Methodology & data dictionary — EM Macro Watch

How the daily credibility score is built, followed by a complete reference of every data point behind it.

Last updated: 2026-07-31

How the score works

Every scored country gets a single composite credibility score from 0 to 100, where higher is better — a more credible, better-anchored monetary regime. The composite is a weighted average of seven dimensions, each itself scored 0–100 and each higher-is-better.

Because the composite blends seven dimensions, one perfect dimension does not make a perfect score. A country can post a credibility gap of 100 — its inflation is inside the central bank's target band (scoring/linex.py returns 100 within band) — while its composite sits only in the moderate range, held down by weaker growth, liquidity, or governance. That is by design, not a contradiction: the headline number is the blend, not any single dimension.

The seven dimensions

• Credibility gap — how far actual inflation sits from the central bank's target, scored with an asymmetric LINEX loss (overshoots are penalised harder than undershoots). Inflation inside the target band scores 100. This is the single heaviest dimension.
• Behind the curve — whether the policy rate is where a Taylor rule says it should be given inflation and the output gap (plus an FX-passthrough sub-signal). A rate far below the rule-implied level — a central bank not tightening enough — scores low.
• Communication — the hawkish/dovish stance read from central-bank statements, and the gap between what the bank says and what it does. Sounding hawkish while sitting on rates is a credibility problem.
• Geopolitical — pressure from the flow of political and conflict events (GDELT), scored through an economic-impact matrix keyed to each country's characteristics.
• Growth — the economy's cyclical position. Gold-tier countries use the OECD leading indicator, business confidence, and the change in unemployment; other countries use a GDP-growth percentile, deviation from the IMF forecast, and an unemployment percentile.
• Liquidity — financial vulnerability. Gold-tier countries use the credit-to-GDP gap, the debt-service ratio, and the real policy rate versus a per-country neutral rate; other countries use the real effective exchange rate, FX reserves, and external debt, plus a sovereign-stress overlay (government debt and the local-minus-US 10-year spread). Two additions layer on top where data allows: short-term external debt as a percentage of reserves (World Bank IDS), and a WEO-derived debt-sustainability gap — the primary balance a country is actually running versus the primary balance that would stabilize its debt ratio, given growth, inflation (capped at 10 percentage points in the calculation), and an interest rate. That interest rate is not directly observed, so we use the policy rate as a documented proxy (falling back to the 10-year sovereign yield when the policy rate is missing) — an approximation, not the IMF's own debt-sustainability analysis.
• Governance — institutional quality from the six World Bank Worldwide Governance Indicators (voice & accountability, political stability, government effectiveness, regulatory quality, rule of law, control of corruption), mapped to 0–100. A FATF grey- or black-listing then applies a flat subtractive penalty on top (−10 for grey, −25 for black, floored at 0) — a discrete external judgment, not a seventh WGI-style estimate averaged in with the rest.

Weights

The dimensions are combined with hand-set base weights:

Every country is scored on that same fixed weight vector. When we have no data for a dimension, it contributes a neutral 50 at its usual weight — its weight is not handed to the dimensions we do have.

That last point matters more than it sounds. Until methodology v3 the missing dimension's weight was redistributed across the survivors, which meant a country escaped being measured on its weak spots and had its strong ones amplified in exchange. Countries with no published policy rate and no English-language central-bank statement were rising up the table on that mechanic alone. A neutral fill makes absent data neither a reward nor a punishment, and keeps every country's score directly comparable. Use the coverage figure on each country page to see how much of a score is measured rather than filled.

Smoothing

The published score is not the raw daily number. It is 0.7 × today's raw composite + 0.3 × the trailing-90-day average, requiring at least 7 days of history. This damps one-day data jumps without lagging genuine trend shifts.

Coverage tiers

Data availability varies enormously across countries, so each sits in one of three tiers that set which inputs its growth and liquidity dimensions use:

• Gold (27 countries, including the synthetic GLOBAL aggregate) — those with an OECD Composite Leading Indicator. Richest inputs: CLI + business confidence for growth, BIS credit / debt-service / real-rate for liquidity.
• Silver (~53 countries) — inflation-targeting regimes without an OECD CLI. Percentile-based growth (GDP + IMF-forecast deviation + unemployment) and REER / reserves / external-debt liquidity.
• Bronze (~92 countries) — annual data only. These typically carry about 5 of the 7 dimensions as measured values, with the remainder taking the neutral fill. Their coverage is usually too low to clear the headline-ranking gate, so most sit in the monitored tier: fully scored and browsable, held out of the league table.

Two dimensions are structurally thin regardless of tier: communication covers only the 15 central banks whose statements are machine-retrievable (of 20 configured — see entry 6), and the OECD leading indicator is now published for only 13 countries.

What unlocks 169 scored countries

Monthly CPI is not available everywhere, so inflation follows a fallback chain: CPI_YOY → CPI_HEADLINE → WEO_INFLATION (the IMF's annual forecast). That annual fallback is what lets the platform score countries with no monthly price series at all, taking coverage to 169 scored countries.

Honest counts

Coverage figures on this page are stated as of 2026-07-30. Several step up on 2026-07-31, and where that is the case both numbers are given rather than the flattering one. The reason for the step is worth stating plainly: an audit on 2026-07-30 found that most of what looked like missing data was being discarded by our own request filters, not withheld by publishers. Monthly inflation, policy rates, business confidence and the leading indicator were all being requested for a fraction of the countries their sources actually serve.

• 169 scored countries; tiers 27 gold (incl. GLOBAL) / ~53 silver / ~92 bronze.
• Monthly inflation: 51 → ~148 countries. Everything else uses the annual fallback chain.
• Policy rate: 56 → 83 countries.
• Business confidence: 7 → 46 countries. Leading indicator: 13 → 17.
• Central-bank statements: 10 of 17 configured banks (see entry 6 for why seven are off).
• Economic Sentiment (Eurostat): 32 countries, first observations 2026-07-31.
• FX: European Central Bank reference rates (~30 currencies) plus 7 official central-bank feeds for EM currencies the ECB list omits — 61 countries.
• Sovereign yields: ~35 countries (after the Wave-1 expansion below).
• REER: 50 → 60 countries.

