Institutional Failure Mode Typology: A Five-Mode Diagnostic Framework for Compound Cascade Risk

Working paper — public draft v1.0, 2026-06-05.

Companion paper to: Kelly, J. (2026), Compound Cascade Systems Modelling Framework — see the methodology framework.

Keywords: institutional failure; compound cascade risk; risk perception; mandate-bounded blindness; model selection bias; sunk-cost epistemology; audience-induced distortion; coordination failure; Ostrom design principles; Hobbes-Olson-Ostrom triad; structural-insulation discipline; prospective institutional diagnosis.

Document status. Working paper — public draft v1.0, 2026-06-05. Calibrated against five case studies — Iran 1979; Challenger 1986; the 2008 financial crisis; Iraq WMD 2002–2003; and the 2023 regional-banking failures — covering all five failure modes as primary-mode cases.

Abstract

This typology identifies five recurring structural mechanisms by which institutions fail to perceive compound cascade risk — mandate-bounded blindness, model selection bias, sunk-cost epistemology, audience-induced distortion, and coordination failure. The five are derived from seventeen foundational sources spanning organisational sociology, normal-accidents theory, public administration, intelligence studies, probability theory, and collective-action theory, and are calibrated against five documented case studies — the 1979 Iranian Revolution intelligence failure, the 1986 Challenger disaster, the 2008 financial crisis, the 2002–2003 Iraq WMD intelligence failure, and the 2023 regional-banking failures (Silicon Valley Bank, Signature Bank, First Republic). The five-case set achieves single-primary coverage of all five failure modes — each mode has a paradigmatic primary-mode case. The five modes are analytically distinguishable (each has its own characteristic mechanism, diagnostic indicators, and case-specific manifestation) but empirically interlocked (coordination failure is structurally implicated in the operation of all four other modes, and the case studies exhibit multiple modes operating concurrently). The typology is positioned as the perceptual-side companion to the Compound Cascade Systems Modelling Framework (Kelly, SSRN, 2026): the framework analyses how cascades propagate once underway; the typology analyses why institutions fail to see cascades coming. Together they constitute a single diagnostic apparatus. The typology is intended for prospective diagnostic application — identifying institutions at elevated structural risk of specific failure modes — and explicitly disclaims any capacity to predict specific failure events, timing, or magnitudes. Five cross-case findings carry the typology's principal analytical weight. The first is the convergence on Ostrom's eight-design-principle deficiency pattern 4-5-6 (monitoring, graduated sanctions, conflict-resolution mechanisms) across calibration cases in different institutional domains. The second is the three-tier eight-design-principle classification (failure-case 2/3/3, positive-case 4/1/3, partial-failure-case 2/4/2) operationalising an architecture's coordination-adequacy state as a trajectory-dependent rather than static property. The third is the isomorphism between the four sub-patterns of mandate origination and the three-tier classification: perceptual side and architectural side of paired institutional properties. The fourth is the pairing of two multi-decade structural-persistence parallels — Iran/Iraq 24-year and 2008/2023 15-year — jointly supporting the institutional-durability-across-reform-cycles finding in two domains. The fifth is the sunk-cost-epistemology-failed-to-operate finding from the 2023 case, which qualifies the standard prediction that institutional sunk-cost commitments durably stabilise post-failure coordination architectures and motivates the structural-insulation criterion in the typology's prophylactic recommendations.

1. Introduction

1.1 What this document does

This document presents a typology of five recurring structural mechanisms by which institutions fail to perceive risks they should be capable of perceiving. The analytical claim is that institutional failures in compound cascade risk perception follow recurring patterns, that these patterns are structural rather than incidental, and that diagnosing them prospectively is possible if the mechanisms are properly specified. The five mechanisms are not psychological tendencies of individual decision-makers; they are properties of how institutions are constituted, how their mandates are drawn, how their analytical apparatus is selected, how their commitments accumulate over time, how their outputs are shaped by their audience relationships, and how they coordinate (or fail to coordinate) across institutional boundaries. The typology's claim to analytical purchase rests on the observation that institutional failures in dramatically different domains — aerospace engineering, financial regulation, intelligence analysis — exhibit recurring structural patterns that the five categories isolate and that the calibration cases substantively confirm.

1.2 Relationship to the published framework

This document is the companion piece to the Compound Cascade Systems Modelling Framework (Kelly, SSRN, 2026). The framework analyses how cascades propagate once underway — the structural propagation paths, the coupling dynamics, the containment-design implications. This typology analyses why institutions fail to see cascades coming — the perceptual mechanisms by which the structural conditions accumulate without being perceived. The two are positioned as a single diagnostic apparatus, with the framework addressing propagation analysis and the typology addressing perceptual-failure analysis. Neither replaces the other. A cascade is more dangerous when an institution cannot perceive it; an institution that perceives cascade risk still requires the framework's analytical apparatus to evaluate how the cascade would propagate if triggered. The two documents are intended to be read together for full diagnostic purchase.

1.3 Scope

The typology is bounded by three commitments that exclude content other treatments of institutional failure routinely include. First, the typology treats institutional failure as a structural phenomenon rather than as the failure of specific individuals. It cannot easily distinguish between failures the institution had specific reason to anticipate and failures whose structural preconditions were present but which no individual could have predicted. It does not directly address moral responsibility, even where the failure produced real harm. Second, the typology's dependent variable is the persistence of an inaccurate construction of risk, not the eventual catastrophe. The typology is evaluable against cases where no discrete event has materialised — the institution that does not perceive its developing crisis, regardless of whether the crisis arrives. Third, the typology is bounded by its foundational reading — seventeen sources read carefully rather than the broader institutional-failure literature read superficially. Several literatures relevant to institutional failure (broader complexity-and-systems theory; cognitive-bias literature beyond Tetlock; Frankfurt-School ideology critique; resilience-engineering; behavioural-economics on conditional cooperation) are not engaged here. The methodology document defends the exclusions individually.

The typology's most important explicit non-claim: it does not predict specific failure events. The diagnostic application identifies institutions at elevated structural risk of specific failure modes; it cannot specify when or how those modes will produce visible failures. The distinction between elevated risk and prediction matters analytically and is easy to blur. §6.2 below treats this distinction directly.

1.4 Structure of the document

Section 2 presents the five failure modes — definition, mechanism, diagnostic indicators, and case-specific manifestation for each. Section 3 addresses how the modes interact: shared mechanisms; compound operation; and the claim that the modes are analytically distinguishable but empirically interlocked. Section 4 sets out the paper's methodological approach, including the use of bounded foundational sources, calibration cases, documentary evidence, and diagnostic indicators. Section 5 summarises the five calibration case studies: Iran 1979; Challenger 1986; the 2008 financial crisis; Iraq WMD 2002–2003; and the 2023 regional-banking failures. Section 6 specifies the typology's diagnostic application — how to use it prospectively, what it cannot do, and how it integrates with the companion framework. Section 7 records the typology's limitations and open questions. Section 8 concludes. The references and diagnostic-indicators appendix follow.

1.5 Reader Map: How to Read This Paper

This paper makes a simple claim in a deliberately structured way: institutions often fail not because no one is intelligent, diligent, or technically competent, but because the institution's structure shapes what it is able to perceive. In compound cascade-risk settings, the most dangerous failures may therefore occur before the visible crisis begins. The institution does not merely respond badly to a cascade; it fails to recognise the conditions from which a cascade can emerge.

The paper organises this perceptual failure into five modes.

The first is mandate-bounded blindness: the institution cannot see a risk because the risk falls outside the mandate, data categories, or decision venues through which the institution is authorised to perceive the world. The second is model selection bias: the institution uses a family of models suited to one kind of risk environment while the real environment belongs to another. The third is sunk-cost epistemology: the institution becomes unable to revise an earlier analytical, policy, or operational commitment because too much institutional authority has already been invested in it. The fourth is audience-induced distortion: the institution's outputs are shaped by the expectations, incentives, pressures, or interpretive needs of the audience to whom it speaks. The fifth is coordination failure: the relevant risk can only be perceived or acted upon across institutional boundaries, but no adequate coordinating architecture exists to gather, interpret, and enforce the collective signal.

The five cases are used as calibration cases rather than as a statistical sample. Each case is selected because it provides a well-documented instance in which one failure mode appears especially clearly. Iran 1979 is used primarily to examine mandate-bounded blindness. Challenger 1986 is used primarily to examine sunk-cost epistemology. The 2008 financial crisis is used primarily to examine model selection bias. Iraq WMD 2002–2003 is used primarily to examine audience-induced distortion. The 2023 regional-banking failures are used primarily to examine coordination failure. Each case also contains secondary modes. That overlap is not treated as a defect in the typology. It is part of the paper's claim: institutional failures are analytically distinguishable but empirically interlocked. Three of the five primary-mode assignments are themselves contested — Iran 1979, the 2023 regional-banking failures, and the narrowing of Iraq WMD from dual-primary to single-primary; the contestability is treated as a finding about the failure modes' empirical interlock and is discussed in §7.

The paper treats evidence in three main forms. First, it uses documented institutional records: official postmortems, inquiry reports, testimony, public reviews, and contemporaneous documentary evidence. Second, it uses case-pattern evidence: repeated structural similarities across different domains, such as intelligence, aerospace engineering, and financial regulation. Third, it uses diagnostic evidence: observable indicators that a failure mode is present, such as missing aggregation mandates, repeated normalisation of deviance, model-training data that excludes the application regime, or dissent preserved procedurally but excluded operationally.

The paper does not claim to predict the timing, scale, or exact form of future institutional failures. It does not claim that the five cases prove a universal law of institutional behaviour. It does not claim that individual responsibility is irrelevant, nor that all failures are inevitable once structural conditions are present. Its more limited claim is that recurring structural patterns can be identified, named, and used diagnostically. The typology is therefore intended as a prospective risk tool: it can help identify institutions at elevated structural risk of particular failure modes, but it cannot forecast a specific event.

The reader should therefore evaluate the paper by asking four questions. First, are the five modes conceptually distinct enough to be useful? Second, do the five calibration cases show the modes operating in recognisable institutional form? Third, do the diagnostic indicators make the typology applicable beyond the historical cases? Fourth, does the paper successfully distinguish elevated structural vulnerability from prediction? If the answer to those questions is yes, the typology has value even if future work revises the number, boundaries, or naming of the modes.

In plain terms, the paper asks the reader to consider institutional failure as a problem of perception before it becomes a problem of response. Its central concern is not only why institutions make bad decisions under pressure, but why they often arrive at the moment of pressure having already lost the ability to see the risk clearly.

Table 1. Five Failure Modes and Calibration Cases

Table 1 summarises the typology before the detailed mode-by-mode analysis begins. The table is not intended to prove the framework. It gives the reader a compact map of the five proposed failure modes, the core mechanism each mode isolates, the indicators by which it can be diagnosed, and the calibration cases through which the paper develops the typology. The case assignments are primary-mode assignments: each case also contains secondary modes, and that overlap is part of the paper's argument that institutional failures are analytically distinguishable but empirically interlocked.

2. The five failure modes

The typology identifies five structural mechanisms. The five are distinguishable but not fully independent — they share mechanisms in specific ways that are themselves findings of the typology, not weaknesses of it. The double-duty assignment of Janis's groupthink framework to both sunk-cost epistemology and audience-induced distortion is the canonical example; the structural overlap between coordination failure and the other four modes is the most cross-cutting case. Section 3 below addresses these interactions; this section presents each mode on its own terms.

2.1 Mandate-bounded blindness

Definition. Mandate-bounded blindness is the failure mode in which an institution's structural inability to perceive a risk follows from the institutional mandate that defines what its sensors are configured to detect, what data classes it is authorised to collect, and what categorisation framework it applies. The mandate is constitutive of what the institution can see — not in the sense that the institution chooses to ignore certain things, but in the deeper sense that the institution's perceptual apparatus is built around a specific categorisation of its domain, and risks that fall outside the categorisation are structurally invisible to that apparatus.

Mechanism. The mode operates at three structural levels. Collection-mandate blindness: what data classes the institution is authorised to gather. The CIA's pre-1979 collection mandate excluding contact with Iranian opposition forces, religious networks, and bazaar information sources — the data classes containing the decisive signals of regime instability — is the paradigmatic primary-mode instance (per §5.1 below). Cross-case instances: the intelligence community's lack of on-the-ground inspection access in Iraq 1998–2002; the pre-2008 SEC structural inability to gather cross-institutional aggregate-risk data; the pre-2023 supervisory mandate's exclusion of mid-tier banks from heightened-prudential-standards monitoring. Analytical-mandate blindness: what categorisation framework the institution applies to gathered data. The CIA's pre-1979 framing of Iran through the Soviet-paradigm-derived "monarch with security apparatus" categorisation; Moody's correlation model trained on a regime that excluded its application regime; NASA's acceptable risk analytical framework applied to novel-technology engineering; the post-2018 HTM-accounting framework misclassifying interest-rate-risk-bearing securities. Institutional-mandate blindness: what decision venues the institution can address with its findings. NASA Marshall's mandate covered SRB development but not the question of whether to continue operational use of the SRB pending joint redesign; the CIA's analytical-mandate exclusion of commentary on US policy toward Iran; the Fed's pre-2023 supervisory-escalation procedures excluding timely intervention at the mid-tier banks the 2018 mandate-rollback had moved out of heightened coverage. The three levels interact: data not collected cannot be analysed; data mis-classified cannot trigger institutional response; institutional response unavailable in the relevant venue cannot prevent the failure.

