In developed countries, 4 times more over-65s in 2020 than in 1950

25Across the 40 countries in our sample, there were almost 4 times more people aged 65 or older in 2020 than in 1950. Their number increased from 54 million to 222 million over the period (Figure 2). The over-85s were 15 times more numerous, increasing from 2 million to nearly 31 million. This growth, sometimes referred to as gerontogrowth, will continue to accelerate until 2050. Under the United Nations medium projection (United Nations, 2019b), the 40 countries of the sample will have almost 324 million over-65s and nearly 72 million over-85s in 2050 (Figure 2).

Components of the numerical increase in older adults

26Given that dynamics of population ageing are governed partly by the size of the birth cohorts reaching old age, it is important to distinguish inflows and outflows, as is the case when calculating net population change. Figure 3 shows the numbers of individuals reaching their 65th birthday and the numbers dying after age 65 in France from 1816 to 2017. Not counting migration flows, the difference between the two curves corresponds to the annual increase in the older adult population. Until the mid-20th century, inflows and outflows followed a parallel pattern, and for 150 years (between 1816 and 1966) the number of older adults increased steadily by around 25,000 per year. If we ignore the birth deficits due to the two world wars, the 1960s marked the beginning of a period of more sustained increase in inflows among the oldest adults, reflecting a marked decrease in mortality before age 65 in the cohorts born from the late 19th century onwards (Meslé and Vallin, 1989). [4]

Figure 2

Increase in numbers of over-65s, over-75s, and over-85s from 1950 to 2050 in 40 high-longevity countries

Increase in numbers of over-65s, over-75s, and over-85s from 1950 to 2050 in 40 high-longevity countries

Note: Beginning in 2020, projections are based on the United Nations medium scenario (dotted lines).

Figure 3

Components of annual increase in the population aged 65 and over (excluding migration) in France, 1816–2017

Components of annual increase in the population aged 65 and over (excluding migration) in France, 1816–2017

27Between 1967 and 1979, when the depleted cohorts born in the decades around the First World War reached age 65, the older adult population grew at a rate of close to 110,000 per year. This growth continued between 1985 and 2010 at an annual rate of nearly 140,000. After 2011, the effects of lower mortality among the cohorts born 65 years earlier, combined with the 65th birthdays of the first cohorts of baby boomers (born after 1945), led to a population explosion of older adults in France; excluding migration, their number increased by 324,500 in 2015 alone. In that same year, their number grew by 1.5 million in the United States and by 830,000 in Japan (Table 1). This growth is an important factor to consider for the planning of needs and infrastructure, as more people are entering the older adult population than are leaving it.

A growing proportion of over-65s but very different national dynamics

28An ageing population is not defined simply by an increase in the absolute number of older adults. It is the increase in older adults relative to the whole population that determines the rate of ageing.

29The most widely used international statistics generally use the threshold of 65 years to compare countries and monitor national trends. In the 40 countries of our sample, the share of over-65s, which stood at just 8.0% in 1950, was estimated at 19.7% in 2020 and, according to United Nations projections (2019b), is set to reach 27.9% by 2050, i.e. more than 1 person in 4. However, while this share will increase in all developed countries over the period 1950–2050, the pattern of change varies from one country to another. In the Asian countries of the sample (Republic of Korea, Hong Kong, Japan, and Singapore), ageing is very rapid over the period; they were the youngest countries in the sample in 1950, alongside those of Southern Europe, but will be the oldest in 2050. The dynamics of different countries can be compared by measuring the time taken by the share of over-65s to rise from one level to the next (Pison, 2009). Figure 4 illustrates this approach for a selection of 37 countries, showing the time taken for the share to increase from 10% to 20% and from 20% to 30%. Some countries, such as France, the United Kingdom, Belgium, and Sweden, already had a share above 10% by the late 1940s and will probably not reach 30% until more than a century later. Conversely, population ageing began much later in other countries, but is advancing much more quickly. According to the latest United Nations projections, in the Republic of Korea and Singapore where the share of over-65s was below 10% in 2010, it is set to reach 30% by around 2040 within just 30 years. [5]

Table 1

Components of the annual increase in the population aged 65 and over in 2015 in 28 developed countries

Number of people who turned 65 in 2015 Number of over-65s who died in 2015 Net increase in over-65s in 2015 (excluding migration) Growth rate (%) of over-65s in 2015 Hong Kong 88,000 36,900 +51,100 4.5 Israel 71,800 36,200 +35,600 4.2 Republic of Korea 449,700 205,600 +244,100 3.8 Canada 409,900 210,300 +199,600 3.5 Slovakia 65,500 39,200 +26,300 3.5 Poland 492,500 290,200 +202,300 3.4 Iceland 3,300 1,900 +1,400 3.4 Australia 244,500 128,200 +116,300 3.2 United States 3,488,400 1,992,400 +1,496,000 3.2 Slovenia 27,200 16,100 +11,100 2.9 Finland 75,600 43,900 +31,700 2.9 France 805,000 480,500 +324,500 2.7 Netherlands 202,700 124,100 +78,600 2.6 Japan 1,980,700 1,147,900 +832,800 2.5 Denmark 66,800 43,900 +22,900 2.2 Switzerland 90,200 58,200 +32,000 2.0 Hungary 130,800 100,000 +30,800 1.8 Sweden 114,000 80,000 +34,000 1.8 Portugal 126,500 92,000 +34,500 1.7 Estonia 15,400 11,900 +3,500 1.6 United Kingdom 694,100 509,400 +184,700 1.6 Belgium 126,300 92,500 +33,800 1.6 Austria 91,600 70,300 +21,300 1.3 Spain 468,300 361,900 +106,400 1.2 Germany 999,300 784,400 +214,900 1.2 Croatia 55,100 44,350 +10,750 1.2 Latvia 23,100 21,500 +1,600 0.4 Lithuania 33,000 31,200 +1,800 0.3

Components of the annual increase in the population aged 65 and over in 2015 in 28 developed countries

Notes: The estimated numbers have been rounded to the nearest hundred. Countries are ranked by the over-65s growth rate. This rate corresponds to the ratio between the net increase in over-65s (excluding migration) and the mid-year population aged 65 and over.

Figure 4

Speed of population ageing from 1930 to 2050 in 37 developed countries (population projections beyond 2019)

Speed of population ageing from 1930 to 2050 in 37 developed countries (population projections beyond 2019)

Note: Data from the Human Mortality Database do not go far back enough in time to determine the year when the share reached 10%. Countries are ranked in increasing order by year in which the share of over-65s reaches 10%, 20%, and 30% of the total population.

Changes in age structure: population pyramid and age-specific rate of numerical increase

30Population pyramids are becoming increasingly top-heavy over time, and they reveal the singular patterns of baby booms and baby busts across birth cohorts in each country. For all 40 countries, from 1950 to 2020 and then from 2020 to 2050, the pyramid grows narrower at its base (ageing from the bottom) and wider at the top (ageing from the top), shifting gradually from a triangle into a chimney shape (Figure 5).

31Figure 6 illustrates the striking contrast between France, the oldest country in 1950 (11.4% of over-65s), and the Republic of Korea where, alongside Hong Kong and Singapore, the proportion of over-65s was below 3% at that time. According to the United Nations, ageing in the Republic of Korea will accelerate over the coming years, such that in 2050, its population will be the oldest by far (38.1%). France should occupy an intermediate position in that year (27.8%; Appendix Table A.1).

32As mentioned earlier, the substantial growth—although heterogeneous from one country to another—in the absolute and relative numbers of older adults is linked to country-specific patterns of births, deaths, and (often to a lesser extent) migration flows (Parant, 1992; Calot and Sardon, 1999) (Box 2). The effects of demographic dynamics are difficult to tease apart because these three factors interact in ways that change over time. A peak in births will rejuvenate the population initially, before ageing it when the large birth cohorts reach old age. Likewise, when a large cohort reaches age 65, it first rejuvenates the population of older adults, before having the reverse effect, known as ‘ageing within ageing’. It is vital to anticipate these waves of peaks and troughs (absolute and relative) in numbers of (very) old people in order to plan for future needs and maintain equilibrium in labour markets, pension systems, health, and long-term care.

33Changes in age structure can also be illustrated by the age-specific rate of numerical increase (Figure 7). From 1950 to 2020, the population of the 40 countries in the sample increased at a mean annual rate of 0.73%, but this figure varies substantially by age group, from just 0.06% for the under-25s to 0.87% for the 25–64 age group, 1.89% for the 65–84 age group, and 3.55% for the over-85s. Under the United Nations medium scenario, these differences will be even more pronounced by 2050, with a population decline among the under-65s (–0.27% per year on average between 2020 and 2050) and an increase among the over-65s (+1.19%). The baby boom cohorts are now reaching old age, and as they die, they will progressively be replaced by the smaller cohorts that follow them. In the countries studied here, this means that the growth rate of the older adult population will slow down over coming decades. These rates will apply to ever larger populations, however, so the numbers of older adults will continue to increase at a rapid pace.

Figure 5

Population pyramids (%) of the 40 high-longevity countries in 1950, 2020, and 2050

Population pyramids (%) of the 40 high-longevity countries in 1950, 2020, and 2050

Note: The projections for 2020 and 2050 are based on the United Nations medium scenario.

Figure 6

Population pyramids (%) of France and the Republic of Korea in 1950 and 2020 and projections for 2050

Population pyramids (%) of France and the Republic of Korea in 1950 and 2020 and projections for 2050

*Based the criterion of the share of people aged 65 and over in the population.
Note: The projections for 2020 and 2050 are based on the United Nations medium scenario.

Figure 7

Mean annual rate (%) of numerical population increase by age group in the 40 high-longevity countries between 1950 and 2020 and projection for 2020–2050

Mean annual rate (%) of numerical population increase by age group in the 40 high-longevity countries between 1950 and 2020 and projection for 2020–2050

Summary indicators of population ageing

34These trends can be summarized by several measures of central tendency (Table 2). One such measure is the mean population age. Over a century, it will increase in the 40 countries by 14.7 years, rising from 31.2 years in 1950 to 45.9 in 2050. Quantiles are used to identify changes in the age structure, and they show that the increase in mean age over the period is due mainly to growing numbers of older adults. [6] The first quartile is forecast to increase by slightly more than 10 years over the period 1950–2050, and the third quartile by 20 years.

