FRBSF Economic Letter
1997-35 | November 21, 1997
In recent years, a debate has re-emerged about whether the Federal Reserve should pay attention to the “NAIRU” in conducting monetary policy. NAIRU is an acronym for “non-accelerating-inflation rate of unemployment” (a closely related concept is the “natural rate of unemployment”). The NAIRU figures prominently in the Phillips curve, which is a relationship that incorporates a temporary trade-off between the unemployment rate and inflation. According to the Phillips curve, an unemployment rate that is below the level identified as the NAIRU (that is, a “tight” labor market) tends to be associated with an increase in inflation; conversely, an unemployment rate that is above the NAIRU tends to be associated with a decrease in inflation. It is well known that the trade-off between inflation and unemployment is only temporary and cannot be systematically exploited by monetary policies aimed at permanently lowering the unemployment rate. In the long run, attempts to do so end up generating higher inflation with no improvement in unemployment. However, the Phillips curve also implies that demand-induced changes in inflation tend to lag behind movements in the unemployment rate, which means that a comparison between the actual unemployment rate and the NAIRU may be helpful in forecasting future changes in inflation.
Tight labor (and product) markets were one reason for the Fed’s “preemptive strike” against inflation in 1994 (see Judd and Trehan 1995). The federal funds rate was raised from 3% in early 1994 to 6% in early 1995 without actual increases in broad measures of inflation, like the CPI. This action was explained as a response to indications that inflation would rise in the future without policy action. Over the past year, however, the funds rate has not been raised despite a fall in the unemployment rate to 4-3/4% – 5, below most estimates of the NAIRU. Some people have argued that policy action should be taken to prevent an upward creep in inflation, while others have asserted that there is no inflation threat on the horizon.
These recent experiences have stimulated the current debate about the NAIRU, with some economists arguing that it provides useful information for monetary policy and others arguing that it is dangerously misleading (Journal of Economic Perspectives 1997). This Letter discusses the key elements in this controversy.
The lags in monetary policy present a problem for central banks, because a policy action taken today may not affect inflation for a year or two. Therefore, in attempting to control inflation, it is dangerous to look only at current rates of inflation. By the time inflation actually begins to rise, inflationary pressures may have been brewing for a year or two, and it may take a substantial tightening of policy (possibly leading to a recession) to head them off. The lag in policy explains why most central banks expend considerable effort in forecasting future economic developments. In fact, some central banks (for example, those in the United Kingdom, Canada, and New Zealand) use publicly announced forecasts as a key element in the formulation of their policies.
According to models of the economy that incorporate a Phillips curve, the unemployment rate plays a role in the transmission process from unanticipated changes in the aggregate demand for goods and services (called “demand shocks”) to inflation. In these models, increases in demand raise real GDP relative to its potential level, which increases the demand for labor to produce the additional goods and services, and therefore lowers the unemployment rate relative to the NAIRU. Excess demand in goods and labor markets leads to higher inflation in goods prices and wages with a lag. Because of this, the unemployment rate can help in generating the inflation forecasts that are crucial in formulating monetary policy.
Critics of using the NAIRU concept to guide policy raise both empirical and theoretical arguments. On the empirical side, they point out that the estimated NAIRU for the U.S. has varied in the postwar period. In the 1960s, the NAIRU commonly was estimated at around 5%. By the mid-1970s, it had climbed to around 7%. And by the mid-1990s, it had fallen back to 5 1/2 to 6% (Staiger, Stock, and Watson 1997). A number of factors can affect the NAIRU, including changes in labor force demographics, governmental unemployment programs, and regional economic disturbances.
A related empirical criticism is that the NAIRU cannot be estimated with much precision. Based upon comprehensive empirical analysis of Phillips curves, Staiger, Stock, and Watson conclude that their best fitting equation yields a 95% probability that the NAIRU falls within a range of 4.8 to 6.6%. Given this kind of uncertainty, the NAIRU can provide misleading signals for monetary policy at various times.
A theoretical objection to the use of the NAIRU for monetary policy is that the short-run trade-off between unemployment and inflation may be unstable over time. This trade-off is sensitive to the way in which expectations about inflation are formed, which in turn will depend upon the nature of the monetary policy regime itself. As noted above, for example, any trade-off would tend to disappear if a central bank attempted to exploit it systematically.
A further theoretic objection –one which has been discussed a lot recently–is that the NAIRU makes sense as an indicator of future inflation only when the economy is hit with demand shocks, like those described above for the Phillips curve model (Judd and Trehan 1990 and Chang 1997). However, the economy also may be affected by supply shocks, or unexpected changes in the aggregate supply of goods and services. An example of a supply shock would be a sudden increase in productivity. Initially, this kind of shock would raise the quantity of goods and services produced relative to the quantity demanded, and thus put downward pressure on prices. At the same time, the increase in real GDP would raise the demand for labor and reduce the unemployment rate. Thus, a falling unemployment rate would be associated with reduced pressure on prices. If a central bank were using the NAIRU to guide policy in this case, it might mistakenly see the lower unemployment rate as a reason to fear higher inflation in the future, and therefore might tighten policy.
