Abstract
The National Study of Protest Events (NSPE) employed hypernetwork sampling to generate the first-ever nationally representative sample of protest events. Nearly complete information about various event characteristics was collected from participants in 1,037 unique protests across the United States in 2010 to 2011. The first part of this article reviews extant methodologies in protest-event research and discusses how the NSPE overcomes their recognized limitations. Next, we detail how the NSPE was conducted and present descriptive statistics for a number of important event characteristics. The hypernetwork sample is then compared to newspaper reports of protests. As expected, we find many differences in the types of events these sources capture. At the same time, the overall number and magnitude of the differences are likely to be surprising. By contrast, little variation is observed in how protesters and journalists described features of the same events. NSPE data have many potential applications in the field of contentious politics and social movements, and several possibilities for future research are outlined.
Protest events occupy a very prominent place in the field of contentious politics and social movements. The roots of their ascension are Tilly and colleagues’ pioneering work of systematically collecting event-level data and using them to analyze the causes and consequences of various forms of collective action (Shorter and Tilly 1974; Tilly 1978; Tilly and Rule 1965; Tilly, Tilly, and Tilly 1975). Drawing on the key methodological insights from this work, research on protest events has thrived over the last several decades. Among other topics, this research has advanced our understanding of protest waves (Koopmans 1993), social movement outcomes (e.g., see King and Soule 2007; McAdam and Su 2002; Soule and Olzak 2004), and state–dissident interactions (e.g., see Earl, Soule, and McCarthy 2003; Davenport, Soule, and Armstrong 2011; Khawaja 1993; Rafail, Soule, and McCarthy 2012). In short, so much of what we know about contentious politics and social movements comes from studying protest events.
Scholars have employed two main methodologies to collect protest-event data. The dominant one is to search newspapers for protest events and to code their reported characteristics. Common features that journalists describe include size, location, and cause. Newspaper studies of protest events tend to cover many decades and focus on papers national in scope (e.g., see Koopmans 1993; Kriesi et al. 1995; McAdam et al. 2009; 1 Rucht and Neidhardt 1999; Tilly 1978; Tilly et al. 1975). Some approaches draw a sample of events from newspapers, selecting only certain days of the week to code (e.g., see Koopmans 1993; Kriesi et al. 1995; Rucht and Neidhardt 1999), while others seek to construct the universe of protests covered by coding all stories that mention them (e.g., see McAdam et al. 2009; Tilly 1978; Tilly et al. 1975). The advent of electronic databases and search engines has greatly facilitated the extraction of stories about protests from newspapers and contributed to the growth of studies based on this method.
Although used less often than newspapers, police records constitute the other main data source for protest events. 2 Researchers must first obtain law enforcement agencies’ logs and permits of public gatherings and decide which events constitute protest action. Ease of access varies by the type of agency and accessibility of records. Police records are often computerized, speeding up the time it takes to search and code for protest events, though sometimes they require a manual reading and marking (Maney and Oliver 2001). Details about protest events from police logs and permits depend a lot on the source, but they generally contain basic information about events (Maney and Oliver 2001; McCarthy, McPhail, and Smith 1996; Oliver and Maney 2000). Given the more arduous nature of this methodology, it has been implemented only in a few cities (Barranco and Wisler 1999; Beissinger 1999; Fillieule 1999; Gentile 1999; Hocke 1999; Maney and Oliver 2001; McCarthy et al. 1996; Oliver and Maney 2000; Oliver and Myers 1999; Titarenko et al. 2001).
The field of contentious politics and social movements owes much to the scholars who have developed, refined, and applied each methodology in an effort to collect the very best possible data on protest events. But no methodology is perfect. Over the years, two main concerns have been raised about studying protest events from newspaper coverage and police records (e.g., see Earl et al. 2004; Olzak 1982; Ortiz et al. 2005). The first focuses on selection bias or how representative newspapers and law enforcement agencies’ records are of the total population of events. Studies show that these sources capture only a small subset of the universe of protests. Concerning newspapers, we know they disproportionately select events deemed to be “newsworthy,” such as those larger in size or involving conflicts (e.g., see Earl et al. 2004). Police permits miss events that do not require them (e.g., those planned for private property) or demonstrations for which activists refuse to file one (Maney and Oliver 2001; McCarthy et al. 1996). Because officers are more likely to be charged with misconduct at events involving contentious interactions, such as when participants are arrested, police logs tend to focus on these events at the expense of more “peaceful” ones (Maney and Oliver 2001). Furthermore, each methodology has spatial constraints. As mentioned above, studies of protest using police records are restricted to a handful of cities (Barranco and Wisler 1999; Beissinger 1999; Fillieule 1999; Gentile 1999; Hocke 1999; Maney and Oliver 2001; McCarthy et al. 1996; Oliver and Maney 2000; Oliver and Myers 1999; Titarenko et al. 2001), and events near newspaper offices have a much higher probability of being covered (Myers and Caniglia 2004; Ortiz et al. 2005).
The second concern deals with description bias or how accurately and thoroughly sources depict protests. As far as objective characteristics of events are concerned (e.g., when or where they occurred), the problem seems to be with the lack of information collected rather than its accuracy (Earl et al. 2004). 3 Police and journalists report only a limited number of all the possible event characteristics. Even for such important characteristics as date and size, roughly half or more of police records and journalists’ coverage of protests do not report them (McCarthy et al. 1996; McCarthy et al. 1999; Oliver and Maney 2000; Rucht and Neidhardt 1999). No doubt other features judged to be less significant are reported at even lower rates, if at all. As a result, police records and newspaper accounts miss a good deal of information about protests.
What all of this means is that the field of contentious politics and social movements lacks a nationally representative sample of protests with thorough data on event characteristics. Consequently, many important questions about what is arguably the field’s fundamental unit of analysis have gone unanswered. Among them are: How many people, on average, turn out to protests across the nation? Are members of minority groups the dominant demographic at events? Where do protests commonly take place? What types of organizations typically sponsor them? For which causes do protests mobilize the most? Do the majority of protests feature radical goals in that they target the state? Are peaceful or disruptive tactics the norm? How often do police and counterdemonstrators show up to events? Is conflict among different actors a rare or regular occurrence?
In this article, we break new ground in the study of protest events by employing the novel methodology of hypernetwork sampling to generate the first-ever nationally representative sample of protest events. 4 McPherson (1982) first introduced this method and it has been successfully used to produce representative samples of work establishments (Kalleberg et al. 1996), religious congregations (Chaves et al. 1999), and voluntary associations (McPherson 1983). When an appropriate sampling frame for a unit does not exist, hypernetwork sampling can generate a representative sample of one. Like work establishments, congregations, and voluntary organizations, there is no comprehensive list of protest events from which to draw a sample.
