Key takeaways
- On pay between £12,570 and £50,270, £1,000 sacrificed escapes 8% employee and 15% employer National Insurance — £230 an ordinary contribution still pays in 2026/27.
- Above the £50,270 upper earnings limit the employee rate falls to 2%, so the same £1,000 leaves £170 uncharged, of which £150 belongs to the employer.
- Statutory Maternity Pay for the first 6 weeks is 90% of average weekly earnings, calculated after the sacrifice, and SMP entitlement stops below £129 a week.
- Cash pay after the sacrifice cannot fall below £12.71 an hour for a worker aged 21 or over, and the pension contribution does not count towards that floor.
- From 6 April 2029 only the first £2,000 sacrificed each year stays exempt. HMRC counts 3.3 million employees already above that line.
The income tax relief is identical. The National Insurance is the whole difference
You want another £1,000 in your pension this year. Your employer can run it as an ordinary employee contribution, or as a salary sacrifice. Does the label change anything?
For income tax, no. For National Insurance, yes — and in the 2026/27 tax year the gap is 23p in the pound.
Employee Class 1 National Insurance runs at 8% on pay between the primary threshold of £12,570 and the upper earnings limit of £50,270, then 2% on everything above that. Employers pay 15% on earnings above the £5,000 secondary threshold, with no ceiling. Those are the rates GOV.UK publishes for 6 April 2026 to 5 April 2027.
Sacrifice £1,000 from pay inside the 8% band and neither charge arises. That is £80 the employee doesn't pay and £150 the employer doesn't pay, or £230 on a £1,000 contribution. An ordinary contribution attracts exactly the same income tax relief. It just comes out of pay that National Insurance has already been charged on. Our guide to pension tax relief under relief at source and net pay sets out how that tax side works.
Why the National Insurance vanishes rather than being refunded
The mechanism is a contract change, not a relief you claim. HMRC's Employment Income Manual puts it plainly: "A salary sacrifice happens when an employee gives up the right to part of the cash remuneration due under his or her contract of employment."
Once the pay has been given up, the money reaching the pension is the employer's contribution, not yours. HMRC's National Insurance Manual then closes the loop. Where an employer pays into a registered pension scheme, "such a payment is disregarded in the calculation of earnings for Class 1 NICs purposes."
Class 1 covers both halves — the employee's primary contribution and the employer's secondary one. Disregard the payment for Class 1 and you have disregarded both at once. That is why nothing gets reclaimed later: the charge never arises.
This survived a cull that killed most of its cousins. From April 2017 the government withdrew the tax and National Insurance advantages of salary sacrifice across nearly every benefit. HMRC's manual records the carve-out: "Employer contributions into registered pension schemes and employer provided pensions advice are excluded from the April 2017 changes."
The arithmetic changes at every National Insurance boundary
That 23% only holds inside one band. Here is the same £1,000 at three points on the scale, at 2026/27 rates.
- Inside the 8% band, on pay between £12,570 and £50,270: £80 for the employee, £150 for the employer. £230 per £1,000.
- Above the upper earnings limit, on pay over £50,270: the employee's marginal rate is 2%, so £20, plus the employer's £150. £170 per £1,000.
- Straddling the limit. On pay of £52,270, a £4,000 sacrifice takes the first £2,000 out of the 2% band and the next £2,000 out of the 8% band. That's £40 plus £160, so £200 for the employee, and £600 for the employer. £800 on £4,000, or 20%.
The chart above plots those per-£1,000 figures. Its shape is the argument. The employee's share of the saving shrinks as pay rises, because the marginal National Insurance rate drops from 8% to 2% at £50,270. Employer relief stays flat at 15% all the way up, so it does most of the work for higher earners — £150 of the £170.
That inverts the usual intuition about who benefits. The employee's National Insurance saving is largest for someone in the middle of the band, not at the top of it.
