We mustn't have it too good. Too much growth too little unemployment is a bad thing. These are not the idle thoughts of economic nail-biters; they are the economic policy of the United States. After real growth of gross domestic product (GDP) hit 4.5 percent in the last quarter of 1994 and unemployment dipped to 5.4 percent in December, the Federal Reserve moved on February 1 to raise interest rates for the seventh time in less than a year. Why? To slow our too rapid rate of growth and stop or reverse the fall in unemployment. Why do that? To fight inflation.
Ordinary people may wonder. Overall inflation, as measured by the GDP implicit price deflator, was down to 2.1 percent, its lowest in three decades. The Consumer Price Index rose only 2.7 percent in 1994 and knowledgeable analysts, including the Fed's chairman, AlanGreenspan, recognize that this measure overstates the rise in consumer costs, perhaps by as much as two percentage points.
Hard-nosed economic analysts and business leaders are also raising questions. They point to technological advances and downsizing in U.S. industry and suggest that productivity and output potential may well be rising more rapidly than the 2.5 percent long-term growth rate that Greenspan and others think marks the outer limit for the economy. Further more, as people lose old, high-paying jobs and look desperately even for lower-paying employment, there is slack in the labor force. Perhaps most important, increasing globalization and world competition may limit the ability of American firms to raise prices and workers to push for higher wages.
These heretical observations have so far failed to dent the dominant dogma haunting economic policy. The central tenet of that dogma is a concept familiarly known among economists as the NAIRU the "nonaccelerating-inflation-rate of unemployment." While unknown to the general public, the NAIRU has become one of the most powerful influences on economic policy this century. My recent work, however, shows that even on the basis of a conventional model used to estimate the NAIRU, there is no basis for the conclusion that low unemployment rates threaten permanently accelerating inflation. And, according to an alternative model more consistent with the data, inflation might actually be lower at lower unemployment levels than we are experiencing today.
The NAIRU Framework
The basic proposition of the NAIRU is simple: Policymakers cannot use deficit spending or an increase in the money supply to reduce unemployment below some "equilibrium" rate, except at the cost of accelerating inflation. This is a sharp departure from the Keynesian view that inflation poses a danger only when increased spending or demand presses against full or near-full employment.
The concept of the NAIRU, derived from Milton Friedman's notion of a "natural rate of unemployment," rejects the assumed trade-off between unemployment and inflation described by the Phillips Curve, named after A.W. Phillips, an innovative economist from New Zealand. The Phillips Curve suggests that maintaining lower unemployment does produce higher inflation, but the inflation is constant. In the NAIRU view, the Phillips Curve is only a short-run relation. Trying to reduce unemployment by increasing spending or aggregate demand may work for a while, but then the higher inflation will cancel out the effects of the stimulus. Increased actual inflation will raise expectations of future inflation; only the excess of actual inflation over what workers, employers, borrowers, and lenders expect will stimulate the economy. At each round, higher spending and inflation will be necessary to maintain the original reduction in unemployment.
Thus, according to the NAIRU, fiscal or monetary policies aimed at reducing unemployment would leave us like a dog chasing its tail. If policy were aimed at keeping total spending sufficiently high to keep unemployment below its "natural rate," inflation would rise more and more rapidly. Ultimately, policymakers would give up in the face of runaway prices. Unemployment would then be back at its natural rate and inflation would stop accelerating, but it would stay at its new, higher level until unemployment rose above the natural rate and the process was painfully reversed.
In this view, the only way to reduce unemployment, except possibly in the short run, is to change conditions affecting the supply of labor for example, by cutting the minimum wage, reducing or eliminating unemployment benefits, or upgrading the skills of workers. If the NAIRU is taken seriously, supply-side measures are the only ways to get unemployment down and keep it down. And if unemployment is at or close to the NAIRU, the monetary authority must take prompt anti-inflationary action to prevent the economy from "overheating." Other wise, inflation will not only be higher but will be launched on its accelerating course, from which it can be diverted only by the medicine of excess unemployment that is, unemployment above the NAIRU.
