An empirical comparison of nonparametric and parametric Engel functions

Abstract

Nonparametric techniques are used to estimate income densities, Engel functions and their derivatives, and income elasticities. Nonparametric kernel methods give less ambiguous estimates of the densities than discrete, maximum penalised likelihood estimation. The nonparametric estimates of the Engel functions, their derivatives, and the income elasticities are compared with some corresponding parametric estimates. The nonparametrically estimated Engel functions provide a better fit to the data. Also, the nonparametric elasticity estimates are qualitatively similar to the parametric estimates. Some statistically significant differences are observed between the parametric versus nonparametric estimates, but not to the extent of having any major policy implications.