Abstract
Protest events affect public opinion on the issue of interest. However, the extent to which an individual’s proximity to protests impacts public opinion is less examined. Does a protest event occurring nearby, i.e., within an individual’s neighborhood, impact their opinion? Do protests that happen further away, perhaps in the next county, have the same impact on public opinion? This study analyzes the impact of exposure to protests by focusing on the Black Lives Matter (BLM) movement in 2020 using public opinion data from Iowa merged with protest locations in Iowa. Specifically, we examine public support for BLM and for defunding the police. We evaluate the role of distance through a discrete mileage cut-off and a distance decay function. Our analysis shows that people living closer to protests show greater support for the BLM movement in general and, to a less extent, for defunding the police. The results suggest that protests may affect public opinion, but only within a very narrow range of a few miles.
Introduction
On May 25th of 2020, George Floyd, a Black man, was murdered by an officer who kneeled on his neck for over 8 minutes. This sparked outrage across the country and a revival of the Black Lives Matter (BLM) movement in the media. From May 26 to August 22 of 2020, nearly eight thousand protests occurred in support of BLM with an estimated 15–26 million participants, making it one of the largest movements in the history of the United States. The BLM movement focuses on the unjust violence inflicted on Black people and communities, with a primary focus on police brutality. The organization originally began in 2013 in response to the acquittal of Trayvon Martin’s murderer and has grown since. The number of Black people killed by police in the U.S. is extremely high, with an estimated 241 killings in 2020 alone (Statista, 2022).
BLM protestors called for justice in 2020 by bringing up the previous murders of Michael Brown, Eric Garner, and Freddie Gray as well as more recent Black civilian killings including Ahmaud Arbery, Breonna Taylor, and Jacob Blake. This movement garnered significant attention in 2020 and led to nationwide conversations about criminal justice reform. While the media focused on the few violent occurrences around BLM protests, the large majority (94%) of these protests were considered peaceful according to the Armed Conflict Location and Event Data Project (ACLED) data (Raleigh et al., 2010). This significant movement was even able to ignite policy change such as the decision for some police departments to ban chokeholds and reallocate funds.
While previous studies have shown that protests have an impact on public opinion (Feinberg et al., 2020; Huff & Kruszewska, 2016; Metcalfe & Pickett, 2022), it is still uncertain whether BLM protest activities shaped public opinion toward the movement itself or on its signature issues. For instance, during a 2017 pipeline protest in North Dakota, Rep. Keith Kempenich said that the “protest was located 40 miles from any population center, so demonstrators were not going to get any attention unless they took dramatic action” (Jackman, 2017). Although this example comes from a separate movement, it clearly raises the question of how distance from protest events may influence their possible effect on public opinion. With thousands of protests occurring all across the United States starting in the summer of 2020, the BLM movement provides an excellent opportunity to evaluate this question.
The present study seeks to answer this question by considering the importance of the proximity of protests to individual opinion. Not all protest events are equally influential or important to the public. As previous studies on different policy areas or political events have discussed, closer events matter more than distant ones because they provide more opportunities to have direct experience, to obtain contextual information, and to increase familiarity with the topic. In addition, closer events can modulate how the public perceives information and have different consequences than more distant events. Protests, as one type of political event, will have some effects on the public based on how close or far the venues are from individuals.
To be specific, we examine the proximity of BLM protests to individuals and whether exposure to greater protest activity affects their opinion about the BLM movement and its goals. We use data from a survey of public opinion in Iowa and merge it with data from ACLED on protests occurring in Iowa. In terms of public opinion we use measures of public support for BLM and for defunding the police. We consider two measures of respondents’ proximity to protests: one that counts the number of protests occurring with a certain distance and a second that uses a distance decay function to weight closer protests more than distant ones. By comparing different distances and decay rates, we are able to get a more precise sense of how close is close enough to matter for public opinion. Our results match Rep. Kempenich’s observation that protests within a narrow range of just a few miles affect opinion, with stronger effects for opinion on the BLM movement than for defunding the police, whereas those further away have little effect.
Spatial Proximity and Public Opinion
Distance Effects on Public Opinion in General
The importance of spatial context is encapsulated in Tobler’s the first law of geography: “everything is related to everything else, but near things are more related than distant things (Tobler, 1970, p. 236).” Stated differently, closer subjects (e.g., political events, political agenda, or even politicians) are likely to have more influence than distant ones. While there are many ways to define the closeness, the current study will focus on physical distance and examine how physical proximity influences public opinion.
