WHAT EXPLAINS TEMPORAL AND GEOGRAPHIC VARIATION ON ISSUES OF RACE IN THE 1960S?

Figure 1 presents data on what Americans, when surveyed, indicated was the “most important problem” facing the country between 1950 and 1980.^{Footnote 2} Looking at issues of race, two trends are noteworthy. First, from 1950 into the early 1960s, the percentage of Americans responding that civil rights was the most important problem remained low. In the early 1960s, however, it spiked from approximately five percent in December 1962 to 48% in mid-1963 and then faded almost as quickly. Second, up until the mid-to-late 1960s, concern about “social control” as the most important problem remained in the single digits and then gripped the country to reach an initial peak of about 41% of respondents in August 1967 before declining rapidly after 1971.^{Footnote 3}

The rapidly shifting attitudes observed in polls were also evident in swings in white voting behavior. In 1964, Republican presidential candidate Barry Goldwater promised “law and order” against “crime in the streets” but lost in a blowout to President Johnson, a champion of civil rights (Flamm Reference Flamm2005). By 1968, though, the tide had turned and Republican presidential candidate Richard Nixon successfully marshaled a “tough on crime” campaign to help win the White House. “Law and order” rhetoric had been popular in the South for decades (Finkelman Reference Finkelman1993; Murakawa Reference Murakawa, Lowndes, Novkov and Warren2008). It was not until the mid-1960s, however, that the slogan took root outside the old Confederacy. In 1966, for example, Ronald Reagan echoed Goldwater and ran on “law and order” to win the California gubernatorial race (Flamm Reference Flamm2005). In short, the 1960s saw rapid temporal and geographic variation in white concern for race-related policy, first with civil rights and later with the rise of a national coalition championing “tough on crime” policies.

ELITE AND PLURALISTIC THEORIES OF INFLUENCE

Whether this variation in elite discourse, public opinion, and voting behavior is better explained by theories emphasizing the role of elite actors or mass movements is unresolved. Fields like history and sociology often attribute a central role to activists but the consensus view in political science is elites dominate mass opinion (Iyengar and Kinder Reference Iyengar and Kinder1987; Lee Reference Lee2002; Lenz Reference Lenz2013; Zaller Reference Zaller1992). Through media, political elites influence what issues are top of mind, how issues are framed, and which stories are most prominently covered (Baumgartner and Jones Reference Baumgartner and Jones2010; Iyengar and Kinder Reference Iyengar and Kinder1987). Carmines and Stimson (Reference Carmines and Stimson1989), Weaver (Reference Weaver2007), and Zaller (Reference Zaller1992) emphasize the role of presidents and presidential candidates like Johnson and Goldwater in driving the public taste for civil rights and “law and order.” Figure 1, however, presents no evidence that trends in public opinion on civil rights and “social control” are shifting appreciably before the four presidential elections between 1960 and 1972 (see dotted vertical lines).

More pluralistic models provide evidence of lateral or bottom-up influences on policy makers and public opinion (Enos, Kaufman, and Sands Reference Enos, Kaufman and Sands2019; Gillion Reference Gillion2013; Lee Reference Lee2002; Luders Reference Luders2010; Mazumder Reference Mazumder2018; Wallace, Zepeda-Millán, and Jones-Correa Reference Wallace, Zepeda-Millán and Jones-Correa2014; Williamson, Skocpol, and Coggin Reference Williamson, Skocpol and Coggin2011). Lee (Reference Lee2002) builds on Dawson (Reference Dawson1994) and Key (Reference Key1949) to posit a theory of political communication in which elites remain influential but that also allows for “activated mass opinion,” in which non-elite counterpublics, in times of social and political contestation, serve as wellsprings of influence on mass opinion. However, Lee only investigates nonviolent protests, and therefore, neither models nor estimates differential effects of nonviolent versus violent protest. Gillion (Reference Gillion2012, Reference Gillion2013) and Gillion and Soule (Reference Gillion and Soule2018) show protests serve as an “informative cue” to Congress, presidents, the Supreme Court, and voters. These works offer compelling evidence that violence serves as a kind of signal amplifier but also do not model possible differential effects of violence. These studies contribute to our understanding of the changes in public opinion and legislative activity on civil rights in Figure 1 but offer limited insight into the possible relationship between protester-initiated violence and trends in public demand for social control. In addition, these studies attribute a central role to media, but the theory and evidence are underdeveloped. Lee, for example, documents year-level correlations between movement activism, New York Times coverage, and citizen letter writing but notes a more thorough examination is beyond the scope of the book.

