This Selected Issues paper on the United States analyzes the measures of potential output, natural rate of unemployment, and capacity utilization. Traditionally, measures of resource utilization have been used as indicators for the potential build-up of inflation pressures, and hence as guides for the formulation of macroeconomic policy. The paper highlights that the most commonly used indicators of resource utilization in the United States are the output gap, the employment gap, and capacity utilization in industry. The paper also analyzes the wage and price determination and productivity trends in the United States.


This Selected Issues paper on the United States analyzes the measures of potential output, natural rate of unemployment, and capacity utilization. Traditionally, measures of resource utilization have been used as indicators for the potential build-up of inflation pressures, and hence as guides for the formulation of macroeconomic policy. The paper highlights that the most commonly used indicators of resource utilization in the United States are the output gap, the employment gap, and capacity utilization in industry. The paper also analyzes the wage and price determination and productivity trends in the United States.

I. Measures Of Potential Output, Nairu, And Capacity Utilization 1

1. Traditionally, measures of resource utilization have been used as indicators for the potential build up of inflationary pressures, and hence as guides for the formulation of macroeconomic policy. The most commonly used indicators of resource utilization in the United States are the output gap (measured as the difference between actual and potential output), the employment gap (measured as the difference between the natural rate of unemployment—or NAIRU—and actual unemployment), and capacity utilization in industry (Figure 1). To varying degrees, all of these indicators are difficult to estimate with a high degree of precision. Over the course of the current expansion, each of them has at times suggested that inflationary pressures might begin to emerge, but inflation has remained remarkably quiescent. In light of such favorable inflation performance, estimates of the traditional measures of resource utilization have been reexamined.

Figure 1.
Figure 1.

United States: Indicators of Resource Utilization

Citation: IMF Staff Country Reports 1999, 101; 10.5089/9781451839579.002.A001

Sources: Bureau of Economic Analysis, Survey of Current Business; Bureau of Labor Statistics, Employment and Earnings; Board of Governors of the Federal Reserve; and Fund staff estimates.1/ Actual GDP minus potential GDP as a share of potential.2/ Based on the average of the range for potential output growth, which is 2.6 percent.3/ NAIRU minus unemployment rate.

2. Using a variety of techniques, revised estimates suggest that the annual growth rate of potential output is in the range of 2½ to 2¾ percent for the period 1990–98, roughly ¼ to ½ of 1 percentage point higher than most previous estimates. With stronger growth in potential than previously estimated, the resulting output gap in 1998 was between −¾ to 1¼ percent. New estimates for NAIRU suggest that it has declined from about 7 percent in the 1980s to about 4¼ to 5¾ percent in recent years. Therefore, with regard to the employment gap, the current rate of unemployment of 4¼ percent lies below nearly all of these estimates. With both the output and employment gaps suggesting that the economy is operating at a very high level of resource utilization, the absence of inflationary pressures is striking. The diminished predictive power of the traditional leading indicators for inflation may in part reflect the large possible margins of error in measuring potential output and the NAIRU. In addition, it may also reflect the effects of the adaptation of new technologies and changes in the structure of the economy that may have boosted (at least on a transitory basis) productivity growth. The supposed “breakdown” in the relationships between these measures and inflation is also related to the favorable price shocks experienced by the United States in recent years that have helped to hold inflation down.

A. Potential Output Growth and the Output Gap

3. There is a wide variety of methodologies for estimating potential output, ranging from atheoretical approaches, involving various detrending techniques, to more structural methods, such as the production function approach. Since none of the commonly used methods is free from difficulties, four different methodologies were used to determine a range of reasonable estimates of potential growth. The results suggest that the growth rate in the recent period is roughly 2½ to 2¾ percent (Table 1).

Table 1.

Estimates of Potential Output Growth

article image

4. Among the detrending techniques, the segmented trend approach attempts to identify points at which the trend rate of growth in GDP may have shifted. The potential rate of growth is assumed to be constant and roughly equivalent to the average growth rate over the interval between break points. Recursive residual tests were used to identify break points in the chain-linked real GDP series over the period 1959 to 1998. Two break points were found; one occurring in the first quarter of 1975, corresponding to just after the first oil price shock, and the other in the first quarter of 1982. Using these breakpoints, and based on log-linear regressions, potential output growth is estimated to have slowed from about 4 percent during the period 1960–75, to about 3¼ percent during the period 1975–82, and to 2¾ percent in the period thereafter.2

5. The Hodrick-Prescott (H-P) filter also was used to detrend GDP. This technique identifies a trend output which minimizes a weighted average of the gap between output and trend output and the rate of change in trend output.3 Although the H-P filter is less restrictive than the segmented-trend approach—in that the growth rate of potential can vary continuously—one disadvantage is that the end points of the H-P filtered trend output series tend to be quite sensitive to the last few observations in the series. To attempt to handle this problem, potential output was estimated to be about 2¾ percent per year over the period from the peak in output in the fourth quarter of 1973 to the end of 1998.

