{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp11_mechanism_context_external20_20260804","research_id":"eoa_inverse_innovation_exp11_external_scrutiny_20260804","cell_id":"invariant_mode_decomposition_design__economics_finance","opaque_id":"invariant_mode_decomposition_design__economics_finance__A","search_lanes":{"direct_problem":{"queries":["bank balance sheet eigenvalue unstable mode liquidity leverage dynamic stress testing","feedback liquidity solvency bank stress testing fire sale management actions control model","official report bank failure capital liquidity ratios above minimum duration funding risk 2023"],"source_ids":["SRC1","SRC2","SRC3","SRC6","SRC7"],"no_result_note":null},"closest_prior_art":{"queries":["dynamic mode decomposition bank balance sheet financial stability stress test","Koopman operator financial systemic risk bank stress eigenvalues","dynamic balance sheet stress model deleveraging feedback management actions"],"source_ids":["SRC3","SRC4","SRC5","SRC7","SRC8"],"no_result_note":null},"historical_terminology":{"queries":["Liquidity Stress-Tester bank feedback effects 2010","older terminology integrated solvency liquidity stress test fire sale spiral","2012 bank enterprise-wide stress testing capital liquidity interaction"],"source_ids":["SRC2","SRC3","SRC7","SRC8"],"no_result_note":null},"products_practices_standards":{"queries":["ECB BEAST top-down workhorse dynamic balance sheet stress model","Basel stress testing principles liquidity capital interaction multiple risks","bank asset liability management software dynamic balance sheet stress testing"],"source_ids":["SRC2","SRC4"],"no_result_note":null},"non_english_regional":{"queries":["site:bde.es prueba de resistencia conjunta solvencia liquidez bancos ventas forzadas","site:banque-france.fr test de résistance solvabilité liquidité bilan bancaire dynamique","site:bundesbank.de Stresstest Solvenz Liquidität Bilanz Rückkopplung Banken"],"source_ids":["SRC6"],"no_result_note":null},"composition_subproblems":{"queries":["dynamic mode decomposition with control financial bank balance sheet stress","action-conditioned transition matrix eigenvalue damping financial risk","coupled leverage liquidity duration funding limits mode drift residual monitoring"],"source_ids":["SRC3","SRC5","SRC7","SRC8"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Lessons from the 2023 banking turmoil - Executive Summary","url":"https://www.bis.org/fsi/fsisummaries/exsum_23912.htm","publisher":"Bank for International Settlements, Financial Stability Institute","date_or_year":"2026 (summarizing the 2023 turmoil and subsequent BCBS work)","source_type":"OFFICIAL_GUIDANCE","language":"English","claims_supported":["Rules-based supervision can miss growing vulnerabilities even while regulatory requirements are met.","Credit Suisse reported an NSFR above 100% before failure, while realized deposit outflows exceeded standard calibrations.","Supervisors need forward-looking, high-frequency indicators beyond regulatory ratios, especially for liquidity, interest-rate risk, and interrelated activities."]},{"source_id":"SRC2","title":"Interagency Supervisory Guidance on Stress Testing for Banking Organizations with Total Consolidated Assets of More Than $10 Billion","url":"https://www.federalreserve.gov/frrs/guidance/interagency-supervisory-guidance-on-stress-testing-for-banking-organizations-with-total-consolidated-assets-of-more-than-10.htm","publisher":"Board of Governors of the Federal Reserve System","date_or_year":"2012","source_type":"OFFICIAL_GUIDANCE","language":"English","claims_supported":["Capital and liquidity can become impaired simultaneously and should be assessed interactively.","Liquidity actions such as loss-making asset sales or expensive funding can damage capital even when capital initially exceeds regulatory minimums.","Boards and senior management are identifiable owners of integrated stress-testing governance and remedial action."]},{"source_id":"SRC3","title":"Taking regulation seriously: fire sales under solvency and liquidity constraints","url":"https://www.bankofengland.co.uk/working-paper/2019/taking-regulation-seriously-fire-sales-under-solvency-and-liquidity-constraints","publisher":"Bank of England","date_or_year":"2019","source_type":"PRIMARY_RESEARCH","language":"English","claims_supported":["Banks' liquidation choices depend jointly on leverage, risk-weighted capital, liquidity coverage, asset liquidity, and regulatory weights.","Funding shocks can create substantial fire-sale losses that solvency-only models underestimate.","Combined funding and solvency shocks exhibit complementarities not reproducible by analyzing either shock separately."]