Cloward and Piven (Reference Cloward and Piven1971) propose an “insurgency thesis” in which mass unrest induces concessions from elite actors attempting to buy off a movement. Empirical tests of the hypothesis have been contradictory. Some work finds evidence in favor of disruption increasing benefits. Gamson (Reference Gamson1975) finds “unruly groups” deploying violence are more successful, as compared with nonviolent groups but, critically, suggests this dynamic might only work in cases wherein “public sentiment neutralizes the normal deviance of the action” (88). Luders (Reference Luders2010) explains variation, particularly in local movement success, through the different types of costs imposed on protest targets, such as businesses and politicians. Other scholars find that political elites responded to the civil unrest through increased investments in social policy and other redistributive policies (Fording Reference Fording1997, Reference Fording2001; Hicks and Swank Reference Hicks and Swank1983). Fording (Reference Fording1997) finds that, under specific conditions, political violence by African Americans led to greater spending on Aid to Families with Dependent Children. Similarly, Skrentny (Reference Skrentny1996) finds that the sense of crisis following violent unrest helped shift white elites in business and politics to support programs like affirmative action. Enos, Kaufman, and Sands (Reference Enos, Kaufman and Sands2019) find violent protest in Los Angeles mobilized a subset of voters and caused a liberal shift in favor of certain education ballot initiatives. Other work points out significant methodological problems in prior research or finds no effect (Albritton Reference Albritton1979; Fording Reference Fording1997). More broadly, due to the focus on violent protest and elite responses, tests of the “insurgency thesis” offer little insight into effects of nonviolent disruption or how protest movements influence mass opinion (e.g., Figure 1) and voting behavior. Also, both Cloward and Piven (Reference Cloward and Piven1971) and Gamson (Reference Gamson1975) published in the first half of the 1970s (with Gamson’s study ending in 1945), precluding significant engagement with how the new mass medium of television might change potential costs and benefits of violent resistance.

Where the “insurgency thesis” suggests states respond with “carrots,” other work finds governments wield “sticks” in which efforts at repression may trump or complement those of redistribution (Button Reference Button1978). A number of studies find evidence that protests and civil disorders were associated with enhanced expenditures on policing and efforts at coercive control (Button Reference Button1978; Feagin and Hahn Reference Feagin and Hahn1973; Fording Reference Fording2001; Sears and McConahay Reference Sears and McConahay1973; Welch Reference Welch1975). More recent work also finds little independent effect of unrest on repression outside strategic use by elites (Weaver Reference Weaver2007). As with evaluations of the “insurgency thesis,” these studies tend to focus on effects of violent protests or emphasize elite responses and, as a result, fail to capture effects of nonviolent resistance or mass public reactions in opinion or voting.

Work in comparative politics and sociology is more attentive to differential effects of protest tactics, particularly with regard to the interaction between tactics and coalition formation. Rojas (Reference Rojas2006) finds the effectiveness of black-led protests in higher education depends on the level of disruption, with more disruptive protests inhibiting campus administration allies from supporting the movement agenda. Looking cross-nationally at social movements, Stephan and Chenoweth (Reference Stephan and Chenoweth2008) find that violent campaigns are successful about 26% of the time, whereas nonviolent efforts succeed about 53% of the time by maintaining legitimacy, attracting broad-based participation and co-opting potential opponents. Although the theory and evidence are attentive to the role of protest tactics in electoral politics, due to limitations of the data the study cannot test within-country variation in the reactions to those campaigns. Wilkinson (Reference Wilkinson2004) investigates violent interethnic protests in India and finds electoral incentives for mono-ethnic versus multi-ethnic coalitions play a significant role in determining whether politicians gin up or attempt to thwart emerging conflicts. Although Wilkinson applies the model to majority-led white-on-black race riots in the United States between Reconstruction and the 1950s, the work does not address how the model applies to minority-led nonviolent and violent protests of the 1960s and 1970s.

An interdisciplinary literature in history and American political development traces the partisan realignment on issues of race and finds evidence that, as early as the 1940s, Democrats were becoming the party of civil rights (Feinstein and Schickler Reference Feinstein and Schickler2008). Flamm (Reference Flamm2005) delineates how the early years of the “war on crime” were initially led by Democrats. Both Flamm and Weaver (Reference Weaver2007) highlight how, by the late 1960s, the issue had been largely captured by strategic conservatives. Hall (Reference Hall2005) traces a “long backlash” beginning in the 1940s, and Murakawa (Reference Murakawa, Lowndes, Novkov and Warren2008, 235) highlights the ways in which national political figures “explicitly and routinely addressed black civil rights in criminological terms” before the 1960s. Joseph (Reference Joseph2009, 708) challenges simple associations between “race riots [and] gun-toting militants” to document the breadth of black radical influence, from local anti-poverty organizing to helping elect a new generation of black mayors. These longer, more complex time horizons offer a helpful corrective but still leave open the puzzles of variation in timing and geography: why, for example, did “law and order” fail for Goldwater yet succeed for Nixon?

AGENDA SEEDING

To reconcile competing elite and pluralistic theories, I propose a model that recognizes elite dominance of political communication but focuses on the ability of activists to overcome asymmetries through tactics like disruption that appeal to news organizations and shape whether and how media cover their concerns. Society is conceived of as a stratified democracy with two groups, one a dominant majority and another a subordinate minority (Sidanius and Pratto Reference Sidanius and Pratto2001). Within this society, two political coalitions compete for power. One coalition, typically more conservative, is broadly aligned with the dominant group and supportive of extending the power of the majority. The second coalition, typically more liberal, is more egalitarian and supportive of the interests of the subordinate group (King and Smith Reference King and Smith2005). The dominant group-aligned coalition has disproportionate capacity to project power through institutions like the government and media, whereas the subordinate-aligned coalition operates under constraints such as discriminatory laws, social stigma, higher rates of poverty, within-group cleavages, state-sanctioned persecution, and greater barriers to collective action (Cohen Reference Cohen1999; Olson Reference Olson1965).

