FRBSF Economic Letter
1997-29 | October 10, 1997
The economy has been performing extraordinarily well recently. Output has grown at a robust rate, and inflation and the unemployment rate are down to levels not seen in several decades. Stock markets are booming, as firms make record profits. All this has led some to suggest that we are now in a new environment, where productivity will grow faster than before. Among the different explanations put forward to support this thesis, one of the most common is based on the growing use of computers and other information processing equipment. An increase in the growth rate of productivity would mean faster output growth–and, consequently, higher living standards–possibly without the higher inflation that might otherwise accompany faster growth. This Letter looks at the case that can be made to support the thesis that the growing use of computers has ushered in an era of higher productivity growth.
What the data show
It is not hard to find evidence of the significant impact that computers have had on our daily lives. Computers help design aircraft, help match buyers and sellers of stocks, and help economists write papers. Investment in computers has exploded. For instance, real investment in information processing equipment grew at rates in excess of 20% in both 1995 and 1996, and has grown at a more than 15% rate so far in 1997. Real investment in computers now accounts for some 3% of real output (though only about 1% of nominal output). And, at $132 billion, the current cost of the private stock of computers and peripheral equipment is about the same as that of the stock of either aircraft or autos. Finally, at $757 billion, the stock of information processing equipment exceeds the stock of transportation and related equipment by some 20%.
The hypothesis that this should raise the productivity trend is intuitively appealing, so it comes as something of a disappointment to realize that the data do not provide us with clear evidence of a sustained increase in productivity growth. To take a specific example, labor productivity (more precisely, output per hour) in the nonfarm business sector increased at a 1.3% rate in the four quarters ending in 1997:Q2 (which is the last quarter for which we have data). Productivity grew at a 1.0% rate over the last ten years (ending in the same quarter), and has grown at a 1.7%
rate since 1960.
Why has measured productivity growth not picked up more markedly in response to the growing use of computers? Some have suggested that the reason has to do with our inability to accurately measure output and, consequently, productivity. One of the big problems is that it is difficult to measure changes in quality. For instance, while today’s new car costs noticeably more than the new car of the 1950s, today’s car is of generally better quality and has many new features. How much of the higher price reflects quality improvements? This is a crucial issue: if we overestimate the increase in price, we also underestimate the value of the quality improvements and, consequently, output. Improvements that lead to increased convenience are especially hard to measure. For instance, I can now use my personal computer to dial into my bank’s computers and manage my finances much more easily and efficiently than before; it is difficult to see how this would ever show up in the data.
A careful look at the relevant statistics does provide some evidence of measurement problems. For instance, Slifman and Corrado (1996) examine productivity growth in the corporate sector (which includes companies like General Motors and IBM, as well as smaller corporations) and the noncorporate sector (which includes sole proprietorships and partnerships such as legal and medical practices, as well as nonprofit institutions such as hospitals) over the last three decades. They find that while there is little discernible change in the average growth rate of productivity in the nonfarm corporate sector since 1960, output per hour in the noncorporate sector grew at a 4 3/4% rate from 1960 to 1973, but fell at a nearly 2% rate over 1973-1980, and has fallen by an average of 1/2% per year since then. And if it is hard to believe that productivity could actually be falling in broad sectors of the economy for decades, the data provide a further puzzle: profits in the noncorporate sector–which appears to be the least efficient in the economy–continue to be robust. This combination of developments suggests that we may be understating output in this sector. To take one example of how measurement error could creep in, income data for part of the noncorporate sector are derived from income tax returns, and it is generally believed that the income shown on these returns is significantly understated.
A different breakdown of the data also provides evidence consistent with this hypothesis. Specifically, the data show that productivity in the service industries has fallen by more than half a percentage point per year since 1977, while growing at about 1% per year when averaged over all nonfarm private industries. Within the services sector the worst performers have been health services and legal services. This pattern of measured productivity growth is consistent with the measurement error story, since productivity appears to be growing more slowly in sectors where output is harder to measure.
Yet such difficulties in measuring output have probably always been with us. Is there any reason to believe that they have gotten worse recently? Griliches (1994) says the answer is yes, based on the fact that the “unmeasurable” sectors (that is sectors where output is difficult to measure) account for an increasing share of output. In the unmeasurable sectors he includes construction, trade, finance, other services, and government, while agriculture, mining, manufacturing transportation, communication, and public utilities are included in the measurable sector. He points out that in the early postwar period nearly half the economy was in the measurable sector; by 1990, this number had fallen to less than a third. As a consequence, “Measurement problems have indeed become worse. [In addition,] … major portions of actual technical change have eluded our measurement framework entirely” (p. 10).
Although the time span over which this shift has taken place may seem too long to be relevant for our purposes, Griliches points out that over three quarters of the recent spending on computing equipment has taken place in the “unmeasurable” sectors; since output in these sectors is hard to measure anyway, it is not surprising that we find no evidence of higher productivity. This view has not gone unchallenged. Sichel (1997a) has argued that even under relatively favorable assumptions, the sectoral shifts in the economy are not large enough to explain most of the productivity slowdown since the 1970s.
