Abstract
The proliferation of gambling opportunities worldwide, including continuous online gambling, has generated concern over how to protect individuals and families from harm caused by excessive spending. In response, researchers and operators have worked with big data to develop risk-identification models to identify indicators of problem gambling. Such models are generally proprietary, non-transparent, and non-generalizable across games, jurisdictions, or player populations, rendering them impractical as regulatory tools. In North America, responsible gambling efforts largely place the onus on players to control their behavior; however, in the UK and elsewhere, regulations have shifted to a model of shared responsibility that targets ‘affordability,’ the amount individual players can afford to lose, instead of indicators of problem gambling. This affordability approach avoids the need for regulators and operators to be clinicians, attempting to identify disorder. Rather, it builds on existing systems to determine creditworthiness and player risk levels.
Using affordability as the key benchmark for responsible gambling, we discuss approaches to operationalizing affordability guidelines in a North American context. Such guidelines will aid in promoting the objective identification of players who are spending beyond their means and facilitate the necessary transition to a shared responsibility model for harm reduction.
INTRODUCTION
Ahallmark of gambling disorder and subclinical problem gambling is spending more than one can afford to lose, typically due to impaired control. In a clinical context, loss of control, tolerance, withdrawal, and chasing are primary symptoms of the addiction, 1 suggesting that gambling outside affordable levels is a direct contributor to gambling-related harms and an indirect cause of harms experienced by individuals, families, and others in society. Impaired control involves the inability to reduce or stop gambling despite repeated attempts. 2 For these reasons, a primary consideration for gambling expansion is whether further proliferation of opportunities will result in more people spending beyond affordable limits, contributing to an increase in gambling-related harm.
Jurisdictions seeking to legalize new forms of gambling debate the potential costs of increasing the availability and number of activities, compared to the economic benefits. Regulators, charged with responsibility for maximizing revenue while minimizing harm, evaluate harm-reduction strategies, such as limit-setting tools, universal deposit limits, and, increasingly, the use of statistical modeling to identify players at risk. These modeling efforts show some promise but are hampered by a lack of consensus around the optimal selection of risk indicators and a standardized approach to operationalizing those indicators across gambling customers, products, and environments.
The purpose of this paper is to propose a complementary, proactive approach to addressing gambling-related harm in North America, a setting that remains heavily focused on individual responsibility in contrast to international jurisdictions with an increasingly institutional focus, such as Western Europe, Australia, and New Zealand.
We will first provide an overview of the challenges in identifying gambling-related harm. Second, we will review efforts to use big data and machine learning models to predict those at risk and explore the limitations of behavioral analytics efforts. We will propose an alternate or complementary strategy, currently required under the UK Gambling Commission's newly released updates to the Social Responsibility Code 3 , which includes the use of affordability assessments to assign risk and identify financially vulnerable players before they experience or engender significant harm. Finally, we will discuss approaches to adopting a working definition of affordability and indicators necessary to design an affordability model in North America.
THE QUANDRY OF HARM
Internationally, the advent of online gambling and sports wagering has virtually eliminated barriers to gambling opportunities, offering continuous access to a range of products for anyone with a computer or smart phone. In the U.S. alone, sports wagering is legal in 35 states plus the District of Columbia, with three states introducing active legislation. 4 According to the American Gaming Association, 2021 was the highest grossing year for the commercial gaming industry, generating $53 billion in revenue, including $4.29 billion from sports wagering, which increased 165% over 2020 figures 5 . Regulators tasked with implementing responsible gambling (RG) safeguards first must disentangle the wide and, often, empirically unproven range of options before deciding how to craft requirements that are likely to be effective but not unduly onerous to operators.
A necessary precursor to this process is to adopt a clear definition of reducible or preventable harms and to identify those at highest risk. This is not a simple task. It requires not only navigating the considerable lack of conceptual clarity around the nature and extent of gambling-related harms but also determining which of those harms could reasonably be identified and addressed by gambling operators.
Prior to 2015, with few exceptions, published research largely characterized and measured “harm” by clinical criteria, that is, the number and severity of cases of problem, pathological, or disordered gambling. This restrictive equivocation of harm meant that harm to families and communities was also restricted to the harm caused by the small percentage of gamblers who developed gambling problems. This was entirely inconsistent with the understanding of harm in other risk sectors such as alcohol, where harms include rates of alcoholism but also many other social and economic costs that arise even from episodic excessive consumption of alcohol.
