Inter-Quartile Age Annex

6. Is there a Limit to Compression?

Regarding the history of mortality compression, we consider the existing evidence sufficient for proving that the transition from high to low mortality - and, thereby, to modern society - has been accompanied by general and massive compression of the bulk of mortality to ever narrower age intervals. This view of Wilmoth and Horiuchi is fully confirmed by our findings. We also confirm the very significant slow-down in this process during the last decades. However, while these authors find that variability in the age at death has been near-constant since the 1950s, we see significant compression well after that time and still on-going in many populations. We also find notable differences between countries in the present level and dynamics of compression. On the broadest, 90-percent level, compression continues as before.

While our analysis, as well as that by Wilmoth and Horiuchi, confirms up to a certain point the notion of Fries of a move towards a more uniform age at death, it can not be overlooked that the recent near-stagnation in this process has occurred at a level where compression is still so far from that visualized by Fries that it seems most unlikely that an even approximately uniform length of life would ever become reality.

If we consider that compression and rectangularity have been essentially completed when only 10 percent of lives do not conform to it, this state is reached when C90 is reduced to one year. At present it is in no population less then 35 years.

A factor influencing the concentration of mortality by age, is population heterogeneity [Vaupel, Manton and Stallard 1979] which is partly genetic but to a greater extent environmental [Christensen and Vaupel 1996, Herskind et al. 1996, Valkonen et al. 1993]. If the data were divided into sub-populations according to occupation, education, marital status, availability of medical care, and life-style factors such as diet, smoking, physical exercise and so on, we would be likely to observe greater compression. However, no sub-population could conceivably die out in a truly short age span. No matter how homogeneous a group of people may be, its members are not going to die at the same approximate age.

Everything that is known about the probability of dying denies the existence of a certain point in age when it would suddenly aggravate so that few could survive it. Death and survival of individuals depend on innumerable and unforeseeable little factors and events and these are not likely to coincide in such a way as to cause such a sudden aggravation. Considering also the recuperative powers of the organism, as well as therapy and surgery, there exists a continuity in the likelihood of further survival which results in a certain flatness in the length-of-life distribution.

The compression of mortality, which we have witnessed, includes actually two separate processes. On one hand, the lives saved through the progressive elimination of premature deaths have been redistributed to older, more usual ages at death. This has been the motor of the compression of mortality through most of the transition. On the other hand, as people survive to ever higher ages, they enter an area where the lengthening of life meets increasing resistance, demonstrated by the fact that the decline in old age mortality in recent decades has been very much faster near age 80 than 100 [Kannisto 1996]. The capability of man to lengthen his life in a favourable environment seems to decline with age. This compresses mortality at ages which are now being reached by larger numbers of people than before.

We think it relevant to the study of mortality to measure compression and to monitor the on-going changes in it. For the purpose we recommend the C-family of indicators, particularly C50, though, as it stabilizes, C90 may become more meaningful. Both of them may acquire new significance as indicators of internal mortality differentials.


Inter-Quartile Age Annex

Measuring the Compression of Mortality
Väinö Kannisto
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871