Introduction

The Black Lives Matter (BLM) movement began as a social media hashtag and protest chant that evolved into a mass protest movement of international renown (Morris, 2021). However, as has been observed in America and other nations with racialized social stratification, notable gains in mobilization for racial justice claims-making by subordinate ethnic groups is often accompanied by backlash from dominant ethnic groups that can evolve into racial justice countermovements (Andrews, 2002; Lee, 2002; Main, 2018; Parker & Barreto, 2014; Weaver, 2007). Such may be the case for BLM: in late 2014, a group of police officers formed “Blue Lives Matter” (BlueLM) for the stated purpose of providing moral and financial support for law enforcement officers and the families of officers killed in the line of duty (Solomon & Martin, 2019). Following its founding, BlueLM gained in membership and popularity in the United States, holding rallies and protests across the country, garnering national media attention, generating prominent symbols and merchandise sales, and pushing for the passage of federal and state laws that classify attacks on police officers as a hate crime (Craven, 2017; Czachor, 2020). A poll conducted in September 2020 found that 68 percent of likely voters were concerned about deadly attacks on the police and 59 percent supported the adoption of a BlueLM law in their state.Footnote 1 In 2021, there were several high-profile BlueLM rallies (Butler, 2021) and a controversial “Back the Blue” law was enacted in Iowa that strengthened legal protections for police officers while augmenting the penalties for rioting and unlawful assembly.Footnote 2

In this article, we explore the sources of public support for BlueLM and the antecedents of BlueLM activity. We explore these issues with an eye toward rendering empirical evidence able to support or oppose the popular claim that BlueLM is a countermovement to BLM. The sociopolitical status of BlueLM has been a debated issue in the American political scene. On one hand, opponents of BlueLM view it as a countermovement motivated by racial prejudice and opposition to BLM (Cooper, 2020; O’Leary, 2020; Smith, 2020; Thusi, 2020). Indeed, critics of BlueLM highlight its founders’ defense of the Ferguson police officer who killed Michael Brown,Footnote 3 as well as the predominately-White composition of BlueLM rallies and the growing association between the “thin blue line” (TBL) flag and White supremacist groups (Rossman, 2017). Myriad reports exist of prejudiced and violent activity at BlueLM rallies, including the use of racist language and physical attacks (Peirera 2020) and a videotaped reenactment of the police killing of George Floyd (Bella, 2020). The TBL flag became a staple image at campaign rallies held by Donald Trump during the 2020 Presidential Election (Schulte, 2020), which added to claims that BlueLM became affiliated with the far-right and took on racial undertones (Valencia, 2020). Finally, scholars have characterized BlueLM as a countermovement (Shanahan & Wall, 2021; Soloman et al. 2021) largely because it possesses definitional features of a countermovement as stipulated in research on movement-countermovement dynamics (Meyer & Staggenborg, 1996).

On the other hand, supporters of BlueLM consistently depict the movement as apolitical and race-neutral. For example, leaders within BlueLM claim that the movement does not oppose the fight for racial justice and that their rallies are not anti-BLM. Rather, BlueLM supporters claim the movement is simply “pro-police, pro-law and order” and motivated to express respect for the difficult and dangerous job of police officers (Offenhartz, 2020). In fact, the pro-police activist behind the TBL flag, who became the president of one of the largest online retailers of pro-police merchandise, stated that, “The flag has no association with racism, hatred, bigotry” and that the flag is “not a direct reaction” to the BLM protests (Chammah & Aspinwall, 2020). And, to the extent that the BlueLM slogan or associated symbols are embraced by far-right extremists, spokespeople for BlueLM argue these are isolated instances that do not embody the original intent of the movement. Indeed, leaders of the movement claim that the majority of BlueLM supporters are friends and family of the nation’s 800,000 law enforcement officers, as well as those in the “silent majority” who are concerned about rising crime rates and appreciative of the service provided by the police (Valencia, 2020). On this point, supporters of BlueLM point to the growing rate of assaults on officers, and ambush-style attacks in particular, as a primary motivation for their continued mobilization. In support of this claim, data from the FBI’s Law Enforcement Officers Killed and Assaulted (LEOKA) program show that the number and rate of civilian assaults on uniformed officers rose from 49,851, or 9.3 percent of every 100 officers, in 2013 to 56,034, or 11.8 percent of every 100 officers, in 2019.Footnote 4 Countering such claims, however, is the argument that in a racially stratified society premised on White supremacy (Alexander, 2020) and White cultural centrality (Aguirre & Turner, 2010), Whites can exercise their power by responding to organized Black challenges to racial injustice with race-neutral statements of “support for the system.” In this way, Whites can leverage a status quo affording them racial privilege to express “know-your-place aggression” (Mitchell, 2018) against marginalized groups while obscuring the existence of White-privileging racial hierarchy.

