Experiential Measures in the Historical Demography of Aging
We have yet to consider the actual impact of the secular shift in aging on the societies in which it has so far occurred. Before we go on to a preliminary assessment of these effects, however, we have to acquaint ourselves with the experiential measures that have been alluded to already. These are numerical measures that it is hoped come closer to the personal and social realities of aging change over time and of aging differences between societies. We shall then find it useful in relation to the secular shift to discuss an illustrative example of the application of the principles of the historical demography of aging and of the measures that will be suggested.
We begin with longevity and meet at the outset an obstacle to understanding that recalls the pons asinorum described earlier. It is natural for someone unacquainted with what the demographers call the life table to suppose that expectation of life—or average expectation of years still to come, to give it its full title—is always highest at the very beginning, at age 0. But this is not so, except under the very special demographic circumstances of the advanced countries at the moment and over the last few decades. Even in their case longevity is at its peak not at birth itself but a few weeks later. In all other populations average expectation of years still to come is higher at exact age 1 than at birth, higher again at age 2, and usually higher still at age 5, the expected peak value in all but contemporary developed countries. It can often be higher at age 15 or 20, or even sometimes 25 or 30. Let us take an example.
In the life table calculated for England and Wales as it was during the years 1989 to 1991, expectation of life for women was 78.67 years at age 0, 78.21 at 1, 75.29 at 5, and 64.41 at 15, all the later values being lower than the value at the beginning. For Canada in 1831, in contrast, it has been reckoned that expectation of life for women was 39.84 years at birth, 46.49 at age 1, 49.48 at 5, and 42.61 at 15, still over three years longer than at birth. The outstanding difference between the two populations, and such is always the case when comparing longevity between the Before and the After, was the very much higher infant and child mortality in Canada during the 1830s. The death rate of females in the first year of life in Canada in 1831 was 162.36 per thousand, whereas in England and Wales in 1989 to 1991, the corresponding figure was 6.78. In the second year in Canada, it was 135.76 and in England and Wales, 5.59.
If the rise in expectation of life in the Canadian figures at successive birthdays after the first seems paradoxical, because it appears to indicate that an individual person had longer to live at 2, 5, or 15 or later than at 0, this is due to the fact that the character of life expectation has not been recognized as an average, an average of the years still to be lived by a specific group of persons of a given age. Since such an average is the result of dividing the pooled total of years of life remaining to all members of the group by the number of members of that group, it must be possible for that average to go up rather than down after an interval during which an appreciable proportion have died. This follows from the fact that such a loss could well reduce the numbers of persons surviving by relatively more, even very much more than it reduced pooled years still to live. Such would be the case in spite of the fact that the comparatively very large number of pooled years still to come would have been lessened by the years lived by those surviving through the interval, or through any part of it. A suitable metaphor might be that those still present at the end of the interval say to each other with some satisfaction, "Now that those unfortunates who were due to die during the interval are out of the way, we can all go on to live longer." Here the word "all" must be taken to indicate the new average of years remaining to each of them, that is, their expectation of life.
There are several implications of this explanation for the comparative study of longevity over time and between populations and for the selection of more realistic longevity measures. It demonstrates that expectation of life at birth has serious disadvantages as a measure, only appropriate for comparison of length of life between the populations of advanced societies in very recent times indeed. In the 1950s in the United Kingdom, longevity at age 1 was still nearly a year greater for females than it was at age 0 and over a year greater for males. The efficiency of expectation of life at age 0 as an indicator of the experience of longevity in other populations is obviously
woefully impaired because it is so heavily affected by high death rates in the earliest years.
