https://orcid.org/0000-0003-2702-0098
![Sample our Health and Social Care journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5938/TOC-Subject-Sample-2021-Health-Social-Care.jpg"})
ABSTRACT
There is a greater need for contemporary research to empirically examine actual internet casino gambling behavior with valid data. To fill this gap in research, the current study examined two years (1 February 2015 to 31 January 2017) of prospective longitudinal data from 4,424 actual online casino gambling subscribers to one of the E.U.’s largest internet gambling services, including novel analyses of depositing and withdrawal behaviors. We found that today’s online casino gambling behaviors are similar to those observed 10 years ago in the E.U., with similarly small proportions of the player pool exhibiting more extreme gambling involvement than the rest. Some unique gambling behaviors (e.g. net loss), depositing behaviors (e.g. credit card use), and withdrawal behaviors (e.g. reversed withdrawals) distinguished more involved bettors from typical bettors. Our results suggest that, like a decade prior, most online casino players in our sample bet modest amounts and play relatively infrequently, yet a small percentage (approximately 5%) engage at disproportionately high and potentially risky levels. These findings are based on data from a single online casino operator and bettors in our sample might have gambled with other operators during the study period.
Acknowledgements
We would like to thank the data teams at Entain for their assistance in providing the data.
Funding sources
This manuscript was supported by a research contract between the Division on Addiction and GVC Holdings PLC. GVC is a large international gambling and online gambling operator. GVC had no involvement with the development of our research questions or protocol, the data collection/analysis, or development of this manuscript.
Constraints on publishing
GVC had no involvement with the development of our research questions or protocol, the data collection/analysis, or development of this manuscript, nor with its submission.
Competing interests
In addition to the funding from GVC supporting this project, the Division on Addiction currently receives funding from the Addiction Treatment Center of New England via SAMHSA; EPIC Risk Management; The Foundation for Advancing Alcohol Responsibility (FAAR); DraftKings; the Gavin Foundation via the Substance Abuse and Mental Health Services Administration (SAMHSA); The Healing Lodge of the Seven Nations via the National Institutes of Health (National Institute of General Medical Sciences and National Institute on Drug Abuse); Health Resources in Action via the Massachusetts Department of Public Health Office of Problem Gambling Services; The Integrated Centre on Addiction Prevention and Treatment of the Tung Wah Group of Hospitals, Hong Kong; St. Francis House via the Massachusetts Department of Public Health Bureau of Substance Addiction Services; and the University of Nevada, Las Vegas via MGM Resorts International.
During the past 5 years, the Division on Addiction has also received funding from Aarhus University Hospital with funds approved by The Danish Council for Independent Research; ABMRF – The Foundation for Alcohol Research; Caesars Enterprise Services, LLC; the David H. Bor Library Fund, Cambridge Health Alliance; Fenway Community Health Center, Inc.; Massachusetts Department of Public Health, Bureau of Substance Addiction Services; Massachusetts Gaming Commission, Commonwealth of Massachusetts; and University of Nevada, Las Vegas via MGM Resorts International.
Preregistration statement
https://osf.io/j6swu/?view_only=077ded103dd744049d1d0898cc83fac4
Data availability statement
The authors agree to make their data available upon reasonable request.
Supplementary material
Supplemental data for this article can be accessed here.
Notes
1. Many casinos distinguish Bingo, Poker, and Sportsbook as unique gambling products, confined to their own separate rooms/venues and away from the main gaming floor. In a similar way, most online gambling operators treat Bingo, Poker, Sportsbook, and Casino as separate products, with their own lobbies and sub-lobbies. Our definition of online casino games reflects these distinctions.
2. Although this exclusion is quite large, it serves the purpose of distinguishing more regular online casino subscribers (who from a public health standpoint, we were more interested in) from subscribers who were likely just experimenting with bwin’s online casino platform, with no intention of engaging with the service in the long term. Without this exclusion, these online casino ‘experimenters’ could substantially bias the results.
3. Subscribers might not have had any withdrawal records either because they lost all of the money they deposited or because they did not withdraw any money from their accounts before the study period ended.
4. Deposits can ‘fail’ for several reasons, including reasons that are benign such as incorrectly entered credit/debit card information, and others that might warrant cause for concern such as deposit amounts that exceed a regulated limit. However, because we did not have access to information related to why the deposit failed, we advise readers to interpret this measure with caution.
