Today, as thirty and forty years ago, economists debate how much unemployment is voluntary, how much involuntary, how much is a phenomenon of equilibrium, how much a symptom of disequilibrium; how much is compatible with competition, how much is to be blamed on monopolies, labor unions, and restrictive legislation, how much unemployment characterizes ‘full’ employment. A concept of full employment more congenial to economic theory is labor market equilibrium, a volume of employment which is simultaneously the amount employers want to offer and the amount workers want to accept at prevailing wage rates and prices. Forty years ago theorists with confidence in markets could believe that full employment is whatever volume of employment the economy is moving toward, and that its achievement requires of the government nothing more than neutrality, and nothing less After Keynes challenged the classical notion of labor market equilibrium and the complacent view of policy to which it led, full employment came to mean maximum aggregate supply, the point at which expansion of aggregate demand could not further increase employment and output. With a deflationary gap, demand less than full employment supply, prices would be declining or at worst constant. Expansion of aggregate demand short of full employment would cause at most a one-shot increase of prices. Postwar experience destroyed the identification of full employment with the economy’s inflation threshold. The natural rate is another full employment candidate, a policy target at least in the passive sense that monetary and fiscal policy makers are advised to eschew any numerical unemployment goal and to let the economy gravitate to this equilibrium. Full employment is once again nothing but the equilibrium reached by labor markets unaided and undistorted by governmental fine tuning. The presumption he challenged is that in competitive labor markets actual employment and unemployment reveal workers’ true preferences between work and alternative uses of time, the presumption that no one is fully or partially unemployed whose real wage per hour exceeds his marginal valuation of an hour of free time.
Orthodox economists found the observed stickiness of money wages to be persuasive evidence that unemployment, even in the Great Depression, was voluntary. Keynes found decisive evidence against this inference in the willingness of workers to accept a larger volume of employment at a lower real wage resulting from an increase of prices. His expected expansion to raise prices and lower real wages, but this expectation is not crucial to his argument. If it is possible to raise employment without reduction in the real wage, his case for calling the unemployment involuntary is strengthened. Why is the money wage so stubborn if more labor is willingly available at the same or lower real wage? Consider first some answers Keynes did not give. He did not appeal to trade union monopolies or minimum wage laws. Instead, Keynes emphasized the institutional fact that wages are bargained and set in the monetary unit of account. Money wage rates are, to use the Keynesian term, ‘administered prices’. That is, they are not set and reset in daily auctions but posted and fixed for finite periods of time. This observation led Keynes to his central explanation: Workers, individually and in groups, are more concerned with relative than absolute real wages. They may withdraw labor if their wages fall relatively to wages elsewhere, even though they would not withdraw any if real wages fall uniformly everywhere. Labor markets are decentralized, and there is no way, money wages can fall in any one market without impairing the relative status of the workers there. A general rise in prices is a neutral and universal method of reducing real wages, the only method in a decentralized and uncontrolled economy. Inflation would not be needed, we may infer, if by government compulsion, economy-wide bargaining, or social compact, all money wage rates could be scaled down together. Keynes apparently meant that relative wages are the arguments in labor supply functions.
