Testing the stability of preferences: a nonparametric approach

A test for preference stability is developed that strengthens existing nonparametric procedures. The test uses indifference curve convexity to restrict (unobservable) compensated consumption bundles. Adding up, noninferiority, and the Slutsky equation are used to limit the range of these compensated consumption bundles. A program is proposed that simultaneously measures the changes in consumption quantities satisfying the theoretical restrictions and the expenditure elasticities minimizing the required changes. The program is applied to meat consumption data and is shown to be capable of detecting small changes in preference.