christian

Eads

Marci

. 2001. “Cued to Coerce: An Exploration of Dissident Rhetoric and its Relationship to Political Repression.” Mobilization 6 (2): 151–71.

Davenport

Christian

Soule

Sarah A.

Armstrong

David A.

II . 2011. “Protesting While Black? The Differential Policing of American Activism, 1960 to 1990.” American Sociological Review 76 (1): 152–78.

della Porta

Donatella

. 1995. Social Movements, Political Violence, and the State: A Comparative Analysis of Italy and Germany. Cambridge: Cambridge University Press.

della Porta

Donatella

. 1996. “Social Movements and the State: Thoughts on the Policing of Protest.” In Comparative Perspectives on Social Movements: Political Opportunities, Mobilizing Structures, and Cultural Framing, edited by McAdam

McCarthy,

J. D.

Zald

M. N.

, 62–92. Cambridge: Cambridge University Press.

della Porta

Donatella

Fillieule

Olivier

. 2004. “Policing Social Protest.” In The Blackwell Companion to Social Movements, edited by Snow

David A.

Soule,

Sarah A.

Kriesi

Hanspeter

, 217–41. Malden, MA: Blackwell.

10.

della Porta

Donatella

Reiter

Herbert

, ed. 1998. Policing Protest: The Control of Mass Demonstrations in Western Democracies. Minneapolis: University of Minnesota Press.

11.

della Porta

Donatella

Peterson

Abby

Reiter

Herbert

, ed. 2006. The Policing of Transnational Protest. Burlington, VT: Ashgate.

12.

Downey

Gary L.

1986. “Ideology and the Clamshell Identity: Organizational Dilemmas in the Anti-Nuclear Power Movement.” Social Problems 33 (5): 357–73.

13.

Earl

Jennifer

. 2003. “Tanks, Tear Gas, and Taxes: Toward a Theory of Movement Repression.” Sociological Theory 21 (1):44–68.

14.

Earl

Jennifer

. 2005. “You can Beat the Rap, but you can't Beat the Ride: Bringing Arrests back into Research on Repression.” Research in Social Movements, Conflicts, and Change 26:101–39.

15.

Earl

Jennifer

Martin

Andrew

McCarthy

John D.

Soule

Sarah A.

. 2004. “The Use of Newspaper Data in the Study of Collective Action.” Annual Review of Sociology 30:65–80.

16.

Earl

Jennifer

Soule

Sarah A.

. 2006. “Seeing Blue: A Police-Centered Explanation of Protest Policing.” Mobilization 11 (2):145–64.

17.

Earl

Jennifer

Soule

Sarah A.

McCarthy

John D.

. 2003. “Protest Under Fire? Explaining the Policing of Protest.” American Sociological Review 68 (4): 581–606.

18.

Fillieule

Olivier

. 1997. Stratégies de la Rue: Les Manifestations en France. Paris, France: Presses de Sciences Po.

19.

Fox

John

. 2000a. Multiple and Generalized Nonparametric Regression. Thousand Oaks, CA: Sage.

20.

Fox

John

. 2000b. Simple Regression: Scatterplot Smoothing. Thousand Oaks, CA: Sage.

21.

Freedman

David A.

2006. “On the So-Called “Huber Sandwich Estimator” and “Robust Standard Errors.”” The American Statistician 60 (4): 299–302.

22.

Gamson

William A.

1975. The Strategy of Social Protest. Homewood, IL: The Dorsey Press.

23.

Gillham

Patrick F.

Noakes

John

. 2007. ““More than a March in a Circle”: Transgressive Protests and the Limits of Negotiated Management.” Mobilization 12 (4): 341–57.

24.

Goldstein

Robert

. 1978. Political Repression in Modern America: From 1870 to the Present. Cambridge, England: Schenkman.

25.

Hastie

Trevor J.

Tibshirani

Robert J.

. 1990. Generalized Additive Models. Boca Raton, FL: Chapman and Hall/CRC.

26.

Keele

Luke

. 2008. Semiparametric Regression for the Social Sciences. Hoboken, NJ: John Wiley.

27.

Koopmans

Ruud

. 1993. “The Dynamics of Protest Waves: West Germany, 1965 to 1989.” American Sociological Review 58 (5): 637–58.

28.

Kraska

Peter B.

Kappeler

Victor E.

. 1997. “Militarizing the American Police: The Rise and Normalization of Paramilitary Units.” Social Problems 44 (1): 1–18.

29.

Kraska

Peter B.

Paulsen

Derek J.

. 1997. “Grounded Research into U.S. Paramilitary Policing: Forging the Iron Fist Inside the Velvet Glove.” Policing and Society 7 (4): 253–270.

30.

McAdam

Doug

. 1983. “Tactical Innovation and the Pace of Black Insurgency.” American Sociological Review 48 (6): 735–54.

31.

McAdam

Doug

. 1999. Political Process and the Development of Black Insurgency, 1930-1970. 2nd ed. Chicago: University of Chicago Press.