If we were to contrast the life expectation of Canadian women in 1831 at its maximum, that is, at age 5, with that of English women in 1989-1991 at its maximum, that is, at birth (49.48 years vs. 78.67 years), this would seem to be a more realistic comparison than that produced by contrasting expectation of life at birth on both sides. The outcome would be that there was a difference of 29.19 years in favor of the contemporary English women rather than 38.83, a substantial reduction of well over a quarter in the superiority of the present day. If both the populations had been on the lower aging plateau, of course, comparisons of maximum longevity would have to be for values at age 5 on both sides. Comparisons of this kind would have to be made, however, in the knowledge that babies and young children who would die before the age of 5 had been left out of account.
Nevertheless, the aging experiences of such young and very young persons might not necessarily be supposed to be of much significance to the subject we are pursuing, though reflecting on their position brings home the important fact that aging is different at different parts of the life course. The experience of proceeding from the fifth to the tenth birthday is certainly not the same as proceeding from the tenth to the fifteenth, let alone from the sixtieth to the sixty-fifth, and these obvious differences are even greater when the historical periods and social circumstances of the populations being compared are allowed for. It has to be reckoned that there is a degree of indeterminacy in comparing the experience of longevity over time and between societies and that there seems to be no easy way of overcoming it. This may be one of the reasons why demographers have not included longevity in their analysis of aging.
The compromise solution recommended here is that life expectation at age 15, not life expectation at age 0, should be taken as the value for comparing longevity between societies at all historical epochs and at all stages of development. This simple expedient has the following advantages over using expectation of life at birth. Taking life expectation at age 15 as the measure of longevity makes unnecessary the reckoning of life expectation at different ages on the different sides of the comparison. Moreover, it is a much better guide to changes in longevity in later life.
Taking up the contrast between Canadian women in 1831 and English women in 1989-1991 once again, the percentage difference in life expectation at birth was 97 percent, that is, English women had 97 percent longer to live than Canadian women as reckoned in this way. But at age 15 the difference was 51 percent in favor of modern English women, while at ages 50, 60, and 70, it was 50 percent, 57 percent, and 63 percent respectively, each value quite evidently considerably closer to the differential figure for expectation of life at 15 than the differential figure for expectation of life at
birth. The value at age 5 would have similar advantages over expectation of life at birth as an indicator for this purpose, and it has the attraction that it is the highest longevity value for a very large proportion of all populations on the lower plateau. But expectation of life at 15 marks the point in the age range at which in many of them the average of years still to come ceases to be higher than expectation of life at birth. It could be said, moreover, that expectation of life at age 15 represents a closer approach to the median experience of longevity in the middle range of the population. But once again it has to be borne in mind that the aging experience of all immature persons is being omitted, and in populations on the lower aging plateau this means a quarter or even a third of the total.
If expectation of life at 15 is adopted as that measure of longevity that is least likely to misrepresent the aging experience of populations that are being compared, the question arises, what change might this imply for the secular shift in aging? Values for expectation of life at age 15 (e15 ) are accordingly graphed in figure 1.7 for England, France, and Sweden, the countries shown in figure 1.3. Values for Canada, the United States, and Japan are also represented for as far back as records go.
Comparison of the curves in figure 1.7 for e15 with those in figure 1.3 for expectation of life at birth (e0 ) makes it immediately apparent that the suddenness and steepness of the secular shift is much more pronounced with respect to longevity when expectation of life is reckoned for age 15. It is also evident that the experience of all the Western countries that appear in the figure has been very close, that England could indeed be taken for each of them without misrepresentation. Interesting, too, is the fact that the two North American, European-descended populations should follow the domestic European populations so faithfully but that their statistics are somewhat higher during the nineteenth century. The graph for Japan, however, though it follows a parallel course from a markedly inferior initial level, climbs from the lower to the higher plateau with an amazing steepness after the 1950s, almost vertically in fact, and rises above all the others. The aging trajectory for that country could be said to be something of a caricature of the secular shift as it has been experienced by Western countries, the extraordinary drop in the 1940s being an outcome of the Second World War. We shall return to these differences later on.
