The toll of past earthquakes

The death toll of a major earthquake is not a measurement but an estimate that is revised for years. Differences between sources are normal and explicable: some people are never found, some lived unregistered, some died later of their injuries and entered a different statistic, and in conflict zones the counting is done by different parties under different rules.

So what follows gives orders of magnitude with an indication of whose estimate it is, without pretending to know an exact number. Where estimates differ several-fold, we say so — that itself is an important part of the picture.

Two earthquakes of magnitude 7.7 and 7.6, nine hours apart. The first struck at 04:17 local time, while people were asleep in their homes. Eleven provinces of Türkiye and five governorates of Syria were affected.

The official death toll: 53,537 in Türkiye. In Syria, 1,414 in government-controlled areas and 4,537 in opposition-held ones. Around 59,500 in total — making it the deadliest earthquake since 1970.

The destruction is measured in hundreds of thousands of buildings collapsed or slated for demolition. This is where the mechanisms described in the article on earthquake resistance appeared en masse: soft ground floors, inadequate transverse reinforcement, poor concrete, buildings extended and altered after construction.

Two circumstances made it worse. The hour — people had no time to get out and were buried inside their homes. And the February cold: hypothermia shortens the window in which someone trapped can still be brought out alive.

A magnitude 7.4 event on the North Anatolian Fault at 03:02, in the densely populated industrial region around the Sea of Marmara. The official death toll exceeds 17,000, and estimates including the missing are higher. A further earthquake at Düzce in November of the same year added hundreds more.

For Türkiye this was a turning point, and not only in scale of loss. Building codes were substantially revised afterwards and the construction inspection system was rebuilt. Much of what is now taken as mandatory dates from then.

The bitter lesson of 2023 is that codes work only where they are followed. Among the buildings that collapsed were some built after the reform.

This is the main conclusion from comparing disasters, and it runs against intuition. The Japanese earthquake of 11 March 2011 had a magnitude of about 9.0 — some thirty times the energy of a magnitude 8, and thousands of times that of a 7. Around 20,000 people died or went missing, and the overwhelming majority drowned rather than being crushed.

The Haiti earthquake of 12 January 2010 had a magnitude of about 7.0 — incomparably less energy. Estimates of the death toll range from several tens of thousands to more than three hundred thousand depending on source and method. Even at the lower bound that is several times Japan's toll from a far larger event.

The difference is explained not by seismology but by construction quality, population density, the hour of day, the readiness of rescue services, and whether a secondary hazard such as tsunami or fire was involved. Magnitude describes an event in the earth's crust. A death toll describes the society that event met.

Tsunami. In Japan in 2011 the overwhelming majority of the dead drowned rather than being crushed. The same in the Indian Ocean in 2004: a magnitude 9.1 event off Sumatra generated a wave that took on the order of 230,000 lives across fourteen countries — most of them hundreds or thousands of kilometres from the epicentre.

Fire. In Kobe in 1995, with a magnitude near 6.9 and around 6,400 dead, a substantial share of the destruction came from fires in dense timber housing with the water mains broken. The historic Tokyo earthquake of 1923 is remembered above all for a firestorm rather than the shaking.

Landslides and liquefaction. In mountainous areas slides bury whole settlements, while on loose saturated ground buildings topple bodily while remaining structurally intact.

The practical conclusion for a reader: the list of threats does not end at "the building might collapse". If you live by the sea, account for tsunami. In dense housing, for fire. In the mountains, for landslides.

Night-time earthquakes are deadlier than daytime ones, all else equal: people are in residential buildings, and housing stock is usually weaker than public buildings; they are asleep and lose seconds waking; they end up in darkness without shoes.

Kahramanmaraş at 04:17, İzmit at 03:02 — both of Türkiye's worst events fell in the dead of night. That is not a coincidence in the sense of cause, but it is a pattern in the sense of consequence.

Daytime shocks redistribute the casualties: more people killed in the street by falling cladding, more in schools and workplaces. Which, incidentally, is why school drills matter — they work in the daytime scenario.

Investigations of different earthquakes in different countries arrive at strikingly similar lists of causes. Soft ground floors for shops and parking. Insufficient or badly detailed transverse reinforcement in columns. Poor concrete and contaminated aggregate. Storeys added beyond the design. Unauthorised alterations affecting structural elements. No gap between adjacent buildings.

Almost nothing on that list is news to engineers. Which is the most uncomfortable part: disasters rarely reveal phenomena unknown to science. They reveal that what was known was not applied.

An earthquake in a wealthy country usually produces enormous economic damage with a relatively small death toll: expensive assets are destroyed but people survive. In a poor country the picture inverts — material losses are smaller in absolute terms and the deaths many times greater.

That follows directly from what actually determines mortality: not the strength of the shaking but the building stock and the state's ability to mount a rapid response. Hence a conclusion uncomfortable for any country: money spent on codes, inspection and strengthening existing buildings is literally the price of human lives, calculated in advance.

In the first hours after an event the death toll is always an undercount: only those found are confirmed. A figure rising through the week does not mean the disaster is worsening — it means the rubble is being cleared.

Then comes a phase where the figure stabilises but the missing remain. Their fate can take months to determine, and the final number depends on whether a source includes the missing in its total.

Finally, years later, research estimates appear that sometimes differ several-fold from the official ones, as with Haiti. When you meet different numbers for one event, look at the date of the estimate, who made it, and whether the missing are included. That usually explains the discrepancy, rather than anyone acting in bad faith.

Not for the statistics. The point is that almost every large figure in this article breaks down into specific causes known in advance: how it was built, what hour it happened, whether there was anywhere to shelter, whether people were reached in time.

No individual controls all of those — but everyone controls some. Where you live and in what building. Whether your wardrobe is anchored. Whether your family knows where to meet. Whether you know that someone pinned for a long time must not simply be pulled free.

That is why the other articles in this section are written the way they are. The figures above are the argument for reading them beforehand rather than afterwards.