Scholars subsequently argued for this focus to extend beyond prevalence rates of problem gambling to more inclusive harms experienced by gamblers, families, and communities. Notably, two research teams in Australia undertook multi-method studies to explore this issue, 6 , 7 spurring much debate in the scholarly community and more deliberate explorations of harm beyond problem gambling rates.
Two definitions emerged from early efforts. Blaszczynski and colleagues 8 characterized harm as: a negative consequence, associated with gambling, that had a significant detrimental interference on the functioning of an individual or societal domain. In contrast, Brown and colleagues 9 added long-term adverse consequences to their definition and broadened consequences beyond individual functioning to the health or wellbeing of an individual, family unit, community or population. Both definitions encompass a wide variety of consequences, both proximate and distal to gambling. They also underscore the difficulty in establishing a clear link between the harms and gambling, a task complicated by the high rate of comorbidity of gambling with substance use and other mental health disorders. 10 , 11
Worldwide, studies focused on gambling-related harm have reflected these difficulties, with a lack of consensus on definition and inconsistent sampling frames and measurement approaches. 12 For example, studies have calculated economic harm using player losses alone 13 and by evaluating losses based on the level of socio-economic disadvantage of neighborhoods. 14 Other scholars have argued for a public health approach to harm, measuring not only financial costs but also adverse effects on work/study, physical health, emotional/psychological, relationships, and social deviance (e.g., criminal activity, neglect of children) 15 , 16 as well as leisure and critical life events, 17 The latter approach calculates the burden of harm as the aggregate impact of gambling harms on quality of life in a population. 18 Measuring harm, therefore, remains elusive, depending on the zone of impact and the likelihood that self-reported consequences are, in fact, proximately related to gambling.
However, irrespective of the scope of harm, it is generally agreed that losses generated by gambling are an essential precursor to future harms, often for those who can least afford to experience them. Financial harm from gambling losses, for example, has been identified as a core harm that leads to others, such as relationship, mental and physical health problems. 19 In a recent analysis of seven years of anonymous customer bank data, for example, Muggleton and colleagues 20 found that gambling was associated with higher financial distress and lower financial inclusion and planning; negative lifestyle, health, well-being, and leisure outcomes; higher rates of future unemployment and physical disability; and, at the highest levels, substantially increased mortality.
Responsibility for limiting player losses to prevent financial harm is a burden that should, arguably, be shared across players, operators, and regulators. 21 , 22 However, the effects of etiological predisposing factors, cognitive distortions, and operant conditioning in a gambling environment 23 , 24 may impair the judgment of individuals, resulting in a loss of control. In a meta-analysis of 104 prevalence studies, researchers found that the highest risk activities are continuous-play formats, such as EGMs and internet gambling, with a high rate of play and short time between wagering and the outcome. 25 Binde 26 found similar results across 18 national prevalence surveys of mostly European countries, whereby interactive Internet gambling, casino gambling, and EGMs were most often associated with problem gambling.
Therefore, it is prudent, particularly in the context of considered expansion, for regulators to require and/or operators to adopt empirically supported strategies to assist players in limiting losses that are associated with gambling harms.
THE COMPLEXITY OF PLAYER ANALYTICS
Internationally, operators increasingly use in-house analytics to identify high-risk gamblers, although it has been suggested by one author that those efforts may be legitimacy-seeking strategies absent genuine engagement with the identified players or full commitment to harm mitigation. 27 In the UK, the regulatory response to this criticism has been to require operators not only to identify players at risk, but also to interact with those players and to evaluate the impact of their interaction in preventing harm. 28
Historically, identifying players at risk has proven difficult, as spending patterns alone do not differentiate between those who can and cannot afford to lose what they spend. In response, RG frameworks increasingly rely on identifying and measuring indicators of problem gambling. Those indicators were historically developed for land-based venues through the codification of behavioral risk indicators such as emotional distress or aggressive behaviors as red flags, 29 to be observed by trained venue staff. 30 Such indicators relied on staff willingness and ability to notice them and their confidence in intervening with the customer when they did notice them.
In jurisdictional debates about the legalization of online gambling, those in favor of legalization touted the unprecedented availability of play-by-play behavior to identify risk more objectively and rapidly than in land-based settings. Hence, the advent of online gambling led to a relatively new strategy to predict at-risk players: using the players' own behavior in “big data” analytics to identify patterns associated with harm.
