8. Mortality of the oldest-old since 1950

This chapter attempts to describe the present levels of oldest-old mortality in lowmortality countries and the changes that have taken place in them during the post-war period. Relying only on data of good or acceptable quality, the description is necessarily centred on Europe. From other continents prime evidence is available on Japan and more tentative results for Australia, New Zealand and Singapore.

        In this study mortality is measured by central death rates (mx) which have been calculated for 5-year age groups and individual calendar years. Age-adjusted death rates for ages 80-99 have been used as an overall indicator of oldest-old mortality1).

        From this indicator, ages of 100 years and over were left out because the necessary data were lacking for eight countries and unreliable for a few others; the effect of this age limitation is negligible because there are more than a thousand 80-99-year-old per each centenarian. For greater stability, moving averages and 5-year period rates are also used. Data for both sexes combined were calculated using weights: 1 for males, 2 for females. The decline of mortality is expressed as percentage decline of the ageadjusted rate in the 30-year period from 1955-59 to 1985-89 and it is also shown per annum. The "present" situation in the text refers to the last-mentioned period.

        The post-war development in old age mortality is shown in Annex Table 2 in the form of annual age-adjusted death rates at ages 80-99 for  all 28 countries for which we have at least conditionally acceptable data. Annex Table 3 gives the same for 5-year periods and Annex Table 4 for selected aggregates of countries.

The time series for 12 countries, selected for their accuracy and characteristic features, and twice smoothed with 3-year moving averages, are shown in Figures 2a and 2b in colour.

A glance at these graphs leaves no doubt about the generalized decline of mortality in the post-war period but also makes clear the existence of widely varying speeds, some sluggish, some precipitous, as well as the occurrence of periods of short-term increase here and there.

        In the tumultuous development the curves of different countries often cross each other, sometimes only to recross soon after, sometimes signifying a more permanent change in ranking. It may further be appreciated that in a general way the curves have been falling more steeply in the second half of the period, roughly after 1970. This testifies to a tendency of the decline to accelerate, and this acceleration is percentagewise even sharper than shown in the arithmetic scale of the figure. A comparison of the two figures shows further that the decline of mortality has been more pronounced among women than men. In fact, a more thorough examination of the figures and the underlying data in the Annex will show that there has been no exception to the faster mortality decline for women and that even in the countries of the former East bloc where male mortality long remained stationary, the female rates showed significant improvement. Among the 26 countries listed in Table 6 and showing death rates for two separate dates, the unweighted mean of mortality decline was 14.4 percent for males and 25.7 percent for females.

        It may be asked whether it is possible to reduce the large variety of national time series into a smaller number of scenarios, each of which has prevailed in several countries. We shall try to do this by considering both the level and the speed of decline in each country.

        Although in the global context all countries of the database are low-mortality countries, there are between them differences large enough to discern the existence of  high-, medium- and low-mortality groups among them. Some countries, however, do not conform to any of these groups while some others have during the observation period moved from one group to another.

        Finally, a few countries defy precise classification because of uncertainty about the data. The six-way classification below is therefore not very clear-cut but may be of some help in the examination of this particular mortality history. In reading the comments the reader may refer to Figure 2(a,b), to country-wise data in Table 6 and aggregate rates in Table 7. Further details are given in Annex Tables 2, 3 and 4.

(i) Iceland

Iceland has had and continues to have the lowest reliably recorded mortality at old age in the world. The advance over other countries is particularly notable among men. The population being small, annual rates are volatile but substantial mortality decline is nevertheless obvious and has amounted to 25 percent in as many years.

        The good quality of the data is highlighted by a list of all centenarians since 1950, prepared by the national statistical office Hagstofa Islands, giving for each one the name, sex, place of residence, date of birth, date of death (if deceased) and the age reached (years, months, days).

(ii) Japan

Japan has the distinction of  having experienced in the post-war era the sharpest decline in old age mortality among all the countries of the database: 39.8 percent in 30 years. Old age mortality was quite high in Japan until about 1960 when it began to fall extremely fast. This has continued ever since so that at the latest date on record Japan is in the forefront of the low-mortality countries though still some distance behind Iceland.

