12  Systemic-Risk & Crowding: Existing Groundwork

Author: Todd B. Adams Reinforces: Proposal §2 — Theoretical Foundations & §3 (network layers) · Reading order: doc 05 of the research-platform package

The claim of this doc. The proposal’s network idea is not foreign to this platform — a version of it is already running. SBFoundation carries correlation-network, factor-network, and crowding machinery in production, and it has already run an empirical spike against the supervisor’s own published connectedness methodology. That prior work both establishes precedent and sharpens the proposal’s distinct contribution.

12.1 Network structure already lives in the platform — BUILT

The proposal treats assets, factors, and funds as nodes across topological layers. The platform already computes network statistics over two of those object types today:

Existing capability What it is, in network terms Package / file Grounded in
Comomentum crowding (FC-1) Abnormal within-decile return correlation among a factor’s constituent names — a correlation-network statistic over the traded cross-section sbcrowding → comomentum_crowding_service.py, _comomentum_math.py Lou & Polk (2012), Comomentum
Factor ρ-network (LH-6y) Pairwise |Spearman ρ| among all active factors — a weighted factor–factor graph, collapsed to effective trial count sbic → pairwise_correlation_service.py Benjamini–Yekutieli FDR over a dependency graph
Orthogonality / confound (LH-6, LH-6z) Residual-IC of a candidate factor against the validated book — an edge-removal / independence test on the factor graph sbic → orthogonality_service.py, factor_confound_service.py rank-space residualization (F-305)
Factor risk model Factor covariance + idiosyncratic variance — the dense dependency matrix the proposal denoises sbriskmodel Barra-style decomposition (F-073)

The network concept is therefore already present at the factor level and in crowding. What the proposal adds is an asset-level, multi-layer read and a spectral phase-transition signal — genuinely new, but built on a substrate that already speaks this language.

12.2 The Di Matteo connection — prior empirical work, already done — BUILT

The proposal’s supervisor, Prof. Tiziana Di Matteo, co-authored the survey the platform already stress-tested: Raddant & Di Matteo (2023), A look at financial dependencies by means of econophysics and financial economics. In July 2026 a read-only research spike ran a Diebold–Yilmaz total connectedness index over the platform’s own 2004–2026 universe (5,698 trading days × 11 GICS sectors, rolling VAR → generalized FEVD) to test whether network connectedness leads market stress.

The honest finding — and why it strengthens, not weakens, the proposal:

  • Connectedness proved coincident, not leading (corr(TCI, forward-vol) fell monotonically with horizon: +0.42 at h=0 → +0.24 at h=42d), with a saturated dynamic range and sensitivity to universe composition. Full write-up: asset-dependency networks findings.
  • This falsified the simplest network-as-early-warning hypothesis before any production code shipped — exactly the empirical discipline a doctoral programme rewards.
  • It also localizes the proposal’s contribution precisely. The proposal does not claim DY connectedness predicts crashes. It claims the eigenvalue spectrum of the Supra-Laplacian (algebraic connectivity / critical slowing down across a multi-layer graph) carries an early-warning signature. That is a different and stronger mathematical object than a single-layer VAR connectedness index — and the platform’s spike is the evidence that the weaker version was tested and found wanting. The PhD picks up exactly where the spike left off.

12.3 How this maps onto the proposal’s layers

  • Layer 3 (institutional ownership / crowding). The platform’s comomentum producer is a return-correlation proxy for crowding today; the proposal’s 13F ownership layer is the holdings-based ground truth. One is an approximation of the other — see the gap analysis in doc 03 and doc 07.
  • Physics evaluation engine (§3). The platform already has the report-only, evidence-first grammar (ops.research_* table → sidecar → dashboard card, no gate at first) that a Supra-Laplacian spectral producer would inherit — the sbcrowding producer is the literal template. See doc 06.

12.4 Convention note

The Raddant & Di Matteo (2023) paper is not yet captured in docs/reference-papers/ (it is open-access, CC-BY). Doc 10 flags it for bundling; Lou & Polk (2012) Comomentum is already bundled there (row 74).

Cross-links: proposal §2 literature · bundled network/GNN papers · DY-connectedness spike