What do gambling prevalence studies really tell us?

Prevalence studies are frequently used in the field of gambling studies and are often seen as the pinnacle of good practice within the field. There are a number of good reasons why prevalence studies are important. In an article I co-wrote with Dr. Richard Wood (GamRes Ltd, Canada) we said that some of the benefits of prevalence studies are that they:

• Provide indicative data on the broad extent of clinical need for the overall population and sub-populations, general population risk factors, and some correlates of a particular disorder. This is useful information for many different stakeholders including those who have responsibility for programmes concerning intervention, treatment and social responsibility.
• Identify groups of people (for example, 18-24 year olds) where apparent needs do not match up with treatment service use. If we just surveyed treatment populations and/or those who attend Gamblers Anonymous, we would almost inevitably conclude that most problem gamblers are primarily white middle-aged men who typically have problems gambling on horse racing and/or casino games because females, various ethnic groups, and youth are disproportionately represented in treatment. It can also provide new research questions such as why such groups are not accessing treatment services.
• Allow comparison of different regions (within country or across counties) in terms of prevalence and their association with game availability, treatment availability, economic prosperity, crime rates, etc.
• Provide a snapshot of the life of a ‘normal’ gambler at a time of our choosing, rather than theirs. In contrast, clinical samples are consistent with people in crisis. We cannot always learn about the “normal” state of gambling, and how individuals can stay that way, from clinical samples.
• Provide attitudes and beliefs and behaviours in the general public (i.e., non-affected people) rather than non- representative groups (like problem gamblers).

However, they have very little explanatory power for understanding the development of problem gambling. This is the case for several reasons. Dr. Wood and I have argued that:

• Problem gambling is non-normally distributed across populations: Prevalence surveys select a sample that is representative of the entire adult population. However, problem gamblers are not equally distributed amongst that population and are therefore under- represented in general population surveys. For example, problem gambling in the UK is usually more prevalent amongst males, 18-24 age groups, those on lower incomes, for instance. Consequently, the actual prevalence of problem gambling may be higher.
• Problem gambling is a ‘sensitive’ issue for participants: Given that gambling is a behaviour that most problem gamblers do not want to talk about, they are much more likely than non-problem gamblers to refuse to agree to participate in any survey. (Conversely, those who do not gamble at all may also be under-represented in gambling surveys as they may feel that the issue is no concern of theirs).
• Non-response from problem gamblers: If problem gamblers happen to be in a household that is surveyed, they are much less likely to return the form than non-problem gamblers. If they happen to be in a household surveyed, they are less likely to return the call or form. Many may make themselves unavailable to answer survey questions if appointments are made to interview them. Furthermore, problem gamblers who agree to be surveyed are more likely to lie about the amount of time and money they spend on gambling, and about the frequency of their gambling – especially if they have not told their family that they have a problem and their family are not aware of the extent of their gambling. They are even more likely to lie during a survey if another family member is at home when they are answering the survey takers questions. No matter what the interview technique, households are not always places that encourage disclosure of information. Furthermore, household telephone interviews may also facilitate non-response as it is harder for problem gamblers to be honest when compared to self-completion methods. PGs are often in denial until they reach a point where they either get discovered or ask for help.
• Small numbers of problem gamblers: One of the real disadvantages of prevalence surveys is that they do not tell us very much about problem gambling. Although prevalence surveys can highlight slight fluctuations in problem gambling rates in comparison with other prevalence surveys, they do not tell us very much about problem gambling itself. The two recent British Gambling Prevalence Surveys (BGPSs) had approximately 55 to 70 people were identified as problem gamblers. Many qualitative studies (including treatment) studies have bigger samples of problem gamblers than that but are classed as unrepresentative.
• Gambling data from diverse groups may be unrepresentative: Some have argued that gambling prevalence surveys rarely capture responses from Culturally and Linguistically Diverse (CALD) groups. Some studies have found that gaming environments such as casinos comprise a disproportionate number of individuals from CALD groups.
• Problem gambling is not uniformly distributed in the population: Given that many prevalence surveys such as the BGPSs are household surveys, it should be noted that problem gamblers are more likely to be homeless and/or to be institutionalized (in prison, in mental hospitals), and therefore not even accessed to survey about their gambling behaviour in the first place.
• Unknown effect of false positives and false negatives on problem gambling estimates: One of the most highlighted problems is that when it comes to the screening instruments used to identify problem gambling, we do not know what effect false positives and false negatives have on the data. Typical survey samples worldwide are rather small (1,000 to 10,000 depending on population size). Therefore, the actual numbers of problem gamblers on which conclusions (and policy decisions) are made are very small (e.g., just over 50 problem gamblers in the case of the latest BGPS). To overcome the problem of small numbers and their analysis, the researchers often collapse sub-clinical and clinically significant cases of interest together. This analysis usually fails to consider the impact of false positive (in the sub-clinical group) on the validity of the conclusions drawn.
• Response rates to national surveys are decreasing: The response rates for national gambling surveys have been decreasing internationally. This may decrease the prevalence of problem gambling as problem gamblers are more likely to be in the group of non-responders.
• Survey response may differ as a function of media exposure to problem gambling: Australian researchers have argued that any given moment in time, the number of people surveyed who will admit to having a gambling problem is dependent on how much media attention has been given to concerns about gambling losses, and the level of problem gambling in the community. Shame and guilt (and therefore lying about gambling involvement) are apt to increase as public concern about gambling and gambling losses increases and as media reports become more prevalent and shocking.
• Random samples are still self-selecting samples: Even though most national gambling prevalence surveys are random it could still be argued that those who are approached still ultimately decide whether or not to participate and in that sense the sample is still self- selecting.
• Self-report methods can be problematic: The use of anonymous self-report methods may allow people to be economical with the truth and/or exaggerate and lie about certain issues. This is coupled with the fact that they may be asked things on which they have to rely on long-term memory (which may not be the most reliable). Furthermore, it is easy for a respondent to exaggerate or lie when they know that they are relatively anonymous and that nobody will question the validity of their answers.
• Actual problematic gambling behaviour is rarely considered in large-scale surveys: In order to overcome question fatigue and to increase participation rates, very few questions in large prevalence surveys actually focus on gambling problems beyond the screen questions used to identify people with problems. This leaves correlational factors only that are often basic demographics (e.g., age, location, etc.) or frequency questions (how often they play, etc.), that by themselves they do not provide much information as to why problems develop.
• Lack of theory-driven and/or model-driven research: In almost all gambling prevalence surveys there is a great emphasis on closed (forced) question responses rather than allowing respondents to explain what the issues are for their specific gambling behaviour (i.e., the studies are more about ‘data trawling’ rather than ‘theory building’). This also means that we are just measuring fluctuations rather than developing and testing theories that help us understand the fundamental issues.
• Understanding severity: There appears to be an assumption that endorsing one or two items on a problem gambling screen indicates a problem at a low level when there is little evidence to support this. Whilst endorsing the specified number of criteria on a diagnostic screen may be a good indicator of a gambling problem, the scores for endorsing one or two items may not have been validated as an indicator of a lesser problem. Answering in this way to one or two items may in fact indicate the extent of ‘normal’ risk inherent in gambling activities.

