Population Change among the Elderly: International Patterns

Modern demography is, above all, about demographic transition. This is to be understood in two distinct, but interrelated senses. First, there is description in the spirit of the narrow, literal definition of demography—a description of quantitative aspects of human populations.… On a more ambitious level, there is the task of explanation.

—Demeny (1968: 502)

In his brilliant editorship of Population and Development Review, Paul Demeny encouraged both types of research on demographic change. This essay is in the spirit of the first type of study. Rather than dealing with the entire population, however, it restricts its attention to ages above 65, where population changes reflect, amplify, and reshape the broad patterns associated with the demographic transition. In this study, we develop and apply a demographic accounting model adapted to the segment of the population above age 65. Through an assessment of the terms in this model, we estimate the contribution of contemporaneous factors and their historical antecedents to rates of growth of the older population.

Background

In a population with fixed age-specific rates of fertility, mortality, and migration, all age groups grow at the same rate (Sharpe and Lotka 1911). Such a demographic regime characterized most of human history over broad sweeps of time and space. Starting in the eighteenth and nineteenth centuries, relatively steady declines in mortality began in Western countries. Such declines typically increased growth rates at ages under five years and above 60, the ages of greatest vulnerability to death (Coale 1972; Coale and Demeny 1983). These declines were followed after some decades by declines in fertility. Initially, fertility declines reduce only the growth rate at age zero, but eventually such reductions become visible throughout the age structure by reducing the number of people at any age relative to the number at any higher age. In the twentieth century, declines in mortality and subsequent declines in fertility have also characterized most of the developing world (Bongaarts 2009; Lee 2003).

Population change at ages 65+ draws from the same demographic sources as population change at other ages, but adds its own filters and lags. Changes in fertility and in infant mortality will affect the population aged 65+ only after a delay of 65 years. Changes in mortality can be registered instantaneously in the death rate and thus in the growth rate at ages 65+, but they may also affect the rate of entrance into the 65+ population by virtue of past improvements in survivorship at ages below 65. Likewise, migration rates at older ages will immediately influence the growth rate at ages 65+, but they can also influence the rate of arrival at age 65 through lifetime migration patterns at younger ages. These factors are formalized in the next section.

Methods

We begin with a demographic accounting framework provided by the balancing equation of population growth as adapted to an age segment of the population rather than applied to the entire population. Through decomposition of the terms in this accounting identity, we estimate the contribution of contemporaneous factors and their historical antecedents to the growth rate of the older population.

The balancing equation of population growth expresses population growth over a time interval in terms of flows into the population via births and in-migration and flows out of the population via deaths and out-migration (Preston, Heuveline, and Guillot 2001). These terms are converted to rates of change by dividing the number of entrances and exits by person-years lived in the interval. When applying the accounting identity to the population aged x and above, the analog to the number of births is the number of xth birthdays that occur in the population during the interval (Brass 1975). The counting of deaths, in-migrants, out-migrants, and person-years of exposure is limited to ages x and above. The population growth rate above age x can be expressed as

$$r_{x+}=b_{x+}-d_{x+}+m_{x+}$$ (1)

In this equation r_x+ is the growth rate of the population, d_x+ is the death rate, and m_x+ is the net rate of migration. All rates refer to the population aged x and above over a specified time interval. For expositional purposes, we assume that the time interval is from t to t+1. The remaining term, b_x+, is the rate of new arrivals at age x by virtue of reaching one’s xth birthday. It is sometimes termed the “birthday rate” by direct analogy to the birth rate, but we use the broader term “arrival rate.” Each term on the right-hand side of the equation has a number of events during the specified period in the numerator and the number of person-years lived above age x during the same period in the denominator. Equation 1 connects the growth rate of the older population to contemporaneous demographic processes of births, deaths, and net migration, permitting evaluation of their respective contributions to the growth rate of the older population.

The number of xth birthdays during the period is in turn a product of the number of births into the cohort x years earlier—i.e., births in the interval t−x to t−x+1 for a one-year period—and any change in the size of the cohort resulting from mortality or migration. Thus the b_x+ term in Equation 1 can be expressed as the birth rate x years earlier, times the probability of survival between ages 0 and x for that cohort, times a factor by which the cohort’s size is modified through migration between ages 0 and x. This decomposition connects the growth rate of the older population to past demographic events, thereby providing insight into future aging processes before they are revealed by the passage of time.

