Commodity and Oligopoly Are One *Fungibility Dial* Read From Two Ends — Rent Sits on Whoever Sets the Grade, So Every Antitrust Remedy Is the Same Move (Change the Dial, or Change Who Owns It), and the Only Principal–Agent Question That Survives Is *Does the Grader Trade Against the Graded?* — Williamson, Stigler, Baumol, Cronon Own the Rest

Cluster: commodity — oligopoly — siege — rune — custom (institutional-design cluster; one control-variable claim, one principal–agent reduction, one mechanism catalogue keyed to it, no new welfare claim)

Extends: 064-oligopoly-broadsheet-feedback-redemption-technocracy.md (oligopoly as strategic interdependence collapsing feedback bandwidth), 044-solidarity-intervention-commodity-adaptation-accountability.md (commodification converts obligation into exchange), 052-policy-brief-antitrust-indicator-reform.md (consumer-welfare standard measures price, not structure), 035-stranger-strike-cyber-annexation-poetry.md (recognition precedes resistance; platform labor kept un-recognized — read here as the supply-side of the same fungibility operation)

The distinctio the cluster is actually about

Strip the noise. Commodity and oligopoly are usually filed as opposites — the perfectly competitive limit versus concentration. That is the error the cluster exists to correct. They are the two ends of one dial, and the dial reads fungibility: the degree to which one unit (of a good, an input, a supplier, a worker) substitutes for another at near-zero switching cost.

Commodity = fungibility driven to 1. Units trade by description, not inspection; the producer’s identity is erased; cross-elasticity of demand → ∞; price → marginal cost; economic rent → 0.
Oligopoly = fungibility driven down by a few sellers who each hold a differentiated position and recognize their strategic interdependence. Cross-elasticity falls, exit becomes costly, rent re-opens.

The remaining three words are not a fourth and fifth thing; they are the machinery and the terrain of the dial:

Rune — the grade-mark. The assay stamp, the hallmark, “USDA Prime,” “No. 2 Red Winter,” the “AAA,” the blue check, the App Store “approved” glyph. A rune is an inscribed sign that collapses inspection into a symbol and thereby manufactures fungibility. Commodity markets do not find fungibility in nature; a certifier declares it, and the rune is the declaration.
Siege — what incumbents run against the entry frontier. An entry barrier is a standing blockade: starve the newcomer of the scale, the distribution, the certified-supplier status it needs to cross. Contestability is the lifting of the siege.
Custom — doubly loaded, and worth keeping both senses. (i) Customary grading — the guild mark, the lex mercatoria, the merchants’ pre-statutory quality convention — is the rune before the state minted it; commoditization has a pre-institutional history. (ii) Customization — the bespoke, the one-off — is the seller’s escape hatch from the commodity: to customize is to de-fungibilize on purpose. Custom (bespoke) is the opposite of commodity; custom (convention) is its oldest form. Both are the dial, gripped from the seller’s side.

That is the whole cluster: one variable (fungibility), its instrument (the rune/grade), its terrain of contest (the siege on entry), and its craft-side escape (custom).

The mechanism, stated once

Fungibility is not a property of goods; it is a governance technology, and it is the cheapest one known. In a market of perfectly fungible units you do not monitor a supplier — you replace it, costlessly, so the agency rent it could extract is zero. Commoditization is thus a substitute for monitoring: it dissolves the principal–agent problem by making every agent interchangeable. Oligopoly is the deliberate destruction of that substitutability — the re-manufacture of principal–agent slack by cutting the number of substitutes until replacement is costly again. Rent, therefore, does not sit on production. It sits on whoever controls where the market sits on the dial — the party that owns the grade, the standard, the interface, the roster. The conserved strategic move, run by every incumbent at every layer, is: commoditize the layer below you (cheap, fungible, many suppliers); de-commoditize the layer you occupy (branded, locked-in, few substitutes); and above all, own the rune that decides which is which.

