We show that two models of the labor market, a Walrasian model and a labor contracting model, both have an approximate dynamic factor structure. We use this result to motivate our empirical approach to estimating the cyclical properties of real wages, which does not impose any structure between real wages and observed cyclical indicators. In particular, we employ a Bayesian dynamic factor model and longitudinal microdata to estimate common latent factors driving real wages. We find that the comovement of real wages is related to a common factor that exhibits a mild correlation with the national unemployment rate. Our findings indicate that overall, roughly half of the wages move procyclically while half move countercyclically. In addition, we find that the estimated common factor can explain only a small portion of wage variability. We conclude that these facts are inconsistent with the prediction of a Walrasian labor market model, but consistent with the prediction of a labor contracting model. Finally, our findings suggest that although skilled and unskilled wages are driven by different common skill factors, these factors cannot explain a significant portion of wage variability.
Wages Wage Differentials Business Cycles Bayesian Analysis