Contemporary Model Life Tables for Developed Countries: An Application of Bayesian Model-Based Clustering

Samuel J. Clark, University of Washington
David J. Sharrow, University of Washington

Using the statistical technique Bayesian Model-based Clustering (Fraley and Raftery, 2006) we identify five typical age patterns of mortality in 844 life tables from the Human Mortality Database (University of California, Berkeley and Max Planc Institute, 2009). Time rather than geography is the most important dimension along which the Human Mortality Database life tables are clustered. Using each pattern as the basis for a 'family' in a traditional system of model life tables, we create a one-parameter model to generate 'levels' of mortality within each family. The result is a new effectively two-parameter system of model life tables for the countries and time periods included in the Human Mortality Database. We demonstrate the use of this model life table system to extrapolate full age patterns of mortality from age-restricted mortality indicators such as 5q0 or 45q15.