At the bottom of the scale the saving runs out before the pay does
Take someone paid £17,570 on part-time hours. Sacrifice £5,000 and their cash pay lands exactly on the £12,570 primary threshold. The employee saving is 8% of £5,000, or £400. The employer saves £750.
Sacrifice a second £5,000 and the employee saving stops dead. There is no employee National Insurance below £12,570 to avoid. The employer keeps saving 15% down to £5,000, so the arrangement still has value — to the employer.
Something else is happening on the way down. At £7,570 of cash pay the worker is still above the lower earnings limit of £6,708 for 2026/27, but not by much. That limit is where credit for the state pension and contributory benefits begins. HMRC's National Insurance Manual defines it exactly, as "the point at which employees start to build up entitlement to contributory benefits."
Cross below it and the year stops counting. You need 10 qualifying years on your record for any new State Pension at all.
The employer's 15% is the employer's, unless the scheme hands it back
Your 8% or 2% is automatic. The employer's 15% is a decision.
HMRC's employer guidance is blunt about who makes it: "The employer decides whether salary sacrifice affects contributions into a workplace pension scheme." Some schemes pass the whole saving through, so £1,000 sacrificed arrives as £1,150. Others keep it and bank the cost reduction. Both are lawful, and the difference is worth more than the entire employee saving above £50,270.
The same guidance flags a quieter one. Pay rises, overtime and other earnings-related payments "can be based on the notional salary or the new reduced cash salary, but this must be made clear to the employee." A pay rise calculated on the reduced figure is a smaller pay rise, compounding for as long as the arrangement runs.
Statutory pay is worked out on the salary after the sacrifice
Here is the trade-off that the sales pitch leaves out.
Statutory Maternity Pay for the first 6 weeks is "90% of the employee's average weekly earnings". Sacrifice reduces average weekly earnings, so it reduces that 90% pound for pound. Statutory Sick Pay for 2026/27 is "£123.25 a week or 80% of average weekly earnings - whichever is lower", again computed on the reduced figure.
Entitlement can disappear entirely rather than merely shrink. To qualify for SMP you must "earn on average at least £129 a week", which is the lower earnings limit in weekly form. HMRC states the consequence directly: "Salary sacrifice can affect the amount of statutory pay an employee receives. It can cause some employees to lose their entitlement altogether."
And, for the employer: "If a salary sacrifice arrangement reduces an employee's average weekly earnings below the lower earnings limit, you don't have to make any statutory payments to them." Maternity Allowance and Additional State Pension sit on the same list of earnings-related entitlements.
Statutory Sick Pay was never a wage replacement in the first place. Our guide to income protection and the definitions that decide a claim covers what sits above it, and how often those claims fail.
The floor: minimum wage is a hard stop, and the pension doesn't count towards it
HMRC's guidance states the limit in one sentence. A salary sacrifice arrangement "must not reduce an employee's cash earnings below the National Minimum Wage" rates, and employers "must put procedures in place to cap salary sacrifice deduction and ensure NMW rates are maintained."
From 1 April 2026 that means £12.71 an hour for a worker aged 21 or over. Cash pay after the sacrifice, divided by hours worked, has to clear it. And you can't argue the pension back into the sum. HMRC's National Minimum Wage Manual is categorical: "Benefits in kind do not count for National Minimum Wage purposes."
HMRC's own research with employers, published in 2025, found this is the binding constraint in practice rather than anything in the tax rules. One large manufacturer's HR officer told the researchers that take-up had fallen "not because they don't want to participate in salary sacrifice, it's because national minimum wage legislation doesn't always seem compatible with salary sacrifice legislation."
Borrowing capacity moves with the number on the payslip
A sacrifice reduces contractual gross pay. That's the figure on the payslip, on the year-end statement and in an employer's reference.
Mortgage underwriting runs on income multiples. The Bank of England's Financial Policy Committee asked regulators to limit lending "with an LTI ratio at, or greater than, 4.5", and that 4.5 is the multiple the market is calibrated around. At it, £2,000 a year of sacrificed salary maps to roughly £9,000 of headline borrowing capacity, if a lender takes the reduced figure at face value.
