Rental market analysis and rental index (module rents_market)
Script: scripts/modules/rents_market.py (idempotent; run from the project root with the project virtualenv).
Source: Dubai Land Department (DLD) rent contracts (Ejari) open data published on data.dubai under the Dubai Open
Data Licence, deduplicated in data/staging_rents.duckdb (see 00-staging.md). Nothing here comes from a commercial
data provider. Data through 2026-11-05; the latest complete month is October 2026.
1. Which contracts are counted
A contract line is in the residential rent universe when all of these hold:
| Rule | Why |
|---|---|
usage = 'Residential' |
residential leases only |
property_type in Flat, Studio, Villa, Complex Villas |
homes; excludes labour camps, staff accommodation, rooms, offices, shops etc. |
no_of_prop = 1 |
one unit per contract. Bulk leases of many units in one contract record the total rent and would inflate averages many times over |
q_rent_ok, q_area_ok, q_date_ok |
staging quality flags (valid dates, rent >= AED 1,000, plausible area) |
| contract start month 2010-01 .. 2026-10 | the handful of contracts dated before 2010 are dropped; months after October 2026 are incomplete (Ejari already holds contracts starting in the future) and are never shown |
- One row per contract. Ejari has one line per property of a contract (
line_number); in this universe (no_of_prop = 1) 4,549 of 5.67 million lines (0.08%, 2010-01..2026-10) are a second line that repeats line 1 field for field (rent, area, type, location, dates), so only the lowest line of eachcontract_idis kept and everycontractscount is a count of distinct contracts (October 2026 excluding short-term: 52,124, not 52,163). - Month = month in which the contract starts (Ejari start date), not the registration date.
- Annual rent = the annualised rent recorded by Ejari (
annual_amount). For contracts shorter than a year this is the annualised figure, not the amount actually paid. - Segment: Flat and Studio = Apartment; Villa and Complex Villas = Villa. DLD does not separate townhouses from villas, so "Villa" always includes townhouses.
- Bedrooms are parsed from the Ejari property sub type ("1bed room+Hall" = 1, "Studio" = Studio). Penthouse, room or missing sub types are "Unknown".
- Contract type: New = a new lease; Renew = renewal of an existing lease (usually a sitting tenant, whose rent tends to be lower than the asking rent for a new lease).
Short-term contracts
A contract is short-term when its end date is earlier than its start date plus 6 calendar months minus 7 days, i.e. it runs for less than 6 months. The 7-day tolerance keeps ordinary 6-month contracts (for example 1 September to 28 February, 180-183 days) out of the short-term group. Short-term contracts are about 1-2% of the universe; their annualised rent is higher than average. They are mostly short renewals (extensions of an existing lease to align dates or bridge a move), not new furnished or temporary lets: among 2026 residential single-unit short-term contracts (January to October) the New share is 6.7% for 0-29 days, 12.1% for 30-89 days and 17.8% for 90-169 days.
- City tables (
city_monthly,latest_by_bedrooms) are published twice:scope = all_contractsandscope = excl_short_term. short_term_monthlygives the monthly short-term count and share and averages/medians with and without them.- Headline figures, community tables and the rental index exclude short-term contracts. One Dubai-wide index
variant (
segment = "Apartment (incl. short-term)") keeps them for comparison.
2. Published tables (data/out/rents_market/, each as .parquet and .csv)
| File | Grain | Content |
|---|---|---|
city_monthly |
month x segment (All/Apartment/Villa) x contract_type (All/New/Renew) x scope | contracts, total / average / median annual rent, median AED per sq ft (apartments only), short-term contracts, 2010-01..2026-10 |
latest_by_bedrooms |
2026-10 x segment x bedrooms x scope | contracts, total, average and median annual rent, New share, median AED/sq ft for apartments, villa AED/sq ft estimate |
community_monthly |
month x community x segment x bedroom group | contracts, median annual rent, median AED/sq ft (apartments), New share (%) |
area_monthly |
month x DLD area x segment x bedroom group | same as above at DLD area level |
rental_index_dubai |
month x segment x measure x variant | Dubai-wide rental index (see section 3) with MoM, QoQ, YoY changes and 2013-01 = 100 |
rental_index_community |
month x community x segment x variant | community rental index with MoM, QoQ, YoY changes |
rental_community_month |
2026-10 x community x segment x variant | contracts and median AED per sq ft of the latest complete month alone (section 3, "One month") |
short_term_monthly |
month | short-term count, share and the effect on average / median rent |
tenant_type_shares |
month x segment | tenant type shares (section 5) |
community_map |
DLD area x master project x segment (x project where a project-name rule applies) | the community each DLD location is assigned to; the project name is listed only for rows where it decided the community (rule = project_rule), otherwise it is blank and the row covers all projects in that area / master project |
villa_mix_weights |
bedroom group | fixed weights of the villa bedroom-mix index |
The site file site/data/rents_market.json holds a compact subset: city monthly series, the latest-month bedroom
table, all Dubai index series from 2013, the latest community index (with the figures of the latest complete month
alone for the site's 1-month period, section 3 "One month"), community trends from 2017 (headline mix-adjusted variant,
communities with at least 20 contracts in the latest window) and the latest community rent by bedrooms.
