Ask people whether child mortality has improved and most say it has stayed flat or got worse. The data says otherwise, and the size of the change is hard to overstate.

Across the 196 countries with figures for both years, deaths before a child's fifth birthday fell from an average of 55.2 per 1,000 live births in 2000 to 24.2 in 2024. That is a 56% reduction in a single generation. 189 of those 196 countries improved.

The countries that moved furthest

• Rwanda — 184.2 → 37.7 per 1,000, a 80% reduction
• Angola — 185.0 → 49.0
• Sierra Leone — 222.8 → 90.5
• Niger — 227.9 → 110.7
• South Sudan — 180.1 → 96.7

Rwanda is the case worth sitting with. In 2000 roughly one child in five did not reach five years old. Today it is fewer than one in twenty-five. That happened in a country rebuilding from genocide, and it happened in twenty-four years.

The quieter story is the big countries

The headline reductions are concentrated in small, very high-mortality states, because that is where the room to move was. But the largest absolute number of lives is in the big populations:

• India — 91.8 → 26.6 per 1,000
• Nigeria — 177.3 → 115.6

India's rate fell by 71%, from roughly one child in eleven not reaching five to one in thirty-eight. Applied to a birth cohort the size of India's, that is the largest absolute reduction in child deaths of any country in the period.

Nigeria shows the other side. It improved — 177.3 to 115.6 is real — but it remains among the highest rates in the world, and its share of global under-five deaths has risen as other countries pulled away.

What "already low" looks like

Finland went from 4.3 to 2.4 per 1,000. Rich countries had little room left in 2000 and made small gains. This is why a global average understates the achievement: the countries that improved most are the ones that were worst off, which is the opposite of how most global indicators move.

Reading this number honestly

Under-five mortality is modelled, not counted. In the countries with the highest rates, birth and death registration is weakest, so these are estimates with real uncertainty bands rather than tallies. The direction and rough magnitude are not in doubt; a specific country's value to one decimal place is more confident than the underlying data warrants.

Six countries did go backwards, and one did not move at all. They are a small minority, and the reasons are what you would expect: conflict, economic collapse, and in a few small states, statistical noise on a tiny number of births.

A falling rate is not the same as fewer deaths. Where the number of births is rising fast, a falling rate can still coexist with a flat death count. Rate is the right measure for how well a health system is doing; it is not the measure of how many children die.

A note on which measure this is. The figures above are under-five mortality: deaths before a fifth birthday. The atlas page below carries infant mortality, deaths before a first birthday, which is a different and lower series — South Sudan tops it at 71.9 per 1,000 against Nigeria's 115.6 here. Most child deaths happen in the first year, so the two track each other closely, but they are not interchangeable and the values will not match.

Explore the map of infant mortality by country, or the relationship between doctors per capita and infant deaths. You can also try to spot this pattern yourself in year mode, which hides the year and asks you to date the map. Source: World Bank Open Data, 2000–2024.