Introduction
The U.S. is a multiracial democracy, with nearly 40% of its population consisting of people of color (PoC) – African Americans, Asian Americans, Latinos, and other non-Whites (Pérez Reference Pérez2021).Footnote 1 This sea change in America’s demographics is steering many social scientists toward better understanding how members of these racial and ethnic groups form coalitions to reach common political goals (Craig and Richeson Reference Craig and Richeson2016; Cortland et al. Reference Cortland, Craig, Shapiro, Richeson, Neel and Goldstein2017; Sirin et al. Reference Sirin, Valentino and Villalobos2021). Prior research suggests that a sense of shared discrimination between racially marginalized groups triggers a sense of solidarity with PoC, which then yields greater downstream support for policies that benefit an ingroup that is not one’s own (for meta-analytic evidence, see Pérez et al. Reference Pérez, Vicuña and Ramos2023). For example, when Black adults feel their ingroup is discriminated against similarly to Latinos, their sense of solidarity with PoC is heightened, leading them to express greater support for pro-Latino policies (e.g., more flexible immigration policies). This solidarity mechanism also operates among Latinos, Asian Americans, and Middle Easterners and North Africans.
While the evidence affirming this solidarity mechanism is extensive, most of it stems from a measurement-of-mediator design (Baron and Kenny Reference Baron and Kenny1986). In this setup, researchers manipulate a putative cause (shared discrimination) but measure the proposed mediator (solidarity) and outcomes (policy support). The putative mediator is observed, not manipulated, because researchers believe it is very difficult and/or psychologically unrealistic to directly manipulate. However, this design choice raises the specter of omitted variable bias, which can undermine causal inferences about a measured mediator (Bullock and Green Reference Bullock and Green2021). Analysts have bolstered claims about solidarity’s mediating effects through sensitivity analyses (Imai and Yamamoto Reference Imai and Yamamoto2013), but these efforts only help to bound these effects away from zero. Thus, some researchers recommend evaluating mediated treatment effects via experimental designs, rather than through statistical adjustments and/or sensitivity analyses (Pirlott and MacKinnon Reference Pirlott and MacKinnon2016; Wayne Reference Wilkinson2023).
Panel A in Fig. 1 below depicts this situation, with dashed lines reflecting the vulnerable path in this framework and the correlated errors implicating a potential third unmeasured variable that might confound this relationship. We revisit this solidarity mechanism and use a blockage mediation design that “neutralizes” solidarity’s downstream influence on policy support (Pirlott and MacKinnon Reference Pirlott and MacKinnon2016). The objective is to address the confounding in Panel A, Fig. 1 by manipulating a mediator’s downstream influence (W), as depicted in Panel B, Fig. 1. If a blockage manipulation reduces solidarity’s downstream influence on policy support (relative to a control), then we have more conclusive evidence about this mediator’s causality.
Figure 1. Solidarity Mediates the Impact of Shared Discrimination on Support for Pro-Outgroup Policy. (a) Mediated Relationship is Highly Vulnerable to Confounding. (b) Mediated Relationship is Less Vulnerable to Confounding.
We employ this design across two pre-registered experiments and a pre-registered internal meta-analysis of two large samples of Black adults (N∼2,692). As the ingroup that is recognized by other non-Whites as the prototypical person of color, Black individuals define the norms and values of the larger mega-group, people of color (Chin et al. Reference Chin, Luna, Huo and Pérez2023), and embrace this category because of their racially liberal ideology (Carter, Wong, and Guerrero Reference Carter, Wong and Guerrero2021). This makes our research setting a “most likely” case (Gerring Reference Gerring2001), allowing us to observe solidarity’s causal effects in a core population of color as it reacts to another racially marginalized population with whom it often encounters tense political relations (i.e., Latinos) (Wilkinson Reference Zou and Cheryan2015; Benjamin Reference Benjamin2017).Footnote 2
We report three results. First, consistent with prior work, we again find that a heightened sense of shared discrimination causes Black adults to express greater solidarity with people of color (d∼.40) (Pérez et al. Reference Pérez, Vicuña and Ramos2023). This elevated solidarity level is strongly associated with greater downstream support for pro-Latino policies. Second, our blockage manipulation consistently weakens solidarity’s downstream relationship with support for pro-outgroup policies. Specifically, calling attention to the unique roots of discrimination against Black and Latino individuals (i.e., slavery versus immigration) modestly reduces solidarity’s downstream impact on support for pro-outgroup policies, but this effect is imprecisely estimated in each study. Third, an internal meta-analysis of both experiments finds this blockage effect is substantively small but statistically reliable (d∼.10) across both studies, suggesting that solidarity’s effects are likely causal and resistant to this divisive threat. We discuss our results’ implications for U.S. inter-minority politics.
