Using the Manufacturing Productivity Distribution to Evaluate Growth Theories

Some multi-sector endogenous growth models make strong predictions about productivity differences across sectors in the form of a distribution or density function. In this paper it is demonstrated that this distribution is left-skewed for a wide range of plausible parameter values. This stands in strong contrast to the right-skewed shape of the respective empirical distribution estimated by kernel methods for a measure of relative productivity for more than 450 four-digit U.S. manufacturing industries during 1958-96. This difference is interpreted as evidence in favor of devoting more emphasis on the effects of structural change on the sectoral level in growth models.