The core idea behind hypernetwork sampling is that a representative sample of a sought-after unit can be derived by randomly sampling people attached to that unit and then having them nominate the particular unit to which they are attached. Chaves et al. (1999), for instance, asked a random sample of American adults who attended religious services at least once during the past year to name the congregation they attended. Our project—the National Study of Protest Events (NSPE)—employed a similar strategy. We surveyed a random sample of U.S. protesters about the events they participated in to generate a nationally representative sample of protests.
The next section describes the background and methodological details of the NSPE. We then present national-level estimates of various characteristics of protest events. These are the first of their kind and they contribute important descriptive information to knowledge about features of protest events that have long been of interest to scholars of contentious politics and social movements. Next we compare NSPE data to events collected from the dominant source in the discipline—newspaper reports. We first analyze the extent to which the characteristics of types of protests (size, location, target, etc.) that appear in the hypernetwork sample differ from those covered in newspapers. Then, drawing exclusively on the set of protests common to both the NSPE and newspaper data, we examine whether protesters and journalists report comparable information about the same events.
The NSPE
Generating the NSPE Sample
Using hypernetwork sampling to generate a representative sample of U.S. protest events first required identifying a sufficiently large representative sample of protesters. This was not easy. Unlike labor force participation, voluntary association membership, or religious service attendance, protest participation in the United States is rare. Less than 10 percent of the U.S. adult population protests annually (American National Election Studies [ANES] 2010). Given that surveys typically used in the social sciences—such as the ANES or the General Social Survey—consist of, at most, a few thousand respondents, relying on such surveys would have yielded a small number of protesters and thus a small number of protest events. As a result, we sought an alternative source from which to draw the representative sample of protesters.
Knowledge Networks’ (KNs) KnowledgePanel (KP) constituted the most viable option. 5 KP is a probability-based online nonvolunteer panel that contains roughly 50,000 respondents aged 18 years and older from all 50 states and the District of Columbia. This panel is increasingly being used to collect data for social science projects, including those focused on political attitudes and behaviors. For example, recent waves of the ANES—the main source of scholarly data on U.S. politics—have relied on the KP. Furthermore, research has shown that the KP is representative of the United States as a whole and that the data collected from it are highly accurate (Callegaro and Disogra 2008; Chang and Krosnick 2009).
KP panelists are recruited through a combination of random digit dial and address-based sampling methods. These methods generate a sampling frame that covers approximately 97 percent of all households in the United States. If a sampled household does not have Internet access, it is given a netbook computer and free Internet service, so that it can participate. To become an “active” panelist, people who are recruited to the KP must first complete a profile questionnaire, consisting of basic demographic information. Once active, KP panelists complete surveys on a variety of topics via the Internet. To minimize attrition and respondent fatigue, panelists are generally asked to complete only one short survey (between 5 and 20 minutes in length) per week. 6
Not everyone in the KP has an equal probability of selection. For instance, the panel oversamples Census blocks with high concentrations of racial minorities. In addition, some people recruited into the KP never join, while others drop out before their designated end date. When pulling samples from the KP for specific projects, such as the one for the NSPE, a poststratification weight that corrects for the panel’s unequal probability of selection of particular groups, nonresponse, and attrition is incorporated into the probability proportional to size selection method, so that these samples are representative of the U.S. adult population. 7
For the NSPE, 25,288 KP panelists were screened between June 6, 2011, and August 8, 2011, to determine whether they had participated or not in a protest in the United States during the last 12 months. 8 Those who had were asked whether they had attended one or more than one protest in the prior year. Protesters of multiple events (38 percent) were then instructed to describe briefly up to four events they attended (single-event protesters were also asked to give a brief description of their attended event). For protesters of multiple events (23 percent attended two events; 8 percent attended three events; and 7 percent attended four events), one event was randomly selected for the NSPE survey. 9 Respondents answered roughly 15 minutes of questions about the event they attended (or in the case of multiple-event attenders, the event that was randomly chosen for them). No financial or other incentive was given to people to complete the NSPE survey.
Looking at Figure 1, we see that of the 25,288 KP panelists who were screened for the NSPE, 1,678 (or 7 percent of them) participated in one or more protests in the United States during the last year. 10 Seventy-eight percent of those who qualified for the NSPE agreed to participate in the study. Over 90 percent of those who agreed to participate completed the entire NSPE survey. 11 Because missing data were extremely low for the NSPE survey questions, this basically gave us complete information on 1,197 protest events that occurred throughout the United States in 2010 to 2011. 12 Not all protesters, however, attended a unique event. 13 After removing the 13 percent of duplicate events, we were left with 1,037 unique protests. 14 The completion rate for the NSPE was 71 percent. 15
Generating the National Study of Protest Events sample.
Due to nonresponse, certain demographics differed between the KP panelists selected to generate the NSPE’s protester sample and the Current Population Study benchmarks. To correct them, a within-gender poststratification weight was created specifically for the NSPE and employed in the event-level weight (discussed in the next section). 16
Correcting for known demographic discrepancies, however, does not completely rule out nonresponse bias. Even after adjusting for them, protesters who responded to the NSPE could still be systematically different in other ways from protesters who did not respond. For instance, perhaps politically-conservative protesters were more likely to participate than were politically-progressive ones. If true and if left uncorrected, protest events organized for progressive causes would be underestimated in the NSPE, given that protesters most likely to be attached to them would not be adequately represented in the protester sample.
Since protesters who opted into and out of the NSPE were part of the KP, we had access to a good deal of information about both groups. As mentioned above, KN collects and regularly updates a range of demographic information for all active KP panelists. Moreover, in the months before the NSPE was fielded, these panelists completed KN’s “Public Affairs” module, which focuses on political attitudes and behaviors. We could therefore compare whether protesters who responded to the NSPE were significantly different from those who did not on numerous variables. Because the outcome was binary (NSPE protester responders/NSPE protester nonresponders), we used logistic regression to identify whether additional nonresponse bias was present. 17 Bivariate and multivariate models were run on over 20 different variables. 18 No significant differences between NSPE protester responders and nonresponders were observed in any of these models. Given these findings, it was not necessary to include additional variables in the poststratification weight (described above) in order for the protester sample to be representative.
