Algorithms for testing the collective consumption model

Conference (2012, November 09)

In this talk, we discuss an extension of the strong axiom of revealed preferences to collective households. The question that we address is whether a set of observed consumption baskets can be decomposed ... [more ▼]

In this talk, we discuss an extension of the strong axiom of revealed preferences to collective households. The question that we address is whether a set of observed consumption baskets can be decomposed in such a way that each of the derived data sets reflects the choices of a “rational” (i.e., utility-maximizing) individual member of the household. Although testing revealed preference axioms on data generated by a single decisionmaker can be done in polynomial time, the extension to two-member households is NP-complete. We propose two algorithms for testing the collective consumption model on large data sets. The first one is an exact algorithm based on a new mixed-integer programming formulation, whereas the second one is a heuristic based on a simulated annealing procedure that solves a global optimization formulation of the problem. Computational experiments are performed on real-life data. [less ▲]