Figure 2 summarizes the model. In the first period, subordinate group activists attempt to raise awareness through agenda seeding that involves activists employing tactics, possibly nonviolent or violent, which might make the news. News organization still set public agendas but social movements, like any group seeking greater attention or publicity, attempt to entice media to prioritize their concerns. Southern civil rights protests, for example, were often planned for mornings so footage could be flown to New York in time to air nationally on the evening news (Torres Reference Torres2018). Similarly, movement leaders selected Selma, AL, as a site for voting rights protests partly for its proximity to television network affiliates in Montgomery (Torres Reference Torres2018).

FIGURE 2. Model of How Activist Agenda Seeding Influences Media and Politics

Media attention, however, can be sympathetic or hostile. Even presented with similar facts, different reporters and news organizations often structure stories in systematically different ways (Davenport and Litras Reference Davenport and Litras2003). Civil disobedience, for example, can be covered as a form of justifiable protest or unjustifiable criminality. Kinder and Kam (Reference Kinder and Kam2010) find that whether an issue can overcome societal biases to become salient depends significantly “on the ability of the issue to command the public’s limited and fickle attention and on how the particular issue is framed” (38). Activists, attuned to the ways in which media can frame an issue, routinely fought to craft narratives through a kind of political theater in which protests were “staged” and injustice “dramatized” (King Jr. Reference King1964; Lewis Reference Lewis2017).

In social science, framing is employed to describe a wide variety of concepts (cf., Benford and Snow Reference Benford and Snow2000). I rely on two definitions from the larger framing literature. First, in keeping with prior work in sociology on social movements, I use the term frame construction to refer to the process by which activist rhetoric and behavior contributes to how outside observers, particularly media, make meaning of an event (Gamson and Wolfsfeld Reference Gamson and Wolfsfeld1993; Gitlin Reference Gitlin1980; Goffman Reference Goffman1974). Though movement-initiated frame construction serves multiple purposes, including mobilizing members of the ingroup, for this paper, I focus on the role of frame construction in “garnering [or dampening] bystander support” (Snow and Benford Reference Snow and Benford1988). Second, I describe media framing, or frames in communication that influence public opinion, as issue framing (Chong and Druckman Reference Chong and Druckman2007; Sniderman and Theriault Reference Sniderman and Theriault1992). In addition, to minimize confusion between the two types of framing, I often refer to social movement frame construction as staging. Table 1 summarizes the terms and concepts.

TABLE 1. Overview of Terms and Concepts

Movement-initiated staging or frame construction consists of the set of actions undertaken by subordinate group activists that influence the issue frames adopted later by the media. These actions could include everything from selecting which figures will embody a particular movement (e.g., Claudette Colvin or Rosa Parks), to recruiting particular types of protesters to be on the front lines of conflict (e.g., the Children’s Crusade in Birmingham) to crafting rhetorical appeals targeted to particular audiences (e.g., “Freedom Now!” or “Black Power!”). Among the most important forms of frame construction occur when movements, leaders, and protesters opt to endorse and/or initiate nonviolent or violent modes of resistance.

Initially, subordinate group leaders develop strategies amid enormous uncertainty. Activists are assumed to operate both with substantial constraints and with significant agency to advocate for particular approaches and tactics. Among the constraints, movement leaders confront incomplete information about which strategies will be most effective. In the 1960s, for example, the question of how to advance minority rights split architects of the black freedom struggle into two competing schools of thought, one more integrationist and the other more nationalist (Walton Reference Walton1971). The integrationist strategy focused on winning over persuadable members of the dominant group through tactics like nonviolent civil disobedience (Higginbotham Reference Higginbotham1994). King emphasized the critical role of the “white moderate” (1963). Bayard Rustin, a key organizer of the 1963 March on Washington, similarly argued for a kind of median white voter strategy (Frymer Reference Frymer1999):

The nationalist strategy rejected the focus on white moderates and emphasized “community control” or building autonomous bases of power. Beginning in the mid-1960s, the influence of traditional civil rights leaders was challenged by black nationalists like Stokely Carmichael and Angela Davis. These leaders advocated for progress through the development of black-led institutions and rejected the contingent power of coalition politics in which black progress depended heavily on appeals to white conscience and funding (Carmichael and Hamilton Reference Carmichael and Hamilton2008; Davis et al. Reference Davis, Olsson, Glover and Carmichael1972; Francis Reference Francis2019). Self-governance and nonviolence were not incompatible, as evidenced by the Black Panther Party Free Breakfast for School Children Program that, at its peak, served about 20,000 children in 19 cities (Araiza Reference Araiza2009). In response to brutal repression, however, nationalist advocates typically favored some version of the idea that “Negroes meet violence with violence as a means of self-defense” (Williams Reference Williams1962, 1–2). Among nationalists, these more militant tactics were considered legitimate and essential tools in the repertoire of resistance against institutionalized racism (Carmichael and Hamilton Reference Carmichael and Hamilton2008).