Table 2

Summary indicators of population ageing in the 40 high-longevity countries, 1950–2050

1950 1990 2020 2050 Measure of central tendency (years) Mean age 31.2 35.7 41.8 45.9 1st decile 4.8 7.5 9.5 10.3 1st quartile 13.5 18.2 23.1 24.1 Median 28.9 34.1 42.4 45.9 3rd quartile 47.4 52.7 60.8 67.4 9th decile 62.3 67.9 73.9 80.9 Relative share of older adults in the population (%) Percentage aged 65+ 8.0 12.6 19.7 27.9 Percentage aged 85+ 0.4 1.1 2.7 6.2 Old-age dependency ratio (a) 12.4 18.8 30.6 48.7 Ageing index (b) 22.5 45.3 91.6 141.6

Summary indicators of population ageing in the 40 high-longevity countries, 1950–2050

(a) Number of people aged 65+ per 100 adults aged 15–64.
(b) Number of over-60s per 100 under-20s.
Interpretation: In 2020, the mean age of the population of the 40 developed countries was 41.8 years; 10% of individuals were under age 9.5, a quarter under 23.1, half were under 42.4, a quarter over 60.8, and 10% over 73.9.

35The most common method uses indicators that measure the relative share of older adults in all or part of the population, such as the share of over-65s, the share of over-85s, etc. But other proportional indicators exist, such as the ageing index (the ratio of older adults to young people; for example, the ratio of over-60s to under-20s, as used by Sauvy [1944]) or the old-age dependency ratio (the ratio of older adults to the working-age population), a measure frequently used to study the impact of ageing on pension systems. [7] Whether expressed as a percentage of the under-20s, the 15–64 age group, or the population at large, the proportion of over-65s has increased sharply since 1950 in all 40 countries. The old-age dependency ratio, for example, which stood at 12.4 over-65s per 100 people aged 15–64 in 1950, had risen to 30.6 in 2020 and is projected by the United Nations to reach 48.7 in 2050. Appendix Table A.4 gives the old-age dependency ratio and median age for each of the 40 countries between 1950 and 2050.

36All these indicators provide confirmation of population ageing. This is illustrated in Figure 8, which shows for each age (x-axis) the proportion of individuals older than this age (y-axis) in 1950, 2020, and 2050. We can see, for example, that the share of over-65s increases from 8.0% in 1950 to 19.7% in 2020 and to 27.9% in 2050. But the growing proportion of older adults is visible whatever the starting age: the curve for 1950 is always below that of 2020, which is below that of 2050. The choice of an age threshold (60, 65, 70, 75 years), sometimes criticized for its arbitrariness, would have no impact on the observed trend. Whatever the indicator used, ageing is a reality and will be relatively rapid in all 40 countries up to 2050.

37More importantly, the use of a fixed age to determine the onset of old age ignores the impact of improvements in health and life expectancy. Alternative measures of population ageing have been developed to consider these factors.

Figure 8

Proportion of individuals aged x and over in the 40 high-longevity countries in 1950, 2020, and projection for 2050

Proportion of individuals aged x and over in the 40 high-longevity countries in 1950, 2020, and projection for 2050

Interpretation: (1) The share of over-65s rose from 8% in 1950 to 20% in 2020 and is forecast to reach 28% in 2050. (2) In 1950, 8% of the population was over 65 years old; in 2020, this proportion corresponds to the population aged 76.3 and over, and in 2050 to the population aged 83 and over (homologous age).

Homologous age

39Alternative measures of ageing adopt a relative approach, rather than a definition linked to a fixed age. Individuals enter the older adult category when, for example, their age places them among the oldest 20% of the population. Under this definition, as pointed out by Ponthière (2017), ‘being old’ means ‘being older than’. A first approach proposed by Calot and Sardon (1999) looked at population ageing using what they call homologous age. They defined a proportion of older adults that remains fixed over time and examined changes in the age at which this proportion is reached in a given population. For example, in 1950, 8% of the people in our sample of countries were aged 65 or over (Figure 8). By 2020, this proportion is reached at age 76.3 and in 2050 at 83.0. Thus, homologous age increases from 65 years in 1950 to 83 in 2050 (Figure 8). Appendix Table A.5 provides homologous ages for each of the 40 high-longevity countries. The increase in this age illustrates population ageing ‘in the vicinity of age 65’ (Calot and Sardon, 1999). Ageing is thus measured as a number of years and corresponds to the horizontal distance between the points of each curve with the same y value (8%, for example, if we are interested in ageing ‘in the vicinity of age 65’). By symmetry, population ageing analysed through changes in the share of older adults in the population is measured by the vertical distance between the points of each curve having the same x value (65 years, for example). This measure via homologous age has the advantage of quantifying population ageing in numbers of years. However, a specific time-invariant quantile of the age distribution must be chosen (8% in our example), along with a specific and time-invariant age (generally 65 years), to use this indicator of the share of older adults in the population (Figure 8).

Threshold age of a variable age category

40A relative conception of age is used by d’Albis and Collard (2013) to measure population ageing. Under their approach, the age of onset of old age is determined endogenously according to the age distribution of the population as a whole and may vary over time as this age distribution evolves. [8] They apply this approach to the United States for the period 1933–2005. As one might expect, given the age structure of the US population, the age of entry into the last age category increased over the period, from 48.7 years in 1933 to 56.6 in 2005. A 55-year-old would therefore be classified as old in 1933 but not in 2005. Applied to a panel of industrialized countries over the period 1955–2005, this approach views population ageing from a new perspective. It suggests that while the share of older adults is trending upward in Australia, Canada, Iceland, and Italy, no significant ageing is occurring in Denmark, France, Norway, Sweden, the United Kingdom, and the United States. In Austria and Switzerland, the population is even getting younger. Ageing measured via the age structure of these two countries has been more than offset by the increase in age at onset of old age.

Prospective age, or remaining life expectancy

41Ryder (1975) proposed a relative measure of age at onset of old age that depends not on the number of years lived since birth but on years left until death: the ‘old’ age category comprises individuals who have x years of remaining life expectancy on average. This approach to population ageing is based on the idea that individuals aged 65 in 1975 are, in a manner of speaking, younger than those aged 65 in 1950, given that the former can expect to live longer than the latter thanks to improvements in longevity. To use the analogy coined by the economist Victor Fuchs (1984), considering a 65-year-old in 1950 as identical to someone of the same age in 1975 or 2020 would be as meaningless as comparing mean wages in these years without correcting for inflation. Population ageing must not, therefore, be viewed in terms of individuals’ nominal age, but rather their real age adjusted for changes in mortality rates.

42The most important and widely disseminated contributions to this approach are those of Sanderson and Scherbov (2005). They introduced the concept of prospective age with the aim of capturing a more physiological dimension of individual age that is liable to vary across countries and over time. Figure 9, similar to the one in Sanderson and Scherbov (2008), shows how prospective age differs from chronological age at the global level.

Figure 9

Remaining life expectancy of females across the world, chronological age, and prospective age, in 1950–1955, 2015–2020, and 2045–2050

Remaining life expectancy of females across the world, chronological age, and prospective age, in 1950–1955, 2015–2020, and 2045–2050

Interpretation: The prospective age of 65 is reached at age 74.1 for females over the period 2015–2020 and at age 76.6 in the period 2045–2050 (based on the reference period 1950–1955).

43Over the period 1950–1955, female life expectancy at age 65 was 12.1 years worldwide. The top of Figure 9 shows that it has increased considerably since then. This increase represents a rejuvenation of women aged 65, which Sanderson and Scherbov express as a prospective age. The bottom of Figure 9 shows that the age at which women have 12.1 years of remaining life expectancy is 65 years in 1950–1955, 74.1 in 2015–2020, and an estimated 76.6 in 2045–2050. Or stated differently, women aged 74.1 in 2015–2020 and 76.6 in 2045–2050 have a prospective age of 65 (based on the reference period 1950–1955). By analogy with Sanderson and Scherbov’s ‘40 is the new 30’, septuagenarians will be the new sexagenarians (Sanderson and Scherbov, 2008).

44Figure 10 shows the change in the relation between female chronological age and prospective age from 1950–1955 to 2045–2050 in the 40 developed countries of our sample. [9] Prospective ages are calculated by reference to United Nations estimates of life expectancies in the period 1950–1955 at the global level. [10] Whatever the chronological age, the prospective age in our sample of developed countries is always lower than the global prospective age, reflecting higher life expectancy in these countries than in the rest of the world. The figure also shows a rejuvenation with respect to chronological age between 1950 and 2050. For instance, a woman aged 80 in 1950–1955 could expect to live for the same number of years as a woman aged 88 in 2015–2020 and, according to United Nations projections, as a woman aged 90 in 2045–2050. The prospective age approach gives a much more nuanced picture of ageing than standard measures, as shown in Figure 11 which compares the share of over-65s in the population of 40 countries by prospective and chronological age.

Figure 10

Change in the correspondence between female prospective and chronological age in the 40 high-longevity countries, from 1950–1955 to 2045–2050

Change in the correspondence between female prospective and chronological age in the 40 high-longevity countries, from 1950–1955 to 2045–2050

Note: Prospective ages are calculated by reference to global life expectancies in the period 1950–1955.
Interpretation: A woman aged 80 in 1950–1955 could expect to live for the same number of years as a woman aged 88 in 2015–2020 and, according to United Nations projections, as a woman aged 90 in 2045–2050.

Figure 11

Change in the share (%) of people considered old, by chronological age (65 years) and prospective age in 40 high-longevity countries, 1950–2050

Change in the share (%) of people considered old, by chronological age (65 years) and prospective age in 40 high-longevity countries, 1950–2050

Note: Prospective ages are calculated by reference to global life expectancies in the period 1950–1955. The curve of chronological age represents the change in the proportion of over-65s. The curve of prospective age represents the change in the proportion of people whose remaining life expectancy is below that of a 65-year-old in 1950–1955 at the global level (12.1 years).

