Decision Accounting: From Nudge to Guardrail

Showing your work improves your work — and a 15-field guardrail formalizes it

The $57B GRC industry manages policies, controls, and risk registers but never requires structured reasoning documentation before a decision is finalized. This paper formalizes that missing architecture.

• Sixteen regulatory regimes on four continents independently converged on 13 of 15 documentation fields — undetected for 60 years.
• The guardrail reduces agency costs: ΔW = p_H·p_L·c·(q_H − q_L)/[(p_H − p_L)(p_H·q_H − p_L·q_L)] > 0 per Holmström (1979).
• Resolves Prat's (2005) conformism problem through multi-audience common agency without restricting information.

Reasoning evaporates before examination arrives — seven cases, $400B+ in fines

Bankers Trust, Boeing's 737 MAX (346 dead), SVB (209B collapse), Purdue Pharma (8.3B criminal resolution), Volkswagen (1,200+ excess deaths), Bhopal (22,000+ dead), Deepwater Horizon (11 dead, 134M gallons spilled) — all had GRC systems. None recorded the reasoning behind consequential decisions.

• Boeing's MCAS: single-sensor design accepted without documented risk trade-off; investigators spent years on forensic reconstruction.
• SVB: 55% of assets in long-duration bonds; no record of who decided, what risks were identified, or what alternatives were considered.
• Purdue Pharma: DOJ spent hundreds of millions reconstructing who knew what — evidence a reasoning record would have provided from the start.

Sixteen regimes on four continents independently converged on the same 15-field architecture

Thirteen of fifteen fields appear in every regime; fourteen base fields appear in every industry corpus; 215 of 224 cells in the regulatory corpus are populated. PREDICTION is the framework's sole addition to the regulatory consensus.

• Monaco Memorandum (US): individual accountability is the 1 priority; corporations must identify every individual involved.
• DORA (EU): traceable decision rationale required; personal fines up to €1M for senior management.
• UK Senior Managers Regime: individual executives personally liable with criminal sanctions including imprisonment.
• APRA CPS 230 (Australia): documented material risk assessments with board-level attestation.

Six literatures predicted the convergence would go undetected — and it did, for 60 years

Hayek (dispersed knowledge), Arrow (organizational limits), Simon (bounded rationality), Abbott (professional jurisdictions), Carlile (knowledge boundaries), and Akerlof/Stiglitz (information asymmetry) converge on a single prediction: this convergence will remain invisible under normal conditions of professional specialization.

• No single regulator or industry body had the cross-sector, cross-continent view to see the pattern.
• The convergence was found in regulated industries because they generate examination records — manuals, enforcement orders, criminal resolutions.
• Falsification: a new regime fails to converge on the 15-field structure.

The guardrail reduces agency costs — welfare gain formula proved from Holmström (1979)

ΔW = p_H·p_L·c·(q_H − q_L)/[(p_H − p_L)(p_H·q_H − p_L·q_L)] > 0 whenever the documentation signal is informative about effort. The proof is direct from Holmström's informativeness principle.

• The guardrail is a mandatory pre-decision documentation requirement — dimension-prescriptive but answer-nonprescriptive.
• It resolves Prat's (2005) conformism problem: under multi-audience common agency with audience uncertainty, conformism is strictly dominated.
• Five general equilibrium theorems: threshold equilibrium (GE1), tipping-point cascades (GE2), under-adoption (GE3), cost-of-capital amplification (GE4), dual-signal amplification (GE5).

200B–4.32T annual welfare gain from cost-of-capital channel alone — a lower bound

Governance quality predicts 88–136 bps (Ashbaugh, Collins, LaFond 2004) to 100–300 bps (GE4 model-calibrated) cost-of-capital spread. At 112 bps, a firm sees a 27% increase in stock value. Capitalized: 2T–43T at corporate rates, 86T–144T at social discount rates.

• The 200B–4.32T range counts only the public-equity fraction of the full DA addressable universe.
• PPP fraud loss alone ($200B) equals the conservative floor — from one government program decision without structured reasoning documentation.
• F-35 cost overrun ($100B+) is larger than most entries in the governance failure catalog; neither appears in the cost-of-capital estimate.

The framework applies wherever consequential decisions are made under uncertainty — households, SMBs, governments, military, civil society

The fifteen fields are not bounded by sector. They describe what any decision-maker requires to reconstruct any consequential choice under subsequent examination — whether the examiner is a regulator, a litigant, a board, a spouse, or a military tribunal.

• Households: job changes, relocation, college choice, marriage, home purchase — irreversible, high-stakes, made without structured reasoning discipline.
• SMBs: capital commitments, hiring, partnership choices, strategic pivots — no board, no regulator, no GRC platform.
• Governments: PPP fraud ($200B) — one pandemic policy decision with no documented alternatives or fraud surface assessment.
• Military: F-35 program — unit cost grew from 69M (2001) to 158M (2023); no contemporaneous record of who accepted each scope change.

Complete open-source specification: DRCS v2.0 with 15 field definitions, cryptographic integrity, and gaming detection

Published under Apache 2.0 license. Includes Merkle tree integrity structure, timestamp anchoring, verification procedures, and Gaming Signature Index (GSI) to detect boilerplate or evasion.

• Field 1 (WHO): named attribution — not committee or process — creates specific deterrence per Holmström (1979).
• Field 7 (AUTHORITY): documents formal vs. real authority (Aghion & Tirole 1997) — what information was available to the principal at the time.
• Field 8 (TRAINING): addresses qualification asymmetry — Boeing's MCAS failure involved personnel without documented training in the relevant software architecture.
• Field 16 (PREDICTION): Brier score accountability — requires calibrated probability before outcome realization, closing the hindsight-bias loop.
• Field 17 (SYSTEM WELFARE): non-derivable — required by W-Independence impossibility result (Postnieks, 2026a).

Eight falsifiable propositions with three immediate conditions that would end the theory

Most working papers commit to nothing. This one specifies its own obituary conditions: exact results that would cause the author to abandon the framework's strongest claims.

• Proposition 1: Documentation improves reasoning quality (behavioral channel) — converging support from medicine and intelligence analysis; governance-domain test specified but unexecuted.
• Proposition 4: Cost-of-capital spread shrinks with DA adoption — identification strategy via natural experiments.
• Proposition 8: Net agency costs decrease — ten research designs specified, three immediately executable by independent researchers.
• Immediate falsification: (i) a new regime fails to converge on the 15-field structure; (ii) a randomized trial shows no improvement in reasoning quality

Decision Accounting moves governance from nudge to guardrail — from influencing probability to requiring reasoning

The existing toolkit — regulatory mandates that prescribe outcomes and behavioral nudges that influence probabilities — has no instrument for the problem these cases share: no one recorded the reasoning. The guardrail fills that gap.

• Nudges change probability without producing a record; mandates remove choice; guardrails require process structure without prescribing the decision.
• The framework is the prerequisite for individual-level deterrence: without a contemporaneous, attributable, unalterable record
• The Pacioli parallel: double-entry bookkeeping took 300 years to become universal.