The Generational Balance Sheet

This is an In Discussion working paper (September 2026, revised October 2026). The underlying data (spending by age band, the historical trend, wealth by age and the size of the debt interest bill) is drawn from primary sources: the OBR, the ONS and the Intergenerational Foundation. What isn't settled is how to combine those layers into a single "generational burden" figure. The methodology section below explains why that is the hardest step, and deliberately stops short of picking one answer. See the note at the end for what is still open.

1. The Numbers

1.1 Current spending by age band

The OBR's July 2026 Fiscal risks and sustainability report breaks spending and tax receipts per person into age bands. ICAEW's summary groups them into five (these figures exclude debt interest):

Source: OBR, Fiscal risks and sustainability, July 2026, Chart 1.2; ICAEW "Chart of the week: tax and spending by age", 16 July 2026. ICAEW gives the 0-19 total as £22,000, although the three components shown add to £22,400; the net figure of -£20,680 matches the £22,000 total.

The OBR's own chart uses ten age bands, not five; this table is ICAEW's five-band simplification. The more detailed series is in the report's Annex C and supporting spreadsheets, which this note hasn't yet used.

1.2 The historical trend, 1999 to 2019

The Intergenerational Foundation tracked spending per person on children, working-age adults and pensioners over nineteen years, using public spending (PESA) and DWP benefit data. The underlying figures come from official sources. The Foundation's interpretation is a separate matter: it campaigns on intergenerational fairness, and language such as "unjustifiably large" pension costs is its own editorial view.

Three things widened the gap. The introduction of the triple lock in 2010 explains 71% of the one-year jump in spending per pensioner in 2010-11. Sure Start funding per child fell from £1,600 to £377 between 2010-11 and 2018-19. And a 2015-16 change in the PESA method started counting interest on public-sector pension liabilities, adding about £1,520 per pensioner to the measured figure, so figures before 2015-16 understate spending per pensioner on a like-for-like basis.

There is no clean way to link this 2018-19 series (three age bands, IF) with the 2026 OBR series above (five bands). The methods differ and there is a seven-year gap. Joining them into one continuous series would need reconciliation work this note hasn't done.

1.3 Wealth by age

The ONS Wealth and Assets Survey, Round 8 (April 2020 to March 2022, published January 2025), gives median household total wealth by the age of the head of household:

Across Great Britain the median is £293,700, or £181,700 excluding private pensions. Of total wealth, 40% is net property wealth, 35% private pension wealth, 14% net financial wealth and 10% physical wealth. These figures exclude state pension entitlement, which would raise the older age bands further if it were counted as an asset.

One caveat on the data itself: the Office for Statistics Regulation removed the survey's accredited official statistics status from Round 8 onwards (June 2025), saying the data was "no longer of sufficient value or quality to meet users' needs". Treat the table as a guide to direction, not a precise measurement.

1.4 The cost of debt

Debt interest spending was forecast at about £111 billion in 2025-26, around 8% of public spending and close to the whole education budget, and it has more than doubled as a share of GDP since just before the pandemic. Assigning that £110bn to particular age groups needs a model of cause and effect, not just a measurement. Section 4 explains why that step is contested.

2. International comparison

Not yet written. The OBR's July 2026 report shows that the UK's old-age dependency ratio has risen relatively little over the past 50 years compared with other G7 countries, thanks to sustained net immigration, but is projected to follow the G7 pattern of substantial increases over the next 50 years. The stronger sources for this section, the EU Ageing Report's projections and the G7 old-age dependency comparison in the OBR report, haven't been used yet.

3. The Steel Man

The case that current patterns are not evidence of unfairness

Pay-as-you-go is a design choice, not a flaw. State pension and NHS spending on today's pensioners is largely funded by taxes that those same pensioners paid through their working lives, on the same reciprocal basis today's workers expect to benefit from in future. Calling this a "burden transferred to the young" assumes the system is one-way extraction rather than a continuing exchange working as intended. Older generations also paid higher tax rates in the past: under the tax systems of the 1970s and 1980s, before thresholds were linked to prices, they paid rates well above what equivalent earners pay today. Any calculation of lifetime net position needs to include that, rather than reading off a single year.

Much of the wealth in Section 1.3 is also unrealised and the product of wider forces rather than deliberate design. A large share of the wealth of 55 to 74-year-olds is rising house prices caused by decades of planning restrictions and low interest rates, a failure of housing supply and monetary policy rather than a transfer that pensioners arranged for themselves. Private transfers also flow the other way and are usually left out of this kind of accounting: inheritances and gifts during life move wealth from old to young, partly offsetting the public spending pattern. McCarthy, Sefton, Lee and Sambt's Generational Wealth Accounts extends the standard generational accounting method to include these private transfers, and asks whether public and private flows offset each other around the financial crisis. That is an open academic question, not a settled one. And policies that favour older people, such as the triple lock and universal benefits, can be read as the normal result of a working democracy in which older people vote more and organise more effectively. That is uncomfortable, but not obviously illegitimate "capture".

