Why Disaster Death Tolls Keep Changing — How the Numbers Are Actually Counted

Direct answer: a disaster death toll rises for weeks or months because the first figure only counts bodies recovered and identified in reachable places. It grows as rescuers reach cut-off areas, as the missing are eventually declared dead, and as people die later of injuries, disease, heat or lack of dialysis and insulin — deaths that are caused by the disaster but happen after it. The largest revisions come from excess mortality studies, which compare deaths in a period against the number normally expected, and these can be tens of times higher than the original count.

The first number is a count of what could be reached

In the hours after an event, the figure being reported is not an estimate of how many people died. It is a tally of confirmed bodies, gathered by whichever authorities could get to whichever places were passable. Everything about that process biases it downwards:

• Roads and bridges are gone. The worst-affected valley is, almost by definition, the one nobody has reached yet.
• Communications are down. A district that cannot report is recorded as zero, not as unknown. See why networks fail after a disaster.
• Counting is local. Figures come up through district and provincial offices, and each layer adds delay.
• Search itself keeps stopping. After an earthquake, every significant aftershock halts rescue work and can collapse already-damaged buildings onto the people searching them.

This is why a toll can double overnight several days in with no new event having occurred. Nothing got worse; the count caught up.

Direct deaths and indirect deaths are not the same number

A direct death is caused by the physical force of the event: a collapsing building, moving water, debris. An indirect death is caused by the conditions the event created — an untreated heart attack because the hospital has no power, cholera from contaminated water, a chronic patient who cannot get dialysis, carbon monoxide poisoning from a generator run indoors.

Different countries and agencies include indirect deaths differently, and some do not include them at all. Two honest, careful authorities can therefore publish very different totals for the same disaster without either of them being wrong — they are answering different questions. When you see two figures far apart, the first thing to check is not who is lying but which definition each is using.

The missing are counted last, and separately

People unaccounted for are usually reported in their own column rather than added to the dead. Moving someone from missing to dead is a legal act in most countries, and it often requires a court declaration months or years later. After tsunamis, landslides and dam failures the missing can outnumber the confirmed dead for a long time, because water and debris move bodies far from where anyone will search — storm surge in particular carries everything it lifts inland and then back out again.

Excess mortality: where the biggest revisions come from

The most reliable way to measure the true human cost is not to count bodies at all. It is to count all deaths in the affected area over a period and compare them with the number that would normally be expected for that place and season. The difference is the excess.

Two well-documented cases show the size of the gap:

• Hurricane Maria, Puerto Rico, 2017. The official toll stood at 64 for months. A study commissioned by the Puerto Rican government from the George Washington University Milken Institute School of Public Health estimated roughly 2,975 excess deaths over the six months following the storm, and the government adopted that figure. A separate Harvard-led survey published in the New England Journal of Medicine produced a higher central estimate with a very wide confidence interval, which is itself a fair illustration of how uncertain this work is.
• The 2003 European heatwave. Almost nobody was recorded as having been killed by heat while it was happening. Later analyses of national mortality records put the excess across Europe at around 70,000 deaths. Heat is the clearest case of a hazard that is nearly invisible to body counting: it kills people who already had a condition, in their homes, and the death certificate names the condition.

Excess mortality is also why heatwave tolls are usually published long after the event and why they are so often described as estimates rather than counts. See why humid heat kills at temperatures that sound survivable.

Why two credible sources disagree

Beyond definitions, four things drive the spread:

• Who is counting. A national government, a UN agency, an academic team and a news wire have different access, different methods and different deadlines.
• What period is counted. Deaths in the first week, the first month and the first year give three very different totals for the same disaster.
• Registration coverage. Where births and deaths are not comprehensively registered, there is no baseline to compare against, so excess mortality cannot be calculated reliably at all.
• Incentives. A low toll can suggest a competent response, and a high one can unlock international assistance. This is worth naming plainly, but it is the last explanation to reach for, not the first — the technical reasons above account for most disagreements.

The 2010 Haiti earthquake remains the widest documented disagreement: the Haitian government reported a figure above 300,000, while a study commissioned by USAID estimated somewhere between roughly 46,000 and 85,000. Both numbers have been in circulation ever since, and the honest position is to give the range and say who produced each end of it.

What a careful figure looks like

A trustworthy death toll tells you four things: who counted it, what period it covers, whether indirect deaths are included, and whether the missing are in it. A bare number with none of those attached is not yet information.

This site records disputed tolls as ranges with a note explaining the disagreement rather than picking a single figure, and every historical record names its sources. You can see that on the historical disasters archive and the deadliest disasters pages, and the rules we follow are set out in our methodology and data sources.

What this means when you are following a live event

If someone you know is in an affected area, the early toll tells you very little about their individual risk, and watching it climb is not new information about them — it is the count catching up with the event. The useful things in those hours are the official national agency for that country, and a direct, low-bandwidth message rather than repeated calls.

Each of our country pages links the official national disaster and meteorological agencies for that country, and those agencies — not aggregators, and not this site — are the authority during an emergency.

This article explains how disaster mortality figures are produced. It is not a source of casualty figures for any specific ongoing event; for that, use the responsible national agency and the sources named on each event page.

Frequently asked questions

Why do disaster death tolls keep rising for days?

Because the first figure counts only bodies recovered in places rescuers could reach. As roads reopen and cut-off districts report, the count catches up with what already happened. A toll doubling on day four usually means access improved, not that the disaster got worse.

What is the difference between direct and indirect disaster deaths?

A direct death is caused by the physical force of the event — a collapsing building or moving water. An indirect death is caused by the conditions it created, such as an untreated heart attack when the hospital has no power, or disease from contaminated water. Agencies include indirect deaths differently, which is one of the main reasons credible sources publish different totals.

What is excess mortality and why is it higher than the official toll?

Excess mortality counts all deaths in an area over a period and subtracts the number normally expected for that place and season. It captures deaths that no body count would attribute to the disaster. After Hurricane Maria the official toll stood at 64 while a government-commissioned study estimated about 2,975 excess deaths.

Are the missing included in the death toll?

Usually not. The missing are normally reported in a separate column, because moving someone from missing to dead is a legal act that can require a court declaration months or years later. After tsunamis, landslides and dam failures the missing can outnumber the confirmed dead for a long time.

Why do heatwave death tolls appear so long after the heatwave?

Heat rarely appears on a death certificate. It kills people who already had a heart or lung condition, usually at home, and the certificate names that condition. The toll therefore has to be reconstructed from national mortality records months later. The 2003 European heatwave is estimated to have caused around 70,000 excess deaths across the continent.

Which disaster has the most disputed death toll?

The 2010 Haiti earthquake. The Haitian government reported a figure above 300,000, while a study commissioned by USAID estimated roughly 46,000 to 85,000. Both figures remain in circulation, and the honest way to report it is as a range that names who produced each end.

Sources

• CRED / EM-DAT — International Disaster Database, UCLouvain
• WHO — Health emergencies: mortality and excess mortality methods
• UN OCHA — Humanitarian data and reporting
• George Washington University Milken Institute School of Public Health — Ascertainment of the Estimated Excess Mortality from Hurricane Maria in Puerto Rico
• New England Journal of Medicine — Mortality in Puerto Rico after Hurricane Maria
• UNDRR — Disaster losses and damage tracking
• USGS — Earthquake casualty estimation (PAGER)

Found an error? See our corrections policy.