United States of America: Selected Issues

This Selected Issues paper on the United States analyzes problems in the measurement of output and prices. The paper examines income versus expenditure measures of national output. Sources of consumer price index and findings of the Boskin Commission are discussed, and mismeasurement of output and productivity is analyzed. Developments in productivity across industries in the United States are described. In particular, the paper focuses on the slowdown in aggregate productivity growth that began in the mid-1970s and examines whether this slowdown has continued in recent years and is common across industries.

Abstract

This Selected Issues paper on the United States analyzes problems in the measurement of output and prices. The paper examines income versus expenditure measures of national output. Sources of consumer price index and findings of the Boskin Commission are discussed, and mismeasurement of output and productivity is analyzed. Developments in productivity across industries in the United States are described. In particular, the paper focuses on the slowdown in aggregate productivity growth that began in the mid-1970s and examines whether this slowdown has continued in recent years and is common across industries.

II. Job Uncertainty and the Phillips Curve1

1. Relatively low rates of increase in wages and prices have been recorded in recent years despite declines in the unemployment rate to levels that in the past have been associated with rising inflationary pressures (Chart 1). It has been suggested that the recent behavior of wages and prices may reflect workers’ increased concerns about job prospects, which limited their demands for wage increases.2 To test this proposition, some empirical measures of job insecurity are used to augment a simple Phillips curve model, and this model is tested to see if it provides a better forecast for inflation performance during 1995 and 1996. The results suggest that job uncertainty does not help to explain the recent behavior of prices and wages.

CHART 1
CHART 1

UNITED STATES: UNEMPLOYMENT RATE, CORE CPI, AND WAGES 1/

(In percent)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Source: Bureau of Labor Statistics, U.S. Department of Labor.1/ Shaded area reflects recession.2/ Employment cost Index for civilian workers.

2. Chart 2 shows a cross-plot of inflation and unemployment over the 1990–97 period. The positive co-movements of unemployment and inflation since the end of 1992 indicate that more than a simple Phillips curve relationship must be at work. In the standard Phillips curve equation estimated here, inflation (the annualized rate of increase of the core CPI (IT)) is expressed as a linear function of lagged inflation and the unemployment gap (Ugap), which is measured as the difference between the actual unemployment rate and a time-varying estimate of the natural rate.3 Table 1 shows the results for an equation estimated using quarterly data over the period 1984–94.4 The static version of the equation is:

Π=3.9-0.89Ugap

Chart 3 presents an out-of-sample forecast for 1995–96 from the estimated Phillips curve. The equation predicts an increase in the rate of inflation over the fore cast period, with inflation rising to about 4 percent during 1996; in the event, inflation declined during the period to around 2½ percent.

CHART 2
CHART 2

UNITED STATES: CROSS PLOT OF INFLATION AND UNEMPLOYMENT RATE

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Source: Bureau of Labor Statistics, U.S. Department of Labor.
Table 1:

Standard Phillips Curve Specification

Regression of core CPI inflation on lagged inflation and the lagged unemployment gap 1/

Sample: 1984:1 to 1994:4

article image

The annualized rate of increase in the core CPI is calculated as the change in the natural logarithm of the index multiplied by 400 to put it in annual percentage terms. The unemployment gap is defined as the civilian unemployment rate less the estimated time-varying natural rate of unemployment from Thomas (1996), in units such that 1 equals 1 percent. The annual data for the natural rate of unemployment were interpolated to quarterly frequency. A lagged error term was included to avoid the need for many lags of inflation. Some experiments with eight lags of inflation and no lagged error term yielded qualitatively similar results.

CHART 3
CHART 3

UNITED STATES: ACTUAL AND FORECAST CORE CPI 1/

(In percent)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Source: Bureau of Labor Statistics, U.S. Department of Labor; and Fund staff estimates.1/ Shaded area reflects recession.

3. One possible explanation for the sharp difference between actual and forecast inflation is that the natural rate of unemployment is overestimated (and the unemployment gap underestimated). However, inserting the inflation and unemployment rates observed in recent quarters (about 2½ percent and around 5 percent, respectively) into the above equation implies a natural rate of about 3¾ percent, which is extremely low in comparison to other estimates and below the lower bound of estimated confidence intervals.5

