The Money Doesn’t Last

Here is the sentence the brochure is built on. You can retire in Thailand on $1,500 a month. It may even be true. The defect is not the number. The defect is the tense. It is a present-tense observation about one month in 2026, dressed as a description of the next twenty-five years, and the two are not the same statement. One is a snapshot. The other is a trajectory. The entire relocation pitch depends on you not noticing that it quietly swapped the first for the second.

A snapshot has no medical inflation in it. No currency drift. No insurance renewal at seventy. No care tail. Photograph a budget on the day it works and it will always look like it works, the way a photograph of someone at forty is an accurate and completely misleading guide to how they will look at eighty. This piece does the thing the pitch refuses to do. It runs the budget forward, with every assumption written down and open to challenge, until it reaches the cell where the margin is zero. That cell has a date. The name of this site is what is in it.

The model, stated plainly

A model is only honest if you can see its assumptions and argue with them. So here are all of them, before any output.

The case. A single retiree, age 65, retiring to Chiang Mai in 2026. Income in pounds, spending in baht. Two income components: a UK State Pension, and a private pot of £300,000 drawn down. Modest, mid-market, not poverty and not comfort. The case nobody photographs because nothing about it is dramatic until the arithmetic finishes.

Income, component one: the State Pension. Take it at the full new rate, about £241 a week, near £12,550 a year, for 2026/27. The structural fact that matters is not the amount. It is the uprating rule. The UK State Pension is frozen, paid forever at the rate you first received with no annual increase, if you live in a country with no reciprocal social-security agreement. That April the full rate rose 4.8% under the triple lock; a pensioner living in Thailand received none of it, and never will. The government’s own guidance lists who is affected. Thailand is a frozen country. The Philippines is not. Hold that. It returns as the single most violent lever in the whole model — and almost nobody buying a Thailand retirement has priced it.

Income, component two: the drawdown. Start at the 4% rule. William Bengen derived it in 1994; the Trinity study reinforced it. Both were built on a US portfolio, spending US dollars, over a 30-year horizon. £300,000 at 4% is £12,000 in year one, and the rule lets you raise that withdrawal with home-country inflation each year. Even the safe starting figure moves with the market’s mood: Morningstar cut it to roughly 3.3% in 2021 on stretched valuations and sequence-of-returns risk, then raised it back to 3.7% by 2024 and 3.9% by 2026 as return expectations recovered. The base run uses 4%. The adverse run uses 3.3%: Morningstar’s own low-water mark, now below even its current figure, used here as a deliberately conservative floor rather than a pessimistic invention.

Spending. A single Western retiree in Chiang Mai, all-in (housing, food, transport, and a baseline private health policy), clusters around THB 60,000 to 90,000 a month across the cost-of-living aggregators, before any health shock. Treat that as order of magnitude, not precision; the aggregators are crowd-sourced and the standard is contested. The model takes THB 75,000 a month as the year-one figure and splits it: a health slice and everything else. They do not inflate at the same rate. That split is where the plan dies.

The two inflations. Everything-else compounds at general inflation, call it 2%. The health slice compounds at the medical trend, which is a different and much steeper number. Aon’s 2026 Global Medical Trend Rates report puts Asia-Pacific around 11.3%. Willis Towers Watson’s survey runs higher, into the 14% region for parts of the region. Mercer’s is in between. Take 11% for the base run, 14% for the adverse. The exact figure barely matters. What matters is that one slice of your spending is growing five to seven times faster than your income can — and that slice is the one that grows as you age.

The cliff. Private international cover surcharges hard, excludes more, or refuses new applicants from around age 70, with the worst renewal steps at the 70 and 75 bands. The model treats this as a step, not a slope: at age 70 the health line takes a one-time jump and then keeps compounding at the medical-trend rate from the new, higher base. The companion essay on the geographic cure describes the same asymmetry in human terms. Here it is just a step function with a cruel argument value.

Currency. This is the input I will not pretend to forecast, and saying so is part of the model’s honesty. But it does not have to be guessed at blind, because the recent history is on the record. Sterling has bought fewer baht each decade: around 70 to the pound before the 2008 crisis, into the low-to-mid forties through the Brexit years, and roughly 43 in 2026.

Date	THB per GBP	Range	Basis	Note
2007	70	65–75	triangulated	pre-2008 sterling peak (GBP ≈ $2.00, USD/THB ≈ 33–36)
2016	45	40–45	triangulated	settled into the ~40–45 band after the Brexit vote
2026-05	42.98	41.69–44.3	sourced	May spot; 2026 range 41.69 (21 Jan) to 44.30 (4 May). Re-read 24 May ~43.9, within range; value held at the representative mid-May figure.