None of the above changed the methodology. The seven dimensions, their weights and their formulas are unchanged, so the published methodology version stays at v3 — this is a coverage change, and conflating the two would make the version number meaningless. What does change is how many countries are scored on measured data rather than on a neutral fill or an annual forecast.

Credibility scoring inputs (daily)

The rest of this page is the data dictionary: every input, where it comes from, and how it is used. These daily inputs feed the credibility-gap, behind-the-curve, communication, and geopolitical dimensions.

1. Consumer Price Index (CPI) — Inflation

What it is: Measures how much more expensive a basket of everyday goods (food, rent, fuel, etc.) has become compared to last year. If CPI inflation is 5%, things cost 5% more than a year ago.

Source (indirect): We don't collect prices ourselves. National statistics offices (e.g., Turkey's TurkStat, India's MoSPI) compile CPI from surveys of thousands of shops. They publish the data, and it flows through intermediaries:

• IMF IFS (International Financial Statistics) collects CPI from member country statistics offices. This is the primary monthly source, covering 164 countries — though see the coverage note below, because not all 164 are alive. Their server does go down; when it does, scoring continues on existing observations rather than stopping.
• FRED (Federal Reserve Bank of St. Louis) repackages OECD's Main Economic Indicators dataset — monthly CPI for up to 27 countries. Kept as a fallback only, and any FRED series whose upstream stopped updating more than 90 days ago is skipped rather than treated as current
• OECD publishes CPI directly from member countries — fills gaps for 5 countries (Colombia, Argentina, Bulgaria, Croatia, Saudi Arabia) not covered by FRED
• IMF Datamapper provides annual CPI as a fallback when no monthly source is available

Coverage: 51 countries as of 2026-07-30, rising to about 148 from 2026-07-31.

That jump is not a new data provider — it is a correction. We had been asking the IMF for a hardcoded list of 27 countries from a dataset that returns 190 economies, on the same URL, host and parser. The rest of the panel was falling back to the IMF's annual inflation forecast when a measured monthly series was available all along.

The same staleness rule as the policy rate applies, and it is what caps this at ~148 rather than 164: Venezuela's last monthly CPI print is 2016, Yemen's 2015, Syria's 2019. A decade-old index scored as current inflation is a fabricated measurement, whereas the annual fallback is at least an honest projection. Series more than 180 days stale are dropped and those countries stay on the fallback chain.

How we use it: We compute Year-over-Year inflation rate and 3-month annualized inflation from the monthly index. Then we compare it to the central bank's inflation target. The gap between actual inflation and target is the core input to the Credibility Gap sub-score. A CB with inflation of 40% and a target of 5% scores terribly. One with inflation of 2.1% and a target of 2% scores well.

Internal indicators: CPI_MONTHLY_INDEX, CPI_HEADLINE, CPI_YOY, CPI_3M3M_ANN

2. Policy Interest Rate

What it is: The interest rate set by the central bank — their main tool to fight inflation. When the CB raises rates, borrowing gets expensive, spending slows, and inflation should come down. This is the CB's main lever.

Source (indirect): Each central bank announces its rate after policy meetings. The Bank for International Settlements (BIS) in Basel collects these announcements from all member central banks and publishes them in a standardized SDMX format. We fetch from the BIS API — so the original source is each central bank, but we access it through BIS.

Coverage: 56 countries as of 2026-07-30, updated within days of rate decisions — rising to 83 as the addition below lands.

Second source (from 2026-07-31): BIS collects policy rates from its 38 reporting central banks, which leaves most of the panel without one. We now also read the IMF's Monetary & Financial Statistics, which adds ~27 more countries. BIS keeps precedence wherever it has a country, because it publishes daily where the IMF publishes monthly. A country's rate is taken from one source, never blended.

Two thirds of what the IMF serves is deliberately discarded: of the 122 countries it returns, 68 stopped reporting long ago — Afghanistan's last observation is 2021, Burundi's 2016, and the eight-member WAEMU bloc has shown the same 2.5% since 2017. A frozen rate is worse than no rate, because the behind-the-curve dimension would read it as a central bank deciding to hold when it is simply absent. Any series whose newest observation is more than 180 days old is dropped, and those countries score on a neutral fill instead.

How we use it: Core input to the Behind the Curve sub-score. We plug the policy rate into a Taylor Rule model: given current inflation and economic activity, what should the rate be? If the actual rate is much lower than the Taylor-implied rate, the CB is "behind the curve" — not tightening enough. We also compute the real interest rate (policy rate minus inflation) and track its historical percentile.

Internal indicator: POLICY_RATE

3. OECD Composite Leading Indicator (CLI)

What it is: An index designed to signal turning points in the business cycle 6-9 months ahead. 100 = on trend. Above 100 = economy expanding faster than trend. Below 100 = slowing. Built from components like order books, building permits, stock prices, and interest rate spreads — things that move before GDP does.

Source (indirect): The OECD compiles CLI from national data sources for each member/partner country. We fetch from the OECD SDMX API.

Coverage: 13 countries as of 2026-07-30, becoming 17 from 2026-07-31.

The OECD did narrow this indicator, but not as far as we long believed. An audit on 2026-07-30 found we were requesting a hardcoded list of 13 countries from a dataset that publishes 17, and then filtering the response through the same list of 13 — so China, Spain, France and Italy were being published and never asked for. The related business confidence survey turned out to be published for 46 countries while we were storing 7, for the same reason. Both are corrected; neither had been discontinued.

How we use it: Proxy for the "output gap" in the Taylor Rule. If CLI > 100 (economy running hot), the Taylor Rule says rates should be higher. If CLI < 100 (economy cooling), rates can be lower. Countries without CLI data get a simpler Taylor Rule that assumes zero output gap.

Internal indicator: OECD_CLI

4. FX Exchange Rates

What it is: How much one unit of each country's currency costs in US dollars. E.g., 1 USD = 32 Turkish Lira means TRY/USD = 32.