Two key factors have driven the growth of player risk analytics. The first is the long-standing collection and use by gambling operators of extensive data on their players, used for purposes such as: assigning complimentary rewards or ‘comps' to engender customer loyalty and increase spending, designing targeted marketing strategies, and assessing financial qualifications for credit purposes and for the prevention of money laundering. Responsible gambling advocates have long asserted that, having collected player data for business or regulatory purposes, the industry should also use that data to identify players at risk and intervene.
The second driving factor, fueled by the continued expansion of mostly online forms of gambling such as sports wagering, is the concern that significant economic growth and revenue should not come at a cost to the most vulnerable citizens. This concern is linked to a range of study findings, indicating that a small proportion of players account for a majority of gambling revenue 31 , 32 . Gambling critics suggest operators rely on this highly engaged group of customers to generate revenue, and this serves as a disincentive for operators to encourage limit-setting, self-exclusion, or other safeguards to reduce harm.
Using big data to determine which players are spending more than they can afford has involved evaluation of play-by-play data to identify not only individual players but also trends and affected subgroups of players to inform higher level harm-reduction efforts. In New Jersey, for example, risky play behavior trended older for a number of years but has shifted downward by age since the advent of sports wagering; 33 this has implications for education and prevention efforts with youth, particularly young boys, who could begin gambling on parents' accounts before progressing to higher levels of gambling intensity.
However, New Jersey is the only state where the legislature mandated the preparation and publication of yearly reports, featuring detailed statistical analysis of bet-by-bet player data, demographics, and use of RG features to assess the relationship of online gambling and sports wagering to problem gambling. Regulators in other states are left to seek out reliable benchmarks that could generalize to their population.
Over the past 15 years, an increasing number of studies have utilized big data and machine learning to generate player risk models. 34 , 35 Studies have attempted to operationalize harm by designing models that predict outcome variables that are proxies for harm. These include self-assessment screens of gambling problem severity, 36 , 37 self-exclusion, 38 , 39 account closure, 40 , 41 exceeding preset limits, 42 , 43 or highly involved/extreme gambling, 44 , 45 , 46 , 47 which attempt to capture: Frequency (active betting days), Duration (average minutes/session or last betting day minus first), Variability (standard deviation of stakes or wagers), Trajectory (increasing wager size), Intensity (total # of bets; # of bets per active day); amount per bet (average and total); # of games; and number of sites wagered.
Despite considerable dedicated resources and promising models based on combinations of the above indicators, there appears to be no published evidence that the resulting models have identified the majority of high-risk players in the real-world environment or, more importantly, that a program of interventions guided by those models has successfully reduced player risk or harm.
Furthermore, establishing standardized requirements for the use of analytics presents several challenges for regulators. A majority of models are proprietary and confidential, using a wide range of play indicators often without evidentiary basis in the research literature; models are typically developed using data from one operator, a limited timeframe of data, and/or specific gambling products in jurisdictions with different regulatory frameworks. There is wide variation in play across casino games and types of sports wagering bets that makes it difficult to apply a common set of indicators across gambling situations. Indeed, even the most widely accepted indicators often fail to identify probable disordered gamblers with reasonable accuracy. 48
Developing generalizable predictive models, then, would require not only identifying the most reliable proxy outcomes and predictive indicators but also adapting those models to a range of environments (product, venue, and channel of delivery), validating them across all operators, and mandating standardized requirements for intervention with players at highest risk. From a regulatory perspective, it is time to shift the focus away from attempting to identify problem gamblers through the inexact science of their play patterns to the more concrete and readily verifiable assessments of what each player can afford.
THE SHIFTING FOCUS: FROM PLAY PATTERNS TO AFFORDABILITY
A primary challenge in regulating gambling products is managing shared responsibility. The Reno Model 49 suggests that government, operators, and players all have a role in minimizing harm. Practically, however, players are typically at a disadvantage, as operators use sophisticated analytics, marketing strategies, and inducements to encourage players to spend more. Moreover, some self-exclusion programs expressly state that individuals who may have gambling disorder, a hallmark of which is loss of control, are solely responsible for controlling their behavior. Given the lack of consistent and reliable strategies to identify and minimize harm, combined with the lack of social responsibility requirements for operators, there is a pressing need for an objective, standardized protocol to identify players at risk of gambling harm and to provide some form of intervention, assistance, or barrier to unaffordable losses. Affordability guidelines could provide such standardization.
The origins of affordability
The concept of affordability has its origin in both socioeconomic policy and financial regulatory contexts. These seemingly disparate approaches, however, are useful to inform the development of standardized affordability protocols to promote RG.
