Abstract

Projecting future mortality at the provincial level is important for China. However, the sparsity and inconsistency of Chinese provincial mortality data pose challenges for the stochastic modeling of Chinese provincial mortality and have not been given adequate attention. In this paper, we build a Common Age Effect (CAE) model that borrows information from the national mortality data to improve mortality modeling at the provincial level. We estimate the model under the Bayesian approach and employ a two-step process selection procedure based on the expected deviance information criterion (EDIC). Applying our model to the national population and 27 provincial populations of China, we find that provincial mortality is coherent with the national population but exhibit different convergence trends and uncertainty levels. These differences have implications for forecasting future provincial life expectancy and other mortality-related valuations.