This box is not visible in the printed version.
This report explores the relationship between participation in individual gambling activities in the past 12 months and Problem Gambling Severity Index Scores (PGSI).
Published: 30 January 2025
Last updated: 6 February 2025
This version was printed or saved on: 2 May 2025
Online version: https://www.gamblingcommission.gov.uk/report/exploring-the-relationship-between-gambling-activities-and-problem-gambling
The Key findings for this report are:
This short report further explores the relationship between participation in individual gambling activities in the past 12 months and PGSI. It examines the associations between engagement in specific gambling activities and PGSI scores when broader gambling behaviours and the socio-economic and demographic profile of people engaging in each activity is considered.
Participants were provided with a list of 18 types of gambling legally available in Great Britain and asked which of these activities they had participated in, in the past 12 months. They were also asked about any private betting or gambling they may have done in the same period. Participants who answered yes to participating in one or more gambling activity are described in this report as participating in gambling in the past 12 months.
The number of discrete gambling activities was calculated for those who had participated in a gambling activity in the past 12 months. Participants who had not participated in any gambling activity in the past 12 months were coded as not having taken part in any activities.
Everyone who had gambled on a particular activity in the past 12 months was then asked to report how often, if at all, they had gambled on this activity in the past 4 weeks. Using this information, frequency of gambling was computed for each participant based on their reported highest frequency of gambling to any activity in the past 4 weeks. For example, if a participant reported purchasing National Lottery tickets once a week but bet on sports events a few times a week, they were categorised as gambling a few times a week. Response categories were: everyday, a few times a week, about once a week, about once a fortnight, about once in the past 4 weeks, and less often than this (that is those who had gambled in the past 12 months, but not in the past four weeks).
The list of gambling activities presented to participants was refined and tested during the development stages of the Gambling Survey for Great Britain (GSGB). The aim was to update the list used in previous surveys to better represent forms of commercial gambling currently available in Great Britain, and to better capture engagement in different online gambling formats. Appendix A shows the full list of gambling activities asked about for the past 12 months.
The PGSI consists of nine questions and is asked of everyone who had gambled in the past 12 months, capturing the current experience of each of these items. Answer options were ‘almost always’, ‘most of the time’, ‘sometimes’, and ‘never’. Responses to the nine questions are summed and a score ranging between 0 and 27 is computed. Scores are grouped into the following categories1:
PGSI score 0
Representing a person who gambles (including heavily) but who does not report experiencing any of the nine symptoms or adverse consequences asked about. In population prevalence analysis, participants who had not gambled in the past 12 months are also given a PGSI score of 0.
PGSI score 1 to 2
Representing low risk gambling by which a person is unlikely to have experienced any adverse consequences from gambling but may be at risk if they are heavily involved in gambling.
PGSI score 3 to 7
Representing moderate risk gambling by which a person may or may not have experienced adverse consequences from gambling but may be at risk if they are heavily involved in gambling.
PGSI score 8 or higher
Representing problem gambling by which a person will have experienced adverse consequences from their gambling and may have lost control of their behaviour. Involvement in gambling can be at any level, but is likely to be heavy.
In July 2024, the Gambling Commission (the Commission) published the first annual report from the Gambling Survey for Great Britain (GSGB) based on data collected in 2023. Further details about the survey methodology including its strengths and limitations are provided in the GSGB Technical report. The annual report contained a chapter about the consequences of gambling, including results from the Problem Gambling Severity Index (PGSI). In addition to examining PGSI scores by age and sex, analyses examined the proportion of people undertaking each gambling activity in the past 12 months who had a PGSI score of 8 or more. Specifically, the proportion of people with a PGSI score of 8 or more taking part in each activity was compared with the average for all people who had gambled in the past 12 months. This analysis indicated which gambling activities had higher than average proportions of people with a PGSI score of 8 or more, which had lower than average, and which had the same as average, among those using each respective product. Data from the GSGB year 1 report is shown in Figure 1, where a relative risk difference of 1 means the results for that activity are the same as average for all people who had gambled in the past 12 months, a difference greater than 1 shows they are higher than average and a difference less than 1 means they are lower than average.
