Gastos Domiciliares nas Regiões Metropolitanas: Uma Análise não Paramétrica Kernel usando Dados de Pseudo-Painel

Abstract

The objective of this paper is to analyze the relationship between the share spent on food (and non-food) items and total household expenditures from non-parametric kernel regression models. To achieve this objective, we used data from a pseudo-panel constructed from Household Budget Surveys from 1987 to 2009. Our main findings suggest that results from parametric models were not significantly different from the average non-parametric estimates for the different consumption groups analyzed. However, when the nonparametric estimates in the terms of percentiles are analyzed, we found an impact substantially different from the total expenditure on the share spent mainly on food and health and personal expenses, but not in other expenditures. Moreover, the hypothesis that the use of panel data models, that takes into account the unobserved and time-invariant characteristics of the heads of households would lead to different results from pooled models, was not corroborated.