6. Another technique for estimating the growth rate of potential output is the Blanchard-Quah bivariate decomposition, in which output is divided into its cyclical and trend components. Rather than use only the information contained in the real output series—as in the case of the H-P filter—this technique also incorporates additional information from highly cyclical aggregate variables such as consumption and the unemployment rate.4 This approach allows for a stochastic trend of output without forcing the trend component to be modeled as a random walk. Thus, it is consistent with the widely held belief that the dynamics of the permanent component of output are driven partly by technological innovation. Based on the Blanchard-Quah decomposition, potential output growth was estimated to be about 2½ percent for the period 1990–98.

7. The main drawback to the detrending techniques is that they are mechanistic, in the sense that the productive limits of the economy are not estimated based on available factors of production. In contrast the production function approach explicitly models output in terms of underlying factors of production, expressing output as a function of capital, labor, and total factor productivity (TFP). This approach requires the assumption of a functional form for the aggregate production function and the construction of series for potential capital, potential labor, and TFP. Following established practice, a constant returns to scale Cobb-Douglas type production function was assumed, with constant shares over time for labor and capital 5

8. The series of potential inputs and TFP were estimated using three different methods. The first method assumed that factor inputs and total factor productivity were at their potential level in the years 1981 and 1990, which were cyclical peaks. The growth rate of TFP, as well as that of potential capital stock was assumed equal to the growth rate between those peak years. Potential labor was estimated to grow at the same rate as the historical population growth rate. The second method estimated the trend growth rate of TFP as in the previous method, but the series for potential inputs were obtained using the H-P filter, Finally, the third method extracted the potential series of factor inputs and trend TFP using the H-P filter. The growth rate of potential was estimated to be about 2.9 percent using the first two methods and 2.6 percent with the third method over the period 1990–98.

9. Based on these four different techniques for estimating the growth rate of potential output, 2½ to 2¾ was chosen as a range of reasonable estimates on which to establish the level of potential output and, therefore, the output gap (Figure 2). The level of potential output was determined by applying the annualized growth rate of potential to the full employment level of output which occurred in the third quarter of 1990. The output gap derived from this potential output series was in the range of -¾ ot 1¼ percent in 1998.

Figure 2.
Figure 2.

United States: Potential Output and Output Gaps

Citation: IMF Staff Country Reports 1999, 101; 10.5089/9781451839579.002.A001

Source: Staff estimates.

B. NAIRU and the Employment Gap

10. The NAIRU is the rate of unemployment that would keep the rate of inflation constant, and the employment gap is simply the difference between NAIRU and the actual unemployment rate. These concepts are derived from the Phillips curve, which captures the inverse relationship between inflation and unemployment. Empirically, the rate of inflation has tended to increase (decrease) when the rate of unemployment lies below (above) NAIRU. For the past several years, the U.S. unemployment rate has remained below most estimates of NAIRU, but inflation has remained quiescent. This puzzle has led economists to hypothesize that changes in the U.S. economy have reduced the NAIRU allowing the economy to reach a lower level of unemployment without sparking inflation. More recent estimates suggest that NAIRU has declined to 4½-5¾ percent, from a peak of about 7 percent in the early 1980s (Table 2). 6

Table 2.

Recent Estimates of NAIRU

article image

Estimate assumes that discrete jump in NAIRU occurred in 1993, and that the level of NAIRU remained unchanged thereafter.

Range represents the tightest of the 95 percent confidence intervals estimates; point estimate is 5.8 percent.

The lower estimate is based on an equation using the personal consumption deflator, whereas the higher estimates is based on the GDP deflator.

11. Following Adams and Coe (1990), NAIRU was calculated using an approach in which cyclical and structural variables were used to explain the unemployment rate. NAIRU is derived from the estimated long-run values of the employment-population ratio (E/P) and the labor force participation rate (L/P), that is:

NAIRU = 1 - [(E/P)/(L/P)]

12. The long-run employment-population ratio was estimated as a function of two structural variables—the unionization rate, and the minimum wage, both of which are negatively correlated with the dependent variable—and two cyclical variables—the output gap and the wage gap (which is the change in real compensation per hour relative to output per hour in the nonfarm business sector), both of which are positively correlated with the dependent variable. The long-run participation rate was estimated as a function of a nonlinear time trend, the child-dependency ratio, which is negatively correlated with the dependent variable, and the same two cyclical variables. Based on this approach, NAIRU was estimated to be about 5 percent in 1998.