Previous studies have shown that spatial proximity to the item of interest matters (Boudet et al., 2018; Pulido et al., 2019). For instance, Branton et al. (2007) examined how voting behavior on nativist California ballot initiatives was affected by voters’ proximity to the US-Mexico border and find evidence that proximity to the border influences voting behavior, although the impact varies by partisanship. Boehmke et al. (2012) examined Indian gaming initiatives in California and found that spatial context and exposure to existing Indian gaming operations and Indian reservations affect how people voted on three ballot initiatives that sought to expand gaming opportunities for Indian nations.
The importance of spatial proximity on public opinion has buttressed other political topics as well. Cortina (2020), for example, found that closer distance to the US-Mexico border influences Republicans’ support for border wall construction. Examining American public opinion toward the military force abroad, Russett and Nincic (1976) found that people are more willing to militarily assist countries that are closer to the border. As highlighted by these studies covering a wide range of policy areas, incorporating the role of spatial context is important for understanding public opinion.
These studies take a couple of approaches to linking proximity to attitudes or behavior. To begin with, spatial distance has an impact on familiarity with the issue. People who live closer to the item(s) of interest, whether an international border or a casino, are more likely to encounter information related to the issue because they can be easily exposed to local media coverage due to the potential higher volume of articles or broadcasting (Branton & Dunaway, 2009).
The physical proximity could further allow individuals to obtain information by direct observation or experience (Cortina, 2020). In contrast, physical distance will hamper direct information acquisition and the ability to link information to the actual context; greater distance could even provide different information than is that obtained closer in (Branton et al., 2007). Regardless of the issue, therefore, residents of the area or residents of nearby areas have more chances to acquire specific information and be familiar with the item of interests.
Spatial proximity also affects information processing. Based on the collected information, individuals evaluate potential costs and benefits and modulate their opinion toward the issue of interest, with the spatial distance working as a weight in this calculation. For instance, with Not In My Back Yard (NIMBY) policies such as building hazardous facilities, people who live closer to the venue will place more weight on the potential cost; on the contrary, individuals who live substantially further away will consider more of the potential benefits if they are broader than the localized costs (Boehmke et al., 2012). While the specific policy context will decide the relative size of a distance-weighted effect, proximity to the item of interest is a key element of individuals’ information process. Distance, in turn, will have therefore have an impact on their evaluation toward the issue.
Third, spatial contiguity is closely related to the degree of consequences. Policy or political events have spillover effects, whether positive or negative. For example, building hospital infrastructure in a region will create more jobs; it will, in turn, improve economic conditions not only in that region but also in nearby areas. These positive spillovers, however, will be diminished as distance increases. With the case of hazardous facilities, for example, negative spillovers will decrease in more distant areas. Even when the issue of interest has administrative boundaries (e.g., state law), neighboring regions may be influenced due to spillover effects. The varying degrees of impact, in turn, can lead to different reaction from the public (Boehmke et al., 2012).
Our research is built upon previous studies which link spatial proximity to public attitudes or behavior. While the issue of interest or political context needs to be further considered, these studies argue that it is important to consider spatial distance when examining public opinion and behavior.
Protests and Public Opinion
Protests are political events that require public support not only to achieve their goals but also to sustain momentum for the associated movement or cause. As such, it is important to understand how protest events affect public opinion. The effects of protests and social movements on public opinion have been studied within different contexts such as the African American Civil Rights movements (Mazumder, 2018; Wasow, 2020), the women’s movement (Costain & Majstorovic, 1994), autism activism (Orsini & Smith, 2010), immigration politics (Branton et al., 2015), and government or domestic actor targeted movements (Murdie & Purser, 2017). While varying in terms of themes and time of occurrence, the findings echo that protests can alter public attitudes and even have lingering effects on public opinion even decades later (Mazumder, 2018).
With regard to BLM, Mazumder (2019) compared how racial attitudes changed pre- and post-protest by comparing countries that had experienced protests with those that had not. Mazumder (2019) found that BLM protests affect public opinion in terms of racial prejudice among whites. In addition, public opinion surrounding BLM movements is polarized. As the public gradually saw the police less favorably, counternarratives such as Blue Lives Matter and blaming brutality on “bad apples” arose to counter this shift in attitude, thereby polarizing public opinion further and restricting change (Banks, 2018). More recently, Updegrove et al. (2020) found that perceived discrimination, mistreatmeant of Blacks by police forces, and political affiliation have an impact on opinions towards BLM protests; specifically, red states and Republicans are more likely to show opposition to the movement.
Protest events differ in their tactics, thereby leading to possibly different impacts on public opinion. Social movements occur under the shared goals of protesters. To achieve these goals, however, it is essential to obtain support from non-participants such as the public or political officials. Previous studies have especially focused on nonviolent and violent tactics and found that the public perceives nonviolent protests as more legitimate (Huff & Kruszewska, 2016; Orazani & Leidner, 2019; Stephan & Chenoweth, 2008); in contrast, employing violent or unlawful strategies negatively affects public perception, leading to decreased popular support for protesters (Feinberg et al., 2020; Simpson et al., 2018). In other words, the level of protest violence has an impact on shaping the public opinion.