Socio-economic policy origins
In policy contexts, the term is used to discuss considerations around financial harms to individual players, families, and communities from excessive gambling. 50 Such evaluations are complex, as significant comorbidity between problem gambling and other mental health and addiction issues make it difficult to attribute health and mental health harms solely to problem gambling.
For that reason, financial harm has emerged as a more objective and measurable construct. It is also an important marker to evaluate in the context of harm to families and children, because individuals with gambling problems can often accumulate significant debt 51 and hide the extent of their losses until a family crisis brings to light the severity of losses and debts from gambling. 52 Muggleton and colleagues 53 examined banking activities over a seven-year period and reported that banking customers with higher levels of gambling exhibited several markers of financial hardship: using an unplanned bank overdraft; missing a credit card, loan, or mortgage payment; taking out a payday loan; and being subjected to debt collection by bailiffs. In contrast to the significant variability in play pattern indicators, financial markers are easily quantifiable through statements or data from a credit reporting agency.
Central to this concept of affordability is the correlation between poverty and problem gambling, whereby socio-economically disadvantaged groups suffer a disproportionate share of the negative consequences. A scoping review of 27 studies found problem gambling was clearly associated with standard poverty measures, including being unemployed, having unstable housing or being homeless, having low income, and living in a disadvantaged neighborhood. 54 In addition, scholars have proposed that gambling contributes to the social disorganization and social deprivation of many communities, particularly low-income communities with indigenous and ethnic minority populations. 55 A Public Health England (PHE) report 56 found that, while the highest rates of gambling participation are among more advantaged populations (educated, employed and in less deprived groups), the relationship is reversed as gambling risk increases, such that, those who are unemployed and living in deprived areas experience the most gambling-related harm.
Gambling harm to those in the lowest socioeconomic groups, therefore, is more likely to exacerbate health inequities and place immediate demands (i.e., costs) on society for financial, health, and social welfare supports. It also challenges the basic tenet in most developed countries that acknowledges core obligations to protect those who are vulnerable or disadvantaged. Affordability as a core RG strategy could potentially reduce the costs of gambling harm to the society by prioritizing protection from harm for those who can least afford to gamble excessively and those close to them.
Regulatory origins
In a regulatory context, the concept of affordability originates in Anti Money Laundering (AML) regulations, where it has long been a central element of compliance. In this context, affordability assessments for individual players help private companies and government bodies determine whether deposits made by individual customers come from legal sources of income or wealth. AML requirements are remarkably consistent worldwide, the result of international agreement by more than 200 countries on the Financial Action Task Force (FATF) 40, International Standards on Combating Money Laundering and the Financing of Terrorism & Proliferation (2012-2022). 57 Pan-national and national proceeds-of-crime laws translate these 40 recommendations into consistent requirements for all risk industries, including gambling.
As a result of AML requirements, operators have existing systems in places to assess player financial data for potential money laundering. In the U.S., for example, complying with the Bank Secrecy Act (BSA) requires operators to report single or aggregated currency transactions in or out of more than $10,000 per day.
To comply with the law, policed by the Financial Crimes Enforcement Network (FinCEN), operators are typically required to: 1) obtain the name, permanent address, and social security number of the player prior to each deposit of funds, account opening or extension of credit and 2) verify the name and address of the person at the time of these events. 58 Operators must also develop and implement compliance programs to manage the risk of illegal activity and file reports on customer transactions that include risk-based procedures to prevent individuals from circumventing the BSA.
In 2021, recognizing unique limitations of online gambling and wagering, FinCEN issued exceptive relief from certain customer identity verification requirements, namely those that require visual verification. In response to operator feedback, FinCEN allowed the use of third-party databases, which pull information from a multitude of publicly available resources to provide comprehensive verification of an online patron's identity, validate the information as legitimate and unsuspicious, and provide an overall risk assessment based on the information obtained. 59
These data and risk assessments from third-party providers also could be used to assess player affordability. It is impossible to determine from play patterns alone who is spending beyond their means. This stands in contrast to other addictions like alcohol, where the combination of weight, height, and sex alone provide a benchmark for the number of drinks an individual can consume before reaching intoxication. In the U.S., for example, a bartender can be liable for injuries to third parties that result when they continue to serve alcohol to someone they knew or should have known was intoxicated.
