Every country dashboard in Chart Lab carries a single 0–100 tailwinds score. This page explains exactly what goes into it, how each input is calculated, why it is built the way it is, and — just as important — what it is not.
The score answers one question: how supportive is this country's macro backdrop for its equity market right now, relative to that country's own history? It is a cross-sectional tilt signal — designed to compare countries against each other over 6–12 month horizons — not a market-timing tool and not an economic health grade. A country in recession with a central bank easing aggressively, a cheap currency, and improving terms of trade can score high; a booming economy with rich valuations and tightening policy can score low. That is by design.
Each pillar is one economic idea, measured with one or a few series, expressed the same way: as a rolling z-score against the country's own 15-year history (minimum 5 years), clipped at ±3. Signs are set so that positive always means “supportive.”
| Inflation | Six-month change in the year-over-year CPI rate, inverted — falling inflation momentum is the tailwind. Why momentum, not level: markets price levels quickly; what changes the policy outlook is the direction of travel. Disinflation gives a central bank room to ease, and easing has historically been kind to equity multiples. |
| Policy | Twelve-month change in the central-bank policy rate, inverted — cutting is supportive, hiking is a headwind. The most direct read on whether monetary conditions are loosening or tightening. |
| FX valuation | The real effective exchange rate's distance from its own five-year average, inverted — a cheap currency is the tailwind. Cheap real exchange rates support competitiveness and, for a foreign investor, add a potential currency-appreciation kicker; our validation found this pillar earns its keep mostly through that currency leg. |
| External balance | The current account (four-quarter average, in USD) versus its own history. Surplus and improving external positions cushion a market against funding shocks and sudden capital-flow stops — the classic emerging-market failure mode. This pillar has the strongest statistical record in our validation. |
| Commodity terms of trade | Twelve-month change in the IMF's commodity net-export price index (CTOT), which weights world commodity-price moves by each country's commodity trade as a share of its economy. Rising commodity terms of trade are an income windfall for exporters and a squeeze for importers — the index measures the windfall directly, and a country that barely trades commodities correctly gets a flat, quiet series. |
| Valuation | Stock-market capitalization as a percentage of GDP (the per-country “Buffett indicator”), versus the country's own history, inverted — cheap is the tailwind. Annual World Bank data, so it moves slowly: this is the anchor pillar, with the lowest turnover of the six. |
Why compare each country to its own history rather than to other countries? Because levels are not comparable across markets — Japan's neutral policy rate, Brazil's normal inflation, and Taiwan's typical current account are different worlds. Z-scoring against a country's own past puts every pillar in the same units: “how unusual is this, for this country?”
Earlier versions of this model included a growth pillar (industrial production, leading indicators, unemployment). We tested it the same way we test everything — and it failed: countries with the strongest measured growth went on to underperform over the following 6–12 months, consistently enough that the pillar was removed from the composite in July 2026. The likely reason: visible growth is already in the price by the time statistical agencies report it. Growth readings still appear on the dashboards as context; they no longer influence the score. We tested a “growth acceleration” variant as a replacement, and it failed too. We would rather show you an honest four-line autopsy than a six-pillar story that doesn't survive contact with data.
The live pillars are averaged with equal weights, and a score is shown only when at least three pillars have data. The average is mapped through the normal CDF to 0–100, so the score reads like a percentile of the country's own historical range: 50 is typical, 80 is a strongly supportive backdrop, 20 is hostile.
Why equal weights? Not for lack of trying alternatives. We ran a walk-forward study in which pillar weights were re-estimated every year from realized predictive power, frozen, and applied out of sample — the disciplined version of “let the data pick the weights.” The learned weights failed to beat equal weights out of sample, so equal weights stayed. In factor modeling, robustness beats optimization more often than intuition suggests.
Before computing any predictive statistic, we froze a written specification: which metrics count, which robustness checks must pass, and what would qualify as validation, non-proof, or rejection. Every deviation since is a dated amendment in the same document. The tests: monthly cross-sectional rank correlations between scores and subsequent 3/6/12-month USD equity returns across ~40 countries of index history; quintile portfolios; subperiod, regional, and crisis-exclusion splits; and a check that results survive on returns of actual single-country ETFs — instruments an investor could really have traded, dividends and delistings included (they do, at essentially full strength). The composite and several pillars passed. Some candidates — a growth pillar, an inflation target-gap factor — failed and were rejected or left out. The two newest pillars (commodity terms of trade and valuation) were adopted only after clearing the same pre-registered bars.
Timing discipline. The validation uses each signal only with the publication lags a real-time investor would have faced — decisions at each month-end use data as it stood a month earlier, CPI gets an extra month, annual valuation data isn't used until the July after publication, and every series must have five years of history before it can produce a score. No country enters the history before its data actually existed.
Three caveats belong in the open. First, the historical validation runs on revised macro data — we simulate publication lags conservatively, but statistical agencies revise history, and no reconstruction fully removes that. True point-in-time evidence accumulates from mid-2026 onward as we archive data vintages. Second, validation statistics are not a live track record; they are evidence about the past, of exactly the kind that sometimes weakens out of sample. Third, the score is one input among many — it knows nothing about politics, earnings, or prices beyond what its six series carry. Updated nightly. For research and education only; not investment advice.