Accounting for Idiosyncratic Wage Risk Over the Business Cycle∗

We demonstrate that wage volatility, measured as the cross-sectional variance of wage changes in PSID data, is counter-cyclical. We quantify this relationship by estimating the regression coefficient of wage volatility on the national unemployment rate in a multilevel Bayesian model, then decompose this coefficient into three main factors. During a recession, wage volatility increases substantially among those workers experiencing spells of unemployment: the cyclical changes in the variance within this group explain about 55% of the cyclical variation in wage volatility. The variance within the group not experiencing unemployment explains 18%. Finally, an increase in the fraction of workers experiencing unemployment explains 25%. We show that a calibrated search-and-matching model of the labor market with on-thejob search gives a good account of the cyclical variation in idiosyncratic wage risk among those experiencing unemployment and of the composition effect over the business cycle. We show that in our model, this result is driven mostly by fluctuations of the reservation wage in response to labor market conditions. JEL classification: C11, E24, E32, J64 ∗We thank Christian Haefke, Nobuhiro Kiyotaki, Per Krusell, Monika Merz, Michael Reiter, Christopher Sims, participants in seminars at the Boston University, the Federal Reserve Banks of Richmond and San Francisco, CERGE-EI, CESifo, NORMAC, UCLA, UCSD, and the Vienna Macro Breakfast. Any remaining errors are our own. This research is supported by the Austrian Central Bank Grant No: 14076. †Boston University. [email protected] ‡Institute for Advanced Studies, Vienna. [email protected]