Source: The European Central Bank euro foreign-exchange reference rates — a free daily mid-market feed (~30 currencies including the major EM). We cross against USD to express every rate as currency-per-USD. (We moved off Open Exchange Rates in the 2026 free-data pivot — OXR was a paid/redistribution-restricted vendor; the ECB feed is free and openly licensed.)

Coverage: ~30 reference currencies covering the scored countries (EUR covers both Germany and Croatia; dollarized economies are pinned to 1.0).

How we use it: Three things:

1. FX returns — we compute 7-day and 30-day percentage depreciation. Shown on the dashboard as a market signal alongside credibility scores. (The 1-day return was removed: at daily ingest cadence it was mostly noise.)
2. FX passthrough signal — rapid currency depreciation makes imports more expensive, feeding inflation. This is a sub-signal in the Behind the Curve score.
3. Backtest validation — we test whether drops in credibility score predict subsequent currency weakness. If our score is useful, low-credibility countries should see their currencies underperform.

Internal indicators: FX_RATE_USD, FX_RETURN_7D, FX_RETURN_30D

5. Geopolitical Events (GDELT)

What it is: A database of every reported political and conflict event worldwide, extracted automatically from news articles in 100+ languages. Each event has: who did what to whom (CAMEO event taxonomy, ~300 types), how conflictual/cooperative it was (Goldstein scale, -10 to +10), how many news articles mentioned it, and the emotional tone of reporting.

Source (indirect): The GDELT Project processes ~300K news articles daily using NLP to extract structured events. The data is stored in Google BigQuery. We query it via BigQuery SQL with our own GCP service account. So the original source is global news media, processed by GDELT's NLP pipeline.

Coverage: Global — every country, every day. We fetch events for every country in our panel.

How we use it: We maintain an impact matrix (a JSON lookup table) that maps CAMEO event types to economic impacts based on country characteristics. For example:

• "Military buildup" (CAMEO 15) → high impact on countries that are net energy importers (could disrupt supply)
• "Trade sanctions" (CAMEO 163) → high impact on trade-dependent economies
• "Diplomatic praise" (CAMEO 05) → low impact, positive

Each event gets a weighted score based on its Goldstein scale, mention count (more mentions = bigger deal), and recency (recent events matter more). The average across all events for a country in the last 30 days produces the Geopolitical Pressure sub-score.

Internal indicator: GEOPOLITICAL_PRESSURE

6. Central Bank Policy Statements

What it is: The official text that central banks publish after monetary policy meetings. These statements signal the CB's thinking — are they worried about inflation (hawkish)? Or focused on growth (dovish)? Markets move on subtle word changes.

Source (direct): We read the central banks' own publications — no intermediary. Every active source is a machine-readable feed or API rather than a scraped web page: RSS for the Fed, Bank of England and the RBI; Atom for the CBRT; the ECB's press feed; and Banco Central do Brasil's site API, whose COPOM minutes exist only as PDF (its statement web pages are a JavaScript shell), so we read the PDF's text layer.

Coverage: 15 central banks, from 20 configured — and the gap is the honest part of this entry.

Currently delivering: Turkey (CBRT), Brazil (BCB), India (RBI), the US Federal Reserve, the ECB (whose stance is applied to every euro-area country), the Bank of England, Sveriges Riksbank, the Swiss National Bank, Norges Bank, the Reserve Bank of Australia, the South African Reserve Bank, the Magyar Nemzeti Bank, the Bank of Canada, the Czech National Bank and the Bank of Israel.

Still off, and not worked around: Chile, Mexico, Poland, Colombia and Indonesia. Four of those refuse automated access through commercial bot management or a registration wall, and Indonesia publishes its releases in an alphabetical list with no chronological index — which is how a 2010 press release once ended up stored as the current one.

Those last four were added on 2026-07-30. Five other candidates were rejected in the same pass despite looking available, because reading them end-to-end showed what they actually return: one served a notice announcing next year's meeting schedule rather than a decision, one returned a genuine statement from 2015 because its archive feed is not ordered newest-first, and one returned a rates page instead of the decision. A source is only enabled here once its extracted text has been read back and confirmed to be a monetary policy decision.

Configured but switched off, because their statements cannot be retrieved: South Africa, Indonesia, Mexico, Chile, Poland, Hungary, Colombia. The reasons split three ways — sites that render their statement list in the browser and publish no feed (SARB, BCCh), a registration wall (MNB), and commercial bot management that refuses automated access outright (NBP, BanRep, Banxico). We do not work around bot protection; a bank that declines automated access is entitled to that, so those countries stay off until an official feed or API exists.

This matters because of what the alternative was. Until 2026-07-30 all thirteen were nominally "covered", and the scraper returned something for most of them — a login form, a press-release index headlining Treasury-bill auction results, a 2010 release about banking crime — which was then read as monetary-policy stance. Six banks that genuinely publish machine-readable statements is a smaller number and a truthful one. Where a bank is off, the country carries no communication score and the dimension contributes a neutral 50 at its 0.10 weight, per the neutral-fill rule above.

How we use it: Each statement is read by Claude Haiku (via a LiteLLM proxy), which returns a hawk/dove score from −1 (most dovish) to +1 (most hawkish), the concerns the bank highlighted, and its forward guidance. That produces the Communication Stance sub-score. We also compute a "stance-vs-action gap" — if a bank sounds hawkish but hasn't been raising rates, that's a credibility problem.

Two corrections worth stating plainly, because earlier versions of this page said otherwise:

• There is no longer a second model. This page previously described a 60/40 blend of a fine-tuned transformer and the LLM. The transformer (ProsusAI/finbert) is a financial sentiment classifier, not a monetary-stance model, and in production it returned near-constant values per country — 0.60 on every Brazilian statement, 0.00 on every Turkish one — which is not a signal. Its weight was set to zero and its daily inference step removed. The LLM is the sole scorer.
• A statement now has to look like a statement before it is scored. Pages shorter than 1,000 characters, and pages carrying a bot-detection or JavaScript-required notice, are skipped instead of scored. Before that check existed, a substantial share of the stored "statements" were scraped error pages being read as monetary policy.

Internal indicator: COMMUNICATION_STANCE

7. Country LLM Summaries

What it is: A daily prose summary for each scored country — an executive summary (2-3 sentences for a PM) and an analyst brief (longer, for research notes).