Some observers argue that a supply shock is currently having an effect on the economy. Over the past couple of years, real GDP has increased rapidly, and the unemployment rate has fallen to a low rate of 4 3/4% – 5%, while inflation has come down a bit. Therefore, standard Phillips curves have over-forecasted inflation recently, although the errors generally have not been outside the historical range of errors. One explanation offered for recent developments is a surge in productivity due to the introduction of new computer-related technologies. While it is still too soon to know for sure what is driving recent developments, the possibility of a supply shock has to be taken seriously. This possibility illustrates the pitfalls in interpreting the implications of the unemployment rate for future inflation. At the same time, however, it is too soon to be sure that the current low level of the unemployment rate does not presage a rise in inflation in the future.
The arguments presented above have been used to criticize what could be called a “trigger” strategy, in which the central bank would compare the unemployment rate to the latest estimate of the NAIRU and change the funds rate according to whether inflation was predicted to rise or fall in the future. This criticism of such a trigger strategy is well founded. However, it is doubtful that any central bank would base policy on such a simple response to any single variable.
A more relevant question is how forecasting models that incorporate the NAIRU concept perform relative to alternative models. Since all forecasting models are subject to error, the practical issue for central banks is which type of model provides the best forecasts. In other words, it is not enough to show that the NAIRU-based models are subject to error. It is also necessary to show that the uncertainties associated with them are bigger than those of alternative models.
The alternative models that have been used for this purpose include monetarist models that rely mainly on a measure of the money supply to forecast inflation, and vector autoregressions (VARs) that produce purely statistical forecasts without relying on any theory concerning what causes inflation. Both of these alternatives have drawbacks.
Since inflation is a monetary phenomenon, monetary models have an obvious theoretical advantage in forecasting inflation. However, the empirical problems with the monetary aggregates over the past 15 to 20 years are well known. In the early 1980s the Fed relied heavily on M1, a narrow aggregate; by the mid-1980s, however, M1’s relationship with real GDP and inflation became too uncertain, and the Fed de-emphasized it in favor of the broader aggregates, M2 and M3. These aggregates retained some reliability until the 1990s when they began to experience serious problems. The main difficulty with all of these aggregates appears to have been the deregulation of the financial system in the 1970s and 1980s and the rapid financial innovation that has been going on in the U.S. and world economies for the past two decades or so. These difficulties help explain the results of studies by Stockton and Struckmeyer (1989) and Tallman (1995), which have found forecasting advantages with Phillips curve models compared with monetarist models, although both approaches involved considerable uncertainty.
With regard to VARs, it is well known that they do a good job of forecasting real GDP, but have more problems forecasting inflation (McNees 1986). The reliability of VARs appears to be particularly vulnerable to major changes in inflation regimes, such as the ones in the U.S. in the 1960s and late 1970s (Webb 1995).
Models that can forecast inflation are valuable to central bankers because monetary policy actions affect inflation with a lag. Models that incorporate a NAIRU concept have problems as forecasting devices, especially if the economy is hit with a supply shock. The current situation may be an example of such a case. Recent forecast errors, though not especially large by historical standards, nonetheless may provide a rationale for some de-emphasis of the unemployment rate in policy deliberations. However, it is not clear that monetary models or VARs provide superior alternatives to NAIRU-based models. This consideration may help to explain the continued use of NAIRU-based models by many policymakers, despite well-known conceptual and empirical shortcomings.
As economists continue to work on these problems, advances in modeling may provide better alternatives. For example, Chang suggests that models along the lines of those developed by Bernanke (1986), which explicitly attempt to decompose relevant data into demand and supply shocks, might be useful. An important test of the usefulness of such models for monetary policy would be whether they offer advantages in forecasting inflation.
John P. Judd
Vice President and Associate Director of Research
Bernanke, Ben S. 1986. “Alternative Explanations of the Money-Income Correlation.” Carnegie-Rochester Conference Series on Public Policy pp. 49-99.
Chang, Roberto. 1997. “Is Low Unemployment Inflationary?” Federal Reserve Bank of Atlanta Economic Review, First Quarter, pp. 4-13.
Journal of Economic Perspectives. 1997. “Symposium: The Natural Rate of Unemployment” (Winter) pp. 3-108.
Judd, John P., and Bharat Trehan. 1995. “Has the Fed Gotten Tougher on Inflation?” FRBSF Economic Letter No. 95-13 (March 31).
_______, and ______. 1990. “What Does Unemployment Tell Us about Future Inflation?” Federal Reserve Bank of San Francisco Economic Review (Summer) pp. 20-37.
McNees, S. K. 1986. “The Accuracy of Two Forecasting Techniques: Some Evidence and an Interpretation.” New England Economic Review (March) pp. 20-31.
Staiger, Douglas, James H. Stock, and Mark W. Watson. 1997. “The NAIRU, Unemployment and Monetary Policy.” Journal of Economic Perspectives (Winter) pp. 33-49.
Stockton, David J., and Charles S. Struckmeyer. 1989. “Tests of the Specification and Predictive Accuracy of Nonnested Models of Inflation.” Review of Economics and Statistics pp. 275-283.
Tallman, Ellis W. 1995. “Inflation and Inflation Forecasting: An Introduction.” Federal Reserve Bank of Atlanta Economic Review (January/February) pp. 13-27.
Webb, Roy H. 1995. “Inflation Forecasts from VAR Models.” Journal of Forecasting pp. 267-285.
Opinions expressed in FRBSF Economic Letter do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco or of the Board of Governors of the Federal Reserve System. This publication is edited by Sam Zuckerman and Anita Todd. Permission to reprint must be obtained in writing.
Please send editorial comments and requests for reprint permission to
Attn: Research publications, MS 1140
Federal Reserve Bank of San Francisco
P.O. Box 7702
San Francisco, CA 94120