National Contributions tax model

This appendix describes the technical structure of the tax model underlying the revenue estimates for National Contributions (NC). It covers the base dataset and sources, the construction of NC percentiles, uprating from base year to operating year, scaling adjustments to reconcile with HMRC administrative aggregates, the treatment of benefit income, and the specific simplifications and design choices that apply to the per-percentile calculation of current-system and NC tax liabilities. It does not describe the main report's policy design, which is set out elsewhere.

Base dataset

The base year is 2022-23. The base dataset consists of 100 income percentiles covering the UK adult population of 52.9 million (aged 16+), sorted by a single individual-level income measure, NC Total Income, defined as the sum of non-benefit regular income, irregular receipts (inheritance and other one-off receipts), and capital gains. Benefits are retained alongside each percentile as a passenger set of columns but do not enter the sort key.

The base year is constructed from four sources. Family Resources Survey (FRS) 2022-23 (SN 9367) provides 28,590 adult records with grossing weights (gross4), yielding a weighted population of 52,922,514 adults. Wealth and Assets Survey (WAS) Wave 5 provides inheritance receipts and other irregular income, uprated from its 2016-18 reference period by earnings growth. HMRC Capital Gains Tax Statistics 2025 Table 3.2 provides the 2022-23 distribution of capital gains by income band. HMRC Survey of Personal Incomes (SPI) Table 3.1 provides percentile points of total income before tax, used to interpolate earnings growth factors for uprating.

Income component definitions

Non-benefit components are taken per individual from FRS and annualised by multiplying weekly values by 52:

• Employment earnings (inearns)
• Self-employment income (max(seincam2, 0) — negative values zeroed, so business losses do not reduce other income in the sort)
• Private pension income (inpeninc)
• State pension income (INRPINC)
• Investment income (ininv — taxable savings interest and related)
• Gross dividend income (DIVIDGRO)

Benefit components, also annualised, comprise:

• Ben_NI = (INNIRBEN − INRPINC). INNIRBEN as recorded by FRS includes state pension, which is treated as non-benefit income in this model; subtracting INRPINC prevents a double count.
• Ben_Oth = INOTHBEN (other means-tested benefits)
• Ben_Dis = INDISBEN (disability benefits)
• Ben_TC = intxcred (legacy tax credits)
• Ben_UC = induc (Universal Credit, treated as income, not as a deduction)

FRS's own total income field (indinc) is not used. Component sums are the authoritative aggregate.

NC percentile construction

Percentiles are constructed in two passes so that capital gains and irregular income are incorporated into the sort key without creating a circular dependency.

1. For each FRS adult, compute non-benefit FRS income (the six components above). Zero negative self-employment.
2. Sort the 28,590 records by FRS non-benefit income, grossed by gross4 weights, and assign provisional percentiles 1–100.
3. Allocate WAS Wave 5 inheritance and other irregular income across these provisional percentiles, distributed uniformly within each percentile.
4. Allocate HMRC CGT by income band from Table 3.2, distributed uniformly within the band.
5. Compute each individual's non-benefit NC Total Income = FRS non-benefit + WAS + CGT.
6. Re-sort all individuals by NC Total Income and assign final NC percentiles 1–100.
7. Compute weighted means for all fourteen income components within each percentile.
8. Lower bounds are the minimum individual NC Total Income in each percentile; P1 is floored at £0.

The re-sort step is essential: including CGT in the sort causes high-gain individuals to displace middle-earnings individuals at the top of the distribution, and ranking by NC Total Income (not by any component) ensures individual-level validity of the percentile boundaries. Monotonicity of NC Total Income across percentiles is by construction.

Benefits are computed separately per household from FRS Household data and the attached as passengers based on the household’s non-Benefit income percentile to derive a marginal NC rate that applies to the Benefit income alone..

Uprating from base year to operating year

Uprating is applied component by component, producing an intermediate (unsorted) block that is then re-sorted by uprated NC Total Income.

