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.

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.

II. Evaluating the Evidence of a Capital Overhang in the U.S. Economy1

1. After nearly a decade of decline, U.S. investment as a share of GDP rose sharply during the 1990s. This increase, which was concentrated in equipment purchases, is widely considered to have played an important role in boosting U.S. productivity growth during the latter half of the decade.

2. However, the subsequent collapse of investment spending in 2001 has reinforced concerns that the earlier surge was a “bubble” and that a capital stock overhang could retard the U.S. recovery from the 2001 recession. These fears have been exacerbated by high profile bankruptcies in the information, technology, and communication sectors, as well as by concern that firms had overinvested in computer equipment in the late 1990s, including efforts in preparation for “Y2K.”

3. This chapter reviews recent investment trends and provides estimates of the extent to which the capital stock has deviated from its long-run equilibrium. The results suggest that a modest overhang of computer equipment did arise in late 2000, but that it has subsequently eroded. An economy-wide overhang in noncomputer-related equipment is not apparent, but several industries may have invested excessively in communications equipment and trucks.

A. Historical Trends in Investment and the Capital Stock

4. Private nonresidential investment as a share of GDP rose sharply during the 1990s, largely reversing the decline that had occurred during the 1980s (Figure 1). On a gross basis (including depreciation), nominal investment fell from a post-war peak of nearly 14 percent of nominal GDP in the early 1980s to around 9¾ percent of GDP in 1992. The decline reflected a response to a range of factors—slower productivity growth, higher real interest rates, overbuilding in the oil and residential sectors in the 1980s, and the recession in the early 1990s. In addition, the relative price of investment goods fell sharply, reflecting the effects of the IT revolution. However, with the robust economic recovery and the decline in the cost of capital, the investment rate recovered strongly during the 1990s, reaching nearly 13¼ percent of GDP in 2000, before falling to around 11 percent during the subsequent economic slowdown.

Figure 1.
Figure 1.

United States: Gross Private Nonresidential Investment (Fixed)

(Percent of GDP)

Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A002

Figure 2.
Figure 2.

United States: Net Private Nonresidential Investment (Fixed)

(Percent of net domestic Product)

Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A002

Figure 3.
Figure 3.

United states: Depreciation

(Percent of net domestic product)

Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A002

Figure 4.
Figure 4.

United States: capital-output Rations

(Net capital as a share of net domestic product)

Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A002

Sources: U.S. Department of Commerce, Bureau of Economic Analysis; and staff estimates. All data are nominal; investment data are for private, fixed nonresidential investment. In Figures 2 and 3, annual data for net investment and capital are not yet available (total, net investment is from staff estimates).

5. The investment boom during the 1990s was associated with significantly higher spending on equipment, which rose from 7 percent of GDP in 1992 to 9¾ percent of GDP in 2000 (Table 1 and Figure 1). Outlays for computers and software were particularly strong, increasing by 1½ percentage points of GDP and accounting for nearly 40 percent of the increase in the business investment rate.2 Structures investment, which had exhibited a secular decline since the early 1980s because of consolidation in the mining industry, rose only marginally during the 1990s.

Chain Weighting and Capital-Output Ratios

  • In 1996, the U.S. national accounts shifted to a system of chain weighting.1 Previously, real magnitudes had been constructed by using fixed-weight deflators—i.e., nominal magnitudes were deflated by prices that were constructed from a fixed set of goods and in fixed proportions. Under the chain-weighted system, the indices are calculated on a rolling basis.

  • The advantage of the chained indices is that they avoid overstating measures of real growth rates when relative prices change rapidly.2 This problem had become particularly acute during the 1990s given the relative decline in the price of investment goods.

  • An important implication of chain-weighted data is that the real components of GDP no longer sum to total real GDP. Similarly, ratios of real magnitudes no longer have economic relevance.

  • As a result, investment/output and capital/output ratios in the text are all expressed in nominal terms. See Whelan (2001) for a detailed discussion.

