Abstract
This Selected Issues paper presents updated IMF staff estimates of potential output growth for the United States, using data through 2001 that incorporates the full cyclical upswing of the 1990s and the subsequent mild recession, as well as taking into account the revisions to the national accounts released in July 2000. The paper also reviews recent investment trends and provides estimates of the extent to which the capital stock has deviated from its long-term equilibrium.
I. Potential Output Growth1
A. Introduction
1. The rapid growth of productivity and output in the United States has undoubtedly been the most salient economic development of the 1990s. The key question going forward is whether these rapid growth rates are sustainable and will continue to support consumer and investment spending. This chapter presents updated staff estimates of potential output growth for the United States, using data through 2001 that incorporates the full cyclical upswing of the 1990s and the subsequent mild recession, as well as taking into account the revisions to the national accounts released in July 2000.2 The individual sectors that have driven the pickup in potential output and total factor productivity (TFP) growth are also examined.
2. These estimates suggest that the annual growth rate of potential output was in the range of 3¼ to 3½ percent for the period 1990-2001, marginally above earlier estimates that were based on data through 1999. The sectors exhibiting the strongest growth in potential output and TFP include communications, wholesale and retail trade, transportation and utilities, and manufacturing. Although the September 11th terrorist attacks are expected to reduce the level of potential output in the near term, recent estimates suggest that they would have only a marginal effect on potential output growth over the medium term.
B. Potential Output Growth: Revised Estimates
3. Four different methods were used to estimate potential output: the segmented trend; the Hodrick-Prescott (H-P) filter; the band-pass filter; and the production function approaches. These methodologies suggest potential output growth during 1990-01 in the range of 3¼-3½ percent, slightly above earlier staff estimates of 3–3¼ percent (Table 1).3
United States: Estimates of Potential Ouput Growth
(Average annual percentage change)
Contribution to growth.
United States: Estimates of Potential Ouput Growth
(Average annual percentage change)
| Revised Estimate | Previous Estimate | ||||
|---|---|---|---|---|---|
| Staff estimates | |||||
| Method/source | Period | Period | |||
| Segmented trend | 1982-01 | 3.1 | 1982-99 | 3.1 | |
| 1991-01 | 3.5 | ||||
| Hodrick-Prescott filter | 1990-01 | 3.2 | 1990-99 | 3.2 | |
| Band-Pass filter (0,32) | 1990-98 | 3.6 | |||
| Production function | 1990-00 | 3.3 | 1990-98 | 3.0 | |
| TFP 1/ | 1.0 | ||||
| Labor 1/ | 1.5 | ||||
| Capital 1/ | 0.9 | ||||
| Other estimates | |||||
| Congressional Budget Office (2002) | 2002-12 | 3.4 | 2000-10 | 3.1 | |
| Office of Management and Budget (2002) | 2002-12 | 3.1 | 2000-05 | 3.0 | |
| Macroeconomic Advisers (2002) | 2002 | 3.0 | 2000 | 3.8 | |
| Council of Economic Advisers (2002) | 2001-12 | 3.1 | 2000-08 | 3.1 | |
Contribution to growth.
United States: Estimates of Potential Ouput Growth
(Average annual percentage change)
| Revised Estimate | Previous Estimate | ||||
|---|---|---|---|---|---|
| Staff estimates | |||||
| Method/source | Period | Period | |||
| Segmented trend | 1982-01 | 3.1 | 1982-99 | 3.1 | |
| 1991-01 | 3.5 | ||||
| Hodrick-Prescott filter | 1990-01 | 3.2 | 1990-99 | 3.2 | |
| Band-Pass filter (0,32) | 1990-98 | 3.6 | |||
| Production function | 1990-00 | 3.3 | 1990-98 | 3.0 | |
| TFP 1/ | 1.0 | ||||
| Labor 1/ | 1.5 | ||||
| Capital 1/ | 0.9 | ||||
| Other estimates | |||||
| Congressional Budget Office (2002) | 2002-12 | 3.4 | 2000-10 | 3.1 | |
| Office of Management and Budget (2002) | 2002-12 | 3.1 | 2000-05 | 3.0 | |
| Macroeconomic Advisers (2002) | 2002 | 3.0 | 2000 | 3.8 | |
| Council of Economic Advisers (2002) | 2001-12 | 3.1 | 2000-08 | 3.1 | |
Contribution to growth.