In the second period of Figure 2, the state and vigilantes react to the subordinate group protest. Activists can control logistical matters like the timing and location of protests but are otherwise assumed to have no direct control over the state response. Within the model, though, activists control the issue frames adopted by the press. State counteractions, such as the use of force, influence the degree of coverage but are predicted to have little effect on the valence of reporting. When state-aligned agents engage in violent repression against nonviolent protesters, in keeping with comparative work, the model predicts media coverage will tend to be sympathetic to the activists (Stephan and Chenoweth Reference Stephan and Chenoweth2008). Conversely, when protesters employ violent tactics against nonviolent majority group actors, coverage will tend to be sympathetic to the dominant group-aligned agenda. When both protesters and the state engage in substantial violence, the “tie” is predicted to go to the dominant group and media coverage will focus on concerns about order maintenance. In short, violence is a double-edged sword. State repression both subjugates activists and is predicted to focus media attention on the concerns of nonviolent protesters. In contrast, violence by protesters can powerfully express discontent and offer a means of self-defense but, in the glare of mass media, is expected to strengthen the coalition of those looking to thwart minority demands.

In the first case, i, nonviolent protest is met with state actors trying to minimize or restrain the government and vigilante capacity for force. In the absence of violence, the agenda seeding effect of a protest may be diminished but the valence of coverage is predicted to remain focused on “rights.” Branch (Reference Branch1998, chap. 15), for example, quotes one reporter’s take on a nonviolent civil rights protest in Hattiesburg, MS, “… even an unprecedented picket line was a dull story. ‘In such situations, blood and guts are news’.” Nevertheless, the staging of nonviolent protesters confronting nonviolent state actors should, on average, help construct frames about collective action as claims for rights. The paradigmatic example of this case is the March on Washington. Roberts and Klibanoff (Reference Roberts and Klibanoff2007, chap. 20) highlight that the New York Times, “in the second paragraph of its coverage, called the [March on Washington] ‘the greatest assembly for a redress of grievances that this capital has ever seen.’ It added: ‘There was no violence to mar the demonstration’.”

In the second case, ii, when nonviolent protest is met by state or vigilante repression, activists might be injured or even killed but the resulting media coverage is predicted to help the larger movement and undermine the dominant group-aligned coalition. Consistent with Gillion (Reference Gillion2013), the volume of coverage is predicted to increase but the media narrative is expected to remain focused on “rights.” Roberts and Klibanoff (Reference Roberts and Klibanoff2007, chap. 16) note “movement leaders and some segregationist leaders were studying the press: how it reacted, what made news, and what did not. One thing was unambiguous: the greater the violence, the bigger the news, especially if it could be photographed or filmed.” Following these insights, civil rights leaders intentionally selected both Birmingham, AL, and Selma, AL, as theaters of battle in which to contest segregation under the expectation that the local police chiefs with “hair-trigger temper[s]” would respond in ways that could shock the consciences of otherwise indifferent white moderates (Roberts and Klibanoff Reference Roberts and Klibanoff2007, chap. 22).

In the third and fourth cases, iii and iv, subordinate group activists use violent tactics. Irrespective of the state response, dominant group media framing is predicted to focus on order maintenance concerns like “riots” and “disorder.” In Albany, GA, for example, black-led largely nonviolent efforts at desegregation occasionally escalated to include protester-initiated violence against the intentionally restrained force led by Chief of Police Laurie Pritchett. On July 24, the New York Times front page covered a protest of “2,000 angry Negroes” and reported that “[b]ricks, bottles and rocks were thrown towards more than 100 city policemen…” (Sitton Reference Sitton1962). Following the violence, King and other leaders called off additional protests and declared “a day of penance” (King Jr. Reference King1962).

Similarly, in the fourth case (i.e., iv), violent resistance confronts violent repression. In line with Gillion’s informative cue model, this scenario is predicted to generate significant attention (Gillion Reference Gillion2013). Unlike the informative cue model, however, staging an event with violence is predicted to nudge the media toward a focus on “riots” rather than “rights.” News coverage of most of the violent urban civil disturbances of the 1960s fits this type, with the 1965 Watts uprising as one high-profile example. In addition to the historical record, these predictions also build on experimental work that finds “extreme” protest tactics reduce popular support for a movement (Feinberg, Willer, and Kovacheff Reference Feinberg, Willer and Kovacheff2020; McLeod and Detenber Reference McLeod and Detenber1999). In short, whether or not the state engages in violent repression, activists using nonviolent tactics are predicted to seed more rights-focused issue frames and activists using violence are predicted to seed issue frames that emphasize disorder.

