In 2007 the University Department of Psychiatry and Oxford Internet Institute conducted a survey of users of online gambling websites in order to learn more about the gambling behavior and its impact. Known as the “Oxford Gambling Survey “, the web-based survey was designed to collect information via a series of questions from a large number of respondents. Direct links were placed on gambling and gambling-related websites in order to get better understanding of the effects of the Internet on the users’ gambling behavior and get a clear picture of the various patterns of online gambling. The results of the survey gave insight about the various ways to help players with problem gambling and controlling their online gambling activities.
The Oxford Gambling Survey was funded by the Responsibility in Gambling Trust and by the Economic and Social Research Council. It got the support of the Remote Gambling Association as well as of the European Gaming and Betting Association. The survey was approved by a National Health Service Research Ethics Committee.
Survey Design and Finding Respondents
The online Oxford gambling study consisted of a series of comprehensive interactive questionnaire displays related to gambling behaviors and health experiences. After completing it respondents were asked to leave their e-mail address if they wanted to take part in the prize draw for the chance to win one of the six iPods that were given to randomly drawn participants.
In order to find respondents, links to the survey were played on 15 gambling websites where people could gamble for real cash. Links were also placed on six sites with gambling related content which didn’t offer real money gambling products. In addition, gambling service providers put links to the online survey in their newsletters which they send to customers. The survey was advertized and got extra advertising from several journalists who put links to the survey on their blogs and news articles.
The Analysis of Data
The data gathered from the survey was analyzed using latent class analysis (LCA) over the iGaming activities of the respondents. First, sub-categories of respondents with different gambling behaviors were identified and then their behavioral and clinical features were compared using parametric and non-parametric statistics.
Furthermore, the survey used a dichotomous coding system where a sore of 1 means that the participant is a regular gambler on any given gambling activity or has participated in gambling at least once a month and a score of zero means that the respondent is not a regular gambler. With this system the LCA was used very effectively in order to group respondents into clusters based on gambling patterns as well as describe those patterns by the chances of a participant from a certain cluster being a regular gambler in each of the activities.
In this study, highly correlated variables were merged, for example, playing lotteries online and buying online scratch cards, and new variables were created, such as lotteries / scratch cards, and models were examined estimating class 1 through class 9 solutions.
For the purpose of this study, three gambling activities were also mutually correlated, such as playing online slots with online casino gaming and with Internet bingo. This, in turn, also correlated with virtual casino gambling. A positive score on at least one of these three activates also shows a positive score on the new casino / slots / bingo variable.
Once the LCA model was selected and all respondents had been allocated to a certain class according to probability, x2 tests were used to compare classes on data frequency. If and when these tests revealed significant differences additional statistics was used to investigate more where these differences are located precisely.
The Important Sections of the Survey
The survey addressed the following sections: gambling related behavior, general mental health (anxiety, mood swings, depressive characteristics and DSH), alcohol and drug use and demographics. In the first section, respondents were asked how often they had taken part in online gambling activities in the past 12 months and whether their parents or guardians had gambled regularly.
The list of “activities” included were playing slots and other casino games including poker, betting on sports and horseracing with both a sportsbook and a betting exchange operator, spread betting, virtual racing, online bingo, online lotteries and playing virtual instant win games. The status of ‘problem gambling’ was evaluated using the 10 DSM-IV criteria. The multiple response scale included the following answers: ‘never’, ‘occasionally’, ‘fairly often’ and ‘very often’. The maximum score in this section was 10 (one point for each question) and those who scored 3 or 4 were indicating problem gambling.
In the second section, the general mental health was evaluated using General Health questions (GHQ) focusing or recent energy, concentration, mood and coping. A score above threshold on these questions indicated the presence of psychological distress or perhaps an inability to carry out normal functions. Participants that scored four or more in these questions also completed the Patient Health questions (PHQ) in order to get more information about the presence of depressive symptoms and panic disorder.
Furthermore, in order to learn more about possible bipolar symptoms, participants also answered the Mood Disturbance questions (MDQ). The sensitivity of these questions in clinical psychiatric populace is high, but in general populace sample the sensitivity is low (0.28). Still the negative predictive value is high in the general sample with a 0.98 specificity. This means that the accuracy of a negative result is good, but getting a positive result might not mean that there is presence of bipolar disorder. In addition, a score above the threshold on these questions doesn’t mean that respondents have bipolar disorder; instead the survey labeled them as having hippomanic symptoms which might be an indicator of bipolar disorder.
Lastly, three more items about deliberate self-harm were also included in this section of the survey. Respondents were asked if they have ever considered harming themselves because of gambling problems. Those that scored positively got additional questions asking if this has led to actually harming themselves. All participants were also asked if they have harmed themselves for other reasons besides problems with gambling.
