Data is built from canonical public datasets: IMF World Economic Outlook (October 2025), World Bank International Debt Statistics (2024 vintage), AidData Chinese Loans & Grants v1.0, Bahaj/Fuchs/Reis swap-line dataset, OECD Trade in Value Added (2022 vintage), and Voeten UN ideal-point data (through session 77).

Risk-tier classifications use quantitative thresholds across debt/GDP, inflation, current account, and real growth. Methodology follows the Manasse-Roubini conditional-cascade framework.

Outbound stress paths (v2). For each country, four trace types surface first-order propagation channels: direct sovereign exposure (who owes this country), co-exposure cluster (countries with structurally similar creditor portfolios, by cosine similarity on share-normalized creditor vectors, computed for sovereigns with at least $2B in tracked external debt and 3+ distinct creditors), currency liquidity outbound (swap-line recipients, with sole-provider dependents flagged), and trade outbound (top foreign-value-added destinations with share-of-destination-imports). Confidence tagging (high/medium/low) reflects both similarity score and portfolio depth. The traces show static first-order exposure, not full dynamic cascade; second-order and cross-layer propagation modeling are outside v2 scope.

Macro time series (v2.5). Each metric card includes a 2010–2030 sparkline drawn from IMF WEO October 2025 vintage. Solid lines are observed annual values through the IMF's latest_actual_data marker; dashed lines are IMF staff forecasts. Sparkline color encodes recent direction: rust for deterioration, moss for improvement, slate for stability, measured over the last five observed years. Anchor dot marks the most recent observed value. Time series help readers see whether a country's structural conditions are building or releasing, rather than just where they currently sit.

Tradable instruments (v2.6). Each country profile includes a reference panel of instruments available for sovereign exposure: USD-denominated bonds, sovereign CDS tenors, local-currency bond market access, FX (with restriction and NDF flags), and country-specific equity ETFs or indices. Trading-depth classification (deep / moderate / thin / minimal) reflects overall market presence. Instrument data is point-in-time and reference-only — not a substitute for a trading terminal. Bond identifiers may include retired or restructured issues for distressed sovereigns.

Scenario engine (v3.0). Users describe a shock at one country, a regional grouping, or a chokepoint; the engine traces structural first- and second-order exposures across the existing dependency layers. Multi-source cascades (Strait of Hormuz, GCC, Eurozone periphery) aggregate exposure across the affected source countries. Free-text input is parsed by Claude API via a Cloudflare Worker proxy; analytical prose is generated dynamically. The engine traces structural exposure only — it does not model commodity prices, bond contagion, EM spread widening, freight rates, central-bank reaction functions, or duration of impact. Channel-specific caveats accompany each scenario output.

Portfolio reallocation channel (v3.1). A sixth cascade channel models scenarios in which a country reduces its foreign portfolio holdings — the canonical case being Japanese institutional repatriation from US Treasuries when yield differentials shift. Bilateral holdings data come from the IMF Portfolio Investment Positions by Counterpart Economy (PIP, formerly CPIS) dataset, vintage end-June 2025, covering 85 reporting economies against 245 counterparts. For US-counterparty rows, the engine overrides PIP with the US Treasury's TIC SLT data (monthly, more accurate, includes foreign exchange reserves which PIP excludes). The methodology follows Judson & Kim (2026), "Measuring Cross-Border Securities Positions: Explaining Asymmetries between U.S. Treasury TIC and IMF PIP Data," FEDS Notes. Custodial-bias countries are flagged at two levels: high-severity (Belgium, Luxembourg, Ireland, Cayman Islands, Bermuda — offshore fund and depository centers whose reported holdings substantially reflect custody booking rather than beneficial ownership) and moderate-severity (France and the United Kingdom — major financial centers with partial pass-through). The cascade de-ranks custodial entries proportionally but always surfaces high-severity flags above the $50B threshold so the reader sees them. The channel does not model foreign exchange reserve reallocation directly (PIP excludes reserves; SEFER country-level data is not released); central-bank reserve dynamics are partially captured via the gold reserves sidebar on country briefs.

Gold reserves (v3.1). Each country brief displays central-bank gold reserves from the World Gold Council, sourced from IFS May 2026. The de-dollarization narrative is most visible in this data: China, Russia, Türkiye, Poland, and others have systematically accumulated gold since 2022 as a structural alternative to dollar-denominated reserve assets. Gold reserves are presented as a country-level data point rather than a cascade channel because they don't propagate through bilateral counterparty relationships the way debt, trade, and portfolio holdings do.

• 184 sovereign profiles (more being added)
• Sovereign debt — 2024 vintage (IDS May 2026 release)
• FX reserves & swap lines — through 2025
• Critical inputs — in development
• Trade (TiVA) — 2022 vintage
• UNGA alignment — through 77th session (2022)

GeoStress is part of a small family of analytical tools by Mickey Fortune, alongside SovRisk, Supply Shocks, FOMC Oracle, and others.

This is a v3.1 beta. Feedback welcome.