Racial cues from unfamiliar sources

Americans often rely on cues to evaluate political messages, including cues based on a source’s identity. For example, George W. Bush’s identity as a born-again Christian may have increased evangelical support for the Iraq War (Jacobson, Reference Jacobson2007), while Hillary Clinton’s gender may have contributed to opposition against her 1993 health care plan (Winter, Reference Winter2008). For racial attributes in particular, a range of studies show that whether a source (at least those who are recognizable political figures) is Black or White influences Americans’ policy positions (Kuklinski and Hurley, Reference Kuklinski and Hurley1994; Tesler, Reference Tesler2012; Knowles, Lowery, and Schaumber, Reference Knowles, Lowery and Schaumber2010; Peffley and Hurwitz, Reference Peffley and Hurwitz2010).

The Political Science literature on racial source cues has primarily examined recognizable public figures, for whom racial or other identities are almost always salient. As Kinder and Dale-Riddle (Reference Kinder and Dale-Riddle2012, pp. 24–25) note, Barack Obama’s racial identity was unavoidable during his 2008 presidential campaign:

Beyond race, Kinder and Dale-Riddle (Reference Kinder and Dale-Riddle2012) explain that other identity cleavages are usually salient for public figures, such as former President John F. Kennedy’s status as a Catholic.

While much of Political Science research focuses on well-known figures, Americans frequently encounter political content from non-recognizable sources. At the local level, voters are often unfamiliar with candidates running for office. Similarly, political messages in city council meetings, school board hearings, union meetings, and religious gatherings often come from speakers whose identities are not widely known.

The same holds for traditional news media. Television and radio programs frequently feature local activists, organizers, and commentators. Print journalism similarly presents political messages from journalists and columnists who may not be widely recognized. In a period before most Americans consumed news from digital devices (Pew Research Center, Reference Center2024a ), Domke et al. (Reference Domke, Lagos, LaPointe, Meade and Xenos2000) showed that the racial identity of an article’s author could influence public attitudes toward race-conscious policies.

Fast forward 25 years, and exposure to non-recognizable political sources is arguably more widespread than ever. As stated in the introduction, research suggests that a majority of Americans now receive their news from social media (Pew Research Center, Reference Center2024b ), often encountering political content incidentally – at unexpected times and from unfamiliar sources (Boczkowski, Mitchelstein, and Matassi, Reference Boczkowski, Mitchelstein and Matassi2018; Fletcher and Nielsen, Reference Fletcher and Nielsen2018). Moreover, social media applications algorithmically curate content to maximize engagement. Consequently, individuals are likely to encounter emotionally charged material (Vicario et al., Reference Vicario, Bessi, Zollo, Petroni, Scala, Caldarelli, Stanley and Quattrociocchi2016), such as discussions of racial bias in policing during the peak of the BLM movement. In this environment, individuals are perhaps more likely to form impressions of political content based on subtle source cues, including racial identity (Banks, Reference Banks2014a ; Banks, Reference Banks2014b ).

Whether through local television broadcasts, print journalism, or radio discussions, audiences have long encountered political content from unfamiliar figures. However, the scale and frequency of such exposure appear to have grown in the social media age. This transformation underscores the importance of understanding how people interpret subtle cues from unfamiliar sources, particularly in politically charged discussions. As a result, this paper focuses on how subtle racial cues shape perceptions of political messages, especially when conveyed by non-recognizable sources.

Name-based racial cues

For non-prominent figures, racial identity cues tend to be subtle, with names being one of the most implicit yet powerful indicators. Research on racial discrimination in political responsiveness and hiring decisions frequently relies on name-based cues to indicate race. For example, studies use contrasts like “Jake Mueller” vs. “DeShawn Jackson” to signal White and Black political constituents (Butler and Broockman, Reference Butler and Broockman2011) or “Emily Walsh” vs. “Lakisha Washington” to signal White and Black job applicants (Bertrand and Mullainathan, Reference Bertrand and Mullainathan2004).

I understand racially distinctive names as prototypical cues of racial category membership. That is, cues can establish someone as prototypically Black, whereby people categorize others’ Blackness based on resemblance to that prototype (Monk, Reference Monk2022; Rosch, Reference Rosch1973). Given their historical significance (Cook, Logan, and Parman, Reference Cook, Logan and Parman2014; Cook, Parman, and Logan, Reference Cook, Parman and Logan2022; Lieberson and Mikelson, Reference Lieberson and Mikelson1995), a name like “DeAndre Jackson,” in contrast to “Connor Smith,” serves as a distinguishing social marker that reinforces racial group boundaries, much like the ethnic distinctions described by Barth (Reference Barth1969).

