695[Journal of Political Economy, 2004, vol. 112, no. 3]䉷 2004 by The University of Chicago. All rights reserved. 0022-3808/2004/11203-0007$10.00Cyclical Dynamics in Idiosyncratic Labor MarketRiskKjetil StoreslettenUniversity of Oslo and Centre for Economic Policy ResearchChris I. TelmerCarnegie Mellon UniversityAmir YaronUniversity of Pennsylvania and National Bureau of Economic ResearchIs individual labor income more risky in recessions? This is a difficultquestion to answer because existing panel data sets are so short. Toaddress this problem, we develop a generalized method of momentsestimator that conditions on the macroeconomic history that eachmember of the panel has experienced. Variation in the cross-sectionalvariance between households with differing macroeconomic historiesallows us to incorporate business cycle information dating back to1930, even though our data do not begin until 1968. We implementthis estimator using household-level labor earnings data from thePanel Study of Income Dynamics. We estimate that idiosyncratic riskis (i) highly persistent, with an annual autocorrelation coefficient of0.95, and (ii) strongly countercyclical, with a conditional standardThis paper circulated previously as a component of the paper “Asset Pricing with Idi-osyncratic Risk and Overlapping Generations.” In addition to participants at seminars andconferences, we thank Dave Backus, Geert Bekaert, David Bowman, Martin Browning,John Cochrane, Wouter den Haan, Mick Devereux, Ron Gallant, Rene´ Garcia, Rick Green,John Heaton, Burton Hollifield, Mark Huggett, Narayana Kocherlakota, Per Krusell, Deb-orah Lucas, Pierre Mella-Barral, Bob Miller, Michael Parkin, Victor Rı´os, Amlan Roy, TonySmith, Paul Willen, Steve Zeldes, Stan Zin, and three anonymous referees for helpfulcomments and suggestions. We have benefited from the support of National ScienceFoundation grant SES-9987602 and the Rodney White Center at Wharton. Valuable re-search assistance was provided by Xiaorong Dong.696 journal of political economydeviation that increases by 75 percent (from 0.12 to 0.21) as themacroeconomy moves from peak to trough.I. IntroductionThe interaction between cross-sectional risk and aggregate risk plays anincreasingly important role in macroeconomics and finance. For ex-ample, Bernanke and Gertler (1989), Carlstrom and Fuerst (1997), Coo-ley, Marimon, and Quadrini (2002), and others argue that financingconstraints are an important propagation mechanism because they gen-erate asymmetric effects of aggregate shocks on the the cross-sectionaldistribution of firms. Davis and Haltiwanger (1992), Caballero and Ham-mour (1994), Boeri (1996), Foote (1998), and others examine cyclicaldynamics in job creation and destruction. Rampini (2000) argues thatincentive constraints associated with entrepreneurial activity give rise tocountercyclical cross-sectional risk. Krusell and Smith (1999), Stores-letten, Telmer, and Yaron (2001b), and Krebs (2002, 2003) ask if coun-tercyclical idiosyncratic risk is important for the welfare costs of businesscycles. Finally, and most closely related to this paper, Lucas (1994),Heaton and Lucas (1996), Krusell and Smith (1997), Marcet and Sin-gleton (1999), Balduzzi and Yao (2000), Alvarez and Jermann (2001),Lustig (2001), Storesletten et al. (2001a), Brav, Constantinides, andGeczy (2002), Cogley (2002), and Sarkissian (2003) follow Mankiw(1986) and Constantinides and Duffie (1996) and study models in whichassets with higher expected returns do badly at times of higher cross-sectional labor income risk.The size of cyclical variation in the cross-sectional distribution of laborincome risk is thus an important question. It is hard to answer, however,because panel data sets are so short. Even the Panel Study of IncomeDynamics (PSID)—the panel with the longest time dimension—coversonly four or five business cycles. We address this problem. We incor-porate macroeconomic information dating back to 1930, in spite of thefact that our panel data do not begin until 1968. To understand whatwe do, consider two “cohorts” of individuals, the first born in 1910 andthe second in 1930. Everyone is subject to idiosyncratic labor marketshocks, some fraction of which are highly persistent. The conditionalvariance of these shocks is countercyclical, increasing during contrac-tions and decreasing during expansions. Suppose that we have laborincome data on each cohort when its members are 60 years old. Thatis, we have 1970 data on the 1910 cohort and 1990 data on the 1930cohort. What we shall observe—given the high persistence and the coun-tercyclical idiosyncratic risk—is more cross-sectional dispersion amongthe 1910 cohort than among the 1930 cohort. The reason is that thecyclical dynamics 697former worked through more contractionary years than the latter, in-cluding the Great Depression. More generally, we shall see variation inthe cross-sectional variance between any two cohorts of similar ages whohave worked through different macroeconomic histories. This is theessence of our procedure. Even though the time dimension of our paneldata is limited to 1968–93, we have a rich cross section of ages in eachyear of the panel. We can therefore use macroeconomic data to char-acterize the working history of each household in the panel and thenuse cross-sectional variation between cohorts of similar ages to identifythe cyclical idiosyncratic-risk effects we are after.More specifically, we focus on the properties of household-level labormarket earnings from the PSID. We model idiosyncratic risk as a classof ARMA(1,1) processes with a regime-switching component in the con-ditional variance. The regime is identified by using macroeconomic datato classify each year between 1930 and 1993 as either a contraction oran expansion. A critical feature is finiteness: we make strong assumptionsabout initial conditions that allow us to base a generalized method ofmoments (GMM) estimator on age-dependent moments.We find robust evidence that idiosyncratic earnings risk is both highlypersistent and countercyclical. Estimates of annual autocorrelationrange from 0.94 to 0.96. Estimates of conditional standard deviationsincrease by roughly 75 percent (from 0.12 to 0.21) as the macroeconomymoves from expansion to contraction. We use graphical methods tomake these GMM estimates transparent. We show that high autocor-relation is driven by a linearly increasing pattern of cross-sectional var-iance with