CHAPTER 1 - INTRODUCTION

This dissertation analyzes the role of information technology in the economy of the United States, through its effects on regional labor productivity. The concept of information technology refers here to the convergence of computing power and communication technology that began in the late 1970s. Information technology (hereafter IT) can be embodied in a certain type of capital stock or in employment inputs or in both.

This study was motivated by the debate over the so-called “productivity paradox,” the oft-cited finding that investment in information technology appears to have had no visible effect on aggregate productivity. Indeed, until the mid-1990s, productivity gains remained sluggish while information technology was booming. Today, even after the recent jump in productivity, the strength of the “new economy” is once again called into question with the “deceleration” of growth and the apparent failure of the “e-economy.” During this last decade, many authors proposed explanations for the productivity paradox:

First, Ives(1994), Brynjolfsson and Hitt (1998) questioned the quality of measurement of national figures. Moving from a “hard” to a “soft” economy, with knowledge and information becoming primary resources, the productivity of difficult-to-measure intangibles has become more difficult to estimate. Second, David (1990) argued that long learning lags are associated with the diffusion of a new technology. The parallel was drawn from previous technological revolutions such as electricity or steam power, which had no significant impact on aggregate productivity figures until several decades after their discoveries. Roach (1998), Powell (2000), Chapman (1996) and Pentland (1989) proposed the “mismanagement” hypothesis, which stated that investors in information technology have underestimated its true cost (hidden costs include maintenance and training). A fourth hypothesis stated that unless IT investment is accompanied by work reorganization, productivity improvement will not occur [Brynjolfsson and Hitt (1995), Bowen (1986)]. Another explanation for the productivity paradox was the “redistribution” hypothesis, which proposed that IT is beneficial for individual firms, but not necessarily for the nation as a whole, as shown by Brynjolfsson and Hitt (1993), Jorgenson and Stiroh (1999, 2000).Finally, Oliner and Sichel (1994, 2000) argued that the income share of information technology capital is too small to have had visible macroeconomic effects, even if it does exhibit excess returns at the microeconomic level. Each hypothesis is a possibly valid explanation for the productivity paradox. However, today there seems to be a consensus around the idea that information technology finally started to increase productivity in the mid-1990s, as more and more firms completed the long reorganization of work process needed to accompany IT investment.

This study re-examines the historical data (1977-1997) in light of the redistribution hypothesis, which may deserve further investigation at the regional level. If information technology has a redistribution effect, then it “redistributes the shares of the pie without making it bigger,” as stated by Brynjolfsson and Hitt (1993), who showed that IT capital does increase an individual firm’s productivity. Thus, the slow diffusion of information technology across firms as well as across space may partly explain the productivity paradox.

The purpose of this dissertation is to further investigate the impact of the spatial diffusion of IT on the validity of the productivity paradox, by analyzing the productivity of IT at the regional level. Because information technology activity tends to be very localized (eight states own more than half of the entire national IT capital stock), there is reason to hypothesize regional redistribution effects regarding the impact of information technology. If this hypothesis is confirmed, then the productivity paradox is shown to have been a problem only at the aggregate level.

This dissertation is composed of three essays, each dealing with a particular feature of the regional relationship between information technology and productivity. First, though, chapter 2 offers a deeper presentation and statement of the problem. Then, the first essay, articulated in chapters 3, 4 and 5, analyzes information technology embodied in the stock of capital. A panel dataset is constructed for the 50 states plus the District of Columbia, covering 52 industry categories from 1977 through 1997. The data come from the U.S. Bureau of Economic Analysis. This dataset is separately analyzed for both industries and states. Using production function regressions and growth accounting techniques, the productive capacity and growth contribution of the IT capital stock is estimated at the state level. The results indicate a positive contribution to state productivity growth that amounts to 10% of the observed growth. Furthermore, decreasing returns to capital accumulation are found to apply to information technology capital, since its growth contribution is lower in states that own the highest shares of IT capital.

The second essay, in chapters 6, 7 and 8, analyzes IT and productivity at the state and county levels for the year 1990. Its purpose is to identify the productivity effects of the location patterns of IT employment across counties. The dataset built in the first essay is used to estimate output and capital stock at the county level. Data on employment by industry at the county level come from the U.S. Bureau of Census. Twenty-one of the 2-digit SIC industries, among those used in the first essay, are identified as IT industries. The concentration, localization and density of IT employment are found to have significant positive effects on labor productivity. Indeed, the agglomeration effects associated with the spatial distribution of IT employment are found to be greater than those associated with traditional employment. Between 5% and 10% of productivity differences can be explained by the spatial distribution of IT employment.

In the third essay, in chapters 9, 10 and 11, the impact of information technology on income inequality across states is estimated for the year 1990. The motivation for this analysis is the simultaneous increase in information technology and income inequality that took place in the 1980s (Greenwood 1999). To investigate empirically this relationship, Gini coefficients and variances of logarithms of median incomes are used alternatively as dependent variables, regressed on various factors explaining income inequality at the state level. According to Lervernier, Rickman and Patridge (1995), these factors include economic, demographic and human capital variables. IT intensity variables are then added to the model and are found to significantly increase income inequality. This result may be explained by the substitution/complement effects of information technology, as described by Krueger (1993), Levy and Murnane (1996) and Autor, Katz and Krueger (1998). Finally, chapter 12 summarizes findings, discusses outcomes and concludes the dissertation