The best current estimate is that 46.2% of adults gambled in the past year, that 8.7% gambled at some level of risk, and that 1.41% met the criteria for problematic gambling. Those figures come from a 2024 systematic review and meta-analysis in The Lancet Public Health that pooled 380 representative samples from 68 countries. The World Health Organization separately estimates that 1.2% of the world’s adults have a gambling disorder.
Those numbers look precise. They are not. Every figure in this page depends on which survey asked, which questionnaire it used, and how it reached people. This guide sets out the gambling statistics worldwide that we can stand behind in 2026, names the survey behind each one, and shows you where the numbers stop being comparable.
The global picture at a glance
Two documents dominate the global evidence. The first is the Lancet Public Health systematic review and meta-analysis of gambling prevalence (2024), which pooled 380 samples covering about 3.4 million people across 68 countries. The second is the Lancet Public Health Commission on gambling, published online on 24 October 2024, which built on that review.
| Measure | Estimate | Source and instrument |
|---|---|---|
| Adults who gambled in the past year | 46.2% (95% CI 41.7–50.8) | Lancet PH meta-analysis 2024; pooled across instruments |
| Adolescents who gambled in the past year | 17.9% (95% CI 14.8–21.2) | Lancet PH meta-analysis 2024 |
| Adults gambling at any level of risk | 8.7% (95% CI 6.6–11.3) | Lancet PH meta-analysis 2024; PGSI, DSM and SOGS pooled |
| Adults with problematic gambling | 1.41% (95% CI 1.06–1.84) | Lancet PH meta-analysis 2024 |
| Adults with a gambling disorder | 1.2% | WHO gambling fact sheet, 2 December 2024 |
| Adults affected by any-risk gambling | about 448.7 million people | Lancet PH Commission 2024 |
The Commission also reports a wide gap between men and women. About 5.5% of women and 11.9% of men gamble at some level of risk. Roughly 80 million adults worldwide are estimated to have problematic gambling or gambling disorder.
Notice that the meta-analysis says 1.41% and the WHO says 1.2%. That is not a contradiction. They are measuring different things. “Problematic gambling” is a score on a screening questionnaire. “Gambling disorder” is a clinical category. We explain the difference in our guide to what problem gambling is and how PGSI, DSM-5 and ICD-11 compare.
What gambling statistics worldwide actually measure
Almost every prevalence figure you will read comes from a screening questionnaire filled in by members of the public. The most common one is the Problem Gambling Severity Index, or PGSI. It has nine questions. Each answer scores 0 to 3. The total runs from 0 to 27.
| PGSI score | Usual label |
|---|---|
| 0 | No problems reported |
| 1–2 | Low risk |
| 3–7 | Moderate risk |
| 8 or more | Often reported as “problem gambling” |
A screening score is not a diagnosis
This is the single most misused point in gambling data. A PGSI score of 8 or more does not mean a person has been diagnosed with anything. It means they answered a questionnaire in a way that suggests harm and that a professional assessment would be worthwhile.
Britain’s regulator is explicit about this. Its guidance on using statistics from the Gambling Survey for Great Britain lists, under how the survey should not be used, “as a measure of addiction to gambling”. So when you see a headline that turns a PGSI percentage into a count of addicts, the headline has broken the rule set by the body that published the data.
Other surveys use other tools. Some use DSM-based checklists. Some use the older South Oaks Gambling Screen, which asks about a person’s whole life rather than the past year. A lifetime measure will almost always give a higher number than a past-year measure. Putting the two side by side without saying so is misleading.
Why one country can report 2.4% and 0.4% in the same year
Great Britain is the clearest example in the world of how much method matters. Two official surveys measured the same thing, with the same questionnaire, in roughly the same period, and produced answers six times apart.
The figures below come from the Gambling Commission’s Gambling Survey for Great Britain Annual Report 2025 and from NHS England’s Adult Psychiatric Morbidity Survey 2023/4. Both are official statistics. Neither is wrong.
| Gambling Survey for Great Britain | Adult Psychiatric Morbidity Survey | |
|---|---|---|
| Latest report | Annual Report 2025, published 16 July 2026 | 2023/4, published 27 November 2025 |
| PGSI 8 or more | 2.4% (95% CI 1.9–2.8) | 0.4% |
| PGSI 3–7 | 3.5% (95% CI 3.0–4.0) | 1.2% |
| PGSI 1–2 | 7.8% (95% CI 7.0–8.7) | 2.8% |
| Past-year gambling | Around 47% of adults | 42.6% of adults |
| How people were recruited | Push-to-web: households invited by post to go online | Interviewers recruited households in person at home |
| How the gambling questions were answered | Self-completed online or on paper | Self-completed, inside the phase-one interview |
A common but wrong explanation is that people admit more to a screen than to an interviewer. That cannot be the main reason here. In the Adult Psychiatric Morbidity Survey 2023/4, the gambling questions sat in the self-completion part of the interview. Respondents filled them in themselves, just as they did in the GSGB.
