The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Measuring Economic Activity and Economic Welfare: What Are We Missing? Charles Steindel Federal Reserve Bank of New York Abstract Major U.S. economic data, most notably GDP and Industrial Production, are undergoing major changes. Proposals have been made for significant alterations in the CPI. The revision process has helped to spur debate on such topics as the proper method of accounting for high technology’s role in the economy, the reported sluggishness of productivity growth in many service industries, and the overstatement of price increases for numerous products. This paper attempts to assess the potential impact of some of these problems on our understanding of basic trends in the economy. It is found that with even fairly generous assumptions as to the time path of errors in price data, the fundamentals of the economy’s broad movements do not change: productivity and real earnings growth were likely still substantially slower in the first half of the 1990s than before 1973. Measuring Economic Activity and Economic Welfare: What Are We Missing? Charles Steindel* Federal Reserve Bank of New York Major U.S. economic data are undergoing fundamental changes. The construction of the two major indicators of aggregate economic activity--Gross Domestic Product (GDP) and Industrial Production (IP)--have been substantially revised. The best-known forecasting gauge--the Composite Index of Leading Indicators (CLI)--has also undergone a significant revision. Finally, as has been well-reported, proposals have been made for significant alterations in our major inflation index--the Consumer Price Index (CPI). This paper starts with a brief overview of the data revisions. The revision process has helped to spur debate on such topics as the proper method of accounting for high technology’s role in the economy, the reported sluggishness of productivity growth in many service industries, and the overstatement of price increases for numerous products. The latter part of the paper attempts an assessment of the potential impact of some of these problems on our understanding of basic * Thanks to Reagan Murray and Hitesh Patel for assistance and to colleagues at the Federal Reserve Bank of New York for comments. The views stated here are those of the author and not necessarily those of the Federal Reserve Bank of New York or the Federal Reserve System. 3 trends in the economy. Essentially, it is found that with even fairly generous assumptions as to the time path of errors in price data, our fundamental view on the economy’s broad movements does not change: productivity growth was likely still substantially slower in the first half of the 1990s than it was before 1973. When we narrow our definition of output to a concept more likely to be directly tied to household welfare the evidence is also strong that the post-1973 slowdown in real earnings growth is continuing. A general implication of this analysis is that the possible discovery and correction of many sectoral biases in the inflation, output and productivity data in many industries is more likely to pay off in the form of a greater understanding of the sources of economic growth rather than in finding a radically different path for the growth of aggregate activity and economic well-being. A further implication is that the hunt for “missing” real output in the aggregate data could well be more fruitful if directed at understatements of nominal output rather than overstatements of inflation. Revisions in the Indexes Traditionally, incoming readings of the major U.S. output series--Gross Domestic Product and Industrial Production were produced using the fixed-base year Laspeyres technique. The aggregate series were constructed by weighting indexes of output in individual components by prices in some past base year. In 4 cases where the contemporaneous price structure was quite different than that of the base year, this procedure had the effect of giving undue importance to components whose relative prices had fallen since the base year. Note that “undue” is meant with both a positive and normative inference--according to basic economic theory, the current relative price structure should reflect the scarcity value of different products. It is fallacious to assign a relative importance to a product higher than the market does. The positive, practical problem with the Laspeyres technique was the tendency for initial estimates of aggregate output to be revised down as base years were moved forward. The positive bias in the initial estimates arose from the natural tendency of output to rise most rapidly in sectors whose relative prices are falling. The Laspeyres technique assigned weights in the computation of aggregate growth to these rapidly-growing areas proportional to their higher relative prices in the far-off base year; when the base year was moved forward the weights on the rapidly-growing sectors fell and overall output growth was reduced. The problems with the Laspeyres method were well-understood. They gained greater urgency, though, because of the spectacular rise in output, and fall in prices, in the computer industry. the late 1980s, as the base years for GDP and IP computations