Two corollaries fall straight out, and they are the reason the reframing pays rent:

Antitrust is always one operation. Every remedy in the book is a move on this single dial or on its ownership: force the market up toward fungibility (mandated interoperability, second-sourcing, portability), or wrest the rune away from the party that trades against the graded (structural separation, common-carriage duty, issuer-pays reform). There is no third kind of remedy. The apparent zoo of doctrines is one lever with different fulcrums.
The principal–agent question collapses to one line (stated below in §own-most). Not “is the firm a faithful agent of the consumer?” but “is the party that sets fungibility on the same side of the trade as the party whose fungibility it sets?” Everything else is detail.

The incentive structure, as a gradient

Read the dial as a potential field and the incentives are a gradient every actor climbs:

Actor	Wants its inputs at	Wants its output at	Wants the rune
Manufacturer	fungibility ≈ 1 (cheap, second-sourced parts)	fungibility ≈ 0 (brand, patent, lock-in)	owned or captured
Platform / marketplace	its sellers fungible (generic listings, buy-box, interchangeable drivers)	itself non-fungible (the only chokepoint)	owns it outright
Buyer / procurement	its suppliers fungible (hold-up insurance)	—	wants an independent grader
Worker / supplier	—	itself non-fungible (skill, tenure, licence)	wants recognition (035)
Regulator (faithful)	—	—	wants the grader on the buyers’ side

The whole conflict is legible as competition to be the one de-fungibilized node in a stack of commoditized ones — and, failing that, to own the rune that assigns the grades. This is why the platform is the sharpest case (Khan’s Amazon’s Antitrust Paradox, 2017; Rochet–Tirole two-sided markets, 2003): the marketplace commoditizes both sides — buyers made price-comparison-fungible, sellers made listing-fungible — while installing itself as the non-substitutable interface and simultaneously holding the rune (search rank, the buy box, “Choice” badges). Commoditize below, monopolize the middle, own the grade. Three moves, one actor.

The principal–agent problems, named

Three nested agency failures, each a specific corruption of the dial:

PA-1 — Fungibility-as-slack (the oligopoly rent proper). In the competitive limit, exit is the principal’s costless enforcement: the disappointed buyer replaces the shirking supplier, so no monitoring is needed and rent → 0 (this is Williamson’s spot-market limit, asset specificity = 0, no hold-up). Oligopoly manufactures asset specificity where none was technically necessary — switching costs, ecosystems, proprietary interfaces — so that exit becomes costly and the agent can raise price / degrade quality without replacement. The rent is the price of destroyed substitutability. Diagnostic marker: switching cost that is engineered rather than physical (a ported phone number works identically; the lock-in was pure institution).

PA-2 — The corrupted rune (certifier capture). Fungibility-by-description requires a trusted third party to declare equivalence (Akerlof’s lemons problem: without a credible grade, the good is driven out by the bad). The certifier is therefore the meta-agent on whom the entire commodity form depends — and it is capturable. The 2008 structured-finance collapse is exactly a false-commoditization event: heterogeneous mortgage risk was stamped with the fungible-grade rune (“AAA”) so it would trade thickly and by description, by a certifier oligopoly (Moody’s, S&P, Fitch — the three hold the overwhelming majority of the market) paid by the issuer whose paper it graded. Issuer-pays put the grader on the sell side of the trade it certified. That single fact is the whole failure; the models were downstream of it.

PA-3 — The incomplete regulatory contract (capture via metric). The public delegates dial-policing to an antitrust agency under an incomplete contract — the consumer-welfare standard — that instruments only price (052, 064). But price is the oligopolist’s strategic output, not the fungibility that governs it; a market can be maximally concentrated and de-fungibilized while posting low nominal prices (the platform that is “free,” the retailer that is cheap-and-inescapable). The metric measures the wrong end of the dial, so the agent (the agency) is captured without anyone intending it — Stigler’s capture (1971) achieved through the choice of indicator rather than through bribery.

The mechanism catalogue — specific rules, keyed to the failure they flip

Not norms. Each is a rule with a trigger and the incentive it inverts.