In the third section of alcohol and drug use, respondents answered the CAGE designed to reveal more about the relationship of the respondents with the use of alcohol. Two or more positive Yes responses on the CAGE mean that the respondent has possible alcohol problems. They were also asked about the use of drugs such as cocaine, marijuana, ecstasy and/or amphetamine in the past 12 months. Respondents who said that they had used illicit drugs within the past year were given the Drug Abuse Screening test (DAST) which tests the presence of a drug abuse problem. This ten item test has a yes/no format. Two or more positive responses mean that there might be a substance misuse disorder.
The final section was about the participants’ demographics, in particular, their gender, birth date, country, marital status, level of education, income estimation, occupation, number of children (if any) and their religion.
Feedback
After completing the survey, participants were given feedback about the signs and symptoms if they had scored high on any of the psychological displays. Clickable links to relevant websites providing help about problem gambling and more information about the symptoms were also provided on the results page. On the same page there was also a caveat pointing out that the online survey was used for research purposes only and not for providing diagnosis to individuals.
Feedback was given in the form of a single sentence indicating the possible areas of concern within the sections of problem gambling, alcohol and substance abuse, uni-polar depression and disturbances in mood.
Results According to Demographics and Representation in Clusters
From all the respondents 79.1% were male and 20.9% were female. Of the men, 41.4% were single, 17.6% were living with a partner and 5.9% were divorced or widowed. Most of the respondents were located in the United Kingdom (68.8%), the United States (8.8%), Canada (2.5%), Ireland and Australia (1.7% and 1.5% respectively). The remaining people were from other European countries and Asia.
41.8% of the survey participants were educated to a degree level or higher, 36% had a college degree and the remaining 22.2% have completed secondary or primary education or both. Approximately three quarters of the people were employed, 55.8% mostly full time, 11.4% self employed and 6.2% part time. About 5% were retired, another 5% were taking care of the home and 6.6% were unemployed.
From the female respondents, 4.2% were represented in the sport betting cluster, while 46.2% were represented in the lotteries cluster. Young respondents (32 years-old) were mostly seen in the casino and sports betting cluster, while the oldest (37 years-old) were seen in the lotteries cluster. Furthermore, there were many single respondents in the casino and sports betting clusters as well as multi-activity gamblers, while many married participants were seen in the lotteries cluster. Respondents with children were found in very low number in the sports betting and casino and sports betting clusters.
The respondents in the non-to-minimal gambling cluster, sports betting cluster, casino and sports betting cluster and lotteries cluster had a university or a college degree, while those in the multi activity gambling cluster had some university or college education. Casino players and sports bettors were most likely to be full time employed rather than part time employed. 12% of the retired respondents were present in the multi-activity cluster, while only 1.5% of them were present in the casino and sports betting cluster. Furthermore, 17% of the unemployed respondents belonged to the multi-activity gambling cluster, while only a few undergraduate students could be seen in the multi-activity cluster and lotteries cluster. Most of the respondents with below average income could be found in the multi-activity gambling cluster, while those with income above average could be seen in the sports betting cluster.
Results Based on the Time Spent Online Gambling
The following are the results of the number of hours spent per day and the number of years spent online gambling in each of the clusters. 48% of the respondents in the non-to-minimal gambling cluster spend zero hours per day gambling compare to less than 2% of casino and sports betting gamblers and multi-activity players. The biggest percentage of respondents that had spent less than 1 hour per day gambling was found in the sports betting cluster, while the biggest percentage of players spending more than 10 hours per day gambling online was found in the multi-activity cluster.
On a yearly basis, the people in the non-to-minimal gambling cluster spent less than a year using online gambling sites, while people in the other clusters used the services and products offered on gambling sites for between 1 and 2 years. Those in the sports betting cluster and casino and sports betting cluster where more likely to report between 3 and 7 years of gambling as well as 7 and 8 years spent online gambling, while the highest portion of respondents to have been engaged into gambling activities for 9 or 10 years belonged to the sports betting cluster and multi-activity gambling cluster.
Flaws in the Study
The self report instruments used in the survey such as DAST, CAGE and MDQ, however, may be too sensitive or too insensitive to the items of interest. They are expected to behave reliably. In other situations, ad hoc questions had to be used that might increase apparent psychopathology. For instance, high levels of DSM-IV problem gamblers were detected in all gambling clusters. The reason for this might be that the simplified DSM-IV interview items are too sensitive. But a big study using the same DSM-IV self-report interview didn’t find high rates of problem gambling among all users which suggests that the high rates in this study are in fact true and not a relic of a scale that is too sensitive.