From a theoretical and empirical standpoint, the contribution of infrastructure capital to aggregate productivity and output has been extensively researched. Public capital has been modeled as an additional input in Ramsey-type exogenous growth models and in endogenous models as well. On the empirical front, the literature has witnessed a proliferation of research over the last 20 years following Aschauer’s (1989) seminal paper on the effects of public infrastructure capital on US total factor productivity. His finding of excessively high returns to infrastructure, however, has not held up. Subsequent research using a large variety of data and more robust econometric techniques has yielded widely contrasting empirical results. For instance, Bom and Ligthart (2008) find that estimates of the output elasticity of public capital range from -0.175 to +0.917 in a wide set of empirical research for industrial countries.
In a joint paper with Enrique Moral-Benito and Luis Servén , we aim to contribute empirically by addressing some of the common problems in the literature: (a) non-stationarity of aggregate output and infrastructure capital, (b) likely simultaneity between infrastructure and income levels (i.e. fast-growing countries would devote more resources to infrastructure development), and (c) heterogeneity in the output elasticity of infrastructure across units (say, countries or states) that may result from technological aspects such as network effects and scale economies, among others. Failure to address these issues may lead to spurious and upward biased estimates of the output elasticity of infrastructure.
We estimate an aggregate production function that includes non-infrastructure physical capital, human capital and infrastructure asset inputs for a sample of 88 countries over the period 1960-2000. Our measure of infrastructure stocks involves physical measures rather than monetary ones—say, investment or capital stock figures. Why? First, there is evidence that public expenditure fails to track the evolution of public capital stock to the extent that inefficiency and corruption plagues project selection and government procurement practices (e.g., Pritchett 2000; Keefer and Knack 2007). Furthermore, our interest lies in the impact of infrastructure capital due to government retrenchment in the sector and the accompanying rising participation of the private sector in infrastructure worldwide since the 1990s.
Our approach to estimating the contribution of infrastructure to output involves:
a. Accounting for the multidimensionality of infrastructure. Instead of focusing on the contribution of a single infrastructure sector, we use principal components analysis to summarize information about three core sectors (telecommunications, power and transportation) in a synthetic infrastructure index.
b. Dealing with the non-stationarity of income and infrastructure capital. We use panel cointegration techniques to estimate our infrastructure-augmented production function and avoid spurious regression problems.
c. Addressing concerns with identification. We find only one cointegrating relationship between output and inputs for all countries in our panel. Exogeneity tests allow us to identify this relationship as an aggregate production function.
d. Accounting for potential reverse causality and heterogeneity of long-run parameters. To accomplish this task we use the Pooled Mean Group estimator, PMGE (Pesaran, Shin and Smith, 1999). This technique imposes homogeneity of long-run parameters and allows cross-sectional heterogeneity of short-term dynamics. Parameter constancy in the long-run production function is then examined using individual and joint Hausman tests.
Our main findings can be summarized as follows: First, we find very precise estimates of the output elasticity of infrastructure that are robust to changes in the econometric specifications and the synthetic index of infrastructure. Our estimates lie in the range of 0.07 and 0.1, in line with the estimates from the meta-study by Bom and Ligthart (2008) after adjusting for publication bias (around 0.086). Second, there is little evidence of cross-country heterogeneity in the output elasticity of the different infrastructure inputs. In fact, there is no cross-country variation in the output elasticity of infrastructure relative to the countries’ income per capita levels, infrastructure endowment, or population size. Third, our estimates are both statistically and economically significant. As an illustration, let us consider an increase in the level of infrastructure provision from the cross-country median in 2000 (an index of -4.65 that roughly corresponds to the level of Tunisia in that year) to the top quartile in our sample (an index of -3.69). This translates to a 7.7 percent increase in output per worker. Finally, homogeneity tests show that there is little cross-country variation in the output elasticity of infrastructure. This finding implies that observed differences in the ratio of aggregate infrastructure to output across countries provide a useful guide to the differences in the marginal productivity of infrastructure.
Aschauer, D. (1989) “Is Public Expenditure Productive?” Journal of Monetary Economics 23, 177-200.
Bom, P.R.D., and J.E. Ligthart (2008) “How productive is public capital? A meta-analysis,” CESIfo Working Paper 2206, January
Calderón, C., E. Moral-Benito, and L. Servén (2011) “Is infrastructure capital productive? A dynamic heterogeneous approach,” The World Bank Policy Research Working Paper No. 5682.
Keefer, P. and S. Knack (2007): “Boondoggles, Rent-Seeking, and Political Checks and Balances: Public Investment under Unaccountable Governments”, Review of Economics and Statistics 89, 566-572.
Pesaran, H., Y. Shin, and R. Smith (1999) “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels,” Journal of the American Statistical Association 94, 621-634
Pritchett, L. (2000) “The Tyranny of Concepts: CUDIE (Cumulated, Depreciated, Investment Effort) Is Not Capital.” Journal of Economic Growth 5 (4), 361–84