ROBERT E. HALL Massachusetts Institute of Technology The Rigidity of Wages and the Persistence of Unemployment IN THE MODERN American economy, fluctuations in unemployment usu- ally persist from one year to the next. Past decades have seen two lengthy deviations from equilibrium in the labor market: 1958 through 1963, when the market was slack for six straight years, and 1964 through 1970, when it was tight for seven straight years. Virtually all forecasters agree today that the sharp recession of 1974-75 will mark the beginning of another extended period of slack, with the official unemployment rate above 6 percent per- haps until the end of the decade. These forecasts are fully consistent with the behavior of the economy after the only comparable postwar recession, in 1957-58. Explanation of the persistence of unemployment has been a major focus of macroeconomic theory since the Great Depression. Though his predecessors had begun to think seriously about the challenge to classi- cal economic theory raised by successive years of high unemployment, John Maynard Keynes was the great pioneer in creating a theory that came to grips with the facts of persistent unemployment, and he remains the dominant figure today. This paper presents a detailed critique of modern explanations of the be- havior of unemployment, both within the Keynesian tradition and outside it. The critique begins with a close look at the modern theory of disequi- librium, which claims to provide a microeconomic foundation for the the- ory of unemployment and wage adjustment. This theory attributes both the Note: This researchwas supportedby the National ScienceFoundation. I am grateful to David Lilien and WynettaMcNeill for assistance,and to membersof the Brookings panel for useful comments. 301 302 Brookings Papers on Economic Activity, 2:1975 rigidityof wagesand the persistenceof unemploymentto the slow diffusion of informationwithin the labormarket. The evidencepresented here, how- ever,suggests that unemploymentin the modernAmerican economy per- sists farlonger than is remotelyplausible if this factoris the maincause of the lag in marketclearing. This hypothesisis stronglyconfirmed by a study of the dataunder the assumptionthat threemonths is long enoughfor both workersand employersto find out what is happeningin the labormarket. The paperthen turnsto a group of theoriesthat are more traditionally Keynesianin flavor.First is the beliefthat laborunions are responsiblefor the rigidity of wages. Keynes' theory of effectivedemand showed con- vincinglythat rigid wages can permit persistent unemployment, and Keynes himself,as well as his successors,appeared to hold laborunions primarily responsiblefor the nonclassicalbehavior of wages. Closelyrelated is the theoryof wagedetermination based on "idiosyncraticexchange." This the- ory emphasizesthe bilateralmonopoly betweenindividual workers and employers,which biases against the adjustmentof wagesin the short run and makesequity an importantelement in wage determination.These two theoriesshare an essentialdefect: however successful they arein explaining wagerigidity in a particularsector, they do not seemcapable of explaining economy-widerigidity. As long as a competitiveresidual sector with a flexiblewage exists,workers who are unableto findjobs in the rigid-wage sectorwill find it in the residualsector and its wage will be depressedas a result. Overall,the labor marketwill clear even if one sector has rigid wages.If unions and other sourcesof rigidityin particularsectors have a role in persistentunemployment, it is more subtle than the traditional Keynesianview supposes. Empiricalevidence on the behaviorof wages, presentednext, suggests that the modernAmerican economy contains an importantsector with rigid wages. Though unionismand idiosyncraticexchange are presentin this sector,its definingcharacteristic is different.The sectorwith inflexible wagesconsists of government,regulated industries, and nonprofitinstitu- tions. These industriesmake up what I will call the "nonentrepreneurial" sector,in whichemployers do not facethe usualincentive to minimizelabor cost. The empiricalresults suggest that the nonentrepreneurialsector is muchthe most importantsource of wage rigidityin today'seconomy. After isolatingthe major source of wage rigidityin the contemporary U.S. economy,the papergoes on to presenta theoryof the transmission, or spillover,of this rigidityto the competitivesector. The moderntheory of RobertE. Hall 303 the supplyside of unemploymentis shownto implythat competitivewages will not fall far enoughto clearthe labormarket at low frictionallevels of unemploymentwhen aggregatedemand falls. Rather,the durationof un- employmentwill rise as the unemployedsearch longer for the high-paying jobsin the