4. The segmented trend approach was used to identify points where the trend rate of growth in GDP may have changed, and in particular to determine whether there was an identifiable trend increase during the last economic expansion. Recursive residual tests were used to identify the break points in the chain-linked real GDP series over the period 1959-2001. In addition to the two break points found in previous studies—one in the first quarter of 1975, and the other in the first quarter of 1982—a new break point, around the first quarter of 1991, was also identified. As a result, potential output growth is estimated to have increased to 3½ percent in the 1990s, from just over 3 percent during the previous decade.4
5. The H-P filter was also used to identify an estimate of trend output that minimizes a weighted average of the difference between output and trend output and the rate of change in trend output.5 Because the H-P filter tends to overweight the end-points in the series being detrended, potential output was estimated over the period from the peak in the level of output in the fourth quarter of 1973 to the third quarter of 2001. On this basis, potential growth was estimated at 3¼ percent in the 1990s.
6. A band-pass filter was also used to remove the high frequency components (business cycle or higher) from the actual data, with the low frequency residual taken as a proxy for the trend component. The band-pass filter is the difference between two low-pass filters (one with a high cutoff frequency and the other with a low cutoff frequency).6 Since approximation requires truncating the series by 12 quarters before the end, the estimated trend goes as far as 1998. For the 1990–98 period, the average rate of growth of the low frequency component in the GDP data was 3½ percent.
7. In addition, a production function approach was used to explicitly model output in terms of underlying factors of production, expressing output as a function of capital, labor, and TFP.7 In this case, the production function was assumed to be of Cobb-Douglas type with constant returns to scale and constant elasticities with regard to capital and labor inputs.8 Potential labor input and trend TFP were estimated by using the H-P filter. Using these detrended series, the growth rate of potential output was estimated to be 3¼ percent over the period 1990–2000.9
8. These estimates of potential output growth are similar to those published by the Office of Management and Budget, the Congressional Budget Office, and the Council of Economic Advisers (Table 1). It is noteworthy that the estimates prepared by Macroeconomic Advisers—a private sector consultancy—suggest that potential growth has fallen sharply from 2000 to 2002—from 3.8 percent to 3.0 percent. Because the particular production function approach used places a heavy weight on short-term developments, this significant downward revision likely reflects the effect of the sharp cutback of investment spending on capital inputs over the last year.
9. The band-pass filter, the HP filter, and the production function all showed similar patterns for the gap between output and potential (Figure 1), with the average gap at about -1.2 percent for the three quarters of 2001. Incorporating the 1991 break point in the segmented trend yields a smaller negative output gap in the recent period than previous segmented trend estimates, but the estimate is broadly similar to the other methods surveyed here.
United states: Measures of the output Gap
(In percent of potential output)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
United states: Measures of the output Gap
(In percent of potential output)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
United states: Measures of the output Gap
(In percent of potential output)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
10. In order to test the relevance of the alternative measures of the output gap for predicting CPI inflation, inflation was regressed on its lagged values and the output gap.10 Using a recursive residual test, several break points were found in the relationship, the most recent being in 1987 and 1996.11 Using the most recent sample estimates, a transitory—one period—1 percentage point increase in the output gap would raise inflation from the estimated zero-gap rate of 2 percent to an estimated 2½ percent, but most of the inflationary effect is estimated to disappear very quickly (Figure 2.)
