Research library

INVESTIGATION 10 / CREDIT & CONTAGION

Risk travels through networks

Can a financial network amplify an initial default?

Graph networksContagionGCN

01 / THE PRACTICAL QUESTION

A decision, before a model.

An institution appears resilient on its own but is exposed to other institutions that can fail together.

The analyst’s decision

Test whether graph learning predicts simulated contagion more accurately than a structural rule.

02 / DATA & COMPARISON

The idea in plain language.

A graph model passes information between connected institutions. A structural baseline explicitly approximates how losses propagate through the same network.

Brier score
The average squared difference between predicted probabilities and observed outcomes.

Data. Seeded financial networks with simulated default cascades.

Baseline. Degree and exposure-based structural scores.

03 / THE EXPERIMENT

What the saved experiment shows.

The graph convolution model has a higher Brier score than the one-round structural baseline in all three held-out seeds. Lower is better: the simpler mechanism-aware baseline wins this saved comparison.

Inspect the supporting result

Evidence record: learning-benchmarks.json#projects/10

Explore the related lab

04 / RESULTS & LIMITATIONS

Evidence with its boundaries attached.

The related lab is a cross-project demonstration. Read this investigation’s evidence and limits before transferring its conclusions.

Inspect numerical checks and validation records

Loading validation evidence…

Loading across-seed evidence…

Interpretation limit

Constructed network labels are conditional on the chosen cascade mechanism.

05 / REPRODUCE

Reproduce and challenge the result.

Code, configuration, and reproduction

The project contains its implementation, configuration, tests, and walkthrough. Download the lab configuration to record the exact parameters used in an interactive run.

Project code and walkthrough