45Once life expectancy gains are taken into account in the choice of age at onset of old age, population ageing only becomes perceptible from the 2020s, and its progression is much slower. Before this threshold, the share of people with a prospective age of more than 65 years remains practically stable across the 40 countries. This tendency is the result of two opposing effects created by the increase in life expectancy. The first is an ageing effect, resulting from the numerical increase in older adults (based on year of birth). The second is a rejuvenating effect via the increase at each age in the average number of years left to live. In the standard approach by chronological age, only the first effect is considered, which explains the trend difference between the two measures.

46The prospective age approach could be extended by setting the age at onset of old age on the basis not of years left to live, but rather of the share of LEB left to live. [11] These refinements of the way the threshold age is determined are largely dependent on life expectancy gains. In this respect, the increase in life expectancy is a key component of population ageing in the early 21st century, a topic the next subsection will address.

Life expectancy at birth

48Life expectancy summarizes the mortality conditions observed on a given date. It corresponds to the mean number of years that would be lived by members of a population who, at each age of their life, were exposed to the same average risks of dying as people of the same age observed in year x (estimated from the population size and deaths recorded in that year). It does not provide prospective information on the age at which the members of a population can expect to die; rather, it reflects the mean age at which deaths occurred in a given year. In 2019, global life expectancy was 75 years; [12] it exceeded 80 years in around 20 countries, reaching as much as 88 years for females in Japan and Hong Kong (Pison, 2019). Past life expectancy gains were attributable mainly to a decline in infant and child mortality, while progress in the last 40 years has been due above all to improved survival at advanced ages. [13] In low- mortality countries, more than 1 in 10 deaths occur after age 90, compared with 1 in 33 in the 1950s.

49Figure 12 illustrates the diverse range of trends in the 40 countries of our sample. It shows LEB and at ages 65 and 85 in Bulgaria, Denmark, Spain, the United States, France, and Japan, chosen because of their distinctive situations. Excepting the fluctuations related to the two world wars, especially visible in France, LEB has increased steadily across decades in Japan, France, and Spain and has exhibited the highest values since the 1970s. In France, LEB has doubled since the late 19th century, rising from around 40 years for both sexes to 85 for females and 80 for males in the late 2010s. Life expectancies at age 65 in France stagnated throughout the 19th century (at around 11 years), then rose rapidly, reaching 14 years in 1950 and 21 years in 2015. Progress at age 85 began later, rising from 4 years in 1950 to 7 years in the first decades of the 21st century. Japanese life expectancy at age 65 has increased at a remarkable pace. It has ranked highest in the world since the 1980s, reaching 22 years in the period 2010–2017, despite tailing off slightly in recent years. Starting from relatively low LEB in the 1910s, Spain has caught up with the leaders and now ranks close to Japan and France. The pattern in the United States is different. While very similar to France and Spain in the 1950s, gains in LEB and at age 65 have been smaller since the 1980s. In Denmark likewise, where life expectancy was higher than that of many other countries from the mid-19th to the mid-20th century, progress also began slowing from the 1950s. Consequently, Denmark ranks behind the United States in the early 21st century, despite a narrowing of the gap in recent years. In Bulgaria, LEB stood at 74.5 years in the 2010s, a relatively low level with respect to other European countries. It stagnated between the 1960s and 1990s, a period of high mortality at working ages and among older adults. Life expectancy at age 85 fell by almost a year during these decades, although progress resumed between 2000 and 2010. Variations among the countries in our sample reflect differences across space and time in living and working conditions and in healthcare and social protection. Large mortality differences are still observed today between these record longevity ‘pioneer’ countries and many countries of Eastern Europe (McMichael et al., 2004).

Survival disparities beyond age 65 by sex…

50Females have longer life expectancy than males in the countries of our sample, with a few rare exceptions (such as life expectancy at age 85 in Bulgaria). However, the gender gap has been narrowing everywhere in recent years. The gender difference in LEB is very large in France, despite a recent improvement. As shown in Figure 12, it fell from around 8 years in the 1980s to 6 years, as in Japan, in the 2010s. At advanced ages, the gender gap increased sharply in the 20th century, from the 1950s especially. The gender gap at age 65 started narrowing in the 1980s in Denmark and the United States, in the 1990s in France, and in the early 2000s in Japan. It is very small in Spain (2 years at age 65), where life expectancies for both sexes are among the highest in Europe, and narrower still in Bulgaria, where they are among the lowest.

Figure 12

Life expectancy at birth, at age 65 and age 85 (both sexes) and gender gap in 6 countries, 19th–21st centuries

Life expectancy at birth, at age 65 and age 85 (both sexes) and gender gap in 6 countries, 19th–21st centuries

51The female advantage is explained partly by distinct physiological traits that protect and expose males and females differently. But it also reflects pronounced differences in practices that increase or reduce exposure to health risks. The decrease among men in risk practices, such as smoking and drinking, and their adoption of healthier practices (prevention, diet, etc.), as women had already done before them, has helped to narrow the gender gap in life expectancy. At the same time, female exposure to certain health risks has increased. This is notably the case for smoking, which, while remaining below the high levels once observed among men, is an issue of concern today.

… and social status

52While most research on differential mortality by social status concerns the working-age population for reasons of data availability, some studies focusing on older adults have revealed substantial life expectancy differences at advanced ages. [14] A study in Great Britain, for example, showed that in the years 2007–2011, life expectancy at age 65 was 4 years higher for males (3 years for females) among the former most qualified workers than among the former least qualified workers and that these differences did not narrow over the period (White, 2015).

53Social inequalities of life expectancy at age 65 partly reflect individual histories of exposure to health risks and differences in healthcare. They may be due, for example, to a lack of resources (poverty), environmental exposure (pollution, poor sanitation), detrimental working conditions, deprivation and insecurity (financial, occupational, etc.), health risk practices, and lack of access to appropriate care. All these factors are prejudicial to health and are linked to social status (income, occupation, education, social background) (Marmot et al., 2008).

54Life expectancy differences across social groups become narrower at advanced ages, in absolute values. This was sometimes interpreted as a selection effect of the most robust individuals with age that weakens the impact of social status on mortality risk among those who have survived to old age. But the mortality differences, both persistent and large in relative values, also suggest that if such selection exists, it does not entirely efface all inequalities, which remain difficult to explain due to a lack of data for these ages (Jylhä and Luukkaala, 2006). These studies also reveal that social inequalities in life expectancy are greater for males. It is likely, however, that occupation or educational level (the most widely used criteria) are less revealing of critical social situations for females than for males. Occupation and educational level provide a more accurate representation of household living conditions for men than for women, at advanced ages especially. When social criteria more directly representative of women’s living conditions (such as household income) can be used, much larger mortality inequalities are found in the female population.

Change in maximum life expectancy

56One way of determining whether further life expectancy gains are still possible for a given population is to observe the records of life expectancy and human longevity, which provide an indication of future survival ‘reserves’. Progress in LEB is tracked via the annual records achieved by females: the countries who hold these records are considered ‘pioneers’. According to statistical records, eight countries were record holders between 1840 and the 2000s, starting with Sweden, followed by Norway, then Australia, and Norway once again. From the 1960s, Iceland, the Netherlands, and Switzerland joined this vanguard group. From the 1980s, these countries were overtaken by Japan—still the world reference 40 years later with a female LEB of 87.5 years in 2019 (Oeppen and Vaupel, 2002).

57The curve of maximum life expectancies shows clear sequences characterized by the pace of gains in years of life (Vallin and Meslé, 2009). A first period of increase (1790–1885) corresponds to the development of the first vaccines and to the decrease in deaths from famine and infectious diseases (trends in causes of death will be detailed later). This period was followed by an acceleration of gains thanks to medical discoveries and notable improvements in health and social protection from the late 19th century (urban sanitation, primary healthcare, disease prevention). Lastly, the gains achieved since the 1960s reflect the impact of the cardiovascular revolution. Whatever the countries considered (depending on the nature and quality of available data), all studies point to a reserve of life expectancy gains.

58The maximum values reached indicate the scope for future progress in countries with still a long way to go (Jasilionis et al., 2014); they highlight the lack of progress in some countries and, in others, the potential for a convergence of mean and maximum life expectancies. Analyses have also focused on ‘second-best’ countries in life expectancy, such as Switzerland in the 1980s, France up to 1995, and Spain since then (Meslé and Vallin, 2019), to identify challengers with the potential to overtake Japan in the coming decade. These countries include the Republic of Korea, which will probably soon join the ranks of the pioneer countries thank to its rapidly increasing life expectancy.

Over-90s, centenarians, and supercentenarians

59According to the United Nations (2019), the number of centenarians in the world, which stood at 150,000 in 2000, reached 573,000 in 2020 (78% women, 22% men), of whom 46% are in Asia, 22% in Europe, and 18% in the United States (312,000 in the 40 countries under study). While these figures sound spectacular, they reflect a minute part of overall population ageing from the top of the pyramid. From 8 million worldwide in the early 2000s, the number of over-90s topped 21 million in 2020, of whom 69% were women, and is set to reach more than 76 million by 2050, according to United Nations projections. Across the 40 countries in our sample, this represents an increase in the share of over-90s from 1% of the population in 2020 to 2.7% in 2050. This proportion is forecast to double or triple between 2020 and 2050 in each of the 40 countries, except the Republic of Korea and Singapore, where the share of over-90s will be multiplied by 5 and 9, respectively. In Japan, the proportion of over-90s will be the highest (1.8% in 2020 and 4.7% in 2050). It is difficult to accurately predict future numbers of very old people, however (Robine and Cubaynes, 2017); the margins of uncertainty are linked, among other things, to the difficulty of measuring mortality at these advanced ages. The International Longevity Database, which includes a large volume of data on people who have reached age 105, will make it possible to analyse this oldest-old mortality and to identify the mortality law (Box 3).

What is the law of oldest-old mortality?

