Systemic Risk as $H^2$: Cohomological Stress Testing, the Unhedgeable Residual, and Why the 2008 Crisis Was a Topological Event

Standard financial stress tests compute the sum of bilateral losses across institutions. This is an $H^0$ computation — it evaluates local data at each edge of the exposure network without asking whether those local data are globally consistent.

This paper argues that financial crises are systemic risk ($H^2$) events: they occur when the triangular risk ($H^1$) classes of individual institutions' portfolios become mutually inconsistent at the system level, producing a non-trivial second cohomology class of the system interaction diagram. When $H^2 = 0$, individual triangular risks are jointly consistent: losses are absorbed without amplification. When $H^2 \neq 0$, individual risks are jointly inconsistent: small shocks cascade into crises through topological amplification.

Three principal results are established. (1) The 2008 crisis as $H^2$: mortgage risk was individually managed as triangular risk ($H^1$) at each institution; the cross-institution correlation of mortgage exposures was an $H^2$ class that no regulator computed; when the $H^2$ class became non-trivial, the Pentagon identity failed at the system level and the cascade began. (2) The cohomological stress test: a three-tier test — bilateral ($H^0$), triangular ($H^1$), systemic ($H^2$) — where only the $H^2$ tier detects the onset of cascades. (3) The SIFI theorem: a financial institution is systemically important if and only if its removal changes the $H^2$ class of the system. Size is neither necessary nor sufficient; topological centrality is.

A dedicated section classifies XVA components by cohomological level. CVA, DVA, FVA, and MVA are triangular risk ($H^1$): computable at the desk level and hedgeable with credit options, funding swaps, and margin agreements. Wrong-way risk — the correlation between counterparty default probability and exposure size — is systemic risk ($H^2$): it cannot be computed from desk-level data and cannot be hedged by any finite collection of triangular instruments. Standard XVA models that sum individual adjustments compute $H^1$ only; the error relative to the true total valuation adjustment is exactly the $H^2$ wrong-way risk class. KVA sits at the $H^1$/$H^2$ boundary because capital surcharges depend on SIFI designation, which is itself an $H^2$ property.

Existing systemic risk measures are identified as special cases: DebtRank is $H^0$ (bilateral propagation only); CoVaR and SRISK are $H^1$ (conditional on one institution's triangular risk). Flood, Kenett, Lumsdaine and Simon (2017) compute Betti numbers — the cohomology of the constant sheaf — on bank holding company ownership graphs; the present framework is the strict generalisation in which the constant sheaf is replaced by the pricing sheaf carrying financial content.

The natural division of responsibility follows from the mathematics: XVA desks compute $H^1$; the CRO and risk management compute the $H^2$ wrong-way risk contribution; regulators and CCPs provide the system-level $H^2$ data. This division is structurally necessary, not organisationally convenient. The $H^2$ component of XVA is not computable at the desk level regardless of model sophistication.

A companion primer (doi:10.5281/zenodo.20642983) develops all concepts from first principles for readers without prior topology.

Keywords

Systemic Risk, H², Sheaf Cohomology, Cohomological Stress Testing, Financial Contagion, Bilateral Risk, Triangular Risk, Pentagon Identity, XVA, CVA, DVA, FVA, MVA, KVA, Wrong-Way Risk, SIFI, Systemically Important Financial Institution, SIFI Theorem, DebtRank, CoVaR, SRISK, Betti Numbers, Financial Networks, Interaction Diagram, Pacioli Manifold, HJM No-Arbitrage, 2008 Financial Crisis, Cascade, Contagion, CCP, Central Clearing Party, Capital Requirements, Basel, Unhedgeable Risk, Origami ISA, Econiac, Financial Gauge Theory, Topological Finance