Employer learning, productivity and the earnings distribution: Evidence from performance measures

Two ubiquitous empirical regularities in pay distributions are that the variance of wages increases with experience, and innovations in wage residuals have a large, unpredictable component. The leading explanations for these patterns are that over time, either firms learn about worker productivity but productivity remains fixed or workers' productivities themselves evolve heterogeneously. In this paper, we seek to disentangle these two models and place magnitudes on their relative importance. We derive a dynamic model of learning and productivity that nests both models and allows them to coexist. We estimate our model on a 20-year panel of pay and performance measures from a single, large firm (the Baker-Gibbs-Holmstrom data). Incorporating performance measures yields two key innovations. First, the panel structure implies that we have repeat measures of correlates of productivity, as opposed to the empirical evidence on employer learning which uses one fixed measure. Second, we can separate productivity from pay, whereas the previous literature on productivity evolution could not. We find that both models are important in explaining the data. However, the predominant effect is that worker productivity evolves idiosyncratically over time, implying firms must continuously learn about a moving target. Therefore, while the majority of pay dispersion is driven by variation in individual productivity, wages differ significantly from individual productivity at all experience levels due to imperfect information. We believe this represents a significant reinterpretation of the empirical literature on employer learning.