Many powerful political movements arise from seemingly insignificant events that set in motion a cascade of consequences. In some cases, the process ultimately results in a change of government or the entire dissolution of a nation. Early theoretical studies struggled to explain the emergence of rebellions, since the reward they provide is a public good, whereas the potentially large costs of participation are borne by the individual (Tullock, 1971). Subsequent literature proposed a variety of explanations consistent with rational choice theory that can reconcile this seeming paradox of revolution. These explanations include bloc mobilization (Oberschall, 1994), uncertainty about the repressive capabilities of the regime (Boix & Svolik, 2013), and social preferences (Shadmehr & Bernhardt, 2011).Footnote 1 In a similar spirit, Kuran (1989) described a framework in which privately held and publicly voiced political preferences can diverge. This results in a bandwagon effect, where individuals hold their political views private until a sufficiently large number of individuals voice similar views.
A recent example of such a movement is Black Lives Matter (BLM), which, although officially founded in 2013, mushroomed into a global movement of an almost unparalleled scale following the death of George Floyd in police custody on May 25, 2020. Driven by concerns about perceived racial injustices, protests occurred across the United States, as well as in many cities worldwide.
Despite the large scale of the movement and its associated protests, little is known about its political consequences. Although the protests primarily targeted perceived racial injustices, they commonly involved calls to get out the vote and emphasized the importance of registering to vote to achieve political change (New York Times, 2020). Moreover, the protests received a large amount of media coverage across the political spectrum. However, this coverage was marked by a deep ideological divide, as some conservative commentators emphasized the occurrence of violent outbursts at some of these protests, seeking to reinforce their narrative that a Democratic government would threaten public safety (FiveThirtyEight, 2020).
These factors suggest that the BLM protests might have contributed in important ways to the record-breaking voter registration levels and turnout observed in the 2020 presidential election by encouraging voters in support of the movement, as well as those opposing it, to cast their vote. In this study, we focus specifically on the impact of local protests on the political mobilization of previously unregistered voters by comparing temporal patterns in voter registration across observationally similar communities with and without large-scale BLM protests.
The vast majority of US states require voters to register to vote, a procedure that has long been acknowledged as potentially detrimental to voter turnout, since it compels prospective voters to expend energy at a time when political interest is relatively low (Highton, 1997; Rosenstone & Wolfinger, 1978). Recent years have seen a variety of efforts to increase political participation, including the abolition of voter registration deadlines (Brians & Grofman, 2001), widespread registration drives (Nickerson, 2015), and automatic voter registration when a citizen engages with government entities (McGhee et al., 2021). Despite these advances, there remains a substantial population of eligible yet unregistered voters (Pew Charitable Trusts, 2012), particularly among low-income Americans (Brians & Grofman, 1999). This gap is highly relevant, since interventions aimed at increasing voter registrations have been shown to translate directly into higher voter turnout (Nickerson, 2015).
The above-cited research thus suggests that drivers of voter registrations are an important factor to study as we seek to understand political participation in the United States.
A distinct advantage of voter registration data over traditional measures of electoral participation is their availability with high frequency. Compared to biennial turnout data, this data availability considerably mitigates potential confounding. One might be concerned about the possibility that protests are endogenous to places where they maximize political mobilization due to unobserved factors, such as the potential for new registrations. As outlined by Azam (2019), such behavior would lead to a biased estimate of the effect of protests in purely cross-sectional regressions. By observing voter registrations in a panel, we can account for such unobserved factors if they are constant during the observed time period (Wooldridge, 2015). We argue that focusing on a short time horizon before and after the protests lends credibility to the assumption that confounding variables did indeed remain constant during our sampling period.
However, the use of voter registrations as an outcome also has some limitations that qualify our conclusions in important ways. First, registrations capture only the political engagement of previously unregistered voters. Although studying this population is interesting in its own right, its non-representative nature limits the extent to which findings can be extrapolated to the electorate as a whole (Jackman & Spahn, 2021). Second, the analysis of timing variation in voter registrations requires assumptions about why individuals prefer to register to vote at one point in time rather than another. While time-varying costs of registration are likely important (Cantoni, 2020; Kaplan & Yuan, 2020), we argue that the salience of political events can be a strong motivating factor, especially considering the availability of online voter registration in most states by 2020.
Our research contributes to several strands of the literature in economics and political science. Most notably, we analyze the impact of political protests on voter mobilization. This question has previously been studied by Madestam et al. (2013), who found that protests by the Tea Party movement led to a local increase in the vote share for the Republican party. We add to this body of knowledge by providing estimates on the local political mobilization effects of another large-scale political movement, using an alternative outcome and identification strategy. To the best of our knowledge, we are the first to estimate the effect of the Black Lives Matter protests on political mobilization.Footnote 2
We further contribute to the vast literature on how voters react to dramatic external events. For example, terrorism has been found to affect voting, even though the violence was committed by independent actors without the support of political parties (Geys & Hernæs, 2020; Montalvo, 2011). The lootings and riots that accompanied some of the BLM protests provided conservative commentators with a powerful narrative contending that under a Democratic government, lootings and riots would be the norm.Footnote 3 Alternatively, the BLM movement can be viewed as an expression of dissatisfaction with prior policy. Voters might be motivated by seeing many citizens openly demand more progress on racial equality and policing. The previous literature on retrospective voting has generally confirmed that voters hold policymakers accountable for failure to control crime (Arnold & Carnes, 2012; Bateson, 2012) or failures of the education system (Holbein, 2016). Experimental evidence has shown that the context and framing in the media matter in determining how voters attribute blame (Healy & Malhotra, 2013; Malhotra & Kuo, 2008). Since conservative and liberal leaning voters consume different news sources (Allcott & Gentzkow, 2017; Bakshy et al., 2015; Gentzkow & Shapiro, 2011), we can expect that they would receive different interpretations of the legitimacy of the Black Lives Matter protests, which might change the extent to which voters are mobilized.
Our results do not support the notion that local political protests affected voter mobilization, either in the aggregate or on either side of the political spectrum. Furthermore, we find similar null effects also for the subset of counties in which protests turned violent, although the considerably reduced sample size does not allow us to confidently rule out meaningful effect sizes.
Although our results stand in contrast to earlier findings in the literature on the mobilization effects of political protests, these differences might be attributed to the scale of the BLM movement and its extensive media coverage. Similar to the prior literature, our analysis cannot identify the overall impact of the BLM movement but, rather, focuses on the differential in the mobilization effects induced by local protests. The vast media coverage of the BLM movement might have reduced the importance of local exposure, thus contributing to the null effect we estimate in this study.Footnote 4 The major national cable TV networks spent almost 2.5 hours per day reporting about the protests on the weekend after George Floyd’s death (FiveThirtyEight, 2020). In addition, we show that despite considerable variation in interest across states, even areas with little exposure to local protests exhibited substantial interest in BLM, as measured by Google Trends data.
Section 2 of this paper describes our data sources and presents descriptive statistics of the sample; Sect. 3 describes our empirical strategy; Sect. 4 indicates our main results, heterogeneity analyses and robustness checks; and Sect. 5 provides a brief conclusion.