Methodology
We track active for-sale and rental apartment listings across Lima's investor districts, refreshed daily, and retain the full history. Here's exactly how the numbers on this site are produced.
Long-term net yield
Gross yield is annual asking rent divided by asking price. Net yield takes off the real running costs of holding a Lima rental — mantenimiento, predial, arbitrios, a vacancy allowance and rental income tax — and loads one-time purchase costs onto the price base. The Airbnb net yield is computed on the same footing, so the two are directly comparable. Every line, and how well we know each one, is set out below. District figures are medians, so a handful of odd listings can't distort them.
Sale-listing coverage
For-sale figures come from Urbania, Lima's largest portal. On 23 July 2026 we deepened the sale scrape to a near-census of each investor district — from a top-of-sort sample of a few hundred listings to several thousand. So the sale medians and their history step on that date for a methodological reason, not a market move: earlier points are a smaller sample, later points a fuller one. We also count each physical unit once — the same apartment relisted by several agents is de-duplicated before the median — though at this scale that changes the numbers by under a fifth of a percent. As always, figures are district medians, so a few unusual listings can't distort them.
Airbnb (short-term) data & gross yield
Short-term-rental figures come from our own census of Lima Airbnb listings — 12,410 active listings as of 31 July 2026, swept weekly and retained as history. Each listing is assigned to a district by its coordinates. District figures are medians, so a handful of unusual listings can't distort them. Note that the median listing performs well below the city-wide average: Lima Airbnb is a power-law market where a minority of professional operators earn most of the money.
How we measure occupancy — and why most published figures are too low
Occupancy is the share of nights a listing is unavailable. The obvious way to measure it is to read a listing's calendar and count. That method understates occupancy badly, and we used to make the same mistake. A calendar shows the future, and a night six months out is nearly always free — not because it won't sell, but because nobody has booked it yet. Averaging across the whole calendar therefore measures how far ahead people book, not how full the flats are.
Measured across our own snapshots on 31 July 2026 (232,200 night-observations), nights 0–6 days ahead are 49.3% unavailable, while nights 60–119 days ahead are only 13.2% — the same listings, the same market, a 3.7-fold difference driven purely by how far out you look. So we measure occupancy on near-term nights only. That puts Lima at roughly 49% (measured 31 July 2026 across 18,725 near-term night-observations), against the ~20% this site published previously. We took that earlier figure from our data provider; computing it the blended way over our own calendars reproduces about the same number, which is why we think the two are the same measurement rather than a disagreement about the market. Treat it as an upper bound: a night can be unavailable because a guest booked it or because the host blocked it, and a single snapshot cannot tell the two apart. We are building a panel of listings watched night after night, which can.
We report the district gross yield — median annual Airbnb revenue over a representative local property value, before costs. We also compute an Airbnb net yield, charged the short-let operating stack and the same ownership costs the long-term net already carries; the table below shows exactly which lines those are. Both net yields load one-time purchase costs onto the price base, so the two are computed on identical footing. The 5% income-tax figure matches the rate we apply to long-term rent and is deliberately conservative: Peru generally treats short-term letting as business income, taxed higher, so the true Airbnb net is more likely below our figure than above it. We decide the winner on a net-vs-net footing. The "×" multiple shown in the index is deliberately re-derivable from the two columns on the page — long-term net divided by Airbnb gross — so it reads conservatively for long-term. On both footings, long-term wins in every district today. A district's Airbnb yield is only scored when it has at least 5 tracked listings; below that we mark it "too thin to call" rather than show a rate.
Coverage & caveats. Two things bound what our Airbnb figures can mean. First, we only ever see the asking price of listings that are still free — a search cannot return a listing already booked for the nights we ask about, so fully booked inventory is invisible to the price sample. Second, our prices are asking prices, not what guests actually paid; Airbnb does not publish transactions, so every provider in this market infers rather than observes. The long-term side is asking rents for the same reason, so the bias runs in the same direction on both sides of the comparison and largely cancels. Airbnb figures are refreshed weekly and appended, so history compounds as we track it.