60The size of the oldest-old population and its evolution depends on the size of the cohorts reaching advanced ages and on survival probabilities. Until recently, data on mortality beyond age 80 were not considered robust, making it difficult to estimate survival probabilities at these ages. In the 1990s, when older populations started to grow, the models of oldest-old mortality had to be revised to improve the fit between estimates and observations (Kannisto et al., 1994). Specifically, researchers began to question the application of the Gompertz law of mortality (1825) to the oldest-old and its assumption that probabilities of dying increase exponentially with age. While it is agreed that the law works well between entry into adulthood and age 85 or so (in low- mortality populations), opinions diverge about its accuracy beyond this threshold (Robine et al., 2020). Some studies have found that death rates increase more slowly or even level off at very advanced ages. A range of hypotheses have been put forward to account for this deceleration: selection of the most robust individuals less sensitive to health risk factors that occur up to advanced ages; a slowing of the process of senescence at these ages; (over-)protective lifestyles or environments that mitigate the risks of disease or injury. But others point up the limits of existing data and modelling methods. To estimate mortality by age, individuals are often grouped by age at death, whatever their birth cohort, to maximize sample size. This produces heterogeneity in the experience of mortality (that varies across cohorts), which may partly explain the existence of an artificial plateau effect that does not necessarily represent actual mortality in a given period. Lastly, analysis at country level is difficult due to the small numbers of people concerned. In an initial study of semi-supercentenarians in Italy based on data from the International Database on Longevity, the authors concluded that mortality risks level off after age 105 (Barbi et al., 2018). Their article prompted numerous responses (presented as comments) questioning the robustness of the chosen assumptions and models due to the inadequate sample size. It was pointed out that modelling constraints and margins of uncertainty are still too large to determine whether the probability of dying by age does or does not level off after age 105.

Indicators of changes in human longevity

61Life expectancy does not capture the full dynamics of oldest-old mortality. While it provides the mean age of death in a given year, it says nothing about the distribution of ages at death. As shown in Figure 13, based on French life-table data for different periods, the proportion of deaths before age 1 has decreased sharply, from 17% of all deaths under the mortality conditions of 1850–1859 to just 0.4% under those of 2010–2017; and from 31% before age 10 in the first period to 0.5% in the second. The curves show a shift in deaths towards advanced ages. In 1850–1859, 31% of deaths occurred after age 65 versus 88% in 2010–2017. They also show that deaths are concentrated within an increasingly narrow age range and that the modal age (the age at which deaths are most frequent) has shifted from 72 years for the two most distant periods to 79 and then 89 for the two most recent. This modal age and the concentration of deaths around it complement the information provided by life expectancy, making it possible to compare countries or population groups. This corresponds to the concept of rectangularization of the survival curve (Fries, 1980), described in Box 4.

Figure 13

Distribution of deaths (%) by age at death, based on life tables for France (averages of the periods 1850–1859 to 2010–2017, both sexes combined)

Distribution of deaths (%) by age at death, based on life tables for France (averages of the periods 1850–1859 to 2010–2017, both sexes combined)

Note: Two points are not represented (17% of deaths occur before age 1 in the decade 1850–1859; 14% of deaths occur before age 1 in the decade 1900–1909).

Life expectancy fluctuations in the early 21st century

62The record life expectancies observed in pioneer countries suggest that further progress is still possible. That said, life expectancy has fluctuated recently in many of the most advanced countries in our sample for reasons that remain unclear. Examples include the United States, where mortality is tending to increase among certain population groups (Case and Deaton, 2015; Barbieri, 2019), and some European countries where unexpected jumps in life expectancy at age 65 are observed in certain years, as illustrated in Figure 14 (in France in 2011 and 2014, for example). In 2015, on the other hand, Europe saw a marked dip from 2014 (slightly more than 7 months in Italy).

Box 4. Rectangularization of the survival curve

The age distribution of deaths can also be plotted on a survival curve, based on the age-specific probabilities of survival given in the life table. These probabilities are calculated by dividing the number of deaths in a given year between ages x and x + 1 by the number of individuals of age x who were alive at the beginning of the year. These probabilities at each age are then applied to a fictitious cohort of, say, 100,000 newborns to determine the numbers who reach each age. The curve represents the decreasing numbers of these 100,000 newborns as they advance in age and all eventually die. With the progressive decrease in mortality before the modal age in recent decades, the curve now remains relatively flat until this age. It then dips downwards, forming an increasingly rectangular shape as deaths become more concentrated around the modal age. Fries (1980) argued that this dynamic should lead to a compression of deaths into the average upper limit of life, set at 85 years in 1980. While a trend towards rectangularization and compression of deaths around a modal age was indeed observed at that time, there is still no sign of an upper age limit in the first decade of the 21st century.

Survival curve based on life tables for France (average of periods 1850–1959 to 2010–2017, both sexes)

Survival curve based on life tables for France (average of periods 1850–1959 to 2010–2017, both sexes)

63These dips may be linked to flu or gastroenteritis epidemics or to heatwaves that are particularly dangerous for the most fragile and oldest populations. These episodes increase respiratory and circulatory mortality, especially among women and the most disadvantaged populations (Khieu et al., 2017). The European heatwave of August 2003 was especially deadly for older adults, with 15,000 excess deaths recorded in France between 1 and 20 August (Toulemon and Barbieri, 2008). Seasonal flu epidemics produce similar mortality peaks. According to Santé publique France, the French public health agency, around 13,000 flu deaths in France occurred in the winters of 2014–2015 and 2016–2017 and around 11,000 in 2017–2018, of which 85% were among people aged over 75. The first (provisional) estimates of the effects of the COVID-19 pandemic on overall mortality in Europe reveal high levels of excess mortality, at advanced ages especially (Dahoo and Gaudy, 2020). During the first wave (March–April 2020), deaths increased by 70% from previous years in Spain, by 47% in Italy, 43% in Belgium and the United Kingdom, and 28% in France, but just 4% in Germany, which fared better than elsewhere. The second wave saw further excess mortality in Europe, with effects that differed across countries. According to INSEE estimates, the excess mortality in France reduced LEB by 5 to 6 months (Papon and Beaumel, 2021). By contrast, no major seasonal flu epidemic occurred in 2020, so flu deaths were much lower than in previous years. Lockdowns and other measures to contain the pandemic also reduced deaths from other causes, such as accidents and injuries. It is still difficult to assess the longer-term consequences of the pandemic, however, as its potential impact on the health of people who contracted the disease or were affected indirectly by the disruption of healthcare and of social and economic life is still unknown.

64Do these life expectancy fluctuations at advanced ages mean that gains are levelling off? To answer this question, trends must be observed over the long term to identify any lasting slowdowns. The strong decrease in cardiovascular mortality has made a substantial contribution to the gains observed in recent years. Scope for progress is now more limited, although further advances are probably still possible through targeted prevention, screening, and healthcare for high-risk subpopulations. Additional gains can also be achieved if we succeed in slowing the progression of old-age mortality linked to mental, neurodegenerative, respiratory, and circulatory diseases.

65Recent studies of the prevalence of dementia in several European countries and in the United States have found overall evidence of a downtrend (for example, Jagger et al., 2016) due perhaps to improvements in health (notably cardiac health, a risk factor) and living conditions over the life cycle. The rise in average educational levels across cohorts is contributing directly to these improvements; duration and level of education are associated with later onset of disabling symptoms of dementia.

Figure 14

Life expectancy at age 65 in the 28 countries of the European Union, 2009–2018

Life expectancy at age 65 in the 28 countries of the European Union, 2009–2018

Causes of death at advanced ages and multimorbidity

66Even though the presence of multiple health problems makes it difficult to identify cause of death at advanced ages, the causes recorded on death certificates can be analysed to understand trends and differences in oldest-old mortality (Meslé, 2006). Figure 15 presents the standardized mortality rates at ages 90 and over between 1979 and 2015 in France (Meslé and Vallin, 2020). It shows the decrease in deaths from circulatory and cardiovascular diseases from the 1980s, among women at least. The decline in mortality from infectious and respiratory diseases is more recent, starting in the 2000s, and less spectacular.

Figure 15

Standardized mortality rates by cause of death at ages 90 and over, France, 1979–2015

Standardized mortality rates by cause of death at ages 90 and over, France, 1979–2015

67Deaths from external causes (accidents [including falls], suicides, etc.) have been decreasing quite steadily since 1980, and slightly faster for women. At the same time, mortality from neurodegenerative diseases and dementia has increased sharply; many deaths previously attributed to senility (a cause in sharp decline) are now diagnosed more accurately and reclassified, often as dementia. But the rise in dementia mortality also reflects the increase in survival to ages where diseases of this kind are more frequent. Cancer mortality among the oldest women has remained stable for around 3 decades. Among men, it increased in the 1980s and 1990s before dipping again to its 1985 level. To study trends in cancer mortality, deaths must be broken down by cancer site. Risk exposure (which differs for males and females) and specific progress in treatment and prevention affect the pace and timing of mortality reduction. And some curves are uneven, notably those of infectious and respiratory diseases, while others have peaks related to the health crises described above (heatwaves such as that of 2003, seasonal epidemics) that may compromise health and expose people to greater risk of illness or complications.

Does a longer life expectancy imply more or fewer years of poor health?

72It is crucial to know whether years of life expectancy gained are spent in good or poor health. During the 1980s, some predicted a pandemic of mental disorders and associated disabilities, frequent among older people and likely to affect an increasingly large population (Kramer, 1980). Others put forward a scenario involving a compression of years of late-life poor health, with premature death virtually non-existent and life expectancy eventually reaching an average maximum life span; for example, 85 years (Fries, 1980). Still others expected a ‘dynamic equilibrium’, produced by a probable increase in poor health associated with increasing age and the development of chronic diseases, while the severity and consequences of these diseases are reduced (Manton, 1982).

73This is a key issue because whether years gained are spent in good or poor health determines people’s quality of life and level of social participation. It determines the needs of a population regarding health care and support for loss of functional independence (Manton et al., 2006). It affects the equilibrium of social protection systems in their current configurations when it comes to health care and support costs and pension funding. Lastly, it provides an indication of older adults’ ability to remain in employment, a vital piece of information in the debate on extending the retirement age. Demographers have therefore developed health expectancy indicators that combine the traditional measure of life expectancy with data measuring certain aspects of health.

What measures of population health are needed to understand the consequences of ageing?

