Estimated Gross Rental Yield Ranking

Estimates a gross rental yield from the town/chome's condo transaction unit price (JPY/m²) and the city/ward-level rent benchmark from e-Stat's Housing and Land Survey (private rental, per-tatami-mat rent converted to JPY/m²). Since a rent benchmark only exists at the city/ward level, we apply a hedonic-regression correction — using land-price ratio, building age, and station walk time as explanatory variables — to bring it closer to a town/chome-level estimate.

① Location premium estimate: from same-city/ward condo transactions (by town/chome × year), fit log(unit price ÷ city average unit price) = intercept + β₁×log(town land price ÷ city average land price) + β₂×(town avg. building age − city avg. building age) + β₃×(town avg. station walk minutes − city avg. station walk minutes) via ordinary least squares (OLS)② Rent correction: town/chome rent estimate = city rent benchmark × exp(town/chome values plugged into ①'s formula, clamped to 0.4–2.5)③ Gross yield (%) = (② rent estimate (JPY/m²/month) × 12) ÷ unit price (JPY/m²) × 100 (the 95% confidence interval is derived from the spread in unit prices)

*No official town/chome-level rent statistics exist. e-Stat also has no rent breakdown by station walk time (its "distance zone" categories measure distance from the metro-area center, a different concept from station distance), so we instead use official/benchmark land prices (National Land Numerical Information L01/L02) as a proxy for the location premium in land value. For station distance, we use the actual "nearest station: minutes" figures from MLIT's transaction data (see the "Transaction data CSV import" section) and include the gap between a town/chome's average station walk time and the city average as a third explanatory variable in the hedonic regression (measured: R²=0.708).*This station-distance correction is applied only to the rent side (the numerator). We do not apply a station-distance correction to the town/chome's average transaction unit price, which forms the denominator of the yield — the unit price is an actual, realized average that already reflects station distance as priced in by the market, unlike the rent side, which requires extrapolating from a city-level average.The regression is an indirect correction based on the simplifying assumption that land price, building age, and station distance move proportionally with rent; it is not an actually-measured town/chome-level rent. If no land-price data is found within 3km of a town/chome, or the regression sample size is insufficient, no correction is applied and the city average is used as-is (indicated by a ⚠ mark in the ranking table). Rent data is only tabulated at the "city/ward" level, so it is unavailable for some municipalities that include unincorporated areas (covering roughly 70% of the 248 target municipalities). Vacancy rate is for reference only (the share of all housing stock in the same city/ward, not limited to rental units). Actual yields do not account for management fees, reserve-fund contributions, vacancy risk, or brokerage fees, so always run your own income/expense simulation for any specific property before purchasing.