Article

Why Germany looks deadlier than Kenya

On the Ledger, Germany's crude death rate1 is 12.3 per thousand and Kenya's is 7.2. Germany's is higher. That is not a measure of danger, and understanding why is the key to reading any country mortality table.

Open the Ledger's Atlas of Lethal Claims, select Germany, and the card reads 12.3 per thousand. Select Kenya, and it reads 7.2. Divide 12.3 by 7.2 and Germany looks 1.7 times as lethal.

Now check life expectancy2 on the same two cards. Germany is 81.0 years, and Kenya is 63.6. The country with the higher death rate is the one where people live 17.4 years longer.

Those two pairs are the United Nations World Population Prospects figures for 2023, as published by the World Bank, rounded to one decimal place: 12.3 per thousand in Germany and 7.2 in Kenya, and 81.0 years in Germany and 63.6 in Kenya. The reversal is in the official series. The cards are using it.

This is not an error in the arithmetic. It is the single most reliably misread statistic in demography3.

What a crude death rate actually measures

The crude death rate is exactly what it sounds like: deaths per thousand people per year, with no adjustment for anything. In particular, there is no adjustment for age, and age is overwhelmingly the dominant predictor of death.

Germany's population is old. In the United Nations estimates for 2023, 22.8% of people in Germany were aged 65 or over, against 2.9% in Kenya. A large share of Germany's population is therefore in the age bands4 where death is common, so a lot of people die each year in a country that is one of the healthiest on earth.

Kenya's population is young, with a median age5 of 19.5 in 2023. Most Kenyans are in age bands where death is uncommon in any country, so relatively few people die each year, even though health is substantially worse at the same ages.

The crude death rate does not measure danger. It mostly measures age structure6.

The fix: age standardisation

Demographers handle this with age-standardised mortality rates7. The method is to take each country's death rate within each age band, then apply those rates to a single shared reference population8. The question becomes: if these two countries had identical age structures, how would their mortality compare?

Standardised that way, the picture inverts. Life expectancy does the same job from a different angle. It is calculated from age-specific death rates9, so it is already immune to the distortion. The age-standardised rate says the same thing. In Global Burden of Disease10 estimates for 2023, the age-standardised death rate was about 860 per 100,000 people in Kenya, nearly twice Germany's figure of about 470 per 100,000.

The rule of thumb: to compare how dangerous two countries are, use life expectancy or age-standardised rates. Crude death rates tell you how many funerals a country holds, which is a different question and mostly a question about demography.

What the cause tables do tell you

The per-cause percentages in the atlas are more robust, because they are shares of that country's own deaths rather than comparisons across populations. Read that way, the tables describe where each country sits in the epidemiological transition11 (the shift from infectious to chronic disease that accompanies rising income).

On the atlas, Nigeria's leading illnesses are neonatal disorders12, lower respiratory infections13, malaria14 and diarrhoeal disease15: a profile dominated by infection and early childhood. The United Kingdom's are cancers as one group, then dementia17, ischaemic heart disease16 and stroke18: chronic conditions of long life. Cancers are kept as one cause on every card. In the official registrations for England and Wales in 2023, where each cancer site is counted separately, the leading named cause was dementia and Alzheimer's disease. Mexico's table has diabetes19 third, just beside cancers, and kidney disease20 inside the five: a country well into that transition and carrying a particular metabolic21 burden.

Those differences are real and meaningful. They also show up under external causes22. Falls lead that list in the United Kingdom, Germany, France and Canada. Interpersonal violence24 leads it in Brazil, Mexico and South Africa. Self-harm leads it in Japan and South Korea. In the United States the leading external cause is road injury, with self-harm second. Drug overdose23 is not in the column. The source counts it as an illness, and in the United States and Canada it is about 3% of deaths: larger than any injury, and still outside the five leading illnesses. Firearm deaths are inside self-harm or interpersonal violence, according to intent. Those are genuine national differences, not artefacts25.

Three more ways to be misled

Better diagnosis looks like more disease

When a country improves its cancer screening, recorded cancer incidence26 rises. When dementia awareness improves, recorded dementia deaths rise. Both look like an epidemic in the data and are partly an improvement in observation. This is a real and persistent confounder27 in any time series28 of causes.

Ill-defined causes hide the real cause

In countries with limited medical certification29, a large share of deaths are coded to ill-defined30 or "garbage" causes. Statisticians redistribute these across plausible real causes using models31. It is the right thing to do, and it means some published country figures rest more on modelling than on observation, which is worth knowing before treating a decimal place as meaningful.

Percentages move when the denominator32 moves

If a country dramatically reduces infant deaths, the share of its deaths attributed to heart disease rises, even if heart disease has not budged. A cause can climb a country's table purely because the causes around it fell. This is progress, and it reads as decline.

How to read the atlas

The country figures on this site are a 2023 extract for a visualisation, rounded to one decimal place. They are not a live feed, and they should not be used for policy, research or any decision of consequence.