Studies of BLM protests produce similar findings. For instance, Metcalfe and Pickett (2022) conducted a survey experiment to investigate how the public would respond to violent tactics being employed at BLM protests during the summer of 2020. Protesters use of strategies such as delaying traffic, carrying firearms, and damaging property elevated fear in respondents. This will, in turn, decrease support for protesters while increasing support for police repression of the protest.
Spatial context also matters in shaping public opinion toward protest. Studies have found that proximity to protest events conditionally affects support or opposition to issues (Branton et al., 2015; Clarke et al., 2016) and influences an individual’s decision whether to participate (Traag et al., 2017). Specifically, Traag et al. (2017) assumed an exponential decay in the probability of participation as distance increases, indicating the unique impact that distance has on protests. In terms of BLM, Klein Teeselink and Melios (2021) found that counties with more BLM protesting activity experienced a shift in support for the Democratic candidate in the 2020 US presidential election.
Collectively, these and other studies show that distance to protests or other relevant locations has an impact on public opinion across various political topics and protest events. As such, it is important to consider the impact of physical or perceived proximity on public attitudes. Having the ability to evaluate from across a multitude of distances will prove potentially valuable for accurately determining how protest exposure influences public opinion regarding protest events.
Political Ideology and Protest Perception
As discussed, spatial proximity has an impact on direct observation, information gathering, information processing, and the degree of consequences; therefore, distance from an event or location can influence public attitudes or behavior. However, there is an important difference between closeness to a protest and closeness to hazardous facilities or other infrastructure. While subjective evaluation is required for individuals to make their own attitude toward all these examples, interpretation of protest events may offer more room for individuals to filter events through their own views, including their political ideology. Literature on frontlash (e.g., Weaver, 2007) further suggests that the impact of spatial proximity might be different based on the party affiliation. When a protest occurs, the event can evoke reactionary or counter-movement from people with different beliefs and purview. In many cases, the differences come from individual’s political identity. Studies on frontlash, therefore, suggest both the importance of political ideology and potentially different impacts of spatial closeness.
Along these lines, previous studies have shown that BLM is a highly partisan issue. In the 2016 presidential election, for example, both Hillary Clinton and Donald Trump explicitly shared their opinions about BLM (Glanton, 2016; Weigel, 2016). Similarly, legislators also differed in their views along party lines during the previous and most recent BLM protests. In general, Democrats were more closely affiliated with BLM protesters than Republicans (Lowery, 2015; Shah & Widjaya, 2020; Weigel, 2016).
Party identification is also an important factor at the individual level. Examining the 2016 BLM movement, Drakulich et al. (2021) and Updegrove et al. (2020) found that the opinion toward BLM differs according to an individual’s party identification. Republicans were much less likely to support the movement compared to Democrats. A Pew Research Center survey further showed that party identification plays a role in the respondent’s attitudes toward the BLM among the White respondents (Horowitz & Livingston, 2016). A more recent study about the partisan impact on the attitudes toward BLM also echoes the previous findings. Using the 2020 BLM movement as a case, for instance, Drakulich and Denver (2022) showed that the partisan impact still holds. These studies as well as literature about frontlash collectively suggest that it is important to consider the effect of partisanship on the BLM opinion. We, therefore, will further analyze how the impact of spatial exposure on the BLM opinion will differ based on the individual’s party identification.
Theoretical Expectations
As discussed above protest can influence public opinion (Feinberg et al., 2020; Huff & Kruszewska, 2016; Metcalfe & Pickett, 2022) in a variety of ways, with more than one of these mechanisms mapping onto proximity. In the context of BLM protests, some of these mechanisms matter more than others. For example, protests are unlikely to influence opinion through considerations related to possible spillover effects from siting a hazardous waste facility or building a new public park. In contrast, protests seem most likely to influence opinion through the information mechanism. Individuals do not necessarily have to participate in or even directly observe a protest; rather, it means that having protest events somewhere nearby, such as in one’s neighborhood, is substantively different from having protests 40 miles away in another city or county. The occurrence of a protest near an individual may draw their attention to the issue in question or shape local discussion about the issue. We discuss these in turn.
First, proximity can affect how individuals process information. This is because closeness, and therefore, exposure to the protest, can provide potential opportunities to learn the central argument of the protest, to have more background and contextual information about the protest, and to empathize with the protesters. As Cortina (2020) and Greenwald and Banaji (1995) suggested, direct and repeated exposures can shift one’s opinion. While we discuss potentially different mechanisms, they collectively highlight how proximity to the protest venue could impact information processing, and thus, evaluation of the protest.
