The SAINT mortality model: Theory and application
Dato: Februar 2008
Skrevet af: Søren Fiig Jarner (ATP) & Esben Masotti Kryger (Laboratory of Actuarial Mathematics, Københavns Universitet)
Abstract (på engelsk):
Projecting the mortality rates for a small population is a challenging enterprise. The development of mortality rates for small populations often show great variability in improvement rates over time and across age groups. Consequently, simple extrapolatory techniques yield unreliable results which are very sensitive to the data period and age bands chosen. In this paper we develop a mortality model which combines small population data with data from a larger reference population. Conceptually, we imagine that the long-term mortality development of the small population will follow that of the reference population but there may be substantial short- to middle-term deviations. Mathematically, we use the reference population to estimate an underlying parametric trend and a 3-dimensional time series to model the deviations from this trend over time and age groups. Mortality projections, including uncertainty assessment, are performed by a combination of trend extrapolation and standard time series methods.