32.

McCarthy

John D.

Martin

Andrew W.

McPhail

Clark

. 2007. “Policing Disorderly Campus Protests and Convivial Gatherings: The Interaction of Threat, Social Organization, and First Amendment Guarantees.” Social Problems 54 (3):274–96.

33.

McCarthy

John D.

McPhail

Clark

. 1998. “The Institutionalization of Protest in the United States.” In The Social Movement Society, edited by Meyer

D. S.

Tarrow

, 83–110. Boulder, CO: Rowman and Littlefield.

34.

McCarthy

John D.

Zald

Mayer N.

. 1977. “Resource Mobilization and Social Movements: A Partial Theory.” American Journal of Sociology 82 (6):1212–41.

35.

McPhail

Clark

Schweingruber

David

D. McCarthy

John

. 1998. “Policing Protest in the United States: 1960-1995.” In Policing Protest: The Control of Mass Demonstrations in Western Democracies, edited by della Porta

Reiter

, 49–69. Minneapolis: University of Minnesota Press.

36.

Meyer

David S.

2007. The Politics of Protest: Social Movements in America. Oxford, England: Oxford University Press.

37.

Meyer

David S.

Tarrow

Sidney

. 1998. “A Movement Society: Contentious Politics for a New Century.” In The Social Movement Society: Contentious Politics for a New Century, edited by Meyer

D. S.

Tarrow

, 1–28. New York: Rowman & Littlefield.

38.

Oliver

Pamela

. 2008. “Repression and Crime Control: Why Social Movement Scholars Should Pay Attention to Mass Incarceration as a Form of Repression.” Mobilization 13 (1): 1–24.

39.

Ortiz

David G.

Myers

Daniel J.

Walls

N. Eugene

Diaz

Maria-Elena D.

. 2005. “Where do We Stand with Newspaper Data?” Mobilization 10 (3): 397–419.

40.

R Development Core Team. 2009. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. ISBN 3-900051-07-0, http://www.R-project.org.

41.

Rafail

Patrick

. 2010. “Asymmetry in Protest Control? Comparing Protest Policing Patterns in Montreal, Toronto, and Vancouver, 1998-2004.” Mobilization 15 (4): 489–509.

42.

Ruppert

David

Wand

M. P.

Carroll

R. J.

. 2003. Semiparametric Regression. New York: Cambridge University Press.

43.

Schweingruber

David

. 2000. “Mob Sociology and Escalated Force: Sociology’s Contribution to Repressive Police Tactics.” The Sociological Quarterly 41 (3): 371–89.

44.

Skolnick

Jerome H.

1969. The Politics of Protest: A Report. New York: Simon and Schuster.

45.

Soule

Sarah A

. 2009. Contention and Corporate Social Responsibility. New York: Cambridge University Press.

46.

Soule

Sarah A

. Forthcoming. “Protest Event Research.” In Blackwell Encyclopedia of Social and Political Movements, edited by Snow

D. A.

della Porta

Klandermans

McAdam

. Malden, MA: Blackwell.

47.

Soule

Sarah A.

Davenport

Christian

. 2009. “Velvet Glove, Iron Fist, or Even Hand? Protest Policing in the United States, 1960-1990.” Mobilization 14 (1): 1–22.

48.

Soule

Sarah A.

Earl

Jennifer

. 2005. “A Movement Society Evaluated: Collective Protest in the United States, 1960-1986.” Mobilization 10 (3): 345–64.

49.

Soule

Sarah A.

King

Brayden G.

. 2008. “Competition and Resource Partitioning in Three Social Movement Industries.” American Journal of Sociology 113 (6): 1568–1610.

50.

Tilly

Charles

. 1978. From Mobilization to Revolution. Reading, MA: Addison-Wesley.

51.

Van Dyke

Nella

Soule

Sarah A.

Taylor

Verta A.

. 2004. “The Targets of Social Movements: Beyond a Focus on the State.” Research in Social Movements, Conflicts, and Change 25: 27–51.

52.

Vitale

Alex

. 2005. “From Negotiated Management to Command and Control: How the New York Police Department Polices Protests.” Policing and Society 15 (3): 283–304.

53.

Vitale

Alex

. 2007. “The Command and Control and Miami Models at the 2004 Republican National Convention: New Forms of Policing Protests.” Mobilization 12 (4): 403–15.

54.

Waddington

P. A. J.

1994. Liberty and Order: Public Order Policing in a Capital City. London: University College of London Press.

55.

Walker

Edward T.

Martin

Andrew W.

McCarthy

John D.

. 2008. “Confronting the State, the Corporation, and the Academy: The Influence of Institutional Targets on Social Movement Repertoires.” American Journal of Sociology 114 (1): 35–76.

56.

Wood

Simon N.

2006. Generalized Additive Models: An Introduction with R. Boca Raton, FL: Chapman and Hall/CRC.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

4.05 MB

0.00 MB