Credibility After the Contract

Credibility depends on whether a stranger can reconstruct the decision

The paper defines public-contract credibility as a function of reconstructability: Cred(C, t+δ) = f(R(D_t)). If the record cannot produce a verifiable narrative of what was decided, by whom, why, under what authority, and what triggers reconsideration, credibility tends to zero over time.

• The Five-Minute/Stranger Test is the operative threshold for reconstructability.
• A failed test means R(D_t) becomes empty as delay δ grows and reconstruction cost rises.
• The claim targets post-contract unraveling, not only contract formation.

The bilateral game drops system welfare from the payoff space

In G = A, B , the state agency optimizes political stability and budget adherence while the firm optimizes profit. System welfare, Dimension C, is outside the payoff function even when both parties depend on it.

• Party A can count a project as on track while system welfare falls.
• Party B can report healthy earnings while future infrastructure users absorb losses.
• MST treats the exclusion of C as game geometry, not agent malice.

The Hollow Win looks successful inside the A-B game

The paper’s 8-outcome taxonomy identifies the Hollow Win as (C=0, A=1, B=1): contract terms appear satisfied, the agency saves face, the firm profits, and the public system degrades.

• In the bilateral payoff space, (0,1,1) appears Pareto-optimal.
• Only G₁ = A, B, C makes the welfare failure visible.
• The system loss appears as degraded infrastructure, trust erosion, and depleted enforcement capacity.

A record that fails FMST has zero credibility over time

The Reconstruction Credibility Theorem states that Cred(C, t+δ) = 0 if and only if FMST(D_t) = False. The missing element can be what, who, why, authority, or trigger.

• Reconstruction cost is αδ + β(1 - D_t/D_max).
• As δ grows, incomplete records become prohibitively expensive to reconstruct.
• Disclosure without reconstructability is classified as Disclosure Futility.

Non-reconstructable sequential decisions converge to the Hollow Win

The Hollow Win Convergence Theorem states that limδ→∞ P((C=0, A=1, B=1) R(D_t)=∅) = 1. Later amendments, waivers, and extensions repeatedly externalize system welfare costs.

• A4: (0,1,1) is a Nash equilibrium when R(D_t)=∅ because C is unpriced.
• A6: enforcement capacity declines as dE/dt = -Σ γ Complexity_i δ .
• A8: once trust falls below T_min, restoration cost exceeds the original welfare loss by factor ρ > 2.

The paper estimates β_W = 8.4 for public infrastructure procurement

β_W measures system welfare destroyed per dollar of annual industry revenue. For a typical $50B revenue sector, the paper attributes the loss to delayed or failed delivery, forensic reconstruction, and trust erosion.

• Delayed or failed service delivery: around $210B per year.
• Forensic reconstruction burden: $85B per year.
• Institutional trust erosion raises future capital costs by an estimated 3-5% and costs $125B per year.

Carillion collapsed after awards the government could not explain

Carillion’s January 15, 2018 liquidation left £7B in liabilities and more than 400 stalled public projects. The paper uses the case to show a complete reconstruction gap around continued awards after financial distress was visible.

• DWP awarded Carillion a £2B contract months before the collapse.
• The Ministry of Defence extended a facilities management contract worth £158M.
• The paper quotes the inquiry’s finding that government was systematically unable to explain its own decisions.

Carillion’s bilateral wins shifted £19.3B of welfare loss to the system

The case maps directly to (C=0, A=1, B=1): agencies gained short-term delivery and political cover, Carillion extracted private gains, and taxpayers, subcontractors, pensioners, and users absorbed the losses.

• Party A gain: £1.8B in off-balance-sheet infrastructure delivery.
• Party B gain: £1.1B in dividends and executive compensation from 2014 to 2017.
• System welfare loss: £7.0B direct liabilities plus £12.3B indirect costs.

A material decision does not count unless it has a 17-field DA record

The proposed rule moves evidence upstream: no material decision affecting a public contract counts as made without a contemporaneous Decision Accounting record, including Field 17 for system welfare impact.

• A7: W(C) enters Π(A, B, C) if and only if DA₁₆(D_t) is not empty.
• The record must document authority, evidence, alternatives, and system welfare impact.
• Low-quality processes face higher DA production costs than high-quality processes under A5.

The mandate changes payoffs by pricing documented welfare degradation

In G₁ = A, B, C , the payoff functions subtract expected sanctions when C=0: π′A = πA - λA(1-C)PenaltyA and π′B = πB - λB(1-C)PenaltyB. With sufficient detection and penalties, (0,1,1) no longer pays.

• λ comes from independent audit and public access to DA records.
• Penalties come from evidentiary presumptions and liability rules.
• Corollary 3.2 states that low λ or capped penalties can leave Hollow Win as a mixed-strategy outcome.

The β_W reduction comes from removing reconstruction and trust costs

4 states that β_W falls from βW₀ to βW₁, with βW₁ < βW₀/2 within one budget cycle under a reconstruction mandate. The proof argues that Cost_{Reconstruction} and Erosion_Trust disappear when records are complete.

• DA production cost is estimated at 0.1-0.3% of contract value.
• Non-reconstructable domains typically lose 8-15% of contract value through welfare loss, trust erosion, and reconstruction cost.
• The global public procurement Reform Dividend is estimated at $2.1T annually.

The framework stands or falls on reconstructable records changing outcomes

The paper gives empirical tests rather than treating the mandate as self-validating. A fully reconstructable record that still reaches Hollow Win at the same rate as a non-reconstructable one would refute the core causal claim.

• Falsification 1: FMST-satisfying records still converge to (0,1,1).
• Falsification 3: low-quality processes can produce indistinguishable 17-field DA records at the same cost.
• Falsification 4: full adoption fails to cut β_W by at least 50% within two budget cycles.