How can Spread Widening be measured?

Direct answer

Spread widening is measurable when you define (1) what “spread” number you will use, (2) when each number is sampled, and (3) how you will compare two points in time. In practice, that means computing or recording the spread from timestamped bid/ask quotes (or from a provider’s displayed spread), then calculating the change using a consistent rule such as an absolute difference or a percentage difference.

What “spread widening” means (measurable definition)

A spread is the difference between the best ask and the best bid at a given moment:

Spread (absolute) = Ask − Bid
Spread (relative) = (Ask − Bid) / Mid, where Mid = (Ask + Bid) / 2

Spread widening means that the spread becomes larger versus a prior reference point. To measure it independently, you need a defined baseline (example: “during a quiet period” vs “during a volatile period”) and a clear measurement approach (example: first quote after a time boundary vs average across a window).

How to measure it (fields, timestamps, comparison rules)

Measure spread widening with one of these data setups:

Raw quote method (bid/ask)

Record Bid(t) and Ask(t) with matching timestamps.
Compute Spread(t) each time you sample.
Compare to a baseline such as Spread(t0).

Common measurable outputs:

Absolute widening = Spread(t1) − Spread(t0)
Percent widening = [Spread(t1) − Spread(t0)] / Spread(t0)
Relative widening vs mid = (Ask − Bid) / Mid, then compare ratios

Displayed spread method

Record the spread value shown by a specific platform/provider at known times.
Use the same provider display settings and quote timing rules each time.
Apply the same absolute or percentage change formulas.

Timestamps: the key requirement

Without consistent timestamps, you can accidentally measure changes caused by quote updates rather than genuine widening. Decide whether you will sample:

at exact times (e.g., every second),
or across fixed windows (e.g., average spread between 10:00:00 and 10:00:10).

Window averaging often reduces noise, but it can also hide short spikes. Choose one approach and state it as an assumption.

Evidence or example approach (calculation template)

Assume you have timestamped quotes at two points, t0 and t1. Let:

Ask(t0) = A0 and Bid(t0) = B0
Ask(t1) = A1 and Bid(t1) = B1

Then:

Spread(t0) = A0 − B0
Spread(t1) = A1 − B1
Absolute widening = (A1 − B1) − (A0 − B0)
Percent widening = [(A1 − B1) − (A0 − B0)] / (A0 − B0)

If you use mid-based measures instead, compute Mid at each time and compare relative spread. This makes the comparison less sensitive to changes in price level.

For a “during vs before” comparison, set an explicit rule such as:

Baseline window: last 30 seconds before a selected event
Stress window: first 30 seconds after the event

State that these are assumptions about what counts as “before” and “after.”

Limitations and failure modes (what can go wrong)

Quote timing vs execution timing You may measure spreads from quotes you observe, but real trade execution can occur a moment later (or through different routing). That can produce realized costs that do not match your sampled spread.
Data source mismatch Raw bid/ask feeds and “displayed spread” may be derived differently. Comparing across sources can produce misleading results.
Volatility and update frequency During fast market moves, quotes can update frequently. If your sampling interval is too slow, you may miss spikes and understate widening.
Provider costs and non-spread costs Spread widening is only one component of trading costs. Other charges (such as commissions or fees) may change when conditions change, affecting total cost even if the measured spread does not.
Non-stationarity Historical widening patterns do not guarantee future behavior. Relationships between liquidity and spread can shift when regimes change.

Verification and next question

To verify your measurement, you should be able to reproduce it from the underlying timestamped data you used. Re-check that:

bid and ask share the same timestamp convention,
your baseline and measurement windows are defined,
your formula is applied consistently (absolute vs percent vs mid-relative).

A useful next question is whether you want to measure observed widening (from quotes) or realized widening (from the actual costs you incur when execution happens). Those two can differ when quote-to-execution delays or routing effects occur.