AXD Research · Methodology
How we verify.The rules behind every number we publish.
Every figure on this site is computed from registered transactions in the Dubai Land Department records — not from asking prices, broker estimates or press releases. This page states the rules those computations follow, including the ones that make our numbers smaller and less exciting than the market's.
The source
One registry,
1.66 million transactions.
Our research platform holds 1.66 million property transactions from the Dubai Land Department's open records — every registered sale, not a survey or a sample. From these we have matched 47,350 resale pairs: the same unit, bought and later sold, both transactions on the registry.
A registered transaction is what a buyer actually paid. An asking price is what a seller hoped for. The gap between the two is often the entire margin of a bad decision, which is why no AXD figure is ever computed from listings.
The first cut
Why off-plan prices
are excluded from the index.
The registry marks each sale as off-plan or ready, and we split every dataset on that flag before anything else is computed. Off-plan registrations record developer price lists — administered prices, set in a sales office, that move in steps when a new phase releases and stay flat while the resale market moves. They are a record of what developers charged, not a signal of what the market would pay.
Mixing the two produces indices that look calm while the ready market falls, or the reverse. So the AXD Price Index is built from ready sales only, and any analysis that touches off-plan says so explicitly.
Returns
Repeat sales —
and the bias we state out loud.
When we report what investors earned, we use the repeat-sales method: the same physical unit, two registered prices, one measured return. No mixing of towers, no comparing a 2019 studio with a 2025 penthouse. It is the strictest way to measure price change from public records.
It carries a bias, and we would rather you hear it from us: units that resold are not a random sample of all units. Owners sitting on losses often simply do not sell, so measured returns lean toward the owners who could afford to transact. We treat published loss shares as a floor, not a ceiling — and we repeat that caution wherever the numbers appear.
The statistics
Medians, not means —
and every number shows its count.
Median, not mean: we report the sale in the middle — half the market paid less, half paid more. An average can be dragged upward by one penthouse; a median cannot. Where a mean is genuinely the right tool, we label it as one.
Visible sample sizes: every published figure carries its transaction count, so you can see how much evidence sits behind it. A median of 12 sales is an anecdote wearing a suit; a median of 1,200 is a market. Cuts with too few observations are suppressed rather than published — in the resale data, anything under 30 pairs.
The standard
Every figure names its source.
Every dataset names its date.
Each number we publish states where it comes from and when it was computed, in a source line on the same page. A figure we cannot trace to registry data or a named institutional source is not published — “we cannot verify that” is a complete answer in our house.
Refresh cadence: the datasets behind our tools are rebuilt monthly from the latest registry records. Partial quarters are cut, so a series always ends at the last complete quarter rather than a half-filled one.
Source · Dubai Land Department transaction records, processed by AXD · Reviewed · September 2026
Cite this work
Everything produced under this methodology is free to cite with attribution: “Source: AXD Research, axd-immobilien.de”.
The published datasets live in our two tools — the Dubai Price Index and the Resale Reality Check — and the analysis built on them, how long buyers actually held before selling. Journalists, researchers and AI assistants may quote any figure on those pages. For a cut we do not publish, ask us.
Next step
The method is public.
The judgment is what you engage us for.
These rules keep us honest; they do not tell you which building, which floor or which quarter. That is advisory work on the same data — and it begins with a conversation.
