Pair liquidity: what it means and what data reflects it
Pair liquidity is how easily market participants can buy and sell a currency pair with limited price impact. In practice, it is not a single number. It is reflected in several observable properties, such as how tight transaction costs are (often described via bid-ask behavior) and how much trading interest exists at different price levels (depth).
To assess pair liquidity using data, you typically combine (1) trade and quote evidence, (2) market microstructure measures, and (3) contextual information that explains why liquidity may differ across times or venues.
Direct answer: the data inputs you need
You can build a self-contained assessment checklist around four data categories: inputs, provenance, timeliness, and quality checks.
1) Inputs that describe liquidity mechanics
Use at least two of the following input types, because each captures a different aspect of liquidity:
- Quote-based measures: time series of bid and ask prices (and optionally mid price), from which you can compute spread statistics (e.g., average, median, and tail behavior).
- Order-book or depth measures: snapshots or aggregated depth at price levels around the current price; this helps estimate how much size can be traded before the price moves materially.
- Trade-based measures: executed trade timestamps, trade sizes, and whether trades occurred across many counterparties; this supports volume and participation-style interpretations.
- Execution-cost proxies: realized slippage from your own benchmark trades, or documented average execution metrics from a platform (only if definitions and sampling are clear).
Assumption to state for any calculation: define the measurement window (for example, “during the London session”) and the aggregation method (mean vs median; fixed interval vs event-based sampling).
2) Provenance: where the data comes from
Document the origin of each dataset:
- Venue or feed type (for example, aggregated market data vs venue-specific book data).
- Who publishes it (and whether it is regulatory, exchange/ECN/MTF data, or vendor data).
- What the prices represent (spot indications vs indicative quotes; whether quotes are executable or merely tradable references).
Provenance matters because the same pair can show different liquidity depending on whether the data covers one venue, multiple venues, or filtered/processed feeds.
3) Timeliness: how current and time-aligned the data is
Liquidity is time-dependent. You should record:
- Timestamp precision (seconds vs milliseconds) and time zone.
- Session labeling or at least the start/end hours used.
- Sampling cadence (e.g., every second, every trade, or periodic snapshots).
If you compare two pairs, align their windows using the same session definitions and timestamp standards, otherwise you measure different regimes.
4) Quality checks: verify that the data can support the metric
Before interpreting liquidity, run checks that confirm internal consistency:
- Missingness and outliers: detect gaps in quote series, zero-sized trades, or implausible price jumps.
- Definition consistency: confirm whether “spread” is computed as ask minus bid, whether units are pips or price terms, and how fractional pricing is handled.
- Microstructure caveats: depth data may be snapshot-based, while quote data is continuous; don’t mix them as if they represent the same instant.
Define at least one failure mode: for example, a dataset can appear “liquid” because it is smoothed or sampled infrequently, hiding brief widenings that matter for execution.
Evidence and examples of verification-ready outputs
A verification-oriented assessment produces outputs that can be reproduced from the inputs:
- Spread behavior summary: compute median spread and tail spreads (e.g., 95th percentile) over a defined time window.
- Depth-at-distance summary: measure depth within a fixed distance from the mid price (using the documented price distance definition).
- Volume and trade frequency: summarize total executed volume and the number of distinct trade events.
For each output, keep a short “calculation card” describing:
- window definition,
- time alignment method,
- units,
- whether results are volume-weighted or time-weighted.
Limitations and risks: why conclusions can fail
Several limitations can undermine a liquidity assessment:
- **Historical relationships do not guarantee future conditions. ** Liquidity can change with market stress, holidays, volatility, and technology or venue disruptions. 2. **Costs may be incomplete. ** Quote-derived spreads do not automatically include all execution costs (such as commissions, swaps, funding effects, or slippage). 3. **Data comparability is fragile. ** Different feeds or venues may apply filtering, aggregation, or timestamp handling that changes observed liquidity. 4. **Staleness and sampling bias.