However, names often signal more than just race; they can also indicate class, gender, and other overlapping social categories (Elder and Hayes, Reference Elder and Hayes2023; Landgrave and Weller, Reference Landgrave and Weller2020; Fryer and Levitt, Reference Fryer and Levitt2004). In some cases, it may be useful to isolate cues of racial membership from cues of other social categories. On the other hand, Elder and Hayes (Reference Elder and Hayes2023, p. 769) point out that researchers need not focus on “isolating the effects of race independent of other relevant attributes.”

This is the premise my experiment builds on, which aligns with what Hu (Reference Hu2023) describes as a “thick constructivist” account of race. Within this framework, prototypical Black cues that also signal other social categories do not confound the role of Black identity, but rather “reveal what it is to be Black or what being Black socially consists in” (Hu, Reference Hu2023, p. 4). In other words, the fact that people may associate prototypical White names with high income and prototypical Black names with lower income is itself part of the social experience of being Black in America.

Mechanism class 1: racial attitudes and “symbolic racism”

The literature on racial attitudes and “symbolic racism” suggests that racial source cues may trigger particular racial attitudes, thereby determining public support for various social policies. As Harris and Lieberman (Reference Harris and Lieberman2015, p. 10) write, “overt expressions of prejudice are almost universally considered illegitimate.” Nevertheless, many Americans hold negative racial stereotypes that constitute a form of “symbolic racism” (Sears, Reference Sears, Katz and Taylor1988). Such stereotypes – defined as “projecting assumptions or expectations about the likely capacities and behaviors of members of a racial or ethnic group onto members of that group” (Bobo, Reference Bobo, Smelser, Wilson and Mitchell2001, p. 275) – are pervasive. For example, in a seminal piece coining the term “symbolic racism,” Sears (Reference Sears, Katz and Taylor1988, p. 57) showed that many Americans agreed with the statements that “the government pays too much attention to blacks” and “blacks who receive welfare could get along without it if they tried.”

Such negative racial stereotypes – for example, that Black Americans are “lazy” or prone to criminality – can be activated by subtle racial cues without, as Hutchings and Jardina (Reference Hutchings and Jardina2009, p. 397) write, “appearing to violate widely held norms of racial equality.” Such subtle racial cues, in turn, affect Americans’ support for a wide array of social policies, those that are both explicitly race-conscious (e.g., affirmative action in higher education) and non-explicitly race-conscious (e.g., welfare reform) (Peffley, Hurwitz, and Sniderman, Reference Peffley, Hurwitz and Sniderman1997; Hurwitz and Peffley, Reference Hurwitz and Peffley2005; Gilens, Reference Gilens1999; Gilliam and Iyengar, Reference Gilliam and Iyengar2000; Hurwitz and Peffley, Reference Hurwitz and Peffley1997; Sears and Citrin, Reference Sears and Citrin1985; Winter, Reference Winter2008; White, Reference White2007). However, Bobo and Kluegel (Reference Bobo and Kluegel1993) suggest that symbolic racism exerts a greater impact on public opinion about race- as opposed to, for example, income-conscious social policies, a point that subsequent research supports (e.g., Rabinowitz et al., Reference Rabinowitz, Sears, Sidanius and Krosnick2009).

Mechanism class 2: proxies for expertise or trustworthiness

As research in cognate fields like social psychology has demonstrated, when the policy under discussion is explicitly race-conscious, a Black as opposed to White source cue may bestow that source with a form of “racialized expertise,” ultimately leading to greater policy support. As Crosby and Monin (Reference Crosby and Monin2013, p. 335) state, for many White Americans making judgments about racial discrimination, “members of target groups appear to be ‘prejudice experts’ by virtue of their group membership.” Hence, a Black source advocating for policies to combat racial discrimination in policing might lead to greater support for that policy compared to if the source had been a White individual. This particular mechanism in which race serves as a proxy for expertise is perhaps underexplored in Political Science but closely related to the large body of research on source credibility, typically understood to encompass both expertise and trustworthiness (Page, Shapiro, and Dempsey, Reference Page, Shapiro and Dempsey1987; Iyengar and Valentino, Reference Iyengar, Valentino, Lupia, McCubbins and Popkin2000; Pornpitakpan, Reference Pornpitakpan2004; Hovland and Weiss, Reference Hovland and Weiss1951).

