The study of socially sensitive issues has always been a challenge for psychologists. AA policies concern a socially sensitive issue, meaning that respondents frequently have incentives to under or over-represent their true attitudes to make them more in line with social norms. Whites, for example, may over-represent their true support for AA policies when asked directly, by fear of appearing prejudiced against non-whites or unconcerned about racial discrimination and inequalities. Non-whites, for their part, may under-represent their true support when asked directly, by fear of being judged as unconcerned about questions of merit and fairness. When social desirability bias is expected, scholars have frequently adopted indirect measures of attitudes like the list and endorsement experiments to circumvent this social desirability bias in attitude report (Kim & Kim, 2016; Sniderman & Grob, 1996; Vidigal, 2018). Turgeon, Chaves and Wives (2014), for example, used a list experiment to measure attitudes about racial quotas in university admissions in Brazil. Their results show that, contrary to expectations, white students do not over-report their support for racial quotas, but that non-white students tend to under-report their support. This study, like others before it, documents the importance of measuring attitudes about sensitive issues adopting indirect measures to avoid over or under-reporting of true attitudes.

In social cognition research, other indirect techniques like the implicit attitudes procedures have been used to measure attitudes on sensitive issues (e.g., Hahn & Gawronski, 2018). Implicit attitude measures are useful to deal with social desirability, but, more importantly, for present purposes, they help in understanding how the implicit processes influence the support for AA policies, especially in populations directly affected by such policies. To this end, we investigate the interplay between university students’ explicit and implicit attitudes on AA in university admission policies. We explore this issue using data from Brazil, a context where there has been little research into social psychological processes and, in particular, about implicit social cognitive processes.

The implicit association test (IAT) is the most commonly used procedure for measuring implicit attitudes and has been used over a broad range of issues (Nosek, Hawkins, & Frazier, 2011). The IAT also has better psychometric properties than the alternatives and has demonstrated its ability to differentiate groups with previously known attitudes (Bar-Anan & Nosek, 2012). We adopt the IAT and make use of its discriminative ability to compare implicit and explicit attitudes on university admissions procedures, namely racial quotas and the traditional entry exam (vestibular).

Scholars have demonstrated that the relationship between implicit and explicit measures is affected by many factors (Hofmann, Gschwendner, Nosek, & Schmitt, 2005). Nosek (2007), for example, shows that the attitude domain influences the relationship between explicit and implicit measures of attitude with higher correlations for attitudes about politics (e.g., party support) and lower correlations in the evaluation of stereotypes or group prejudice (e.g., white’s attitudes about blacks). Nosek (2007) shows that more socially sensitive issues, like AA policies, tend to exhibit greater implicit-explicit discrepancy. This finding is important because predictive attitude validity is affected by the strength of the association between implicit and explicit measures, with greater discrepancies decreasing attitude predictive validity over behavior (Karpen, Jia, & Rydell, 2012). Consequently, a greater discrepancy between implicit and explicit attitudes on AA may reduce our ability to predict behaviors related to AA, more generally.

Attitudes on AA policies are thought-out and firmly held by university students because such policies are essential to them. Precisely, Brazilian students are directly concerned about the policy because it affects them directly in significant ways by benefiting non-white students and potentially jeopardizing the access of other racial groups to public universities, namely white students. To be sure, racial quotas in university admissions make racial identity central to understanding support for such policy because students’ racial identity (white/non-white) directly affects their likelihood of being admitted to the university. The salience of race is thus undeniable to understand students’ explicit and implicit support for university admission rules. We expect, like past work in the area, that now-whites will be more supportive of AA policies like racial quotas (Schmermund, Sellers, Mueller, & Crosby, 2001).

Beyond race, implicit and explicit attitudes on AA policies can also be explained by political beliefs and predispositions (Sniderman, Crosby, & Howell, 2000). Specifically, there are many reasons why one may support or oppose AA. For example, some value greater equality and diversity and want governments to enact policies like AA to achieve such goals. Others, on the other hand, believe it is best for governments to stay put and let people’s fortune be defined solely by merit and self-reliance. Admittedly, such political beliefs and preferences are expected to correlate in similar ways with both implicit and explicit support for AA policies.

