In February 2011, the Gambling Commission announced that the British Gambling Prevalence Survey 2010 had found ‘an increase in the number of problem gamblers in Britain’ and that ‘The proportions increased from 0.5% of the adult population in 2007 to 0.7% in 2010 (which is not statistically significant) on one measure and from 0.6% in 2007 to 0.9% in 2010 (which is at the margins of statistical significance) on the other measure used’[i], On the same day, the gambling Minister, John Penrose MP, told the Daily Mail; ‘‘The increase in problem gambling is a direct result of Labour’s reckless Gambling Act…The Labour government liberalised gambling laws but failed to implement the safeguards needed to protect the public and as a result the number of problem gamblers has risen to almost half a million”[ii].
Problem gambling behaviour had become ‘front page’ news for the first time as the previous Prevalence Survey in 2007 had shown no increase in the level of problem gambling from the previous one of 1999. The salient issue was that the 2010 Survey was the first since the introduction of the Gambling Act 2005 (which came into force in September 2007). Not surprisingly it was the Daily Mail trumpeting the increase, as it was they who had launched a vociferous campaign against the reforms which the Gambling Act 2005 brought in. At the time, their villain of the piece was Tessa Jowell MP, the then Secretary of State at the Department of Culture, Media & Sport (DCMS) who had sponsored the Bill and had categorically stated that gambling liberalisation wouldn’t happen if it caused a single new problem gambler.
This may have seemed like a rash promise to make by any politician considering liberalising Britain’s archaic gambling laws, but six years later, her successor would seem to be upholding her promise. In March 2011, when questioned by the Culture, Media & Sport Select Committee, the then Secretary of State for DCMS, Jeremy Hunt MP, argued that the introduction of the Gambling Act 2005; “hasn't worked, you are absolutely right. There have been a whole series of problems with it, but we have said to the gambling industry that the Government's policy on this is very clear. We won't consider any moves to liberalise gambling until we are confident that we have a coherent strategy in place to tackle problem and addiction gambling, and the latest prevalence study that only came out a few weeks ago said that problem gambling had increased to 0.9% of the population, which is a significant number of people, and we have to be very careful that we aren't letting the genie out of the bottle. In principle, we will look at any measures that people would like us to make, but we do need to address the issue of problem gambling and make sure that we are not embracing a policy that will make that worse.”[iii]
While the increase in those categorised as problem gamblers increased at only a statistically insignificant amount (according to the Gambling Commission) between 2007 and 2010, the numbers of people actually gambling increased by a relatively small amount. Between the 2007 and 2010 Surveys, it is estimated that those gambling increased from 68% of the adult population to 73% or an increase of 2.4 million people[iv], a similar level to the 1999 level of 72%[v]. So according to the Prevalence Surveys, the decade which saw probably the biggest liberalisation of gambling legislation in British history, saw a minimal increase in the numbers of people gambling normally and a possible small increase in those gambling abnormally.
This paper will look at the issue of problem gambling, or, as we prefer to call it; abnormal gambling behaviour, as this terminology is far more accurate. Abnormal gambling behaviour is something which afflicts a very small proportion of the population and is still not properly understood or researched. However, the estimated numbers of people who suffer from this condition has now been intrinsically linked with the fortunes of the British gambling industry and so how the numbers of sufferers are determined has become of great importance both politically and commercially. This is not, in any way, to diminish the seriousness of this condition, or its effects on the individuals and their loved ones, but this paper will be examining the issue of abnormal gambling behaviour from a policy perspective and taking a critical view as to the current level of understanding of abnormal gambling and how this has impacted on the gambling industry’s relationship with government.
To do this we will examine the British Gambling Prevalence Surveys (BGPS) of 1999, 2007 and 2010. These Surveys have become the main source of data for assessing the level of abnormal gambling in the UK and have so far been the primary data source for government. The 1999 and 2007 Surveys represent British gambling before the enforcement of the Gambling Act 2005 on September 1st 2007. The 2010 Survey represents the first survey conducted since the Act came into force.
The 1999 and 2007 Surveys pertain to a time period which includes some minor liberalisations in British gambling law but also probably some of the biggest changes in the gambling environment, the introduction of internet gambling and the roll out of fixed odds betting terminals in betting shops.
Internet gambling had started in the mid 90’s and became main stream within the UK by about 2001 when the population started getting broadband (1.6% of households in 2001 raising to 30.6% in 2010[vi]). Growth in internet gambling was assisted by the fact that the advertising of foreign internet gambling sites was considered legal and so there had been gambling advertising proliferating (though not via broadcast media) since before the 1999 Survey. According to the Gambling Commission’s Industry Statistics 2009/10 10.8% of the adult population have gambled online[vii], during the period of 1999-2007 this penetration would have been in increasing single digits.
