Top-slicing relief after 2021: where HMRC’s own calculator went wrong

Top-slicing relief runs in a fixed five-step shape (ITTOIA 2005 s.535–537; IPTM3820). The 2021 change left the steps alone. What it changed sits inside them: for gains arising on or after 6 April 2021, the Personal Savings Allowance and the savings starting-rate band are recalculatedat each step's own income level — in particular at Step 4, where income includes only one year's slice of the gain. The old method fixed them once at the taxpayer's actual income and carried them through.

The effect is concrete. A taxpayer who is additional-rate with the whole gain in income (PSA nil) can be higher-rate with only the slice (PSA £500). The old method carried the nil PSA into Step 4; the current method gives Step 4 the recalculated £500. More allowance inside the relieved-liability leg means more relief.

The same case, both methods

An offshore bond is fully surrendered in 2025/26: a £90,000 gain before any TAR over six complete years (slice £15,000), other income £45,000. With the full gain stacked, income is £135,000 — above the additional-rate threshold, PSA nil. With only the slice, notional income is £60,000 — higher rate, PSA £500.

Worked-example basis: No time-apportioned reduction is due for any gain or policy shown. TAR is not calculated or applied. Raw gains are shown before any TAR.

Six hundred pounds on a single, unremarkable case — £500 of extra PSA inside the slice computation. Larger gains and longer policy terms move more.

IPTM3820 · HMRC Agent Update 83 (April 2021) · ITTOIA 2005 s.535–537

The documented record — including HMRC's own calculator

HMRC announced the change in Agent Update 83 (April 2021) and reflected it in IPTM3820 — and the update itself is a frank document about how hard the implementation had been. It declares that “the Top Slicing Relief fully automated approach on life insurance policy gains is active from 6 April 2021”, describes an automated recovery run for affected 2018/19 returns carried out on 28 August 2020, and records that for 2019/20 an online-filing exclusion applied — in HMRC's words, “affected customers are required to complete paper returns to get the correct calculation”. For a period, the Self Assessment calculation could not produce the right top-slicing figure for the affected cases at all. HMRC's own scale estimate in the same update: of the 45,000 customers who incur gains, 2,000 would benefit from the changes.

Nor did the announcement settle it. Weeks later, on 7 May 2021, AccountingWEB reported tax lecturer and Absolute Software director Tim Good's position that HMRC “had not resolved all of the longstanding issues and was still getting TSR calculations wrong in a number of cases”, according to the Topslicer tool he built — the residual dispute being over the allocation order FA 2020 s.37 wrote into the relief computation. And the record carries a concrete number: Lawrence Rose Ltd, a firm of accountants, published a first-person account of resubmitting a client's return and finding “HMRC have overtaxed him by £1,267”, adding the practitioner's warning that “almost all the main brands clone the HMRC calculation”. A lagging tool isn't slightly stale. It answers a different year's law.

The one-line test for any tool: run a case where the full gain pushes a basic-rate taxpayer into higher rate, and check whether the PSA used in the Step-4 leg is £1,000 or £500. If it's £500, the tool is applying the pre-2021 method — and understating relief on every case with this shape.

The common error underneath

The mistake is easy to make. The PSA looks likea fact about the taxpayer (“she's higher-rate, so £500”), when inside top-slicing it is a fact about each step's notional income. The personal allowance carries the same trap: at Step 4 the PA is recomputed at the notional income level, so a gain that withdraws the PA in the real computation can leave it intact inside the slice computation. Recalculate everything from scratch at each step's own income, and carry nothing forward.

Our engine implements the recalculation per IPTM3820 and regression-tests the case above at zero-pence tolerance on every commit. The live calculator below opens on the engine's offshore regression case — a £60,000 gain before any TAR over six complete years with £35,000 of other income, whose own Step-3/Step-4 PSA recalculates from £500 to £1,000 — and prints both step-level PSA figures side by side; enter the £90,000 case above to reproduce the £600 swing.

Worked-example basis: No time-apportioned reduction is due for any gain or policy shown. TAR is not calculated or applied. Raw gains are shown before any TAR.

For gains before 6 April 2021 the pre-change method remains correct — the engine branches on gain date.