ExchangeVolatilityIndex.com
Methodology
Construction choices that define what an EVI number actually means.
Goal: define a construction method that is precise enough to reproduce, but flexible enough to support different datasets (spot, forward, options) and different “baskets” of currencies.
Step 1 — Choose the Underlying
- Single pair: e.g., EUR/USD, USD/JPY, GBP/USD.
- Basket: trade-weighted set (major pairs) or region-specific set (EM FX).
- Rate type: spot, forward, or an index rate (trade-weighted exchange rate).
Step 2 — Choose Volatility Type
| Type | Input | Interpretation |
|---|---|---|
| Realized | Historical FX returns over a lookback window | “What happened” volatility (backward-looking) |
| Implied | FX option prices (ATM, variance swap, or surface) | “What the market prices” volatility (forward-looking) |
| Hybrid | Blend of realized + implied or model-based estimates | Stability + responsiveness tradeoff |
Step 3 — Specify the Horizon
- 1-week, 1-month, 3-month, or 1-year horizons are common.
- Implied-vol horizons typically align to option tenors; realized-vol horizons align to return windows.
Step 4 — Normalize & Annualize
To compare across horizons, volatility is commonly annualized. A minimal realized-vol definition:
Let rt = ln(St / St-1). Over window N, realized vol ≈ sqrt(252) · stdev(r) for daily data (adjust the factor for your sampling).
Step 5 — Aggregate (If Basket-Based)
- Equal-weighted: simple average across pairs.
- Trade-weighted: weights reflect trade shares or exposure.
- Risk-weighted: weights scaled by long-run volatility so one pair doesn’t dominate.
Reproducibility Checklist
- Data source, timestamps, and rate definition
- Return definition (log vs simple), sampling frequency, and missing-data handling
- Window, horizon, annualization convention
- Weighting scheme and rebalancing cadence (if basket)