The heat paper measures whether the market prices heat-resilient housing, across five margins built from listings, the energy-certificate register, and grid temperature. This page sets out the analysis variables, the sources, their summary statistics, and the estimators. The variable roles are those of the estimation code: the workhorse is a cross-sectional hedonic whose dependent variable is the log asking price and whose regressor of interest is the high-EPC indicator, with the covariates below held fixed.
Variables
Dependent variable
- Asking price: the listed sale price or monthly rent, used in logs. This is the outcome the hedonic explains.
Independent and control variables (the inputs)
- High-EPC indicator: whether the home carries an A to C energy-performance band. Its coefficient is the premium the paper tracks.
- EPC efficiency score: the continuous 1 to 100 score behind the band.
- Air conditioning: whether the listing text advertises cooling, the most direct marker of heat readiness.
- Floor area, bedrooms: size controls (bathrooms, living rooms, new-build status, property type, and region also enter as controls).
- Summer heat anomaly: the June-to-August maximum-temperature departure from the 1991 to 2020 normal, by area and year, the chronic-exposure regressor.
Data source
Rightmove listings from 2018 to 2025 (a hedonic sample of about 10.1 million with a valid energy band and positive price), matched to the English domestic Energy Performance Certificate register by unique property identifier; the cooling indicator from the listings text; Met Office HadUK-Grid temperature summarised to the summer anomaly; and the 2019 English Index of Multiple Deprivation for the equity cut.
Summary statistics of the analysis variables
Sale listings (N = 6,122,310):
| Variable | Role | Mean | SD | Min | Max |
|---|---|---|---|---|---|
| Asking price (£) | dependent | 401,828 | 475,526 | 10,000 | 82,500,000 |
| Log asking price | dependent | 12.64 | 0.70 | 9.21 | 18.23 |
| Floor area (m²) | input | 115.8 | 116.0 | 15 | 10,556 |
| Bedrooms | input | 2.91 | 1.07 | 0 | 10 |
| High-EPC (A–C) indicator | input | 0.437 | 0.50 | 0 | 1 |
| EPC efficiency score (1–100) | input | 65.5 | 11.8 | 1 | 100 |
| Air-conditioning indicator | input | 0.00055 | 0.023 | 0 | 1 |
Rent listings (N = 4,002,380):
| Variable | Role | Mean | SD | Min | Max |
|---|---|---|---|---|---|
| Asking rent (£/month) | dependent | 1,509 | 1,511 | 100 | 50,000 |
| Log asking rent | dependent | 7.10 | 0.61 | 4.61 | 10.82 |
| Floor area (m²) | input | 87.9 | 77.6 | 15 | 9,870 |
| Bedrooms | input | 2.35 | 1.23 | 0 | 10 |
| High-EPC (A–C) indicator | input | 0.557 | 0.50 | 0 | 1 |
| EPC efficiency score (1–100) | input | 68.6 | 10.4 | 1 | 100 |
| Air-conditioning indicator | input | 0.00109 | 0.033 | 0 | 1 |
Summer heat anomaly, the chronic-exposure regressor, by area and year (N = 54 area-years, 2017–2025):
| Variable | Role | Mean | SD | Min | Max |
|---|---|---|---|---|---|
| Summer heat anomaly (°C) | input | 0.86 | 0.92 | −0.55 | 2.54 |
Floor area is populated for 5,893,314 sale and 3,871,459 rent listings; the air-conditioning share is computed over all listings (8,376,744 sale, 5,439,449 rent). The indicator standard deviations are those of a 0/1 variable.
Method
The dependent variable is the log asking price. Five margins build on the same hedonic: a monthly cross-sectional regression of log price on the high-EPC indicator with the size, type, and region controls, read as the premium and summarised into a seasonal profile and an interrupted time series around the July 2018 and July 2022 heatwaves; a two-step regression of the area-by-year premium on the summer anomaly with area and year fixed effects; a pooled hedonic interacting the anomaly with the high-EPC and air-conditioning indicators; and an energy-band boundary regression discontinuity, whose interaction with heat gives a difference-in-discontinuities. Standard errors are clustered by local authority.