This paper maintains that the commonly used measure of the aggregate stock of public capital is conceptually divergent from the "true" public capital input in private production technologies. Consequently, the published measure has the potential to misrepresent the historic growth profile of the public capital input and the productivity of government purchases in general. As a test of this hypothesis, the identifying assumption of a standard stochastic growth model is combined with observations on output, consumption, labor hours and government purchases to deduce public capital paths that are mutually consistent with observed flows and economic theory. It is found that the inferred series grows more rapidly than the published series during the 1973-1993 period thereby providing evidence that government under-investment is not an important source of declines in U.S. productivity growth.
