Open Collections

Deng, Xiaolei

University of British Columbia

The internet has increased the accessibility of gambling, with a corresponding increase in the risk of gambling problems. To date, little research has examined the risk profile of different forms of online gambling, or considered financial data to predict gambling risk. This dissertation describes a series of secondary analyses looking at the ‘concentration’ of online gambling activity, i.e. to what extent a small set of highly-engaged gamblers accounts for the majority of gambling activity. This is explored via the Pareto (20:80) rule and Gini coefficients. To link these effects to likely gambling problems, enrolment in Voluntary Self-Exclusion (VSE) is used; VSE programs enable a gambler to bar themselves from online and physical casinos. The dataset is from a Canadian provincial online casino for British Columbia from 2014-2015, consisting of 30,902 gamblers making over half a billion bets. Study 1 quantified the overall concentration effects for the eCasino, based on the number of bets and the net loss. Study 1 also examined Pareto values over time, finding that concentration values accumulated at longer timeframes. In the full year dataset, the 20% most active gamblers accounted for roughly 90% of gambling activity, and this elevated concentration (compared to the traditional 20:80 Pareto rule) was associated with elevated VSE rates among most active gamblers. Study 2 extended these findings by examining concentration metrics among four categories of gambling products. Video poker had the highest concentration values, and online slot machines had the lowest values. Study 2 proposes a novel measure of personal risk, combining product engagement with the concentration estimates. These personal risk scores were found to be higher among VSE gamblers. Study 3 found that gamblers with a VSE record showed higher levels of payment behaviours, including deposits and withdrawals and via more payment channels, compared to non-VSE gamblers. Random Forest models were able to classify VSE status at AUROC = 0.8 using only payment variables, and performance further improved to AUROC = 0.87 when personal risk scores were included. These results support utility of concentration effects as an indicator of product risk, and the utility of payment variables in risk modelling.

Text

2024-12-20

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/89963

Psychology

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/