Figure 3 in Chapter
3 demonstrated how mortality curves of successive periods moved generally to the right
while essentially maintaining their characteristic form. In other words, mortality
shifted to higher age. The magnitude of this shift is measurable in years of age. The
shift may be considered indicative of a similar shift in the physical health or robustness
of a population. It may mean a commensurate delay
in the aging process.
At young ages, the age shift
has few if any practical applications but starts to have meaning in middle age when
mortality begins to rise monotonically. It is perhaps most meaningful in old age.
The age shift could be
measured in relation to given mortality levels expressed in mx or qx
or mx or other parameters but
is more easily visualized when expressed in relation to age itself. In the following, we
therefore start with mortality by age in a given period and observe at which ages the same
mortality levels are reached in a later period. We shall call the difference the age
shift in mortality. When mortality declines, the shift is positive; when it rises, it
is negative.
To illustrate the procedure,
we find in the life table for Austrian females in 1950-60 the value q(85) = 0. 172 1.
Assuming that mortality rises from age x to x +1 linearly, we can pinpoint the exact age
at which q equals 0.1721 in later life tables and obtain the following:
In the same way, two
contemporary populations can be compared by observing at which age the death rates found
in a population of higher mortality are reached by a population of lower mortality. The
difference may be called age delay in mortality. Likewise, we may measure the age
delay of female mortality in relation to male.
It is advantageous to
calculate the age shift or age delay from smoothed - or originally smooth - age series in
order to avoid meaningless irregularities. The smoothing can be considered fully justified
at ages where mortality is generally linked to aging.
As an introduction, the age
shift is illustrated in Figure 14 for the aggregate of thirteen
countries. The curves for males rise with advancing age. In the first period, it
progressed between 0.7 years at age 80 and about twice as much at 95-97. In the combined
20-year period, the age shift was more than twice as large but the rise by age relatively
and even absolutely more modest, namely from 2.2 to 2.7 years.
The age shift for women had
the opposite, descending profile with age. Among octogenarians it was much greater for
women than men but declined thereafter. Around age 95 the shift was approximately equal
for both sexes, namely 2.7 years in twenty years' time. The generally ascending curve for
men means that for them the ascent of mortality by age became less steep while the
opposite form for women indicates a steepening of the mortality curve by age.
Figure
15 demonstrates a great variety in the shifting of age between countries. For men,
both the level and the form contrast widely. In some countries, the shift is very steady
over the entire age range, declining only slightly while in some others it is very low or
negative near age 80 and then climbs gradually. To these belong on one hand the
Netherlands, Norway and Denmark (not shown), on the other, the former East bloc,
exemplified here by Hungary. Though the level of mortality was very different in these two
groups, common to both was that they recorded little progress among the relatively younger
men.
The curves for women vary
widely between the limits of one and five years but display a common tendency to a slow
descent by age. The rise recorded at oldest ages is not very significant in Finland
because of small numbers, nor in Japan because of uncertain age information on the very
oldest in the 1960s.
The actual data for 22
countries and 4 aggregates are given in Annex Table 6 and
summarized by 5-year age groups in Table 12. Only the age range
80-94 (as of 1960-70) can be effectively observed because for the older the comparable
1980-90 rates are found at ages where data are not available or in some cases more
erratic. The largest shift, surpassing 4 years, is registered for Finnish women and for
Japanese men and women, shifts of over 3 years for Icelandic and Swiss women and Finnish
men. In Eastern Europe, the shift was generally less than one year.
That the age shift in
mortality is an indicator independent from change in life expectancy is shown by the large
difference in their magnitude and a less than strict correlation be-tween the two. The two
indicators are juxtaposed in Figure 16 for the 22 countries in Table 17 plus Ireland which was omitted from it because heavy age
heaping made the details questionable.
The age shift proves to be
throughout much larger than the gain in life expectancy. This can be well understood from
the fact that if in a life table we move a certain number of years up or down, the
corresponding life expectancy changes much less.
The second observation is the
looseness of the correlation. When the countries are ranked by increase in life
expectancy, the profile of the age shift is jagged. The former is a more robust measure
while the latter is more sensitive to the age pattern of mortality and also to small
numbers. Above all, however, they measure mortality frorn different points of view.
Updated by V. Castanova, 1 November 1999