Misallocation or Misspecification? The Effect of “Average” Distortions on TFP Gains Estimations

In recent years a vast literature has been devoted to estimate the degree of misallocation in different countries, sectors and time periods using Hsieh & Klenow (2009) -henceforth HK- framework. Even if we take the HK model at face value, such estimations still depend (heavily) on the assumed production technologies and elasticity of substitution. How much of the estimated TFP gain from eliminating distortions is due to actual TFPR dispersion among firms and how much is related to the specific parameterization of the model? We propose a decomposition of the inferred distortions that allows us to isolate the effect of “average” distortions (which depend on the parameterization and are defined at the industry level) from that of “dispersion” distortions (which are unaffected by the parameters and operate at the firm level). Using a newly available administrative dataset with the universe of Chilean firms between 1999 and 2015, we show that TFP gains from eliminating misallocations using the standard HK parameterization are 58% in the manufacturing sector (68% for the entire economy), but are reduced to 28% (44%) once the “mean” components of the distortions are removed. We find that the fraction of TFP gains explained by ”average” distortions increased significantly in Chile between 2000 and 2013, which is mainly explained by a sustained increase in average markups. We verify the robustness of our results using different datasets from Chile and Colombia.