Essays on Diversity

Abstract

This dissertation consists of three chapters. All papers focus on diversity and use a mix of econometrics and machine learning techniques. My first chapter focuses on police discretion. Using a novel dataset from Washington D.C. that contains extensive administrative data, we estimate the degree to which individual police officers vary in their responses to calls for service. We find that officer identity accounts for approximately 10% of the explainable variation in stop and arrest outcomes and that officers vary widely in their propensity to make a stop or arrest. We find that officer level propensity trends are persistent over time and that a one standard deviation increase in officer propensity to make a stop and arrest increases odds of stops and arrests by approximately 20-25% in the six months following the sample period. These results are robust to a variety of specifications. Rather than officer race or gender, we find that officer rank and supervisor are the most predictive features of officer propensity to arrest or make a stop. This finding is consistent across various specifications and machine learning methods. My second chapter focuses on machine learning and studies of racial disparities. It is well documented that individuals of different races have disparate experiences in the criminal justice system in the United States. As most studies focus on Black-white interracial differences, intraracial disparities are often overlooked. Analyzing intraracial disparities is further complicated by the fact that it has been historically difficult to accurately and consistently measure skin tone and Afrocentric features. Utilizing convolutional neural networks and photos as data, this study creates a consistent, running measure of perceived race. Using this new measure and data type, new types of analysis are possible. This study presents photographic summary statistics, shows the positive relationship between perceived race and sentence length is robust to the inclusion of various controls and, reevaluates traditional Black-white gaps, measuring both inter- and intraracial disparities. This study also evaluates how new advances in race inference methodology perform throughout the perceived race spectrum and the implications of misclassification for metrics of racial disparities. My third chapter focuses on ideological diversity in classroom settings. Surprisingly, little is known about the impact of ideological diversity on classroom outcomes given the amount of attention paid to the role of ideological diversity on higher education outcomes such as critical thinking and academic performance, scant causal evidence exists. We use a lab-in-the-field experiment to test whether the presence of ideologically more conservative students in academic discussion groups, as compared to groups of students who all slanted ideologically liberal, would improve academic outcomes in terms of the quality of each student’s individual academic work. The complete population of an incoming cohort of policy graduate students (N = 78) took part in the experiment. Results demonstrate that students assigned to the ideologically heterogeneous discussion groups subsequently wrote individual assignments that received significantly more negative grades by a professional grader blind to experimental condition and to student identity. Survey results from participating students also suggest that students in the ideologically heterogeneous discussion groups were also significantly more likely to perceive interpersonal conflict and to dislike their group dynamics—a result that was not driven by students of a particular ideological slant. As a small pilot, this study provides questions to resolve with future research, including the role of pedagogy in managing ideological diversity, and provides a template for future experimental designs.