In were moved from 1982 to 1987 the reductions in aggregate growth 5 estimates were quite noticeable. The obvious solution to this problem was continually moving forward the base year, and constructing the output aggregates as “linked” or “chained” indexes tied to some base year and backed out from the computed annual growth rates (this was already done on a five-year basis for IP, but the long time interval meant for an awkward transition around the base years). This change was made in 1995-97 for both GDP and IP; in the course of the transition the decision was also made to drop the Laspeyrses technique for computing aggregate growth in favor of the Fisher Ideal method.1 The Fisher Ideal method combines (using the geometric average) aggregate growth computed by the Laspeyres technique with that computed by the Paasche procedure (which involves assigning current prices to sectors to calculated aggregate growth). Research has established the theoretical attractiveness of this long-proposed method of computing aggregate output and price indexes (Diewert, 1976, 1983a, 1983b, 1987). The revisions of the aggregates brought into sharp focus the apparent failure of trend output and productivity to improve in the 1990s, despite the extraordinary growth in spending on high- 1 For overviews of the switch in GDP, see Young, 1989; Triplett, 1992; Motley, 1992; Landefeld, 1995;, Landefeld and Parker, 1995; and Steindel, 1995; for the IP change see Corrado, Gilbert, and Raddock, 1997. 6 technology items.2 with price indexes. Debate began to focus on potential problems If inflation is overstated, real output and productivity growth are being understated. In support of the argument that there may be significant problems with the price data in the service sector, in particular, has been the observation that the published productivity data for much of the service sector of the economy continues to be very disappointing.3 Observation has also been made of the rather limited coverage the existing industrial classification system gives to the service sector--in the sense that there are fewer classifications at a disaggregated level--of rapidly growing tertiary services in such areas as health care, finance, and business services. There appears to be a presumption that improvements in the data collection system might result in upward revisions of real growth for the service sector and overall GDP, perhaps because of improvements in the construction of price indexes (Baily and Gordon, 1988). The growing importance of the service sector to the economy-or, at least, the relative decline in importance of goods 2 See Business Week, 1995; Farrell, 1995; and Spiers, 1995, for criticisms of the revisions. McNamee, 1997, discusses the continuing puzzle of low productivity growth. 3 Slifman and Corrado (1996) note the oddly slow growth of the noncorporate sector of the economy. While there are many services in this sector it is by no means synonymous with services as a whole, nor is it clear that all the data problems are on the price side. 7 production--has been underscored by changes in the Composite Index of Leading Indicators (CLI). In the recent revision of the index, long-standing components that were presumably most closely related to the manufacturing sector--commodity prices and unfilled orders for manufactured goods--were dropped (Conference Board, 1996). One of the new components placed in the index was the spread between the yield on the 10-year Treasury Bond and the Federal funds rate, in line with recent studies suggesting its value as a generalized forecasting variable (Estrella and Mishkin, 1996). An alternative, non-exclusive view, is that this variable may also foretell expansion in the very large financial sector of the economy. Of course, these changes in the index were prompted by the statistical need to maintain a good forecasting tool, rather than by a priori notions about tying individual components to individual sectors. Still, it is reasonable to think there should be some rough correspondence between the composition of such an index and that of the economy. What are We Missing? There is then a fairly widespread impression that the published data could be missing significant movements in the economy--possibly because of underestimation of the effect of high technology in the service sector (Business Week, 1995; Farrell, 1995; Spiers, 1995; and Quinn and Baily, 1994; discuss the impact of high technology on the service sector; Baily and Gordon, 1988; and Gordon, 1995, discuss low reported service 8 sector growth). At one extreme, it has been argued that aggregate productivity growth might be as high or higher today than in the 1950s and 1960s (Nakamura, 1997). This section attempts to quantify the amount that aggregate real growth may be higher than we observe due to understatements of service sector activity resulting from overstatements of service sector inflation, examining a number of sets of data: real GDP, real productivity growth, and a measure of average real earnings growth. Although discussions of understatements of overall growth often focus upon the oddly low productivity growth rates in many service industries, it is unreasonable to estimate understatements of aggregate