The bigger driver is who takes part at all. A postal invitation to complete a survey online is answered by a different slice of the population than a doorstep invitation from a trained interviewer. People with a strong interest in the subject are more likely to respond to a survey they have to opt into. That pushes the measured rate up.
This is not a fringe view. The Office for Statistics Regulation review of the GSGB, published on 22 May 2025, records Professor Patrick Sturgis’s conclusion that policymakers must treat the estimates with due caution because there is a “non-negligible risk that they substantially over-state the true level” of gambling and gambling harm. The Gambling Commission’s own guidance says that only comparisons with earlier waves of the GSGB itself are appropriate for tracking change over time.
Can you turn a percentage into a number of people?
Cautiously, and only for the survey it came from. The Gambling Commission used to advise against scaling its percentages up to population totals at all. It relaxed that advice in October 2025, removing the instruction not to gross estimates up to whole numbers in the population.
That change makes headline counts of people more defensible than they were. It does not make them diagnoses. A figure like “1.4 million adults” is still a count of people who scored 8 or more on a questionnaire, carrying all the uncertainty of the survey that produced it, including the confidence interval. The ban on reading those scores as addiction has not been lifted.
The practical rule for readers: never place a figure from one survey mode next to a figure from another and call it a trend. Britain did not become six times worse or six times better. Two different methods measured the same country.
What national surveys report, country by country
Read the table down the rows, not across them. Each row is sound on its own terms. Comparing one row to another is exactly the mistake described above.
| Country | Survey and year | Headline figure | Instrument and mode |
|---|---|---|---|
| Great Britain | GSGB Annual Report 2025 (pub. 16 Jul 2026) | 2.4% scored PGSI 8+ | PGSI; push-to-web self-completion |
| England | APMS 2023/4 (pub. 27 Nov 2025) | 0.4% scored PGSI 8+ | PGSI; self-completed in an interviewer-recruited home visit |
| United States | NGAGE 3.0, fieldwork 26 Jan–20 Mar 2024 | 71% gambled in the past year; 8% reported at least one problematic behaviour “many times” | Not a PGSI or DSM prevalence measure; online panel, n=3,013 |
| New South Wales, Australia | NSW Gambling Survey 2024 | 53.5% gambled; 21.0% experienced gambling harm | Telephone interviews, n=10,000 |
| Sweden | National public health survey, tracked since 2014 | 3–4% have some degree of gambling problem | PGSI-based; national health survey |
The United States figure is widely misread
The National Council on Problem Gambling’s NGAGE 3.0 key findings report that 8% of American adults, close to 20 million people, said they had experienced at least one of four potentially problematic gambling behaviours “many times”. That is down from 11% in 2021 and up from 7% in 2018.
The report states plainly that the survey “is not designed to assess the prevalence of gambling disorders, as defined by the American Psychiatric Association’s Diagnostic and Statistical Manual, Fifth Edition (DSM-5)”. Any page that presents 8% as a US rate of gambling addiction has ignored the authors’ own warning. The United States has had no national prevalence study of the kind Britain or Sweden run since the late 1990s.
Australia measures harm, not just scores
The NSW Gambling Survey 2024, commissioned by the NSW Responsible Gambling Fund and run by Central Queensland University, interviewed 10,000 adults by telephone between March and May 2024. Participation was 53.5%, close to the 53% recorded in 2019.
Its most useful finding is about harm rather than scores. One in five respondents, 21.0%, reported experiencing gambling harm: 7.8% from their own gambling and 12.7% from someone else’s. The survey also concludes that electronic gaming machines are the source of most gambling harm in the state, followed by wagering, casino table games, online casinos and online poker.
Sweden shows what a stable series looks like
Sweden’s Public Health Agency reports that roughly 3–4% of the population aged 16 to 84 has some degree of gambling problem, and that this has stayed broadly flat since measurement began in 2014. The agency also notes, from its Swelogs study, that about 130,000 people in Sweden share a household with someone whose gambling is a problem.
Harm spreads well beyond the person gambling
Prevalence rates count gamblers. They miss most of the people harmed.
The WHO gambling fact sheet of 2 December 2024 states that for every person gambling at high-risk levels, an average of six other people are affected, and that these are usually people who do not gamble themselves. The NSW survey gives a direct measurement of the same effect: more people there reported harm from someone else’s gambling (12.7%) than from their own (7.8%).
