Estimating Multiple-Discrete Choice Models: An Application to Computerization Returns

Abstract

Buyers in many markets face multiple-discrete choices: they can purchase multiple-units as well as multiple-brands at the same time. This paper presents a multiple-discrete choice model for the analysis of differentiated products demand. Users maximize profits by choosing the number of units of each brand they purchase. I estimate the model using micro-level data on the demand for personal computers. I use the estimated demand structure to assess the welfare gains from and technological innovation in peripherals. The estimated return on investment in personal computers is 92%. Moreover, a 10% increase in the performance-to-price ratio of microprocessors leads to a 2.2% gain in the estimated user surplus. The personal computer (PC) market is of interest to industrial organization economists because of its size, the continual technological change, and above all, because of the farreaching implications of PCs on the economy as a whole. Several distinct features make PC demand difficult to model. PCs are a differentiated durable good in constant technological evolution. Moreover, buyers make multiple-discrete choices. Firms, according to our data, spread their purchases of PCs over different brands. At the same time they buy PCs in blocks, several units of each brand that they have chosen. The aim of the paper is to present a model that captures these features and to use this model to estimate the welfare gains from computerization. Classical discrete choice models allow one choice among mutually exclusive alternatives. In contrast, the proposed framework allows users to demand multiple units and multiple brands. The model can be applied to other cases of multiple-discrete choices. e.g. the composition of airline fleets, car holdings per household, the fields of the faculty in economics departments, household cereals consumption, or any other good consumed in more than one variety during a relevant time period. Although pervasive, multiple-discrete choice problems have received little attention because of their complexity. A central issue regarding the PC market is the current controversy in the economics literature about the computerization This puzzle stems from the empirical finding that has had no effect on firm productivity (Griliches and Siegel (1991) and Morrison and Berndt (1990)). Contrary to those results, Hendel (1994) and Lichtenberg (1995), using more disaggregate data than most previous studies, report excess returns to computerization. This indicates that data aggregation may have created the puzzle. In this paper, I present a model consistent with the basic