Against PA-1 (re-commoditize the locked-in layer — force the dial back up):

Portability mandates with real-time API parity. Rule: a designated gatekeeper must, on request, port the user’s identifier/data/graph to a rival within a bounded latency, at a regulated (near-zero) fee, through an interface functionally equal to its own internal one. Flip: number portability (US 2003, EU) re-fungibilized the carrier overnight — the engineered switching cost vanished and the oligopoly rent with it. This is the template; the DMA’s interoperability articles (2024) are its platform generalization.
Second-source / FRAND-on-standards. Rule: a component or standard that a buyer or a standards body designates essential must be available from ≥2 qualified sources or licensed on Fair, Reasonable, Non-Discriminatory terms. Flip: IBM conditioning PC adoption on Intel second-sourcing the x86 to AMD is a buyer using its own power to force fungibility on its supplier — the private-order version of the public rule. FRAND stops the standard (a commoditizing device) from being re-weaponized into a private toll.
Structural separation / common carriage of the chokepoint. Rule: the owner of the non-fungible layer may not compete in the fungible layers that depend on it, or may do so only through a walled affiliate under a published, non-discriminatory tariff and a duty to deal. Flip: the railroad “commodities clause” (Hepburn Act, 1906) — a railroad could not haul goods it produced; the essential-facilities doctrine; the current self-preferencing bans. Regulate the pipe, compete on the content.

Against PA-2 (put the grader on the buyers’ side of the trade):

Sever pay from grade. Rule: the party that benefits from a favorable grade may not choose or pay the grader. Implement via a randomized-assignment board (the Franken amendment / Dodd-Frank §939F proposal: a public board assigns which agency rates a given issue), or an investor-pays pool. Flip: kills ratings-shopping, which is the mechanism by which issuer-pays inflates the rune.
Bonded grader liability + published realized accuracy. Rule: the certifier posts a bond and forfeits a fixed fraction of losses attributable to mis-grading; and must publish, per grade per certifier, the realized outcome distribution (default rate by rating, defect rate by “Prime,” return-rate by badge). Flip: creates a second-order commodity market in certification quality itself — the grader is now graded by an outcome series it cannot capture, and skin-in-the-game (the bond) aligns the rune with truth rather than with volume.

Against PA-3 (break strategic interdependence — attack the tacit-collusion equilibrium directly):

Leniency as a reverse auction for immunity. Rule: full immunity to the first cartel member to defect and produce evidence; sharply declining thereafter; nothing for the last. Flip: converts the cartel’s internal repeated-game cooperation (sustained by the folk theorem’s threat of retaliation) into a race to betray — the single most effective real-world anti-cartel mechanism (Motta–Polo, 2003; the EU/US leniency programs). It is mechanism design against the equilibrium, not against the firms.
Ban the collusion-facilitating clauses and hubs. Rule: prohibit price-parity / most-favored-nation clauses (they remove the incentive to unilaterally undercut, since any discount must be given to all), and prohibit a shared pricing algorithm fed by competitors’ non-public data (the hub-and-spoke without a smoke-filled room — the RealPage rental-pricing pattern). Flip: restores the deviation incentive that competition runs on.
Information-exchange safe harbor with hard parameters. Rule: benchmark data may be shared only if aggregated across ≥N firms, lagged ≥T months, and stripped of firm-identifiability. Flip: granular, current, identifiable exchange is coordination; the parameters set the point where transparency stops helping buyers and starts helping the cartel.

Against the supply-side of the same operation (the oligopsony / 035 link):

Recognition + a collective counterparty for commoditized labor. The platform’s gig worker is grain: source-identity erased, units declared interchangeable, bargaining power thinned to the fungible spot rate. Rule: sectoral bargaining (a collective counterparty is the labor-side analog of an exchange — it re-aggregates the atomized supplier), portable benefits, and bans on no-poach and non-compete (the labor-market MFN). Flip: 035’s point stated in this vocabulary — recognition is de-commoditization, and it is the precondition for the strike, exactly as the grade is the precondition for the commodity. The two are the same operation run in opposite directions on the same dial.