(iii) Low-mortality group

This group includes four countries with traditionally low mortality. Norway and the Netherlands recorded already before 1950 particularly low death rates which were almost as low for men as for women. In the period that followed, the male rates have made only slight progress (3.7 and 8.0 percent respectively) and have been overtaken by several other countries. The development of the female rates has been more favourable with declines of 25.2 and 30.2 but even these have lost some ground in relation to a few other countries.

Sweden and Denmark, the latter not shown in the figures, belong naturally and even geographically to the same group. Beginning with slightly higher, though internationally quite low rates, they have recorded more sutained decline amounting to about 14 percent for men and 33 percent for women. The result has been that the mortality in the four countries has converged to a solid low level.

(iv) Rapid decline group

Switzerland and France are characterized by a very regular, rapid and sustained decline of mortality over the entire post-war period which has amounted to 34.4 and 31.8 percent respectively, results exceeded only by Japan. The improvement has been remarkable also for men: more than 25 percent. Although starting at moderately high level in the 1950s, their death rates are now among the lowest on record. For women they are already consistently below the 100 mark, only slightly above those of Iceland and Japan.

(v) Medium-mortality group

This group, larger and less uniform than the others, is headed by England and Wales where the mortality of women has been approximately on the Scandinavian level while that of men, considerably higher, has kept the combined rate above that of the low-mortality group. The decline, though regular and substantial (26.0 percent), has been slower than e.g. in France causing the English rates to fall gradually behind.

        The position of  England and Wales is closely matched by Italy which, however, has a lower death rate for men but a higher one for women. Italy is not shown in the figures because its data for the first post-war decades were apparently affected by age overstatement and the actual decline may have been faster than the calculated 22.3 percent.

        Next in this group are Belgium (not shown in figures), West Germany and Finland, their respective rates of decline having been 25.3, 26.4 and 30.9 percent. Most of this decline took place in Finland after 1970 and in West Germany after 1975, being therefore remarkably rapid for a short period and showing now signs of slowing down.

        Slightly higher death rates are recorded for Scotland, Austria and Luxemburg which are not shown graphically. Scotland, like England, is characterized by an unusually high excess mortality of men. The three countries have posted good average declines of close to 25 percent or roughly one percent per year. Among the men the progress gained momentum only in the late 1970s but has been rapid since then.

        In the same group, though with again somewhat higher rates, we have included Portugal and Ireland. In these countries the apparent progress has been slower (19 and 17 percent respectively) but this may be an artifact caused by improving data quality.

(vi) High mortality group

This group is composed by six countries of the former East bloc. The data for Czechoslovakia, East Germany and Hungary are of proven good quality and can be used for describing the entire group. The death rates calculated for these countries -closely similar for all three - are the highest in the entire study. The apparent rates for Estonia, Latvia and Poland are lower but unreliable due to considerable age overstatement which in the case of Estonia and probably also of  Latvia has been caused by Soviet immigrants.

        The group is represented in Figures 2a and 2b by only Czechoslovakia and Hungary because the curves for East Germany would be almost indistinguishable from them. The male curves show stagnation and even increase until a very recent turn for the better. For women, more sustained but sluggish improvement can be noted. The overall declines are the lowest in the database except for the questionable data for Latvia and Poland which would indicate a slight increase.

(vii) Other countries

The scenarios for a few other countries are more difficult to assess and classify. Australia seems to belong at present to the low-mortality group even though the exact level of mortality there may not be accurately represented by the calculated death rates. The past developments in Australia are also somewhat in doubt, not least because we have the requisite data only from 1965. What can be considered certain is that a declining trend has been operating there too. New Zealand non-Maoris display medium-level death rates and belong no doubt to the medium-mortality group of populations even if the calculated rates would not be very precise. More uncertain is the past course of mortality which according to the data has been one of only slow decline. The data for the Singapore Chinese indicate low mortality but the series available is still too short for a definitive assessment. The use of the Chinese calendar is hoped to assure reliable information on age.