Source (direct): Generated by Claude Sonnet via LiteLLM. The prompt includes the country's current scores, recent events, and latest CB statement. So the source is our own LLM generation, grounded in our data.

Coverage: Gold- and silver-tier countries (~81), generated at weekends rather than daily — the summaries are prose context, not a scored input, so they don't warrant a daily LLM bill.

How we use it: Displayed on the country detail page. Not used in scoring — purely for human consumption. Gives context that numbers alone can't convey.

DB table: country_summaries

Macro fundamentals (annual)

Annual structural indicators for ~171 countries. Several now feed the composite directly — GDP, unemployment, and the IMF forecast drive the growth dimension; government and external debt, reserves, and REER drive liquidity; and the World Bank governance estimates (entry 34) drive the governance dimension. The remainder are shown as country context on the country and coverage pages.

8. GDP Growth (annual %)

What it is: How much bigger (or smaller) the entire economy got compared to last year, in real terms (adjusted for inflation). 3% growth is healthy for an EM. Negative means recession.

Source (indirect):

• World Bank WDI — compiles from national accounts data published by each country's statistics office. Annual, covers 171 countries. Typically 6-18 months lag.
• IMF WEO — the IMF's own estimates + forecasts. Published twice yearly (April and October). Covers 169 countries, includes 5-year forward projections. So "2027 GDP growth" from WEO is a forecast, not a measurement.

How we use it (today): Displayed on the Macro Data dashboard page. Users can compare GDP growth across all 171 countries, filter by region, sort by value.

How we'll use it (future): Core input to the Growth Fragility vulnerability index (Phase 5). Countries with declining or volatile GDP growth are more fragile.

Internal indicators: GDP_GROWTH_ANNUAL, WEO_GDP_GROWTH

9. GDP per Capita (USD)

What it is: Total economic output divided by population. A rough measure of average living standards. Luxembourg at $130K vs Burundi at $230 tells you very different stories.

Source (indirect): World Bank WDI, which gets it from national accounts (GDP) and UN population estimates. 171 countries.

How we use it: Context for comparisons. A 5% GDP growth in India means something different than 5% in Switzerland. Not directly scored.

Internal indicator: GDP_PER_CAPITA_USD

10. Unemployment Rate

What it is: The percentage of people who want to work but can't find a job, out of the total labor force.

Source (indirect):

• World Bank WDI (170 countries) — compiles from ILO modeled estimates and national labor force surveys
• ILO ILOSTAT (167 countries) — the International Labour Organization's own estimates, harmonized across countries for comparability
• IMF WEO (107 countries) — IMF staff estimates + forecasts

All three are modeled/estimated — true unemployment is hard to measure, especially in developing countries with large informal sectors.

How we use it (future): Input to Growth Fragility index. Rising unemployment signals economic weakness and social pressure on the CB to cut rates (even if inflation is high).

Internal indicators: UNEMPLOYMENT_RATE, UNEMPLOYMENT_RATE_ILO, WEO_UNEMPLOYMENT

11. Labor Force Participation Rate

What it is: What percentage of working-age people (15+) are either employed or actively looking for work. If only 50% participate, half the working-age population is neither working nor job-hunting (students, retirees, discouraged workers, homemakers).

Source (indirect): ILO ILOSTAT (166 countries), based on modeled estimates from national labor force surveys.

How we use it: Context indicator. A country can have low unemployment but also low participation — meaning many people have given up looking. True labor market health needs both numbers.

Internal indicator: LABOR_FORCE_PARTICIPATION

12. Trade Openness (% of GDP)

What it is: (Total Exports + Total Imports) / GDP. Measures how intertwined a country's economy is with the rest of the world. Singapore at ~300% is extremely open. Brazil at ~30% is relatively closed.

Source (indirect): World Bank WDI (163 countries), compiled from balance-of-payments data reported by national central banks to the IMF.

How we use it (future): Key input to External Vulnerability. Highly open economies are more exposed to global trade shocks — a worldwide recession hits them harder.

Internal indicator: TRADE_OPENNESS

13. Current Account Balance (% of GDP)

What it is: The broadest measure of a country's international transactions. Includes trade in goods and services, investment income, and transfers. Negative means the country is spending more internationally than it earns — it needs foreign capital to fill the gap.

Source (indirect):

• World Bank WDI (165 countries) — from IMF Balance of Payments data
• IMF WEO (169 countries) — IMF staff estimates + forecasts

How we use it (future): Critical input to External Vulnerability. A country running a large current account deficit (say, -8% of GDP) depends heavily on foreign investors continuing to lend. If confidence drops, capital flees, the currency collapses, and inflation spikes. This is the classic EM crisis pattern (Turkey 2018, Argentina repeatedly).

Internal indicators: CURRENT_ACCOUNT_PCT_GDP, WEO_CURRENT_ACCOUNT_PCT_GDP

14. Government Debt (% of GDP)

What it is: How much the government owes in total, relative to the size of the economy. Japan at 250% is an outlier. Below 60% is generally considered manageable for EMs.

Source (indirect):

• World Bank WDI (77 countries only — sparse coverage)
• IMF WEO (167 countries — much better, includes forecasts)

The two sources can differ because they use different debt definitions (gross vs net, central govt vs general govt).

How we use it (future): Core input to Fiscal Vulnerability index. High debt means the government has less room to stimulate during downturns, pays more interest, and faces higher default risk. Markets price this into bond yields and CDS spreads.

Internal indicators: GOVT_DEBT_TO_GDP, WEO_GOVT_DEBT_PCT_GDP

15. External Debt (% of GNI)

What it is: How much the country (government + private sector) owes to foreign creditors, relative to national income. Unlike government debt, this includes corporate and bank borrowing from abroad.

Source (indirect): World Bank WDI (111 countries), compiled from the World Bank's Debtor Reporting System and IMF data. Many high-income countries don't report to this system, hence the coverage gap.

How we use it (future): Key input to External Vulnerability. If external debt is high and denominated in foreign currency (usually USD), a currency depreciation makes the debt harder to service. This is the "original sin" problem — borrowing in dollars when you earn in pesos.

