A forecast can have a precise headline number and still be deeply uncertain. In 2017, an Imperial College London-led team projected that life expectancy at birth for South Korean women could reach 90.8 years in 2030. The same analysis put the probability of actually crossing 90 at 57 per cent.

Those two numbers are not contradictory. The 90.8 was the central result from a distribution of possible futures generated by several mortality models. The 57 per cent described how much of that distribution lay above a particular line. Crossing 90 was more likely than not, but only narrowly.

ScienceBlog previously examined why South Korea led the 2030 forecast. This follow-up focuses on the uncertainty inside it and on what newer Korean statistics now say as the target year approaches.

The projection was built in 2017 from data ending in 2014

The study, published in The Lancet, covered 35 industrialised countries with reliable death registration and populations of at least one million. Vasilis Kontis, James Bennett, Colin Mathers, Guangquan Li, Kyle Foreman and Majid Ezzati used national death counts and population estimates to project age-specific mortality through 2030.

For South Korea, the available series ran from 1985 to 2014. The country had moved rapidly up the longevity table: female life expectancy increased by an average 3.7 years per decade from 1985, when South Korean women ranked 29th among the countries compared.

Rather than select one forecasting equation, the researchers built an ensemble of 21 models. They tested each model by withholding 13 years of historical data, asking it to forecast those missing years, then assigning greater weight to models with smaller errors. The final projection pooled draws from all 21 in proportion to those weights.

This was a serious attempt to include uncertainty about model choice. It was not a crystal ball. Even the best-performing historical relationship can change after the last observed year.

If you enjoy learning about Korean culture in general, we came across a great video that dives into the Korean way of raising children. It’s an interesting watch for any parent:

What the 90.8-year estimate actually described

Life expectancy at birth is often mistaken for the age a baby is personally expected to reach. In a period life table, it is a summary of mortality rates at every age during a given year.

The 2030 estimate asks a hypothetical question: how long would a newborn live if she experienced the mortality rate projected for infants in 2030, then the projected rate for one-year-olds, two-year-olds and every later age, with that entire 2030 pattern held constant over her life?

No real baby will live under permanently frozen 2030 conditions. Medicine, behaviour, infectious disease, climate, inequality and health policy will continue changing. A person’s actual lifespan belongs to a birth cohort moving through future decades, while period life expectancy compresses one year’s mortality environment into a single number.

Nor does an average of 90.8 mean most women die at 90 or that every woman receives an extra four years. National life expectancy averages together very different lives. It can rise because deaths decline in infancy, middle age, old age or some combination. In this study, more than half the projected gains for women were expected to come from improved survival after 65.

Why a central estimate above 90 produced only a 57 per cent chance

The model did not output one fixed future. It produced a posterior distribution reflecting uncertainty within each model and disagreement between models. Its central estimate for South Korean women was 90.8, but a substantial part of the distribution remained below 90.

The result assigned a 57 per cent probability to exceeding 90 and a 43 per cent probability to landing at or below it. That is enough to make the threshold the most likely side of a two-way question, but nowhere near certainty.

The broader probabilities show how cautiously the forecast should be read. The researchers calculated a 90 per cent chance that South Korean female life expectancy would exceed 86.7 years, the highest national female figure recorded anywhere in 2012. They gave South Korea a 45 per cent chance of ranking first for women in 2030 and a 27 per cent chance of ranking second.

France, Japan and Spain followed in the central ranking, but their probability distributions overlapped. South Korea was the leading candidate, not a guaranteed winner. The Imperial College summary published with the paper reported both the 90.8 estimate and the 57 per cent chance, even though many later headlines retained only the more dramatic number.

The latest observed female figure is 86.6 years

There is no need to wait until 2030 to check the direction of travel. South Korea’s official 2024 life table, released in December 2025, put female life expectancy at 86.6 years. That was 0.2 year above 2023. Male life expectancy was 80.8, leaving a female advantage of 5.8 years.

South Korean women remained among the longest-lived national populations in the OECD comparison, behind Japan at 87.1 and Spain at 86.7 in the years available for each country. The observed Korean figure is nevertheless 4.2 years below the Lancet model’s central estimate for 2030.

Closing that entire gap in the six calendar years from 2024 to 2030 would require an average increase of 0.7 year for each passing year. That arithmetic is not an updated probability calculation, and life expectancy does not have to rise in a straight line. It simply shows that the 2017 central forecast now demands a much sharper late acceleration than recent data has delivered.

A newer Korean projection points to 88.1, not 90.8

South Korea has since produced a national forecast using more recent information. Its 2022-based population projection used sex- and age-specific death rates from 1970 through 2021 and an extended Li-Lee-Gerland mortality model.

Under the medium assumption, female life expectancy reached 88.1 in 2030. The high-longevity assumption produced 89.0, while the low assumption gave 87.0. Even the high path remained below the 90-year threshold for that date.

This later forecast does not retroactively make the Lancet analysis careless. The studies used different models, time windows and treatments of uncertainty. The Korean projection also overshot the subsequently observed 2024 female figure, placing it at 87.1 rather than 86.6. Forecasts can miss in either direction.

What the comparison does change is the language appropriate now. The 90.8 figure is a notable 2017 projection, not a settled current expectation shared by every model. The latest life table cannot be turned into a revised probability without rerunning the original ensemble with the new data.

Unexpected events are not a technical footnote

The original authors explicitly identified one limitation shared by all forecasts: they cannot account for completely unexpected changes in social conditions, technology or health systems. Three years after their analysis appeared, Covid-19 supplied an unusually large example.

Future surprises need not all reduce longevity. New treatments, better control of cardiovascular risk or further declines in smoking could improve survival faster than expected. Other shocks could slow it. Forecast uncertainty is not merely statistical fuzz around an otherwise predetermined path. It represents the fact that history after 2014 had not happened yet.

The ensemble did perform well in historical tests. For 22-year projections, its average absolute bias was 0.68 year for women across the countries and periods tested, better than the strongest single models. But an average error across past tests cannot guarantee the outcome for one country at one future date.

What becoming the first would mean

If South Korean female period life expectancy exceeds 90 in 2030, it would be the first recorded national female average to cross that threshold. It would not mean the entire South Korean population averaged 90, because men were projected lower. It would not mean most deaths occurred at exactly 90, and it would not erase inequalities within the country.

The milestone would still matter. Population life expectancy above 90 was once treated by some researchers as virtually unattainable. Crossing it would demonstrate how far national mortality can move when survival improves across many ages, especially later life.

But the study’s most useful lesson may be the less spectacular number in the title. Fifty-seven per cent is not a caveat pasted onto a confident prediction. It is the prediction stated honestly. South Korean women were the leading candidate to break 90, while the model left substantial room for a different future. With 2030 approaching, that room matters more than ever.