In Pursuit of Racial Equality: Identifying the Determinants of Support for the Black Lives Matter Movement with a Systematic Review and Multiple Meta-Analyses

The Black Lives Matter (BLM) movement arose to put a much-needed spotlight on police brutality and systemic racism. In two comprehensive studies, we sought to investigate the determinants of support for the BLM movement. First, in a systematic review 1,588 records were identified and findings from twenty-four studies (Npooled =27,691) were narratively synthesized along five categories relating to demographics, race, partisanship and ideology, discrimination and prejudice, and psychology. Second, we exhaustively examined the determinants of BLM support across thirteen probability-based nationally representative datasets (Npooled =31,779), finding thirty-seven common predictors for which individual meta-analyses were conducted to estimate the strength and robustness of their associations. Our results suggest a near perfect match between BLM opposition and positive attitudes towards political actors and institutions rooted in systemic racism in the United States. The present work contributes to a broad categorization of correlates of support for BLM across social, psychological, and political domains.

T he continued deaths of African Americans in the hands of law enforcement have reignited public debate on how racial minorities are unfairly and unjustly treated in a country built on the premise of equal opportunities. The killings of Willie Ray Banks in 2011, Trayvon Martin in 2012, Eric Garner and Michael Brown in 2014, and Breonna Taylor and George Floyd in 2020 to name a few, are symptomatic of the ingrained racism imbued in the institutions that comprise American society A list of permanent links to Supplemental Materials provided by the authors precedes the References section. Flavio Azevedo is a Fulbright fellow and Associate Researcher at Social Decision-Making Lab, Cambridge University (fa441@cam ac.uk, UK). His research focuses on ideology, how to measure it, its role in political behavior and justifying social and economic injustices. He is the principal investigator of the Psychology of Political Behavior Studies (PBBS), a series of studies informing psychological explanations of political attitudes, values, voting, and participation. He co-founded and directs FORRT-Framework for Open and Reproducible Research Training-an award-winning interdisciplinary community of more than three hundred early-career researchers integrating open scholarship principles into higher education and advancing research transparency, reproducibility, and ethics through pedagogical reform and meta-scientific research.
Tamara Marques is a Research Assistant currently doing a Masters in Sociology at University of Minho,Portugal (pg40934@alunos.uninho.pt;Portugal). She is interested in racial inequality and how social movements such as Black Lives Matter impact political behavior. (López 2010;Petersen-Smith 2015). It is under this context that Black Lives Matter (BLM) emerged as a social movement, to put a much needed spotlight on the systemic racism against Black people in the United States (Campbell 2021), and to bring the legacy of police brutality against people of color front and center of civic and political discourse.
The dawn of the movement is attributed to the backlash following George Zimmerman's 2013 acquittal for the killing of Trayvon Martin, an unarmed seventeen-yearold, and the ensuing widespread support for the online hashtag activism of #BlackLivesMatter (Clayton 2018). The BLM movement became a loose confederation of grassroots groups advocating for racial justice at both local and national levels. The movement has since become one of the largest in American history, rivaling that of the 1960s Civil Rights Movement (Buchanan, Bui, and Patel 2020). BLM demonstrations were organized in every major U.S. city (Sawyer and Gampa 2018), gathering more than twenty-five million people in the aftermath of George Floyd's brutal killing (Hamel et al. 2020). BLM protests sparked not only across the country (Faust et al. 2019;Sinanan 2020;Swarns 2016) but also internationally (Beydoun and Ocen 2015;Khan 2015;Saric 2021). BLM's plea for racial equality reached far beyond its borders as the international community witnessed the extent of racial injustices unfolding in the United States, and began to recognize similar structural symptoms at home (Bricker 2020;King 2020;Strong 2017). Indeed, the BLM movement has been expanding its scope to embrace the fight for the rights of other marginalized racial groups, such as refugees (De Genova 2018) and indigenous people (Scott 2021). Now best described as a transnational social movement (Porta, Kriesi, and Rucht 2009;Tarrow 2005), BLM demonstrations spanned across the world-from Pretoria to Reykjavík, Bangkok to Buenos Aires, Auckland to Vancouver-and became tantamount to a global rallying cry against racism and police brutality (Kirby 2020). In 2021, the BLM movement was nominated for the Nobel peace prize, reinforcing BLM's societal relevance in both confronting racially motivated violence and leading the fight towards racial equality (Belam 2021).
Yet despite its well-recognized importance, the BLM movement has been met with mixed public support in the United States (Reinka and Leach 2017). While George Floyd's death embodied a moment of racial awakening in America-increasing individuals' perceptions of racial discrimination against Blacks, and decreasing favorable attitudes towards the police (Curtis 2021)-a Pew Research Center report from June 2020 indicated that 67% of American respondents expressed support towards the BLM movement, 30% of the population still opposed it (Parker, Horowitz, and Anderson 2020). Support for the movement later dropped to 55%, as indicated by a September follow-up Pew poll (Thomas and Horowitz 2020). This decline occurred for all ethnicities except for Black people, and across partisan lines where Republican support more than halved. Analyzing trends in time, Chudy and Jefferson (2021) find a similar pattern and show that support for BLM has declined after George Floyd's killing, mostly driven by conservative Republicans and white Americans. Race has rarely mattered more in American politics than it does now. It is thus not surprising that conservative Republicans dismiss the legacy of slavery as affecting black people's position in the American society today, say the country has already given Black people equal rights, and believe a big problem for the country is that people see discrimination where none exist (Horowitz, Brown, and Cox 2019). Another factor swaying public opinion on BLM is one's attitudes toward the fourty-fifth U.S. president and his uniquely racially charged campaign and time in office-and by contrast, towards Barack Obama and his legacy, which Trump sought to undo. Trump's racialized campaign and presidency have paved the way for the rise of white nationalism (Jardina and Piston 2021), encouraged whites to embrace their whiteness as a social identity (Jardina 2019), reinforced white Americans' xenophobic sentiments (Hooghe and Dassonneville 2018), strengthened prejudice toward Latinos and Asian Americans (Louie and Viladrich 2021), and Muslims (Lajevardi and Oskooii 2018). Racism, xenophobia, and prejudice are the tripartite pillars of white nationalism as evidenced by spikes in hate crimes following Trump's 2016 campaign rallies (Feinberg, Branton, and Martinez-Ebers 2019). It is particularly telling that Trump's history of explicit racism appears not to be a bug that conservative Republicans had to overlook but a feature which would ultimately deliver the nomination, the presidency, and control over American conservatism and the GOP (Abramowitz 2018; Sides, Tesler, and Vavreck 2019).
This divide in public support for BLM across racial and political (partisan and ideological) lines buttresses the need to better understand which factors underlie individuals' proclivities towards social movements in search of racial equality. The existing literature, however, focuses on vastly different elements associated with BLM support, providing important but ultimately disconnected insights. As support for BLM in America is likely a multifaceted issue influenced by a plethora of factors, and single studies are "limited in the generalizability of the knowledge they produce about concepts, populations, settings, and times" (Cook et al. 1992, chap. 3), research synthesis can provide a stable foundation for an authoritative account on the subject matter (Dacombe 2018). This is especially the case since the BLM literature appears to be proliferating exponentially (Campbell 2021), and-to date-there has not yet been an attempt to integrate existing findings.
In the present research program, we sought to uncover the demographic, political, and psychological bases of 1306 Perspectives on Politics BLM support by conducting a systematic literature search and multiple meta-analyses. The goal was to build an analytical infrastructure paving the way for future research. We a) uncovered four main themes in the extant literature; b) contextualized the literature on collective action and political behavior into that of the BLM movement; c) consolidated published research and empirical evidence (surveys and polls) on the determinants of BLM support; d) pit the most reliable predictors of BLM against one another, compared the magnitudes and their relative importance; and e) explored disparities in BLM support across societal groups and subgroups. This research program contributes in numerous ways to theory development and testing because it combines evidence into a broader, more generalizable, framework; identifies sources of variability among the different components of evidence; generates new hypotheses; and uncovers understudied areas and potential gaps.