Alchian and other theorists of search activity have offered a somewhat different interpretation, namely that workers whose money wages are reduced will quit their jobs to seek employment in other markets where they think, perhaps mistakenly, that wages remain high. Keynes’s explanation of money wage stickiness is plausible and realistic. All equilibrium prices, including money wage rates, should differ in the same proportion, while all real magnitudes, including employment, should be the same in the two equilibriums. The resistance of money wage rates to excess supply is a feature of the adjustment process rather than a symptom of irrationality. The other side of Keynes’s story is that in depressions money wage deflation, even if it occurred more speedily, or especially if it occurred more speedily, would be at best a weak equilibrator and quite possibly a source of more unemployment rather than less. What relevance has this excursion into depression economics for contemporary problems of unemployment and wage inflation? The issues are remarkably similar, even though events and Phillips have shifted attention from levels to time rates of change of wages and prices. Phillips curve doctrine is in an important sense the postwar analogue of Keynesian wage and employment theory, while natural rate doctrine is the contemporary version of the classical position Keynes was opposing. Let us adapt Keynes’s test for involuntary unemployment to the dynamic terms of contemporary discussion of inflation, wages, and unemployment. Associated with it is a path of real wages, rising at the rate of productivity growth. Consider an alternative future, with unemployment at first declining to a rate one percentage point lower and then remaining constant at the lower rate. Associated with the lower unemployment alternative will be a second path of real wages. Eventually this real wage path will show, at least to first approximation, the same rate of increase as the first one, the rate of productivity growth. The growth of real wages will be retarded in the short run if additional employment lowers labor’s marginal productivity. In any case, the test question is whether with full information about the two alternatives labor would accept the second one whether, in other words, the additional employment would be willingly supplied along the second real wage path. For Keynes’s reasons, a negative answer cannot necessarily be inferred from failure of money wage rates to fall or even decelerate. Actual unemployment and the real wage path associated with it are not necessarily an equilibrium. Rigidities in the path of money wage rates can be explained by workers’ preoccupation with relative wages and the absence of any central economy-wide mechanism for altering all money wages together. According to the natural rate hypothesis, there is just one rate of unemployment compatible with steady wage and price inflation, and this is in the long run compatible with any constant rate of change of prices, positive, zero, or negative. Only at the natural rate of unemployment are workers content with current and prospective real wages, content to have their real wages rise at the rate of growth of productivity. Along the feasible path of real wages, they would not wish to accept any larger volume of employment. Lower unemployment can arise only from economy-wide excess demand for labor and must generate a gap between real wages desired and real wages earned. The gap evokes increases of money wages designed to raise real wages faster than productivity. This intention is always frustrated, the gap is never closed, and money wages and prices accelerate. Older classical economists regarded constancy of money wage rates as indicative of full employment equilibrium, at which the allocation of time between work and other pursuits is revealed as voluntary and optimal. Their successors make the same claims for the natural rate of unemployment, except that in the equilibrium money wages are not necessarily constant but growing at the rate of productivity gain plus the experienced and expected rate of inflation of prices.
There are two conflicting interpretations of the welfare value of employment in excess of the level consistent with price stability. One is that additional employment does not produce enough to compensate workers for the value of other uses of their time. The alternative view, which I shall argue, is that the responses of money wages and prices to changes in aggregate demand reflect mechanics of adjustment, institutional constraints, and relative wage patterns and reveal nothing in particular about individual or social valuations of unemployed time vis-a-vis the wages of employment. On this rostrum four years ago, Milton Friedman identified the noninflationary natural rate of unemployment with “Equilibrium in the structure of real wage rates”. “The natural rate of unemployment”, – he said, “… is the level that would be ground out by the Walrasian system of general equilibrium equations, provided that there is embedded in them the actual structural characteristics of the labor and commodity markets, including market imperfections, stochastic variability in demands and supplies, the costs of getting information about job vacancies and labor availabilities, the costs of mobility, and so on”. Presumably this Walrasian equilibrium also has the usual optimal properties; at any rate, Friedman advised the monetary authorities not to seek to improve upon it. In fact, we know little about the existence of the Walrasian equilibrium that allows for all the imperfections and frictions that explain why the natural rate is bigger than zero, and even less about the optimality of such an equilibrium if it exists. In the new microeconomics of labor markets and inflation, the principal activity whose marginal value sets the reservation price of employment is job search. It is not pure leisure, for in principle persons who choose that option are not reported as unemployed; however, there may be a leisure component in job seeking. Members of our own profession are adept at seeking and finding new jobs without first leaving their old ones or abandoning not-in-labor-force status.
In surveys of job mobility among blue collar workers in 1946-47, 25% of workers who quit had new jobs lined up in advance. A considerable amount of search activity by unemployed workers appears to be an unproductive consequence of dissatisfaction and frustration rather than a rational quest for improvement. Reynolds found that quitting a job to look for a new one while unemployed actually yielded a better job in only a third of the cases. Lining up a new job in advance was a more successful strategy: two-thirds of such changes turned out to be improvements.