By highlighting some of the problems of prevalence surveys, Dr. Wood and I are not saying that these should not be carried out (as they clearly have a use as outlined at the start). However, there are lots of other methodologies for examining and understanding problem gambling. We need to look at the lives of the problem gamblers in far more detail than the data collected from prevalence surveys. Future prevalence surveys should be complemented with other more ‘in-depth’ methodologies including interviews, focus groups, Q-sorts and online discussions

Dr Mark Griffiths, Professor of Gambling Studies, International Gaming Research Unit, Nottingham Trent University, Nottingham, UK

Further reading

Griffiths, M.D., Wardle, J., Orford, J., Sproston, K. & Erens, B. (2010). Gambling, alcohol consumption, cigarette smoking and health: findings from the 2007 British Gambling Prevalence Survey. Addiction Research and Theory, 18, 208-223.

Griffiths, M.D., Wardle, J., Orford, J., Sproston, K. & Erens, B. (2011). Internet gambling, health. Smoking and alcohol use: Findings from the 2007 British Gambling Prevalence Survey. International Journal of Mental Health and Addiction, 9, 1-11.

Griffiths, M.D. & Wood, R.T.A. (2009). Prevalence studies: What do they really tell us? Casino and Gaming International, 5(4), 102-104.

Orford, J., Griffiths, M.D., Wardle, J., Sproston, K. & Erens, B. (2009). Negative public attitudes towards gambling: findings from the 2006/07 British Gambling Prevalence Survey using a new attitude scale. International Gambling Studies, 9, 39-54.

Sussman, S., Lisha, N. & Griffiths, M.D. (2011). Prevalence of the addictions: A problem of the majority or the minority? Evaluation and the Health Professions, 34, 3-56.

Wardle, H., Moody. A., Spence, S., Orford, J., Volberg, R., Jotangia, D., Griffiths, M.D., Hussey, D. & Dobbie, F. (2011).  British Gambling Prevalence Survey 2010. London: The Stationery Office.

Wardle, H., Sproston, K., Orford, J., Erens, B., Griffiths, M.D., Constantine, R. & Pigott, S. (2007). The British Gambling Prevalence Survey 2007. London: The Stationery Office