13. Another approach to deriving NAIRU is to use Okun’s Law, which establishes an empirical relationship between the output gap and the employment gap.7 Based on the staffs estimate of an output gap of -¾ to 1¼ percent in 1998, an actual unemployment rate of 4½ percent, and an Okun coefficient of 2½, NAIRU is computed to be in the range 4¼ to 5 percent,

14. For purposes of comparison, recently published estimates of NAIRU based on different methodologies are included in Table 2. The Congressional Budget Office (CBO) estimate of NAIRU is based on a Phillips-curve equation, which relates the inflation rate to lagged inflation, lagged levels of the unemployment rate, productivity growth, and variables to control for changes in food and energy prices. To derive NAIRU, the estimated equation was solved for the rate of unemployment that would deliver constant inflation. This method yields an estimate of 5½ percent in 1998. From its own augmented Phillips curve, the staff derives an estimate for NAIRU of 4½ percent in 1998.8

15. Other Phillips-curve estimation approaches allow NAIRU to vary over time, rather than to produce a single point estimate, and are therefore designed to track changes in NARIU. For example, Murphy (1998) uses a discrete jump approach. Phillips curves are estimated for subperiods during 1960–97. For each subperiod, a level of unemployment that delivers constant inflation is derived. Murphy finds that NAIRU was about 5¼ percent for the period 1960–72, increased to 6½ percent during 1973–85, and to 7 percent in 1986–92, and then declined to about 5¾ percent for the subperiod 1993–97.

16. Staiger, Stock, and Watson (1997) and Gordon (1998) use a more flexible approach, which allows NAIRU to vary continuously over time, rather than at specific break points. The Phillips curve is estimated jointly with a second equation that allows NAIRU to vary over time. The Staiger, Stock, and Watson results confirm that NAIRU declined from about 7 percent in the mid-1980s, to about 5¾ percent in 1994. For this estimate, the confidence interval ranges from 4¾ to 6½ percent in the best case, and between 2¾ to 7¾ percent, in the worst case. Gordon’s results indicate that NAIRU declined from about 6¼ percent in the mid-1980s to between 5 to 5¼ percent in 1998.

17. These recent estimates provide compelling evidence that NAIRU has fallen during the 1990s. A number of factors have been identified as contributing to this decline.9 First, as the baby-boom generation has aged, the United States now has a more mature labor force and older workers tend to experience less frequent spells of unemployment. Second, the unexpected pickup in productivity growth over the past few years may have temporarily depressed NAIRU, as workers accept wages that are lower than what their higher rate of productivity would indicate. Third, product and labor markets have become more competitive since the early 1980s, as international trade has increased and unionization has declined.

C. Capacity Utilization

18. Another measure of resource utilization commonly used to assess potential inflationary pressures is the rate of capacity utilization published by the Federal Reserve Board. Capacity utilization is the ratio of the actual level of output to an estimated sustainable maximum level of output. The actual level of output is based on the monthly industrial production indexes. The capacity data are based on survey data collected at the plant level for the fourth quarter of each year and alternative sources of data on capacity change (such as, growth in an industry’s available capital input, or in the case of some industries, capacity measured in physical units). The annual capacity estimates are then interpolated to a monthly frequency.

19. As capacity utilization reaches a high level, inflation rises because the marginal cost of producing goods increases and leads to higher prices. Empirical evidence suggests that inflation begins to accelerate when capacity utilization exceeds a threshold near 82 percent, and this relationship has remained fairly consistent over the last 30 years.10 Despite the empirical robustness in explaining the acceleration in inflation, there are a number of shortcomings in using capacity utilization as an indicator for economy-wide inflationary pressures. Capacity utilization is based primarily on the goods-producing sector and ignores the rapidly growing service sector.11 With the rapid adoption of new technology, gains in productivity may not be adequately captured by measured capacity. The capacity utilization rate also does not capture the effect that foreign-produced goods may have on inflation,

20. During the current economic expansion, the capacity utilization rate on an annual basis increased from a low of about 78 percent in 1991, to a peak in 1994 of nearly 83 percent, before falling off to 80½ percent in May 1999. The decline in the capacity utilization rate reflects, in part, the rapid pace of business investment in recent years. The growth of industrial capacity averaged 2¾ percent over the period 1967–90; yet since 1994 growth in capacity has sutstantially exceeded this average, peaking in 1997 at about 51/2 percent. The decline in capacity utilization since 1994 also reflects the weakness in activity in the goodsproducing sector owing to appreciation of the dollar and the Asian financial crisis.

List of References

  • Adams, Charles, and Coe David, 1990, “A Systems Approach to Estimating the Natural Rate of Unemployment and Potential Output in the United States,” IMF Staff Papers, Vol. 37, June, pp. 232293.