Internal indicator: EXTERNAL_DEBT_TO_GNI

16. FDI Inflows (% of GDP)

What it is: Foreign Direct Investment — when a foreign company builds a factory, acquires a local firm, or makes a long-term capital commitment. Unlike portfolio flows (buying stocks/bonds), FDI is "sticky" — it doesn't flee overnight.

Source (indirect): World Bank WDI (170 countries), from IMF Balance of Payments and UNCTAD data.

How we use it: Positive signal — countries attracting FDI are seen as stable and growing. But high FDI dependence also means the economy relies on foreign confidence. Displayed in macro dashboard.

Internal indicator: FDI_PCT_GDP

17. Remittances (% of GDP)

What it is: Money sent home by citizens working abroad. For some countries (Nepal ~25%, El Salvador ~24%), this is a huge part of national income.

Source (indirect): World Bank WDI (163 countries), estimated from central bank and IMF data on cross-border personal transfers.

How we use it (future): Countries heavily dependent on remittances are vulnerable to economic downturns in the host countries where their diaspora works. A US recession hits Central American remittance flows hard.

Internal indicator: REMITTANCES_PCT_GDP

18. Foreign Reserves (Months of Imports)

What it is: How many months the central bank's foreign currency reserves could pay for the country's imports, if all other income stopped. The IMF considers 3 months the minimum safety level.

Source (indirect): World Bank WDI (157 countries), computed from IMF reserve data and import volumes.

How we use it (future): Critical External Vulnerability signal. A country with 1 month of reserves can't defend its currency if capital flees. One with 12 months can weather a crisis. This is the first thing investors check during EM stress.

Internal indicator: RESERVES_MONTHS_IMPORTS

19. IMF Inflation Forecast

What it is: The IMF's projection for consumer price inflation in each country, including 5-year forward estimates.

Source (indirect): IMF WEO (169 countries). Published April and October.

How we use it (future): Forward-looking complement to actual CPI. If the IMF expects inflation to stay elevated, the credibility gap may persist even if the CB is tightening.

Internal indicator: WEO_INFLATION

Global context & financial indicators

Higher-frequency daily / monthly / quarterly series that sit alongside the annual fundamentals — global risk gauges, commodity prices, sovereign yields, and BIS / IMF financial-stability statistics.

20. VIX (Volatility Index)

What it is: The CBOE Volatility Index — often called the "fear gauge." Measures expected stock market volatility over the next 30 days. When VIX spikes above 30, markets are panicking. Below 15 = calm.

Source (indirect): FRED (VIXCLS). CBOE computes it from S&P 500 options prices. Daily.

How we use it (future): Global risk appetite proxy. When VIX spikes, capital flees EM. Every country's vulnerability score should incorporate global risk conditions.

Internal indicator: VIX (stored as country_iso = GLOBAL)

21. Crude Oil Prices (WTI + Brent)

What it is: Spot price of crude oil in USD/barrel. WTI (West Texas Intermediate) is the US benchmark; Brent is the international benchmark. Oil is the single most important commodity for macro analysis.

Source (indirect): FRED (DCOILWTICO for WTI, DCOILBRENTEU for Brent). Daily.

How we use it (future): Commodity shock input. Oil importers (India, Turkey) suffer when oil rises; exporters (Saudi Arabia, Russia) benefit. The impact engine will propagate oil shocks through country-specific import/export exposures.

Internal indicators: WTI_CRUDE_OIL, BRENT_CRUDE_OIL (stored as country_iso = GLOBAL)

22. USD Index (DXY)

What it is: Trade-weighted USD index — measures the dollar's strength against a basket of major currencies. When DXY rises, the dollar is strengthening.

Source (indirect): FRED (DTWEXBGS). Federal Reserve Board computes it. Daily.

How we use it (future): Dollar strength drives EM FX pressure. A strong dollar means EM countries pay more (in local currency) for dollar-denominated imports and debt service. Critical input for external vulnerability scoring.

Internal indicator: USD_INDEX_DXY (stored as country_iso = GLOBAL)

23. Gold Price

What it is: London AM gold fixing price in USD per troy ounce. Gold is both a commodity and a safe-haven asset.

Source (indirect): FRED (GOLDAMGBD228NLBM). ICE Benchmark Administration sets the fixing. Daily.

How we use it (future): Safe-haven demand indicator. Gold spikes during crises. Also relevant for gold-exporting countries (South Africa, Peru, Ghana).

Internal indicator: GOLD_PRICE (stored as country_iso = GLOBAL)

24. Copper Price

What it is: Global copper price in USD per metric ton. Copper is called "Dr. Copper" because its demand tracks industrial activity — it's in wiring, plumbing, electronics, construction.

Source (indirect): FRED (PCOPPUSDM). World Bank commodity data. Monthly.

How we use it (future): Industrial activity proxy. When copper prices drop, global manufacturing is slowing. Especially relevant for copper exporters (Chile, Peru, Zambia).

Internal indicator: COPPER_PRICE (stored as country_iso = GLOBAL)

25. Sovereign Bond Yield (10-Year)

What it is: The annual interest rate a government pays to borrow money for 10 years. This is the market's real-time verdict on country risk. Higher yield = market demands more compensation for lending to that country.

Source (indirect): FRED (IRLTLT01 series from OECD). Monthly. ~35 countries (expanded from the original 15 in the Wave-1 work below; dead series were dropped and live ones added).

How we use it (future): The single most important signal for institutional users. Yield spreads (country yield minus US yield) are the market's risk premium for each country. A widening spread means investors are getting nervous.

Internal indicator: SOVEREIGN_YIELD_10Y

26. Credit to GDP Ratio

What it is: Total credit to the private non-financial sector as a percentage of GDP. Measures how leveraged the private sector is relative to the economy's size. High and rising credit-to-GDP is the best known early-warning of financial crises (per BIS research).

Source (indirect): BIS (WS_TC_PRIV dataflow). Quarterly. ~45 countries.

How we use it (future): Core input for Fiscal Vulnerability and Growth Fragility indices. A credit-to-GDP ratio above its long-term trend (the "credit gap") has historically preceded banking crises by 1-3 years.

