lakilé.
Home

Standards

Every number on this site had to be built.

Panama's public registry is a filing system. It is organized one property at a time under a folio number, so you can pull a single file if you already know which file you want, and that is where it stops. It has no concept of a building, no comparable set, no price index, and every entry is a wall of legal prose written for notaries. Nobody can ask it what apartments in a given tower sold for last year, because it was never built to answer that. So we built what could: 76,878 property files read and re-read, 580,769 entries parsed into structured fields, matched against real buildings and corrected where they were wrong. This page is how that works, and where it runs out.

What we read, and what we make of it

Two inputs come in, and they are not the same kind of thing. What we do with them is the third piece, and it is the one that took the years.

  • Everything we publish starts at a registered deed. A compraventa names the parties, the notary, the escritura number, the date and, when it is stated, the amount. We take that from the Registro Público, which is its official source. What the registry will not do is organize it: there is no building in it, no comparable set, no index, and the same event can reach us more than once. If a deed states no price, we do not publish one.
  • The asking market is collected separately and stored apart: 64,590 listings across three portals, 35,502 of them active, with what was advertised, at what price, for how long, and which ones vanished without a sale ever being registered. An asking price is what somebody hoped for. A registered sale is what somebody paid. The two never meet in the same calculation.
  • Then our own layer, which is the product. Parsing legal prose into structured fields, resolving duplicates down to a single event, matching units to the building they actually sit in, and deciding which sales can carry weight. We audit that layer against itself on a schedule, and it is how we have caught and repaired our own extraction errors. How each of those steps works is the part we keep.

What the three counters actually count

The numbers beside the link that brought you here are narrower than they look, and the narrowing is deliberate. Here is what sits inside each one.

  • Buildings tracked counts residential buildings whose files we have matched and verified in the registry, not every building standing in the city and not every building somebody has advertised. Coverage inside that set varies, which is why only a subset carries enough history for us to sell a report on it.
  • Sales counts registered compraventas, once for each time one happened. An apartment sold four times contributes four sales, because four sales is what happened. It spans the whole history we have read rather than a recent slice: the oldest deed we hold is from 1973.
  • Sales volume sums the prices written on those deeds, inside a sanity band that keeps typing errors and isolated outliers from distorting a headline total. Deeds recording no amount contribute nothing, and we never estimate a missing price to make the number bigger.
  • None of the three counts a listing, and none counts an asking price. An apartment advertised for a year at a figure nobody paid adds nothing to any of them.

What is a fact and what is our estimate

Some numbers are written on a document. Others are ours, computed from the ones that are. You should always be able to tell which of the two you are reading.

  • Verified: sale prices, dates, parties, mortgages, liens and seizures (a debt or claim registered against an apartment), and the chain of owners. Each one traces to a deed we have read. Present ownership is derived from the complete chain of transfers rather than from any single summary field, because a summary can lag the chain behind it.
  • Estimated: fair value, negotiating range, rental yield, anything projected forward. These come from verified sales run through our own models. They are informed estimates, not a certified avalúo, and they are labeled as estimates wherever they appear.
  • Traced: every paid report shows its own working. How many comparable sales it used, over what period, and what it left out. You see the set the model saw, which means you can disagree with it on the evidence.
  • Reviewed: more than 40 automated quality checks run across five stages before anything reaches a customer, and a person reads every paid report before it goes out. One that fails that read is not sent.

How the comparable set is built

A median price per m² is worth exactly what the sales behind it are worth, so most of this work is deciding what to leave out. Those exclusions are computed from the record rather than asserted.

  • Sales between parties who were never really negotiating do not belong in a median. A company selling to itself, a transfer inside a family, one co-owner buying out another: the amount on those deeds is a formality rather than a price the market agreed on. We identify them by rule and drop them before anything is computed, and the rules are ours.
  • Forced auctions are handled on their own. We hold 1,545 remates, where a lender forecloses and the apartment is sold at judicial auction. That price reflects the pressure of the process rather than what the building is worth, so it never enters a market median. Where it matters we show it in its own section, which is where it says something useful.
  • Size is matched before anything is compared. Price per m² falls as apartments get larger, so a studio and a penthouse on two levels in the same tower are not comparables. The set is held inside a band around that building's typical unit size.
  • Medians, never averages. One outsized sale drags an average somewhere the market never went, and leaves it describing nobody.
  • When too few qualifying sales remain, the report says so in the place where the number would have been, and prints the comparable count beside any range it does show.

How current this is

None of this refreshes itself. The files are re-read on a cycle, and whatever is still moving through the registry gets named as such.

  • Every file we track is re-read on a rolling monthly cycle rather than pulled once and left. Of the 76,878 files we hold, 76,207 are fully verified and 75,976 were re-read within the last 30 days.
  • Registration runs behind reality. Weeks pass between a deed being signed and the registry showing it, sometimes months, which makes the most recent period always the thinnest part of the record. A report leaning on it says so rather than presenting it as settled.
  • Dates in the source are not written one way. The same six digits can mean 6 March or 3 June, and guessing manufactures sales in months where nothing sold. We resolve each date against the document it came from rather than assuming one house format, and nothing is ever dated later than the day we read it.
  • No counters ticking upward on screen to suggest motion. The figures here and on the home page come from the database as the last cycle left it. If something has not moved in a while, we would rather you noticed.

Where the record is thin

Panama's record is uneven. Some files run complete back to the 1970s, others skip a decade, and plenty of deeds register a sale without registering a price. We would rather name that than paper over it.

  • 41,628 of the 130,657 recorded sales we hold carry no amount at all, roughly a third. The transfer happened and the deed exists, but the price is not in the document. Those sales count as sales and contribute nothing to any price median or volume total.
  • Older entries were scanned from paper, and paper degrades. Ink runs, a stamp lands on the figure, a page will not resolve. Where an amount cannot be read it stays empty. We do not infer it from the unit next door or from that year's average, however reasonable that would look.
  • Coverage is not even. Only a subset of the buildings we track carry enough sales across enough years to support a paid report, and on the rest we will tell you what we have rather than sell you a range the history cannot hold up.
  • Uncertainty is disclosed on the specific building, inside the report, rather than as a blanket line at the foot of a page. A general warning is not something anyone can act on.

Four things this is not

Worth saying plainly, so nobody has to guess where we stop.

  • Not a certified appraisal. Our valuation work is built from verified comparable sales and is meant for deciding what to offer and what to ask. When a bank disburses a mortgage it requires an avalúo signed by a licensed appraiser, and this does not replace that.
  • Not legal advice. We can show you that a folio carries a mortgage or an open seizure. What it means for your position, and how it gets cleared, is your lawyer's answer rather than ours.
  • Not investment advice. We publish what sold, at what price, and what that implies about a building. We do not promise a return and we do not rank one building above another.
  • Not a brokerage. We do not list properties, take commission, or represent either side of a deal. We are paid for the report, which is why the report can tell you the apartment is overpriced.

The method, applied to a whole neighborhood.

Costa del Este, Bella Vista, Avenida Balboa and Punta Pacífica have each had the full method run over them: every figure traced to deeds, every exclusion applied, every thin spot named where it matters. Free to read, and nothing in them is gated.

Read the Costa del Este report