日本を含めた工業国における年齢別死亡率の時系列的変動: テイラーの法則とゴンペルツ・メイカム・シラーの邂逅
-
National Institute of Population and Social Security Research
国立社会保障・人口問題研究所 特別講演会
下記の要領で国立社会保障・人口問題研究所の特別講演会を開催いたしました。
日時 : 2017年5月15日(月)13:30 ~ 15:00
場所 : 日比谷国際ビル(内幸町)6階 国立社会保障・人口問題研究所 第4会議室(地図)
講演者: ジョエル E. コーエン(米国ロックフェラー大学・コロンビア大学 教授)
題目 :「日本を含めた工業国における年齢別死亡率の時系列的変動 : テイラーの法則とゴンペルツ・メイカム・シラーの邂逅」
ご講演の概要
The variability over time in age-specific mortality in industrialized countries (including Japan) follows a reproducible pattern known to ecologists as Taylor's law: the variance (over time) of mortality at a given age is approximately proportional to a power of the mean (over time) of mortality at that age (Bohk, Rau & Cohen, Demographic Research 2015). Specifically, in 12 countries (including Japan) with reliable estimates of mortality for each age x = 0, 1, ..., 100, in each year from 1960 to 2009, separately for women and men, we plotted the log10 variance (vertical axis) of the force of mortality at each single year of age (over these 50 years) as a function of the log10 mean of the force of mortality at each single year of age (over these 50 years). The plot was very nearly linear, as Taylor's law predicts.
What explains this striking empirical regularity? We proved recently (Cohen, Bohk-Ewald, Rau, under review) that the Gompertz model of adult mortality predicts exactly the form of the observed Taylor's law when the modal age at death in the Gompertz model changes linearly in time and other parameters remain constant in time. However, the Gompertz model does not describe accurately mortality below 30 years of age and its other parameters are not constant over time. The Makeham and Siler models, which are more accurate descriptions of mortality below 30 years of age, predict the form of Taylor's law about as well as the data approximate Taylor's law. This work provides the first theoretical link between classical demographic models of mortality and a classical power law of ecology. The patterns discovered may prove useful in producing and evaluating projections of future mortality.
This is joint work with Christina Bohk-Ewald, Max Planck Institute for Demographic Research, Germany, BohkEwald@demogr.mpg.de, and Roland Rau, University of Rostock, Germany. roland.rau@uni-rostock.de.
* ご講演は英語で行われました。