57 1 population. Abramson et al. (2022) point out that null-effects can coexist with racist responses. Additionally, negative responses can average-out to null-effects when accompanied by some highly positive results. One could argue that this might be the case in these studies. As we will point out in the next section, sharing the same race/ ethnicity causes more positive assessments. That means that the white population might be more negative than the general population (see Figures 1, 2, 3 and 4). In the articles of which we have data on which respondents are white we can disentangle whether white respondents produce more negative effect sizes. Appendix 6 demonstrates that this is not the case; white subsets do not produce more negative responses than the general population. It could still be the case, however, that some respondents respond more intensely, but this does not apply to white respondents on average. A third possible reason the null-effects could be misleading is that voters are more racist in their preferences than these results suggest, but that the effect sizes are toned down because of social desirability bias. When voters feel that their answers might be socially undesirable, Krupnikov et al. (2016) find that social desirability bias is mitigated if the design provides participants the option to explain their answers. In the condition where respondents were allowed to explain their answers, the negative bias towards Black candidates was much stronger (idem). This suggests that other studies which do not include the option to explain oneself, might underestimate the effect of racism against racial/ethnic minority political candidates. At the same time randomizing multiple attributes can could already mitigate social desirability bias. Petsko et al. (2022) find that respondents of conjoint experiments tend to single out one attribute they find very important, instead of letting all attributes weigh together evenly. Each extra attribute serves as a potential explanation a respondent can hide behind while the respondent is focusing on one specific attribute (idem). Shockley (2020) finds clear evidence that conjoint experiments indeed garner less socially desirable results that simple survey questions. She asks whether respondents prefer candidates from a certain group and finds null-effects, while the outcomes of the conjoint reveal they do favor one group over the other (idem). Moreover, social desirability bias only is an explanation for outcomes in the unjust stereotypes school of thought, not if you understand stereotypes to be a useful heuristic to base voting decisions on. Voters may very well prefer white over Black candidates because they expect the Black candidate to stand for policies they disagree with, such as universal healthcare or taxing the rich (Crowder-Meyer et al., 2018). In Europe, racial/ethnic minority citizens are most likely to vote and run for traditional left-wing parties (Aktürk and Katliarou, 2021: 391). Voters therefore might expect more left-leaning policy positions from racial/ethnic minority candidates than their majority counterparts. It is difficult to disentangle whether voters prefer certain candidates because of their preferences (unjust stereotypes) or because of the policy they are deducing from
RkJQdWJsaXNoZXIy MjY0ODMw