Analytics

What the market is doing this week

The Central Bank publishes what the market did last quarter, with a 45–90 day lag. We measure the flow of listings weekly. These are different things, and the second one is not computed anywhere else.

Stocks and flows — why these are different data

Official statistics measure states and levels: how many transactions closed, how much mortgage lending was issued. A listings panel measures flow: what was posted, what was withdrawn, what was discounted and how fast. The second turns before the first.

WhatOfficial statisticsListings panel
Unit of measurementa completed transactiona listing and what happens to it
Frequencyquarterlyweekly
Publication lag45–90 daysdays
Answerswhat has already happenedwhere the market is turning

Neither replaces the other. A transaction is a fact; a listing is an intention. But the intention is visible earlier, and you can act on it while the fact is still unpublished.

Why the published median misleads

This is not a theoretical caveat. On our data the raw median and the composition-adjusted index diverge as far as the sign — they answer "did prices rise or fall" in opposite directions.

Market and periodRaw medianComposition-adjusted
Tashkent apartments, March 2026−2.4%+0.7%
Tashkent apartments, April 2026+0.3%+5.1%
Used cars, May, year on year+1.2%−2.7%

The cause is the same in all three rows: what gets listed changes faster than what things cost. In March the raw median says prices fell and the index says they rose. For cars the raw number says growth and the like-for-like number says decline. The same effect shows up geographically: Tashkent city looks 62% more expensive than the surrounding region, but its median car is six model years newer.

Data: our own listings panel, February–April 2026 (apartments) and 2025–2026 (cars). All figures in US dollars. The index is a Laspeyres with February weights; for cars, like-for-like matching on model, year and mileage band.

What we compute

None of these appear in the standard Central Bank or international housing-indicator sets — and none can be computed without panel data on listings.

Fixed-weight price index

A Laspeyres index over district × segment × room-count cells. Separates price movement from changes in the composition of supply.

Days on market

Median time from posting to withdrawal. The first thing both an agent and a bank ask for.

Survival curve and absorption

Liquidity as a curve rather than a single point: what share of listings clears within 30, 60 and 90 days.

Price-cut rate and depth

A direct measure of negotiating power: how many listings cut their price, and by how much.

Cuts-to-raises ratio

A leading indicator — it turns before medians do.

Seller concentration

The share held by the top 100 sellers, plus a concentration index. Without this correction a median is agency-weighted, not market-weighted.

Duplicate and re-listing rate

The same property posted under several listings inflates every count-based statistic.

Price at development level

The primary market broken down by individual residential development. Nobody publishes this.

GeoScore-adjusted price

What a property costs relative to the quality of its location. There is no regional analogue.

Mortgage-mention share

A weekly demand proxy, against quarterly mortgage volume in the official statistics.

Urgency and negotiability share

Seller sentiment, measured from listing text, weekly.

Data freshness

The last-updated date shown next to the number. Against a quarterly report with a three-month lag, that is the product.

Market context

The official data we build on and reconcile against.

319,500
property transactions in 2025, +15.8% on 2024
21.2 trln
UZS of mortgage lending in 2025, +29%
~22%
of deals are mortgage-backed — about 70,000 collateral valuations a year
3–7 days
and $200–500 is what one manual appraisal costs

Source: Central Bank of Uzbekistan, 2025. We reconcile our coefficients against the Central Bank's own hedonic regression and publish the divergences rather than hiding them.

What we publish, and how often

One study a month with a specific finding, a method note and a downloadable data extract. Not product press releases.

A monthly study

One finding, computed on our data, with the method and the limitations described.

A method note

Exactly how it was computed, on what data, and what was left out. Without this the number is not worth citing.

A data extract

The slice the result can be reproduced from — or argued with.

Need a specific slice or a recurring extract?

Tell us which indicator and which segment — we will send the method and a data sample