Corrigendum

Abstract

Craig, M. A., & Richeson, J. A. (2014). More diverse yet less tolerant? How the increasingly diverse racial landscape affects White Americans’ racial attitudes. Personality and Social Psychology Bulletin, 40(6), 750–761. doi:10.1177/0146167214524993

It was brought to the authors’ attention that an issue with the base SPSS package leads the “weight by” command to utilize probability weights improperly (treating them as frequency weights; for a more comprehensive description of this issue, please see http://www.ats.ucla.edu/stat/stata/faq/weights.htm).¹ Mis-specifying the weighting variable as frequency weights instead of as probability weights resulted in underestimated standard errors and p values, both of which were reported in Study 3 of the above article. The authors report the revised and accurate results of this study below.

In the interest of full transparency, the authors report both the unweighted results and the results if weighted with a statistical software package designed for complex survey data that correctly specifies probability weights (Stata v13.1). Included is a corrected Table 1; Table 2 from the original article should be disregarded. The authors also report a supplemental study (see article’s online Supplemental Materials) that replicates the basic effect of the racial shift manipulation on attitudes toward Blacks and Latinos.

Corrected Table 1.

Study 3: Descriptive Statistics of the Potential Mediators by Experimental Condition.

Potential mediator	Control condition	Racial shift condition
Potential mediator	M (SD)	M (SD)
Perceived status threat	4.83 (1.37)	5.29 (1.37)***
Perceived uncertainty	4.25 (1.33)	4.31 (1.39)
Racial identification	3.37 (1.52)	3.62 (1.65)^†
System threat	−0.03 (0.85)	0.03 (0.96)
System justification	3.33 (1.41)	3.38 (1.41)

†

p < .10. ***p < .001.

Study 3 Corrected Results

Unweighted Results

These analyses were conducted with SPSS by selecting only White participants from the sample. The weighting variable was not utilized.

Racial attitudes

Separate feeling thermometer ratings

To test the effect of the racial shift information on racial attitudes, the authors conducted a 4 (target group: White, Black, Latino, Asian American) × 2 (experimental condition: racial shift, control) mixed-design ANCOVA with target group as the within-subjects factor and experimental condition as the between-subjects factor, controlling for participant age, gender, and education level. Results revealed a main effect of target group, F(3, 1,323) = 54.66, p < .001, $η_{p}^{2}$ = .11; a marginal main effect of experimental condition, F(1, 441) = 3.46, p = .064, $η_{p}^{2}$ = .01; and a marginal interaction, F(3, 1,323) = 2.19, p = .088, $η_{p}^{2}$ = .01. Examination of the effects of the racial shift manipulation on attitudes toward each racial group separately revealed that participants in the racial shift condition expressed more negative attitudes toward Asian Americans, F(1, 449) = 6.45, p = .011, $η_{p}^{2}$ = .01, compared with participants in the control condition. Attitudes toward Blacks, F(1, 450) = 2.28, p = .132, $η_{p}^{2}$ = .01; Hispanics, F(1, 450) = 2.16, p = .143, $η_{p}^{2}$ = .01; and Whites, F(1, 452) < 1, p = .822, did not differ reliably by condition.

Racial minority composite rating

The authors averaged the ratings of the three racial minority groups (i.e., Black Americans, Latinos, Asian Americans) to create a feelings toward racial minority groups index (α = .87). A 2 (target group: White, racial minority) × 2 (experimental condition: racial shift, control) mixed-design ANCOVA—controlling for participant age, gender, and education level—revealed a main effect of target group, F(1, 451) = 98.97, p < .001, $η_{p}^{2}$ = .18, moderated by a marginal interaction, F(1, 451) = 3.48, p = .063, $η_{p}^{2}$ = .01. Participants in the racial shift condition expressed more negative feelings toward racial minority groups, compared with participants in the control condition, F(1, 455) = 4.31, p = .038, $η_{p}^{2}$ = .01.

Mediation testing

Group status threat was the only potential mediator to significantly differ by experimental condition (controlling for participant age, gender, and education level), F(1, 461) = 15.74, p < .001, $η_{p}^{2}$ = .03. Participants in the racial shift condition expressed more agreement that racial minorities would reduce White Americans’ societal status, compared with control condition participants (see corrected Table 1). Furthermore, using the PROCESS macro (Hayes, 2013), the authors conducted analyses to examine whether perceived loss of status (i.e., group status threat) mediated the effects of experimental condition (coded 1 = racial shift condition, 0 = control condition) on racial attitudes. The authors calculated the indirect effect of the experimental condition on the feeling thermometer ratings regarding the racial minority groups through group status threat (with 5,000 bootstrap samples). In addition to the feeling thermometer ratings, the models included the effect of the experimental condition, the proposed mediator—group status threat—and covariates of participant age, gender, and education level.

Separate feeling thermometer ratings

Group status threat served as a mediator of the non-significant effects of the U.S. racial shift information on feelings toward Black Americans, 95% confidence interval (CI) = [−2.97, −0.64], and Latinos, 95% CI = [−2.75, −0.51], but was not a mediator for the significant effect of condition on feelings toward Asian Americans, 95% CI = [−1.75, 0.03].

Racial minority composite rating

Group status threat served as a mediator of the effect of the U.S. racial shift information on the feelings toward racial minority groups composite, 95% CI = [−2.29, −0.40].

Weighted Results

The authors utilized Stata v13.1, treating the weighting variable as a probability weight and conducting subpopulation analyses including Whites as the subpopulation of interest.

Racial attitudes

To test the effect of the racial shift information on racial attitudes, the authors conducted a series of regressions, regressing each feeling thermometer rating on racial shift condition and participants’ age, gender, and education level. Participants in the racial shift condition expressed more negative attitudes toward Asian Americans, b = −5.09, p = .028, r_partial = −.13, compared with control condition participants. Attitudes toward Blacks, b = −3.21, p = .182; Hispanics, b = −3.57, p = .144; and Whites, b = −1.67, p = .469, did not differ reliably by condition. Regressing the racial minority group composite on racial shift condition and participants’ age, gender, and education level revealed a marginal effect, b = −3.87, p = .068, r_partial = −.11.

Mediators

The only potential mediator that significantly differed by experimental condition (controlling for participant age, gender, and education level) was perceived group status threat, b = 0.47, p = .001, r_partial = .17. The PROCESS macro is not available for this statistical program.

Footnotes

Notes

References

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York, NY: The Guilford Press.