We theoretically and empirically adjudicate the status of BlueLM as a potential countermovement to BLM. BLM is commonly understood as an anti-racist Black-led social movement engaging in claims-making activity aimed at achieving racial justice by ending systemic racism and holding police accountable for violence against Black civilians (Taylor, 2016). Drawing upon the movement-countermovement literature, the large corpus of research on White racial attitudes, and recent work on public opinion toward BLM, we develop and test the following propositions related to the status of BlueLM as a potential countermovement: (1) BlueLM is predominately a White-backed movement, (2) support for BlueLM is systematically linked to White racial prejudice, and (3) BlueLM activity is caused by BLM protest. To test these propositions, we employ correlational and causal analyses on a wide array of data, including public opinion, news and social media, internet search interest, and corporate merchandise sales.

First, we find that White survey respondents report higher levels of support for BlueLM than non-Whites, and that BlueLM merchandise sales—specifically, TBL flags—were more numerous in locales with a higher density of White residents. These findings lend support to the claim that BlueLM is a White-backed movement. Second, using the universe of over 2.7 million comments on Twitter containing the hashtag “#BlueLivesMatter,” we find that expressions of support for BlueLM are more prevalent in locales with more racially prejudiced Whites. This finding is reinforced with survey data, which reveals that support for BlueLM among White respondents is higher among those reporting higher levels of racial prejudice. These findings lend support to the claim that BlueLM activity and support derives in part from racial prejudice. Third, using a regression discontinuity approach, we find that a random shock of BLM protest, triggered by the unanticipated police killing of a Black civilian, caused drastic increases in BlueLM internet searches and news coverage, expressions of support for BlueLM on social media, and purchases of TBL flags. In contrast, we fail to find consistent evidence that six different randomly-timed ambush-style attacks on police officers caused upticks in BlueLM activity. These findings lend support to the claim that BlueLM is reactionary to BLM protest, and comport with the conceptualization of BlueLM as engaging in a process of “competitive victimhood” (Solomon & Martin, 2019; Young & Sullivan, 2016) with the BLM movement.

Taken together, the findings presented in this article position BlueLM as a new incarnation of a deep historical pattern of White countermobilization against Black racial justice claims-making activity in the U.S. Indeed, mobilization around the issue of “law and order” is the predominant form that White resistance to Black claims-making has taken since the success of the Civil Rights Movement in ending formal segregation (Alexander, 2020; Weaver, 2007). Thus, our findings suggest that the stated focus on “law and order” by BlueLM renders it a new movement employing a decades-old strategy: assemble racially resentful Whites in response to organized Black claims-making activity using a mobilizing frame and set of goals that are not explicitly anti-Black but instead arguably race-neutral. To be sure, while it is beyond the scope of our findings to conclude that any support for BlueLM derives from racial prejudice, our findings demonstrate strong and consistent associations between Whiteness, racial prejudice, and support for BlueLM. At the very least, our findings suggest that support for BlueLM represents a new chapter in the constantly-morphing “coded” expression of prejudice in America, where support for BlueLM has evolved into a vehicle for the channeling of racial animus akin to support for the “Official English” movement (Citrin et al., 1990), “three strikes” laws (Johnson, 2008), birther conspiracy (Pasek et al., 2015), Tea Party movement (Parker & Barreto, 2014), and “Southern heritage” (Strother et al., 2017).

Black and Blue Lives: Race, Prejudice & Competitive Victimhood

In this section, we draw on relevant journalistic and scholarly work to generate testable propositions surrounding the potential status of BlueLM as a countermovement. We begin with the racial composition of BlueLM. If BlueLM is a countermovement to BLM, we would first expect to find that it is a White-backed movement, with Whites comprising the majority of active participants at BlueLM rallies and the bulk of supporters among the general public. Investigative reports using law enforcement personnel data reveal that contemporary police forces are predominantly made up of Whites, with rank-and-file staff in most major cities being considerably Whiter than the civilian population being policed (Ashkenas & Park, 2014; Keating & Uhrmacher, 2020) and command staff being predominately White—even in major cities with more racially diverse rank-and-file staff (Smith, 2019). Journalistic accounts of BlueLM rallies have suggested that the movement is led by White officers, and largely made up of White participants.Footnote 5

