Long Retirement
Retirement is the part of Keynes’s leisure that the rich countries actually took. For thirty years people left work earlier and lived longer; since about 2000 the exit age has been rising again, and in Portugal it is now tied by law to how long people live. This essay asks how long the last part of life lasts, how much of it is spent in good health, whether the pension that pays for it is enough, and who carries the adjustment.
For most of history old age was not a stage of life with its own income. People worked until they could not, and then depended on their children, their savings or charity. Retirement as a mass condition, a period of years that most people reach and live through on a pension, is little more than a century old. Germany’s old-age insurance law of 1889, pushed through by Bismarck, paid a pension from the age of 70, when few workers lived that long; the age was lowered to 65 in 1916.[1] Britain’s Old-Age Pensions Act of 1908 paid a means-tested five shillings a week from 70. The United States’ Social Security Act of 1935 settled on 65, which became the conventional age across much of the rich world.[2] Portugal built its own system later and in stages, from the corporatist previdência funds of the 1930s to the universal social security of the democracy, which extended pensions to farm workers and to people who had never been able to contribute.
The first essay in this series, The Fifteen-Hour Week, found that the rich countries took only about a fifth of a century’s productivity growth as shorter working weeks. Its retirement section added the qualification that matters here: counted over a whole career rather than a single year, much of Europe took close to a third or more of its gains as time, and a large part of that time came at the end of life. The second, The Two-Thirds Country, found that Portugal, with less productivity to share, had taken less of it as shorter weeks than almost anyone. This essay follows the time that was taken at the end: how long it lasts, how healthy it is, what it costs and what it pays.
Down for thirty years, up for twenty-five
The OECD has estimated, for most of its member countries since 1970, the average age at which people actually leave the labour market, which is often quite different from the official pension age.[3] Figure 1 shows it for men. Every line has the same shape, a V. In 1970 men left work between 65 and 69; Portuguese men, many of them farmers and self-employed without a pension worth stopping for, at 69. Over the next thirty years the average fell by three to eight years, to between 58 and 63 around 2000: 58.4 in France, 59.5 in the Netherlands, 60.7 in Portugal. Then it turned. By 2024 Portuguese men were leaving work at 66.5, Dutch men at 65, American men at 67.3. Women followed the same V, in Portugal from 68.9 in 1970 to 58.0 in 2000 and back to 64.5.
The fall had several causes. Pensions matured and became worth retiring on. Governments in the 1970s and 1980s used early retirement, often through disability and unemployment schemes, to take older workers out of labour markets that had too few jobs, on the mistaken theory that it would make room for the young. And the systems themselves rewarded leaving early: a cross-country study led by Jonathan Gruber and David Wise found that the implicit tax on working an extra year after the early-retirement age explained much of the difference in exit ages between countries.[4] The rise since 2000 is mostly the undoing of those incentives: early-retirement routes closed, penalties added for leaving early, and official ages raised. Meanwhile life expectancy at 65 kept rising, by one to two years every decade. The combination, as the first essay showed, took the number of years a Portuguese man could expect to live after leaving work from under ten in 1970 to almost nineteen in 2000, and it has stayed close to eighteen since. For Portuguese women it went from 12 to 23.
It is already happening
The question in the title of so many policy reports, whether people will have to work longer, has already been answered by the data. Figure 2 shows the share of people aged 55 to 64 in work. In 2000 it was under 30 per cent in France and Italy and under 40 in Germany and Spain. By 2025 it was over 60 per cent in all four and 75 per cent in Germany. Portugal, which started higher, at about 51 per cent, fell back during the crisis to 47 per cent and then rose to almost 70.[5]
Portugal is unusual at the older end too. A quarter of Portuguese aged 65 to 69 still work, a share that was already 27 per cent in 2000, when in France and Spain it was two to four per cent. In most of northern Europe the share has climbed to around 30 per cent from very low levels. In Portugal it has been high for a long time, and not always by choice: much of it is small farmers, self-employed people and those whose pensions are too small to live on. Eurostat’s estimate of the expected length of a working life puts the average Portuguese at 39.7 years in 2025, two years more than the EU average, and more than in Spain, France or Italy.