On the land-based front; fixed odds betting terminals (high stakes gaming machines that offer casino type games) began rolling out in betting shops from about 2001 and rapidly reached a maximum penetration level of just under 4 per shop or 30,000+ nationally. Their introduction has arguably changed the nature of betting shops, but very much in response to customer demand. These have always been a contentious issue as they originated through a loophole in the law and have been criticised for having too high stakes and prizes for their premises. But their popularity is evidenced by the fact that they now represent half a betting shops income and are now a mainstay of the betting shop product range.
So although Britain has undoubtedly become a more gambling aware nation with the introduction of the Gambling Act 2005, the actual liberalisations that the Act brought in were not as enormous as certain commentators (e.g. faith groups, newspapers, politicians) have made out as some of the biggest changes had already happened and without a perceivable impact on abnormal behaviour.
The Gambling Act 2005 brought in a reduction in the number of gaming machines available (by removing machines from unlicensed premises such as minicab offices and cafes and by removing section 16 and section 21 machines from bingo halls, casinos and arcades), brought in for the first time some of the most stringent abnormal gambling behaviour prevention measures ever legislated for in any jurisdiction.
It did, however, remove the membership requirement for casinos and for bingo halls. Whether this liberalisation increased the number of people attending them is another matter. Casinos have seen a 5.5% increase in attendances since 2007[viii] but this needs to be considered as part of a trend as casino attendance have been increasing for a number of years, possibly due to the growth in the popularity of poker. Statistics aren’t readily available for bingo attendance but industry statistics show the number of bingo clubs has continued declining as has the amount of Gross Gaming Yield, so this would suggest attendance has been falling over the years.
Probably the biggest variable the Act introduced was the legalisation of gambling advertising, including television and radio. As stated above, there had been widespread non-broadcast advertising of internet gambling for the previous decade, but for the first time would we see adverts on terrestrial and satellite channels offering gambling products (mostly online). The Advertising Association state that advertising expenditure on gambling in the UK in 2010 was £150M (0.9% of total advertising spend)[ix]. This has undoubtedly led in some way to the increased normalisation of gambling but there is no hard evidence that this would directly cause any increase in abnormal gambling behaviour considering that the UK has a lengthy history of gambling outlets on the high street so access and awareness of gambling products has always been high.
This unfortunately is the state of the data we have when considering abnormal gambling. The only method we have used for all three British Gambling Prevalence Surveys is known as DSM IV and that shows a very large increase in abnormal gambling numbers and yet has some major issues with its utility as a method for estimating numbers (of which more later). When we consider the results of SOGS v PGSI screening methods, we see a more believable increase but then have to consider that they are two completely different screening methods and so can’t really be compared. Even if we consider the differences between 2007 and 2010, where there are two consecutive years of similar methodologies, we still see an enormous increase using DSM IV;
This just adds to the confusion, as during this period the number of people gambling reduced from 72% to 68%[xv] of the adult population, yet according to DSM IV there was a 10% increase, but using two different methods of screening for abnormal gambling behaviour, we see a logical reduction, but much larger than the reduction in the number of gamblers would suggest. All during the time period when, arguably, some of the biggest changes were happening in the British gambling environment.
So if we follow the data from the British Gambling Prevalence Surveys of 1999, 2007 and 2010 we have a mildly discredited methodology (DSM IV) showing a major increase in abnormal behaviour, after the major changes in gambling have happened (maybe abnormal gambling behaviour takes time to exhibit itself) and a more credible method (PGSI) showing an increase in numbers, which the Survey’s authors consider to be at the margins of statistical significance because of the small sample size. Such confusion obviously merits further investigation and so to consider abnormal gambling behaviour in the policy context, we need to first consider certain factors; what it is, how it’s identified and how the size of the problem is calculated.
The first question is what is abnormal gambling behaviour or problem gambling. The Prevalence Survey’s uses one of the most widely cited definitions that of Lesieur & Rosenthal’s and found in their 1991 article in the Journal of Gambling Studies; Pathological Gambling: A Review of the Literature, they define problem gambling as; ‘gambling to a degree that compromises, disrupts or damages family, personal or recreational pursuits’[xvi].
This to the layman could seem as quite a broad, almost an amorphous approach to the issue. Flippantly, one could almost replace the term gambling the ‘common cold’ and we would have to consider such a minor ailment as a ‘problem’. The fact is that unlike some chemical addictions, abnormal gambling behaviour takes different forms and has different outcomes and so an extremely broad definition is required and this is where some of the policy problems (lack of definition) derive from.
The Australian Productivity Commission’s 1999 Report on Gambling; provides a number of alternate definitions:
Academics that specialise in abnormal gambling behaviour tend to agree that such behaviour exhibits itself along a continuum which will include such characteristics as:
This is where some critics of the abnormal gambling academic process have made the point that while no one has any issues with those at Level 1 and everyone has concerns about those at Level 3, those at Level 2 form a group that could arguably inflate abnormal gambling numbers. The question posed is that while those in Level 2 are obviously of significant interest to medical practitioners (as they have problems which affect their lives and more importantly, could move on to become Level 3), should they be included in abnormal gambling numbers from a government policy perspective?