State of the Art
The rise of BLM has been accompanied by increased scholarly interest on the impacts of racial violence in the United States. Despite its recent emergence, it is possible to already identify some of the major themes around which discussions have centered. Initial academic discussions have concentrated mostly around the topic of police violence, exploring how policing poses harm to marginalized communities in the United States (Gaber and Wright 2016), documenting the lived experiences of Black Americans with the law enforcement (Brooks et al. 2016), discussing the impact of the episodes of police brutality to different academic fields (Pratt-Harris et al. 2016), and the need of interventions aiming to increase the psychological and physical health of Black students, employees, and organizations McCluney et al. 2017;Opie and Roberts 2017).
As the BLM matured, a second stream of literature emerged focusing on the characteristics of the movement and how they might have shaped its public image. Studies have emphasized how BLM stands out in comparison to earlier Black social movements for its intersectional framing, particularly among Black communities (Ray 2020). Differently from the Civil Rights movement, which arguably centered more around a Black heterosexual male leadership (Matthews and Noor 2017, p.8), BLM activists deliberately sought to be inclusive of all Black lives across divisions of gender, socioeconomic status, and sexuality, providing a much-needed forum for diverse voices and marginalized groups (Clark, Dantzler, and Nickels 2018). BLM was also considered unique in standing in solidarity and building coalitions with other oppressed groups such as Palestinians and Indigenous people (Clark, Dantzler, and Nickels 2018), refugees (De Genova 2018), and aborigines (Scott 2021). Another identified feature distinguishing BLM is its active use of social media to a) inform of news seldom covered by traditional outlets (e.g., racist incidents; Graham and Smith 2016); b) generate mainstream media coverage (Freelon, McIlwain, and Clark. 2018); c) galvanize public support and political mobilization (Casas and Williams 2019); and perhaps more significantly, to d) foster discussions from the point of view of Black people (Graham and Smith 2016); and e) enable organizers and supporters to continuously shape their message and voice criticism against insidious attacks from-and pervasive (white) framing of-mainstream media (Nummi, Jennings, and Feagin 2019).
As BLM gained space in the media, it became imperative to investigate the role of news media in determining individuals' attitudes towards the movement. Portrayals of BLM were found to be often racialized, with numerous news outlets expressing explicit disapproval of protests, failing to contextualize the movement within the historical struggles of Black people, attributing blame for acts of violence to Black protesters, and framing peaceful protests derogatorily as riots (Fabregat and Beck 2019;Leopold and Bell 2017;Lane et al. 2020). Research also revealed how media contextual and visual frames can affect individuals' perceptions of protests. For example, legitimizing frames portraying the movement's goals and complaints increased support for protesters (Kilgo and Mourão 2021), whereas framing BLM protests as being composed solely of Black participants (versus more racially diversified depictions) elicited lower support (Wouters 2019) and increased perceptions of the protest as being violent (Peay and Camarillo 2021). Other elements such as the race-ethnicity of individuals expressing support for BLM movement on open letters (Arora and Stout 2019), messaging strategies framing the movement as intersectional (Bonilla and Tillery 2020), and consumption of conservative media outlets (Kilgo and Mourão 2019) were also found to impact attitudes towards the BLM movement.
More recently, scholars have started to examine political and psychological characteristics influencing support for-and participation in-BLM protests. Published studies have found a range of factors contributing to participation in BLM protests including intergroup contact (Meleady and Vermue 2019), empathy-mediated contact , and perceived societal inequality (Lake, Alston, and Kahn 2018). Support for BLM has been linked to identifying as Democrats (Arora and Stout 2019), endorsement of progressive-liberal politics (Merseth 2018), recognizing society lacks equal opportunities , as well as to psychological tendencies such as lower levels of both authoritarianism (Barker, Nalder, and Newham 2021), and social dominance orientation (Holt and Sweitzer 2018). Additionally, Sawyer and Gampa (2018) found that during BLM protests, liberals tend to display less pro-white implicit and explicit bias than conservatives.
We dive deeper into the published literature on the factors influencing support for BLM as part of presenting results of Study 1. Next, to expand the determinants of BLM support beyond the published literature, in Study 2 we conducted systematic multiple meta-analyses of predictors of BLM support using thirteen nationally representative public opinion surveys. In an effort to contextualize reported findings theoretically, results from both studies are discussed integrating insights from the collective action and political behavior literature.