Adding to this, research finds that longstanding gaps in attitudes toward the police between Whites and Blacks notably widened following the 2014 Ferguson uprising (Drake, 2014). Indeed, public views toward the BLM movement and police shootings are significantly divided along racial lines, with White Americans most supportive of the police, least supportive of BLM, and most likely to view police shootings of Black civilians as justified (Doherty et al. 2014; Jefferson et al., 2021; Thomas & Horowitz, 2020). While existing scholarship has not evaluated how race is linked to support for BlueLM, a poll by Rasmussen found greater support among White respondents for BlueLM legislation relative to Black Americans.Footnote 6 Together, this work provides a strong foundation for the expectation that a putative pro-police movement, as well as any movement standing in opposition to BLM, would be predominately comprised of, and supported by, White Americans. In sum, our first proposition is that BlueLM is a White-backed movement. Empirical support for this proposition would constitute the first prong of evidence for BlueLM as an instance of White countermobilization and thus its status as a countermovement to BLM.

Our second proposition relates to the motivations for participation in, and support for, BlueLM. Historically, each major movement to improve the position of Black people in American society has been met with White countermobilization and backlash (Alexander, 2020). Importantly, while pre-Civil Rights era White countermovements were explicitly and transparently motivated by White supremacy and racism, the norm of racial equality in the post-Civil Rights era inaugurated a shift to more subtle and race-neutral or “coded” expressions of racism by elites (Gilens, 1996, 1999; Jamieson, 1993; Mendelberg, 2001) and the public (Kinder & Sanders, 1996; Schuman et al., 1997). In the post-Civil Rights era, scholarship firmly documents the ever-morphing coded vehicles through which Americans express racial animus, including support for the “Official English” movement (Citrin et al., 1990), “three strikes” laws (Johnson, 2008), “Southern heritage” (Strother et al., 2017), the birther conspiracy (Pasek et al., 2015), the Tea Party movement (Parker & Barreto, 2014), and opposition to the Colin Kaepernick NFL protests (Stepp & Castle, 2021). In each of these instances, the attitude object in question is argued to be race-neutral yet scholarship has established a systematic link between racial prejudice and opinions.

This same dynamic could apply to BlueLM, as the organization claims that the movement is not anti-Black or opposed to efforts to achieve racial justice but is instead simply pro-police and pro “law and order.” However, an analysis of the role of the police in the history of Black insurgency renders repeated evidence of Whites’ use of law enforcement and concern over “law and order” as a means of suppressing Black claims-making activity and reinforcing the racial hierarchy (Alexander, 2020; Stockley et al., 2020; Weaver, 2014). Given long-standing evidence of the racialization of “law and order” (Stephens-Dougans, 2021) and attitudes toward the police, especially in the era of BLM protests (Reny & Newman, 2021), it is distinctly possible that support for BlueLM may serve as a vehicle for the channeling of racial prejudice among Whites. Indeed, expression of support for BlueLM may involve a defense of the “privilege of violence” (Philips, 2022) of particular appeal to racist Whites who view police killings of Black Americans as an allowable or necessary extra-legal use of violence wielded by “natural enforcers of justice and order” (pg. 471). Such possibilities are supported by qualitative research documenting the prevalence of White ethnonationalist sentiments in BlueLM documents, media interviews, social media posts, and opinions reported by a small group of BlueLM supporters in Ohio (Soloman et al., 2021). Therefore, our second proposition is that support for BlueLM among White Americans is associated with racial prejudice. Empirical support for this would constitute the second prong of evidence for BlueLM as a countermovement to BLM rooted in White racial prejudice.

Our third proposition relates to the temporal dynamics between BLM and BlueLM protest activity. If BlueLM is primarily a White countermovement in response to BLM, we should observe not only that the movement is largely supported by White Americans and motivated by White racial prejudice but also that its political activities should appear in reaction to claims-making activity by BLM. As noted earlier, throughout U.S. history, there appears to be a clear temporal pattern and a great deal of qualitative evidence that key episodes of White mobilization are preceded by outbreaks of Black protest. In the case of BlueLM, many observers draw attention to the fact that the movement was formed several months after the 2014 Ferguson uprising and evolution of BLM into a nationwide social movement engaging in demonstrations throughout the U.S. (Taylor, 2016). Adding to this, scholarly observers suggest that BlueLM is engaging in a process of “competitive victimhood” (CV), which is conceptualized as an intergroup dynamic that emerges in contexts of intractable conflict and/or structural inequality whereby groups compete with each over claims to relative victim status (Young & Sullivan, 2016). The CV literature highlights how high-status or dominant social groups may evoke claims of their own group victimhood in response to claims of victimization made by low-status or subordinate groups. In theory, this process is motivated by the threat to a dominant group’s moral identity created by subordinate group claims of being victimized by the dominant group. Thus, dominant group members engage in a motivated social cognition process emphasizing their own experiences of suffering and deprivation as a means of reducing negative stigma and feelings of responsibility for harmdoing (Noor et al., 2012).

