Essays on charter student mobility and machine learning applications in empirical microeconomics
OA Version
Citation
Abstract
This dissertation consists of three chapters in empirical microeconomics. The first two chapters apply machine learning tools to labor market discrimination in professional basketball and demographic sorting in American politics. The third chapter examines the effect of charter school attendance on subsequent student mobility.
In the first chapter, I analyze whether conformity to racial stereotypes is rewarded or penalized in the labor market for first-year players in the National Basketball Association. Given anecdotal evidence of racial stereotypes about the playing style and characteristics of white and black basketball players, I test the hypothesis that these stereotypes, and specifically whether a player conforms to the relevant stereotypes for his race, may have an effect on the outcome of the NBA draft, where incoming players are assigned to NBA teams. I apply a random forest model to analyze players' relative conformity to racial stereotypes and a conditional logit model to test whether there are observable effects on team assignments. I find that players who conform to their racial stereotype are positively rewarded, with much of the effect concentrated in more recent draft cohorts and along the extensive margin.
In the second chapter (joint with James Feigenbaum), we use a machine approach to test the hypothesis that post-1950s American political polarization has taken the form of increasing sorting along demographic cleavages. We find that while the predictive power of Americans' demographic profiles for the political camps was relatively stable from 1952 through the mid-2000s, it has subsequently intensified. We also use our model to confirm prior findings that racial resentment has become an increasingly important predictor of political sorting.
In the third chapter (joint with Marcus Winters), we study the causal effect of charter school attendance on subsequent student mobility under a common enrollment system for charter and traditional public schools in Newark, New Jersey. Exploiting information from the deferred acceptance assignment mechanism and applying a novel correction for the proportion of mobility attributable to differences in assigned students' preferences, we find that charter school attendance leads to significantly less student mobility, with no evidence of disproportionate mobility within historically disadvantaged categories.