Methods
We synthesized evidence in a transparent and accessible manner, in line with the outlined principles for good evidence synthesis for policy (Donnelly et al. 2018). To ensure a transparent and complete reporting of the literature search and screening, we adopted the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA, Page et al. 2020). A flow diagram describing each step in the identification, screening, eligibility, and inclusion criteria is presented in figure 1. For further details on search methodology and data extraction, see sections 1.1 and 1.2 of the online appendix. We abided to open and reproducible research practices: replication materials can be found at https://osf.io/pvk69.

Results
The systematic review yielded twenty-four studies from the original twenty-one records. The findings of the systematic review are shown in figure 2. The reviewed records comprised a total of 27,691 participants. Sample size across quantitative studies ranged from 202 to 3,769 participants per sample. Qualitative studies included in the systematic review had sample sizes of 41 (S4) and 115 participants (S9). Except for one study, which was conducted with British participants (N=202; S12), all other studies included U.S. participants (N=27,489). Most articles were published in 2020 (k=12) and 2018 (k=5) while only three articles were published in 2019 and one article was published in 2021. A total of twentyone studies were conducted online while four studies employed in person interviews. Studies used diverse methodologies, with some studies using experimental research (S2, S3) while most employed observational surveys (k= 20). Sixteen studies investigated specific racial/ethnic groups, while the remainder investigated the general population. The most prevalent race investigated was whites (k=9) followed by African Americans (k=6), whereas the least frequent ones were Asian Americans (k =3), Afro-Latinx (k=1), and Latino Americans (k=1).
We employ a narrative approach (Baumeister and Leary 1997) to systematize the reviewed literature. We clustered findings into five categories: demographics; race and ethnicity; partisanship and ideology; discrimination and prejudice; social attitudes and psychology. We followed a vote counting procedure to integrate patterns of studies' results, extracted the direction of the associations ensuing from all statistical models, and assessed the degree of consistency (cf. Cooper, 2017, 225). Results are reported per racial/ ethnic group whenever studies investigated different groups. Additional (infrequent) predictors are reported in the online appendix (section 1.3).

Demographics
Age. Ten studies out of the twenty-four included in the systematic review investigated the association between age and support for BLM. The majority of studies (k=7) report non-significant associations between age and support for BLM (S1.1, S1.2, S8, S10, S13, S14, S21). Nevertheless, in line with previous studies showing that older individuals hold more positive attitudes towards the police (Weitzer and Tuch 2002) and are less likely to support the Civil Rights Movement (Jones 2006) three studies find that older individuals are less prone to support the BLM movement (S3, S5, S19).
Education. Eight studies examine the relationship between educational attainment and support for BLM, all of which consistently found non-significant associations between education and BLM support (S1.1, S1.2, S3, S5, S8, S13, S19, S21). This is surprising in light of previous literature considering education as an important factor determining political participation and activism (Hall, Rodeghier, and Useem 1986;Dalton, Van Sickle, and Weldon 2010).
Gender. Results for gender are inconsistent across eleven studies. Four studies find non-significant associations between gender and BLM support (S8, S11, S13, S21), whereas seven studies report that-compared to menwomen are more prone to support the movement (S1.1, S1.2, S3, S5, S10, S14, S19). These inconsistencies in BLM support reinforce the need to further consider how social movements might be gendered. Feminist theory has been crucial in bringing to light the contributions of women to many social movements (e.g., Civil Rights; West and Blumberg 1991). Black women compose the majority of registered voters among Blacks (Noe-Bustamante and Budiman 2020) and were fundamental in turning the course of the 2020 election (Herndon 2020). As such, the role of women -and especially Black women-in supporting BLM should be a subject of further research.
Income. Out of the twenty-four studies, eight investigate income, all of which report a non-significant association between income and BLM support (S1.1, S1.2, S3, S5, S8, S13, S14, S19). This consistent result across studies suggests that income does not appear to predict BLM attitudes.