Revisiting solidarity’s downstream effect through a blockage mediation design
Prior work reveals a robust downstream association between heightened solidarity with PoC and increased support for pro-outgroup policies (d = 0.79; Pérez et al. Reference Pérez, Vicuña and Ramos2023). In a measurement-of-mediator design, this downstream path is vulnerable to confounders. One way to minimize this threat by design is to knock off course solidarity’s impact on pro-outgroup policies (Pirlott and MacKinnon Reference Pirlott and MacKinnon2016). If an intervention increases (decreases) its relationship with pro-outgroup policies, we gain more confidence that solidarity’s downstream effects are causal, rather than merely correlational.
This is the essence of a blockage mediation design, depicted in Panel B in Fig. 1. There, X = shared discrimination, M = our mediator, solidarity between PoC, and Y = support for pro-outgroup policies. To evaluate whether M’s impact on Y is causal, our blockage design introduces W = an additional manipulation intended to “block” solidarity’s downstream influence. Insofar as W moderates M’s downstream influence on Y, we have more diagnostic evidence that solidarity between PoC is a causal mediator (Pirlott and MacKinnon Reference Pirlott and MacKinnon2016). Our pre-registered hypothesis (https://aspredicted.org/TKM_D31) is that the interaction between M and W will be negatively signed, suggesting a downstream reduction in solidarity (H1).
We innovated the typical design used to measure solidarity’s downstream effects (Pérez et al. Reference Pérez, Vicuña and Ramos2023) by adding a new manipulation in the path connecting this mediator with support for pro-outgroup policy. Specifically, after we measure solidarity between PoC, we randomly assign Black participants to read about how it is very difficult to compare the discriminatory experiences of Black people with those of Latinos, since each community faces discrimination for unique reasons (i.e., slavery versus immigration)(cf. Zou and Cheryan Reference Zou and Cheryan2017). We will describe this manipulation in more detail in the next section of the paper, but here we note that this treatment is known to induce distinctiveness threat (Pérez Reference Pérez2021) – the sense that the unique attributes and experiences that comprise one’s ingroup are in jeopardy (Brewer Reference Brewer1991; Branscombe et al. Reference Branscombe, Ellemers, Spears, Doosje, Ellemers, Spears and Doosje1999). This threat operates by undermining the perceived similarity between marginalized ingroups (e.g., Black and Latino people), which shared discrimination induces (Cortland et al. Reference Cortland, Craig, Shapiro, Richeson, Neel and Goldstein2017). In the context of people of color, this distinctiveness threat motivates individuals to focus on their own specific ingroup (e.g., Black people) and away from the larger shared group, people of color (Craig et al. Reference Craig, Rucker and Richeson2018). This implies that inducing distinctiveness threat in the downstream path should reduce solidarity’s influence on support for pro-outgroup policies (H1), our primary hypothesis.