This is the view that underlies the otherwise inexplicable policy of the Federal Reserve. Most of our central bankers believe that we are at the natural rate of unemployment or below it, and we need more unemployment before it is too late. The main difference among macroeconomists today is that conservatives tend to put the NAIRU higher, at say 6-plus or 7 percent, while liberals put it at 6 or perhaps 5-plus percent. A few brave souls suggest that since our estimates of the NAIRU are imprecise, we should cautiously try to bring down unemployment until we have signs of inflation. But others say by then it will be too late.
Few economists have challenged the basic concept of the NAIRU. Keynes observed six decades ago that economists could stubbornly stick to their assumptions in the face of crushing reality, as when they argued in the depths of the Clutch Plague that there could be no involuntary unemployment. Another such episode of professional obstinacy may well be unfolding. Business leaders report, and national statistics confirm, that despite unemployment falling below the conventional NAIRU, accelerating inflation is nowhere in sight. But many economists are unmoved by mere evidence.
The available data do not, in fact, show that the NAIRU has much to do with historical levels of unemployment. In the United States, as shown in the first figure, actual unemployment has bounced all around a NAIRU that was altered only slightly to keep up with it. Why, for example, did unemployment dip well below the NAIRU through most of the 1960s? The theory does not tell us why it was possible then but impossible now.
The conventional model could simply be ignoring many factors affecting inflation or the interaction of unemployment and inflation. These factors may also have a different impact when unemployment is high and when it is low. (See second figure.) When unemployment is high, workers may indeed hesitate to press for higher wages because they are worried about losing their jobs. Still higher unemployment and falling demand may lead to more competition for limited markets, which may further check inflation.
But when unemployment is low, inflation may also be held in check. Low unemployment is usually associated with more efficient use of all resources. Persistent low unemployment rates that might lead to higher wages may encourage the substitution of capital for labor and raise anticipated future productivity, which would curb inflation. And with profits high and overhead costs spread broadly, firms may keep down prices to discourage others from entering their markets. Firms that are flush with profits may consider moving into new areas. Firms already there may well hesitate to raise prices and thus offer greater invitation to would-be interlopers.
This is, of course, just a sketch of why low unemployment and the high profits usually associated with it may inhibit inflation. Thus, the relationship may be different from what is usually assumed. It may be true that high unemployment reduces inflation, while it is false that low unemployment raises inflation.
The Conventional Formulation
Two crucial assumptions are necessary to arrive at the usual concept of the NAIRU. The first is that, left to itself, any given rate of inflation is self-perpetuating; the second, that unemployment is a key factor in changing inflation rates specifically, that higher unemployment lowers inflation, and lower unemployment raises inflation.
There has been something of a cottage industry in estimating the NAIRU over the years. An exemplary case is the formulation by the Congressional Budget Office (CBO) in its August 1994 Economic and Budget Outlook: An Update, which is similar to influential work a decade earlier by Robert Gordon. The general idea is that inflation is a function of a number of variables such as presumably independent food and energy price movements, changes in productivity, the imposition and removal of price controls, and, most important, past inflation and current and past unemployment. The idea that inflation is self-perpetuating is embodied in the assumption that past inflation enters the equations with a coefficient of one. The formulation then has an estimated constant term which is positive, pushing inflation up and negative coefficients of unemployment to hold inflation down. The size of those negative coefficients determines how much unemployment will be necessary to keep inflation from increasing. The rate of unemployment just sufficient to do this is the NAIRU.
I have replicated the CBO estimates and have confirmed the agency's results using its own model. The sum of the past inflation coefficients is at or above that crucial value of unity necessary for inflation to be self-perpetuating unless stopped. The constant terms are positive and the sums of the unemployment coefficients negative. My estimates yield a NAIRU at just about CBO's figure of 5.8 percent. (The measure of unemployment used by the CBO is the unemployment rate for married men, which it then adjusts to estimate the general rate of unemployment.)
However, even this model does not support some of the implications usually drawn for policymaking. Many economists argue that we must never let the genie of inflation out of the bag because even a brief, inflation-accelerating experience of low unemployment will be disastrous and difficult to correct. Testing that proposition, I found that a one percentage point reduction in the married-male unemployment rate to 2.55 percent (one percentage point below the CBO estimate of the married-male NAIRU) generates a sharp increase and fluctuation in CPI inflation for several quarters, which subsides quickly if unemployment goes back up to the NAIRU. Even permanent unemployment of 2.55 percent does not, after five years, get inflation past 7 percent.