Finally, we believe that prejudice against minority groups, in general, may also carry substantial influence on both the implicit and explicit support for AA. Specifically, prejudiced people are generally less supportive of policies targeting minority groups (e.g., Kuklinski et al., 1997), and this pattern should also be observed among prejudiced students.

The implicit and explicit scores were measured on different metrics. We converted them to a 0-10 scale to facilitate comparison. For both measures, higher scores indicate a preference for the vestibular over the quota system. The mean for the explicit and implicit measures is similar at 5.10 (SD = 2.44) and 5.19 (SD = 1.62), respectively. Both measures are moderately positively correlated (r = 0.37, p < 0.001). Both the means and the correlation indicate moderate convergence between the two measures.

We find that the implicit and explicit measures are both affected by race, political preferences, and prejudice, and in ways expected. First, as shown in Figure 1, we find that non-whites demonstrate more support for AA policies, as compared to whites. Specifically, race has a small effect on implicit attitude (F (1, 476) = 4.59, p = 0.033, r = 0.10), with white students expressing a greater preference for the vestibular (M = 5.31, SD = 1.63) than non-white students (M = 4.99, SD = 1.61). There is also a small effect for race on explicit attitude (F (1, 476) = 7.20, p = .008, r = 0.12), with white students expressing a greater preference for the vestibular (M = 5.36, SD = 2.48) than non-white students (M = 4.76, SD = 2.39). Admittedly, the differences are small, but still explain an important difference between whites and non-whites over the issue of AA. The differences between explicit and implicit measures among whites and non-whites, however, are small and do no reach a conventional level of statistical significance. This result is surprising and does not suggest, contrary to past research in the area, much of an effect for social desirability among white participants.

Figure 1 Figure 1 Comparison of the explicit and implicit measures about the traditional entry exam (vestibular) among white and non-white students; data shown are means and 95% confidence intervals.

Figure 2 shows a similar analysis but concerns the effect of political preferences about AA on our explicit and implicit measures. Here, we find larger differences than for race. Specifically, support for AA has a moderate effect on the implicit measure (F (1, 490) = 47.16, p <0.001, r = 0.30), with non-supporters of AA showing a greater implicit preference for the vestibular (M = 5.65, SD = 1.44) than supporters (M = 4.69, SD = 1.66). The effect of support for AA is even larger for the explicit measure (F (1, 490) = 470.23, p <0.001, r = 0.70), with non-supporters showing a much stronger explicit preference for the vestibular (M = 6.75, SD = 1.66) than supporters (M = 3.34, SD = 1.82). The within-group analysis also reveals important differences between our implicit and explicit measures. Specifically, we find that support for AA was strongly related to the magnitude of the difference between the implicit and explicit measures. We find that supporters of AA tend to under-report their explicit attitude about the vestibular while non-supporters tend to over-report it.

Figure 2 Figure 2 Comparison of the explicit and implicit measures about the traditional entry exam (vestibular) among supporters and non-supporters of racial quotas; data shown are means and 95% confidence intervals.

Finally, Figure 3 shows the effect of prejudice on our explicit and implicit measures. We find a small effect for prejudice on the implicit measure (F (1, 490) = 13.19, p < 0.001, r = 0.16), with high-prejudiced participants showing a greater implicit preference for the vestibular (M = 5.56, SD = 1.70) than low-prejudiced participants (M = 5.00, SD = 1.55). We also find a moderate effect for prejudice on the explicit measure (F (1, 490) = 71.43, p < 0.001, r = 0.36), with high-prejudiced participants also reporting a greater explicit preference for the vestibular (M = 6.35, SD = 1.96) than low-prejudiced participants (M = 4.50, SD = 2.42). The within-group analysis also reveals that there are significant differences between implicit and explicit measures. Specifically, high-prejudiced participants tend to over-report their explicit support for the vestibular while low-prejudiced participants, for their part, tend to under-report it.

Figure 3 Figure 3 Comparison of the explicit and implicit measures about the traditional entry exam (vestibular) by level of prejudice; data shown are means and 95% confidence intervals.