United States: Inflation Response to a 1 Percentage point Fall in the Output Gap 1/
(In Percent)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
Source: Staff estimates.1/ Using an output gap based on the Hodrick -Prescott filter.United States: Inflation Response to a 1 Percentage point Fall in the Output Gap 1/
(In Percent)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
Source: Staff estimates.1/ Using an output gap based on the Hodrick -Prescott filter.United States: Inflation Response to a 1 Percentage point Fall in the Output Gap 1/
(In Percent)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
Source: Staff estimates.1/ Using an output gap based on the Hodrick -Prescott filter.C. Sectoral Trends
11. During the 1990s, some sectors—the technology sectors, in particular—grew at exceptionally rapid rates, while others lagged. This suggests considerable differences in sectoral trends in potential output growth. In order to explore this possibility, the H-P filter was applied to sectoral output data.12
12. In the 1990s, sectors with trend potential GDP growth above that of the total economy included communication, wholesale and retail trade, transportation and utilities, and manufacturing (Figure 3). Over the 1990s, these sectors grew by 3-18 percent more than the economy-wide aggregate. This confirms the conclusion of other studies that suggest strong productivity growth in the wholesale and retail sectors, such as McKinsey (2001). Within the manufacturing sector, the electronic equipment aggregate exhibited a trend growth rate 50 percent higher than the economy-wide rate. The relatively slow-growing sectors included the mining sector as well as the finance and construction sectors.
United States: Sectoral Output Trend Relative to Overall GDP Trend Hodrick-Prescott Filter
(1990=100)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
United States: Sectoral Output Trend Relative to Overall GDP Trend Hodrick-Prescott Filter
(1990=100)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
United States: Sectoral Output Trend Relative to Overall GDP Trend Hodrick-Prescott Filter
(1990=100)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
13. TFP calculations on sectoral data suggest that in the second half of the 1990s those sectors that exhibited the highest output growth also experienced the fastest TFP growth. 13 These sectors included transportation, communications and utilities, wholesale and retail trade, and manufacturing (Table 2 and Figure 4). In the construction sectors, where output growth had been modest relative to the economy-wide norm, TFP growth was relatively weak. In contrast, the service sector grew at the economy-wide trend rate, despite below average TFP growth; conversely the mining sector exhibited trend growth well below the average, but TFP growth in the first part of the 1990s was well above the economy’s average.14
United States: Sectoral Productivity Growth
(Average annual percentage change)
United States: Sectoral Productivity Growth
(Average annual percentage change)
| Total Factor Productivity Growth | Labor Productivity Growth | ||||||
|---|---|---|---|---|---|---|---|
| (Staff Estimates) | Bailey (2002) | ||||||
| 1977-1989 | 1989-1995 | 1995-2000 | 1989-1995 | 1995-2000 | |||
| Real gross domestic product | 0.9 | 0.9 | 1.1 | ||||
| Mining | 0.1 | 2.5 | 0.8 | 4.6 | -1.8 | ||
| Construction | 0.1 | -0.4 | -0.4 | -0.1 | -0.7 | ||
| Manufacturing | 2.5 | 2.6 | 3.3 | 3.2 | 4.5 | ||
| Transportation, public utilities, and communications | 2.1 | 2.5 | 2.2 | ||||
| Transportation | 2.5 | 1.5 | |||||
| Public utilities | 2.5 | 2.3 | |||||
| Communications | 5.1 | 2.2 | |||||
| Wholesale trade | 0.5 | 2.2 | 2.9 | 2.8 | 5.9 | ||
| Retail trade | 0.8 | 1.2 | 2.7 | 0.7 | 4.7 | ||
| Finance, insurance, and real estate | -0.2 | 0.6 | 1.6 | 1.7 | 3.5 | ||
| Services | -0.6 | -1.0 | -1.1 | -1.1 | 0.1 | ||
United States: Sectoral Productivity Growth
(Average annual percentage change)
| Total Factor Productivity Growth | Labor Productivity Growth | ||||||
|---|---|---|---|---|---|---|---|
| (Staff Estimates) | Bailey (2002) | ||||||
| 1977-1989 | 1989-1995 | 1995-2000 | 1989-1995 | 1995-2000 | |||