Second, it is also important to consider potential media impact. Studies have shown that local and national media diverge in terms of how they cover local protest events. For instance, Neveu (2002) examined coverage of a farmers’ protest in Brittany during 1998 and found local news media tend to express more about the underlying social issues in their coverage. The proximity to the protest allows local journalists to incorporate their personal knowledge about the region; and thus, it leads to provide more comprehensive and engaged coverage. Local news coverage also regularly includes community-based protest events (Cottle, 2008). While BLM protests dominated media coverage, previous studies suggest that media stories (at least local media) will provide more detailed information about protests occurring within a community; this, in turn, implies that the public will have better understandings of (nearby) neighborhood protests.
Lastly, proximity to protest events can lead opinions to shift in various ways. First, spatial closeness can evoke sympathy, and therefore, make the protest be more morally persuasive. Closer distance allows more exposure to protest in general and repeated exposures could lead to positive evaluation toward the protest (see Greenwald & Banaji, 1995 p.10). In their study of the civil rights protests in the United States, for example, Andrews et al. (2016) found that white Southerners who were physically close to the sit-in movements showed greater sympathy for minorities and tended to acknowledge the legitimacy of the movement. Given that BLM protest fundamentally speaks to anti-racism, we could expect the similar impact of proximity.
Overall, our focus on the link between information and opinion in the context of BLM protests in Iowa leads to our first hypothesis:
As previously mentioned, the BLM movement centers on anti-racism. On top of that, the George Floyd case in 2020 and many other instances of unarmed Black people killed by police further brought another important agenda: defunding the police (Cobbina-Dungy et al., 2022; Jean, 2020). Protesters and other activists believe that the killing of unarmed civilians, especially Black people, occurred because law enforcing institutions are overpowered (Jean, 2020). Therefore, they support decreasing funding to these institutions (i.e., police). Given that defunding the police is discussed as a way to ameliorate racial disparities in terms of law enforcement, it is expected that the public’s opinion toward defunding the police will parallel their opinion towards BLM. In other words, the proximity to BLM protests will also affect the public’s opinion toward defunding the police. This leads to the second hypothesis:
Beyond these overall effects, the highly politicized nature of the BLM movement and calls to defund the police means that opinion on them was substantially determined by a respondent’s partisanship. Republicans were strongly opposed to both whereas Democrats were much more supportive of the BLM movement and, while still somewhat mixed on defunding the police, they were still more open to it than Republicans. Prior work on spatial exposure provides evidence for such heterogeneous effects by partisanship. Branton et al. (2007) argued that individuals closer to the Mexican border will perceive a greater threat of immigration and therefore be more supporting of nativist ballot measures. Further, the effect will be greatest for Democrats. Republicans likely support such measures no matter where they reside whereas Democratic opposition to such measures will decline with proximity to the border as the perceived threat from immigration increases. Applying this logic to the case of BLM suggests that exposure to protests will increase support among Republicans. Democrats will generally be supportive regardless of exposure to protests while Republicans begin with a baseline of opposition that could be reduced by protests. The effect for Independents will fall between the two partisan groups.
The effect of exposure to BLM protests will vary according to partisan identification.
Research Design
While these hypotheses can be tested using any movement or context (e.g., other states or countries), we focus here on the BLM movement. Because the BLM movement produced thousands of protests all over the country in summer 2020, individuals will experience widely varying exposure to protest activity based on where they reside. In the rest of this section we provide details on the data, protest exposure measurement, and methods used in this study.
Opinion Data on BLM
We collected data on public opinion regarding BLM as part of a broader survey of political attitudes targeting Iowans aged 18 and over. The survey was fielded online through a respected web panel vendor.
1
Quotas were set for eight age-by-sex groups as well as a separate quota based on urbanicity.
2
Responses were collected between January 13 to February 3 in order to reach a total sample size of 1000 respondents. We included two questions to capture attitudes towards BLM: (1) the degree of support for the BLM movement and (2) the degree of support for defunding the police in Iowa. Both questions use a 5-point scale: strongly support, support, neutral, oppose, and strongly oppose. We use these questions as our dependent variables in our analysis.
3
Figure 1 shows the distribution of responses, with those to the defund question shaded by the treatment condition. With the exception of the roughly 25 percent of responses to the “defund the police” version to the defund/shift resources question that were strongly opposed the responses are fairly evenly spread across the five-point scale.
4
Distribution of Opinions on BLM and Defund/Shift Resources. Note: Responses to defund question separated by treatment condition, i.e., “defund the police” versus “shift resources.” See text for details.