NSPE Probability Proportional to Size and Multiple-event Attender Features
Because of the nature of hypernetwork sampling, not all protest events in the NSPE had an equal probability of selection. Recall that this method generates events through protesters attached to them. Protests larger in size thus have a greater likelihood of appearing in the NSPE. 19 If unaddressed, protest events with a larger turnout would be overrepresented. 20 By weighting the data inversely proportional to event size, we were able to account for the fact that hypernetwork sampling yields a sample of protest events that are disproportionately larger in size.
We had to make another weighting adjustment. As discussed in the prior section, the NSPE included protesters who attended more than one event in the last year. Excluding more-than-one event attenders would have biased the NSPE sample since they are part of the protester population. But including them also introduced a complication in that not all NSPE protesters’ events had an equal probability of selection. The probability of an event being included in our study is inversely proportional to the number of protests attended. For example, the probability of an event appearing in the NSPE for a protester of two events is 1/2 (50 percent), while that same probability is 1/4 (25 percent) for a protester of four events (the probability is 1/1 [100 percent] for protesters of a single event). Left uncorrected, events from multiple-event attenders would have had a smaller probability of being selected, and thus the kinds of events they attended would be underrepresented. To address this issue, we weighted the NSPE data by the number of protests respondents attended in addition to the study-specific poststratification adjustment and the inversely proportional to size correction. 21
NSPE Data Collection Mode and Questionnaire Content
Information about protests in the NSPE was collected from event participants. Some other studies employing hypernetwork sampling have relied on leaders of the desired unit to collect data about it. Chaves et al. (1999), for instance, asked clergy or people holding other religious leadership positions to report on various characteristics about their congregations. For several reasons, we decided against using organizers of events. First, protests do not always have a formal organizer, especially a clearly identifiable one like a congregation has a clergyperson. Second, unlike congregations, other voluntary organizations, or businesses, protest events are not physical entities that have addresses, phone numbers, or websites. As a result, even if an event organizer was known, it would have been extremely difficult to reach and interview this person. Last, as long as the characteristics of the sought-after unit are directly observable, rank-and-file reports of them tend to be just as accurate as those of leaders. In their study of voluntary groups, McPherson and Rotolo (1995) obtained information on group size, number of women, number of members younger than 30, and number of college graduates from a regular group member, an officer of the group, and direct observation of group meetings. They found that information regular members gave was generally consistent with that obtained from officers and direct observation, leading them to conclude that “at least for [voluntary] face-to-face groups, we can measure salient compositional characteristics such as group-size reliably by asking respondents in a probability sample to estimate them” (McPherson and Rotolo 1995:1114).
This finding had an important implication for the content of the NSPE questionnaire. If we were to collect valid information about protest events from people in attendance, the questions had to focus on those characteristics that were directly observable. Doing so meant limiting the NSPE survey to items such as the number of participants, the location of the protest, the presence or absence of police or counterdemonstrators, and protesters’ overt tactics. We intentionally did not ask respondents, for instance, to report on the motivations of other activists in attendance. Such information is certainly valuable to know, but collecting it requires a different methodological approach, like surveying a large number of protesters at the same event (e.g., see van Stekelenburg et al. 2012).
The NSPE instrument began with questions about when, how long, and where (state, city, and specific location, such as in a street, in front of a building, or on a college campus) the protest took place. Respondents were then asked about the cause(s) and target(s) of the protest; whether counterdemonstrators were present (and if so whether there was any conflict between them and the main protesters); whether one or more organizations or groups sponsored the event and what types of organizations or groups were sponsors; whether any one spoke at the protest, and if so how many people spoke and their identity (e.g., celebrities, community members, or elected officials); the number of protesters; and the gender, racial, and age makeup of event participants. The survey concluded with questions about various protester tactics that may have been present at the event (such as participants carrying signs, marching in a circle, or joining arms), whether the police showed up, and if so, whether they arrested or took other actions against the protesters. Questions about all of these topics were closed ended and answer categories were designed to be as exhaustive as possible. For instance, respondents were presented with a list of over 30 possible protester tactics. Moreover, an open-ended “other” option was given for the majority of questions in an effort to collect the most comprehensive information about events from participants.
Results
NSPE Descriptive Statistics
This section presents the first-ever nationally representative estimates of the characteristics of protest events. We focus on characteristics that have long been of interest to scholars of contentious politics and social movements: size, demographic composition, location, speakers, organizational sponsors, causes, targets, tactics, counterdemonstrators, and law enforcement officers. Beginning with size, the average number of protesters at U.S. events in 2010 to 2011 was 61 people. Table 1 provides more details about events’ size distribution. The median American event featured only 15 protesters and a mere 5 percent of protests across the nation mobilized 200 or more individuals. In other words, turnout for the majority of events in the United States circa 2010 was very modest.
Number of Protesters at Events Falling on or Below the Specified Percentile.
In addition to size, the NSPE collected information about the gender, racial, and age distribution of events. Looking at the first two columns in Figure 2, we see that the average protest had slightly more male (52 percent) than female (48 percent) participants. The typical racial composition of U.S. protest events in 2010 to 2011 was 66 percent White, 19 percent Black, 9 percent Latino, 4 percent Asian, and 2 percent another race. For age, the two middle categories stand out. On average, 27 percent of events featured protesters who were between 18 and 29 years old, while 42 percent had participants between 30 and 60 years old.
National Study of Protest Events—Protester demographic composition.
Where did protests take place? As Figure 3 shows, sidewalks, at 47 percent, were the most common location of U.S. protest events in 2010 to 2011. 22 More than a third of events were organized in front of buildings, while about a fifth occurred in both streets and parks. Building interiors were the sites of 1 of every 10 protests. Designated free speech areas and college/school campuses hosted the fewest number of events during this time period.
National Study of Protest Events—Location.
Thirty-four percent of protests in 2010 to 2011 had an organizational sponsor. Figure 4 shows the distribution of different types of sponsoring organizations. 23 Among sponsored protests, civil rights/advocacy groups were featured the most, at 23 percent. Environmental and political associations were not far behind. Unions and human/social service groups were present in more than 1 of every 10 events with an organizational sponsor. Less than 10 percent of sponsored protests involved the following groups: public benefit, international, community, religious, educational, health, (non-union related) work, recreational, cultural, youth, and philanthropic.
National Study of Protest Events—Sponsoring organizations.