How many of those years are healthy
Adding years of work at the end makes sense if the extra years of life are healthy ones, and the case for tying the pension age to life expectancy assumes that they are. Eurostat estimates, from surveys that ask people whether health problems limit their usual activities, how many of the years a 65-year-old can expect to live will be free of such limitations. Figure 3 shows the answer for 2024.[6]
Total life expectancy at 65 is remarkably similar across these countries, between about 20 and 22 years. The healthy part is not. A Swede of 65 can expect 14.2 healthy years, an Italian 13.1, a Spaniard 12.5. A Portuguese can expect 9.8, slightly below the EU average of 10.3, and fewer than half of his or her remaining years. Asked directly, 28 per cent of Portuguese over 65 describe their health as bad or very bad, the highest share in this group of countries and ten points above the EU average. The Portuguese figures are improving: the share reporting bad health was 49 per cent in 2010, and healthy years at 65 rose from about six in 2015 to 9.8 in 2024, though that series has breaks in it, reading 9.5 in 2013 before a change in the survey and jumping by more than a year in 2024 alone. But the starting point was low, and it matters for the rule Portugal has chosen.
That rule raises the pension age by two-thirds of every gain in life expectancy at 65.[7] It uses total life expectancy, not healthy life expectancy, and the average, not the life expectancy of the people who will actually be asked to work longer. The averages hide large differences. In the United States, the richest one per cent of men live about 15 years longer than the poorest one per cent.[8] Europe has no estimate by income built the same way; Eurostat’s figures by education show a gap in life expectancy at 65 of about a year in Portugal and Sweden. But lifespan is the wrong yardstick here. What differs between a builder and a professor is less how long they will live than how healthy they are at 65 and how physically demanding the last working years are. If both retire at 67 they are not being asked for the same thing.
The numbers can be put to the rule directly. Suppose Portugal had tied the pension age to healthy life expectancy at 65 rather than to total life expectancy, by the same two-thirds formula. On Eurostat’s series, life expectancy at 65 rose by 1.2 years between 2013 and 2024, which the formula turns into a pension age of 66 years and 10 months, close to the 66 years and 11 months the law has set for 2027. Healthy life expectancy rose by 0.3 years over the same period, which would give 66 years and 2 months. Start the clock a year later, on the other side of the 2014 change in the survey, and the healthy series rises by 3.6 years, for a pension age of 68 and a half.[6] A measure that moves the answer by more than two years depending on the year you start from cannot carry an indexation rule. Between 2014 and 2024 the share of Portuguese aged 65 to 74 who reported a long-standing limitation fell from 60 to 42 per cent, far too fast to be a change in health alone. That is the first result: healthy life years are a survey measure, and no country indexes its pension age to one, for good reason.
What the measure can do is compare groups in the same year, where the survey’s quirks largely cancel. Figure 4 uses the Portuguese life tables and the same survey question to estimate, for 2024, how many healthy years a person of 66 years and 9 months, the pension age for 2026, can expect, and the age at which each group would have the same healthy years ahead as the average person has at that age, 8.9.[16] Men, who at 65 can expect to live three years less than women, can expect 9.4 healthy years after the pension age; women 8.4. A rule that equalised healthy years would let women retire at 65 years and 10 months and keep men at work until 67 years and 10 months, the reverse of what a lifespan rule implies. Education matters more than sex. Life expectancy at 65 barely varies with schooling in Portugal, about a year between the least and the most educated, but healthy years vary a great deal: 8.3 after the pension age for people with basic education or less, 11.3 for those with secondary education and 12.2 for those with a degree. Equal healthy years would mean a pension age of about 65 and a half for the first group and 72 for the last. By income, the poorest fifth has 8.3 healthy years ahead at the pension age and the richest fifth 10.7, four and a half years apart in pension age. Combine the two: a woman with basic education can expect 7.9 healthy years after the pension age, a man with a degree 11.5. To reach retirement with the same healthy years ahead, she would leave at 64 years and 9 months and he at 70 years and 9 months. The builder and the professor are six years apart.
These are estimates, and rough ones. They rest on one survey question, on small samples of older Portuguese with secondary or tertiary education, and on the assumption that people with basic education and people with degrees share a life table, which Eurostat’s figures by education say is nearly true in Portugal. A more careful calculation would move the numbers by months, not years. The point stands. A pension age indexed to the average lifespan asks the least educated and the poorest to spend a larger share of a smaller stock of healthy years at work, and asks the most educated for years they will mostly have anyway. Nobody will write a pension age that varies by education, and no country trusts its healthy-years measure enough to index to it. What the arithmetic supports is the older and blunter idea, that a long contribution record, which usually belongs to people who started work young and with little schooling, is a fair proxy for the difference, and that the early-retirement exceptions Portugal already makes for very long careers are a correction, not a favour.