Community and area tables cover single months (not rolling windows) and exclude short-term contracts. Bedroom groups there are Studio, 1, 2, 3, 4, 5+ and Unknown, always within a segment (an apartment and a villa with the same bedroom count are not mixed). "Median AED/sq ft" is published for apartments only (see villa caveat below).
3. Rental index
Measure. For each window-end month m the index is the median annual rent per sq ft (annual rent / Ejari area in sq ft) of all contracts in the universe that start in months m-2, m-1, m (a pooled rolling 3-month window), excluding short-term contracts. Rents per sq ft outside AED 10-1,000 are treated as recording errors and left out of per-sq-ft statistics. The series starts with the window ending January 2013 (Ejari coverage before 2013 is thin and patchy).
Changes. MoM, QoQ and YoY compare the window ending m with the windows ending 1, 3 and 12 months earlier. A change is only shown when both windows have at least 5 contracts. Note that consecutive windows overlap by two months, so MoM changes are smoothed.
Variants.
New- new leases only;Renew- renewals only.Combined- all contracts pooled. Because New contracts carry higher rents per sq ft than renewals and the New share swings strongly with the season (in 2026 roughly 32% of contracts in spring and about 46-52% from August to October), the pooled Combined median moves partly because of the mix. Its month-on-month and quarter-on-quarter changes move with the seasonal mix, and its year-on-year change is distorted whenever the New/Renew mix differs from a year earlier, including when renewals are registered late. Example: in October 2026 New contracts were 51.8% of the month (27,014 New, 25,110 Renew, short-term contracts excluded) against 42.6% in October 2025 (24,673 / 33,191); the Dubai Combined index (All residential) rose 4.59% year on year, the mix-adjusted index 1.00%, New -4.27% and Renew +5.10%. The site marks the latest three months as provisional (meta.provisional_frominsite/data/rents_market.json), because late renewal registrations still change them.Combined (mix-adjusted)- the headline variant (meta.headline_variant; owner decision of 7 October 2026): w x New median + (1 - w) x Renew median, where w is the New share of contracts over the 12 complete months ending three months before m (the four non-overlapping 3-month windows ending m-3, m-6, m-9, m-12). This removes the seasonal mix swing, and because the weights leave out the latest three months, renewals that are still being registered late do not move the index through the mix. (Until 7 October 2026 the weights used the windows ending m, m-3, m-6, m-9; for September 2026 that gave +2.19% year on year instead of +1.70%.) The New and Renew medians of the latest months can still change as late registrations arrive. Published only when both New and Renew have at least 5 contracts in the window. The site's hero tile, overview row, Dubai chart and community table default to it; community trend lines use it too.
Segments.
Apartment- the headline per-sq-ft index; Ejari areas for flats are close to unit sizes.Villa(measuremedian_aed_sqft) - experimental estimate. For many villas and townhouses the Ejari area is the plot, not the built-up area. Only villas whose recorded area is 1,000-8,000 sq ft are used. This drops large plots but cannot remove plot areas that fall inside that range, so villa AED/sq ft levels are not comparable with built-up-area figures; treat changes over time as more reliable than levels.Villa(measuremix_adjusted_median_annual_rent_aed, Dubai level only) - the primary villa measure: the median annual rent of villas in each bedroom group (2, 3, 4, 5, 6+) in the rolling 3-month window, combined with fixed weights equal to each group's share of villa contracts January 2023 - October 2026 (villa_mix_weights). Shown only when every bedroom group has at least 5 contracts. It uses no area at all.All residential- apartments plus the area-filtered villas, pooled.
Levels. index_2013_01_100 (Dubai rows) rebases each series to its own January 2013 window = 100.
