The Compliance Game Never Ends

Danske Bank is the paper's canonical system-welfare failure

The paper defines the Danske Bank Estonia scandal as a Hollow Win: C=0, A=1, B=1. Regulators won by accepting AML procedures as adequate on paper. Danske won through fee income from the non-resident portfolio. The system lost through illicit finance, tax-base damage, and institutional trust loss.

• About €200 billion in suspicious transactions flowed through the Estonian branch between 2007 and 2015
• Danske paid about €2 billion in fines in 2022
• Executives retained bonuses and severance packages exceeding €10 million
• The Danish FSA had conducted multiple inspections and found AML procedures adequate on paper

The missing player is system welfare

The Missing System Theorem says the bilateral game G only contains regulator A and firm B. System welfare W is outside both payoff functions, so the system cannot win inside the current game form.

• A1: U_A and U_B are defined only over the bilateral action space (a, b)
• A6: W(d) is not observable inside G and appears only in G1 = A, B, C
• A7: dW/dΠ is undefined inside G because revenue and welfare are decoupled
• The (1,1,1) outcome is structurally invisible until C is added to the game

The Hollow Win is stable because every active player is already paid

1 claims that, under A1-A7, the unique stable equilibrium is (C=0, A=1, B=1). The regulator can show visible activity. The firm can preserve revenue. The compliance industry earns more when rule complexity grows.

• A wins by issuing mandates, conducting audits, and levying fines for procedural violations
• B wins by maintaining Π while avoiding sanctions through procedural compliance
• CI wins because compliance revenue increases with regulatory complexity (A4)
• C loses because no player has W in its payoff function

Firms can move from conduct x to conduct x' before regulation catches up

A2 and A3 explain why the compliance game keeps restarting. A firm adopts revenue-generating conduct x. The regulator classifies x as harmful only after a positive lag τ. The firm then shifts to x' that preserves nearly the same revenue while falling outside the current definition.

• τ is bounded below by administrative rulemaking, notice-and-comment periods, and judicial review
• A3 allows x' to generate Π' ≥ Π - δ, where δ can be arbitrarily small
• Danske's non-resident portfolio generated about €1.2 billion in annual fee income at peak from 2013 to 2015
• Boeing designed MCAS before FAA certification requirements caught up to its failure modes

More information does not change payoffs

2 says disclosure-only interventions cannot move the game out of Hollow Win. They add observations about W, but they do not make W an argument in U_A or U_B.

• Danske produced thousands of pages of KYC documentation, transaction monitoring reports, and regulatory filings
• Howard Wilkinson identified specific red flags in 2013
• Liability remains tied to a procedurally complete compliance narrative under A5
• Disclosure does not close τ or prevent conduct reshaping from x to x'

The paper estimates financial crime β_W at 0.7

β_W measures system-welfare destruction per dollar of industry revenue. For financial crime, the paper estimates β_W = 0.7, below the corpus mean of 5.0, while still implying about $2.1 trillion per year in welfare loss.

• Financial crime β_W = 0.7
• Implied annual welfare loss is roughly $2.1 trillion, about 3.6% of global GDP
• Compliance deadweight loss is $206.1 billion per year
• The paper classifies the constraint as institutional, not physical

Danske's paperwork passed while the suspicious-flow pipeline stayed open

Danske Bank Estonia shows procedural compliance coexisting with aggregate loss. KYC files and transaction reports satisfied the bilateral game, while the source of funds was systematically ignored.

• Period analyzed: 2007-2018, with detection in 2018 and whistleblower report in 2013
• About €200 billion in suspicious transactions moved through the branch from 2007 to 2015
• Peak non-resident portfolio fee income was about €1.2 billion annually from 2013 to 2015
• Case β_W is about 42, using €50 billion dW divided by €1.2 billion Π

Volkswagen used a defeat device to win against a static test

The Volkswagen case shows conduct reshaping against a fixed regulatory protocol. Cars passed EPA and CARB tests on paper while emitting far more NOx in real driving.

• Period analyzed: 2008-2015, with detection in 2015
• US diesel vehicle sales generated about $15 billion in annual revenue at peak from 2012 to 2014
• The clean diesel campaign aimed to help Volkswagen capture 25% of the US market
• NOx emissions were 10 to 40 times the legal limit in real driving

Purdue and the DEA quota system shared incentives to under-disclose

The Purdue case applies Conflictoring: formally separate regulated parties had aligned incentives to keep the supply-chain narrative intact. Purdue projected demand. DEA approved quotas. Both could say the supply chain was being managed.

• OxyContin generated about $35 billion in revenue from 1996 to 2019
• The paper assigns system-welfare loss of 500,000 deaths and more than $500 billion in direct costs
• Private penalties were about $10 billion
• Case β_W is about 14.3, using 500 billion dW divided by 35 billion Π

Delegated certification let MCAS stay outside the safety record until two crashes

The Boeing case turns on delegated certification under the FAA ODA program. Boeing had incentives to keep MCAS from becoming a major certification issue, while the FAA relied on delegated review.

• The 737 MAX was expected to generate $30 billion in annual revenue
• Two crashes killed 346 people
• The paper estimates system-welfare loss at 35 billion to 45 billion
• The FAA grounded the MAX in March 2019 and recertified it in November 2020

Field 17 makes system welfare part of the decision record before action

3 proposes mandatory 17-field Decision Accounting as the rule change R. Field 17, SYSTEM_WELFARE , requires a quantitative or bounded estimate of dW before a material decision is taken.

• DA replaces narrative compliance with structured decision records
• The record is cryptographically hashed and timestamped to block post-hoc revision
• U' includes penalties for Field 17 inaccuracy through α·Field16 - Field16
• U' includes penalties for Field 17 inaccuracy through β·Field16 - Field16

The paper's lower bound for the compliance reform dividend is $836.1 billion per year

The corollary defines RD_min as the compliance industry's deadweight loss plus the avoidable share of unmitigated harm. For financial crime, the paper uses 206.1 billion plus 30% of 2.1 trillion.

• DWL_CI = $206.1 billion per year
• H = $2.1 trillion per year
• λ = 0.3 based on corpus evidence from similar domains
• RD_min ≥ 206.1B + 0.3·2.1T = $836.1B per year