In this paper, we estimate a mixed logit model for the U.S. processed cheese market and implement cost pass-through simulations and related consumer welfare analysis under different regimes of competition. In the model, the curvature of demand is determined to a significant extent by the empirical distribution of consumers. This property reduces the prediction errors of cost pass-through caused by an erroneous assumption as to the functional form of demand. We find that, under collusion, the pass-through rates for all brands fall between 23% and 37% and under Nash price competition, the range of pass-through is 67% ~ 98%.