4. The potential role played by job uncertainty in explaining the recent behavior of wages and prices also was examined. There are a variety of possible indicators of job uncertainty. Some relevant measures have been suggested by Greenspan (1997) and Valletta (1996 and 1997). Greenspan (1997) emphasizes survey measures of uncertainty, low wage increases in collective bargaining arrangements, and low levels of work stoppages as potentially important factors in explaining recent low levels of inflation (though he expresses doubts that any one measure predominates). Valletta (1996) suggests that the distinction between employer- and employee-initiated job separations is important. For the same unemployment rate, a higher rate of employer-initiated separations (e.g., layoffs, as opposed to voluntary quits) would imply greater job uncertainty and less pressure on wages and hence on prices. Valletta also presents data suggesting that over the last few decades, employers have become more likely to shed employees through permanent dismissals rather than layoffs and that dismissals have become more sensitive to the business cycle, which would add to job insecurity. Valletta (1997) also shows that displaced workers as a share of total employment has remained high in recent years because of corporate downsizing.6

5. Table 2 lists various measures of job uncertainty, and Chart 4 shows the evolution of each measure in the period 1984–96. Chart 5 compares the path of each job-uncertainty measure over the current economic expansion with its path during the previous expansion. Most measures are close to their levels at the same point in the previous expansion. The exception is the unemployment rate among job leavers, which currently is relatively low. Also, the number of job leavers as a percent of civilian employment has been stagnant in the current expansion, while it increased dramatically in the previous one. These facts suggest some continuing reluctance of workers to leave their current jobs to search for new work.

CHART 4
CHART 4

UNITED STATES: MEASURES OF JOB UNCERTAINTY

(In percent)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Sources: Bureau of Labor Statistics, U.S. Department of Labor; Conference Board; and National Association of Business Economists.
CHART 5a
CHART 5a

UNITED STATES: JOB UNCERTAINTY 1/

(In percent)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Sources: Bureau of Labor Statistics, U.S. Department of Lobor; Conference Board; and National Association of Business Economists.1/ T represents the troughs of the cycle; positive (negative) numbers represent quarters after (before) the troughs. The trough of the current cycle occurred in 1991Q1, and the trough of the previous cycle occurred in 1982Q4.
CHART 5b
CHART 5b

UNITED STATES: JOB UNCERTAINTY 1/

(In percent)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Sources: Bureau of Labor Statistics, U.S. Department of Labor; Conference Board; and National Association of Business Economists.1/ T represents the troughs of the cycle; positive (negative) numbers represent quarters after (before) the troughs. The trough of the current cycle occurred in I991Q1, and the trough of the previous cycle occurred in 1982Q4.
CHART 5c
CHART 5c

UNITED STATES: JOB UNCERTAINTY 1/

(In percent)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Sources: Bureau of Labor Statistics, U.S. Department of Lobor; Conference Board; and National Association of Business Economists.1/ T represents the troughs of the cycle; positive (negative) numbers represent quarters after (before) the troughs. The trough of the current cycle occurred in 1991Q1, and the trough of the previous cycle occurred in 1982Q4.
Table 2.

Measures of Job Uncertainty 1/

article image

All data are converted to quarterly frequency by either averaging or interpolation. Figures in parentheses below mnemonic (+ or -) indicates the expected relation of the variable to job uncertainty (positive or negative). For example, a higher level of work stoppages may indicate more confidence among workers and lower job uncertainty.

6. Augmented Phillips curves were estimated over the period 1984–94 including the variables measuring job uncertainty. Table 3 summarizes these results. The percent of days idle due to work stoppages (WSTOPP) and the NABE survey measure of falling employment (NABEMF) were significant; however, in both cases, the coefficient had the wrong sign. Regressions using the change in job uncertainty, rather than the level, performed no better.7

Table 3.

Regression Results: Augmented Standard Phillips Curve 1/

article image

Asterisk indicates significance at the five percent level, ρ denotes the coefficient on the lagged error term.

First principal component The principal components serve as composite indexes of the common forces driving the job-uncertainty variables. The first principal component explains more of the common variation among the series than any other single principal component (see Greene (1990), pp 283–5 for technical details).

First principal component, excluding work stoppages (WSTOPP).

First principal component, excluding collective bargaining variables (WF1A and WL1 A).

First principal component, excluding collective bargaining variables (WF1A and WL1 A) and work stoppages.

7. Alternatively, in line with Chairman Greenspan’s conjecture that multiple measures may be needed, equations were estimated with more than one measure of job uncertainty. The job-uncertainty variables in these equations also were not significant, possibly because of the inclusion of many variables in a fairly short sample period. For this reason, some regressions using principal components of the job-uncertainty variables were performed (results are shown in the last four rows of Table 3).8 These regressions showed mixed results for the job-uncertainty variables. Only in one case was job uncertainty significant; however, in every case, equations incorporating the principal components fix job uncertainty had implausibly large coefficients on the unemployment gap variable.