Speakers were also fairly common at NSPE protests, with 46 percent of events featuring at least one. In Figure 5, we see that protest organizers were speakers in over half of events that had talks. Community members (42 percent) and religious leaders (28 percent) were the next two most likely spokespeople at protests with a speech. One-fourth of protests with presentations included a celebrity address. Business leaders and elected officials took part in roughly one of every five events with speakers. Among protests in which someone gave a talk, people running for office were least likely to be present, at 6 percent.
National Study of Protest Events—Speakers.
Turning to Figure 6, the NSPE data show that protests in the United States were organized for over 25 different causes. But there was considerable variation in how often a particular cause was the focus. Governmental issues led the way. Between 2010 and 2011, over a fifth of all protests mobilized to contest some aspect of the government. Protests about abortion (18 percent), education (17 percent), health care (17 percent), workers’ rights (16 percent), human rights (14 percent), neighborhood issues (14 percent), gun regulations (11 percent), crime (11 percent), and drug policy (10 percent) rounded out the top 10. At the other end of the spectrum, less than 1 percent of all U.S. protests rallied around globalization. A number of other issues were also featured in less than 5 percent of all events: foreign policy, death penalty, racial issues, and religion. Given popular perception about the Tea Party’s prominence when the NSPE was in the field, it is noteworthy that this movement accounted for only 8 percent of all protests that occurred across the nation at this time.
National Study of Protest Events—Causes.
Next, we focus on the targets of U.S. protest events in 2010 to 2011. Figure 7 shows them in descending order. State or local governments were targeted the most, with almost a third of all protests directed at them. The national government (28 percent) and a group of specific people (25 percent) were, respectively, the second and third most likely targets. Eighteen percent of events targeted a medical facility/organization and 13 percent a business/corporation. Less than 1 of every 10 protests in America had as its target a particular person, research/scientific lab, university/school, or foreign government/state.
National Study of Protest Events—Targets.
Figure 8 ranks tactics employed at NSPE events from most to least frequent. Alone at the top was carrying a banner or sign. Three-fourths of all U.S. protests in 2010 to 2011 featured this activity. Participants distributed flyers or displayed pictures in about one out of every two events. The following 10 tactics were present at least 30 percent of the time, in descending order of occurrence: praying, chanting phrases, shouting, yelling or whistling, marching from one location to another, collecting petition signatures, waving U.S. flags, reading something together, marching in a circle or line, and singing. On the other hand, disruptive actions were among the least frequent. Eleven percent or less of protests involved participants throwing an object, going on a hunger strike, performing a die-in, or damaging property. Carrying crosses and performing skits joined these contentious behaviors as being the least likely tactics to occur at U.S. protest events.
National Study of Protest Events—Tactics.
Lastly, Figure 9 describes how often counterdemonstrators and police officers showed up at events. Neither actor was present the majority of the time. At 43 percent, counterdemonstrators were more likely to appear at events than were law enforcement agents. Among events featuring these agents (35 percent), about a fifth involved arrests of participants. Considering all events in the hypernetwork sample, the arrest rate drops to 7 percent.
National Study of Protest Events—Counterdemonstrators and Police.
Given that the NSPE is the first-ever nationally representative sample of protest events, the results in this section provide new information about the distribution of various event characteristics. But did this distribution vary within the NSPE’s collection period (summer 2010 to summer 2011)? Thinking about this period from a protest scholar’s point of view, two times stand out. The first half of the collection period included the peak of the Tea Party movement and growing anti-Obama sentiment, culminating in many Republican victories in the November mid-term elections. The other half, starting in February 2011, saw the rise of the pro-union, anti-austerity protests. State employees resisting deunionization kicked these off and union-sponsored solidarity protests soon followed (the emergence of the “Occupy” movement in September 2011 fell just outside the end point of the NSPE’s data collection [first week of August 2011]).
We divided the NSPE data into an “early” (June 2010–January 2011) and “late” (February 2011–August 2011) period to investigate whether event characteristics varied along the lines described above. For instance, union presence at events would be expected to be higher in the “late” period. Figure 10 shows the results of our two-time period analysis. This figure utilizes Wickham’s (2009) ggplot2 graphics package in R to display results of independent sample t-tests for various event characteristics in the form of point range plots. Black-ringed circles indicate a significant mean difference in a particular characteristic between the NSPE “early” and “late” period. Ringed circles marked with an “X” indicate no significant mean difference. The t-test plots are ordered by the size of the test statistic (lowest to highest). Line ranges represent the 95 percent confidence interval. Gray-filled circles represent the “early” period, while white-filled circles represent the “late” period. Looking at Figure 10, we see only three significant differences. As expected, NSPE protests from the “late” period had greater rates of union sponsorship. On the other hand, protests rallying around gun regulations and crime were more likely to occur in the “early” period.
National Study of Protest Events—Time t-tests: earlier versus later.
Comparing the NSPE to Newspaper Coverage of Events
In this section, we compare events from the NSPE to those covered in newspapers. Two different comparative analyses were conducted. The first evaluates differences in the types of events that these sources capture. Drawing on the full set of protests from the hypernetwork sample and newspaper reports (described below), we investigate variation in event characteristics—such as size, targets, or tactics—between the two sources. Then, we focus exclusively on the subset of events the sources shared to examine the extent to which protesters and journalists report comparable information about the same events.
To minimize media source bias (Earl et al. 2004; Myers and Caniglia 2004; Ortiz et al. 2005), protest events were extracted from four different newspapers—the Chicago Tribune, Los Angeles Times, New York Times, and Washington Post. These newspapers constitute the highest-circulated daily paper in each of the major U.S. Census regions. Plus, the Chicago Tribune, the Washington Post, and especially the New York Times have been the primary newspapers used for protest-event research in the United States. Like other scholars (for a recent example, see Ebert and Okamoto 2013), we used keyword searches to identify protest events. Using LexusNexis and ProQuest databases, the above papers were searched using a variety of keywords such as protest, rally, march, or demonstration (and their conjugations) for the same time period the NSPE covered (summer of 2010 to summer of 2011). 24 This yielded over 18,000 articles, all of which were downloaded and read to determine whether the story contained a valid protest event. 25 Of the original articles flagged, 978 contained a unique protest event. 26
Each of the unique events was then coded for numerous “hard news” characteristics, including date, location, size, cause, and police or counterdemonstrator presence. With the exception of size, if an article did not mention a characteristic for which we were coding, it was assumed to be absent, as is customary in newspaper protest-event research. Cases containing no information about the number of participants in the newspaper accounts of events—523 of the 978 articles (or 53 percent)—were dropped from the size comparative analysis. Because of the extremely high rate of missingness for protester demographics in the newspaper data—90 percent for age, 94 percent for race, and 96 percent for sex—it was not feasible to compare media reports for these variables to those from the NSPE.
