Why reported volume can mislead

Reported volume can mislead when datasets contain duplicated records or economically non‑distinct activity, so volume should be cross‑checked rather than accepted at face value. Researchers should perform cross‑venue reconciliation, deduplication, and audit research inputs.

Reported trading volume is often treated as a straightforward proxy for activity, liquidity, or market interest. But volume can be distorted by how venues record trades, how data vendors aggregate them, and whether the underlying prints reflect genuine economic activity. For quantitative traders and researchers, that matters because volume is frequently used in signal design, liquidity screens, execution models, and market-quality diagnostics.

The core problem is simple: venue-reported volume is not always the same thing as economically distinct trading activity. Two common failure modes are wash trading and double counting.

Wash trading refers to activity that inflates reported turnover without representing independent trading interest. In practice, this can make a venue appear more active than it is. Double counting refers to the same underlying trade or economic event being recorded more than once across feeds, venues, or aggregation layers, which can inflate totals when datasets are merged naively.

Market data should be audited, cross‑checked, and treated as potentially biased or duplicated rather than accepted at face value. Researchers should reconcile records across venues and account for duplication and inconsistencies when building datasets. Volume metrics need sanity checks against independent or cross‑validated sources rather than blind reliance on venue‑reported figures.

A practical sanity‑check workflow is:

1. Reconcile trades across venues and feeds. If the same event appears multiple times under different identifiers, timestamps, or venue labels, aggregated volume can be overstated. 2. Compare venue‑level totals with independently compiled datasets. If a venue reports materially higher volume than a separate consolidated source, the discrepancy should be investigated before using the data in models. 3. Test for abnormal concentration and spikes. Sudden volume bursts that are not corroborated by broader market activity, related venues, or independent datasets may indicate recording artifacts or non‑economic trading. 4. Audit feature sensitivity. If a strategy or model depends heavily on a venue's reported volume, rerun the analysis with alternative vendor inputs, deduplicated prints, or excluded venues to determine whether the result survives basic data‑quality controls.

Research conclusions can look stronger than they are when inputs are noisy, biased, or selected after the fact. Statistical metrics should be adjusted for known distortions and multiple‑testing effects rather than interpreted naively. Applied to volume, the analogous lesson is that raw reported figures should not be assumed valid without adjustment and validation.

Within the limits of the sources, the defensible conclusion is: volume is a useful input, but only after cross‑venue reconciliation, deduplication, and independent validation. For quantitative research and execution, treating reported volume as an audited variable rather than a raw truth helps reduce the risk of modeling on inflated or duplicated activity.

Drafted with AI assistance from cited sources. Reviewed and approved by Sonar Sciences Quant & Research Team.