[An updated version of this article can be found at Phillips Curve in the 2nd edition.]
The Phillips curve represents the relationship between the rate of inflation and the unemployment rate. Although several people had made similar observations before him, A. W. H. Phillips published a study in 1958 that represented a milestone in the development of macroeconomics. Phillips discovered that there was a consistent inverse, or negative, relationship between the rate of wage inflation and the rate of unemployment in the United Kingdom from 1861 to 1957. When unemployment was high, wages increased slowly; when unemployment was low, wages rose rapidly. The only important exception was during the period of volatile inflation between the two world wars.
In Phillips's analysis, when the unemployment rate was low, the labor market was tight and employers had to offer higher wages to attract scarce labor. At higher rates of unemployment there was less pressure to increase wages. Phillips's "curve" represented the average relationship between unemployment and wage behavior over the business cycle. It showed the rate of wage inflation that would result if a particular level of unemployment persisted for some time. Significantly, however, the relationship between wages and unemployment changed over the course of the business cycle. When the economy was expanding, firms would raise wages faster than "normal" for a given level of unemployment; when the economy was contracting, they would raise wages more slowly than "normal."
Economists soon estimated Phillips curves for most developed economies. Because the prices a company charges are closely connected to the wages it pays, economists also frequently used Phillips curves to relate general price inflation (as opposed to wage inflation) to unemployment rates. Chart 1 shows a typical Phillips curve fitted to data for the United States from 1961 to 1969. The individual observations appear to lie closely along the fitted curve, indicating that the cyclical behavior of inflation and unemployment is similar to the average behavior. That is, the relationship between inflation and unemployment does not seem to change much over the course of the business cycle.
Chart 1. The Phillips Curve: 1961-69
Enlarge in new window
This observation encouraged many economists, following the lead of Paul Samuelson and Robert Solow in 1960, to treat the Phillips curve as a sort of menu of policy trade-offs. For example, with an unemployment rate of 6.5 percent, the government might stimulate the economy to lower unemployment to 5.5 percent. Chart 1 indicates that this would entail a cost, in terms of higher inflation, of less than 0.5 percentage point. But if the government initially faced lower rates of unemployment, the costs would be considerably higher: a reduction of unemployment from 4.5 to 3.5 percent is associated with an increase in the inflation rate of about 2 percentage points.
At the height of the Phillips curve's popularity as a guide to policy, Edmund Phelps and Milton Friedman independently challenged its theoretical underpinnings. They argued that well-informed, rational employers and workers would pay attention only to real wages—the inflation-adjusted purchasing power of money wages. In their view, real wages would adjust to make the supply of labor equal to the demand for labor, and the unemployment rate would then stand at a level uniquely associated with that real wage. This level of unemployment they called the "natural rate" of unemployment.
In Friedman's and Phelps's view the government could not make a permanent trade-off between unemployment rates and inflation rates, as the Phillips curve in chart 1 suggests. Imagine that unemployment is at the natural rate. The real wage is constant: workers who expect a given rate of price inflation insist that their wages increase at the same rate to prevent the erosion of their purchasing power. Now imagine that the government uses expansionary monetary or fiscal policy in an attempt to lower unemployment below its natural rate. The resulting increase in demand encourages firms to raise their prices faster than workers had anticipated. With higher revenues firms are willing to employ more workers at the old wage rates, and even to raise those rates somewhat. For a short time workers suffer from what economists call money illusion: they see that their money wages have risen, and willingly supply more labor. Thus the unemployment rate falls. They do not realize right away that their purchasing power has fallen because prices have risen more rapidly than they expected. But over time, as workers come to anticipate higher rates of price inflation, they supply less labor and insist on increases in wages that keep up with inflation. The real wage is restored to its old level, and the unemployment rate returns to the natural rate. But the price inflation and wage inflation brought on by expansionary policies continue at the new, higher rates.
Friedman's and Phelps's analysis provides a distinction between the "short-run" and "long-run" Phillips curves. So long as inflation remains fairly constant, as it did in the sixties, inflation is inversely related to unemployment. When the average rate of inflation changes, however, unemployment returns after a period of adjustment to the natural rate. That is, once worker expectations of price inflation have had time to adjust, the natural rate of unemployment is compatible with any rate of inflation. This long-run relation could be shown in chart 1 as a vertical line above the natural rate of unemployment. In other words, once unemployment falls to the natural rate, expansionary policies will not push it any lower except for brief, transitional periods. These long-run and short-run relations can be combined in a single "expectations-augmented" Phillips curve. The more quickly worker expectations of price inflation adapt to changes in the actual rate of inflation, the more quickly unemployment will return to the natural rate, and the less successful the government will be in reducing unemployment through monetary and fiscal policy.
The seventies provided striking confirmation of Friedman's and Phelps's fundamental point. The average inflation rate rose from about 2.5 percent in the sixties to about 7 percent in the seventies, while average unemployment rose from about 4.75 percent to about 6 percent. Thus, contrary to the original Phillips curve, higher inflation was associated with higher—not lower—unemployment.
Most economists now accept a central tenet of the Friedman-Phelps analysis: there is some rate of unemployment that, if maintained, would be compatible with a constant rate of inflation. Many, however, prefer to call this the "nonaccelerating inflation rate of unemployment" (NAIRU), because unlike the term "natural rate," it does not suggest an unchanging unemployment rate to which the economy inevitably returns, which policy cannot alter, and which is somehow socially optimal.