These results are based on the conventional formulation, but that is only the beginning of the story. The conventional model constrains the unemployment and inflation parameters in ways that are in fundamental conflict with the data. Freeing the model from those constraints leads to dramatically different conclusions; this calls into question the use of the NAIRU as a justification for blocking fiscal and monetary policies that might bring "full employment," or distinctly lower unemployment than what is now widely viewed as acceptable.
An Alternative Model
My reformulation of the conventional model suggests that the effect of unemployment on inflation is different when unemployment is low compared to when it is high. The key question, then, is what happens to the estimated values of the unemployment coefficients when unemployment is low. Do they differ consistently from the coefficients when unemployment is high?
First, estimates of separate relations for high and low unemployment show that differences between the unemployment coefficients are clearly statistically significant.
Second, the unemployment coefficients in the low-unemployment regressions are generally positive, though usually modest in size. This suggests that, whatever the effect on inflation of unemployment below the NAIRU, once below the NAIRU, lowering unemployment further may reduce inflation.
Third, under low unemployment, the sums of inflation coefficients were below unity, contradicting a critical assumption underlying the NAIRU. Inflation left to itself would not be self- perpetuating, and low unemployment would not cause accelerating inflation. Even if unemployment below the NAIRU did raise inflation, it would raise it by a finite amount the old Phillips-Curve relation, not permanently accelerating inflation.
Simulations and Forecasts
One way to reveal the effects of the various interacting coefficients is to simulate or forecast ahead. I show results based on a single equation for inflation in the consumer price index. The high-unemployment inflation paths in the figure on page 61 fit the conventional view. Unemployment above the NAIRU drives inflation down, although the implicit NAIRU is closer to 6.8 percent in my simulations based only on high-unemployment observations. It takes still higher unemployment to break the back of inflation. But high enough unemployment does eventually turn inflation negative; that is, it drives prices down.
The low-unemployment paths shown, however, offer quite a different picture. At 5.8 percent unemployment, contrary to Alan Greenspan's fears, there is no accelerating inflation. By the end of the century, inflation settles at about 4.4 percent. Strikingly, at lower unemployment rates, inflation is no higher. At 4.8 percent unemployment, the simulation shows inflation coming down to 3.6 percent. At 3.8 percent unemployment, inflation comes down to 2.9 percent. At 2.8 percent unemployment, inflation at the end of 1999 is down to 2.1 percent.
I would not bet the family farm or the nation's economy on any set of econometric estimates, even my own. But promoters, defenders, and practitioners of the conventional NAIRU have done exactly that, with increasingly dogmatic assertion. They have paralyzed macroeconomic policy that should be aimed at the "high" and "full" employment targets set by the Employment Act of 1946 and the Humphrey-Hawkins Full Employment and Balanced Growth Act of 1978.
If accelerating inflation is not our fate, some might think a few extra percentage points of constant inflation might offer a pretty good bargain. Lower unemployment would generate large increases in output. According to the robust Okun's Law, named after Arthur Okun, the late Yale economist, each percentage point of unemployment costs at least two percentage points of output. That would amount to more than $130 billion of GDP this year.
Those committed to the concept of a NAIRU cannot easily dismiss the evidence of asymmetry that I have presented. I am not proposing a new dogma that lowering unemployment will reduce inflation. Even if my formulation is right, my standard errors are often too high as, I should add, are those of practitioners of the conventional model to permit any precise conclusions. There may be no stable, universal relation among unemployment and all the various factors contributing to inflation.
But the results reported here should clearly show the lack of empirical support for the NAIRU and the policies based upon it. They suggest that we have no sound basis for deliberately raising unemployment. On the contrary, we ought to be trying to reduce it, not only by supply-side measures, but by ensuring that the economy is not starved for adequate aggregate demand or productivity- increasing public investment. These measures should aim at reducing both underlying structural unemployment and the unemployment caused by misguided anti-inflation policy. The fight against inflation can then be focused where it should be on promoting the greatest measure of domestic and international competition.