Survivorship Bias in Expat YouTube

The channels you are watching to decide whether to move are made, without exception, by people for whom the move has not yet failed. This is not a criticism of any of them. It is a property of the medium. A retire-abroad channel requires a living, solvent, present person to film it, and that requirement quietly removes, before you ever press play, everyone the move broke. The retiree who ran out of money does not post a final video explaining the arithmetic. The one who got the diagnosis does not narrate the decline. The one who flew home, defeated and quiet, does not upload a closing episode titled Why I Left. They simply stop appearing, and the feed closes over the gap as though they were never there.

So the genre is a sample, and the sample is selected on survival before a single frame is edited. That is a specific, named statistical error — and it has a body count attached.

The missing denominator

The error is survivorship bias, and the cleanest version of it is a wartime story worth getting exactly right because the shape recurs everywhere. During the Second World War the US ran a Statistical Research Group, and the military brought it a problem: bombers were coming back riddled, armour was heavy, and they wanted to know where to put it. The obvious answer was to armour the places the returning planes were most hit. The statistician Abraham Wald gave the counterintuitive one. Armour the places the returning planes were not hit. The planes shot through the engines were not in the sample, because planes shot through the engines did not come back. The damage you could see was, by definition, the survivable damage. The information that mattered was carried entirely by the aircraft that were absent.

Finance learned the same lesson with a number on it. A mutual-fund performance database flatters itself automatically, because the funds that do badly are quietly liquidated or merged out of existence and vanish from the record. Strip the dead funds out and the survivors’ average return rises. The measured inflation is not trivial: over the decade to 2003 the Center for Research in Security Prices found the surviving US stock funds had returned 8.8% a year against the 7.2% you get counting the funds that no longer exist — about 1.6 percentage points a year, conjured purely by deleting the dead. The funds did not get better. The losers left the building.

Now apply the structure to the thing you are actually doing, which is forming a picture of late-life relocation from a feed. The feed has a numerator and no denominator. It shows you the people for whom it is working, the numerator, and it has no mechanism whatsoever to show you the denominator, meaning everyone who attempted the same move, including the ones who left and the ones who died. That denominator is not unknowable. It is partly measured, elsewhere, by people not in the business of selling the move: roughly 9% of Dutch pensioners abroad returned home within three years, against fewer than 5% who said they intended to, and a cohort-decay model built on that hazard and the ordinary mortality table leaves about eighteen of every hundred arrivals still living in place at eighty-five. The other eighty-two went home or died. None of them is making content.

Why the camera points one way

Three separate forces aim the lens at the front of the curve, and it is worth keeping them distinct, because each is enough on its own and together they are nearly total.

The first is the one already described: construction. A channel cannot be made by a person who is not there. Survival is a precondition of production, so the sample is filtered on the outcome before any choice about what to film is made.

The second is money, and this is the one the genre’s defenders skip. It is sometimes said that creators could show the hard parts if they were honest enough. The platform makes that expensive. YouTube’s advertiser-friendly content guidelines limit or remove ad revenue on a list of categories that reads, almost exactly, like the contents of a failed retirement abroad. “Controversial issues”, a category whose own examples are suicide, self-harm, eating disorders and domestic abuse. “Shocking” content, content that may “upset, disgust, or shock”. A walking tour of a market and a clean apartment is fully monetisable. A frank account of drinking alone through the long afternoons, a marriage going silent in a country with no one else in it, the slow recognition that the money will not last, the decision to go home beaten — that is the demonetised list. Nobody has to suppress the failure story. The platform simply declines to pay for it, and pays well for its opposite, and creators do more of what they are paid for. The result is not a lie. It is an economy that prices the honeymoon and docks the obituary.

The third force operates even with no money involved at all. People present themselves positively as a default. A 2025 registered experiment across Facebook, Instagram and X confirmed what the prior literature already suggested: users craft and share more positive than negative self-presentations, the edited self over the actual one. Strip out the algorithm and the ad revenue entirely and you would still get a feed tilted toward the good day, because that is how people hold a camera. The monetisation rule and the positivity reflex push in the same direction the construction filter already points, and the three of them compound.

