Real Estate Transactions in Saudi Arabia: The Deed File

The public record of real estate transactions in Saudi Arabia is a quarterly file of registered deeds, a flow of registrations rather than a picture of the market. The Ministry of Justice's twenty-four files, Q1 2020 to Q4 2025, hold 1,412,132 rows, of which 47,518 sit under a 2025 label and carry only April to June 2024 dates, leaving 1,364,614 usable deeds worth SAR 1,424.2 billion across 8,516.0 million sqm (MOJ). Ten columns say where a deed was registered, when, for how much and over what area. Only 9.6% say what was sold, and the ministry's own transaction lookup releases five records per district and no more.

Key figures at a glance

Where real estate transactions in Saudi Arabia are published

One CSV per quarter from the Ministry of Justice, under the KSA Open Data Licence, which permits derivative works provided the ministry is credited. Twenty-one files carry ten columns; three carry more. Grouped on the date inside each row rather than on the file name, the six years read like this:

Across those six years Riyadh region registers 467,343 deeds, 34.2% of the national flow, and Riyadh city 323,556 against Jeddah's 156,595 (MOJ, 2020 to 2025). Take the bulk CSVs for anything you intend to compute on. REGA's real estate indicators platform is the reachable official aggregate, sourced to "Ministry of Justice and Real Estate Registry" and quarterly from Q3 2023 (REGA): the right place to check a total, the wrong place to build one. The Saudi source map sets it beside the other series.

Count, value and area mislead in three different directions

Each of the three headline columns is honest in its own terms and wrong as a market number, in a different direction. Count is bounded by documentation, not by activity: the transaction tax regime in force from 10 April 2025 charges 5% on a disposal "whether or not it has been documented" (سواء تم التوثيق أم لم يتم), and Article 6 of its implementing regulation is headed "sham or concealed transactions" (ZATCA, regulation issued 24 March 2025). The regulator legislates for undocumented disposals and understated prices. Neither reaches the file.

Value carries a defect that survives into any series built by summing it. On 14 May 2020, nine separate Riyadh deed references each carry the identical price of SAR 12,996,688,997, across Al Aqiq and Al Nakheel, on areas from 173.10 sqm to 961,068.59 sqm, every row declaring a single property (MOJ, 2020 Q2). The ten largest deeds are 60.9% of the Q1 2020 value total and 74.9% of Q2 2020, and mean consideration per deed falls from SAR 3,180,301 in Q2 2020 to SAR 466,764 in Q3. Registration events, not the market, move that average.

Omran estimate. Rows sharing a date, a city and a price of SAR 50 million or more, counted beyond their first occurrence, carry SAR 139.9 billion across 39 rows, 9.8% of the SAR 1,424.2 billion registered between 2020 and 2025 (Omran measurement on the MOJ files). Strip that surplus and registered consideration falls to roughly SAR 1,284 billion. Treat the corrected figure as a floor on one visible pattern, not a cleaned series: bundles under the threshold escape it, and the file publishes no flag that would catch them.

Area gets quoted as building activity and is nothing of the sort. Agricultural deeds are 4.0% of the count and 81.1% of the area across 2020 to 2025, 6,907.5 million sqm against 914.7 million residential, and the 2021 spike of 3,569.4 million sqm is 3,225.9 million sqm of farmland, spread widely enough that the year's ten largest rows carry 2.0% of its area (MOJ, 2020 to 2025). A deed is not a property either: 1,364,614 rows cover 1,415,073 properties, and one row covers 326.

Nothing in the file says villa, apartment or plot

Unit type is published on 131,447 rows, 9.6% of the usable set, every one in the first three quarters of 2023. On those rows land plots are 104,947 (79.8%), apartments 21,322 (16.2%) and villas 231, or 0.18% (MOJ, Q1 to Q3 2023). The column was dropped from Q4 2023 and has not come back.

GASTAT's workaround confirms the hole. Its methodology box states that "a geospatial artificial intelligence (Geo AI) model is utilized to process various types of real estate transactions", which "improves the accuracy of real estate classification" (GASTAT, Q4 2025). The statistics office built a classifier because the record does not carry the class.