Working longer also collides with another demand on the same people. In 2019, 22 per cent of Portuguese women aged 45 to 64 cared for or helped someone old, ill or disabled at least once a week, without pay, against 12 per cent of men of the same age; across the EU the figures were 28 and 20 per cent, and in France and Denmark more than a third of women in that age group did so.[9] These are the years in which the pension age is rising, and the people being asked to stay at work until 67 are, disproportionately, the people looking after their parents. As the number of the very old grows, the care they need will be paid for either in taxes and wages for care workers or in the unpaid hours of their daughters.
A generous formula and poor pensioners
On paper, Portugal has one of the most generous pension systems in the OECD. The organisation’s standard calculation, for a man who starts work at 22, earns the average wage throughout and retires at the normal age, gives a Portuguese net pension of 93 per cent of his last net earnings, against 70 in France, 53 in Germany and 51 in the United States.[3] The reality of Portuguese pensioners looks different, because hardly any of them had that career. Most of today’s retired Portuguese worked in an economy of low wages, informal work and short contribution records, and many draw the minimum or social pension. Eurostat’s measure of what pensioners actually receive, the median pension of people aged 65 to 74 as a share of the median earnings of those aged 50 to 59, is 0.68 in Portugal, somewhat above the EU’s 0.60. The two measures are built differently, gross rather than net and across generations rather than for one career, so they are not expected to match; but the distance between them is a measure of how far today’s pensioners are from the formula.[10]
The clearest measure is poverty. Figure 5 compares the share of people below the poverty line, 60 per cent of the national median income, among those aged 65 and over and those aged 18 to 64. In Spain, France and Sweden older people are less likely to be poor than working-age adults. In Portugal it is the other way round: 17.8 per cent of over-65s were at risk of poverty in 2025, against 13.9 per cent of working-age adults, and among older women the figure was 19.4 per cent. The trend is not steady. Old-age poverty in Portugal fell from 27.6 per cent in 2005 to 15.1 per cent in 2014, as the relative incomes of the young collapsed in the crisis and minimum pensions were protected; it rose again to 21.1 per cent in 2024, largely because wages, and with them the poverty line, rose faster than pensions, and eased to 17.8 in 2025; survey incomes refer to the year before, so the 2025 figure describes 2024.[11]
Poverty lines are relative, so a pensioner whose income is fixed becomes “poorer” on this measure when everyone else gets a raise; the jump in 2024 is partly that. But the relative measure is the one that captures what the first essay called Keynes’s relative needs, and it tells a consistent story: in Portugal the time taken at the end of life is time on a small income, and the smallest incomes are those of the oldest women, who had the shortest paid careers.
The gap between women’s and men’s pensions is the bill for the unpaid shift, arriving late. Across the EU, women aged 65 to 79 received pensions 24 per cent lower on average than men’s in 2025, down from 34 per cent in 2015. In Portugal the gap was 23 per cent, down from 32; in the Netherlands, where women’s part-time work is the norm, 35; in Denmark, with its flat basic pension, 12.[12] The fall reflects cohorts of women with longer paid careers reaching retirement. Much of what remains reflects the years spent on unpaid work at home and part-time jobs, the work that the second part of the Keynes essay, Where the Hours Went, found to be missing from every official count of working time.
The adjustment is already written in
Portugal spent about 12.5 per cent of its GDP on pensions of all kinds in 2023, down from a peak of 15.6 per cent in 2013, when GDP had shrunk and the number of pensioners had not.[13] The official projections, compiled by the OECD, show public pension spending rising from about 12 per cent of GDP to a peak of about 15 per cent around 2045, as the baby-boom generations retire, and then falling back to under 12 per cent by 2060 (Figure 6). Spain’s spending is projected to rise to about 17 per cent and stay there; Italy’s to peak around 17 and fall; France’s and Germany’s to stay roughly flat.[3]
The Portuguese curve falls after 2045 not because the country gets younger but because the rules adjust. The reforms of 2002 and 2007 moved the calculation of new pensions to the whole career rather than the best years, tied their annual updating to growth and inflation rather than to wages, and introduced a sustainability factor, which since 2014 applies to those who retire before the normal age, an age that has itself followed life expectancy since the 2013 reform took effect.[7] The factor itself has grown as life expectancy has: it cut early pensions by 16.9 per cent in 2025 and by 17.6 per cent in 2026. Each of these mechanisms moves the cost of longer lives from the budget to pensioners, either as more years of work or as a smaller pension relative to wages. That is the choice most European countries have made. It keeps the systems solvent. It also means that the generation now in its thirties and forties, the one that emigrated in the crisis and pays today’s rents, will work longer than its parents for a pension that replaces less of its pay.
