Name: Larissa Carvalho
Institution: Ence, IBGE, Brazil
E-mail: larissaalves.ufrj@gmail.com
Co-authors: Victhor Sartório, Renata Bueno
Abstract:
The way in which mortality rates have fallen has changed in the last decade and certain hypothesis made by popular statistical models used to study this behavior may need revision. The model from Lee and Carter (1992) and its Bayesian approach by Pedroza (2006) assume that the relative speed in which mortality decreases for different ages is kept constant through time, which does not seem like a reasonable assumption when considering 21st century data from the American male population. This work presents an extension of the dynamical Bayesian approach presented in Pedroza (2006) in such a way that temporal dynamics are incorporated in these relative speeds, which can be estimated using dynamic linear models with Taylor approximations. This work also contains an application of the proposed methodology to recent American male population mortality data, comparing results with existing models.