The paper introduces preference persistence into a dynamic discrete choice model of demand for durables. This persistence may arise, for example, when the products can be categorized into a few number of formats, which involve special knowledge, maintenance and upgrade. The standard optimal stopping problem of when to buy a new product is completed by the upgrade problem. Customers who already have a product may choose to upgrade it, but this upgrade is format specific. Hence, the expected future upgrade qualities of different formats must be taken into account already at the purchase decision of a new product. The model is estimated on a data set of low-end computer servers, where formats are represented by operating systems. For this application, the model can be considered as a proxy of a computer network building customer who cares not only about the likely future quality of the individual computers, but also about the direction of evolution of the network. That induces even stronger forward looking behavior than a simple optimal stopping problem. The results suggest that the model is better able to capture main tendencies in the segment than a static or a simple optimal stopping model.
preference persistence differentiated durables Markov process computer servers