The Tel Aviv Property Report Index — Methodology

The Tel Aviv Property Report Index combines two sources: 1,037 tracked active asking listings (a July 2026 snapshot from Hebrew-language classifieds, PII-stripped) and 1,100 registered sales from the government registry (nadlan.gov.il, January 2023–May 2026). We aggregate both by neighborhood, publish a figure only where the sample reaches 20 listings, and state the sample size (n) and month.

What we track

We track asking listings — apartments actively offered for sale in Tel Aviv-Yafo — not inventory we own or represent. We never describe these as “holdings.” As of July 2026 the pipeline holds 1,037 tracked active listings across roughly 30 Tel Aviv-Yafo neighborhoods, posted by private sellers and by 329 tracked agencies. Every published statistic is computed from this dataset and the closed-transaction registry below, and carries the number of records it is drawn from.

The asking-side pipeline

The asking side is a current snapshot of live listings tracked from Hebrew-language classifieds, loaded into our own pipeline and stripped of personal information before any aggregation. It captures the price a seller is asking, the apartment’s size, and its neighborhood, from which we compute the median asking price and the median asking price per square metre (₪/sqm). Because it is a snapshot of what is on the market now (July 2026, n = 1,037), it reflects seller expectations rather than achieved prices — which is why we pair it with the closed side.

Price per square metre is computed on apartment-scale listings only, on both the asking and the closed side, and carries its own n ≥ 20 gate independent of the neighborhood’s overall sample. The reason is that a private house is typically listed with its plot area rather than its built area, so mixing houses into a ₪/sqm median silently changes what the denominator measures — and because land is cheaper per square metre than floor space, it drags the figure down and can push a genuinely expensive neighborhood to the top of a “cheapest per square metre” table. We separate the two by floor area per room, not raw size: a genuine apartment — even a large penthouse — runs about 20–45 sqm per room, while a house on a plot runs 70–110 (a 320 sqm listing on three rooms is 107), so we exclude any listing above 50 sqm per room (falling back to a 200 sqm cap when the room count is missing). Keying on density this way keeps large apartments and drops houses, where a raw size cap would do neither. Where a neighborhood has fewer than 20 apartment-scale listings we publish no ₪/sqm for it and no asking-versus-closed spread, and say so on the page. Prices per apartment — medians, percentiles, ranges — are computed over the full tracked panel and are unaffected by this rule.

The closed side (government registry)

The closed side is registered sale transactions from nadlan.gov.il, the Israeli government’s official real-estate transaction registry (served through the govmap backend). These are prices that were actually paid and legally recorded. Our current closed sample is 1,100 transactions spanning January 2023 to May 2026. Unlike the asking snapshot, the closed set is a trailing window — deals accumulate over time and are reported with a lag — so it is not a like-for-like match to the current asking month. For a plain-English overview of why Israel has no MLS and how to look up a given address’s recorded price on nadlan.gov.il yourself, see how the Tel Aviv property market works.

The join: the neighborhood dimension

The two sides meet on the neighborhood. We map every asking listing and every registered transaction to a Tel Aviv-Yafo neighborhood, then compute medians within each neighborhood. This lets us compare what sellers ask against what buyers actually paid in the same area, and it is the axis on which we publish every per-neighborhood metric.

The publication gate: n ≥ 20

We publish a per-neighborhood figure only where that neighborhood has at least 20 records on the relevant side. This threshold suppresses noisy, unreliable medians drawn from a handful of listings. As of July 2026, of the tracked neighborhoods 11 meet the asking threshold and are published, 28 are suppressed for having fewer than 20 tracked listings, and 6 non-Tel-Aviv-Yafo neighborhoods are excluded from the city index entirely.

The asking-vs-closed spread

The spread measures how far current asking prices sit above or below recently recorded sale prices in the same neighborhood. In plain language:

Spread (%) = (median asking ₪/sqm − median closed ₪/sqm) ÷ median closed ₪/sqm × 100.

A positive spread means sellers are asking more per square metre than the registry shows was recently paid; a negative spread means asking prices sit below recent recorded sales. We show a spread only where both the asking and closed samples reach n ≥ 20, which currently resolves for the following neighborhoods:

Spread rows as of July 2026 (asking snapshot) against registered transactions through May 2026.

That table is a neighborhood statistic on two different samples, not a licensed appraisal (shamaut) of one apartment. A mortgage bank will not lend against this spread.