Level 2 gamblers suffer problems that may interfere with their own well being and that of their immediate circle, but so far have not reached a level which impacts on society as a whole by either breaking the law or incurring a cost to the NHS or their employer. They have problems but they are not (arguably) serious ones and they may disappear with time (equally they may turn into serious ones). An analogy would be that many an employee has turned up late for work and/or may not have been ‘on top of their game’ during the day when on the odd occasion they have over indulged the night before but this is quite different from those who are repeatedly late or absent from work and/or are ineffectual at work due to an alcoholism problem. One is usually considered the symptom of high spirits, generally found in youth and generally not thought a major problem while the other is a completely different matter, deserving of serious attention, both medically and by government. Critics have argued that when considering methods of determining problem gamblers, those who suffer from Level 2 issues have been included in the numbers and that this may make the abnormal gambling issue seem far greater than it actually is from a policy perspective (and so garner more funds and possibly, persuade government to restrain gambling activities). No one is dismissing the need of those with Level 2 problems to be provided with help if they need it, but should government be basing regulatory policy on estimates of the number of people considered to show abnormal gambling behaviour if a proportion of those who suffer are suffering what are arguably only minor problems? Those who consider such debates from an incentives viewpoint would argue that academics and medical practitioners would argue yes, Level 2 people need to be included because they have a problem that warrants research and medical support and consequently funding, while some gambling industry people would argue no, as they argue that these are people who only have minor and/or transitory problems so government shouldn’t change its gambling policy to accommodate them (i.e. tighten regulations). To consider this debate further let us look at the way that the number of problem gamblers is calculated.
The British Gambling Prevalence Survey’s (BGPS) use screening questionnaires to determine the number of problem gamblers:
The 1999 Survey[xx] started with a sample survey questionnaire which was trialled on a small group of abnormal gamblers and a slightly larger group of ordinary citizens. 7,000 households were then randomly selected from 850 postcodes across the country. These were sent a letter introducing the survey and informing them that an interviewer would be calling on them. Those households that allowed the interviewer access (4,619 households) had their household categorised by the socio-demographic data of the highest income earner in the household and then all members of the household over 16 were invited to fill in a questionnaire about their gambling behaviour. In these 4,619 households this meant 8,584 eligible adults, of whom 7,680 filled in the questionnaire. The questionnaire asked a series of questions about what gambling activities they participated in, how frequently and how much they spent. As part of these questions were a series designed to screen for abnormal gambling behaviour. The BGPS 1999 used two methods; known as the South Oaks Gambling Screen (SOGS) and DSM-IV.
The South Oaks Gambling Screen was designed by Leseiur & Blume and presented in their 1987 paper in the American Journal of Psychiatry entitled “The South Oaks Gambling Screen (SOGS): a new instrument for the identification of pathological gamblers.”. They describe it as “a 20-item questionnaire based on DSM-III criteria for pathological gambling. It may be self-administered or administered by nonprofessional or professional interviewers. A total of 1,616 subjects were involved in its development: 867 patients with diagnoses of substance abuse and pathological gambling, 213 members of Gamblers Anonymous, 384 university students, and 152 hospital employees. Independent validation by family members and counsellors was obtained for the calibration sample, and internal consistency and test-retest reliability were established. The instrument correlates well with the criteria of the revised version of DSM-III (DSM-III-R). It offers a convenient means to screen clinical populations of alcoholics and drug abusers, as well as general populations, for pathological gambling.”[xxi] DSM means Diagnostic and Statistical Manual of Mental Disorders which is published by the American Psychiatric Association and provides a common language and standard criteria for the classification of mental disorders.[xxii]
The BGPS 1999, as published by the Gambling Commission, does not provide an example of the questionnaire used in their 1999 survey but the Australian Productivity Commission Report on Gambling 1999 gives an example which is very similar, if not, identical (albeit with some questions in a different order). These are the SOGS twenty questions with their possible answers[xxiii]:
The scores for all twenty questions are added up and then the total score dictates if the respondent is considered to exhibit abnormal gambling behaviour. The contentious issue is at what level (or threshold) should this be considered.
