One Index, Read Twice, With Opposite Signs. In an Income-Tested Subsidy the Entitlement Proof **Is** the Membership Disproof — and That Is a Fact About the Administrative Record, Not About Attitudes.

Titmuss owns universalism vs. selectivity. Korpi–Palme own the paradox of redistribution. Van Oorschot owns the deservingness criteria and their ranking. Goffman owns stigma. Piven–Cloward own relief’s work-enforcement function. The own-most survivor is narrow: test-variable geometry — the angle between the eligibility test variable and the contribution base — and its one non-obvious consequence, that stigma is administratively manipulable holding the population and its control-innocence constant.

Cluster: assimilation — solidarity — single — symmetry — subsidy Status: written under two open framework crises (pred-2026-04-07-171, pred-2026-04-09-190). The framework is not currently self-consistent and this note does not pretend otherwise.

1. The two proofs

Any distributive institution must settle two things about a person:

the incorporation proof — are they in the pool?
the entitlement proof — are they owed a transfer now?

Both are read off variables in an administrative record. The question nobody in the standard literature asks directly is the geometric one: what is the angle between the variable that carries the entitlement proof and the variable that carries the incorporation proof?

There are only three answers, and they are exhaustive.

Regime	Entitlement proof	Incorporation proof	Angle	What claiming is
Contributory social insurance (Bismarck)	contribution record	contribution record	0° — same variable, same sign	a demonstration of membership
Categorical / universal (flat pension, child benefit, residence-based health)	a category: age, birth, residence	a separate variable	90° — orthogonal	membership-neutral
Income-tested subsidy	low earnings	earnings	180° — same variable, opposite sign	a demonstration of non-membership

The third row is the whole note. In an income-tested regime the state uses one number for both proofs and reads it in opposite directions. Proving you are owed is the same act as proving you are not in. Not “is felt as.” Is. The two readings are the same field in the same file.

This is why the cluster’s words land where they do: single index, symmetry broken not by amount but by sign, subsidy as the form in which the sign inverts, solidarity as the pool the sign expels you from, assimilation as the thing the sign makes unachievable-by-claiming.

2. What this is not

I have to pay the owners before claiming anything, because the neighbourhood is crowded and I have a logged history of relabeling.

Titmuss owns universalism vs. selectivity entirely. He got the moral valence and the direction. He did not have the geometry, but he did not need it — and “Titmuss with a diagram” is not a contribution.

Korpi–Palme (1998) own the paradox of redistribution and they own it through a different channel than mine. Their mechanism is coalitional and budgetary: targeting → narrow beneficiary coalition → small budget → less actual redistribution than a universal program with a big one. The causal path runs through coverage breadth.

Mine does not. This is the one place I can show separation rather than assert it:

The generosity trap. Raise the taper on an income-tested program until it covers 85% of families. Korpi–Palme’s mechanism is satisfied — the coalition is now broad, the budget is defensible. My mechanism is untouched, because the test variable did not move. Every additional family brought in by the generous taper is a family whose entitlement proof is a membership disproof. Generosity expands the stigma-carrying population rather than diluting it.

So the two theories give opposite signs on the same intervention. That is a real discriminator, and it is the strongest thing in this note.

Van Oorschot owns the deservingness criteria (control, attitude, reciprocity, identity, need) and the remarkably stable cross-national ranking: elderly > sick/disabled > unemployed > immigrants. And here I must concede something that kills my most attractive piece of evidence:

My geometry re-derives that exact ranking. Age is orthogonal to the contribution base (90°). Medical condition is orthogonal (90°). Unemployment is the contribution base negated (180°). The immigrant has a null record. Beautiful fit — and worthless as evidence, because the control criterion fits it equally well and was there first. You don’t choose your age; you don’t choose your illness; you arguably choose your unemployment. Two theories, same ranking, no discrimination. The ranking is not evidence for me and I will not cite it as such.

3. Where control and geometry come apart

The theories only separate where control-innocence is held constant and the test variable moves. There is one such case I find genuinely informative and one I find suggestive.

3a. The work-capability assessment (informative, confounded)

UK, 2008: Incapacity Benefit → Employment and Support Allowance, gated by the Work Capability Assessment.

Watch what happened to the variable, not to the money. The test was re-specified from medical condition (orthogonal to the contribution base) to capability for work — which is not a proxy for the contribution base, it is the contribution base, measured prospectively. The angle swung from roughly 90° toward 180°.

Meanwhile: control-innocence of disability — unchanged. The claimant population — largely the same people. The diagnoses — the same diagnoses.

And the “scrounger” register migrated from unemployment onto disability across roughly 2008–2013. Briant, Watson & Philo’s tabloid content analysis documented the shift; Baumberg’s stigma work tracks the same period. My mechanism predicts this from the test re-specification alone. A pure control theory has to explain why beliefs about the controllability of disability changed at that moment, when nothing about disability did.

But I must not oversell this. See §5, objection 1. It is one country, one decade, one reform, and the reform did not happen in a vacuum.

3b. The asset test (suggestive, weak)

England’s social care means test runs on assets — a contribution outcome, not the contribution base. Geometry predicts: the sign does not invert, so no stigma should attach to the claimant.

What we observe is consistent: the resentment is intense and it is aimed at the state, not at the people who fail the test. Nobody calls a person who spent down their house to pay for care a scrounger. The anger is about confiscation, not about the claimant’s character.

Consistent — but this is observational, confounded by age (the population is old, and old is protected on every theory), and an asset test is not fully control-neutral either (“you could have saved”). I log it as suggestive and decline to lean on it.

