Measuring Gambling Harm: The Influence of Response Scaling on Estimates and the Distribution of Harm Across PGSI Categories.

Abstract

Recent research has shown that harm is not just a feature of problem gambling, but can also be observed in other lower risk categories. Some debates exist, however, as to the distribution of harm across these categories and how harm should be best measured. This study was designed to examine how estimates of self-reported harm are affected by the methodology used. A particular focus was on how harm estimates for low and higher risk gambling (as classified by the PGSI) varied when respondents were able to make more graded attributions of their harm to gambling. An online panel sample of 554 gamblers responded to a brief survey that included the PGSI, measures of gambling harm drawn from Browne et al. (Assessing gambling-related harm in Victoria: a public health perspective, Victorian Responsible Gambling Foundation, Melbourne, 2016) as well as questions about demographics and gambling habits. The recruitment was designed to obtain good representation of each PGSI group, with 23% found to be problem gamblers; 36% moderate risk and 21% low risk gamblers. In support of Browne et al. (2016), the findings showed that higher proportions of harm in low risk gamblers is likely to be identified when one uses binary or 'any harm' scoring, but that this effect mostly disappears when more graded scoring or attribution of harm measures are used. Higher risk PGSI groups consistently reported more harms and more serious harms than lower risk groups. It was concluded that the measurement of gambling harm and its estimated distribution over PGSI categories is quite sensitive to how it is measured.