1 See Whelan (2000) for a discussion of the methodological issues regarding chain-weighted data.

2 Components of chain-weighted data can grow faster than the overall total forever, which is impossible with fixed-weighted data. The further away from the base year, the greater the difference between the components and the chain-weighted total. For example, in 1970 the sum of the real values of the individual components of the information-processing capital stock (computers and peripheral, software, communications, instruments, photocopies, and other office equipment) was 54 percent larger than the chain-weighted total.

Table 1.

United States: Private Investment Rates

article image
Source: U.S. Department of Commerce, Bureau of Economic Analysis. Disaggregated net investment data are not yet available for 2001. Net investment data are not available on a disaggregated basis in the NIPAs.

6. The decline in the investment rate during 2001-02 was led by a sharp drop in spending on equipment.3 However, cutbacks in computers and software spending played a less important role than during the boom years, falling by only ½ percentage point of GDP, less than one-quarter of the decline in the business investment rate. Instead, the decline primarily reflected lower outlays on communication equipment, machinery, and trucks.

7. The net investment rate (i.e., excluding depreciation) exhibited broadly similar trends from the 1980s to 2000. Net private nonresidential investment declined sharply from the early 1980s to the early 1990s, reaching a low of around 1½ percent of net domestic product (NDP), before rebounding in the early 1990s and reaching around 5 percent of NDP in 2000 (Figure 2). Equipment investment led the increase—during 1995–2000, net investment in these categories averaged about 2¾ percent of NDP, nearly twice the rate during 1986–94 (Table 1). The net investment rate for structures also rebounded somewhat during the 1990s but remained well below its average in previous decades.

8. The share of gross investment devoted to offsetting depreciation has remained broadly unchanged since the mid-1980s, averaging around 9¾ percent of NDP.4 This was in contrast to the sharp increase in depreciation that occurred during the late 1960s and 1970s (Figure 3). The earlier increase was mostly accounted for by just nine industries—representing less than one-third of GDP.5

9. The growth of the capital stock also accelerated during the 1990s. The growth of the net stock of real, private fixed nonresidential assets had slowed to an average annual rate of around 2¼ percent during 1986–94 from 3–4 percent per year earlier (Table 2). However, during 1995–2000, the growth rate of the capital stock accelerated to 3½ percent, mainly reflecting investment in high-technology equipment—computers, software, and communication equipment—as well as transportation equipment. Again, the stock of nonresidential structures in real terms grew relatively modestly.

Table 2.

United States: Private, Fixed Nonresidential Capital Stock

article image
Source: U.S. Department of Commerce, Bureau Economic Analysis.

10. High investment rates in the 1990s arrested the sharp decline in the capital stock to NDP ratio that began in the 1980s (Figure 4). The capital stock fell from over 150 percent of NDP in the early 1980s to around 125 percent of NDP in the late 1990s, reflecting the effects of the lower investment rate during the first half of the 1990s, which was insufficient to maintain the capital/NDP ratio at the high level reached in the late 1980s. The drop in the price of investment goods also worked to lower the ratio, since the capital stock data are measured at the current cost of replacement.

11. Despite increased investment in high-tech capital, its share in the total stock remains relatively modest. Rapid increases in the quantity of investment in computer equipment were largely offset by falling computer prices, and the share of computer equipment in the total capital stock increased to only about 1 percent during 1995-2000 (Table 2). Similarly, price declines for communications equipment helped keep its share at only around 4 percent.

12. Sectoral data also do not indicate excessive increases in capital stocks during the late 1990s (Table 3). The ratio of private capital to business output actually declined, on average, from above 135 percent of business output in 1965–94 to just under 130 percent of business output in 1995–2000. The decline was evident in nearly all industries except for mining, whose capital-to-output ratio rose over the past 35 years, partly in response to preferential tax breaks given to the oil industry and the sharp rise in the price of oil during the 1970s. The capital-to-output ratios in the wholesale and retail trade sectors increased significantly during earlier periods, but increased only modestly during the late 1990s.

