FX Fixing: A System Asset Pricing Model

Each 1 of dealer benchmark-service revenue is priced against 2.54 of welfare cost

The paper applies SAPM to the WM/Reuters 4 p.m. London fix using annual dealer benchmark-service revenue, not profit, as the denominator. The central calibration is 8.9 B/yr of system welfare cost against Π = 3.5 B/yr, producing Π = -$5.4 B/yr.

• Π = 3.5 B/yr: 1.0 B/yr legitimate benchmark revenue plus $2.5 B/yr manipulation rent
• Π_C = $8.9 B/yr across five monetized deadweight and real-resource channels
• β_W = 2.54 deterministic; Monte Carlo median β_W = 2.55 with 90% CI [1.4, 4.5]

The welfare estimate excludes transfers, fines, settlements, and client cash-flow losses

The paper uses finance-safe welfare accounting: cash-flow shifts between counterparties are redistribution unless they create deadweight loss, real-resource cost, or allocative distortion.

• Enforcement fines enter as conduct evidence, not as welfare cost
• Settlement amounts and client damage claims are excluded from Π_C
• The counted channels are execution distortion, integrity premium, remediation, litigation resource cost, and benchmark transition cost

Prior studies measured conduct, price pressure, compliance, litigation, or governance in isolation

SAPM joins five partial literatures into one revenue-denominated welfare ratio for the benchmark regime.

• Enforcement records document chat-room conduct and legal violations but not deadweight loss
• Microstructure studies estimate fix-window price pressure and spreads but not remediation or reform costs
• Litigation records describe damages disputes and settlements but mix redistribution with real legal costs
• Benchmark governance reports name methodology fixes but do not price private revenue against system cost

Fix execution markup contributes $5.0 B/yr, more than half of the welfare stack

The channel decomposition shows where the $8.9 B/yr composite welfare cost comes from. All amounts are annualized welfare costs, not penalties or transfers.

• C1 fix execution markup: $5.0 B/yr, 56% of total, from Evans (2017) and Ito & Yamada (2017)
• C2 market integrity premium: $1.5 B/yr, 17%, from FSB post-scandal spread data
• C3 regulatory remediation: $1.0 B/yr, 11%, from CFTC, FCA, and DOJ enforcement records
• C4 litigation real costs: $0.5 B/yr, 6%, from class-action dockets and legal fee estimates
• C5 benchmark reform transition: $0.7 B/yr, 8%, from FSB/IOSCO implementation costs

The clean benchmark counterfactual leaves only $0.8–1.2 B/yr of legitimate revenue

The paper defines Π₀ using post-reform, transaction-anchored benchmark designs and alternative execution methods rather than assuming a fix-free world.

• Central Π₀ = $1.0 B/yr under transaction-based, reconstructable governance
• Manipulation rent R = $2.5 B/yr above competitive execution
• Total revenue denominator is Π₀ + R = $3.5 B/yr

Only 0.48% of 100,000 simulations put β_W below 1.0

The robustness analysis propagates uncertainty across the five welfare channels with correlated inputs and seed 42. The code path is python -m pipeline.monte_carlo fx-fixing.

• Draws: 100,000; correlation ρ = 0.3; dependencies: numpy and scipy
• Median β_W = 2.55; mean β_W = 2.70
• 5th percentile = 1.41; 95th percentile = 4.49
• P(β_W < 1.4) = 5.00%

Client inelasticity, dealer coordination, and thin intent records create the fix-window floor

The Fix Window Manipulation Floor Theorem says voluntary market correction cannot push β_W below 1.0 when three benchmark-specific conditions hold together.

• FIX-1: client order flow is inelastic at the fix window, giving dealers a stable rent base
• FIX-2: information sharing and coordination incentives lower the cost of implicit collusion
• FIX-3: non-reconstructable trading intent records block after-the-fact verification
• Falsification requires a jurisdiction where voluntary mechanisms alone reduce β_W below 1.0 without sovereign benchmark governance

Channels 2–5 nearly clear the rent-denominated lower bound before execution markup is counted

The proof is revenue-grounded. Channels 2–5 equal $3.7 B/yr, or 1.48R at the central rent estimate, so a de minimis positive slice of C1 pushes D above 1.5R.

• Rent lower bound: D ≥ 1.5R
• Revenue condition: Π₀ is at most half of R
• Therefore β_W = D / (Π₀ + R) ≥ 1.0 under the theorem's assumptions

The 2015 WM/Reuters reform moved the game but did not make the regime welfare-positive

The paper treats FSB/IOSCO principles and the 2015 WM/Reuters reform as the documented G to G' transformation: from a 1-minute fix with thin records to a 5-minute transaction-anchored window with stronger governance.

• Ito & Yamada (2017): transaction-based reform reduced price impact by 40–60%
• UK FCA Final Notice, November 2014: £1.1 B penalty and mandatory conduct reconstruction
• G: 1-minute fix window, thin records, dealer coordination incentives
• G': 5-minute transaction-anchored window, audit trail, personal liability

Five enforcement jurisdictions supply the conduct record behind the calibration

The paper maps documented FX benchmark misconduct into welfare channels rather than treating enforcement totals as the damage estimate.

• United States: CFTC and DOJ records anchor manipulation conduct and remediation evidence
• United Kingdom: FCA records include the November 2014 final notice and conduct reconstruction requirements
• European Union, Swiss FINMA, and Singapore MAS extend the cross-jurisdictional record
• The enforcement record identifies the regime; SAPM prices its welfare footprint

FX fixing is measured on the same revenue-denominated scale as other SAPM domains

The paper uses β_W to compare institutional harm per dollar of industry revenue across domains, while keeping FX fixing's own calibration tied to 8.9 B/yr over 3.5 B/yr.

• FX fixing: β_W = 2.54 deterministic; 2.55 simulated median
• Comparison set: tobacco 6.5, Bitcoin 5.1, PFAS 35.2, AMR 1.60, nuclear power 0.53, monoculture agriculture 7.36, Frontier AI 7.4
• FX fixing classification: Infrastructure Hollow Win, (C,A,B) = (0,1,1)

The paper's remedy stack targets transaction anchoring, audit trails, and personal liability

Because the theorem's floor depends on inelastic order flow, coordination incentives, and unreconstructable intent, the proposed interventions map directly to those mechanisms.

• Benchmark administrators: transaction-anchored benchmark construction to reduce C1 execution markup
• Regulators: mandatory audit trails and conduct reconstruction to weaken FIX-3
• Dealer boards: surveillance, compensation clawbacks, and personal liability to reduce coordination payoff
• Buy-side clients: transaction cost analysis and record access to reduce exploitable fix-window flow
• Supranational standard-setters: FSB/IOSCO-style governance rules to move G toward G'