Further decomposition of Equation 1 is possible through elaboration of the d_x+ term into components representing the intensity of mortality and the effects of age composition on the death rate. (See Appendix.)

The intensity of mortality is the age-standardized death rate above age x, i.e., the death rate that would prevail if the age composition were that of the standard population. The age-compositional effects on the crude death rate at ages x+ are represented by the component of the death rate above age x that is a result of age-distributional differences between the actual population and the standard population. This term will have positive values where the population above age 65 is exceptionally old.

Data

In our primary analysis, we focus on growth rates above age 65, a common age of retirement and entitlement in many countries. We examine growth in the older population and its sources between 2005 and 2010 in a sample of 13 high-income countries drawn from North America, Western, Northern, and Southern Europe, and Asia. Countries were selected on the basis of the availability of detailed historical data series and are broadly representative of variation in demographic conditions among high-income countries. In a second stage, we explore growth at ages 65+ in the aggregates of more developed (MDCs) and less developed (LDCs) countries.

The growth rate of the population aged 65 and above, the arrival rate at age 65, and the death rate above age 65 are calculated using population and death counts by five-year age groups drawn from the United Nations (2011). The arrival rate at age 65 is calculated using data on population size in adjacent five-year age groups centered around 65 at the beginning and end of the time interval. In particular, the number of 65th birthdays occurring in the interval 2005–2010 is estimated by observing the number of people aged 60–64 in 2005 and the number aged 65–69 in 2010. Linear interpolation between these numbers provides an estimate of the number of people who experienced their 65th birthday between 2005 and 2010. The remaining term in Equation 1, the net migration rate, is derived as a residual.

The three historical components of the arrival rate at age 65 are calculated using country-level data on the number of births, deaths, and exposure-to-risk from the Human Mortality Database (HMD) (2011). The relevant birth cohort for persons experiencing their 65th birthday in the interval 2005–2010 is persons born in 1940–45. The “historical birth rate” term, b₀, is calculated using the number of births in this period divided by person-years lived at ages 65+ during 2005–2010. For details on the construction of the series, see the Appendix in Preston and Stokes (2012).

To estimate the probability of survival between birth and age x, we combine age-specific death rates for the relevant birth cohort from each consecutive five-year period and age group. For example, for analysis of the population 65+, we combine deaths and exposure at ages 0–4 during 1940–44 with deaths and exposure at ages 5–9 during 1945–49 and so on until ages 60–64 during 2000–2005. We then apply standard life-table techniques to the cohort age-specific mortality rates to estimate the proportion of original cohort members who live to exact age 65 (using the ratio of life-table values l₆₅/l₀) (Preston, Heuveline, and Guillot 2001).

For decomposition of the death rate above age x into components representing the intensity of mortality and the effect of age composition, the standard population used combines the age distribution of countries in Europe (excluding Eastern Europe) and North America as well as Australia, New Zealand, and Japan (standard population values available upon request).

For analysis of the growth of the older population and its sources in MDCs and LDCs, we use the UN classification system, under which MDCs are defined to include countries of Europe, North America, Australia, New Zealand, and Japan and LDCs comprise the remainder. We remove Eastern Europe from the set of MDCs because its demographic history is both atypical and hard to recover from available data. We use different procedures to identify the contributions of historical births and cohort survival to arrival rates between the two regions because of differences in data availability. The methods used for constructing demographic series for MDCs and LDCs are described in the Appendix of Preston and Stokes (2012).

Results

High-income countries

Table 1 presents growth rates of the 65+ population and the sources of growth in 13 high-income countries for 2005–2010. The mean and variance of each column are presented at the bottom of the table.

TABLE 1.

