The Evidence That Decision Accounting Works: A Falsifiable Answer to the Investigative Objection

The ProPublica-style test asks for a real record that changed an outcome

The paper answers a narrow empirical objection: show one institutional case where a Decision-Accounting record caught a bad decision or prevented harm, rather than another governance framework that sounds good after failure.

• The objection comes from reporters who exposed Wells Fargo, Purdue Pharma, Volkswagen, and the UK Post Office Horizon scandal
• The requested evidence is concrete: a record, a decision, a harm prevented or detected
• The paper answers with four absence cases and two presence cases

The evidence claim has a built-in way to be wrong

The paper defines Decision Accounting as a tamper-evident, point-in-time reconstruction of the decision process: prediction made, welfare metric used, evidence considered, outcome observed, and accountable decision-maker.

• Proposition 1: records have caught bad decisions or prevented harms in real institutions
• Proposition 2: missing records were necessary conditions for large governance failures
• Proposition 3: the pattern follows from governance under uncertainty, not one regulator or sector
• Falsification requires a fully implemented system failing at comparable scale

A failed fully implemented system would refute the framework

The paper's falsification condition is specific enough to test. A counterexample must involve full implementation, not partial documentation or weak compliance.

• Mandatory recording of all material decisions with predictions and welfare metrics
• Tamper-evident storage with Merkle chain integrity
• Periodic independent audit of the record
• Regulatory access to the record
• Comparable magnitude measured by welfare loss, victims, or regulatory penalties

The Reconstruction Game lets bad decisions survive after the fact

Organizations make material decisions under uncertainty without records that can be reconstructed years later. Once harm appears, each player faces an evidence problem.

• Decision-makers can build selective ex post narratives around known outcomes
• Regulators cannot separate justified ex ante decisions from rationalized ones
• Victims can show bad outcomes, but not process failure
• Decision Accounting changes the information structure at the time the decision is made

The cross-selling metric lacked a welfare prediction before fake accounts spread

From 2002 to 2016, Wells Fargo used accounts per household as the primary retail performance metric. The metric rose from 4.2 to 6.3, but the decision was not recorded with a customer-welfare prediction or pilot evidence.

• Approximate period revenue: $380 billion
• System welfare cost: 3.7 billion in fines and settlements, about 2.5 billion in customer remediation, and an estimated $8 billion market-cap loss
• The unrecorded metric let the bank frame the fraud as rogue employee behavior
• A record would have shown no welfare analysis, no pilot study, no alternatives, and the decision-maker's authority

OxyContin marketing treated addiction risk as outside the welfare metric

From 1996 to 2019, Purdue marketed OxyContin as low-risk for chronic pain. The paper identifies the missing record as the executive decision to promote that claim without a reconstructable prediction about addiction consequences.

• OxyContin revenue: about $35 billion
• Settlements and fines: about $8 billion
• Estimated opioid-crisis welfare cost: $1.5 trillion
• Over 500,000 Americans died from opioid overdoses between 1999 and 2020
• A record would have exposed reliance on a single flawed retracted study and ignored evidence of oxycodone addictiveness

The defeat-device decision optimized redesign cost while hiding legal risk

From 2006 to 2015, Volkswagen installed software that detected emissions testing and reduced emissions during the test while allowing higher emissions in normal driving. The decision was not recorded with a prediction, welfare metric, or accountable owner.

• Costs paid: about $33 billion in fines, settlements, and buybacks
• Market capitalization drop after disclosure: about $30 billion
• Environmental cost: about 1 million tons of excess nitrogen oxide emissions
• A record would have named the cost-savings metric, the emissions-standard problem, the ignored legal risk, and the management-level decision path

Horizon data became prosecution evidence without a recorded reliability decision

From 1999 to 2015, the UK Post Office relied on Horizon accounting data as evidence in prosecutions even though sub-postmasters reported bugs and errors. The paper treats the missing decision record as central to the scandal.

• Compensation paid: about £1.5 billion
• Over 900 sub-postmasters were convicted of theft, fraud, or false accounting
• At least four sub-postmasters died by suicide before convictions were overturned
• A record would have shown whether bug reports were considered before relying on Horizon as sole evidence

MiC records caught an undocumented high-risk AML approval

Under Singapore's Manager-in-Charge regime, MAS requires senior managers to be accountable for material decisions. In 2020, records at a major Singapore bank showed a senior manager approved a high-risk AML transaction without documenting the reasons.

• MAS imposed about SGD 30 million in fines during 2018 to 2023 for governance failures, including record failures
• The specific enforcement action imposed a SGD 1.5 million fine and required remediation
• The record did not stop the approval, but it caught the undocumented decision
• Without the record, MAS would have had only a general non-compliance finding instead of individual accountability evidence

A recorded acquisition analysis stopped a capital-ratio breach

Under the UK Senior Managers and Certification Regime, a senior manager considering a significant acquisition maintained a record of predictions, welfare metrics, and evidence. The record showed the acquisition would reduce the bank's capital ratio below regulatory minimums.

• SM&CR has operated from 2016 to 2024 in the paper's case window
• FCA and PRA fines under the regime total about £500 million for individual accountability failures
• The 2022 acquisition decision was abandoned because the record made the capital impact reconstructable
• The mechanism was discipline before approval: the senior manager knew the decision could be audited

The paper shifts the dispute from theory to counterexample

The paper does not claim Decision Accounting prevents every bad outcome. It claims that records block the specific failure mode that depends on unreconstructable decisions, and it gives a falsification test for skeptics.

• Absence cases show missing records around metrics, marketing claims, software fraud, and prosecution evidence
• Presence cases show records catching an undocumented AML approval and stopping an acquisition below capital minimums
• A skeptic must produce a comparable failure inside a fully implemented Decision-Accounting system
• The practical implication is regulatory access to material decision records, not generic compliance paperwork