Two-Sided Learning with Applications to Labor Turnover and Worker Displacement

We make several extensions to the recent literature on job loss while modernizing the very early job-displacement literature. After constructing a dynamic model of two-sided learning between a firm and its workers, we estimate it using personnel data from Fokker Aircraft that cover the path of layoffs and quits through its bankruptcy in March 1996. We find that the firm learns about workers' loyalty (demonstrating the role of information in repeated cooperative principal- agent relationships), while workers do not learn (consistent with earlier empirical results on American workers). The type of data that we use also generates information on the value of learning and on whether and how the characteristics of workers who remain until the firm's death differ from those of all affected workers. It thus allows us to measure the increases in the firm's value from learning about its workers' behavior and to infer the extent of biases in estimating losses from displacement from samples restricted to displaced workers.