We show that the phenomenon of Increasing Returns to Scale (IRS) can artificially emerge in the absence of any type of sorting or agglomeration effects, in a systematic, predictable and measurable way. This is in contradiction to the convention where the null hypothesis is the absence of IRS. We show the null hypothesis should instead be the presence of IRS when the variance of log-productivity is of the same order of magnitude than the log-size of the smallest observation in the sample. Our analytical results are validated through simulations and through their application to real data of wages across municipalities in Colombia. This effect can obscure the real size productivity premium, and we provide a methodology to statistically test it.
↧