Where does a square metre cost what in Iran, and where is Tehran's price heading?
Built on 516,947 real Divar listings across 420 cities, 3,447 individual Tehran ads, and the Central Bank of Iran's monthly transaction prices. Click any district, neighbourhood or province: every indicator on the page follows your selection.
1Tehran, district by district
Median asking price per m² for each of the 22 municipal districts and their neighbourhoods. The year bar sets the period: c. 1400 (a 3,447-ad sample), 1403 (91,836 listings) or the latest Divar collection; every figure carries the period it belongs to. Switch the indicator, drop to neighbourhood level, or set a budget.
The north-south gradient
All 22 districts ranked on the selected indicator. Hover to find it on the map, click to select.
Cheaper districts rose faster
Price per m² c. 1400 (x) against how many times it multiplied by 1403 (y). Districts with fewer than 15 ads in the older sample are left out.
2Price a home
A hedonic model estimated on 3,447 individual listings sets the premiums for size, rooms, parking, storage and elevator; the price level comes from each neighbourhood's listings in the period chosen on the year bar.
The estimate uses the price level of the period on the year bar. The slider moves it by hand; in 1403 the Kilid mark rolls it to Mordad 1405.
3Iran, province by province
Median asking price per m² by province, from 514,282 listings in the 349 cities with at least 30 ads (mostly 1403). Turn on cities to see each one.
Five largest markets, month by month
Composition-adjusted asking-price index (Ordibehesht 1403 = 100): the listing mix is held fixed, so the line shows price, not what happened to be listed.
Satellite towns vs Tehran
How many times the median asking price per m² multiplied between the older Divar sample (c. 1400) and 1403.
4Tehran's price over time, and the next 12 months
The Central Bank's monthly average transaction price (1395–1403) and Kilid's listing indicator (1404–1405). The forecast method was chosen by testing five methods on 35 historical forecast origins.
Year-on-year change, official series
Tehran's price moves in jumps (1397, 1399, 1402) separated by plateaus, which is why momentum forecasts fail.
Which forecasting method actually works?
Mean absolute % error by horizon, rolling-origin backtest on the official series. Lower is better; best per column in bold.
5How good is the pricing model?
Five-fold cross-validation; the neighbourhood price level is encoded out-of-fold so no listing sees its own price.
What a feature is worth
Price premium holding location and size fixed, with 90% bootstrap intervals.
Predicted vs listed price
Out-of-fold predictions for 700 random listings, billion toman, log scales.
What drives price
Permutation importance in the gradient-boosting benchmark.
6Data, method and limits
Every number traces to a public file. What is real, what is derived, and what it cannot tell you.