Table 3.

United States: Capital-Output Ratio, by Industry

article image
Source: U.S. Department of Commerce, Bureau of Economic Analysis.

13. Investment rates in many other OECD countries also rose during the 1990s (Figure 5). Although the pickup was less pronounced than in the United States, a broad-based increase was apparent, especially in the EU and Australia.6 The investment boom abroad was also associated with an accelerating decline in the relative prices of investment goods in most countries. However, the relative prices of investment goods has fallen the fastest in the United States during the late 1990s, possibly reflecting the higher share of high-technology goods in investment than in other countries.7

Figure 5.
Figure 5.

International Comparisons: Rates and Relative prices of Investment

Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A002

Sources: OECD; and staff calculations. Private fixed nonresidential investment as a percent of GDP. The relative price of investment is measured as the ratio of the investment deflator to the GDP deflator, normalized to 1970=1.

B. Estimating the Extent of a Capital Overhang

14. To investigate the extent to which the increase in investment may have resulted in the capital stock overshooting its equilibrium level, an investment model is developed and estimated. The model allows the capital stock to deviate from its optimal level in the short run, but it requires that capital revert to the optimal level over time. In this framework, investment occurs as firms adjust stocks to the long-run equilibrium. In the econometric analysis described below, separate equations were considered for computer and noncomputer capital in order to take into account the possibility that rapid price declines and high depreciation rates for computer equipment resulted in different adjustment patterns.

15. The model is similar to an approach suggested in Tevlin and Whelan (2000), but it also accounts for the long-run relationships between capital, output, and the cost of capital, while including short-run dynamics.8 Specifically, an error-correction model was estimated, using a long-run cointegrating vector between capital, business output, and the Hall-Jorgenson cost of capital.9 Following Hubbard (1998), the short-run adjustment model also includes a cash-flow variable (from the National Income and Product Accounts) to reflect the effect of short-run constraints on firms’ investment decisions.

16. The estimated results are broadly consistent with economic theory (Table 4 and Table 5). In both the long-run equation and error-correction model, higher output (and lower cost of capital) leads to higher levels of long-run capital. Furthermore, the estimates suggest that the computer capital stock is more sensitive to output and the cost of capital than the aggregate, noncomputer stock. This is consistent with anecdotal evidence for the computer industry—that is, its short product cycles, and the sensitivity of computer prices to demand. The coefficient on the error-correction term is also consistent with theory; when the capital stock is above equilibrium (CVt-1 > 0), economic forces adjust to reduce the growth rate of capital (ΔK1<0). The speed of adjustment—measured by the coefficient on the error-correction term—is 3½ times larger on the computer equation. This implies that excess computer capital is eliminated much more quickly than noncomputer equipment.

Table 4.

United States: Estimates From Cointegrating and Error-Correction Models for the Stock of Equipment and Software Capital

(Excluding Computer Equipment)

a. Cointegrating equation:

kt=2.5060.564+1.119yt0.0280.035ct0.105

b. Vector error-correction model (VECM):

article image
Source: Staff estimates.Note: standard errors in italics, The CV term is the residual from the cointegrating vector; that is, CVt-1 = kt-1 - 1.119 yt-1 + 0.035 ct-1.
Table 5.

United States: Estimates From Cointegrating and Error-Correction Models for the Computer-Equipment Capital Stock

a. Cointegrating equation:

kt=15.76013.547+1.710yt0.8680.867ct0.179

b. Vector error-correction model (VECM):

article image
Source: Staff estimates.Note: standard errors in italics, The CV term is the residual from the cointegrating vector; that is, CVt-1 = kt-1 - 1.710 yt-1 + 0.867 ct-1.

17. The results of the model for noncomputer capital indicate that there is little evidence of a widespread capital overhang. Indeed, the forecast results from the error-correction model suggest persistent underinvestment in capital, starting in the late 1980s and corresponding to the period of low net investment in the United States (Figure 6). However, the estimates should be interpreted with caution, since the fit of the cointegrating model is not very strong, and since the cost of capital term has the correct sign but is insignificant.

