Índice INGAR methodology — v1.4

Universe and data sources

The index considers sale offers published openly in the Montevideo market, obtained through our own systematic market survey, for apartments (AP), houses (CA) and land (TE), that were active at some point during the closed calendar month (listed before month end and delisted after month start, or still active). Neighborhoods follow a normalized list of 71 neighborhoods within the department of Montevideo: that is the surveyed universe. Of that universe, only neighborhoods that pass each month's quality gates are published (currently 62 neighborhoods with at least one publishable month); the rest are surveyed but not published. Bordering localities of other departments that offers often label as Montevideo (e.g. parts of Ciudad de la Costa, Canelones) are excluded from the universe.

Estimated capital appreciation additionally uses official sources: the INE IAI sale-price index for Montevideo and rental contracts from the DGI (tax authority).

Each property is counted once

The same property is often publicly offered more than once. Offers are consolidated so each property counts exactly once, and all statistics are computed over unique properties. Supply volume is published as a relative level (oferta: alta / media / baja — high / medium / low), defined by terciles within each month and property type, instead of absolute counts.

Per-offer filters

Before computing statistics, each offer must pass plausibility filters: price between USD 20,000 and 5,000,000, surface between 15 and 2,000 m², and price per m² between USD 300 and 8,000. Medians and percentiles use continuous interpolation (percentile_cont).

Asking prices, not transaction prices

All prices in the index are asking prices: what the seller requests in the offer, not the final sale price. The last known price of each offer is used. Declared limitation: price changes within the same offer are only partially captured, because re-publication of an offer resets its history.

The only exception is estimated capital appreciation, which incorporates transaction data (INE + DGI) and is therefore labeled and published as a separate metric.

Rent, yields and market metrics

The segment's median market rent (neighborhood level, USD) is computed from collected rental offers, using the month's latest calculation. Yields are published only with confidence A or B and within a 2–15% gross range. Net yield is estimated: it applies a total deduction of 35.16% on gross rent, itemized in the next section.

Months of supply and absorption rate come from the month's snapshot of active stock and delistings, and are only published when data is available. Days on market (DOM) are not published yet: see limitations.

Net yield: deduction breakdown

The estimated net yield applies a total deduction of 35.16% on annual gross rent. Components (as % of gross rent):

• Management: 12.4% — professional property management (6% monthly fee + VAT, plus the amortized placement cost).
• Income tax: 10.06% — see detail below.
• Maintenance: 5.0% — annual maintenance assumption.
• Vacancy: 4.2% — assumption of one vacant month every 24.
• Owner taxes: 3.5% — Contribución Inmobiliaria and Primaria (property taxes).

The 6% + VAT monthly fee is a market assumption (midpoint of the surveyed market range, 5–7% + VAT), not INGAR's own rate: INGAR's property management service charges 7% + VAT. Recomputed with that fee, the total deduction rises from 35.16% to 36.38% of gross rent, and the estimated net yield is 0.05–0.12 percentage points lower depending on the segment (for example, a 5.75% gross yield nets 3.66% instead of 3.73%).

Income tax: the index assumes an individual owner opts to treat the 10.5% withholding on accrued gross rents as final (applicable to residents under IRPF, Decree 148/007, and to non-residents under IRNR, Decree 149/007, arts. 24 and 33 lit. a). Alternatively, the general assessment of 12% on net income is available, with the expressly admitted tax deductions; which option is more convenient depends on each owner's deductible expenses. The tax is applied on rent adjusted for the modeled vacancy — which is why it weighs 10.06% of gross rent rather than 10.5%. In projects with Vivienda Promovida (housing-incentive) benefits this tax can be exempt for up to 10 years; the general breakdown does not incorporate that exemption.

Maintenance, repairs and insurance are not admitted tax deductions: they enter the breakdown as the real cost of operating the property, not as a tax deduction.

Estimated capital appreciation (INE + DGI)

It anchors on the INE IAI sale-price index for Montevideo (full-series CAGR, auto-updating), multiplied by a relative neighborhood factor built with rent-tilt over DGI rental contracts (76 neighborhoods with their own factor; those 76 refer to the coverage of this appreciation metric, not to the number of neighborhoods in the price index). The scope field indicates whether the neighborhood has its own factor or inherits the city average. It is NOT the change of the index's own series: it is a different metric and is labeled as such.

