Each tournament, Goldman Sachs, the Metaculus superforecasters, and the bookmakers forecast the football itself, forecasting who wins each match and lifts the trophy. Their odds are our starting point. We run each result as a what-if scenario on a causal model, an intervention traced through to a country's exports and its stock market.
Does it hold up? Scored on tournaments the model never saw, its 90% ranges covered 90% of held-out export moves and 87% of market moves. PolyBridge Live Scoreboard
Exports
A World Cup win lifts a country's exports — by a median of +0.3 to +0.8pp of additional two-quarter growth, with a 90% range reaching about +2.0pp. Below, for each contender: the market's current chance it wins the trophy, and our model's expected export lift if it does.†
† Each country's lift scales with its manufactures-export share — branded, differentiated goods (cars, machinery, design) benefit from a nation's rise in global visibility; commodities do not. The amplification is estimated on 168 countries (a more manufacturing-heavy exporter's exports react ~18% more per standard deviation) and applied to each contender by its own export mix: commodity exporters (Argentina, Brazil) get less, branded-goods exporters (Belgium, Mexico, Italy) more. Dots are the central estimate; lines span the 90% range. Method, data, and calibration in the appendix. Championship odds as of 30 Jun 2026.
Equity markets
A country's equity market runs about 15% more volatile the trading day after its national team plays a World Cup match, fading back to normal within a month. There's no read on direction — a win, an ordinary loss, and an elimination all land within a fraction of a percent of zero.
How Polybridge reads the tournament
A football result reaches the economy through a few channels — global attention, trade demand, market mood. The Polybridge model maps these pathways as a causal graph and fits the hidden channels that link football to the economy from data. Its structure maps the causal pathway the effect actually follows — a result first moves attention and sentiment, which in turn move markets and trade — rather than a single black-box correlation. Prediction-market odds inform the football side; real export and market data inform the outcome side. To answer a question, we apply a causal intervention and trace it through the graph to each country's exports and markets.
Confidence differs by question. The volatility result is directly measured — 400+ matches, with price swings and trading volume moving together. The export effect is positive and calibrated, with a wider interval, because exports move for many reasons and the global-attention signal behind the channel exists only for recent tournaments. Full method, data, and calibration follow in the appendix.
Method & evidence
Data
Exports: seven European economies across seven World Cups (1998–2022), Eurostat quarterly real exports. Volatility: 400+ country-matches across 17 countries — realized volatility and trading volume around each match. The channels also draw on foreign attention (Wikipedia, recent tournaments) and macro covariates (World Bank).
The measured effect, partial pooling & do-calculus
Winning raises two-quarter export growth by about +0.5–0.6 points per standard deviation of upset. The model learns this effect from the data — a prior centred at zero is pulled to it by the evidence — then each figure is an intervention on the fitted model (do(win) through the trade channel), carrying its learned structure: a concave response to upset size, full posterior uncertainty. We report a branded-export per-country effect: the loading scales with each country's manufactures-export share — identified on a 179-country external panel (×1.18 per SD; wild p<0.001) and ported as a prior, so it transports to every contender by its own export mix. Residual country variation is partially pooled under a skeptical prior.
Calibration & likelihood
Every lane uses a heavy-tailed Student-t likelihood, fit to the data's own fat tails (10–18 degrees of freedom). Held out one tournament at a time and scored on events never seen in fitting, the model's 90% ranges cover 90% of export outcomes and 87% of volatility outcomes.
What is measured, carried, or taken as input
| Channel | Status | Basis |
|---|---|---|
| Prediction markets | Input — taken as given | live championship and match win-odds feed the football side of the model; we forecast the consequences, not the football |
| Market attention (volatility) | Directly measured | two signals — price swings and trading volume — co-move across 400+ matches |
| Trade (exports) | Positive & calibrated; precision partly prior-assisted | direct on the export panel; the attention signal behind it is recent-only |
| Foreign attention | Carried (literature-prior) | Wikipedia / GDELT spike on the winning nation; recent tournaments only, so the salience channel falls back to a cited prior |
| Domestic mood | Placebo / discriminant | consumption + consumer confidence (Eurostat / OECD) — a near-zero response supports export-specificity |
| Branded-export factor | Data-derived prior (external) | manufactures-export share, identified on a 179-country World Bank panel; scales each country's export lift |
Volatility direction
Direction uses the same machinery — do(win) / do(loss) / do(elimination) → the next day's market reaction across 400+ matches, full posterior, no forced sign. We fold the published Edmans–García–Norli elimination effect (−0.5% next-day; 2007) in only as a weakly-informative prior, so the data dominates: all three land within a fraction of a percent of zero — no reliable directional edge, matching Gatto (2026).
Explore PolyBridge causal world model foresight on cyber, geopolitical and macro here: polybridge.ai/situation-rooms