Fig 1.
Dividing changes in mortality rates between compression and postponement.
Table 1.
Description of data.
Fig 2.
Extinct cohorts with highest (left) and lowest (right) R-squared in our dataset.
Fig 3.
Box-and-whisker plot of the distribution R-squared across extinct cohorts, by decade of birth.
Fig 4.
Box-and-whisker plot of the distribution of the RMSE across extinct cohorts, by decade of birth.
Fig 5.
Observed mortality hazard rates (dots) and estimated base mortality hazard rates (lines) by age for Swedish males born 1880–1950.
Note: Estimated base mortality rates are calculated using modal estimates and
for each cohort.
Fig 6.
Bayesian estimates of slope and intercept parameters of male and female cohorts for Sweden (panel A); implied changes in remaining cohort base life expectancy at age 50 over ten year birth cohorts, divided into the portion due to postponement and portion due to compression (panel B), and 95% confidence intervals for the age at which cohort mortality hazard first reaches 2/3 (Λc) and the maximum observed lifespan (Mc) in each birth cohort (panel C). Note: Confidence intervals are not shown in Panel A for ease of interpretation. In the other panels, bands or bars show 95% confidence intervals for the true base value. In panel B, bar heights show median estimates. Black dots show the actual change in life expectancy for extinct cohorts. Note that under the assumption that the model is correct, this would equal the true base value plus random error due to calendar-year effects and random variation, plus the change in bias due to these. In panel C, black dots show the longest-lived person in each birth cohort from IDL [22] and GRG [23] data (green if they are still alive in 2021). The mean of Mc increases in line with the logarithm of cohort size at age 50 even if there are no changes in mortality rates. The shaded area to the right of the diagonal line in panel C indicates censoring in 2021.
Table 2.
Statistical tests for δc = δ ∀ c.
Fig 7.
Changes in remaining base life expectancy at age 50 across 10-year birth cohorts separated into the portion due to postponement and the portion due to compression for males and females.
Note: These show the median changes in remaining cohort life expectancy at age 50 due to compression and postponement estimated from our posterior sample for each country. Confidence bands for earlier cohorts are small, but widen for the most recent two decades shown. Detailed results for each country, along with confidence intervals, are available in the supplementary materials.
Fig 8.
95% confidence intervals for the length of maximum life (Mc, grey) and the Gompertzian maximum age (Λc, red).
Note: Dots in the figures show the age at death (black dots) and current age (green triangles) of the longest-lived person in each birth cohort, from the IDL [22] and the GRG [23]. IDL data contains people over the age of 105 or 110 (depending on country); GRG data only show people over the age of 110. The mean length of maximum life increases due to compression and population increase (because more people are expected to reach age Λc) as well as postponement. Confidence intervals for Mc are conditional on at least one person reaching age Λc. Unconditional upper confidence intervals for Mc are always the same as the conditional ones, but the unconditional lower confidence interval may be smaller than shown in countries with small populations, especially for males. See methods section M3 for details.
Fig 9.
Confidence intervals for Mc for the USA using combinations of use of data ages and assumption of the height of the plateau.