Figure 1: Relative difference between activities in proportion with PGSI score of 8+, compared with overall proportion with PGSI score of 8+
Gambling activity in the past 12 months | All participants: Gambled in the past 12 months (relative difference ratio) (number) |
---|---|
Betting on non-sports events (in person) | 9.9 |
Online fruit and slots | 5.9 |
Casino games on a machine and or terminal | 5.5 |
Casino games at a casino | 5.2 |
Online casino games | 5.2 |
Betting on non-sports events (online) | 5.1 |
Online bingo | 4.9 |
Non-National Lottery online instant wins1 | 4.8 |
Football pools | 4.5 |
Fruit and or slots machines | 3.7 |
Non-National Lottery scratchcards2 | 3.4 |
Betting on sports and or races in person | 3.1 |
Online betting on sports and or races | 2.1 |
National Lottery online instant wins | 2.0 |
Bingo in person | 2.0 |
Private betting | 1.9 |
National Lottery scratchcards | 1.8 |
Charity lottery draws | 1.1 |
National Lottery draws | 0.9 |
For those buying tickets for the National Lottery, the proportion with a PGSI score of 8 or more was similar to the average for all people who had gambled in the past 12 months.
Among those gambling on fruit and or slots machines, online bingo, betting on non-sports events online, football pools, online casino games, casino games at a casino, casino games played on a machine or terminal, and non-National Lottery online instant wins, the proportion with a PGSI score of 8 or more was between 4 to 5 times higher than average.
Finally, both online slots and betting on non-sport events in person had substantially higher than average proportions of people with a PGSI score of 8 or more. For online slots, rates were nearly 6 times higher than average and for betting on non-sports events in person, they were over 9 times higher than average.
Whilst these estimates are informative, showing which specific activities have a greater proportion of people with a PGSI score of 8 or more among their user base (itself an important metric for the identification of risk among different types of people who gamble), the strength of association may be influenced by a range of other factors.
It has previously been suggested that the association between specific gambling activities and problem gambling is driven not by engagement in specific types of gambling but rather by the fact that people who experience gambling problems engage in many different activities. This was first postulated by LaPlante et al (opens in new tab), who analysed data from the British Gambling Prevalence Survey 2007, observing that the relationship between specific gambling activities and problem gambling attenuated or became statistically insignificant once the number of other gambling activities undertaken was accounted for. They concluded that “greater gambling involvement better characterises disordered gambling than does any specific type of gambling”. Several studies have found similar results, though others have noted ongoing associations between some forms of gambling and problem gambling even after wider gambling involvement was taken into account. For example, Binde et al (2017) (opens in new tab) examined data from Sweden, finding that “that some forms of gambling are more closely associated with problem gambling than other forms, and that gambling policy and regulation, as well as the development of responsible gambling initiatives, should focus on these forms.” Likewise, analysis of data from Massachusetts found that gambling involvement mediates the relationship between gambling formats and problem gambling, but also that casino gambling was most associated with gambling problems across all levels of gambling involvement (Mazar et al, 2020 (opens in new tab)).
The aim of this short report is to examine whether the associations between engagement in specific gambling activities and PGSI scores observed in the GSGB year 1 (2023) report hold once broader gambling behaviours and other factors are accounted for.
1Non-National Lottery online instant wins are those products available on operator websites often marketed as online scratchcards for example ‘Rainbow Rewards Scratchcard’
2Non-National Lottery scratchcards are those scratchcards bought in person but that are not sold on behalf of the National Lottery
To explore this, 3 binary logistic regression models were run for each of the 18 commercial gambling activities undertaken in the past 12 months. For each model, the outcome variable was whether someone had a PGSI score of 8 or more.
Unadjusted binary logistic regression model. This shows the odds of having a PGSI score of 8 or more among those taking part in each individual activity (see Table 1 in the Excel file).
Adjusted binary logistic regression model. This shows the odds of having a PGSI score of 8 or more among those taking part in each activity, with the number of other gambling activities undertaken in the past 12 months and the highest frequency of gambling across any activity also entered into the model. If a gambling activity remains significant in this model, it means the association persists even after the broader gambling involvement is taken into account, separating out the impact of the activity from an individual’s overall gambling (Table 2 in the Excel file).
Further adjusted model. This repeats model 2 but also includes the following variables: age, sex, education level, employment status, household income quintile, area deprivation, marital status and ethnicity, all of which have been previously identified as being associated with PGSI scores. As with model 2, if a gambling activity remains significant in the further adjusted model, it means the association persists even after the broader gambling involvement and other factors are taken into account, separating out the impact of the activity from an individual’s overall gambling and their demographic and economic characteristics (Table 3 in the Excel file).
All logistic regression models were based on those who had gambled in the past 12 months to identify differences among those who gamble.
