In this section, we specify an education production
function model linking high school outcomes to whether a
student attends a STEM or Non-STEM high school. We are
particularly interested in the contrast between raw,
unconditional differences and those conditioning on
student characteristics and prior performance. We then
modify this model to allow for heterogeneous STEM effects
by gender and illustrate how coefficients in this model help
us answer questions about females’ performance in STEM
schools. Finally, we generalize our analysis to a model that
allows for heterogeneous effects by race. We conclude this
section with a discussion of possible remaining selection
issues and our efforts to address them