Measuring market power from plant-level data

Abstract

Measuring the degree of market power has been the object of many applied studies in the industrial organization field. But, as this paper points out, identification problems arise when we estimate markups from production function regressions using data sets that do not report firm or plant-level physical quantities of output. In a differentiated product industry, the lack of such information introduces an unobserved (price) heterogeneity term. In this paper, I set up an econometric model that controls for this unobserved term and shows that failing to do so leads to spurious markup estimates. I illustrate this result using data from Colombian plants. The results reveal that, if we do not control for (price) heterogeneity, we will find misleading evidence of firms with little or no market power.