How Many Expats Die Abroad — Reading the Consular Data

There is exactly one number, and it is the wrong one.

A peer-reviewed analysis of US State Department records, published in the Journal of Travel Medicine in 2025, counted 15,549 non-natural deaths of US citizens abroad over 2002 to 2022. That is an average of 654 a year, a peak of 1,065 in 2010, and a pre-pandemic low of 650 in 2020. It is the most rigorous public figure that exists for how Westerners die abroad. It is also, by the statute that created it, a count that cannot include the way most older expats actually die.

That sentence is the whole piece — everything below is what it means and why it holds.

What the dataset is, exactly

The US figure is not a survey or an estimate. It is a legal obligation. The Foreign Relations Authorization Act for Fiscal Year 2003 requires the State Department to collect and publish, by country, the date, place, and cause of each US-citizen death abroad from a non-natural cause. The table covers the preceding three calendar years and is refreshed roughly every six months. It is precise, mandated, and public.

The defining word is non-natural. Homicide, vehicle crash, drowning, suicide, accident. In the Journal of Travel Medicine breakdown, four causes account for 81 percent of the recorded deaths: vehicular accidents at 29 percent, homicides at 21 percent, non-vehicular accidents at 17 percent, and suicide at 14 percent, with drowning adding roughly another 13.5 percent. That is a profile. It is the profile of a death that is sudden, external, and disproportionately young. A road at night. A rip current. A fall.

It is not the profile of how a 74-year-old retiree dies in a condo in Chiang Mai or a coastal town in Cebu. That death is a cardiac event, a cancer, a stroke, the long arithmetic of decline. Natural-cause. And natural-cause death is not collected for this table at all. Not undercounted. Not collected.

The matrix: what each dataset can and cannot see

Three Western governments publish something. None of them publishes a death rate. Read what each one structurally can and cannot show.

Dataset	What it counts	What it cannot see	A death rate?
US State Dept (FRAA 2003)	Non-natural deaths only: homicide, vehicle crash, drowning, suicide, accident.	Every natural-cause death. The cardiac event, the cancer, the stroke — the way most older people die.	No, a case list
FCDO consular	All-cause death cases it assisted with.	Deaths never notified to it; the resident population to divide by.	No, a caseload
DFAT consular	All-cause overseas-death cases.	Unnotified deaths; the resident denominator.	No, a caseload

Three Western governments publish something; none publishes a rate. The all-loss right-hand column is the artefact: every dataset is structurally blind to the modal aging death, and none has a denominator.

Source: US State Dept (FRAA FY2003); FCDO (House of Commons Library CDP-2021-0210); DFAT Consular State of Play 2024-25 · checked 2026-05

The British and Australian numbers are all-cause, which sounds like the fix. It is not. The FCDO supports roughly 4,000 families affected by a death abroad each year. In Thailand specifically there were 597 new death cases in 2022 and 430 between January and mid-October 2023. Australia’s DFAT recorded 2,054 death cases in 2024-25, up about 7 percent on the year. Counted a different way, by destination across every kind of consular case, Thailand was its single busiest at 1,017 and the Philippines third at 502, or 5.5 percent of the total. Those last two are not death-only figures, which is itself the point: even where the consulate publishes a country breakdown, it is a breakdown of cases it touched, not of deaths and not of residents. These count cases the consulate touched. They do not count deaths the consulate never heard about, and they cannot be divided by a resident-expat population, because no agency holds that denominator. A caseload that rises tells you the cohort is growing. It does not tell you the chance any one person in it dies this year.

The cohort the data is built not to see

Put the matrix together and one figure is missing from all three datasets at once.

The retirement-age expat who dies of a natural cause, at home, alone, is absent from the US table by statute, a single undifferentiated line in the FCDO and DFAT caseloads, and invisible to any rate because the denominator does not exist anywhere. The most predictable death in the entire cohort, the old person abroad dying of an ordinary illness, is the one the public data is least able to show. The deaths that are over-represented are the ones the systems were built to record: the sudden, the violent, the young. The death that defines aging abroad is under-represented by design, not by accident.

This is why the headline number misleads even when it is exactly correct. The 654-a-year US figure is not too low because someone miscounted. It is structurally scoped to a population that is not the one this site is about. A reader who takes “around 650 Americans die abroad each year, mostly in accidents” and concludes that dying abroad is rare and external has read a true number and drawn a false picture, because the dataset answered a different question than the one they asked.

Why the dataset is shaped this way

The exclusion is not an oversight, and it helps to be precise about why it exists rather than to imply concealment. The statute that built the US table was an accountability instrument. Its purpose was to make visible the deaths a government can be asked about: citizens killed by violence, by terrorism, in transport, in incidents a foreign state might be answerable for or a traveller warned against. Those are the deaths with a policy lever attached. A retiree’s heart failing at 78 in a rented room has no foreign state to answer for it and no advisory that would have prevented it. It is not a diplomatic event. It is an actuarial one, and the instrument was not built to count actuarial events.

So the design is internally coherent and externally misleading at the same time. It answers “where are our citizens being killed” well. It was never meant to answer “how does the aging diaspora die,” and it is now read, by journalists and by retirees, as if it were. The failure is not in the data — it is in the question it gets asked to settle. Every country-ranking listicle built on the non-natural table is a correct answer to a question almost nobody is actually asking, presented as the answer to the one they are.

