Between Compression and Shifting Mortality the Longevity Revolution

The Human Mortality Database shows fluctuations around age 105 from 1750 to 1950: 105 years is the expected maximum life span according to a normal distribution. From 1950 to today, we observe a huge increase in the maximum life span predicted by this normal distribution, from 105 to a value close to 113 years.
Some countries, like Switzerland have very good detailed data. From 1876 onwards, we have a record of the maximum reported age at death. Thus, we can compare the expected normal values coming from the Human Mortality Database, to the actual distribution of maximum age reported at death in Switzerland. The same has been done for Sweden, using the data published by John Wilmoth and his colleagues in Science a few years ago (Wilmoth et al., 2000). The two series match perfectly (Figure 5). The series made up of the exceptional people dying year after year at the oldest age in a country occupies exactly the expected spot according to a normal distribution of individual life duration. So these exceptional individuals reaching the highest age year after year are in fact reaching a totally normal age.

Figure 5: Maximum reported age at death (MRAD) and (M + k*SDM+) by gender, Switzerland and Sweden
click to enlarge

To conclude, I will present the empirical data we have gathered concerning human mortality from the first empirical data, gathered by Halley (1693), when he computed the first life table in 1693 to the most recent data in Japan. All the distributions have been built for 100 000 persons at birth. This means that 1000 deaths represent 1% of all deaths. The first distribution of adult life durations is totally flat from about the age of 40 to the age of 74, with about 1000 deaths per year, which is 1% of people dying at every age. Of course in this case it is absolutely impossible to think about an indicator like the modal age at death (Figure 6).

Figure 6: Distribution of adults life durations: Empirical data 1693-2004
click to enlarge

The second series (Sweden 1754-1756) already shows a very little mode around the age of 70. This series was gathered by Wargentin who sent the data by single age to the French demographer Deparcieux (1760). Wargentin published the very long series starting in Sweden in 1751, but the data were assembled by age group. So it was no longer possible to nicely distribute the death by single age as on Figure 5.
The third series is the first Swiss complete life table (1876-1880). This distribution is exactly the same as the Swedish one, some 125 years later: the modes are exactly at the same age. There are more people dying at the mode, 2000 in Switzerland in place of 1000 in Sweden, because the infant mortality fell, and there are many more people reaching adult age in the nineteenth century in Switzerland than in Sweden during the mid eighteenth century.
The next distributions are the modern series for Japan in 1950-1951, some 125 years later. We are after the demographic transition and after World War II, when all the developed countries from Europe, to North America and Japan, are very similar with a mode close to age 80.
Thirty years later, in 1980 James Fries proposed his famous theory on the rectangularisation of the survival curve and the compression of mortality (Fries, 1980). He was proposing an ultimate distribution for adult life duration, centred on the modal length of life of 85 years, with about 10% of people dying at the mode, and a very narrow distribution with a standard deviation of about 4 years above and under the mode. No more people reached ages above 100. The calculation (85 years + 3.5* 4 years) is close to 100 years, so the distribution is perfectly normal. Of course, when Fries published this famous paper in the New England Journal of Medicine in 1980, he could not know the actual values of Japan, but now we do.
In 1980, Japan was already at least at the ultimate value proposed by James Fries with a mode of 85 years for female. But, in the 1980-1984 distribution, Japan was witnessing less than half the number of deaths proposed by James Fries for the mode, less than 5%. Interestingly, many more people were reaching ages above 100. Above age 95 there are almost no people left in Fries’ distribution, but there are still a lot of people alive and a lot of people dying in the Japanese distribution.
Now, with the most recent distribution in Japan, for the years 2000-2004, we are suddenly observing something totally new and totally unexpected. Until now, all the series were following the general scheme of compression of mortality: the higher the modal length of life, the more concentrated the distributions of individual life durations. With the last distribution in Japan, we are observing the sliding to the right of the whole distribution. Japan is now displaying a modal age at death about 6 years higher than 20 years before – more than 90 years – but the full distribution is presenting exactly the same scheme and there is exactly the same number of people dying at the modal age at death. We are no longer in the phase of compression of mortality. The phase of compression of mortality which has been true for almost all the period since World War II seems to be no longer true in Japan for the last 10 or 20 years. Japan is moving to a new scenario that we can call following Bongaarts the ‘shifting mortality scenario’ where the modal length of life goes on increasing extremely regularly but the standard deviation, which has been decreasing from 1950 to 1980, is now remaining at the same level.


Bongaarts, J. (2005): “Long-range Trends in Adult Mortality: Models and Projection Methods”, Demography, 42(1):23-49.

Cheung, S.L.K., Robine, J.M. (2007): “Increase in Common Longevity and the Compression of Mortality: The Case of Japan”, Population Studies, in press.

Deparcieux (1760): “Addition à l’essai”, in: Essai sur les probabilités de la durée de la vie humaine (1746) et Addition à l’essai (1760). Paris, Institut National d’Etudes Démographiques, 2003.

Fries, J.F. (1980): “Aging, Natural Death, and the Compression of Morbidity”, The New England Journal of Medicine, 303(3): 130-135.

Halley, E. (1693): “An Estimate of the Degrees of Mortality of Mankind, Drawn from Curious Tables of the Births and Funerals at the City of Breslaw”, Philosophical Transactions of the Royal Society of London, 1693 Vol. XVII, No. 196, pp. 576-610.

Robine, J.M., Cheung, S.L.K., Thatcher, R. and Horiuchi, S. (2006): “What Can be Learnt by Studying the Adult Modal Age at Death?” PAA Paper, Population Association of America Annual Meeting, March 30-April 1.

Wilmoth, J.R., Deegan, L.J., Lundström, H. and Horiuchi, S. (2000): Increase of Maximum Life-span in Sweden, 1861-1999”, Science, 289:2367-8.

Pages: 1 2

Tags: ,