The Australian Productivity Commission point out that “The difficulty of identifying the ‘right’ threshold for problem gambling stems from the fact that ‘cases’ are not clearly defined where the severity of the problems varies along a continuum. In some areas of public health it is easy to define a case. For example, someone either has HIV or they do not. But with problem gambling (and a range of other possible areas, such as obesity and diabetes) it is not clear where along the continuum people can be said categorically to have a ‘problem’. If the threshold for defining problems is set low then obviously a lot of people are said to be ‘problem’ gamblers, in the same sense that there will be a lot more ‘obese’ people if obesity is defined as being 10 per cent overweight rather than 20 per cent overweight”[xxv]
The BGPS 1999 uses 5 as its SOGS threshold. The Survey acknowledges that this level has been disputed; “While the original thresholds for classification on the SOGS are 3 to 4 to indicate a ‘problem gambler’ and 5 or more to indicate a ‘probable pathological gambler’, there has been recent consensus that these cut-offs are too low (see the Australian Productivity Commission (APC) report for a useful discussion of this issue). These arguments have fuelled criticism that the SOGS overestimates the prevalence of problem gambling by including too many false positives [someone being identified as being an abnormal gambler when they aren’t] in its classification. Nevertheless, some studies continue to use a threshold of 3 or more to identify ‘problem gamblers’.” and continues “In contrast, a number of Australian studies (eg Dickerson et al 1996) use 10 or more as the threshold for SOGS. This had its genesis in the first major Australian survey, which raised the SOGS threshold to 10 or more, after data analysis, apparently because the estimate of problem gambling prevalence according to the threshold of 5 or more was considered too high. The recent APC report questions this rationale and concludes that 5 or more is the most appropriate cut-off.”[xxvi]
Upon reading the cited part of the Productivity Commission’s Report (pp. 6.33 – 6.34.3), one can only say that the authors of the BGPS 1999 have reached a unique interpretation of what the Report concludes. The Productivity Commission acknowledges that a SOGS threshold of 5 (SOGS 5+) incorporates too many false positives; while a SOGS threshold of 10 (SOGS 10+) incorporates too many false negatives (abnormal gamblers who aren’t identified). When considering the research data using additional surveys they found, unsurprisingly, that those identified as abnormal gamblers with SOGS 10+ had severe problems and those with SOGS 5+ had lesser problems, which brings us back to the policy issue of whether government should be interested in people who feel guilty about their gambling or may have lied about it rather than concentrating the limited resources on those who need clinical treatment. The Commission points out:
“The SOGS 10+ group have a very similar pattern of SOGS responses to those gamblers who seek help from specialist problem gambling agencies — evidence that the SOGS 10+ threshold provides a measure of people suffering severe problems requiring assistance. They also have similar responses for clearly adverse harms except that the group seeking help have a higher prevalence of job loss, suicide ideation and crime. The false positive [normal gamblers who are wrongly classified as abnormal] rate among SOGS 10+ is probably very small.
The SOGS 5+ group has a lower prevalence of self-assessed harmful impacts than the SOGS 10+ group, but nearly all of such gamblers suggest that they spend more than they intended, around 90 per cent say they feel guilty about their gambling, about 70 per cent feel they have a problem and 70 per cent indicate that they have control problems.”[xxvii]
The Commission had considered the issue of severe problems versus less severe problems and came to the conclusion that it is viable to use a SOGS 5+ score if you intend to incorporate Level 2 or sub clinical problem gamblers in your numbers. The main thing is to be open about what your motives are when designing the survey and presenting the results. The conclusion of the Productivity Commission was (which is arguably a different interpretation to that given in BGPS 1999):
“the SOGS can be legitimately used to look at the prevalence of people whose problems do not require individual intervention, but which are of concern for public health reasons. Walker (1998b, p. 44), for example, notes:
Similarly, Shaffer et al. (1997, p. iii) observe:
... scientists and public policy makers have paid insufficient attention to level 2 gamblers (ie those with sub-clinical levels of gambling disorders). While extremely diverse, level 2 gamblers experience a wide set of problems from their gambling.
In this instance, it is clearly appropriate to use lower SOGS scores to determine the number of Australians whose gambling behaviour entails significant risks (level 2 gambling problems using Shaffer et al.’s terminology), so long as the purpose of this prevalence rate is made clear[Commission’s italics],namely:
Though they may be well-intentioned, it is clear that many parties have a strong career interest in exaggerating the problem gambling phenomenon and in seeing that the reported incidence is never below some threshold (sub. 155, p. 71).
But, similarly, industry groups who wish to minimise the perception of apparent harms created by gambling, will tend to set the bar high to achieve this objective. Some of the criticisms by industry of the draft report’s findings in relation to the prevalence of problem gambling (for example, the AHA NSW sub. 208, p. 28) reflect their view that someone must have severe problems to be termed a problem gambler. There is a clear need for any test of gambling problems to set thresholds which have known risks of harms, and to explain the purposes of each of the thresholds that may be selected. The Commission considers it useful to employ a number of different benchmarks for ‘problem’ gambling — which suit the different possible purposes of such a test — in the same way that different benchmarks are now used to assess problematic alcohol use or degrees of weight problems.”[xxviii]
Ignoring the obvious response of why someone doesn’t do a SOGS 7.5+ (Australia has started using SOGS 8+), this argument embodies the whole policy debate about the numbers of abnormal gamblers, where do you draw the line?
Using the SOGS questionnaire, a person could be in the SOGS 5+ category of abnormal gamblers by claiming that they had won money when they had lost (a lie), gambling more than they intended to (an extravagance), feeling guilty about their gambling (regret), hidden evidence about their gambling (from a spouse perhaps) and argued with someone they lived with about spending money on gambling. While these could all be signposts of a potential descent into abnormal gambling as seen by the medical profession or problem gambling academia, they could also be evidence of a fairly stereotypical existence of a gambler who has a spouse who is not a fan of his/her hobby.