In the third period of Figure 2, reporters, editors, and media outlets make choices through headlines, prose, photographs, audio, and video that not only convey facts but also help audiences organize information about protests into larger conceptual buckets (Chong and Druckman Reference Chong and Druckman2007; Gamson and Wolfsfeld Reference Gamson and Wolfsfeld1993). Framing typically draws on preexisting cultural motifs, stories, and beliefs (Tarrow Reference Tarrow1992), and two recurring issue frames for black resistance were as either a legitimate effort to fight injustice or an illegitimate act of lawbreaking within a long tradition of representing black life as pathological (Murakawa Reference Murakawa, Lowndes, Novkov and Warren2008). Though majority-oriented media are likely to exhibit systematic bias against minority-led movements (Davenport Reference Davenport2009), news organizations are also guided by a range of incentives including a desire to increase profits and to tell noteworthy stories that help attract audiences. Gamson and Wolfsfeld (Reference Gamson and Wolfsfeld1993) note, “social movements often make good copy for the media. They provide drama, conflict, and action; colorful copy; and photo opportunities” (116–7).

In the fourth period of Figure 2, both elites and masses are exposed to media coverage that, in turn, shapes their attitudes and behavior. Media agenda setting raises protest activity into the public sphere and issue framing contributes to more “rights”-focused or more “order maintenance”-focused shifts among moderates in relevant outcomes like public opinion, elite discourse, and voting. Through tactics like disruption, subordinate groups are predicted to upend the normal dominant group focus of the media to achieve a form of punctuated pluralism in which their concerns temporarily come to the fore of the press and, by extension, public opinion and policy making.

EMPIRICAL STRATEGY

My goal is to estimate a causal effect of protests on voting, with attention to possible differential effects by tactic type. In addition, I aim to demonstrate a plausible mechanism via media coverage. I begin the analysis using three statistical methods to test for a relationship between disruption strategies and the ultimate outcome of interest, voting behavior. First, with county-level data, I use panel models with county fixed effects to control for any time-invariant characteristics. In addition, to address potential spatial correlation of errors, I run spatial panel models that control for characteristics of neighboring counties (Millo and Piras Reference Millo and Piras2012). Second, to address heterogeneity of units, I use various matching and weighting techniques to approximate treated and control groups that are similar on all observed covariates (though may still differ for unobserved covariates). Third, to further address endogeneity, I use rainfall in April 1968 as an instrumental variable that approximates random assignment of violent protests. Finally, to assess whether the estimated shifts in voting mattered in electoral politics, I run a counterfactual simulation of the 1968 presidential election.

The agenda seeding model makes specific predictions about the effects of subordinate group protests on dominant group voting behavior. Aggregate measures of county-level demographics and voting may not accurately reflect the individual-level behavior of white voters (King Reference King1997). Consequently, I conduct a variety of “homogeneous place” and placebo tests. The hyper-segregation of Americans by race offers one solution to the ecological inference problem through a form of extreme case analysis (Duncan and Davis Reference Duncan and Davis1953; Lau, Moore, and Kellermann Reference Lau, Moore and Kellermann2007). By limiting some statistical analyses to counties that are 90% white, county-level data on Democratic vote share serve as reasonably good estimates of individual-level white voting behavior.

To test the mechanisms by which protests are influencing politics, I again use multiple methods. First, I present plots to show the punctuated patterns of protest activity correspond closely to changes in front page news coverage, elite discourse, and public opinion. I also present results from Granger causality tests that estimate whether trends in protest activity, front page news coverage, congressional speech, and public opinion are able to forecast one another. As a further test of agenda seeding, I show that violence by activists, even when controlling for violence by police, increases coverage and prominence in the New York Times. In addition, I show that protest tactics influence the language and frames employed by reporters, again, even when accounting for state violence. Across all the tests, I show differential effects by disruption tactic. Though the model predicts protests influence politics through news coverage, I do not use a mediation model as I lack county-level measures of news in the 1960s. Table 2 presents an overview of the questions, methods, and data. Main findings are presented below in coefficient plots or figures. Regression tables and additional robustness checks appear in the Online Appendix.

TABLE 2. Overview of Questions, Methods, and Data

Note: All data from 1960–72 except Q1a OLS & IV and Q1b, which use data from 1968.

DATA AND VARIABLES OF INTEREST

To empirically test for agenda seeding, I use data on protests, news coverage, public opinion, congressional speech, and voting behavior. I also use a variety of data sets from government sources to provide county-level demographic information and historic rainfall.

MECHANISMS

Are Activists Leading or following Elite Agendas?

How could protests change voting behavior? Almost no one directly observes a protest, and a large body of evidence suggests elites drive mass opinion (Canes-Wrone Reference Canes-Wrone2010; Lee Reference Lee2002; Lenz Reference Lenz2013; Zaller Reference Zaller1992). Might activists and voters simply be following elite cues? Or, alternatively, could protests—through indirect mechanism like media—co-opt typically elite channels of political communication to amplify marginal voices? To evaluate who is leading whom, I present descriptive plots and a summary of 48 Granger causality tests to assess the sequencing of black-led protests, news coverage, elite behavior, and mass opinion (i.e., Table 2, Q2). Figure 6 presents a scatter plot with loess-smoothed trend lines of nonviolent protest activity, front page news coverage, and public opinion on “civil rights” from 1960 to 1972. The left y-axis represents the scale for the percentage of respondents in a national survey saying that “civil rights” is the most important problem in America (i.e., a subset of data from Figure 1). In addition, the left y-axis presents the scale for the number of front page headlines, per month, that mentioned civil rights-related terms (as outlined above). The right y-axis also presents the scale for black-led nonviolent protest activity, calculated as the sum of the logged number of protesters, by month.