Perspectives on Politics
Urbanicity. Only three studies investigate urbanicity. Results are mixed. One study finds the relationship between urbanicity and BLM support to be nonsignificant (S19) while the other two report that living in urban-compared to non-urban areas-is associated with higher support for the movement (S5, S8). This is in line with studies showing that rural consciousness is related to racial resentment and stereotyping (Nelsen and Petsko 2021) and that the racial attitudinal shift following George Floyd's death was much smaller in rural than urban areas (Curtis 2021). Of interest, the two studies reporting a significant and positive association also suggest there are differences in this result depending on ethnicity, such that the relationship is mostly significant and positive for whites but non-significant for Latinos (S5) and African Americans (S8). These results suggest that lower levels of support for BLM in non-metropolitan areas may ensue from the higher proportion of whites in rural areas rather than urbanicity itself. Predictors of support for BLM

Race and Ethnicity
As can be expected for a movement addressing issues of racial inequality, race and ethnic identity play an important role in shaping attitudes towards BLM. Results are largely consistent across studies. The three studies comparing participants who self-identify as African Americans versus white find positive and significant associations (S8, S19, S21). These findings support previous evidence reporting that self-identifying as white was a strong predictor of negative attitudes towards the Civil Rights Movement (Jones 2006). In addition, S5 finds that Latinos who self-identify as Blacks are more prone to support the movement than Latinos who self-identify as whites, suggesting that other minorities may support BLM to the extent that they identify with Black people or their struggles. Indeed, perceptions of commonalities with African Americans appear to shape political alliances between minority groups (Sanchez 2008).
Compared to whites, however, racial minority groups are largely and consistently more supportive of BLM across three studies (S10, S11, S16). This might be due to the spike in hate crimes against minorities in the United States in the last years, which may have contributed to a shared identity among people of color (Pérez 2021).

Partisanship and Ideology
For both partisanship and ideology, findings are quite consistent among studies, both showing large effect-sizes predicting BLM support.
Identification with the Democratic party (versus other parties or Independents) is significantly and positively associated with BLM support in four studies (S1.1, S1.2, S13, S21). Similarly, identification with the Republican party (versus other parties or Independents) is significantly and negatively associated with BLM support across four studies (S1.1, S1.2, S5, S19). These findings echo long established patterns (Olsen 1968) whose enduring nature are likely due to the GOP's increasingly racially-tinted politics (Jardina 2019)-arguably as an effort to satisfy Republicans' growing white constituency (Kinder and Kalmoe 2017). Democrats, on the other hand, are comprised of a more racially and ethnic diversified membership (Mason and Wronski 2018) and have progressively shown a meaningful and genuine decrease in racial resentment (Engelhardt 2019(Engelhardt , 2021. Ideology was found to be a significant and negative predictor of BLM support across all four studies, suggesting that the more conservative individuals are, the lower their support for the social movement whose mission is to eradicate white supremacy (S2, S13, S14, S19). This was the case even after controlling for partisanship (S13, S19) echoing the findings of McCright and Dunlap (2008) who showed ideological asymmetries in the endorsement of collective action such that conservatives-as opposed to liberal/progressives-tend to show lower levels of support for (progressive) social movements. These findings are also largely in agreement with research linking conservatism and anti-Black racial sentiment (Federico andSidanius 2002a, 2002b;Sears et al. 1997;Sears and Henry 2003;Sidanius and Pratto 2001). Taken together, liberalconservative differences appear to be of importance for BLM support insofar as it impacts the legitimation of existing forms of inequality (Azevedo, Jost, and Rothmund 2017;Hall 1996;Jost 2021;Wilson 2021) and the desire to change society in a more egalitarian direction (e.g., Bobbio 1996;Womick et al. 2019).

Discrimination and Prejudice
Prejudice. Four studies examined the effect of participants' racially motivated prejudice on support for BLM. Even if measures of racial attitudes differed across studies, findings were consistent in that the more individuals are prejudiced, the less they support the BLM movement. Using feeling thermometers on African Americans, a negative and significant association between favorability/warmth against African Americans and support for BLM was found in S12. Both S14 and S21 used the racial resentment scale finding the more participants resent Blacks, the less they support the movement. S10, too, found consonant results by adapting the measure of symbolic racism to the context of individuals' perceptions about those living in unprivileged neighborhoods ("the policed"). These findings resonate with literature showing that prejudice reduces support for the Civil Rights Movement (Jones 2006) and for political activities benefiting Blacks (James et al. 2001), Aborigines (Leach, Iyer, and Pedersen 2006), and immigrants (Shepherd et al. 2018).
Perceptions of racial discrimination. Three studies investigated individuals' general perceptions of racial discrimination in America. Findings are consistent, suggesting that the more individuals recognize racial discrimination to be a pervasive issue in America, the more they support the BLM movement (S5, S13, S21). Previous literature has shown that witnessing racial discrimination motivates whites to engage in collective action for racial justice, an effect that is mediated by enhancing awareness of white privilege (Uluğ and Tropp 2021). Accordingly, recognition of one's own privilege is thought to be an important factor motivating members of advantaged groups to engage in collective action supporting disadvantaged groups (S4; Radke et al. 2020).
Personal experience with racial discrimination. Four studies included questions about participants' previous personal experiences with racial discrimination. The relationship between personal experience with racial discrimination and BLM support is not significant for the general population (S19), for whites and Latinos (S5), African Americans (S20), and Asians (S1.2). Although personal experiences with discrimination are associated with increased psychological distress (Broman, Mavaddat, and Hsu 2000), they do not necessarily lead to engagement in collective action. As shown by Foster and Matheson (1998), negative emotions elicited by experiences of discrimination may actually inhibit participation. As such, individual differences in how the discrimination episode is internalized and reappraised might be more important than the actual experience in determining support forand participation in-collective action (Foster 2000). Experiencing discrimination as a racial-ethnic minority did not lead to BLM support, and whites who report to have experienced racial discrimination were less inclined to support the BLM movement (S1.2). This perception of "reverse racism" is likely associated with whites' understanding racism as a zero-sum game, such that perceived decrease in racial biases against Blacks is associated with a perception of increased bias against whites (Norton and Sommers 2011).

