Production function analysis of the rate of return on public capital.

摘要

The purpose of my thesis is to shed light on the controversy surrounding the rate of return on public capital. Spurred on by the work of Aschauer (1989), a number of studies have examined the relationship between infrastructure capital and private sector output and productivity. In studies carried out with aggregate data, the estimated elasticities imply that the rate of return on public capital is on the order of 70%, considerably larger than the rate of return on private capital. This suggests that the US has grossly underinvested in public capital. In my thesis I build a case for the hypothesis that the actual rate of return on public capital is equal to, or less than, the rate of return on private capital, and that results based on aggregate data reflect various econometric problems. I conclude that the aggregate regression results constitute, at best, very weak evidence for the proposition that the US should expand public sector investment.; My thesis is divided into two parts, reflecting the two econometric issues that I investigate. In the first part, I attempt to quantify the impact on the aggregate regressions of simultaneity bias. In the second part I attempt to quantify the impact of distortions arising from aggregation and the fact that aggregate regressions are based on a very short sample of data.; To quantify the likely distortions from simultaneity bias, a real business cycle model with productive public sector capital is constructed. Using this model as a data generating mechanism, Monte Carlo experiments confirm the presence of simultaneity bias in aggregate regressions but its magnitude is not large enough to explain rates of return on the order of 70%.; One widely known fact which suggests there may be other econometric problems, in addition to simultaneity bias, is that rate of return estimates based on state level output and input data tend to be much smaller, and insignificantly different from zero. I use the information in state-level data to investigate the small sample distribution of regression coefficient and standard error estimators in aggregate regressions. The Monte Carlo experiments articulate the view that existing aggregate regression analyses overstate the precision with which the output elasticity of public capital is estimated.