The generalizations that have been made about the secular shift, then, are emphatically confirmed for longevity by the use of the more realistic measure of life expectation at age 15. It might perhaps be asked why this indicator was not adopted at the outset. The answer has to be that it has a grave disadvantage in practice. Because the only statistic for longevity usually published for any population is expectation of life at birth, any other measure but this is very seldom likely to be available. For further discussion of this and other circumstances in the measurement of longevity in experiential
terms, see the appendix to this chapter together with table 1.A1 and figure 1.A1. The outcome of our consideration of longevity to this point might have to be that we must continue to accept expectation of life at 0 as the standard measure but that this measure must be used bearing in mind all the drawbacks and obliquities that have been set out.
We are not quite at the end of the complications about the reckoning of longevity for comparative historical purposes. There is an important difference between what is called period or generation life expectation, which has been used so far, and cohort life expectation, which has not been mentioned.
Period life expectation is nearly always the one that is used. But it is a synthetic construct. In reckoning the total years still to live, which have to be averaged out among every member of the population in question, it assumes that the death rates being experienced at all ages during the year at issue re-
main the same over the whole time during which the people concerned will go on living. It is almost as if everyone had his or her whole life experience in the current year. In reality, of course, these mortality rates will undoubtedly change as the years go by, perhaps not by very much but by enough to make this synthetic calculation a hypothetical estimate. The enormous advantage of the period calculation, however, is that it can be done for any chosen year or, more usually, for any two or three years, provided only that mortality by age is completely known for the year or years concerned.
Cohort life expectation makes use of the actual number of deaths experienced by a cohort (persons born in the same year) from the time of its appearance until the last member of that cohort has died, that is, something like one hundred years later, or even more. Cohort life expectation, then, is about as realistic as such a statistic could be. But it has the enormous disadvantage that it can only be calculated after every member of the collection of people in which we are interested has ceased to exist. It is for this reason entirely impractical to recommend cohort life expectation as being more realistic and true to experience than period life expectation. Only for historians studying populations existing at least a century ago would cohort life expectation be of use. Though such a statistic could in theory be used for comparisons between past populations, all of which existed a century or more ago, it cannot be used for comparisons involving any more recently existing population. What is more, the necessary data for such calculations so far into the past only survive for a small number of historic populations. Nevertheless, cohort life expectation can usefully be calculated for some purposes on occasion, such as reckoning of the Third Age Indicator, which will concern us later on. It is available for English cohorts born between 1541 and 1781 and again for those born between 1841 and 1876.
This brings us almost to the end of the considerations necessary in measuring expectation of life for historical purposes, though in the appendix to this chapter we shall look at yet a further possible indicator. This is the age at which there are a given number of years still to live (say, two years more or five years more), an age that differs in a most interesting way between the Before and the After. It goes down rather than up during and after the secular shift (compare especially Bourdelais 1993). As with all measures, that which is the most satisfactory, or perhaps in this case the least unsatisfactory, for general purposes may not be of much use for particular purposes. In describing the emergence of the Third Age as an outcome of the secular shift, a somewhat different statistic from that of e15 or of expected years still to live will be suggested in relation to longevity. This will be the chances of reaching age 70 from age 25, the Third Age Indicator, or 3AI. It would be possible to argue that from the point of view of the interests of most students of aging—and certainly of most of the contributors to this volume—the 3AI, reckoned either in the period or the cohort mode, might
be even more realistic as a general measure than expectation of life at age 15, and closer to the experience of past people. I shall suggest later on that a low level of the 3AI might have disposed those people to write off the possibility of ever becoming old. These points are illustrated and enlarged in the appendix to this chapter.
Interesting as this possibility, might be thought to be, the major concern of those who now study aging is in the numbers, experience, condition, and prospects of those already in late life and their relations with their juniors and their juniors with them. It is this concern that informs the suggestion of a revised measure of proportions in late life, our second aging variable, a measure more realistic than a simple fraction of older persons in the whole population. The realistic, experiential measure suggested here is the share of those over 60 of all adults, of all those over 25, in symbols 60+/25+. A more detailed impression of the significance of older persons with respect to their numerical size in relation to that of other age bands, together with changes over time in such relationships, can be gained of course from figure 1.6.
