**Summary:** This paper investigates the concept of statistical discrimination in the labor market, focusing on how employers learn over time and how this affects wage determination. The authors propose a model where firms initially rely on easily observable characteristics, such as education and race, to estimate worker productivity. As firms gain more experience, they become better informed about actual productivity, leading to a shift in wage determination from observable traits to actual productivity. The authors use data from the National Longitudinal Survey of Youth (NLSY) to test their hypothesis. They find that the wage coefficient on education decreases over time, while the coefficient on unobservable productivity variables increases. This supports the idea of statistical discrimination, where firms use observable characteristics to make decisions about workers. However, the results are inconsistent with the hypothesis that firms fully utilize information on race. The paper also explores the implications of statistical discrimination on the basis of race. It shows that if race is correlated with productivity, the wage differential between racial groups may widen as experience accumulates. However, the results do not support the idea that firms fully utilize race information. The authors also consider the role of on-the-job training in the model. They find that the presence of training effects can complicate the interpretation of wage trends, as training can influence both the productivity of workers and the wage response to observable characteristics. Overall, the paper provides a framework for understanding how employer learning and statistical discrimination affect wage determination in the labor market. The results have implications for the analysis of wage growth, productivity, and statistical discrimination in other contexts.**Summary:** This paper investigates the concept of statistical discrimination in the labor market, focusing on how employers learn over time and how this affects wage determination. The authors propose a model where firms initially rely on easily observable characteristics, such as education and race, to estimate worker productivity. As firms gain more experience, they become better informed about actual productivity, leading to a shift in wage determination from observable traits to actual productivity. The authors use data from the National Longitudinal Survey of Youth (NLSY) to test their hypothesis. They find that the wage coefficient on education decreases over time, while the coefficient on unobservable productivity variables increases. This supports the idea of statistical discrimination, where firms use observable characteristics to make decisions about workers. However, the results are inconsistent with the hypothesis that firms fully utilize information on race. The paper also explores the implications of statistical discrimination on the basis of race. It shows that if race is correlated with productivity, the wage differential between racial groups may widen as experience accumulates. However, the results do not support the idea that firms fully utilize race information. The authors also consider the role of on-the-job training in the model. They find that the presence of training effects can complicate the interpretation of wage trends, as training can influence both the productivity of workers and the wage response to observable characteristics. Overall, the paper provides a framework for understanding how employer learning and statistical discrimination affect wage determination in the labor market. The results have implications for the analysis of wage growth, productivity, and statistical discrimination in other contexts.