Cookies on the Gambling Commission website

The Gambling Commission website uses cookies to make the site work better for you. Some of these cookies are essential to how the site functions and others are optional. Optional cookies help us remember your settings, measure your use of the site and personalise how we communicate with you. Any data collected is anonymised and we do not set optional cookies unless you consent.

Set cookie preferences

You've accepted all cookies. You can change your cookie settings at any time.

Skip to main content

Report

Gambling Survey for Great Britain Year 2 topic report: Investigating the profiles of those who gamble more frequently

This topic report uses data from Year 2 of the GSGB to explore the association between engagement in gambling activities, frequency of gambling, and risk.

  1. Contents
  2. Analysis of weekly activities and PGSI scores

Analysis of weekly activities and PGSI scores

The association between gambling at least weekly on certain product types and Problem Gambling Severity Index (PGSI) scores was explored further to identify which product types had a stronger relationship with PGSI scores of 8 or more.

Multivariate analysis

A series of regression models were run to examine whether there were associations between weekly engagement in specific gambling activities and PGSI scores, whilst taking into account broader gambling behaviours and socio-demographic characteristics.

For each activity two models were run:

  • negative binomial model where the outcome variable was PGSI score
  • a binary logistic regression models, where the outcome variable was whether someone had a PGSI score of 8 or more.

In both instances the models included the same set of demographic characteristics used in the profile comparisons of gambling frequency groups, namely: age, sex, ethnicity, tenure, qualifications, economic activity, household income, marital status, whether there were children present in the household, deprivation indicators, and region.

In addition, the models included the number of additional activities that the individual had participated in weekly (that is, the total number of activities, excluding the specific activity being tested in the model).

The two models measure different aspects of the relationship between activity and PGSI scores:

  • negative binomial regression models3 the relationship across the whole range of PGSI, from 0 to 27. A significant relationship here implies the activity, when done weekly, is significantly associated with incremental increases across the whole range of PGSI score

  • the logit model4 is focused on the relationship between activity and a PGSI score of 8 or more. A significant relationship here implies the weekly activity is associated with an increased risk of having a score of 8 or more.

Models were run on weighted data. For each model, the base is the people who had participated in that activity (or sets of activities5) in the past 12 months.

Results

In an accompanying set of data tables, table 11 shows the frequency of gambling for each activity and the mean PGSI score of people taking part in each activity, by frequency. The regression output is summarised in tables 12 and 13, with full output presented specifically in tables D1 and D2. These tables show Incidence Risk Ratios (IRR) from the negative binomial regression models and the Odds Ratios (OR) from the logistic regression models. IRR are interpreted in a similar manner to OR. An IRR of 1.25 means that activity is associated with a 25 percent increase in PGSI score, whereas an OR of 1.25 would mean the activity is associated with a 25 percent higher likelihood of having a PGSI of eight or more.

The models show weekly gambling on non-National Lottery instant wins, weekly online and in person fruits or slots, and in-play betting are significantly associated with higher PGSI scores. Additionally, it shows these weekly activities are associated both with increases across the full range of PGSI scores, and with having a PGSI score of 8 or more. The relationship between having a PGSI of 8 or more and both weekly in play betting and gambling weekly on fruits and/or slots in person, are particularly strong; for both activities the likelihood of having a PGSI of 8 or more is over 3 times higher than it is for individuals who gamble less frequently on these activities.

There is also a significant, but negative, relationship in both models for weekly National Lottery online draws, which suggests lower PGSI scores amongst those who participate in weekly lotteries online, regardless of how many other gambling activities the individual takes part in. Similarly, there are two activities, National Lottery draws in person and bingo in person, that are also associated with lower PGSI scores, but only in the negative binomial models. Whilst both activities are associated with lower PSGI scores, there is no evidence of a relationship with PGSI scores of 8 or more. These results suggest participation in these activities is not associated with higher levels of harms.

Finally, there were 3 activities where there was a significant association in the logit but not the negative binomial, meaning weekly participation in each activity was associated with a PGSI score or 8 or more, but not associated with PGSI scores overall. One of these (other charity lotteries in person) had a negative relationship. This means weekly purchase of charity lotteries in person is significantly associated with having a PGSI lower than 8, and unrelated to PGSI more widely when PGSI is treated as a continuous score.

