Guidance on using statistics from the Gambling Survey for Great Britain
The guidance set out here is designed to help anyone who wishes to use data from the Gambling Survey for Great Britain (GSGB) to ensure it is reported correctly.
Firstly, we set out some key principles about how the survey findings should and should not be used, including some worked examples. For those who require more information we have provided a more detailed section about reporting survey estimates and further information about the methodology can be found in the GSGB technical report (opens in new tab).
This guidance is produced in accordance with the Code of Practice for Statistics, Value 3.4 Clarity and Insight (opens in new tab).
If you wish to get in touch about the GSGB, or would like some advice on how best to use or communicate our statistics please email statistics@gamblingcommission.gov.uk
The guidance is important because the official statistics from the GSGB are collected using a different methodology than previous official statistics. The guidance was published in 2024 and has since been updated (October 2025) to take on board the latest evidence to understand the causes of differences in gambling estimates in general population surveys (opens in new tab). We will continue to keep this guidance updated where further clarity is needed, or as a result of further work undertaken.
The GSGB is of course one source of evidence, and we encourage users to consider the findings in the context of the wider evidence base. Users should also keep in mind that the GSGB, in common with other surveys, collects information from a sample of the population. Consequently, the statistics based on the survey are estimates and are subject to sampling error.
How the GSGB can be used
To look at patterns within the data of gambling participation, Problem Gambling Severity Index (PGSI), and consequences amongst different demographic groups, across nations and regionally where sample sizes allow.
To assess trends and changes in gambling participation, PGSI scores, and consequences of gambling, measuring changes against the 2023 baseline.
To describe the range of consequences that someone may experience due to a person’s own gambling and due to someone else’s gambling.
To provide estimates of gambling participation and PGSI scores amongst adults (aged 18 and over) in Great Britain, including expressing these estimates as numbers of people in the whole population.
For example, to report participation estimates you could say:
“The Gambling Survey for Great Britain estimates 48 percent of adults in Great Britain aged 18 and over have gambled in the last 4 weeks.” (GSGB Year 2, 2024)
“Approximately 26 million adults in Great Britain have gambled in the past 4 weeks.” (GSGB Year 2, 2024).
You could also use 95 percent confidence intervals* to provide the range in which the true value is likely to fall: For example, you could say: “Between 25 million and 26 million adults in Great Britain have gambled in the past 4 weeks” (GSGB Year 2, 2024).
*95 percent Confidence Intervals can be calculated using the formula: p ± 1.96 × √[p(1−p)/n], where p is the observed proportion and n is the base size.
When reporting PGSI scores, you could say:
“Estimates suggest approximately 2.7 percent of adults in Great Britain aged 18 and over had a PGSI score of 8 or more.” (GSGB Year 2, 2024).
You could improve the way you report this by showing the range within which the true value is likely to fall. For example: “Estimates suggest approximately 2.7 percent of adults in Great Britain aged 18 and over had a PGSI score of 8 or more. Confidence intervals indicate that the true value within the population is likely to fall between 2.2 percent and 3.2 percent (GSGB Year 2, 2024)
When reporting PGSI scores, you could also say:
“Estimates suggest approximately 1.4 million adults in Great Britain score 8 or more on the PGSI” (GSGB Year 1, 2023). To improve reporting, you could use the 95% confidence interval to provide the range in which the true figure is likely to fall “Estimates show that between 1.2 and 1.7 million adults in Great Britain score 8 or more on the PGSI” (GSGB Year 2, 2024)
How the GSGB should not be used
As a measure of addiction to gambling.
To provide direct comparisons with results from other gambling or health surveys. This is because differences in survey design, methodology, and sampling can lead to misleading comparisons. Only comparisons with previous waves of the GSGB are appropriate when examining changes over time.
However, it is acceptable to highlight differences between surveys, provided you make it clear that different methodologies were used, and avoid suggesting that these differences indicate changes over time.
For example, you could say:
“The Gambling Survey for Great Britain estimates that 2.7 percent of adults have a PGSI score of 8 or more (GSGB Year 2, 2024). This is higher than estimates produced by other studies which use different methodologies.”