This realistic age-proportional measure, relative weight of the elderly and old among all adults, passes over an even greater number of the nonadult, the immature, than the realistic longevity measure just discussed, life expectation at age 15. The justification for doing so is just as strong in my view, or perhaps even stronger. When attempting to sense the presence of those in late life in a society, it is not easy to see why every individual member of that society should weigh as much as every other. The young and very young are undeniably of significance in the social structure. We in our own day are not the first body of people to be conscious of the immature and maturing members of our society and of the crucial character of our relationships with them and theirs with us. The young and the very young are also of evident importance because they are largely dependent and their numbers and proportions must be known to study support relationships and transfers between age groups. In this respect, the young are in a position very similar to that of the dependent old. These are highly significant examples of generational interchanges and of age structuring. Where changes in the proportions of age over time are accompanied by changes in the flow of support, with the result that some cohorts are privileged over other cohorts, a fascinating set of issues to do with intergenerational justice comes into view.
These issues can certainly be classed as consequentially related to the historical demography of aging, but relationships as to aging experience between the various age groups are scarcely affected. Figure 1.8 traces the course of the ratio recommended, proportion of all adults who are over 60, for the four hundred fifty years of known English aging experience and the two hundred fifty years of French and Swedish experience, along with the briefer periods observable in Canada, the United States, and Japan. In my
view, it presents the clearest and most revealing comparative numerical account of the weight and importance of the elderly over the stretches of time which it is possible to observe and complements for aging experience of this kind the results we have just surveyed for the experience of longevity. The figure includes, it will be noticed, a line for the threshold of the Third Age, which, along with the 3AI itself, will be discussed below.
The features that stand out in figure 1.8 are once more the similarity with regard to the abruptness and the shape of the secular shift to what is to be seen for all the populations included in figure 1.3 and the somewhat greater unevenness of the lines, none of which shows quite the same smooth progression that marks the long English graph for proportions over 60 in the whole population portrayed in figure 1.2. The pronounced peak in the English profile in figure 1.8 during the early eighteenth century is somewhat disconcerting, since it reaches a height not seen again until the secular shift was well under way in the 1930s. At 22.4 percent, the proportion that those over 60 made of all English adults in the five-year period 1706 to 1710 was over two-thirds of its level now (1991 = 31.2 percent). It might be thought that the share of the elderly in the population was not as constant as has been made out, or that the experience of the English was less than representative.
Closer examination shows, however, that more modest peaks of this kind are present in the graphs for the other populations represented and that the general resemblance to the lines in figure 1.2 is quite pronounced. Once again the relative shortness of the lines in the figures, the woeful lack of temporal depth in our data, makes judgment difficult. Inspection of figure 1.6 and of the quinquennial figures in table 1.3 reveals that, during the
early decades of the eighteenth century, proportions over 60 were at their highest in the whole English series before the 1930s and that the age band 25-69 fell into a trough at the same time. Though neither movement was particularly sharp, their coincidence seems to have produced the effect we are examining.
The adoption of a more realistic, experiential measure for proportions of elderly scarcely confirms the arguments that have been presented using the whole population as a divisor in quite the decisive way in which the substitution of expectation of life at 15 does for longevity. The rise during the secular shift in the proportion of adults who were over 60 was only 57 percent in England. But the dissonance cannot be said to affect in any great degree the theory of relative constancy on the lower plateau, the abruptness of the secular shift, or of the general typicality of the English data. The reexamination of the aging processes and particularly of the secular shift by the use of experiential measures is highly illuminating, nevertheless. We shall find this point confirmed as we take up an illustrative example in the aging history of two neighboring European countries whose comparative development has preoccupied us so much.