At least since the early 1970's, it has been apparent that the cyclical variations in inflation summarized by the short-run Phillips curve are only one part of the inflation problem that confronts modern industrial economies. Another part is the relative persistence of an established rate of inflation. There is a good deal that we do not understand about this persistence. But I find the most useful way to model it is to start with the concept of a relatively stable wage norm, by which I mean a norm for the rate of wage increase. The model distinguishes sharply between the cycle and the trend in inflation, with the wage norm determining the trend. The variations in inflation of the typical business cycle take place around the existing wage norm and generate the empirical short-run Phillips curve. The wage norm itself is affected little if at all by the typical business cycle. Historically the wage norm has been shifted by prolonged departures from typical business cycles or by other extreme economic developments. Figuring out more precisely what it takes to shift wage norms, or what might keep them from shifting, is a central challenge for understanding inflation better. Before turning to its implications, let me sketch the behavioral underpinnings of the wage norm model and the empirical evidence about wage norms. The norm rate of wage increase has no allocational significance and describes the trend of nominal wages independent of real aggregate demand or relative demand effects. In this respect, it is like the anticipated rate of inflation in many familiar models. Wages are not determined in an auction-like labor market that clears over any reasonable interval of time. Rather they are established by wage-setting firms with a profit-maximizing interest in their long-run relation with their employees, in some cases in a bargaining situation with unions. Under both the implicit and explicit contracts that thus dominate wage setting, keeping up with the norm is the neutral standard for firms. An individual firm that raises wages in line with the norm neither improves nor worsens its relative position as an employer. A firm that wants to expand employment will, typically, offer a higher wage than would be required just to keep up with the norm. Relative wages and relative employment levels are thus codetermined in this process. When most firms want to expand employment, as in a cyclical upturn, the same behavior is part of the process producing the modest cyclical rise in inflation that we observe as the short-run Phillips curve. Thus the onset of cyclical inflation is not a sign that capital and labor resources are being overutilized. Nor is it a sign that inflation is on an accelerating path or even that wage norms are shifting up. In analyzing U.S. postwar data, I have found the wage norm shifted up substantially by the end of the 1960's and down again, though not by as much, by the end of the 1980-82 recession (see my 1983 paper). The first episode was a period with a historic expansion that ended with several years of very low unemployment rates. The second was a recession of unusual length and severity that ended with the highest unemployment rates since the 1930's. There is also evidence of a small shift down in the wage norm after the weak economic performance of 1957-61, which featured two recessions with only an aborted recovery in between. I also found evidence for Germany, the United Kingdom, and Japan of upward shifts in wage norms in manufacturing industries after the 1960's and downward shifts in the early 1980's (see my 1986 paper). All these episodes suggest the kinds of extreme cyclical developments that have shifted wage norms in the postwar period. *The Brookings Institution, 1775 Massachusetts Avenue, NW, Washington, D.C. 20036. In preparing this paper, I benefited from discussions with Charles Schnl t7.e

The turmoil that has characterized macroeconomics for at least a decade originated with the inflation that emerged in the late 1960's and persisted stubbornly throughout the 1970's. The inability of the neoclassical synthesis to model inflation convincingly spawned the new classical models. Because they infer macroeconomic results more directly from principles of maximizing behavior, they appeal to some as more rigorous. Many others reject them as irrelevant because neither the microeconomic behavior that these models postulate nor their macroeconomic implications are realistic. The central postulates are that all agents are price takers and that all markets clear in the sense of Walrasian auction markets. In a dynamic, or multiperiod, context this means markets clear in rationally expected future prices. The most controversial implications are policy ineffectiveness and, in some versions, the proposition that inflation could be eliminated with little cost in real output. As one of those who was unimpressed with these more provocative postulates and implications, I was distressed that anyone took them seriously, and that the profession became so divided over important policy issues. But it is also true that some more interesting ideas, that were originally linked with the auction-market model by Robert Lucas and other authors, might not have been developed without the controversy. One is that the rational expectations methodology should be applied in a thorough way to macroeconomic relations. Another is that private agents' behavior will depend on the rules governing the conduct of policy, or the policy regime. I am optimistic that we are on the verge of generating more light and less heat than we have been recently. The interesting new ideas and methodology that have developed in the course of the past decade's debates will continue to be explored. But inevitably, I believe, these and other ideas will be examined within the framework of an economy that in crucial ways does not operate with auctionlike markets. Both the lack of empirical success with the classical postulates and the intellectual challenge of developing a more general micro foundation to supplant the auction model are pushing in this direction. With this development, the gap between rigor and reality should narrow as researchers differ less about the basic postulates underlying macro models.