Internal indicator: CREDIT_TO_GDP

27. Debt Service Ratio

What it is: How much of a country's income goes to servicing debt (interest + principal payments). A ratio above 20% is considered dangerous — the country is spending so much on debt that it can't invest or absorb shocks.

Source (indirect): BIS (WS_DSR dataflow). Quarterly. ~30 countries.

How we use it (future): Crisis precursor. Countries with high debt service ratios are fragile — any interest rate increase or income shock can push them into distress.

Internal indicator: DEBT_SERVICE_RATIO

28. Real Effective Exchange Rate (REER)

What it is: A trade-weighted exchange rate adjusted for inflation differences between countries. An index above 100 means the currency is "overvalued" relative to its trading partners — the country's goods are expensive, hurting competitiveness.

Source (indirect): BIS (WS_EER dataflow). Monthly. ~60 countries.

How we use it (future): Competitiveness measure. An overvalued REER signals devaluation risk. Countries with persistently overvalued REERs tend to experience sudden corrections (currency crises).

Internal indicator: REER

29. Current Account Balance (Quarterly)

What it is: The sum of a country's trade balance, net income from abroad, and net transfers — measured quarterly in USD. A negative current account means the country is spending more abroad than it earns, requiring capital inflows to finance the gap.

Source (indirect): IMF IFS (BCA_BP6_USD). Quarterly. ~80 countries.

How we use it (future): Core External Vulnerability signal. Countries with large current account deficits depend on foreign capital. When global risk appetite drops (VIX spikes), these countries are most vulnerable to "sudden stops."

Internal indicator: BOP_CURRENT_ACCOUNT

30. Reserve Assets, Net Acquisition (Quarterly)

What it is: The change in reserve assets over the quarter — a signed balance-of-payments transaction, in USD. Positive = the central bank added to reserves that quarter; negative = it ran them down. This is a flow, not the reserve stock. Measured 2026-07-30: TUR −42.05bn, BRA +4.44bn, IND +7.22bn, against a Turkish reserve level of ~$210bn.

Source (indirect): IMF IFS balance-of-payments, .A_T.R_F.USD.Q. Quarterly, ~148 reporting economies. (The monthly RAFA_USD level series this entry once described no longer exists upstream.)

How we use it: Not consumed by any score. Exposed for inspection only.

Do not use it as a reserves level or adequacy measure. The reserves-adequacy quantity the liquidity dimension consumes is #18 RESERVES_MONTHS_IMPORTS (World Bank WDI FI.RES.TOTL.MO, months of import cover, 172 countries).

Internal indicator: BOP_RESERVE_ASSETS_FLOW (renamed from BOP_RESERVES in #134)

31. Trade Balance (Quarterly)

What it is: Exports minus imports of goods and services, measured quarterly in USD. Positive = trade surplus (exporting more than importing).

Source (indirect): IMF IFS (BGS_BP6_USD). Quarterly. ~80 countries.

How we use it (future): Input to trade channel in the impact engine. Countries with trade deficits are net importers — they're vulnerable to commodity price shocks and currency depreciation.

Internal indicator: BOP_TRADE_BALANCE

32. Net Portfolio Investment Flows (Quarterly)

What it is: Net flows of portfolio investment (stocks + bonds) into and out of a country, measured quarterly in USD. Positive = foreign investors buying more of the country's assets than they're selling.

Source (indirect): IMF IFS (BFPF_BP6_USD). Quarterly. ~60 countries.

How we use it (future): Investor confidence signal. When portfolio flows turn sharply negative, foreign investors are pulling money out — a precursor to currency pressure and rate hikes.

Internal indicator: BOP_PORTFOLIO_FLOWS

33. Bilateral Trade Flows (10-Year History)

What it is: Annual bilateral trade data — which country exports/imports how much to/from which partner, by commodity group. Now with 10 years of history (2014-present) instead of just the latest year.

Source (indirect): UN Comtrade (200+ country pairs). Annual. Free preview endpoint.

How we use it (future): Foundation for the trade linkage matrix — who depends on whom for trade. "If China's imports drop 10%, who loses the most?"

Internal indicators: TRADE_EXPORT_TOTAL_USD, TRADE_IMPORT_TOTAL_USD

34. Worldwide Governance Indicators (WGI) — Institutional Quality

What it is: Six World Bank governance estimates — Voice & Accountability, Political Stability & Absence of Violence, Government Effectiveness, Regulatory Quality, Rule of Law, and Control of Corruption. Each runs roughly −2.5 (weak) to +2.5 (strong) and reflects the institutional environment a central bank operates within.

Source: World Bank WGI (~207 countries, annual).

How we use it: The six estimates are mapped to a 0–100 scale and averaged into a Governance score — the seventh composite dimension (w7 = 0.11). Institutional quality is slow-moving but matters: a credibility regime sits on top of the rule of law, policy independence, and corruption control that WGI measures. High-governance economies (e.g. Norway, Germany) score in the high 70s–80s; weak-institution economies score in the teens–low 30s. A country on the FATF grey or black list then gets a flat penalty subtracted from that average — 10 points for grey, 25 for black, floored at 0 — reflecting a discrete external judgment on financial-integrity controls rather than a seventh number averaged into the WGI blend. A FATF flag with no WGI data at all does not by itself produce a governance score.

Internal indicators: WGI_VOICE_ACCOUNTABILITY, WGI_POLITICAL_STABILITY, WGI_GOVT_EFFECTIVENESS, WGI_REGULATORY_QUALITY, WGI_RULE_OF_LAW, WGI_CONTROL_CORRUPTION, FATF_STATUS, GOVERNANCE_SCORE

Coverage summary

Total: 47 indicator groups across 169 scored countries + global context, 18+ data sources.

Data Expansion Wave 1 — Fast pulse signals (2026-07)

New free, keyless sources that add faster per-country and global signals on top of the slow annual fundamentals above. All were verified free and programmatically accessible on 2026-07-23. Attribution/citation lines below are required wherever the data is displayed.