Procedures, methods, and estimation
We test (H1) across two pre-registered experiments with Black participants in the context of Black-Latino relations. Both studies shared an identical design but were run on different survey platforms: 1) Dynata (Study 1; N = 1,719; November, 2023); and 2) Cloud Research (Study 2; N = 973; February 2024). Section 1 in the supplementary material (SM.1) reports demographics and balance tests for these samples. SM.2 reports our pre-registrations.Footnote 3
In each experiment, Black adults completed a brief pretreatment schedule of items measuring demographic (e.g., age, education) and political attributes (e.g., ideology) to help characterize our samples. We then informed Black participants that they would be reading some developing news stories, which they would be asked to give their feedback on. At this point, participants were assigned to a control or treatment condition (i.e., manipulation 1). In the control, participants read an article, attributed to the Associated Press (AP), about the declining number of giant tortoises throughout the globe. In the treatment condition, participants read an AP article of comparable length about continued discrimination against Latinos in the United States and how this discrimination is like the one encountered by Black people in the United States. Specifically, the treatment article was titled, “ Despite Their Presence in the United States for Decades, Many Latinos are Still Treated as Second Class Citizens, As Evidenced by Hate Crimes Data,” with the article noting trends in hate crimes toward Latinos. Thus, the structure of this manipulation is one where the title conveys the thrust of our treatment, while the body of the article provides some brief reasoning for that claim. Drawing on a similarity principle (Cortland et al. Reference Cortland, Craig, Shapiro, Richeson, Neel and Goldstein2017; Pérez et al. Reference Pérez, Vicuña and Ramos2023), the article concludes by briefly noting how these discriminatory trends toward Latinos are like those of Black people, “many of whom experience a similar sense of exclusion.” The full wording and visuals used in the manipulation are reported in SM.3.Footnote 4
After reading their assigned article, participants completed a manipulation check, which consisted of a true/false item about the thrust of the article they read (Pérez et al. Reference Pérez, Vicuña and Ramos2023). Passage rates for the shared discrimination manipulation check were 91.27% (Study 1) and 96.12% (Study 2). Immediately following this check, participants then completed three validated items measuring solidarity between people of color, which is the proposed mediator of shared discrimination in this framework (Pérez et al. Reference Pérez, Goldman, Huo, Nteta and Tropp2024). Using a scale from 1-strongly disagree to 5-strongly agree, participants completed each item below:
-
1) I feel solidarity with people of color, which includes Asian, Black, and Latino people.
-
2) The problems of Black, Latino, Asian, and other people of color are similar enough for them to be allies.
-
3) What happens to people of color in this country has something to do with what happens in my life as a Black person.
We scale replies to these items (Study 1, α = 0.728; Study 2, α = 0.781) and transform each one to a 0–1 interval (Study 1, M = 0.613, SD = 0.242; Study 2, M = 0.648, 0.248). This lets us interpret all coefficients as percentage-point shifts.
After assessing solidarity with PoC, Black participants were then randomly assigned to our blockage manipulation before completing our outcome variables (see SM.4). Our second manipulation also consisted of a short article characterized as a developing story attributed to the Associated Press, with the title conveying the gist of our treatment and the rest of the brief providing additional reasoning behind the claim in the title.Footnote 5 This article is an adaptation of a treatment that effectively manipulates distinctiveness threat among Black, Asian, and Latino adults (Pérez Reference Pérez2021). Specifically, Black participants were randomly assigned to a control group (no information) or a treatment condition where they read a new article titled “With a Unique History and Set of Political Goals, Black Alliances with Latinos Don’t Always Make Strategic Sense.” As such, this article induces distinctiveness threat by explaining that:
“…it is very hard to compare African Americans’ experience with slavery and its aftermath to the social and political exclusion faced by Latinos. Indeed, the United States continues to marginalize many Blacks as second-class citizens, even though African Americans have been in this country since its founding. Other people of color are not treated in this peculiar way.”
After this blockage manipulation, participants responded to a second true/false manipulation check that captured the gist of the article they read. 86.73% (Study 1) and 95.89% (Study 2) of participants passed the check. Then, Black participants reported their support for three pro-Latino policy proposals, each answered from 1-strongly disagree to 5-strongly agree, which we scale (Study 1, α = 0.651; Study 2, α = 0.632) and transform to a 0–1 range interval (Study 1, M = 0.681, SD = 0.216; Study 2, M = 0.718, SD = 0.214).