By implication, these shifts are not large enough to conceal a recent computer-related surge in productivity growth, either. While this might appear surprising to some, it does not surprise David (1991), who believes that what we see today may be similar to what happened following the introduction of the electric dynamo in the late nineteenth century. As he points out, new technological paradigms are not implemented instantaneously; instead, they require changes and adjustments that are likely to take a substantial amount of time. For instance, the Edison central generating station was introduced in London and New York in 1881; yet it is generally agreed that British productivity growth over the period from 1900 to 1913 was the lowest it had been since the late eighteenth century. U.S. productivity data for that period show a slowdown as well. (By comparison, Intel introduced the silicon microprocessor in 1970.)
According to David, commentators worried about the lack of higher productivity from computers suffer from “technical presbyopia.” This condition is marked, on the one hand, by a conviction that computers will “…swiftly and inexorably…” lead to more efficient production, and on the other, by a “…depressing suspicion that something has gone terrible (sic) awry, fostered by the disappointment of premature expectations about the information revolution’s impact upon the conventional productivity indicators…” (p. 336).
David’s main point–that significant changes in the technological paradigm do not take place overnight–is well taken. However, that by itself does not imply that the growing use of computers represents a change that is big enough to raise the growth rate of productivity. Indeed, another group of economists believes that the measured productivity data show no evidence of faster growth because computers have not had–and are unlikely to have–a significant impact on productivity growth.
Sichel (1997b) points out that computer hardware makes up only about 2% of the nation’s capital stock. As a consequence, any contribution that it can make to output growth will be limited (even if investment in computers leads to unusually high returns). Adding software and other computer related items raises the contribution computers make to growth in his calculations, but not by very much. For instance, under the assumption that investment in computers earns competitive returns, Sichel estimates that computers (inclusive of software) helped raise the rate of real output growth by only about 0.3 percentage points per year over the period from 1987 to 1993.
Gordon (1996) agrees with these arguments and then goes further to state that he does not think that computers are as important as the inventions of the early twentieth century. Among the latter are the “…pervasive spread of electric motors and appliances into all aspects of production and consumption,…the use of the internal combustion engine in motor transport and air transport, with the derivative inventions of the suburb, interstate highway and supermarket, …the telephone and its derivatives,…the range of entertainment and information industries, including radio, movies, television, and recorded music” (p. 267). Consequently, there is little reason to expect that computers will lead to the kind of productivity growth seen in the 1960s.
Who is correct?
Clearly, economists are divided over what role computers and other information processing equipment have played–and will play–in the growth of the economy. Which side is correct? At this point, it is hard to tell. For instance, if computers represent a change in the technological paradigm as David suggests (if computers transform production the way the dynamo did, for example), accounting exercises based on their share in the nation’s capital stock may miss the point.
There is some indirect evidence suggesting that computers are likely to be important. As mentioned above, firms have invested large amounts of money in computers. It is hard to imagine them doing so without the expectation of a substantial return. Some of the effects of this massive investment appear to be showing up in labor markets as well. In particular, the wages of highly skilled workers have gone up relative to those with fewer skills, and many labor economists believe that the growing use of computers, and related technology, has something to do with this.
However, this evidence is not enough to establish that the rate of productivity growth has accelerated recently or that it is about to accelerate. It is difficult to make judgements or predictions about productivity growth at least partly because we do not have good theories about what determines its long-run trend. Was the rapid productivity growth during the 1960s normal? Or was that the aberration? Is there even any reason to believe that productivity should keep growing in the long run? In other words, maybe productivity would not have grown at all without computers. There are no easy answers to these questions.
Overall, it looks like measurement error may be causing some understatement of productivity growth. While the size of this error may be in dispute, there is not a lot of evidence suggesting that it has grown significantly larger in the last few years. The large sums of money that firms are spending on computers certainly attest to their importance, but that is not enough to establish that we are now in a regime where productivity will grow substantially above the rates experienced over the last decade or two.
Corrado, C., and L. Slifman. 1996. “Decomposition of Productivity and Unit Costs,” mimeo, Board of Governors of the Federal Reserve System, November.
David, Paul A. 1991. “Computer and Dynamo: The Modern Productivity Paradox in a Not-Too-Distant Mirror,” in Technology and Productivity: The Challenge for Economic Policy, OECD, pp. 315-345.
Gordon, Robert J. 1996. “Comment” on “Can Technology Improvements Cause Productivity Slowdowns?” NBER Macroeconomics Annual, pp. 259-267.
Griliches, Zvi. 1994. “Productivity, R&D, and the Data Constraint,” American Economic Review, pp. 1-23, March.
Sichel, Daniel E. 1997a. “The Productivity Slowdown: Is a Growing Unmeasurable Sector the Culprit?” The Review of Economics and Statistics, pp. 367-370.
_____. 1997b. The Computer Revolution: An Economic Perspective, Washington, D.C.: Brookings Institution Press.
Opinions expressed in FRBSF Economic Letter do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco or of the Board of Governors of the Federal Reserve System. This publication is edited by Sam Zuckerman and Anita Todd. Permission to reprint must be obtained in writing.
Please send editorial comments and requests for reprint permission to
Attn: Research publications, MS 1140
Federal Reserve Bank of San Francisco
P.O. Box 7702
San Francisco, CA 94120