This box is not visible in the printed version.
Impact of new gambling activities list on the Gambling Survey for Great Britain participation and consequences estimates
Published: 9 October 2024
Last updated: 9 October 2024
This version was printed or saved on: 2 May 2025
Online version: https://www.gamblingcommission.gov.uk/report/sturgis-review-recommendation-3-update
In February 2024, Professor Patrick Sturgis published an independent review (opens in new tab) of the Gambling Survey for Great Britain (GSGB).
The independent review assessed the GSGB’s methodological approach and made recommendations for improvements to ensure the quality and robustness of the statistics. A key emphasis of the review was the need to examine how methodological differences between previous Health Surveys1 and the GSGB may have impacted estimates of gambling participation and harm. In particular, Sturgis recommended examining the effect of transitioning from in-person interviews to self-completion surveys, and establishing the potential impact of changes to the list of gambling activities provided to participants.
Recommendation 3 was as follows:
“The stage 3 experimental survey found significantly higher Problem Gambling Severity Index (PGSI) scores than the 2022 pilot. This might have been a result of the use of an updated list of gambling activities on the 2023 survey but it might equally have been due to an increase in gambling harm in the population. In order to assess the impact of the updated gambling activity list, an experimental design is necessary.
Recommendation 3: the Gambling Commission should undertake a randomised experiment to evaluate the effect of the updated list of gambling activities on estimates of gambling prevalence and harm.”
1 There are three separate health surveys: the Health Survey for England (HSE)(PDF)(opens in new tab), the Scottish Health Survey (SHeS)(PDF)(opens in new tab), and the National Survey for Wales (NSF)(PDF)(opens in new tab) which has a health component to it. All contain the same gambling participation activity list questions. For brevity we will be using the term “health surveys” going forward.
In March 2024, the Gambling Commission (the Commission) asked YouGov to conduct an online survey experiment to assess the impact of different gambling activity lists on estimates of gambling participation and prevalence of potentially problematic gambling behaviours. In total, 2,064 participants aged 18 and over and living in Great Britain took part in the study.
Participants were randomly allocated to one of two conditions:
In both conditions, participants were asked to indicate which, if any, of the listed activities they had engaged in over the past 12 months. The key difference between the 2 lists was that the GSGB list included a broader range of online activities, which better reflects the current gambling landscape. For the full list of activities see Appendix - Activity lists.
Participants who had taken part in any gambling during the past year were then presented with the full 9 item Problem Gambling Severity Index (PGSI) screen which measures behavioural symptoms of gambling disorder and certain adverse consequences from gambling. Answer options for the 9 PGSI statements are ‘almost always’, ‘most of the time’, ‘sometimes’, ‘never’. Responses to the 9 questions are summed up and a score ranging between 0 and 27 is computed. Scores are grouped into the following categories:
Participants were roughly evenly allocated to Health Survey and GSGB Conditions (Health Survey Condition = 1,051; GSGB Condition = 1,012), and groups were matched in terms of age and gender. The survey was distributed to a nationally representative sample, and responses were weighted to reflect the national profile of adults aged 18 and over. The fieldwork for the survey took place between Thursday 28 March and Friday 29 March 2024.
The list of gambling activities provided in the GSGB covers a broader range of online activities than those listed in the health surveys. For this reason, we hypothesised that the GSGB Condition would identify a higher proportion of people who had gambled online, relative to the Health Survey Condition. Given that online gambling activities tend to be associated with higher PGSI scores1, we also hypothesised that the GSGB Condition would yield a higher proportion of people scoring 8 or more on the PGSI, compared with the Health Survey Condition.
Significance testing (using chi-squared tests of independence) was used to assess whether there were any statistically significant differences in gambling participation and/or Problem Gambling Severity Index (PGSI) scores between those who completed the Gambling Survey for Great Britain (GSGB) activities list compared with those who completed the health surveys activities list1.
Approximately 63 percent of participants reported gambling in the past 12 months, and this did not differ significantly between conditions (as shown in Table 1: Gambling participation in the past 12 months by experimental condition). However, consistent with our hypothesis, the GSGB Condition identified a statistically significant higher proportion of people who had gambled online, compared with the Health Survey Condition (19.8 percent versus 15.3 percent). The GSGB Condition also identified a statistically significant higher proportion of people who had only gambled online, compared with the Health Survey Condition (3.1 percent versus 1.6 percent) (as shown in Table 2: Online gambling participation in the past 12 months, excluding National Lottery players).
| Experimental condition | Health Survey Condition (percentage) | GSGB Condition (percentage) | Difference (percentage) |
|---|---|---|---|
| Gambling participation | 63.2% | 62.6% | -0.4% |
Base: All participants in survey (Health Survey Condition: 1,051; GSGB Condition: 1,013, unweighted).
| Experimental condition | Health Survey Condition (percentage) | GSGB Condition (percentage) | Difference (percentage) |
|---|---|---|---|
| Online gambling participation | 15.3% | 19.8% | +4.5%* |
| Online only participants | 1.6% | 3.1% | +1.5%* |
Base: All participants in survey (Health Survey Condition : 1,051; GSGB Condition: 1,013).
