Social Identity and Achievement Gaps: Evidence From an Affirmation Intervention

Abstract

One provocative explanation for the continued persistence of minority achievement gaps involves the performance-dampening anxiety thought to be experienced by minority students in highly evaluative settings (i.e., “stereotype threat”). Recent field-experimental studies suggest that modest, low-cost “buffering” interventions informed by this phenomenon may be highly effective at reducing minority achievement gaps. This field-experimental study evaluates such an intervention in which students complete a self-directed “self affirmation” exercise that encourages them to identify and reflect upon their core personal values. This within-classroom randomized trial was conducted among 2,500 7th and 8th graders from six Philadelphia-area middle schools during the 2008–09 and 2009–10 academic years. Although this study failed to replicate the earlier findings indicating that the affirmation generated large increases in the academic performance of minority students, this treatment did lead to statistically significant improvements in the performance of the minority students in more supportive classroom environments. However, the treatment contrast also reduced the performance of female students in those settings.

Keywords:

• Stereotype threat
• achievement gaps
• field intervention

Notes

Experiment 2 slightly simplified the affirmation, which is described in more detail below. The standard deviation in this course-grade measure was roughly 1.0, so the effect sizes were 0.26 to 0.34.

As described next, the randomization procedure blocked on race and ethnicity within classrooms so this sample construction does not influence random assignment to condition.

This script was also used in the study by Cohen et al. (2006) and was generously provided by Geoffrey Cohen.

I followed the materials used in Experiment 2, which generated larger treatment effects. The first intervention excluded the values “Being Smart or Getting Good Grades,” but these values were included in the intervention fielded later in the year.

The limited noncompliance with the experimental assignment occurred because students switched classrooms after their baseline assignment or because they were absent and did not complete the intervention worksheet on their return.

To improve the efficiency of these classroom-level regressions, I use the inverse of the standard error for each treatment estimate as a weight. There are fewer than 139 observations in this analysis because some classrooms did not have within-classroom variation with regard to ethnicity, gender, and treatment status.

The treatment estimates for African American students in this study were sufficiently precise to reject effects half as large as those reported by Cohen et al. (2006).

However, it should be noted that this study generates similar null findings using the one school-year observation that also used active consent.