},{"source_id":"SRC4","title":"Simulating dynamic balance sheet reactions and macroprudential policy using the 2025 EU-wide stress test","url":"https://www.ecb.europa.eu/press/financial-stability-publications/macroprudential-bulletin/html/ecb.mpbu202511_02.en.html","publisher":"European Central Bank","date_or_year":"2025","source_type":"FIRST_PARTY_PRODUCT","language":"English","claims_supported":["The ECB's operational BEAST framework models dynamic bank balance sheets, behavioral equations, policy measures, and bank-macroeconomy feedback.","Bank deleveraging can improve individual capital ratios while contracting credit and intensifying the downturn.","Buffer release is already evaluated as a candidate intervention with heterogeneous bank and credit-supply effects."]},{"source_id":"SRC5","title":"Evaluation of systemic risk in a financial system using dynamic mode decomposition","url":"https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6031928","publisher":"SSRN Electronic Journal","date_or_year":"2026","source_type":"PRIMARY_RESEARCH","language":"English","claims_supported":["Dynamic mode decomposition has already been proposed for financial-system risk evaluation using banking statements and market data.","The paper formulates instability using eigenvalues of an approximated Koopman operator.","The reported application analyzes major US financial institutions before the 2007-2008 crisis and compares the metric with existing methods."]},{"source_id":"SRC6","title":"Liquidité, solvabilité bancaire et crise financière : quelle relation ?","url":"https://www.banque-france.fr/fr/publications-et-statistiques/publications/liquidite-solvabilite-bancaire-et-crise-financiere-quelle-relation","publisher":"Banque de France","date_or_year":"2020; updated 2024","source_type":"PRIMARY_RESEARCH","language":"French","claims_supported":["A simultaneous-equations model was estimated on 725 French banks over 1993-2015 to study solvency-liquidity interactions.","Financial stress variables have nonlinear effects concentrated in crisis regimes.","Solvency and liquidity responses have different persistence, supporting joint dynamic rather than purely coordinate-level analysis."]},{"source_id":"SRC7","title":"Liquidity at Risk: Joint Stress Testing of Solvency and Liquidity","url":"https://www.imf.org/en/publications/wp/issues/2020/06/05/liquidity-at-risk-joint-stress-testing-of-solvency-and-liquidity-49325","publisher":"International Monetary Fund","date_or_year":"2020","source_type":"PRIMARY_RESEARCH","language":"English","claims_supported":["Traditional solvency and liquidity stress tests have often been separate and based on inconsistent scenarios.","A structural joint balance-sheet framework already models external solvency shocks, endogenous liquidity shocks, mitigating actions, and loss amplification.","Solvency-liquidity interaction can amplify equity losses through funding costs and fire sales."]},{"source_id":"SRC8","title":"Dynamic Mode Decomposition with Control","url":"https://epubs.siam.org/doi/abs/10.1137/15M1013857","publisher":"Society for Industrial and Applied Mathematics","date_or_year":"2016","source_type":"PRIMARY_RESEARCH","language":"English","claims_supported":["DMD with control is an established method for separating autonomous dynamics from the effects of actuation.","The method learns low-order input-output models from observed state and action snapshots without known governing equations.","Action-conditioned modal modeling is therefore not novel as a general method, although this source does not apply it to bank balance-sheet governance."