Source: Bank of England spot series (via poundsterlinglive GBP-THB history) · latest 42.98 THB per GBP (2026-05) · as of 2026-05-24

That is a fall of tens of per cent across the very window the brochure quietly assumes is flat. The dollar tells a different story — USD/THB has stayed broadly range-bound near 32 to 33 (32.68 in May 2026, FRED series DEXTHUS), which is exactly why a US-Social-Security retiree and a sterling pensioner are running different models in the same town. Nobody can honestly forecast the next 25 years of either, and anyone who tells you the direction is selling something. So FX is run as a scenario band, not a number: 0% real drift in the base case, minus 2% a year in the adverse, plus 1% in the benign. You are free to reject all three. The model is built so you can.

The care tail. From the point a dependency event occurs, a care line is added. In Thailand a full-time private nurse is commonly quoted near THB 37,500 a month; facilities range from about THB 30,000 to over THB 100,000 depending on need. The benign run never triggers it. The base run triggers a moderate version in the early eighties. The adverse run triggers it at 80. Note one thing before the output: that line, on its own, is roughly the size of the entire starting income.

That is the whole machine. Eight inputs, all sourced, all arguable. Now turn it on.

What it outputs: three runs

The model does one calculation per year. Income in baht, after the FX assumption. Spending in baht, with the two slices inflating at their two rates and the cliff step at 70 and the care line when triggered. Income minus spending is the year’s surplus or deficit. Deficits are drawn from the pot. The output is the age at which the pot hits zero and income no longer covers spending. The age at which there is no margin left. Not a metaphor. A row.

Plotted as the pot over time, the three runs are a cone, not a line. That is the honest shape of a projection: an envelope of outcomes, with a floor where the adverse path hits zero years before the actuarial table says it should.

Source: Model output; inputs and sources in §Contest the model · checked 2026-05

The benign run

4% rule holds. Medical trend moderates to 8% as global pressure eases. FX moves a gentle 1% a year in the retiree’s favour. No early care event; the tail arrives late and brief. Under these assumptions the plan substantially works. The pot thins but the income, helped by currency and a softer medical curve, mostly keeps pace, and the margin survives into the early nineties, around the edge of the actuarial table. This is the run the brochure is implicitly selling. It is internally consistent. It is also the run in which every uncertain variable breaks the right way at once, which is not a forecast. It is a wish with arithmetic attached.

The base run

4% rule. Medical trend 11% (Aon, 2026). FX flat in real terms. The cliff step lands at 70. A moderate care event in the early eighties. Nothing here is pessimistic. Every input is the central, mainstream figure.

One mechanism deserves naming before the run, because it is the part the 4% rule’s own authors flagged and the brochure never repeats. It is called sequence-of-returns risk. Two retirees with the identical average return over 25 years can end very differently depending only on when the bad years fall. A market drop in the first few years of drawdown is structurally far more damaging than the same drop later, because you are selling more units to fund the same withdrawal while the pot is largest, and those units never recover. Morningstar’s reassessment leaned on exactly this to cut the safe rate. Now layer the medical trend on top: abroad, the early bad years are not only a market risk, they are also the years the health slice begins compounding. The relocation pitch implicitly assumes average returns arriving smoothly. The model does not get to assume that, because retirements are not lived in averages. They are lived in order.

The first decade looks fine, which is exactly the problem, because it is the decade the photograph is taken in. Spending and income track within a manageable deficit, covered comfortably by the pot. Then the medical slice, compounding in the low teens against a frozen pension and a 2%-indexed drawdown, stops being a slice and starts being the budget. The cliff step at 70 resets it higher. By the late seventies the annual deficit is no longer a trim off the pot. It is eating it. The care line arrives into an account already running down. In the base run the margin reaches zero in the early-to-mid eighties, inside the period a healthy 65-year-old today is actuarially expected to still be alive. The plan does not fail because anything went wrong. It fails because nothing did — and the central case was always insolvent on a long enough row.

The adverse run

3.3% start, deliberately below even the 3.9% Morningstar’s current work supports. Medical trend 14% (the Willis Towers Watson end). FX minus 2% a year as sterling drifts against the baht. The cliff at 70 hardens into partial refusal, pushing more cost fully out of pocket. The care event at 80.

