Let's separate this into three different effects:

1. Immediate accounting effect - the cost of awarding new pension rights falls straight away.
2. Long-term liability effect - the £1.4tn liability gradually falls because new, lower benefits replace the old accrual.
3. Cash effect - actual pension payments do not fall immediately, because pensions already earned still have to be paid.

The latest Whole of Government Accounts give us the starting point: £1,400.9bn of unfunded pension liabilities, £23.8bn of unfunded current-service cost, and £60.4bn of unfunded pension benefits actually paid in 2024-25.

1. Starting point

Using 2027 as the reform starting year and model 10, 20 and 30 years afterwards:

2027 2037 2047 2057

The £1.4tn is a present-value accounting measure, not £1.4tn sitting in a bank account. HM Treasury explicitly warns that the liability is highly sensitive to the discount rate and other assumptions.

2. If only future accrual is reduced

I'll retain the earlier illustrative assumption that approximately 70% of the eventual liability relates to benefits already accrued, leaving about 30% associated with future service.

That isn't an official Treasury estimate; it is simply a modelling assumption so that we can examine the mechanics.

Effect on the £1.4tn liability

So there is an important distinction from simply cutting existing pensions:

A 40% cut in future accrual does not produce a 40% reduction in the £1.4tn liability.

Under our 70/30 assumption it produces roughly a 12% reduction in the eventual liability.

3. What happens to the annual cost?

The current-service cost for the unfunded schemes is £23.8bn.

If future pension accrual were immediately reduced:

So, purely from the accounting perspective, a 40% reduction could reduce the annual cost of new pension promises by about £9.5bn initially.

But that doesn't mean the government immediately saves £9.5bn in cash.

The £60.4bn of pensions actually paid in 2024-25 is overwhelmingly determined by pensions already being drawn and benefits earned previously.

4. The transition is the really interesting part

Suppose the reform starts in 2027.

Someone retiring in 2037 might have:

• 10 years of reduced pension accrual after reform, but
• decades of unreduced pension rights from before reform.

Someone retiring in 2057 could have substantially more of their pension based on the new system.

Therefore the cash saving builds slowly.

We can illustrate this with a 20-year transition assumption: the proportion of pension payments affected by the new accrual rules increases approximately linearly from 0% to 100% between 2027 and 2047.

This is not a Treasury forecast. It is a deliberately simple illustration of the transition.

Using the current £60.4bn annual unfunded pension payments as the starting point:

Illustrative annual cash saving

But these figures should not be interpreted as a forecast of actual government cash savings.

They answer a narrower question:

What would the savings look like if the proportion of pension payments affected by the reform gradually increased over 20 years?

Actual savings would depend on the age distribution of public-sector workers, retirement dates, mortality, salary progression, inflation, pension indexation and how the reform affected different schemes.

5. A more useful comparison

We can therefore put the three policies side by side.

A. Cut pensions already earned

A mechanical 40% reduction would produce:

£1,400.9bn -> £840.5bn

A £560.4bn reduction in the accounting liability.

And because existing pension payments are immediately reduced, the cash effect could be relatively rapid.

But this would involve changing accrued pension rights, rather than simply changing the pension people can earn in future.

B. Protect existing pensions, cut future accrual by 40%

Our illustrative model gives:

£1,400.9bn -> approximately £1,233bn

So the eventual liability reduction is only about:

£168bn

But the reform would affect people primarily through future pension earned from 2027 onwards, rather than reducing pension already earned.

C. Protect existing pensions, cut future accrual by 20%

The corresponding figures are approximately:

£1,400.9bn -> £1,317bn

or a reduction of about:

£84bn

There is another important number

The WGA shows how the £1.4tn liability actually moves.

For 2024-25:

• Current service added £23.8bn
• Interest added £65.6bn
• Contributions from scheme participants were £16.8bn
• Benefits paid reduced the liability by £60.4bn
• Other actuarial/revaluation effects changed it further.

That is why simply saying 'there is a £1.4tn liability, therefore a 40% cut saves £560bn' is misleading.

The liability is a long-term discounted projection.

In fact, the WGA reports that the unfunded liability would be £1,531.7bn using the National Accounts 2% real discount rate, versus £1,400.9bn under the WGA's 2.4% rate. At a 0% real discount rate it is £2,274.9bn.

So a change in accounting assumptions alone can move the apparent liability by hundreds of billions without changing anybody's pension.

If the objective were reducing future government pension expenditure rather than cutting pensions already earned, the numbers look roughly like this:

*Using our illustrative 70% accrued / 30% future-service assumption.

Protecting accrued rights dramatically changes the economics of the proposal.

A 40% reduction in future pension accrual might eventually save something on the order of £168bn of today's measured liability, rather than £560bn.

And the government would not receive £168bn in cash. It would mostly represent lower future pension obligations.

Meanwhile, the £60.4bn annual pension-payment bill would initially remain broadly intact, because those payments relate largely to pensions already earned.

If we take this one step further, the really useful calculation is to model 2027-2057 year by year, showing the cumulative government cash saving and the remaining pension liability under 20%, 30% and 40% future-accrual reductions. That would make the difference between accounting savings, cash savings and eventual liability reduction very clear.