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The core problem is externally supported: capital, liquidity, funding, interest-rate, and fire-sale mechanisms interact; combined shocks can be complementary; dynamic deleveraging can amplify credit contraction; and regulatory ratios or satisfactory assessments can remain reassuring while material vulnerabilities grow. The stronger assertion that a persistent eigen-direction specifically provides earlier warning than limits and transition-based stress models remains unvalidated.","source_ids":["SRC1","SRC2","SRC3","SRC4","SRC6","SRC7"],"uncertainty":"The evidence establishes coupled and nonlinear stress, but not that one institution's weekly data contain a stable, spectrally separated mode or that such a mode precedes conventional warnings. SVB is not a clean confirming example because several internal limits were breached, while the Credit Suisse NSFR example shows only that one regulatory coordinate remained compliant."},"adopter_evidence":{"status":"SUPPORTED","finding":"An institution's board, senior management, treasury/risk functions, and risk committee are identifiable analysis owners, while prudential supervisors are identifiable authorizers or reviewers for consequential capital and liquidity actions. Official guidance assigns governance and remedial responsibilities, and ECB practice shows supervisors using dynamic models for policy analysis.","source_ids":["SRC2","SRC4"],"uncertainty":"The sources establish role categories, not a named institution willing to supply the proposed weekly state, action, and incidence data."},"implementation_evidence":{"status":"PARTLY_SUPPORTED","finding":"All major ingredients exist separately: joint solvency-liquidity balance-sheet models, fire-sale constraints, dynamic behavioral stress models, financial DMD/Koopman instability metrics, and DMD with control. No retained source implements the full package of institution-level rolling modes, bootstrap separation, locked-holdout detection, identified or bounded action-conditioned maps, worst-case gain reduction, drift retirement, and affected-party incidence budgets.","source_ids":["SRC3","SRC4","SRC5","SRC6","SRC7","SRC8"],"uncertainty":"Public sources cannot reveal confidential in-house bank models. Twelve plausibly exogenous episodes per action class may be too few for reliable quasi-experimental identification, and reflexivity may invalidate a weekly local operator."},"prior_art":{"disposition":"ADJACENT_PRIOR_ART","closest_analogues":[{"name":"Financial-system dynamic mode decomposition / Koopman instability metric","source_ids":["SRC5"],"same_problem":false,"same_causal_lever":true,"overlap":"Uses bank statements and market data, approximates financial dynamics with a Koopman operator, and classifies instability from its eigenvalues.","remaining_difference":"It is a system-level retrospective detection proposal; the retained source does not show a single-bank weekly balance-sheet mode, action-conditioned damping, robust worst-case control, incidence constraints, or operational retirement rules."},{"name":"Liquidity at Risk joint solvency-liquidity stress testing","source_ids":["SRC7"],"same_problem":true,"same_causal_lever":false,"overlap":"Models a granular balance sheet, endogenous liquidity shocks, funding-cost and fire-sale amplification, and mitigating actions in one coherent framework.","remaining_difference":"It uses structural stress mechanisms rather than learned invariant modes and does not select actions by worst-case post-action modal gain with locked-holdout drift and separation tests."},{"name":"ECB BEAST dynamic balance-sheet framework","source_ids":["SRC4"],"same_problem":true,"same_causal_lever":false,"overlap":"Operationally models bank behavior, dynamic balance sheets, deleveraging, credit contraction, macroeconomic feedback, and policy interventions.","remaining_difference":"It is a system-wide behavioral/VAR framework rather than a local institution-level eigendecomposition and does not use modal gain as the action criterion."},{"name":"Dynamic Mode Decomposition with Control","source_ids":["SRC8"],"same_problem":false,"same_causal_lever":true,"overlap":"Already supplies the generic data-driven method for separating autonomous dynamics from action effects and learning an input-output modal model.","remaining_difference":"It does not address financial balance sheets, prudential limits, deleveraging spirals, causal identification from bank actions, protected-party incidence, or supervisory authority."