A 1973 start empties my pension. A 2008 start does not.

1973 was not a bad year to be alive in Britain. Sunderland, in the Second Division, beat Leeds 1–0 in the FA Cup final. Red Rum won his first Grand National, catching Crisp on the run-in. Pink Floyd released The Dark Side of the Moon in March and Mike Oldfield released Tubular Bells in May. Jackie Stewart won his third world title. Last of the Summer Wine started, the Wombles arrived on television, and Slade had the Christmas number one. On 1 January the country joined the Common Market.

It was also, if you had just retired on a pot of shares, the worst ten-year start in the UK data I hold. The oil shock arrived in October. UK shares fell about 28% that year and about 52% the next. Inflation ran 9%, 16%, 24%, 16% and 16% in the first five years. A £100 pot of shares was worth about £35 two years in, while the cost of living had gone up by a quarter.

My calculator can replay that decade as the first ten years of a plan, equity returns and inflation together, then go random. I had never run it properly on my own plan. This week I did, as part of building the scenario pack, and the result was 0 of 100.

The plan, and the two replays

My plan as it stands works to age 85 in 98 of 100 futures. It is not a reckless plan. Half the pot goes into a 20-year inflation-linked annuity at 62, there is a three-year cash buffer, spending steps down at 70 and 80, and a guardrail cuts spending by a quarter after any year the pot falls more than 2%.

Replay 2008 as the first decade (a 30% fall in year one, then a slow recovery) and it works in 100 of 100. The buffer and the annuity do their job.

Replay 1973 and it works in 0 of 100. Not 20, not 5. Every single future runs out before 85.

The year-by-year trace shows why, and it is not only the crash. By the fourth year the pot has fallen from about £980,000 to about £370,000. In the same four years, because the plan keeps spending in real terms, the money I need each year has gone from £60,000 to about £112,000. By 62 there is about £280,000 left to buy the annuity with, which buys £8,800 a year. The crash is half of it. The other half is five years of inflation that doubles what a year costs while the pot is on the floor.

What I tried, and what each bought

Same plan, same random draws, 2,000 futures, each defence on its own. "To 85" is how many of 100 futures pay every year to age 85 in ordinary random futures; "1973" is the same with the 1973 decade replayed first.

Read the first six rows and the lesson looks bleak: the levers that make an ordinary retirement safer do nothing against 1973. A bigger annuity bought at 62 is bought with a pot that has already fallen two thirds. A guardrail cuts spending by a quarter while inflation is adding a quarter a year. Retiring later moves the crash to a bigger pot. And a conventional gilt ladder, the thing I wrote about in September and had been planning to build, pays fixed pounds, so it loses to 24% inflation almost as badly as the shares do.

The index-linked ladder on its own barely moves it either, and that surprised me until I looked at why. It covers nine years of what I would need in today's money, net of the annuity and the State Pension. Under 1973 the need it is sized to keeps rising with that decade's inflation, and it carries the early years while the rest of the pot is falling, but it does not stop the rest of the pot being drained to cover the gap.

Pair it with spending 10% less, and the two together reach 87 of 100. The ladder keeps the first years whole in real terms; the lower spend stops the rest of the pot being sold while it is down. Add inflation-linked income bought on the day rather than four years later, and 1973 becomes an ordinary retirement: 100 of 100. None of these three is enough alone. Together they are the whole answer, in this model.

What index-linked gilts cost, and why they return less

An index-linked gilt pays a coupon and a redemption that are both uplifted by RPI. The price you pay for that is a real yield, set by the market, rather than a nominal one. On the Bank of England's curves for 14 October, a nine-year index-linked ladder for my plan yields about 1.5% a year above RPI, against 5.3% fixed for the conventional one. The gap is the market's view of inflation over those years: about 3.7% a year. If inflation averages less than that, the conventional ladder ends up paying more; if more, the index-linked one does. Only one of them has fixed spending power whatever happens.

In pounds today the index-linked version of my bridge costs £430,000 against £413,000. Not a big premium for the one thing that would have survived 1973.

What it costs if 1973 never comes

Insurance has a premium, and the fair question is what these three cost in an ordinary retirement. The same engine, random futures, no replay:

The ladder's premium is about one future in a hundred and roughly £100,000 less at 85: nine years of 1.5% above inflation instead of 5% on £474,000 carved out on the day. Spending £6,000 a year less costs exactly that, £6,000 a year of living, about £160,000 over the years to 85 in today's money, and it leaves more behind than the ladder takes. Buying the inflation-linked income at 58 rather than 62 costs nothing in the odds and a little in what is left. None of these is free. Whether the premium is worth paying for a decade that has happened once in the data is the kind of question a model can size and only a person can answer.

What I built

Two things, both live today.

The free gilt ladder calculator now has a switch between conventional and index-linked. Index-linked gilts are priced in today's money off the Bank's real curve, shown at the nominal price and face your broker will quote, with the other ladder's cost and the implied inflation rate beside it. The gilts in issue come from the Debt Management Office, the curves from the Bank of England, the RPI index from the ONS, refreshed every weekday.

The scenario pack, the paid part of the site, opens today. It runs your own plan through everything above and more: the levers, losing the job, a fall before the date, the annuity shapes, the 1973 and 2008 decades, care, one of you dying first, and your ladder in both kinds of gilt, each run through the engine so you can see what it buys. £49 for a year, no account, no login, and the data never leaves your browser.

Where the model is wrong, or at least thin

Four things to hold in mind. The replay applies the equity series to the whole invested pot, so a plan that already holds bonds gets no credit for them in those years; 1974 was a bad year for gilts too, but not that bad. The replay is ten years and then random, so a 1973 start is one path, not a distribution. Index-linked gilts did not exist in 1973 (the first came in 1981), so "what would have worked" is a model's answer, not history's. And from February 2030, RPI will be calculated the same way as CPIH, which has run about a point a year lower, so a rung maturing after then is uplifted by less than today's RPI suggests.

What I take from it is narrower than "buy index-linked gilts". It is that the two defences everyone reaches for, a bigger annuity and a cash buffer, protect against a crash and not against a crash with inflation on top, and that the thing which does is inflation-linked income plus the willingness to spend a little less. 1973 is the worst start in the data. It is also the one that makes you check what your plan is actually protected against.

To replay 1973 on your own plan in the free calculator: on the Assumptions step, choose the historical scenario "1973", and compare the result with "none". The gilt ladder page sizes the index-linked bridge; the pack runs it through your plan.

Sources: FTSE All-Share total returns and RPI, Barclays Equity Gilt Study and ONS (series MM23), 1973 to 1982, as held in the calculator; Bank of England nominal, real and implied-inflation spot curves, 14 October 2026; DMO gilts in issue, 14 October 2026; ONS RPI (CHAW) to August 2026. All Open Government Licence. The 1973 events are from memory and general reference; if I have one wrong, tell me and I will fix it.