What none of the three is, anywhere, is a filter that randomly drops cases. They drop a specific kind of case. They drop the ones it went wrong for.

The dataset with the deaths deleted

This last point is one the genre structurally cannot make, because it is a statement about the genre rather than anything that fits inside a frame.

The camera is pointed at the winners because the losers are, definitionally, off-camera. Not by editorial choice, not by conspiracy, but by the same logic that kept the engine-shot bombers out of Wald’s inspection hangar. The retiree whose plan failed is removed from the visible sample by precisely the event a prospective mover most needs to see, and removed because of it. The man who went home broke at seventy-eight is not in the feed for the same reason his money is not in the bank. The woman whose husband died and who could not stay is absent from the cost-of-living spreadsheet for the same reason she is absent from the country. And the hundreds of Western nationals who die in Thailand each year, in numbers that rise mechanically as the cohort ages, are the most absent of all. They did not move home, and they will not be updating a channel.

You are not watching a representative account of late-life relocation. You are watching the output of the survivors, sampled at the front of a curve that bends down hard later, with every adverse outcome silently removed from the denominator before the query ran. And you are using it to plan your last two decades. You are planning them off a dataset with the deaths deleted.

That is the entire mechanism, and it runs in one direction only. There is no equal and opposite genre of channels made by the failures, because failure, here, is the precise condition that ends the channel. Paradise is the place you visit, and the footage is shot by the people for whom it is still a visit.

What the survivors cannot tell you, even sincerely

The honest version of this is not that the creators are dishonest. Most are showing you something true about the day they filmed. The error is not in any video. It is in the inference the viewer draws across them.

A man at three years in, filming his pleasant morning, is reporting accurately on a man at three years in. He cannot report on himself at year twelve, because he has not been there, and the second-year accounts and the drawdown that reaches its zero year are not things he can pre-experience and narrate for you. By the time the late shocks arrive, the widowhood, the disabling diagnosis, the fixed income hollowed out by twenty years of medical inflation and currency drift, he will either still be coping, in which case the channel continues and the genre’s tilt holds, or he will not be, in which case the channel ends and he leaves the sample. There is no third path where the channel documents its own author’s defeat at length and stays monetised and active. The medium has no way to file that report. The structure of the sales pitch and the structure of the medium that appears to scrutinise it are, in the end, the same structure: both show the front of the asymmetry and neither shows the back, because the back arrives after the audience, and the author, have stopped watching.

So the sincerity is real and it changes nothing. A thousand accurate reports from people for whom it is currently working still does not contain the one number a person deciding actually needs, which is the fraction for whom it does not, measured over the length of a real retirement rather than the length of a honeymoon. That number is in the denominator. The denominator is the part the feed was built, by construction and by economics and by the ordinary vanity of self-presentation, never to contain.

What would have to be true

For the feed to be a usable guide to the decision, one of three things would have to hold. The people the move broke would have to keep filming through the breaking, against the construction filter that removes them and the monetisation rule that defunds them. Or the genre would have to be balanced by an equal volume of content from the return-and-death cohort, which cannot exist because that cohort is, definitionally, gone from the place the content is made. Or the viewer would have to mentally reinsert the missing denominator while watching — to look at every cheerful tour and supply, unprompted, the eighteen-in-a-hundred-at-eighty-five, the engine-shot bombers that never reached the hangar, the funds that left the database.

None of the three happens. The first two are structurally impossible and the third is not how anyone watches anything. So the feed remains what it is, which is the survivors, talking, at the front of the curve, sincerely, for as long as it is still working for them. The information you need is carried by the people who are not in it, and they are not in it because of what the move did to them. The brochure at least admits it is selling. The genre that looks like the brochure’s critic is selected by the same forces and deletes the same people, and the deletion is the part you cannot see, which is why it works.

---

This piece treats expat and retire-abroad media as a genre and a sample; it does not name or disparage any specific creator or channel, and quotes none. The return and deaths-abroad figures are sourced and dated, label their own populations, and are bounds and proxies, not a measured rate. This article is analysis, not financial, medical, immigration or legal advice; verify anything actionable with a licensed professional.