We reexamine Robert Hall's recent finding of increasing returns in U.S. manufacturing. With gross output data at roughly the two-digit SIC level, we estimate that returns to scale are close to constant. We show why, with imperfect competition, value-added data lead to biased estimates of returns to scale, and hence why Hall's results are easily explained. We show how to control for value-added bias by combining Hall's data with data on intermediate input use: using Hall's data with this correction, returns to scale again appear to be constant. We also estimate that the average markup of price over marginal cost is about 15 percent.
Center for Research on Economic and Social Theory, Department of Economics, University of Michigan
http://deepblue.lib.umich.edu/bitstream/2027.42/101020/1/ECON045.pdf