The lack of any context in the questionnaire (i.e. how big a lie, how many lies, how many arguments) or the lack of weighting between the questions (feeling guilty about gambling is less severe than selling household possessions to pay off gambling debts) does make this author believe that SOGS (and other similar screening methods) are blunt instruments looking at a complex topic, which in some respects fail to reflect the behaviours of the passionate gambler. This was part of the reason that some Australian studies have used SOGS 10+; “A central concern in Australian studies has been that many people with SOGS scores of between 5 and 10 may, in fact, be highly motivated regular gamblers who face little real risks from their gambling (Dickerson et al. 1996a, p. 61) and would, therefore, scarcely require individual intervention to help them”[xxix] It is interesting to note that if BGPS 1999 had used SOGS 10+, the numbers of people exhibiting abnormal gambling behaviours would be so small as to be immeasurable and even at a SOGS score of 9 it would show an abnormal gambling rate of just 0.1%.[xxx]
Unfortunately many of the same questions apply to the second form of abnormal gambling screening used in the BGPS 1999, DSM IV. “The DSM-IV screening instrument is taken from the fourth edition of the manual used by the American Psychiatric Association. It has been used much less commonly than the SOGS and, unlike the SOGS, does not exist in a validated questionnaire format. The DSM-IV consists of 10 diagnostic criteria, and a person who answers ‘yes’ to 3 or more criteria is classified as a ‘problem gambler’, with a score of 5 or more indicating a ‘probable pathological gambler’[xxxi]. In layman’s terms this means that DSM-IV is in fact a 10 question multiple choice survey, which the BGPS research team developed themselves based on a system used for diagnoses in a clinical setting. Most surprisingly is the additional caveat; “the DSM-IV screen has not been validated in terms of its prevalence estimates in the general population.”[xxxii] The BGPS 2007 explains “The DSM IV was created as a tool for diagnosis and not as a screening instrument for use in the general population. Since there is no single gold standard questionnaire version of the DSM IV criteria, as part of the development work for the 1999 survey we adapted the criteria and developed and pre-tested a DSM IV based screen”[xxxiii]and repeats the caveat “The DSM IV was developed as a diagnostic tool, and has not been validated for general population use.”[xxxiv]. To the uninitiated it may seem strange that the screening method that is found in all three BGPS surveys and the one that has recently grabbed the most headlines as its numbers shows the biggest increase in abnormal gambling behaviour, was never designed for the job it has been used for in the BGPS.
The questionnaire used in the BGPS 2007 (which we assume is identical to the BGPS 1999) was as follows[xxxv] with the possible answers provided:
Yet again this screening approach arguably suffers from issues about question comprehension, context and severity (although this approach does appear to weight symptoms more than SOGS). Someone could still be designated an abnormal gambler if they had occasionally (why no ‘rarely’ as a category?) thought about their gambling (which is quite common with sports betting), gambled while depressed or anxious (none of which may be due to gambling) and lied to hide the extent of their gambling (even if this is a white lie to say a bet was £2 instead of £20?). As stated before, these could all be signposts to far more serious psychological problems, but could also just be run of the mill experiences for a passionate gambler. To solve this, the threshold arguably needs raising, but, as with the Australian Productivity Commission’s conclusions about the threshold with SOGS, the point about the threshold level is what do you want to achieve?
It would appear from both DSM IV and the SOGS screening methods that the intent was to include as many people as possible, which optimistically, would be due to the belief that with any public health screening policy, it is always better to overestimate the problem than under-estimate it. The difference though, is that when screening for problems that actually have a medical impact (e.g. alcoholism, drug addiction, obesity etc.) the need to overestimate is because those not picked up by the screen will end up with more severe medical problems that will cost more to the state in the long run. The same cannot be said for those who have Level 2 abnormal gambling behaviours, the evidence is just not there yet to suggest that those not picked up by screening (false negatives) will go on to be a major cost to society – some will obviously do so, but many wont. The paucity of evidence in this field means such assumptions can only be questioned. It is worth noting that if the DSM IV screening method had a threshold of 5, then the percentage of abnormal gamblers would be 0.2% as opposed to 0.6% using the BGPS 1999 data. [xxxvii]
The BGPS 2007 built on the BGPS 1999, introducing new questions to incorporate new gambling activities, an attempt to assess gamblers’ net expenditure, questions about health and lifestyle and some attitudinal questions. Questionnaire design was similar to BGPS 1999, with a pilot questionnaire trialled on a small sample of abnormal gamblers, then on a sample of ordinary citizens. This time 10,144 households were chosen from 350 postcodes and yet again an interviewer visited the house, took the socio-demographic information of the highest income earner and offered every person over 16 within the household the questionnaire to self-complete (either using a booklet or online). 5,832 households responded, meaning 11,052 eligible respondents of whom 9,003 completed the questionnaire. The overall response rate was 52%. This low level of response prompted the authors of BGPS 2007, the National Centre for Social Research, to investigate, as getting a representative sample of the population is key to the accuracy of the survey’s results. They concluded that by weighting the data that they received and the fact that low responses from young men (who tend be abnormal gamblers) were cancelled out by low responses with those with low gambling frequency (who tend not to be abnormal gamblers) there was not an issue[xxxviii]. This report did not, however, question why interviewers were physically sent to households as opposed to doing the questionnaire online or by telephone as is common practice for most population surveys these days (although both phone and internet were used minimally as back up methods).