While Black identity may serve as a proxy for expertise, this same identity might also serve as a proxy for a lack of trustworthiness, particularly when the message pertains to a race-conscious policy. Evidence in social psychology suggests that Black sources delivering messages about racial discrimination are often perceived as less objective in that they are thought to be characterized by in-group bias (Torrez, Dupree, and Kraus, Reference Torrez, Dupree and Kraus2024; Czopp and Monteith, Reference Czopp and Monteith2003; Schultz and Maddox, Reference Schultz and Maddox2013; Kaiser and Miller, Reference Kaiser and Miller2001; Rasinski and Czopp, Reference Rasinski and Czopp2010). In perhaps the most relevant antecedent to this paper, Domke et al. (Reference Domke, Lagos, LaPointe, Meade and Xenos2000) vary the author’s race of a magazine article about race relations and show that Black source cues lead to decreased support for government policies to improve race relations. Consistent with the social psychology literature, Domke et al. (Reference Domke, Lagos, LaPointe, Meade and Xenos2000) attribute this result to perceptions that Black sources advocating for race-conscious policies are, as Burkhart, Sigelman, and Frith (Reference Burkhart, Sigelman and Frith1997, p. 313) write, “lacking in objectivity” due to having a “vested interest in what they are saying.”

Predicted effects of racial cues on policy support

The theories above yield predictions about what we should expect to observe under Black vs. White source cues for different policy types:

• If the mechanism of anti-Black stereotypes were at play, then we would expect lower levels of support for both policy types when the source name cues Black as opposed to White identity. Nevertheless, it could be the case that the policy must be explicitly race-conscious in order for a subtle racial source cue to be salient enough to activate respondents’ anti-Black stereotypes (Bobo and Kluegel, Reference Bobo and Kluegel1993; Rabinowitz et al., Reference Rabinowitz, Sears, Sidanius and Krosnick2009). In this case, we would also expect a more negative effect of the racial source cue, given the race- as opposed to income-conscious policy.

• Two additional mechanisms are those of racial expertise (in which race proxies for expertise on race-conscious policies in the minds of respondents) and in-group bias (in which a Black source cue leads respondents to view that source as less objective about race-conscious policies). If the racialized expertise mechanism has high explanatory power, then we would expect the effect of the Black source cue to be positive for the race-conscious policy, but not for the income-based policy. The mechanism of in-group bias, by contrast, suggests that the Black relative to White source cue should yield less support for the race-conscious policy, but this effect need not be negative for the income-based policy.

Experimental conditions

Each respondent observed a vignette explaining five policy proposals. In following the typology of Bobo and Kluegel (Reference Bobo and Kluegel1993), distinguishing between explicitly race- and income-conscious policies, I randomly assign respondents to either a policy that aims to combat racial inequality in policing or a policy that aims to combat income inequality. The first policy type is explicitly race-conscious, referring directly to police violence against Black civilians. The second policy type pertains only to income inequality with no explicit references to race. All explicit racial references are underlined in Table 2 below, though they were not emphasized in the actual texts shown to respondents.

Table 2. Policy conditions presented in the experiment

The source in both vignettes is a community organizer, rather than an unknown political candidate or elected official, for two principal reasons. First, a community organizer better aligns with the theoretical mechanisms outlined above – particularly those in which racial source cues may shape perceptions of racialized expertise and trustworthiness grounded in lived experience. In contrast, political candidates or elected officials may be perceived as guided more by party affiliation than personal experiences. Second, using a fictional candidate or official could raise methodological concerns: Highly engaged participants may seek more information or recognize the deception, potentially triggering Hawthorne effects and compromising the experiment’s validity. Nevertheless, a valuable direction for future research is experiments that incorporate candidates or elected officials as sources, offering additional insight into the effects of racial source cues.

This experiment’s primary target is not the average treatment effect (ATE) of the policy type (either race-conscious or race-neutral). Instead, the experiment’s primary target is the impact of racial source cues, which may vary depending on whether the policy context is race-conscious or race-neutral. To be sure, the income-conscious policy may not be race-neutral since Americans often interpret income-based social policies through a racial lens (Gilens, Reference Gilens1999; DeSante, Reference DeSante2013). However, the income condition addresses economic inequality without explicitly addressing racial disparities or the absolute level of poverty among specific racial groups. The policing condition, by contrast, focuses on racial disparities in law enforcement, as well as the absolute level of police violence experienced by Black people.

Moreover, this experiment took place during the peak of the BLM movement, a period when activists and media prominently framed police violence in racial terms. By contrast, public debates over income inequality arguably focused more on class divisions than on racial ones. That is, unlike in the Reagan era – when political rhetoric invoked racially coded tropes like “welfare queen” – prominent figures like Bernie Sanders and others tended to emphasize class rather than race in their framing of economic inequality.

Table 3 presents evidence supporting the claim that the policing policy is interpreted by respondents as race-conscious, while the income policy is not. Nearly 60% of respondents perceive the policing policy as race-conscious, which may not seem particularly high on its own. However, this percentage is markedly greater than the roughly 25% who view the income policy as race-conscious. This stark contrast indicates that the policing policy is far more race-conscious than the income policy.