Days on market (publication to detected removal)

Our days-on-market metric counts the calendar days from a listing’s first publication date to the date the tracker detected the listing had been removed. It is a proxy — a removal can be a sale, a relisting, a switch to rental, or a withdrawal, and detection runs on a cycle so lag can add days — so we never present it as verified time-to-sell; it is the closest observable proxy. As of July 2026, 94 of the 159 removed listings in the tracker (59.1%) carried a first-publication date; those 94 form the sample (citywide median 33 days, 25th–75th percentile 16–42 days). Listings whose dates imply a negative interval are excluded as data errors (0 this month), undated removed listings could differ systematically from dated ones, and per-neighborhood medians are published only at n ≥ 20 — the same gate as every other figure on this site.

Gross rental yield (two panels, joined per square metre)

Our rental-yield metric is gross: (median asking rent ₪ per sqm per month × 12) ÷ (median asking sale price ₪ per sqm) × 100. It excludes arnona, va’ad bayit, management, home insurance, vacancy, and rental income tax, so a landlord’s net return is materially lower. The rent side is a separate panel of tracked rental listings (1,529 citywide, July 2026) — different apartments from the sale panel, which is why we join the two per square metre rather than dividing a median rent by a median price: a per-metre basis removes any size-mix difference between the panels. Both sides are asking snapshots, not signed leases or registered sales.

Per-listing rent per sqm outside a sane band is dropped as a data error or non-residential noise, and a neighborhood is published only where both sides reach n ≥ 20 — 9 of the 11 sale-side neighborhoods clear that gate as of July 2026, and the rest are suppressed, never zeroed. One boundary caveat we state rather than hide: for the large neighborhoods the listing source splits into sub-hoods (City Center, Old North, American-German Colony / Sarona, New North), the rent panel is the union of those sub-hoods while the sale panel follows our own neighborhood labels — same neighborhood name, not a guaranteed-identical polygon. Citywide gross yield as of July 2026: 2.77%.

Feature premiums (an association, published with its confound)

Our feature-premium page compares the median asking ₪/sqm of tracked listings whose amenity flags report a feature against those that do not. Its basis is narrower than the asking aggregates above: 1,028 active listings, July 2026, drawn from the tracked panel only — the same-day classifieds snapshot rows that widen the asking basis elsewhere carry no amenity flags, and including them would silently push every unflagged listing into the “without” arm. Both arms of every published figure must reach n ≥ 20; below that the figure is omitted, never shown as zero.

The honest problem with any such comparison is that the features travel with building age — in our panel 75.7% of listings reporting a safe room also report an elevator and parking, against 27.9% of those without one. So a raw gap is largely a new-building gap. Rather than caveat that in prose we publish two corrections next to every raw figure: a neighborhood-adjusted gap, where each listing is expressed as a ratio to its own neighborhood’s median ₪/sqm before the medians are compared (removing the location component), and a matched-group gap, recomputing each of the elevator / parking / safe-room gaps inside strata that hold the other two constant. Where a gap shrinks under those controls we say so on the page — parking’s does. These are differences in asking prices across different apartments, never a valuation of a feature on any individual apartment.

Update cadence

The asking-side data refreshes weekly: the tracked-listings snapshot is re-pulled and re-aggregated, and published figures move with it. The closed-transaction set is backfilled from the registry as new deals are recorded and reported. When a page’s figures change we bump its machine-visible last-updated date.

External context (sourced, not our data)

For interpretation only, we reference two official indices alongside our own figures: the Central Bureau of Statistics dwelling-price index stood at 593.6 as of May 2026, and the Bank of Israel reported an average mortgage rate of 6.69% as of February 2026. These are national published series, cited to frame our Tel Aviv figures — they are not part of the Index and are not computed by us.

Limitations

We state our gaps plainly rather than paper over them:

• Snapshot vs trailing. The asking side is a current snapshot (July 2026); the closed side is a trailing accumulation (January 2023–May 2026). A spread is therefore an indicative comparison, not a paired, like-for-like measurement of the same apartments.
• per-row median_days_on_market is still null (the listings 'item_time' field is empty); city + per-hood days-on-market now ships separately in days_on_market.json, computed from the Airtable publishTime export (59% dated coverage of removed listings -- see its methodology_note).
• asking-vs-closed spread currently resolves for 11 neighborhoods: it needs both asking n>=20 and closed n>=20. Neighborhoods where either side is below the gate are suppressed; a wider nadlan backfill or deeper classifieds coverage on those hoods unlocks more spread rows.

Corrections

We correct errors quickly and transparently. If you believe any figure, sample size, or neighborhood mapping is wrong, contact us from the homepage or via the About page — corrections to any factual data point are welcome.

Frequently asked questions