Key takeaways
- A single person aged 55 aiming at the Moderate Retirement Living Standard of £32,700 a year needs £20,152 a year on top of the full new State Pension.
- Priced on the Bank of England's real yield curve for 25 August 2026, that income stream costs £280,345 to match with index-linked gilts.
- The same stream cost £856,954 at the end of 2021, when the 25 year real yield was minus 2.35%. The price fell 67.3% in under five years.
- At the Minimum standard the shortfall is £1,352 a year, and the matched holding is £18,814, or 6.7% of the same pot rather than 100%.
- UK defined benefit schemes cut equities from 61.1% of assets in 2006 to 15.1% in 2025, and now hold 70.6% in bonds.
Liability matching starts with a price, not a percentage
You want to know how much of your portfolio belongs in bonds. The familiar answers arrive as percentages: 100 minus your age, or a glide path that lands somewhere near it by retirement.
Liability matching answers in pounds instead. It asks what income you have promised yourself, prices that promise at today's real yields, and calls the answer the matched holding. Whatever sits above that number is the growth portfolio. The percentage is an output of the sum, not the input to it.
Here is that sum on one household, using figures published this year. Take a single person who turns 55 in 2026. Someone born between 6 March 1961 and 5 April 1977 reaches State Pension age at 67, so the income has to start there. Pensions UK put the Moderate Retirement Living Standard at £32,700 a year for a one-person household in its 3 June 2026 update, and the Minimum at £13,900. The full new State Pension is £241.30 a week in 2026/27, which annualises at 52 weeks to £12,548. The gap between the Moderate standard and the State Pension is £20,152 a year, in today's money, for as long as the person lives.
Discount that stream on the Bank of England's implied real spot curve for 25 August 2026, one rate for each year of the payment, from age 67 to age 90, and it prices at £280,345. That is the liability. It is also, on this method, the entire index-linked allocation. No percentage was consulted to produce it.
The same retirement income cost three times as much four years ago
The number moves, and it moves violently, for reasons that have nothing to do with the household.
Run the identical cash flows through the Bank of England's real curve for 31 December 2021 and the same promise prices at £856,954. Real yields were negative across the curve then: minus 3.26% at five years, minus 2.35% at 25 years. On 25 August 2026 the same points read plus 1.03% and plus 2.51%. The cost of the promise fell 67.3%, a factor of 3.06, in under five years.
Nothing about the person changed. Their age moved by four years, their spending plan by an inflation adjustment. What moved was the price of buying certainty, and no age-based rule contains a term for it. A 55-year-old following 100 minus age held 55% in bonds at the end of 2021 and 55% in bonds today, through a period in which the thing those bonds exist to fund got two thirds cheaper.
That is the case for liability matching in one line. It does not claim to earn more. It claims to measure the right thing, and the right thing has a price that moves.
Three households, one age, three different answers
Hold the age fixed at 55 and the pot fixed at £280,345, and change only the spending plan. Every payment runs from 67 to 90 and is discounted on the same 25 August 2026 curve.
| Spending target | Shortfall after the State Pension | Cost to match | Matched share of the pot | 100 minus age says |
|---|---|---|---|---|
| Minimum, £13,900 a year | £1,352 a year | £18,814 | 6.7% | 55% |
| Moderate, £32,700 a year | £20,152 a year | £280,345 | 100% | 55% |
| Comfortable, £45,400 a year | £32,852 a year | £457,018 | 163% | 55% |
The chart above plots the same three bills on one axis, which is the clearest way to see how little the age of the household has to do with them. Three people of identical age, identical wealth and identical health get 6.7%, 100% and an answer that does not exist. The age rule gives all three 55%. The evidence behind the 100 minus age rule is worth reading on its own terms, but the arithmetic above is the structural objection to it: age is not in the liability.
The third row is the useful one. A pot of £280,345 cannot match a Comfortable standard at these yields, because the matching holding would be 163% of it. The liability calculation does not fail quietly there. It returns a number larger than the assets, which is the same information a pension scheme reads as a funding ratio below 100%. What follows from that is a question about spending, retirement date or contributions, not about the bond weight.
The cost of the same promise rises with age because the discounting runs out
Age does enter the sum, but through the back door, and not as a risk preference. The nearer the first payment, the less discounting stands between you and it.
| Age now | Years to the first payment | Cost to match the Moderate shortfall |
|---|---|---|
| 50 | 17 | £248,696 |
| 55 | 12 | £280,345 |
| 60 | 7 | £320,921 |
The table stops at 60 because the Bank publishes its real spot curve from 2.5 years out, and a 65-year-old's first payment falls inside that. The bill rises with age here, which looks superficially like a glide path. It isn't one. The rising number is the cost of the same promise, not a growing appetite for safety, and it rises whether or not the household has the money. An age rule and a liability calculation can agree on direction while disagreeing completely on what they are measuring.
UK pension schemes rebuilt their portfolios around liabilities, not horizons
This is not a thought experiment. It is what the UK's defined benefit schemes did, in public, over nearly two decades, and the Pension Protection Fund counts it every year.
In The Purple Book 2025, covering 4,838 schemes as at 31 March 2025, the weighted average equity allocation was 15.1% and the bond allocation 70.6%. In the first Purple Book, at 31 March 2006, the equivalent figures were 61.1% in equities and 28.3% in bonds. Within the bond holding, index-linked bonds were the largest single category at 44.4%, and 12.9% of all assets sat in annuities, a record high.