Table 3. Percent of respondents who view policing and income policies as Black-centered issues

Another crucial feature of the two policy conditions is that they reflect actual debates at the time of the experimental survey. The race-conscious policy is based on the BLM network’s actual proposals to mitigate racial bias in policing. The income-conscious policy comes from the Hamilton Project, a policy initiative within the Brookings Institution.

Because the policy conditions are based on actual civil society platforms, both conditions encompass multiple issues rather than isolated topics. As a result, the two policy conditions lack the precise control a researcher might achieve by designing a platform focused on neatly demarcated, standalone issues. This choice is intentional: Citizens rarely encounter policies in isolation; rather, media, partisan groups, and civil society organizations present policies as interconnected. For example, commitments to raising the minimum wage are often bundled with Medicaid expansion, and legislative bills themselves usually combine multiple related issues.

While the ATE of the policy condition is not the primary focus of this experiment, it is worth noting that the bundling of issues would have implications for inference of the effects of the policy conditions: As Clifford, Leeper and Rainey (Reference Clifford, Leeper and Rainey2024, p. 1233) write, “[r]esearchers frequently develop a general theory and hypotheses (e.g., about policy attitudes), then conduct a study on a specific topic (e.g., environmental attitudes).” To address this gap between general theory and single-topic treatments, Clifford, Leeper and Rainey (Reference Clifford, Leeper and Rainey2024) define a key causal target as the average across the ATEs of single-topic treatments, which can be estimated through the random selection of single-topic conditions from a full set of topics (Clifford, Leeper and Rainey, Reference Clifford, Leeper and Rainey2024; Clifford and Rainey, Reference Clifford and Rainey2024).

However, despite its advantages, topic sampling may not capture the ATE of policy platforms as they exist in the real world, where issues are typically presented as bundles rather than in isolation. The ATE of a set of bundled topics may differ from the average of single-topic ATEs because survey responses may be non-separable (see, e.g., Lacy, Reference Lacy2001). As a result, the overall ATE marginalized over the ATEs of single-topic experiments may not match the ATE of a real-world policy bundle, potentially making the latter more substantively meaningful.

The two policy types may condition the effects of either a Black source cue (signaled by the name DeAndre Jackson) or a White source cue (signaled by the name Connor Smith). Hence, within the two randomly assigned policy types, I randomly expose participants to either the Black or White source cue. The number of respondents in each of the experimental conditions is in Table 4 below.

Table 4. Number of respondents by experimental condition (source name × policy type)

To increase compliance (i.e., subjects’ actually reading the treatment or control text), I disabled the “Next” button for 90 seconds – the approximate time it takes to read through the text – on the Qualtrics page displaying the text. This likely increased the proportion of participants who fully engaged with both conditions. While I did not conduct a manipulation check, although Table 3 could be interpreted as such, I emphasize that both policy conditions are nearly identical in length and layout, differing only in content. If content alone does not affect whether a respondent reads the text, then randomization ensures that, in expectation, reading rates are equal across policy types. If such a difference were to exist, it would be an insightful feature (rather than a bug) of how respondents engage with race-conscious versus race-neutral policy platforms.

The primary outcome of interest is respondents’ answers to the following question: “How willing are you to support police-reform (poverty) policies? Support could mean attending a meeting, signing a petition, or voting for a candidate that supports such policies.” The answer choices were “Very willing” (1), “Moderately willing” (2), and “Not at all willing” (3). Before analysis, I recoded these responses to 1, 0, and $ - 1$ , respectively. My primary causal targets of interest are the average difference in policy support if the experimental subjects were to receive the policy message from a Black (relative to White) source given each of the two policy types.

As an anonymous reviewer insightfully noted, this outcome measure cannot distinguish between respondents who oppose a policy and those who support it but are unwilling to take concrete actions to express that support. Therefore, this experiment should be interpreted as speaking to how racial source cues influence respondents’ stated intentions to mobilize into action. This measure aligns with the depiction of the source in both policy conditions as a community organizer whose role, by definition, is to rally the community into action.

Nevertheless, this outcome measure may also align closely with the notion of support distinct from the willingness to act on that support: The questionnaire includes examples of actions, from higher-cost ones like “attending a meeting” to lower-cost ones like “signing a petition.” Consequently, respondents who select “Not at all willing” are plausibly only those who oppose the policy rather than those who support it but are reluctant to act. However, even though policy support and stated willingness to take action may closely align, they are not necessarily the same, making this distinction important to emphasize.