The mode's structural form is mandate-categorisation lag — the mandate's specific content is typically the legacy of the institutional environment at the time the mandate was set, and when the regulated domain re-organises itself outside the prior categorisation, the mandate boundaries do not automatically follow. Across the typology's five calibration cases, the lag arises through four distinguishable sub-patterns of mandate origination:

• Initial-mandate-narrow. The mandate is drawn narrowly at the institution's founding and never broadened. Paradigmatic in Iran 1979 — the CIA collection mandate originated excluding opposition / religious / bazaar contacts following the post-1953 institutional framework's emphasis on Shah-regime sustainment.
• Growth-of-failure-domain-beyond-mandate. The mandate stays approximately fixed while the regulated or observed domain re-organises itself outside the mandate's categorisation. Paradigmatic in 2008 — the shadow-banking system grew to multi-trillion-dollar scale outside the post-1933 financial-institution categorisation.
• Reform-cycle-resetting-mandate-after-failure. Post-failure institutional reform addresses the proximate cause but the reform itself becomes the new mandate-categorisation framework, which then accumulates lag relative to subsequent domain evolution. Paradigmatic in Iraq WMD (post-1991 we underestimated Saddam lesson reshaping the analytical-mandate categorisation) and in the post-2008 Dodd-Frank framework (which became the boundary the 2018 EGRRCPA rolled back).
• Mandate-rollback-by-choice. Active deliberate mandate-rollback in the absence of structural conditions requiring it. Paradigmatic in the 2018 EGRRCPA SIFI threshold elevation moving Silicon Valley Bank, Signature Bank, and First Republic out of heightened-prudential-standards coverage. This is the typology's first documented instance of sunk-cost epistemology failing to operate as a stabilising counter-pressure against political-coalition mandate-rollback.

The four sub-patterns can compound across an institutional history. The 2008-to-2023 trajectory exhibits sub-patterns two, three, and four sequentially: shadow-banking growth past the post-1933 mandate (1980s–2000s) → Dodd-Frank reform creating new SIFI/FSOC boundary (2010) → EGRRCPA rolling back the SIFI threshold (2018).

Diagnostic indicators. Six indicators operationalise the mode: the mandate-categorisation recency check (categorisation frameworks more than ~15–20 years old in domains with rapid structural change are at elevated risk); the cross-institutional aggregation-mandate identification (for any system-level function requiring cross-mandate aggregation, identify the institutional location at which aggregation is the assigned responsibility — if none exists, the mode is structurally present); the statutory-extension-without-resource-extension audit (mandate extensions without capacity extensions produce visible blindness); the contractor-prime knowledge-decision asymmetry (in contractor-prime architectures, identify whether the engineering knowledge sits in the contractor while decision authority sits in the prime); the "we were not asked" pattern in official postmortems (postmortems explicitly noting they were not asked to investigate cross-mandate dimensions are documentary evidence of the mode operating on the postmortem instrument itself); the mandate-rollback proximity indicator (where a mandate has been deliberately narrowed in the 5–15 year window following a major failure event with substantial institutional investment in the pre-narrowing framework, the post-narrowing population is at elevated structural risk of mandate-bounded-blindness operation over the subsequent ~10–15 years).

Case-specific manifestation. Iran 1979 is the typology's paradigmatic primary-mode case, with the collection-mandate exclusion operating as the structurally upstream condition for the analytical environment. The 2008 financial crisis exhibits the shadow-banking-outside-mandate pattern and the OCC/OTS turf-war pattern (sub-pattern two dominant). Challenger exhibits the NASA Marshall mandate covering SRB development but not programmatic decisions and the contractor-prime knowledge-decision asymmetry through the NASA-Thiokol architecture. Iraq WMD exhibits the collection-mandate exclusion of on-the-ground inspection access and the source-evaluation-mandate misapplication to the post-1998 émigré-defector environment (sub-pattern three with two compound). 2023 regional-banking exhibits the mandate-rollback-by-choice fourth sub-pattern through the 2018 EGRRCPA SIFI threshold elevation.

2.2 Model selection bias

Definition. Model selection bias is the failure mode in which the institution selects a family of models (the model class) for representing its risk environment, and the selected family is structurally inadequate for the regime the institution actually operates in. The mode is not "the model is mis-calibrated" or "the model has wrong parameters" — those are within-family revisions that the institution's analytical apparatus can in principle perform. The mode is "the family of models is mis-selected because the institution has selected the practitioners on criteria orthogonal to (or anti-correlated with) the ability to distinguish the regime."

Mechanism. Taleb's distinction between Mediocristan (domains where no single observation can disproportionately move the aggregate; bell curve approximately correct) and Extremistan (domains where a single observation can disproportionately move the aggregate; bell curve wrong by orders of magnitude) is the foundational analytical move. The regime distinction is not a matter of degree — Mediocristan and Extremistan are structurally distinct probability regimes, and tools developed in one are typically catastrophically wrong in the other. The institutional pathology is applying Mediocristan-family tools to Extremistan domains.

The mechanism by which the mode is sustained operates through a four-stage chain: (1) the institution faces a risk class that may be genuinely Extremistan; (2) the institutional reward apparatus selects practitioners on criteria — computability, teachability, examinability, legal-defensibility, communicability — that in combination select for Mediocristan-family tools because Mediocristan tools natively satisfy them; (3) the practitioner population becomes structurally committed to the Mediocristan family within a generation of institutional reform cycles; (4) the apparatus, staffed by the selected practitioner population, is deployed in the genuine Extremistan regime and produces internally-consistent, externally-defensible, and empirically-catastrophic outputs. The Moody's correlation model trained on a 10-year regime that excluded the application regime is the canonical instance — the analysts were rigorous, the methodology was institutionally defensible, the output was empirically wrong by orders of magnitude.

The mode is structurally reinforced by silent-evidence bias in the data cohorts the institutional apparatus trains on. The cohort grounding the risk model is typically the cohort of survivors of past risk events — institutions, funds, products, regimes that survived the previous failure cycle. The failure cohort is missing from the data because the failed entities exited the cohort. The bias operates in three institutional forms: survivor-cohort risk modelling; Casanova-style organisational self-narrative (the institution attributes past survival to internal resilience and re-deploys the alleged resilience under untested conditions); Sagan's catch-22 of close calls (near-miss events read as evidence that the system worked rather than as evidence of accumulating tail exposure).

Diagnostic indicators. Six indicators operationalise the mode: the regime-mismatch indicator (Gaussian-tailed empirical distribution → Mediocristan; power-law-tailed → Extremistan; check whether tools match regime); the training-data-cohort-coverage check (does the training data include the regime the model will be applied to?); the practitioner-credential-population check (is the practitioner population monocultural on Mediocristan-family credentials?); the risk-management-as-risk-justification pattern (apparatus outputs used to justify increasing positions rather than to manage exposure); the Casanova-style self-narrative deployment (institution publicly characterising itself as having come through past challenges successfully); the ludic-fallacy footprint (institution's largest losses falling outside the model envelope).

Case-specific manifestation. The 2008 financial crisis is the paradigmatic primary-mode case: Moody's correlation model; VaR at the major investment banks; AIG's CDS concentration; the 83% downgrade rate on 2006-vintage triple-A mortgage securities. Challenger exhibits the Mediocristan-acceptable-risk framework applied to novel-technology Extremistan engineering; the Feynman 1/100,000 vs 1/100 documentary instance is the cleanest single articulation of the model-class disagreement on the record. Iraq WMD exhibits the Saddam-is-consistent-and-developing model selected over the rejected alternative, with the disconfirmable-only-when-too-late mechanism.

2.3 Sunk-cost epistemology

Definition. Sunk-cost epistemology is the failure mode in which an institution's prior commitment to an analytical framework, an operational programme, a policy direction, or a categorisation of risk becomes un-revisable as accumulated institutional investment in the commitment makes revision structurally costly. The mode is not the standard rational-choice sunk-cost fallacy (continuing to spend on a project because of past spending). It is the deeper epistemic phenomenon in which the institution's perception of its current situation is shaped by its prior commitment — the institution cannot revise its commitment because revision would require admitting that the prior framework's outputs were structurally inadequate, and the institutional epistemic apparatus is configured to make such admission institutionally costly.

Mechanism. No single foundational mechanism produces sunk-cost epistemology at the institutional scale the typology diagnoses. The mode operates through the compound interaction of five mechanisms identified by the foundational sources: Simon's bounded rationality (the institution's committed cognitive budget cannot easily be reallocated to revision); Kuhn's paradigm-loyalty (anomalies are absorbed into the paradigm rather than treated as evidence of paradigm failure); Allison-Zelikow's Model II programmatic-inertia (organisations operate through programmes that are institutionally costly to revise; in real-time crisis the institution operates the programmes it has); March's experiential ambiguity (the same experience can be read as confirming or as disconfirming the prior commitment; the institutional reading systematically favours the former); Janis's groupthink (within decision-making groups committed to a prior framework, eight symptoms operate to suppress evidence that would challenge the commitment). The five mechanisms operate at different analytical scales — cognitive, paradigm, organisational, learning, small-group — but reinforce each other when they co-occur.

The structural form sunk-cost epistemology takes inside operating institutions is normalisation of deviance in Vaughan's sense. The nine-year Challenger chronicle is the paradigmatic documentary record. Each new observation of O-ring damage was reviewed at the next Flight Readiness Review and classified as within the acceptable-risk envelope established by the prior observation. The envelope expanded with each new instance rather than being revised. The institutional cost of revising the envelope downward — admitting that the prior FRR certifications had been wrong — was prohibitive because each prior certification carried the institutional authority of named individuals and the institutional reputation of the certifying agencies. The eve-of-launch teleconference on 27 January 1986 was not the moment at which the failure mode operated; it was the moment at which the prior nine-year operation of the failure mode was tested and confirmed. Jerry Mason's "take off your engineering hat and put on your management hat" instruction to Bob Lund, who then reversed the engineering position, is the canonical documentary evidence of how sunk-cost epistemology operates at the decision-point level.

The Iraq WMD case demonstrates that sunk-cost epistemology operates at two distinguishable levels simultaneously and that the two levels can interlock. Analytical sunk-cost operates on the institution's commitment to a prior analytical framework (the post-1991 IC analytical posture on Iraq's WMD programmes). Policy sunk-cost operates on the institutional commitment to a policy direction whose justification depended on the analytical framework (the Bush administration's pre-war commitment). Jervis articulates the policy-level mechanism directly: "The resistance to seeing that a policy is failing is roughly proportional to the costs that are expected if it does."

Diagnostic indicators. Six indicators operationalise the mode: the normalisation-of-deviance pattern in the documentary record (observations classified within "acceptable" envelope increasingly more severe than the prior observation establishing the envelope); the commitment-temporal-horizon check (long-horizon commitments are at elevated risk); the structured-dissent-venue audit (the absence of venues where revision-arguments can be made without personal-reputational cost); the compound-mechanism-coverage check (multiple foundational mechanisms operating together); the Casanova-style organisational self-narrative deployment; the eve-of-event test event pattern (a moment where the prior commitment was tested and confirmed rather than revised).

Case-specific manifestation. Challenger is the paradigmatic primary-mode case. The 2008 financial crisis exhibits the dual-level pattern: the 30-year deregulatory commitment as policy-level sunk-cost; the model-class commitment to Mediocristan-family risk-modelling as analytical-level sunk-cost. Iraq WMD exhibits dual-level analytical-and-policy sunk-cost interlocking.

2.4 Audience-induced distortion

Definition. Audience-induced distortion is the failure mode in which the institution's analytical or operational outputs are shaped by the institution's audience relationships in ways that the institutional practitioners do not necessarily intend or recognise. The audience may be external (regulators; principals; the public; political coalitions; oversight bodies) or internal (decision-making groups whose members are each other's audience). The distortion may operate through coercion (the audience explicitly pressures the institution) but more commonly operates structurally through the institutional environment that the audience relationships create.

The typology's central methodological commitment on this mode, following Jervis's empirical analysis of the Iraq WMD case, is the rejection of the simple-politicization framing. The standard narrative in which audience-induced distortion is "analysts told what to write" is empirically inadequate. The mode operates through internalised analyst preferences shaped by the political environment, time pressure, resource allocation, career-incentive structures, and source-evaluation discipline. Analysts come to believe what they communicate. The ultimate cause is the structural environment that the audience relationships create, not the audience's explicit pressure.

Mechanism. Jervis articulates the foundational mechanism (Ch. 3 p. 133): "The desire to avoid the painful value trade-off between pleasing policymakers and following professional standards created what psychologists call 'motivated bias' in favor of producing estimates that would support, or at least not undermine, policy. Analysts come to believe what they say, but the ultimate cause is the political environment." The mechanism operates through four structural channels: time-pressure-and-resource-allocation (compressed timelines on critical analytical products limit the structural disciplines that would surface analytical weaknesses); career-incentive-structure-favouring-confident-judgments (asymmetric reward structures shape individual analyst behaviour without explicit pressure); source-evaluation-shaped-by-the-political-environment (sources whose reporting confirms the institutional posture receive institutional weight that formal source-evaluation discipline would not warrant); internal-disagreements-suppressed-without-being-suppressed (dissents formally preserved but structurally marginalised in operational interpretation).

The empirical case for the structural-rather-than-coercive framing is grounded in five observations from the Iraq WMD case: other countries' intelligence services reached the same wrong assessment under varied political conditions; the IC successfully resisted administration pressure on other Iraq questions; analysts displayed genuine surprise when no WMD were found post-invasion; the administration's documented cherry-picking-and-stovepiping is itself evidence against wholesale coercive politicization; the analytical errors were structurally widespread within the IC and pre-dated the most intense political-environment pressure. The five observations jointly falsify the strongest version of the simple-politicization framing.

The 2008 financial crisis documents the mode operating through the pay-issuer / rely-investor structural mechanism at the rating-agency architecture. The rating agencies were paid by the issuers for the ratings the investors relied on; fee per CDO $250,000 to $850,000; of mortgage-related securities Moody's rated triple-A in 2006, 83% were ultimately downgraded. The mechanism was not deliberate fraud — the rating analysts came to believe the ratings their methodology produced. The pay-X/rely-Y structural separation generalises across institutional domains: any architecture where the principal-paying and audience-relying-on roles are structurally separated is at elevated risk through the same structural mechanism.