The old are also the owners. Three-quarters of Portuguese people living as a couple with at least one partner over 65 own their home outright, without a mortgage, against fewer than one in five of those living as couples with children, many of whom own with a mortgage; across the EU the figures are 72 and about 30 per cent.[14] The rise in house prices described in The Two-Thirds Country is, among other things, a transfer of paper wealth to older owners and a cost to younger buyers and renters. It complicates the generational picture in both directions: a pensioner below the poverty line may still live in a home worth more than a young family will save in decades, and a young worker paying today’s rent is also paying, through the pension system, for the retirement of the people who own the flat.
The next quarter-century
Several things are predictable. The Portuguese pension age will go on rising by about two months a year while life expectancy rises: 66 years and 11 months in 2027, and, on the OECD’s calculation, 68 for someone who entered the labour market at 22 in 2022. Other countries are going further, to 70 in Italy, the Netherlands and Sweden and 74 in Denmark. The number of people over 65 for every hundred aged 20 to 64 in Portugal will rise from about 43 today to about 69 by 2050 on Eurostat’s baseline projection,[15] one of the highest in Europe; the large immigration since 2022, discussed in the second essay, will lower that somewhat. The need for long-term care will grow faster than the population over 65, because the fastest-growing group is the over-80s.
Less predictable, and more important, is whether the extra working years will be healthy and whether they will be shared fairly. Three things are worth watching. Whether healthy life years at 65 in Portugal keep rising as fast as the survey series says they have since 2015; if they do, a later pension age will cost much less than it seems. Whether old-age poverty, especially among women, falls as the cohorts with longer and better-paid careers reach retirement, or rises as pensions fall behind wages. And whether any country moves from tying the pension age to average life expectancy towards something that accounts for how long, and how hard, people have actually worked. Several already allow earlier retirement after very long contribution records; Portugal does, with conditions. The broader principle, that the time dividend at the end of life should be shared according to health and work rather than averaged, is still mostly a debate.
The leisure we took
Keynes imagined his grandchildren working three hours a day. They did not. What they got instead, in most of Europe, was something he barely mentioned: a decade or two at the end of life with no work at all, paid for by the productivity he predicted. It was one of the largest forms in which the twentieth century took its gains as time, and the one people value most; nobody campaigns to work until they die. It is now being shortened, for sound demographic reasons, by rules that treat everyone as average.
The argument of all three essays comes back to the same point. How the gains of productivity are split, between income and time, between the top and the middle, between the working years and the years after, is not decided by technology or by demography alone. It is decided by rules, and rules can be written with the builder and the professor, the old woman on a minimum pension and the young emigrant paying London rent, in mind. Whether they will be is the part of Keynes’s question that remains open.
Where this could go
The same calculation for other countries. Whether the reversal between men and women, and the gap of six years by education, hold across Europe, and how each country’s long-career exceptions compare with what the healthy-years arithmetic would give.
The pensions of the emigrants. Hundreds of thousands of Portuguese who worked abroad in the 2010s will draw pensions from several countries. How their contribution records are pieced together, and what they will receive, is an open and very Portuguese question.
This piece was written collaboratively with Claude Opus 5.5 (Anthropic): human specification, editorial direction and critical review; machine data research, analysis and drafting. The numbers are computed by scripts/retirement.py from the OECD’s Pensions at a Glance database and Eurostat’s labour-force, income, health and social-protection statistics; results are in docs/retirement-results.json and the figures in docs/retirement-figures.html. Healthy life years rest on self-reported limitations and the series has breaks; poverty rates are relative to each country’s median income in the same year. The OECD’s replacement rates are for a hypothetical full-career worker under current rules, not for today’s pensioners. The healthy-years pension age uses the Sullivan method: person-years from Eurostat’s Portuguese life tables for 2024, multiplied in each age band from 55 to 85 and over by the share of people in EU-SILC who report no long-standing limitation in their usual activities, for each sex, education level and income fifth; it reproduces Eurostat’s own healthy life years at 65 to within a tenth of a year. Education and income groups use the life table of their sex, or of the whole population, since Portuguese life expectancy at 65 differs by about a year across education levels; the samples of older people with secondary or tertiary education are small, and where an age band is missing the 75-and-over figure is used. The photographs are from Wikimedia Commons and credited with their licences in the captions; local copies are in assets/images/2026/.