74In the context of an ageing population, one of the major public health concerns is chronic disease (often involving long-term or even life-long treatment), which can be disabling. With large numbers of people now living to advanced ages, we have witnessed the significant emergence of such diseases, rare before the age of 80, and particularly of neurodegenerative diseases with serious consequences leading to disabilities, such as Alzheimer’s and Parkinson’s (Box 5). These disabilities can be categorized as involving either functional limitations or activity restrictions. The former relate to functional impairments that cause difficulties in moving about, seeing, hearing, remembering, reasoning, etc. They can often be reduced through the use of equipment, such as mobility aids, glasses, hearing aids, etc., or through the adaptation of the environment, particularly to facilitate mobility. Where they cannot be (adequately) mitigated, functional limitations may lead to activity restrictions. These are situations in which individuals have difficulty carrying out basic activities of daily living: seeing their family and friends, doing the shopping, doing housework, or, in the most critical cases, performing basic self-care activities such as washing and feeding themselves. When people can no longer perform these activities, they lose some of their functional independence and require support. When diseases affect cognitive functions, there is a risk of a loss of ‘decisional’ autonomy. Alzheimer’s disease and related disorders have been a public health priority since the turn of the 21st century, as the functional limitations they cause are difficult to manage or compensate for (confusion, disorientation in time and space, etc.) This leads to significant care requirements, particularly because even if the person can carry out many of their activities, frequent loss of bearings means they require constant supervision and assistance, even to the point that decisions are made for them.

75In addition to severely disabling diseases, moderately disabling conditions are common in adults and especially older adults. Of these, chronic musculoskeletal disorders and anxiety or depression are particularly significant. Because these diseases lead to the restriction of certain social, family, domestic, or personal care activities, they are responsible for substantial deterioration of the individual’s quality of life and social participation, and ultimately contribute to the frailty of the population. For example, a study based on French data from 2008 indicates that, among the 50+ age group, 32% of all situations of health-related activity limitation in women were associated with musculoskeletal disorders and 23% in men, 11% and 19% were associated with cardiovascular disease, and 4% and 6% with chronic non-specific lung disease. The contribution of neurodegenerative diseases (Alzheimer’s and Parkinson’s) is around 3%, less than that of accidents (in men) and cancers (Nusselder et al., 2019). Far more situations generate functional deterioration than we might initially think. They could be viewed as an area for significant preventive action. In addition to medical and prevention issues, it is important to consider the various situations of loss of functional independence and the associated needs. In the context of ageing, using a disability-based approach provides a comprehensive view of these risks. Looking at this functional dimension of health also captures the dimensions of quality of life, social participation, and loss of functional independence. Health expectancy indicators have been developed using these measures of disability.

Describing and monitoring the health status of the population

76In the 1970s, the concept of health expectancy was proposed to add a ‘quality’ dimension (years lived in good or poor health) (Sullivan, 1971) to the estimate of ‘quantity’ of years lived (life expectancy). Although developed and often used with measures of disability (disability-free life expectancy, active life expectancy), there are, in fact, as many indicators of health expectancy as there are measurable dimensions of health. From a technical perspective, measuring health expectancy involves incorporating the probability of having a health problem (chronic illness, disability, dependency, etc.) into the calculation of life expectancy; the years of life expectancy are then divided into years with or without this problem. Repeated over time or calculated for different populations, these estimates enable us to answer questions such as: Are the life years gained years with or without disability? Do population groups that live shorter lives also spend fewer years in poor health?

Life expectancy with and without disability in low-mortality countries

77Disability-free life expectancy has been calculated in many countries since the 1980s. In 2004, the European Commission included the indicator in its list of structural indicators for each country under the label ‘Healthy Life Year’ estimated by Eurostat [16] using data from the Study on Income and Living Conditions, or EU-SILC (Jagger et al., 2008). The International Research Network on Health Expectancies (REVES) has produced two international summaries on trends, disparities, and determinants of healthy and unhealthy life years, with and without disability (Robine et al., 2003; Jagger et al., 2020). [17]

Box 5. Measuring health: disease, disability, and loss of functional independence

Since the 1980s, the literature has proposed models explaining how illnesses can lead to disability, which in turn can lead to dependence (people are then dependent on the assistance of another person to perform essential daily activities) (Wood and Badley, 1978; OMS, 1980; Verbrugge and Jette, 1994). The diagram below is an illustration of one of these models. The benefit of the approach is that these various dimensions can be broken down to explore separately the risk of having diseases (prevention and primary care), the risk that these diseases will lead to functional limitations (secondary care), and the likelihood of being able to prevent these limitations from generating activity restrictions, through the use of effective technical aids, for example, or a sufficiently adapted environment (Verbrugge et al., 1997; Agree, 1999). Lastly, it can be used to measure the situations and needs of individuals who have lost some of their functional independence or (decisional) autonomy.
Question grids were developed in the 1980s to incorporate functional health into health surveys and provide information that could be used at the population level: grids identify functional limitations (going up and down a flight of stairs, clearly recognizing the face of someone on the other side of the road, clearly hearing what is said in a conversation in a quiet room, etc.) (Nagi, 1976), and other restrictions in elementary activities of daily living ([ADLs] grooming, feeding, etc.) (Katz et al., 1963) or in instrumental activities of daily living ([IADLs] shopping, laundry, administration, etc.) (Lawton and Brody, 1969). These functional health models have been widely used to describe the health status of populations. At the individual level, they have also been used to understand trajectories of functional decline. The aim is to estimate the risks of moving from one level of the disability development process to another (deterioration or improvement) in order to identify the associated determinants and disparities.

Illustration of the disability development model

Illustration of the disability development model

78The following conclusions emerge. The most complex (and rarest) situations of self-care activity restrictions represent about 5 years of total life expectancy, in all countries. [18] These years occur at advanced ages, and this figure appears to be fairly stable over time, from the 1980s to the early 2000s. These results, which are relatively consistent from one country to another, indicate that years of life expectancy gained have not been years of so-called severe disability, since the latter have not increased along with life expectancy. However, the most recent studies show less homogeneous trends (Robine et al., 2020). In England and the United States, the number of years lived with severe disabilities (i.e. difficulty or need for assistance with basic personal and domestic care activities) has increased slightly.

79As for less severe disabilities (restrictions in activities other than the essential ones), levels and trends are much more varied from country to country. This is particularly noticeable with the measure used for the European indicator: ‘activity limitations’ are situations in which people perceive they have been limited in usual activities for more than 6 months due to a health problem. These disabilities cover a variety of situations as thought by the respondents: difficulty doing shopping or going out in general, doing housework, working (for those in the labour market), managing everyday administrative tasks, etc. Across Europe, in 2018, women who had reached 65 years of age could expect to live a further 10 years without activity limitation (called Healthy Life Years), and men a further 9.9 years (Figure 16); but as life expectancy for women is higher, they spend, on average, longer living with these limitations (11.5 years) than men (8.3 years). Between countries, we see considerable variation: about a 10-year difference in Healthy Life Years at age 65 in Sweden or Norway (almost 16 years for women and men) and those of Latvia, Slovakia, or Croatia (around 5 years). Overall, activity limitations affect between 8 and 15 years of life expectancy for women and between 5 and 12 years for men.

80Some countries saw an increase in the indicators for life expectancy with moderate disability around the turn of the 21st century (Japan, Belgium, Spain, Sweden, France, and USA), particularly in the latter three countries where this disability expansion reached people in their 50s and 60s.

81Lastly, regarding life expectancy with and without functional limitations (sensory, motor, or cognitive), the last 3 decades have seen an upward trend in the number of years lived with these difficulties, which are frequent at advanced ages (Robine et al., 2020). In the mid-2010s, we also saw a decline in the number of years lived with cognitive functional limitations (remembering, being oriented in time and space, etc.), which illustrates the trend towards lower prevalence of the signs of dementia, for example in the United Kingdom (Jagger et al., 2016).

Explaining the differences between countries and their trends

82Levels of disability-free life expectancy vary significantly between European countries (Figure 16), particularly when considering disabilities that are not the most severe. These differences are partly related to the way health and disability are measured: the wording of questions, range of responses, cultural variations in the understanding of one’s own health, etc. (Cambois and Robine, 2017; Robine et al., 2020). Even when the measures are relatively similar, differences remain not only between the countries of Europe, but also between the 50 US states (Karas Montez et al., 2017). These differences raise questions about the impact of health and social contexts, with health and social protection systems resulting in situations that may be more or less conducive to the prevention and management of diseases and their consequences (Rechel et al., 2013). Two analyses of the association between the level of health expectancy and macroeconomic indicators in EU countries around 2010 showed that variations in the level of poverty and material deprivation partly explained these disparities, particularly between older and newer EU countries (Fouweather et al., 2015). [19]

Figure 16

Life expectancy at age 65 with and without activity limitation in European countries in 2018

Life expectancy at age 65 with and without activity limitation in European countries in 2018

83These macroeconomic indicators, such as the poverty level, reflect the living conditions that affect disability-free life expectancy. These indicators undoubtedly have a strong social dimension and are linked to public policies that can limit the consequences of social situations that are critical to health, specifically through poverty reduction, universal access to care, education, unemployment benefits, etc. Therefore, we also find connections between public policies and inequalities in disability-free life expectancy within countries.

Health inequalities among older people within each country

84As for total life expectancy, differences in life expectancy with and without disability can be observed between men and women and between social groups. Women live for longer than men, but for many women, some or all of these additional years are lived with disabilities, regardless of how it is measured. Figure 16 shows, for example, that while, for the 28 European countries, average life expectancy for women is 3.3 years longer than for men (21.5 vs. 18.2), they also experience 3.2 additional years with disabilities (11.5 vs. 8.3). This female disadvantage, sometimes described as a paradox if we consider that longevity and health go together (Luy and Minagawa, 2014), is the result of disease ‘profiles’ that appear specific to each sex. On average, women are more likely than men to have chronic disabling diseases, such as anxiety or depression and musculoskeletal conditions, while men are more likely to experience life-threatening disabling diseases, such as heart disease, cancer, or traumatic injuries (Nusselder et al., 2020). [20] Differences in the life circumstances and trajectories of men and women may also contribute to these disparities in the risk of disability at advanced ages (Cambois et al., 2017): standard of living, career, circumstances, and family backgrounds are all factors that impact health and that differ between men and women at advanced ages.