FIGURE 6. Nonviolent Protest Activity, Headlines and Public Opinion on “Civil Rights,” by Month

Note: Scatter plot with smoothed trend lines of nonviolent protest activity, news coverage of “civil rights” and public opinion about “civil rights” as the most important problem, aggregated by month.

The data in Figure 6 show that front page headlines mentioning civil rights covary closely with periods of nonviolent protest mobilization. For example, news coverage of civil rights rises and falls coincident with waves of nonviolent protest in 1960 during efforts to integrate southern lunch counters and in 1963 during the buildup to the March on Washington (8/27/63). Similarly, the spikes in 1965 co-occur with the “Bloody Sunday” march in Selma, AL, in March (3/7/65) and precede the Voting Rights Act, which was enacted on August 6, 1965. Figure 6 also provides evidence that news coverage of civil rights rises coincident with periods in which civil rights legislation is being considered by Congress. The Civil Rights Acts of 1960 (5/6/1960), 1964 (7/2/1964), 1966 (died in the Senate on 9/19/66), and 1968 (4/11/68) also affirm the more elite-centered models of news coverage and public opinion. Although year-over-year changes in the three trends are substantial, Figure 6 also shows the significant within-year variation obscured in analyses limited to annual data.

Several prominent elite-centered theories suggest that presidential campaigns were critical in driving support for racial liberalism in the early 1960s and counter-mobilizations in the latter part of the decade (Carmines and Stimson Reference Carmines and Stimson1989; Weaver Reference Weaver2007; Zaller Reference Zaller1992). Figure 6 includes vertical dashed lines for dates of presidential elections between 1960 and 1972. The trend in news coverage and public opinion preceding a presidential election is either fairly flat or declining, suggesting that presidential candidates and campaigns were not moving mass opinion on civil rights. In short, the patterns in Figure 6 offer suggestive if not conclusive evidence that subordinate group protest activity is helping to seed news agendas on civil rights and that elite-led legislation also drives coverage. Figure 6 does not, however, support the proposition that presidential campaigns are playing an important role in increasing the salience of civil rights in the media or with the public.

Figure 7 presents trend lines for time series data between 1964 and 1972 on monthly violent protest activity, monthly front page headlines that mention “riots” (and related terms), and irregular monthly polling data on public concern about “social control.” From about 1964 to 1968, the periodicity and the magnitude of violent protest activity covaries closely with newspaper headlines that mention “riots” and the percentage of opinion poll respondents who say that that issues of “social control” are the most important problems facing America. From a low in 1964, we see a trend of rising concern in the summer followed by a decline in the winter. On April 4, 1968, King was assassinated and violent protest activity and newspaper headlines spiked dramatically. Public opinion also increases but with a significant lag which may be an artifact of the absence of polling data on this question between January 1968 and July 1968. In short, consistent with the proposed model, from 1964 to the peak in 1968 and decline through 1970, violent protest activity, headlines, and public concern for “social control” covary quite closely. From 1970 onward, as black protest activity declines significantly and anti-war protests become more prominent (e.g., Kent State shootings, 5/4/1970), the three trends are less synchronized except around the Attica prison uprising in September 1971.

FIGURE 7. Violent Protest Activity, Headlines on “Riots” and Public Opinion on “Social Control,” by Month

Note: Scatter plot with smoothed trend lines of violent protest activity, news coverage of “riots” and public concern about “social control,” aggregated by month. Violent protest activity in June 1967 and April 1968 exceeds 200 on arrests scale. For visualization, those two data points plotted at half of measured values.

The results in Figures 6 and 7 include very limited data on elite behavior (presidential election dates) and therefore cannot directly disambiguate between top-down and bottom-up models of media coverage and public opinion. To assess whether elite action is preceding or following protest activity, Figure 8 presents a plot of the relationship between weekly violent protest activity and weekly congressional speech mentioning the word “riot” and “antiriot” (and related terms). The pattern in 1964, 1965, 1966, and 1967 shows an initial spike in violent protest activity, followed by a spike in congressional speech mentioning “riots.” In 1968, during the week of March 3rd, the pattern is slightly different as Congress considered anti-riot legislation proposed by President Johnson. Johnson’s proposal failed to gain traction and congressional discourse about “riots” quickly drops. On April 4, 1968, King was assassinated and violent protest activity then reached record levels. The following week, in keeping with the pattern of earlier years, elite discourse spikes and, substantively noteworthy, an anti-riot proposal backed by conservative legislators was incorporated into the 1968 Civil Rights Act, which passed on April 11. In short, congressional discourse about “riots” mostly follows violent protest activity. The relationship between nonviolent protest activity and Congressional speech about “civil rights” and “voting rights” is presented in the Online Appendix.

FIGURE 8. Violent Protest Arrests and Congressional Speech on “Riots,” 1964–72, by Week

Note: Line plot of weekly counts of congressional speech using phrase “riot,” “antiriot,” and weekly, black-led violent protest arrests from 1964 to 1972. Data sources: Congressional Record and Carter (Reference Carter1986).

