Methodology
Raw data pulled from 6 institutional sources via bulk download APIs. Each source has a dedicated ETL adapter that handles schema differences, missing values, and encoding.
All values standardized to common units. Currency values converted to current USD. Population-dependent metrics normalized per capita. Temporal alignment to calendar year.
Year-over-year changes (absolute & percentage), global rankings per indicator per year, moving averages, and volatility indices computed for every country-indicator pair.
5 pattern templates scanned across all country-year observations. 1,597 historical matches identified with outcome tracking at 12–24 month horizons.
Processed data stored in SQLite for sub-millisecond reads. No external database dependency. Full dataset ships with the application — zero API latency for end users.
| Source | Type | Coverage | Series |
|---|---|---|---|
| World Bank WDI | Macro indicators | 1960–2026 | 100+ |
| IMF WEO | Projections | 2025–2030 | 47,778 |
| World Bank CMO | Commodity prices | 1960–present | 71 |
| FRED | Indices, FX, Crypto | Varies | 15+ |
| ECB | Exchange rates | 1999–present | 22 |
| BIS | REER indices | 1964–present | 60+ |
Patterns are multi-condition templates applied across all country-year observations. Each pattern defines threshold conditions on 3–6 indicators simultaneously. When all conditions are met, a match is recorded and the outcome is tracked over a 12–24 month horizon.
Pattern-based projections are descriptive extrapolations from historical precedent. They are not econometric forecasts. Confidence intervals are derived from the variance of historical outcomes. Past outcomes do not guarantee future results.
HistorySaid displays two types of projections, clearly labelled in data tables:
- IMF — Official forecasts from the IMF World Economic Outlook. Covers GDP growth, inflation, unemployment, current account, gross savings, government debt, budget deficit, and government revenue.
- Trend — Linear extrapolations computed from the last 5 available data points. These are simple directional estimates, not econometric forecasts. Confidence intervals represent ±1 standard deviation of the regression residuals. Trend projections should not be interpreted as forecasts.
- Coverage spans 216 economies. Some indicators have sparse data for earlier decades.
- All data is subject to revisions by the original institutional sources.
- Projection accuracy depends on the stability of historical patterns, which may not hold under novel conditions.
- Rankings are computed across all available countries, not just G20.
- HistorySaid does not produce original data. We organize, contextualize, and detect patterns in publicly available datasets.
Programmatic access to HistorySaid data is available for research and commercial use. The API provides JSON endpoints for all indicators, rankings, projections, and pattern data.
GET /api/v1/{country}/{indicator}?year=2024
To request API access, email hello@kavela.pro with your use case.
Get notified when we add new indicators, patterns, or projection models.