Time-Weighted Versus Trade-Weighted Spreads

A stock's quoted spread doesn't sit still — it widens and narrows all day as liquidity providers adjust to volatility, inventory, and news. Reporting a single "average spread" for the day requires deciding how to weight all those different spread values, and the two natural choices — by how long each spread lasted, or by how much volume traded while it was in effect — don't generally agree.

Two ways to average

A time-weighted spread treats every second of the trading day as equally important: it multiplies each observed spread by the fraction of the day it was quoted, and sums. A spread of 10 cents that held for an hour counts for a full hour's worth of weight, regardless of whether a single share traded during that hour. A trade-weighted spread instead weights each observed spread by the volume that traded while it was in effect: a spread that happened to be quoted during a burst of heavy trading counts far more than the same spread quoted during a quiet stretch with no trades at all. The two answer genuinely different questions — time-weighted asks "what did the market look like, on average, moment to moment," trade-weighted asks "what spread did an average traded share actually face."

Worked example: a day with one wide, low-volume period

A stock quotes a 5-cent spread for 90% of the trading day, but widens to 25 cents for the remaining 10% of the day — specifically during a lunchtime lull when almost no one trades. Volume data:

Period	Spread	Share of trading time	Share of daily volume
Normal hours	5 cents	90%	97%
Lunchtime lull	25 cents	10%	3%

Time-weighted average spread: $0.90 \times 5 + 0.10 \times 25 = 4.5 + 2.5 = 7.0$ cents. Trade-weighted average spread: $0.97 \times 5 + 0.03 \times 25 = 4.85 + 0.75 = 5.6$ cents.

The time-weighted number makes the day look noticeably more expensive (7.0 cents) than the trade-weighted number (5.6 cents), because the wide lunchtime spread gets full credit for the time it lasted even though almost nobody actually paid it. An investor asking "what did it typically cost to trade this stock today" is much better served by the trade-weighted figure; a market-quality regulator asking "how good was liquidity provision throughout the day, independent of who showed up to trade" wants the time-weighted one.

What this means in practice

Any published "average spread" statistic is meaningless without knowing which weighting was used — the two can differ by 20% or more on an ordinary day and considerably more around scheduled news or thin periods, and comparing a time-weighted figure from one source against a trade-weighted figure from another will produce a spurious conclusion about which venue or stock is more liquid. Transaction-cost analysis almost always wants trade-weighted, because it's estimating what real orders actually paid; academic market-quality studies more often report time-weighted, because they're characterizing the state of the market itself.