2 The Substitution of Information Technology for Other Factors of Production: A Firm Level Analysis Authors Sanjeev Dewan and C. Min Working Paper #ITR-127 199631. Introduction The phenomenal price and performance improvements in information technology (IT), that fuel the "information revolution," have also led to the overarching displacement of non-IT inputs in the production of goods and services. Over the past few decades, IT has accounted for an increasing share of new investments by the U.S. business sector. Based on the BEA asset category "Office, Computing and Accounting Machinery," IT investment has increased at the average rate of 7.4% per annum (1990 dollar basis) in the period from 1960 to 1994. By comparison, the overall output of the economy (GNP) has increased at only about 3% per annum. Figure 1 shows that the share of IT in nonresidential fixed investment has grown from 0.4% in 1960 to 15.5% in 1994, while the ratio of IT investments to labor costs, has increased from 0.4% in 1980 to 4.7% in 1994.1 IT intensity, as measured by both IT investment per employee, and IT as a percentage of total fixed investments (Figures 2 and 3, respectively), has steadily increased in every sector of the economy.2 The fact that the financial sector appears to be a clear leader in the growth of IT intensity is not surprising. It was among the first to incorporate electronic data processing in its operations, through check handling, bookkeeping, credit analysis and automated teller machines (Franke 1987). Across all industry sectors, the earliest applications of IT were directed toward the reduction of personnel costs in such labor-intensive operations as accounting, purchasing and payroll. Since then, computing and communications technologies have affected all functions of the modern corporation, gradually displacing traditional labor and capital inputs, through such applications as electronic order entry, electronic data interchange, office automation, telecommuting, expert systems, robotics, object-oriented programming, and point-of-sale settlement, among others. The recent trend of IT-led restructuring and reengineering has accelerated the transformation of business processes. By substituting IT for labor or ordinary capital, organizations seek to capitalize on the vastly superior price and performance improvements in IT relative to these other inputs. According to Lau and Tokutsu (1992), the quality-adjusted price of computer assets has _____________________________ 1The BEA Office, Computing and Accounting Machinery (OCAM) asset category primarily includes computers and peripheral equipment, while nonresidential fixed investments include, in addition to OCAM, durable and transportation equipment, nonresidential structures, furniture and fixtures. These are available in both current and constant dollars from BEA (1995). Labor cost is taken to be the aggregate wage and salary disbursements by the private sector, deflated by the Employment Cost Index, which is only available for the period 1980-1994. 2The graphs in Figures 1-3 would be substantially flatter in later years if the quantities were measured in nominal terms (i.e., in current dollars).4decreased at the average annual rate of 20%, relative to the price of other capital goods, over the period 1960-1992. The ability to take advantage of the economic opportunities created by improvements in IT is determined, in part, by the substitutability of IT for other inputs. The primary purpose of this paper is to estimate IT substitutability, relating it to structural firm characteristics on the one hand and to the productivity of inputs on the other. We develop a flexible production function framework to jointly estimate output and substitution elasticities, using a recent sample of firm level data. A pioneering work in the context of input factor substitution is the development of the constant elasticity of substitution (CES) production function by Arrow et al (1961). They used the CES production function to demonstrate that the elasticity of substitution between capital and labor can be substantially different from unity, as assumed by the prevalent Cobb-Douglas production function. This has important implications for international trade and for the returns and factor shares of inputs. Later, more flexible production functions were developed, including the translog specification of Christensen et al (1970) and the CES-translog introduced by Pollak et al (1984). These flexible functional forms have been used to examine the degree of substitutability or complimentarily between various inputs, including capital, labor, materials, and energy (see Berndt 1991 for a review). Our contribution to this line of research is to formally test whether IT is a net substitute or complement for the traditional inputs of ordinary capital and labor, and to examine the implications of our findings for the growth in IT intensity in different sectors of the economy. The "productivity paradox" literature, surveyed by Brynjolfsson (1993) and Wilson (1993), generally failed to reject the null hypothesis of zero IT contribution. However, there is growing new evidence that IT investments in fact generate large positive returns. Brynjolfsson and Hitt (1995,1996) and Lichtenberg (1995) analyze firm level data to show that IT investments generate returns that are often in excess of the returns on other types of investments. While these studies go a long way toward settling the debate on the productivity paradox, a number of questions regarding IT investments and productivity remain unanswered. For example, what types of firms and industries are in a better position to take advantage of the "information payoff," which is in part the focus of our study. We use the same Computerworld survey data as Brynjolfsson and Hitt, and Lichtenberg, to jointly estimate output and substitution elasticities. Our analysis confirms earlier findings of positive marginal returns to IT investments. We also find evidence of excess returns on IT investments, more so relative to labor than with respect to ordinary capital. Our analysis suggests that IT capital is a net substitute for both ordinary capital and labor, in all sectors5of the economy. One implication of this finding is that the slow improvements in labor productivity in many sectors of the economy may be due to the offsetting effects of growing IT capital but shrinking non-IT capital per worker.