A policymaker might wish to place a value on NAIRU. To obtain a simple estimate, chart 2 plots changes in the rate of inflation (i.e., the acceleration of prices) against the level of unemployment from 1974 to 1990. The regression line (i.e., the straight line that best fits the points on the graph) summarizes the rough, inverse relationship. According to the regression line, NAIRU (i.e., the rate of unemployment for which the change in the rate of inflation is zero) is about 7 percent. The slope of the regression line indicates the speed of price adjustment. Imagine that the economy is at NAIRU with an inflation rate of 4.5 percent, and that the government would like to reduce the inflation rate to zero. Chart 2 suggests that contractionary monetary and fiscal policies that drove the average rate of unemployment up to about 8 percent (i.e., 1 point above NAIRU) would be associated with a reduction in inflation of about 1.5 percentage points per year. Thus, if the unemployment rate were held at about 8 percent, the inflation rate of 4.5 percent would, on average, be reduced to zero in three years.
Chart 2. Acceleration of Prices versus the Unemployment Rate: 1974-90
Enlarge in new window
Using similar methods, estimates place NAIRU at about 5.25 percent for the twenty years before 1974 and sharply higher—at about 7 percent—after 1974. Clearly, NAIRU is not constant. It is consistent with the natural rate hypothesis for NAIRU to vary (even by considerable amounts) as a result of changes in demographics, technology, the structure of labor markets, the structure of taxation, relative prices (e.g., oil prices), and other so-called "real" factors affecting the supply of and demand for labor. But monetary and fiscal policy, which affect aggregate demand without altering these real factors, should not change the natural rate of unemployment.
The expectations-augmented Phillips curve is a fundamental element of almost every macroeconomic forecasting model now used by government and business. Nonetheless, two criticisms of the expectations-augmented Phillips curve deserve notice.
First, economists of the new classical school argue that people form expectations rationally. According to the new classicals, people use information efficiently, so that they find ways to eliminate every systematic mistake in their predictions. For example, if people systematically underpredicted inflation by, say, 2 percent for several years, they could simply add 2 percentage points to their forecasts to obtain more accurate results. After eliminating every systematic source of error, any remaining mistakes must be unsystematic. That is, they must be random, or inherently unpredictable, so that people are as likely to overpredict as to underpredict inflation.
The "rational expectations" hypothesis says, in effect, that people's expectations adapt so rapidly that a government using expansionary monetary and fiscal policy to engineer a higher rate of price inflation cannot consistently push the unemployment rate below the natural rate. People will catch on too fast and demand higher wages whenever policy becomes more expansionary.
Although the rational expectations hypothesis enjoys wide currency among economic theorists, it has not affected the manner in which the expectations-augmented Phillips curve is incorporated into large macroeconomic forecasting models. This is not surprising. An expectations-augmented Phillips curve with rational expectations implies that the economy is always very close to NAIRU. But as is obvious from chart 2, unemployment has fluctuated widely over the years. This could be consistent with the new classical view only if NAIRU itself fluctuated as much as actual unemployment. And no convincing explanation has been given of why such fluctuations would occur.
Some "new Keynesian" economists offer a second criticism. They argue that there is no natural rate of unemployment in the sense of a rate to which the actual rate tends to return. Instead, when actual unemployment rises and remains high for some time, NAIRU rises as well. The dependence of NAIRU on actual unemployment is known as the hysteresis hypothesis. One explanation for hysteresis in a heavily unionized economy is that unions directly represent the interests only of those who are currently employed. Unionization undermines the ability of those outside the union to compete for employment. After prolonged layoffs employed workers inside the union may seek the benefits of higher wages for themselves, rather than moderating wage demands to promote the rehiring of unemployed workers. The downside to the hysteresis hypothesis is that once unemployment becomes high—as it did in Europe in the recessions of the seventies—it is relatively impervious to monetary and fiscal stimulus, even in the short run. The hysteresis hypothesis appears to be more relevant to Europe, where unionization is higher and where labor laws create numerous barriers to hiring and firing, than it is to the United States, with its considerably more flexible labor markets.
The Phillips curve was hailed in the sixties as providing an account of the inflation process hitherto missing from the conventional macroeconomic model. After three decades the Phillips curve, as transformed by the natural rate hypothesis into its expectations-augmented version, remains the key to relating unemployment and inflation in mainstream macroeconomic analysis.
Kevin D. Hoover is an economics professor at the University of California at Davis. In 1991 and 1992 he was a fellow at the National Humanities Center. He has been on the board of editors of the American Economic Review since 1990.
Cross, Rod, ed. Unemployment, Hysteresis, and the Natural Rate of Unemployment. 1988.
Friedman, Milton. "The Role of Monetary Policy." American Economic Review 58, no. 1 (1968): 1-17.
Lucas, Robert E., Jr. "Econometric Testing of the Natural Rate Hypothesis." In The Econometrics of Price Determination, edited by Otto Eckstein. 1972. Reprinted in Lucas, Robert E., Jr. Studies in Business Cycle Theory. 1981.
Phelps, Edmund S. "Phillips Curves, Expectations of Inflation and Optimal Employment over Time." Economica NS 34, no. 3 (1967): 254-81.
Phillips, A. W. H. "The Relation between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861-1957." Economica NS 25, no. 2 (1958): 283-99.
Samuelson, Paul A., and Robert M. Solow. "Analytical Aspects of Anti-Inflation Policy." American Economic Review 50, no. 2 (1960): 177-94.
Sheffrin, Steven M. Rational Expectations. 1983.