Consistent nonparametric hypothesis tests with an application to Slutsky symmetry

A general consistent, nonparametric testing technique is applied to U.K. survey data to test Slutsky symmetry. This paper first provides a set of low-level technical assumptions that yield root-N convergence of averages of functions of both kernel estimators and their derivatives. The results are then used to construct nonparametric moment-based tests, including the extension to consistent tests i.e., tests that asymptotically have correct size and have power against all relevant alternatives. The nonparametric tests are compared to ordinary parametric tests from the QUAIDS (Quadratic Almost Ideal Demand System) functional form.