None of these is a tail risk. The medical and currency figures are each a documented central estimate from a different reputable source, and the draw is set below current guidance, not above it; the adverse run is just what happens when they co-occur, which they are not forbidden from doing. Here the deficit turns structural in the early seventies. The pot, started lower and drawn into a falling exchange rate, is gone within a few years of the cliff. The margin reaches zero in the mid-to-late seventies. The person is, in the actuarial sense, very likely still alive, in a country selected on the assumption that the money would not do this, with the scaffolding already gone for the reasons the companion essay sets out. This is not the worst case. The worst case has an early diagnosis in it, and the model does not need the worst case to make the point.

The base run on one page

Prose hides arithmetic, so here is the base run as a table. It is rounded, it is illustrative, and that is the point: the structure is what travels, not the decimals. Income is the frozen State Pension plus the rule’s inflation-uprated drawdown. Spend is split into the everything-else slice growing at 2% and the health slice growing at 11% with a one-time step at the age-70 cliff. The care line is added from the early eighties. The pot column isolates the relocation-specific drain: it assumes the £300,000 funds the rule’s own drawdown by construction, and tracks only the additional principal consumed by the gap medical trend, the cliff and the care tail open up. State that assumption out loud and you can argue with it. That is more than the brochure offers.

Age (year)	Income £/yr	Spend £/yr	of which health	Annual gap	Pot remaining
65 (2026)	24,000	21,000	6,000	+3,000 surplus	300,000
70 (2031)	25,000	31,700	15,200 (cliff step)	−6,500	~300,000
75 (2036)	26,600	43,800	25,600	−17,200	~231,000
80 (2041)	28,100	63,200	43,100	−35,100	~101,000
82 (2043)	~28,700	~78,000	~53,000	~−49,000	~12,000
85 (2046)	29,800	117,000	72,600	−87,000	0, insolvent

Source: Model output; inputs and sources in §Contest the model · checked 2026-05

Read the health column down. It starts as a fifth of the budget and ends as the budget. Read the income column: it barely moves, because the largest component is frozen by treaty and the rest only tracks home inflation. The two columns are running at different speeds on purpose, and the gap between them is not noise. It is the model’s entire content.

Source: Model output; inputs in §Contest the model · checked 2026-05

Around age 82 the pot crosses zero. After that the gap does not stop. It just has nothing left to draw on, and the retiree is, on the actuarial table, expected to live years beyond that row.

Note what is not in this table to make it look bad. No market crash. No early dementia diagnosis. No scam, no divorce, no failed business, none of the things the persona channels treat as the only ways it goes wrong. The base run is the version where nothing goes wrong. The failure is endogenous. It was in the snapshot the day it was photographed; it just had not finished compounding yet.

What the calculators show you instead

There is an entire genre of retire-abroad calculator, and it is worth being precise about what it does. It takes today’s rent, today’s price of a restaurant meal, today’s cost of a domestic helper, sums them, and returns a comfortable monthly figure. Every input is a present-tense price. Not one of them is a 25-year rate of change. The tool is structurally a snapshot generator that calls itself a planner.

It is also subject to the same survivorship the rest of this market is. The cost-of-living aggregators are populated by people currently abroad and currently coping; the retiree who ran out and went home does not update his entry. So the crowd-sourced baseline is drawn from the surviving cohort, at the front of the curve, which is the exact bias the companion essay documents in the happiness numbers. A median built only from people for whom it is still working is not a median of the decision. It is a median of the winners.

None of those tools contains a medical-trend field. None has an insurance-cliff step. None asks which country you are in for State Pension uprating purposes, which this model shows is the highest-leverage variable of all. They are not lying. They are answering a smaller question accurately and letting the reader mistake it for the larger one. This piece exists to ask the larger one.

The one variable nobody priced: the country itself

Now the result the model exists to produce. Run the base case twice. Change exactly one thing. Not the pot, not the spend, not the medical trend. The country.

In Thailand the State Pension is frozen. In the Philippines, under the reciprocal agreement, it uprates each year. Everything else identical. Roughly 2% annual uprating, compounding across a 25-year retirement, is not a rounding error. It is the difference between an income that is being slowly euthanised by inflation and one that is, weakly but really, defending itself. In the model the Philippines run reaches zero years later than the Thailand run on identical money, identical health, identical everything, because of a checkbox on a Department for Work and Pensions list that the relocation pitch never mentions because it is not on the brochure and it is not on the beach.

That is the counter-intuitive finding, and it is the moat. Every retire-abroad calculator on the internet will optimise the variables it shows you: rent, a meal out, a beer, a maid. The variable with the most leverage over whether you run out of money before you run out of life is one almost none of them model, because it is not a price. It is a treaty status. The market sells you the snapshot and never shows you the switch.