35. Port trade activity (PortWatch)

What it is: Daily satellite-AIS estimates of import/export tonnage and port-call counts per country, plus chokepoint transit volumes. We derive Port trade momentum (30d, YoY) — trailing 30-day trade tonnage versus the same window a year earlier — which is robust to the AIS coverage drift the PortWatch FAQ warns against (always year-over-year on windows, never raw levels).

Source: IMF PortWatch (satellite AIS via Kpler + the UN Global Platform), ArcGIS FeatureServer. Country-day aggregates, ~180 coastal countries, 2019→, ~5-day lag, refreshed weekly.

Cadence: daily.

License / attribution (REQUIRED verbatim citation): Sources: Kpler; UN Global Platform; IMF PortWatch (portwatch.imf.org). IMF terms are restrictive — raw tonnage series are derived-only (never on the customer API); only the derived momentum score is shown.

Internal indicators: PORT_IMPORT_TONS, PORT_EXPORT_TONS, PORT_CALLS (raw, restricted), TRADE_MOMENTUM_30D (derived, shown).

36. Commodity terms of trade (IMF CTOT)

What it is: A net-export commodity price index weighted by each country's commodity exposure relative to GDP — pre-computed by the IMF. We derive Commodity terms-of-trade, 12m change; a positive value is a terms-of-trade windfall (export prices rising faster than import prices).

Source: IMF Research (CTOT dataset) via the new api.imf.org SDMX 2.1 platform. 182 countries, 1962→, ~2-month lag.

Cadence: monthly.

License / attribution: IMF terms (restrictive). Raw CTOT index is derived-only; only the derived 12-month shock is shown. Cite IMF.

Internal indicators: CTOT_NX_GDP (raw, restricted), CTOT_SHOCK_12M (derived, shown).

37. Electricity demand (Ember)

What it is: Monthly electricity demand in TWh and its year-over-year percentage change per country — a fast, high-frequency proxy for real economic activity.

Source: Ember (monthly electricity data). 87 countries with strong EM overlap, 2018→, ~7-week lag.

Cadence: monthly.

License / attribution: CC BY 4.0 — commercial use permitted with attribution. Credit: Source: Ember (CC BY 4.0).

Internal indicators: ELECTRICITY_DEMAND_TWH, ELECTRICITY_DEMAND_YOY.

38. Geopolitical Risk index (GPR)

What it is: The Caldara–Iacoviello Geopolitical Risk index — a newspaper-based measure of adverse geopolitical events and threats, per country and globally.

Source: Dario Caldara & Matteo Iacoviello, "Measuring Geopolitical Risk," American Economic Review 2022. 44 countries monthly (1985→) plus a global daily series.

Cadence: monthly (country) + daily (global).

License / attribution: explicit CC BY (clean) — cite Caldara & Iacoviello (2022).

Internal indicators: GPR_COUNTRY (monthly per country), GPR_GLOBAL_DAILY (global daily).

39. Economic Policy Uncertainty (EPU)

What it is: The Baker-Bloom-Davis Economic Policy Uncertainty index — a newspaper-based measure of policy-related economic uncertainty, per country and a global aggregate.

Source: Baker, Bloom & Davis, policyuncertainty.com. 20 countries monthly (1985→) + a global EPU series.

Cadence: monthly.

License / attribution: CC BY 4.0 (clean) — cite Baker, Bloom & Davis.

Internal indicators: EPU_INDEX (per country), GLOBAL_EPU (global).

40. World Uncertainty Index (WUI)

What it is: The Ahir-Bloom-Furceri World Uncertainty Index — a text-based measure of economic and political uncertainty derived from Economist Intelligence Unit country reports.

Source: Ahir, Bloom & Furceri, worlduncertaintyindex.com. 143 countries quarterly (plus 71 monthly, 2008→).

Cadence: quarterly.

License / attribution: no explicit open license — treated as derived-only (like BIS/IMF), no raw passthrough on the customer API. Cite Ahir, Bloom & Furceri.

Internal indicator: WUI_QUARTERLY.

41. FAO Food Price Index (FFPI)

What it is: The FAO Food Price Index — a monthly global measure of international food commodity prices, with five sub-indices (meat, dairy, cereals, oils, sugar).

Source: Food and Agriculture Organization of the United Nations (FAO). Global, 1990→, ~3-week lag.

Cadence: monthly.

License / attribution: CC BY 4.0 — cite FAO. Food Price Index. Food and Agriculture Organization of the United Nations.

Internal indicator: FOOD_PRICE_INDEX (stored as GLOBAL).

42. Expanded global context & yields (FRED / OECD)

Additional clean-license gauges, all stored as GLOBAL unless noted:

• EM equity volatility EM_VOLATILITY (FRED VXEEMCLS, daily)
• Inflation breakevens INFLATION_BREAKEVEN_5Y, INFLATION_BREAKEVEN_10Y, INFLATION_5Y5Y_FWD (FRED, daily)
• Financial conditions / stress FINANCIAL_CONDITIONS_NFCI, FINANCIAL_STRESS_STLFSI (FRED, weekly)
• Commodity option vol OIL_VOLATILITY (FRED OVXCLS), GOLD_VOLATILITY (FRED GVZCLS), daily
• Business confidence BUSINESS_CONFIDENCE (OECD Key Economic Indicators, monthly; ~13 gold-tier countries) — resurrects the gold-tier growth path
• Sovereign yields SOVEREIGN_YIELD_10Y — expanded from 15 to ~35 countries (dead FRED series removed, live IRLTLT01* series added)

Cadence: daily to monthly. License / attribution: FRED and OECD are clean-license — no restriction gate. All FRED series cite the Federal Reserve Bank of St. Louis (FRED); CBOE/academic series retain their upstream provider credit.

Data Expansion Wave 2 — Activity & structural signals (2026-07)

Two more free, keyless, CC BY 4.0 sources: Eurostat monthly activity indices (a faster real-activity signal for Europe incl. Turkey and the CEE that Wave 1 missed) and World Bank / NASA Black Marble nighttime lights (a batch-published structural activity series). Attribution lines below are required wherever the data is displayed.