real growth by simply adding-up understatements of productivity growth and sectoral output, at least in any simple fashion (Baily and Gordon, 1988; and Gordon, 1995). The problem is that in the U.S. data, industry output is measured as value-added, not sales. If an industry’s output growth is understated, it is possible that the output growth of some supplier or customer industry is overstated; getting a fix on the understatement in aggregate output would then involve working through industry linkages in an input-output framework. A much simpler way to compute possible output growth understatement is to use estimates of individual price overstatements to adjust the real expenditure data. exercises below, this procedure will be followed. In the The working 9 assumption is that the price overstatement issue is of particular importance in services. This is not to deny that there may well be overstatements in the prices of goods or structures--indeed, in the Boskin commission report (Boskin, et al, 1996), most of the hard data in support of inflation overstatement came from studies of goods prices. However, there is little reason to believe that problems with the overstatement of goods price inflation have increased as time has passed, while the dramatic changes in the nature of many services gives us some reason to think that biases may have grown over time.4 Table 1 gives the basics on the composition of aggregate spending. A bit over half of GDP consists of spending on services. About 20% of the spending on services consists of government compensation of employees plus depreciation of government capital, and more than 10% consists of the imputed services provided by the housing stock. Thus, only two-thirds of total spending on services--about one-third of GDP, but a higher share of nonfarm business output--consists of actual purchases of services from the private sector. Another significant portion of services consists of those whose prices are probably measured fairly well, or, at least probably no worse than in the past-items such as utilities and transportation. 4 This leaves only The issue fundamentally involves adjusting posted price increases for quality improvements. Hulten, 1997; and Moulten and Moses, 1997; discuss the theory and practice of quality adjustments in the CPI. 10 about 40% of spending on services--about 20% of GDP--in categories where there are significant questions about pricing-areas such as financial and business services, medical care, and educational and charitable expenses. (Griliches, 1994, notes the rising importance of “hard-to-measure” products, though his list is somewhat larger than the one used here.) Note that spending on these services enters directly into GDP when it is made by households or governments or is a part of foreign trade. Assuming that the national accounts correctly measure current-dollar spending in these service categories, we can readily compute the effect on real growth of alternative, reduced estimates of inflation in these hard to measure sectors. presents such estimates. Table 2 They are made under two assumptions: 1. Inflation in these sectors is uniformly overestimated by 2 percentage points a year. 2. Inflation in these sectors is reduced to equal that in the rest of the economy. Alternative measures are presented of overall GDP growth and nonfarm business productivity growth. apply is 1. The periods to which these alternatives 1960 and thereafter. 1983 and thereafter. 4. 2. 1974 and thereafter. 3. 1992 and thereafter. In general, reducing the rate of inflation in the rapidlygrowing “hard-to-measure” service categories does raise recent growth rates. GDP growth over the last generation would have averaged .3 percentage point higher, and nonfarm business productivity growth about .4 percentage point higher, if the 11 inflation numbers in these categories had been lower along the lines of the alternatives.5 However, the more interesting implication is what these adjustments would have meant to the long-term dynamics of the economy. The answer is, not too much: comparing the growth rates shown for the later periods in the alternative lines with the published numbers for the period before 1973 indicates that real GDP growth and real productivity growth would still have been significantly lower in recent years than in the 1960-1973 period even if the overstatement to inflation in these categories is assumed to have started in recent years (if the inflation overstatement began earlier the conclusion is even more valid). One cannot meaningfully assume that reduced measured aggregate output and productivity growth is an artifact of problems in pricing certain types of services unless one believes that the problems are very large indeed relative to those of the past (There could, of course, be large persistent overstatements of price changes throughout the economy, but the point at issue here is whether the comparison of growth today with the past changes much if the inflation overstatements in certain sectors have grown over time). Another set of issues involving misstatement of prices 5 The productivity adjustment is almost surely exaggerated, since it was made assuming that all outlays in the hard-measurecategories were produced in the nonfarm business sector. Many medical, educational, and religious and welfare services are produced directly in the nonprofit sector of the economy. 