A Simple Globally Consistent Continuous Demand Model for Market Level Data

Abstract

This paper considers a new method of uncovering demand information from market level data on differentiated products. In particular, we propose a globally consistent continuous-choice demand model with distinct advantages over the models currently in use and describe the econometric techniques for its estimation. The proposed model combines key properties of both the discrete- and continuous-choice traditions: i) it is flexible in the sense of Diewert (1974), ii) it is globally consistent in the sense it can deal with the entry and exit of products over time, and iii) incorporates a structural error term. In order to encompass different possible real-world applications, we consider two alternative specifications of the baseline model depending on the degree of flexibility the researcher is willing to accept for the substitution patterns between inside and outside goods. The estimation procedure follows an analog to the algorithm derived in Berry (1994), Berry, Levinsohn and Pakes (1995). Depending on the specification considered, the contraction mapping for matching observed and predicted budget shares may be analytical or not. The case for which the contraction is analytical is relatively simple and fast to estimate which can prove a key advantage in competition policy issues, where time and transparency are typically crucial factors. For the case it is not, we propose an alternative to Berry, Levinsohn and Pakes (1995)'s contraction mapping with super-linear rate of convergence. The final sections provide a series of Monte Carlo experiments to illustrate the estimation properties of the model and discuss how it can be extended to cope with consumer heterogeneity and dynamic behaviour.