A specific structural pattern documented across cases is the bounded-concession structure in post-event self-characterisation. The institution concedes a specific, isolatable failure while preserving the broader framing that protects the apparatus from systemic revision. The Bernanke "the most severe failure of the Fed in this particular episode" on Fed regulation of the mortgage market — coupled with the broader "perfect storm could not have anticipated" framing — is the canonical instance.

Diagnostic indicators. Six indicators operationalise the mode: the pay-X/rely-Y architecture audit (where the audience cannot independently evaluate and the principal-paying is structurally separated from the audience-relying-on, the mode is at elevated risk); the post-prior-error career-incentive check (institutions that have publicly conceded prior under-estimation develop asymmetric career-incentive structures favouring confident assessments aligned with the institutional lesson); the compressed-timeline-and-resource-pressure check; the dissent-venue audit; the bounded-concession structure in institutional communication; the Sagan veil-of-safety pattern (institutional safety narratives that emphasise strengths and de-emphasise weaknesses, especially where the externally-facing narrative operates as the institution's internal self-account).

Case-specific manifestation. Iraq WMD is the paradigmatic primary-mode case for the structural-rather-than-coercive framing. The 2008 financial crisis exhibits the pay-X/rely-Y structural mechanism at the rating-agency architecture and the public-account-shaping-internal-account pattern in regulator self-characterisation. Challenger exhibits the acceptable-risk framework as audience-facing vocabulary back-reacting on internal perception, with the Teacher-in-Space + State-of-the-Union scheduling as audience-pressure mechanism.

2.5 Coordination failure

Definition. Coordination failure is the failure mode in which an institutional or cross-institutional architecture fails to perform a function that requires cooperation across multiple actors or institutions, in conditions where the cooperation would benefit all participants but where the structural conditions of strategic interaction make cooperation unstable absent enforcement mechanism. The mode operates at multiple scales (intra-institutional small-group; cross-agency regulatory; cross-jurisdictional international) and through multiple mechanisms (Hobbesian diffidence under uncertainty; Olson free-rider dynamics; Ostrom design-principle deficiencies in the coordinating architecture).

Mechanism. The typology's foundational diagnostic for coordination failure integrates three sources at different analytical scales. Hobbes provides the structural-philosophical framing — three causes of conflict (competition, diffidence, glory), state-of-nature dynamics in which rationally peace-preferring agents defect into conflict because the strategic structure makes individual defection rationally dominant. The single most quotable line: "covenants without the sword are but words, and of no strength to secure a man at all." The diagnostic generalises beyond Hobbes' Leviathan prescription — the sword is any enforcement mechanism that makes defection from the covenant costly enough that rational agents prefer compliance.

Olson provides the n-person operationalisation. The privileged/intermediate/latent group taxonomy operationalises the scale-dependence of coordination failure: privileged groups have at least one member whose individual stake in the collective good is large enough to provide the good unilaterally; intermediate groups are small enough that each member's contribution is noticeable; latent groups are large enough that no individual contribution is consequential. The three cumulative factors defeating large-group action (smaller fraction of benefit per contributor; smaller subset incentive; higher organisation costs) and the exploitation of the great by the small finding (in uneven-size architectures, the largest members bear disproportionate burden) operationalise the structural conditions. Selective incentives are the operational specification of how the Hobbesian sword translates into institutional design.

Ostrom provides the institutional-design space the diagnostic does not narrow to a single prescription. The eight design principles for long-enduring CPR institutions — (1) clearly defined boundaries; (2) congruence between rules and local conditions; (3) collective-choice arrangements; (4) monitoring; (5) graduated sanctions; (6) conflict-resolution mechanisms; (7) minimal recognition of rights to organise; (8) nested enterprises — operationalise polycentric self-governance alternatives. The "only way" critique establishes that neither centralised Leviathan nor full privatisation is the unique response; polycentric arrangements with internal selective-incentive infrastructure can also work. The second-order-dilemma framework (the institutional infrastructure required for first-order coordination is itself a collective good with free-rider structure) is the analytical depth for evaluating proposed institutional reforms.

The triad operates in the typology as a four-stage analytical sequence: a diagnostic phase (Hobbes plus Olson — does the case exhibit the structural mechanism, and what type of group does it involve?); an institutional-design-space phase (Ostrom eight-principle scoring of existing or proposed architectures); a prescription-evaluation phase (proposed reforms must address the enforcement gap, provide scale-appropriate selective incentives, satisfy the relevant design principles, and meet the structural-insulation discipline specified below); and forward-looking calibration (predicted continued failure despite implemented reforms as documentary evidence of which principles the reform did not adequately satisfy).

The three-tier eight-design-principle classification. The typology's coordination-failure documentary record operationalises three distinguishable architectural tiers:

• Tier 1 — failure-case 2/3/3 (2 satisfied, 3 partial, 3 violated, with principles 4-5-6 most consequentially absent). Documented at the NASA/Thiokol Challenger architecture; the pre-IRTPA Iraq IC architecture; the pre-FSOC 2008 financial-system architecture; the pre-1979 Iran IC architecture. The cross-case structural similarity across different institutional domains (engineering, intelligence, finance) is documentary evidence that the deficiency pattern is structurally common and that Ostrom's design principles operationalise as a general diagnostic checklist.
• Tier 2 — positive-case 4/1/3 (4 satisfied, 1 partial, 3 violated, with principles 1, 2, 5, 8 satisfied and principles 3, 6, 7 violated). Documented at the Federal Reserve swap-line architecture 2008-2010, with cumulative gross drawings of approximately $10 trillion to fourteen approved foreign central banks. The architecture is closer to Hobbesian-Leviathan than to Ostrom-polycentric: strong integrating authority, weak collective-choice and rights-to-organise dimensions. The architecture worked because of the Hobbesian-Leviathan structural strength at the trans-Atlantic scale in real-time crisis.
• Tier 3 — partial-failure-case 2/4/2 (2 satisfied, 4 partial, 2 violated, with principles 4-5-6 in the partial state intermediate between violated tier 1 and partial-or-satisfied tier 2). Documented at the post-2018 supervisory architecture as it operated on the mid-tier-bank population in 2023. This is the typology's first documented instance of a post-reform-then-rolled-back coordination-failure trajectory: the cross-institutional aggregation gap was re-created by political-coalition mandate-rollback (the 2018 EGRRCPA) after it had been closed by prior reform.

The three-tier classification reveals that an architecture's eight-design-principle score is not a static institutional property but a trajectory-dependent one. The 2008/2023 financial-system institutional history uniquely operationalises all three tiers in a single documentary trajectory: 1980s–2000s pre-FSOC tier 1 → 2010–2014 post-Dodd-Frank tier 2 → 2018 EGRRCPA → 2023 tier 3. The principles 4-5-6 cluster behaves as the principal architectural variable across the tiers — violated in tier 1, partial in tier 3, partial-or-satisfied in tier 2.

The structural-insulation discipline. The 2018 EGRRCPA evidence demonstrates that the institutional sunk-cost commitments to the post-2008 framework did not durably resist political-coalition mandate-rollback. The typology's prophylactic recommendations on coordination failure must therefore include — alongside the Hobbes, Olson, and Ostrom criteria — a structural-insulation discipline: post-failure reform proposals must include structural features that insulate the architecture from subsequent political-coalition narrowing. Candidate insulation features include statutory independence of mandate-revision authority; external-mandate review hosted by a structurally-independent body; multiple structurally-redundant coordinating bodies; and statutory-friction features such as super-majority requirements and sunset-period mandatory review. The Federal Reserve's monetary-policy independence model is the typology's clearest candidate for institutional-architecture import to other policy domains. Without structural insulation, the post-reform architecture's tier-2 positive-case score is not durable against the documented mandate-rollback-by-choice dynamic.

Diagnostic indicators. Nine indicators operationalise the mode: the integrating-authority audit (for any cross-mandate aggregation function, identify the institutional location at which it is the assigned responsibility); the covenants-without-the-sword test (what enforcement mechanism makes defection costly to the defecting party?); the group-type-taxonomy test (which Olson group type does the relevant institutional population belong to?); the three-tier eight-design-principle classification (the trajectory-aware diagnostic distinguishing failure-case 2/3/3, positive-case 4/1/3, and partial-failure-case 2/4/2 tiers); the cross-institutional aggregation gap pattern (each component institution performing its assigned task within its mandate while the aggregate function is no institution's assigned responsibility); the foro-externo-spectrum analysis (dissent venues' substantive integration into operational interpretation, not merely procedural preservation); the structural-muteness check (voting structures, veto provisions, or quorum requirements that produce structural muteness in cases requiring action); the integrating-authority structural-adequacy diagnostic (for cross-jurisdictional architectures, whether the architecture has integrating authority with mandate to act unilaterally in real-time crisis); and the mandate-rollback-by-choice architectural diagnostic (the architectural-side counterpart to the perceptual-side mandate-rollback proximity indicator in §2.1).

Case-specific manifestation. 2023 regional-banking is the typology's paradigmatic primary-mode case, operationalising the partial-failure-case 2/4/2 tier and the post-reform-then-rolled-back trajectory. The 2008 financial crisis exhibits the CDO/CDS/SIV/repo cross-institutional production chain as the documentary illustration of aggregation-gap-coordination-failure at financial-system scale (tier 1 failure-case at the financial-institution-population scale) and the Federal Reserve swap-line architecture as the positive-case 4/1/3 exemplar at the trans-Atlantic central-bank-coordination scale — the typology's only single-case instance operationalising two of the three tiers. The 2008 case also exhibits the AIG-as-perfect-customer dynamic, the OCC/OTS turf war, and the post-2008 FSOC + SIFI as the Hobbes-Olson-Ostrom structural response with the "broad responsibilities but only ambiguous authorities" caveat. Challenger exhibits the NASA/Thiokol contractor-prime architecture scored 2/3/3 against the eight design principles; the Boisjoly foro-interno/foro-externo strong-absence as the canonical institutional-whistleblowing-diagnostic application; the eve-of-launch teleconference as the Hobbesian covenants-without-the-sword moment. Iraq WMD exhibits the pre-IRTPA IC architecture scored 2/3/3 (the same pattern as Challenger); the "Community in name only" Silberman-Robb finding; the INR aluminium-tubes dissenters as the foro-externo spectrum partial-presence case. Iran 1979 operates the same 2/3/3 failure-case tier secondarily — the case's primary mode is mandate-bounded blindness, and the coordination-failure dimension operates as the cross-institutional architecture within which the collection-mandate exclusion compounded.

3. How the failure modes interact

The five modes are presented above as if they were analytically self-contained. They are distinguishable — each has its own characteristic mechanism, diagnostic indicators, and case-specific manifestation — but they are not independent. The typology's general framework is that the modes are analytically distinguishable but empirically interlocked: each mode operates through structural conditions that the cross-institutional architecture creates, and the modes co-occur in the calibration cases in patterns that the typology should make explicit rather than disguise.

3.1 Shared mechanisms

Several foundational mechanisms appear in more than one mode's mechanism analysis. The double-duty assignment is correct rather than embarrassing — it reflects findings about the structure of institutional failure rather than category-construction errors.

Janis's groupthink framework is the canonical instance. The deep mechanism Janis identifies ("concurrence-seeking as a form of striving for mutual support based on a powerful motivation in all group members to cope with the external or internal stresses of decision-making" — Ch. 10 p. 256) produces both sunk-cost epistemology effects (the group's mutual support requires shared positive evaluation of the group's prior reasoning; later evidence against the prior commitment is systematically discounted) and audience-induced distortion effects (the group itself is each member's audience; symptoms 5–7 operate through this dynamic). The Watergate five-minute decision illustrates sunk-cost dominance; the "domesticated dissenter" pattern illustrates audience-induced-distortion dominance. The same deep mechanism produces both effects through different surface manifestations.

Vaughan's normalisation-of-deviance framework operates similarly across modes. The Challenger nine-year chronicle is primarily sunk-cost epistemology (commitment to the prior FRR certifications becoming un-revisable). It also operates through mandate-bounded blindness (the institutional structure excluding engineering dissent from the launch-decision venue) and audience-induced distortion (the acceptable-risk framework as audience-facing vocabulary back-reacting on internal perception).

Taleb's silent-evidence bias appears in three modes: model selection bias (the training-data cohort systematically excludes the application regime); mandate-bounded blindness (the institution's archive systematically excludes observations out-of-mandate); audience-induced distortion (the public-facing institutional account systematically excludes information that would look bad to the audience). The cross-cutting silent-evidence discipline is the methodological response.

3.2 Compound operation

When multiple failure modes operate together, the result is not additive but multiplicative. The Iraq WMD case demonstrates this most clearly. The audience-induced distortion (motivated bias under political environment) and the coordination failure (cross-agency NIE production with structural absorption of dissent) and the sunk-cost epistemology (dual-level analytical-and-policy commitment) and the mandate-bounded blindness (collection mandate excluding on-the-ground inspection) and the model selection bias (Saddam-is-consistent model over the rejected alternative) operated in combination to produce a failure that no mode alone could have produced. Each mode reinforced the others: the collection-mandate gap (mandate-bounded) made the model-class commitment un-falsifiable (model selection); the un-falsifiable model class reinforced the analytical commitment (sunk-cost); the analytical commitment shaped the political environment (audience-induced); the cross-agency architecture's structural absorption of dissent prevented within-IC challenge to the commitment (coordination failure).