So the number most often quoted as Saudi house prices is largely a land index. In the Q2 2026 basket residential plots weigh 45.849% and commercial plots 22.836%, so the two plot lines alone are 68.685%, against 10.455% for villas (GASTAT, Q2 2026). That quarter's headline is 106.34 on a 2023 base, up 1.3% year on year, and underneath it residential plots rose 6.3% while villas fell 9.7%. Read the headline as the price of a Saudi villa and you have the sign backwards. The price index page carries the rest of the basket.

The same gap caps district work. A median from this file is a median of whatever mix registered there that quarter, over rows 90.4% untyped, so it needs n beside every cell, a suppression floor and the mix disclosed, which is how our Riyadh land price table is built.

The duplicated quarter, the reversed dates and the five-record cap

Two of the twenty-four files are broken, and no reachable page states a release calendar or a revision policy for any of them. One carries a whole quarter under the wrong year. Another writes its dates backwards. Below both sits a lookup that hands out five records per zone and stops.

MOJ-Sales-2025-Q2.csv holds 47,518 rows, every one dated between 1 April and 30 June 2024, 44,831 of them byte-identical to rows in the 2024 Q2 file (MOJ). Load the files as labelled and you count April to June 2024 twice, then report a Q2 2025 of 47,518 deeds and SAR 57.46 billion that never happened. The invented quarter sits plausibly beside its neighbours, so nothing downstream catches it.

The second defect hides in the most valuable file of the set. MOJ-Sales-2023-Q1.csv is the only quarter publishing a cadastral address, plan and parcel, on 34,656 of its 44,091 rows (78.6%) (MOJ, Q1 2023). It is also the only file writing 1/1/2023 where the other twenty-three write 2024/04/01, so a loader assuming one date order drops roughly one day in two into the wrong quarter. And it spells three of the thirteen regions its own way, منطقة مكة المكرمة where the rest write منطقة مكة المكرمه, with its own variants for the Eastern Province and Madinah, so grouping on the raw region string makes 16,867 rows vanish. Both traps live in one file. Neither gets caught.

That pushes readers to the dashboards, where REGA warns that the daily feed is "displayed as received from its source (MOJ and the Real Estate Registry), without any cleansing or processing of prices or classification of real estate unit types" (REGA, 29 August 2026), and means it. Of the 656,775,001 sqm its daily area series carried over the thirty days to 28 August 2026, 422,563,684 sqm is the single 17 August print, flagged +3,726.65% by the platform's own change field, leaving 234,211,317 sqm across the other twenty-nine days (REGA). Two thirds of a month of national traded area is one uncleansed record. That window closes on a Saudi weekend rather than a market signal: 28 August 2026 prints 95 deeds and a 91.4% one-day fall, on a Friday.

Below the aggregates the door closes. Both REGA services that would show individual deeds sit behind a Nafath sign-in requiring a national ID or residence number, so an analyst without a Saudi identity has aggregates and bulk files and nothing between them. SREM, the ministry's platform for property trading, mortgage services and deed issuance (SPA, 3 September 2024), is the surface people assume fills that gap. It does not. The service refuses our own infrastructure, so Omran measured it in August 2026 from a leased node inside the Kingdom, across 8,774 zone queries covering 284 cities and 8,490 districts. It never returned more than five distinct transactions for a zone: exactly five for 5,207 zones (59.3%), six for none. Nor are those five a recent window, since in the saturated zones they span a median of 249 days and a maximum of 6,258 days. The same service's annual city aggregates declare 4,869,841 transactions between 2006 and 2026 against 32,321 distinct records released, so the public surface hands out 0.66% of what it says it holds (Omran measurement, internal route). Every district panel of recent transactions built on it is a panel of five: a lookup for a citizen checking one neighbourhood, not a sample of a market.

What the deed record leaves out, and the estimate we publish instead

Four holes matter to professional work and the public record fills none of them. Unit type is absent on 90.4% of rows and on every row after Q3 2023. Plan and parcel were published once, for Q1 2023, so the finest public grain is a free-text district, and that cell reads "other" (أخرى) on 145,338 of the 1,412,120 rows that carry one, 10.3% (MOJ, 2020 to 2025). Q2 2025 is missing, with nothing published to say so. And nothing below the deed is recorded: no buyer, no seller, no financing flag, no marker for whether a transfer is a sale, an inheritance, a gift or a court partition, distinctions the tax law makes and the file does not. Joining it to the Real Estate Registry works, but no official crosswalk between the two geographies exists, so the join is an editorial act and has to be disclosed as one.