Data on BLM Protests
We obtained data on BLM protest activity from ACLED. 5 ACLED tracks political protests and violence worldwide and details factors such as actors involved, dates, locations, fatalities, and violence. ACLED finds that there were over 7,750 BLM-related demonstrations in more than 2,440 locations across all 50 states between May 26 and August 22 of 2020. For the purposes of our research we filtered the data to only include protests that occurred from January 2020 to May 2021 in the state of Iowa. We also collected the same information for the six states that share a border with Iowa to capture protests close to Iowans that occur outside the state. To ensure that the resulting data pertained solely to BLM, we reviewed the description of each protest. Because the survey data we use was collected in starting in January, 2021, we only included protests through January 8, 2021.
In total we identified 145 BLM protests in Iowa and 1,297 in all seven states. Out of these protests, 12 in Iowa and 73 total were labelled as violent. Compared to other states, Iowa has a slightly greater than average number of protests given its population size, with about 1% of the total population of the 50 U.S. states and about 1.5% of all BLM-related protests in the ACLED data. These protests occurred in 43 different different cities spread widely across the state of Iowa, with the highest numbers occurring in Des Moines (38 protests), followed by Iowa City (24 protests) and Cedar Rapids (13 protests).
As the counts in these cities indicate, BLM protests tend to occur in larger cities. To explore the location of protest events further, we aggregated them to the county level to compare to county-level features. Notably, we find that protests occurred more often in counties with higher black population and with higher vote share for the Democratic candidate in the 2020 Presidential election. 6 For instance, the five counties with the highest number of BLM protests (Polk, Johnson, Linn, Black Hawk, and Scott) are also ranked among the top six of all 99 Iowa counties for both Democratic candidate vote share and the percentage of Black population in the county. 7 In terms of correlations, the number of protest events strongly positively correlates with both Democratic vote share (ρ = 0.632) and the proportion of Black population (ρ = 0.692).
Measuring Exposure to Protests
Given the large number of protests in our database we need a way to succinctly capture their influence on respondents’ opinion on the BLM movement. To do so we begin with the assumption that protests that occur closer to where a respondent lives will have a greater influence than those that occur further away. To measure the distance between respondents and protest activity we use the geocoded locations of protests and respondents. Protest latitude and longitude are included in the ACLED data based on the city in which the protest occurred. For survey respondents, we use their self-reported zip code, which allows us to merge in information on the latitude and longitude of the centroid of each zip code. With this information we can then get an approximate measure of the distance between every respondent and every protest. Figure 2 shows the locations of protests and respondents in Iowa, with cities marked by circles sized and shaded according to the number of protests and counties shaded by the number of survey respondents. These show that both are spread broadly across the state but more common near major population centers. BLM Protest Locations by City and Survey Respondents by County. Note: Created using ArcGIS. Circles sized according to number of protests occurring in a city. Counties shaded by number of survey respondents.
Because we have many protests that happen in Iowa we require a measure that summarizes an individual’s experience of BLM protests. If there were one protest, we could just measure everyone’s distance to that location. But with multiple protests we want to capture more information. We could just use the distance to the closest protest, but that would discard a lot of information and heterogeneity across individuals. Some individuals might have multiple protests that occur near them whereas others may have just one in their immediate area. We therefore rely on a measure that incorporates all protests but which conforms to our notion that closer protests matter more. Consistent with earlier work (e.g., Boehmke et al., 2012) we construct a weighted measure of overall protest exposure. This measure adds a decay component so that more distant protests contribute less to the measure than nearer ones. Further, the researcher can control the rate of decay in order to adjust the relative importance of more distant protests.
Formally, our measure of exposure for individual i to protest j, E
i
j
is calculated as follows:
In order to see how this measure works, Figure 3 shows how protests at different differences contribute to the overall measure of exposure for different values of the decay parameter. For this example we assume five total protests at distances of 1, 3, 4, 7, and 12 miles. The plot on the left sets the decay parameter, δ, equal to 0.4 while the one on the right sets it to 0.15. Thus the one on the left counts more distance protests relatively less than the one on the right. The black line in each plots shows the decay function for the corresponding decay parameter, with the one on the left much lower than the one on the right. For each of our five protests we indicate their contribution to decay with a dropped-line plot and report the value of the decay function at each distance. Thus in the left-hand plot protest 1 contributes 0.67 to exposure, protest 2 contributes 0.301, and so on until protest 5, which contributes a meagre 0.008. By 12 miles, then, protests contribute effectively nothing to an individual’s measure of exposure. The total exposure for this value of the decay parameter is 1.24. Illustration of Exposure Measure, varying δ.
In contrast, the total exposure in the right-hand plot rises adds up to 2.56 due to its smaller decay parameter. For example, the closest protest’s exposure rises from 0.67 to 0.86 and the furthest protest’s value increases from 0.008 to 0.165. These changes also indicate that the relative contributions of protests also change. For example, when δ = 0.4 the second protest adds less than half of the first protest whereas when δ = 0.15 it adds nearly three quarters of what the first protest contributes. Given these differences we need to pay attention to our choice of the decay parameter since it affects the relative importance of each protest to overall decay as well as determining whether distant protests contribute meaningfully.