Social Attitudes and Psychology
Social Dominance Orientation (SDO). Three studies reported associations between SDO and support for the BLM movement. Social dominance was significantly and negatively related to support for BLM (S2, S8, S12). These results are consonant with the literature showing that individuals holding views that endorse group-based dominance and inequality are less likely to support-and engage in-a range of progressive social movements (Ho and Kteily 2020), such as in favor of refugees (Thomas et al. 2019), environmentalism (Milfont et al. 2018), Arab uprisings (Stewart et al. 2016) and racial equality (Stewart and Tran 2018). These findings also square with extant literature linking anti-Black attitudes and racism with SDO and authoritarianism (Duriez and Soenens 2009;van Hiel and Mervielde 2005;Knowles et al. 2013). Taken together, individuals desiring the ingroup to dominate out-group members appear to display low support for the BLM movement.
Intergroup contact. Two studies investigated whether intergroup contact with Blacks affects one's predisposition to support the BLM movement (S5, S17.1). Contrary to expectations (Pettigrew and Tropp 2006), results are inconsistent. While non-significant associations between intergroup contact and support for the BLM movement were found in S5, S17.1 reported the more individuals experience contact with Blacks, the more they support BLM. We found more consistent results when the quality of the intergroup interaction is taken into account. All three studies investigating positive intergroup contact with Blacks find that the more individuals report having positive interactions with Blacks, the greater is their support for the BLM movement (S12, S17.2, S17.3). Moreover, the relationship between negative intergroup contact with Blacks and support for BLM is consistently found to be non-significant across two studies (S12, S17.2).
Police misconduct. Two studies investigated whether individuals' perceptions of police misconduct were related to support for BLM (S10, S21). Alternatively, one study investigated perceptions of police misconduct in relation to African Americans, in specific (S19). Results are mixed with two studies finding that individuals who believe police misconduct occurs are more prone to support the BLM movement (S10, S19), while one study reported a non-significant association between perceptions of police misconduct and BLM support (S21).
Empathy, anger, and willingness for collective action. Three studies belonging to the same article investigate the relationship between support for BLM, empathy for Black people, anger in the context of racial injustices, and willingness for collective action (S17.1, S17.2, S17.3). All three studies find significant and positive bivariate correlations between empathy, anger, willingness for collective action, and support for BLM. Interestingly, as these studies are applied to the context of intergroup contact, all three studies consistently show that positive intergroup contact with Blacks is associated with support for BLM through a sequential process of increasing empathy for Blacks and subsequently anger towards racial injustice.

Taking Stock
The systematic review revealed informative patterns. Race and ethnicity, as well as partisanship and ideology are reliable predictors of support for the BLM movement. Whites, Republicans, and conservatives consistently opposed BLM. Social psychological variables such as low social dominance orientation and authoritarianism, positive intergroup contact, empathy for protestors, anger, and willingness to participate in collective actions in response to social injustices are all consistently associated with BLM support. Similarly, prejudice towards African Americans and lack of recognition of existing racial discrimination against Black people consistently predict opposition to BLM. Education and income show a consistent pattern of non-significance. The remaining demographics (i.e., age, gender, and urbanicity), as well as personal experiences with discrimination and perceptions of police misconduct, are inconsistent predictors of BLM support across studies. Being female, from urban areas, and perceiving police misconduct are found to be positively associated with BLM support in most studies, even if other studies found these associations non-significant. In contrast, age and personal experiences of discrimination are mostly nonsignificant, even if a few studies find these variables to be negatively associated with BLM support.
While relatively minor, these inconsistencies could arise from cross-sample heterogeneity and differences in the racial and ethnic sample composition of reviewed studies.

Perspectives on Politics
Additionally, the use of convenience, non-representative, and small-N samples, raises further concerns about potential existing biases on the published record.
To address these concerns, in Study 2, we aim to confirm the earlier-described patterns and explore a wider range of predictors of BLM support. To this end, we conduct multiple meta-analyses of thirteen public opinion datasets (N=31,779) which offer several advantages towards the consolidation of a profile of BLM supporters. First, the use of large probability-based nationally representative samples enables precise and generalizable estimates of BLM support. Second, results obtained from the multiple meta-analyses are not dependent on the (idiosyncratic) analyses reported in the current literature, nor are they subject to publication biases. Third, the comparison of the same BLM predictors across datasets allows us to investigate the robustness of prospective correlates of BLM support across different samples while assessing its heterogeneity. Finally, by comparing the magnitude of effect-sizes of a diverse set of determinants of BLM, we provide a comprehensive view of the most important factors determining support for BLM.

Methods
To identify suitable public opinion datasets containing questions on support for the BLM movement, we conducted a systematic search using the Roper Center for Public Opinion Research (iPoll), on December 23, 2020. The process is akin to that of a systematic review, except that the unit of observation is a dataset as opposed to a scientific article. We detail our search methodology and inclusion criteria in the online appendix (Sections 2.1-2.3).