Changes

The index publishes month-over-month change (MoM of the median price per m²) and cumulative change vs. March 2026, the series start. No year-over-year change is published until the series is 12 months old (February/March 2027).

Publication thresholds

We prefer declaring “insufficient data” over publishing a noisy number. A neighborhood-type segment is published in a given month only if it meets:

• Minimum sample: at least 30 unique properties offered in the month (n ≥ 30).
• Stability: month-over-month change within ±20%; larger changes exclude the segment that month, with the reason recorded.
• Yields: only with confidence A or B and a value between 2% and 15% gross.
• Final review by an AI model (see next section).
• Excluded segments are declared each month with their reason — the gap is part of the data and is never filled or interpolated.

AI review

Before publishing, each monthly run goes through an automated review with a language model (claude-opus-4-8; the model used is recorded with each publication) verifying internal consistency — ordered percentiles, ranking, and the per-m² yield identity (renta_m2_yield_usd × 12 ÷ precio_m2_yield_usd ≈ gross yield; see changelog v1.3) — and plausibility of the figures. Its exclusions are recorded with their reason. The summaries shown on this site are the reviewed texts; the website does not recompute or alter any figure.

Typical volume

Verified in the March–June 2026 backfill: about 200 segments computed per month, of which ~120 are publishable (~50 neighborhoods for apartments, ~50 for houses and ~15 for land).

Declared limitations

• Short series: the index starts in March 2026; no year-over-year comparisons until 12 months are complete (Feb/Mar 2027).
• Days on market (DOM): not published yet. The observation window starts in February 2026 and is still too short to measure them without bias (left-truncation). They are computed internally and will be published once the series accumulates at least 12 months of observation (2027).
• Relisting: a re-published offer for the same property resets its history; price changes within the same offer are only partially captured despite the consolidation.
• Asking prices: all price levels reflect seller expectations, not closings.

License and attribution

Índice INGAR data is published under a Creative Commons Attribution 4.0 International licence (CC BY 4.0): free to use, including commercially, with mandatory attribution. Required citation: “Source: Índice INGAR — ingar.com.uy/indice”. Full downloads at /indice/datos.csv and /indice/datos.json. The official series we republish (INE rents, City of Montevideo construction permits) are free to use citing the original source stated in each dataset.

Changelog

• v1.4 (August 2, 2026): the per-m² rent and price medians feeding the yield are now computed with 2 decimals (previously, integers). Reason: in 98 of ~172 publishable segments the median rent is ≤ 11 USD/m², and integer rounding introduced artificial jumps of up to ±10% relative in the yield of low-rent neighborhoods. The change applies forward only: the previous series keeps its original values and may show a micro-jump of up to ±5% relative in those segments in the first month computed under v1.4 (August 2026).
• v1.3 (August 2, 2026): additive change, no published figure is modified. (1) The dataset gains the columns renta_m2_yield_usd and precio_m2_yield_usd: the exact numerator and denominator behind each gross yield. The yield is NOT computed from the published median rent and median price (which are totals, drawn from different listing universes — rental listings and sale listings), but as median rent per m² × 12 ÷ median sale price per m² of the same segment; with the new columns the identity is reproducible with a verified maximum error of 0.01 percentage points across every row with a yield. Rent per m² was published at integer precision up to v1.3; resolved as of v1.4 (August 2, 2026), which computes it with two decimals. (2) The oferta_nivel column is added (high/medium/low by terciles within each month and typology), formalizing in the dataset the supply level the data pages already displayed. Both back-filled from the frozen source records — no recalculation.
• v1.2 (July 12, 2026): net yield is now calculated with an explicit deduction breakdown — 35.16% of gross rent: management, income tax, maintenance, vacancy and owner taxes (detailed above) — applied retroactively to the whole series. Previously (v1.1) the deduction was a flat 25% on gross and did not include income tax. Reason for the change: net yield now deducts income tax and itemizes its components. Gross yield, median rent and prices are unchanged.
• v1.1 (July 2026): first published version of the index.

Versioning

Current version: v1.4, applied in the calculation engine since August 2, 2026. Rows published through July 2026 carry metodologia_version v1.3 (their figures are unchanged); the first monthly publication under v1.4 is August 2026, published on September 1. Any methodological change is published as a new version on this page, recorded in the changelog above and in the metodologia_version field of every data row.