Comments on "Racial Resentment, Prejudice, and Discrimination"

Forthcoming in the Journal of Politics is Peyton and Huber 2021 "Racial Resentment, Prejudice, and Discrimination". Study 1 estimated discrimination among White MTurk workers playing a game with a White proposer or a Black proposer. The abstract indicated that:

Study 1 used the Ultimatum Game (UG) to obtain a behavioral measure of racial discrimination and found whites engaged in anti-Black discrimination. Explicit prejudice explained which whites discriminated whereas resentment did not.

I didn't see an indication in the paper about a test for whether explicit prejudice predicted discrimination against Blacks better than racial resentment did. I think that the data had 173 workers coded non-White and and 20 workers with missing data on the race variable, but Peyton and Huber 2021 reported results for only White workers, so I'll stick with that and limit my analysis to reflect their analysis in Table S1.1, which is labeled in their code as "main analysis".

My analysis indicated that the discrimination against Black proposers was 2.4 percentage points among White workers coded as prejudiced (p=0.004) and 1.3 percentage points among White workers coded as high in racial resentment (p=0.104), with a p-value of p=0.102 for a test of whether these estimates differ from each other.

---

The Peyton and Huber 2021 sorting into a prejudiced group or a not-prejudiced group based on responses to the stereotype scales permits assessment of whether the stereotype scales sorted workers by discrimination propensities, but I was also interested in the extent to which the measure of prejudice detected discrimination because the non-prejudiced comparison category included Whites who reported more negative stereotypes of Whites relative to Blacks, on net. My analysis indicated that point estimate for discrimination was:

* 2.4 percentage points against Blacks (p=0.001), among White workers who rated Blacks more negatively than Whites on net on the stereotype scales,

* 0.9 percentage points against Blacks (p=0.173), among White workers who rated Blacks equal to Whites on net on the stereotype scales, and

* 1.8 percentage points in favor of Blacks (p=0.147), among White workers who rated Blacks more positively than Whites on net on the stereotype scales.

The p-value for the difference between the 2.4 percentage point estimate and the 0.9 percentage point estimate is p=0.106, and the p-value for the difference between the 0.9 percentage point estimate and the -1.8 percentage point estimate is also p=0.106.

---

NOTES

1. I have blogged about measuring "prejudice". The Peyton and Huber 2021 definition of prejudice is not bad:

Prejudice is a negative evaluation of another person based on their group membership, whereas discrimination is a negative behavior toward that person (Dovidio and Gaertner, 1986).

But I don't think that this is how Peyton and Huber 2021 measured prejudice. I think that instead a worker was coded as prejudiced for reporting a more negative evaluation about Blacks relative to Whites, on net for the four traits that workers were asked about. That's a *relatively* more negative perception of a *group*, not a negative evaluation of an individual person based on their group.

2. Peyton and Huber 2021 used an interaction term to compare discrimination among White workers with high racial resentment to discrimination among residual White workers, and used an interaction term to compare discrimination among White workers explicitly prejudiced against Blacks relative to Whites to discrimination among residual White workers.

Line 77 of the Peyton and Huber code tests whether, in a model including both interaction terms for the "Table S1.1, main analysis" section, the estimated discrimination gap differed between the prejudice categories and the racial resentment categories. The p-value was p=0.0798 for that test.

3. Data. Stata code for my analysis. Stata output for my analysis.