This dynamic is also highlighted in Jardina’s work on White racial identity (2019). She argues that the increasing hostility we observe between Whites and minoritized groups in today’s politics has emerged from a context in which Whites feel a sense of threat from losing their majority status given the changing racial landscape. While Jardina does not frame her theory as competitive victimhood, it is an element to her theory: “Amidst these changes, many whites have described themselves as outnumbered, disadvantaged, and even oppressed” (p. 3, 2019). She argues that the implication of this is that it has led to a greater sense of commonality and group attachment among Whites, as well as to a greater sense of racial solidarity, whereby Whites are motivated to protect the interests of their group. If BlueLM is a White-backed movement, as we contend, then Jardina’s framework would also suggest that we should observe increased support for the movement in response to claims-making activity by BLM.

The CV framework has been applied to the BlueLM movement by qualitative scholarship as a quintessential case (Solomon & Martin, 2019). Using content from BlueLM webpages, this research highlights efforts by BlueLM to shift the national narrative about police brutality and systemic racism that arose in response to the deaths of Michael Brown and Eric Garner and ensuing BLM protests. These efforts involved the introduction of counter-narratives emphasizing the difficult and dangerous job performed by the police, the loss of life of officers in the line of duty, the suffering experienced by their friends and families, and the mis-portrayal of the police as victimizers rather than as victims (Solomon & Martin, 2019). Inherent to this process is that the subordinate group’s victimhood and claim-making activity is causally prior to the respective victimizer’s competing claims of victimhood. Indeed, Philips contends that, when certain groups (e.g., Whites) are afforded the “privilege of violence,” their efforts to defend this privilege are “mobilized precisely in those moments when unequal deployments of force are exposed or challenged” (2022, 471).

Extant work situating BlueLM within the CV framework relies on qualitative evidence and journalistic accounts, leaving it open to question whether BlueLM activity over the past 7 years is systematically tied, or is exogeneous, to Black victimhood and ensuing BLM claims-making activity. Thus, our third proposition is that BLM protest activity should be causally prior to BlueLM activity. Empirical support for this would constitute the third prong of evidence for BlueLM as a countermovement to BLM.

Overview of Data and Methods

To test the above propositions, we utilize a diverse array of data, including internet search interest data, news and social media data, public opinion surveys, and corporate merchandise sales. Our goal was to collect a range of indicators of BlueLM activity and support for BlueLM, including: news media coverage of BlueLM as a proxy for BlueLM rally activity; revealed interest in BlueLM using daily internet search volume from Google Trends; self-reported support for BlueLM from surveys of American adults; discussion of BlueLM by users of the social media platform Twitter; and revealed support for BlueLM via purchases of the TBL flag.

We see multiple benefits in using this eclectic set of data. First, each of our propositions can be tested using more than one dataset. This allays concern over evidence in support of any given proposition being confined to a single test or outcome variable. Second, we have data that can be analyzed at the individual-level and in aggregate form. This enables us to potentially demonstrate consistent results across multiple levels of analysis. Third, our data include attitudinal and behavioral measures of support for BlueLM, which stands to enhance external validity if the findings from surveys are backed by findings using real-world behaviors. Finally, some of our data are cross-sectional (e.g., survey data) while others span several years (e.g., tweets, flag sales, internet searches), which allows us to employ multiple analytic strategies, including model-based (e.g., multivariate regression) and design-based (e.g., regression discontinuity in time) approaches. Since we bring multiple datasets and tests to bear for each of our three propositions, we organize the presentation of results by tests for each proposition.

Proposition #1: BlueLM is Associated with Whiteness

To test our first proposition, we analyze the relationship between White racial self-identification and self-reported support for BlueLM using two national samples of American adults. We complement this analysis with data on a central behavioral indicator of support for BlueLM: geocoded sales of TBL flags from the top vendor of the flag on Amazon.com. We utilize this data to analyze the relationship between the prevalence of non-Latino White residents in a zip code and TBL flag sales, while controlling for key zip-level confounders.