Methods
To evaluate the effect of our measure of protest exposure on opinions regarding BLM and defunding the police/shifting resources we include it as an independent variable in a regression analysis. Because our outcome variable is an ordered opinion measure, we estimate an ordered logit model. We adjust for survey weights in our analysis and estimate robust standard errors. All analyses and interpretation were done in Stata.
We further include a series of variables drawn from the survey to account for other features of respondents that likely predict responses to our two questions. Income is a categorical variable with nine groups of total family income in 2020. Education captures the respondent’s education level across seven ordered categories. We include a binary variable female, where male is coded as 0 and female is coded as 1. Given the centrality of race in the BLM movement, our race variable focuses on three categories: White, Black, and other. For the respondent’s party affiliation, we differentiate between Democrats, Republican, and other, which includes Independents, no preference, and other party preference; given that BLM is politically polarized issue, it is essential to include party affiliation (Updegrove et al., 2020). Lastly, we consider the respondent’s community type (urban, suburban, rural), age group (18–34, 35–54, 55–69, and 70 and older, based on their reported birth year), and how much they follow the news. To capture the effect of the question wording experiment on average support for the defund item, we also add an indicator variable for which treatment respondents received.
Results
Our results include estimates from four different sets of models using the two outcome variables and the two measures of exposure. To recap, we expect that survey respondents with more BLM protests occurring near them will express lower levels of opposition to BLM and defunding the police/shifting resources (hereafter we drop explicit references to shifting resources for brevity). Since we don’t know how close is close enough in terms of protest distance, we run multiple models for each measure. For the count of protests occurring within a given radius, we set the radius to one, two, three, four, and five miles. For the continuous distance-decay measure, we varied the decay parameter from 0.5 to 0.9 by steps of one-tenth for the BLM model and from 0.3 to 0.7 for the defund model. We considered a wider range of values for each model, but present these here since we these ranges bracket the best-performing values (in log-likelihood terms) while also showing how the results change with different values.
Opposition to BLM by Number of Protests within a Given Distance.
Note. Standard errors in parentheses. ∗p < .1, ∗∗p < .05, ∗∗∗ p < .01. Base categories as follows: white (race), other party or no party (party ID), urban (community type), and 18–34 (age).
Opposition to Defunding the Police by Number of Protests within a Given Distance.
Note. Standard errors in parentheses. ∗p < .1, ∗∗p < .05, ∗∗∗p < .01. Base categories as follows: white (race), other party or no party (party ID), urban (community type), and 18–34 (age).
Turning to our other variables, opposition generally increases among older age groups, as respondents move from urban to suburban or rural communities, and for those that follow the news more often. Partisanship also plays a strong role, with Republicans exhibiting greater opposition and Democrats lower opposition compared to voters not associated with either of the two major parties. Female respondents show significantly lower levels of opposition only for the BLM item. In contrast, education only increases opposition to defunding the police. Notably, we do not find significant effects for race, including among respondent who identified as Black. While the coefficients estimates are sizeable, the standard errors are even larger, quite likely due to the relatively small proportion of Black respondents in Iowa and in our data. Finally, our treatment indicator for the defund item shows a large increase in opposition for respondents who were asked about explicitly about defunding the police as opposed to the alternate phrasing of shifting resources. 8
Opposition to BLM by Protest Exposure Decay Parameter (δ).
Note. Standard errors in parentheses. ∗p < .1, ∗∗p < .05, ∗∗∗p < .01. Base categories as follows: white (race), other party or no party (party ID), urban (community type), and 18–34 (age).
Opposition to Defunding the Police by Protest Exposure Decay Parameter (δ).
Note. Standard errors in parentheses. ∗p < .1, ∗∗p < .05, ∗∗∗p < .01. Base categories as follows: white (race), other party or no party (party ID), urban (community type), and 18–34 (age).