Today, according to the dual labor market hypothesis, the basic reason for frequent and long spells of unemployment in the secondary labor force is the shortage of good jobs. In any event, the contention of some natural rate theorists is that employment beyond the natural rate takes time that would be better spent in search activity. Why do workers accept such employment? An answer to this question is a key element in a theory that generally presumes that actual behavior reveals true preferences. The answer given is that workers accept the additional employment only because they are victims of inflation illusion. One form of inflation illusion is overestimation of the real wages of jobs they now hold, if they are employed, or of jobs they find, if they are unemployed and searching. If they did not underestimate price inflation, employed workers would more often quit to search, and unemployed workers would search longer. The force of this argument seems to me diluted by the fact that price inflation illusion affects equally both sides of the job seeker’s equation. He over-estimates the real value of an immediate job, but he also overestimates the real values of jobs he might wait for. As a first approximation, inflation illusion has no substitution effect on the margin between working and waiting. It does have an income effect, causing workers to exaggerate their real wealth. In which direction the income effect would work is not transparent. Does greater wealth, or the illusion of greater wealth, make people more choosy about jobs, more inclined to quit and to wait? Or less choosy, more inclined to stay in the job they have or to take the first one that comes along? I should have thought more selective rather than less. Natural rate theory must take the opposite view if it is to explain why underestimation of price inflation bamboozles workers into holding or taking jobs that they do not really want. Workers are myopic and do not perceive, that wages elsewhere are, or soon will be, rising as fast as the money wage of the job they now hold or have just found. Consequently, they underestimate the advantages of quitting and searching. This explanation is convincing only to the extent that the payoff to search activity is determined by wage differentials. The payoff also depends on the probabilities of getting jobs at quoted wages, therefore on the balance between vacancies and job seekers. Quit rates are an index of voluntary search activity. The new microeconomics of job search, is nevertheless a valuable contribution to understanding of frictional unemployment. Does the market produce the optimal amount of search unemployment? Is the natural rate optimal? I do not believe the new microeconomics has yet answered these questions. An omniscient and beneficent economic dictator would not place every new job seeker immediately in any job at hand. The hypothetical planner would prefer to keep a queue of workers unemployed, so that he would have a larger choice of jobs to which to assign them. He would not make the queue too long, because workers in the queue are not producing anything. Of course, he could shorten the queue of unemployed if he could dispose of more jobs and lengthen the queue of vacancies. With enough jobs of various kinds, he would never lack a vacancy for which any worker who happens to come along has comparative advantage. Because of limited capital stocks and interdependence among skills, jobs cannot be indefinitely multiplied without lowering their marginal productivity.
Our wise and benevolent planner would not place people in jobs yielding less than the marginal value of leisure. Given this constraint on the number of jobs, he would always have to keep some workers waiting, and some jobs vacant. He certainly would be in efficient if he had fewer jobs, filled and vacant, than this constraint. This is the common sense of Beveridge’s rule-that vacancies should not be less than unemployment. Is the natural rate a market solution of the hypothetical planner’s operations research problem? According to search theory, an unemployed worker considers the probabilities that he can get a better job by searching longer and balances the expected discounted value of waiting against the loss of earnings.
A worker deciding to join a queue or to stay in one considers the probabilities of getting a job, but not the effects of his decision on the probabilities that others face. He lowers those probabilities for people in the queue he joins and raises them for persons waiting for the kind of job he vacates or turns down. Too many persons are unemployed waiting for good jobs, while less desirable ones go begging. Empirically the proposition that in the United States the zero-inflation rate of unemployment reflects voluntary and efficient job-seeking activity strains credulity. If there were a natural rate of unemployment in the United States, what would it be? It is hard to say because virtually all econometric Phillips curves allow for a whole menu of steady inflation rates. Estimates constrained to produce a vertical long-run Phillips curve suggest a natural rate between 5 and 6 % of the labor force. So let us consider some of the features of an overall unemployment rate of 5 to 6 %. First, about 40% of accessions in manufacturing are rehires rather than new hires. Their unemployment is as much a deadweight loss as the disguised unemployment of redundant workers on payrolls. This number declines to 25-30%, when unemployment is 4% or below. A 5-6% unemployment rate means that voluntary quits amount only to about a third of separations, layoffs to two-thirds. The proportions are reversed at low unemployment rates. Second, the unemployment statistic is not an exhaustive count of those with time and incentive to search. An additional 3% of the labor force are involuntarily confined to part-time work, and another 3-4% are out of the labor force because they ‘Could not find job’ or ‘Think no work available’ – discouraged by market conditions rather than personal incapacities. Third, with unemployment of 5-6% the number of reported vacancies is less than 1/ 2 of 1%. Vacancies appear to be understated relative to unemployment, but they rise to l2% when the unemployment rate is below %. At 5-6% unemployment, the economy is clearly capable of generating many more jobs with marginal productivity high enough so that people prefer them to leisure. The capital stock is in limitation, since 5-6% unemployment has been associated with more than 20% excess capacity. When more jobs are created by expansion of demand, with or without inflation, labor force participation increases; this would hardly occur if the additional jobs were low in quality and productivity. As the parable of the central employment planner indicates, there will be excessive waiting for jobs if the roster of jobs and the menu of vacancies are suboptimal. In summary, labor markets characterized by 5-6% unemployment do not display the symptoms one would expect if the unemployment were voluntary search activity. The only reason anyone might regard so high an unemployment rate as an equilibrium and social optimum is that lower rates cause accelerating inflation.
