Production function regressions, returns to scale, and externalities

A number of recent papers have used simple linear regressions in an attempt to identify market structure, the extent of returns to scale, and possible external effects in U.S. manufacturing industries. The results obtained from these regressions have important implications for several branches of modem macroeconomics. As a result, the macro literature frequently cites specific numerical evidence from Caballero and Lyons (1992) and Hall (1990), which suggests that there are quantitatively significant increasing returns to scale, or external effects in U.S. manufacturing. In contrast, it is the argument of this paper that this evidence is not convincing. The most robust evidence suggests that the typical U.S. manufacturing industry displays constant returns with no external effects. On the other hand, there is significant heterogeneity across industries.