We begin with individual-level survey data to assess the relationship between self-reported race/ethnicity and support for BlueLM. We utilize two national samples of adult Americans benchmarked to national population demographics fielded with Lucid between January and March 2021 (Surveys 2 and 3 listed in Table 1). In these surveys, we measured support for the BlueLM movement using a 6-pt Likert scale. General sample details for all surveys are provided in Table 1 and more detailed information on the survey vendors, as well as descriptive statistics and relevant survey question wording, can be found in Appendix A. We fit OLS models regressing support for BlueLM in both surveys on a dummy variable coded “1” for non-Latino White respondents alone (“Bivariate” model) and then controlling for education, income, age, gender, partisanship, ideology, and racial prejudice (“ + Controls” models). We estimate a bivariate model to mitigate the possibility that an observed relationship between White racial self-identification and support for BlueLM is due to suppression effects (Lenz & Sahn, 2021). Additionally, if Whites are more likely than non-Whites to identify with the political right and harbor anti-Black prejudice (Davis & Wilson, 2021), then inclusion of partisanship, ideology and prejudice as controls may bias estimated coefficients for White racial identification toward zero. As such, our bivariate models render a view of the relationship between Whiteness and BlueLM support that rules out suppression effects and post-treatment bias. Setting aside the question of causal ordering, given known correlations between Whiteness, partisanship, ideology, and racial prejudice, we estimate multivariate models to rule out the possibility that estimated coefficients for White racial identification are capturing these related factors.

Table 1 Original survey datasets

The results from these analyses are presented in Fig. 1 panel A (Table B1). Starting with bivariate models, we see that Whiteness is associated with a 0.142 point (95% CI: [0.118,0.166]) or 0.46 standard deviations (Lucid S1) and 0.125 point (95%CI [0.084,0.165]) or 0.42 standard deviations (Lucid S2) increase in support for BlueLM. Turning to the models with controls, we see that Whiteness remains a statistically significant predictor, associated with a 0.050 point (95% CI [0.027, 0.074]) or 0.16 standard deviations and 0.073 point (95%CI [0.030, 0.117]) or 0.25 standard deviation increase in support for BlueLM. The results of our multivariate models indicate that being White is predictive of support for BlueLM above and beyond factors associated with Whiteness, such as partisan identity, ideology and racial prejudice. This said, the reduction in the size of the coefficients for Whiteness in the multivariate models is unsurprising given the inclusion of these associated variables, which are arguably causally downstream from race (Montgomery et al., 2018).

Fig. 1
figure 1

Whiteness and Support for Blue Lives Matter. Note: Panel A displays predicted change in support for BlueLM for White versus non-White respondents in two samples. Estimates calculated from bivariate (top) and fully controlled models (bottom) with all controls held at their means. Panel B displays predicted count of zip-level Thin Blue Line flag sales conditional on zip code percent White with all controls held at their means. 95% confidence intervals. Full results in Appendix B

Having established evidence across two samples of American adults of a systematic link between race and attitudinal support for BlueLM, we now investigate the relationship between race and a powerful behavioral indicator of support for BlueLM—purchases of the TBL flag. The top vendor of TBL flags listed on Amazon.com is ANLEY Flags Inc (www.anley.com). The CEO of ANLEY provided us with customer-anonymized sales data, including the zip code of the recipient and date of sale, for all standard-sized TBL flag sales completed directly through ANLEY’s purchasing and distribution center between August 4, 2018 and August 4, 2020. Using zip code of sale, we merged zip-level demographic variables from the 2015–2019 five-year American Community Survey (ACS) with the ANLEY sales data. Using these data, we compiled a count of sales by zip code (mean = 0.27, sd = 0.71). We estimate whether or not the percent of a zip code that is non-Latino White is predictive of zip-level aggregated sales of TBL flags, controlling for other zip-level socio-economic and demographic controls (e.g., median income, educational attainment, percent Black, employment in protective services, total population, percent Republican presidential vote in 2016,Footnote 7 and population density). In Fig. 1 panel B (Table B2) we plot the predicted count of TBL flag sales across the percentage of the White population at the zip code level. The estimates, simulated from a zero-inflated negative binomial model, correspond to a move from about 1.8 flag sales per 100,000 people in non-White communities to 6.2 flag sales per 100,000 people in completely White communities. Perhaps unsurprisingly, we also observe a significant increase in the count of TBL flag sales associated with an increase in the presence of zip code residents employed in protective service professions (e.g., law enforcement). This finding could serve as a validity check on the sales data, as we would expect a relationship between the presence of persons potentially working in law enforcement and TBL flag sales. What is notable is that, even after accounting for the potential presence of law enforcement workers—as well as other potential confounders—we observe a significant relationship between the presence of White people and TBL flag sales.

The findings in this section suggest that Whiteness is a strong predictor of support for BlueLM. Employing different data and units of analysis, as well as attitudinal and behavioral indicators, we find reoccurring evidence of an association between White racial status and embracing BlueLM. Each piece of presented evidence has strengths and weaknesses: the survey data is representative of the White population but only has a relatively low-cost form of BlueLM support (i.e., a survey response); the TBL flag sales data capture behavioral support for BlueLM but are anonymized and can only be analyzed in the aggregate (i.e., zip code level). Taken together, the findings in this section establish the first prong of evidence positioning BlueLM as a countermovement to BLM, as the evidence suggests it is a White-backed movement.