In other words, the economy has an inflationary bias: when labor markets provide as many jobs as there are willing workers, there is inflation, perhaps accelerating inflation. One rationalization might be termed a theory of stochastic macro-equilibrium: stochastic, because random inter sectorial shocks keep individual labor markets in diverse states of disequilibrium; macro- equilibrium, because the perpetual flux of particular markets produces fairly definite aggregate outcomes of unemployment and wages. It is an essential feature of the theory that economy-wide relations among employment, wages, and prices are aggregations of diverse outcomes in heterogeneous markets. The myth of macroeconomics is that relations among aggregates are enlarged analogues of relations among corresponding variables for individual house-holds, firms, industries, markets. Money wages do not adjust rapidly enough to clear all labor markets every clay. Excess supplies in labor markets take the form of unemployment, and excess demands the form of unfilled vacancies. At any moment, markets vary widely in excess demand or supply, and the economy, as a whole, shows both vacancies and unemployment.
Higher aggregate demand means fewer excess supply markets and more excess demand markets, accordingly less unemployment and more vacancies. In any particular labor market, the rate of increase of money wages is the sum of two components, an equilibrium component and a disequilibrium component. The first is the rate at which the wage would increase were the market in equilibrium, with neither vacancies nor unemployment. Why aren’t they eliminated by the very adjustments they set in motion? Workers will move from excess supply markets to excess demand markets, and from low wage to high wage markets. The overlap of vacancies and unemployment-say, the sum of the two for any given difference between them is a measure of the heterogeneity or dispersion of individual markets. The amount of dispersion depends directly on the size of those shocks of demand and technology that keep markets in perpetual disequilibrium, and inversely on the responsive mobility of labor. A central assumption of the theory is that the functions relating wage change to excess demand or supply are non-linear, specifically that unemployment retards money wages less than vacancies accelerate them. Nonlinearity in the response of wages to excess demand has several important implications. Linear wage response, on the other hand, would mean a linear Phillips relation. Second, given the overall state of aggregate demand, economy-wide vacancies less unemployment, wage inflation will be greater the larger the variance among markets in excess demand and supply. Of course, the rate of wage inflation will depend not only on the overall dispersion of excess demands and supplies across markets but also on the particular markets where the excess sup- plies and demands happen to fall.
An unlucky random drawing might put the excess demands in highly responsive markets and the excess supplies in especially unresponsive ones. Full long-run equilibrium in all markets would show no unemployment, no vacancies, and no unanticipated inflation. With unending sectorial flux, zero excess demand spells inflation and zero inflation spells net excess supply, unemployment in excess of vacancies. Both criteria automatically allow for frictional unemployment incident to the required movements of workers between markets; the no-inflation criterion requires enough additional unemployment to wipe out inflationary bias. I turn now to the equilibrium component, the rate of wage increases in a market with neither excess demand nor excess supply. It is reasonable to suppose that the equilibrium component depends on the trend of wages of comparable labor elsewhere. A ‘competitive wage’, one that reflects relevant trends fully, is what employers will offer if they wish to maintain their share of the volume of employment. TI his will happen where the rate of growth of marginal revenue product the com pound of productivity increase and price inflation is the same as the trend in wages. In some markets the equilibrium wage will be rising faster, and in others slower, than the economy-wide wage trend. A ‘natural rate’ result follows if actual wage increases feed fully into the equilibrium components of future wage increases. Phillips tradeoffs exist in the short run, and the time it takes for them to evaporate depends on the lengths of the lags with which today’s actual wage gains become tomorrow’s standards. Suppose there is a floor on wage change in excess supply markets, independent of the amount of excess supply and of the past history of wages and prices. Suppose, for example, that wage change is never negative; it is either zero or what the response function says, whichever is algebraically larger.
