Essays on Microeconometric Analysis

Abstract

This dissertation comprises three essays in micro econometrics that explore issues of social justice and individual decision-making.In the first chapter, I examine how judges’ decisions affect defendants' reoffending behavior. I consider the issue of selection when a judge decides on sentencing and probation and builds a novel method using many instruments to identify the compound effects of the verdict (including sentence and probation) on recidivism. In the second, more theoretical chapter, I propose a novel method for evaluating treatment effects when both the instrument and controls have high dimensionalities. In the third, more empirical chapter, I assess the moral hazard issue in HIV preventative medication. In the first chapter, I consider an important social issue - the problem of reoffending and its determinants. I address the question, ``How do judges make a series of decisions, and how does this affect the future criminal behavior of defendants?`` using judge randomization with a selection model. In my model, the judge makes numerous decisions, such as whether or not to impose probation, the duration of probation, and the (potential) length of the sentence. To address the selection issue inherent in these decisions, a novel methodology is developed that use multiple instruments to identify the compound effects of the verdict on reoffending behavior. Using data on new offenders in North Carolina from 1995 to 2010, the analysis reveals that neglecting the selection process for probation can introduce severe estimation biases. The findings suggest that longer probation periods lead to a marginal increase in recidivism, while both active and suspended sentences exhibit a marginal deterrent effect. This chapter contributes to the literature by providing a comprehensive understanding of the criminal population and identifying the effectiveness of judicial decisions.

The second chapter proposes a new method for estimating treatment effects when both the instrument and control variables have high dimensionalities. This approach is particularly relevant in many empirical settings where exogenous shocks interact with covariates to capture heterogeneity. Existing solutions, such as Lasso, only address high dimensionality in either instruments or controls and rely on sparsity assumptions that may not be compatible with various econometric applications. The proposed technique can simultaneously handle both high-dimensional controls and instruments, making it especially useful when researchers are uncertain about asymptotic constraints. The inference protocol developed in this chapter yields more robust results compared to alternative methods.

The third chapter investigates the potential moral hazard associated with HIV pre-exposure prophylaxis (PrEP) medication among high-risk individuals. While PrEP is highly effective in preventing HIV infection, concerns have been raised regarding its potential to encourage riskier sexual behaviors. Using panel data from the Multicenter AIDS Cohort Study and employing a difference-in-differences approach, the analysis examines the impact of PrEP on hazardous sexual behaviors and the incidence of other sexually transmitted diseases (STDs). The findings indicate that PrEP increases the likelihood of engaging in unprotected sexual activity, having a higher number of sexual partners, and contracting syphilis.