Who’s Leading Whom?

To further assess whether subordinate group staging of nonviolent or violent protests is influencing elite behavior, I conduct bivariate Granger causality tests. In simple terms, Granger causality tests evaluate whether data in one time series improve our ability to explain variation in another time series. For example, if rainfall yesterday is a useful predictor of rainfall today, a Granger causality test assesses whether including one additional variable in the model, such as yesterday’s temperature, significantly improves our prediction for rainfall today. Following Granger (Reference Granger1981), I regress lagged values of X against present values of Y while also controlling for lagged values of Y. If the full model with lagged X explains significantly more of the variation in Y than the reduced model (with just lagged Y), X is said to “Granger-cause” or forecast Y.^{Footnote 11}

Figure 9 presents summary results for 30 bivariate Granger causality tests that assess whether trends in protest activity, front page headlines, congressional speech, and public opinion are predictive of each other. Arrows indicate the direction of the forecast and line weights indicate the degree of statistical significance (see legend). Contrary to models in which elites dominate political communication, Figure 9(A) also shows strong evidence of bottom-up effects in which subordinate group protests Granger-cause elite behavior. Put simply, nonviolent protest activity today is highly predictive of front page headlines and congressional speech about civil rights tomorrow. At the same time, consistent with an elite model, the results presented in Figure 9(A) suggest congressional speech and news coverage remain influential.

FIGURE 9. Two Diagrams Summarizing the Results of 48 Granger Causality Tests

Note:Granger causality test results summarized for (A) nonviolent protest activity and (B) violent protest activity. Arrows indicate whether one time series improves forecasting of the other. Results suggest both nonviolent and violent protest activity predict front page headlines and discussion in Congress. Nonviolent protest also predicts public opinion on civil rights. Line weights indicate statistical significance (see legend). No line between nodes indicates we fail to reject the null hypothesis of no additional predictive power. All tests used daily time series with lag of one day except for those involving public opinion which use monthly data (with interpolation). Estimates for effects of violent protests did not differ substantively across DCA and Carter data. Consequently, a third panel of six tests is not presented. In addition, no panels are presented for 18 placebo tests as only one result is significant.

Figure 9(B) tests the same relationships for violent protests and issues of social control like concern about “riots.” Again, consistent with the proposed model, trends in violent protest activity are predictive of trends in elite discourse and news coverage. Also, again, we observe some elite effects with congressional speech and news coverage forecasting other trends. Though Figure 7 suggests a strong relationship between patterns of violent protest activity and public opinion on “social control,” the Granger causality tests fail to reject the null hypothesis. This result might reflect the true absence of a relationship but could also be an artifact of issues such as that the two trends move together so closely that a lag of one month is insufficiently granular to detect how one trend predicts the other. In sum, the results of these Granger causality tests suggest that subordinate group activists can seed the agenda for news media and political elites.

I also conduct 18 placebo Granger tests to assess whether nonviolent protest activity forecasts news coverage of “riots,” public opinion about “social control,” and congressional speech about “riots.” Similarly, I test whether violent protest activity forecasts of news coverage, public opinion, and congressional speech about “civil rights.” Results in 17 of the 18 placebo Granger tests are not statistically significant (p > 0.05).^{Footnote 12} In other words, nonviolent and violent protest activity are only reliably predictive of language and attitudes that the dominant group associates with the particular tactics. This suggests the findings presented in Figure 9 are not simply spurious correlations. In addition, these placebo results suggest the specific pattern of nonviolent protest elevating concern about “civil rights” and violent protest driving concern about “riots” reflect meaningful relationships between tactics and terms employed by the press (full results in Online Appendix).

These findings are consistent with the predictions in Figure 2 that subordinate group agenda seeding helps to set mainstream news agendas and that the tactics employed significantly influence the issue frames adopted by the media. Cohen (Reference Cohen1963) argued the press “may not be successful much of the time in telling people what to think, but it is stunningly successful in telling its readers what to think about,” (13). The results presented in Figure 6 through Figure 9 broadly support Cohen’s insight. In the 1960s, however, this evidence suggests it was not just elites influencing media and mass opinion. Rather, the concerted efforts of black activists played a critical role in making the news.

Do Protest Tactics Shape Media Issue Frames?

Do different types of tactics produce different media issue frames (i.e., Table 2, Q3a)? Protests are typically contentious events in which media have considerable latitude to emphasize a range of interpretations (Davenport Reference Davenport2009). For example, both nonviolent and violent protests often involved illegal acts that could be described in terms that suggest either civil disobedience or criminal behavior. Similarly, no bright line distinguishes nonviolent and violent disturbances. Many nonviolent events were suffused with violence initiated by police and white civilians, while many events that escalated to protester-initiated violence often began as relatively peaceful, nonviolent affairs. This ambiguity allows for reporting to vary even for factually similar situations. Framing theory suggests that as journalists compose news stories, they rely on recurring organizing ideas and themes that structure how the public makes sense of a story and, often, imply particular public policies (Chong and Druckman Reference Chong and Druckman2007; Gamson and Wolfsfeld Reference Gamson and Wolfsfeld1993).