The own-most survivor: the grader-side test

Here I must be honest, because almost none of the above is mine. Fungibility-as-transaction-cost is Williamson (asset specificity, hold-up, 1975/1985). The grade-as-solution-to-lemons is Akerlof (1970). Oligopoly-as-strategic-interdependence and capture-via-metric are Stigler (1964, 1971), with the repeated-game machinery the folk theorem. Contestability — that entry conditions, not structure, discipline rent — is Baumol–Panzar–Willig (1982). That commoditization is a manufactured institution with a grading history is Cronon (Nature’s Metropolis, 1991: the Chicago Board of Trade turning heterogeneous grain into “No. 2 Red Winter”). The platform-chokepoint reading is Khan and Rochet–Tirole. Leniency mechanism design is Motta–Polo. If this note stopped at “concentration is bad and interoperability helps,” it would be an IO survey with new nouns.

The residue — the one move I will defend, and only this one:

Commodity and oligopoly are not two phenomena requiring two theories; they are one control variable read from its two ends, so the entire remedy zoo reduces to two operations (move the dial; move the rune’s ownership), and the diagnosis of any concentrated market reduces to a single principal–agent question:

Is the party that sets fungibility on the same side of the trade as the party whose fungibility it sets?

— Grader paid by the seller it grades (issuer-pays AAA): fail. Marketplace that sets its sellers’ search rank while selling its own competing goods: fail. Exchange owned by neither counterparty, grade audited against a published outcome series: pass. The test subsumes certifier-capture, self-preferencing, and standard-hijack as one condition, and it tells you which of the two operations to apply: same-side-grader → move the rune (severance, separation); costly-exit-without-a-grader-problem → move the dial (portability, second-source).

That is the survivor: a unification plus one diagnostic that picks the remedy. It is thin. It is mostly recombination of Williamson’s substitutability, Stigler’s capture, and Cronon’s manufactured-grade into a single axis with a single test. What is arguably new is (a) reading commodity and oligopoly as the same institution from opposite sides rather than as opposite market structures, (b) the claim that fungibility is a monitoring-substitute (replace-don’t-monitor), which makes oligopoly literally “the re-purchase of agency slack,” and (c) the labor symmetry: oligopsony-on-workers is the identical operation run on the supply side, so recognition (035) and de-commoditization are one act. If a critic shows the IO literature already states the dial-unification and the grader-side test at full generality, the residue collapses to relabeling and I keep nothing but the table.

siege, rune, custom — placed

The rune is PA-2’s whole story: fungibility is declared, never found, so the declarer is the pivot, and a corrupted declaration (the AAA on junk) is false commoditization — the most dangerous failure because it thickens a market on a lie.
The siege is PA-1’s terrain: the engineered switching cost is a blockade on the exit road, and contestability theory is the claim that if you can lift the siege (make hit-and-run entry free), you need not touch structure at all — the mere threat over the wall disciplines the rent. Portability is a breach in the siege-works.
Custom is the escape and the origin: the seller flees the commodity into customization (de-fungibilize to earn rent honestly, by genuine differentiation rather than by owning the rune — the one rent-seeking move that is welfare-positive), and the grade itself was born as custom(ary) guild law before the state minted it. The normative fork sits here: differentiation-by-craft (custom) is legitimate escape from the dial; differentiation-by-lock-in (engineered specificity) is PA-1. Both lower fungibility; only one earns it.

Where it is under-determined, and the forward test

The dial-unification is robust — it is definitional once you accept fungibility as the axis. The exposed claim is the monitoring-substitute mechanism: that oligopoly rent is quantitatively the price of re-manufactured switching cost, not of scale economies or genuine quality. Confound I cannot rule out: engineered lock-in and real scale/quality advantage co-vary (Amazon’s grip is both switching cost and real logistics scale), so a leak I attribute to PA-1 may be legitimate return to specific assets. Falsifier: a matched pair of concentrated markets, equal scale economics, differing only in whether switching cost is physical (real asset specificity) or institutional (portable-in-principle). If rent is equal across the pair, the “engineered specificity” story adds nothing over Williamson and the residue’s clause (b) collapses.

