Management Differences and Productivity: A Simulated Investigation into Dummy Variables in Two-Stage Data Envelopment Analysis

This paper tests the ability of two-stage DEA to pick up on efficiency differences between two groups. Using the context of franchising, I simulate establishment-level data and give franchisee-owned establishments an artificial output bump of various sizes. Using an ownership dummy variable in the second stage regression, I find that two-stage DEA is able to detect that franchisee-owned establishments are more efficient than their franchisor-owned counterparts. The results suggest that the estimated ownership effect is a lower bound on the true value. I find that franchisee-owned establishments are 4.4 percentage points more efficient than franchisor-owned establishments when a 5% ownership effect is used. Similar results are found for effects of 10%, 15%, 25%, and 50%. The results hold up remarkably well to various robustness checks. I test the results using the DEA inputs in the second-stage regression, logged DEA scores in the second-stage regression, and alternate DEA specifications. I also check my results against the commonly-used average product measure of productivity. In all of these robustness checks, the estimated effect from ownership stays steady.