In behavioral research, discrete choice models (DCM) are widely used to analyze and predict choices between two or more discrete alternatives. A central component of any discrete choice modelling is the specification of the set of alternatives that determines a decision agents consideration set. In stated preference settings, factorial design algorithms and efficiency statistics can assist in choice set design. In revealed preference settings, where actual decisions form the basis for behavioral modelling, the researcher cannot directly observe the alternatives an individual actually considered on a particular choice occasion. Beyond that, in many revealed preference decision settings, choice sets can be very large. Often this is the case in spatial decision contexts, such as travel route choices, selection of residential locations, or destination site decisions. In particular, recreational site choice studies are frequently faced with hundreds or thousand of feasible alternatives, such as lakes, forests, and angling sites. Considering the complete set of alternatives for model estimation in such situations can be computationally burdensome and sometimes even impossible. This rises the problem of how to specify choice sets in the presence of a very large number of alternatives; narrowing choice sets can be necessary but misspecification will yield inefficient and biased parameter estimates, and lead to inaccurate behavioral predictions and policy metrics.
