International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
Market Structure and Eliciting Preferences in Choice Experiments: Choice of choice sets
Matthew John Beck, Simon Fifer, John Matthew Rose

Last modified: 28 March 2017


There is an extensive literature exploring the role of market structure on consumer choices. Most applications consider a two-step characterization of the decision making process first suggested by Manski (1977), first modelling choice set formation and then choice model estimation, which has been widely applied within the choice modelling literature (e.g., Gilbride and Allenby, 2004; Swait, 2001; Andrews and Srinivasan, 1995; Roberts and Lattin, 1991; and Gensch, 1987). Previous evidence on the role of market structures and choice set formation is based on statistical estimations of econometric models, which estimate respondent choices and market structure at the same time. However, estimating a two-stage decision process is not straightforward. On the one hand, available market alternatives may not sufficiently varied enough to completely understand choice determinants. On the other hand, most stated preference studies have not been explicitly designed to estimate product preferences and market structure simultaneously, which also reduces the validity of the available estimations.

In this paper, we present the results of an empirical experiment related to vehicle choice in which respondents are first presented with four forced-choice choice sets consisting of varying numbers of alternatives, after which they are asked to complete a fifth choice set which involves indicating a preference for one of the previously shown choice sets. Finally, respondents are presented with sixth unforced choice set consisting of the four preferred alternatives from the original four choice sets.

As is to be expected, we find a positive relationship between higher levels of the expected maximum utility obtained from the original choices sets and a respondent preferring a choice set. We further find that whilst there significant heterogeneity exists, on average respondents prefer choice sets with more alternatives all else being equal; having potential ramifications for the way in which choice experiments are designed. Of particular interest however is the finding that for 57 percent of respondents, they would choose none of the vehicles selected in the force choice task suggesting that preferences obtained from such tasks are unlikely to reflect true market preferences. Further, we find the preferred vehicle is located in the preferred choice set in only 14 percent of cases, increasing to 33 percent when the no choice option is accounted for. These findings indicate that market structure preference is likely to be only moderately correlated with produce choice.

Unlike the majority of previous studies which examine error variance in the process of choice set formation and choice estimation, this paper models is novel in that we use expected utility to understand market structures and find that there is additional utility generated, above the utility of the chosen alternative, by making a choice in a market that is preferred. The results of the proposed experiment have several implications for the validity of choice experiments in different context like predicting market shares and eliciting individuals’ willingness to pay for improving product characteristics. For the policy relevance, the results may help to provide more accurate estimations of price elasticity parameters and also widen the battery of policies available for decision makers to actually influence final choice.  Additionally, the methodology for data collection that will be described within this paper can easily and parsimoniously be applied to other choice contexts.


Andrews, R.L. and T.C. Srinivasan (1995) Studying Consideration Effects in Empirical Choice Models Using Scanner Panel Data, Journal of Marketing Research, 32(1), 30-41.

Gilbride, T.J. and Allenby,  G.M. (2004) A Choice Model with Conjunctive, Disjunctive, and Compensatory Screening Rules, Marketing Science, 23(3), 391 – 406.

Gensch, D.H. (1987) A Two-Stage Disaggregate Attribute Choice Model,Marketing Science, 6(3), 223 – 239.

Manski, C.F. (1977) The structure of random utility models, Theory and Decision, 8(3), 229–254.

Roberts, J. H. and J.M. Lattin (1991) Development and Testing of a Model of Consideration Set Composition,Journal of Marketing Research, 28(4), 429-440.

Swait, J. (2001) Choice set generation within the generalized extreme value family of discrete choice models, Transportation Research Part B, 35(7), 643-666.

Conference registration is required in order to view papers.