Earnings-linked components (employment, self-employment, private pension, investment, dividends) are uprated using growth factors interpolated from HMRC SPI Table 3.1 percentile points. Nine anchor points are read from Table 3.1 for both base year and update year (P1, P5, P10, P25, P50, P75, P90, P95, P99); a growth factor is computed at each anchor and linearly interpolated between anchors to give a per-percentile earnings growth factor. This captures the different pace of income growth at different points in the distribution — a useful refinement over a single uniform factor, particularly when fiscal drag is material.

State pension is uprated by an aggregate factor reflecting cumulative triple-lock uprating since the base year.

Benefit components are uprated by an aggregate factor reflecting cumulative CPI-linked uprating since the base year.

WAS-sourced income (inheritance and other irregular) is uprated by an aggregate factor. No annual administrative series exists for these components, so the default uses earnings growth as a proxy; the published HoC Library research briefing on wealth distribution can be used for sensitivity testing.

CGT is uprated in real terms. The update-year nominal total gains is divided by the base-year nominal total (giving a nominal ratio), then divided by cumulative CPI since the base year to give a real factor. This real factor is applied to each percentile's base-year CGT amount. Setting CPI to 1 in Parameters recovers pure-nominal uprating.

The realisation-cycle sensitivity in this comparison is two-sided. The current-system CGT comparator is computed on the same uprated gains flows that feed NC, and reconciles to the HMRC 2024-25 outturn of £13.70 billion rather than to the OBR 2025-26 forecast of £22.00 billion, which assumes a recovery in realisations that is not present in the model's data on either side. Substituting the forecast would set NC calculated on trough-year realisations against a current system calculated on recovered realisations, mixing two different years' gains flows in a single subtraction. Were realisations to recover, both sides would rise together: current-system CGT receipts would increase, and so would NC revenue on the same incremental gains, taxed as income at marginal rates of up to 46 per cent against CGT's 24 per cent top rate, partially offset by the inflation deduction and NS&I deferral. The net effect on the NC uplift is therefore second order and ambiguous in sign, and the comparator is held at outturn.

After per-component uprating, the intermediate block is re-sorted by uprated NC Total Income. The sort is implemented with SMALL/MATCH/INDEX to maintain compatibility across spreadsheet engines. Lower bounds in the sorted block are reconstructed as N[n−1] (i.e., each percentile's lower bound equals the previous percentile's NC Total Income mean), ensuring monotonicity by construction.

FRS coverage scalars

FRS under-records several income components relative to HMRC and DWP administrative aggregates. Two gaps are material for the income-side tax calculation. Dividends: FRS captures approximately £24 billion of dividend income at uprated values against HMRC SPI estimates around £90 billion, largely due to under-reporting by limited-company directors and small shareholders. Taxable savings interest: FRS captures around £22 billion against published figures closer to £45 billion, with the gap widening during periods of rising interest rates.

The position on benefit income is more nuanced. The model's uprated benefit aggregate is £201.6 billion (excluding state pension, which is treated separately as non-benefit income). The comparable published figure — total UK social security minus state pension for 2025-26 — is approximately £185–190 billion. The direction and size of this apparent over-statement reflects two offsetting factors:

• Child Benefit (approximately £12 billion per year) is captured in FRS benefit fields but is HMRC-administered and does not appear in DWP benefit expenditure tables. It is a genuine cash transfer to households and is retained in the model base.
• Cost of Living Payments during 2022-23 (approximately £15 billion nationally) were one-off payments that show up in the FRS 2022-23 base data and are uprated to 2025-26 even though the programme has been discontinued. This is a base-year artefact that will disappear when the model's base year rolls forward to FRS 2023-24 or 2024-25.

Separately, DWP's Family Resources Survey Transformation research (2024) documents average 37 per cent undercoverage of benefit caseload in raw FRS responses relative to administrative records. The benefit aggregate here uses survey-only FRS 2022-23 and is therefore subject to baseline undercoverage. The upward pressure from Cost of Living Payment inflation and the downward pressure from survey undercoverage partly cancel, which is why the raw aggregate sits in a reasonable neighbourhood of administrative totals despite both known distortions.

