Mixed Logit Analysis of Carrier Market Share with Stated-preference Data

Abstract

Stated-preference survey techniques have become established as a means to elicit consumer preferences, particularly when a product is new or only recently exposed to competitive pressure. This development has proved particularly helpful for studying consumer behavior, both actual and potential, in telecommunications markets newly open to competition. The approach is well suited to econometric modeling via discrete choice methods—particularly when data reflect consumer ranking of alternatives in order of preference. The rank-ordered logit model has become a standard tool, particularly the mixed logit form for rank-ordered data. The mixed logit model accommodates variation in consumer response to product attributes and correlation among unobserved factors in consumer choice, e.g., when consumers make repeated choices. Mixed logit models are flexible and permit the estimation of random coefficients, and reduce to a standard logit model should empirical tests show that estimated coefficients possess degenerate distributions. This feature helps to provide insights into certain aspects of consumer behavior that hitherto have only been analyzed by standard logit models. In Particular, the mixed logit model implied much larger market share elasticity estimates for incumbent service providers. Therefore, the combination of ranked stated-preference data and mixed logit models can be a valuable tool for market analysis as the dramatic transformation of the telecommunications industry continues.