2008 Financial Crisis
AIG is the #1 derivatives counterparty hub. Fannie Mae is the #1 direct credit hub. The Functional Proximity Law is denied — derivatives and credit exposure reflect different institutional operating models. The denial has a named mechanism.
Denied results with named mechanisms are not failures. They define the law's boundary conditions. The finance denial is the third structural constraint on the IRDME law.
What was measured
16 major financial institutions. Three independently defined exposure layers, sourced from FCIC report (2011), BIS data, and FDIC filings.
derivatives_counterpartyInstitution A is a significant OTC derivatives counterparty to institution B (CDS, interest rate swaps, forex derivatives). Encodes trading-desk exposure — the operating model of investment banks.
direct_credit_exposureInstitution A holds significant direct lending or credit exposure to institution B (mortgage-backed securities, direct loans, repo agreements). Encodes balance-sheet credit risk — the operating model of commercial banks and GSEs.
equity_ownershipInstitution A holds significant equity ownership stake in institution B (stock, warrants, equity tranches of structured products). Encodes cross-ownership structure.
Cross-layer hub correlation
derivatives ↔ creditDENIEDNo correlation. Derivatives hubs (investment banks) and credit hubs (GSEs, commercial banks) are structurally separate populations.
derivatives ↔ equityPARTIALWeak, non-significant correlation. Some overlap between derivatives and equity exposure networks (investment banks appear in both).
credit ↔ equityDENIEDNo correlation. Credit exposure (GSEs, commercial banks) and equity ownership (investment banks) are entirely separate institutional classes.
The named mechanism
This is the third formal law boundary condition. It states when IRDME correctly should not find correlation.
The 2008 financial system had two structurally distinct institution classes operating simultaneously:
AIG, Goldman Sachs, Lehman, Deutsche Bank, Morgan Stanley. Their systemic risk was in derivatives counterparty networks — the contracts they traded, not the loans they held.
Fannie Mae, Freddie Mac, WaMu, Citigroup (commercial arm). Their systemic risk was in direct credit exposure — the mortgages and loans they originated and held.
These two classes participated in the same crisis but operated in structurally different relational regimes. A trading-desk hub and a balance-sheet hub are not measuring the same kind of coupling. The law's prerequisite — both layers encoding the same relational regime — is not satisfied. r = 0.042 at p = 0.824 is the correct result. The law working as intended.
Hub shadows — what the regulators missed
10 of 16 institutions have a rank gap ≥ 4 between layers. The most informative divergences:
Fannie Maehub_shadowgap = 12The largest hub shadow in the dataset. Fannie Mae was the dominant credit exposure hub but was virtually absent from derivatives counterparty networks. Risk frameworks that tracked derivatives exposure — the standard post-LTCM practice — were structurally blind to it.
Lehman Brothershub_shadowgap = 5High derivatives hub, mid-tier credit hub. Lehman's failure propagated primarily via derivatives counterparty chains — exactly where its hub rank was highest. The credit exposure channel was secondary.
What this means
- ·A single-layer risk map — derivatives-only or credit-only — will miss the institutions that are hubs in the other layer. In 2008, this meant missing Fannie Mae's credit centrality if you were tracking derivatives, and missing AIG's derivatives centrality if you were tracking credit.
- ·The 10 hub shadows are the quantitative version of "too big to fail in ways regulators weren't measuring."
- ·The law denial is informative: when r ≈ 0 between two financial exposure layers, the layers encode structurally separate risk channels. A multi-layer view is not optional — it's the only way to see the full picture.
A hub shadow in software is a module with low import rank and high co_change rank — behaviorally central, declared as peripheral.
Fannie Mae is structurally identical: declared peripheral in derivatives (rank #13), actual hub in credit (rank #1). Same archetype, different domain, same IRDME label.
This is what the Universal Layer Grammar demonstrates: hub_shadow in software is systemic risk in finance — the same structural pattern applied to different material.
Hub ranking by layer
| institution | derivatives | credit exp. | equity | archetype | note |
|---|---|---|---|---|---|
AIG | #1 | #2 | #11 | universal_hub | Derivatives counterparty to every major bank |
Citigroup | #4 | #3 | #5 | universal_hub | Persistent hub across two layers |
Goldman Sachs | #2 | #7 | #1 | chameleon | Dominant in equity ownership, mid-tier in credit |
Lehman | #3 | #8 | #12 | hub_shadow | High derivatives rank, invisible in credit |
Fannie Mae | #13 | #1 | #14 | hub_shadow | Largest hub shadow: credit hub, derivatives peripheral |
Freddie Mac | #14 | #4 | #15 | hub_shadow | Same pattern as Fannie Mae |
JPMorgan | #8 | #5 | #8 | relay | Consistent mid-tier across all three layers |
Deutsche Bank | #5 | #11 | #4 | relay | Derivatives/equity hub, mid credit |
8 of 16 institutions shown.
Pre-registered hypotheses
r(derivatives ↔ credit) > r(derivatives ↔ equity)
0.042 < 0.183. Mechanism: institutional operating model mismatch (trading vs lending). Law boundary condition #3.
AIG is top hub in both derivatives and credit layers
Derivatives: #1. Credit: #2 (Fannie Mae is #1). Near miss — AIG is persistent but credit top hub is Fannie Mae.
≥ 5 institutions have rank gap ≥ 4 between derivatives and credit layers
10 of 16 institutions diverge by ≥ 4 ranks. Top divergents: Fannie Mae (gap=12), Freddie Mac (gap=10), WaMu (gap=9).
r(derivatives ↔ credit) > 0, p < 0.05
r = 0.042, p = 0.824. Direction positive but not significant — n=16 too small to detect weak correlation, and true r is near zero.
r(derivatives ↔ equity) > 0, p < 0.05
r = 0.183, p = 0.473. Weak positive correlation, not significant at n=16.
Related case studies
post.php as hub shadow. r = 0.5156, p = 0.012.
PVCL/PVCR confirmed from topology alone. r = 0.7774, p = 0.004.
Transformer #1 in structural layers. Law confirmed, one denial with named mechanism.