Report
Gambling behaviour and harm among people from different ethnic groups
This report examines how gambling participation, gambling risk, adverse consequences and use of support services vary among people from different ethnic groups.
Methods
This section references information that can be found in the ethnic group topic report findings excel files that accompany this report.
Bivariate analyses
Bivariate analyses were used to look at how gambling behaviours, Problem Gambling Severity Index (PGSI) scores, adverse consequences and help seeking vary among people from different ethnic backgrounds. The gambling behaviours evaluated included overall participation, the total number of activities and participation in specific activity types. Both potential and severe adverse consequences were assessed. To account for the complex survey design of the Gambling Survey for Great Britain (GSGB), all bivariate analyses used weighted cross-tabulations, with chi-squared tests used to assess statistical significance between groups.
Overall participation rates can be influenced by the popularity of lottery draws and mask wider variations in gambling participation among sub-groups. As such, additional analyses were performed excluding participants who had only taken part in lottery draw-based activities, to provide a clearer picture of the variation in gambling behaviour.
Multivariable analyses
Negative binomial models were used to explore the reasons why PGSI scores and adverse consequences vary by ethnic background. All models accounted for the complex survey design.
PGSI score
To examine PGSI scores among participants who gambled in the past 4 weeks, 3 sequential models were tested:
Model 1 (unadjusted): Compares PGSI scores across ethnic groups using those from White backgrounds as the reference baseline (as shown in Table 6 in the ethnic group topic report findings).
Model 2 (demographic and socio-economic adjustment): Controls for sex, age, area-level deprivation, income, employment status and level of education. Statistically significant results indicate that differences in PGSI score across ethnic groups persist regardless of demographic and socio-economic factors (as shown in Table 7 in the ethnic group topic report findings).
Model 3 (participation in gambling activities): This repeats model 2 but also includes whether a participant reported participating in any of 21 gambling activities in the past 4 weeks. Statistically significant results indicate that differences in harm scores persist regardless of demographic, socio-economic and both the volume and specific types of gambling activities reported across the 21 gambling activities (as shown in Table 8 in the ethnic group topic report findings).
Adverse consequences
Potential and severe adverse consequences were combined into a single harm score to capture the overall severity and breadth of gambling-related harm in one outcome measure (see definitions for a full description). The 3 sequential models tested were:
Model A (unadjusted): This compares the harm scores across ethnic groups, relative to those from White backgrounds (as shown in Table 10 in the ethnic group topic report findings).
Model B (demographic and socio-economic adjustment): Controls for sex, age, area-level deprivation, income, employment status and level of education. Statistically significant results indicate that differences in harm score across ethnic groups persist regardless of demographic and socio-economic factors (as shown in Table 11 in the ethnic group topic report findings).
Model C (participation in gambling activities): This repeats model B but also includes whether a participant reported participating in any of 21 gambling activities in the past four weeks. Statistically significant results indicate that differences in harm scores persist regardless of demographic, socio-economic and the number/type of activities an individual engages in (as shown in Table 12 in the ethnic group topic report findings).
Interpreting model results
Model results are reported as Incidence Rate Ratios (IRRs) relative to participants from White backgrounds (the baseline reference group):
- an IRR above 1 indicates a higher average score (higher PGSI score or harm score) than the reference group
- an IRR below 1 indicates a lower average score than the reference group
- an IRR equal to 1 indicates no difference between groups.
Each IRR is presented with a 95% Confidence Interval (CI) to indicate estimation precision. If the confidence interval does not overlap 1, the difference between that ethnic group and the reference baseline is statistically significant.
Sample approach
Because overall gambling participation varies considerably by ethnic group, PGSI scores and adverse consequences were examined among those who had gambled in the past 4 weeks. This ensures a consistent and comparable baseline population across analyses.
Use of support services was examined among those who had gambled in the past 12 months, as this is the only timeframe for which this measure was collected.
Previous sectionDefinitions - Gambling behaviour and harm among people from different ethnic groups Next section
Results - Gambling behaviour and harm among people from different ethnic groups
Last updated: 27 October 2026
Show updates to this content
No changes to show.