Racial Egalitarianism - The Disparity Inference

One inference organises most public argument about group outcomes. Groups differ measurably in income, education, and crime. Every claim here is about averages and none of it reaches an individual, who can sit anywhere in a distribution. If the groups themselves are equivalent in every relevant respect, then something external must be producing the gaps, and discrimination is the candidate. The conclusion follows cleanly from the premise, which is why the premise carries the weight and why it is rarely examined.

Six defenses are offered for that premise. They vary considerably in quality, and one of them is answered with a policy argument rather than with a reply. What produces the gaps, once the inference is set aside, is not this page’s job.

Intelligence is not heritable

If cognitive ability were not genetically influenced at all, differences between groups could not be genetic and the inference would go through by default.

The premise does not hold. Twin and adoption studies show substantial heritability within populations, and any account of how the trait arose at all requires that it was selectable. Within-species variation in cognitive and behavioural traits is uncontroversial elsewhere in biology, where working breeds differ from companion breeds in trainability without anyone attributing it to circumstance.

What this establishes and does not. Heritability within a population does not by itself establish that a difference between populations is genetic — the two are separate measurements and the second does not follow from the first. The argument reaches the weaker conclusion that a genetic contribution is possible and cannot be ruled out on the grounds offered. That is enough to defeat the defense, which claimed impossibility, and it is short of establishing the alternative.

Humans are 99.9% genetically identical

The figure is correct and the inference from it is not. Quantitative similarity in genetic material does not imply qualitative equivalence in outcomes: roughly the same order of similarity separates domestic dogs from grey wolves, populations that diverged more recently than some human groups did and that differ substantially in temperament. The traits at issue are carried in the fraction that varies, so a statistic reporting the size of the shared fraction reports nothing about them.

There is more variation within groups than between them

The premise is true and is granted. Individual variation within any population exceeds the difference between population averages.

Averaged variance conceals near-fixed differences. Specific variants approach universality in one population and absence in others: the Duffy-negative variant, conferring resistance to a form of malaria, is carried by nearly all West and Central Africans and effectively nobody without African ancestry; the EPAS1 variant that supports function at low oxygen is carried by about nine in ten Tibetans and by no Africans or Europeans; a variant governing body odour is carried by most East Asians and few others. A statistic that averages across all loci cannot see a locus where one group is at one hundred per cent and another at zero.

The defense fails as stated. The average does not establish equivalence, and it does not establish a difference in any trait it did not measure.

Race is a social construct

That a category is socially constructed does not make what it tracks unreal. Every category is socially constructed in the sense of being named and agreed on, including the ones for planets and molecules, and the naming is downstream of the thing rather than constitutive of it. Categories are built around distinctions because the distinctions were noticed.

The defense would carry weight if it argued something narrower — that the particular boundaries drawn are arbitrary, or that they track ancestry poorly. As deployed it argues that the referent is invented, which is a stronger claim and one it does not support.

IQ tests are unreliable

Here the objection is conceded and the subject changed.

That the measure has limitations is granted outright. The reply offered is that it remains the best predictor of success available, ahead of parental wealth, which answers a claim of uselessness but not a claim of unreliability, and these are different objections.

The rest of the section takes up a different claim. The reframing — that the measure predicts success in Western societies rather than intelligence as such — draws one real reply: no society exists where high cognitive ability is not an advantage, so what is being measured is not local to the West. What comes after that is not a reply but an argument that the reframing carries policy consequences its advocates would reject. That may be true and it does nothing to the defense. The reliability objection is not answered. An argument that a position is inconvenient for those holding it leaves the position where it was.

Any culture would transfer

One defense of the premise says that any group could adopt a culture that has served another, so the groups are equivalent in the respects that matter and remaining gaps must be imposed from outside. The transfer is doing the work. That a set of practices works for the population that generated it carries no guarantee it transfers, and a claim that it would is the equivalence premise restated rather than shown.

What this does not touch is whether practices explain the gaps in the first place. That is a different claim, and it is not a defense of equivalence. The page does not take it up.

What survives

Four defenses fail as stated. Two of those failures are clean: the identity statistic and the within-group variance statistic both report something true and are used to license something the numbers cannot reach.

One defense is untouched. The reliability objection is conceded on its own terms and then answered with a consequence rather than a rebuttal.

The sixth fails as a defense of equivalence and only as that. A transfer that has not been shown cannot carry the premise.

What the six establish is smaller than either side wants. Showing that the equivalence premise is assumed rather than demonstrated removes the licence for concluding discrimination from disparity alone. It does not replace the inference with another account of the gaps, and it does not leave a residual candidate standing by default. The inference fails. The cause of the gaps is not settled here.

Links

• DEI as Anti-Discrimination — the policy built on the inference, and what its own defenses inherit from it.
• Validity and Truth — why a true statistic can license nothing, and why premise truth and inferential validity are separate questions.
• Per Capita — the rate-versus-total distinction that group comparisons depend on.
• Who Gets Called Racist — what happens to the argument once the accusation enters it.
• Positions — where cause is located once this page has finished with the inference.

Open questions

If the equivalence premise is the load-bearing step, what evidence would actually establish it rather than assume it?

Where a transfer of practices is offered as the reason groups should be treated as equivalent, what would count as a test that the transfer had occurred?

Sources

Racial Egalitarianism: All Arguments Ranked & Debunked, 2026-07-24 — https://www.youtube.com/watch?v=aVZ2I4lzUlI. Supplied the six defenses, their ordering, and the population-genetic examples.