Heterogeneous Exposure to Labor Earnings Risk

We analyze labor earnings dynamics based on a large panel of French individuals issued from administrative records. We use the non-parametric approach of Guvenen, Karahan, Ozkan, and Song (2016), and first show labor earnings shocks to exhibit both strong asymmetry and high peakedness, so that (log)normality assumptions do not hold, and then show labor earnings responses to differ substantially between positive and negative shocks, so that non-linearities are at stake. Moreover, both earnings shocks and earnings dynamics are heterogeneous across the earnings distribution. We then get one step further by disentangling the risk related to annual working time from the risk specific to hourly wage. We prove non-Gaussian features and heterogeneity of the earnings shocks distribution to stem mostly from working time instability, rather than from wage shocks. Both wage and working time dynamics display non-linearities, but they do not vary much across the earnings distribution; heterogeneity in labor earnings dynamics arises from large working time shocks explaining a larger share of large labor earnings changes for low earnings workers than they do for high earnings workers. Our results imply that unemployment risk is a key factor of labor earnings risk.