Results
For each dataset, we ran a bivariate Pearson's correlation between each identified predictor and BLM support. We then pooled correlations of the same predictor and conducted a random-effects meta-analysis, with the Sidik-Jonkman estimator, to assess the heterogeneity of the effects (Harrer et al. 2021). The choice for random, as opposed to fixed, meta-analyses was based on the need to account for the study effect showing more variance than when drawn from a single population and to control for the effect of statistical heterogeneity (Schwarzer, Carpenter, and Rücker 2015). To interpret the magnitude of meta-analytical effect-sizes and contextualize it in relation to the published literature, we use empirically derived guidelines for effect-size interpretation (Lovakov and Agadullina 2017). According to this system, correlation coefficients of 0.12, 0.24 and 0.41 correspond, respectively, to the twenty-fifth (small effect), fiftieth (medium effect), and seventy-fifth (large effect) percentiles of the distribution of effect-sizes in the literature. This study design yields high statistical power which can detect even very small effect-sizes. We also conducted an individual respondent meta-analysis (Stewart and Tierney 2002)-also known as integrative data analysis (Curran and Hussong 2009)-obtaining largely equivalent results.
We summarized 234 associations between support for Black Lives Matter and its predictors in figure 3. We describe the found thirty-seven predictors in seven clusters-namely, demographics, race and ethnicity, partisanship and ideology, voting behavior and political attitudes, attitudes towards immigration, racial attitudes and economic attitudes. In addition, two measures of heterogeneity are provided. Tau-squared (t 2 ) represents the variation between the effects observed across datasets with smaller values indicating lower variation (Deeks, Higgins, and Altman 2011). I 2 quantifies the percentage of variation across datasets that is due to heterogeneity between the datasets rather than sampling error (Higgins 2003) with values above 50% indicating substantial heterogeneity (Deeks, Higgins, and Altman 2011). Although I 2 is more commonly used than other heterogeneity measures, it becomes less adequate for high-powered meta-analyses, as the sampling error tends to zero and I 2 tends to 100% (Harrer et al. 2021). Given the high sample size of the meta-analyses reported here, for 92% of predictors heterogeneity as measured by I 2 is considered to be substantial (>50%) even if the estimate's confidence interval is comparatively small. Furthermore, we note that t 2 , which is insensitive to both k and N, tended to zero for all predictors, indicating low heterogeneity between datasets.
The purpose of Study 2 is to provide confirmatory evidence for-and expand the conceptual breadth ofpredictors investigated in Study 1. For brevity, we deferred to section 2.4 of the online appendix, a detailed statistical description of results, while summarizing main takeaways here.
Results of the multiple meta-analyses largely corroborate the results of Study 1. We find that BLM support is negatively associated with age, being white, Republican, and conservative, and positively related with being female (versus male), African American, and believing Blacks are more likely to experience both police violence and discrimination in the United States. Another consistent effect is that experiencing discrimination is not associated with BLM support. Two predictors, however, showed divergent results in our two studies-namely, education and income-which were consistently found to be non-significative across reviewed studies, but significative in Study 2: education was positively related to BLM support while income showed the inverse association.
With Study 2, we build on the scope of demographics, voting behavior, and political, racial, immigration and economic attitudes. Results show that not identifying with a religious denomination, being single (versus married), employed (versus retired), and Asian American is associated with BLM support. Being Hispanic American, however, appears to bear no association with support for BLM. Political behavioral predictors display some of the highest effect-sizes. Voting intention for the Republican nominee in 2016 and 2020 as well for Republican House of Representatives members are highly related to BLM opposition. Similarly, political attitudes such as approval of Trump's presidency and disapproval of Obama's, favorability towards Trump and antagonism for Clinton, and believing Trump would handle race relations better than Biden are all strongly and negatively related to BLM support. We also expand the findings of Study 1 to attitudes about immigration, race, and economics. All three immigration-related attitudes display the same pattern, the more one believes immigrants take jobs away from Americans, holds the view that illegal immigrants living in the U.S. should be required to leave the country, and favors building a wall along the United States and Mexico border, the lower the support for BLM. A similar pattern is found regarding racial attitudes. Thinking Black (versus white) Americans are more likely to experience police violence, that generations of slavery and discrimination have created conditions that make it difficult for Blacks to work their way out of the lower class, that race relations in the United States are getting better (during the Obama presidency), and that racial discrimination against Blacks is a serious problem, are all consistent predictors of support for BLM. Economic attitudes all suffer from the president-in-power-effect (e.g., correlations between BLM support and being optimistic about future and personal finances flip direction once Trump is elected).

Subgroup Analyses
We were also interested in the predicted probability of support for the BLM by various societal groups and Overview of multiple meta-analyses Notes: The N denotes the number of respondents across datasets used to calculate the meta-analytical correlation between support for BLM and a given predictor. Note that this number may differ across predictors using the same datasets as it is contingent upon the predictors' amount of missing data. The total number of datasets is denoted by k. Two measures of heterogeneity of the meta-analytical Pearson's r are provided: t 2 and I 2 . Bars to the right of the dashed line denote positive associations between the predictor and BLM support, whereas bars to the left of the dashed line denote negative associations with BLM support. Lastly, on the right, the meta-analytical Pearson's r is provided accompanied by its 95% confidence interval.
subgroups, exploring differences in how political cleavages may shape support for BLM, and identifying group-based disparities that could potentially motivate targeted interventions. We present results in percentages to facilitate interpretation. All discussed results are statistically significant (p < 0.001), even when correcting for multiple comparisons. Refer to Section 2.5 of the online appendix for additional details on methodology and results (e.g., Ns in each cell).
Results are displayed in figure 4 and reveal that out of the eleven societal groups, symbolic ideology appears to cleave support for BLM the most, with a difference of more than 62 perceptual points between liberals and conservatives. Party identification follows with a gap of 54% in BLM support between Democrats and Republicans. Race represents the third largest cleavage on BLM support: while 82.3% of African Americans support BLM, only 47.3% of whites endorse the movement. More mild differences are observed across religious affiliation, where we observe Catholics and Protestants being substantially less prone to support BLM than non-religious individuals. Disparities in age are also modest, with young adults (18-29 years) showing 21% more support for BLM than adults older than 65 years. An identical gap is observed between single/never married and married individuals, with the former showing more support than the latter. Across remaining groups, differences between subgroups become less pronounced.
These analyses reveal that there is substantial heterogeneity within societal groups, suggesting stronger appreciation for subgroup analyses will likely yield illuminating insights into contemporary racial attitudes and BLM support.

