Research

Abstract: We examine the effects of a criminal record on labor market outcomes and the mediating role of job sorting using Swedish register data. Prime-age adults with criminal records earn about 30% less than observably similar adults without records and are concentrated in specific employers and occupations. To estimate the causal effect of a criminal record, we use an event study design that compares outcomes for adults charged with an offense for the first time to matched adults who were suspected of a similar offense but not charged. Acquiring a criminal record reduces months employed by 2% and annual earnings by 5%. These negative effects are: twice as large for more serious or subsequent charges, not driven by job displacement or incapacitation, and not mitigated by automatic record expungement, which typically occurs 5 or 10 years after case disposition. We classify firms by their propensity to hire workers with criminal records, holding suspected offense history fixed. A criminal record reduces employment at firms classified as less likely to hire workers with criminal records, increases employment at other firms, and decreases monthly wages across all firm types. Firm propensity to hire workers with criminal records varies substantially---even within industries---and is linked to firm size and managers' prior exposure to people with records. Leveraging manager moves across small firms, we find that when a firm hires a new manager with greater prior exposure to people with criminal records, it hires more people with records, with no detectable effect on productivity.

Abstract: We examine class disparities and discrimination in police searches and stops using data on traffic stops conducted by Texas Highway Patrol. Low-income motorists are more likely to be searched for contraband, less likely to be found with contraband when searched, and more likely to be stopped for infractions associated with pretext stops. We measure class-based discrimination in searches per potential stop, accounting for both the search and stop margins. Our research design leverages motorists stopped in multiple vehicles conveying different class signals. Motorists are more likely to be searched when stopped in a low-status vehicle, and evidence suggests that they are also more likely to be stopped when driving one. Marginal searches triggered by vehicle status are also less likely to yield contraband when the motorist is low-income. We argue that lower hassle costs associated with arrests of low-income motorists help explain trooper behavior. 

Abstract: In Brazil, firms' later hires are more likely to be nonwhite than early hires for the same job. We argue that this pattern reflects racial disparities in entrepreneurship and co-racial hiring: firms are more likely to hire from groups already well-represented at the firm, though with some decay. At entry, firms with white founders are about 30% less likely to hire nonwhite employees than comparable firms with nonwhite founders. After 400 hires, these firms nearly converge in their composition of subsequent hires. Yet few firms reach this scale. Within-firm racial differences in dismissal rates follow an analogous pattern. We provide suggestive evidence that referral hiring can at least in part account for our findings.

Abstract: A growing body of evidence shows that differences in firm-specific pay premiums account for a large share of the gender pay gap. This paper asks how a common form of pre-labor market skill specialization, college major, mediates access to high-paying firms, and what this means for the gender earnings gap. Using employer-employee tax data from Chile matched to educational records, we show that differences in college major account for more than two-thirds of the firm contribution to the gender earnings gap among college admits. Degrees in Technology, which are numerous, male-dominated, and associated with high firm premiums, drive these effects. 

Abstract: This paper examines whether job suburbanization caused declines in black employment rates from 1970 to 2000. I find that black workers are less likely than white workers to work in observably similar jobs that are located further from the central city. Using evidence from establishment relocations, I find that this relationship at least in part reflects the causal effect of job location. At the local labor market level, I find that job suburbanization is associated with substantial declines in black employment rates relative to white employment rates. Evidence from nationally planned highway infrastructure corroborates a causal interpretation.

Abstract: During traffic stops, police search black and Hispanic motorists more than twice as often as white motorists, yet those searches are no more likely to yield contraband. We ask whether equalizing search rates by motorist race would reduce contraband yield. We use unique administrative data from Texas to isolate variation in search behavior across and within highway patrol troopers and find that search rates are unrelated to the proportion of searches that yield contraband. We find that troopers can equalize search rates across racial groups, maintain the status quo search rate, and increase contraband yield. Troopers appear to be limited in their ability to discern between motorists that are more or less likely to carry contraband.

Abstract: In settings where social norms promote gender segregation, firms may find it costly to employ both men and women. These integration costs may hinder women's employment. We develop a methodology to test for the presence of fixed integration costs and estimate counterfactual women's employment at all-male firms where these costs bind. We apply our approach in Saudi Arabia and find that integration costs bind for the majority of firms.  We show that Nitaqat, a gender-neutral quota program that incentivized the hiring of Saudi nationals at private sector firms, induced firms to integrate and dramatically increased Saudi women's employment.

Abstract: The US criminal justice system is exceptionally punitive. We test whether racial heterogeneity is one cause, exploiting cross-jurisdiction variation in punishment severity in four Southern states. We estimate the causal effect of jurisdiction on arrest outcome using a fixed effects model that incorporates extensive charge and defendant controls. We validate our estimates using defendants charged in multiple jurisdictions. Consistent with a model of ingroup bias in electorate preferences, the relationship between local severity and black population share follows an inverted U-shape. Within states, defendants are 27%-54% more likely to be incarcerated in `peak' heterogeneous jurisdictions than in homogeneous jurisdictions. We estimate that confinement rates and race-based confinement rate gaps would fall by 15% if all jurisdictions adopted the severity of homogeneous jurisdictions within their state.

Abstract: I estimate the dynamic effects of federal affirmative action regulation, exploiting variation in the timing of regulation and deregulation across work establishments. Affirmative action increases the black share of employees over time: in 5 years after an establishment is first regulated, the black share of employees increases by an average of 0.8 percentage points. Strikingly, the black share continues to grow at a similar pace even after an establishment is deregulated. I argue that this persistence is driven in part by affirmative action inducing employers to improve their methods for screening potential hires.