This is not financial advice, and the uprating rules can themselves change; verify your own position with the DWP and a licensed adviser before treating any of it as settled. The point is structural, not prescriptive. The most important number in your retirement-abroad plan may not be a number you have ever been shown.

What would have to be true for it to be fine

The model is not an argument that nobody should do this. It is an argument against doing it on the snapshot. There is a minority for whom the runs come out solvent, and they are specific. Name them honestly.

The plan survives if the pot is materially larger than the 4%-rule minimum, so the drawdown is a low single-digit percentage with room to absorb the medical curve. It survives if the income is genuinely inflation-linked rather than nominally frozen: a public-service or defined-benefit pension with real uprating, not a State Pension stranded in a non-agreement country. It survives better in an uprating country than a frozen one, which is the cheapest large improvement available and costs only a different visa queue. It survives if the spending currency and the income currency are the same, removing the FX leg entirely, which is structurally why a US-Social-Security retiree spending in a dollarised cost base faces a different model than a sterling pensioner in Thailand. And it survives if the care tail is a funded line, provisioned for, not a thing hoped against.

Those people exist. They are the well-capitalised, the holders of real inflation-linked pensions, the currency-matched, the ones who moved toward a funded plan instead of away from a cold country. For them the geographic move is a lifestyle decision with the arithmetic underneath it actually checked, which is what running the pre-move reality check in advance establishes. That is allowed. It is also not who the pitch is mostly sold to.

For everyone else the model does not say don’t. It says something colder and more useful. It says: here is your date. It is a specific year, derivable from inputs you can look up this afternoon, and it arrives whether or not you have looked at it. The only choice the model leaves you is whether you find out the year now, while you can still act on it, or in the year itself, when you can’t. The brochure is a photograph. This is the negative, and it was always in the same envelope.

Contest the model

The model’s credibility is not that it is right. It is that you can see exactly where it is wrong for you. Here is every input, its source and date, and which way the failure year moves if you change it. Argue with the column on the right. That is the document working as intended.

• Drawdown rate, 4% base / 3.3% adverse. Source: Bengen (1994); Trinity study; Morningstar’s State of Retirement Income series (3.3% in 2021, 3.7% in 2024, 3.9% in 2026). A bigger pot or a lower rate pushes the failure year out, possibly off the table entirely. A higher rate or a smaller pot pulls it forward fast. This is the input with the most legitimate dispute attached, and the literature itself keeps moving: 3.3% then, 3.9% now, which is why the adverse draw is a deliberately conservative floor rather than a current estimate.
• Medical trend, 11% base / 14% adverse. Source: Aon 2026 Global Medical Trend Rates; Willis Towers Watson; Mercer, all Asia-Pacific, 2026. This is the engine. Halve it and the plan often survives; the brochure implicitly assumes it equals general inflation, which no medical-trend survey in the region has ever reported. If you change one number, change this one and watch the year move.
• State Pension uprating. Source: gov.uk guidance on the State Pension if you retire abroad, current 2026. Binary by country and the single highest-leverage lever here. Frozen pulls the year in; uprated pushes it out by years on otherwise identical inputs. Confirm your own country’s status directly; the list is the government’s, not ours, and it can change.
• Currency, 0% / −2% / +1% real drift. Anchor: USD/THB ≈ 32.7, May 2026 (FRED DEXTHUS); GBP/THB ≈ 43 in 2026, down from about 70 before 2008 (ledger above). Run as a band, not a forecast, because nobody can honestly forecast it. If your income and spending are the same currency this input is zero and the model is materially kinder. If they are not, the adverse band is not pessimism. It is one of the two directions the rate has historically taken.
• Spending baseline, THB 75,000/month. Source: cost-of-living aggregators, cross-referenced, 2026, treated as order of magnitude. Lower it and you buy a few years; the health slice still eventually dominates regardless, because it compounds and the rest does not.
• The cliff step and the care tail. Source: international-insurer age-banding behaviour around 70 and 75; Thai and Philippine care-cost ranges, 2026. Move the cliff age or the care-onset age and you move the failure year directly. Remove the care line entirely and the base run still fails, just later. The tail is what makes it brutal, not what makes it true.

Change any of these and rebuild it. If your rebuilt year is comfortably past your actuarial life expectancy with the adverse band switched on, the plan is real. If it is not, you have learned the most useful thing this site can tell you, and you have learned it while there is still time to use it.

An illustrative model, not a recommendation or a forecast of any individual outcome. Inputs are sourced and dated to 2026 and will drift; rebuild it with your own figures and confirm them with a licensed financial adviser regulated in your jurisdiction before acting.