查看文章

After this brief was published in August, we noticed an error in the statistical code used to create the concept scales. This error underestimated scale values for respondents who did not answer all survey questions for a concept, such as perceptions of police legitimacy. It was especially problematic when a respondent did not answer any questions for a concept: the summated value and the scale value were zero, which is outside the possible range of response options. To correct this problem, we have recalculated all the scales. We created a scale for a respondent only when they answered more than half the items about that concept. Then, we used the average of the valid responses rather than the average of the summated responses. These steps better ensure the scale for each respondent is measuring the intended concept. These changes primarily affected the data tables and the text summarizing the tables …