Infrastructure Productivity with a Long Persistent Effect

Appropriate investment and management of infrastructure is an important issue in national or regional planning. Aschauer (1989) reported two important findings: that infrastructure productivity was significantly higher than expected, and that fewer investments in the USA after 1970 would result in less growth of national productivity. Since then, measurement of infrastructure productivity has become a focal issue in national or regional management policy, and a large number of empirical studies have accumulated; see Sturm (1998). An aggregate production function approach for time-series applied in Aschauer’s study, however, was criticized by economists, or econometricians. Among the assumptions inherent in the production function approach, “constant return to scale” and “competitive input factor market” are often viewed with suspicion from the viewpoint of the endogenous growth theory by Romer (1986) and Lucus (1988) or from an uncompetitive structure of infrastructure market, respectively. Basu and Fernald (1997) empirically tested these assumptions with the production function approach, based on data from 34 sectors of US industries. They concluded that in some segmented sectors of industry, the constant return to scale assumption was violated but that it held true in most sectors. Holtz-Eakin and Lovely (1996) analyzed the effect of infrastructure on economic activities by using a general equilibrium system explicitly considering scale economies. They showed that infrastructure decreases the input factor cost, increases the variety of industries, and increases the number of newly founded companies. Haughwout (2002) applied the spatial equilibrium theory considering regional monopolistic power over input factor market, and measured infrastructure productivity. This study showed that infrastructure provided significant marginal benefit on regional economic activities. Most studies have reported that there is a positive production/cost reduction effect caused by infrastructure, but unfortunately, some studies have paid little attention to the data generation process that significantly influences findings through model specifications. Recent American, European, and Japanese studies about the production function approach have been widely reviewed by Ejiri et al. (2001).