5.2Excess returns from IT capital

In this section I present some evidence on the “excess” return hypothesis, which states that returns to IT capital are greater than those to traditional capital. In order to test this hypothesis, I estimated the following equation (based on equation 4.9 and 4.15, respectively):

ln(Y)its = ln(A) + α.ln(K)its + αθ(IT%)its + (1-α)ln(L)its + εits(5.3)

Introducing fixed effects and taking logarithms:

ln (Y)its = ln(A) + Σγt-1 + Σλi + Σνs + αln(K)its + αθ(IT%)its + (1-α) ln(L)its + εits(5.4)

Table 5.5 reports estimates of equations 5.3 and 5.4. The coefficients for capital and labor reach their expected constant return to scale values of 1/3 and 2/3 respectively. When no fixed effects are accounted for, θ has a value significantly greater than 5 (7.54), which leads to the conclusion that IT capital exhibits excess returns over traditional capital. Regressions with fixed industry effects show a value of θ =9, also significantly higher than 5, which means that IT capital has a return higher than that of traditional equipment. Finally, regression results at the state level also indicate excess returns to IT capital (θ =7.91), but not when state and time effects are introduced. Thus, the excess returns of IT capital may be partly due to differences across states and time.

Equation 5.3 is then estimated for selected sectors, years and states. Results appear in Tables 5.6, 5.7 and 5.8, respectively. Table 5.6 shows that, except for F.I.R.E. and transportation industries, the value of θ is significantly greater than 5, which confirms the hypothesis of excess returns to IT capital. The highest value was found in the service industry (θ = 22.2). Equation 5.3 is also estimated for each year between 1977 and 1997. Results appear in Table 5.7. First, the coefficients for capital increased over time (from 0.271 to 0.433), and the coefficient for labor decreased (from 0.693 to 0.535), but these coefficients remained close to their expected values of 0.33 and 0.66, respectively.

The increase in the coefficient for capital is probably mostly due to the increase in the returns to IT capital over time. Between 1980 and 1993, the estimated value of θ is significantly greater than 5, indicating excess returns to IT capital for these years.

Figure 5.2 represents the evolution of θ over the period 1977-1997. The first and last three years of the period do not seem to exhibit excess returns to IT capital because of a low value of θ. This is explainable by the heavy fixed costs associated with the introduction of IT capital in the economy in the late 1970s, preventing excess returns. Finally, in the early 1990s the excess returns capacity of IT capital may have been exhausted after its important price (and marginal return) declined.

Equations 5.3 and 5.4 are finally estimated for each state at the detailed and aggregated industries levels, by state. However, the elasticities estimates do not indicate excess returns to IT capital for any of the states, with a value for θ not significantly different or even lower than 5. Therefore, IT capital does seem to exhibit excess returns at the national aggregated level and at the sectoral level, but not at the state level.