Summary
Regional household_land_value in the Enhanced FRS is effectively locked to each region's share of property wealth. This flattens the London land premium and overstates rural/commuter-belt land, because in reality the land share of property value varies strongly by region (roughly ~80-85% in London vs ~45-55% in the South West and parts of the North).
Where this is baked in
1. The PolicyEngine-UK formula uses a single national scalar (household_land_value.py):
property_wealth_intensity = (
wealth.land.value.aggregate_household_land_value
/ wealth.property_wealth
)
return property_wealth * property_wealth_intensity + owned_landproperty_wealth_intensity is a national scalar, so regional variation in household land comes only from regional variation in property_wealth.
2. The regional calibration targets are built from property wealth too (mhclg_regional_land.py):
df["property_wealth"] = df["dwellings"] * df["avg_house_price"]
total = df["property_wealth"].sum()
return dict(zip(df["region"], df["property_wealth"] / total))
Each region's share of the national household-land total is set proportional to its property wealth, so the loss-matrix targets cannot pull the reweighted FRS toward a higher London land intensity than the national average.
Why this matters
For any LVT analysis that cuts by region, the regional distribution of burden is biased:
- London is understated as an LVT base.
- South East / East of England / rural areas with high property wealth but lower land share are overstated.
- Winners/losers by region under a council-tax → LVT swap are distorted in both directions.
Suggested direction
- Source regional land-to-property ratios from:
- ONS National Balance Sheet regional non-financial assets / AN.211 splits where available.
- Savills residential land value estimates by region.
- VOA / MHCLG data on dwelling values vs reconstruction cost as a land-share proxy.
- Replace the proportionality assumption in
_compute_regional_shares() with region-specific (land value / property wealth) ratios applied to regional property wealth.
- Companion change (separate issue in
policyengine-uk): make wealth.land.value.aggregate_household_land_value and/or property_wealth_intensity region-aware so the formula can consume a regional parameter rather than a national scalar.
Context
Surfaced while modelling a council-tax → flat-rate LVT swap for the UK (uk-land-value-tax). Switching our own aggregation from land_value to household_land_value removes the corporate-land-via-shareholdings flattener, but the remaining regional bias is inherited from the upstream calibration described above.
Summary
Regional
household_land_valuein the Enhanced FRS is effectively locked to each region's share of property wealth. This flattens the London land premium and overstates rural/commuter-belt land, because in reality the land share of property value varies strongly by region (roughly ~80-85% in London vs ~45-55% in the South West and parts of the North).Where this is baked in
1. The PolicyEngine-UK formula uses a single national scalar (
household_land_value.py):property_wealth_intensityis a national scalar, so regional variation in household land comes only from regional variation inproperty_wealth.2. The regional calibration targets are built from property wealth too (
mhclg_regional_land.py):Each region's share of the national household-land total is set proportional to its property wealth, so the loss-matrix targets cannot pull the reweighted FRS toward a higher London land intensity than the national average.
Why this matters
For any LVT analysis that cuts by region, the regional distribution of burden is biased:
Suggested direction
_compute_regional_shares()with region-specific(land value / property wealth)ratios applied to regional property wealth.policyengine-uk): makewealth.land.value.aggregate_household_land_valueand/orproperty_wealth_intensityregion-aware so the formula can consume a regional parameter rather than a national scalar.Context
Surfaced while modelling a council-tax → flat-rate LVT swap for the UK (uk-land-value-tax). Switching our own aggregation from
land_valuetohousehold_land_valueremoves the corporate-land-via-shareholdings flattener, but the remaining regional bias is inherited from the upstream calibration described above.