Estimating County-Level Life Expectancy and Infant Mortality Rate In China

Yong Cai, University of Washington

Using data collected in China’s 2000 census, we construct sex-specific life tables at county-level. One challenge to this effort is to provide reliable and robust estimates of infant mortality rate, which often suffer severe underreporting. In this paper, we construct life tables for each of China’s 2367 county-level units and compare three different methods of estimating infant mortality rate: conventional infant mortality estimates based on reported births and deaths, Brass logit model based on mortality rates of other age groups and a standard mortality schedule, and empirical Bayes estimates based on infant mortality rates reported in the neighboring counties. The results show a good correspondence among these three estimates, particularly between the empirical Bayes estimates and the Brass model estimates. The study not only confirms the overall quality of Chinese mortality data, but also demonstrates the power of empirical Bayes method as a powerful tool for small area population estimation.