One could argue—and some European regulators maintain—that operators are similarly responsible for “serving” too many gambling activities to individuals they knew or should have known were spending beyond their means. Lawsuits have arisen around land-based casinos, alleging that slot machines are inherently deceptive and addictive products; these suits have met with little success, 60 largely because they concern land-based venues where players make a series of choices, move about the venue, and have access to informed choice material that should, in theory, put them on notice about the odds of winning. However, the growing volume of individual player data as well as algorithms, developed by private companies to assess player debt and ability to pay, will make it increasingly difficult for operators to claim they had no information that players were experiencing increasing amounts of financial harm.
Adopting an affordability perspective renders more objective the process of addressing harm and removes the burden on regulators or operators to determine who is or is not a problem gambler based on models constructed from risk indicators that vary in reliability. It also shifts the function of risk-analytics away from predicting problem gambling to a complementary function of informing risk levels for affordability.
This policy shift is already evident in the UK, where a majority of regulatory settlements in recent years have involved dual AML/social responsibility failures whereby the information operators obtained in the course of AML due diligence was treated as information that should also be used to identify players at risk and inform RG responses by operators. Consistent with research that highlights the overlap between criminal acts and problem gambling, 61 recent regulatory decisions in that country have treated AML assessments of affordability as part of operator due diligence for RG as well; this essentially eliminated the longstanding wall between AML and RG, a move that will be difficult to undo. Operators with licenses in the UK were called upon to integrate assessments of affordability across AML and RG in June 2019, 62 with more specific guidelines provided thereafter and forthcoming. In the U.S., risk assessments from third-party data providers, recently sanctioned by FinCEN for AML purposes, could likewise be used to identify players spending beyond their means.
Operationalizing affordability in a North American context
Applying affordability guidelines in a North American context will require regulators to reconcile a number of important issues: defining affordability, establishing thresholds and triggers; determining who will conduct, communicate, and monitor assessments of affordability; codifying action that operators are required to take when customers exceed thresholds; and specifying the role of individual customers in setting personally affordable limits.
Defining affordability and thresholds
Given the wide variation in income among patrons, seemingly huge expenditures may represent entertainment for one player but bankruptcy for another. In anticipation of the UK Government's revisions to the 2005 Gambling Act, various stakeholders have grappled with the concept of affordability. Noyes and Shepherd suggested the following definition: “Gambling is only affordable when it does not impede other financial commitments that a household must fulfill in order to achieve a socially acceptable standard of living (p. 8).” 63 They added that, once these commitments are met, individuals should be free free to spend as much of their disposable income on gambling as they see fit. In this way, individuals can spend within their means and limit opportunities for harm. 64
Under that definition, it is important to consider whether an individual's gambling is impeding other financial household commitments and/or impairing their standard of living. In the UK, operators have long opposed universal spend or loss limits, contending they have the data to identify those at risk of harm. However, proponents of limits say operators, despite having the data, have failed to actually limit spending by those who are gambling beyond their means or to adopt standardized criteria across operators. 65
Noyes and Shepherd reviewed affordability calculations for the housing, food and healthcare sectors and proposed that the UK adopt a “thresholding system,” based on household disposable income, contextualized by minimum income standard (i.e., poverty level); they asserted this system should be used to establish a soft cap limit (e.g., £100 or $123 USD per month). To spend above that level, individuals would need to submit to an affordability check by an independent third-party data depository. Under these proposed guidelines, gambling compliance data companies could map individual customer data against sociodemographic and economic data sources in the public domain to assess risk.
In the U.S. context, it first would be important to make the additional distinction between disposable income (i.e., net pay after taxes, social security, and Medicare) and discretionary income, or the amount remaining after taxes and household expenses are paid. We would argue that affordability should be based on discretionary rather than disposable income. Discretionary income could be calculated using strategies analogous to those used to guide income-driven Federal student loan repayment plans, which generally calculate discretionary income as the difference between annual gross income and 100% to 150% of the poverty guideline for family size and state of residence.
For example, to calculate affordable spending for someone living in a family of four in New Jersey making $50,000 per year, we would reference the 2022 U.S. Department of Health and Services state poverty guidelines, which estimates the family would need $27,750 in income to remain above the poverty line. To calculate discretionary income using 150% of the poverty guideline, we would multiply $27,750 by 1.5 ($41,625), then subtract that amount from gross income. This would leave $8,375 per year ($698/month or $161/week) in discretionary income to spend on non-essential items and activities (e.g., movies, vacations, etc.), including gambling.