Demographic components of growth in the 65+ population, 2005–2010

	Growth rate of the population at ages 65+	Rate of arrival at age 65	Death rate at ages 65+	Net migration rate at ages 65+
Country
Australia	0.0255	0.0668	0.0412	0.0000
Belgium	0.0075	0.0531	0.0460	0.0004
Finland	0.0205	0.0642	0.0448	0.0011
France	0.0100	0.0519	0.0420	0.0001
Italy	0.0136	0.0543	0.0424	0.0016
Japan	0.0269	0.0621	0.0350	−0.0002
Netherlands	0.0210	0.0676	0.0465	0.0000
Norway	0.0106	0.0639	0.0508	−0.0025
Spain	0.0144	0.0555	0.0445	0.0035
Sweden	0.0186	0.0678	0.0495	0.0003
Switzerland	0.0171	0.0637	0.0431	−0.0035
United Kingdom	0.0135	0.0617	0.0486	0.0005
United States	0.0203	0.0666	0.0484	0.0022
Mean	0.0169	0.0615	0.0448	0.0003
Variance	0.0035	0.0033	0.0018	0.0003
Region
MDCs	0.0179	0.0611	0.0441	0.0010
LDCs	0.0250	0.0845	0.0595	0.0000

NOTE: Means reported under each column are unweighted. Variances are multiplied by a factor of 100 to reduce the number of reported decimals. MDCs are defined to include countries of Europe, North America, Australia, New Zealand, and Japan and LDCs comprise the remainder. Eastern Europe is removed from MDCs for the above calculations. Description of the methods used to construct historical demographic series for MDCs and LDCs are included in the Appendix to Preston and Stokes (2012).

SOURCES: United Nations (2011); Human Mortality Database (2011).

By far the fastest growth of the older population (2.6 percent annually) occurred in Japan and Australia. Different demographic factors are responsible for the rapid growth of the older population in these two countries. The proximate cause of the more rapid growth of the older population of Australia was its high rate of arrival into the 65+ population combined with slightly better-than-average mortality in 2005–2010. As shown in Table 2, the very high arrival rate in Australia is a product of an extraordinarily high rate of immigration at ages below 65 in the past. The size of birth cohorts reaching age 65 in Australia between 2005 and 2010 was inflated by approximately 61 percent as a result of a huge inflow of migrants at younger ages.

TABLE 2.

Demographic components of the rate of arrival at age 65, 2005–2010

	Rate of arrival at age 65	Ratio of births in 1940–45 to person-years lived above age 65	Probability of survival from birth to age 65 for the 1940–45 birth cohort	Factor by which cohort born 1940–45 has changed through migration by age 65
Country
Australia	0.0668	0.0507	0.8201	1.6083
Belgium	0.0531	0.0628	0.7521	1.1235
Finland	0.0642	0.0861	0.7481	0.9972
France	0.0519	0.0544	0.7484	1.2744
Italy	0.0543	0.0778	0.7288	0.9574
Japan	0.0621	0.0841	0.7258	1.0174
Netherlands	0.0676	0.0830	0.8064	1.0092
Norway	0.0639	0.0774	0.8138	1.0145
Spain	0.0555	0.0771	0.7025	1.0240
Sweden	0.0678	0.0708	0.8297	1.1549
Switzerland	0.0637	0.0633	0.8197	1.2276
United Kingdom	0.0617	0.0784	0.7866	1.0005
United States	0.0666	0.0753	0.7694	1.1495
Mean	0.0615	0.0724	0.7732	1.1199
Variance	0.9386	2.8192	0.3011	2.0418
Region
MDCs	0.0611	0.0761	0.7574	1.0607
LDCs	0.0845	0.2225	0.3799	1.0000*

^*By assumption

NOTE: Means reported under each column are unweighted. Variances are based on the logarithm of the data since the terms are multiplicative and are multiplied by a factor of 100 to reduce the number of reported decimals. See Table 1 for definition of MDCs and LDCs.

SOURCES: United Nations (2011); Human Mortality Database (2011).

The exceptional rate of growth in Japan’s older population in 2005–2010, on the other hand, is principally attributable to the unusually low rate of mortality in older ages in that country. The death rate in Japan is only 78 percent of the mean value of the 13 countries. Table 3 shows that none of the exceptionally low mortality of the 65+ population in Japan is attributable to an unusual age distribution. If Japan had the same death rate at ages 65+ as the mean of the 13 countries (44.8 per 1000 instead of 35.0 per 1000), its growth rate would actually be slightly below the mean rather than the highest. Japan is a world leader in longevity (National Research Council 2011); one consequence is an exceptionally rapid growth of the 65+ population.

