The antitrust logit model: An application to spaghetti sauce mergers

This paper provides a derivation, description and econometric explanation of how to estimate the Antitrust Logit Model used to simulate price changes in mergers. It tests the predictions of the ALM against those from the Rotterdam model, which does not impose the Independence of Irrelevant Alternatives assumption. Bootstrap confidence intervals are used to determine the significance of the predicted price changes. The ALM is compared to the Herfindahl-Hirschman index to see if there is any advantage to using the ALM to determine which mergers should be challenged. The ALM does not provide any gain in information when compared to the structural approach of the HHI. Comparing the ALM to the Rotterdam model I find that the Rotterdam model is a better predictor of price changes.