Bargaining, searching and price dispersion in consumption good markets

Abstract

In consumption goods markets, we observe both bargaining and searching. However, in this literature, very little work has been done to incorporate both features into one model. This study addresses this problem. In my first chapter, I add a bargaining parameter to a traditional sequential search model and solve for the new equilibrium in this set-up. Then, I do some comparative statics, changing the distribution of the bargaining parameter to see what happens to the equilibrium. Finally, I use the model to explain two seemingly contradicting empirical works in the literature of discrimination in the auto market. Ayres and Siegelman (1995), using data they collected from a controlled experiment, found that the initial offers for the minorities are higher. Yet Goldberg (1996), using consumer expenditure survey data (CES), reported that there is no significant difference between the final prices for minorities and non-minorities. My model reconciles these two results and shows that if minorities have a more dispersed bargaining parameter distribution and if the final transaction prices are the same at the mean level, then the initial offer distribution for the minorities first-order stochastically dominates that for the non-minorities. In my second chapter, I investigate how the bargaining process affects firms’ offer distribution and thus the final price distribution. Based on Varian (1980), I add a bargaining parameter into the model, and solve for the new equilibrium in this set up. Then, I do some comparative statics, changing the distribution of the bargaining parameter to see what would happen to the equilibrium. This model yields the same results as the first chapter. In the third chapter, I applied my theoretical model to the automobile market, and empirically test the model. I used CES data, and my findings support the theoretical model. The minority dummies are not significant in determining the mean level of consumers’ bargaining ability distribution, but are significantly positive in determining the variance of the distribution.