Does Merger Simulation Work? A Natural Experimentin the Swedish Analgesics Market

We exploit a natural experiment associated with a large merger in the Swedish market for analgesics (painkillers). We confront the predictions from a merger simulation study, as conducted during the investigation, with the actual merger e¤ects over a two-year comparison window. The merger simulation model is based on a constant expenditures speci…cation for the aggregate nested logit model (as an alternative to the typical unit demand speci…cation). The model predicts a large price increase of 34% by the merging …rms, because there is strong market segmentation and the merging …rms are the only competitors in the largest segment. The actual price increase after the merger is of a similar order of magnitude, but in fact even larger: +42% in absolute terms and +35% relative to the “control group”of non-merging rivals. These …ndings at …rst sight suggest strong support for merger simulation and structural models more generally. But a closer look at a wider range of merger predictions leads to more nuanced conclusions. First, both merging …rms raised their prices by a similar percentage, while the simulation model predicted a larger price increase for the smaller …rm. Second, the merging …rms’market shares dropped, as predicted by the model, but some of the outsider …rms’market shares also dropped (in favor of other …rms). We discuss the possible reasons for the divergence between the predicted and actual e¤ects.