Abstract: The pure stochastic NBD-Dirichlet model can be applied to study consumers' retail store choice behavior(store choice and shopping frequency).However,due to lack of covariate variables in the model,it cannot explain consumers'store choice behavior.This article extends the pure stochastic model by incorporating explanatory variables,and applies the extended model to empirically study consumers'supermarket choice behavior in Beijing.The major findings include:(1) Market penetration is the critical indicafor of supermarket performance;(2) Promotion has no effect on consumers'supermarket choice,but it has positive significant effect on consumers'shopping frequency;(3) Household income has significant effects on both supermarket choice and shopping frequency;(4) Amount of expenditure per store trip has no effect on store choice,but it has negative significant effect on shopping frequency;(5) There is little price variation among supermarkets so price has no effect on consumers'shopping frequency;and(6) Product assortment has tremendous impact on consumers'supermarket choice.

Key words: NBD-Dirichlet model, retail store choice, buying behavior