The BGPS 2007 introduced a new screening method called PGSI, or the Canadian Problem Gambling Severity Index. “The PGSI was developed by Wynne et al, over a three year period (1997-2000). This period included a development phase which was followed by a testing phase in order to validate the screen in a general population survey in Canada (among a sample of over 3,000). The PGSI was launched in 2001 and refined in 2003. The PGSI constitutes nine items of a larger screen (more than 30 items) - the Canadian Problem Gambling Inventory (CPGI). The full screen assesses gambling involvement, gambling problems, correlates and demographics. The PGSI items include chasing losses, escalating gambling to maintain excitement, and whether gambling has caused health problems. The full CPGI has been used in general population surveys in seven Canadian provinces, as well as in Denmark and Iceland. The subset of problem gambling items has been used in a national survey in Canada, smaller-scale surveys in the Canadian provinces and in general population surveys in Queensland, Victoria, Tasmania, and the Northern Territory, Australia.”[xxxix]
The BGPS had decided to discontinue using the SOGS screening method as it tended to overstate the number of abnormal gamblers. BGPS continues about the new screen; “Though the development and testing work on the PGSI is not yet complete, indications suggest that it is likely to become widely used and we decided to include this in the 2007 survey in preference to the SOGS. In order to allow comparison with prevalence rates in 1999, we kept the DSM IV as well [BGPS’s own developed screen]. So, as in 1999, we have two separate prevalence rates of problem gambling, allowing us to capitalise on the advantages of each, and to correlate and compare the results of the two screens.”[xl] BGPS obviously saw the advantage of being able to compare one screen with another (as they had done in the BGPS 1999 where they found that different screens ‘pick up’ different people) and seemed confident that their own screen, DSM-IV, was the best screen to keep so as to provide a longitudinal element to the data even though it hadn’t been validated for use as a population survey tool.
The PGSI questionnaire as used in the BGPS 2007 is as follows[xli]:
In the past 12 months, how often
The PGSI is scored as follows:
PGSI classification category PGSI score
PGSI shows, to this author’s view, the most sensible approach to screening for abnormal gambling yet developed, as it is quite difficult to work out how one would become a false positive, where it was quite simple to do so with SOGS and DSM-IV (see above). Minor criticisms would be in the following questions:
The Australian Productivity Commission Report on Gambling 2010 has a number of problems with the PGSI screen as used in the surveys that BGPS 2007 cite in support of its usage. Their first issue is the use of the terms ‘Almost always’, ‘Most of the time’, ‘Sometimes’ and ‘Never’ to denote frequency as they argue these terms are subjective and will differ in meaning from one person to another[xliv]. This is undoubtedly true but difficult to eradicate in a questionnaire scenario as asking respondents to provide numerical answers to frequency questions is beset with recall problems. The Productivity Commission also found that a number of Australian surveys had used slightly different terminology and scoring system (never=0, rarely=1, sometimes=1, often=2 and always=3)[xlv] which was not how the developer intended; fortunately the BGPS 2007 used the correct method. The Commission also found a threshold issue (though arguably less of a one than with SOGS and DSM-IV), they appear to be happy that the 8 threshold represents those with abnormal gambling behaviour but are concerned that those with lesser scores could be being mislabelled;
“However, using a term ‘problem gambler’ to encompass a set of problems that range from the moderate to the major is not appropriate. For instance, a person could score three by sometimes betting more than they could afford, sometimes feeling guilty, and sometimes being criticised for gambling. These may still be worrying signs — but they suggest risk, more than significant harm — which is why the actual classification of CPGI 3–7 is ‘moderate risk’ not ‘moderate problem gambling’.”[xlvi]
BGPS 2007 doesn’t make the differentiation with PGSI (but does with DSM-IV) but it is worth reiterating that when discussing those with abnormal gambling behaviours it is all too easy to bundle those with severe problems in with those with lesser ones without a declaration that this is the intent. This would be most apparent with DSM-IV and it being the only longitudinal screen (as in it’s been used in all three BGPS surveys) when it would appear superficially to produce more false positives than PGSI (see above) has got to be a problem that needs addressing in any future prevalence surveys.