The remaining two (betting on outcomes of a non-sporting event online and other non-National Lottery scratch cards) have a positive relationship with PGSI, meaning weekly gambling on these activities is associated with a higher chance of having a PGSI score of 8 or more. Betting on outcomes of non-sporting events online have odds that are nearly three times as high, and other non-National Lottery scratch cards nearly twice as high.

The distribution of PGSI scores in accompanying data table 11 is useful in interpreting this finding; for both these activities the mean PGSI score of those participating fortnightly is similar or marginally higher than the PGSI score for those participating weekly, which means there is no clear association between activity frequency and increasing PGSI score that would be picked up by the negative binomial modelling. However, in the logit model, the PGSI scores are collapsed into two groups; the model indicates that both betting on outcomes online and other non-National Lottery scratch cards, when carried out weekly, are associated with a higher likelihood of a PGSI score higher than 8, and are therefore associated with greater harms.

The remaining activities were not significantly related to PGSI score in either model. These were: Other charity lotteries online, National Lottery scratch cards, National Lottery Instant Win, Betting on outcomes of non-sporting events in person, Bingo online, Casino games online, Casino games at a casino, Casino games at a machine, and any sports betting not in-play (online or in-person).

Summary

Previous research has identified those who gamble on a weekly basis as having increased risk for the experience of gambling harms6. Analysis presented here examined whether engagement in certain kinds of gambling activities on a weekly basis was associated with increased PGSI scores.

Findings show that weekly gambling on fruit or slot machines, both in person or online, are significantly associated with elevated PGSI scores and having a PGSI score of 8 or more. Equally, betting in play on a weekly basis was also significantly associated with elevated PGSI scores and having a PGSI score of 8 or more, whereas betting on sports excluding in play betting was not associated with either.

These findings are both commensurate with existing knowledge of the types of products more likely to be associated with harms, being continuous, rapid reward products7 8 that bridge the gap between online casinos and/or slots and online betting9.

It is notable, however, that in this analysis other continuous gambling forms, like weekly engagement in online casino products, were not associated with PGSI scores. This may be because engagement in other activities or the characteristics of those who engage better explains this expected association. It may also be because less frequent engagement (that is, less often than weekly) is also strongly associated with PGSI scores, meaning the differences between weekly and less frequent engagement are obscured. In short, there is a need to look at risk curves for frequency of engagement in specific activities to better understand the frequency levels at which risk increases for people who take part in each specific activity.

These data are cross-sectional with attendant issues for causality. The models control for wider gambling involvement (measured by engagement in a number of other gambling activities on a weekly basis) but also the demographic and socio-economic profile of participants. There may, however, be some other unmeasured factor influencing results. Irrespective of the causal direction, there is a strong association between weekly in-play betting and weekly gambling on fruit and slots machines and their online equivalents. Operators providing these products should be aware of the enhanced risk among those gambling on these products most frequently.

References

3 Negative binomial regression models are well-suited to modelling outcomes, such as PGSI, that are non-negative (0 or greater), skewed, and contain a lot of 0 values.

4 Logit models are suited to outcomes that are binary (have two possible values), in this instance whether the individual has a PGSI score of eight or more, or not.

5 Because of specific filtering used in the online questionnaire, the bases for each activity varies. For example, the base for the model exploring PGSI and weekly participation in online casino games is anyone who played any casino games (online, at a machine, or in person) in the past 12 months. This reflects how the questionnaire was filtered on grouped activities.

6Currie SR, Hodgins DC, Wang J, El-Guebaly N, Wynne H. In pursuit of empirically based responsible gambling limits. International Gambling Studies. 2008;8(2):207–227. doi: 10.1080/14459790802172265

7Allami, Y., Hodgins, D. C., Young, M., Brunelle, N., Currie, S., Dufour, M., Flores-Pajot, M., & Nadeau, L. (2021). A meta-analysis of problem gambling risk factors in the general adult population. Addiction. https://doi.org/10.1111/add.15449

8Wardle H., et al (2024) The Lancet Public Health Commission on gambling, The Lancet Public Health, Volume 9, Issue 11,2024, Pages e950-e994,ISSN 2468-2667,https://doi.org/10.1016/S2468-2667(24)00167-1.

9Killick, E. A., & Griffiths, M. D. (2018). In-play sports betting: A scoping study. International Journal of Mental Health and Addiction 17(2) DOI:10.1007/s11469-018-9896-6

Previous section
Investigating the profiles of those who gamble more frequently - Composition of the gambling frequency groups
Next section
Investigating the profiles of those who gamble more frequently - Appendix A
Is this page useful?
Back to top