Omran estimate. Applying the only type mix the ministry ever published, the three 2023 quarters, to the whole usable set puts the six-year deed flow at roughly 1,089,000 land plots, 221,000 apartments and 2,500 villas (79.8%, 16.2% and 0.18% of 1,364,614 rows). The assumption is loud and we would rather state it than bury it: it rests on three quarters of 2023, the thinnest full year in the series at 183,788 deeds, and it holds that mix fixed across six years, which nothing in the public record confirms. What survives is the order of magnitude, and it is what readers get most wrong. On the only rows the ministry ever typed, villas are under a fifth of one per cent of registered deeds.

One asymmetry sits under most of the wrong numbers in circulation: the aggregate is public while the file behind it is not reachable from outside the Kingdom, since the national open-data portal and the SREM service both refuse this network and REGA's latest open-data page carries rental indicators only. Tabulations of that file circulate in the Saudi press that no outside reader can audit against the source. One test catches most of them and costs nothing: a region cannot register fewer deeds than its own capital city. On the Q2 2026 cards, covering 1 April to 30 June 2026, the Kingdom shows 53,663 sale transactions and SAR 72,293,066,106.11, and Riyadh city alone shows 13,637, 25.4% of the national count, on SAR 29,074,353,496.04 (REGA, Riyadh city). Any table putting the whole Riyadh region below its own capital has read a partial extract, whatever the byline.

Omran holds the twenty-four files parsed, dated and deduplicated: mislabelled quarter removed, bundled considerations flagged rather than summed, region strings folded to one spelling, the ministry credited as its licence requires. Request the cleaned quarterly deed series or scope a district-level brief against your own city list.

FAQ

Does the Ministry of Justice publish every property sale in Saudi Arabia? It publishes every sale that becomes a registered deed, a narrower population than every sale. The transaction tax regime in force since 10 April 2025 charges the 5% "whether or not it has been documented" and gives sham and concealed transactions their own article, which says plainly that undocumented disposals exist and never enter the file.

Where do I download the data, and how far back does it go? The quarterly CSVs sit on the national open-data portal under the Ministry of Justice, from Q1 2020 to the latest published quarter, licensed for derivative works if the ministry is credited; the set used here was read from a public mirror that republishes those CSVs unmodified. To Q4 2025 it runs to 1,412,132 rows, 47,518 of them a duplicate quarter.

Is the traded area indicator a measure of building activity? No, on two counts. In the 2020 to 2025 files agricultural land is 4.0% of deeds and 81.1% of area (MOJ). On the live feed, the 17 August 2026 print alone is 422.6 million sqm against a median day of 6.7 million sqm, tagged +3,726.65% by the platform itself (REGA).

My quarterly totals do not match the published aggregates. Which one is wrong? Probably neither, because you are almost certainly comparing two populations. SREM's own description covers property trading, mortgage and financing services, and deed issuance for splits and merges (SPA, 3 September 2024), so a headline built on that platform counts operations rather than sales; the quarterly files count sale deeds only, and REGA's indicators platform counts sale transactions, at 53,663 for the Kingdom in Q2 2026 (REGA). Then check your loader: one published file is mislabelled, so a naive load double-counts April to June 2024 and invents a Q2 2025 of 47,518 deeds.

Sources

• Ministry of Justice, quarterly real estate sales open data (KSA Open Data Licence), read on the public mirror: link
• Ministry of Justice, 2025 Q2 file (the duplicated quarter): link
• Ministry of Justice, 2020 Q2 file (nine deeds at one price, 14 May 2020): link
• REGA, Real Estate Indicators Platform (Q2 2026 Kingdom card, daily series, source label): link
• REGA, Riyadh city indicators (13,637 transactions, Q2 2026): link
• REGA, live daily transactions view (the no-cleansing disclaimer): link
• REGA, deals service (Nafath sign-in): link
• ZATCA, RETT implementing regulation, Articles 2 and 6 (Arabic PDF, issued 24 March 2025): link
• ZATCA, real estate transaction tax law page (5% rate, law effective 10 April 2025): link
• GASTAT, Real Estate Price Index Q2 2026 tables (index 106.34, basket weights): link
• GASTAT, Real Estate Price Index Q4 2025 bulletin (methodology, Geo AI classifier): link
• Saudi Press Agency, Ministry of Justice statement on the SREM platform (3 September 2024): link