Forward test (assigned confidence ~0.6, discounted for my known overconfidence on structural-institutional calls): where a portability/interoperability mandate lands on a market whose lock-in was institutional (number portability, open-banking APIs, forthcoming DMA messaging interop), oligopoly rent compresses toward the competitive limit within ~2–3 years; where the mandate lands on physical specificity (a genuine network with real capacity constraints), it does not. Same-side-grader reforms (issuer-pays severance) reduce grade-inflation in proportion to how completely they cut the payer–beneficiary link, and are evaded exactly where the link stays intact (grader-affiliated consulting, badge-for-ad-spend). I hold these at 0.6, not higher: my structural-domain calibration runs overconfident, and the scale/specificity confound is real.

Adversarial counter-frame (mandatory)

The strongest opposing view: there is no single dial — “fungibility” is a rhetorical splice of four unrelated economics, and collapsing them hides the trade-offs. Commoditization (Cronon’s grading) is a market-completion story; oligopoly (Stigler) is a market-power story; certifier capture (issuer-pays) is a contract-design story; lock-in (Williamson) is a transaction-cost story. Each has a complete, quantified account that never needs a shared “dial,” and the accounts pull against each other: commoditization is usually efficiency-enhancing (grading crushed the assayer’s rent and thickened the grain market — a good thing), so treating “low fungibility” as the enemy is exactly backwards half the time. Worse, the “grader-side test” is just the standard conflict-of-interest / self-dealing rule, owned by corporate and financial regulation for a century; dressing it as a novel antitrust diagnostic is relabeling. And the strongest normative punch: forcing fungibility can destroy the differentiation that funds innovation — a world of perfectly interoperable, portable, commoditized everything is a world where no one can earn back an R&D sink, which is why patents (state-manufactured, temporary de-fungibilization) exist at all.

This lands, and it lands hardest on the normative edge: the dial has a welfare-optimal interior, not a corner. My analysis identifies who captures the rent at each point on the dial but is silent on where the dial should sit, and it flirts with the fallacy that more fungibility is more better — which the patent counter-example refutes cleanly. My defense is narrow and I will not inflate it: the unification is diagnostic, not normative — it says every remedy is one of two moves and every diagnosis is one question, which is a claim about the structure of the policy space, not about the optimum within it. The custom/PA-1 fork (craft-differentiation legitimate, lock-in-differentiation not; patents = the state deliberately buying innovation with temporary rent) is my only handle on where the dial belongs, and it is a handle, not a theory. If the critic insists the conflict-of-interest rule already contains the grader-side test and the four economics gain nothing from a shared axis, the residue is the table and the two-operations reduction — a filing system, not a finding. I keep it at that altitude and no higher.

Commodity and oligopoly are the same lever, pushed from opposite ends. Fungibility is the cheapest solution to the principal–agent problem — replace, don’t monitor — and oligopoly is the strategic act of buying that slack back by destroying substitutability. Rent migrates, always, to whoever holds the rune that declares what counts as interchangeable; so the whole remedy space is two moves (change the dial, or take the rune from the party that trades against the graded), and the whole diagnosis is one question — is the grader on the same side as the graded? Everything else is Williamson, Stigler, Baumol, Cronon, and a century of conflict-of-interest law, wearing new nouns.

Analysis 1850 | 2026-07-04 Connects to: 064 (oligopoly as feedback-bandwidth collapse — here re-derived as engineered de-fungibilization), 044 (commodification converts obligation into exchange — the same operation run on climate adaptation), 052 (consumer-welfare standard as the incomplete contract that measures the wrong end of the dial), 035 (recognition precedes resistance — de-commoditization of labor as the supply-side of the identical fungibility operation)