12 involves the growth of real income. How much of the reported post-1973 decline in growth of average real income may be due to problems in deflating nominal income? The answer to this question involves recomputing the growth of the cost of living. A natural way to do this is to apply the above alternatives to the consumer portion of hard-to-price services and recompute the growth of the chain-weight deflator for personal consumption. This procedure was done, with the partial modifications that 1. spending on personal business services was removed from consumption, on the grounds that these expenditures do not directly add to household well-being; and 2. spending on consumer durables--which is a form of asset accumulation--was removed from consumption and replaced by estimates of the gross imputed rent from the existing stock of durables (here measured as depreciation6 plus a 3 percent additional rate of return). Table 3 presents alternative estimates of the growth of average real income, with nominal income being compensation per full time equivalent worker in private nonfarm business. As was the case for real output and productivity, alternative estimates are presented on the basis of inflation for “hard-to-measure” services being equal to that for the rest of consumption, and 2 percentage points lower than published, for various time periods 6 Current-dollar data on the depreciation of durables are available in the flow-of-funds tables. It was assumed that the deflator for depreciation was the same as that for the stock of durables. 13 since 1960. The top lines of the two parts of the table present the real income growth data for the conventional measure of consumer inflation and that for the restated measure, which removes spending on personal business activity and restates durables. The results of this exercise are much the same as those for GDP and productivity: reducing inflation rates in selected service categories does markedly raise estimates of real income growth over the last few years--from roughly .5% to 1% or more over the last decade. However, at best, all we can conclude is that if either of these alternatives are correct--and only correct for recent years, not for the whole period since 1960-real income per employee is now growing about half as fast as it was in the 1960s. While this is more favorable than the more conventional calculations that real income per worker is growing much less that half as fast as in the pre-1973 era, it hardly changes the observation that real income growth remains well below its earlier standards. Conclusion These exercises suggest that one is unlikely to explain away continued historically slow growth of output, productivity, and real incomes through faulty pricing of various hard-to-price services,7 unless the pricing problems have widened to such a 7 Sichel, 1997a and 1997b, reaches a similar conclusion. 14 magnitude that there has been, in reality, outright deflation for a wide variety of products in these and other categories. One implication of these exercises is that the payoff from reengineering the statistical system toward services, with emphasis on improved pricing, may not involve a radically different picture for aggregate growth, unless we think that recent growth of nominal output will be raised, along with reductions in published inflation. One way a redesign of the statistical system may produce higher nominal output growth is a through a redefinition of capital to include more intangible items, such as software and some measure of accumulated R&D knowledge. Even this change, though, will not necessarily increase recent growth relative to the distant past (conceivably, the “stock” of, and the output generated by, R&D capital grew even more rapidly in the 1960s than today). The real payoff from an improved statistical system could well be a greater understanding of the industrial sources of U.S. growth. Such an improvement could well guide policymaking and possibly improve longer-term forecasting. It seems much less likely that we would see a major revamping of the stylized facts suggested by the existing data. 15 References Baily, Martin Neil, and Robert J. Gordon, “The Productivity Slowdown, Measurement Issues, and the Explosion of Computer Power,”, Brookings Papers on Economic Activity, 1988 no. 2, pp. 347-420. Boskin, Michael J., Ellen R. Dulberger, Robert J. Gordon, Zvi Griliches, and Dale Jorgenson, “Toward a More Accurate Measure of the Cost of Living. Final Report to the Senate Finance Committee from the Advisory Commission to Study the Consumer Price Index.” December 4, 1996. Business Week. “Suddenly, the Economy Doesn’t Measure Up.” July 31, 1995, pp. 74-76. Conference Board. Indicators.” “Forthcoming Revisions to the Composite Business Cycle Indicators. November 1996, pp. 3-4. Corrado, Carol, Charles Gilbert, and Richard Raddock, “Industrial Production and Capacity Utilization: Historical Revision and Recent Developments,” Federal Reserve Bulletin, vol. 83, no. 2, Feb. 1997, pp. 67-92. 