| Real gross domestic product | 0.9 | 0.9 | 1.1 | ||||
| Mining | 0.1 | 2.5 | 0.8 | 4.6 | -1.8 | ||
| Construction | 0.1 | -0.4 | -0.4 | -0.1 | -0.7 | ||
| Manufacturing | 2.5 | 2.6 | 3.3 | 3.2 | 4.5 | ||
| Transportation, public utilities, and communications | 2.1 | 2.5 | 2.2 | ||||
| Transportation | 2.5 | 1.5 | |||||
| Public utilities | 2.5 | 2.3 | |||||
| Communications | 5.1 | 2.2 | |||||
| Wholesale trade | 0.5 | 2.2 | 2.9 | 2.8 | 5.9 | ||
| Retail trade | 0.8 | 1.2 | 2.7 | 0.7 | 4.7 | ||
| Finance, insurance, and real estate | -0.2 | 0.6 | 1.6 | 1.7 | 3.5 | ||
| Services | -0.6 | -1.0 | -1.1 | -1.1 | 0.1 | ||
United States: Sectoral TFP Relative to Overall TFP
(1990=100)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
Sources: U.S. Department of commerce, Bureau of Economic Analysis, and staff estimates.United States: Sectoral TFP Relative to Overall TFP
(1990=100)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
Sources: U.S. Department of commerce, Bureau of Economic Analysis, and staff estimates.United States: Sectoral TFP Relative to Overall TFP
(1990=100)
Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A001
Sources: U.S. Department of commerce, Bureau of Economic Analysis, and staff estimates.D. The Impact of the Terrorist Attacks
14. The September 11th terrorist attacks dealt a significant blow to an already weak U.S. economy.15 In the immediate aftermath of the attacks, consumer and business confidence and equity prices fell sharply, and economic activity was disrupted—including the shutdown of the air transport system and financial markets. The economy, however, demonstrated remarkable resilience in the face of this shock, with confidence and equity prices returning to pre-attack levels by year-end. Nonetheless, increased security costs and greater uncertainty have raised the question of whether the attacks will adversely affect potential output and TFP growth over the medium term.
15. In analyzing the impact of the terrorist attacks on the productive capacity of the economy, it is important to distinguish between one-time effects on the level of TFP and potential output, and their growth rates. A one-time increase in the cost of doing business—for example, higher security costs—is likely to contribute to a one-time decline in the level of TFP as businesses spend more on capital and labor without an increase in production.16 Other added costs include business travel delays; irradiating mail; the need for higher levels of inventories to guard against supply disruptions; more elaborate backup systems in case of another attack; and cross-border delays. Accordingly, CBO (2002) estimates indicate that such costs together would total around $20 billion in 2002, and would reduce the level of TFP by about 0.3 percent after five years.17 However, some of these negative effects could dissipate as businesses adapt—for example, by using video conferencing instead of air travel.
16. Over the long term, as firms divert resources away from investment spending and toward security equipment, the terrorist attacks could also affect the growth rate of total factor productivity. However, CBO (2002) suggests that TFP growth would be reduced by just 0.03 percentage point over the medium term.
List of References
Baxter, M., and R. G. King, 1999, “Measuring Business Cycles: Approximate Band-Pass Filters for Economic Time Series,” Review of Economics and Statistics; 81, No. 4: 575-93, November.
Bailey, M., 2002, “The New Economy: Post Mortem or Second Wind?” unpublished paper.
CBO, 2002, The Budget and Economic Outlook: Fiscal Years 2003-2012 (Washington, D.C.: U.S. GPO).
CBO, 2001, CBO’s Method for Estimating Potential Output: An Update (Washington, D.C.: U.S. GPO).
Council of Economic Advisors, 2002, Economic Report of the President (Washington, D.C.: U.S. GPO).
De Masi, P.R., J. Chan-Lau, and A. Keenan, 1999, “Measures of Potential Output, NAIRU, and Capacity Utilization,” in United States: Selected Issues, IMF Staff Country Report No. 99/101, September.
De Masi, P. and M. Kaufman, 2000, “Potential Output Growth: Revised Estimates,” in United States: Selected Issues, IMF Staff Country Report No. 00/112, August.
