Measuring volatility: definition first
Volatility describes how much a price series moves over time. In currency-pair context, a common approach is to measure volatility from returns, not directly from the observed exchange rate levels.
For GBP/USD and GBP/JPY, measuring volatility means quantifying how variable the exchange rates (or their returns) are over a chosen time horizon. Importantly, volatility measurement is descriptive: it tells you how “choppy” or variable the past was under specific assumptions; it does not predict future direction.
Core mechanics: from exchange rates to returns and variability
1) Choose what “GBP USD vs GBP JPY” means numerically
You must fix the data representation for each pair:
- Which quotation you use (typically the spot or a consistent reference price).
- Which timestamp you sample (e.g., end-of-minute, end-of-day).
- Which price type you use (for example, closes vs midpoints). If you include bid/ask effects, measured volatility can rise even when “true” mid-price movement is modest.
2) Convert prices into returns
A standard choice is log returns or simple returns between two consecutive observations. If (P_t) is the exchange rate at time (t), then for log returns:
- (r_t = \ln(P_t/P_{t-1}))
Returns help normalize percentage movement, which is useful when comparing pairs that may trade at different absolute price levels.
3) Quantify variability over a window
A common volatility measure is the rolling standard deviation of returns:
- compute returns over time,
- pick a window length (for example, N observations),
- calculate the standard deviation within each window.
To compare across different time scales, you often annualize volatility using a scaling factor that depends on your sampling frequency (this step requires a clear assumption about how many periods correspond to a year).
4) Use at least one alternative metric
Range-based measures can capture a different aspect of movement. For example, you can estimate a volatility proxy from high–low ranges within each period (if your dataset provides highs and lows). This may respond differently to intraperiod spikes than return-based standard deviation.
Evidence or example: how the comparison is made
A straightforward, self-contained comparison workflow looks like this (no live data required):
- Pick a sample interval (e.g., daily closes) for both GBP/USD and GBP/JPY.
- Compute returns for each series using the same return definition.
- Choose a rolling window length (N).
- For each day (or each window), compute rolling volatility = standard deviation of returns over that window.
- Compare the resulting volatility series:
- A pair with higher rolling volatility has shown larger average variability in that window, under your chosen sampling and definitions.
Assumptions matter. If GBP/USD volatility spikes during certain periods (a “volatility regime”), but GBP/JPY remains smoother, your rolling comparison will reflect that. If you change (N) or switch from closes to midpoints, the values can change.
Limitations and risks: what can fail in volatility measurement
1) Window choice changes the answer
Short windows can react strongly to sudden moves and may exaggerate transient noise. Longer windows can smooth variability and hide regime changes. Two analysts using different (N) may report different “volatility levels” even with the same underlying data.
2) Data source and price type can dominate results
Bid/ask effects, sampling frequency, and how missing observations are handled can affect computed volatility. If one dataset uses mid-prices and another uses last-traded closes, measured volatility may not be comparable.
3) Volatility is not stable across time
Historical volatility is conditional on past conditions. Even if GBP/USD has been more volatile than GBP/JPY recently, there is no guarantee that this relationship persists.
4) Costs and execution are separate from volatility
Volatility measurement does not include trading costs, execution quality, or settlement frictions. Those factors can materially affect realized outcomes even when volatility estimates are accurate as a description of price variability.
Verification and next question
To independently verify your measurement, check these items:
- Use the same return formula for both pairs.
- Recompute volatility with two different window lengths to see sensitivity.
- Confirm the exact price inputs (close vs mid vs bid/ask) and sampling timestamps.