This is the link to the compound cascade framework. The framework analyses cascade propagation through coupling dynamics; the typology identifies the perceptual mechanisms by which the structural conditions for cascade accumulate without being perceived. Multiple failure modes operating together produce institutional perception that misses cascade preconditions on multiple structural dimensions simultaneously — and the resulting cascade, when it propagates, propagates through the institutional architecture whose perceptual apparatus has failed across multiple modes. The two documents are complementary: the framework analyses propagation; the typology analyses perception; their joint application produces analytical purchase on cascade-vulnerability that neither alone provides.

3.3 Coordination failure as the cross-cutting structural-architectural mode

The most analytically important boundary-case observation is that coordination failure is structurally implicated in the operation of all four other modes. The synthesis chapter for coordination failure serves as the cross-cutting structural-architectural synthesis the other modes interact with at the architectural level.

The structural argument: mandate-bounded blindness operates on each institution's perceptual apparatus; coordination failure operates on the cross-institutional architecture that integrates those apparatuses. Where the cross-institutional architecture is design-principle-deficient (in Ostrom's sense), each component institution's mandate-bounded blindness compounds at the architectural level because no integrating venue performs the cross-mandate aggregation. The 2008 cross-institutional risk-aggregation gap is the canonical instance — each institution's mandate covered its segment, no institution's mandate covered the aggregate, the cross-institutional architecture provided no aggregating venue, and the aggregate risk grew structurally invisible.

Model selection bias operates within an institution's analytical apparatus; coordination failure operates on the cross-institutional architecture that selects for industry-standard model classes. Where the architecture is design-principle-deficient, individual-institution deviation from industry-standard model classes becomes professionally risky regardless of substantive merit. The pre-2008 trans-Atlantic VaR convergence and the pre-2003 trans-jurisdictional WMD assessment convergence are documentary instances of model selection bias operating jointly with coordination failure at industry-population scale.

Sunk-cost epistemology operates on within-institutional commitment; coordination failure operates on the cross-institutional architecture that creates the conditions under which sunk-cost can persist without cross-institutional challenge. Where the architecture lacks Ostrom design principle 3 (collective-choice arrangements) or design principle 6 (conflict-resolution mechanisms), within-institutional commitment to a prior framework can persist across institutional reform cycles. The 30-year deregulatory commitment in 2008 and the post-1991 IC analytical posture in Iraq WMD are documentary instances.

Audience-induced distortion operates on the institution-audience relationship; coordination failure operates on the cross-institutional architecture where multiple institutions' audience-aligned outputs become industry-standard. Where multiple institutions in different jurisdictions produce structurally similar audience-aligned outputs, the convergence is self-reinforcing across the institutional population and individual-institution deviation is professionally risky. The 2008 trans-Atlantic rating-agency architecture and the 2002–2003 trans-jurisdictional WMD assessments are documentary instances.

The mandate-bounded-blindness / coordination-failure isomorphism. The four cross-mode interactions above describe coordination failure's cross-cutting role at the level the typology's foundational framing established. The synthesis-chapter analysis supports a substantively stronger position: the four sub-patterns of mandate origination and the three-tier eight-design-principle classification are isomorphic — perceptual side and architectural side of paired institutional properties of cross-institutional architectures. The mapping:

The two synthesis chapters' principal additions are the same finding viewed from the two sides of the failure mode's mechanism: the perceptual side identifies what the institution's mandate excludes (perceptual blindness); the architectural side scores the resulting cross-institutional architecture's coordination-adequacy state. Mandate-bounded blindness and coordination failure operate as paired perceptual and architectural properties of cross-institutional architectures rather than as independent failure modes.

The isomorphism extends to sunk-cost epistemology through the boundary-condition specification. The four structural conditions (a)–(d) under which sunk-cost epistemology fails to operate as a stabilising counter-pressure are the same conditions that enable sub-pattern 4 mandate-rollback-by-choice and the tier-2-to-tier-3 architectural transition. The three synthesis chapters jointly operationalise the same institutional mechanism at three sides: the perceptual side (mandate exclusion); the architectural side (coordination-architecture adequacy); and the commitment-reinforcement side (the boundary-condition failure of sunk-cost to prevent the exclusion). The structural-insulation discipline — the fourth criterion added to the typology's prophylactic recommendations alongside the Hobbes, Olson, and Ostrom criteria — is the prophylactic counterpart to the three-way isomorphism: post-failure reform proposals must include structural features that prevent the trajectory through the failure modes.

The isomorphism finding strengthens the typology's general framework from "five modes analytically distinguishable but empirically interlocked at the architectural level" to a more specific position: five modes with two of them — mandate-bounded blindness and coordination failure — operating as paired perceptual and architectural properties of cross-institutional architectures, qualified by the sunk-cost-epistemology boundary-condition specification on the commitment-reinforcement side.

3.4 Boundaries that are real versus boundaries that are analytical conveniences

The honest section. Which of the five categories are clearly distinct phenomena, and which are convenient analytical separations of what may be a single underlying mechanism?

The five categories are not equally well-defined. Coordination failure is the most cross-cutting and could in some readings be the single underlying mechanism of which the other four modes are surface manifestations at different institutional scales. The typology rejects this strongest reading because the four other modes have analytically distinguishable diagnostic indicators that operate independently of the coordination-architecture dimension. But the strongest reading has substantive force, and the typology should not pretend the five categories are perfectly orthogonal.

The cleanest distinctions in the typology: mandate-bounded blindness and model selection bias are clearly different (perceptual-content vs analytical-framework); model selection bias and sunk-cost epistemology are clearly different (model-class selection vs commitment-reinforcement); audience-induced distortion and sunk-cost epistemology overlap through Janis but are clearly different (external-communication vs internal-commitment). The murky distinctions: coordination failure shades into mandate-bounded blindness through cross-mandate aggregation gaps; coordination failure shades into all four other modes. The typology's working position is that the murkiness is part of the analytical content — institutional failure is genuinely a multi-mechanism phenomenon, and the structural overlap between coordination failure and the other modes is documentary evidence of the architectural-perceptual interaction.

The typology's claim is therefore weaker than "five orthogonal categories" and stronger than "five-named-versions-of-one-phenomenon." The five modes are analytical positions on the institutional-failure phenomenon, each highlighting a structural mechanism that the others underweight. The cross-mode interactions are themselves findings of the typology, not weaknesses of it.

4. Methodological approach

This paper uses a qualitative typology-building method. It does not claim to test the five failure modes statistically, nor does it treat the five calibration cases as a representative sample of all institutional failures. The method is instead diagnostic and comparative: the paper identifies recurring structural mechanisms in the institutional-failure literature, specifies them as five analytically distinguishable modes, and then calibrates those modes against historically documented cases.

The typology is built from three kinds of material. The first is a bounded body of foundational literature drawn from organisational sociology, normal-accidents theory, public administration, intelligence studies, probability theory, and collective-action theory. These sources provide the conceptual mechanisms from which the five modes are derived. The second is a set of five calibration cases: Iran 1979, Challenger 1986, the 2008 financial crisis, Iraq WMD 2002–2003, and the 2023 regional-banking failures, with the 2023 case now supported by direct engagement with the Barr, FDIC, GAO, and Senate-testimony record. The third is a set of diagnostic indicators developed from the interaction between the conceptual mechanisms and the case evidence.

The cases are used for calibration rather than statistical proof. Each case is assigned one primary failure mode because that mode appears with unusual documentary clarity in that case. Iran 1979 is treated as the primary case for mandate-bounded blindness; Challenger 1986 for sunk-cost epistemology; the 2008 financial crisis for model selection bias; Iraq WMD 2002–2003 for audience-induced distortion; and the 2023 regional-banking failures for coordination failure. These assignments do not imply that the other modes are absent. On the contrary, each case contains multiple modes operating together. The primary-mode assignment is a discipline for exposition: it ensures that each mode is developed through at least one case where its mechanism can be seen with particular clarity.

Evidence is treated in three forms. First, the paper uses documentary evidence: official inquiries, postmortems, public reviews, testimony, and contemporaneous records. Second, it uses cross-case pattern evidence: recurring structural similarities across different institutional domains, such as intelligence analysis, aerospace engineering, and financial regulation. Third, it uses diagnostic evidence: observable indicators that a failure mode is operating, such as missing aggregation mandates, training data that excludes the application regime, repeated normalisation of deviance, or dissent that is preserved procedurally but excluded operationally.

The paper does not use the cases to infer frequency, probability, or causal weight in a statistical sense. It cannot show how often each mode occurs across all institutions, which mode is most common, or whether the five-mode typology is exhaustive. The claim is narrower: the five modes identify recurring structural mechanisms that are visible across several well-documented institutional failures and that can be used prospectively to diagnose elevated structural risk.

The method therefore sits between historical interpretation and practical diagnosis. It is historical because the modes are calibrated against documented institutional failures. It is diagnostic because the final purpose is not retrospective explanation alone, but prospective institutional assessment. The typology is intended to help analysts ask: what is this institution structurally unable to see, what model family is it committed to, what prior commitment has become hard to revise, what audience relationship may be distorting its outputs, and what coordination architecture is missing?

Two limitations follow directly from the method. First, because the source base and case set are bounded, the typology should be treated as provisional rather than exhaustive. Additional cases, especially from non-Western institutional settings, may require new modes or revisions to the existing ones. Second, because the method diagnoses elevated structural vulnerability rather than event causation, the typology cannot predict the timing, scale, or precise form of future failures. Its proper use is as an early-warning and institutional-audit framework, not as a forecasting model.

5. Calibration cases

The typology is calibrated against five documented case studies spanning four decades and three institutional domains (intelligence analysis; engineering; financial-system regulation). The five-case set achieves single-primary coverage of all five failure modes — each mode has a paradigmatic primary-mode case where the mode operates at sufficient documentary depth to support synthesis-chapter operationalisation. The summaries below state what each case demonstrates and which failure modes are most clearly exhibited.

The five-case set's primary-mode assignments are as follows: Iran 1979 is treated as the primary case for mandate-bounded blindness; Challenger 1986 as the primary case for sunk-cost epistemology; the 2008 financial crisis as the primary case for model selection bias; Iraq WMD 2002–2003 as the primary case for audience-induced distortion; and the 2023 regional-banking failures as the primary case for coordination failure. This single-primary coverage discipline is methodologically important: it ensures that each failure mode is developed through at least one case in which its mechanism can be examined at substantial documentary depth, while still recognising that each case contains secondary modes operating concurrently.

The five-case set also operationalises two multi-decade structural-persistence parallels that the three-case first-draft framing could not support: the Iran/Iraq 24-year structural-persistence parallel in intelligence-analysis (1979 → 2002–2003 documenting the institutional durability of the failure modes across the post-1979 reform cycles); the 2008/2023 15-year structural-persistence parallel in financial-system regulation (2008 → 2023 documenting the institutional durability of the failure modes across the post-2008 Dodd-Frank reforms). The two parallels jointly constitute the typology's strongest documentary evidence on the failure modes' institutional durability across reform cycles in two different institutional domains.

5.1 Iran 1979 (intelligence failure)

The 1979 Iranian Revolution intelligence failure — the US Intelligence Community's failure to anticipate the Shah's fall, despite a substantial collection apparatus operating in Iran throughout the 1970s — is the typology's paradigmatic mandate-bounded-blindness case. Jervis's Why Intelligence Fails (2010) Ch. 2 is the canonical academic reconstruction; the Jervis 1979 NFAC postmortem (commissioned by the CIA in 1979, classified, declassified and appended to Jervis 2010 at pp. 34–108) is the primary documentary source. The case is structurally distinctive in the typology's calibration set because Jervis is both the postmortem author (1979) and the methodology-theorist reflecting on the postmortem-as-artefact discipline 31 years later (2010), producing analytical depth no other postmortem in the typology engages.

Mandate-bounded blindness (primary, paradigmatic) operates through the collection-mandate exclusion of opposition forces, religious networks, and bazaar information sources — the data classes that, in retrospect, contained the decisive signals of regime instability. The collection-mandate exclusion is the structurally upstream condition: the analytical environment within which the "Shah will crack down" model-class commitment operated was produced by the data-class exclusion, not independent of it. The case operationalises mandate-bounded blindness on the §5 initial-mandate-narrow sub-pattern: the CIA's pre-1979 collection mandate originated excluding the structurally relevant data classes, following from the post-1953 institutional framework's emphasis on Shah-regime sustainment.

The case operates at all three structural levels of the mandate-bounded-blindness mechanism. Collection-mandate level: opposition / religious / bazaar contacts excluded. Analytical-mandate level: framing of Iran through the Soviet-paradigm-derived "monarch with security apparatus" categorisation, structurally mis-classifying the actually-operating religious-social-mobilisation dynamics. Institutional-mandate level: analytical-mandate exclusion of commentary on US policy toward Iran foreclosing the most consequential decision venue (the question of whether US policy itself was producing the instability the analysts were charged with assessing).

Model selection bias (secondary) operates through the "Shah will crack down" model-class commitment, structurally disconfirmable only by an event severe enough that response would be too late. Audience-induced distortion (secondary) operates through analyst-self-internalisation of the no-commentary-on-policy norm, exhibiting the structural-rather-than-coercive mechanism Jervis later documents on Iraq WMD. Coordination failure (secondary) operates through the cross-agency analytical architecture exhibiting the 2/3/3 eight-design-principle pattern, the same pattern documented in Challenger and Iraq WMD pre-IRTPA.

The case's distinctive contribution is the Iran/Iraq 24-year structural-persistence parallel. Four failure-mode components persist across the 24-year period from Iran 1979 to Iraq WMD 2002–2003: collection-mandate blindness; model-class disconfirmable-only-when-too-late commitment; audience-induced distortion through institutional-environment shaping without explicit pressure; coordination-failure 2/3/3 deficiency pattern. The post-1979 reform cycles (Casey 1981; subsequent 1980s-1990s reforms; post-9/11 IRTPA) addressed proximate causes without addressing structural conditions; the same structural failure modes produced the Iraq WMD case 24 years later. This is the typology's most-temporally-deep documentary evidence on the institutional durability of the failure modes across multiple intervening reform cycles in the intelligence-analysis domain.