-
1) Introducing harsher penalties for hate crimes committed against Latinos.
-
2) Renewing temporary relief from deportation for undocumented Latino immigrants brought to the United States as children.
-
3) Supporting the use of affirmative action for Latinos in jobs and education.Footnote 6
Using these data, we estimate the model in panel B in Fig. 1 in a structural equation modeling (SEM) framework, where we simultaneously estimate the effect of the first treatment (shared discrimination) on expressions of Black solidarity with people of color and the downstream influence of solidarity on Black support for pro-Latino policies moderated by the second manipulation (distinctiveness threat) (Hayes Reference Hayes2022). Our design resembles a factorial experiment with varied levels of two manipulated variables (Shadish et al. Reference Shadish, Cook and Campbell2002). However, unlike a standard 2 x 2, our blockage experiment has two outcomes of interest (i.e., PoC solidarity, policy support), with one of those outcomes sandwiched between our pair of manipulations. Thus, one can consider our blockage experiment an adaptation of a 2 x 2 with the express goal of altering our mediator’s downstream influence. As such, the quantity of interest in both experiments is the interaction term between solidarity and our distinctiveness manipulation, which we expect to be negatively signed, indicating a reduction in solidarity’s downstream effect on support for pro-Latino policy. Following prior work (Pérez et al. Reference Pérez, Vicuña and Ramos2023), all estimates include liberal ideology as a covariate to better estimate the impacts of solidarity, which is positively associated with liberal ideology and influences PoC’s political attitudes in similar directions (Kam and Trussler Reference Kam and Trussler2017). We further evaluate our results through a pre-registered internal meta-analysis, which evaluates any systematic trend(s) across both conceptually similar experiments (Goh et al. Reference Goh, Hall and Rosenthal2016). We describe this analysis after discussing Study 1–2’s results.
Results
Table 1 reveals that exposure to shared discrimination with Latinos heightens Black adults’ expression of solidarity with PoC. In Study 1, this effect increases solidarity by nearly 10 percentage points (.099, SE = 0.001, p<0.01). Study 2 produces a similar effect (.088, SE = 0.015, p<0.001). Transforming these coefficients to Cohen’s d values, the average effect across both studies is d∼0.385, which is considered a medium-sized effect (d values reflect standardized mean differences). These patterns replicate prior work on shared discrimination’s effects on expressions of solidarity with PoC (Pérez et al. Reference Pérez, Vicuña and Ramos2023).
Table 1. Shared discrimination boosts Black solidarity with PoC, while distinctiveness threat reduces its downstream influence on support for pro-Latino policies
Note: Entries are coefficients from a structural equation model, with standard errors in parentheses. Coefficients reflect percentage-point shifts. Shaded entries represent the effects of our blockage manipulation on solidarity’s downstream influence. Following prior work (Pérez et al. Reference Pérez, Vicuña and Ramos2023), liberal ideology is included as a covariate to optimize the robustness of solidarity, which is positively associated with liberal ideology and influences PoC’s political attitudes in similar directions (Kam and Trussler Reference Kam and Trussler2017).
* p <0 .05, two-tailed.
Next, we examine the downstream association between a heightened sense of solidarity with PoC and Black support for pro-Latino policies. Consistent with prior work, heightened solidarity is significantly and strongly associated with support for pro-Latino policies in Study 1 (.424, SE = 0.027, p<0.001) and Study 2 (.355, SE = 0.034, p<0.001). Converting these associations to Cohen’s d values, the average relationship between solidarity and our outcome is strong across studies (d∼.850), consistent with prior published studies (Pérez et al. Reference Pérez, Vicuña and Ramos2023).