Worthy of note is the comparison in the BGPS 2007 of the PGSI screen and the DSM IV screen;
“0.8% of the sample were classified as problem gamblers according to one or other screen; 0.4% were classified as problem gamblers according toboth. The vast majority of people (99.2%) were classified as ‘non-problem gamblers’ on both screening instruments.”[xlvii]
BGPS do not appear to think that their own screen, DSM IV has a lower threshold than PGSI but do acknowledge it provides a slightly higher prevalence rate;
“64% of people who were classified as problem gamblers by the DSM IV, were also problem gamblers according to the PGSI. 74% of those who were classified as problem gamblers according to the PGSI were also classified as problem gamblers by the DSM IV. Conversely, 36% of those who were classified as non-problem gamblers according to PGSI were problem gamblers according to DSM IV; and 26% of DSM IV non-problem gamblers were classified as problem gamblers according to the PGSI. This suggests that it is not simply the case that the DSM IV has a lower sensitivity for measuring problem gambling than the PGSI. Rather, it seems that the two screens are capturing slightly different groups of people, and therefore different types of problems”[xlviii]
They go on to explain how the two screens could have such different results;
“A weighted kappa statistic showed that the agreement between the two problem gambling screens is moderate (0.68; confidence interval 0.57-0.79). (No agreement would be expressed as a value of 0 and perfect agreement as a value of 1.). A number of conclusions can be drawn from the comparison of the two screening instruments:
Considering the caveats about both methods (neither being designed for the use they have been put to), the fact that they have produced, arguably quite different results when they would appear (to the layman) to use a similar methodology and style and content of question has got to raise some questions about the effectiveness of these screens as an approach to identifying the number of abnormal gamblers in the population. To argue that they the two screens just pick up different sectors of the population just doesn’t provide any confidence in their accuracy. Especially, when these numbers will become headlines and ‘dog whistles’ to government policy – all sides of the gambling debate demand and deserve credible research and these screens don’t even accommodate for the fact that the biggest problems for all gambling survey is that a) abnormal gamblers have a tendency to not respond to gambling survey requests and that b) gambling is a stigmatised behaviour and that both normal and abnormal gamblers may under-report their behaviour to conform with social norms. So while every BGPS survey comes with the caveat that “No screen to measure problem gambling is perfect. A best estimate of any population sub-group endeavours to minimise both ‘false positives’ and ‘false negatives’”[l] maybe there should be more emphasis on it being a ‘best guess’ and more explanation of what numbers are being sought and why. For the casual observer, it would appear that abnormal gambling numbers are either too high as thresholds have been set too low, or are too low because the methodology of screening doesn’t pick up all of those it’s intended to.
The BGPS 2010 uses both DSM IV and PGSI screens for considering the prevalence of abnormal gambling. It does this in an identical manner to the BGPS 2007 although this time respondents to the Survey were given a lap top to self-complete the questionnaire. One major difference between the two surveys is the inclusion in the results of a whole chapter on ‘at risk gamblers’ or those who don’t pass the thresholds to be considered abnormal gamblers or ‘problem gamblers’. Obviously aware of the concerns cited above (which have mostly been found in the Australian Productivity Commission inquiry into gambling), BGPS give their reasoning for the inclusion of the ‘at risk’ chapter as:
“Many health behaviours are matters of degree and are therefore most accurately described and measured as continua. For example, many people describe their own general health to varying degrees using terms such as excellent, good, fair, bad, very poor and so on, rather than stating that they are in bad health or not ‘bad’ health. There is increasing recognition within the field of gambling research that gambling problems, like alcohol dependence, also lie upon a continuum of risk or problems. Examining the profile of those who experience some problems but are classified below the threshold of problem gambling, or those who may be at-risk of developing problems is a public health concern.”[li]
BGPS fail to explain why someone who may have over indulged on their betting, been criticised by their spouse for it and consequently felt guilty about it (which is enough to be considered a moderate gambler under a PGSI screen) should be a public health concern. They then rather confusingly state;
“it is possible that the greatest volume of harm from gambling is associated with those at low to moderate risk, simply because this group greatly outnumbers those who are at the highest risk, problem gamblers”[lii]
which surely invites the analogy that the biggest medical problem facing the country today is the common cold and cephalgia (headaches). The closest the BGPS comes to declaring why study of those considered ‘at risk’ is necessary, is their statement that;
“these groups are at elevated risk of experiencing adverse consequences from gambling”[liii]
“examining the prevalence of at-risk gambling is important as the greatest volume of harm from gambling may be associated with those at low to moderate risk, simply because this group greatly outnumbers those who are at the highest risk, problem gamblers. Looking at the prevalence estimates presented above and applying these to population data, shows that the number of moderate risk gamblers in Britain is around 879,000 people and the number of low risk gamblers in Britain is around 2,686,000 people.”[liv]
is that there is no explanation of what ‘at risk’ means in the context of abnormal gambling behaviour other than they have answered the screening questionnaire in a less positive/ frequent way than abnormal gamblers. For the newspapers and the politicians there are just two factors here; large numbers and the term ‘at risk’ and so the debate will get swayed unfairly in one direction. To state that those ‘at risk’ demand more attention than those who have abnormal gambling behaviours just purely because there are more of them makes as much sense as saying that everyone who has ever got drunk is ‘at risk’ of being an alcoholic.