Instead of using the keyword-search strategy to collect information about protest events from newspapers, another method compiles all pages for a particular time period and engages in “a full-text read” of them to identify relevant protest events and their characteristics (Maney and Oliver 2001). We employed this method for the New York Times and then compared the results from it to those from the keyword search to make sure we were adequately capturing media-reported protests for the NSPE time period. Over 32,000 New York Times articles from the summer 2010 to summer 2011 were extracted and manually read to see if they contained coverage of a valid protest in the United States. 27 Doing so yielded 398 unique protest events, which were then coded for the same characteristics as in the keyword searches. 28 Comparing results from the manual-reading method to those from the keyword-search method showed no substantive differences in protest characteristics for events reported in the New York Times. As a result, we used event data the keyword searches generated from all four newspapers to reduce media source bias.
Analyzing Differences in Types of Events
We begin by comparing the geographic distribution of protests from the newspaper data to that in the NSPE. Figures 11A and B plot the city locations of the 978 newspaper-based and 1,037 hypernetwork-sample-generated events. 29 The results are striking. Comparing the two figures, NSPE events cover a far greater number of cities. Moreover, looking state by state, we see that the newspaper reports of events are relatively sparse. For instance, while Florida is filled with many points in the map based on the hypernetwork sampling method, the newspaper map for this state only has a few points on it.
(A) Newspapers—Reported events. (B) National Study of Protest Events—Reported events.
Both sources show a clustering of events in larger cities, but as Figures 12A and 12B make clear, the nature of the clustering varies considerably. The circle area in these figures represents the relative number of events in each city. The larger the circle area, the greater the number of events concentrated within that city. Circles are drawn to the same scale in both figures. Looking at the event densities from the newspaper data in Figure 12A, we see that the largest circles are centered on New York, Washington, D.C., Chicago, and Los Angeles, while a handful of medium-sized circles hover over Madison, Phoenix, and San Francisco. Much smaller circles are scattered across this map. This indicates that New York, Washington, D.C., Chicago, and Los Angeles had the highest number of newspaper-reported events, followed by medium event-dense cities, while the additional cities (many of which are relatively high in population) had only one or two reported events.
(A) Newspapers—Reported events density. (B) National Study of Protest Events—Reported events density.
The hypernetwork event densities displayed in Figure 12B show that this method also captured a higher number of events in New York, Washington D.C., Chicago, and Los Angeles. However, the circles for these four cities in the hypernetwork sample are smaller relative to the newspaper-event densities map, revealing a lower concentration of NSPE events in them. Furthermore, medium-sized circles appear in many more cities in Figure 12B: Austin, Houston, Phoenix, San Francisco, Atlanta, Denver, Portland, and so on. And small circles cover 10 times as many smaller cities as compared to the newspaper data. This means that newspapers disproportionately cover events occurring in the cities where they are located, which confirms prior research showing that journalists are more likely to report protests near their offices (Myers and Caniglia 2004; Ortiz et al. 2005). After that, newspapers cover events primarily from other large cities. In contrast, the hypernetwork sample of events shows a distribution of events that is more evenly distributed across larger and medium-sized cities and produces a more diffuse distribution of events across many more small cities. 30
Figure 13 provides additional support for this observation. It shows the percentage of events falling into different city population sizes. 31 Considering the least populated cities (25,000 or fewer people), we see that almost twice as many events from the NSPE occurred in them relative to the newspaper data (15 percent and 8 percent, respectively). The reverse is true for the other end of the distribution. Over a third of events from newspaper reports were located in cities with a million or more people. But less than a fifth of NSPE protests come from them. Looking at the other population categories, events from smaller cities generally appear in higher numbers in the hypernetwork sample, while those from larger cities have greater representation in the newspaper data. These findings, along with those above, show clear and meaningful geographical and spatial differences in events that “make the news” and those that the NSPE capture (for interactive maps that enable visual comparisons of key event characteristics between the NSPE and newspaper data, go to nspe.nd.edu).
Differences in event location population size between newspapers and national study of protest events.
In what other ways do protests covered in newspapers differ from those in the hypernetwork sample? Figures 14 and 15 answer this question. These figures utilize the same Wickham’s (2009) ggplot2 graphics package described above to display the results of independent sample t-tests for various event characteristics from the NSPE and newspaper data in the form of point range plots. Figure 14 tests for differences in the continuous measure of event size, while Figure 15 tests for differences in the dichotomous measures, such as mobilizing for a particular cause (or not) or targeting a certain source (or not).
Newspaper versus National Study of Protest Events t-tests: All events, event size.
Newspapers versus National Study of Protest Events t-tests: All events, event characteristics.
Figure 14 shows the results of the size comparison between the NSPE and the newspaper data. Recall from above that the average number of protesters in the hypernetwork sample was 61, which the gray-filled circle on the left-hand side of this figure represents. In contrast, the white-filled circle shows that the mean turnout in the newspaper reports was nearly 3,000 participants. Additionally, as indicated by the white-filled circles farther to the right-hand side at the top of Figure 15, causes related to elections, religion, and race had significantly greater representation in newspaper reports of protests compared to those in the NSPE data. For instance, 22 percent of all protests that the newspapers covered in 2010 to 2011 focused on elections, while only 5 percent of events in the NSPE did. The newspaper data were also more likely to capture events taking place in both the nation’s capital and state capitals than was the hypernetwork sample. While events occurring in Washington, D.C. were, at 1 percent, few and far between in the NSPE, the nation’s capital was the site of more than 1 of every 10 protests reported in newspapers. Nearly a fifth of events from the media were located in state capitals, which is about double the number of them that appeared in the hypernetwork sample. Local or state government targets were also more likely to be featured in the newspaper data (43 percent) than they were in the NPSE (30 percent).
The bottom of Figure 15 shows that the NSPE was more likely to capture 10 types of events. Protests about the following eight causes showed up in greater numbers in the hypernetwork sample: war, neighborhood, abortion, drug policy, (non-election related) government, 32 crime, healthcare, and human rights. With respect to healthcare, for instance, 18 percent of events in the NSPE focused on this issue relative to only 2 percent in the newspaper data. Protests that featured police and counterdemonstrators also had greater representation in the hypernetwork sample of events. Law enforcement agents showed up to more than a third of protests in the NSPE data, while journalists reported them to be present only 13 percent of the time in the events they covered. The difference in counterdemonstrator attendance at events was larger still (43 percent for the NSPE and 6 percent for the newspaper data).