16 Diewert, W.E. “Exact and Superlative Price Indexes,” Journal of Econometrics, vol. 4, no. 2, May 1976, 115-146. _______. “The Theory of the Cost-of-Living Index and the Measurement of Welfare Change.” In W.E. Diewert and C. Montmarquette, eds., Price Level Measurement: Proceedings from a conference sponsored by Statistics Canada, Minister of Supply and Services Canada, 1983 (1983a). _______. “The Theory of the Output Price Index and the Measurement of Real Output Change.” In W.E. Diewert and C. Montmarquette, eds., Price Level Measurement: Proceedings from a conference sponsored by Statistics Canada, Minister of Supply and Services Canada, 1983 (1983b). _______. “Index Numbers.” In John Eatwell, Murray Ligate and Peter Newman, eds., The New Palgrave: A Dictionary of Economics. New York, Stockton Press, 1987. Ehrlich, Everett. “The Statistics Corner: Notes on Chain-Weighted GDP.” Business Economics, October 1995, pp. 61.-62. 17 Estrella, Arturo, and Frederic S. Mishkin, “The Yield Curve as a Predictor of U.S. Recessions,” Federal Reserve Bank of New York Current Issues in Economics and Finance, vol. 2, no. 7, June 1996. Farrell, Christopher. “Why the Numbers Miss the Point.” Business Week, July 31, 1995, p. 78. Gordon, Robert J. “Problems in the Measurement and Performance of Service-sector Productivity in the United States.” In Productivity and Growth, Reserve Bank of Australia, 1995. Griliches, Zvi. “Productivity, R&D, and the Data Constraint,” American Economic Review, vol. 84, no. 1, March 1994, pp. 123. Hulten, Charles R. “Quality change in the CPI.” Federal Reserve Bank of St. Louis, April 1997. Landefeld, J. Steven. “BEA’s Featured Measure of Output and Prices,” National Association of Business Economists News, no. 113, Sept. 1995. _______, and Robert P. Parker. “Preview of the Comprehensive Revision of the National Income and Product Accounts: BEA’s 18 New Featured Measures of Output and Prices,” Survey of Current Business, vol. 75, July 1995. McNamee, Mike. “The Productivity Boom is Still a Mystery,” Business Week, August 25, 1997, p. 42. Motley, Brian. “Index Numbers and the Measurement of Real GDP,” Federal Reserve Bank of San Francisco Economic Review, 1992 No. 1, pp. 3-13. Moulton, Brent R., and Karin E. Moses, “Addressing the Quality Change Issue in the Consumer Price Index,” Brookings Papers on Economic Activity, 1997 no. 1, pp. 305-349. Nakamura, Leonard I. “Is U.S. Economic Performance Really that Bad?” Federal Reserve Bank of Philadelphia, mimeo, Feb. 1997. Quinn, James Brian, and Martin Neil Baily. “Information Technology: Increasing Productivity in Services.” Academy of Management Executive, 1994, vol. 8, no.3, pp. 26-47. Sichel, Daniel E. The Computer Revolution: An Economic Perspective. Washington (Brookings) 1997 (1997a). 19 . “The Productivity Slowdown: Is a Growing Unmeasurable Sector the Culprit?” Review of Economics and Statistics, vol. 79, Aug. 1997, pp. 367-370 (1997b). Slifman, L., and C. Corrado, “Decomposition of Productivity and Unit Costs.” Federal Reserve Board, Occasional Staff Study 1, November 18, 1996. Spiers, Joseph. “Why Can’t the U.S. Measure Productivity Right?” Fortune, Oct. 16, 1995, pp. 55-56. Steindel, Charles. “Chain-weighting: The New Approach to Measuring GDP.” Federal Reserve Bank of New York Current Issues in Economics and Finance, Vol. 1, no. 9, December 1995. Triplett, Jack E. “Economic Theory and BEA’s Alternative Quantity and Price Indexes,” Survey of Current Business, vol. 72, April 1992. Young, Allan H. “Alternative Measures of Real GDP,” Survey of Current Business, vol. 69, April 1989. 20 Table 2 Alternative Estimates of GDP and Productivity Growth 1960-1996 1960-1973 1974-1996 1983-1996 1992-1996 3.2 4.2 2.5 3.0 2.7 Alternative 1:8 3.4 4.3 2.8 3.3 2.8 Alternative 2:9 3.4 4.4 2.8 3.2 2.9 1.7 2.9 1.0 1.2 1.0 Alternative 1:1 2.1 3.1 1.4 1.6 1.2 Alternative 2:2 2.1 3.2 1.4 1.5 1.3 GDP Growth Published Productivity Growth Published 8 Prices of hard-to-price services grow at same rate as other products. 9 Prices of hard-to-price services grow at 2 percent less than published. 21 Table 3 Alternative Estimates of Real Income Growth 1960-1996 1960-1973 1974-1996 1983-1996 1992-1996 Published 1.1 2.5 .5 .3 .4 Restated 1.3 2.5 .5 .5 .4 Alternative 1 1.6 2.7 .9 .9 .9 Alternative 2 1.6 2.6 .9 .9 .9 Published: Growth of compensation per full-time equivalent employee, divided by personal consumption deflator. Restated: Deflator restated to eliminate personal business expenditures and durables spending replaced by estimate of service flow. Alternative 1: Prices of medical, educational, and religious and welfare spending assumed to grow at same pace as other items. Alternative 2: Prices of medical, educational, and religious and welfare spending assumed to grow 2 percent a year less than published. 22 Table 1 1996 Composition of GDP Billions of Dollars Total Goods Structures Services Government Employee Compensation Government Depreciation Space Rent Other Hard-to-Measure Components Consumer Medical Care Personal Business Educational Religious and Welfare Government Net Exports All Other Services 7636.0 2785.2 663.6 4187.3 759.9 125.1 564.4 2737.9 1774.9 1499.0 808.1 421.1 119.6 150.5 179.0 96.6 963.0 Percent 100.0 36.5 8.7 54.8 10.0 1.6 7.4 35.9 23.2 19.6 10.6 5.5 1.6 2.0 2.3 1.3 12.6