Behind the Bets: How payments data identifies potential at-risk gamblers

Magnifying glass in front of coins illustrating a data graph
Image: Shutterstock

As Problem Gambling Awareness Month draws to a close, it is worth looking into less obvious and traditional ways of identifying problem gambling behavior. New Jersey is already mandating operators to identify and act on problem gambling markers if they want to operate in the state, and more states could soon follow suit.

With that in mind, Sightline’s SVP of Strategic Development and Government Affairs Jonathan Michaels and Dr. Kasra Ghaharian, a Senior Research Fellow at the UNLV International Gaming Institute, have put together this piece on research around how payment-related behavior can help in flagging problem behavior before it gets out of control.

Dr. Kasra Ghaharian

The potential rise in problem gambling in the United States is undeniably a popular subject of discussion—and rightfully so. Today, legal mobile gaming products are available to consumers in 24 states, reaching almost 60% of the US population. This Problem Gambling Awareness Month, we’d like to highlight a new stream of gamblers’ behavioral data that could help inform operators, regulators, and advocates about ways that payment transactions can be used to help identify potential gambling-related harms.

Problem gambling is a clinically diagnosed addiction, so behavioral data alone cannot determine whether an individual has a gambling problem. But markers of harm illustrate atypical behavior in customers that can help indicate when a person may be at risk and struggling to gamble responsibly.

The question is: can payments-related data identify additional markers­—and if so, how? That’s what Sightline and the University of Nevada Las Vegas (UNLV) sought to answer when creating the Payments Research Collaborative at UNLV’s International Gaming Institute. Sightline provided UNLV with data from millions of transactions across the full spectrum of operators—online casino, mobile sports betting, and cashless gaming at casinos—with the goal of understanding if payments data can help identify whether a customer may be at risk of experiencing gambling harms.

Researchers at UNLV computed behavioral variables for thousands of customers based on their deposits and withdrawals. They then used a machine learning algorithm to cluster similar customers together based on these variables, ultimately revealing five clusters of customers with distinct payments behaviors.

Jonathan Michaels

The results indicated that most customers, around 88%, exhibited patterns of behavior that were not suggestive of any unsustainable or harmful behavior. This finding aligns with other studies and the notion that most people who gamble do so responsibly and within their own personal affordability limits.  

Within the remainder of the customers studied, there were three smaller clusters that exhibited payment behaviors that may represent customers who are at potential risk of experiencing harm. Notably, the data alone cannot indicate that these customers are problem gamblers­—merely that their behavior patterns were outside the norm.

The smallest cluster, which comprised 1.2% of customers, exhibited a high volume of activity in the number of their deposits and withdrawals, as well as high variability in their deposited amounts. With this high variability in the amount deposited and overall high net spend, we need to ask whether these customers could be engaging in unsustainable gambling behaviors, for example, loss chasing.    

The second smallest cluster, representing 2.5% of customers, exhibited the highest deposit activity, averaging almost 15 deposits per week. These customers, on average, also had the largest net balance deficit of any cluster and exhibited the highest number of declined transactions—a potential red flag that might indicate their frequency or amount limits were reached. Investigating the efficacy and optimal structure of these and other limits (e.g., amount limits) is an important avenue for future research.

For a cluster containing roughly 8% of the customers, transaction frequencies, amounts, and variability were higher than the majority of customers in the dataset, but in most cases, not as high as the two smaller clusters. These customers had the lowest number of withdrawals per deposit­—approximately one withdrawal for every 53 deposits. This meant that while the average net loss of these customers was not as high as the two smaller clusters, they recuperated the smallest fraction of their total deposited amount.

Identifying these three clusters of atypical behaviors demonstrates that researchers can use payments data to identify potential markers of harm and explore interventions to reduce gambling harms. This research could help inform the development of a machine learning-driven responsible gambling technology at the payment level.

Zooming out, payments have helped fuel the incredible growth of the legal US online gambling market over the past several years. And as the market continues to grow, the need for responsible gambling practices becomes even more crucial. Through the power of data, payments can also help identify where players might need additional help to ensure they gamble responsibly.

Data alone cannot prevent problem gambling, but it can provide critical indicators that help operators, regulators, and advocates take proactive steps to combat it. This Problem Gambling Awareness Month, the importance of identifying these potentially vulnerable gamblers can’t be overstated. This research was a vital first step towards achieving that goal—but for the sake of those battling gambling addiction, the industry must go further.

Jonathan Michaels is Sightline’s Senior Vice President, Strategic Development and Government Affairs. Dr. Kasra Ghaharian is a Senior Research Fellow at the UNLV International Gaming Institute. During his PhD, Kasra led and supervised the Payments Research Collaborative activities, including data curation, conceptualization of overarching research goals, methodology development, analysis and model development, and creation of published works. 

For questions about the research, contact Dr. Kasra Ghaharian. For the full research article, go to https://doi.org/10.1016/j.chb.2023.107717.