Modelos de regressão bivariada: uma aplicação em equações mincerianas de rendimento
Nenhuma Miniatura disponível
Data
2018-02-08
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Goiás
Resumo
In this work, bivariate regression models based on the bivariate normal, t and Birnbaum-Saunders
distributions are used to analyze labor market data. In special, the objective is to model the dependent
variable of the Mincerian earnings equation separately, namely, the variable hourly earnings (which is
obtained by dividing gross monthly earnings by hours worked) is modeled in two parts, earnings and hours
worked. The bivariate regression models are used to model these two parts in order to try to capture the
correlation between them and the different effects, that is, remuneration or premium for labor effort, and the
labor supply or the time that the worker offers to the market. In order to accomplish this, data from the
Brazilian National Household Sample Survey (PNAD) for the years 2013, 2014 and 2015 are used. The
parameters of the models are estimated using the maximum likelihood method. The results show that the
bivariate regression model based on the bivariate t distribution has the best fit for the data, and that the
presence of correlation between earnings and hours worked indicates that the bivariate model is more
adequate than the univariate model.
Descrição
Citação
CUNHA, Danúbia Rodrigues da. Modelos de regressão bivariada: uma aplicação em equações mincerianas de rendimento. 2018. 72 f. Dissertação (Mestrado em Economia) - Universidade Federal de Goiás, Goiânia, 2018.