Choice Probability Generating Functions
This paper considers discrete choice, with choice probabilities coming from maximization of preferences from a random utility field perturbed by additive location shifters (ARUM). Any ARUM can be characterized by a choice-probability generating function (CPGF) whose gradient gives the choice probabilities, and every CPGF is consistent with an ARUM. We relate CPGF to multivariate extreme value distributions, and review and extend methods for constructing CPGF for applications.
We are grateful for comments from Rosa Matzkin, Richard Blundell and the participants at the Conference on Demand Analysis and Welfare Measurement held in London, 2011. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.