Modeling stochastic mortality with O-U type processes
Modeling log-mortality rates on O-U type processes and forecasting life
expectancies are explored using U.S. data. In the classic Lee-Carter model of mortality,
the time trend and the age-specific pattern of mortality over age
group are linear, this is not the feature of mortality model. To avoid this disadvantage, O-U type processes will be used to model the log-mortality in this paper. In
fact, this model is an AR(1) process, but with a nonlinear time drift term.
Based on the mortality data of America from Human
Mortality database (HMD), mortality projection consistently indicates
a preference for mortality with O-U type processes over those with
the classical Lee-Carter model. By means of this model, the low bounds of
mortality rates at every age are given. Therefore, lengthening
of maximum life expectancies span is estimated in this paper.
关键词:
mortality,
stochastic forecasting,
O-U type process