As these results demonstrate, the types of events that make the news are clearly different from those that a hypernetwork sample generates. All in all, we observed 17 event characteristics for which there was significant variation between the two sources. Of these characteristics, only seven were more likely to be reported in the media: larger events; protests rallying around elections, religion, or race; those taking place in the nation’s capital or state capitals; and those with local or state government targets. The remaining differences in event characteristics—such as mobilizing for abortion or human rights or the presence of police or counterdemonstrators—all favored the NSPE.
Analyzing Differences in Descriptions of the Same Events
The second part of our comparative analysis examines the extent to which protesters and journalists vary in how they describe features of the same events. Using the 55 protests that appeared in both the NSPE and newspaper reports, 33 we ran paired sample t-tests on various event characteristics and created point range plots to summarize the results. Unlike the findings reported in the section above, we only see a few black-ringed circles—indicating a significant mean difference—in Figures 16 and 17. For instance, no significant differences were observed in the reports of the number of participants, tactics employed, or protest targets between journalists and NSPE respondents for the same events. The only black-ringed circles are those in which a gray-filled circle is farther to the right (bottom third of Figure 17). This means that when there was a significant difference in the description of an event characteristic for the same protests, it was always present at a higher rate in the NSPE data.
Newspapers versus National Study of Protest Events t-tests: Matched events, event size.
Newspapers versus National Study of Protest Events: t-tests: Matched events, event characteristics.
For the set of protests that the sources shared, NSPE respondents (relative to journalists) were more likely to report that events focused on causes related to health care, gender/women’s issues, education, neighborhood, and human rights. Sponsoring organizations, counterdemonstrators, and police officers were also all more likely to be present in the hypernetwork sample of events common to both sources. Among the same events, for instance, NSPE participants indicated that law enforcement agents showed up to more than three of every four protests, while journalists reported them to be present less than a quarter of the time.
Discussion
For decades, research on protest events has provided important insights into the dynamics of collective action. But compiling event-level data has not been easy. A critical mass of scholars has spent countless hours collecting, assembling, and coding police agencies’ records and especially newspaper reports to build valuable data sets on protest events. Without such efforts, we would know far less today about contentious politics and social movements, particularly how their processes and outcomes unfold over time.
As scholars who draw on these sources acknowledge, however, they were not designed to study protest events. It thus should not be surprising to learn that they have limitations when used for this purpose. Because newspaper coverage and police records of protests disproportionately select certain types of demonstrations—such as those large in size—they capture only a small subset of the universe of events. Moreover, journalists and law enforcement officers have particular agendas for what they record, and thus their descriptions of events miss some important details. In sum, standard methodologies paint an incomplete picture of protest events.
The NSPE employed hypernetwork sampling to produce the first-ever nationally representative sample of protest events that gathered nearly complete information about a range of event characteristics directly from participants. Doing so allowed us to answer many previously unknown questions about protest events. Among the new facts learned about U.S. protests were their average size, how they were geographically distributed, what types of speakers and organizational sponsors were most common, and the extent to which they featured radical goals in the form of targeting the government.
Given that newspaper reports are the go-to source for data on protest events, we extracted events from the highest-circulated daily paper in each of the major U.S. census regions to compare to the NSPE. Variation in the types of protests was first assessed between the full set of events from these sources. Protests that appeared in both the NSPE and newspaper data were then analyzed to investigate how consistent journalists’ and protesters’ accounts were of the same events.
Regarding descriptions of characteristics for events found in both sources, similarity was the general theme. For instance, newspapers and the NSPE reported roughly equivalent number of participants as well as rates of protester arrest, property damage, employed tactics, and selected targets. Considerable agreement was also observed for the specific cause(s) around which protests rallied. In fact, of the over 30 event characteristics compared for protests common to both sources, only 8 significant differences were identified. These results are reassuring for the hypernetwork sample and newspaper coverage of protests. Had journalists and NSPE participants reported considerably different information about the same events, we would wonder about the validity of both sources.
When we found significant differences in information about the same events, they were all about the presence or absence of a particular characteristic—such as whether protests had an organizational sponsor or involved counterdemonstrators—and NSPE participants always reported greater presence of the characteristic than did media representatives. Why did protesters provide more details about identical events? For people to report something, they obviously have to be aware of it. For various reasons, journalists have access to significantly less information about events than do protesters. While organizers seem to be increasingly reaching out to the media in hopes of gaining attention (e.g., see Sobieraj 2011), reporters are still unlikely to learn about many events in advance. By showing up after the start of events, they are bound to miss important details. Moreover, because of deadlines, journalists do not have the time to investigate every nook and cranny of an event, nor to stay for its duration. In their systematic observation of a March for Life event in Washington, D.C., McPhail and Schweingruber (1999) found that journalists were more likely to be present at the destination (in front of the Supreme Court building) relative to the beginning and especially during the procession of the protest. Finally, given the goal of capturing “newsworthy” details, media representatives may self-filter information and hone in on characteristics meeting this criterion.
But even if journalists observed everything that protesters did, we would not expect them to report matching information. Rather than writing a story that appeals to a wide audience, the goal of the NSPE was to collect the most exhaustive data about events as possible. Respondents were asked roughly 15 minutes of detailed questions about the protest they attended. Questions focused not only on characteristics that would grab journalists’ attention (such as event size and whether conflict occurred) but also on less noteworthy ones. Together with the advantage of gathering information from a source with greater awareness about the protest, it makes sense that NSPE descriptions of events were more complete than those reported in newspapers. Whether they were unknown to journalists or not significant enough to print given “news holes” and other professional constraints, certain details about the same events were less likely to appear in the pages of newspapers relative to protesters being surveyed about them.
By contrast, analyses comparing the types of protests from the full set of events revealed many differences between the NSPE and newspaper reports. That such differences were observed is consistent with several studies that have evaluated local police records of protests against newspaper coverage of events (McCarthy et al. 1996; Oliver and Maney 2000; Oliver and Myers 1999). Confirming expectation that big crowds make for good news, McCarthy et al. (1996) found that among permitted demonstrations in Washington, D.C., during 1982 and 1991, those with more participants appeared in greater rates in the two newspapers studied (the New York Times and Washington Post). Our research also showed that protests in the newspaper data were larger in size than those in the NSPE. Unlike studies using police records, however, the events that the hypernetwork sampling method generated were not limited to a particular city. As a result, we were able to document for the first time the size differential in newspapers and a representative sample of events from across the nation. The differential was considerable. The average NSPE event size was 61 participants compared to nearly 3,000 in the newspaper data. Furthermore, the media sources documented a median turnout of 200 protesters, which was over 13 times that of the hypernetwork sample (see Table 1). Since the NSPE does not reach protests with 200 people until the 95 percentile, half of the events that newspapers cover in terms of size are in its top 5 percent.