85Social inequalities, already significant for LEB, increase still further when we look at years of good and poor health. In France, in the early 2000s, manual workers aged 35 lived on average 6 years less than management-level workers and experienced on average 10 more years of life with functional disorders, a phenomenon referred to as the ‘double penalty’ for manual workers (Cambois et al., 2008). Such social inequalities in years of disability exist in both men and women, among the oldest (65 and over) and younger groups (50–65). They are found in all countries, regardless of how social status is defined (occupation, level of education, etc.), as shown by a recent literature review (Cambois et al., 2020). Social differences in health are linked to living and working in harmful conditions, more common among the less qualified, less educated, and less financially well-off groups. These conditions expose people to higher risks of illness, particularly disabling and/or life-threatening diseases such as cardiovascular and musculoskeletal disorders (Palazzo et al., 2019). Precarious living conditions also limit options for adapting an individual’s environment or lifestyle to functional impairments, thereby accelerating the risk that functional difficulties will translate into activity restrictions and lead to a loss of functional independence.

The Global Burden of Disease

86This programme aims to incorporate the full spectrum of diseases into summary measures to describe the health of different regions of the world (Murray et al., 2000). [21] The prevalence, incidence, and mortality risks associated with a number of diseases are combined to assess the health status of a population. In the context of population ageing, this programme aims to compare different age groups, countries, country regions, or population groups. One of the summary indicators is the disability-adjusted life year (DALY), or life expectancy minus the years of disability associated with the various diseases prevalent in the populations. To estimate this, a weighting system is used for each disease, based on the opinion of experts who quantify the ‘loss of good health’ each disease is likely to cause. By employing an approach based on prevalent diseases rather than disabilities directly, these indicators provide insight into the ‘exposure and health care’ dimension. They supplement disability-free life expectancies, which shed light on the long-term care side (by analysing the nature and severity of disabilities). Calculating DALYs requires different data, such as the prevalence of diseases and their mortality risk. Data that are not available for some countries are modelled on a regional basis. The measurement of the ‘quality’ of years lived, based on expert opinion on a disease-by-disease basis, may change over time or from country to country. This programme reveals the difficulty of providing internationally comparable measures. It nevertheless benefits from a high capacity to produce estimates based on large data sets and powerful modelling, but this should not obscure the role of the modelling and assumptions used to arrive at the indicators (Mathers, 2020).

Frailty

87In parallel with conceptual advances in measuring the health of older population groups, particularly with the notion of disability, the concept of frailty has gained importance in the field of gerontology: it is based on the observation of situations of multimorbidity, functional decline, loss of physiological reserves, and reduced chances of protecting oneself or recovering from an illness (Rockwood et al., 1994). The frailty of individuals is measured based on a combination of many medical, functional, and psychosocial characteristics. From a prevention perspective, the aim is to detect situations before functional deterioration and loss of functional independence occur. When these dimensions are measured using population-based surveys, it is possible to produce a frailty indicator for a population. However, the large number of characteristics required in the approach proposed by Rockwood et al. (1994) makes it difficult to measure through surveys. Another approach, used by Fried et al. (2001), is based on the functional motor dimension, which requires less information that can be more easily collected in general population surveys. These variables include grip strength or walking speed, weight loss, fatigue, or sedentary lifestyle. Along with the measurement of disabilities, which aims to break down the various stages of functional decline, the notion of frailty summarizes an ‘at-risk’ situation by incorporating all dimensions of the functional decline process.

Ageing and biodemography in health studies

88Biodemography is an area of study that has emerged strongly over the last 20 years with the rise of chronic diseases and frail states of health (Robine, 2011). It was introduced in the eighth edition of the Handbook of Aging and the Social Sciences (Crimmins and Vasunilashorn, 2016). In the field of health, biodemography uses demography, epidemiology, and biology, adding the biological component to the disability development model described above. The human body has resources and reserves that determine whether a disease occurs, whether it can be recovered from, and the risk of experiencing functional decline. These resources and reserves are measured by markers (or biomarkers) identified via biological samples (blood samples, urine samples, etc.); these include, for example, indicators of inflammatory processes closely linked to cardiovascular diseases. These markers vary depending on living conditions throughout the life course and the accumulation of harmful situations endured. Biomarkers enable us to analyse the links between age, gender, and social situation, and health and mortality risks. The data available in this area include the Health and Retirement Study and the sister studies in Europe (SHARE): in the latter, a biological sample was collected during Wave 6 of the survey in 12 countries, providing a number of biomarkers among the individual variables.

How much support is provided by families?

98As previously mentioned, particular attention is paid to family caregivers looking after their dependent spouse or parents. In the early 1980s, in the introduction to a special issue of the French journal Gérontologie et société devoted to ‘family and generations’, Paul Paillat emphasized the lack of data on the ‘hidden families’ caring for frail or disabled people. For a long time, support for loss of functional independence remained confined to the domestic sphere, excluded from any recognition as socially useful (Maisonnasse, 2016). Today, the situation is quite different. For nearly 20 years, various national and international surveys conducted in the general population have highlighted the major role of family support for older people who can no longer perform certain ADLs alone. [26] As well as extensive research, family caregivers now form the subject of specific public policies in many countries.

99Research on caregivers has shown that in all OECD countries, regardless of the specific social protection models and family support norms of each country, family and friends (family in particular) are the main source of assistance for older people unable to live independently (OECD, 2019). This ‘informal’ caregiving is, in most countries, more frequent, more time-intensive, and more diversified than the ‘formal’ caregiving provided by professionals. In 2015 in France, 3 million people aged 60 and over living at home reported they were regularly assisted with ADLs because of their age or a health problem (Brunel et al., 2019); 48% of them received help from family and friends only, 34% from family and friends and from professionals, and 19% from professionals only. [27] The (median) duration of informal caregiving is 5 hours per week compared to 55 minutes of professional care. The latter is generally limited to housekeeping, bathing, dressing, and meal preparation, while informal caregiving also includes shopping, medical appointments, travel, administrative tasks, coordinating professional interventions and, very often, supervising medical treatment as instructed by medical professionals.

100Informal caregiving is a major pillar of the social protection system for dependent older people in all European countries, although it may take different forms (Colombo et al., 2011). Fontaine (2017) shows that almost 80% of dependent older people receive informal care, across Germany, Spain, France, Italy, the Netherlands, and Sweden. Informal caregiving increases with the severity of disabilities in a similar way in all six countries. Modes of care, however, vary significantly. In Southern Europe (Spain and Italy), older people unable to live independently are more likely to cohabit with one of their children than elsewhere (27% in Italy and 35% in Spain compared to around 10% in France and Germany and less than 5% in the Netherlands and Sweden in 2004). When care is organized remotely, it mostly occurs at least once a week and very often daily. In Northern Europe, more dependent older people live alone (63% in Sweden and 58% in the Netherlands compared to 46% in Germany, 40% in France, 37% in Italy, and 32% in Spain). Remote support is more frequent than in Southern Europe, but in most cases it is occasional. The countries of ‘continental’ Europe (Germany and France) are in an intermediate situation.

101The significance of family support has led to the development of social science research aimed at better understanding helping behaviours, their determinants, and their implications. How is support organized within the family? Why do some individuals in the same family become involved in care and others not? How is assistance shaped by family configurations, the characteristics of potential caregivers, the needs of the parent, or the available professional assistance? What are the consequences for caregivers? In the context of an ageing population, a better understanding of family caregiving patterns is crucial to our comprehension of two issues: the role that social protection systems attribute to family support in care provision and the possible mismatch between the supply of and demand for informal care.

The role of family support within the social protection system

102The first issue concerns the role that social protection systems attribute to family support for older adults provided alongside state assistance. Because family support generally takes the form of unpaid services, there is a risk that informal caregiving is equated with costless production and that the burden of dealing with activity restrictions in advanced age is placed, to an excessive extent, on families (Fontaine, 2017). Research focuses on the consequences of informal caregiving by looking at two aspects of caregivers’ quality of life: their health status and their participation in the labour market. Economic and epidemiological studies have highlighted the negative effects of caregiving on the mental health of caregivers, such as stress, anxiety, depression, etc. (Coe and Van Houtven, 2009; Do et al., 2015). According to the literature, [28] there is a substitution effect between working hours and informal caregiving, but it mainly affects relatives closely involved in caregiving—and mostly women. Most often, professional life is preserved at the cost of a reduction in family and social time (Bihan-Youinou and Martin, 2006).

103These results, as well as findings pointing to the social inequalities created by intra-family transfers, [29] have contributed to making family caregivers a target of social policies. There are two possible public policy approaches to reducing the burden on family and friends. The first would aim to support informal caregivers by reducing the indirect costs they bear, such as leave, compensation, specific medical monitoring, and development of respite services. [30] This line of public action is not intended to reduce family support but to ‘help carers to provide care’ by reducing the cost of caregiving on their professional life, health, and overall quality of life. The second approach would aim to reduce the involvement of caregivers by significantly increasing the accessibility of professional care.

Mismatch of supply and demand for informal care

104The second issue associated with demographic ageing is the potential for increasing divergence between demand for and supply of informal support. On the one hand, population ageing results in an increase in older people requiring daily assistance. [31] On the other hand, the supply of informal assistance may decrease due to rising divorce and separation rates, a fall in the number of children, the increasing participation of women in the labour market, and the growing geographical distances between children and parents (Colombo et al., 2011).

105The assumption that informal caregiving is decreasing should certainly be put into perspective. If we look at family environment, the projections made in Europe tend to contradict conventional wisdom. In the medium term, in Europe, care-dependent older people will have a larger family network than they do today (Gaymu, 2008). Froment et al. (2013) for France and Carrière (2008) for Canada reach similar conclusions. The reduced risk of widowhood resulting from the increase in life expectancy would more than compensate for the increasing frequency of divorces and separations, and would mitigate the impact of fewer children per family on the size of the family circle.

106However, having a larger family only offers potential resources for informal caregiving. As mentioned above, various socio-economic developments may reduce the actual involvement of potential caregivers, such as the increasing female employment rate or increasing geographical distances between children and their parents. There is relatively little prospective research aimed at anticipating trends in actual informal caregiving. While the research that does exist points to a decrease in family support, socio-economic changes seem to have only a very limited effect on the ability of family members to be closely involved in caregiving (Pickard, 2008; Janus and Doty, 2018). Demographic and socio-economic changes in potential caregivers may have only a modest effect on the overall amount of help actually provided by the family group if, within the same family group, the lower involvement of some members is compensated by the higher involvement of others.

