The coastal-erosion paper tests whether a certain, dated, total loss is priced, using erosion projections, full listing histories, and repeat sales along the English coast. 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 sharpest test is a repeat-transaction panel whose dependent variable is the log price and whose regressor of interest is the interaction of erosion exposure with a calendar trend, with property and year fixed effects absorbing the level.
Variables
Outcome variable (dependent)
- Price: the sale price of a property, in logs. The repeat-transaction test follows the same property over time, so its view, street, and character are held fixed.
Independent and control variables (the inputs)
- Erosion exposure: whether a home sits inside a predicted land-loss polygon, with a remaining-life band (doomed, near-cliff, set-back, protected).
- Erosion exposure × trend: the exposure indicator interacted with a calendar trend, the coefficient that would be negative if a shortening life were marked down.
- Erosion rate: the predicted coastal recession in metres per year, the continuous exposure measure.
- Sea-view proximity: closeness to the cliff on a defended coast, which captures the amenity that must be separated from the erosion penalty in the cross-section.
- Floor area, bedrooms: size controls.
- Deprivation: the 2019 English Index of Multiple Deprivation, a place control.
Data source
Rightmove listings and enquiries for the English coast, 2018 to 2025, with the repeat-transaction test extended back to 2010 because it uses prices rather than search; the National Coastal Erosion Risk Map 2024 from the Environment Agency, giving predicted-loss polygons and a metres-per-year recession rate under intervention and no-intervention scenarios; HM Land Registry repeat sales; the corporate and overseas ownership registers together with the Rightmove landlord ledger for the end-of-chain owner; and the 2019 English Index of Multiple Deprivation.
Summary statistics of the analysis variables
The repeat-listing panel (N = 13,015 listing-years across 5,630 repeat-listed properties; eroding 1,691, protected 11,324):
| Variable | Role | Mean | SD | Min | Max |
|---|---|---|---|---|---|
| Price (£) | dependent | 418,400 | 481,800 | 20,000 | 9,750,000 |
| Log price | dependent | 12.62 | 0.78 | 9.90 | 16.09 |
| Eroding indicator | treatment | 0.130 | 0.34 | 0 | 1 |
| Calendar trend (years from 2018) | input | 3.62 | 2.30 | 0 | 7 |
| Eroding × trend | input | 0.493 | 1.51 | 0 | 7 |
Place and property inputs, released as means for the exposed (treated) and protected (control) groups:
| Variable | Role | Treated (eroding) | Control (protected) |
|---|---|---|---|
| Erosion rate (m/year) | input | 1.67 | 2.45 |
| Deprivation (IMD score) | input | 20.95 | 28.65 |
| Floor area (m², sale listings) | input | 118 | 116 |
| Bedrooms (sale listings) | input | 2.79 | 2.63 |
| Postcodes | — | 517 | 2,796 |
The erosion rate, deprivation, floor area, and bedrooms are released as group means rather than a property-level distribution, so their standard deviation, minimum, and maximum are not shown. There is no continuous remaining-life variable in the released data; remaining life enters categorically through the exposure bands.
Method
Four passes plus ownership. A naive eroding-versus-defended hedonic; a proximity difference-in-differences whose interaction term isolates the erosion penalty net of the sea-view amenity, clustered by frontage; a continuous erosion-rate hedonic; an exit analysis by remaining-life band; and the sharpest test, a repeat-transaction panel of the log price with property and year fixed effects and a band-specific calendar trend, from which a degree-of-capitalisation coefficient is read against the fundamental discount a certain loss implies. An end-of-chain ownership decomposition asks who is left holding the exposed homes. Comparison to protected hold-the-line segments controls the amenity that correlates with exposure.