UK: data fixes to model the High Value Council Tax Surcharge (mansion tax) correctly, and the related housing and council-tax gaps #1240

Summary

We need reliable national and local (constituency and local-authority) estimates of the High Value Council Tax Surcharge (HVCTS) and of a lower threshold. A £1.5m threshold is reportedly being considered for the 28 October 2026 Budget.

Today the national totals land in a plausible range, but for the wrong reasons: the band mix is wrong and local estimates aren't usable. This issue:

• collects the data fixes needed, in dependency order;
• links the open issues that already cover parts of them;
• lists the other housing and council-tax gaps found along the way.

It assumes Chronicle's Land Registry and house-price facts (PolicyEngine/chronicle#343) land shortly.

What's wrong now

These findings come from the 11 October 2026 local candidate uk-local-candidate-f100-s42-20261011T200435Z-108c1225 (staged in policyengine/populace-uk-private; local run uk-k25-h100000-ngfes-f0.5-u0.0216), compared with the incumbent eFRS (policyengine-uk-data 1.58.0). They are aggregates only. The details are in a private, disclosure-controlled analysis repo (PolicyEngine/uk-mansion-tax-microdata).

• The top tail is truncated and thin. See UK national release: main_residence_value keeps the WAS-imputed tail, so the surcharge's top bands stay empty #1112.
  • Owner-occupied homes above £2m in England come from a few dozen survey households.
  • All but 3 of the 543 English constituencies have fewer than 10 such households behind them.
• Nothing anchors where high-value homes are.
  • No target uses property values.
  • The English local-authority band-H targets are a signed exclusion.
  • Property values are imputed before households are given an area, then rescaled to regional means. UK national release: main_residence_value keeps the WAS-imputed tail, so the surcharge's top bands stay empty #1112 describes the stacked regional blocks this creates.
  • Constituency estimates correlate only weakly with VOA band-H stock (rank correlation 0.37).
• Value and council-tax band don't match. Only 15% of the homes above £2m carry band H.
• Coverage. Only owner-occupiers' main residences are modelled. The roughly 40% of liable properties that are second homes, let or company-owned are outside the model (see the notes on policyengine-uk#2164).
• Valuation and uprating.
  • Property values are uprated with GDP per capita, not house prices: Uprate main_residence_value with the OBR house price index, and value the High Value Council Tax Surcharge at April 2026 house prices policyengine-uk#2392.
  • UK national release: main_residence_value keeps the WAS-imputed tail, so the surcharge's top bands stay empty #1112 asks for a valuation-date contract.
• Deferral. The consultation's proposed test (household income under £35,000, capital under £16,000, plus disabled owners) depends on how income, capital and home value combine among high-value owners. That rests on the same thin tail.

Fixes, in dependency order

1. Diagnose and extend the top tail: #1112, steps 1–2

#1112's plan stands. This issue adds one requirement: support as well as values.

• Extending values for the same few dozen households would still leave most seats with fewer than 10 records.
• The fix needs more distinct high-value households. One option is a high-value support channel, like the SPI support copies, drawing values from the extended tail with consistent incomes, wealth, tenure and band.
• Size it so that the seats where Land Registry places most £2m+ homes clear the 10-household floor.

2. Local high-value-home targets from Chronicle's Land Registry facts (new; needs chronicle#343)

Chronicle stores only publisher values, so the targets are built here:

1. Take the latest Price Paid transaction per property (category A; types D, S, T, F).
2. Uprate it to the valuation date with UK HPI by local authority and property type.
3. Map postcode to output area with NSPL, then to 2024 constituency and local authority with oa21_pcon_jul2024_lookup.
4. Count homes above £1m, £1.5m and £2m by local authority and constituency, plus £2.5m and £3.5m nationally.

Then:

• Owner-occupied adjustment. Use census tenure by local authority (census2021_ts054_tenure_lad), with the OBR's roughly 40% non-owner-occupied share as a national control. Record the assumption, and prefer value-specific evidence if any appears.
• Stock corrections. Account for properties not sold since 1995 and for new builds. Check totals against the VOA dwelling stock, and seat sales counts and medians against ONS HPSSA by constituency.
• Binding. Bind the counts as a target family at local-authority and constituency grain, with tolerances that reflect the stock estimate's uncertainty. National band totals stay diagnostics against OBR Table 1.1: no back-casting, as UK national release: main_residence_value keeps the WAS-imputed tail, so the surcharge's top bands stay empty #1112 says.
• Acceptance.
  • Rank correlation by seat with the Land Registry stock and with VOA band H.
  • Regional £2m+ shares close to Land Registry's.

3. Impute property values after area assignment: #931, item (b)

Move was_wealth (for main_residence_value) and regional_uprating after geography assignment. Condition the draw on the assigned area, using the local price distribution from fix 2 and the HPSSA medians, instead of rescaling each region to its mean. Do the same for private rents (#1123, item 4).

4. Make council-tax band consistent with value: #796 (CT-B per-clone band redraw)

• Redraw or constrain the band jointly with value and area. A home worth over £2m in bands A–E should be rare.
• Use VOA CTSOP by LSOA (band mix by area) together with Land Registry values to define plausible band–value pairs.
• Then revisit the English band-H exclusion council_tax_band_h_spine_support_absent, which expires 2026-11-26 (WS-E spine deferrals: known bugs and signed inconsistencies awaiting their increment #796), and the micro-cell rule (UK local build (K=25, 2026-10-06): data-quality defects on the target surface and the spine #1123, item 7).

