CHAPTER 12 - CONCLUSION

This dissertation supports the view that the productivity paradox was only a problem at the national aggregate level. Evidence presented in the preceding chapters shows that redistribution certainly played a role in producing the paradox. Analysis of the redistribution hypothesis in the regional dimension shows that there were redistribution effects across states.

Using a panel dataset covering the U.S. states and a comprehensive set of industries during the period 1977-1997, production function regressions show that the elasticity of output with respect to IT capital was positive and significant throughout the period, although it was sensitive to industry fixed effects. This elasticity was the highest during the early 1980s, and varies across states. Information technology was also found to have exhibited excess returns across industries. Because of a small income share (around 10%), the output growth contribution of IT capital goes up to 15 percent per year from 1977 to 1997. This value is very close to the 16 percent estimated by Oliner and Sichel (1994). However, my estimates vary between 5 and 15 percent across states, supporting the regional redistribution hypothesis. Furthermore, the contribution to labor productivity growth from information technology capital varies from 5 to 10 percent across states. The surprising fact is that states that own the largest shares of the national stock of IT capital also exhibit some of the smallest contributions of this type of capital to productivity growth. The highest contributions were found for Colorado, Delaware, Georgia, Washington and New Mexico, while the bottom-ranked states included New York, Washington D.C., North Carolina and Indiana. Hence, the paradox at the aggregate level can be attributed to the fact that the states that accounted for the largest volume in the aggregation process also had the smallest contributions. This fact can be explained by convergence theory, which states that as capital accumulates, the speed of convergence is reduced.

Considering information technology as a special type of employment, this study shows that there are agglomeration externalities associated with the spatial distribution of IT employment. These externalities of IT employment are usually higher than those associated with traditional non-IT employment. Indeed, in a given state the concentration of IT employment can explain up to 10% of the differences in state labor productivity, holding the concentration of traditional non-IT employment constant. Similarly, at the county level, 5% of labor productivity differences can be explained by the location and density of IT employment relative to traditional non-IT employment. The strongest agglomeration effects of IT employment come from localization economies. Therefore, since information technology employment tends to be very localized across states, the effect of agglomeration economies varies across space, further supporting the regional redistribution hypothesis.

Finally, evaluating the effects of information technology on income inequality across space indicates a state Gini coefficient elasticity of 7% with respect to the percentage of IT employment. The density of IT employment is also found to increase state income inequality, whereas the density of traditional non-IT employment decreases it.

Therefore, on one hand I found that the stock of IT capital and IT employment have positive and significant effects on productivity across states. On the other hand, the intensity of information technology in a given state is associated with higher income inequality, showing that there is indeed a “digital divide” regionally.

Consequently, policy recommendations at the state level would be (1) to facilitate investment in information technology capital, (2) to favor the concentration and the density of employment in IT intensive industries as opposed to traditional non-IT industries, and (3) to control the negative effects on income inequality with various training and social programs.

Future research should be oriented towards further investigations at various disaggregated levels such as the firm, industry or city level, using panel data analysis. For data availability reasons, this dissertation addresses a period running only through 1997, but extended research over more recent periods should be undertaken. Indeed, productivity has picked up only since 1996, starting an astonishing period of growth similar to the 1960s. Following Oliner and Sichel (2000) as well as Jorgenson and Stiroh (2000), more investigations on the contribution to growth from information technology during these last five years must take place. Furthermore, the role that the stock market (especially the NASDAQ) has played in influencing the availability of IT capital during this period of growth should be evaluated. Finally, the recent “deceleration” of growth observed in the last six months (since the last quarter of 2000), combined with the semi-collapse of the “dot com economy,” constitutes a challenge for the most enthusiastic IT researchers.