The cut runs with liability maturity, not with the age of anyone involved. Schemes still open to new members held 42.1% in equities. Schemes closed to new benefit accrual held 8.8%. Sorted by how much of the liability belongs to people already drawing a pension, schemes under 25% pensioner liabilities held 18.7% in equities and those between 75% and 100% held 6.5%. Same regulator, same tax treatment, same markets. Different promises outstanding.
The funding ratio moved while the portfolio fell
Liability driven investing changes what counts as a good month, and the PPF's monthly index shows it plainly. In the month to 31 July 2026, total scheme assets fell 1.7%. Liabilities fell 3.1%. The aggregate surplus rose to £271.3bn and the funding ratio reached 133.0%. The index explains why both sides move together: s179 liabilities are "sensitive to the yields available on a range of fixed-interest and index-linked gilts", which are the same yields that price the bonds the schemes hold.
A household that measured only the left-hand side of that would have recorded a bad month. The wider version is starker. Between 31 March 2021 and 31 March 2022, scheme assets fell from £1,720.7bn to £1,666.9bn while s179 liabilities fell from £1,673.8bn to £1,473.9bn, and the aggregate funding ratio rose from 102.8% to 113.1%. Assets down, position better. That only makes sense if the liability is on the page. Portfolio trackers, this one included, show the assets continuously and the liability not at all, which is a gap worth knowing about rather than a reason to ignore either number.
The strongest objection is that matching is what broke in 2022
The case against runs like this: the UK runs liability matching at national scale, and in September 2022 it required an emergency central bank intervention. That is a fair description of what happened.
After 23 September 2022, nominal gilt yields "across the medium and long maturity spectrum rose by more than 100 bps" in a matter of days, which Gabor Pinter's Bank of England staff working paper calls "unprecedented in recent economic history of the UK". The paper is research rather than Bank policy, and it says so on its cover. Leveraged liability driven investment funds faced collateral calls and sold into the fall. Pinter puts net gilt sales by the LDI, pension and insurance sector at over £36 billion between 23 September and 14 October, with three firms accounting for over 70% of sales to primary dealers. Sir Jon Cunliffe told the Treasury Committee on 18 October that "the five largest daily moves in the 30 year inflation-linked gilt, in data that dates back to 2000, have all been since the 23 September".
Two things separate that from the arithmetic above, and one does not. The separations are leverage and pooling: those funds ran what the paper calls "highly leveraged" gilt positions financed in the repo and swap markets, and a household buying index-linked gilts outright posts no collateral and receives no margin call. The part that does not separate is the mark. An unlevered matched holding fell hard in price too: the same 24 payments, revalued on the curve of 27 September 2022, priced at £333,302 against £856,954 nine months earlier, a fall of 61.1%. It was only a hedge for anyone who valued the liability on the same curve on the same day, and it was not a hedge at all for someone who needed to sell that year for a reason unrelated to the plan.
What locking in a real yield gives up
A matched holding at these prices earns about 2.5% real and no more. The Dimson, Marsh and Staunton database behind the UBS Global Investment Returns Yearbook 2026 puts the US real equity return at 6.6% a year from 1900 to 2025, against 1.6% for bonds. Matching the liability converts an expected 6.6% into a contracted rate close to the 2.51% the curve offered at 25 years, and the difference over 24 years of payments is the honest cost of removing the uncertainty.
The Yearbook is also the best evidence that the choice is not binary. Since 1900, equities and bonds have each lost more than 70% in real terms on several occasions, and a 60:40 blend has never declined more than 50%. Partial matching is available at every point in between, which is what the middle row of the first table describes and what our note on how much you need to retire takes from the spending side.
What this arithmetic cannot tell you
Start with the index. The Bank of England’s 2001 method paper describes "the retail prices index (RPI), to which IGs are indexed", and states that "estimates of the real yield curve are derived from the prices of index-linked gilts". Every figure here is therefore a real yield relative to RPI. The UK Statistics Authority has said the change aligning RPI with CPIH "can legally and practically be made by the Authority in February 2030". The Retirement Living Standards basket is not RPI either. A matched portfolio hedges a promise indexed one way against spending indexed another, and the residual is real.
Then the horizon. Paying to 90 is a choice, not a fact. ONS period life tables for 2022 to 2024 put life expectancy at 65 at 18.7 years for men and 21.2 for women, so the mean lands closer to 84 and 86. Planning to a mean leaves roughly half the outcomes unfunded, and stretching the horizon raises every figure in both tables.
Then the day. One curve, one date. The 67.3% swing between 2021 and 2026 is the measure of how little a single day's pricing describes anything permanent. The Retirement Living Standards exclude housing costs entirely, so a household with a mortgage or rent in retirement carries a liability that appears nowhere above. Tax is ignored throughout, as are annuities, which price the same promise with longevity pooling included and are covered in our piece on annuity vs drawdown. And the DB schemes cited here match with an employer covenant behind them, professional hedging desks and no capital gains tax, none of which a household has.
What would change the conclusion
If real yields returned to where they sat at the end of 2021, the entire method inverts. At minus 2.35%, matching the Moderate standard costs £856,954 rather than £280,345, and the answer for most households becomes that the liability is unfundable at any allocation. Liability matching does not tell you to buy the hedge. It tells you what the hedge costs, and at some prices the reply is no.
If your spending is genuinely flexible, the liability shrinks to whatever floor you would defend in a bad decade, and the matched holding shrinks with it. The gap between the 6.7% and 100% rows is entirely a statement about which spending is treated as fixed.
And if RPI and your own inflation part company after February 2030, a matched portfolio stops matching while still looking matched, which is the failure mode worth watching for. The number to keep an eye on is not the bond weight. It is the ratio of the pot to the cost of the promise, on the same day, on the same curve.