The Identification of the Dynamic Discrete-Choice Model with Stockpiling - with Jean-Pierre Dubé and Xinyao Kong, **Revise and Resubmit at Marketing Science**

Abstract

We prove that the discount factor is set-identified for dynamic discrete-choice models of demand with stockpiling estimated using choice panels. Our key theorem generalizes the identification result in Abbring and Daljord (2020) to the case where the outside good cannot be normalized to zero, as in the case here where there is additional consumption from existing inventory. This result generates a test for consumer forward-looking behavior and the extent to which purchases during discount periods are incremental or diversion away from future periods. Out theorem nests the auxiliary restrictions in Ching and Osbourne (2020) and Akca and Otter (2015), which are testable and can point-identify the discount factor. Using Ching and Osbourne (2020)’s case study of laundry detergent, we fine a mean discount factor estimates of 0.92, rejecting myopic models of choice in favor of the dynamic model. In counterfactual simulations, we show that some of the promotional lift during a discount period is indeed due to forward-buying. However, we predict considerable more incremental sales compared to a standard discrete-choice model, which predicts that discounts mostly cause substitution between brands. These results contribut to the long-standing discussion regarding the effectiveness of price promotions.