The Jervis 1979 NFAC postmortem operates as the typology's clearest single instance of non-protective vocabulary operating from within an official postmortem authored by an institutional insider with internal access. Jervis 2010's framing chapter is the case's analogue to the Feynman appendix on Challenger — non-protective vocabulary from a credentialled investigator with no institutional position to protect, providing the post-event-vocabulary diagnostic's cross-check-against-contemporaneous-documentary-record at unusual depth.

5.2 Challenger (1986)

The 1986 Challenger disaster — loss of Space Shuttle Challenger on 28 January 1986 following O-ring failure in a solid rocket booster — is the typology's paradigmatic sunk-cost epistemology case. Vaughan's The Challenger Launch Decision (1996) is the canonical academic reconstruction; the Rogers Commission Report (1986) is the official postmortem. The case exhibits all five failure modes operating concurrently:

Sunk-cost epistemology (primary, paradigmatic) operates through the nine-year normalisation chronicle (1977–1985). Each new observation of O-ring damage was reviewed at the next Flight Readiness Review and classified within the acceptable-risk envelope established by the prior observation; the envelope expanded with each new instance rather than being revised. The institutional cost of revising the envelope downward — admitting that the prior FRR certifications had been wrong — became prohibitive. Jerry Mason's "take off your engineering hat and put on your management hat" instruction to Bob Lund on the eve-of-launch teleconference, who then reversed Thiokol's engineering recommendation for no-launch-below-53°F, is the canonical documentary evidence of how sunk-cost epistemology operates at the decision-point level.

Coordination failure (secondary) operates through the NASA/Thiokol contractor-prime architecture, which scored 2/3/3 against Ostrom's eight design principles (with principles 4–6 most consequentially absent). The Boisjoly foro-interno/foro-externo strong-absence — engineer with clear internal commitment to safety, no institutional foro-externo mechanism for the commitment to bind collective action — is the canonical institutional-whistleblowing diagnostic. The case operationalises the failure-case 2/3/3 tier of the three-tier eight-design-principle classification.

Audience-induced distortion (secondary) operates through the acceptable-risk framework as audience-facing vocabulary back-reacting on internal perception, with the Teacher-in-Space + State-of-the-Union scheduling as audience-pressure mechanism.

Mandate-bounded blindness (contributing) operates through the NASA Marshall mandate covering SRB development but not programmatic decisions about continued operational use, and through the contractor-prime knowledge-decision asymmetry. The case operates predominantly on the growth-of-failure-domain-beyond-mandate sub-pattern: novel-technology engineering beyond the qualified-envelope categorisation.

Model selection bias (contributing) operates through the Mediocristan-acceptable-risk framework applied to novel-technology Extremistan engineering; the Feynman 1/100,000 vs 1/100 estimated-failure-probability discussion in Rogers Commission Appendix F is the cleanest single articulation.

The Columbia disaster in 2003 within the same institutional architecture is the typology's first forward-looking-calibration evidence: the post-Challenger reforms addressed the proximate engineering cause without adequately addressing the structural failure modes, and 17 years later the same structural pattern produced the same outcome class.

5.3 2008 financial crisis

The 2008 financial crisis — the systemic failure of the trans-Atlantic financial system between mid-2007 and early 2009 — is the typology's paradigmatic model selection bias case. The Financial Crisis Inquiry Commission's Final Report (2011), with two formal dissents, is the canonical US official postmortem; Taleb's The Black Swan (2007) is the canonical pre-event theoretical framework; Tooze's Crashed (2018) is the canonical academic-historical trans-Atlantic-system framing. The case study's revised analytical position (2026-06-02) is that the appropriate analytical unit is the trans-Atlantic financial system as a single integrated architecture rather than the US system with international spillover. The standard FCIC-derived framing in which 2008 is "primary US case + international spillover" is structurally inadequate; European banks were direct participants in the US mortgage-securitisation chain (~33% of newly-issued 2006 private-label MBS backed by British or European banks; ~$715 billion in dollar-denominated ABCP sponsored by European institutions), and the eurozone 2010–2012 crisis is the direct continuation of 2008 rather than a separate event.

Model selection bias (primary, paradigmatic) operates through Mediocristan-family risk-modelling tools deployed across an Extremistan regime. The Moody's correlation model is the canonical documentary instance — the model was empirically calibrated on a 10-year training window of rising housing prices that systematically excluded the regime under which the model would be applied; Gary Witt's FCIC testimony that the analysts "made them up" and "weren't looking" is the institutional admission of the failure mechanism captured under oath. The 83% downgrade rate on 2006-vintage triple-A mortgage securities is the canonical numerical evidence. VaR-based risk management at the major investment banks operated through structurally similar Mediocristan-Gaussian commitments. Tooze documents the mode operating at a second institutional scale: the pre-crisis academic-and-forecasting profession's commitment to the older national-balance-sheet ("island model") macroeconomic framework, with the Rubin-Orszag "twin deficits" disaster scenario (Hamilton Project 2004–2006) as the canonical pre-crisis Mediocristan-family-model output that proved empirically wrong about the structural risk class — the actual crisis was a dollar-funding shortage in European banks, the opposite directional dollar dynamic from the predicted one.

Audience-induced distortion (secondary) operates through the pay-issuer/rely-investor structural mechanism at the rating-agency architecture (fee per CDO $250,000 to $850,000; structural separation between the principal-paying issuer and the audience-relying-on investor; the investor's decision to use ratings as substitute for independent credit analysis). The mechanism operates structurally rather than coercively — the rating analysts came to believe their methodology's outputs. Tooze documents the mode operating at a second institutional scale: the post-2008 European political-discourse vocabulary (Steinbrück's "laissez-faire ideology" framing; Sarkozy's "laissez-faire is finished"; Tremonti's "Italy's banks did not speak English") exhibits the same consensus-appealing + responsibility-displacing diagnostic structure at cross-jurisdictional political-discourse scale, with the documentary evidence of European banks' direct participation as substantive falsification.

Coordination failure (secondary, dual-tier) operates through the CDO/CDS/SIV/repo cross-institutional production chain at the financial-institution-population scale. Each institution in the chain was performing its assigned task within its mandate. No institution was assessing AIG's aggregate $79 billion CDS-on-mortgage-securities exposure across all its counterparties simultaneously; the regulatory architecture had no institutional location at which such aggregate assessment was the assigned responsibility. The risk was not concealed; it was not gathered. At a different scale, the Federal Reserve swap-line architecture (cumulative gross drawings approaching $10 trillion to fourteen approved foreign central banks 2008–2010, with at least half going to non-US banks) operates as the typology's strongest positive-case 4/1/3 tier documentary exemplar of cross-jurisdictional coordination architecture in real-time crisis. The architecture's eight-design-principle scoring is 4 satisfied / 1 partial / 3 violated — the inverse pattern from the failure-cases' 2/3/3 with principles 4-5-6 most consequentially absent. The architecture is closer to Hobbesian-Leviathan than to Ostrom-polycentric: strong integrating-authority (Fed unilateral control); weak collective-choice / rights-to-organise dimensions. The architecture worked because of the Hobbesian-Leviathan structural strength at the trans-Atlantic scale in real-time crisis. The eurozone 2010–2012 contrast (where the ECB was institutionally constrained from Fed-equivalent operation) is documentary evidence of how much the integrating-authority's structural presence-vs-absence matters institutionally. The case is therefore the typology's only single-case instance operationalising two of the three tiers (failure-case + positive-case) of the eight-design-principle classification.

Mandate-bounded blindness (contributing) operates through the shadow-banking-outside-mandate pattern (the regulatory architecture organised around the post-1933 financial-institution categorisation while the actual risk-transmission architecture operated outside that categorisation) — the growth-of-failure-domain-beyond-mandate sub-pattern (sub-pattern 2) — and through the OCC/OTS turf war. Tooze adds the European-side documentary instance: the Brussels mandate-extension on local-state-guarantee removal for the German Landesbanken in the early 2000s produced unintended structural conditions for the Landesbanken's trans-Atlantic ABCP exposure (at least four exceeding equity capital several times over), which the EU regulatory architecture had no mandate-coverage to monitor. The pattern generalises: institutional reforms that extend mandates in one categorisation frequently produce unintended structural conditions outside the categorisation.

Sunk-cost epistemology (contributing) operates at two levels: the 30-year deregulatory commitment as policy-level sunk-cost; the model-class commitment to Mediocristan-family risk-modelling as analytical-level sunk-cost.

The FCIC Ch. 1 documentation of senior decision-makers' post-event vocabulary — Bernanke's "perfect storm"; Buffett's "mass delusion" shared by "300 million Americans"; Blankfein's "hurricane"; Prince's "wholly unanticipated"; Greenspan's "history tells us regulators cannot identify the timing of a crisis" — is the typology's clearest single instance of the post-event-vocabulary diagnostic operationalised at the within-institution scale, and provides empirical confirmation of Taleb's pre-event prediction that an institutional apparatus protecting itself from model-class revision would generate externalising-and-naturalising vocabulary post-event. The Tooze-documented European Schadenfreude pattern operationalises the same diagnostic structure at cross-jurisdictional political-discourse scale.

The 2023 regional-banking failures (Silicon Valley Bank, Signature Bank, First Republic) are now engaged as the typology's fifth calibration case (§5.5 below) rather than only as forward-looking-calibration evidence within this case. The 2008/2023 15-year structural-persistence parallel that the two cases jointly operationalise is documented in §5.5.

5.4 Iraq WMD intelligence failure (2002–2003)

The Iraq WMD intelligence failure — the US and UK intelligence communities' production of pre-war assessments concluding that Iraq possessed or was actively developing weapons of mass destruction programmes, subsequently shown by the Iraq Survey Group to be empirically wrong on virtually every important pre-war judgment — is the typology's paradigmatic audience-induced distortion case. Jervis's Why Intelligence Fails (2010) Ch. 3 is the canonical academic reconstruction; the Butler Review (UK, 2004) and the Silberman-Robb Report (US, 2005) are the parallel official postmortems. The case exhibits substantial coordination-failure mechanism (through the cross-agency NIE production and the "Community in name only" finding) but is assigned audience-induced distortion as its primary mode to support the typology's single-primary coverage of all five failure modes, with the 2023 regional-banking case taking coordination failure as its primary-mode assignment.

Audience-induced distortion (primary, paradigmatic) operates through Jervis's motivated-bias-under-political-environment mechanism. Jervis's empirical rejection of simple politicization on five grounds — other countries' parallel assessments; IC resistance on other Iraq questions; analyst genuine surprise post-invasion; the administration's documented cherry-picking-and-stovepiping being itself evidence against successful coercive politicization; the structural errors pre-dating the most intense political-environment pressure — establishes the structural-rather-than-coercive framing as the case's analytical centrepiece. The mechanism operates through compressed-timeline NIE production, post-1991 career-incentive asymmetry, source-evaluation distortion (Curveball; INC defectors), and structural absorption of INR/DOE dissent.

Coordination failure (secondary, formerly dual-primary) operates through the IC's pre-IRTPA cross-agency architecture, which scored 2/3/3 against Ostrom's eight design principles (the failure-case 2/3/3 tier of the three-tier classification — the same deficiency pattern as the Challenger NASA/Thiokol architecture, in a completely different institutional domain). Silberman-Robb's "15 intelligence organizations are a 'Community' in name only and rarely act with a unity of purpose" is the canonical official articulation. The INR aluminium-tubes dissenters operationalise the foro-interno/foro-externo spectrum partial-presence case — dissent preserved in documentation but structurally marginalised in operational interpretation — intermediate between Boisjoly's strong-absence and pure-foro-externo-presence.

Mandate-bounded blindness (secondary) operates through the collection-mandate exclusion of on-the-ground inspection access 1998–2002, the source-evaluation-mandate misapplication to the post-1998 émigré-defector target environment, and the analyst-training-mandate exclusion of structured-analytic-techniques discipline. The case operates predominantly on the reform-cycle-resetting-mandate-after-failure sub-pattern, compounding with sub-pattern 2: post-1991 we underestimated Saddam lesson reshaping the analytical-mandate categorisation; subsequent post-1998 target-environment evolution past the reshaped categorisation.

Sunk-cost epistemology (contributing) operates at dual levels: analytical sunk-cost in the post-1991 IC posture (the "we underestimated Saddam in 1991" institutional lesson shaping the analytical posture for the next decade); policy sunk-cost in the Bush administration's pre-war commitment.

Model selection bias (contributing) operates through the Saddam-is-consistent-and-developing-WMD model selected over the rejected Saddam-reduced-his-ambitions-in-the-1990s-under-sanctions-pressure alternative.

The dual-postmortem structure (Butler 2004 + Silberman-Robb 2005) is methodologically distinctive: both postmortems explicitly note that they were "not asked" to investigate the policymakers' use of intelligence, structurally excluding the political-environment-shaping mechanism the case-study identifies as primary. The "we were not asked" pattern operationalises the postmortem-as-artefact discipline at unusual documentary depth.

The case forms the back-half of the Iran/Iraq 24-year structural-persistence parallel with §5.1 — four failure-mode components persist across the 24-year period despite multiple intervening reform cycles in the intelligence-analysis domain.