Figure 6.
Figure 6.

United States: Stock of Noncomputer Equipment and Software

(In billions of 1996 dollars)

Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A002

Sources: U.S. Department of Commerce, Bureau of Economic Analysis; and staff estimates.

18. The model for the computer capital suggests that the stock fell below its equilibrium level starting in 1991 (Figure 7). However, as a result of the rapid increase in the investment rate in the late 1990s, overinvestment of computer equipment totaling $37 billion had emerged by end-2000. The apparent overinvestment was equivalent to 9½ percent of the computer stock, or about ½ percent of the overall stock of equipment and software on a nominal basis.10

Figure 7.
Figure 7.

United States: Stock of computer Equipment

(In billions of 1996 dollars)

Citation: IMF Staff Country Reports 2002, 165; 10.5089/9781451839548.002.A002

Sources: U.S. Department of Commerce, Bureau of Economic Analysis; and staff estimates.

19. Although the estimated overhang is apparently significant, the rapid depreciation rate of computer equipment would mean that this overhang would be quickly erased. The median age of computer equipment is less than two years, and BEA assumes that its economic depreciation rate is 32 percent.11 As a result, most of the excess capital would have been erased by the end of 2001.12

20. While the results above do not suggest a significant capital overhang, the analysis of broader aggregates leaves open the possibility that excess stocks might have been accumulated in specific types of capital or in specific sectors. This uncertainty is underscored by the fact that the estimation results for the equation for the noncomputer capital stock were relatively poor. The possibility that pockets of excess stocks may have arisen in recent years is explored below.

C. Industry- and Equipment-Specific Overhangs

21. Analysis of industry-level data point to a range of assets where excessive stocks may have been built up. In particular, there were six categories of equipment that had a substantial (more than 5 percentage points) swing in the annual rate of growth of capital accumulation from the first to the second half of the 1990s (the change in the growth rate is in parentheses, see Table 6 for greater detail):13

  • Computers and peripheral equipment (27 percent)

  • Mining and oilfield machinery (11½ percent)

  • Construction tractors (8¾ percent)

  • Trucks, buses, and truck trailers (8½ percent)

  • Communication equipment (6 percent)

  • Construction machinery (excluding tractors) (5½ percent)

Table 6.

United States: Growth Rates of Equipment and Software, by Type

(Average annual growth)

article image
Source: U.S. Department of Commerce, Bureau of Economic Analysis.

22. In order to explore further the possibility that specific pockets of overinvestment may have occurred, unpublished industry-level data on capital stocks from the Bureau of Economic Analysis (BEA) were examined. These data also suggest the possibility of overinvestment in a small number of sectors and types of equipment—most notably: communication equipment; construction machinery (except tractors); custom software; mainframe computers; trucks, buses, and truck trailers; and wire and cable structures (Table 7).14 Moreover, rapid investment was concentrated in industries representing only 12 percent of private output and in assets representing 5 percent of the nonresidential capital stock. The apparent overinvestment in the telecommunications sector is well known, and industry analysts have ascribed the overinvestment in trucking to preferential leasing arrangements that have encouraged purchases of trucks.

Table 7.

United States: summary of Industries with Excessive changes in the Capital-Output Ratio and Rapid Capital Acumulation1/2/

article image
Sources: U.S. Department of Commerce, Bureau of Economic Analysis; and staff calculations.

The analysis is based on the following types of capital: mainframe computers; integrated systems; personal computers, computer storage devices, computer terminals; mining and oilfield machinery, computer printers, construction tractors, trucks, buses, and truck trailers, direct access storage devices; wire and cable structures, communication equipment; construction machinery, except tractors.

In each of the cases, the real capital stock grew at least 5 percent faster during 1995-2000 than 1990 -1994.