Hodrick, R. J. and E. C. Prescott, 1997, “Postwar U.S. Business Cycles: An Empirical Investigation,” Journal of Money, Credit, and Banking, Vol. 29, pp. 1-16.
Jorgenson, D., M. Ho, and K. Stiroh, 2001, “Projecting Productivity Growth: Lessons from the U.S. Growth Resurgence,” paper prepared for the conference Technology, Growth and the Labor Market, sponsored by the Federal Reserve Bank of Atlanta and Georgia State University, December 31.
Macroeconomic Advisers, 2002, Economic Outlook, March.
McKinsey Global Institute, 2001, U.S. Productivity Growth, 1995-2000 (Washington D.C., McKinsey Global Institute)
Office of Management and Budget, 2002, Budget of the United States Government, Analytical Perspectives, Fiscal Year 2003.
OECD, 2002, Economic Outlook, No. 71 (Paris: OECD).
Prepared by Paula De Masi and Martin Kaufman.
Previous staff estimates had already incorporated comprehensive revisions to the national accounts released in October 1999. These revisions showed that real GDP over the last three decades grew at a faster rate than the previous data indicated, especially in the 1990s. Several factors contributed to the stronger rates of growth in the revised data, including revisions to the source of data, new methods to adjust for inflation, and updated definitions of spending categories, the most important being the inclusion of computer software in investment. See De Masi and Kaufman (2000).
These results were presented in (De Masi and Kaufman 2000). For methodological details see also De Masi, Chan-Lau, and Keenan (1999).
The estimation periods for these log-linear regressions were specified from cyclical peak to cyclical peak in an attempt to eliminate the distorting effects associated with end-points that are at different points in the business cycle. Earlier estimates are in De Masi and Kaufman (2000).
The specification used to obtain the below-business-cycle frequency in the approximated band-pass filter used 0 and 36 quarters for the high and low cutoff frequencies of the low-pass filters, and a truncation point of K=12 for the symmetric moving average representation. See Baxter and King (1999).
See Jorgenson and Ho (2001) and CBO (2001).
The capital and labor elasticities are based on their shares in national income and are assumed at 30 and 70 percent, respectively; these are also the assumptions used in CBO (2001). The labor input series was constructed from average weekly hours and employment in the private nonfarm sector, while the capital stock was defined as the stock of fixed private nonresidential equipment and structures.
It is worth noting that the estimates of TFP growth accelerated in the second half of the 1990s to 1.2 percent, compared to 1 percent for the whole 1990s and 0.9 percent since the 1960s.
Although inflation and the output gap exhibit significant persistence, they are stationary processes. Unit root tests performed rejected the 1(1) hypothesis. The measure of output gap used in the regression was estimated from the HP filter. Other measures yielded similar results.
The change in parameter estimates observed between the 1988-2001 and 1997-2001 samples is concentrated on the autoregressive component of inflation and not on the parameter estimates for the output gap, which were remarkably stable across samples.
Data limitations precluded replicating all of the other methodologies used in the previous section with sectoral data.
Sectoral TFP growth, similar to the economy-wide case, was calculated using a Cobb-Douglas production function with constant coefficients for labor and capital. These coefficients were estimated using long-term averages of the share of inputs in nominal output. The series for the potential labor input and trend TFP were estimated using the H-P filter. Jorgenson, Ho, and Stiroh (2001) also estimate sectoral TFP growth, and find a wide variation of productivity growth patterns across sectors. Their estimating procedure differed in that they used production function coefficients that changed over time, and were based on a rolling two-period average of the share of inputs in nominal output.
These results are broadly consistent with labor productivity estimates presented by Bailey (2002).
For a detailed discussion on the economic impact of the attacks, see OECD (2002).
This assumes that businesses purchase increased security as an intermediate good. To the extent that firms contract with a security firm, then some offsetting increase in output could occur.
Based on a similar methodology, CEO (2002) estimates that increased security costs would reduce the level of TFP by about 14 percent after five years.
