A KERNEL REGRESSION WITH MIXED DATA TYPE INVESTIGATION OF THE KUZNETS HYPOTHESIS

Abstract

This paper reexamines the Kuznets hypothesis by taking into account a pool of data provided by Iradian (2005). The empirical strategy is based on: i) the use of the test for parametric specification developed by Hisao et al. (2007); ii) the use of the nonparametric estimation method with mixed data proposed by Racine and Li (2004) and iii) the use of the likelihood ratio test devised by Fan et al. (2001). Results indicate inconsistency of the linear parametric model for the dataset under analysis. Nonparametric inferences produced arguable results. While the bivariate model corroborates the Kuznets hypothesis, the multivariate estimation does not support this hypothesis. Finally, the likelihood ratio tests showed statistical superiority of nonparametric models over linear ones.