Whether it does is lender policy, and it varies. Some add the sacrifice back where the employer confirms it in writing; some don't. We could not find an authoritative cross-market source on how many do which, so £9,000 is an upper bound on the effect rather than a measurement of it.
The strongest objection: this relief is already legislated smaller
Everything above describes an arrangement the government has decided to shrink.
HMRC's impact note, published on 4 December 2025, sets out the design. From April 2029, "only the first £2,000 of employee pension contributions through salary sacrifice each year will be exempt from NICs." Above that, "employee contributions above this amount will be subject to employer and employee NICs like other employee workplace pension contributions."
The same document sizes it. HMRC estimates "7.7 million employees currently use salary sacrifice to make pension contributions", and "of these, 3.3 million sacrifice more than £2,000 of salary or bonuses" — 44% of users above the line, 56% below it. For those affected, "the average additional employee National Insurance contributions liability is estimated to be £84 in the first year of impact".
The driver is cost. The relief "has increased markedly from £2.8 billion in forgone National Insurance contributions in tax year 2016 to 2017, rising to £5.8 billion in tax year 2023 to 2024", and HMRC expected it to "nearly triple to £8 billion by tax year 2030 to 2031" if left alone.
£84 a year is not a large number, and the objection isn't really about its size. It's that the 23p in the pound at the top of this piece has a partial expiry date attached to it, and the part expiring is the part that applies to larger contributions.
What this evidence can't tell you
The rates are current and the mechanism is documented in statute and manual. The behaviour around it is much thinner evidence.
HMRC's employer research is qualitative — interviews, not a survey — and it was carried out in 2023, when the employee National Insurance rate was 12% rather than 8%. Its worked examples are stale for that reason. Only its structural findings, such as the minimum wage constraint, carry across.
The prevalence figures inside it are older still: "In 2019, 30% of private sector and 9% of public sector employees used a salary sacrifice arrangement for their pension contributions." That's a 2019 sample describing an arrangement whose economics have moved since, and it is the most recent split that report offers.
The 2029 cap isn't finished either. HMRC's own note says "Details on the design and operation of the £2,000 contribution limit will be set out in secondary legislation in due course." The exchequer numbers attached to it are forecasts of behaviour under a rule that does not exist yet.
Two further limitations. Scottish income tax bands differ from the rest of the UK, but National Insurance does not — so the arithmetic here is UK-wide while the tax relief wrapped around it isn't. And none of this covers defined benefit schemes, where a sacrifice buys accrual rather than a pot, and the pound-for-pound comparison doesn't hold.
What would change the conclusion
If the employer's rate moves. £150 of the £230 is the employer's 15%. Most of what makes the arithmetic work isn't the employee's saving at all, which means most of it is exposed to a rate the employee has no visibility of.
If the employer keeps its share. Where the saving isn't passed on, the pot receives £1,000 rather than £1,150, and the employee's side of it is £80 — or £20 on pay above £50,270.
If the year's sacrifice crosses £2,000 after April 2029. On HMRC's own count that is already 3.3 million people, and the threshold is a cash figure with no stated uprating.
If cash pay approaches either floor. £12.71 an hour and £129 a week are the two lines where this stops being a question about tax and becomes a question about entitlement — sick pay, maternity pay, and whether the year counts on your National Insurance record.
Above all of it sits the annual allowance, £60,000 for 2026/27, which caps what goes in regardless of the route. How much marginal relief the contribution attracts still depends on which tax band the pay came from; our piece on UK dividend and savings tax and the marginal rates behind it works that through.
The figure worth watching isn't the percentage saved. It's the distance between your contractual gross pay and the two floors underneath it — and, from 2029, how much of the year's sacrifice sits above £2,000. LedgerTouch tracks contributions against the annual allowance; the floors are on your payslip.