UNITED NATIONS STATISTICAL COMMISSION and ECONOMIC COMMISSION FOR EUROPE CONFERENCE OF EUROPEAN STATISTICIANS STATISTICAL OFFICE OF THE EUROPEAN UNION (EUROSTAT) Joint Eurostat/UNECE Work Session on Demographic Projections

In recent years we have seen an increasing demand for indicators of mortality for smaller (sub-national, sub-regional) areas, either to examine geographic inequalities in mortality, to monitor the effects of Public Health policies, to inform local strategies or to prepare long-term sub-national population projections. The usual way to obtain life expectancy indicators involves the construction of complete or abridged life tables. Attempts to calculate mortality rates directly from small numbers of counts and deaths often results in highly erratic schedules that are very difficult to interpret. In this paper we give an in-depth overview of the method adopted by Statistics Portugal for estimating life expectancy in small population (sub-national) areas (NUTS II and NUTS III). The method uses parametric graduation techniques to smooth crude age-specific mortality rates in order to construct a survival model presented in the form of a life table. We give an overview of parametric and non-parametric graduation methods and revisit the graduation methodology developed by the Continuous Mortality Investigation Bureau (CMIB) and its extension to generalized linear models, recently adopted by Statistics Portugal. The method uses a family of parametric (generalised Gompertz-Makeham) functions estimated by means of generalized linear models in order to graduate crude mortality estimates. We discuss the statistical tests and procedures used to evaluate the goodness-of-fit of the models. The methodology is empirically tested using data for the Portuguese sub-national region of Lisbon and for the period 2006-2008. We conclude that the Gompertz-Makeham functions estimated by means of generalized linear models offer a good alternative for estimating life expectancy in small population areas. The method is flexible and applicable to mortality data for a wide range of ages from any geographical conditions.