The decline in fertility

109In France, the mothers of early generations of baby boomers (born in the 1930s) had an average of 2.6 children, and nearly a quarter of them had four or more. Thirty years later, their daughters (born in the 1960s) had an average of 2.0 children, and 9% of them had four or more. This decline in the number of large families has been accompanied over the generations by a convergence of family size around two children, with the share of women with one or three children remaining relatively stable at around 20% (Brée, 2017).

110The decline in the number of children per family is not necessarily associated with a reduction in informal caregiving since the latter is impacted not only by the number of children but also by their relationships and how they organize themselves to assist their parents. The problem of future informal care arises particularly for individuals without children, whose proportion is increasing. Childlessness increases the likelihood of institutionalization (Freedman, 1996). It also seems to be associated with poorer well-being at older ages (Dykstra, 2009). Over the period 1998–2010, 6.6% of Americans aged 55 and over had no spouse or biological children, and this proportion is increasing in the most recent cohorts (Margolis and Verdery, 2017).

111More specifically, in France the proportion of women without children fell slightly between the 1930s and 1940s before rising again to 14% in the 1960s (Brée, 2017). This upward trend from the 1940–1945 generations onwards can be observed in many neighbouring countries, although the rate of growth and the levels reached vary between generations. For cohorts born in the late 1960s, the proportion of women without children is similar in France and Sweden (around 14%) and slightly lower in Norway, Denmark, and Portugal (around 12%). The ratio is 1 in 5 women in Italy, Spain (16%), Switzerland, and Austria, and it is almost 1 in 4 in the former West Germany (Sobotka, 2017). In the United States, the proportion of childless women is 10% for the 1943 cohort and roughly 16% for the cohorts born in the late 1950s (Frejka, 2017). At the global level, in an analysis of 34 countries comprising 70% of the world’s over-50 population, Verdery et al. (2019) also highlight the wide variability in the share of adults without a biological child or partner, from 10% in Canada, Ireland, the Netherlands, and Switzerland to less than 2% in the Republic of Korea.

112Finally, while the generations now entering retirement may have fewer traditional sources of support (spouses or biological children), they have more surviving ex-spouses, stepchildren, and siblings (Agree, 2018; Seltzer, 2019). The question of how support is organized in these new family configurations remains largely open and poses new research challenges.

More coexisting generations within families

113The fall in numbers of siblings is accompanied by an increase in the number of coexisting generations within families because of longer life expectancy (Bengtson, 2001). More individuals approaching retirement age now have children and grandchildren as well as parents who are still living. In France, the coexistence of four generations increased from 26% for women aged 50 born in 1920 to 44% for those born 30 years later in 1950 (Pennec, 1996). In the United States, over the period 1998–2010, 40% of adults in their 50s had a four-generation family (Margolis and Wright, 2017). Also in the United States, the share of children aged 10 with four living biological grandparents rose from 6% in 1900 to 41% in 2000 and was expected to reach 48% by 2020 (Uhlenberg, 2005). This vertical extension of families leads, among other things, to the potential diversification of support systems within them. Interactions and support can be multidirectional: between grandparents and grandchildren, and between parents and adult children, and not solely between adult children and dependent parents or parents and young children (Margolis and Verdery, 2017).

Higher prevalence of late-life couples

114If behaviours remain unchanged, the decline in mortality and the closing of the gender gap in life expectancy increase the likelihood of living with a partner, at all ages. While in France in 1980, 2.1% of married men aged 75 became widowers in that year, 20 years later the risk was only 1.2%. For women, at the same age, this risk fell from 7.0% to 5.1% (Delbès and Gaymu, 2003). The phenomenon is accentuated by a greater fall in mortality among married individuals than others (Valkonen et al., 2004; Kravdal et al., 2018). However, changes in certain marital behaviours, such as the rise in divorce, could work in the opposite direction. This could be the case for men aged 75–84, among whom the proportion of married men is predicted to fall slightly by 2030 (Kalogirou and Murphy, 2006; Gaymu et al., 2008a). On the other hand, for older men aged 85 and over and for all women aged 75 and over, the significant decline in widowhood is likely to more than offset this behaviour change and would therefore result in a significant increase in the proportion of married individuals. Married individuals benefit from the presence of their spouse in situations of dependency. These developments might therefore reduce the strong gender imbalance in the care of dependent spouses. In 2000, for example, in people aged 85 and over, the age at which the gender differences in marital status are greatest, men were almost 5 times more likely than women to be living with a partner (50% vs. 9%). By 2030, this ratio is expected to decrease to 2.5 (Gaymu et al., 2008a).

A significant increase in divorce rates

115One of the key features concerning partnerships over the last 2 decades is the sharp increase in divorce at late ages, prompting demographers to coin the term grey divorce revolution (Brown and Lin, 2012). Brown and Lin (2012) showed that in the United States, more than a quarter of individuals who had divorced in 2010 were over 50 compared to only 10% in 1990. The explanation lies not only in large generations reaching these ages but also in the increased risk of divorce. The divorce rate has doubled in the last 20 years, from 5 to 10 divorces per 1,000 married individuals aged 50 or over. In France, while the risk of divorce remains lower after age 50 than at younger ages, it is in this age group that it has increased the most over recent years. During the 2000s, the relative increase in risk of divorce was correlated with advancing age, up to the age of 70 (Prioux and Barbieri, 2012). Risk of divorce is 2.8 per 1,000 and 2.0 per 1,000 for men and women aged 60 and over. This appears to be relatively low but has increased by a factor of 1.5 at ages 50–54 and by a factor of 1.9 after age 60 since 2000 (Solaz, 2021).

116This strong increase in divorce is prompting new research on the determinants and implications of marital behaviours at advanced ages (Brown et al., 2016, 2019). Divorces at older ages appear in many respects quite different from those at younger ages: fewer dependent children, greater difficulty in returning to the labour market, less repartnering, and more often associated with poor health, concomitant transition to retirement, etc.

Rise in partnering and repartnering

117Divorce among older people has increased, as has partnering and repartnering. Recent literature has focused on the type of union (cohabitation or marriage), on the determinants of such (Vespa, 2012; Brown et al., 2019) in a context where cohabitation is now more widespread (De Jong Gierveld, 2004), and on whether these new relationships are a way of mitigating the consequences, particularly the financial consequences, of divorce, especially for women. [36]

118Here again, partnering behaviour at these ages is different from that at younger ages (Bonnet et al., 2019a) and markedly different in women and men. In particular, due to longer female life expectancy, the ratio of men to women in the marriage market becomes more favourable to men over time. This may be one of the reasons why they are more likely to repartner. Women are also more reluctant to repartner due to the persistent asymmetry in family and domestic activities; they are much more involved in domestic or caregiving tasks (Brown et al., 2018; Lewin 2018). Those who repartner after age 50 are more likely not to cohabit than at intermediate ages, primarily to maintain their independence (Régnier-Loilier et al., 2009; Liefbroer et al., 2015). [37] In France, around 6% of men and 5% of women in a relationship aged 55–64 are in non-cohabiting relationships (Régnier-Loilier, 2019). This proportion is 4 times higher than in Italy for women aged 60 and over (Régnier-Loilier and Vignoli, 2018).

Older people in institutions

120Making international comparisons of the proportions of older people living in institutions is not straightforward. The definition of care institutions for older people can vary between countries, primarily due to the varying diversity of housing types and the scope of residents considered. OECD data (OECD Health Statistics, 2020), for instance, discloses the social model of long-term care for older adults rather than other types of collective residences with services. For France, the proportion of over-65s living in institutions is estimated at 4% by the OECD, which only takes dependent older people into account. [38] If we expand the population to include all institutions for old people, we conclude that 5.6% of the over-65s live in an institution. In the United States and Spain, depending on the scope of the population, the proportion of older people in institutions may double. [39] According to the OECD definition, the share of over-65s in institutions is lower in countries such as Spain and Italy (around 2%) and slightly higher in the United States (2.5%). The share in Sweden, Denmark, and Germany is close to that of France (around 4%), as shown in Figure 17. A notable phenomenon is the downward trend in this share in northern countries (Denmark, the Netherlands, and Sweden), which have long been ‘deinstitutionalizing’, a phenomenon introduced in Sweden in the post-war period.

Figure 17

Trends in the share (%) of over-65s in institutions in 8 high-longevity countries, 2000–2016

Trends in the share (%) of over-65s in institutions in 8 high-longevity countries, 2000–2016

Note: The OECD reports a number of series breaks in the definition of institutionalized persons. This justifies taking only isolated points into account for some countries.

121While the proportion of over-65s in institutions is relatively low, it increases rapidly with age. In France, according to the census, in 2013, 3% of 70- to 74-year-olds, 12% of 80- to 89-year-olds, and 30% of 90- to 99-year-olds lived in institutions for old people, and 1 in 2 centenarians lived in an institution (Blanpain and Buisson, 2016a). While there is a great deal of research on the informal caregiving provided by families in later life, most of it focuses on ordinary households. Very few studies address family connections and informal help in the context of institutionalization, mainly due to the lack of data (Agree, 2018).

Family configurations of older persons in ordinary households

122The family configurations of over-65s vary greatly by gender, country (Figure 18), and age (Figure 19). In the EU-28, in 2017, 4 in 10 women aged 65 or over live alone, and the same proportion live with a partner. The other women live with children or in complex households. These proportions are very different for men, who are twice as likely to live alone (22.5%), and nearly 6 in 10 live with a partner. Differences between countries are significant. The share of women aged 65 and over living alone is lowest in Spain (29.9%) and highest in Denmark (54.7%), followed closely by Lithuania and Estonia.