In Canada, the Market Basket Measure (MBM) calculates the cost of a specific basket of goods and services (food, clothing, shelter, transportation, and other items for a family) that represent a modest, basic standard of living. Families fall below the poverty line if the MBM is higher than their disposable income. The MBM for a family of four, earning $73,000 per year in a large Ontario city, would be $46,306 CDN. 66 Using the same formula to calculate discretionary income of 150% of the poverty line, we would multiply $46,306 by 1.5 ($69,459 CDN). Subtracting that amount from their gross earnings ($73,000 minus $69,459) would leave the family $3,541 per year ($295/month or $68/week) to spend on gambling and other non-essentials.
Communicating to customers
Once defined, affordability guidelines should be communicated to customers, so they understand why and how operators need affordability assessments. This would reduce the friction operators fear when they must request personal financial information. All customers should be asked to calculate their own “affordability” and be presented with tools to set limits on time and money spent. In New Jersey, for example, researchers have long advocated for an opt-out rather than an opt-in system for setting limits; this would require patrons to review guidance on how to set personal limits at sign-up, then have the opportunity to set each type of limit or opt out. 67 In this way, RG limit-setting options, which are often overlooked or buried in fine print, would become a natural part of the onboarding process for online play.
Conducting assessments
After adopting a discretionary income standard, it also would be important to consider how affordability assessments would be triggered and managed. One approach is to set a universal spend threshold after which patrons must submit to affordability assessments to continue gambling. This requires setting a generalizeable threshold. Alternatively, given that personal data is widely used and accessible, all patrons could be assessed regularly through an affordability risk profile as long as they are active players. With either approach, regulators must determine who will administer assessments, what access operators will have to patron credit and other financial information, and the purposes for which such data can be used, including in their well-established role in preventing individuals from laundering money or gambling with proceeds of crime. 68
In the UK, the All-Party Parliamentary Group for Gambling Related Harm recommended that third-party services, such as Experian Open Banking service, oversee affordability checks, providing a “wall” between risk assessments and operators, who stand to lose substantial revenue if all problem gambling patrons begin spending within their means. 69 The House of Lords report states: “The people most at risk are also the most profitable to the industry: the greater the problem, the bigger the profit”. 70 The group also recommended that all operators use one single sign-on platform, operated by a third-party; patrons would create a user profile, have their identity verified, set optional play limits, and receive an affordability check and guidelines for what they could spend, ideally tied to deposit limits.
Based on this argument, creating an arms-length centralized system to assess affordability in each state would likewise be important in North America. Data from New Jersey, for example, suggest that about 5% of online players place 65% of the bets and wager 55% of the money, 71 highlighting the need to ensure that high-intensity players, irrespective of platform or location, are held to the same standard of affordability. This centralized platform could be served by a single or multiple vendors that would conduct affordability checks and provide limits based on affordability guidelines for individual customers or sub-groups based on income ranges. Affordability platforms could establish a risk warning system that would notify operators when a customer's spending was approaching or exceeding affordability thresholds to trigger operator-specific interventions.
It is beyond the scope of this paper to detail the specific actions required of operators if intervention is necessary to prevent financial harm. For some operators, existing protocols for intervening with at-risk customers may be sufficient once affordability has been fully implemented as a fundamental risk-identification requirement. However, such protocols should include preventing a customer who has lost control from spending far beyond affordable limits. The UK Gambling Commission has suggested that operators should enact interventions that are tailored to individual players and tiered sufficiently to minimize future harm, which may ultimately include terminating the business relationship. 72 Enacting affordability guidelines is a first step in this process.
CONCLUSION
Absent a federal system of oversight, regulators in the U.S. and Canada are in the best position to provide player protection and promote responsible and non-predatory gambling practices. The rapid expansion of gambling has provided significant revenue to operators, states, and provinces. However, many players may be spending more than they can afford to lose. At this time, developing game-, venue-, and jurisdiction-specific machine learning models is impractical and will provide only partial information relevant for harm reduction. Guidance from the UK Gambling Commission specifies that, in addition to gambling behaviors, operators must use a range of additional indicators to identify harm or potential harm associated with gambling: customer spend, patterns of spend, time spent gambling, customer-led contacts, use of gambling management tools, and account indicators. 73
Affordability, which can be objectively and extrinsically measured, represents the crucial but currently missing information needed to to identify players who cannot afford what they are spending, in contrast to those whose excessive spending is supported by their income. Adopting such guidelines will be a substantial step toward a holistic approach to prevention as well as harm reduction in a North American context.