rigid-wagesector. During a contraction,the payoffto job search rises because the wage premiumfor these good jobs rises. Conversely, searchis less usefulto the unemployedwhen the labormarket is unusually tight, as competitivewages rise close to rigid wages. The existenceof the rigid-wagesector lessens the ability of the competitivesector to push the wage down to its market-clearinglevel duringa contraction.The paper does not containdirect evidence of this mechanism,but does show that the observedrelation between unemployment and relativewages in the two sectorsaccords roughly with the predictionsof the theory. All of the theoriesdiscussed here are Keynesianin the sense that they explainwhy the mechanismof effectivedemand can operatewithout the immediatecorrective force of market-clearingwage adjustments.They agree that the basic cause of an increasein unemploymentis contrac- tionaryaggregate policy or othershocks in the economy,and similarlythat expansionarypolicy can eliminateexcessive unemployment. With the pos- sible exceptionof some versionsof the theory of disequilibrium,none of these theoriescan fairlybe accusedof makingpersistent unemployment a voluntaryphenomenon arising from the supply side of the labor market. Rather,all of the theoriesthat appearto have any explanatorypower con- cur that unemploymentis the resultof inadequatedemand for labor. The interactionof demandand supplyis describedin greaterdetail in the next section. Theoriesof Unemployment A simple two-equationstructural model of the economy capturesthe theoreticalrelationships that are importantfor the issuesaddressed by this paperand can encompassthe alternativetheories considered here. (1) effective demand: ut = f(wt, xt); (2) wage adjustment: wt - wt- = g(ut) + t - wt-. Hereut is the unemploymentrate and the functionfembodiesthe effective- demandmechanism for the determinationof real gross nationalproduct 304 BrookingsPapers on EconomicActivity, 2:1975 and Okun'slaw for translatingit into the unemploymentrate. Effective aggregatedemand depends negatively on the nominalwage, wt, throughthe Keynes and Pigou effects, and depends on various policy variables,xt (monetaristsand Keynesiansdisagree about the presenceof fiscalvariables in xt, but that does not matterhere). In the wage-adjustmentequation, g(ut) is the disequilibrium component and w-t- Wti is the expected or equilibriumcomponent.' Monetarists usually write the equationin a pre- ciselyequivalent form, (3) Ut = g(wt- , and call it "aggregatesupply."2 A convenientalgebraic specification of the systemis (4) effective demand: ut = - i/i(mt - wt) + nt; (5) wage adjustment: wt - wti = -k1(ut - u*) + Wt - Wt-i + Et, whereu* is the equilibriumvalue. H-ereI have switchedto takingw as the log of the nominalwage, m as the log of the moneysupply, and u as the log of the unemploymentrate. I assume,for simplicity,that otherpolicy vari- ables, representedby x in the previousform of the equation,do not enter in determiningeffective demand. The equationsalso containrandom shifts, 77and e. In a structuralmodel, the fundamentalexplanation of the behavior of the endogenousvariables resides in the reducedform, which gives the impact of the predeterminedvariables on the endogenousvariables after takingaccount of the interdependenceof the endogenousvariables in the equations.Statements dealing with causal relationsbetween endogenous variables-such as "high wages cause excessive unemployment"-are meaninglessor contradictoryin a structuralmodel. High wages are asso- ciated with high unemploymentin the aggregate-demandequation, but with low unemploymentin the wage-adjustmentequation. The reduced- formequation for unemploymentsorts out the interactionof the two vari- ables in both equations: - t (6) Ut = u* + -o -1/(mt - Et) + 77t 1. More generally, wi - Wt embodies the inertia that is widely believed to be presentin wage determination.The evidencesuggests that expectationsare only a part of the explanationof inertia,but the use of the term is so widespreadthat I will adopt it here. 2. Thomas J. Sargent studied exactly this two-equationsystem in monetaristform in "Rational Expectations,the Real Rate of Interest, and the Natural Rate of Unem- ployment,"BPEA, 2:1973, pp. 429-72. RobertE. Hall 305 The majorissues of the theoryof unemploymentthat areconsidered in this paperconcern the valuesof the parametersin this equation,the specifica- tion of the expectedwage, and the behaviorof the randomshifts. First is the role of the nominalwage in determiningreal aggregatedemand, con- trolledby the parameter01. In the crudeKeynesian model with the liquidity trapand no Pigoueffect, aip = 0, wageadjustment is irrelevant,and unem- ploymentremains at a level determinedby the effective-demandprocess: (7) crudeKeynesian reduced