The substitution trap

The substitution trap can be written as a sequential-move game with three stages. In stage one the government selects a regulatory target T, which is either a product feature or a specific molecule. In stage two the industry observes T and selects a substitute s from the substitution set A'(T). In stage three the government observes s and may update T. The structural property that drives the result is that A'(T) is non-empty whenever T names a product feature rather than a functional class. The industry's payoff is increasing in the potency of s, because potency concentrates value per unit of the regulated attribute. The government's payoff depends on whether system welfare W is preserved. When the government targets a product feature, the industry's best response is to choose the most potent available member of A'(T), and that choice generically lowers W. Targeting the function instead removes the move: with A'(T) empty, stage two offers no within-class substitute.

Cowan's Iron Law of Prohibition holds that regulatory pressure on a substance drives the market toward more potent, more dangerous forms. The mechanism is a cost calculation. Prohibition raises the cost per unit of volume of the regulated substance, since each unit moved now carries enforcement risk. The industry responds by concentrating the active ingredient so that each unit of volume carries more value. The market therefore shifts from less potent to more potent forms. Read this way, the Iron Law is the special case of the substitution trap in which the regulated attribute is volume or weight. The generalization follows directly: any regulation that targets a product feature rather than a functional class can induce substitution toward more concentrated versions of that feature, in any product class that has a concentration dimension.

The three classical traditions of externality economics — Pigovian taxation⁴, Coasean bargaining⁵, and Ostromian self-governance — are not rival theories but distinct operations on a single structural defect: the exclusion of system welfare from the payoff space of every strategic interaction⁶. Functional regulation maps onto the Pigovian switch. The regulator imposes the system price on the function rather than on the molecule, so the regulated target cannot be redefined by a reformulation. The capture analysis still applies: the Pigovian switch is predicted to fail and restore the Hollow Win when capture intensity rises above a threshold. The relevant difference is that a function is harder to redefine than a molecule, so the functional version of the switch is more resistant to capture, even though it is not immune to it. Ostromian self-governance enters as the endogenous repair that folds the system coordinate back into the participants' own incentives rather than imposing it from outside⁷.

The substitution trap resolves to a Hollow Win, the outcome written as (C,A,B) = (0,1,1). The industry gains private profit from selling the substitute. The buyers of the substitute gain access to the function they wanted. Yet the system coordinate degrades, because the substitute the market selected is more potent and harder to control than the product it replaced. Standard bilateral analysis cannot display this result, because its payoff space carries only the two parties and not the system coordinate W on which both depend. That is why the trap reads, in conventional terms, as a private success and a public surprise: the same transaction that satisfies both parties lowers the welfare of the system that made the transaction worth completing.

The framework carries four boundary conditions worth stating plainly. First, it applies only when the substitution space A' is non-empty; where a product is a natural monopoly (one producer can serve the whole market more cheaply than several) or has no functional substitute, a product-level rule may hold. Second, it assumes the industry can identify and produce substitutes faster than the government can update its target. Where the government can pre-empt the substitution space and write the scope ahead of the reformulation, the trap can be avoided. Third, the framework does not resolve the political economy of functional regulation: because functional scope closes the substitution escape, exposed industries have a strong incentive to resist it. Fourth, the framework does not specify how to identify the welfare-relevant function F in a given domain, which remains empirical work.

The protected narrative uses the 2010 abuse-deterrent reformulation⁸ as a substitution exercise. The instructional question is whether making one misuse route harder changes the underlying function or shifts activity into substitutes; this analysis does not assert an observed outcome until an external case card supplies a source, population, time window, and locator. Treat this as a qualitative teaching reconstruction. It is a mechanism exercise, not a domain-wide numerical calibration.

RECALL — see the theorem chapter: W is not a function of the parties' payoffs. Pigou imposes a price on harm; its boundary is capture of the price setter, where the setter's incentives or funding depend on the activity being priced. Coase assigns a boundary and enables bargaining; it requires manageable transaction costs and standing for the system, meaning a recognized actor can bargain for the affected system. Ostrom organizes monitoring and sanctions; it needs a bounded community able to govern itself. The switches are distinct operations on the system effect missing from the parties' payoff reports.

A product rule targets a molecule, formulation, or delivery mechanism. A functional rule targets the welfare-relevant function across substitutes. Here the function is obtaining the opioid effect; A′ is every other route that can still provide that effect. Product regulation can produce a substitution trap when A′ remains available and a more potent or less governable substitute becomes attractive. It can still work when A′ is absent, covered, or unable to preserve the harmful function.

Start by specifying the outcome to be changed. “Reduce misuse,” “protect consumers,” and “make markets safer” are intentions; they do not identify the activity whose system cost must be reduced. State the product target, welfare-relevant function, people who bear the residual, and substitute set A′. Only then select a switch. Pigou is suitable when a defensible price can be set and the authority setting it has an independent mandate. Coase is suitable when an identifiable claimant can bargain and transaction costs do not overwhelm the remedy. Ostrom is suitable when participants form a bounded, repeat-interaction community capable of monitoring and graduated sanctions—penalties that escalate from a warning to restriction or removal when violations repeat. Complex cases may require several switches, but adding switches does not remove their separate boundary conditions.

The protected OxyContin reconstruction teaches a practical review discipline⁹. Ask what feature was constrained, what function users and suppliers were still attempting to perform, and which substitutes can perform it. Mark each substitute as documented, plausible, or unknown. A policy can reduce use of the named formulation and still fail its system objective if activity shifts into an uncovered channel. The failure signal is functional: evidence of persistence or migration of the residual harm, not merely a sales decline in the named product. Functional regulation is not a magic generalization. A function can be hard to define, can include legitimate activity, and can create new enforcement and distributional burdens. The analyst must show why the chosen boundary captures the harm mechanism and identify who bears the cost of the new rule.

Worked. A rule bans one delivery mechanism. The substitute map identifies two available mechanisms serving the same function. The correct output is a substitution-risk finding and a request for system-level outcome measures before declaring success. Faded. A warning label is prominent, but the reader cannot change the available option at the moment of choice. Which disclosure condition fails, and what intervention could change the answer? <details><summary>Answer key</summary> The missing condition is actionable choice: the information is present, but it arrives without a usable decision point, alternative, or ability to alter the relevant action. A bounded intervention could move the warning to the choice point, add a meaningful alternative or cooling-off step, or give the reader a mechanism to refuse the default. The answer must specify the changed feasible action and the boundary under which the intervention works; a visible label alone does not establish a system- welfare result. </details> Independent. Choose a product rule. Define its function, list A′, choose one or more switches, state each boundary, and give a success and failure signal. A complete answer may conclude the information is insufficient to choose a remedy. <details><summary>Answer key</summary> Any defensible answer names the function the rule is meant to constrain, lists the substitutes A′ that can perform that function, and selects a switch whose operator and authority are explicit. It then states the activity, geography, period, affected groups, and implementation boundary, plus one success signal and one failure signal that would detect substitution or residual harm. A conclusion that the evidence is insufficient is complete when it identifies the missing source or boundary and the observation needed to choose a remedy. </details>