"Do Grade Signals Drive the Gender Gap in STEM? Evidence From a Regression Discontinuity" (job market paper!)
A common hypothesis for the gender gap in STEM, which is believed to drive much of the gender gap in earnings, is that women are more responsive to the negative grade signals that are common in STEM. I test this hypothesis by applying a regression discontinuity design to the underlying numerical scores of more than 21,000 university students in feeder courses in STEM and economics. Results indicate that letter grade thresholds (with no plus/minus modifiers in this context) have no effect on STEM major choice for either women or men. This is true even for susceptible subgroups, within particular majors, or at particular grade thresholds.
Racially charged rhetoric often surrounds layoff events, with specific minorities often blamed for the loss of "American jobs.'' We examine whether information about impending mass layoffs causes racial animus. Our data consist of information on mass layoff notices linked to Google Search Trends and FBI Hate Crime Statistics. We compare outcomes across areas that vary in the timing of news of impending layoffs. Results indicate an increase in both racist internet searches (1.5 percent) and hate crimes (23 percent) following layoff notices.
Media ownership has become more concentrated in recent years, leading to concerns over media integrity and the nature of the information being passed on to the public. In this paper, we study the impact of broadcast television ownership consolidation on ideological preferences. To do so, we use a difference-in-differences design to examine the impact of within-market consolidation on election outcomes. Results show that within-market consolidation shifted vote share towards Democrats by 3-4 percentage points for both presidential and senate elections, and that this effect persists for at least 12 years.
Debate continues over the advantages and drawbacks of voter ID laws. In this paper, I consider the impact of one of the strictest such laws, enacted just prior to the 2014 election in Texas. To identify effects, I use a difference-in-differences approach comparing turnout for people with and without permissible photo IDs over time. I do so using individual-level data matching voter history and state-issued IDs, accounting for ID issue date. Results show that registered voters who had not been issued a photo ID were no more or less likely to vote in the 2014 midterm election. In the 2016 presidential election, voters lacking ID were 7.6 percentage points less likely to turn out. I find no evidence of differential effects for minority groups.
Works in progress
"Racial Bias in Police Investigations" with Abigail Peralta
"Exposure to Racial Diversity and Judge Behavior" with CarlyWill Sloan