This paper by Joseph G. Altonji and Charles R. Pierret explores the hypothesis of statistical discrimination and employer learning in the labor market. The authors propose a test for statistical discrimination, which occurs when firms use easily observable characteristics, such as education or race, to make hiring decisions based on statistical regularities rather than actual productivity. They argue that as firms gain more information about a worker's productivity over time, wages should become more dependent on actual productivity and less dependent on these observable characteristics. The paper uses a wage equation that includes interactions between experience and both observable and unobservable variables to estimate the time-profile of these coefficients. The authors find that the coefficient on the unobservable productivity variable should rise with time, while the coefficient on education should fall. They use data from the National Longitudinal Study of Youth (NLSY) to investigate these propositions, focusing on education and race as examples of observable characteristics. The results support the hypothesis of statistical discrimination but are inconsistent with the idea that firms fully utilize information on race. The authors also discuss the implications of their findings for wage growth, productivity, and statistical discrimination in other contexts. The paper concludes by highlighting the importance of understanding how employers learn about worker productivity and how this affects wage dynamics and labor market outcomes.This paper by Joseph G. Altonji and Charles R. Pierret explores the hypothesis of statistical discrimination and employer learning in the labor market. The authors propose a test for statistical discrimination, which occurs when firms use easily observable characteristics, such as education or race, to make hiring decisions based on statistical regularities rather than actual productivity. They argue that as firms gain more information about a worker's productivity over time, wages should become more dependent on actual productivity and less dependent on these observable characteristics. The paper uses a wage equation that includes interactions between experience and both observable and unobservable variables to estimate the time-profile of these coefficients. The authors find that the coefficient on the unobservable productivity variable should rise with time, while the coefficient on education should fall. They use data from the National Longitudinal Study of Youth (NLSY) to investigate these propositions, focusing on education and race as examples of observable characteristics. The results support the hypothesis of statistical discrimination but are inconsistent with the idea that firms fully utilize information on race. The authors also discuss the implications of their findings for wage growth, productivity, and statistical discrimination in other contexts. The paper concludes by highlighting the importance of understanding how employers learn about worker productivity and how this affects wage dynamics and labor market outcomes.