Importantly, the benefit figures in the main model are used only as "passenger columns" alongside each percentile for distributional display and NC effective-rate denominators. They are not used to derive NC on benefits revenue, and they are not used for the NC on benefits distributional analysis. Both of those outputs come from a separate microdata pass documented in the "NC on Benefits" appendix, which operates on the same FRS 2022-23 base but at household rather than percentile level. The Cost of Living Payment artefact affects both passes identically; the effect on the headline NC-on-benefits figure, and the decision not to attempt a model-level correction, are documented in the NC on Benefits appendix.

Three coverage scalars are provided in the Parameters sheet to correct these gaps. The scalars apply in the Uprating sheet's intermediate block, after growth factors and before the re-sort, so scaled values feed the re-sort and propagate consistently to all downstream calculations. The base-year Combined Percentiles sheet is never modified: it remains the firm FRS/WAS/HMRC data foundation.

Default scalar values are calibrated to tax-yield-match rather than to income-match. That is, scalars are set so the model's implied tax yields on each component match published HMRC receipts at the operating year, rather than scaling income totals to match published income aggregates directly. The two approaches differ because the FRS-captured subset of dividend recipients is already concentrated at high incomes and therefore faces higher-than-average marginal rates; scaling the income aggregate to HMRC level would over-apply these high rates to the full dividend base and overstate dividend tax yield. Tax-yield-match scalars deliver the correct revenue figure while acknowledging that the distributional shape of the captured subset is not corrected.

Defaults at the current operating year:

• Dividend scalar: 1.79 — targets ~£15bn dividend tax yield under current rates.
• Investment scalar: 1.56 — targets ~£12bn Income Tax contribution from savings.
• Benefit scalar: 1.00 — unscaled. The current system does not tax most benefits, so scaling benefits would only inflate the passenger-column figures in this model without affecting the current-system comparator. NC on benefits revenue comes from the separate microdata pass (see NC on Benefits appendix) and is unaffected by this scalar. The scalar is retained as a parameter for sensitivity analysis.

Additionally, a yield scalar is applied directly to the current-system Employee NIC line in Tax Comparison (Parameters row 37, default 1.17). This reconciles the FRS-based Employee NIC calculation (£46 billion at scalar 1.00) with published HMRC receipts (£54 billion in 2024-25), which reflect administrative rather than survey-based earnings data. Unlike the income-level scalars, this operates only on the current-system comparator; NC itself is unaffected because Employee NICs are replaced by NC. A similar surgical adjustment (×2) is applied to Class 4 NICs at the top percentile in Tax Comparison to compensate for the percentile-mean model under-capturing self-employment income spread across multiple upper percentiles.

A limitation of the scalar approach is that it corrects aggregate levels but not distributional shape. If FRS captures 25 per cent of dividend income, scaling 4× restores the aggregate but the relative distribution across percentiles still reflects the FRS-captured subset. For aggregate revenue estimation this is acceptable; for fine distributional analysis it is a limitation that is noted explicitly.

Tax Comparison sheet design

Tax Comparison applies current UK tax rules and the NC progressive schedule to the uprated, scaled, sorted percentile data. The following design choices shape the comparison.

Current-system thresholds applied at frozen 2025-26 levels

Current-system taxes (Income Tax, Employee NICs, CGT, IHT) are computed on the uprated income figures using current frozen thresholds: personal allowance £12,570, higher-rate threshold £50,270, additional-rate threshold £125,140, and the corresponding NIC, CGT and dividend thresholds. Applying current thresholds to uprated incomes reproduces the fiscal-drag effect of the current UK freeze policy, which is the correct counterfactual for a reform comparison.

CGT and dividend allowances not applied at percentile level

The £3,000 CGT Annual Exempt Amount (AEA) and the £500 Dividend Allowance are retained as reference parameters but are not applied in the per-percentile calculation. The reasoning is that both allowances are designed to exempt small incidental amounts at individual level: real-world CGT has approximately 348,000 taxpayers, so total AEA consumption is around £1 billion against £81.8 billion of gains (1.2 per cent). Applying the AEA per-adult across all 53 million adults in the model over-counts the allowance by roughly two orders of magnitude and wipes out CGT in most percentiles. The same logic applies to the Dividend Allowance: dividend income is concentrated in a minority of adults, and a per-adult application of the £500 allowance zeros out dividend tax in all but the top percentiles.

