Gambling

the frame

The law tells the regulator to permit gambling

The Gambling Act 2005 has never been amended on this point. It gives the Gambling Commission three licensing objectives and a duty with two limbs: to pursue those objectives, and to permit gambling so far as that is reasonably consistent with them. Preventing harm to adults in general is not among the objectives, and no council may refuse a betting shop on the ground that the town already has enough of them. That is the frame every number below sits in: an apparatus built to license and to tax, and never asked to count what happens next.

The 2023 white paper, High stakes: gambling reform for the digital age, proposed more than 60 measures and did not propose changing the objectives. It affirmed them.

counted exactly

£16.8 billion, reported sector by sector

Operators report to the Gambling Commission and pay duty to HM Revenue and Customs, so the size of the industry is one of the better measured quantities in British public data. It is published quarterly, split by sector, and the components reconcile with the totals. The seven duties below sum to £3,616.3 million before rounding, against the £3,616 million that HMRC publishes separately as the year's receipts. Adding the rounded figures as printed gives a total a hundred thousand pounds lower, which is what rounding seven numbers does and is left visible rather than hidden.

FY2024-25, HMRC UK Betting and Gaming Statistics. Total receipts reached £3,831 million in 2025-26, up from £2,116 million in 2014-15. Remote Gaming Duty rises from 21 per cent to 40 per cent on 1 April 2026 and a new 25 per cent remote rate of General Betting Duty follows on 1 April 2027; HMRC scores the package at £810 million in 2026-27 rising to £1,155 million by 2030-31.

Betting premises have fallen from 9,111 at 31 March 2014 to 5,669 at the end of December 2025, a fall of 37.8 per cent. That comparison begins in 2013 rather than at the higher figure of 9,128 recorded for 2012. The Commission's methodology page says premises data came from local authorities up to March 2012 and from operator regulatory returns from March 2013, that casino premises stayed on a separate return until 31 March 2015, and that it does not recommend comparing the data across the different data sources. So there are two breaks, and no casino trend should cross April 2015 either. Even the estate count, one of the better measured things here, has seams in it that the publisher tells you not to cross.

The latest 12 months of Gambling Commission statistics are provisional, and the figures on this page carry that status: the FY2024-25 yield, the premises and the machine counts. For the annual report the Commission apportioned 961 of 6,539 regulatory returns, 15 per cent, and estimated a further 70, with 52 operators overdue on 70 returns. The quarterly workbook the premises and machine counts come from publishes no equivalent figures. Well measured here means better measured than the harm, not exact.

the one clean measurement

What happened when a stake limit met a tax line

In April 2019 the maximum stake on a fixed odds betting terminal fell from £100 to £2. Those machines were taxed at the higher rate of Machine Games Duty, so the effect landed in a tax series that HMRC publishes to five decimal places. It is the clearest picture in British official data of what a stake limit does to a revenue stream, and it exists only because the thing being measured was money.

counted twice, differently

Three official surveys, a sixfold gap, and no agreed explanation

The Problem Gambling Severity Index scores nine questions from 0 to 27, and a score of 8 or more is the standard threshold for the most serious category. Two sets of official statistics measure how many British adults are above it. They do not agree, and the difference is not small.

The Gambling Commission set the same three surveys side by side itself, in a note to editors of 14 August 2025, using its own 2023 reading of 2.5 per cent against the same two England figures. The reading shown above is its latest, from 2025. Its guidance says the survey should not be compared directly with other gambling or health surveys, and that highlighting the difference is acceptable as long as the different methods are made clear and the gap is not read as change over time. That is what is being done here.

1. July 2023Main stage fieldwork begins on the Gambling Survey for Great Britain, a push to web survey replacing the interviewer administered series.
2. February 2024Professor Patrick Sturgis assesses the new survey as exemplary in all respects, and warns of a non-negligible risk that it substantially overstates the true level of gambling and gambling harm. The Commission's press release announcing the assessment does not mention the warning.
3. May 2025The Office for Statistics Regulation publishes a compliance review with nine recommendations, finding that the Sturgis conclusion on overstatement is absent from the statistical releases themselves, and asking for it to be communicated within them.
4. August 2025A randomised experiment finds that a gambling framed invitation and online self completion together add roughly 5 to 6 percentage points to estimates of a PGSI score above zero. The authors conclude that approximately a third of the difference remains unaccounted for. The Commission says the findings greatly increase its confidence.
5. October 2025The Commission relaxes its own guidance, removing the advice against grossing estimates up to whole population numbers. Its worked examples now do exactly that.
6. July 2026The third annual report is published. It carries the comparability warning and a pointer to the technical report. It does not state the overstatement conclusion the regulator asked for.