The harms the WHO lists are not only financial. They include relationship breakdown, family violence, financial distress, stigma, income-generating crime such as theft and fraud, and neglect of children.
A second limitation matters just as much. Most harm does not sit in the highest-scoring group, because that group is small. In Great Britain the GSGB found 2.4% scoring 8 or more but 3.5% scoring 3 to 7 and 7.8% scoring 1 to 2. A small amount of harm spread across a much larger group can add up to more total damage than a severe problem in a few.
If any of this is describing your life
Support is free and confidential in most countries. Our step-by-step guide to getting help for gambling explains what happens when you make contact, and our directory of gambling helplines by country lists the official service where you live. If you want to make gambling harder to reach today, our guides to self-exclusion and to responsible gambling tools set out the practical options.
The health data, handled carefully
The WHO fact sheet records that gambling can threaten health, with a higher incidence of mental illness and of suicide. It cites a Swedish study in which people with a gambling disorder were around 15 times more likely to die by suicide than the general population, and an estimate that at least 4.2% of suicides in Victoria, Australia were gambling-related.
Read those figures with care. They describe raised risk across a population. They do not describe what will happen to any one person, and they are not a forecast. People recover from gambling disorder, including people who have been very unwell. If you are struggling, the services above are the right place to start, and a doctor can help too.
Our page on gambling addiction, its signs, causes and treatment covers what treatment actually involves.
The money, and where it comes from
Market data is more solid than prevalence data, because operators must report it to regulators and tax authorities.
The WHO fact sheet cites industry analysts’ projection that global gambling revenue will reach US$700 billion by 2028. The Lancet Commission puts online gambling yield on course for US$205 billion by 2030. “Revenue” here means net consumer losses: the money players lose, which is the industry’s income.
For a single regulated market, the Gambling Commission’s industry statistics for 2025 to 2026, published on 17 September 2026, report gross gambling yield in Great Britain of £17.5 billion for April 2025 to March 2026, a 4.4% increase on the year before. Remote casino, betting and bingo accounted for £8.3 billion of that, up 6.9%. The land-based arcade, betting, bingo and casino sectors accounted for £4.9 billion, up 1.1%.
One figure links the money to the harm. The WHO states that people gambling at harmful levels generate around 60% of losses. If that holds, the commercial model does not depend mainly on people gambling small amounts safely. It depends on a minority losing a great deal.
Risk is concentrated in particular products
Prevalence varies enormously by what people play. The Lancet meta-analysis found problematic gambling among 15.8% (95% CI 10.7–21.6) of people who gamble on online casino games or slots. That is more than ten times the 1.41% pooled rate for adults generally.
| Product | Problematic gambling among adults who use it | Among adolescents who use it |
|---|---|---|
| Online casino games or slots | 15.8% | 26.4% |
| Sports betting | 8.9% | 16.3% |
These are rates among people who already use the product, not shares of the whole population. They cannot tell us on their own whether the product caused the harm or attracted people already at risk. But the pattern is consistent with decades of research: fast, continuous, always-available products carry more risk than slow ones. The Commission also reports that 10.3% of adolescents had gambled online.
What the numbers still cannot tell us
- Most of the world is barely measured. The meta-analysis covered 68 countries, and most of the studies came from high-income countries. Africa, the Pacific Islands and Central Asia are close to blank.
- Direction of travel is hard to read. When a country changes its survey method, the series breaks. Britain’s does.
- Screens are not diagnoses. No national survey diagnoses anyone. They flag people who should be assessed.
- Affected others are under-counted. Few surveys ask them directly. NSW is an exception.
- Harm is not the same as a score. Someone can be badly harmed and still score below 8.
How to read any gambling statistic in five questions
- Which survey, and what year? No survey name means no way to check it.
- Which questionnaire? PGSI, a DSM checklist and SOGS give different answers. A lifetime screen gives higher numbers than a past-year one.
- How did it reach people? Push-to-web, telephone and doorstep recruitment produce different results from the same population.
- Who is the denominator? All adults, all people who gamble, or only users of one product? The 15.8% slots figure means nothing without that.
- Is it being called a diagnosis? If a percentage has been turned into a number of “addicts”, the number has been misused.
Applied honestly, the 2026 data picture supports a simple summary. Most adults who gamble are not harmed. A substantial minority are. Harm reaches far more people than the gamblers themselves, it clusters in the fastest online products, and it generates most of the industry’s income. The precise percentage is less certain than most headlines suggest, and anyone telling you otherwise is not reading the surveys.
For the principles behind these debates, see what responsible gambling means. How we choose and check sources is set out in our editorial policy and our funding and independence statement. If you think a figure here is wrong, our corrections policy explains how to tell us.
Last reviewed: 24 September 2026. Sources checked on this date.