TABLE 3.

Demographic components of the death rate at ages 65+, 2005–2010

	Death rate at ages 65+	age-standardized death rate in the 65+population	age-compositional effects on the crude death rate
Country
Australia	0.0412	0.0409	0.0003
Belgium	0.0460	0.0473	−0.0013
Finland	0.0448	0.0473	−0.0025
France	0.0420	0.0392	0.0028
Italy	0.0424	0.0424	0.0000
Japan	0.0350	0.0350	0.0000
Netherlands	0.0465	0.0489	−0.0024
Norway	0.0508	0.0464	0.0045
Spain	0.0445	0.0446	−0.0001
Sweden	0.0495	0.0462	0.0033
Switzerland	0.0431	0.0408	0.0022
United Kingdom	0.0486	0.0476	0.0010
United States	0.0484	0.0464	0.0020
Mean	0.0448	0.0441	0.0008
Variance	0.0018	0.0017	0.0005
Region
MDCs	0.0441	0.0441	0.0000
LDCs	0.0595	0.0764	−0.0168

NOTE: Means reported under each column are unweighted. Variances are multiplied by a factor of 100 to reduce the number of reported decimals. The standard population is the 65+ population of more developed countries in 2005–2010 (according to UN classification) excluding Eastern Europe. See Table 1 for definition of MDCs and LDCs.

SOURCES: United Nations (2011); Human Mortality Database (2011).

Countries with the lowest rates of growth in the 65+ population between 2005 and 2010 are France and Belgium. These countries had exceptionally low arrival rates at age 65 in 2005–2010. Table 2 shows that the low arrival rate reflects very low birth rates during World War II (Dublin 1945). Italy and Spain had the third and fourth lowest arrival rates, but their position is not a result of a deficiency of births but rather of relatively low survivorship to age 65 combined with relatively low lifetime immigration into the cohorts reaching age 65. In Italy’s case, 4.3 percent of the births during 1940–45 left the country before reaching age 65, the greatest loss suffered by any country.

Among the three contemporaneous demographic processes shown in Table 1, the greatest contributor to international differences in the 65+ growth rate is the arrival rate. The variance in the arrival rate is roughly double that of the death rate, while contemporary migration into the 65+ population makes only a small contribution to international differences. Spain is an exception with a relatively high immigration rate into the elderly population, but this is probably a temporary phenomenon related to an amnesty program in 2005 that applied to illegal immigrants (Preston and Wang 2007). Norway and Switzerland experience moderate out-migration among the elderly.

The arrival rate at a given age is a function of the relative size of the original birth cohort, the proportion of those births who survived to age 65, and the rate of migration into the birth cohort before reaching age 65. The relative size of the original birth cohort is the largest source of variance in the arrival rate, although the migration factor is also powerful, in part because Australia is such an outlier (Table 2). France also experienced very rapid immigration, which swelled the size of the original birth cohort by 27 percent by age 65. Cohort survival probabilities also varied, ranging from a low of 70 percent in Spain to a high of 83 percent in Sweden. These rankings reflect comparative mortality levels among children and young adults in the period 1940–60. For example, 13 percent of births died before reaching age 10 in the Spanish life table of 1950, compared to 3 percent of Swedish births in the same period (Keyfitz and Flieger 1968: 456, 504).

Table 3 shows, perhaps surprisingly, that the age structure of the 65+ population has a relatively minor influence on the death rate of the 65+ population. The largest effects are in Sweden and Norway, where relatively old age structures raise the death rate by 3–5 per 1000.

In terms of the components of demographic growth among the elderly, the United States stands out for its normalcy. Its values are not exceptional on any of the eight indexes in Tables 1–3. That circumstance is about to change as baby boomers reach age 65. How results vary with the initial age is considered in the next section.

Extension of results to other ages

In our primary analysis, we have focused on age 65 because of its social significance. Had we chosen another age, the ranking of growth rates across countries would be different. Figure 1 (a) shows growth rates in the five largest developed countries during 2005–2010 at age x+, where x varies from 60 to 80. The patterns in this figure are erratic, with lines that connect the growth rates of particular countries often intersecting. For example, the growth rate above age 60 in US is well above that of France and Italy and on par with Japan, while above age 80, the US growth rate is well below that of Japan, France, and Italy.