Finally, we evaluate the effectiveness of our blockage manipulation in reducing the downstream association between solidarity and support for pro-outgroup policies. The relevant coefficients are shaded in grey in Table 1. As hypothesized, the interaction between solidarity and distinctiveness threat is consistently negative but falls short of statistical significance. For example, in Study 1, our blockage manipulation appears to reduce solidarity’s downstream relationship with pro-Latino policy about four percentage points (−0.041, SE = 0.037, p<0.271), although this trend is imprecisely estimated. Similarly, Study 2’s blockage manipulation decreases solidarity’s relationship with pro-Latino policy by about 3 percentage points, but again, this trend is imprecisely estimated (−0.028, SE = 0.047, p<0.550). These patterns suggest our downstream manipulation slightly knocks solidarity’s influence off its course, but these effects are statistically insignificant at the 5% level in each sample. Despite this, the lower bound of the confidence interval for each blockage effect suggests we cannot rule out substantively larger effects than those uncovered here (Study 1: −0.041, 95% CI: [−0.113, .032]; Study 2: 95% CI: [−0.121, 0.064]. This pattern is inconsistent with a negligible result (Rainey Reference Rainey2014) and steers us toward further investigating this modest blockage effect by capitalizing on the enhanced statistical power of combining both of our experiments.Footnote 7
To this end, we draw on Goh et al.’s (Reference Goh, Hall and Rosenthal2016) template and use a fixed-effects regression that re-evaluates whether the negative interaction between solidarity and distinctiveness threat is reliable in our pooled sample. Given our directional prediction, we pre-registered a one-tailed test for this. Our analysis uncovers a negative interaction term between solidarity and distinctiveness threat that is modest in size and statistically significant (d=−0.070, p<0.029 one-tailed). More specifically, our meta-analyzed blockage manipulation significantly decreases solidarity’s downstream relationship with pro-Latino policy by nearly one-tenth of a standard deviation. This design-based reduction in our mediator’s downstream effect aligns with an interpretation of solidarity as causal, rather than simply correlational (Pirlott and MacKinnon Reference Pirlott and MacKinnon2016).
Implications
In comparison to prior work (Cortland et al. Reference Cortland, Craig, Shapiro, Richeson, Neel and Goldstein2017; Pérez et al. Reference Pérez, Vicuña and Ramos2023), our results further bolster the view of solidarity between people of color as one causal mechanism behind coalition-building efforts between PoC. In this way, our results strengthen the available evidence in favor of solidarity as a mechanism whose effects can be increased and decreased systematically. This does not mean, however, that solidarity is the only mechanism connecting shared discrimination to support for pro-outgroup policies: a proposition that requires additional and careful research.
Our findings also suggest that, insofar as solidarity is important to understanding inter-minority politics in the U.S., solidarity’s influence is not invulnerable to divisive threats. Although shared discrimination unifies people of color under a banner of heightened solidarity, inter-minority politics is characterized by a dense and cross-cutting information environment where messages to unify these motley groups are often countered by messages to drive a wedge in PoC coalitions (Vaca Reference Vaca2004; Brilliant Reference Brilliant2014, Zou and Cheryan Reference Zou and Cheryan2017). Footnote 8,Footnote 9 Thus, our findings relay the point that solidarity’s downstream effects depend, in a significant way, on the communication environments that people of color encounter. The experimental design we used here provides a flexible template to theorize and evaluate additional threats to the unity that solidarity produces between PoC, thereby providing additional evidence about its causal nature.
Acknowledgements
This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2024S1A3A2A07046269).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/XPS.2025.3
Data availability
The data, code, and any additional materials required to replicate all analyses in this article are available at the Journal of Experimental Political Science Dataverse within the Havard Dataverse Network at https://doi.org/10.7910/DVN/UWJYCB.
Competing interests
We have no conflicts of interest to report for this research.
Ethics statement
These studies were reviewed by the Institutional Review Board at the University of California, Los Angeles (protocol number: 23-001615). This study adheres to APSA’s Principles and Guidance for Human Subjects Research.
Open access