Proposition #2: Support for BlueLM is Associated with White Racial Prejudice

To test our second proposition, we analyze data from three original national surveys: the two Lucid surveys analyzed in the previous section and a YouGov survey of a representative sample of non-Latino White adult Americans (details summarized in Table 1). To complement these data, we collected the universe of tweets that included the hashtag “#BlueLivesMatter” from Crimson Hexagon, yielding a dataset of N = 2,744,762 tweets posted between November 24, 2014 and October 19, 2020. Crimson Hexagon estimates the location of tweets using geo-tag metadata, profile location, and time zones. Tweets were cross-walked and aggregated to the county-level and merged with mean county-level White racial resentment scores estimated from the Cooperative Congressional Election Survey (CCES; 2010–2014) by Acharya et al. (2020), and county-level demographics from the 2006–2010 five-year ACS (including population, population density, median income, percent non-Latino White, college educated, 65-years old or older, and employed in protective services). We use the 2006–10 ACS to ensure that our controls are temporally prior, or “pretreatment,” to our prejudice measures. Our main dependent variable is a logged count of county-level BlueLM tweets per 1,000 residents (mean = 1.86, sd = 2.01).

In Fig. 2 panel A (Table B3) we present the simulated change in predicted support among White adults for BlueLM moving prejudice from its in-sample minimum to maximum values. The models underlying panel A include controls for education, income, age, gender, partisanship, and ideology. Given the correlation between racial prejudice, partisanship and ideological self-identification (Conover & Feldman, 1981; Tesler, 2016), controlling for partisanship and ideology lessen the concern that our estimates for prejudice are capturing the effects of partisan and ideological orientations. The results in panel A offer consistent evidence, with two different operationalizations of prejudice and across three different surveys, that prejudice is the single strongest predictor of support for BlueLM among Whites, even exceeding in all models the strength of the association between partisanship and support for BlueLM. All else equal, White adults highest in racial prejudice are between 0.48 and 0.52 standard deviations more supportive of BlueLM than those lowest in prejudice. Looking at the analysis with our Twitter data yields strikingly similar findings. We show in Fig. 2 panel B (Table B4) that residents in counties that are highest in racial resentment tweet 3.6 standard deviations more about BlueLM than those residing in counties lowest in racial prejudice, all else equal. Importantly, we find that these results are robust to analyzing only tweets with positive or joyful sentiment scores, using alternative measures of anti-minority orientation such as anti-immigrant sentiment, and several other modeling specifications choices (see Appendix B).

Fig. 2
figure 2

White Racial Prejudice and Support for Blue Lives Matter. Note: Panel A displays the predicted change in White support for BlueLM moving prejudice (racial resentment and old-fashioned prejudice) from its minimum to maximum in-sample values across three samples with all controls held at their means. Panel B displays the predicted count of county-level per capita tweets using the hashtag “#bluelivesmatter” moving county-level racial resentment from its minimum to maximum values and holding all other controls at their means. Full results in Appendix B

Overall, the findings in this section offer robust evidence that prejudice is strongly predictive of support for BlueLM. Indeed, using both individual-level survey data with self-reported support and aggregate county-level social media data with expressed support, we find reoccurring evidence of a systematic relationship between racial prejudice and BlueLM support. While the analytic approaches employed in this section are not causal in design, uncovering a systematic association between prejudice and BlueLM support may be sufficient in light of the overarching claim (e.g., Proposition #2) we seek to test, as our findings show that Americans who harbor the most racial prejudice also tend to be those who most strongly support BlueLM. This association reappears using different data, outcomes, and units of analysis. In sum, our findings in this section establish the second prong of evidence positioning BlueLM as a countermovement to BLM.

Proposition #3: BlueLM Activity is Caused by BLM Protest

In this final section, we test our third proposition using regression discontinuity in time (RDiT) analyses (Hausman & Rapson, 2018). This approach leverages the unanticipated eruption of BLM protest in response to the random timing of the police killing of George Floyd in May 2020 to estimate the causal shift in attention to and support for BlueLM. The 2020 George Floyd protests have been shown to generate considerable shifts in public awareness of anti-Black discrimination and depress public favorability toward the police (Reny & Newman, 2021). Moreover, as one of the largest episodes of social protest in response to Black victimhood in American history, the 2020 Floyd protests represent a most-likely case (Gerring & Cojocaru, 2016) for observing a reactionary response to BLM protest activity. In addition to the datasets of TBL flag sales and daily counts of #bluelivesmatter tweets described above, we also collect Google Trends data for daily searches of “Blue Lives Matter” and national media coverage of “Blue Lives Matter” in the nation’s top 50 media sources via Media Cloud. For each of these analyses, the running variable we use is time, the number of days before and after the eruption of massive BLM protests following these two major moments of Black victimization by the police. Following Reny and Newman (2021), we set the cutpoint to May 28th for the 2020 analysis, the day after the nationwide outbreak of protest following the killing of George Floyd. We model the running variable using a parametric polynomial regression with Imbens-Kalyanaraman optimal bandwidth.