Results from logistic regression models are presented as odds ratios. Results for each activity are interpreted relative to a reference category. Those who did not gamble on a specific activity are given a reference of 1, and the odds ratios for those who did gamble on this activity are interpreted relative to this group. This shows whether odds ratios are higher or lower among those that gambled on this activity than those who did not (but did gamble on other things). An odds ratio of more than 1 indicates higher odds of having PGSI score of 8 or more, an odds ratio of less than 1 indicates lower odds of having a PGSI score of 8 or more. Confidence intervals at the 95 percent level (95 percent CI) for each odds ratio are also shown. If this confidence interval does not include 1, there is a significant association between the gambling activity being considered and having a PGSI score of 8 or more.
The unadjusted odds ratios showed that engagement in every gambling activity, with the exception of National Lottery tickets, was significantly associated with having a Problem Gambling Severity Index (PGSI) score of 8 or more. The pattern was similar to that reported in the Gambling Survey for Great Britain (GSGB) year 1 (2023) report, whereby activities like betting on other events, either in person or online, gambling on casino games (in person or online or via a machine and or terminal), online gambling on fruit and or slot games and other online instant wins had substantially elevated odd ratios (greater than 10) (see Model 1, Table 1 in the attached Excel file).
When gambling frequency and number of gambling activities were taken into account (Model 2, Table 2 in the attached Excel file) the odds ratios for most activities substantially attenuated, though the association between specific gambling activities and a PGSI score of 8 or more remained significant for most. For example, for betting on non-sports events in person, odd ratios reduced from 24.42 (95 percent CI: 15.41-38.67) to 5.95 (95 percent CI: 3.17-11.18). Likewise odd ratios for gambling on online fruit and slot games reduced from 14.68 (95 percent CI: 10.28-20.96) to 3.72 (95 percent CI: 2.12-6.54). Nonetheless, participation in both activities in the past 12 months were still significantly associated with a PGSI score of 8 or more.
Activities which became non-significant between model 1 and model 2 were charity lotteries, football pools, bingo played in person, betting on sports and or races online and National Lottery online Instant Wins.
Tickets for the National Lottery was the only activity where the odds ratios of a PGSI score of 8 or more was significantly lower (0.41; 95 percent CI: 0.27-0.63) among those that purchased these in the past 12 months than those who did not.
Finally, comparisons between the unadjusted model (model 1) and the further adjusted model (model 3) showed 3 distinct patterns (illustrated in Figures 2-4):
Figure 2: odds ratios of a PGSI score of 8+ among those purchasing tickets for the National Lottery in the past 12 months
Logistic regression model type | Odds Ratio | Lower 95 percent confidence interval | Upper 95 percent confidence interval |
---|---|---|---|
Model 1: Unadjusted | 0.93 | 0.66 | 1.30 |
Model 2: Adjusted for number of gambling activities and gambling frequency | 0.41 | 0.27 | 0.63 |
Model 3: Adjusted for gambling and socio-economic and demographic characteristics | 0.62 | 0.39 | 0.98 |
Figure 3: odds ratios for betting on sports or racing online
Logistic regression model type | Odds Ratio | Lower 95 percent confidence interval | Upper 95 percent confidence interval |
---|---|---|---|
Model 1: Unadjusted | 4.18 | 2.98 | 5.85 |
Model 2: Adjusted for number of gambling activities and gambling frequency | 0.96 | 0.59 | 1.56 |
Model 3: Adjusted for gambling and socio-economic and demographic characteristics | 0.78 | 0.47 | 1.28 |
Figure 4: odds ratios of a PGSI score of 8+ among those gambling on fruit and or slot games online in the past 12 months
Logistic regression model type | Odds Ratio | Lower 95 percent confidence interval | Upper 95 percent confidence interval |
---|---|---|---|
Model 1: Unadjusted | 14.68 | 10.28 | 20.96 |
Model 2: Adjusted for number of gambling activities and gambling frequency | 3.72 | 2.12 | 6.54 |
Model 3: Adjusted for gambling and socio-economic and demographic characteristics | 3.23 | 1.94 | 5.37 |
The Gambling Survey for Great Britain (GSGB) Year 1 report (2023) showed a substantially high proportion of people who gambled on activities like casino games (online or in person), online fruit and or slots and betting on other sports events (online or in person) had Problem Gambling Severity Index (PGSI) scores of 8 or more. These data are useful for understanding the risk profile of people who engage with these activities. Similar data have been used previously to better understand the number of interactions gambling operators should make with higher risk consumers, irrespective of whether the relationship between the activity and PGSI score is driven by other factors.