Skipping these allowances introduces small overstatements — approximately £0.2 billion on a £12 billion CGT yield, and similar magnitude on dividend tax — which are acceptable and offset by other modelling simplifications (flat rate split by percentile-mean income band rather than individual-level band calculation). A more refined treatment would apply the allowances at aggregate level (taxpayers × allowance) or scale per-percentile by the fraction of adults in each percentile who are actual CGT or dividend payers; these refinements are deferred.

IHT approximation

Under the current system, inheritances are taxed through IHT on the estate; under NC, they are deposited into National Savings Bonds and NC is paid on withdrawal at the recipient's marginal rate. For ordinary inheritances, the nil rate band allowances in the current IHT system are replaced by delayed income using tax sheltered NS&I accounts, then taxed at in-year NC rates that will be below the current 40% IHT rate. NC on inheritance at steady state taxes the full annualised flow at each recipient's NC marginal rate, reflecting a 10-year withdrawal window cohort structure at equilibrium.

Inheritance is treated simply in this model. Inheritance and other irregular receipts are taken from the Wealth and Assets Survey and taxed as income at each recipient's National Contributions rate, rather than modelled at the level of the estate. This introduces two errors of opposite sign. The survey basis overstates the ordinary flow of inheritances, and it barely reaches the largest estates, so it understates the top-tail transfers the reform is designed to catch. On initial inspection the two are of similar size, so they broadly offset at the level of the flow. The residual is left to a later, more granular treatment of the top of the distribution. Looking at the tax rather than the flows, the two do not cancel out from a revenue perspective. The inheritances overcounted are ordinary ones, taxed at low rates. The large estates undercounted would be taxed at high rates. So putting both right would add more tax than it takes away. If anything the model undercounts the revenue from inheritance rather than overstating it. Reconciliation is left to the next iteration of the model, but the direction is not in doubt.

Two kinds of deferral

The first kind defers recognition. Funds sheltered in designated National Savings accounts have not yet arisen as income: the income arises on withdrawal and is taxed at the recipient's NC rate for that year. Until then nothing is due, nothing is owed, and no interest runs, for the same reason that no interest runs on a pension not yet drawn. Sheltering defers the point of charge, not the liability.

The second kind defers payment. A deferred Property Tax bill, or a real-gain charge being settled by instalments, is tax that has fallen due and remains owed. It is held as a loan from the state: the full amount stands, simple interest accrues, the debt is secured by a first charge on the property, and it is settled without fail at the next sale or transfer. Nothing is reduced, relieved, or forgiven. Only the timing of the cash moves, and it moves at a price.

The distinction answers the two criticisms usually made of deferral in a single breath. Recognition deferral is not avoidance, because there is no liability yet to avoid; the charge travels with the sheltered balance and is collected as the income arises. Payment deferral is not an escape, because the debt survives in full, senior and interest-bearing, and every route out of it, sale, transfer, or death, is a collection point.

The programme offers no value reliefs: no agricultural relief, no business relief, no trust wrapper, no discretionary hardship reduction. Where the current system forgives value, the programme extends time.

The reason is structural rather than austere. Every relief in the current system began as a principled protection and matured into a planning industry, because a relief that shrinks a liability is worth structuring into. Time to pay at full freight is not. The entire reward for qualifying is a neutrally priced loan, repaid in full, secured on the asset; there is nothing behind the boundary worth reaching for, so the boundary needs no policing. That is why the programme can afford generous access to deferral while refusing every reduction: the generosity is safe precisely because the prize is only time.

Two rules govern who may defer what.

In-life deferral of the annual Property Tax is elective and open-ended, and is therefore tightly scoped: it is available only for the owner-occupied principal residence. The test is not the asset but the consequence of a forced sale. A household stripped of its home loses one of the basic services, shelter, care, food, education, information, transport, and legal recourse, that the programme exists to secure, and no other asset class fails that test. A second home, an investment property, a portfolio, or a business can be sold to meet a tax without depriving anyone of a basic service, and so each pays as it falls due.

