Radar Perene / Archive / science

Two versions, one revision: the ILI finding that did not survive

◦ Index methodology v2.2 (working papers with DOI). See the methodology.

Science

There is a document almost no indicator provider wants you to find: the old version. Not because it contains scandal — because it contains differences. Comparing what was claimed before with what is claimed now is the cheapest audit there is, and the reason so many market methodologies live on editable pages, undated, without history, forever in the present tense. In this house, the comparison is not merely available: it has been done, in writing, inside the public document itself.

The current version of the Housing Liquidity Index working paper — The Brazilian Housing Liquidity Index, version 2.0 — documents, in its own text, two findings from earlier versions that did not hold up under re-examination. The old versions remain preserved in the repository; the revision is dated and explained in the new document. None of this became public through a leak or outside pressure: it came through the house's audit protocol, working as designed — detect, correct, record.

Which findings they were and which specific test brought them down is detail the public paper dosed at the right measure, and this article does not go beyond it. What matters here is the anatomy of the process — because it is rare, and because its rarity explains a great deal about the indicator market.

The full cycle, no shortcuts

The road to a revision like this has three stations, and each has a more comfortable alternative. The first station is detection: the original findings were submitted to re-examination by the house itself, well after publication — the ruler Track 1 describes in robustness testing, pointed inward. The comfortable alternative was not to re-examine: finding published, page turned. It is the silent choice of much of the market, where an indicator never returns to the laboratory once it starts selling.

The second station is the decision: having established that two findings did not hold, the house removed them from what the index claims — instead of keeping them under softened language, which is comfortable alternative number two. The third station is the public record: the revision entered the paper's text, with the earlier versions preserved for comparison. Comfortable alternative number three: edit in silence and let the past disappear — impossible, here, by construction, because the DOI deposit freezes each version beyond the reach of whoever wrote it.

Why this strengthens the index rather than weakening it

The hasty reading sees in this episode a stain on the ILI. The informed reading sees the opposite, through an argument that comes from the very logic of measurement: every instrument that measures a difficult object produces, over its lifetime, claims that do not survive — that is a property of the problem, not of the provider. The difference between providers is not in having or not having fragile findings; it is in what happens to them. In a process without audit, the fragile finding stays in the product, silently working against whoever uses it. In a process with audit, it is found, removed and documented.

The reader who encounters a declared revision learns two things at once: that this specific finding fell, and that there is an active process hunting the ones that have not fallen yet. The reader who never encounters any revision in an old indicator learns something too — only about the process, and nothing reassuring. An instrument that never revised anything is not an instrument without errors; it is an instrument without an audit.

The detail that changes the incentive

The mechanism that makes this honesty sustainable is worth recording, because it is replicable: public versioning with a DOI turns revision from a cost into an asset. Under an editable-page regime, admitting error is expensive — the admission is the only proof the error existed, and the temptation to erase is permanent. Under a frozen-version regime, the error is already preserved forever in the old version; the only real choice is between revising it in the open or leaving it there, unanswered. The immutable deposit does not make the house more virtuous — it makes virtue the rational option. It is incentive design, not heroism.

That is also why the ruler applies to reading third parties: faced with any indicator, ask where the old versions are and what changed between them. The answer — and the existence of an answer — says more about the process than any promotional material.

Frequently asked questions

Why not detail which findings fell and why?

Because the public paper already records the revision at the dose that serves it: what changed between versions is there, verifiable. Re-amplifying the detail outside the document would add no auditability — the citable record is the deposited text itself, in both versions.

Aren't two revised findings a lot for a young index?

It is what re-examining a difficult object produces when it is actually done. The relevant number is not the count of revisions — it is the ratio between what gets re-examined and what gets published. Zero revisions in an instrument nobody re-examines is not a record; it is a gap.

Is the ILI's current reading affected by the revised findings?

Version 2.0 is the contract in force: what the index claims today no longer includes what fell. That is precisely the function of revision — the current product carries the survivors of the re-examination, not the original full inventory.

Is this protocol exclusive to the ILI?

No — the same ruler produced declared revisions elsewhere in the collection, including one case in which a paper's published conclusion is the refutation of the very thesis that named it. The trail reaches it two articles ahead.

---

Continue the trail: One index, many regimes: why Structural Ânima and Tactical Ânima are not the same index →

House reading: today's reading is in the Diário; the episodes that tested the house's measures, in the Atlas.

Submitting a finding of one's own to the same kind of re-examination — before time does it — is work the house carries out on request, in conversation.