Motivation: It is widely documented that women and minorities are underrepresented in certain science, technology, engineering, and mathematics (STEM) fields. Potential reasons include overt discrimination, a lack of mentors, and teaching practices. Improving STEM opportunities for women and minorities is vital because if they are misallocated away from STEM, society loses valuable and diverse contributions, while they end up in suboptimal careers. I propose to study how grading policies in STEM classes discourage women and minorities from further study and identify potential solutions. STEM classes tend to have difficult grading standards, and low grades in introductory courses are especially likely to discourage women and minorities from further …show more content…
I will restrict each regression sample to the set of students in an introductory course that is a prerequisite for major $i$, who had the same ``adjusted grade,'' but whose actual grades fell immediately above or below a given grade threshold $j$. The first stage will regress a dummy variable indicating whether the student graduated with a major in field $i$ on a dummy for having an actual grade above the subgroup threshold $j$ (e.g., receiving a B- versus a C+ among the set of students with an ``adjusted grade'' of C+) and generate fitted values. The second stage will regress future earnings or alternative measures of economic outcomes from SSA administrative data on the fitted dummies for majoring in $i$. This parameter will capture the causal earnings premium for a particular major for students at a certain grade threshold relative to other students at the same threshold who selected into other majors. I will estimate this model separately for as many grade thresholds, majors, and demographic groups as my data allow in order to account for multiple dimensions of