Among participants who had gambled in the past 12 months, the GSGB Condition yielded a slightly higher percentage of people scoring 8 or more on the PGSI (5.1 percent), compared with the Health Survey Condition (4.0 percent). However, due to low numbers of participants in this category, our analyses lacked sufficient statistical power to determine whether this difference was significant.
To address this limitation, we expanded our analysis to include a broader range of PGSI scores. We combined the percentage of people who had scored 3 to 7 on the PGSI, with the percentage of people who had scored 8 or more on the PGSI. This approach provided a larger sample size and therefore allowed for more robust statistical comparisons. Subsequent analyses of combined PGSI categories showed that the GSGB Condition yielded a significantly higher proportion of participants scoring 3 or more on the PGSI, relative to the Health Survey Condition (12.9 percent versus 11.0 percent). There were no significant differences between conditions in the proportion of participants who scored 0 or 1 to 2 on the PGSI (as shown in Table 3: PGSI scores amongst participants who had gambled in the past 12 months, by experimental group).
| PGSI category | Health Survey Condition (percentage) | GSGB Condition (percentage) | Difference (percentage) |
|---|---|---|---|
| 0 | 72.0% | 72.8% | +0.8% |
| 1 to 2 | 16.0% | 14.3% | -1.7% |
| 3 to 7 | 7.0% | 7.8% | +0.8% |
| 8 or more | 4.0% | 5.1% | +1.1% |
1 It is important to note that statistical significant doesn’t guarantee practical importance. It only assesses the likelihood of results occurring by chance.
In response to Professor Sturgis’ independent review of the Gambling Survey for Great Britain (GSGB), we conducted an experiment to assess the impact of changes in gambling activity lists on estimates of gambling participation and the prevalence of people scoring 8 or more on the Problem Gambling Severity Index (PGSI). Consistent with our hypothesis, we found that the GSGB activity list identified a significantly higher proportion of people who gambled online compared to the Health Survey list. There was also a slightly higher percentage of participants scoring 8 or more on the PGSI in the GSGB condition, however this difference was not statistically significant due to low sample sizes. When combining PGSI risk categories of 3 to 7 and 8 or more, we found that the GSGB condition yielded a significantly greater proportion of people scoring within higher risk PGSI categories.
Our findings suggest that the list of activities provided in the GSGB more effectively identifies people who gamble online, compared with the list of activities used in the health surveys of England, Scotland and Wales. This enhanced ability to capture people who gamble online may contribute to the higher prevalence of people scoring 8 or more on the PGSI in the GSGB relative to previous health surveys. Specifically, by including a wider range of online gambling activities, the GSGB likely captures more people displaying ‘at-risk’ patterns of gambling who may have been missed by previous gambling surveys. This improved identification of people who gamble online is crucial for developing a comprehensive understanding of gambling behaviours, particularly given the recent growth of online gambling platforms.
As with all research, this experiment has several limitations that should be considered when interpreting the findings. Firstly, the sample sizes for some subgroups were relatively small, particularly for those scoring 8 or more on the PGSI. Generally, cell counts below 30 are considered insufficient for robust statistical analysis, and so we were unable to test for significant differences between these groups. We addressed this issue by combining PGSI risk categories, but further research using larger samples is needed to understand how differences in activity lists between the GSGB and health surveys affect the prevalence of people scoring 8 or more on the PGSI. Secondly, the online panel service used to recruit participants may have introduced a bias in the sample. Participants recruited through online channels may be more likely to engage in online gambling, and so the generalisability of our findings to the general population may be limited.
We will continue to investigate possible explanations for the differences in PGSI estimates provided by the GSGB and previous health surveys, and have outlined how we plan to improve the survey and implement Professor Sturgis’s recommendations. Given the high priority of these recommendations, we have issued an Invitation to Tender (ITT) to invite external organisations to undertake 3 experimental recommendations on our behalf, including replicating this experiment with larger samples. This is important to maintain the independence of the experiments and to progress in a timely manner.