23. While overinvestment in these areas seems too small to pose a systemic risk to the U.S. economy, it suggests challenges to the affected sectors, especially since the depreciation rates on trucks and communications capital are far lower than for computers. According to Fraumeni (1997), the BEA uses a depreciation rate of 31 percent for computer equipment. For communications equipment it is 11 to 15 percent; for trucks, the rate is 12 to 19 percent.

List of References

  • Collecchia, A. and P. Schreyer, 2001, “ICT Investment and Economic Growth in the 1990s: Is the United States a Unique Case? A Comparative Study of Nine OECD Countries,STI Working Paper, DSTI/DOC(2001)/7, 25 October 2001 (Paris: OECD).

    • Search Google Scholar
    • Export Citation
  • Caballero, R., 1999, “Aggregate Investment,” in Handbook of Economics, Taylor, J.B. and M. Woodford, eds. (New York: NorthHolland).

  • Fraumeni, B., 1997, “The Measurement of Depreciation in the U.S. National Income and Product Accounts,Survey of Current Business, July (Washington, D.C.: Department of Commerce) 7-23.

    • Search Google Scholar
    • Export Citation
  • Hubbard, R.G., 1998, “Capital-Market Imperfections and Investment,Journal of Economic Literature, March (Nashville, TN) 193-225.

  • Kirova, M. and R. Lipsey, 1998, “Measuring Real Investment: Trends in the United States and International Comparisons,Federal Reserve Bank of St. Louis Review, January/February (St. Louis, MO: Federal Reserve) 3-18.

    • Search Google Scholar
    • Export Citation
  • Kopcke, R., 1993, “The Determinants of Business Investment: Has Capital Spending Been Surprisingly Low?New England Economic Review, January/February (Boston: Federal Reserve) 3-31.

    • Search Google Scholar
    • Export Citation
  • Lum, S. and B. Moyer, 2001, “Gross Domestic Product by Industry for 1998-2000,Survey of Current Business, November (Washington, D.C.: Department of Commerce) 17-33.

    • Search Google Scholar
    • Export Citation
  • Macroeconomic Advisers, 2002, Economic Outlook (St. Louis) February.

  • OECD, 2001, “Saving and Investment: Determinants and Policy Implications,” in Economic Outlook, No. 70 (Paris: OECD).

  • Oliner, S., G. Rudebusch, and D. Sichel, 1995, “New and Old Models of Business Investment: A Comparison of Forecasting Performance,Journal of Money, Credit and Banking, Vol. 27, No. 3, August, 806-826.

    • Search Google Scholar
    • Export Citation
  • Plegrin, F. S. Schich, and A. de Serres, 2002, “Increases in Business Investment Rates in OECD Countries in the 1990s: How much can be explained by Fundamentals?,Working Paper ECO/WKP(2002)13 (Paris: OECD).

    • Search Google Scholar
    • Export Citation
  • Tevlin, S. and K. Whelan, 2000, “Explaining the Investment Boom of the 1990s,mimeo. (Washington, D.C.: Federal Reserve Board of Governors).

    • Search Google Scholar
    • Export Citation
  • Whelan, K., 2001, “A Two-Sector Approach to Modeling U.S. NIPA Data,Finance and Economics Discussion Series, Working Paper 2001-4 (Washington, D.C.: Federal Reserve Board of Governors) January.

    • Search Google Scholar
    • Export Citation
  • Whelan, K., 2000, “A Guide to the Use of Chain Aggregated NIPA DataFinance and Economics Discussion Series, Working Paper 2000-35 (Washington, D.C.: Federal Reserve Board of Governors) June.

    • Search Google Scholar
    • Export Citation
1

Prepared by Christopher MacDonagh-Dumler.

2

The balance was largely due to higher investment spending on: communications equipment, industrial machinery, and trucks (which together accounted for 40 percent of the increase) and office building and other commercial, nonfarm, and natural gas structures (20 percent of the increase).