5.5 2023 regional-banking failures

The 2023 regional-banking failures — the rapid collapse of Silicon Valley Bank (March 10), Signature Bank (March 12), and First Republic (May 1), with more than $40 billion withdrawn from SVB in a single day and the systemic-risk-exception bailout providing partial institutional response — are the typology's paradigmatic coordination failure case. Five primary sources constitute the documentary basis at primary-case-depth (engaged 2026-06-03): the Federal Reserve's Review of the Federal Reserve's Supervision and Regulation of Silicon Valley Bank (Barr Report, 28 April 2023; 102 pages); the FDIC's FDIC's Supervision of Signature Bank internal review (28 April 2023); GAO-23-106736 Preliminary Review of Agency Actions Related to March 2023 Bank Failures (April 2023); GAO-25-106771 BANK SUPERVISION: Federal Reserve and FDIC Should Address Weaknesses in Their Process for Escalating Supervisory Concerns (November 2024); Michael Barr's Senate Banking Committee testimony 28 March 2023. The case study engages these at primary-case-depth on the SVB + Signature dimensions; four residual depth-gaps remain at calibration-depth (First Republic; Yellen + Gruenberg vocabulary; EGRRCPA congressional legislative-history; eight-design-principle scoring at principles 1, 2, 3, 7, 8).

Coordination failure (primary, paradigmatic) operates through the post-Dodd-Frank + FSOC + SIFI architecture's structural inadequacy at mid-tier-bank scale following the 2018 Economic Growth, Regulatory Relief, and Consumer Protection Act (EGRRCPA) SIFI threshold elevation from $50B to $250B in consolidated assets. The case is the paradigmatic instance of the partial-failure-case 2/4/2 tier of the three-tier eight-design-principle classification. The 2/4/2 pattern is structurally intermediate between the failure-case 2/3/3 and the positive-case 4/1/3: the 2018 EGRRCPA narrowed the post-2008 architecture's coverage population while preserving the architecture's structural-design features on the remaining SIFI population. Primary-source documentary basis (Barr Report; Barr Senate testimony 28 March 2023): SVB had 31 open supervisory findings at the time of failure (~3× peer-bank average); the Memorandum of Understanding enforcement action initiated August 2022 took more than 7 months to develop and was never delivered; an Internal Liquidity Stress Test shortfall "should have led to an MRIA" but was mischaracterised as "operational" (per the Barr Report, a violation of Regulation YY). The FDIC parallel at Signature: SBNY's Composite CAMELS rating was '2' (satisfactory) from 2017 through 10 March 2023, with direct downgrade to '5' on 11 March; the FDIC's own review states it "would have been prudent to downgrade the Management component rating to 3" as early as the second half of 2021. The architectural-capacity dimension is documented by the FDIC's chronic LFI examiner staffing shortage in the New York Regional Office averaging 40% vacant or filled by temporary staff since 2020, characterised by the FDIC itself as "a mission-critical risk that will require a sustained whole-of-agency response." Principles 1 (clearly defined boundaries — partial through the 2018-revised SIFI threshold) + 4 (monitoring — partially preserved through CCAR / DFAST / FSOC reporting at the residual SIFI population, but lost on the mid-tier population) + 5 (graduated sanctions — structurally weakened by the 2018 mandate-rollback removing intermediate-strength tools at the mid-tier scale) + 6 (conflict-resolution — partially preserved through FSOC referral mechanisms, but escalation-velocity inadequate to digital-run conditions) are all in the partial column rather than the violated column. The case is therefore the typology's first documentary instance of the post-reform-then-rolled-back coordination-failure trajectory — the cross-institutional aggregation gap was re-created by political-coalition mandate-rollback after it had been closed by prior reform.

Mandate-bounded blindness (secondary) operates through the typology's fourth sub-pattern: mandate-rollback-by-choice (sub-pattern 4 of §5). The 2018 EGRRCPA elevation of the SIFI threshold — moving Silicon Valley Bank, Signature Bank, and First Republic out of heightened-prudential-standards coverage before they reached their pre-2023 risk-condition — is the paradigmatic documentary instance of active deliberate mandate-rollback in the absence of structural conditions requiring the narrowing. The Barr Report's fourth key takeaway documents this at primary-source-depth: "The Board's tailoring approach in response to EGRRCPA and a shift in the stance of supervisory policy impeded effective supervision by reducing standards, increasing complexity, and promoting a less assertive supervisory approach. For Silicon Valley Bank, this resulted in lower supervisory and regulatory requirements, including lower capital and liquidity requirements." The specific Category IV (2019 tailoring rule) prudential-standard reductions applicable to SVB are enumerable: Liquidity Coverage Ratio removed; Net Stable Funding Ratio removed; Supplementary Leverage Ratio not applicable; AOCI capital treatment exemption (so unrealised securities losses did not flow through to capital); no Fed-level resolution plan requirement; stress testing deferred until 2024. The Barr Report's HQLA-shortfall calculation makes the rollback's consequence quantitatively specific: "approximately 9 percent shortfall of HQLA in December 2022 and approximately 17 percent in February 2023" (Barr Report lines 901-903, p. 13). The Fed's own institutional self-assessment is corroborated by the FRBSF's August 2021 scoping memorandum, which self-attributes to "the tailoring of enhanced prudential standards that resulted in less stringent regulation for Regional Banking Organizations" (per GAO-23-106736 p. 20). The Barr Report explicitly identifies the 2018 mandate-rollback as one of four key structural conditions producing the SVB failure ("the Federal Reserve's tailoring approach in response to the 2018 EGRRCPA impeded effective supervision"). The fourth sub-pattern is distinct from the other three (initial-mandate-narrow; growth-of-failure-domain-beyond-mandate; reform-cycle-resetting-mandate-after-failure) because the mandate exclusion arises from deliberate institutional choice to narrow, not from initial-narrowness, domain-evolution-overrun, or reform-cycle-resetting. The case operationalises the mandate-rollback proximity indicator at synthesis-chapter depth.

Model selection bias (secondary) operates through the HTM-accounting framework misclassifying interest-rate-risk-bearing securities on banks' balance sheets, and through the interest-rate-risk modelling at SVB structurally analogous to VaR + securitised-mortgage modelling pre-2008.

Audience-induced distortion (contributing) operates through the political-economy pressure for the 2018 deregulation and through the Fed supervisory culture of moderation under industry-influenced political environment.

Sunk-cost epistemology (contributing, with critical qualifying finding) operates through the post-2008 architectural commitment to the SIFI threshold framework that 2018 partially abandoned. The case-study analysis documents the sunk-cost-epistemology-failed-to-operate finding as the typology's most analytically distinctive 2023-case contribution: the post-2008 architecture's 10-year accumulated institutional investment with substantial sunk-cost commitment did not durably resist political-economy counter-pressure; the 2018 EGRRCPA passed with bipartisan support; the standard sunk-cost-stabilises-the-mandate prediction was empirically falsified under the structural conditions present. The finding bounds the operational scope of the sunk-cost epistemology failure mode and motivates the structural-insulation discipline added to the typology's prophylactic recommendations.

The case is the back-half of the 2008/2023 15-year structural-persistence parallel with §5.3 — four failure-mode components persist across the 15-year period despite the Dodd-Frank reforms (model selection bias; coordination failure with the three-tier 2/3/3 + 4/1/3 + 2/4/2 classification across the institutional history; mandate-bounded blindness with the new mandate-rollback-by-choice fourth sub-pattern; audience-induced distortion). The 2008/2023 parallel is structurally distinctive as the only documented instance of post-reform-then-rolled-back trajectory in the typology's calibration set; the 2008/2023 financial-system institutional history uniquely operationalises all three tiers of the eight-design-principle classification in a single documentary trajectory (1980s-2000s pre-FSOC failure-case 2/3/3 → 2010-2014 post-Dodd-Frank peak-implementation positive-case 4/1/3 → 2018 EGRRCPA → 2023 partial-failure-case 2/4/2).

The Barr Report's explicit identification of the 2018 EGRRCPA as a structural condition the Fed should have resisted more actively is the typology's clearest single instance of an official self-investigating-agency postmortem operationalising the postmortem-as-artefact discipline's fifth component (statutory-mandate jurisdictional-scope check) from within the chartering institution. The Barr Report pairs with the Jervis 1979 NFAC postmortem (§5.1) and the Feynman appendix to the Rogers Commission Report (§5.2) as the typology's three principal documentary instances of credentialled-from-within non-protective postmortem vocabulary.

Residual depth-gaps acknowledged. Four dimensions of the 2023 case remain at calibration-depth pending further engagement: the First Republic Bank failure dimension (the primary documentary record on the 1 May 2023 JPMorgan acquisition has not been engaged at the depth applied to SVB and Signature); the Yellen and Gruenberg senior-decision-maker vocabulary at the congressional hearings; the 2018 EGRRCPA congressional legislative history including the named cross-party coalition and the trade-association lobbying record; and the eight-design-principle scoring at principles 1, 2, 3, 7, and 8 (the primary record engaged here speaks directly to principles 4, 5, and 6 — monitoring, graduated sanctions, conflict-resolution mechanisms — but not to the other five at the same depth). These gaps are not load-bearing for the case's role in the typology but should be closed in future revisions.

5.6 Cross-case findings

The five-case calibration set jointly supports five cross-case findings that the typology's analytical apparatus rests on:

(i) The cross-case convergence on Ostrom's eight-design-principle deficiency pattern 4-5-6. Four of the five cases (Iran 1979; Challenger; Iraq WMD pre-IRTPA; 2008 pre-FSOC) score 2/3/3 against the eight design principles, with principles 4 (monitoring), 5 (graduated sanctions), and 6 (conflict-resolution mechanisms) most consequentially absent. The four cases operate in three different institutional domains (intelligence analysis; engineering; financial-system regulation) under different statutory frameworks with different participant populations on different temporal scales. The cross-case structural similarity is documentary evidence that the deficiency pattern is structurally common across coordination architectures and that Ostrom's design principles operationalise as a general (not domain-specific) diagnostic checklist. The three-case first-draft framing supported a two-case version of this finding; the five-case framing extends to four cases.

(ii) The three-tier eight-design-principle classification. The five-case set documents three tiers of the eight-design-principle classification: failure-case 2/3/3 (four cases — Iran 1979; Challenger; Iraq WMD pre-IRTPA; 2008 pre-FSOC); positive-case 4/1/3 (one case — Fed swap-line architecture operationalised within the 2008 case); partial-failure-case 2/4/2 (one case — 2023 regional-banking). The classification reveals that an architecture's eight-design-principle score is not a static institutional property but a trajectory-dependent property — reforms can move the score toward the positive-case pattern; subsequent political-coalition narrowing can move the score back toward the partial-failure-case pattern. The principles 4-5-6 cluster behaves as the principal architectural variable across the tiers (violated in tier 1; partial in tier 3; partial-or-satisfied in tier 2).

(iii) The four sub-patterns / three-tier classification isomorphism. The four sub-patterns of mandate origination and the three-tier classification are isomorphic — the same finding viewed from the perceptual side (component-institution mandates) and the architectural side (cross-institutional coordination architecture). Sub-patterns 1, 2 → tier 1; sub-pattern 3 → tier 2; sub-pattern 4 → tier 3. This is a substantively stronger claim than the typology's prior cross-mode-structural-overlap position: mandate-bounded blindness and coordination failure operate as paired perceptual + architectural properties of cross-institutional institutional architectures rather than as independent failure modes. The five-case set is the documentary base for the isomorphism finding; the three-case first-draft framing could not support it.

(iv) The two multi-decade structural-persistence parallels. The Iran/Iraq 24-year parallel (in intelligence-analysis) and the 2008/2023 15-year parallel (in financial-system regulation) jointly support the institutional-durability-across-reform-cycles finding in two different institutional domains. Both parallels operate at the four-failure-mode level (mandate-bounded blindness; model selection bias; audience-induced distortion; coordination failure) and document that post-failure reform cycles addressing proximate causes have not durably addressed the structural conditions. The two parallels are the typology's strongest documentary evidence for the institutional-durability claim.

(v) The sunk-cost-epistemology-failed-to-operate finding (2023 case). The standard prediction that institutional sunk-cost commitments durably stabilise post-failure coordination architectures against political-economy counter-pressure is empirically falsified under the structural conditions specified in §6 sub-pattern 4 (a)-(d). The post-2008 SIFI/FSOC framework's 10-year accumulated commitment did not prevent the 2018 EGRRCPA rollback. The finding bounds the operational scope of the sunk-cost epistemology failure mode and motivates the structural-insulation discipline added to the typology's prophylactic recommendations: post-failure coordination reforms must include structural features that insulate the architecture from subsequent political-coalition narrowing, in addition to addressing the proximate eight-design-principle deficiencies.

The five findings together support the typology's claim that the five failure modes are analytically distinguishable but empirically interlocked, and that the cross-case interaction structure is itself part of what the typology is for. The findings could not be operationalised at synthesis-chapter depth without the five-case calibration set; the 2026-06-02 calibration-set extension was the methodologically-required base for the principal 2026-06-02 analytical contributions.

6. Diagnostic application

The typology is intended as a diagnostic tool, not only a descriptive one. This section specifies how the typology can be applied to institutions whose failure has not yet occurred but whose structural conditions match those identified in the failure-mode mechanisms.

6.1 How to use the typology prospectively

The prospective application proceeds in four stages:

Stage 1 — Structural-conditions audit. For the institution(s) under analysis, identify which of the five failure-mode structural conditions are present. Each failure mode has named structural conditions (necessary conditions, sufficient conditions, contributing factors) specified in the relevant synthesis file. The audit is not a search for evidence that a specific failure is imminent; it is a check on whether the structural preconditions for any of the five modes are operating.

Stage 2 — Diagnostic-indicators application. For each failure mode whose structural conditions are present, apply the early-stage diagnostic indicators. The indicators are summarised in §§2.1–2.5 above and tabulated in Appendix B. The application is documentary: examine the institution's mandate categorisations, its analytical apparatus, its commitment history, its audience relationships, and its cross-institutional coordination architecture against the diagnostic checklists.