There is also a quieter structural reason the aging death stays invisible: it is often not witnessed. The sudden death has a scene, a report, a foreign authority, a notification. The slow one can go undiscovered for a period, be certified locally as natural, and close without ever generating the kind of record that travels back to a home government as a flagged case. The data is thinnest exactly where the cohort is largest, and the mechanism of that thinness is the unwitnessed natural death of someone who lived alone. That is not a number this site can supply, because no honest one exists. The absence itself is the datum.

Why every number here is a floor

Even within its own scope, the count understates. The authors of the Journal of Travel Medicine analysis state it plainly: deaths are missed when they are not notified to an embassy, when privacy restrictions block the record, and when the person is medically evacuated and dies back in the US, where the death is no longer “abroad.” Over 2,500 records in the dataset could not even be classified by cause. The published number is the part that surfaced. It is a floor with the natural-cause cohort sawn off the bottom of it.

There is a sibling reading of the same gap in how many actually go home: the people who never returned because they died in place are a cohort no return statistic can see. The two pieces describe the same blind spot from opposite sides. One asks how many leave. This one asks how many of the deaths are even counted. The answer in both cases is the same shape: the cheerful number is the visible subset, and the invisible part is the one that matters to someone planning the last decade.

What the count does not include: the bill

For the person actually searching this, usually a relative and after the fact, the dataset has a second silence. It records that a death happened. It says nothing about what it costs the living.

Consular services are procedural, not financial. The published guidance is consistent across the FCDO and its peers: they confirm the death, attempt to trace and notify next of kin, liaise with local police, hospitals and the morgue, and provide a list of local funeral directors or international repatriation firms. They do not pay for any of it. The cost of repatriation or a local cremation or burial falls on the next of kin or the estate, and in practice the body is not released until someone has agreed to and usually fronted that cost.

That sequence is the part the headline number erases entirely. The dataset’s unit is a death. The family’s unit is an invoice arriving in a currency they do not hold, for a person in a mortuary in a country they may never have visited, often with no insurance because the deceased aged out of the market years earlier. Add the complications the guidance hints at but does not quantify: an estate that cannot be touched until probate while the morgue charges by the day, a death during a visa overstay, no traceable next of kin at all. None of that is counted anywhere. The figure for what an uninsured death abroad costs the family sits in the insurance cliff at 70, because it is the same failure viewed earlier. The decade when cover lapses is the decade this dataset’s missing cohort dies in, and the unfunded repatriation is the bill at the end of it.

Age exists in the data only when someone asks

There is one place the British data is broken down by age: FOI release 0064-17, covering British citizens who died in Thailand from 2014 to 2016 by age and circumstances of death. It exists. It is also the tell. Age-resolved death-abroad data is not a standing series anyone publishes; it is a freedom-of-information disclosure for three years and one country, released because a person specifically asked. The routine, recurring, public figures, the ones that become news headlines and country rankings, carry no age at all.

That asymmetry matters for this cohort more than any other. The entire question of aging abroad is an age question. A dataset that publishes counts continuously but age only on request is, in practice, a dataset that does not let you see the age structure of who is dying without doing the work yourself, one FOI at a time, never comparably across countries or years. The information is not secret. It is just not assembled, and the not-assembling falls hardest on the one variable that defines the subject. A reader cannot pull “deaths of British residents aged 70-plus in Thailand, by year” from anywhere, because nobody publishes the thing that would answer it.

And no denominator exists to turn even a perfect count into a rate. No government maintains a clean register of resident expatriates by age. Embassy registration is voluntary and sparse, visa categories do not map to permanent residence, and the long-stay retiree on serial extensions is counted differently in every system that touches them. So the deepest reason there is no expat death rate is not squeamishness or thin reporting. It is that both halves of a rate are missing at once: the numerator is scoped away from natural causes, and the denominator was never built. You cannot divide a censored count by a number that does not exist.

What would have to be true for the headline to mean what people think

For “around 650 a year, mostly accidents” to be a fair picture of dying abroad, one of two things would have to hold. Either Western expats abroad die overwhelmingly of non-natural causes, which is false for any aging population anywhere and there is no reason an expat cohort inverts ordinary mortality. Or the natural-cause deaths that are missing from the US table are captured, denominated, and published somewhere else. They are not. The FCDO and DFAT figures include them but only as undifferentiated caseload, with no cause split released, no age split released, and no resident population to divide by.

So neither condition holds, and the headline cannot carry the meaning it is routinely given. The honest reading is narrow and it is the artefact of this page: the most authoritative public number describes the least representative deaths, the more complete numbers are not rates, and the cohort this site exists for is the one all three are weakest at seeing. That is not a hedge bolted onto a conclusion. It is the conclusion.

The honest statement

No public dataset measures how likely a Western retiree is to die in Thailand or the Philippines. One dataset measures non-natural deaths precisely and excludes the aging cohort by law. Two more count consular cases without a denominator. All three are floors. The single signal that leaks through the design is DFAT’s own attribution of its rising caseload to more older Australians retiring in both countries. That is the data admitting, in the one place it can, that the invisible cohort is growing.

The number you will be quoted is 650, or 4,000, or some country ranking built on a caseload. Each is real. None is the rate, and the one that is mandated and exact is the one most carefully scoped away from the death this whole subject is about. Rebuild this the day a natural-cause, denominated dataset exists. None does, and the absence is not a gap in the reporting. It is the finding.

---

This piece cites deaths-abroad and suicide data and is analytical, not advice or counsel. If you or someone you know is struggling, contact a local crisis line or, internationally, findahelpline.com. Figures are sourced and dated to 2026 and label their own populations and limits; they are scoped case counts and floors, not measured death rates, and should be read as such. Verify any procedural or financial specifics with the relevant consular service and a licensed professional.