To further systematically test for a differential effect of protest tactics on how majority-oriented media frame these events, I selected 500 black-led protest events at random from the DCA data. Those protests included 385 events previously human-coded as nonviolent and 115 human-coded as occurring with some protester-initiated violence. A single event could be covered on multiple days and in multiple papers. On average, each query generated eight news articles for a total of 3,213 full-text documents after removal of any referencing the New York Times (2,461 nonviolent, 752 violent).^{Footnote 13} Those documents were converted to a machine readable format and the 25 most common terms from each corpus of articles was identified. Term frequencies were then calculated separately within the nonviolent or violent subsets and a ratio of nonviolent to violent term frequencies was estimated.

Figure 10 presents the ratios of 30 top stemmed term frequencies between the events previously human-coded as including nonviolent or violent protest tactics.^{Footnote 14} The top 15 rows of Figure 10 are terms that appear with greater frequency in articles about nonviolent events relative to violent events and the bottom 15, the inverse. The upper row of Figure 10 indicates that the term “march” is among the most common terms in articles covering nonviolent events and that it is about two-and-a-half times as common in articles about nonviolent versus violent events (2^1.3 ≈ 2.5).

FIGURE 10. Ratio of Term Frequencies in Articles about Protests Coded as Protester Nonviolent or Protester Violent

Similarly, the bottom row of Figure 10 indicates that the term “shot” appeared about three times as often in articles in which protester tactics were coded as violent as compared with those coded as nonviolent (2^1.6 ≈ 3). Stemmed terms like “presid[ent]” and “student” appeared with only slightly different frequencies in both sets of articles. Figure 10 also indicates that articles about nonviolent events were considerably more likely to include terms like “demonstr[ation],” “civil_right[s],” and “protest.” Likewise, articles about events that included protester violence were much more likely to include terms like “night,” “violenc[e],” and “polic[e].”

Term frequencies, however, only offer a crude sense of the content of the articles. Looking at the top terms in context further reinforces the results presented in Figure 10. For example, a seemingly neutral stemmed term like “store” is employed very differently depending on the event. In articles about events coded as nonviolent, about 46% do so when a store is the object of a boycott, sit-in or lunch counter protest. For events coded as including protester violence, “store” co-occurs in about 77% of the articles with “riot” and in about 54% with the stem “loot.” Similarly, the stem “nation” is part of the two-word phrase “national guard” in about 24% of the articles about nonviolent events and 52% of the articles about violent events.

What About Police Violence?

Do protester tactics influence media coverage even when accounting for police violence (i.e., Table 2, Q3b)? I assess whether events that include protester violence (compared with nonviolent protests) increase media attention even when controlling for the presence of serious violence by police. Violence by protesters is predicted to seed news agendas whether or not the state engages in repression. Figure 11 uses variation in the DCA data to test the relationship between violence and three measures of New York Times coverage of black protests: articles per event, paragraphs per article, and page number. Using Poisson count regression, each outcome is modeled as a function of a binary variable for protester violence, a measure of police violence (on a scale of zero to two), an interaction of the two violence variables, and year fixed effects to account for possible time trends in coverage. In line with civil rights leaders, Gillion (Reference Gillion2013), and the agenda seeding model, violence by activists, even when accounting for police violence, is significantly associated with more articles, longer articles, and stories positioned closer to the front page (full results in Online Appendix).

FIGURE 11. Marginal Effects of Police and Protester Violence on New York Times Coverage

Do Police Tactics Shape Media Issue Frames?

The previous results show police violence associated with increased coverage of protests, but does repression change how protests get framed by media (i.e., Table 2, Q3c)? The agenda seeding model outlined in Figure 2 predicts protester tactics, not police response, drive issue framing. To test for an effect of police violence, Figure 12 presents a subset of the news data in Figure 10 and focuses on the two cases in which only protesters or police use violence (i.e., cases ii and iii from Figure 2). Consistent with predictions, the results in Figure 12 suggest violent repression against nonviolent protesters remained associated with “rights” framed coverage, whereas violent protests met by a restrained state were associated with “riots” framed articles. In short, independent of police repression, black-led nonviolent events were able to repeatedly construct frames conveying a fight for rights. Despite widespread anti-black racism, media routinely covered these disruptions as legitimate civil disobedience within dominant group norms. While individual nonviolent activists were harmed by police violence, the joint effect of increased coverage and greater “rights” framing actually often hurt the white supremacist cause and helped the larger civil rights movement. Conversely, events in which protester-initiated violence occurred, irrespective of police response, were much more likely to construct frames that played to dominant group biases and invoke language associated with disorder and social control.

FIGURE 12. Ratio of Term Frequencies in Articles (Protesters Nonviolent and State Violent/Protesters Violent and State Nonviolent)

SUPPLEMENTARY MATERIAL

To view supplementary material for this article, please visit https://doi.org/10.1017/S000305542000009X.

Replication materials can be found on Dataverse at: https://doi.org/10.7910/DVN/HVRCKM.