8. In using the various equations for forecasting inflation over the period 1995–96, only the equation with the unemployment rate among job losers (LRJL) produced a decline in inflation during the forecast period. However, the coefficient on LRJL was insignificant. Typical results showed a forecast for core inflation increasing to about 3.5–4 percent over 1995–96, about the same results as derived from the standard Phillips curve.

9. None of the ten measures of job uncertainty helps to explain why lower rates of inflation have been associated with lower rates of unemployment in the past few years. There is no consistent relationship between measures of job uncertainty and inflation in the standard Phillips curve equations estimated. It may be that job uncertainty is simply not at high levels, despite anecdotal evidence, which would be consistent with the impression given by the data in Chart 5. Indeed, Valletta (1997) notes that most recently displaced workers are quickly finding jobs.

10. Underlying the standard Phillips curve are two equations: one that relates wage inflation to the unemployment gap (a wage Phillips curve) and another that relates price inflation to wage inflation (a markup equation). To examine whether the source of the forecasting errors in the standard Phillips curve is wage or pricing behavior, a wage Phillips curve was estimated. Wage inflation (as measured by the change in the wages and salaries component of the employment cost index) was expressed as a linear function of the unemployment gap and lagged price inflation (as a proxy for inflationary expectations), with an equation estimated over the period 1984–94 (Table 4).

Table 4.

Wage Phillips Curve Specification

Regression of wage inflation (wages and salaries in the Employment Cost Index) on lagged core CPI inflation and the lagged unemployment gap 1/

Sample: 1984:1 to 1994:4

article image

The annualized rate of increase in the Employment Cost Index for wages and salaries is calculated as the change in the natural logarithm of the index multiplied by 400 to put it in annual percentage terms. The unemployment gap is defined as the civilian unemployment rate less the estimated time-varying natural rate of unemployment from Thomas (1996), in units such that 1 equals 1 percent, with annual data for the natural rate interpolated to quarterly frequency.

The coefficient on the unemployment gap is significant at the ten percent level when the lagged error term is omitted.

11. A forecast of wage inflation for 1995–96 produces more accurate results than a forecast of price inflation from the standard Phillips curve reported above (Chart 6). While the standard-error bands are wide and wage inflation varied substantially over the period, the actual data are well within two standard errors of the forecast and the root mean squared error (RMSE) of the forecast is 0.5 compared to 1.1 for the forecast from the standard Phillips curve. The addition of the job-uncertainty variables to the wage Phillips curve equation generally did not add significantly to the equation’s explanatory power, or forecasting ability. The exception was when a principal component of the job-uncertainty variables was used. In this equation, the principal component was significant, but the RMSE of the forecast was slightly higher (0.484 compared with 0.482).

CHART 6
CHART 6

UNITED STATES: ACTUAL AND FORECAST ECI WAGES AND SALARIES

(In percent)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Source: Bureau of Labor Statistics, U.S. Department of Labor; and Fund staff estimates.1/ Shaded area reflects recession.

12. These results suggest that in large part the recent favorable performance of inflation may be attributable to factors influencing production costs other than wages and salaries, and to the pricing behavior of firms for a given increase in costs. In particular, firms have moved aggressively to rein in benefits costs in recent years, successfully slowing the growth in their total labor costs (Chart 7). Since the end of 1994, the employment cost index has increased at an average annual rate of around 2¾ percent; while increases in wages and salaries have risen to an annual rate of 3½percent during this period, the rise in benefits costs has slowed to well below the rate attained at the same point in the previous expansion. At the same time, labor productivity growth may be higher than measured, holding down the rate of increase in unit labor costs and helping to account for the subdued behavior of prices.9 Moreover, materials costs have been held down by the appreciation of the U.S. dollar and general weakness in commodity prices, owing in part to sluggish economic activity in other major industrial countries. In addition, the pricing behavior of firms may have been restrained by the competitive effects of the dollar’s appreciation and more broadly by the ongoing process of globalization of markets. However, the rise in the profitability of U.S. firms would seem to suggest that competitive pressures may not have had a substantial restraining effect on prices.

CHART 7
CHART 7

UNITED STATES: UNIT LABOR COST AND EMPLOYMENT COST INDEXES

(Annualized growth rate)

Citation: IMF Staff Country Reports 1997, 097; 10.5089/9781451839494.002.A002

Sources: Bureau of Labor Statistics, U.S. Department of Labor; Bureau of Economic Analysis, U.S. Department of Commerce.1/ T represents the troughs of the cycle; positive (negative) numbers represent quarters after (before) the troughs. The trough of the current cycle occurred in 1991Q1, and the trough of the previous cycle occurred in 1982Q4.

