Employer Learning, Biased Beliefs, and Labor Market Discrimination

My dissertation proposes and tests a new theory of labor market discrimination based on employers developing persistent negatively biased beliefs about the productivity of worker groups through their hiring experiences with these groups. My first chapter presents a statistical discrimination model in which employers are initially uncertain about the productivity of groups and endogenously learn about it through their hiring experiences. An employer's hiring history determines their beliefs about group productivity, but also shapes their subsequent incentives to hire from the group and learn more about their productivity. Positive experiences create positive biases which correct themselves by leading to more hiring and learning. Negative experiences create negative biases which decrease hiring and therefore learning, leading to the persistence of negative biases. Differential hiring and learning across employers thus generates a negatively-skewed belief distribution about worker group productivity. Endogenous employer learning disproportionately affects workers from minority or underrepresented groups if there is less initial information available about their productivity in the labor market, making employers more reliant on their own experiences to assess these groups. I show that discrimination in the form of a wage below these groups' expected productivity can arise and persist from this initial information asymmetry even with market competition and without true productivity differences between groups, prior bias, or prejudice. The model generates analogous predictions to taste-based discrimination, in a statistical framework with beliefs replacing preferences, providing a new way to understand prejudice as the result of "incorrect" statistical discrimination. The model helps explain the persistence and pervasiveness of discrimination, also generating new implications for policies like affirmative action which can induce employers to hire from specific groups and learn about their productivity. My second chapter tests how hiring experiences of employers with worker groups impact hiring and beliefs about group productivity. I design an experiment where employers hire a worker from one of two groups each period, with one group framed as a minority about whose productivity employers are initially given less information. Employers are incentivized to hire productive workers, observe their hire's productivity after they perform a real-effort task, and then report their beliefs about group productivity. The results show that negative experiences with the minority group, captured through the hiring of low productivity workers, lead to negatively-biased beliefs about the group's productivity by decreasing subsequent hiring and learning. In contrast, positive biases which arise from positive experiences are mitigated through increased hiring, leading to a negatively-skewed belief distribution across employers. My third chapter joint with Alan Benson at the University of Minnesota uses employment records of a large retail firm to study how hiring experiences of managers with worker groups influence their hiring. We study the hiring of black and white workers, relating current hiring decisions of a manager to measures of their previous experiences with these groups. We find that negative experiences with previous hires of a group, measured by a higher fraction quickly being fired or quitting, decrease subsequent hiring of the group. More positive experiences, measured by a higher fraction of previous hires achieving long tenure, increase subsequent hiring of the group. These impacts are substantively larger for black workers, and early negative experiences with them lead to particularly persistent decreases in relative hiring of the group.