UCLA

This collection of essays considers three issues regarding the performance of the structural merger simulation. The first chapter addresses the underprediction problem of merger simulation by considering more flexible demand models. To be as flexible as possible in modeling the demand, I estimate the full variance--covariance structure of brand preferences and let the brand intercepts to be different across cities to exploit the multi--market structure of data. Compared to a simple logit demand model, a flexible setting allows the predicted price changes to be closer to the actual price changes. In Chapter 2, because the underprediction problem still exists with the flexible structural demand model, I consider the Almost Ideal Demand System as an alternative demand model to determine how the simulation performance differs from that of Chapter 1. I also address the firm-side modeling issues in Chapter 2. I examine, whether incorporating the retailer side in the pricing game would produce improved merger simulation results. In Chapter 3, I separately estimate the effect of several factors compounded in the actual price changes in the post--merger period. I explain that part of the reasons the merger simulation generally fails to result in accurate predictions is that both demand--side factors and supply--side factors considerably affect the post--merger prices.