Demand Estimation with Social Interactions and the Implications for Targeted Marketing

This paper develops a model for the estimation and analysis of demand in the context of social interactions. Decisions made by a group of customers are modeled to be an equilibrium outcome of an empirical discrete game, such that all group members must be satisfied with chosen outcomes. The game-theoretic approach assists estimation by allowing us to account for the endogeneity of group members' decisions while also serving as a managerial tool that can simulate equilibrium outcomes for the group when the firm alters the marketing mix to the group. The model builds upon the existing literature on empirical models of discrete games by introducing a random coefficients heterogeneity distribution. Monte Carlo simulations reveal that including the heterogeneity resolves the endogenous group formation bias commonly noted in the social interactions literature. By estimating the heterogeneous equilibrium model using Bayesian hierarchical Markov chain Monte Carlo, we can also recover some parameters at the individual level to evaluate group-specific characteristics and targeted marketing strategies. To validate the model and illustrate its implications, we apply it to a data set of groups of golfers. We find significant social interaction effects, such that 65% of the median customer value is attributable to the customer and the other 35% is attributable to the customer's affect on members of his group. We also consider targeted marketing strategies and show that group-level targeting increases profit by 1%, whereas targeting within groups can increase profitability by 20%. We recognize that customer backlashes to targeting could be greater when group members receive different offers, so we suggest some alternatives that could retain some of the profitability of within group targeting while avoiding customer backlashes.