![Sample our Sports and Leisure journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5949/TOC-Subject-Sample-2021-Sports-Leisure.jpg"})
Abstract
The present study investigated the impact of survey administration format, survey description and gambling behaviour thresholds on obtained population prevalence rates of problem gambling. A total of 3028 adults were surveyed about their gambling behaviour, with half of these surveys administered face-to-face and half over the telephone, and half of the surveys being described as a ‘gambling survey’ and half as a ‘health and recreation’ survey. Population prevalence rates of problem gambling using the CPGI were 133% higher in ‘gambling’ vs ‘health and recreation’ surveys and 55% higher in face-to-face administration compared to telephone administration. If people with less than Can$300 in annual gambling expenditures are not asked questions about problem gambling, then the obtained problem gambling prevalence rate is 42% lower. When all of these elements are aligned they result in markedly different problem gambling prevalence rates (4.1% vs 0.8%). The mechanisms for these effects and recommended procedures for future prevalence studies are discussed.
Acknowledgements
The authors would like to thank the Survey Research Centre at the University of Waterloo for their skilful and thoughtful implementation of this study as well as the Ontario Problem Gambling Research Centre for their generous financial support for this work.
Notes
1. This is not an unusual result, as response rates are usually higher in central government surveys relative to academic or nonacademic surveys (Groves & Couper, Citation1998).
2. It is interesting to note that the Quebec 2002 provincial study (Ladouceur, Jacques, Chevalier, Sevigny & Hamel, Citation2005) also utilised a high gambling threshold before asking problem gambling questions (i.e. person had to have spent more than Can$520 annually on gambling or have ‘played too much’, ‘spent too much money’, or ‘spent too much time gambling’), and it is one of the few provincial studies that obtained problem gambling rates comparable to the CCHS.
3. ASDE Survey Sampler (from whom the sample was purchased) indicated that approximately 2% of listed numbers had no accompanying address.
4. In the case of the CPGI, only the nine items that comprise the scored Problem Gambling Severity Index (PGSI) were included.
5. The impact of interviewer characteristics is not reported in the present paper because these variables were not balanced across conditions. In any case, analysis of gender and ethnicity failed to show a consistent pattern of effects on problem gambling prevalence rates, although there was a tendency for higher rates of problem gambling with male interviewers.
6. The ANCOVA procedure permits all six confounding demographic variables to be controlled for whereas conducting ANOVA and χ2 tests after SPSS data weighting only controls for the three demographic variables that have been weighted.
7. Fifty-one per cent of adults in the Kitchener CMA voted in the 2007 provincial election.
8. The SRC was asked to record any stated reasons for nonparticipation.
9. This is not an optimal statistical approach, as there is severe and uncorrectable skewness, kurtosis, and heterogeneity between the groups, which violates the statistical assumptions of ANOVA. However, violation of these assumptions may still produce valid results with large sample sizes, as we have in this study.
10. Obviously, this assumption may not be well founded, as these individuals have not been clinically assessed. Furthermore, there is still considerable debate as to the clinical significance of the CPGI moderate problem category.
11. The point is sometimes made that telephone surveys likely under-represent problem gamblers because they are less likely to be home, less likely to answer the telephone and more likely to be incarcerated, in residential treatment, or in military service. While there is some truth in this, it is not as strong an effect as one might expect. The first author has conducted two large scale prevalence studies (Williams & Wood, Citation2007, n = 6654; Wood & Williams, Citation2009, n = 8450) that used exhaustive telephone contact attempts (up to 36 in the first study and 48 in the second study) over several months (12 months in the first study and 18 months in the second study) and found that the average number of telephone calls to establish contact is only somewhat higher for problem gamblers compared to nonproblem gamblers (5.5 vs 4.9) and that 95% of contactable problem gamblers are contacted by 16 attempts (vs 14 for nonproblem gamblers). Furthermore, when adjustments are made for problem gamblers who are incarcerated, hospitalised, in residential treatment, or in the military, the problem gambling prevalence rate is only marginally changed (i.e. increases by ∼0.03%).