Methodology
Both series are monthly and use official Dubai Land Department sales of Dubai residential flats and villas. They are price indexes, not valuations and not investment advice. Base for both is January 2010 = 100.
Paimio Dubai Matched Floorplate Index (PAIMIO-DXB-RSI)
A Bailey-Muth-Nourse monthly dummy regression on deed-to-deed pairs: Sell of Existing Properties only. Off-plan / Oqood transfers stay out of the headline. A floorplate is project, building, room band and procedure area. This is not a title-deed match. Same-size units in one tower can still collide. Floorplates with more than 12 prints are dropped as collisions. Dubai Pulse publishes a units register with property_id; the transaction tapes we use do not carry that id, so the series is not joined to it.
Pairs use total price, not AED per square metre. Holding period must be at least six months. Trim is on annualized return only, outside −50% to +100%. A five-year doubling stays in. Stage-two residuals are fit on holding period and the index is re-estimated with Case-Shiller interval weights. The published series is not smoothed. From March 2026 the official tape has no building name. A building is attached only when the historical floorplate or the project maps to exactly one tower. Sales that stay ambiguous or unseen are left out.
Repeat-transacting stock over-represents investors. Renovation between legs is unobservable and biases measured appreciation up. A new-build first sale as a first leg biases it down. Those are why this series and the hedonic are co-published: they fail differently. The 2014–2020 drawdown on this series is steeper than broker ready-stock prints of about −30%; that gap is noted, not forced out of the estimator.
Paimio Dubai Hedonic Index (PAIMIO-DXB-HED)
A chained adjacent-period hedonic. Each month is a new regression on that month and the two before it; only the last step is chained, then frozen. The model is log AED per square metre on size and size squared, off-plan, room, flat versus villa, age band (years since first project sale, with a missing-age dummy) and community. Standard errors are clustered by community. The published band is the one-step error on that month's link, not stacked history. Thin months look thin.
Sample is arm's-length residential flats and villas. Hotel apartments, offices and shops are dropped. AED per square metre must sit between 2,000 and 150,000 and inside the community 1st–99th percentile over the trailing year. Same-project same-day same-price blocks of five or more collapse to one print.
What the indexes are not
They are not a listing-price index, a unit-level valuation, or a forecast. They do not use asking prices. Repeat-sales matching is not a title-deed match. Machine IDs and citation wording are on How to cite.