FIGURE 1.

Growth rates above various ages (a) for the period 2005–10 and (b) for the birth cohort of 1925–29

SOURCE: United Nations (2011).

The fluctuation in growth rate rankings is explained primarily by variability in the arrival rates, reflecting erratic births series in the period from 1940 to 1960. As an illustration of this point, in Figure 1 (b) we plot the growth rate above age x in the same five countries included in Figure 1 (a). However, in this case we follow a particular set of birth cohorts through time, i.e., those aged 60+ in 1985–90, 65+ in 1990–95,…80+ in 2005–10. Because observations for a country pertain to the same set of birth cohorts, fluctuations in the number of births cannot influence the series. And it is obvious in Figure 1 (b) that when such fluctuations are removed, the series become much more orderly than in Figure 1 (a), with only one cross-over occurring. The growth rate series rise smoothly with age for a particular cohort because mortality improvements are occurring more or less continuously as the cohort ages (Preston and Stokes 2012). This demonstration helps to confirm the evidence in Table 1 that the arrival rate is the principal source of variation in growth rates at older ages.

World regions

We now examine the growth rate of the older population for the aggregate of more developed and less developed countries, using the United Nations’ classification of these groups. Tables 1–3 present the growth rate of the 65+ population and its decomposition for the two regions. Note that the elements of growth for MDCs ins Tables 1–3 are quite similar to the mean of the 13 developed countries.

The older population in LDCs is growing more rapidly than that in MDCs. This difference is entirely a result of higher arrival rates in LDCs. The death rate of the older population in LDCs is much higher than that in MDCs (59.5 per 1000 compared to 44.1 per 1000) despite the fact that its 65+ population is much younger than that in MDCs. As shown in Table 3, once age-distributional differences are controlled, the 65+ death rate in LDCs exceeds that in MDCs by 32.3 per 1000 rather than by 15.4 per 1000.

The decomposition of the arrival rate shows pronounced differences between MDCs and LDCs (Table 2). The term reflecting birth rates in the period 1940–45 is three times higher in LDCs than in MDCs. However, the probability of surviving from birth to age 65 for this birth cohort was only half as great in LDCs (38 percent) as in MDCs (76 percent). If LDCs had maintained the same cohort survival rates as MDCs and all other elements of the growth rate equation remained the same, the annual growth rate of the 65+ population in LDCs would have been 10.9 percent, more than four times its value for 2005–2010. While this is an implausible scenario, it is nevertheless evident that a history of extremely high mortality in LDCs has left a massive imprint on contemporary demographic processes at older ages.

Thus, despite much higher historic birth rates in LDCs, their significantly lower cohort survival, together with the high contemporary death rate of its 65+ population, has produced a growth rate of the 65+ population in LDCs that is only 0.007 higher than that of MDCs.

The future

Our findings on the determinants of older population growth at the beginning of the twenty-first century primarily reflect twentieth-century demographic history. The distinction between arrival rates and death rates is also useful for clarifying elements of the future growth of the 65+ population. Figures 2 (a) and (b) show the arrival rates, death rates, and growth rates that are implied by UN medium-variant population and mortality projections for more developed and less developed countries. In both cases, the death rate declines until about 2030–35, where it levels off. Because projected mortality levels continue to improve, the leveling of death rates is a consequence of an increasingly older age structure among the 65+ population.

FIGURE 2.

Growth, arrival, and death rates, population 65+, 1995–2000 to 2050–2055 in (a) more developed countries and (b) less developed countries

NOTE: Population and mortality data after 2010 are based on UN medium-variant projections. Observations are located at the beginning of the 5-year interval over which each rate is calculated. Projections for MDCs are based on all countries classified as more developed under UN classification, including countries of Eastern Europe.

SOURCE: United Nations (2011).

