2026-11-18
How to read a mortality rate without lying to yourself
Life expectancy, infant mortality, and under-five mortality are three different numbers that get used interchangeably. They should not be.
本文为英文。
Most arguments about how long people live are actually arguments about which number is on the slide. Here are the four that matter, what each one measures, and the specific way each one misleads.
Life expectancy at birth
The average number of years a newborn would live if current age-specific death rates stayed constant for their whole life.
Two things follow. First, it is a period measure, not a cohort measure — it describes a snapshot of today’s rates, not the actual lifespan of anyone alive. Second, it is dominated by infant mortality. Because the average includes deaths at age zero, a country that halves its infant mortality can gain several years of life expectancy without a single adult living one day longer.
This is why life expectancy jumped so fast in the twentieth century in places that did not get richer: they got vaccines and clean water, and fewer people died before their first birthday.
Infant mortality rate
Deaths before age one, per 1,000 live births.
The subtlety: “live births” excludes stillbirths, and the definition of “live” varies. A baby who breathes once and dies is a live birth in some registries and a stillbirth in others. In countries with weak registration, early neonatal deaths are undercounted, which makes the IMR look better than it is.
A rate of 3 per 1,000 (Japan, Finland) and a rate of 70 per 1,000 (parts of West Africa) are not points on the same spectrum of quality. They are different worlds.
Under-five mortality rate
Deaths before age five, per 1,000 live births. Always higher than IMR, since it includes it.
This is the number that best captures the experience of childhood in a country, because the second through fifth years are where diarrhoea, pneumonia and malaria do most of their work — all treatable, none expensive.
U5MR is also the number that responds fastest to public health investment, which makes it the favourite of development agencies and the reason it appears in so many reports.
Healthy life expectancy (HALE)
Years lived in full health. Always lower than life expectancy, by a margin that varies enormously.
This is the number that matters most for the last twenty years of a life, and it is the one almost nobody quotes. Japan has a life expectancy around 84 and a HALE around 74. Ten years of life in less than full health, on average. In countries with weaker health systems the gap is often larger, because the same chronic conditions are managed worse.
A country can have a respectable life expectancy and a poor HALE. It means people live, and live unwell.
The three lies
The first lie: comparing a period measure to a personal expectation. “Life expectancy is 84, so I should plan for 84.” You are not a newborn and the rates will not stay constant. If you are 60 today, your expected additional years are computed from age 60, and they are not 24.
The second lie: reading a national average as a personal fate. The distribution around a national mean is wide. In a country with a life expectancy of 55, a substantial minority reach 80. In a country with 84, a substantial minority die before 60.
The third lie: using a single year’s figure. Life expectancy moves. COVID took more than a year off several rich countries in 2020–2021, and some of that has not come back. A figure quoted without a year is a figure quoted without meaning.
What this simulator does with them
It uses life expectancy at birth and infant mortality as the two inputs, then derives under-five mortality from the ratio between them, then builds a full annual hazard curve from age five using a Gompertz form with one solved parameter.
It does not model HALE at all. That is a real limitation: the game will let you reach 85 in good health when the realistic version of that life includes a decade of managed chronic illness. The health stat captures some of it — health decays with age and events subtract from it — but it is a game abstraction, not a health-adjusted life year.
Saying so plainly is part of the design. A simulation that does not state its own limits is not a simulation, it is an advertisement.