In Fig. 3 (Table B5) we document the drastic increases in Google search behavior, media coverage, tweet activity, and TBL flag sales following the Floyd protests in May 2020. We find statistically significant spikes in flag sales (β = 21.8, p < 0.001), Tweets (β = 3811.8, p < 0.001), media coverage (β = 7.4, p < 0.001) and Google trends searches (β = 38.8, p < 0.001). These spikes in behavior are substantively large and meaningful, corresponding to a 2 standard deviation increase in flag sales, 2.3 standard deviation increase in Tweets, 1.2 standard deviation increase in media stories, and 4.9 standard deviation increase in Google searches. In order to show that BlueLM activity is driven primarily by collective claims making (social protest) in the wake of police killing of an unarmed Black civilian and not Black victimization alone, we estimate the same RDiT models for ten other cases where unarmed Black civilians were killed by the police between July 2019 and March 2020 but that did not generate large-scale BLM protests. Appendix Table B6 documents the consistent absence of statistically significant positive RDiT effects for these other killings across our four outcome measures.

Fig. 3
figure 3

Regression Discontinuity Analysis of the Effect of the 2020 Black Lives Matter Protest on Blue Lives Matter Activity. Note: Change in Google searches for “Blue Lives Matter”, media mentions of “Blue Lives Matter,” tweets using the “#bluelivesmatter” hashtag and “thin blue line” (TBL) flag sales (2019–2020 only), following the outbreak of George Floyd protests (May 28, 2020) in Panel. Full results in Appendix B

Last, as a point of comparison, we also estimate the effect of six of the highest-profile recent ambush-style civilian killings of police officers on BlueLM activity. If BlueLM activity were driven by the motive to “Remember the Fallen” and “help Law Enforcement Officers and their families during their time of need”,Footnote 8 we would expect these deadly attacks on officers to generate BlueLM activity. In Table B7, we find little consistent evidence of substantively meaningful and statistically significant increases in our outcomes. In key instances where an effect is statistically significant, the sign of the coefficient is often negative—indicating the ambush killing caused a decrease in BlueLM activity. We only find 3 out of 20 tests where an ambush killing triggered statistically significant and substantively meaningful increases in BlueLM activity. This finding is important because it suggests that, rather than representing a response to incidents of police victimhood, BlueLM activity instead represents reactionary activity primarily in response to organized claims-making surrounding incidents of Black victimhood.Footnote 9

In sum, across several different behavioral measures and time periods, we see robust evidence that BLM protest following instances of Black victimization by the police triggered spikes in BlueLM activity, suggesting that BlueLM is reactionary to BLM and recurrently ramps up claims of police victimhood in competitive response to claims of Black victimhood by the police. As such, the results in this section establish the third prong of evidence positioning BlueLM as a countermovement to BLM.

Conclusion

The leaders of the BlueLM movement portray the movement as being apolitical, race-neutral, and concerned with supporting the needs of police officers who face risks on the job (Offenhartz, 2020). Opponents counter that BlueLM is in fact motivated by racial prejudice and opposition to efforts by the BLM movement to call out systemic and unaccountable police violence against Black Americans (Cooper, 2020; O’Leary, 2020; Smith, 2020; Thusi, 2020). Our goal in this article was to assess the sources of public support for BlueLM, and the extent to which the movement can be considered a countermovement to BLM. Using an extensive array of data and analytic methods, we offer evidence that BlueLM is a White-backed movement motivated by racial prejudice that is reactionary to BLM protest. In short, we offer strong empirical evidence suggesting that BlueLM is a racial justice countermovement.