What do these results mean substantively in terms of the effect of exposure on support for BLM or defunding the police? To convey this we create a figure to show how individual protests contribute to exposure through our decay measure and then create a series of figures to show how exposure translates into support. Figure 4 starts by combining the decay effect with the coefficient on exposure to show how a protest at a given distance contributes to the exposure measure. For illustration we set the decay parameter to 0.8 for the BLM question and 0.5 for the defund question. We then calculate the decay function for protests located up to 10 miles from the respondent since the decay effect has effectively reached zero by then. We then multiple each decayed effect by its estimated coefficient to compare how they contribute to a respondent’s latent opinion on BLM or defund support. The coefficient for BLM is greater in magnitude, so this curve starts out roughly twice as large as the one for defunding the police. Its decay parameter is larger, however, so it decreases more quickly, crossing the one for defund at a distance of two miles. At the point the contribution of a protest to latent opposition to BLM has fallen by 75% and to defund by 50%. Protests as little as six miles away contribute nearly nothing. Recall that this figure shows the contribution of a single protest at a given distance. Since respondents may have multiple protests near them, the total effect comes from the sum of each of those individual protests’ effects. For example, two protests at two miles will change opposition to BLM by the same about as one protest at one mile. Estimated Contribution of a Protest to Latent Opinion by Distance to Respondent. Note: This figure plots the value of 
Figure 5 makes the switch from describing the contribution of individual protests to illustrate the total effect of exposure on opinions. Here we plot the probability that a respondent strongly supports BLM or strongly supports defunding the police across the range of the protest count or exposure measures for a single value of the decay or distance parameters. Two notable features emerge. First, the increase in probability over the range of the variables is greater for the exposure models than for protest count models. As noted earlier, the decay measure seems to better capture the combined effect of protests as varying distances. Second, the probability of strong support increases more for the BLM question than for the defund the police question. These plots illustrate the cumulative substantive effect of the two curves shown in Figure 4 when accounting for all protest activity. The results show that the probability of strong support for BLM increases from 22% with no exposure to about 42% for respondents near its maximum observed value. For defunding the police the same probability increases from a small, but not insubstantial 14–22%. Effect of protests on support for BLM or defund.
Even smaller increases occur for protests within three miles: strong support for BLM increases from 22 to 33% and for defunding the police it increases from 14 to 16%. Despite the wide range of values for exposure, it’s worth noting that most respondents have small values of exposure: nearly one-third less than 0.01 and four in five less than one. Thus a more realistic change in this variable would produce a much smaller effect than we see over the reported range.
Effects of Protests by Party Identification
Our first set of models provides support for our first and second hypotheses about the overall effect of protest exposure on opinions. We now evaluate Hypothesis 3 to determine whether the effect of protests differs according to a respondent’s partisan affiliation. We estimate variations on the models just reported in which we interact our measures of exposure with our three-valued partisanship variable. This estimates distinct effects for Republicans, Democrats, and other partisans (who are mostly independents). Rather than report five models for each of our two dependent variables and two measures of protest exposure, we report two models for each outcome using one value of our exposure measure. Since they performed the strongest earlier, we use a two-mile radius for our count measure and a decay parameter of 0.7 for our decayed exposure measure. The reported results tend to be representative of those for other values of these parameters, though as in the first four tables the p-values tend to increase as the parameter gets further from what seem to be the best values. We include a full set of models in our replication materials.
Opposition to BLM and Defunding the Police with Effect of Exposure by Respondent Partisanship for Select Values of Protest Distance and Exposure Decay.
Note. Standard errors in parentheses. ∗p < .1, ∗∗p < .05, ∗∗∗p < .01. Base categories as follows: white (race), other party or no party (party ID), urban (community type), and 18–34 (age).
In contrast, the two models for defund show smaller effects overall. The only significant result emerges for Democrats using decayed exposure. While providing some evidence of distinct effects by partisanship, the results differ from those for the BLM question and therefore do not support a general pattern across both questions. One possible explanation is that opposition to defunding the police starts out much higher for Republicans and other partisans, with 83% of the former being opposed or strongly opposed to the “defund the police” treatment for this question and 54% for the “shift resources” version. A much lower 51 and 19% of Democrats fall into those two categories, so perhaps members of that party have more potential for influence from protests.
Overall, then, our results provide some evidence for heterogeneous effects of protest exposure by respondent partisanship. Yet the patterns differ for the BLM and defund questions, providing no clear pattern of results. Further, while we find differences in the magnitude and significance of the coefficients by partisan group, in none of the models did a joint F-test for the null hypothesis that the baseline model without interactions was correct. Given this we conclude that the evidence for heterogeneous effects by partisanship are suggestive at best.
Robustness Checks
In addition to the model reported here, we also considered a variety of other specifications in order to evaluate the robustness of our results. As noted before, we separated the exposure effect for the defund/shift resources question by the treatment condition and found similar effects in both groups. Our first set of robustness checks focuses on features of protests themselves, e.g., whether they were violent or when they occurred. We considered protests that involved violence by including an additional exposure variable for violent protests only. 9 While there were just 73 violent protests (12 of which were in Iowa) the correlation between total exposure and exposure to violent protests was over 0.93 for all values of the decay parameter. The results indicated a more negative effect for violent protests, but the high correlation meant that the effect could not be distinguished from the effect of all protests. Similarly, we also separated out recent protests to see if they had a distinct effect. Because most protests happened during summer 2020, we considered two definitions of recent protests: those after August 1 (22% of all protests) and those after June 15 (45% of all protests). The results for exposure to those after August 1 showed a larger negative effect relative to total protest exposure, but high correlations (all above 0.89) again meant that the differences were not significant. The results for the June split differed, with the effect of all protests still negative, but a positive coefficient for protests after June 15 that was occasionally significant and that produced one net positive and significant (p < .05) effect for recent protests for the defund analysis when delta was 0.3. Simultaneously, the negative effect for exposure to all protests became significant at the .05 level for values of delta below 0.5.