It is, however, also important to explore whether other factors could explain these associations to better understand which products are associated with greater risk. The analysis presented here shows that casino gambling (either online or in person), fruit and slots machine gambling (either online or in person), betting on other sports events (either online or in person), betting on sports and or races in person, gambling on non-National Lottery online instant wins and gambling on non-National Lottery Scratchcards are all significantly associated with PGSI scores of 8 or more when broader gambling involvement and demographic and socio-economic status were taken into account.
Prior analysis of the British Gambling Prevalence Survey 2007 (opens in new tab) data suggested that measures of gambling involvement better characterised gambling disorder than engagement in specific activities. Other studies disagreed with this conclusion, noting that some gambling forms are more strongly correlated with the experience of problem gambling and that these formats require additional regulatory attention. The evidence presented here supports the latter conclusion, with findings from the GSGB year 1 (2023) data showing that activities like casino gambling and fruit and slot games, undertaken either online or in person, were significantly associated with PGSI scores of 8 or more.
Notably, many of the gambling activities most closely associated with a PGSI score of 8 or more represent those which are faster, continuous gambling formats (casino, slots, online instant wins). However, the results also highlighted betting on non-sports events and betting on sports/races at in person as being significantly associated with a PGSI score of 8 or more. It is less clear what is driving these associations. Betting on non-sports events is a low prevalence activity and previous studies, such as the British Gambling Prevalence survey series, typically demonstrated relatively high proportions of people experiencing problem gambling among those betting on non-sporting events. Low prevalence among the population and increased popularity among those experiencing adverse consequences may explain some of these associations.
Future replication of this analysis with subsequent years of the GSGB would be beneficial to explore the stability and consistency of the findings reported here. Supplementing year 1 data with that from future years will also allow analyses to be examined separately for men and women and will allow further examination of the association between specific gambling activities and PGSI scores among those who gamble more regularly (that is those who gambled in the past 4 weeks). Base sizes for year 1 data did not permit this level of analysis.
The results presented here are associations and whilst controls include wider gambling activity and certain demographic and economic characteristics, there may be other, unmeasured, factors underlying the observed associations. This requires further investigation, and the replication of analysis from different surveys with different available measures.
Nonetheless, the analysis presented in this report replicates the analytical approach used by LaPlante et al, extending it to include controls for gambling frequency and demographic and socio-economic status. The results challenge their conclusion that “greater gambling involvement better characterises disordered gambling than does any specific type of gambling”. Gambling involvement is clearly an important factor, as evidenced by the sharp attenuation of odds between models 1 and 2, but some activities remain strongly associated with PGSI scores of 8 or more even once this was taken into account, highlighting potential increased risk of adverse consequences among those who engage in these activities.
The list below shows the full description of gambling activities shown to participants when reporting past 12 month participation:
Wardle H and Tipping S. (2025) Exploring the relationship between gambling activities and Problem Gambling Severity Index (PGSI) scores. Gambling Commission: Birmingham.
In the last 5 years, Heather Wardle (HW) discloses grant funding for gambling-related research by the Economic and Social Research Council, National Institute for Health Research, Wellcome Trust, the Commission (including their regulatory settlement fund), Office of Health Disparities and Improvements; Public Health England; Greater London Authority; Greater Manchester Combined Authority; Blackburn with Darwen Local Authority and the Department of Digital Culture Media and Sport.
HW declares consulting fees from the Institute of Public Health, Ireland and the National Institute for Economic and Social Research.
HW declares payment for delivery of seminars from McGill University, the University of Birmingham, John Hopkins University and from the British Broadcasting Corporation.
HW has been paid as an expert witness by Lambeth and Middlesborough Borough Councils.
HW declares travel costs paid by Gambling Regulators European Forum, the Turkish Green Crescent Society, Alberta Gambling Research Institute; the REITOX Academy (administered through the Austrian National Public Health Institute) and the University of Helsinki.
HW served as Deputy Chair of the Advisory Board for Safer Gambling between 2015 and 2020, remunerated by the Commission; is a Member of the WHO panel on gambling (ongoing) and provided unpaid advice on research to GamCare for their Safer Gambling Standard (until mid-2021).
HW runs a research consultancy for public and third sector bodies only. She has not, and does not, provide consultancy services to gambling industry actors.
In researching the gambling industry and their practices, HW declares occasional attendance at events where gambling industry actors are present (including industry-sponsored conferences).
As part of her work on the Gambling Survey for Great Britain (GSGB), HW is required by the Commission (the funder) to participate in events disseminating research findings to their stakeholders, which includes the industry. Her attendance at events where industry is present is independently funded and does not involve collaborations or partnerships with industry.
In the last 5 years, Sarah Tipping (ST) declares funding for gambling projects from the National Institute of Health Research and the Gambling Commission.