This research aims to shed some light on the interplay between implicit and explicit attitudes on AA policies and explore how such attitudes relate to race, political preferences, and prejudice in a context where AA policies are essential, but little understood. We find that race, political beliefs and preferences about AA, and prejudice are all correlated, and in ways expected, with both our implicit and explicit measures about the traditional entry exam (vestibular). Specifically, 1) whites, as compared to non-whites, show slightly higher implicit and explicit support for the vestibular; 2) participants holding beliefs and preferences more supportive of AA tend to show lower implicit and explicit support for the vestibular, as compared to those holding less supportive beliefs and preferences; and, 3) high-prejudiced participants, as compared to low-prejudiced participants, exhibit higher implicit and explicit support for the vestibular.

We find that the relationship between implicit and explicit attitudes is strongly influenced by political beliefs and preferences about AA and prejudice, but, surprisingly, race does not exert a strong influence. Previous research has shown that attitudes on AA policies depend upon whether or not the individual is a member of the affected group, i.e., is a beneficiary of the policy (Telles & Bailey, 2002). Our results are in line with this finding, but, surprisingly, we find only a minimal effect for race, as compared to other determinants. The differences in effect sizes between race, political beliefs and preferences, and prejudice are considerable and reinforce the importance of grounding statistical reasoning on more than alpha values (Cumming, 2014). The smaller effect of race can be a result of the high level of miscegenation in the Brazilian population, leading Brazilians who share common characteristics (e.g., skin tone) to self-categorize themselves in different racial groups. As a result, it dilutes the effect of race, at least, as compared to other societies with higher (past) racial segregation.

Our results provide evidence that other variables are more important than race to explain the interplay between attitudes about the admission system. In particular, we find that participants more supportive of AA tend to under-report their explicit support for the vestibular, as compared to their implicit support. Among those participants less supportive of AA, we find the opposite pattern, with an over-reporting of the explicit support, as compared to the implicit one. We believe that this pattern is a way for participants to explicitly demonstrate their political point of view because participants have less control over the implicit measure. This finding raises some concerns about the use of explicit measures to gauge attitudes about AA.

A similar pattern is observed among high and low-prejudiced participants. Precisely, we find that high-prejudiced participants tend to over-report their explicit support for the vestibular, as compared to their implicit support. Among low-prejudiced participants, on the other hand, we find an under-reporting of the explicit support for the vestibular, as compared to the implicit one. This result suggests that low and high prejudiced individuals tend to explicitly demonstrate their position more strongly than their implicit support show. These patterns of under and over-reporting are potentially the results of the IAT measure that is less vulnerable to attempted manipulation (Steffens, 2004), at least as compared to explicit attitudes measures. Our findings are in line with the work by Gawronski et al. (2015) that note that the IAT is particularly helpful when dealing with socially sensitive issues like AA. To be sure, the discrepancies observed between the implicit and explicit measures suggest that the IAT technique can help to improve our understanding of attitudes on affirmative action.

Past research has shown that the object domain is one important determinant of the correlation between implicit-explicit attitudes (Nosek, 2007). We also know that groups with previously known attitudes about a specific object may show different explicit and implicit attitudes. Our findings of AA policies in Brazil are in line with these previous findings. The issue of AA policies represents a complex object domain because it involves considerations beyond race like values about meritocracy, access to resources, social justice, prejudice, and, of course, concerns about social desirability. The important discrepancies we find between our implicit and explicit measures on the vestibular is an illustration of this domain complexity.

Our results, although suggestive, should be interpreted cautiously, given two important limitations of our research design. First, our study recruited a convenience sample of volunteers. Presumably, these participants may hold different attitudes about racial quotas and the vestibular than the average university student. They may also hold strong attitudes about the admission process, contributing to their decision to take part in our study. Second, our recruited participants all come from the same university, a university where the issue of racial quotas has been on the agenda and discussed extensively for more than 10 years. Our results could potentially be different in other contexts where AA policies are a novelty. Thus, we must be cautious about the generalizability of our results and encourage the replication of our study in other contexts and adopting probabilistic samples.

Finally, our study adds to others in the evaluation of AA policies (e.g., Debell, 2016) and provides paths to improve them. Our study has contributed to the debate about AA by demonstrating the need also to explore the implicit attitudes of people affected by such policies. Our findings show that implicit and explicit attitudes on AA are related to race, as past work has shown, but also to other factors like political beliefs and preferences and prejudice. Taking into account these factors help in understanding how support for AA policies is developed and maintained, especially in a context where AA policies have already been widely implemented.