Predicting Earnings Distributions across Cities: The Human Capital Model vs the National Distribution Hypothesis

Abstract

Microdata are used to examine the relative ability of the human capital model and of an alternative national distribution hypothesis to generate predicted distributions of earnings that are close to actual distributions within 48 SMSAs. Surprisingly, the national distribution hypothesis is found to be relatively more robust in predicting earnings distributions than the fixed-parameter human capital model. Earnings functions are then estimated separately within each SMSA, and it is found that the parameters of the human capital model vary significantly across labor markets. Further analysis examines the relationship between earnings distributions, the estimated parameters of the model, city size, and region.