A flood is the most legible of climate signals: it is acute, dated, and mapped. If any hazard should move the housing market, this is it. Yet after a flood the market barely stirs. This paper reads six major fluvial floods between 2019 and 2024 through Environment Agency inundation outlines, and compares postcodes just inside a flood outline with their neighbours in the same local authorities just outside it, before and after the water arrived.
What the search data show
Searching for safer areas does not rise. Residents of freshly flooded neighbourhoods do not send more of their own enquiries elsewhere; the outbound response is a clean null in both tenures, and it survives every robustness check. The intent to flee, if it forms at all, does not show up as search.
Demand for the flooded homes themselves does dip, but only briefly. Holding the stock of active listings fixed, enquiries per listing fall by about a sixth in the first two months after a flood, then revert within a quarter. The effect is larger for major floods than for minor ones, which is the dose-response a real signal should show. It is a transient flinch, not a lasting reappraisal.
Prices do not follow at all. Once the composition of what sells is accounted for, sold prices show no post-flood change, and the pre-trends are flat. The gap between a brief wobble in demand and an unmoved price is the intent-to-move wedge: the market registers the event for a moment and then forgets it.
The little response there is leans regressive. The demand dip is larger in the less-deprived flooded neighbourhoods, so the households with the most options react most, while the most exposed and least mobile stay put. At coarser geography this signal washes out, so it is suggestive rather than settled.
Headline estimates
| Outcome | Effect at months +1 to +2 | Reads as |
|---|---|---|
| Outbound search for safer areas | No change (joint tests insignificant) | The intent to move does not appear in search |
| Enquiries per active listing | About −15% (95% CI roughly −23% to −5%), reverting within a quarter | A transient dip in demand for flooded homes |
| Composition-adjusted sold price | No change; flat pre-trends | Risk is not capitalised: the wedge |
| Demand dip by deprivation | Larger in less-deprived areas | The response leans regressive |
The estimates come from a within-town stacked event study across six flood events; inference uses event-by-authority clusters and an event-level wild cluster bootstrap, which is conservative with only six events. The demand effect is significant at the ten-percent level and marginal at five. See the data and methods for definitions, sources, and the full descriptive statistics.