Our work serves as an important complement to the growing literature examining the BLM movement. Existing work has traced the origins of BLM and its deep historical roots (Taylor, 2016; Thurston, 2018) and explored the diversity and success of different movement frames (Jackson, 2016; Tillery, 2019; Bonilla & Tillery, 2020). Scholars have also examined the consequences of BLM protest media coverage for legislative reform efforts (Arora et al., 2019) and the impact of BLM protest on public opinion (Reny & Newman, 2021). Here, we examine a common consequence in American political history of the success of Black protest movements—the emergence of White countermovements. The findings in this article echo a long historical pattern in which Whites who harbor deep-seated antipathy toward Black Americans develop countermovements when Black communities organize to improve their social, economic, and political position. While for much of American history such White countermobilization efforts have been explicitly racist (Alexander, 2020; Chalmers, 1987; Gordon, 2017; Stockley et al., 2020), a shift to race-neutral justifications became the norm following the Civil Rights Movement. We see this with organizations such as the John Birch society and the Tea Party, which highlighted themes of anti-communism and a more limited federal government (Parker, 2018). We also observe this quite clearly in the example of BlueLM, whereby leaders of the movement disavow racism and highlight the danger and threats faced by police on the job. It is important to study these distinct reactionary movements because, while they all may be associated with Whiteness, racial prejudice and threats to the White racial hierarchy, they may employ different types of non-racial justifications to justify their movements.

While countermovements that make implicit or coded racial appeals are still the norm, there has been increasing use of more explicitly racialized language, which pre-dated Donald Trump’s presidency but gained further momentum with his candidacy and during his administration (Newman et al., 2021; Reny et al., 2019; Valentino et al., 2018). A number of far right extremist groups, that typically are outside of the mainstream, were proactive in encouraging their supporters to vote for Trump during the 2016 primary and general election (Long, 2022). Group such as the Oath Keepers, Proud Boys, and 3 Percenters, were active in the “Stop the Steal” protests and the riot at the U.S. Capitol on January 6, 2021.Footnote 10 According to the Southern Poverty Law Center, which tracks extremist activity, while the number has declined since their peak of 155 in 2019, White nationalist groups have increasingly concentrated their efforts on more mainstream spaces since January 6th.Footnote 11

How does BlueLM compare to some of these other more explicitly racially-motivated, right-wing groups? Movements that are motivated by racial prejudice but appear to rely on coded appeals, such as BlueLM, are potentially more pernicious; these movements allow those who hold prejudice to hide behind a non-racial cover and build what appears on the surface to be credible claims of victimhood. Such movements may be more successful than those that are more explicitly racialized since they may also attract support from some who do not hold negative racial attitudes but are persuaded by some of the non-racial justifications. This therefore builds higher levels of public acceptability and credibility to those movements, and in turn may increase their strength and ability to counter reform efforts.

Understanding BlueLM as a countermovement is also particularly important in the current political context. The murder of George Floyd in 2020 at the hands of police sparked the largest and most sustained protests for criminal justice reform in recent history. In the spring and summer of 2020, public support for BLM increased dramatically,Footnote 12 as did negative attitudes toward the police and perceptions of discrimination against Black Americans (Reny & Newman, 2021). More than any other time since the Civil Rights Movement, there is greater motivation and incentive to enact meaningful reform, and there are some signs of progress. The Department of Justice has launched investigations into two major city police forces, with a particular focus on use-of-force policies. Even though it is a rare event for police officers to be convicted when they use deadly force, Derek Chauvin was convicted of murdering George Floyd. Some states, such as California, have passed legislation to address the use of force by police, and the chief of the Los Angeles Police Department recently ordered the “thin blue line” flag removed from police station lobbies given its use by far-right extremist groups, such as the Proud Boys.Footnote 13 The U.S. House passed the George Floyd Justice in Policing Act in March of 2021, which would tie federal funding to police departments adopting use-of-force standards set out in the legislation, which require that police officers first try to de-escalate situations and only use deadly force as a last resort.

Though there are signs of progress, police unions have historically fought against such reform efforts.Footnote 14 Furthermore, even if meaningful legislation is enacted at the federal and state level, to be successful, reform efforts will require buy-in on the ground, in police departments around the country, and past experience suggests that police unions often make implementation of reform efforts difficult.Footnote 15 Our results speak to and highlight the power of the police as a political group in America. We show that support for BlueLM is fairly high among White Americans, especially those high in racial resentment, and activity in support of the movement increases in direct response to greater activity by BLM. This pattern of increased support for the (counter) movement suggests that BlueLM may have a greater ability to resist various reform efforts being proposed, and even if they are unable to stop such efforts, they may have the political capital to make it difficult to implement reforms.

Our work also opens up new areas of exploration for future research. First, it would be useful to dig deeper into the psychology underlying support for racial justice countermovements. For example, do individuals justify their support for BlueLM using the language of competitive victimhood, and deflect accusations that their support is rooted in prejudice? Future work might also consider additional ways of capturing BlueLM activity in reaction to instances of BLM protest. And, while we have explored factors associated with support for BlueLM activity, it is an open question of whether such activity impacts public opinion. Does increased BlueLM activity lead to greater support for the police and more opposition to policing reform efforts?