The data for Spain suggest quite low mortality already some time ago but are obviously biased by considerable age overstatement. This is to an even greater extent the case of Canada, Chile and the New Zealand Maoris.

        During the last 30 years the ranking of the countries regarding mortality level has undergone an extensive transformation as is evident from Table 8 and Figure 3.

        In this comparison we have included only the 19 countries with good quality data plus Ireland and Portugal in which the death rates calculated for at least the 1985-89 period can be considered fairly precise.

        The change has been so deep-going that most of the countries have moved down to levels apparently never reached before. The lowest rate of the 1950s would now be one of the highest in Western Europe. At the same time the ranking of the countries has changed thoroughly. Only the leading position of Iceland has not been seriously challenged but Norway and the Netherlands have ceded their places to rapidly advancing Japan, Switzerland and France.

        Finland, Scotland and Luxemburg have moved from their former lagging positions close to the middle while Portugal, Ireland and Austria have been pressed relatively more back in spite of large reductions in mortality.

        The shake-out has left East Germany, Czechoslovakia and Hungary at the end although even there, reductions in death rates, mainly of women, have taken place. The rates calculated for Poland (160.5), Estonia (153.3) and Latvia (151.0) would place them in the next lowest ranks but they are believed to be affected by age errors and are not included in the table.

        The development has not led to greater similarity but, on the contrary, to wider differentiation. While the unweighted mean for the 21 countries fell from 172.8 to 132.7, the standard deviation rose from 14.3 to 19.2 and the coefficient of variation from 8.3 to 14.5. This was mainly caused by the East -West divergence but even if the former East bloc countries are excluded, the coefficient of variation has grown from 8.4 to 9.3.

        The development of mortality by sex is given for each country in Table 9 and in Figure 4. In all countries without exception the development has been more favourable to females than males, thus increasing the already existing gender gap of mortality. In Portugal and Spain this additional element has been very small while in the other extreme the Netherlands and Norway have seen widely different development for the two sexes and in Eastern Europe the progress has been essentially limited to women only.

        The data for aggregates summarize this development: (i) hardly any improvement for men in the high-mortality group; (ii) sharply differential progress in favour of women in the traditional low-mortality countries; (iii) rapid decline has been possible only when also men have fully taken part in it: the case of Switzerland and France is strengthened by evidence from Japan and Finland.

        The decline of mortality has up to now been mainly discussed in relative terms which is perhaps the more common viewpoint of observers in general. The demographic impact, however, can be better judged by  the changes in relation to population. In these terms, per 1000 population of both sexes, aged 80-99, the following declines were recorded between 1955-59 and 1985-89 (as given in Table 9):

Japan 72.2 Italy 35.3
Finland 59.3 Australia 34.3
Switzerland 58.6 Portugal 34.0
France 52.8 Netherlands 33.7
Germany, West 47.0 Ireland 29.0
Luxemburg 47.0 Norway 25.6
Scotland 46.2 New Zealand 21.5
Sweden 43.9 Hungary 21.1
Belgium 43.6 Czechoslovakia 19.7
England & Wales 42.6 Estonia 15.0
Denmark 42.1 Germany, East 13.2
Austria 42.0 Latvia -0.3
Spain 41.8 Poland -4.1
Iceland 35.8

        The decline that has taken place in Japan is also in this comparison outstanding. Finland, Switzerland and France have undergone the greatest reductions in Europe while also the gains made in West Germany, Luxemburg and Scotland are impressive. Many other changes in ranking when compared with percentage decline have also taken place. The six countries of the former East bloc remain nevertheless at the end of the list.

1) The weights, calculated according to the Swedish population in the database in 1950-89, were the following: 80-84: 0.6360; 85-89: 0.2743; 90-94: 0.0774; 95-99: 0.0123. For the detailed standard population, see Annex Table 12.

Rates for both sexes combined have been calculated using weights 1/3 for males and 2/3 for females based on the fact that in the combined population aged 80 and over in the 19 countries with good quality data the percentage of males was 38.8 in 1950 and 31.2 in 1990. It is practical to have constant weights for all periods, countries and ages.


Updated by V. Castanova, 1 March 1999