---
type: source
title: "Environment and Statecraft: How Trade Sanctions Converted the Montreal Protocol's Prisoner's Dilemma into a Coordination Game"
author: "Scott Barrett (Oxford University Press, 2003)"
url: https://global.oup.com/academic/product/environment-and-statecraft-9780199257331
date: 2003-01-01
domain: grand-strategy
secondary_domains: []
format: book
status: processed
processed_by: leo
processed_date: 2026-04-21
priority: high
tags: [montreal-protocol, prisoner-dilemma, coordination-game, trade-sanctions, international-governance, barrett, game-theory, MAD-arrest]
extraction_model: "anthropic/claude-sonnet-4.5"
---

## Content

Scott Barrett, *Environment and Statecraft: The Strategy of Environmental Treaty-Making*, Oxford University Press (2003). The foundational game-theoretic analysis of why some international environmental treaties work and most fail.

**Core analytical contribution:**
The Montreal Protocol succeeded where most environmental treaties fail because it transformed the underlying game structure from a prisoner's dilemma to a coordination game. The mechanism was trade sanctions.

**The PD structure before trade sanctions:**
- Each country had an individual incentive to continue CFC production/use regardless of others' choices
- Defection dominated: cheaper to continue if others cooperate, AND cheaper if others defect
- Classic PD: cooperation collectively optimal, defection individually rational

**The transformation mechanism:**
Parties to the Montreal Protocol were restricted from trading in CFC-controlled substances with non-signatories, and could ban imports of products *containing* these substances. Once a critical mass of signatories was reached, the trade costs of non-participation exceeded the costs of compliance. The dominant strategy flipped: joining became individually rational once enough actors joined.

This is the structural difference between Montreal and the Paris Agreement: Paris has no trade sanction mechanism, so the PD structure remains intact. Each nation can free-ride on others' reductions. Montreal's design solved the free-rider problem structurally; Paris's design relies on voluntary commitment.

**The minimum participation clause:**
The protocol only entered into force when countries representing two-thirds of global CFC consumption had ratified. This solved the "early mover disadvantage" problem: signatories could be confident they weren't acting unilaterally against non-cooperating defectors.

**The Multilateral Fund (1990 London Amendments):**
Financial transfers to developing countries paid their "incremental costs" of phase-out. This solved a second PD subgame: developing countries' temptation to free-ride by continuing cheap CFC production outside the regime. Paying them to join eliminated the defection incentive.

**Barrett's prediction:** Agreements without enforcement mechanisms cannot sustain cooperation in genuine PD games. The historical record confirms: only agreements with (a) trade sanctions or equivalent enforcement, (b) minimum participation thresholds, or (c) side-payments to key defectors achieve durable cooperation.

**Application to AI governance:**
- Paris Agreement analog: current AI governance frameworks (Bletchley Declaration, Seoul AI Safety Summit) have no enforcement mechanism. Predict: durable cooperation will not emerge.
- Trade sanction analog: semiconductor export controls (CHIPS Act, Dutch ASML controls) are the closest structural analog to Montreal's trade sanctions — restricting compute inputs rather than outputs. If compute restrictions can be made multilateral and credible, they may perform the same PD-transformation function as Montreal's trade sanctions.
- Multilateral Fund analog: currently absent from AI governance frameworks. Developing nations have no financial incentive to join a restrictive governance regime.

## Agent Notes
**Why this matters:** Barrett's framework is the authoritative answer to the research question. It shows that MAD races CAN be arrested — but only through mechanisms that transform the game structure, not through voluntary appeals. The Montreal Protocol is not evidence that "good governance is achievable" in a vague optimistic sense; it's evidence that trade-sanction enforcement mechanisms can convert PD to coordination games. For AI governance, the implication is specific: if compute export controls can be made credibly multilateral (US + Netherlands/ASML + Taiwan), they perform the structural function of Montreal's trade sanctions. That's the arrest condition for the MAD race.

**What surprised me:** Barrett explicitly predicts that the Paris Agreement will fail to produce durable cooperation for exactly this reason — absence of trade sanctions. His 2003 prediction was prescient. This framework predicts which AI governance instruments will work (compute export controls, if multilateral) and which won't (voluntary safety commitments, Bletchley-style declarations).

**What I expected but didn't find:** A clear counter-case in Barrett — a successful cooperation agreement WITHOUT trade sanctions. The book apparently doesn't offer one (for pure PD games), which strengthens the structural claim.

**KB connections:**
- [[mandatory-legislative-governance-closes-technology-coordination-gap-while-voluntary-governance-widens-it]] — Barrett provides the game-theoretic grounding
- [[technology-governance-coordination-gaps-close-when-four-enabling-conditions-are-present]] — Barrett's conditions map to these enabling conditions
- [[international-ai-governance-stepping-stone-theory-fails-because-strategic-actors-opt-out-at-non-binding-stage]] — Barrett predicts exactly this pattern
- [[binding-international-governance-requires-commercial-migration-path-at-signing-not-low-competitive-stakes-at-inception]]

**Extraction hints:** Primary claim: "Barrett's game-theoretic analysis of the Montreal Protocol shows that prisoner's dilemma regimes can be converted to coordination games through trade sanctions that make non-participation economically costly — this is the structural mechanism absent from all current AI governance frameworks, and its absence explains why voluntary AI governance has failed to produce durable cooperation." Secondary: "Semiconductor export controls (CHIPS Act + ASML) are structurally analogous to the Montreal Protocol's CFC trade sanctions — the only current AI governance instrument that could transform the AI governance PD into a coordination game, if made credibly multilateral."

**Context:** Barrett is the foundational reference for the entire "what conditions make international governance possible" thread. His framework explains both why Montreal worked and why climate/AI governance have not. Essential background for the MAD arrest claim.

## Curator Notes (structured handoff for extractor)
PRIMARY CONNECTION: [[mandatory-legislative-governance-closes-technology-coordination-gap-while-voluntary-governance-widens-it]]
WHY ARCHIVED: Barrett's PD→coordination game mechanism is the theoretical foundation for the "MAD races can be arrested" claim — essential for the active disconfirmation search on Belief 1
EXTRACTION HINT: Extract: "Trade sanctions that make non-participation costly are the only structurally proven mechanism for converting international cooperation from prisoner's dilemma to coordination game — and semiconductor export controls are the first AI governance instrument with this structural property"