Examining the gender and minority test score gap on the MFT-B: A Blinder-Oaxaca decomposition approach

Abstract

This paper examines performance differences among demographic groups on the ETS Major Field Test in Business. The study employs the Blinder-Oaxaca decomposition technique to analyze the test score differentials by gender and racial minority status. This technique decomposes the difference into two parts: an endowment effect (or explained portion) and a returns effect (or unexplained portion). The results demonstrate that the endowment effect fully explains the gap between white and racial minority students but virtually none of the gender gap. This large unexplained gap between male and female test performance suggests the need for further study of potential gender bias of the exam.

Keywords:

• Blinder-Oaxaca decomposition
• gender bias
• Major Field Test in Business
• MFT-B
• racial bias
• standardized test performance
• test score gap

Notes

1 Throughout this paper, minority refers to racial minority, i.e., that the student identifies as a race other than white.

2 We use county level social capital information from roughly the middle of our sample period (2014). This dataset is available at: https://aese.psu.edu/nercrd/community/social-capital-resources and its construction is described more fully in Rupasingha, A, Goetz, S.J., and Freshwaterm D. (2006).

3 The social capital index is formed by standardizing the cross-county distributions of each of the four items to have zero means and unit standard deviations, and then summing the four standardized items within each county. The values of this standardized index for every student are based on their hometown county using the 2014 index value, the closest to our sample period.

4 In the interest of simplicity, we conduct a simple twofold decomposition. Jann (2008) discusses a three-way decomposition that accounts for cross-group differences in explanatory variables and coefficients that can occur at the same time. Our twofold results are qualitatively similar, so in the interest of simplicity we present two-way decomposition results.

5 Ketcham et al. (2018) find that student “passion” impacts test score.

6 We ran a two-way ANOVA test to examine the impact of gender, minority status and their interaction on the test scores. The F-test shows that both gender and minority status are significant; the interaction is not significant.

7 As a robustness check, we gathered data on first-generation college students, who scored 1.0 points lower on the exam. The difference was not statistically significant. When adding a first-generation indicator variable to the model, the results did not change (the indicator variable is insignificant). We included a control variable of the average income in the student’s hometown zip code, and that variable was not significant.

8 We conduct a decomposition to compute the individual contributions of the predictors to the components of the decomposition. This detailed decomposition finds that SAT- Math score explain .923 of the differential, while SAT verbal scores explain 1.53. Thus, SAT score alone explain (.923 + 1.54/3.87) 63% of the test score differentials.