Arctic Tracker / What changed / Finding

Finding: pre-2001 Red List categories are not points on the modern scale

Status: this project's own reasoning. No adopted standard was located that covers comparison across the pre-2001 Red List systems, and this document exists so that the reasoning can be argued with rather than taken on trust.

This is the justification for Gate 3 on /effectiveness. It is the largest single exclusion the site applies: 25 of 43 species, against 2 at Gate 1, 4 at Gate 2 and 2 at Gate 4, leaving 10 that pass gates 1 to 4.

What the gate does

Gate 3 excludes a species from category-movement scoring when its baseline Red List category is not one of LC NT VU EN CR, the ordinal scale introduced with the 2001 criteria revision.

The 25 excluded baselines carry these codes:

Code	Meaning	Species
`K`	Insufficiently Known (pre-1994; not a risk category)	6
`LR/lc`	Lower Risk, least concern (1994 system)	6
`V`	Vulnerable (pre-1994 criteria)	4
`E`	Endangered (pre-1994 criteria)	3
`T`	Threatened (pre-1994 grouping)	3
`NR`	Not recognised	1
`NA`	Not assessed	1
`LR/cd`	Lower Risk, conservation dependent (1994 system, discontinued 2001)	1

Those baselines were published between 1986 and 2007. Gate 3 tests the code, not the year; the year is Gate 4's concern, and the two gates catch different species.

The argument

A category is a label produced by applying criteria. Two labels are comparable only if the criteria behind them are the same, or if the difference between them has been reconciled. Neither holds here.

Three of the codes are not risk categories at all. K, NR and NA state that the assessors did not know, did not recognise the taxon, or had not looked. Placing them on a risk gradient would require inventing a risk level that the assessment explicitly declined to assign. Eight of the 25 are in this group, and for them the gate is not a judgement call.

The rest are risk categories from a different system. V and E come from the pre-1994 system, LR/lc and LR/cd from the 1994 one. A 1994 LR/lc and a 2001 LC share a name and an intuition, not a definition. If they were treated as the same point and the species were later assessed LC, the site would report "no change", which is a claim about extinction risk derived from an artefact of nomenclature. If they were mapped to adjacent points instead, it would report a change that no assessor ever made.

Both errors run in the direction that flatters the analysis: they manufacture scoreable movement, or manufacture reassuring stability, out of a revision to the instrument. The paper is about whether a listing changed an outcome. An exclusion that is too strict costs coverage. A mapping that is too generous costs the finding.

The reconciliation that would fix this is retrospective reassessment, re-scoring the old assessment under current criteria, and it is exactly what this dataset does not contain. That work is done by assessors, from evidence, not by a lookup table applied downstream. This project is not in a position to do it and does not pretend the mapping is available.

What this document does not claim

It does not claim an external authority requires this exclusion. The standard cited for gates 1, 2 and 5 (SDG Indicator Metadata 15.5.1, section 4.c) governs what the Red List Index counts; it assumes assessments already on one consistent scale and does not say what to do when they are not. That silence is why this gate is marked as the project's own reasoning on /effectiveness rather than attributed to a standard.

It also does not claim the excluded species are unimportant, or that nothing can be said about them. Their assessment history is shown in full on each species page, including the legacy code and the year, with the gate stated in plain words. What is withheld is a *movement score*, not the record.

The cost, stated plainly

25 of 43 species are excluded here, and the cascade leaves 10 passing gates 1 to 4. A reviewer is entitled to read that as the analysis being unable to say much, and this site does not present it otherwise: /effectiveness reports the cascade as counts rather than reporting the survivors as a result.

The alternative was a category crosswalk that would have kept most of the 43 and produced numbers that looked like findings. That option was available and was not taken.

How to check this

The gate logic is gate3 in src/lib/effectiveness.ts, and ORDINAL_SCALE there is the list a baseline must be in. The legacy code labels are in iucnLabel in src/components/format.ts.

The counts above were measured against the deployed database, not transcribed:

`` classifyAll(client) .filter(r => !r.eligible && r.gate.gate === 3) .length // 25 ``

Every excluded species names its own gate on its species page, so the 25 can be enumerated from the site without access to the database.

Application: /effectiveness, gate 3 in the cascade
Adjacent gate: gate 4, same class of problem at the level of the year rather than the code, and argued inline on /effectiveness because it takes two sentences
Related finding: /findings/FINDING-INCOMMENSURABLE-QUANTITIES.md, which is the same failure in a different dimension: adding quantities whose units differ, rather than comparing categories whose criteria differ