5. Reverse withdrawals are instances when online gamblers cancel a pending cashout of winnings.
6. As a supplementary exploratory analysis, we ran a series of scatterplots between our variables of interest. This analysis confirmed that the relationship between variables tended to be non-linear and heavily influenced by outliers, thus justifying our use of the Spearman correlation.
7. Our prior experience with online gambling data suggests that these data do not typically have underlying distributions that easily facilitate parametric testing, even with corrections for unequal variances (i.e. Welch’s t-test). For this reason, we planned to conduct median tests for assessing differences between groups.
8. These MIB groups are not mutually exclusive. Subscribers can be most involved on multiple metrics (e.g. both total stake and number of bets).
9. Although both Fisher tests and Chi-square tests usually yield comparable statistics, we opted for Fisher tests over Chi-square tests because of their ability to handle low expected cell counts (i.e. cell counts of 5 or less).
10. We excluded gambling metrics related to the MIB group of interest from their respective analyses. As an example, to assess membership in the MIBbets group, we did not look at interactions between gender/age and number of bets, regardless of whether there were significant differences on gender or age for number of bets in the full sample gender comparative analyses.
11. When doing these same comparisons, LaBrie et al. (Citation2008) did not detect any significant gender differences, except for bets per day. However, they used means tests with corrections for unequal variances for these comparisons. As part of an unplanned exploratory analysis, we also tested these same gender differences using means tests with corrections for unequal variances. As with LaBrie et al., we did not detect any significant gender differences, except for bets per day.
12. Bivariate F.e.t’s looked at the proportion of a country’s representation in each MIB group (compared to other countries) and compared it to the proportion of its representation in the LIB (compared to other countries).
13. Subscriber age was not skewed. Therefore, we also conducted unplanned exploratory Welch’s unpaired t-tests for examining age differences between these groups. The results of these tests were extremely similar to the median tests (i.e. all differences were statistically significant; estimated confidence intervals of mean differences mirrored our estimated confidence intervals of median differences).
14. This could be partially a function of the LIB group comprising a larger proportion of subscribers who lose a few bets and then quit entirely. This has the potential to overinflate percent lost.
15. Differences between LaBrie et al.’s (Citation2008) results and our own regarding demographic differences might be attributable to differences in measurement. LaBrie et al. (Citation2008) used mean comparison tests (i.e. t-tests) and we used median comparison tests (i.e. Mann-Whitney U tests) in the present study. Although not part of our planned analyses, when we tested gender differences on gambling metrics using mean comparison tests we arrived at the same results as LaBrie et al. (i.e. the only significant difference for gender was bets per day). It is still an open debate which test (means or medians) should be used with actual gambling data, given that variable distributions are often highly skewed within these data. We believe these comparative results demonstrate that the distributions of actual gambling data (in their raw un-transformed state) do not readily facilitate parametric statistical testing, even with corrections for unequal variances, and can lead to type-2 errors. That being said, our median test effect sizes for all demographic comparisons were small, meaning that while our results do point to meaningful demographic differences, these results should still be interpreted as exploratory instead of confirmatory.
Additional information
Notes on contributors
Timothy C. Edson
Timothy C. Edson, PhD is a research associate in Psychiatry at Harvard Medical School and a Research & Evaluation Scientist at the Cambridge Health Alliance Division on Addiction. His research is mainly focused on deviant and criminal behaviors, as well as the intersections between these behaviors and mental health.
Matthew A. Tom
Matthew A. Tom, Ph.D. is a research associate in Psychiatry at Harvard Medical School and a Research Data Analyst at the Division on Addiction, Cambridge Health Alliance, a Harvard Medical School teaching hospital. Dr. Tom’s research interests include the study of gambling behavior and the detection of problems with gambling.
Eric R. Louderback
Eric R. Louderback is a Research & Evaluation Scientist at the Division on Addiction, Cambridge Health Alliance, a Harvard Medical School teaching hospital. He has expertise in diverse areas including quantitative risk assessment models for online gambling, responsible gambling program evaluation, geospatial mapping approaches using GIS, and open science methods.
Sarah E. Nelson
Sarah Nelson is the Director of Research at the Division on Addiction, Cambridge Health Alliance, and an Assistant Professor at Harvard Medical School. Her research includes studies of internet gambling and daily fantasy sports, and the examination of early online play patterns to detect gamblers at risk for gambling problems.
Debi A. LaPlante
Debi LaPlante is Director of the Division on Addiction at Cambridge Health Alliance and an assistant professor at Harvard Medical School. Dr. LaPlante is interested in the epidemiology of gambling and gambling-related problems, and addiction studies.