Time to pay at a transfer is the opposite case. The trigger is a death, not a choice; the liability is fixed at that moment; and the accommodation ends, at the latest, at the next transfer. Because it is bounded and involuntary, it is universal: the real-gain charge arising at a death may be settled by instalments over up to ten years wherever the asset is land or property, on the model of the existing statutory instalment provisions for tax on land, with no test of the owner's character, use, or purpose. Chosen, open-ended deferral is confined to the home; imposed, self-terminating liabilities get time, on any land, for anyone.

The refusal of agricultural and business reliefs is not indifference to stewardship. Land held, worked, and cared for across generations is a tradition with real value, and the households that raise their children inside it produce some of the country's best land managers. The tax system is simply the wrong instrument for honouring it: a relief pays for the label rather than the practice, favours the incumbent over the newcomer, and is invisible when it fails.

Stewardship is a public good, and public goods are purchased. Nature-outcome agreements, land-management contracts, and conservation covenants pay for stewardship directly, verifiably, and at a price the public sets, to whoever delivers it, whether the heir raised into the tradition or the newcomer who arrives at it. The tax system, meanwhile, is neutral on lineage: patience for anyone, discount for no one. A family whose stewardship is real can meet the charge from the land's own income across the time the programme allows, and in doing so demonstrates the tradition rather than claiming it. Nothing in this structure prevents a family keeping its land across generations; it only ends the subsidy for doing so.

Every deferred liability settles at its own owner's transfer. Any transfer, a sale or a death alike, accelerates outstanding balances: deferred Property Tax vintages, unpaid real-gain instalments, and any arrears are collected from the proceeds or the estate before anything passes on. What passes to the next holder is only what has not yet arisen as anyone's income, and it passes as a fresh receipt taxed in their hands. Debts do not travel; receipts do.

The consequence is that deferral is confined to a single ownership. One life's accommodations cannot be stacked on another's, no balance compounds across generations, and at every death the slate is settled before it is passed. The programme's answer to dynastic accumulation is not a prohibition but an arithmetic: each holding pays its own way, in full, once per owner.

All deferred balances are held as loan accounts at National Savings and Investments, which already holds the sheltered receipts on the recognition side. One institution services every deferral in the programme, and one construction prices them: simple interest, fixed per vintage at the rate prevailing when each year's liability arose, so that any balance can be verified with basic arithmetic. Liabilities attaching to the owner-occupied principal residence accrue at Bank of England base rate plus one percentage point for the elective Property Tax deferral and at base rate for charges arising at a transfer; liabilities on any other land or property carry the one point premium throughout. The premium is the margin that keeps deferral from becoming cheap credit for owners who have market alternatives; the base rate is the state's own cost of funds, at which patience is neither a subsidy nor a penalty.

Each deferred liability is registered against the property as a restriction: the title cannot be sold or remortgaged without settlement through NS&I, and the restriction is visible on any standard search. The balance itself is not published. A household's deferred tax is its own affair until it transacts, at which point the party who needs the figure, a lender advancing against the property or a buyer completing on it, obtains it as every lender already obtains a borrower's financial position: by requiring sight of the NS&I statement as a condition of the advance, and by redemption at completion. The public register records that the state must be settled; it does not announce what a family owes.

Up to three records can stand against a property, and they rank in a fixed order: the real-gain charge first, then charges on receipts, then Property Tax, all ahead of any private charge. A mortgage lender left short at a sale by the senior public charges is protected by the same mechanism the Right to Sell provides, conversion of the uncovered balance into long bonds, so that seniority reorders the queue without destroying the lender's claim.

Where deferred balances come to equal the value of a home, the accommodation ends in acquisition rather than forgiveness: the property transfers into public ownership in satisfaction of the debt, the occupier is offered a secure tenancy, and any junior lender is made whole in bonds. The state's claim concludes in an asset added to the social housing stock, not a write-off, and the occupant keeps their home on new terms rather than losing it.