43. Industrial production (Eurostat)

What it is: Monthly industrial-production volume index (2021 = 100, seasonally + calendar adjusted) and its year-over-year percentage change per country — the standard monthly proxy for real activity in industry. This is a scored input: industrial-production YoY joins Ember electricity and PortWatch trade momentum as a third monthly pulse in the growth dimension.

Source: Eurostat short-term business statistics (sts_inpr_m, indic_bt=PRD, nace_r2=B-D), keyless JSON-stat API. ~37 European countries incl. TUR, UKR and silver-tier PL/HU/CZ/RO/BG/HR/RS, ~2-month lag.

Cadence: monthly.

License / attribution: CC BY 4.0 — commercial use permitted with attribution. Credit: © European Union, Eurostat, licensed under CC BY 4.0.

Internal indicators: INDUSTRIAL_PRODUCTION_IDX, INDUSTRIAL_PRODUCTION_YOY.

44. Retail turnover (Eurostat)

What it is: Monthly retail sales-volume index (2021 = 100, deflated = real activity; the nominal NETTUR variant is not used) and its year-over-year percentage change. Ingested and exposed via the API but not scored (industrial production is the activity proxy).

Source: Eurostat short-term business statistics (sts_trtu_m, indic_bt=VOL_SLS, nace_r2=G47), keyless JSON-stat API. ~36 European countries, ~2-month lag.

Cadence: monthly.

License / attribution: CC BY 4.0 — credit © European Union, Eurostat, licensed under CC BY 4.0.

Internal indicators: RETAIL_TURNOVER_IDX, RETAIL_TURNOVER_YOY.

45. Nighttime lights (World Bank / NASA Black Marble)

What it is: Monthly nighttime-lights radiance per country — the sum of gas-flaring-masked VIIRS radiance (ntl_nogf_5km_sum) over each country's ADM2 units, plus its 12-month change. A satellite-measured structural proxy for real economic activity, independent of every other feed. Not scored (deliberate): the file is batch-published with a ~7-month lag, so it lands as a structural / backtest series (like the annual WB/WGI feeds), not a live pulse.

Source: World Bank Space2Stats dataset 0066940, derived from NASA Black Marble VIIRS. Worldwide (all 169 scored countries incl. bronze tier), monthly 2012→2025, ~7-month publication lag.

Cadence: monthly (batch-published; ingested one year-file per run on a monthly schedule).

License / attribution: CC BY 4.0 — commercial use permitted with attribution. Credit: World Bank via NASA Black Marble VIIRS, CC BY 4.0.

Internal indicators: NIGHTLIGHTS_RADIANCE (raw radiance sum), NIGHTLIGHTS_YOY (12-month % change, derived).

47. Economic Sentiment Indicator (Eurostat)

What it is: The European Commission's Economic Sentiment Indicator — a composite of business and consumer survey balances across industry, services, retail, construction and households, indexed so 100 is the long-run average. It is a purpose-built leading indicator, which makes it the closest free substitute for the OECD leading indicator that is no longer published broadly.

Source: Eurostat Business & Consumer Surveys (ei_bssi_m_r2), keyless JSON-stat API. 32 countries, including Turkey, Serbia, North Macedonia, Montenegro and Albania. Monthly, ~1-month lag — when added on 2026-07-30 its newest observation was one month ahead of every other activity series we hold.

Cadence: monthly. First observations land 2026-07-31.

Licence / attribution: CC BY 4.0 — commercial use and redistribution permitted with attribution. Credit: © European Union, Eurostat, licensed under CC BY 4.0. This matters beyond compliance: it is the first confidence-type series we hold that we are permitted to republish raw, rather than only as a derived score.

Not yet scored. It is stored and exposed, but deliberately not merged into the business-confidence input, because the OECD and Eurostat indices are differently constructed — switching a country's source midway would put a discontinuity inside its own history.

Internal indicators: ECONOMIC_SENTIMENT, ECONOMIC_SENTIMENT_YOY.

Global gauges

46. Global market-context gauges (displayed on the /global page)

What it is: The complete set of market-wide, GLOBAL-scoped series shown together on the public /global page and summarised in the site's global-context strip. These are already documented individually above (entries 20–24 and 41–42); this entry is the single reference for exactly what is displayed and under what attribution.

Displayed series, by group:

• Volatility & financial conditions — VIX (VIX), EM equity volatility (EM_VOLATILITY), oil volatility / OVX (OIL_VOLATILITY), gold volatility / GVZ (GOLD_VOLATILITY), the US dollar index (USD_INDEX_DXY), Chicago Fed financial conditions / NFCI (FINANCIAL_CONDITIONS_NFCI), and St. Louis Fed financial stress / STLFSI (FINANCIAL_STRESS_STLFSI).
• Commodities — WTI crude (WTI_CRUDE_OIL), Brent crude (BRENT_CRUDE_OIL), gold (GOLD_PRICE), copper (COPPER_PRICE), FAO food price index (FOOD_PRICE_INDEX).
• Inflation expectations — 5-year, 10-year and 5y5y-forward TIPS breakevens (INFLATION_BREAKEVEN_5Y, INFLATION_BREAKEVEN_10Y, INFLATION_5Y5Y_FWD).
• Uncertainty & risk — global economic policy uncertainty (GLOBAL_EPU), geopolitical risk monthly and daily (GPR, GPR_GLOBAL_DAILY).
• Shipping chokepoints — the derived 30-day-transits-vs-1-year-baseline stress for Suez, Panama, Hormuz, Bab el-Mandeb and Malacca (CHOKEPOINT_STRESS_*). The raw PortWatch transit counts are restricted (derived-only); only the derived stress percentage is shown.

License / attribution (required wherever displayed):

• Volatility, financial-conditions, commodity and inflation-breakeven series: FRED (Federal Reserve Bank of St. Louis); VIX/OVX/GVZ retain CBOE provenance.
• Food price index: FAO (CC BY 4.0).
• Economic policy uncertainty: Baker, Bloom & Davis, policyuncertainty.com (CC BY 4.0).
• Geopolitical risk: Caldara & Iacoviello (2022) (CC BY).
• Chokepoint stress: Sources: Kpler; UN Global Platform; IMF PortWatch (portwatch.imf.org).