The cover photograph shows card players in the Jardim de São Lázaro, Porto, in 2015; photograph by Vysotsky, via Wikimedia Commons, licensed CC BY-SA 4.0, cropped.
Authored by: Luis Matos Ferreira — Physicist, Developer, Writer
- The Fifteen-Hour Week — Part I: Keynes’s prediction, productivity and the hours.
- Where the Hours Went — Part II: who got the income, what it bought, and where the hours of work actually go.
- The Two-Thirds Country — the same questions for Europe and Portugal.
- The Sustainability Story — Portugal’s pension system: a contributory surplus, a closed civil-service scheme and thirty years of gloomy forecasts.
- The Duty to Work — how work became a duty, an identity and a source of status, and what it does to happiness.
- Gesetz betreffend die Invaliditäts- und Altersversicherung, 22 June 1889 (German Reich), in force 1891; the pension age was lowered from 70 to 65 in 1916.
- Old-Age Pensions Act 1908 (8 Edw. 7 c. 40), United Kingdom, pensions payable from 1 January 1909; Social Security Act of 14 August 1935, United States.
- OECD, Pensions at a Glance database (OECD Data Explorer, DSD_PAG@DF_PAG): average effective age of labour market exit and expected years after exit, by sex, 1970–2024; net pension replacement rates for an average earner; current and future normal retirement ages; projections of public expenditure on pensions.
- Gruber & Wise (eds.), Social Security and Retirement around the World, University of Chicago Press, 1999.
- Eurostat, Labour Force Survey: employment rates by sex and age (lfsa_ergan); duration of working life (lfsi_dwl_a).
- Eurostat, healthy life years and life expectancy at age 65 (hlth_hlye); self-perceived health by age (hlth_silc_01); life expectancy by educational attainment (demo_mlexpecedu).
- Decreto-Lei n.º 35/2002, de 19 de Fevereiro (whole-career pension formula); Lei n.º 53-B/2006, de 29 de Dezembro (indexation of pensions); Decreto-Lei n.º 187/2007, de 10 de Maio (sustainability factor); Decreto-Lei n.º 187/2007, art. 20, as amended by Decreto-Lei n.º 167-E/2013, de 31 de Dezembro (normal pension age moves by two-thirds of the change in life expectancy at 65); Portaria n.º 358/2024/1, de 30 de Dezembro, and Portaria n.º 476/2025/1, de 29 de Dezembro (ages for 2026 and 2027; sustainability factors for 2025 and 2026).
- Chetty, Stepner, Abraham, Lin, Scuderi, Turner, Bergeron & Cutler, “The Association Between Income and Life Expectancy in the United States, 2001–2014”, JAMA 315, 1750 (2016).
- Eurostat, European Health Interview Survey 2019: persons providing informal care or assistance at least once a week, by sex and age (hlth_ehis_ic1e).
- Eurostat, EU-SILC: aggregate replacement ratio (ilc_pnp3); relative median income ratio of people aged 65 and over (ilc_pnp2).
- Eurostat, EU-SILC: at-risk-of-poverty rate by poverty threshold, age and sex (ilc_li02).
- Eurostat, EU-SILC: average gender pension gap by age group (ilc_pnp13), ages 65–79.
- Eurostat, ESSPROS: pension expenditure (spr_exp_pens), all pension schemes, % of GDP.
- Eurostat, EU-SILC: distribution of population by tenure status and type of household (ilc_lvho02), 2025; rates within each household type computed by scripts/retirement.py.
- Eurostat, EUROPOP2023 population projections, demographic balances and indicators by type of projection (proj_23ndbi): old-age dependency ratio, 3rd variant (65+ to 20–64), baseline; the same series as Figure 11 of The Two-Thirds Country.
- Eurostat, life tables by age and sex (demo_mlifetable), Portugal, 2014 and 2024; EU-SILC, self-perceived long-standing limitations in usual activities due to health problem, by sex, age and labour status (hlth_silc_06), by educational attainment (hlth_silc_07) and by income quintile (hlth_silc_12); life expectancy by educational attainment (demo_mlexpecedu), 2017. Healthy years by the Sullivan method, computed in scripts/retirement.py.
Comentários
Enviar um comentário