The principles stand: tax inheritance as income to recipients, remove the seven year escape route used by large estates, look through trusts, do not treat inter-spousal transfers as inheritance until passed on, and treat everyone equally across the income spectrum. This treatment removes the penalty the current system imposes on unexpected death.

Employer NICs excluded from the headline comparison

NC replaces Income Tax, Employee and Self-Employment NICs, Capital Gains Tax, Dividend Tax and Inheritance Tax. Employer NICs continue unchanged under NC. A memo column shows Employer NIC liabilities under both systems (identical in value) for completeness but excludes them from the NC-versus-current headline comparison.

NC progressive schedule

NC is parameterised by two anchor rates: a Base rate (the marginal rate at P50) and a Top rate (the marginal rate at P90 and above). Marginal rates at intermediate percentiles are linearly interpolated:

• For p ≤ 1: marginal rate = Base / 50.
• For 1 < p ≤ 50: marginal rate = (p / 50) × Base.
• For 50 < p ≤ 90: marginal rate = Base + ((p − 50) / 40) × (Top − Base).
• For p > 90: marginal rate = Top.

The specific Base and Top rate values are policy choices set out in the main report body and are not fixed by this appendix. For each percentile the per-adult NC liability on income comprises two parts: tax accumulated on all income slices below the percentile's lower bound (computed recursively), plus tax on the partial slice from the lower bound to the percentile's mean NC Total Income. NC on benefit income is computed separately at household level using the microdata pass documented in the NC on Benefits appendix.

VNC voluntary-revenue treatment

The Voluntary NC (VNC) threshold applies to individuals whose NC Total Income falls below the threshold. Such individuals are voluntary remitters: they are assessed a liability but choose whether to pay. Above-threshold individuals pay compulsory NC on every income slice from £0 up to their total income. The specific VNC threshold value is a policy parameter set in the main report.

The implementation exploits the fact that NC Total Income is individual-level valid at percentile level — each percentile's lower and upper bounds describe every individual's income range within the band — while component-level values are not. A threshold applied against a component mean (for example, VNC subtracted from mean benefit income) misclassifies individuals by averaging recipients and non-recipients together before the non-linear threshold test. The VNC threshold is therefore applied against the sort variable, not against components.

For each percentile, a voluntary fraction f is computed: 1 if the whole income band lies below VNC, 0 if it lies above, and proportional for the single percentile straddling the threshold (f = (VNC − lower bound) / (upper bound − lower bound)). The straddle approximation assumes uniform distribution across the band; its error is confined to one percentile and is numerically small. At-risk revenue = percentile's total NC revenue × f. A remittance rate parameter (default 0 per cent) determines what fraction of at-risk revenue is actually collected. Net NC revenue = total NC revenue − at-risk × (1 − remittance rate). At 100 per cent remittance, VNC is fully disabled and everyone pays in full; at 0 per cent, below-VNC individuals pay nothing.

NC on benefits — referred to the NC on Benefits appendix

NC on benefits cannot be computed accurately at percentile level because the household-level threshold structure of the NC-on-benefits design depends on household composition (dependent children in particular) that varies within each percentile. A per-percentile calculation that applies a marginal rate to mean benefit income per adult produces a diagnostic figure only, not a revenue estimate.

The authoritative treatment of NC on benefits — including the principled basis for the taxable benefit base, the dependent-child allowance structure, the household aggregation method, and the resulting revenue and distributional figures — is set out in the companion NC on Benefits appendix. That appendix uses a separate microdata pass against the same FRS 2022-23 base year used by this tax model, with household-level calculations that the per-percentile model cannot replicate.

Within this model, the per-percentile NC-on-benefits figure in Tax Comparison column Q exists only as a diagnostic. The headline NC on benefits revenue used in Results is supplied from the microdata pass via a manual override in Parameters (currently rows 63–64). Any question about the NC on benefits treatment, aggregate figure, distributional breakdown, or methodology should be resolved by reference to the NC on Benefits appendix, not this one.

