Abandoned homes in Japan, akiya, are usually sold on price and charm. Almost nobody sells them on risk. So we did the opposite: we took every one of the 549 empty homes in our catalogue and ran its exact coordinate through Japan's official government hazard datasets, one home at a time. This is what the data says about how dangerous, or how safe, these houses actually are.
The short version: most catalogued akiya are not in a designated hazard zone, the real headline is the age of the buildings, and, contrary to what buyers assume, homes in a hazard zone are not sold at a discount.
How many akiya are in a designated hazard zone?
We checked each home against three separate national inundation and slope datasets: fluvial flood zones, landslide-warning zones, and tsunami-inundation zones. A home counts as "in a zone" only if its coordinate falls inside a formally designated polygon.
| Hazard | Homes in a designated zone | Homes with usable data | Share |
|---|---|---|---|
| Landslide warning zone | 56 | 459 | 12.2% |
| Fluvial flood zone | 37 | 459 | 8.1% |
| Tsunami inundation zone | 20 | 459 | 4.4% |
The denominators matter, and we are deliberate about them. Not every home can be scored: some coordinates are too coarse (a listing pinned to a town centre, not a plot) and some sit outside a dataset's published coverage. We exclude those homes from the denominator entirely rather than quietly counting them as "safe". So "12.2% in a landslide zone" means 56 of the 459 homes we could actually evaluate, not 56 of all 549. That is the honest number, and it is the one we publish.
What about earthquakes?
Every point in Japan carries earthquake risk, so the question is not "is there any" but "how much". We use the national seismic-hazard model's 30-year probability of strong shaking (intensity 6-lower or worse, the level at which older houses start to fail).
Across the 483 homes with a seismic score, the median 30-year probability of strong shaking is 7.3%, and the mean is 10.2%. The distribution is heavily skewed toward the low end, but it has a real tail.
| 30-year probability of strong shaking | Number of homes |
|---|---|
| 0 to 10% | 261 |
| 10 to 20% | 167 |
| 20 to 30% | 41 |
| 30 to 40% | 9 |
| 40 to 50% | 2 |
| 70 to 80% | 2 |
| 80 to 90% | 1 |
Most catalogued akiya sit in relatively low-probability ground. But three homes carry a 30-year strong-shaking probability above 70%, and the single worst reaches 81.8%. Those are precisely the homes a buyer should never buy without a structural assessment, and precisely the homes a glossy listing never mentions.
The real risk is the age of the building
The most consequential hazard fact is not a flood polygon. It is the calendar. Japan tightened its seismic building code in 1981, and a house built before that ("old code", pre-1981) is materially harder to insure, harder to mortgage, and more expensive to make safe.
Of the 548 homes where we could establish a build year, 177 (32.3%) are pre-1981 old-code buildings. Nearly a third of the catalogue was built to a seismic standard the country decided was not good enough over forty years ago. For a buyer, that is a retrofit budget, not a detail.
The finding buyers do not expect: risk is not priced in
Here is the counter-intuitive result. You might assume a house in a flood or landslide or tsunami zone would be cheaper, the market pricing in the danger. It is not.
Among the 459 homes we could fully score across all three inundation hazards, we compared the median asking price of homes in at least one hazard zone against homes in none:
| Group | Homes | Homes with a price | Median asking price |
|---|---|---|---|
| In at least one hazard zone | 99 | 65 | 3,800,000 yen |
| In no hazard zone | 360 | 247 | 3,610,000 yen |
The hazard-zone homes are, if anything, marginally more expensive, not cheaper. The rural akiya market is thin, fragmented, and opaque enough that hazard status simply is not reflected in price. That is the single most important thing this study shows: you cannot infer a home's safety from its sticker price, and a cheap akiya is not cheap because it is dangerous. You have to check the hazard data yourself, which is exactly why we publish it free on every listing.
Methodology
We analysed the 549 akiya in the Engawa catalogue as of the data snapshot on 8 July 2026. Each home was geolocated and scored independently against these official government datasets:
- Flood: national fluvial-flood inundation zones (MLIT).
- Landslide: designated landslide-warning zones (MLIT / prefectural).
- Tsunami: tsunami-inundation zones (MLIT / prefectural).
- Earthquake: 30-year probability of strong shaking from the Japan Seismic Hazard Information Station (J-SHIS, published by NIED).
- Building era: derived from the recorded build year against the 1981 seismic-code boundary.
Coverage is reported honestly. A home is excluded from a hazard's denominator when its coordinate is too coarse to score reliably or falls outside the dataset's published coverage, never counted as safe by default. Every figure here is reproduced by a committed analysis script (`backend/seo/research_stats.py`) that reads the same database the site serves from, so any journalist can re-run it. Absence of a designated zone at a coordinate is not proof of zero risk; it means no formally designated hazard polygon covers that point in the national data.
Related reading: our guide on whether it is safe to buy an akiya, how to read a Japanese hazard map, and the subsidy league table for the money side. Browse the homes themselves, each with its own hazard panel.