We also identified significant variation in 17 other event characteristics. Protests organized for causes related to elections, religion, and race were more likely to show up in newspaper reports than they were in the hypernetwork sample. One potential explanation for the differences is that these causes occupied a high point in the U.S. “media issue attention cycle”—the rise and fall of certain issues in the news world (Downs 1972; McCarthy et al. 1996)—during the NSPE’s data collection period (summer of 2010 to summer of 2011). News Coverage Index (NCI) data from the Pew Research Center’s Project for the Excellence in Journalism confirm that this was the case for elections: the cause for which the largest discrepancy in representation between the NSPE and the newspapers was found. The NCI gathers and combines stories from some 50 newspaper, television, radio, and online news sources. Based on various analyses, they create a “Big Story” measure. Table 2 lists the 10 biggest stories when the NSPE was in the field. 34 Looking at this table, we see that two of the top 10 categories were related to elections.
Biggest News Stories for NSPE Time Period.
Source: News Coverage Index (NCI) data from the Pew Research Center’s Project for Excellence in Journalism (PEJ).
Note: NSPE = National Study of Protest Events.
It is also possible that newspapers covered more protests about elections because they constitute newsworthy “pegs” around which stories can be crafted. According to McCarthy et al. (1996:480), events that are notorious (e.g., well-known participants are involved or “trendy” actions are taken), consequential (e.g., presence of powerful participants or activities in which they are involved have wide impact), extraordinary (e.g., large crowds, spectacular actions, or unusual happenings that make for broad human interest), or culturally resonant (e.g., actors, actions, or topics reflect those which are familiar) represent desirable news pegs. Protests rallying around elections certainly fit the second and fourth criteria, and depending on whether candidates are present could also satisfy the first one. The news peg argument seems relevant as well for explaining why protests rallying around religion and race showed up more often in newspaper reports than they did in the NSPE. Given its “privileged legitimacy” in the United States (e.g., see C. Smith 1996), religion fulfills most of the requirements of a good news peg. Moreover, race has long been and continues to be a salient topic in American society, rendering it a “culturally resonant” peg around which to construct a media story.
Unlike newspapers, the NSPE was not sensitive to the 2010 to 2011 U.S. media attention issue cycle or features of events that qualify as attractive news pegs. It thus follows that protests focused on causes outside of this cycle’s peak or not considered to be worthy news pegs would have shown up more frequently in the NSPE. Among the causes appearing in higher rates in the hypernetwork sample were those related to war, neighborhoods, abortion, drug policy, (non-election related) government, crime, health care, and human rights. As shown in Table 2, none of these were top 10 news story in 2010 to 2011. But protests about war were more likely to show up in the NSPE than they were in the newspapers, and the war in Afghanistan was among the biggest news story when our project was in the field. What explains this finding? By 2010 to 2011, the heyday of the antiwar movement (as the February 15, 2003, massive demonstrations marked) had long passed. While the war itself continued to be important news, protests mobilizing for it would not be appealing to journalists.
In addition to differences in size and causes, significant variation in the location of events was identified between the two sources. Relative to the NSPE, protests from newspaper reports were more likely to be found in the nation’s capital and state’s capitals. These locations are common “beats” among reporters since they are epicenters of legislative and other important activities. Beats structure how journalists collect news and what stories they follow (e.g., see Ryfe 2009). Stationing reporters in and around the various capitals throughout the United States provides easier access to events that occur there. Since the hypernetwork sampling method drew a nationally representative sample of protesters to generate its events, it was not subject to journalists’ beats or routines and events specific to them.
Events featuring police and counterdemonstrators were also more likely to appear in the NSPE relative to the newspaper data. At first glance, these findings may be surprising. After all, research shows that such characteristics increased the probability of protests in the past gaining media attention (e.g., see Earl et al. 2004). Contrary to prior eras, however, today’s protests are generally routinized in that activists work with law enforcement agencies to facilitate peaceful demonstrations (e.g., see McCarthy and McPhail 1998). Given this situation, we would not expect police presence in itself to be a major draw for journalists. Similarly, events in which counterdemonstrators are present but not acting aggressively toward the main protesters should not be very attractive to the media. By contrast, events involving arrests of activists or disruptive actions (e.g., property damage) were no more likely to appear in the NSPE than they were in newspaper reports. Protests with conflict still make for good news, and thus why they were present in journalists’ coverage of events and the hypernetwork sample at similar rates.
Another possibility for the greater representation of police officers and counterdemonstrators in NSPE events is that journalists routinely attend protests featuring them but do not regularly report their presence. In our analyses of events common to both newspaper data and the hypernetwork sample, differences in the presence of law enforcement agents and counterdemonstrators were among the few observed. For the same events, NSPE participants reported these actors to be in attendance much more often than did media personnel. Along with the previous argument that simple police or counterdemonstrator presence may not make for “good news” these days, this finding helps explain why the 1,037 cases from the NSPE captured considerably more protests featuring them than did the 978 events reported in the newspapers.
Our comparative analyses relied on newspaper coverage from four major national newspapers: the Chicago Tribune, Los Angeles Times, New York Times, and Washington Post. Had we used local newspapers, differences in the types of events would not likely have been as stark. We would expect less of a gap in population size and turnout of events (Figures 13 and 14), for instance, given that local newspapers provide better cover of protests occurring in smaller places. Collecting protest data from local newspapers for cities all over the United States would be a very expensive and labor-intensive endeavor; hence, why most protest-event studies draw on major newspapers (usually the New York Times). Compared to efforts and resources required to assemble local newspaper data that are national in scope, hypernetwork sampling is a more cost-effective approach for identifying protests across the country.
Although used less often than newspapers, police records are another data source for studying protest events. It thus would have been ideal to compare the NSPE with these records, especially since they show some important differences in event characteristics relative to media coverage (McCarthy et al. 1996; Oliver and Maney 2000; Oliver and Myers 1999). But gathering and coding police records of protests for a large number of cities (the NSPE contains 461 unique ones) was simply not possible. Doing so for a single city is hard enough, and thus why studies of protest events based on police records are so rare. Consequently, we can only speculate about what differences may exist between the NSPE and police records of protests.
