Hassle Costs and Price Discrimination: An Empirical Welfare Analysis: Online Appendix

In this appendix, I discuss the differences in the estimates of the model that I use in this paper (baseline model) with the estimates of a random coefficient logit model. Table 7 shows the results of the estimation of both models using the subsample of customers who always purchased Diet Coke products. The table shows that the coefficients on price and the refillable dummy are not statistically different across models. It also shows that most of the interaction effects between the customer characteristics and both the refillable dummy and price preserve the signs reported in Table 5. One exception is the negative coefficient on the interaction between trip frequency and the refillable dummy in the baseline estimates, which is inconsistent with Tables 3 and 5. In spite of this inconsistency, the estimates of both these models preserve the negative relationship between the price coefficient and the taste for refillables that was discussed above (see Figure 3). An interesting difference between models is how the models rely on unobserved heterogeneity to explain choice heterogeneity, which is reflected in the relative magnitude of coeffi∗University of Illinois at Urbana-Champaign, Department of Economics, 214 David Kinley Hall, 1407 W. Gregory, Urbana, Illinois 61801 (e-mail: [email protected]). Acknowledgements: I am grateful to Liran Einav, Igal Hendel, Fernando Luco, Aviv Nevo, Álvaro Parra, Esteban Petruzzello, Tiago Pires, Rob Porter, Alberto Salvo, Anthony Wray, and three anonymous referees for comments and suggestions.