Previsão do gap de mortalidade entre sexos via modelos dinâmicos multivariados

Abstract

The mortality rates of males and females have been changing in the course of the years due to changes in quality of life and medicine, and in certain age groups these populations can be similar in some aspects. Forecasts should consider the common features from both these populations to better capture uncertainty. The model proposed in Li e Lee (2005) assumes a demographic hypothesis of convergence between these populations, modeled through a parameter that takes into consideration these common features. An application to American mortality data illustrates the importance of this parameter for greater forecast accuracy. An application to lung cancer mortality data of the population in Brazil is presented using univariate and multivariate dynamic models, in addition to the Lee-Carter and Li-Lee models in order to verify the convergence between both sexes, as the population’s behaviour regarding tobacco addiction has been suffering changes in the last decades. A Bayesian approach is considered for inference and forecast.