Communities. The same calculation is run per community for Apartment, Villa (area-filtered estimate) and All
residential. Windows with fewer than 5 contracts are not listed. Windows with fewer than 20 contracts are flagged
low_sample = true: their medians can move by several percent from one month to the next on composition alone.
One month. The site's community table can also show the latest complete month alone instead of the 3-month window
(rental_community_month; in the site file the fields contracts_1m, median_aed_sqft_1m, low_sample_1m and
median_na_1m of community_index_latest, with the two periods in meta.community_windows). It counts the same
contracts and uses the same rent per sq ft as the index, but only those that start in that month: contracts with a
usable rent per sq ft, so for villas only those with a recorded area of 1,000-8,000 sq ft. The count can therefore be
lower than in the bedroom table of the site's detail panel, which counts all contracts of the month. New, Renew
and Combined are plain medians of the month; Combined (mix-adjusted) is w x New median + (1 - w) x Renew median of
the month with the same weight w as the index of that month. Its median needs at least 5 New and 5 Renew contracts in
the month and a weight (the weight exists whenever the community had New or Renew contracts in the weight window);
its count is New + Renew, the same as the month's Combined count. When the month has 5 or more contracts but the
median cannot be formed, the row carries the count, no median and median_na_1m = true (the site shows the count and
"–"); a mix-adjusted month whose 3-month cell is not published has no row on the site and is left out of
rental_community_month. Cells with fewer than 5 contracts are not published (the site shows "<5"), cells with fewer
than 20 are flagged as small samples. A single month is noisier than the 3-month window and, being the latest month,
it is provisional: renewals are often registered late, which can tilt the month towards new leases. The changes MoM,
QoQ and YoY shown next to it always refer to the 3-month index.
4. Community names
Communities use the same mapping as the sales price index (scripts/modules/price_index.py, function
resolve_community): project-name rules first, then DLD master project aliases, then DLD area aliases, then the DLD
area name as-is. Ejari records have no building name, so building-name rules cannot fire. community_map lists, for
every DLD area / master project combination in the rent universe (split by project only where a project-name rule
applies), the community it is assigned to and the
rule used. About half of Ejari residential contracts have no master project; those fall back to the DLD area (for
example "Al Nahda Second" or "Al Karama"), so some rental communities are DLD areas that do not appear in the sales
index. area_monthly gives the same statistics by raw DLD area for users who prefer the official geography.
5. Tenant type
Ejari records the tenant as Person or Authority; we label Authority as company or organisation (it covers
companies and public bodies leasing homes, typically for staff). Tenant type is missing for a large and changing part
of the data (above 80% of contracts in 2012-2014, about 11% in 2026), so tenant_type_shares shows only:
tenant_type_recorded_pct- share of contracts where the tenant type is recorded;person_share_pct,company_share_pct- shares among contracts where it is recorded.
No counts or rents by tenant type are published.
6. Confidentiality and suppression
- Every published statistic is at city, community, DLD area or project-location level; no row describes a single unit or contract, and no tenant, landlord or company is named.
- Any cell built from fewer than 5 contracts is suppressed: its count is null with
suppressed = trueand all its statistics are null. In the community index, windows below 5 contracts are omitted entirely. - The same threshold applies inside a cell: a median AED/sq ft is published only when at least 5 apartment (or, for the
villa estimate, 5 area-filtered villa) rates are behind it; short-term contract counts below 5 are blank; in
short_term_monthlystatistics are blank for months with fewer than 5 contracts (or fewer than 5 non-short-term contracts for the "excl. short-term" columns); tenant-type shares are blank when fewer than 5 contracts are behind them;community_mapcontract counts below 5 are blank. - Medians, averages and indices are estimates of market rents from registered contracts.
7. Caveats
- Not quality-adjusted. Medians reflect which homes were let; a month with more large or more central units moves the median.
- Recent months may be revised. Ejari data are published as registered; contracts (renewals in particular) can be registered after their start date, so the most recent months may be revised upward and the New share of the most recent months may be overstated.
- Villa areas. See section 3. Do not compare villa AED/sq ft with built-up-area rates from other sources.
- Bedroom counts come from the Ejari sub type text and can differ from the unit register.
- Townhouses cannot be separated from villas in DLD data.
- Contract dates. Month = contract start. A contract registered today for a lease starting next month is counted next month; months after October 2026 are excluded as incomplete.
- Totals. Total annual rent is the sum of annualised rents of contracts starting in the month; it is a measure of contracted rent value, not cash collected.