The Retrofit Gap Index measures how far the housing stock of England & Wales sits from the EPC band-C standard, what it would cost to close that gap, and what the below-standard stock is worth. This is the full account of how it is built — the sources, the joins, the models and, just as importantly, the limits.

18.1m

de-duplicated dwellings

48.9%

below EPC band C

8.0m

dwellings matched to a sale

Sources

The Index fuses two open national datasets, both under the Open Government Licence:

  • EPC Open Data (DLUHC) — every domestic Energy Performance Certificate for England & Wales: current band, floor area, address, tenure, local authority and constituency.
  • HM Land Registry Price Paid — every residential sale, used to attach a price (and a price per square metre) to dwellings.

From certificates to dwellings

A property can hold several certificates over time. We keep the latest certificate per dwelling (by UPRN), filter to England & Wales domestic property types, and de-duplicate — collapsing 23 million raw certificates to 18.1 million distinct dwellings. “Below band C” means a current rating of D–G. The national gap of 48.9% is below-C dwellings over all rated dwellings, and the same ratio is computed for every local authority, constituency and postcode district.

The cost to reach band C

Each below-C dwelling is assigned a modelled cost to reach at least band C, by current band, from the English Housing Survey's Energy Report:

  • Band D → C: £6,472
  • Band E → C: £13,285
  • Band F or G → C: £18,858

These are cost-to-band-C figures, deliberately not the higher “all recommended measures” (cost-to-potential) numbers, which would inflate the headline. Summed across the stock, the modelled national bill is around £76 billion.

The price fusion

Land Registry carries no UPRN, so price is joined to EPC by address. This is the part that most affects accuracy, and where we have invested the most.

The matcher

Within each postcode, a dwelling is matched to a sale on the building number — taken from the Land Registry SAON where present (named terraces and flats record the unit there), otherwise the PAON, otherwise the leading number of the EPC address — corroborated by a shared street or building-name token, so a bare number can never grab the wrong property. A postcode that yields more than one candidate number is left unmatched rather than guessed. This lifts the share of dwellings carrying their own sale from a naïve 18% to 44%.

Uprating

Sale prices are uprated to the current data vintage by a national annual index (median sale price per year against the trailing-twelve-month median), so a 2015 sale and a 2025 sale are compared in the same money. Price per square metre is the uprated price divided by the certified floor area.

What we derive

  • Median £/m² per area — £3,295 nationally.
  • Observed C+ price spread — the matched-sale £/m² spread of C-or-above over below-C stock; +4.6% nationally.
  • Matched-sale exposure — the modelled total value context for below-C stock.
  • Sell-off ratio — below-C turnover relative to C-or-above.

Limitations

Aggregate figures are published openly in the data room and are free to cite with attribution. The same pipeline, pointed at a user's own addresses, powers the per-property exposure tool. Headline figures and area tables are on the Index and the map; the data licence and source notes are on the methodology page.