What comes out of each yield
A yield is a subtraction, so the honest question about any of them is what was taken off, and how well each of those figures is actually known. Here is the entire stack, both strategies side by side. Some lines are measured from our own data or carry a statute behind them; others are numbers we chose because no source exists. Every row says which it is.
| Charged against the yield | Long-term | Airbnb |
|---|---|---|
| Mantenimiento (HOA) Measured The district median, taken from the listings that publish the figure on the card. Charged to Airbnb only. Urbania quotes rent and mantenimiento as two separate fields, and listings advertise "con mantenimiento incluido" as a selling point — nobody advertises the default. So a long-term tenant pays it directly, and it was never inside the advertised rent we use as the landlord's revenue; charging it there was wrong twice over. An Airbnb host pays it every month and no guest reimburses them. This is the one genuine asymmetry in the table, and it is worth 0.5–1.4 points of long-term yield. | tenant pays | $59–208/mo |
| Predial (property tax) Assumed A flat rate standing in for a progressive schedule of roughly 0.2–1.0% that we have not yet modelled. The tax is real and unavoidable; the rate is our approximation of it. | 0.3% of value | 0.3% of value |
| Arbitrios (municipal services) Measured Median of 22 Lima listings that disclose it, corroborated independently at S/50 a month. Every municipal portal blocks automated access, so this is one figure standing for eight districts — which is wrong in principle, since arbitrios are set district by district by ordinance. | $176 / yr | $176 / yr |
| Rental income tax Verified Ley del Impuesto a la Renta, Art. 54(e): "Otras rentas provenientes del capital: 5%". Rent is renta de capital, and the rate is the same for a non-domiciled owner. Summaries of Peruvian tax quote 30% on gross income for non-residents; the statute is what we followed. | 5% of rent | 5% of revenue |
| Vacancy Assumed The long-term figure is chosen, not measured. Airbnb's equivalent is not a line at all: empty nights are already inside the occupancy we measure, at roughly half of near-term nights across 18,725 night-observations. So the softest number in this table sits on the side that wins the comparison. | 3% of rent | in occupancy |
| Platform fee Verified Airbnb's published host service fee (help article 1857). | — | 3% of revenue |
| Utilities & internet Measured A fixed bill per flat, scaled by size — not a share of revenue. Anchored to Numbeo Lima (7 Jul 2026, 958 entries from 74 contributors): S/228.42 a month of basic utilities for an 85 m² flat, scaled by floor area, plus S/84.90 of broadband, which does not scale. Charged as 8% of revenue until 1 Aug 2026, which billed $430 in Pueblo Libre and $969 in San Isidro for the same water meter and the same router — right on average, wrong in shape, and wrong in the direction that flattered Airbnb where its returns are weakest. Floor area comes from the bedroom count, since Airbnb listings do not publish m². A long-term tenant pays their own. | tenant pays | $573–1304/yr |
| Furnishing / fit-out Not charged Left out from 1 Aug 2026. Furnishing is a one-time capital outlay, not an annual operating cost, and the $1,200/yr we had been charging was the one line in this table with no source at all behind it. Saying so plainly: it is not free. A furnished flat costs more to buy into than an empty one, so its honest home is the denominator, beside purchase costs — and it is in neither today. That omission flatters Airbnb. | — | excluded |
| Cleaning & consumables Not charged Airbnb pays the guest's cleaning fee to the host (help article 459), and we never see that fee, so it is absent from the revenue side too. Charging a cost against revenue that excludes the matching income would run one error twice. Consumables are genuinely uncovered: no Lima figure exists in any source we reached, and zero is the honest placeholder. | — | none |
| Management Not charged Both columns are self-managed, so there is no management fee on either side; anyone paying for one should subtract their own quote from what we publish. | none | none |
| Purchase costs Assumed Notary, registry and alcabala, loaded onto the price base rather than the income, identically on both sides. A round figure, not a quote. | 5% of price | 5% of price |
Measured — computed from our own data. Verified — a statute or a published rate, read at source. Assumed — a number we chose, with nothing published behind it. Not charged — deliberately zero, for the reason given in the row.
Three lines are missing from both sides
Nothing above covers repairs, insurance, or a reserve for the eventual kitchen, bathroom and appliances. "Buy it and collect rent" is not a real cost model, and we would rather print that sentence than let a flat appear to maintain itself. Those are asset costs — the roof does not know how the flat is let — so they fall on both strategies alike and the gap between the two survives them intact. The levels do not. At the common rule of thumb of 1% of value a year for upkeep, insurance and capex combined, long-term drops by about a point and Airbnb turns negative in the weakest districts. Read every net yield on this site as an upper bound on that account.
