Essays in Crime, Personnel, and Health Economics

This dissertation consists of three chapters. In Chapter 1, I estimate the causal effect of drug testing on drug crime. While many researchers have tied liquidity to drug crime, this paper presents evidence of a new mechanism that also propagates drug use: employer-sponsored drug testing. Using the United States Army as a natural experiment, I exploit the conditional random testing of soldiers to estimate the effects of drug testing on drug crime. Controlling for liquidity and individual soldier characteristics, I find that soldiers respond sharply, and drastically to their own individual drug tests. The probability of a soldier's arrest for a drug charge, not caught by testing, within a week they are tested increases 15.7% over the outcome sample mean. My findings suggest that soldiers structure their drug use around the monthly testing cycle within the Army and that the Army's current drug testing policy displaces, and minimally deters soldier drug use.

In Chapter 2, I estimate the causal effect of personality on hiring preference. I exploit the exogenous assignment of 1,038 applicants to 46 different hiring boards to assess whether there is a "similar-to-me'' preference for applicant personality and other observable characteristics. I match data from a unique assessment within the United States Army to administrative records to control for applicant quality, previous interactions with board members, reputation effects, and similar experiences of both hiring board members and applicants. I find that an applicant being similar in personality to a hiring board member in certain traits significantly decreases their probability of earning a hiring vote by 2.5%; however, a hiring board member having prior knowledge of an applicant counteracts these negative effects. These findings help convey which information is most salient among hiring managers.

In Chapter 3, I estimate the effect of patient panel size on provider practice style in a primary care setting. I address panel size endogeneity using unique features of the US Army healthcare system and exogenous variation in provider absences. For every one provider absent, patient panel size increases by 10.8%. Leveraging this variation through an instrumental variables framework, I estimate that an increase of 100 patients to a provider's panel causes a 4.6% increase in daily appointments, a 3.1% increase in referral rate, increases to 30,60, and 90-day readmission rates, and a decrease in treatment intensity by 7.1%. These estimates are all statistically significant and suggest that among larger patient panels, providers treat patients less intensely which results in higher uses of healthcare. I find no statistically significant evidence to suggest that providers use resources more efficiently or that the quality of care is reduced, and provide evidence that neither own nor peer provider quality mediate these effects.