    • Search Google Scholar
    • Export Citation
  • Blanchard, Olivier L, and Quah Danny, 1989, “The Dynamic Effects of Aggregate Demand and Supply DisturbancesAmerican Economic Review, Vol. 79, September, pp. 655673.

    • Search Google Scholar
    • Export Citation
  • Cochrane, John, H., 1994, “Shocks,” Carnegie-Rochester Conference Series on Public Policy, Vol. 41, pp. 295364.

  • Corrado, Carol and Mattey Joe, 1997, “Capacity Utilization,” Journal of Economic Perspectives, Vol. 11, Winter, pp. 151167.

  • Hodrick, Robert J. and Prescott Edward C., 1997, “Postwar U.S. Business Cycles: An Empirical Investigation,” Journal of Money, Credit, and Banking, Vol. 29, pp. 116.

    • Search Google Scholar
    • Export Citation
  • Eisner, Robert, 1998, “The Decline and Fall of the NAIRU,” in The Keynesian Revolution, Then and Now: The Selected Essays of Robert Eisner, Volume 1, (Cheltenham, United Kingdom: Elgar), pp. 45487.

    • Search Google Scholar
    • Export Citation
  • Gordon, Robert, 1998, “Foundations of the Goldilocks Economy,” Brookings Papers on Economic Activity, Vol. 2.

  • Murphy, Robert G., 1998, “Accounting for the Recent Decline in the NAIRUBoston College Working Paper No. 414.

  • OECD, 1999, OECD Economic Surveys: United States (Paris: OECD).

  • Office of Management and Budget, 1999, Budget of the United States Government, Fiscal Year 2000, Mid-Session Review; June.

  • Staiger, Douglas, Stock James H., and Watson Mark W., 1996, “How Precise are Estimates of the Natural Rate of Unemployment?National Bureau of Economic Research Working Paper 5477, March.

    • Search Google Scholar
    • Export Citation
  • Staiger, Douglas, Stock James H., and Watson Mark W., 1997, “The NAIRU, Unemployment and Monetary Policy,” The Journal of Economic Perspectives, Vol. 11, Winter, pp. 3349.

    • Search Google Scholar
    • Export Citation
  • Stiglitz, Joseph, 1997, “Reflections on the Natural Rate Hypothesis,” The Journal of Economic Perspectives, Vol. 11, Winter, pp. 310.

    • Search Google Scholar
    • Export Citation
  • U.S. Congressional Budget Office, 1999, Economic and Budget Outlook: Fiscal Years 2000–2009, January and July.

  • U.S. Congressional Budget Office, 1995, CBO’s Methodfor Estimating Potential Output, CBO Memorandum, October.


Prepared by Paula R. De Masi, Jorge Chan-Lau, and Alex Keenan.


The estimation periods for the log-linear regressions were specified from cyclical peak to cyclical peak in an attempt to eliminate the distorting effects associated with end-point years which are at different points in the cycle. Earlier work suggested that there were break points in the fourth quarter of 1973 and the fourth quarter of 1989. Accordingly it was estimated that potential GDP growth slowed to 2¾ percent after the first break point, and slowed further to 2¼ percent after the second break point.


More specifically, the Blanchard-Quah approach assumes that there exist two types of uncorrelated shocks associated with a structural VAR model that includes the growth rate of output and a cyclical variable (in this case the consumption-output ratio was used). In addition, the variance of the shocks is assumed equal to one. The long-run restriction imposed on these shocks is as follows: the first type of shock is permanent and has a long-run effect on output while the second type is temporary and does not have a long-run effect. In this framework, potential output is the component related to the permanent shock series. The structural shocks are unobserved but they can be recovered from a reduced VAR representation under the long-run restriction assumptions, and be used to construct the potential output series.


The shares for labor and capital were based on their share in national income, 70 and 30 percent, respectively.


Some economists view the absence of wage inflation given tight labor market conditions as confirmation that the NAIRU concept is flawed both empirically and theoretically. For example, Eisner (1998) established empirically an asymmetric relationship: when unemployment is above NAIRU inflation accelerates, but when unemployment is below NAIRU there is little impact on inflation.


More specifically, Okun’s Law has been estimated to be that the output gap tends to be 2½ times larger than the employment gap.


The methodology underlying the staffs augmented Phillips curve is discussed in Chapter II.


For a more detailed discussion, see Stiglitz (1997).


It has been observed, however, that inflation in the goods and services sectors follow similar cyclical patterns, so that cost pressures in the goods-producing sector may be a reasonable proxy for economy-wide pressures.

United States: Selected Issues
Author: International Monetary Fund