Examining the Sensitivity of the Production Approach to Markup Estimation

Under the production approach for markup estimation, using different flexible inputs should measure the same markup. I test the production approach by comparing markups using the output elasticities of labor and materials for four manufacturing censuses and internal store-level data from a major nationwide US retailer. Across all five datasets, I find that markups using labor are negatively correlated with those using materials, exhibit much greater dispersion and different trends over time, as well as different correlations with the degree of competition faced by the establishment. I examine several mechanisms for these findings, and conclude that these findings could be consistent with plant level heterogeneity in production technology. These differences thus raise questions about whether the production functions estimated in the standard implementation of this methodology can successfully be used to measure markups.