WAITING FOR WORK*George A. Akerlof, Andrew K. Rose and Janet L. YellenUniversity of California at BerkeleyApril, 2002* This paper was prepared for Imperfect Economics: Essaysin Honor of Joseph Stiglitz. It is adapted from our earlierunpublished paper with the same title, NBER Working PaperNo. 3385, May 1990.I. IntroductionIt is more than a pleasure to say “Happy Birthday” to JosephStiglitz. To two of the co-authors of this paper, he has beenthe finest of friends for as long as we can remember. GeorgeAkerlof first met Joe when Joe, aged 20, came to MIT as aprecocious graduate student. Janet Yellen has known him since1967, when she, aged 21, arrived at Yale for her Ph.D. One ofJoe’s first papers, with George, concerned the relation betweenwages and unemployment; that paper motivated Janet Yellen’sthesis, which Joe advised. The relation between wages andunemployment has been a prime focus of Joe’s research throughouthis career (see, for example, Shapiro and Stiglitz (1984) andGreenwald and Stiglitz, (1988) and (1993)). His contributionshave been central to the development of efficiency wage theoryand the New Keynesian macroeconomics. This paper presents adifferent perspective on the relation between wages andunemployment and thus is a fitting contribution to celebrate ourlongstanding collaboration with Joe. Joe’s interest inunemployment is, of course, just one facet of his broader concernwith poverty and its cure, both at home and abroad. We admirethe passion, intellect and purposefulness with which Joe haspursued this important research and policy agenda as well as hisconstant good humor, grace, and ever-present smile.2This paper develops a model of “wait unemployment” designedto accord with three well-documented empirical regularities:high-skilled workers suffer more moderate cyclical fluctuationsin employment and unemployment than low-skilled workers;1low-skilled workers gain substantially improved access to “good jobs”during expansions;2and job-changers experience larger procyclicalreal wage movements than workers who remain in the same job.3Wedo not attempt to model the shocks that cause cyclicalfluctuations in job opportunities, treating both the flow of newjobs and the wages associated with them as exogenous. Our focusinstead is on the process governing skill patterns of wages,employment and unemployment during a cyclical recovery.Although our model takes the supply of jobs in the aggregateas “exogenous”, we assume that more skilled workers can, shouldthey choose, bump less skilled workers for available jobvacancies since firms prefer to hire the most able workersavailable. Consistent with the evidence concerning jobdowngrading and upgrading, such bumping occurs during recessionsin our model; but the extent of cyclical downgrading isendogenously limited by the willingness of workers who are laidoff in a downturn to rationally wait to accept jobs untilbusiness conditions improve. The unemployment experienced byskilled workers in our model during recessions thus reflects3their decision to “wait for work”: these workers find it rationalto hold out for the “good jobs” which will appear later in anexpansion rather than “locking-in” the lower wages paid by the“less good” jobs that are available to them during the initialstages of a recovery.Our model is motivated in part by the observation that laboris not the only factor of production that experiences periods ofidleness. Office buildings sometimes stand unoccupied forextended periods of time and oil reserves sit idly underground.In the case of oil (and other exhaustible natural resources), awell-developed theory (Hotelling 1931) explains why the ownerswait to extract their resource. In the equilibrium of theHotelling model, the owners of oil reserves are compensated forwaiting by an increase in the price of oil at the rate ofinterest.In contrast to oil, the use of office buildings in one perioddoes not preclude their use in other periods. In this respect,workers more closely resemble office buildings than oil. Yetoffice-space gluts are fairly common: in Houston during the1990s, for example, vacancy rates were extraordinarily high in amany completed office buildings. The Hotelling model can beadapted to explain the existence of vacant office buildingsprovided that a significant fixed cost must be borne when office4space is occupied or vacated. If such costs are sufficientlylarge, there is a "lock-in" effect: a building owner who rentshis office space today to one tenant forgoes the possibility ofrenting the same space in later periods to other tenants. Iflong-term rental rates increase more rapidly than the rate ofinterest, it pays the owner of an unoccupied building to leavethe space vacant and wait until conditions improve to rent outspace4. This is true even if there are tenants willing to pay tooccupy the space now. In contrast, if the rental rates on long-term leases increase at less than the rate of interest, thebuilding owner maximizes the present value of his income byrenting all available space now, since the reward to waiting inthe form of higher rents in the future, does not make up for theloss in rentals today. In analogy to the market for oil, inequilibrium, long-term rental rates will rise at precisely therate of interest with the stock of excess office space beinggradually eliminated over time.The theory of wait unemployment developed in this paper isexactly analogous to Hotelling's model as it would be applied tovacant office space. The cyclical unemployment of workersseeking long-term (primary-sector) jobs is analogous to thevacancies in office buildings whose owners seek long-termtenants. The "labor supply" function in our model is perfectly5elastic; thus the model can rationalize the finding that largevariations in employment are accompanied by small procyclicvariations in wages concentrated among those workers who changejobs. Our model thus accounts for large aggregate fluctuationsin employment without empirically implausible elasticities ofsubstitution between leisure in different time periods. Itoffers an alternative rationale for a high elasticity of laborsupply with respect to transitory wage movements: if the wagewere rising more rapidly than the rate of time preference, arational worker seeking a new, long-term job would optimally waitfor work, rather than commit to the best job currently available.This behavior occurs even if workers place no value at all