So long as there are markets where this floor is effective, there can be determinate rates of economy-wide wage inflation for various levels of aggregate demand. Markets at the floor do not increase their contributions to aggregate wage inflation when overall demand is raised. Nor is their contribution escalated to actual wage experience. The frequency of such markets diminishes, it is true, both with overall demand and with inflation. The floor phenomenon can preserve the Phillips tradeoff within limits, but one that becomes ever more fragile and vanishes as greater demand pressure removes markets from contact with the zero floor. It will be objected that any permanent floor independent of general wage and price history and expectation must indicate money illusion. The answer is that the floor need not be permanent in any single market. It could give way to wage reduction when enough unemployment has persisted long enough. With stochastic inter sectorial shifts of demand, markets are always exchanging roles, and there can always be some markets, not always the same ones, at the floor.
During the great contraction of 1930-33, wage rates were slow to give way even in the face of massive unemployment and substantial deflation in consumer prices. Finally in 1932 and 1933 money wage rates fell more sharply, in response to prolonged unemployment, layoffs, shutdowns, and to threats and fears of more of the same. I have gone through this example to make the point that irrationality, in the sense that meaningless differences in money values permanently affect individual behavior, is not logically necessary for the existence of a long-run Phillips trade-off. In full long-run equilibrium in all markets, employment and unemployment would be independent of the levels and rates of change of money wage rates and prices. But this is not an equilibrium that the system ever approaches. The economy is in perpetual sectorial disequilibrium even when it has settled into a stochastic macro-equilibrium. I suppose that one might maintain that asymmetry in wage adjustment and temporary resistance to money wage decline reflect money illusion in some sense. Such an assertion would have to be based on an extension of the domain of well-defined rational behavior to cover responses to change, adjustment speeds, costs of information, costs of organizing and operating markets, and a host of other problems in dynamic theory. These theoretical extensions are in their infancy, although much work of interest and promise is being done. Meanwhile, I doubt that significant restrictions on disequilibrium adjustment mechanisms can be deduced from first principles. Why are the wage and salary rates of employed workers so insensitive to the availability of potential replacements? One reason is that the employer makes some explicit or implicit commitments in putting a worker on the payroll in the first place. The employee expects that his wages and terms of employment will steadily improve, certainly never retrogress. He expects that the employer will pay him the rate prevailing for persons of comparable skill, occupation, experience, and seniority. He expects such commitments in return for his own investments in the job: arrangements for residence, transportation, and personal life involve set-up costs which will be wasted if the job turns sour. The market for labor services is not like a market for fresh produce where the entire current supply is auctioned daily. It is more like a rental housing market, in which most existing tenancies are the continuations of long-term relationships governed by contracts or less formal understandings. Employers and workers alike regard the wages of comparable labor elsewhere as a standard, but what determines those reference wages? There is not even an auction where workers and employers unbound by existing relationships and commitments meet and determine a market-clearing wage. If such markets existed, they would provide competitively determined guides for negotiated and administered wages, just as stock exchange prices are reference points for stock transactions elsewhere. In labor markets the reverse is closer to the truth. Wage rates for existing employees set the standards for new employees, too. The equilibrium components of wage increases, it has been argued, depend on past wage increases throughout the economy. In those theoretical and econometric models of inflation where labor markets are aggregated into a single market, this relationship is expressed as an autoregressive equation of fixed structure: current wage increase depends on past wage increases. The same description applies when past wage increases enter indirectly, mediated by price inflation and productivity change. The process of mutual interdependence of market wages is a good deal more complex and less mechanical than these aggregated models suggest. Reference standards for wages differ from market to market. The equilibrium wage increase in each market will be some function of past wages in all markets, and perhaps of past prices too. But the function need not be the same in every market. Wages of workers contiguous in geography industry, and skill will be heavily weighted. Imagine a wage pattern matrix of coefficient describing the dependence of the percentage equilibrium wage increase in each market on the past increases in all other markets. The coefficients in each row are non-negative and sum to one, but their distribution across markets and time lags will differ from row to row.