Notes

  1. Crude death rate. Deaths in a year for every thousand people, with no allowance for whether the population is old or young. ↩
  2. Life expectancy. The average years a newborn would live if today's death rate at each age never changed. It is a snapshot of one year, not a forecast for anyone alive now. What life expectancy actually measures sets the calculation out in full. ↩
  3. Demography. The study of populations: how many people there are, how old they are, and how many are born and die. ↩
  4. Age band. A group of ages counted together, such as under-fives, or people aged 65 or over. ↩
  5. Median age. The age that splits a population in half. Half the people are younger than it, and half are older. ↩
  6. Age structure. How a population is spread across ages: how many children, how many adults, and how many older people. ↩
  7. Age-standardised mortality rate. A death rate worked out as if every country had the same mix of ages, so one country does not look worse merely because it is older. ↩
  8. Reference population. A population with a fixed age mix, used as the common yardstick. Each country's own age-by-age death rates are applied to it. ↩
  9. Age-specific death rate. The death rate inside one age band, rather than for the whole population at once. ↩
  10. Global Burden of Disease. A research project, led by the Institute for Health Metrics and Evaluation, that estimates deaths by cause for every country, including places where many deaths are never certified. ↩
  11. Epidemiological transition. The shift, as a population grows wealthier and more children survive, from deaths caused by infection to deaths caused by chronic disease at older ages. A chronic disease is one that lasts a long time, such as heart disease, rather than an infection that runs its course. ↩
  12. Neonatal disorders. Illness and injury in the first 28 days of life, including premature birth and complications during birth. ↩
  13. Lower respiratory infections. Infections of the lungs and the lower airways. Pneumonia is the largest of them. ↩
  14. Malaria. A parasitic infection spread by some mosquitoes. It remains a leading cause of death in parts of Africa, mostly among young children. ↩
  15. Diarrhoeal disease. Severe diarrhoea, usually from infection. It kills mainly by dehydration, and most of the deaths are in young children. ↩
  16. Ischaemic heart disease. Disease in which the heart muscle receives too little blood, usually because the arteries that supply it have narrowed. A heart attack is the best-known form. ↩
  17. Dementia. A group of illnesses, including Alzheimer's disease, that progressively damage memory and thinking. Almost all of the deaths are in old age. ↩
  18. Stroke. A sudden stop in the blood supply to part of the brain, or a bleed inside it. Either can kill or leave lasting injury. ↩
  19. Diabetes. A lasting failure of the body to keep sugar in the blood within a safe range. Over years it damages the heart, the kidneys and the blood vessels. ↩
  20. Chronic kidney disease. Long-term damage that leaves the kidneys less able to filter the blood. It often follows diabetes or high blood pressure. ↩
  21. Metabolic. To do with how the body handles food and energy. Here it means diabetes and the diseases that follow it, including kidney disease. ↩
  22. External causes. Deaths from injury rather than from disease. On the atlas the group is road injuries, falls, drowning, fire, unintentional poisoning, natural disasters, self-harm, interpersonal violence and collective violence. Overdose is not in it. A firearm death is counted with self-harm or interpersonal violence, by intent. ↩
  23. Drug overdose. In the WHO estimates used by the atlas, fatal overdose is filed under drug use disorders, which is an illness. The 2023 share is 3.4% of deaths in the United States and 3.1% in Canada. Unintentional poisoning is a separate, much smaller line. ↩
  24. Interpersonal violence. Homicide and fatal assault between people, as distinct from deaths in war or from self-harm. ↩
  25. Artefact. A pattern created by the way the figures were collected or calculated, rather than by a real difference in the world. ↩
  26. Incidence. The number of new cases of a disease in a period. It is not the same as the number of deaths. ↩
  27. Confounder. Some other change that moves the figures and can be mistaken for the change you meant to measure. ↩
  28. Time series. The same measurement repeated over many years, so the movement can be followed. ↩
  29. Medical certification. A doctor recording the cause on a death certificate. Where that rarely happens, the recorded cause is often vague or missing. ↩
  30. Ill-defined causes. Deaths recorded under a vague label, such as "unknown" or "old age", because no precise cause was certified. The least usable of these are called garbage codes, and analysts estimate how they should be reassigned. ↩
  31. Model. A calculation that estimates a figure the records do not state directly, using patterns from places where the records are better. ↩
  32. Denominator. The total a percentage is taken from. Here it is all deaths in that country, so if other causes fall, a cause that has not changed takes a larger share. ↩

More from the Ledger

Why 1.95 deaths per second? It shows how the global death rate is derived, and why the answer moved. The number that says the most about a country Why demographers ask for infant mortality first, and how far it has fallen. What life expectancy actually measures It is a snapshot of one year, dressed up as a lifetime. How the Ledger counts It lists every rate the site uses, and exactly which parts are dramatised.

Last reviewed: 5 October 2026