List of References

  • Debelle, Guy, and Douglas Laxton, 1996, “Is the Phillips Curve Really a Curve? Some Evidence for Canada, the United Kingdom, and the United States,IMF Working Paper WP/96/111.

    • Search Google Scholar
    • Export Citation
  • Gordon, Robert J., 1997, “The Time-Varying NAIRU and its Implications for Economic Policy,Journal of Economic Perspectives, pp. 1132, Vol. 11 No.1 (Winter).

    • Search Google Scholar
    • Export Citation
  • Greene, William H, 1990, Econometric Analysis (New York: MacMillan).

  • Greenspan, Alan, 1997, testimony before the Committee on the Budget, United States Senate, January 21, 1997.

  • King, Robert G. and Mark W. Watson, 1994, “The Post-War U.S. Phillips Curve: A Revisionist Econometric History,Carnegie-Rochester Conference Series on Public Policy. pp. 157220, Vol. 41 (December).

    • Search Google Scholar
  • Staiger, Douglas, James H. Stock, and Mark W. Watson, 1997, “The NAIRU, Unemployment and Monetary Policy,Journal of Economic Perspectives, pp. 3349, Vol. 11 No.1 (Winter).

    • Search Google Scholar
    • Export Citation
  • Stiglitz, Joseph, 1997, “Reflections on the Natural Rate Hypothesis,Journal of Economic Perspectives, pp. 310, Vol. 11 No.1 (Winter).

    • Search Google Scholar
    • Export Citation
  • Thomas, Alun, 1996, “The Cyclical Position of the U.S. Economy and Its Relationship With Inflation,Chapter III in United States: Recent Economic Developments, IMF Staff Country Report Number 96/93.

    • Search Google Scholar
    • Export Citation
  • Valletta, Robert, 1996, “Has Job Security in the U.S. Declined?”, Federal Reserve Bank of San Francisco Weekly Letter, Number 9607, February 16, 1996.

    • Search Google Scholar
    • Export Citation
  • Valletta, Robert, 1997, “Job Loss During the 1990s,Federal Reserve Bank of San Francisco Weekly Letter, Number 9605, February 21, 1997.

    • Search Google Scholar
    • Export Citation
1

Prepared by Charles Kramer and Yutong Li.

2

See Greenspan (1997).

3

The time-varying estimate of the natural rate used here is taken from Thomas (1996). Alternative specifications and functional forms for the Phillips curve are discussed in Debelle and Laxton (1996); Gordon (1997); and Staiger, Stock, and Watson (1997).

4

Results for regressions using the GDP deflator or the deflator for personal consumption expenditure to measure inflation were similar, as were results using the output gap (the difference between actual and potential output) in place of the unemployment gap. Regressions including the change in the unemployment gap gave mixed results; the change in the output gap was not significant when added to equations that included the level of the output gap. Also, a few experiments with the results did not yield evidence of nonltnearities in the Phillips curve relationship.

5

See for example, Staiger, Stock, and Watson (1997). The authors have a central estimate for the natural rate of 5.8 percent, with a 95 percent confidence interval of 4.5–6.9 percent.

6

Displaced workers are persons who lost or left jobs because of the closure or relocation of a plant, because of insufficient work, or because their position or shift was eliminated. Valletta argues that the implications for wages of an increase in the number of displaced workers are likely to be limited since the trend toward higher displacement rates in the 1990s largely affected white-collar workers, and this group of workers represented only a small portion of the labor force.

7

Estimation of the effect of job uncertainty directly on the natural rate of unemployment, following the method used in Thomas (1996), showed qualitatively similar results: the effect of job uncertainty on the natural rate was rarely significant, and when it was significant, it had the opposite of the expected effect.

8

The principal components are composite indexes formed from multiple time series. Li this context, they are indexes of the common forces driving the job-uncertainty variables. Greene (1990, pp. 283–5) provides technical details.

9

Estimates of labor productivity growth, derived from the income side measure of GDP suggest that productivity has grown at roughly a 1½ percent annual rate over the past two years, compared to the traditional expenditure-side measures of output which show annual productivity growth of a little more 1 percent. See Section I of this selected issues paper for additional information on problems in the measurement of output.

United States: Selected Issues
Author: International Monetary Fund