On the one hand, it is possible that less variation in the types of protests would have been observed since the police are not subject to journalistic standards of reporting (i.e., covering newsworthy events). On the other hand, officers are more invested in documenting certain types of events—contentious ones, for example, since they carry greater threat of misconduct charges (Maney and Oliver 2001). Perhaps, then, we would have found more variation in the types of events that appeared in the NSPE and police records. Regarding differences in descriptions of the same events, we would predict a similar pattern to that identified in the newspapers. Law enforcement agents and protesters should provide comparable information about most details of events they both attend. Like journalists, however, police officers do not fill out a survey about the protests they watch over. As a result, we would expect NSPE respondents to provide more thorough information about the same events.
Limitations and Future Recommendations
While the NSPE is not a comprehensive list of protest events, its hypernetwork sampling method generated the first-ever nationally representative sample of events and collected nearly complete information about a range of event characteristics directly from participants. Doing so advanced the study of protest events since well-known biases associated with other methods (such as selection on “newsworthy” features) were sidestepped. With that said, the NSPE’s events are specific to its collection period (summer 2010 to summer 2011). If our project was fielded more recently, some of its findings would be different. For instance, given the rise of the Black Lives Matter movement over the past couple of years, we would expect more protests about police brutality to be captured. The cross-sectional nature of the NSPE also means that certain topics fall outside its scope, including the diffusion and consequences of protest over time. Longitudinal event data—like those in the Dynamics of Collective Action (DCA) project—are needed to address such topics.
Future waves of protest events, however, could be collected using the hypernetwork sampling method and merged with the NSPE to examine changes in collective action across recent time periods. Such efforts could expand on the work of the NSPE. One area of development should focus on multiple-event attenders. We limited the number of protests respondents could list attending in the last year to four and randomly selected one for them to answer questions about. As a result, events from multiple-event attenders had a lower probability of being selected. We thus weighted the NSPE by the number of protests respondents attended, so that events from multiple-event attenders were not underrepresented. Ideally, information about each protest would have been collected and incorporated into our descriptive statistics. Moreover, by cutting off the number of protests, some imprecision was introduced into the events-attended weight adjustment. All 7 percent of protestors who attended four events were assigned the value of four. But some of these protesters would have participated in more than four events in the last year. Additionally, it is possible that the more than four event attenders listed their biggest or most “interesting” events. If so, this problematizes capturing their smaller and more pedestrian protests.
The following improvements could be made in future protest-event research using the hypernetwork sampling method. First, ask protesters about the total number of events they attended in the last year. Then, for those who attended all but a very high number (say over 10), gather basic information about each protest (size, date, location, and cause). Collect the same information for over 10 event attenders’ most recent protest and probe to get some information about the rest (e.g., “Thinking about your other protests, what was the size of the smallest one, the largest?” “In which cities did they take place?” “For what causes did they mobilize?”). Like the NSPE, randomly select one event to ask the full set of questions about for both the less than 10 and over 10 event attenders to minimize respondent fatigue. Data from such collection efforts would allow analyzing events from multiple-event attenders in greater detail.
Conclusion
In his study of the nonprofit sector, D. H. Smith (1997) found that extant data on U.S. voluntary organizations contained a great deal of “dark matter” since they significantly underreported grassroots associations. He argued that extrapolating from McPherson’s (1982) hypernetwork sampling of civic groups in a specific city provided the best national estimate of these associations. By employing the hypernetwork methodology, the NSPE shed new light on protests. It generated the first-ever nationally representative sample of events with nearly complete information about a variety of characteristics collected directly from participants.
These qualities constitute unique data for studying protest events. Compared to the full set of events reported in newspapers—the most relied on source for protest event research to date—the hypernetwork sampling method was more likely to capture protests considerably smaller in size; distributed across a far greater number of cities; located in less populated areas; taking place outside the nation’s capital and state capitals; mobilizing for such causes as crime, drug policy, abortion, and health care; and featuring police officers and counterdemonstrators. Since selection bias in media coverage of events has been well documented (e.g., see Earl et al. 2004), differences in the types of events that the NSPE generated and those appearing in the pages of newspapers were expected. But until now, we could only guess as to the total number and magnitude of these differences, given the absence of a nationally representative sample of events in the field of contentious politics and social movements.
Validating the accuracy of information for both sources, descriptions of event characteristics were generally found to be similar for protests the hypernetwork sample and newspaper coverage shared. NSPE respondents and journalists reported comparable data for size, tactics, organizational sponsorship, and over 30 other event measures. The handful of differences observed always favored the NSPE in that protesters reported more details about the same events than did media personnel. But this is not surprising since the NSPE directly asked participants a range of questions about the characteristics of events they attended.
The NSPE’s novelty opens the door for many new discoveries in the field of contentious politics and social movements. For instance, NSPE cases could be linked with census data. By doing so, we could identify the contextual characteristics that explain why protests take particular forms and encounter certain responses in some U.S. communities but very different ones in others. Because of the geographic clustering of newspaper coverage of events and the fact that police records of protests have been compiled for only a handful of cities, such research was not possible before the NSPE. In addition, classic questions could be revisited, including why certain protests, but not others, are policed and policed aggressively. Studies based on newspaper data specify a number of important factors including activists’ use of disruptive tactics and the presence of counterdemonstrators (e.g., see Earl et al. 2003). But do these variables also predict protest policing in the NSPE since it is a nationally representative sample of events that contains very different types of protests (e.g., those considerably smaller in size and located in less populous cities)? Plus, new variables for understanding variation in police presence and action at events are likely to emerge from analyzing the hypernetwork sample. These ideas only scratch the surface for how the NSPE could be put to use in the future to advance scholarship on protest events and collective action more broadly.
Footnotes
Authors’ Note
Acknowledgment
For helpful comments on prior drafts, we thank Kenneth Andrews, Christian Smith, participants in the Studies in Politics and Movements seminar at the University of Notre Dame, especially Christian Davenport, Rory McVeigh, and Guillermo Trejo, participants in the Center for the Study of Democracy colloquium at the University of California at Irvine, and anonymous reviewers.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The NSPE was supported by grants from the Science of Generosity and the University of Notre Dame’s Kroc Institute for International Peace Studies as well as the Institute for Scholarship in the Liberal Arts.