Further information about reporting survey estimates from the GSGB
Survey estimates and confidence intervals
The GSGB, in common with other surveys, collects information from a sample of the population. Consequently, the statistics based on the survey are estimates and are subject to sampling error. The intuition of a confidence interval is that, were we to repeat the survey in exactly the same way many times the true value of the statistic in the population would be within the range given by the 95 percent confidence interval in 95 samples out of 100. Confidence intervals are affected by the variability of the concept being measured, the size of the sample and other features of the sample design, such as stratification and weighting. Generally, the larger the sample, the smaller the confidence interval and, therefore, the more precise the estimate.
Confidence intervals should be taken into consideration by users. This is particularly true for PGSI estimates where base sizes can be small. We have provided confidence intervals for PGSI estimates within the data tables. Where differences are commented on in the annual report, these reflect the same degree of certainty that these differences are real, and not just within the margins of sampling error. Such differences can be described as statistically significant.
Latest research on the impact of new methodology
The GSGB uses a push to web methodology, and in his independent review of the GSGB methodology (opens in new tab) Professor Sturgis found that the GSGB enables better understanding of patterns and trends in gambling behaviour compared to periodic in-person interview surveys. However, Professor Sturgis also emphasised the need to conduct further research to examine the impact of the new methodology on estimates of gambling participation and PGSI rates (opens in new tab).
To address some of these recommendations, we commissioned the London School of Economics and an independent team at NatCen to examine how the GSGB’s methodology impacts reported gambling behaviours. Using an experimental design, the study (opens in new tab) tested whether estimates of gambling participation and PGSI scores varies based on:
- whether the survey invitation explicitly mentioned gambling
- the mode of administration (online self-completion vs. telephone interview)
- the comprehensiveness of the gambling activity list included in the survey.
The study found that mentioning gambling in the survey invitation significantly increased estimates of gambling participation but did not significantly affect PGSI estimates. However, the study could not conclude whether estimates of gambling participation were more or less accurate as a result of mentioning gambling in the invitation letter.
The study also found that participants who completed the survey online had a significantly higher score on the PGSI, compared with those who completed the survey via telephone. This finding suggests that responses to questions about the negative impacts of gambling are under-reported in interviewer-administered surveys, due to social desirability bias (the tendency for people to respond to surveys in a way that they believe will be viewed favourably). While in-person surveys mitigate this through self-completion during the interview, this is unlikely to be entirely successful. In contrast, the GSGB’s self-completion methodology largely avoids this measurement bias and encourages more accurate reporting of gambling behaviours.
Finally, the study showed that providing participants with a more extensive and up-to-date list of gambling activities (as in the GSGB) did not have a detectable impact on survey estimates.
Overall, this experimental research helps us to understand why interviewer-administered surveys produce lower estimates of gambling and its impacts compared to online self-completion.
As a result of the research we can have more confidence in the accuracy of the estimates produced by the GSGB. In addition, the sample size in Year 2 data is larger than in Year 1. Based on these developments, we have updated our guidance for using the GSGB. The main changes in the guidance published in October 2025 are to remove some of the caution we previously advised whilst we waited for the findings of the experimental research. This includes removing the advice not to gross estimates up to whole numbers in the population.
Be careful reporting base numbers
To ensure we can include all relevant content within the GSGB, core questions are asked on both the online and paper version of the survey whereas some topical or modular questions are only asked on the online version of the survey. The Gambling Commission will clearly label any statistics which are based on online responses only, and users should do the same.
The GSGB asks a range of questions, some of which are applicable to all participants, some which are only applicable to people who have gambled and some which are only asked in the online version of the survey.
It is important to correctly reference whether statistics are based on all participants, or whether they are a subset of all participants, such as people who have gambled in the past 12 months or participants who completed the online version of the survey to set the findings in the correct context.
Through our stakeholder engagement we know that stakeholders are interested in multiple ways of presenting the data, for example at a population level including people who do not gamble as well as a focus on people who have gambled.