Figure 18

Family configurations of men and women aged 65 and over in EU countries, 2017

Family configurations of men and women aged 65 and over in EU countries, 2017

123All developed countries experienced an increase in the proportion of older persons living either alone or with a partner, during the 20th century. In the United States, for example, in the mid-19th century, nearly 70% of over-65s coresided with their adult children; by the end of the 20th century, less than 15% did so (Ruggles, 2007). In France, in the 23 municipalities (communes) studied by Bourdieu et al. (2013), the coresidence of over-60s (in couples or alone) with at least one person under 60 (whether a family member or not) fell from 68% in 1846 to 53% in 1931. [40] In France, the proportion of over-65s living alone has increased by almost 30% since 1960, and the share of over-65s in couple-only households has risen by more than 60% (Figure 19). These changes are even more marked among the over-80s. There has been a sharp decline in coresidence with children.

Figure 19

Trends in the living arrangements of over-65s and over-80s in France, 1962–2011

Trends in the living arrangements of over-65s and over-80s in France, 1962–2011

124Various determinants explain the changes in these cohabitation patterns: changes in marital behaviour, the longer survival of couples (associated with the closing of gender gaps in life expectancy), lower number of children, and improvements in the health of older adults (McGarry and Schoeni, 2000). But economic factors also play a part. The greater financial independence of older people, particularly because of the development of pension systems, is also a factor in the decline in intergenerational coresidence (Costa, 1999; McGarry and Schoeni, 2000). Ruggles (2007) attributes part of this decline to the decline of farming families, an environment in which intergenerational coresidence was widespread.

Poverty rate among retirees

126The 20th century was marked by the development and progress of pension systems, both in coverage and pension levels. The poverty rate among older people fell sharply in the 1960s and 1970s in many countries, as seen in the United States and France (Figure 20). In the early 1970s, poverty affected more than a third of retirees in France compared to just over 20% of the general population. By 2016, it had dropped to 8% compared to nearly 14% of the general population. Retired people are now in the same situation as employed people, 6 points below the national average. This situation has been broadly stable since 2000. The decline in poverty among the over-65s is of roughly the same magnitude in the United States, where the rate fell from 35% in 1959 to 10% in 2014.

Figure 20

Changes in poverty rate

Changes in poverty rate

Note: In the United States, the poverty rate is defined in absolute terms based on a minimum consumption basket using data from the Current Population Survey Annual Social and Economic Supplements produced by the Census Bureau. In France, it is defined by INSEE as the proportion of ordinary households (before 1996) or of individuals living in ordinary households (after 1996), whose standard of living is below the poverty line (equal to 60% of the median standard of living of the entire French population) based on the Tax and Social Income Surveys from 1970 to 2016. As such, the poverty levels of the two countries are not directly comparable.

127The relative position of retirees compared to the national average is not favourable everywhere, however, and situations vary widely across OECD countries (Figure 21). In half of the 32 countries, people aged 66 and over are slightly less poor than the national average; in around 10 their situation is comparable; and in six countries they are significantly worse off (Australia, Estonia, Republic of Korea, Switzerland, Lithuania, and Latvia).

Figure 21

Poverty rates for people aged 66 and over and the total population in 32 OECD countries, 2016

Poverty rates for people aged 66 and over and the total population in 32 OECD countries, 2016

Note: The poverty rate threshold here is 50% of the median standard of living.

128While poverty among retirees has generally declined, some groups remain economically disadvantaged. Smeeding (2003), using data from the Luxembourg Income Study for European countries and the United States, notes that poverty is generally higher at older ages and higher among women and individuals living alone than among those living with a partner. Marital status is also associated with different levels of poverty. From the European Community Household Panel (1994–2001), de Santis et al. (2008) observe that among women aged 65 and over, those who have divorced and separated and, to a lesser extent, those who have remained single have a higher probability of being poor than widows. [41] The latter often benefit from a survivor’s pension which limits the reduction in their resources upon the death of their spouse (Thompson and Carasso, 2002).

129Several factors are behind this generally favourable trend in the situation of retirees. The increase in the share of retirees in a union may be one of them, but there is no doubt that the development of pension systems has played an important role. Public transfers (including earnings-related pensions) represent the most important component of the income of over-65s in many countries (OECD, 2019b), although the proportion may vary between countries, depending mainly on the organization of social protection systems. For the over-65s in Germany, Italy, and France, it represents more than 70%, while it only constitutes 40% of their income in the United States. Other resources come from work, income from assets, or company pensions. However, this rather favourable situation for retirees could be under threat. Given pressures on public spending, particularly on pensions, it is likely that developments will be less favourable. The European Commission’s projections indicate a likely downward trend in the ratio of average pension to average wage in many European countries, primarily because of the reforms implemented to ensure the financial balance of pension schemes (European Commission, 2018). As pensions represent a significant proportion of older people’s resources, this would normally result in a decline in the relative standard of living of retirees compared to the rest of the population. [42]

Work longer?

135Since the early 1970s, the labour force participation rates of people aged 55–64 and the over-65s have followed a U-shape, falling until the mid-1990s and then showing an upturn (Figure 24). These trends are similar in many countries (Blundell et al., 2016).

136This U-shaped effect is more pronounced in men than women (Figure 24), with the decline observed for men offset by a significant generation effect for women (Goldin and Katz, 2018). Women have been participating massively in the labour market since the 1970s thanks to rising educational attainment, changing gender norms, and the transformation of careers.

137The literature has identified determinants of these changes in labour market participation at older ages. First, the decline in the participation rate observed until the early 2000s was linked to the development of pension systems and in particular to the increase in the ‘generosity’ of these systems and the absence of incentives to raise the retirement age (Gruber and Wise, 1999, 2004). The fall in the participation rate is also linked to mechanisms that enable people to leave the labour market before they actually retire (e.g. unemployment, early retirement, or disability). Next, the determinants of the reversal in participation rate of older persons have been analysed, but quantification of the effects remains complex (Blanchet et al., 2019). Besides the effects of institutional reforms [43] on pension systems and early retirement schemes, rises in levels of education and changing retirement age norms also have an impact (Seibold, 2019), as do the growing labour market participation of women, health improvements at later ages, companies’ demand for labour, and age discrimination. Over the past 30 years, understanding retirement behaviour has become a major field of research (Coile, 2015). The objective has been to produce an ex ante evaluation of the implications and effectiveness of any pension system rules change on these behaviours.

Figure 24

Labour force participation rate (%) in the over-55s in 12 OECD countries, 1965–2019

Labour force participation rate (%) in the over-55s in 12 OECD countries, 1965–2019

Note: The labour force participation rate is calculated as the labour force divided by the total working-age population.

138Shifting the retirement age is one of the preferred paths of reform to balance the funding systems in almost all countries. With longer retirement periods ahead, it may seem justified to divide life expectancy gains between longer working lives and longer retirement periods. However, this raises questions about the individual’s ability to continue working, particularly from a health perspective. A recent international comparison [44] evaluated health-related work capacity after age 55 (Coile et al., 2017): in general, health status does not hinder the extension of working life in older adults. This result becomes less conclusive, however, when the subpopulations are disaggregated by level of education or social category (Cazenave-Lacroutz and Godet, 2016). As such, policies to raise the retirement age should take this heterogeneity into account (Blanchet et al., 2017b).

139The issue of the inequalities associated with an extension of working life echoes the question raised in recent years by the observation that inequalities in life expectancy associated with socio-economic status remain, or have even increased (Blanpain, 2016; Chetty et al., 2016), possibly resulting in unequal periods of time spent in retirement. Social groups with higher pension levels benefit, on average, from longer periods of retirement, which may reduce, to varying degrees, the redistributive nature of the pension system (Liebman, 2002; US National Academy of Sciences, 2015; Ponthière and Pestieau, 2016; Auerbach et al., 2017).

140This finding shows that the entire life course has an impact on the health and retirement situations of the older population. Understanding these situations provides a complete picture of the public policy systems associated with the various stages of life.

Generational accounting

143Generational accounting, first implemented in the United States 30 years ago by Auerbach et al. (1991) and Kotlikoff (1992), calculates net transfers to or from the state for each generation: the difference between all benefits received (retirement, health, unemployment, family, etc.) and taxes and contributions paid. Assuming that social and tax legislation remains unchanged for all generations currently living, we can calculate what future generations will have to pay to ensure the financial sustainability of public spending in the long term. [45] Generational accounting generally leads to the conclusion that future generations will have a strong disadvantage, with a particularly high tax burden. While this approach has been applied in many countries (Raffelhuschen, 1999), it has been met with a number of criticisms [46] (Masson, 2002; d’Albis and Moosa, 2015). It is less widely used today. It this context, a second methodology was developed: national transfer accounts.

National transfer accounts: a measure of economic flows between age groups

144Initiated in the United States, the purpose of the national transfer accounts (NTA) is to quantify flows of production, consumption, savings, and resource sharing by age. The method is based on an accounting equation that equates private and public resources and consumption. An individual’s resources (income from work and capital, and public and private transfers received) must be equal to the uses made of them (consumption, savings, or public and private transfers paid). The various elements of this equation are calculated for each age group, with total flows being consistent with the national accounts aggregates. We can also identify the channels through which flows pass: family, state, or individual. Comparing age-group profiles allows us to identify periods/ ages of surplus (when labour income exceeds expenditures) and deficit (when expenditures exceed labour income) and their nature. The funding of consumption that is not based on individual income can be calculated. The global and harmonized framework (forming part of the national accounts) of NTAs allows comparisons over time and across countries (Mason and Lee, 2011). This methodology has become widely used: 60 countries, including France (d’Albis et al., 2017), currently have NTAs. [47] Comparison between countries shows how periods of youth, working age, and retirement are organized in different countries, in connection with education, labour market, and retirement age policies and the various social protection systems. As Figure 25 shows, while consumption at older ages is mostly financed by public transfers in France and Germany, this is far less the case in the United States, where individual financing plays a more important role (d’Albis et al., 2019). In Germany, one-third of consumption is financed by public transfers between the ages of 63 and 64, but in France, where people leave the labour market earlier, this level of state funding occurs slightly sooner (from age 62).

Figure 25

Sources of funding for consumption by age in France, Germany, and United States

Sources of funding for consumption by age in France, Germany, and United States

Interpretation: The consumption of 65-year-olds in France and Germany is mainly financed by the state (public transfers). In the United States, at these same ages, individuals finance nearly the main part of their consumption from their individual income, with the government contributing only a limited fraction.