}],"contrastive_claim_remaining":"For one institution and a prespecified regime, a regularized rolling balance-sheet operator may yield a stable joint mode that adds locked-holdout stress-detection value beyond separate limits and a transition-based rival; where action responses are independently identified or tightly bounded, selecting the smallest action that reduces the worst-case modal gain may improve held-out outcomes without exceeding incidence budgets.","contrastive_claim_falsifier":"The claim is eliminated by finding a routine implementation of that same institution-level detection-and-robust-damping package, or empirically if modes fail prespecified separation, residual, drift, or incremental-detection tests, or if every admissible action-response set contains no gain reduction relative to no action and the rival.","confidence":"MODERATE","search_limitations":"This was a bounded public-web review, not a patent, procurement, code, or confidential bank-model search. The newest finance-DMD item is a lightly downloaded 2026 SSRN paper whose direct page resisted full-text fetching, although its indexed abstract was opened through search. Commercial ALM documentation is usually high-level. The search cannot exclude unpublished supervisory prototypes or internal treasury models."},"researchability_gates":{"externally_supported_problem":{"status":"PASS","rationale":"Official guidance, primary research, and observed banking turmoil support the existence of interacting balance-sheet stresses and late or incomplete signals from isolated ratios. Spectral early warning remains the testable mechanism rather than an assumed fact.","source_ids":["SRC1","SRC2","SRC3","SRC4","SRC6","SRC7"]},"identifiable_adopter_or_authorizer":{"status":"PASS","rationale":"Board and senior-management governance, treasury/risk ownership, and supervisory authority are externally identifiable; the first step is analysis-only and can be authorized internally without delegating live prudential powers to the model.","source_ids":["SRC2","SRC4"]},"distinct_testable_incremental_claim":{"status":"PASS","rationale":"Close prior art covers financial eigenmode detection, generic DMD control, and dynamic joint stress separately, but leaves a checkable increment: locked-holdout institution-level modal detection plus independently identified or bounded worst-case damping with residual, drift, and incidence criteria.","source_ids":["SRC4","SRC5","SRC7","SRC8"]},"bounded_next_evidence_step":{"status":"PASS","rationale":"A no-action, preregistered retrospective study is bounded and reversible. It can first test data sufficiency, mode stability, and incremental detection; damping analysis proceeds only for action classes with adequate exogenous episodes or defensible response bounds.","source_ids":["SRC2","SRC4","SRC5","SRC8"]},"no_unresolved_safety_or_authority_stop":{"status":"PASS","rationale":"The authorized next step does not trade, sell assets, restrict customers, or alter capital or liquidity policy. Existing controls remain in force, and official governance practice keeps live remedial decisions with management and applicable supervisors. Incidence and halt criteria address borrower, depositor, and credit-supply harm before any later action study.","source_ids":["SRC2","SRC4"]},"adequate_search_evidence":{"status":"PASS","rationale":"All six lanes were searched adversarially, including historical language, official guidance and operational models, French/Spanish/German terminology, and component combinations. Exactly eight retained direct sources span seven independent publishing organizations and include multiple official, primary, and first-party sources; each retained source was opened or directly fetch-attempted.","source_ids":["SRC1","SRC2","SRC3","SRC4","SRC5","SRC6","SRC7","SRC8"]}},"strict_success":true,"screen_survival":true,"remaining_research_value":"MODERATE","recommended_next_step":"Run a data-feasibility checkpoint before fitting: verify at least 260 consistently defined weekly observations, action timestamps, affected-party outcomes, and enough genuinely exogenous episodes. Then preregister a detection-only comparison against separate limits, joint structural stress testing, and a non-spectral transition rival on the locked final 20%. Estimate action-conditioned damping only for action classes passing identification or conservative response-set requirements; otherwise stop at detection.","world_novelty_boundary":"This bounded search supports only an adjacent-prior-art judgment and a remaining falsifiable research claim. It does not establish world novelty, patentability, freedom to operate, market size, adoption, superiority, or realized impact."}