Mantenimiento may not be the landlord's to pay
Lima portals quote rent and mantenimiento as two separate lines, and some listings advertise "mantenimiento incluido" as a selling point — which nobody would bother advertising if it were the norm. If the tenant customarily pays it, charging it to the long-term landlord is wrong twice: not their cost, and not inside the asking rent that stands in for their revenue. Dropping it would add roughly 0.5 to 1.4 points to long-term net yield. We have not dropped it, because it would widen the lead of the side already winning, and that is a reason for more scrutiny rather than less. It is stated here instead of quietly banked.
Neither figure pays anyone to run it
Neither yield subtracts a management fee. We had been charging Airbnb ~20% of revenue against long-term's ~8% of rent — a real difference, since an Airbnb turns over perhaps fifty times more often, but a difference whose size we had invented on both sides. Charging one unsourced number against another is not a comparison. So both columns show what an owner earns running the place themselves, and anyone paying for management should subtract their own quote. That figure is larger for Airbnb, because the work is larger — which means the published numbers flatter Airbnb more than they flatter long-term.
The middle listing is not the typical outcome
Long-term returns cluster within about half a percentage point across Lima. Airbnb returns do not: the gap between a median listing and a top-quarter one is several points, in the same district, in the same month. Each district page shows both, because publishing only the median would hide the most decision-relevant thing we measure. The top-quarter figure is the 75th percentile of per-listing annual revenue: for every listing that carries both a quoted nightly rate and a week of its own calendar, we multiply that listing's rate by that listing's own occupancy, and read the quartile off the resulting distribution. Rate and nights filled therefore count together, as they do in reality — a listing that out-earns its neighbours usually does both, and the two multiply. Until 1 August 2026 this varied price alone and held occupancy at the district average, which described a listing priced like the better quarter rather than one run like them, and understated the spread. Two limits remain, and neither is small: it is a cross-section of different apartments, so it says what the better quarter earns and not that any particular flat can be operated into it; and a forward calendar cannot distinguish a night a guest booked from one the owner blocked, so a dormant listing would read as fully occupied. We tested that directly. The 1,394 that show a fully booked week are 28.1% occupied 60–119 days out, against 7.4% for everyone else — a dormant flat is unavailable at every horizon, and these have plenty of free nights four months ahead and nearly four times the forward bookings. They are not blocked; they are the strongest listings, selling out first. Dropping them, which we considered, would have deleted the best performers and pushed the top quarter down by a third. Some genuine blocking certainly remains, so the figure is still an upper bound — but a much narrower one than that caveat usually implies.
The per-size table shows its working
Both columns of that table are net, on the same cost model as the district figures above, so the winner it names is decided the same way as the verdict at the top of the page. It compared gross to gross until 1 August 2026, which is not like-for-like: Airbnb carries a platform fee, a utilities bill and mantenimiento that a long-term let does not, so gross-vs-gross awarded Airbnb rows the same page said it lost.
Every cell carries the sample behind it — sale and rent comps on the long-term side, listings priced and listings with a measured calendar week on the Airbnb side — because segmenting eight districts by four sizes turns one large sample into thirty-two small ones. We show a small number and mark it rather than hiding it: a marked figure can be discounted by the reader and improves as coverage grows, while a blank tells them nothing and never improves. Below 10 apartments of that size with calendar data we publish nothing, because one blocked week among nine flats moves the answer more than the market does — several districts have fewer than five four-bedroom listings in total. Where a cell reads too thin, we looked and there is not enough market to measure; that is a different statement from missing data, and we do not substitute another provider's smaller sample to fill the gap.
Comparables & the plausibility gate
Comparable districts are the nearest by price band, so you can jump between similar markets. Listings whose implied gross yield falls outside a plausible range — above 15%, or on a sale price below $30,000 — are treated as data artifacts and excluded before the median. A district with fewer than 5 tracked sale listings is omitted entirely; per-bedroom figures need at least 3 sale and 3 rent comps (long-term) or 5 listings (Airbnb) to show, and a seasonality month needs 20. So a thin sample never headlines a number.
What's free, and what's not advice
Everything here is free, aggregate market data. The free weekly newsletter adds our read on specific listings. Nothing on this site is personalized investment advice — always verify a specific property (title in SUNARP, the real mantenimiento, the building's short-term-rental rules) before you act.