Run a hedonic regression on enough cities and a few patterns turn out to be genuinely global. The clearest one in our four-city panel: the implicit price the market pays per extra square metre (or per extra unit in a building, for New York) has fallen in every market since 2020. Bigger isn't winning anywhere.
| City | Size factor (regression) | Cumulative since Jan 2020 | Trough |
|---|---|---|---|
| New York | Residential Units (log, building scale) | −9.3% | −12.4% |
| London | Floor Area (sqm) | −4.5% | −4.5% |
| Paris | Log Floor Area | −1.9% | −3.1% |
| Singapore | Non-linear Floor Area | −0.5% | −0.7% |
The factor encodings are different in each city — London's regression has total floor area in square metres with no transform, Paris uses log floor area, Singapore uses a non-linear deviation from a typical-size baseline, New York doesn't have unit-level square footage and so uses the log of residential units in the building (a building-scale rather than unit-scale measure of "bigger"). Despite those mechanical differences, the sign is the same in all four.
London's size factor is the simplest to interpret. The regression has total floor area in metres squared with no transform, and the cumulative coefficient return is −4.5% since 2020 — i.e. an additional square metre commands roughly 4.5% less premium than it did five years ago, all else equal. The unwind has been monotonic, with no recovery, and the latest 12-month read (−3.8% per the October 2025 model) shows the trend is still active.
New York's factor is at building scale rather than unit scale (number of residential units in the building, in log), so it isn't directly comparable to London's "per extra sqm" interpretation. But it's the largest swing of the four (−9.3% cumulative, −12.4% at the trough), and the direction is consistent: per-sqm price growth in larger buildings has lagged smaller buildings. Combined with the simultaneous +4.1% cumulative return on the "elevator building" dummy, the picture in New York is: smaller, older, lower-density buildings have outperformed the bigger high-rise stock.
Paris's log-floor-area factor has moved less in magnitude (−1.9%) but the path is the same: a drawdown to −3.1% during the 2023–24 correction, then a partial recovery. Singapore's non-linear size factor is the smallest absolute move (−0.5%) — the HDB resale market is also the most regulated of the four, with relatively standardised unit sizes per flat type, so there is structurally less variance in size for the factor to express.
This is a four-city snapshot, not a causal model, but the obvious story is post-2020 affordability. Higher mortgage rates compress purchasing power, and the marginal buyer in every one of these cities is more rate-sensitive than the marginal seller — so demand rotates down the size ladder. Once the cross-city sign agreement is as clean as it is here, "it's just one market's quirk" is hard to argue.
Floor-area, room-count, and unit-scale factors with monthly returns and p-values.
Open the global comparison →Methodology: rolling 3-month cross-sectional OLS of log(price/sqm) on hedonic property characteristics per city. The "size factor" is each city's coefficient on its size variable; cumulative return is the cumulative geometric change since January 2020. See About for the full spec.