Arrival rates also show a declining trend in both sets of countries, but the trend is interrupted during the present decade. In the case of more developed countries, the interruption is clearly a result of the postwar baby boomers reaching age 65. In less developed countries, on the other hand, the impending rapid increase in arrival rates appears to be primarily a consequence of large declines in infant and child mortality. Growth rates at ages under 10 accelerated rapidly from 1940–45 to 1950–55, coinciding with the widespread introduction of anti-malarial programs and other public health efforts, including the use of antibiotics. The accelerated growth rates were maintained as these cohorts aged and are now beginning to affect growth rates at ages 65+ (Preston and Stokes 2012).

The sudden burst in arrival rates in both sets of countries will have ended by around 2015. After that date, arrival rates are projected to decline, echoing earlier declines in birth rates, and to do so faster than declines in death rates. Accordingly, growth rates of the older population in both sets of countries are projected to show a persistent pattern of decline after 2015.

The classic graphical representation of the demographic transition shows death rates initially declining, birth rates declining later, the difference between birth rates and death rates first widening and then narrowing, and growth ultimately declining to low levels or ceasing altogether. Figure 2 shows such a pattern in both developed and developing countries for the population aged 65+. While in MDCs this pattern is a somewhat coincidental result of the timing of the baby boom, the projected pattern of growth in LDCs at ages 65+ is an authentic echo of the earlier demographic transition pertaining to all ages combined. And with annual growth rates at ages 65+ still as high as 2 percent in 2050, it is clear that this age-distributional tail will be hosting its own version of a demographic transition for most of the twenty-first century.

Appendix

For a discrete time interval τ the population growth rate above age x, r_x+(τ), can be expressed as

$$r_{x+}(\tau )=b_{x+}(\tau )-d_{x+}(\tau )+m_{x+}(\tau )$$ (A1)

where d_x+(τ) is the death rate, m_x+(τ) is the net rate of migration, and b_x+(τ) is the rate of new arrivals reaching their xth birthday (the arrival rate). Each term on the right-hand side of the equation has a number of events during the period τ in the numerator and the number of person-years lived above age x during period τ in the denominator.

The number of xth birthdays during the period τ is in turn a product of the number of births into the cohort over the period τ − x and any change in the size of the cohort resulting from mortality or migration. Thus the b_x+(τ) term in Equation 1 can be expressed as

$$b_{x+}(\tau )=b_0{(\tau -x)}_x{p}_0^{\tau -x}{{}_{x}f}_0^{\tau -x}$$ (A2)

where b₀(τ − x) is the number of births in the population over the interval τ − x divided by person-years lived above age x during period τ; ${{}_{x}p}_0^{\tau -x}$ is the probability of survival between ages 0 and x for a member of the birth cohort born during the interval τ − x; and ${{}_{x}f}_0^{\tau -x}$ is the factor by which the birth cohort’s size is modified through migration between ages 0 and x. The term b₀(τ − x) is related to the actual birth rate in the interval τ − x by a factor equal to the ratio of the population aged x and above in period τ to the total population in the interval τ − x.

Further decomposition of Equation A1 is possible through elaboration of the d_x+(τ) term into components representing the intensity of mortality and the effects of age composition on the death rate. The death rate above age x can be expressed as

$$\begin{array}{l}{d}_{x+}(\tau )=\sum _x^{\beta }{{}_{1}c}_a(\tau ){{}_{1}m}_a(\tau )+\sum _x^{\beta }{{}_{1}c}_a^S{{}_{1}m}_a(\tau )-\sum _x^{\beta }{{}_{1}c}_a^S{{}_{1}m}_a(\tau )\\ =\sum _x^{\beta }{{}_{1}c}_a^S{{}_{1}m}_a(\tau )+\sum _x^{\beta }[{{}_{1}c}_a(\tau )-{{}_{1}c}_a^S]{{}_{1}m}_a(\tau )\end{array}$$ (A3)

where ₁c_a(τ) is the proportion of the population aged a to a+1 in the interval τ; ₁m_a(τ) is the death rate between age a and a+1 during the period; and ${{}_{1}c}_a^S$ is the proportion of the population ages a to a+1 in an arbitrarily chosen standard population. Each product is summed between ages x and β, the highest age attained. The first term in Equation A3 is the age-standardized death rate above age x, representing the intensity of mortality. The second term is the component of the death rate above age x that is a result of age-distributional differences between the actual population and the standard population.