Testing the efficacy of three informational interventions for reducing misperceptions of the Black–White wealth gap

Significance An intervention study exposed a US community sample to messages about Black–White racial inequality. Interventions including data bearing on Black–White wealth inequality elicited higher estimates of that inequality that persisted for at least 18 mo, aligning with federal data measuring the Black–White wealth gap. The data interventions also increased acknowledgment of White Americans’ structural advantage and reduced beliefs in personal achievement as the remedy for racial inequality. In contrast, a narrative-based intervention, including information on a single Black family contending with racial inequality, did not shift inequality estimates or change respondents’ explanations. This study suggests how social science data can be used to create more realistic perceptions of racial inequality—a prerequisite to enacting equity-enhancing policy.


Additional Study Procedure Details
Respondents received $5 each for completing surveys at time 1, 3, and 4 and an additional $15 for completing the laboratory visit. Respondents consented to video recording during the nonjudgmental listening portion of the study but were also allowed to withdraw consent after completing the session. Video recordings ended after respondents had a chance to answer our four interview questions. As respondents were escorted out of the laboratory, experimenters gave a brief debriefing and made sure that respondents were willing to be contacted again by the behavioral lab. At this time, respondents did not know they would be recontacted for this study.
For recruitment to our two follow up surveys, respondents received three emails before they were considered non-responsive. Following data analysis for our final follow-up survey, respondents received a thorough written debriefing about the study and a preliminary report detailing the results of the intervention and how their data would be summarized for future scientific reports. Respondents were encouraged to provide further commentary about the research at this time, and several did so.

Perceptions of inequality in other domains.
We also asked respondents about current levels of Black-White income inequality using the same methodology. Respondents were asked, "For every $100 earned by an average white family, how much do you think is earned by an average black family?" (M T1 = 58.58, SD T1 = 27.9). Respondents were reminded that selecting 100 represents equality for this item and responded on the 0-200 slider scale.
Reducing Misperceptions of the Black-White Wealth Gap 3 We also measured perceptions of general wealth inequality based on prior research (62).
Respondents were asked "If the US consisted of 100 individuals with 100 units of wealth, how much wealth do each of the following groups [top 20/the second 20/the middle 20/the fourth 20/and the bottom 20] have?" Respondents then allocated these 100 units of wealth among the quintiles (responses were forced to sum to 100). We used estimates of the wealth owned by the top 20% as a measure of general estimates of class-based wealth inequality (M T1 = 60.11, SD T1 = 26.23). As an additional measure of wage inequality, respondents were asked to estimate how much an average Chief Executive Officer (CEO) earns in income for every $1 the average unskilled worker earns, on a 0 to 600 scale (M T1 = 365.78, SD T1 = 177.32; 1).

Individual difference measures.
For exploratory purposes as part of a larger study, we also assessed several beliefs about economic opportunities and society, as well as several individual difference measures. An 8-item scale assessed respondents' beliefs regarding how easy it is to change one's social class (2). Four statements asked respondents to indicate their agreement with a number of statements (e.g., "It is common for people who are motivated enough to go 'from rags to riches'"; 1 = Strongly disagree, 4 = Neither agree nor disagree; 7 = Strongly agree) and two asked respondents to indicate how easy or difficult it is to change one's social class (e.g., "These days, how easy is it to change one's social class?"; 1 = Very hard, 7 = Very easy). This scale formed a reliable index (α = 0.89; M T2 = 3.15, SD T2 = 1.00).
Social dominance orientation (SDO) assesses support for inequality among social groups and was measured using an 8-item scale (SDO 7(S) ; 3; α = .84). Respondents rated their opposition or support for several ideas on 7-point scales (1 = Strongly oppose; 4 = Neither oppose nor favor; 7 = Strongly favor). Example items include, "Some groups of people are simply inferior to other groups" and "We should do what we can to equalize conditions for different groups" (reverse-coded).

Supplementary Analyses
Intervention impact on estimates of inequality in other domains. Interestingly, though our manipulation did not include specific information on income disparities between White and We also examined both general wealth inequality and CEO pay, which respondents reported at three time points, as a function of our three intervention conditions. Unlike estimates of racial wealth equality, estimates of general wealth equality showed a significant effect of time Moderation by respondent race. We also explored whether respondent race moderated the effects of the intervention in a 2(race)X3(intervention)X4(time) ANOVA. Our results were similar to those reported without accounting for respondent race in that the significant time by intervention interaction emerged F(6,568)=3.45, p=.002, with no effect of time F (3,568) Speech differences by condition. A supplementary analysis comparing speech differences between the data/combined and narrative intervention conditions found a significant effect for achievement related words, aligning with our expectation that the data/combined Reducing Misperceptions of the Black-White Wealth Gap 6 intervention would reduce the use of achievement related speech surrounding racial inequality (Table S1). No statistically significant differences emerged for sentence length, word length, or the negativity and positivity of words. Interestingly, the data/combined intervention conditions elicited more speech than the narrative condition, which could be indicative of respondents processing the data they were exposed to in that condition.   Table S2).

Moderation by SDO and political ideology.
Theoretically, our manipulation is likely to be most effective for respondents who tend to be most liberal and in particular, most egalitarian, and by contrast the intervention will be least effective for those high in conservatism and in particular, social dominance (5-6). We tested this with two linear regression analyses predicting Black-White wealth equality estimates with the structural/combined intervention and its interaction with Social Dominance Orientation or political ideology. In the first analysis, we adjusted for time 1 estimates of Black-White wealth equality, race, income, education, ideology, gender, and age. The results of this analysis find some support for a reduced effectiveness of the intervention among high SDO respondents (see Table S3). As the Table shows, the structural/combined intervention again reduced time 2 estimates of Black-White wealth equality, and time 1 estimates were again a significant predictor of time 2. In this model a significant effect of SDO emerged such that high SDO respondents reported higher time 2 Black-White wealth equality. When relaxing criteria for statsitical significance to p < .10, the intervention and SDO effects were qualified by a significant interaction that was in line with the above expectations: the structural/combined intervention was less effective at moving respondents who were high versus low in SDO.  Moderation by age. We contend that interventions such as this should be conducted as early and as often as possible. Nevertheless, one reasonable hypothesis given the results of our study is that an educational intervention, like the one we use in this study, would be best suited to younger audiences with less deeply entrenched beliefs about racial inequality. However, we believe there are at least a few reasons to be skeptical of this hypothesis. According to a large and growing body of research, racial socialization processes happen at very early ages in American families (7). As a result of this socialization process, entrenched beliefs about racial inequality are likely to be present at the time when children are able to understand more complex Reducing Misperceptions of the Black-White Wealth Gap 10 components of societal structures through data-such as those we used in our data interventions (8). As well, though children may have more malleable views about racial inequality, these views are also likely to be actively shaped by adults, as part of the curriculum of primary education institutions, who are likely to adhere to narratives of racial progress in their pedagogical choices (9-10).
Analysis of our own data, capitalizing on the age range of our adult sample, also suggests caution about the heightened effectiveness of this sort of intervention with younger populations: We conducted three linear regression analyses examining the extent that age moderated the effect of the intervention on respondent estimates of the Black-White wealth gap. None of these regression analyses found a significant effect of age (ts < 1.38, ps > .168) or a statistically significant interaction between the intervention and age (ts < 1.53, ps > .128). Importantly, because these analyses are with adults, these data cannot directly answer the question of the differential effectiveness of the intervention among children.