5. Valuation date and uprating: #1112 step 5, policyengine-uk#2392

• Declare the date main_residence_value refers to.
• Uprate it with the house price index, and value homes for the surcharge at April 2026 house prices.

6. Coverage beyond owner-occupied main residences (scope decision)

• Second homes and let property held by households. other_residential_property_value is a single total, so a banded charge per property can't apply. The options are:
  • carry a count of additional residential properties and values per property from WAS (new variables, plus engine support);
  • a documented approximation.
• Company-owned and non-UK-resident-owned homes. These are outside any household survey. Model them as an external block (counts and band mix from VOA, the OBR, or Land Registry's corporate ownership data) that adds to totals but not to households.
• Decide which of these is in scope before the Budget. Until then, label results "owner-occupied main residences only".

7. Deferral inputs

• Check how household income, capital and home value combine among high-value owners against WAS directly, including asset-rich, income-poor owners. The OBR estimated inability to pay from WAS.
• Check the disability flags.
• Expose both tests as parameters: the consultation's proposal (income under £35,000, capital under £16,000) and the OBR costing's assumption (net income under £30,000).

8. Gates and receipts for high-value homes

Add release diagnostics:

• national band counts against the OBR (diagnostic);
• regional £2m+ shares against Land Registry;
• rank correlation by seat against Land Registry and VOA;
• distinct source households per seat above £1.5m and £2m, and the largest household's share.

Receipts follow UKDS disclosure control: aggregates only, no maxima or minima, no cells under 10 source households, no differencing.

Other housing and council-tax gaps

Open issues still unaddressed, or only partly addressed:

• Council-tax level and liability. This is the rules-based liability: band × the local authority's Band D × the statutory ratios × discounts (WS-C deferred adjudications and the reconciliation layer #736, CT-C). It needs the engine's local_authority input (UK: supply the engine's local_authority input from the ladder's local_authority_code, plus council tax follow-ups from policyengine-uk#1855/#1864 and uk-data#482/#484 #953). Any Budget-day council-tax reform needs it: band changes, revaluation, second-home premiums.
• Council-tax micro-cells and the band-H deferral: UK local build (K=25, 2026-10-06): data-quality defects on the target surface and the spine #1123 item 7, WS-E spine deferrals: known bugs and signed inconsistencies awaiting their increment #796.
• Private rents in high-rent districts can't be reached from region-level FRS rents: UK local build (K=25, 2026-10-06): data-quality defects on the target surface and the spine #1123 item 4, fixed through UK graph: move geography assignment ahead of enrichment (geography-first reordering) #931(b).
• Working-age Housing Benefit, including supported and temporary accommodation, is missing: UK spine: housing benefit £2.7bn against OBR £12.1bn — pensioner and supported-accommodation HB missing after the UC migration #867. Its exclusions expire 2026-12-08.
• Rental income. SPI amounts undercount rental income against HMRC's property rental statistics (WS-C deferred adjudications and the reconciliation layer #736, reconciliation item).
• Migrating the primary residence value imputation: Migrate primary residence value imputation to Populace #140.

New, with no issue yet:

• Second-home and empty-home council-tax premiums. The MHCLG council taxbase counts are in the pinned feed, but a household file can't carry non-household dwellings (WS-C deferred adjudications and the reconciliation layer #736 notes this). Decide on a report-only block.
• Stamp duty base. Land Registry transaction counts by price band (chronicle#343) can also check the stamp-duty base (property_purchased and the first-time-buyer flag), if the Budget changes SDLT.

Dependencies

• High-value housing facts for the UK local calibration: HM Land Registry Price Paid Data, UK HPI, ONS house price statistics by constituency, the National Statistics Postcode Lookup, and the OBR High Value Council Tax Surcharge costing chronicle#343: Land Registry Price Paid, UK HPI, HPSSA, NSPL and the OBR costing. Its port is expected to merge shortly.
• Uprate main_residence_value with the OBR house price index, and value the High Value Council Tax Surcharge at April 2026 house prices policyengine-uk#2392 (uprating and valuation index); policyengine-uk#2164 (owner-only liability, merged).
• UK graph: move geography assignment ahead of enrichment (geography-first reordering) #931 (geography first), UK national release: main_residence_value keeps the WAS-imputed tail, so the surcharge's top bands stay empty #1112 (tail), WS-E spine deferrals: known bugs and signed inconsistencies awaiting their increment #796 and WS-C deferred adjudications and the reconciliation layer #736 (council tax), UK: supply the engine's local_authority input from the ladder's local_authority_code, plus council tax follow-ups from policyengine-uk#1855/#1864 and uk-data#482/#484 #953 (local_authority input).

Done when

• Owner-occupied band counts and the regional split of £2m+ homes are plausible against the diagnostics in fix 8.
• Constituency estimates track Land Registry stock.
• The seats that hold most high-value homes clear the 10-household floor.
• The coverage scope is documented, and release receipts report the fix-8 diagnostics under disclosure control.