TFL: Measured not managed

Part two of two

Part one argued that TfL's income is not what it is described as. This part concerns the other side of the ledger, and the finding is worse. The costs are not what they are described as either — and in the most consequential case, the movement everyone treated as evidence of discipline was produced by an actuarial assumption.

Begin with the accusation everybody already believes, because it is wrong, and saying so is what makes the rest credible.

There is no wage explosion

The received view is that TfL is being eaten alive by the salaries of its own staff. Thirteen years of statutory accounts say otherwise. What follows is drawn from the staff costs note in each TfL Annual Report and Statement of Accounts from 2012/13 to 2024/25, with earned income taken as fares plus other operating income on the London Assembly Research Unit's basis.

Staff costs are the lowest share of earned income in the entire series: 32.0 per cent, against a pre-pandemic average of 37.8 per cent. Over twelve years staff costs rose 40.2 per cent in cash terms, which is roughly CPI, while earned income rose 58.7 per cent. Headcount moved 4.2 per cent, from 28,020 to 29,206. Whatever is wrong with Transport for London, the wage bill of the people it directly employs is not it.

Which raises the obvious question of where the money is going instead. The 2026-27 cost base answers it.

Gross expenditure is 139 per cent of earned income; TfL earns about 72 pence of every pound it spends. And the largest single line is not wages. It is £2,823.8m paid to private bus companies — more than the entire in-house payroll. Add the two and 72 per cent of everything TfL earns is gone before maintenance, electricity, debt interest or a single station is cleaned.

The £370m that nobody moved

Now the central finding, and it is the reason this article exists.

Between 2022/23 and 2023/24, TfL's reported staff costs fell by £120m. That reads as restraint. It has been treated as restraint. Decompose it.

Wages and salaries rose by £152.7m — an increase of ten per cent in a single year. Pension costs fell by £282.6m. The entire reported reduction, and rather more besides, is an IAS 19 measurement effect. Gilt yields rose through 2022, the discount rate applied to the defined benefit obligation moved with them, and the current service cost of the pension promise collapsed on paper.

Nothing about the pension became cheaper. No benefit was reduced, no contribution rate changed, no member was moved to a different scheme. An actuary altered an assumption in response to a bond market, and £282.6m left the income statement. Meanwhile the actual cost of employing people was growing at ten per cent, and the accounts reported a saving.

The pension component swings from 29.0 per cent of staff costs in 2021/22 to 12.8 per cent in 2024/25 — a range of some £370m inside a £2.2bn line, larger than any operational movement in the period and entirely invisible in the headline. TfL's 2025-26 budget books a further £181m of savings from a pension revaluation. The mechanism is identical. The word used is savings.

This is not fraud and it is not even bad accounting. IAS 19 requires exactly this treatment and the auditors were right to sign it. The failure is one of description: an organisation reporting to the public that its costs fell, when what fell was the measurement of its costs, and declining to distinguish between the two because the distinction is unflattering and nobody is obliged to draw it.

The scheme that escaped both reforms

The volatility has a cause, and the cause is a pension scheme that no one has been able to touch.

The TfL Pension Fund is a final salary defined benefit scheme, accruing at sixtieths, with benefits available from sixty. It remains open to new members and continues to accrue future service. Members pay a fixed five per cent of pensionable salary, which the fund rules do not permit them to vary. The employer pays the balance of cost, reported to the London Assembly at 33.3 per cent of payroll — five to six times the member contribution.

Set that against what happened to everybody else.

Private sector DB did not close because employers were mean. It closed because FRS 17 and then IAS 19 put the liability on the sponsor's balance sheet at market value, so a listed company's reported equity began swinging on discount rates it did not control — precisely the effect described above — and because the Pensions Act 2004 brought statutory funding objectives, a regulator with anti-avoidance powers and levies to the Pension Protection Fund. Finance directors closed their schemes because they could not defend the volatility to shareholders. None of those forces operates on a statutory corporation with no share price, no analysts and no possibility of insolvency.

The public sector did not escape either. Lord Hutton's 2011 review reformed the major public service schemes rather than closing them: final salary became career average, member contributions rose, and normal pension age was tied to the state pension age. The NHS, Teachers', Civil Service and Local Government schemes all went through it. An NHS consultant today pays up to 12.5 per cent into a career average scheme and works to sixty-seven.

