Optimising survey design for a national gambling prevalence study

Key messages

• Australia does not currently have a routine national gambling prevalence study. Existing state, territory and occasional national studies use different methods and are conducted at different times, limiting the ability to track gambling participation and harm consistently across Australia over time. Recognising this gap, the parliamentary inquiry into online gambling ‘You win some, you lose more’ recommended ongoing funding for prevalence studies to understand the extent of gambling harm in Australia (House of Representatives Standing Committee on Social Policy and Legal Affairs, 2023).
• The National Gambling Prevalence Study Pilot is the first attempt to establish a nationally representative gambling prevalence study in Australia. Conducted in 2024 by the Australian Gambling Research Centre (AGRC) at the Australian Institute of Family Studies (AIFS), the Pilot was designed to test whether consistent national data on gambling participation and harm can be collected in a rigorous, affordable and repeatable way.
• Survey design matters. Methodological choices about survey methods, recruitment strategies and survey names can significantly affect estimates of gambling participation and harm.
• A mixed-mode approach offers the best balance of rigour and affordability. Combining survey modes can reduce bias and improve representativeness while costing substantially less than traditional telephone-only approaches.
• The Pilot provides a practical blueprint for an ongoing national gambling prevalence study. It demonstrates how Australia could collect consistent national data to monitor trends, evaluate policy reforms and better understand gambling-related harm over time.

Background

Australia’s gambling landscape is changing rapidly, particularly with the growth of online wagering, changing gambling products and ongoing harm-reduction reforms. An ongoing, methodologically consistent national gambling prevalence study is needed to monitor these changes, evaluate policy impacts and provide comparable evidence to guide future reforms.

In 2024, the Australian Gambling Research Centre (AGRC) at the Australian Institute of Family Studies (AIFS) conducted a National Gambling Prevalence Study Pilot (‘the Pilot’). The Pilot examined the most cost-effective and methodologically rigorous approach to running an ongoing gambling prevalence study. The subsequent prevalence data collected using the Pilot are reported in a separate Key Findings report (Tillman et al., 2025a). These cover gambling participation, harm, mental health, intimate partner violence and finances.

This report identifies a methodologically rigorous and cost-effective design for an ongoing national gambling prevalence study. While the report does not seek to harmonise all existing state and territory studies, it provides evidence on survey design features that could support greater consistency and comparability in future prevalence research.

Methods

The Pilot used a partially crossed factorial survey design to test how different sampling frames, survey methods and survey names influenced gambling prevalence estimates. Respondents were recruited through 3 probability-based sampling frames: random mobile numbers, address-based sampling and the Life in Australia™ online panel. Respondents completed the survey either by telephone or online, depending on the recruitment pathway (see Methodology section for further details).

For the Pilot, 3,881 Australians over the age of 18 were recruited via 4 survey modes:

• dialling random mobile numbers to invite potential participants to a phone survey (CATI)
• sending SMS invitations to random mobile numbers to complete the survey online (SMS)
• sending invitations in the mail to a random sample of addresses using address-based Sampling (A-BS), where participants complete the survey online
• sending invitations to an existing probability-based online panel (Life in Australia^TM) to complete the survey online (LinA).

All respondents were randomly assigned to receive an invitation to a survey named as either a ‘gambling’ or ‘health’ survey, with the survey content identical.

Key findings

Methodological choices about survey methods and names can impact the accuracy and quality of gambling prevalence data, which guides gambling policy and public understanding.

• There are differences in the overall prevalence estimates based on how participants were recruited and how participants provided their responses.
• Respondents who received a ‘gambling’ framed invitation reported higher levels of gambling participation and gambling harm rates than those who received a ‘health’ framed invitation.
• Compared with CATI, A-BS produced higher gambling participation, while SMS and A-BS produced higher gambling harm scores.
• Compared with CATI, online surveys produced higher online gambling participation.
• Survey method and name effects are likely to reflect both participation and response effects.
• The ‘gambling’ framed invitation modes were more likely to include people with higher gambling involvement and gambling-related risk compared to ‘health’ survey invitations.
• Online completion modes were more likely than CATI to include people who gamble online.
• CATI may increase social desirability bias because it involves a human interviewer, potentially contributing to lower reporting of gambling harm.¹
• Attribution of harm to specific gambling activities was not substantially affected by survey method. Electronic gaming machines (EGMs) accounted for the largest share gambling harm, followed by wagering activities.
• EGMs consistently accounted for the largest share of explained variance (34.2%–46.2%) in gambling harm, followed by wagering activities (27.9%–36.5%).
• Poker, keno, scratch tickets, casino table games and bingo ranked next but added relatively smaller contributions. Lottery had negligible contributions to harm.
• Although effect sizes varied by survey mode, the overall ranking remained consistent.
• Mixed-mode reduced overall bias and lowered costs.
• A-BS, SMS and panel-based survey modes were more cost-effective than CATI but each mode introduced its own unique biases. The cost per complete (CPC) for the online panel was only 14% that of CATI, SMS was 20% and A-BS was 28%.
• A blend of 60% CATI, 29% SMS and 11% A-BS achieved the lowest average absolute bias (AAB = 2.64 percentage points) at 67% the cost of CATI alone.
• The best value for money was a blend of 40% CATI and 60% A-BS at a cost of 57% of CATI alone and an AAB of 2.67.
• A future ongoing national gambling prevalence study could use mixed-mode to balance cost, representativeness and methodological rigour.
• Among the survey modes modelled in the Pilot, mixed-mode provided the best balance of cost and representativeness.
• The Pilot findings can be used to help researchers and policy makers undertake prevalence surveys within tight budget constraints.

¹ See Heirene et al., 2025; Sturgis, 2020; Sturgis et al., 2025.