Revenue aggregation — two views

NC revenue can be aggregated two ways that sum to the same total (adjusted for the benefits-NC add-on described above):

Per-adult aggregation attributes revenue to the adults classified in each percentile. For percentile p: V[p] = B[p] × S[p], where B[p] is the adult count and S[p] is the per-adult NC liability (accumulated tax across every slice of NC Total Income from £0 up to the individual's income, plus NC on benefits where applicable). Each adult is counted once, in the percentile where their total income falls. Per-adult aggregation is used for the VNC voluntary treatment, because voluntary-versus-compulsory is a property of the taxpayer (determined by whether their total income falls below VNC), not of the income slice.

Per-slice aggregation attributes revenue to each income slice (the band between adjacent percentile lower bounds), counting all payers who are taxed on that slice. For slice p: revenue = L[p] × [(upper[p] − lower[p]) × pop_above + B[p] × (D[p] − C[p])], where pop_above is the total adult count in all higher percentiles, D[p] is the percentile's mean NC Total Income, and C[p] is the lower bound. The first term is the full-slice contribution from payers whose income sits in a higher percentile; the second is the partial-slice contribution from payers whose income is within percentile p itself.

Both views are complete in themselves. They differ by total NC on benefits, which is included in the per-adult view (because that liability is a property of the adult) but has no counterpart in the per-slice view (because benefits do not participate in the NC Total Income sort and therefore do not correspond to any slice). The per-slice view is useful as a memo to show how the progressive schedule generates revenue across income bands; the per-adult view is the operationally relevant aggregation for revenue forecasting and for voluntary-compliance analysis.

Importantly, below-VNC slices are not inherently voluntary. An adult whose total income exceeds VNC pays compulsory NC on every slice of their income, including any slices below the VNC threshold. Only the total liability of below-VNC adults is at risk. This is the conceptual reason the VNC calculation operates on per-adult aggregation rather than per-slice aggregation.

Parameters and configurability

The Parameters sheet concentrates all editable inputs:

• NC rates: Base, Top, VNC threshold.
• NC switches: benefits-taxed on/off, voluntary NC remittance rate.
• FRS coverage scalars: dividend, investment, benefit.
• Current-system rates and thresholds: personal allowance and taper, basic/higher/additional bands, dividend rates and allowance, Employee NIC thresholds and rates, Class 4 NIC thresholds and rates, Employer NIC threshold and rate, CGT AEA and rates, IHT effective rate.

Uprating inputs (Table 3.1 percentile points, benefit and state pension uprating rates, WAS factor, CGT update-year total, CPI factor) are held in Section A of the Uprating sheet alongside source-note references.

Changes to any Parameters cell propagate automatically through the Uprating intermediate block, the re-sort, the sorted block, Tax Comparison, and the Results summary.

Aggregate validation

Base-year aggregates (£ billion) before scaling, for reference:

After uprating and coverage scaling, the model produces an operating-year NC Total Income base in the region of £1.9 trillion and a taxable benefit base whose detailed treatment is set out in the NC on Benefits appendix. The current-system comparator reproduces published HMRC receipts for Income Tax, CGT, Dividend Tax and Inheritance Tax (after the dedicated scalars and surgical adjustments described under "FRS coverage scalars" above). Employer NICs are shown as a memo column but fall outside the NC comparison since they are unchanged under NC.

Specific revenue figures under NC depend on the Base and Top rate choices made in the main report and the VNC threshold setting, and are therefore reported in the main report rather than here. The aggregation mechanics described in the preceding "Revenue aggregation — two views" section are rate-independent and reconcile across any rate configuration.

Key design choices — summary

• NC Total Income is the sort key; benefits are passenger columns. This avoids benefit-income variation disturbing the ranking used for the progressive schedule and is consistent with the observation that benefit-recipient status is not the intended rate-determining signal.
• Thresholds that apply to nearly every adult (personal allowance, NIC Primary Threshold, Class 4 LPL, Employer NIC ST) are applied at percentile-mean level because the per-adult approximation holds. Thresholds that apply only to a subset of taxpayers (CGT AEA, Dividend Allowance) are skipped at percentile level and flagged in methodology.
• VNC is applied against the sort variable (NC Total Income) because it is individual-level valid at percentile level. Component-level thresholds (notably household-VNC on benefits) require microdata treatment.
• Scalars for FRS coverage gaps are applied once, in the Uprating intermediate block, so they flow through the re-sort and all downstream calculations consistently. Defaults are calibrated to tax-yield-match.
• Revenue can be aggregated per-adult or per-slice; the two views reconcile and serve different analytical purposes. Per-adult is used for voluntary-compliance analysis because voluntary status is a property of the taxpayer.