All three readings are official statistics and only the two lower ones are accredited, which the regulator cautions is not in itself a ranking of quality. The point is not that one is right. It is that Britain has been measuring the same quantity two ways since 2023 and cannot say which is closer to the truth. On the wider measure of any PGSI score above zero, where the 2023 survey gave 14.3 per cent against the accredited survey's 4.4 per cent, the reviewer who ran the experiments concluded it was plausible that survey framing and online self completion together add 5 to 6 percentage points, an approximate estimate rather than a precise one, and that about a third of the difference remained unaccounted for. The published experiments address that wider measure rather than the PGSI 8 or above gap shown here, and the benchmarking exercise meant to settle it is not out yet.

measured

Where the harm is concentrated

Whatever the true level, the shape inside the survey is consistent and steep, and the gradient across the bands is far larger than the gap between the surveys. The two charts that follow cover people who gambled in the past 12 months, grouped by their PGSI score and reporting on themselves. The deprivation chart below them has a different base again, and says so.

Reported one or more severe consequence in the past 12 months due to their own gambling: losing something of significant financial value, relationship breakdown, experiencing violence or abuse, or committing a crime to finance gambling. Base: people who gambled in the past 12 months, Great Britain. Unweighted bases 9,750, 1,440, 576 and 336.

not counted

What Britain does not count

Each entry below is a negative finding, checked at the source that would hold the answer if it existed. Some of these quantities are genuinely hard to measure, and the coroner's record is the clearest case. Others were counted once and stopped, or are collected and not published, or have never had a field defined for them. One of the five is now changing: NHS England has said providers will be required to submit treatment data to the Mental Health Services Data Set, in the future tense and without a stated start date.

estimated

The official cost of harm rests on the lower prevalence figure

The current government estimate of what gambling harm costs in England is an evidence update published by the Office for Health Improvement and Disparities in January 2023. It puts the annual excess direct financial cost to government at £412.9 million and the societal value of health impacts at between £635 million and £1,355.5 million, a combined £1.05 billion to £1.77 billion in 2021-22 prices.

Office for Health Improvement and Disparities, evidence update 2023, in 2021-22 prices. The suicide component is shown at the top of its range.

• 1.It is built on the 2018 Health Survey for England, which found 0.4 per cent of adults at PGSI 8 or above. That is the lower of the two contested prevalence figures. Nobody has redone the costing on the higher one.
• 2.The suicide component is not a count of deaths. It applies standardised mortality ratios from a Swedish cohort study to English suicide rates, giving an estimated excess of 117 to 496 deaths a year. The report states that applying the Swedish figures to the English gambling population is a limitation of the analysis. The median age of suicide death in the Swedish sample was 32.5 against 47 in England.
• 3.The report says explicitly that it cannot say the excess cases and costs are caused by harmful gambling, only that this is the best current evidence on the size of the association. In December 2025 the statistics regulator wrote to the department warning that users repeatedly misread that association as causation.
• 4.Its own exclusion list is long: financial harms to gamblers and their families, costs to affected others, healthcare costs of suicidality, rough sleeping, indirect crime costs including victims, lost productivity, relationship harms, and the direct cost of gambling treatment. The report says the total is likely to be underestimated.

the state's own verdict

Asked to spend the money, the NHS would not use the data

From April 2026 NHS England commissions the gambling treatment pathway funded by the statutory levy, with regional allocations of £15.90 million in 2026-27 and an indicative £38.62 million and £39.90 million in the two years after. Allocations between regions are not based on gambling harm prevalence. The reason is given in a footnote to the annex.

There are significant concerns about the quality of existing data on the prevalence of gambling harms, either due to the methodology used or the fact that studies have been industry funded/facilitated.

Weighted mental health populations are used instead. Treatment takes 50 per cent of the levy under a departmental policy split rather than under the regulations, shared between NHS England and the Scottish and Welsh governments. This is the clearest available statement of how far the state trusts its own gambling harm numbers, and it comes from the body that has just been handed the English half of that work.

questions this raises

Questions this raises