Our second set of robustness checks focused further on the role of race given the high proportion of white respondents in Iowa. To account for local context that might explain protests and opinion, we added a measure of local racial diversity with the percentage of the respondent’s county that was Black (in lieu of a zip-code level measure). That variable never had a p value below 0.75. We then restricted our analysis to self-identified white respondents, which reduced our regression observations by 62. The coefficients for the effect of exposure on opposition to the BLM movement increased by a little over 20 percent and were still significant. In contrast, the coefficient estimates for opposition to defund showed a slight decrease that, couple with a mod-est increase in the standard errors, resulted in insignificant effects. Lastly, we estimated a model that added in a measure of how often respondents received news from social media sites. Adding that variable to our base model led to insignificant effects of exposure among the full sample of respondents, but did not change the findings just reported for the subsample of white respondents.
Conclusion
Distance matters in public opinion across various political areas. Physical proximity to the issue or activity of interest provides individuals more chances to obtain relevant information, affects how individuals perceive information, and influences the degree of consequences. Either separately or collectively, each of these elements has an impact on individuals’ attitudes or behavior toward the issue. In the present study, we evaluate the effect of exposure to BLM protests in 2020 as the political context in which to examine the role of spatial proximity in public opinion.
Our analysis of public opinion on BLM shows that people with greater exposure to protests show greater support for the BLM movement in general and, to a lesser extent, more support for defunding the police, one of the movement’s main goals during the summer of 2020. We find an effect whether we count the number of protests within a fixed radius or whether we measure exposure with a continuous decayed function. In both cases, however, we find an effect only for protests occurring within just a few miles of our survey respondents. Thus protests may affect public opinion, but only within a very narrow range.
This finding suggests an important caveat for our findings. Our measure of distance to a protest is based on the centroids of the respondent’s self-reported zip code and the city in which the protest occurred. At the range of relevant distances found in our results, i.e., within a three miles at most, the accuracy of this measure will vary with the size of the zip code or city and where a respondents lives relative to the centroid. Further complicating our measure is that, as those of us living in Iowa City can attest, the protests often involved marching over fairly long distances, so that they might encompass more than one mile and more than one zip code. A further caveat is that our results include only respondents to a survey of Iowans. This might affect our results in unknown ways. Further work might therefore identify data in other states. National survey data would allow us to better explore this, though we would require sufficient data to examine state-specific effects and would also require comparably fine-grained detail on respondent location to create our distance measure.
Building on the current study, which shows the importance of spatial proximity, future work could explore the possibility of heterogeneous effects of exposure based on respondent characteristics, features of respondent community, or features of the protests. Our final analysis shows that the effect of BLM protests varied with respondent partisanship, though the affected groups differ for opinion on BLM and defunding the police. While the differential effects for BLM opinion line up with previous work on immigration opinions, the effects may still depend on the policy and protest context. Further, in our robustness analysis, we find that the effects generally increased when we limited our analysis to respondents who identified as white. The limited number of non-white respondents in our data preclude us from systematic conclusions here. Moving from respondent to context effects, media coverage of the protests in the respondent’s community could be considered. Undoubtedly, media is an important source of information to the public; and previous studies have shown the type of media coverage could be different based on spatial proximity. For instance, when examining immigration coverage, Branton and Dunaway (2009) found that news outlets closer to the issue of interests (i.e., the U.S.–Mexico border) published more articles about illegal immigration. In a similar vein, protest coverage could be different based on the distance, and the information about the protest that respondents receive could further affect their opinion on the protests. We also suspect that racial context of the respondent community will influence opinion because having more Black neighbors could lead to sympathy to the BLM movement. Examining whether community features lead to racial sympathy or racial threat will be another fruitful next step. Lastly, the type of protest could affect public attitudes. While the current study could not distinguish a distinct effect for peaceful and violent protests, future work with greater variation in protest type may be able to better test the finding that the public is less supportive of violent or illegal protests (Gutting, 2020; Metcalfe & Pickett, 2022). Adding how proximity plays a role in opinion toward peaceful or violent protests will help extend our understanding how distance matters in protest and public support to the protest.
Footnotes
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 University of Iowa Public Policy Center provided support for this poll as part of its student success mission, intended to expand experiential policy-relevant research for Unversity of Iowa students.