TfL's fund went through neither process. It is not a public service pension scheme under the Public Service Pensions Act 2013 — it is a trust-based occupational scheme with a statutory corporation as sponsor — so it fell outside the Hutton settlement while remaining immune to every pressure that closed private schemes. No career average conversion. No contribution increase. No linkage of retirement age to longevity. It is, so far as one can establish, the last significant open final salary scheme in Britain whose members contribute a flat five per cent.

Not for want of trying. Government made pension reform an explicit condition of the pandemic funding settlements and commissioned an independent review. The review reported. The reform did not happen. What arrived instead was a revaluation.

The fastest-growing cost is the outsourced one

If the wage bill is not the problem, the bus contract is. And it is growing at a rate that ought to be the story.

London buses are not run by TfL. They are tendered on gross cost contracts: TfL specifies the route, the frequency, the fare and the vehicle standard, then pays a private operator a fixed sum to deliver the mileage. TfL keeps every penny of fare revenue and carries all demand risk. The operator owns the buses, runs the garages and employs the drivers — around 25,000 of them, on considerably less than their Underground counterparts.

Bus contract payments run £2,823.8m in 2026-27, then £3,062.2m, then £3,259.4m — a rise of 26.3 per cent across the settlement. Over the same period TfL's own staff costs rise 5.4 per cent. The outsourced arm is inflating at five times the rate of the directly employed one, and TfL names the cause: increased tender costs from third-party providers as routes come up for re-tender.

Consider what that means. London bus services have been competitively tendered since the mid-1980s — the model held up for forty years as the disciplined alternative to an in-house unionised workforce. The verdict is now in the budget. The competitively tendered arm is the fastest-inflating cost in the organisation, and the fare box covers well under half of it.

Part of the increase is a capital programme in disguise. TfL mandates zero-emission fleets. The operator buys the vehicles, builds the depot charging infrastructure, and recovers the cost through the bid price over a contract far shorter than the asset's life. TfL obtains the infrastructure without capitalising a penny of it, and pays for it through an operating line nobody scrutinises as investment. At the end of the term the operator keeps the asset — and holds an incumbency advantage at the next re-tender, which erodes the only discipline the model ever had. The economics rhyme with PFI closely enough to be uncomfortable, and arrive without any of the scrutiny PFI eventually attracted, because they are buried in a line item called bus contract payments.

Four hundred and forty-two million pounds nobody intends to pay

One further line deserves attention. The 2026-27 budget provides £442.0m for bad debt — 6.5 per cent of earned income written off before it is collected, principally against unrecovered road user charging penalties. TfL's fare evasion rate stands at 3.5 per cent against a target of 1.5 per cent by 2030.

Nearly half a billion pounds a year is booked in the expectation that people will not pay. That is more than half the annual DfT capital grant, and it is a number that would end careers in any organisation with an owner.

What the critics of this argument would say

Two objections, and the first is fair.

That IAS 19 volatility is a feature of every DB sponsor in the country and it is unreasonable to single out TfL for an accounting standard it did not write. True. The criticism is not that TfL applies the standard; it is that TfL reports the output as a management achievement without disclosing the driver, in headline communications aimed at people who will not read note 35.

That the bus network is deliberately loss-making because it serves the Londoners least able to afford alternatives, and rising contract costs partly reflect a mandated transition to zero-emission vehicles that Londoners voted for. Also true, and it is the strongest defence available. But a policy choice financed through an operating line, with the asset accruing to a private counterparty and the cost curve rising 26 per cent in three years, is a choice that ought to be argued for explicitly rather than absorbed silently.

 The pattern across both parts is the same, and it is not really about transport. An organisation with no owner, no economic regulator and no minister answerable for it has been left to describe its own performance, and has described it favourably. Income that was taken by statute is reported as earned. Costs that fell because a bond market moved are reported as controlled. A pension scheme that survived every reform imposed on everybody else is managed through revaluation rather than reform. And the fastest-growing cost in the building is the one that was outsourced forty years ago to make it disciplined.

None of this required anyone to lie. It required only that nobody was appointed to ask.