The case that current patterns are a real imbalance

Spending per pensioner grew faster between 1999 and 2019 than the pensioner population did, which demographic change alone can't explain. Education spending per child fell in real terms after 2010-11 while spending on pensioner benefits rose over the same period, a policy choice rather than a demographic inevitability. The OBR's July 2026 baseline projects health spending rising from 8% to 13% of GDP and state pension spending from 5% to around 9% of GDP by 2075-76 under unchanged policy, so today's snapshot understates what today's young people will face by the time they are old themselves. The gaps also grow within generations, not just between them: millennials with property-owning parents are set to inherit roughly four times as much as those without, so a "the young lose out" framing on its own hides a sharper divide between those with assets and those without, which cuts across age.

4. Why This Is Harder Than Stacking Three Datasets

The three layers of data above are not the same kind of thing. Spending by age and wealth by age are both snapshots, and mostly uncontested (with the caveat on the wealth survey in 1.3). Assigning the debt bill to generations needs a model of cause and effect, which makes it contested rather than simply measured. And any single combined "generational burden" figure needs a value judgement on top of that model, which is where advocacy comes in, on every side of the argument.

There are three defensible ways to assign the debt interest bill to generations, and they reach very different conclusions from the same debt:

1. Backward-looking: assign the interest to the generations that benefited from the spending the debt paid for, such as the 2008 bank rescues, Covid furlough and decades of borrowing for day-to-day spending.
2. Forward-looking: assign the interest to whoever's working-age taxes will pay it from now on, which automatically loads the cost onto the young and future generations, whoever benefited from the original spending.
3. Lifetime accounting by birth year (Kotlikoff's approach): model each birth year's total lifetime taxes minus benefits received, including its share of debt interest across its whole life.

These aren't interchangeable, and any published figure needs to say which one it used. Some technical problems apply whichever method is chosen:

• Discount rates. Present-value estimates of lifetime burdens swing substantially with the discount rate chosen, and that needs stating openly rather than leaving it in a footnote.
• Double counting. The wealth survey values a final salary pension by capitalising its expected future payments. Counting the future spending that funds those same payments as well counts the same value twice.
• Different people. Comparing today's 20 to 39-year-olds with today's over-80s doesn't compare the same people over time. It compares different generations at different stages of life in a single moment, and today's net contributors become tomorrow's net recipients.
• Fixed assumptions. The OBR says explicitly that its 50-year scenarios hold policy and behaviour fixed. They illustrate pressures; they aren't forecasts.
• Labels. What counts as a "tax" and what counts as a "transfer" is, in Kotlikoff's own phrase, a labelling problem: a fiscal convention, not an economic fact. Different conventions produce different headline burden figures from the same underlying cash flows.

As soon as a document picks one method of assigning debt, one discount rate or one framing of pay-as-you-go ("burden" versus "exchange") without setting out the alternatives, it has moved from measurement into contested ground presented as neutral analysis. That is the biggest risk to this project's credibility in developing this note further. The data in Section 1 is largely undisputed; a combined "impact" story is where a hostile reader would get real traction, and this note deliberately doesn't attempt that step yet.

Links to the Rest of the Site

The method for assigning the debt bill feeds directly into the separate work on tax receipts drafted for the Economy pillar. Whichever method this note eventually adopts, that work should use the same one, rather than the two developing incompatible conventions. Any future triple lock reform in the Welfare pillar needs to engage directly with the pay-as-you-go argument in Section 3, not just assert that the pension is unsustainable. The fall in Sure Start and per-pupil funding in Section 1.2 overlaps with the Education pillar's figures and should be checked for consistency rather than sourced twice. And the property part of Section 1.3 is largely a story about housing supply and planning rather than demographics alone, which the Housing pillar should take into account whenever it returns to generational wealth.

Sourcing note: the data in Sections 1.1 to 1.4 comes from the OBR's July 2026 Fiscal risks and sustainability report, the Intergenerational Foundation's Age Bias (2021) and the ONS Wealth and Assets Survey Round 8, subject to the caveats in the text, particularly the loss of accredited status for the wealth survey and the lack of a clean link between the 1999-2019 and 2026 spending series. Section 2 is not yet written. The method for assigning debt in Section 4 is deliberately left open rather than chosen quietly; the McCarthy, Sefton, Lee and Sambt work on private transfers and the Resolution Foundation's Intergenerational Audit are the next sources to read in full before this note takes a firmer position. If this ever moves from background material to a policy proposal, it will need the full falsification framework (prediction, magnitude, time horizon, counterfactual and falsification condition) that this project's numbered pillars require.

For public discussion. Not affiliated with any political party. | generationalreset.org

The Generational Reset | In Discussion: The Generational Balance Sheet | For public discussion. Not affiliated with any political party. | generationalreset.org