This is a serious shortcoming of the BGPS 2010 as it not only gives the implication that this is a political move (one of the authors of the BGPS 2010 in a self-acclaimed anti-gambling proponent) but also does a disservice to those who may well be ‘at risk’. By not considering the need to explain properly why those below the abnormal gambling threshold are worthy of consideration and why they are a public health concern, this chapter gives the impression of being purely there to aggrandise the issue unfairly.
The final critique in this paper about the way abnormal gambling numbers are collated is to look at the actual sample sizes used. As we have stated above, the BGPS 1999 used a sample of 7680 respondents, BGPS 2007 used 9,003 respondents and the BGPS 2010 used 7,756 respondents (BGPS 2010 had an overall response rate of 47%, down on BGPS 2007’s 52% and still a cause for concern). These are significantly large numbers when considering commercial market research survey’s which usually use 1,200 respondents to reflect the national population but as we’re talking about tenths of a percentage when considering abnormal gambling such large numbers are necessary.
As stated at the very beginning of this paper, according to the BGPS, the level of abnormal gambling has increased from 2007 to 2010 from 0.6%[lv] to 0.9%[lvi] using the DSM IV screen and from 0.5%[lvii] to 0.7%[lviii] using PGSI. To add clarity, these figures show that for DSM IV there was a 50% increase in abnormal gamblers between 2007 and 2010, if we take the median numbers of the estimated number of abnormal gamblers from the range provided in the Surveys, as we did at the beginning of this paper, we get a 52% increase and for PGSI, these percentages give a 40% increase, while the increase in the median of the range is 34.5%. The differences between the two are probably due to rounding and also the small numbers involved.
If we apply the percentages to the actual sample numbers used in the Surveys we get the following:
One conclusion that can be drawn is that inherent in the BGPS Surveys is the fact that actual numbers of people exhibiting abnormal gambling behaviours are very small in actual numbers (<70 for DSM IV and <55 for PGSI) and the increases shown from 2007 and 2010 involve such a small number of actual people that there is a major potential for sampling error. BGPS has covered this to some extent by urging caution when considering these increases.
BGPS 2010 goes further by stating that the increases in DSM IV were at the margins of statistical significance. It should be noted that their numbers and sample sizes differ from those provided in the tables in the report; they appear to be using total sample sizes for the whole survey as opposed to actual samples used in the screening and applying the prevalence rates:
“In 1999, DSM-IV problem gambling estimates were 0.9% among men and 0.2% among women and 0.6% overall. In 2007, equivalent estimates were 1.0%, 0.2% and 0.6%. In 2010, estimates were 1.5% for men, 0.3% for women and 0.9% overall. When looking at problem gambling estimates for all adults aged 16 and over, the difference between 2010 and earlier surveys is significant at the 5% level, but not at the 1% level (the p-value is 0.049). This is thus at the margins of statistical significance and caution should be taken when interpreting this result. Firstly, as noted above, the number of cases identified in each survey sample is small and therefore very sensitive to relatively small changes in responses. For example, in 2007, 47 people (out of 9003) were categorised as problem gamblers according to the DSM-IV screen. In 2010, 64 people (out of 7756) were categorised the same, the difference in absolute numbers is just 17 people. However, when sample sizes for each survey year and weighting for non-response are taken into account, the net effect is an increase in prevalence from 0.61% to 0.92%; an increase of 0.31 percentage points (pp).”[lxiv]
For increases in the numbers screened by PGS, BGPS 2010 also urges caution and considers the increases to be statistically insignificant:
“Problem gambling prevalence estimates as measured by the PGSI were 1.0% for men, 0.1% for women and 0.5% overall in 2007. In 2010, comparable estimates were 1.3%, 0.2% and 0.7%. These changes were not statistically significant (p=0.23). As discussed in section 5.7, it appears that the DSM-IV and PGSI screens are capturing slightly different groups of people and may be measuring slightly different types of gambling problems. Problem gambling prevalence rates when measured by the PGSI did not increase significantly between 2007 and 2010 whereas increases at the margins of statistical significance were evident when problem gambling was measured by the DSM-IV.”[lxv]
So in conclusion, this investigation into how abnormal gambling figures have been identified shows, arguably, that the British Gambling Prevalence Studies do not, in this author’s view, represent the hard evidence that policy making about gambling deserves in the UK. It is not within the scope of this paper to consider why the academic community believes that when assessing such a complex issue as abnormal gambling behaviour, a ‘tick box’ exercise on a population survey is considered sufficient for the task, especially when the number of problem gamblers becomes the key issue in any policy making. Nor is it within its scope to consider that if there really are up to 593,400 abnormal gamblers with problems so severe they need treatment and 879,000 gamblers deemed at moderate risk and 2,686,000 gamblers deemed low risk, then where are they? One would have thought they would be appearing in the Courts (having stolen money to gamble) or with the NHS (because of the extreme stress/depression) in numbers far greater (by multiples of thousands) than they do now. What is certain is that all sides of the gambling debate need and deserve data on the incidence of abnormal gambling that can be trusted as realistic and neutral.