Figure 4
Percentage of endorsement of BLM across societal groups and subgroups

Multivariate Regressions
So far, we sought to analyze the demographic, political, social, and psychological bases of support for BLM, focusing especially on zero order associations, as is the standard practice when the goal is to synthesize empirical research (Donnelly et al. 2018) and maximize the generalizability of presented results against idiosyncratic controls, covariates, and statistical models (Gelman and Loken 2014;Breznau et al. 2021). This bivariate approach was also used due to the limitation that common predictors are seldom found across several datasets, and thus merging individual datasets to conduct an omnibus regression would not be feasible except for a few demographics and partisanship. For this reason, to assess which predictors are comparatively stronger and robust to controls in a multivariate setting, we resorted to applying a series of regressions-maximizing the use of available variables per dataset-to identify a subset of predictors that consistently and reliably exhibited significant effects (Zou and Hastie 2005). This approach minimizes the influence of false positives and spurious relationships and complements the multiple meta-analyses against the inferential pitfalls of separately modeling closely related constructs.
In figure 5 (panel A), predictors of BLM are presented and broken down by the number of times each predictor is available across datasets and found to be statistically significant in the multivariate regressions. Overall, twenty-seven out of the thirty-seven predictors were significantly related to BLM support controlling for other variables at least once (eighteen at least twice, ten at least trice), suggesting BLM in America is a complex, multifaceted issue influenced by various competing and intersectional factors. Notwithstanding, symbolic ideology (conservatism) is the most frequent predictor significantly associated with BLM support (ten out of the eleven datasets) whereas party identification is only a significant predictor of BLM support in six out of the thirteen datasets. Attitudes towards immigrants, support for Trump, and attitudes like believing BLM protests are legitimate and perceiving racial discrimination are also often significant predictors of BLM support. Furthermore, even after adjusting for other variables, Blacks versus others, female versus male, and education remained often significant predictors, suggesting support for BLM is cleaved across race, gender, and education levels. An overview of variables present in each model is provided in section 2.6 in the online appendix.
To identify the most robust and strong predictors, we standardized regression slopes (Gelman 2008) and proceeded to calculate the average effect of each predictor across datasets. We display the standardized regression coefficients in figure 5 (panel B) where we applied a transparency filter to highlight the number of times each predictor was significant as a proxy for reliability of the estimated effect-size. Among the most frequently significant predictors, symbolic ideology and party identification have the largest, most robust effect-sizes, followed by race (Blacks versus others) and gender. Attitudes towards immigrants, political attitudes (i.e., approval of Trump, vote intention for the House of Representatives) and attitudes like BLM legitimacy, protesting for racial equality, perceptions of racial discrimination, and perceived systematic racism all had a moderate effect on BLM support. Even if vote intention for Trump in 2016 and Republicans in the 2020 election have the largest effectsize in predicting BLM opposition, we note these variables were only significant twice and once, respectively. We also provide a dot-and-whisker plot of standardized averaged regression coefficients per dataset in figure 6. It can be observed that the direction and magnitude of most predictors is consistent across datasets. For example, despite some small variations in effect size, across all datasets in which symbolic ideology is a significant predictor of BLM support (after controlling for other variables), conservatism always predicts lack of BLM support. The same is true for party identification: self-identifying as a Republican always predicts lack of BLM support. Gender and race are always found to be positive predictors of BLM support, with females and Blacks showing greater support for the movement compared to males and whites, respectively. Education and experienced discrimination are the only predictors for which the direction of the effect changes across datasets, which explains why their confidence bands include zero in figure 5 (panel B). This variation across datasets might suggest that the effect of these variables on BLM support is contingent on available controls, sample characteristics, model covariates, or measurement differences.
The goal of these multivariate analyses was to tease apart the effects of the most reliable predictors of BLM support found in Studies 1 and 2. When predictors were entered simultaneously into multivariate regression models, endorsement of politically conservative ideology was the most reliable predictor. Other factors including race, gender, and partisanship-while important-were less reliable in comparison with ideology. These patterns of associations are indicative of conservatism being a key factor explaining variance in support for the BLM movement, and of racial inequality more broadly.
These findings are hard to square with the ideological innocence hypothesis (Kinder and Kalmoe 2017). According to Kalmoe (2020, 18), "partisan identification has broader and stronger empirical and theoretical foundations for guiding public opinion" and that "mass partisanship routinely outperforms ideology." We found the opposite. Not only ideology vastly outperforms partisanship, but 1316 Perspectives on Politics Figure 5 Relative proportion of significant predictors when controlling for other variables (panel A) and averaged standardized effect-sizes (panel B) for multiple linear regressions across datasets Notes: Panel A compares the number of times a predictor is available across datasets and the times it is a significant predictor of BLM support when accounting for all other available predictors. Panel B shows the meta-analytical standardized regression coefficients (Gelman, 2008), with a 95% confidence interval, across 13 datasets. On panel B, the grayscale highlights the number of times each predictor is statistically significant, controlling for other variables, with darker colors representing higher frequency of statistical significance.

Figure 6 Standardized regression coefficients
Notes: This dot-and-whisker plot of standardized regression coefficients (Gelman 2008) summarizes, per dataset, the significant predictors of BLM support, controlling for other variables. partisanship is outperformed by race and gender (at least as far as consistency is concerned). It is worth noting that since Converse (1964), the explanatory and theoretical value of ideology for the study of political behavior-and that of racial attitudes and racism specifically-is often downplayed, when not omitted completely (see Azevedo et al. 2019, Azevedo andJost 2021;, for a critique and counterexamples).

