Computers And Growth

Abstract

The purpose of this paper is to describe the impact of investment in computers on the growth of the U.S. economy. The economic literature on computers is relatively rich in information on the decline in computer prices and the growth of computer investment. Constant quality price indices for computers have been included in the U.S. National Income and Product Accounts (NIPA) since 1986. These indices employ state of the art methodology to capture the rapid evolution of computer technology. While the annual inflation rate for overall investment has been 3.66 percent for the period 1958 to 1992, computer prices have declined by 19.13 percent per year! Similarly, overall investment grew at 3.82 percent, while investment in computers increased at an astounding 44.34 percent! These familiar facts describe growth in the output of computers. The objective of this paper is to complete the picture by analyzing the growth of computer services as inputs. In a pioneering paper Bresnahan (1986) has focused on pecuniary externalities arising from the rapid decline in computer prices. Griliches (1992, 1994) has emphasized the distinction between pecuniary and nonpecuniary externalities in the impact of computer investment on growth. This paper is limited to pecuniary externalities or the impact of reductions in computer prices on the substitution of computer services for other inputs. As Griliches (1992) points out, this is an essential first step in identifying nonpecuniary externalities or ‘spill-overs’ through the impact of a decline in computer prices on productivity growth. * *Brynjolfsson (1993) has proveded a detailed survey of studies of nonpecuniary externalities or ‘spill overs’. Recent studies include those of Brynjolfsson and Hitt (1994a, 1994b) and Lichtenberg (1993). In two important papers Stephen D. Oliner (1993, 1994) has introduced a model of computer technology that greatly facilitates the measurement of computer services as inputs. In this paper we estimate computer stocks and flows of computer services for all forms of computer investment included in NIPA. We construct estimates of computer services parallel to NIPA data on computer investment by combining these data with information on computer inventories. For example, the International Data Corporation (IDC) Census of Computer Processors includes an annual inventory of processors in the U.S. In Section 1 we present data on investment in computers and constant quality price indices from NIPA. These data incorporate important innovations in modeling computer technology stemming from a joint study by IBM and the Bureau of Economic Analysis (BEA) completed in 1985. This study utilized a ‘hedonic’ methodology for constructing an econometric model of computer prices that accurately reflects rapid changes in computer technology. This methodology generates an index of computer prices that holds the quality of computers constant. In Section 2 we present the model of computer services originated by Oliner (1993,1994). This differs in important respects from the model of capital services used in the previous studies of U.S. economic growth surveyed by Jorgenson (1989,1990). The model employed in previous studies is based on the decline in productive capacity with the chronological age of a capital good. Oliner assumes that computers maintain their productive capacity until they are retired. Decline in productive capacity occurs only through removal of used computers from the inventory through retirement. In Section 3 we construct estimates of stocks of computers that incorporate IDC data on computer inventories and derive the implied flow of computer services. While output of computer investments has grown very rapidly, the input of computer services has grown even faster. The price of these services has declined at 23.22 percent per year over the period 1958 to 1992, while the input of these services has grown at 52.82 percent! This is prima facie evidence of an important role for computer price declines as a source of pecuniary externalities. In Section 4 we combine computer services with the services of other types of capital to produce a measure of capital input into the U.S. economy. We link this with labor input to obtain the contributions of both inputs to U.S. economic growth, arriving at the growth of productivity as a residual. We find that the contribution of computer services to input into the U.S. economy is far more important than the contribution of computer investments to output. This is a significant step toward resolution of the Solow paradox: ‘We see computers everywhere except in the productivity statistics. * *Robert M. Solow, quoted by Brynjolfsson (1993). Declines in computer prices generate very sizable pecuniary externalities through the substitution of computer services for other inputs. By contrast Solow focuses on nonpecuniary externalities that would appear as productivity growth. In Section 5 we conclude that information on inventories of computers is critical in quantifying the role of computer services as inputs. The constant quality price indices for computers incorporated into NIPA are also essential. A price index for computers that reflects only general trends in inflation would result in a highly distorted perspective on the growth of GDP and capital services, especially during the past decade. To capture the contribution of all forms of investment to U.S. economic growth, similar price indices should be included in NIPA for capital goods with rapidly evolving technologies, as proposed by Gordon (1990). The long term goal should be a unified system of income. product, and wealth accounts, like that proposed by Laurits Christensen and Jorgenson (1973) and Jorgenson (1980). This incorporates capital stocks, capital services, and their prices. Achieving this goal will necessitate much greater elaboration of the accounting system described in Section 3. These accounts would incorporate data on prices and quantities of investment, stocks of assets, and capital services for all forms of capital employed in the U.S. economy.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 195 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.