Three waves of the Gambling Survey for Great Britain, reviewed as an evidence base
The GSGB now has three comparable annual waves and 20,775 respondents in the latest one. What it can carry into a policy argument, what it cannot, and why one figure keeps being misused.
One number out of this survey has done more work in British gambling policy than any other statistic of the past three years, and most of it badly. The 2.4% — adults scoring 8 or more on the Problem Gambling Severity Index, per the Commission’s published key facts — turns up in committee evidence, trade-body rebuttals and campaign material, almost always parked beside an older figure from a different survey and invited to look like a trend. The publisher says, in terms, that this is not a comparison the data supports.
So: the instrument, rather than the arguments hung on it. The Gambling Survey for Great Britain annual report reached its third edition on 16 July 2026 — roughly where a series stops being a pilot.
The file
What three waves buy
An internally comparable trend line, which is what a policy argument needs. The Commission’s framing is that the correct comparison is between successive GSGB waves — three years now on one methodology. On that basis participation sits at 47% in the past four weeks, and the Commission’s announcement calls the PGSI 8-plus share stable across the three years, putting it at 2.4% for 2025 — stable being the regulator’s word, not one figure repeated unchanged. Stability is an unrewarding finding for everyone holding a position, and the survey reports it without dressing it up.
Cadence is the second purchase. Replacing an occasional prevalence module with an annual series of quarterly waves changes what a regulator can ask of its own evidence: a question raised in one year is answered the next, not in the next parliament.
Breadth is the third. Per the published findings, 2.7% of past-12-month gamblers reported severe consequences, 6.4% cut everyday spending and 5.9% lied to family. The survey also reaches affected others, the publication’s most under-read pair: 3.3% of them sought support over someone else’s gambling, against 3.4% of gamblers seeking support for their own.
The break in the series is the whole story
The instrument’s largest limitation is not a defect in the instrument. The report states plainly that its estimates are not directly comparable with results from prior gambling or health surveys, and that such comparisons should not be used to assess trends over time. The announcement repeats it. That warning is correct, responsible and near-universally ignored: the most-quoted figure gets set against pre-GSGB estimates produced by a different mode of collection, and the movement read as a public-health event rather than a methodological artefact.
The practical consequence is a hard floor. Britain has decades of participation data and, usably, a three-year trend line — arriving exactly when the white-paper programme wanted long-run evidence. Our mid-2026 scorecard filed several measures as still pending. Evaluating them needs a series holding both a before and an after, and that series has only just started counting.
Where the caveats live is the second problem. The Commission acknowledges that all surveys have strengths and limitations, points readers to the technical report, and notes a margin of error on its estimates. All of it honest; all of it one click from the page people quote. The number travels; the qualifications stay home.
Freshness is the last of it. Fieldwork ran January 2025 to January 2026 and published in July 2026: the newest data point is roughly six months old on publication day, the oldest closer to eighteen. Fine for direction; far too slow for reacting to a stake limit or a marketing rule.
What it cannot be asked to do
It is a survey of people, not of products or operators. It cannot attribute harm to a game mechanic, one licensee’s controls or a campaign, and it cannot price a stake limit; self-report on stigmatised behaviour carries its own asymmetries. Mistaking the context-setter for the evidence is how a good survey ends up blamed for a bad argument.
Everything here concerns measurement of the 18+ licensed market, not participation in it. Anyone wanting support rather than statistics will find it free, and free of promotion, at begambleaware.org.
The balance sheet
Verdict
The right way to read GSGB is as an evidence base that begins in 2023 and improves with every wave it survives; the wrong way is the common one, and the survey gets marked down for the misuse rather than the design. An eight would need two things, only one in the publisher’s gift: the limitations discussion carried inside the publication people actually cite, and enough elapsed waves to assess a measure enacted in one year against the next. The first is an editorial decision available now; the second only by waiting.