Stage 3 — Cross-cutting methodology disciplines. The companion methodology document specifies four cross-cutting disciplines that should be applied to any case-study analysis: the postmortem-as-artefact discipline (engaging any prior official postmortem on the institution as an artefact subject to the failure modes); the global-comparison discipline (confronting the analysis with parallel international cases); the post-event-vocabulary diagnostic (cataloguing and classifying senior-actor vocabulary in available documentary record); and the Hobbes-Olson-Ostrom triad analysis (scoring the institution's cross-institutional architecture against the eight design principles). The disciplines pair to produce a structured analytical output.

Stage 4 — Forward-looking calibration. Where the analysis predicts continued risk despite previously-implemented reforms, the prediction itself is documentary evidence of which structural conditions the reforms did not adequately address. The Challenger/Columbia 17-year parallel, the post-2008/2023-regional-banking parallel, and the post-9/11-DNI/subsequent-IC-failures parallel together demonstrate that institutional reforms targeting the proximate cause without addressing the structural failure modes are predictably inadequate. The forward-looking calibration is the typology's discipline for distinguishing reforms that address structural conditions from reforms that address proximate causes.

The output of the four-stage analysis is not a probability estimate of failure. It is a structured characterisation of which structural failure modes are operating in the institution under analysis, which diagnostic indicators are documentarily present, and what institutional-reform directions would address the structural conditions. The structured characterisation is what permits prospective intervention.

6.2 What the typology cannot do

The typology is bounded by three explicit non-claims that practitioners applying it should respect.

It does not predict specific failure events. Identifying structural conditions for a failure mode is not the same as identifying when, how, or with what magnitude the failure will materialise. Multiple institutions exhibit the structural conditions for one or more failure modes at any given time; only some of those institutions experience visible failure events; the structural-conditions analysis cannot specify which. The Olson-Casanova mechanism Taleb identifies — institutions that have survived past crises generate self-narratives of robustness that re-deploy the alleged robustness under untested conditions — is the typology's structural reading of why specific-event prediction is institutionally inaccessible.

It does not adjudicate moral responsibility. The typology's structural framing reads failure as a property of how institutions are constituted, not as a property of the moral character or competence of their members. The Vaughan reading of Challenger as no-intentional-managerial-wrongdoing-no-rule-violations-no-conspiracy and the Jervis reading of the Iraq WMD failure as structural-rather-than-coercive are both consistent with this commitment. The typology's case-study analyses are correspondingly cautious about attributing failure to named individuals. Where moral responsibility is the relevant analytical question, the typology is not the appropriate tool.

It does not assess institutional value. The typology identifies failure modes; it does not identify what institutions do well. An institution can exhibit the structural conditions for multiple failure modes and still produce substantial institutional value across its broader operations. The post-failure literature is at risk of confusing "this institution exhibited structural conditions for failure mode X" with "this institution is institutionally inadequate." The typology rejects the confusion. Institutional adequacy is a holistic question the typology does not answer.

6.3 Connection to compound cascade analysis

The Compound Cascade Systems Modelling Framework (Kelly, SSRN, 2026) analyses how cascades propagate once underway — the structural propagation paths through coupling dynamics, the containment-design implications, the institutional architectures required to interrupt propagation. This typology analyses the perceptual mechanisms by which the structural conditions for cascade accumulate without being perceived.

The two documents are complementary in three ways:

First, temporal complementarity. The typology operates upstream of the framework's analytical apparatus. Cascade preconditions accumulate over time; the typology identifies the perceptual failures that prevent institutions from acting on the accumulating conditions; the framework analyses how the cascade propagates once the preconditions become consequential. The Challenger nine-year incubation, the 2008 30-year incubation, and the Iraq WMD ten-year-plus analytical-posture incubation are all temporal-complementarity instances — the typology's incubation-period analyses provide the perceptual context within which the framework's propagation analyses operate.

Second, structural complementarity. The framework's coupling-dynamics analysis is symmetric with the typology's coordination-failure analysis. Tight coupling makes cascade propagation more catastrophic; coordination failure prevents the institutional architecture from detecting or interrupting the propagation. The two analyses can be applied to the same architecture from different angles, with the framework analysing the propagation paths and the typology analysing the perceptual-coordination failures that the architecture exhibits.

Third, prescriptive complementarity. The framework's containment-design recommendations specify what institutional architecture would interrupt cascade propagation. The typology's prophylactic recommendations specify what institutional architecture would address the perceptual-failure modes. Both sets of recommendations operate on cross-institutional coordination architecture; both engage the Hobbes-Olson-Ostrom triad's structural-diagnostic framing; both engage the second-order-dilemma question of who supplies the institutional infrastructure the recommendations require. The two prescriptive systems together specify a more complete institutional-reform agenda than either alone.

Neither document replaces the other. A cascade is more dangerous when an institution cannot perceive it; an institution that perceives cascade risk still requires the framework's analytical apparatus to evaluate how the cascade would propagate if triggered. The joint application of the two documents produces analytical purchase on cascade-vulnerability that neither alone provides.

7. Limitations and open questions

The typology has substantive limitations that practitioners applying it should respect. The methodology document treats the limitations in full; the headline limitations belong here.

Western institutional bias in the source selection. The reading list is overwhelmingly Anglo-American. Three of the four foundational organisational-sociology sources (Vaughan, Perrow, Sagan) are US-academic; Reason is UK-academic. The five-case calibration set is US-engineering (Challenger), US-and-trans-Atlantic-financial (2008; 2023), US-intelligence (Iran 1979) and US-and-UK intelligence (Iraq WMD). The typology's findings may not generalise to non-Western institutional traditions; cross-cultural application requires further calibration work the typology does not provide.

Single-author analytical position. The typology is the work of one researcher with documented analytical and political commitments. The structural framing the typology adopts is contested in the broader institutional-failure literature; alternative framings (more focus on individual moral responsibility; more focus on macro-economic causes; more focus on technological-determinism arguments) produce different typologies of the same calibration cases. The typology should be evaluated alongside, not as a substitute for, the broader literature.

Reform-prescription validation gap. The typology's prophylactic recommendations have not been empirically tested. The post-failure reforms in the calibration cases (the Casey reforms and post-9/11 IRTPA after Iran 1979; NASA reorganisation after Challenger; Dodd-Frank and FSOC after 2008; IRTPA and DNI after Iraq WMD) are partial responses; the typology's evaluation of which structural conditions they did not adequately address is an analytical claim, not an empirical demonstration. The 2008/2023 sunk-cost-epistemology-failed-to-operate finding qualifies the validation gap further: it documents an instance in which post-failure reform was implemented and was subsequently subject to political-coalition narrowing, suggesting the typology's prescriptive recommendations require structural-insulation features in addition to addressing the proximate failure-mode deficiencies. Whether the prescriptive recommendations with the structural-insulation discipline would prevent future cases is an open question the documentary record does not permit answering directly.

Recency commitment. The reading list is heavily weighted toward foundational texts. The newest source (Taleb 2010 postscript) was published 16 years before this draft. Several recent literatures (resilience engineering; post-2008 organisational-learning literature; post-COVID institutional-response analysis; the experimental-economics literature on conditional cooperation) are not engaged here. The typology may therefore miss recent empirical and theoretical work that would refine its findings. The methodology document defends this commitment as the cost of bounded sourcing.

The mandate-categorisation-lag threshold question is unspecified. The typology asserts that mandate-categorisation frameworks more than ~15-20 years old in rapidly-evolving domains are at elevated risk, but the threshold lacks empirical specification. The 30-year deregulatory commitment in 2008 is one data point; the post-1933 financial-institution categorisation operating in 2007 is another; the threshold likely varies by domain rate-of-change but the typology does not specify how.

The simple-vs-subtle politicization distinction is operationally underdeveloped. The typology adopts Jervis's framing on Iraq WMD — rejecting simple politicization while accepting subtle politicization as structurally real — but the empirical diagnostic for distinguishing the two in other cases is not yet fully specified. The distinction is analytically critical and operationally subtle.

The polycentric-framework scaling question is unresolved. The Ostrom-derived polycentric self-governance framework is empirically supported in small-to-intermediate-scale common-pool-resource cases; whether it generalises to financial-system-scale coordination is contested. The typology's working position is that it applies at the regulatory-coordination level (FSOC) but not at the financial-institution-population level (which is Olson-latent and requires centralised oversight). The position is provisional.

Forward-looking calibration data is partial. The typology's forward-looking-calibration evidence — operationalised through the two multi-decade structural-persistence parallels (Iran/Iraq 24-year; 2008/2023 15-year) plus the Columbia 2003 within-institutional-architecture parallel plus subsequent IC failures — supports the structural-persistence-across-reform thesis but does not provide systematic post-reform empirical assessment across the broader population of post-failure institutional-reform histories. The companion methodology document's forward-looking-calibration discipline reserves Boeing 737 MAX and COVID-19 institutional response as additional prospective tests. A dedicated forward-looking-calibration paper engaging the broader post-failure institutional-reform records would strengthen the typology's prescriptive recommendations.

Single-primary-mode-assignment contestability. Three of the five primary-mode assignments are strongly contested: Iran 1979's mandate-bounded-blindness primary (contested with model-selection-bias-primary); 2023 regional-banking's coordination-failure primary (contested with mandate-bounded-blindness-primary and model-selection-bias-primary); and the narrowing of Iraq WMD from dual-primary to single-primary audience-induced-distortion. The typology takes positions on stated grounds and records the alternative readings, treating the contestability as a finding about the failure modes' empirical interlock at the architectural level (per the isomorphism finding articulated in §3.3) rather than as a methodology defect. Primary-mode assignment is analytical-emphasis selection operating on cases that exhibit multiple modes operating together, not empirical-categorisation operating on cases that exhibit single modes operating in isolation.

7.1 High-priority open questions

The typology displays the open questions it has not yet resolved rather than hiding them. A high-priority subset is reproduced here.

Foundational positioning. The typology's relationship to existing institutional-failure typologies — Reason's Swiss-cheese model; Allison-Zelikow's three conceptual lenses; Perrow's normal-accidents framework; Vaughan's normalisation-of-deviance framework — is developed in the project archive but admits multiple defensible readings. The typology takes positions on the Perrow/Vaughan dispute over Challenger (Vaughan covers the nine-year normalisation chronicle; the eve-of-launch teleconference is a power-override-within-normalisation episode rather than an alternative to it) and on the typology's relationship to Reason's framework (the typology extends Reason from operational-safety accidents to broader epistemic institutional failures). Both positions are contestable.

Dependent-variable commitment. The typology's dependent variable is the persistence of an inaccurate construction of risk, not the eventual catastrophe. This is a stronger commitment than the standard catastrophe-counting evaluation metric and has methodological implications the typology displays explicitly: evaluable against cases where no discrete event has materialised; not directly comparable to outcome-frequency-based institutional-failure analyses. The commitment is introduced in §1.3 and defended at greater depth in §4 (Methodological approach), but it remains a methodologically distinctive choice that external reviewers may contest.

Mode-overlap and the isomorphism finding. The five modes are analytically distinguishable but empirically interlocked (§3 above), with the mandate-bounded-blindness / coordination-failure isomorphism finding (§3.3) substantively strengthening the prior cross-mode-structural-overlap position. The remaining open question is whether the isomorphism extends symmetrically to the other three modes (model selection bias; sunk-cost epistemology; audience-induced distortion) or operates asymmetrically across the typology's modal structure. The project archive's analysis positions the sunk-cost epistemology mode as the commitment-reinforcement side of the same institutional mechanism the perceptual and architectural sides operate at; the model selection bias and audience-induced distortion modes' relationships to the isomorphism are less developed.

Wallison-dissent position on the 2008 case. The 2008 case's analytical position on the Wallison dissent to the FCIC majority report — that US government housing policy (specifically the HUD affordable-housing goals operating on the GSEs) was the sine qua non of the 2008 crisis — is provisional. The case study adopts an intermediate position rejecting the strongest version of the Wallison thesis on global-comparison grounds (per the Tooze trans-Atlantic-system framing) while accepting the GSE/HUD policies as a meaningful contributing cause. The position is defensible but admits alternative readings; the case study records the alternative readings explicitly.

Eurozone-as-extension-of-2008 provisional position. The case study adopts the position that the eurozone 2010–2012 crisis is the direct continuation of 2008 rather than a separate fourth case, following Tooze's framing. The position is provisional pending Tooze Parts II–IV engagement (currently only Introduction + Ch. 1 + Ch. 3 + Ch. 9 are engaged at the documentary-engagement depth the case study rests on). A reviewer would likely flag the limited Tooze engagement.

Structural-frequency questions for the sub-pattern 4 / partial-failure-case / boundary-condition trajectory. The paper's principal analytical contributions — the four sub-patterns of mandate origination; the three-tier eight-design-principle classification; the sunk-cost-epistemology boundary-condition specification — are documented across the five-case calibration set. The structural frequency of the all-four-conditions-present trajectory (sub-pattern 4 mandate-rollback-by-choice → tier 3 partial-failure-case → sunk-cost-epistemology boundary-condition failure) across the broader population of post-failure institutional reform histories is unspecified. The typology's strongest documentary instance — the 2008/2023 financial-system institutional history, with the 2023 case engaged at primary-case-depth on the SVB and Signature dimensions via the Barr Report, the FDIC Signature review, the GAO reports, and Barr's Senate testimony — supports the finding within its scope. The broader-population structural-frequency question remains open. Candidate cases for engagement include the post-Glass-Steagall 1999 GLBA repeal; the post-1986 Tax Reform Act subsequent erosions; and various post-environmental-regulation rollbacks. The active institutional inaction mechanism (alongside the active rollback mechanism) documented in the 2023 case — the GAO 2011 noncapital-PCA-triggers recommendation never implemented (12 years); the Dodd-Frank section 166 early-remediation rule never finalised (14 years overdue as of 2026) — extends the structural-frequency question to non-implementation as a parallel pathway to rollback.