This distinction is important as the first group includes people who have not gambled on any activity in the past year, whereas the second group is based only on people who have gambled in the past 12 months. In the annual report we have also included a third group which excludes people who have only taken part in lottery draws. This is because lotteries are so much more popular than any other form of gambling with a large proportion of people only participating in this activity, therefore, it can mask patterns of what is going on with other types of gambling. For this reason, in the report we sometimes present findings excluding the people who have only taken part in a lottery draw and not taken part in any other type of gambling. Where findings excluding those who have only taken part in a lottery draw are used, they should be clearly labelled.
Care should be taken when reporting statistics relating to the PGSI to make sure you are correctly stating if the results are based on the responses of all participants, or if they are based on people who have gambled. This is an area where we have previously seen misreporting.
It is also worth noting that new questions in the GSGB about the wider consequences of gambling are all presented as a proportion of participants who have gambled in the past 12 months or as a proportion of participants who know someone close to them who gambles, so we think it’s best practice to report data in this way. Based on Professor Sturgis’ experimental research, we do not expect methodological factors to materially affect population-level estimates for these questions, but we are taking a cautious approach because his study did not specifically examine this. It is also important to note that these questions only capture experiences within the past year and therefore do not capture adverse consequences that persist after someone has stopped gambling.
Annual versus wave specific data
In a typical year there will be four wave specific publications from the GSGB plus an annual publication. Where possible, the annual data should be used as the priority with wave specific data being used when you want to look at patterns of gambling participation within a year, or where modular questions have only been asked in certain waves.
The GSGB collects data continuously throughout the year. Survey data will be available:
- on a quarterly basis via wave specific publications
- annually where data for the calendar year will be combined to provide a more detailed breakdown.
Annual datasets will be published to UK Data Service (opens in new tab).
We recommend using annual data as the default as this will be based on a large sample size and will allow for more analysis at sub population level. This is also how we will track trends over time.
Wave specific data should be used if you need data for a specific time period, and to track trends or patterns within a calendar year. These publications will focus predominately on participation in gambling in that time period.
Language
Use a person centric approach when reporting statistics about gambling. Do not stigmatise or victimise those people experiencing adverse consequences from gambling. Do not describe PGSI as a measure of gambling addiction.
The language we use matters. People who gamble are defined by more than their actions when they gamble. That is why we recommend a “person-centric” or “person first” approach. Whilst taking this new approach may use more words, it is important in lowering stigma and barriers to people seeking help for gambling addiction.
For example, instead of writing “x percent of gamblers…”, you can write “x percent of people who gamble…”.
There is more information available on why language matters from organisations including the University of Glasgow (opens in new tab), GambleAware (PDF) (opens in new tab) and Manchester Combined Authority (PDF) (opens in new tab).
The Problem Gambling Severity Index (PGSI) consists of nine questions which measure both behavioural symptoms of gambling disorder and certain adverse consequences from gambling. The PGSI should not be confused with a measure of gambling addiction. More information on how the PGSI is measured can be found in our problem gambling screens report.
Wider evidence base
The GSGB is one source of data in the Commission’s wider evidence base.
The Gambling Commission uses a range of data, research and insights to inform the decisions that we make and provide advice to the Government about gambling behaviour and the gambling market. To be the most effective regulator possible, we require a robust evidence base. The GSGB forms one source of evidence for our evidence base and should be considered alongside a wealth of other evidence and information which we use to fill our evidence gaps and priorities 2023 to 2026.
If statistics are used incorrectly
We encourage people to use our statistics to support understanding of important issues related to gambling.
We expect that anyone using our official statistics should present the data accurately and in accordance with the guidelines presented here. This includes ensuring that the data is not taken out of context, manipulated, or presented in a way that could materially mislead others.
We have set out further information on the action we will take if we spot misuse of official statistics .
If you wish to get in touch about the GSGB, or would like some advice on how best to use or communicate our statistics please email statistics@gamblingcommission.gov.uk
Last updated: 2 October 2025
Show updates to this content
Updated for GSGB Annual report (2024).