How QGI Finds Geopolitical Risk
QGI surfaces economically-grounded risk candidates by finding historical precedents in World Bank data spanning 214 countries and 58 indicators. Here is how it works.
The Core Idea
Countries often walk the same economic path, just at different times. When Country A's economic trajectory today mirrors Country B's trajectory from years ago — and Country B subsequently experienced a significant geopolitical event — QGI flags that as a precedent worth examining.
QGI does not predict the future. It identifies structural similarities between countries' economic trajectories and surfaces historical precedents that analysts should investigate.
Three Layers of Analysis
Layer 1: Indicator Relationships (SCDIs)
A Single Dyadic Correlating Index (SCDI) is a statistically significant correlation between Country A's trajectory on one economic indicator and Country B's trajectory on the same indicator, possibly offset in time. We compute Pearson correlations across every country pair, every indicator, and every possible time window.
Layer 2: Patterns
A Pattern aggregates all SCDIs that share the same country pair, time window, and segment length across multiple indicators. It answers: “How many indicators, and how strongly, does Country A's trajectory match Country B's?”
Layer 3: Precedents (Recipes)
A Precedent matches a pattern to a known geopolitical event. If Country B's economic trajectory was followed by a fiscal policy change, and Country A is currently walking that same trajectory, the system flags it: “Country A is following a path that preceded fiscal_policy_change in Country B.”
Data Sources
QGI currently uses 58 economic indicators from the World Bank Open Data platform (CC BY 4.0 license), covering:
- Macroeconomic: GDP, GNI, inflation, trade balance
- Fiscal: government debt, tax revenue, expenditure
- Financial: credit to private sector, reserves, FDI
- Military: military expenditure, armed forces, arms trade
- Social: unemployment, education enrollment, health
- Governance: World Governance Indicators (6 dimensions)
- Infrastructure: electricity access, internet, mobile
- Environment: CO2 emissions, renewable energy, forest area
Additional providers (V-Dem governance indices, FAOSTAT food security, UCDP conflict data, UNHCR displacement data) are being integrated to expand coverage to 150+ indicators.
How Scoring Works
Z-Score Normalization
For each country, we compute how unusual each event category's precedent count is compared to that country's own average. This removes the data-completeness bias — countries with more complete World Bank data don't automatically score higher.
Salience Score
The z-score is multiplied by a category relevance weight that reflects the geopolitical consequence of each event type. Civil war and economic crisis carry full weight. Infrastructure projects and education reforms carry minimal weight. This ensures the ranking surfaces consequential signals, not statistically unusual but low-impact events.
Rank Fusion
The salience-weighted scores are blended with a persistence baseline (did this country experience similar events recently?) using Reciprocal Rank Fusion. This ensures both structural trajectory signals and recent history contribute to the final ranking.
Risk Tiers
Strong, consequential signal detected. Multiple historical precedents converge.
Notable signal present. Historical precedents exist but are less concentrated.
Weak or low-consequence signals only. Baseline monitoring.
Validation
QGI V1.8.1 was backtested against 403 ground-truth country-event pairs across 88 curated countries with a 2020 cutoff (trained on pre-2020 data, tested on 2020–2025 events):
Lift@50 of 2.24× means the top 50 predictions are 2.24 times more likely to match a real event than a random selection. The system beats a persistence baseline (predicting last year's events will repeat) on F1 score.
Pattern Length Tiers
Patterns are grouped by how many years they span, which reflects fundamentally different types of structural dynamics:
Short-term: policy cycles, commodity shocks, fiscal crises.
Medium-term: leadership eras, generational shifts, reform cycles.
Long-term: Cold War legacies, development trajectories, structural eras.
Institutional: civilizational similarities, deep structural parallels.
Limitations
- Annual resolution: World Bank data is updated annually. QGI detects slow-moving structural trajectories (5–10 year patterns), not fast-moving crises.
- Correlation is not causation:A statistical match between two countries' trajectories does not mean one causes the other. QGI identifies historical analogues, not causal mechanisms.
- Data lag: World Bank indicators are typically 1–2 years behind the current year. The most recent complete data year is usually 2 years ago.
- Economic indicators only: QGI does not currently incorporate political event data, social media sentiment, or conflict databases. Events driven by personality politics, external military intervention, or sudden natural disasters may not show in economic trajectories.
- Category breadth:Some event categories (e.g., “government formation”) are broad. A government formation could mean a routine election or a post-coup transition. Category specificity is being improved.
How to Read QGI
When QGI shows “Argentina — Fiscal Policy Change — HIGH,” it means: Argentina's current economic trajectory statistically resembles the trajectories of multiple countries that subsequently experienced fiscal policy changes. Whether Argentina will actually experience a fiscal policy change depends on many factors QGI does not measure.
The value of QGI is in the evidence trail: which countries, which indicators, which time periods form the historical analogy. That evidence is what analysts should examine.