Pension tax relief cost statistics – a peek under the bonnet

On 30 September, HMRC published new estimates of the cost of income tax and National Insurance Relief in relation to pension saving.

Table summarising the provisional estimates for 2019-20

Most ‘tax relief’ is not taxing employees on employer contributions

Strikingly, 78% of income tax relief on contributions relates to the fact that individuals are not taxed upfront on the value of contributions that employers pay into their pensions for them (instead, most of the associated withdrawal in retirement will be subject to income tax). If National Insurance relief is included, 86% of relief on contributions relates to money paid into pensions by employers. Yet many proposals for “reforming” tax relief on pension contributions fail to mention how employer contributions would be treated.

National Insurance relief looming larger

At £19.7bn, the cost of employers’ and employees’ National Insurance relief in respect of employer pension contributions is smaller than the gross cost of income tax relief on all contributions (£34bn). However, whereas pension withdrawals are subject to income tax, no National Insurance is due.

The £19.2bn collected on pensions paid in 2019-20 is a poor proxy for the tax that will be ultimately be collected on pensions derived from 2019-20 contributions – for example, because there are more contributors than pensioners. But National Insurance relief could plausibly be worth more, end-to-end, than income tax relief, despite being much less discussed. And its value should rise from April 2022, when employer and employee NI rates both increase by 1.25 percentage points.

A cost relative to what?

Measuring the cost of tax relief is not like asking HMRC to tot up how much tax was collected last year – this task involves estimating the tax that was not collected but which would have been due under different laws, if behaviour did not change. HMRC has to start by selecting an alternative set of rules which it treats as having zero cost. It chose one where contributions come from taxed and NIC-ed income (with employer contributions taxed and NIC-ed as a benefit-in-kind), where investment income is taxed along the way, and where withdrawals are tax-free. Discussing an earlier version of these statistics, the Institute for Fiscal Studies explained that the estimates were:

“…relative to a benchmark where individuals are not able to benefit from tax-rate smoothing by only paying tax on pension income when it is received and where the system encourages individuals to spend rather than to save. A better estimate suggests the true cost of income tax and NICs relief on pension saving is less than half the official estimate. Taking into account the impact of taxes at the corporate level – corporation tax on normal returns and stamp duty on purchases of shares and property – would reduce this figure further.”

If the numbers exaggerate fiscal support for pension saving relative to a system where it was neither rewarded nor penalised, do they at least tell policymakers how much revenue they could get their hands on by switching to HMRC’s baseline system?

One problem is that some of the National Insurance relief will have been given to public sector employers, so just represents a transfer from one part of government to another. Making a hospital trust return more of its budget to the Exchequer in the form of higher employer NICs is a backdoor spending cut – hence public sector employers being compensated for next year’s rise in NICs.

Taxing Peter when Paul’s ex-employer clears a deficit

Another difficulty is highlighted in a note HMRC published alongside these figures:

“For funded DB [defined benefit] schemes HMRC’s modelling allocates DRCs [deficit reduction contributions] to members in private DB schemes based on the weighting of their employer contributions, therefore a member with larger employer contributions will be allocated a larger value of DRCs. Employer contributions are then treated in a member’s income stack ‘sitting’ above their individual contributions.

“Estimated DRCs are split among active members of private DB schemes. A closed DB scheme with no active members will not be included in the modelling, their share of estimated DRCs will be included in the ONS’ MQ5 estimate and will be allocated across active members of private DB schemes in HMRC’s modelling.”

If policymakers were to decide that pension contributions should come from taxed income, they would want to avoid encouraging employers to keep taxable contributions to fund new DB accrual low, with tax-free deficit contributions making up the shortfall later. But that does not mean the Government could just start taxing individuals now on the value of contributions that shore up pensions promised under a different regime. And even if that were thought reasonable, the way of doing it modelled by HMRC surely is not.

HMRC appears to assume that active members in defined benefit schemes could be taxed on the value of contributions that sponsoring employers pay to repair deficits in these schemes – even though most of the members whose pensions are made more secure by these contributions will be deferred and pensioner members; in 2020, only 11% of members in DB schemes eligible for the Pension Protection Fund were accruing benefits.

And if a scheme has no active members, the tax charge would fall on active members of other schemes, who do not benefit from the contribution at all! This amounts to taxing Peter when Paul’s ex-employer clears a pension deficit. Although some pensions tax changes have had perverse consequences (think of the Annual Allowance taper discouraging NHS consultants from covering extra shifts), it is hard to imagine any Chancellor making retrospective taxation of the wrong people a principle of pensions taxation.

As well as using an implausible policy baseline, HMRC’s approach seems likely to overstate the cost of tax relief. Dividing DRCs between a small subset of the members who benefit from them gives a bigger per member value, putting more of the contribution above the 40% tax threshold (and sometimes the 45% threshold). This will especially be the case when the members selected still have employment income for the pension contribution to sit on top of.

HMRC’s estimates are often cited by politicians and pundits. The numbers, and their limitations, should be much better understood.