debtrank-globe

An interactive, model-driven visualization of sovereign external debt contagion risk.

Live application: lastch1ld.github.io/debtrank-globe

Instead of just charting external debt statistics (à la JEDH), this project builds a real cross-border debt exposure network from public World Bank and BIS data, then runs published financial systemic-risk algorithms on it:

• DebtRank (Battiston, Puliga, Kaushik, Tasca & Caldarelli, 2012) — iterative distress propagation through a weighted exposure network, used by central banks for systemic risk assessment.
• Eisenberg–Noe clearing model (Eisenberg & Noe, 2001) — fixed-point clearing vector for network default cascades, used as a secondary/comparison model.

The result: pick a country, dial in a shock magnitude, toggle between DebtRank and Eisenberg-Noe, and watch distress propagate through the real global debt network on an interactive 3D globe — for any year from 2005 to 2025 via the year scrubber.

Structure

• data-pipeline/ — fetches and normalizes World Bank (external debt, reserves, GDP) and BIS (bilateral cross-border banking exposures) data into a network snapshot.
• model/ — Python implementation of DebtRank and Eisenberg–Noe, with correctness tests reproducing the toy examples from the original papers.
• web/ — React Three Fiber static site: 3D globe visualization and interactive shock simulation.

Use it yourself

The model is on PyPI, and the per-year networks it runs on are public static JSON:

pip install debtrank-model
debtrank-simulate 2020.json --shock GRC=1.0 --shock PRT=0.6

• model/README.md — library and CLI usage.
• docs/data-api.md — the snapshot schema, its URLs (2005-2025, CORS-enabled), and the attribution terms for the underlying World Bank / BIS / IMF data.
• notebooks/reproduce.ipynb — fetch a year, run the model, print the ranking.

Data attribution

• External debt, reserves, and GDP indicators: World Bank Indicators API, International Debt Statistics database.
• Cross-border bilateral banking exposures: BIS Locational Banking Statistics, used under the BIS's terms of permitted use. Only a small, derived, aggregated network snapshot is redistributed in this repo — not bulk BIS data.

Status

Functional end-to-end: real data pipeline, correctness-tested DebtRank and Eisenberg-Noe models (Python reference + matching TypeScript ports), and an interactive globe with a model toggle, shock-magnitude control, market comparison, and a 2005–2025 historical year scrubber. The static application is deployed publicly through GitHub Pages.

License

MIT — see LICENSE.