3

A slowdown in structures investment—as reflected in rising vacancy rates in office building and other commercial and nonfarm structures—accounted for 35 percent of the decline in the investment rate (0.2 percentage point of GDP).

4

As a share of the capital stock, depreciation has steadily increased, however. In 1980, BEA’s measure of depreciation in the capital stock data was about 6¼ percent of the capital stock. By 2000, it reached 8 percent.

5

The industries were: industrial machinery and equipment; electronic equipment, motor vehicles; petroleum and coal products; telephone and telegraph; electric, gas and sanitary services (public utilities); wholesale trade; retail trade, and business services.

6

As Figure 5 indicates, and Kirova and Lipsey (1998) note, the United States has had a relatively lower investment rate through much of the post-war period. OECD (2001) and Pelgrin, Schich, and de Serres (2002) note that the recent increase in investment is largely in line with fundamentals.

7

Collecchia and Schreyer (2001) illustrate that although high-technology investment provided a boost to growth in all countries, the United States benefited the most, followed by Australia, Finland, and Canada. Technology contributed the least to economic growth in Germany, France, Italy, and Japan. Prices of high-technology goods declined most rapidly in the United States, although after accounting for different statistical methodologies, the U.S. decline is not as pronounced.

8

See Caballero (1999). By including a cointegrating term, this model differs from many traditional models Oliner, Rudebusch, and Sichel (1995) provide a comprehensive survey of the performance of common investment models, all of which fit the data poorly, and most of which suffer from theoretical and methodological problems. Tevlin and Whelan (2000) show that the perpetual inventory method used to derive the traditional estimating equation is not valid because of the chain-weighting methodology used by BEA to aggregate the capital-stock data.

9

Pelgrin, Schich, and de Serres (2002) estimate a cointegrating relationship between investment, output, and the cost of capital in nine OECD countries and find some support for cointegration.

10

In contrast to the overall capital-stock equation, the fit of the cointegrating relationship in the computer equipment model is significantly better, and the coefficient on the cost of capital is negative, significant, and appropriately close to -1.

12

These results are similar to recent research by Macroeconomic Advisers (2002) that estimates a capital overhang in the technology sector (computers and software) of about $55 billion by the end of 2000, declining to $25 billion by the end of 2001.

13

BEA data cover 62 types of capital (equipment, software, and structures) for 58 industries and include the average age of capital, real capital stocks (in 1996 chain-weighted dollars), and nominal capital stocks and are combined with industry output data from Lum and Moyer (2001). Both data sets are available on BEA’s website http://www.bea.gov).

14

Specifically, industries and capital types were selected as follows: industries were ranked by the ratio of the nominal stock of capital (for each asset type) to nominal industry output, and by the growth rate of the real stock of capital (for each type) between 1995 and 2000. Individual asset types in an industry were selected if: (i) the change in the capital-output ratio between 1995 and 2000 was disproportionately large (that is, the change in the ratio was in the lowest or highest 2½ percentile, measuring extreme changes), and (ii) if the real stock of capital grew by more than the industry average (to measure rapid capital accumulation). Attention was focused on only types of capital that grew significantly faster in the late 1990s than the early 1990s (the annual growth rate of real capital was more than 5 percentage points higher in the latter 1990s than in the early 1990s).

United States: Selected Issues
Author: International Monetary Fund
  • View in gallery

    United States: Gross Private Nonresidential Investment (Fixed)

    (Percent of GDP)

  • View in gallery

    United States: Net Private Nonresidential Investment (Fixed)

    (Percent of net domestic Product)

  • View in gallery

    United states: Depreciation

    (Percent of net domestic product)

  • View in gallery

    United States: capital-output Rations

    (Net capital as a share of net domestic product)

  • View in gallery

    International Comparisons: Rates and Relative prices of Investment

  • View in gallery

    United States: Stock of Noncomputer Equipment and Software

    (In billions of 1996 dollars)

  • View in gallery

    United States: Stock of computer Equipment

    (In billions of 1996 dollars)