Increasing Returns to Scale in U.S. manufacturing industries: evidence from direct and reverse regression.

In this paper, I compare the OLS and IV estimators for the direct and reverse regression models in the context of estimating returns to scale and technical progress. It shows that the direct and reverse OLS estimators are inconsistent, that the direct OLS is always more precise than the reverse OLS under the normality assumption, and that the direct IV estimator and its reverse counterpart are consistent and asymptotically equivalent. Working with data from U.S. manufacturing industries over the last half-century, the estimation results show that in most industries increasing returns to scale are important and technical progress is small when it comes to explaining productivity growth.