Tax Lab

Personal mode. Uses the actual UK (non-Scotland) Income Tax banding — Personal Allowance £12,570, then basic/higher/additional rate bands — and employee National Insurance between the Primary Threshold and Upper Earnings Limit. Does NOT apply the Personal Allowance taper above £100,000, and does not model Scotland’s separate income tax bands — both stated simplifications, not oversights.

UK mode.Uses HMRC’s own published “Direct effects of illustrative tax changes” ready-reckoner directly — the same figures HMRC uses in Budget scorecards — rather than grossing up the personal calculation, which would need a full income-distribution microsimulation this simulator doesn’t have. HMRC states these “account for taxpayers’ behavioural responses”, not a static costing. Scaling beyond 1 percentage point is a disclosed linear extrapolation of HMRC’s own per-point figure.

Employer National Insurance. Never affects your personal take-home pay directly — but who ultimately bears an employer NI change (wages, prices, or profits) is a genuinely contested economic question, deliberately left unquantified rather than assumed.

Wealth tax.A hypothetical ONE-OFF tax — assessed once on wealth at a fixed date, paid in 5 annual instalments — modelled directly on the Wealth Tax Commission’s own six published threshold/revenue scenarios (a flat 5% cumulative rate, net of their own 10% non-compliance deduction). The Commission explicitly recommends a one-off tax over an annual one: efficiency, avoidance resistance, administration. Values between two published thresholds are our own disclosed linear interpolation, never an extrapolation beyond their published £250,000–£10,000,000 range.

Corporation Tax.The revenue figure uses HMRC’s ready-reckoner directly — a single combined figure for the main rate (25%) and small profits rate (19%, on profits under £50,000), since HMRC does not publish them separately. The business-investment effect is NOT scaled from that figure — it quotes OBR’s own published assessment of the 2023 19%→25% rise directly, a single historical case study rather than a per-percentage-point elasticity, and is explicitly weaker (“our own inference”, not an OBR finding) for a rate cut than for the rise OBR actually studied.

Behavioural response (Income Tax & NI, UK mode).HMRC’s ready-reckoner figures are NOT a naive, no-behaviour calculation — they already assume people’s taxable income responds to the rate they face. Click “Follow this consequence” on the revenue card for four separate unintended-consequence channels — the taxable-income elasticity itself, incorporation/ dividend-shifting, migration, and forestalling — each quoting named primary studies directly, never rescaling a historical figure to your chosen rate. Where the evidence genuinely disputes the size of an effect (it often does, even between specialists), this simulator says so rather than picking a number.

“Zero the bottom, surcharge the top” exemplar. HMRC publishes Income Tax percentile shares only at 1%, 5%, 10%, 25% and 50% — there is no published top-2% point. We tested two interpolation methods against a real intermediate HMRC data point (additional-rate payers, ≈top 2.43%, actual liability share 37.6%): linear interpolation was off by 3.1pp, a Pareto-tail (log-log) interpolation — the standard method for extrapolating fine top-income percentiles — was off by only 0.5pp. Pareto is used as the central estimate; linear is disclosed as the more conservative alternative. The bottom-10% side needs no interpolation — HMRC publishes that share directly (0.4% of total liability, ≈£1.1bn). Every static figure is this site’s own derived estimate, never presented as an HMRC or OBR costing — the behavioural-context card sets it directly alongside HMRC’s own post-behavioural analogue and a published high-response bound from IFS, rather than picking one number.

Full source-by-source detail is recorded in the repository’s docs/policy-simulator/tax-research-audit-2026-09-04.md, docs/policy-simulator/tax-behavioural-response-research-2026-09-04.md and docs/policy-simulator/tax-exemplar-distributional-data-research-2026-09-04.md research notes.