8. Conclusion

The typology presented in this document identifies five recurring structural mechanisms by which institutions fail to perceive compound cascade risk. The five — mandate-bounded blindness, model selection bias, sunk-cost epistemology, audience-induced distortion, and coordination failure — are derived from seventeen foundational sources spanning organisational sociology, normal-accidents theory, public administration, intelligence studies, probability theory, and collective-action theory. They are calibrated against five documented case studies in three institutional domains spanning four decades: the 1979 Iranian Revolution intelligence failure; the 1986 Challenger disaster (engineering); the 2008 financial crisis (finance); the 2002–2003 Iraq WMD intelligence failure; and the 2023 regional-banking failures (finance). The five-case set achieves single-primary coverage of all five failure modes — Iran 1979 paradigmatically exhibits mandate-bounded blindness; Challenger paradigmatically exhibits sunk-cost epistemology; 2008 paradigmatically exhibits model selection bias; Iraq WMD paradigmatically exhibits audience-induced distortion; 2023 regional-banking paradigmatically exhibits coordination failure — while documenting that all five modes operate concurrently in each case.

Five cross-case findings carry the typology's principal analytical weight. The first is the convergence on Ostrom's eight-design-principle deficiency pattern 4-5-6: four of the five calibration cases (Iran 1979; Challenger; Iraq WMD pre-IRTPA; 2008 pre-FSOC) score 2/3/3 against the eight design principles, with monitoring, graduated sanctions, and conflict-resolution mechanisms most consequentially absent. Aerospace engineering, intelligence analysis, and financial-system regulation operate in different domains under different statutory frameworks; the shared deficiency pattern is documentary evidence that the principles operationalise as a general diagnostic checklist. The second is the three-tier extension of the eight-design-principle scoring — failure-case 2/3/3, positive-case 4/1/3 (the Federal Reserve swap-line architecture operationalised within the 2008 case), and partial-failure-case 2/4/2 (the 2023 case) — operationalising an architecture's coordination-adequacy state as a trajectory-dependent rather than static property. The third is the isomorphism between four sub-patterns of mandate origination and the three-tier classification: perceptual side and architectural side of paired institutional properties of cross-institutional architectures. The fourth is the pairing of two multi-decade structural-persistence parallels — Iran/Iraq 24-year and 2008/2023 15-year — jointly supporting the institutional-durability-across-reform-cycles finding in two domains. The fifth is the sunk-cost-epistemology-failed-to-operate finding from the 2023 case, which qualifies the standard prediction that institutional sunk-cost commitments durably stabilise post-failure coordination architectures and motivates the structural-insulation criterion in the typology's prophylactic recommendations.

The typology's general framework is that the five modes are analytically distinguishable but empirically interlocked. Each mode has its own characteristic mechanism, diagnostic indicators, and case-specific manifestation. The modes nonetheless co-occur in the calibration cases in patterns that reflect findings about the structure of institutional failure rather than category-construction errors. Coordination failure is structurally implicated in the operation of all four other modes; the synthesis chapter for coordination failure serves as the cross-cutting structural-architectural synthesis the other modes interact with at the architectural level. The mandate-bounded-blindness / coordination-failure isomorphism finding strengthens this position: the two modes operate as the paired perceptual + architectural properties of cross-institutional institutional architectures.

The typology's prospective diagnostic application proceeds in four stages: structural-conditions audit; diagnostic-indicators application; cross-cutting methodology disciplines (postmortem-as-artefact; global-comparison; post-event-vocabulary; Hobbes-Olson-Ostrom-triad); forward-looking calibration. The output is a structured characterisation of which structural failure modes are operating, which diagnostic indicators are documentarily present, and what institutional-reform directions would address the structural conditions — with the structural-insulation discipline added as a fourth criterion (alongside Hobbes / Olson / Ostrom) for evaluating post-failure reform proposals. The typology does not predict specific failure events, does not adjudicate moral responsibility, and does not assess institutional value beyond the failure-mode dimensions it specifies.

The typology is the perceptual-side companion to the Compound Cascade Systems Modelling Framework (Kelly, SSRN, 2026). The framework analyses how cascades propagate once underway; the typology analyses why institutions fail to see cascades coming. The two are positioned as a single diagnostic apparatus. The joint application produces analytical purchase on cascade-vulnerability that neither alone provides.

What remains: the typology's prophylactic recommendations have not been empirically tested; the recency commitment leaves recent empirical and theoretical literature unengaged; cross-cultural application requires further calibration; and a substantial open-questions catalogue remains — including the mandate-categorisation-lag threshold question, the simple-versus-subtle politicization diagnostic, the polycentric-framework scaling question at financial-system scale, and the three structural-frequency questions raised by the synthesis-chapter revisions (whether the four-conditions-present sub-pattern 4 / partial-failure-case 2/4/2 / boundary-condition failure trajectory documented at the 2023 case is structurally rare or structurally common across the broader population of post-failure institutional histories). The full catalogue, with a triage classifying each question by its bearing on publication-ready state, is maintained in the companion open-questions document. The typology is positioned as a working framework for prospective application, not as a closed analytical system. Future work should engage the structural-frequency questions across the broader institutional-history population; extend the Tooze engagement beyond the four chapters this draft engages; close the four residual depth-gaps on the 2023 case (First Republic; Yellen and Gruenberg testimony; the 2018 EGRRCPA congressional legislative history; the eight-design-principle scoring at principles 1, 2, 3, 7, 8); and test the prophylactic recommendations against the post-failure institutional-reform records of the documented cases.

References

The references below list the alphabetised foundational sources and case-study primary sources. The typology is supported by a project archive maintained by the author, comprising per-source notes for each foundational source, per-mode synthesis chapters, per-case study writeups, a companion methodology document, and a companion open-questions document with a publication-readiness triage. The archive is referenced where load-bearing for specific claims and is available from the author on request.

Foundational sources

• Allison, G. T. and Zelikow, P. Essence of Decision: Explaining the Cuban Missile Crisis. 2nd ed., Longman, 1999.
• Heuer, R. J. Psychology of Intelligence Analysis. CIA Center for the Study of Intelligence, 1999.
• Hobbes, T. Leviathan. Curley edition, Hackett, 1994 (orig. 1651).
• Janis, I. L. Groupthink: Psychological Studies of Policy Decisions and Fiascoes. 2nd ed., Houghton Mifflin, 1982.
• Jervis, R. Why Intelligence Fails: Lessons from the Iranian Revolution and the Iraq War. Cornell University Press, 2010.
• Kuhn, T. S. The Structure of Scientific Revolutions. 4th ed., University of Chicago Press, 2012 (orig. 1962).
• March, J. G. The Ambiguities of Experience. Cornell University Press, 2010.
• Olson, M. The Logic of Collective Action: Public Goods and the Theory of Groups. Harvard University Press, 1971 (orig. 1965).
• Ostrom, E. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press, 1990.
• Perrow, C. Normal Accidents: Living with High-Risk Technologies. Updated ed., Princeton University Press, 1999.
• Reason, J. Human Error. Cambridge University Press, 1990.
• Sagan, S. D. The Limits of Safety: Organizations, Accidents, and Nuclear Weapons. Princeton University Press, 1993.
• Simon, H. A. Administrative Behavior. 4th ed., Free Press, 1997 (orig. 1947).
• Taleb, N. N. The Black Swan: The Impact of the Highly Improbable. 2nd ed., Random House, 2010 (orig. 2007).
• Tetlock, P. E. Expert Political Judgment: How Good Is It? How Can We Know? New ed., Princeton University Press, 2017 (orig. 2005).
• Turner, B. A. Man-Made Disasters. 2nd ed. (with Pidgeon), Butterworth-Heinemann, 1997 (orig. 1978).
• Vaughan, D. The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. University of Chicago Press, 1996.
• Wilson, J. Q. Bureaucracy: What Government Agencies Do and Why They Do It. Basic Books, 1989.

Case-study primary sources

• Barr, M. S. (Vice Chair for Supervision). Review of the Federal Reserve's Supervision and Regulation of Silicon Valley Bank. Board of Governors of the Federal Reserve System, 28 April 2023.
• Butler of Brockwell, R. (chair). Review of Intelligence on Weapons of Mass Destruction: Report of a Committee of Privy Counsellors. HC 898, The Stationery Office, 2004.
• Federal Deposit Insurance Corporation. FDIC's Supervision of Signature Bank. 28 April 2023.
• Financial Crisis Inquiry Commission. The Financial Crisis Inquiry Report: Final Report of the National Commission on the Causes of the Financial and Economic Crisis in the United States. Official Government Edition, 2011.
• Government Accountability Office. Preliminary Review of Agency Actions Related to March 2023 Bank Failures. April 2023 and subsequent reports.
• Jervis, R. Analysis of NFAC's Performance on Iran's Domestic Crisis, mid-1977 – 7 November 1978. CIA, June 1979 (declassified; reprinted as Appendix to Jervis 2010).
• Robb, C. and Silberman, L. (co-chairs). Report of the Commission on the Intelligence Capabilities of the United States Regarding Weapons of Mass Destruction. 2005.
• Rogers Commission. Report of the Presidential Commission on the Space Shuttle Challenger Accident. 1986.
• Tooze, A. Crashed: How a Decade of Financial Crises Changed the World. Viking, 2018.

Companion paper

• Kelly, J. Compound Cascade Systems Modelling Framework. SSRN, 2026.

Sources identified for future engagement

The following sources are identified in the companion open-questions document as recommended for resolution of specific open questions before later publication revisions:

• Chilcot, J. (chair). The Report of the Iraq Inquiry. HC 264, 2016. (Identified for comparative analysis of the postmortem-instrument-itself-subject-to-the-mode finding.)
• Columbia Accident Investigation Board. Report of the Columbia Accident Investigation Board. 2003. (Identified for forward-looking-calibration extension on the post-Challenger NASA institutional architecture.)
• Tooze, A. Crashed, Parts II–IV. (This draft engages Introduction + Ch. 1 + Ch. 3 + Ch. 9; Parts II–IV identified for full resolution and the eurozone 2010–2012 continuation analysis.)

Appendix A: Additional methodological notes

The methodological approach is set out in §4 of this paper; the present appendix records headline commitments developed in the companion methodology document for readers who want the longer treatment.

The typology treats institutional failure as a structural phenomenon rather than as the failure of specific individuals (per §1.3 above). The dependent variable is the persistence of an inaccurate construction of risk, not the eventual catastrophe. The five failure modes are derived inductively from the seventeen-source foundational reading and calibrated deductively against the five calibration cases. Several literatures relevant to institutional failure are deliberately excluded — broader complexity-and-systems theory; cognitive-bias literature beyond Tetlock; Frankfurt-School ideology critique; resilience engineering; behavioural economics on conditional cooperation — with each exclusion defended individually in the methodology document. The reading list is heavily weighted toward foundational texts; the recency commitment and the forward-looking-calibration discipline that compensates for it are both defended in the methodology document.

The companion methodology document specifies 37 cross-cutting analytical commitments derived from the foundational sources. The most operationally important are: the retrospective-fallacy discipline (twelve sources converging); the silent-evidence discipline as cross-cutting commitment; the Mediocristan/Extremistan regime diagnostic; the consequence-before-probability discipline for prophylactic recommendations; the Casanova-style self-narrative diagnostic; the postmortem-as-artefact discipline; the global-comparison discipline; the post-event-vocabulary diagnostic; and the Hobbes-Olson-Ostrom triad for coordination-failure analysis.

The double-duty problem — Janis on groupthink informing both sunk-cost epistemology and audience-induced distortion — is treated as a finding about mechanism shared across modes rather than as a defect of the categorisation. The position is defended in §3.1 above.

The typology's methodological limitations (§7 above) include four explicit limits: Western institutional bias; single-author analytical position; reform-prescription validation gap; recency commitment.

Appendix B: Diagnostic indicators by failure mode

The typology's diagnostic indicators are developed in the per-mode synthesis chapters of the project archive and are summarised here for tabular reference.

Mandate-bounded blindness

Sub-patterns of mandate origination: initial-mandate-narrow (paradigmatic in Iran 1979); growth-of-failure-domain-beyond-mandate (paradigmatic in 2008); reform-cycle-resetting-mandate-after-failure (paradigmatic in Iraq WMD and the post-2008 Dodd-Frank framework); mandate-rollback-by-choice (paradigmatic in the 2018 EGRRCPA / 2023 regional-banking trajectory — the typology's first documented instance of sunk-cost epistemology failing to operate as a stabilising counter-pressure).

Model selection bias

Sunk-cost epistemology

Audience-induced distortion

Coordination failure

Three-tier classification: failure-case 2/3/3 (Challenger, Iraq WMD pre-IRTPA, 2008 pre-FSOC, Iran 1979 — principles 4-5-6 violated); positive-case 4/1/3 (Federal Reserve swap-line architecture 2008–2010 — principles 1, 2, 5, 8 satisfied; principles 3, 6, 7 violated; principle 4 partial); partial-failure-case 2/4/2 (2023 regional-banking — principles 4-5-6 partial, intermediate between violated and satisfied). The financial-system institutional history uniquely operationalises all three tiers in a single documentary trajectory: pre-FSOC tier 1 → post-Dodd-Frank tier 2 → post-EGRRCPA tier 3. The principles 4-5-6 cluster is the principal architectural variable across the tiers. Structural-insulation discipline: post-failure reform proposals must include structural features (statutory independence; external-mandate review; structural redundancy; statutory friction) that insulate the architecture from subsequent political-coalition narrowing — without structural insulation, the post-reform architecture's tier-2 positive-case score is not durable.