University Prestige, Performance Evaluation, and Promotion: Estimating the employer learning model using personnel datasets

The employer learning model postulates that employers form employees' prior ability distribution from educational credentials and update its distribution by observing workers' performance on the job. This paper estimates the employer learning model for university-graduate white-collar workers using personnel datasets from two large manufacturers that contain rich information, including the name of the university from which the worker graduated, annual performance evaluations, and position in the promotion ladder. The estimates indicate that employers learn workers' ability relatively quickly through observing their performance on the job. The initial expectation errors on ability decline by a half in about three to four years in the two companies. Companies promote graduates of elite schools quickly mainly because they tend to perform better on the job.