Corrigendum to “How partisanship and sexism influence voters’ reactions to political #MeToo scandals”

Costa M, Briggs T, Chahal A, et al. (2020). How partisanship and sexism influence voters’ reactions to political #MeToo Scandals. Research and Politics 7(3): 1–8. doi: 10.1177/2053168020941727

In the above referenced article, the below corrections have been made:

An error in the recoding of the sexism variables resulted in the sexism scale being comprised of two items. It was reported in the article that respondents’ levels of hostile sexism was measured using their mean agreement on four statements from the hostile sexism battery. Due to a typo in the code (using `pre_sexism_2` instead of `pre_sexism_3`), the authors created the scale using the second statement twice (once reverse-coded so it effectively cancels itself out) and the first and fourth statement, without the third statement. The authors are grateful to L.J. Zigerell for notifying them of this error.

Using the full, corrected scale affects the reported reliability of the four items (alpha = .80 instead of 0.95 for pre-test sexism and 0.81 instead of 0.96 for post-test sexism) and standard deviation of the combined scale (1.0 instead of 1.2 for both pre-test and post-test) and slightly adjusts the findings shown in Figure 3. The correct Figure is reproduced below. The authors note that, as reported in the original article, this was not a pre-registered analysis and thus should be considered exploratory and the substantive size of the effects are very modest. The authors apologize for any inconvenience or misinterpretation this error may have caused. The authors have updated the replication code in the Harvard Dataverse to reflect this correction.

Figure 3.

Treatment effects on percent change in sexism from pretest to posttest.

Note: Figure shows the percent change in sexism from the pretest to the posttest by experimental condition compared to the control group. Circles represent treatment effect for subjects who share the legislator’s party ID; triangles represent treatment effect for subjects that do not share his party ID. Independents/Not sures excluded from analysis. Horizontal lines represent 95% confidence intervals.