Abstract
Mortality improvement has been a common phenomenon since the 20th century, and the human longevity continues to prolong. Postretirement life receives a lot of attention, and modeling mortality rates of the elderly (ages 65 years and beyond) is essential because life expectancy has reached the highest level in history. Mortality models can be divided into two groups, relational and stochastic models, but there is no consensus on which model is better in modeling mortality rates of the elderly. In this study, instead of choosing either a relational or stochastic model, we propose a synthesis model, selecting and modifying appropriate models from both groups, which not only has a satisfactory estimation result but also can be used for mortality projection. We use the data from the United States, the United Kingdom, Japan, and Taiwan (data were from the Human Mortality Database) to evaluate the proposed approach. We found that the proposed model performs well and is a possible choice for modeling mortality rates of the elderly.
ACKNOWLEDGMENTS
We greatly appreciate the insightful comments from the editor and two anonymous reviewers, which helped us to clarify the context of our work.
Discussions on this article can be submitted until July 1, 2020. The authors reserve the right to reply to any discussion. Please see the Instructions for Authors found online at http://www.tandfonline.com/uaaj for submission instructions.
Notes
1 Source: Ministry of Interior, Department of Statistics, http://www.moi.gov.tw/stat
2 Notice that the Gompertz parameters now have subscripts t. Instead of setting a fixed Gompertz parameter for the whole period, we estimate the parameter on a year-to-year basis.
3 We did not apply the transition age process on the Lee-Log and the CBD-Log models because estimating parameters for the logistic model requires regression on all ages.