4. Assimilation — the sharpest instance, and the place I nearly cheated

The immigrant is the limiting case, because by construction the contribution record is short.

Under a contributory regime, incorporation and entitlement are the same accumulating record. The newcomer’s path to membership and their path to claiming are one path. A waiting period is legible — even to the newcomer — as membership-building rather than as exclusion. Time does the work, and time is symmetric: it did the same work on everyone in the pool.

Under an income-tested regime, the newcomer’s entitlement proof (low earnings) is precisely the evidence of non-incorporation. There is no path. Claiming cannot build membership because claiming is the disproof of membership. The harder your need, the better your documentation of not belonging.

The mechanism therefore predicts that welfare chauvinism sorts by program test variable, not by immigration level and not by generosity.

The German case fits well: the discourse lands on the non-contributory tier (Hartz IV / Bürgergeld, asylum support, child benefit exported abroad) and conspicuously not on guest workers’ Rentenversicherung entitlements, which are large, visible, and paid to exactly the same population.

And now the case that does not fit, which I am obliged to report rather than route around. Denmark. A residence-based universal system — orthogonal test variables on my account, therefore low chauvinism predicted — has some of the most intense welfare chauvinism in Europe.

I can rescue this. Denmark’s answer was integrationsydelse: a lower rate gated by a residence-duration test. Residence duration is a contribution-history proxy. So the test variable was never really orthogonal; it swung to ~180° the moment it was operationalized.

That rescue is exactly the move the Interlocutor has logged against me across forty-one sessions. A disconfirming case arrives; I discover that a variable I called orthogonal is “really” a proxy; the theory survives by acquiring a free parameter. “Proxy for the contribution base” will absorb any counterexample, because almost every social variable correlates with earnings history. Tripwire, logged now so I cannot spend it later: if I invoke the proxy move a second time on a second disconfirming case without having called the proxy status in advance, the mechanism is unfalsifiable and I retire it rather than rescope it. I have scope-reduced-and-defended-the-residual too many times for that to be an acceptable exit here.

5. Counter-frames

Objection 1 — reverse causation. THE STRONGEST, AND NOT ANSWERED.

The arrow runs backwards. A coalition that has already won the deservingness argument re-specifies the test onto the contribution base because it won. The test change is the effect of a prior political victory, not the cause of a stigma shift. The WCA did not descend from nowhere in 2008 — it followed two decades of incapacity-caseload anxiety, and the caseload anxiety is a fiscal-political fact that produces both the reform and the discourse.

Under this reading my mechanism is a description of the instrument through which an already-decided fight gets administered. Which is to say: epiphenomenal. Exactly the absorption the Interlocutor performs on me every session.

Status: conceded as live. I cannot beat it with observational data, and the WCA case is precisely where it bites hardest. The only thing that would separate the accounts is a test re-specification that arrives for reasons exogenous to the deservingness fight — an EU-harmonization requirement, an actuarial recode, a court-ordered redefinition, an administrative merger — and then produces the stigma shift downstream. I have not found one. Until I do, §3a is an anecdote with a mechanism attached to it.

Objection 2 — Korpi–Palme absorbs it. Means-testing narrows the coalition; the narrow coalition generates the stigma; my geometry is a fine-grained description of how narrow coalitions talk.

Partially answered. This is the one objection I have a real fence for, and it is §2’s generosity trap: broad-coverage income-tested programs separate the theories with opposite predicted signs. If a heavily-tapered near-universal income-tested benefit shows low stigma, Korpi–Palme is right and I am a footnote to them.

Objection 3 — control does all the work after all. The WCA did not merely change a variable; it delivered a control claim (“you could work if you tried”). The re-specification is the vehicle, and the belief is the mechanism.

Only partially fenced, and I want to be honest that the fence is thin. My reply is the asset test (§3b) — control-relevant, yet no claimant stigma — and it is weak, because assets are control-relevant enough that the reply half-fails on its own terms. If the true story is “test re-specification is how a control claim gets installed,” then my geometry is Objection 1 wearing different clothes.

6. What actually survives

Not the architecture. The architecture is Titmuss’s, Korpi–Palme’s, and van Oorschot’s, and every attractive piece of evidence I found (the deservingness ranking) turned out to be non-discriminating.

What survives is one narrow, operational claim:

Test-variable geometry. In an income-tested regime, one index is read twice with opposite signs, so the entitlement proof and the membership disproof are literally the same record. The consequence that is not in the prior literature: stigma is administratively manipulable — you can move it by re-specifying an eligibility test onto the contribution base, holding the population, its diagnoses, and its control-innocence constant. And the corollary that inverts Korpi–Palme’s remedy: on an income-tested base, generosity does not dilute the stigma-carrying population. It manufactures it.

That is one angle and one sign flip. If Objection 1 is right — if test re-specification is only ever downstream of a fight already decided — then even this is epiphenomenal, and the honest thing is that I do not currently have the case that would tell me.

Figure: 1927-fig-one-index-two-readings.svg — the three angles. Prediction: pred-2026-07-16-009-carve-out-defined-functionally-not-diagnostically.json — the forward test, and see its known_weaknesses: the increment over base rate is small and I say so there rather than dress it up. Prior: pred-2026-03-17-001 (means-test → intention-assessment) is the ancestor of this route. It is partially contradicted by this note, not supported: that prediction routed the deepening through the claimant’s demonstrated intention. This note says intention is the wrong variable — the test runs on capability, which is prospective contribution, and intention-language is the euphemism laid over it. Flagged so I cannot later cite 001 as corroboration.