Discussion
I carry the memory of living under terror-the terror of knowing that I, or any member of my family, could be killed with impunity-in my blood, in my bones, in every step I take.
- Khan-Cullors and Bandele 2018, 15 Police brutality against Blacks is deeply endemic to the fabric of America. Black individuals are 3.5 times more likely to be met with police brutality than white people (Geller et al. 2021), and Black males-in particular-are 5 times more likely to be killed by the police than their white counterparts (Robinson 2017). Adding insult to injury, African Americans are also systematically incarcerated at much higher rates than white people (Carson 2020), more prone to be convicted and receive harsher sentences (Anwar, Bayer, and Hjalmarsson 2021), and vulnerable to the U.S. misdemeanor system impeding their social-economic mobility. The U.S. criminal justice system is riven by racial bias (Bronner 2020). When it comes to wealth inequality, the median white household owns 7.8 times more wealth than Black households (Moss et al. 2020) and these economic disparities are likely to persist for generations as Black Americans are significantly less likely to achieve upward economic mobility than white Americans (Chetty et al. 2020). In the context of access and quality of care, Black patients are at an increased risk of not receiving proper care for chronic diseases (Chin, Zhang, and Merrell 1998) and of not being rated as suitable candidates for organ transplantation (Epstein 2000). The COVID-19 pandemic is no exception, having disproportionately affected Black communities both in terms of infection and death rates (Millett et al. 2020;Tai et al. 2021). These facts are but a few pieces of the jigsaw puzzle depicting the various systems that detrimentally and disproportionately affect Black lives. The evidence of systemic racism against African Americans in the United States is so overwhelming it should be incontrovertible.
Against this backdrop, we sought to uncover the demographic, political, and psychological bases of BLM support by conducting a systematic literature review and multiple meta-analyses. Study 1 uncovered important patterns in the published literature for which Study 2 provided independent, high quality confirmatory evidence, as well as an expansion of its conceptual breadth. We found the specific arrangement of associations between support for BLM and demographic, social, and political constructs is indicative of a system of interconnected beliefs whose goal is to hinder-if not impede-racial progress. In interpreting the theoretical breadth and prevalence of investigated constructs, there is a near-perfect correspondence between opposition to BLM and positive attitudes towards American institutions that are deeply rooted in systemic racism (e.g., the police, Republican Party, Conservatism), the attitudes disseminated by these actors, and the psychological predispositions habitually associated with endorsing them. This interpretation is furthered by the multivariate analyses, which pitted predictors against one another, showing that liberal-conservative differences prevailed over other robust predictors of BLM attitudes. We argue for a greater appreciation of the ideological character of contemporary racial attitudes and behaviors on the part of whites and Blacks in the United States and contend that political ideology matters insofar as it underlies both the legitimation of existing forms of inequality and the desire for societal change. Hence, the inclusion of ideology in models of collective action is paramount .
Our results have theoretical and methodological implications for the study of BLM, especially regarding promising directions for future research. First, the nomological network of BLM support is powerfully interwoven and should not be studied disjointedly-or be relegated to accessorizing partisanship-but as a system of interrelated attitudes toward social, racial, and political "others". Second, results suggest that American public opinion on BLM is largely cleavaged across race and gender, demonstrating a fertile territory for intersectional and critical approaches to the study of racial inequality, and political behavior in general. Similarly, as cross-racial coalitions are essential to the fight for racial equality, research on support for BLM across different racial-ethnic groups is sorely needed. Third, as we uncovered insightful patterns diving deeper into societal subgroups such as in the case of religious affiliation and urbanicity, research providing more granularity on religious denominations, sexual and gender orientations, levels of income, education, and urbanicity might uncover patterns our approach was unable to. Fourth, when available, psychological factors were found to be highly robust predictors of BLM support suggesting that an in-depth systematic examination of the psychological bases of BLM support would shed light into its underlying motives and dispositions, beyond that of partisanship or ideology. Lastly, as BLM ventures overseas, it would be exceedingly valuable to investigate whether the findings presented in this work generalize cross-nationally.

Conclusion
BLM demonstrations rose against police brutality, one of many institutions of American society that "prioritizes the comfort of white Americans over the lives of people of color" (Oliver 2020). The prerogative of the BLM movement is, and has always been, about bringing relief to those who had long lived under a regime of racial oppression (Jefferson and Ray 2022). In the words of its co-founder: "We have a right to live" (Khan-Cullors and Bandele 2018, 14). Yet despite the rather elementary mission, white backlash has been unrelenting. To the privileged-when all you have ever known is unfettered, unimpeded privilege-other people's right to live can feel like oppression. It is to this sentiment that rancid, racist rhetoric resonates. To Trump, GOP supporters, and contemporary conservatism in the United States, the goal of the BLM movement is "to achieve the destruction of the nuclear family, abolish the police, abolish prisons, abolish border security, abolish capitalism, and abolish school choice" (Villarreal 2020). This conservative reactionarism (Parker and Barreto 2014;Robin 2011) leads to viewing the oppressed as "terrorists," the victims as "thugs," and those who are demonstrating for racial equality as "rioters". BLM's very existence and pursuit for racial equality have been subverted to galvanize support for counter movements ) and racist causes (Taylor 2019), further entrenching systemic racism in the American political ethos (Jefferson and Ray 2022). To conclude, as America grapples with the extent of its systemic racism, it is fundamental to broaden the scope of research on race and ethnicity in general and on the BLM movement more specifically to amplify the voices of racial minorities, inform policies that challenge American institutions, and reshape them towards racial equality.

Supplemental Materials
To view supplementary material for this article, please visit http://doi.org/10.1017/S1537592722001098.