This paper presents an empirical analysis of individual earnings and hours data from three different longitudinal surveys. In the first part of the paper we catalog the main features of the covariance structure of earnings and hours changes. We find that this structure is very similar across data sets, and may be adequately summarized by a simple components-of-variance model, consisting of (i) serially uncorrelated measurement error, (ii) a shared component of earnings and hours with a second-order moving average covariance structure, and (iii) a nonstationary component that affects only the variances and contemporaneous covariances of earnings and hours. In the second part of the paper we offer an interpretation of this model in terms of a simple life-cycle labor supply model. On the assumption that we can identify individualproductivity growth with the shared component of earnings and hours variation, we obtain estimates of the intertemporal substitution elasticity. The results are not favorable to the life-cycle model: most of the covariation of earnings and hours occurs at fixed hourly wage rates.