Known limitations

• The per-percentile NC-on-benefits figure (Tax Comparison column Q) cannot apply the household-level allowance structure of the NC design and is therefore a diagnostic, not a revenue estimate. The headline NC on benefits revenue comes from the microdata override supplied by the NC on Benefits appendix. All questions about the NC on benefits methodology, aggregate figure, and distributional breakdown are properly resolved by reference to that appendix.
• The model's base year is FRS 2022-23, which coincides with a period of exceptional one-off Cost of Living Payments (approximately £15 billion nationally). These payments flow through the benefit passenger columns via the standard benefit uprating factor. The effect on NC on benefits revenue is documented in the NC on Benefits appendix. The income-side calculations (Income Tax, NIC, CGT, IHT and NC on income) are unaffected. A rebuild on FRS 2023-24 or 2024-25 (the latter incorporating DWP's new administrative-data-linked variables) would remove this artefact.
• FRS coverage scalars correct aggregate tax yields but not the distributional shape of under-recorded components. Percentile-level effective rates for dividend and savings income reflect the captured subset, not the full population.
• Small ripples in per-percentile NC effective rate on NC Total Income plus benefits (up to about 0.1 percentage points over roughly 15 percentiles) arise from benefits being a passenger column whose mean varies non-monotonically across percentiles. These do not affect aggregate revenue.
• The straddle approximation in the VNC voluntary calculation assumes uniform distribution of individuals within one percentile's income band; error is confined to that single percentile.
• Treatment of pension contributions. National Contributions retains the current tax treatment of pensions: employee contributions are deducted from NC income before the rate applies, employer contributions are not taxed as income, and pensions in payment are taxed as income at NC rates. The distributional model is built on FRS gross earnings, which include employee contributions made through net pay and relief at source arrangements (salary sacrifice contributions are excluded at source). Both the NC and current-system calculations are therefore performed on the same unrelieved base, and all comparisons between the two systems are like for like. Reflecting the deduction explicitly would reduce modelled NC revenue from earned income by around £10 billion a year. HMRC reports £14.6 billion of individual contributions to personal pensions in 2023 to 2024 (relief at source, including the basic-rate top-up), and employee contributions to occupational net pay schemes add approximately £20 billion, of which around £15 billion sits in the public service schemes. Because the current-system comparator relieves the same contributions at lower marginal rates on the same base, the net effect on the additional revenue raised is of the order of £3 billion a year at steady state, within the programme's £38 billion of annual fiscal space.
• Capital gains are allocated to base-year percentiles by income band from HMRC Table 3.2, and real-value uprating strips inflation from nominal growth. A full behavioural model of CGT realisation under NC rates is not attempted and would require separate analysis.
• Inheritance realisation behaviour. The largest behavioural margin in the NC model is the timing of withdrawals from National Savings Bonds. The modelled treatment is conservative on rates: the 10-year withdrawal assumption spreads each inheritance across years, taxing it at moderate marginal rates rather than the top rates a lump-sum assessment would produce, and the cohort structure defers full recognition of the revenue to year 10 of operation. Slower withdrawal than assumed defers revenue further but does not extinguish it: there is no uplift at death, the NC liability travels with the sheltered balance, and deferred balances sit on deposit with NS&I, funding the Exchequer until withdrawn. Faster withdrawal front-loads revenue at higher marginal rates. The five-year programme window in the Macro Cashflow leans primarily on PAYE income, with only the first withdrawal cohorts inside the window, so the fiscal space presented is comparatively insensitive to this margin; realisation timing shifts steady-state revenue between years rather than changing its total.

Further iterations of the model are planned for later in 2026 to use more recent admin linked FRS data as the base and to incorporate more detailed treatments of the known limitations.