Gambling availability has increased during previous decades because of growth in internet sites which have enabled gambling from home, work and nearly anywhere. Gambling is a common form of entertainment in many western countries, and most of adults have participated in it at least sometime of their life (1,2). However, for some people gambling can cause serious health, financial and interpersonal harms (3,4). Financial harms, such as lost savings and debt problems, are the most common harms reported by gamblers (5). Gambling-related financial harms can in turn increase psychological distress, substance use, relationship problems, crime, and even suicidality (4,6). Based on the Lotteries Act, the aims of the Finnish gambling monopoly system are to prevent and reduce gambling-related financial, social and health-related harm (7).
There is strong evidence that different types of health problems are linked with problem gambling, and its most severe form, gambling disorders (GD) (e.g. 8,9). It is clear that more severe morbidity tends to cause higher expenses. Accordingly, the need to calculate gambling costs has been identified in several countries (10). Browne and colleagues (11) found in their systematic literature review on gambling-related harms and costs that only three out of 36 gambling-prevalence studies published in 2010–2016 reported gambling costs. The most cited and emulated research in the field of gambling costs studies found that the overall gambling costs in Australia were 1.8–5.6 billion Australian dollars (AUD) between 1997 and 1998 (12). A more recent study from Australia estimated cost to be $7 billion in 2014–2015 in Victoria (11). Other studies conducted in Europe (8,13–18) and Asia (19) have also identified significant costs to society. However, due to a range of differences in gambling environments and policies, and different methods behind these calculations, comparison of costs between countries is not straight-forward.
There is also a lack of consensus on what should be included when calculating gambling costs (20,21). There has been also discussion whether intangible (pain and suffering) costs, which are more difficult to evaluate in monetary form, should be included in calculations. While these questions are not easily resolvable, it is nevertheless advisable to pay attention to the methods how gambling costs are calculated. Regardless of past research showing a strong association between gambling and harms, relatively few studies have examined costs of these harms to society. Thus, we will start by reviewing methods used in the gambling cost studies and evaluate them critically. Our emphasis is on tangible costs, where a monetary value can be most readily applied. This is not to deny the importance of other costs, but rather to make our evaluation more tractable. This review informed our evaluation of alternative methods for calculations that follow later in this paper.
Gambling and health issues
There is strong evidence on that psychiatric and substance used disorders are linked with gambling problems (e.g. 22–28). A Swedish registry-based study indicated that 73 percent of those who had a diagnosed GD had other co-occurring psychiatric diagnosis (23). A meta-analysis showed that the highest mean prevalence for co-occurring psychiatric disorders were for nicotine dependence (60.1%), followed by a substance use disorder (57.5%), mood disorders (37.9%) and anxiety disorders (37.4%) (25). Overall, self-rated health is found to be lower among those with problem gambling than non-gambling counterparts (29,30).
There is evidence of comorbidity between problem gambling and poor physiological health (9,29–32). Problem gambling is linked to poor diet, low physical exercise, and obesity (9,29). Moreover, those with gambling disorder are more likely than low-risk individuals to be diagnosed with tachycardia, angina, cirrhosis, and other liver disease (31). In addition, problem gambling is associated with headache, fatigue, and sleeping problems (33). Among women, problem gambling is linked to bronchitis, fibromyalgia, and migraine (34) and among older adults, with heart conditions (32). There is also evidence that individuals with GD have an increased risk of receiving sickness allowance (35) and increases risk of work disability (36).
In Finland the most common reasons for receiving sickness allowance and disability pension are mental and behavioral disorders. Mental and behavioral disorders account for over half of disability pensions and over one third of sickness allowances (37). The likelihood of transitioning to disability pension increases significantly after the first year of receiving sickness allowance (38), with the risk escalating further with the duration of the sickness absence (39). Disability pension poses a substantial financial burden on society due to the low rates of recipients returning to work (40).
Cost of illness (COI) studies
Cost of illness (COI) is defined as the value of the resources that are lost because of a health problem. COI studies assess the economic burden of health problems on the population overall (41). COI studies can be used to draw public attention to particular health issues and to provoke policy debate (42). They can also guide planning of healthcare and preventive services, interventions and the evaluation of different policies (43–45).
COI studies can be based on the incidence or prevalence of the disease. In an incidence-based approach, the new (incident) cases are measured, while the prevalence-based approach measures the existing (new and pre-existing cases) cases over a specified period, usually one year. It is considered that prevalence-based approach is more appropriate for assessing total current economic burden of a health problem whereas an incidence-based approach is more useful for estimating the expected impact in the future (46). In this study we applied prevalence-based approach.
COI studies commonly include related healthcare costs and other resources used (direct costs), losses of productivity related to morbidity and mortality (indirect costs), and the losses in quality and length of life (intangible costs) (43). Traditionally these effects of health problem are converted into monetary values wherever possible (41,44). However, intangible costs are not usually monetized; instead they are expressed measures, such as disability-adjusted life-years (DALYs) or quality-adjusted life-years (QALYs) (41).
Although debate regarding the most appropriate approach to calculate productivity losses is still on-going, we opted to use the human capital approach (47). The value of the human capital is estimated based on the value of an average individual's future earnings. On the other hand, the fractional costs approach attributes only 80% of losses to avoid potential overestimation of indirect costs (48). This discounting is intended to account for the fact that it cannot be assumed that the condition plays a 100% causal role in impacting earnings. However, given the true causal role is unknown, any such discounting is unavoidably somewhat arbitrary.
Methods commonly used in gambling cost studies
Causality adjustment factors
As mentioned earlier, the most emulated method is from the Australian Productivity Commission (12), and it is still widely used. The main principle behind this method is that cost of the harm per gambler is multiplied by the number of people experiencing the harm. Because the causality between gambling and harms is often unknown, costs have been discounted with a ‘causality adjustment factor’ (CAF) (12). This based on expert opinions suggesting that approximately 20% of individuals struggling with gambling issues would have encountered similar personal and family-related consequences even in the absence of gambling problems. In practice, this entails that costs are discounted by 20 percent (multiplied by 0.8), similar to the fractional cost approach described above. In some studies costs were discounted by as much as 50 percent, as there were no or only little evidence on the direction of causality (8). The costs are calculated as following,
Cost of harm = NG * CAF* C, where (1.1)
NG= estimated number of gamblers with a particular harm
CAF= causality adjustment factor
C=unit cost of individual with harm.
Despite acknowledging the limitations of discounting costs by an arbitrary CAF, many studies have used the method (18).
Excess costs
Three research reports conducted in Britain (15,17,49) quantified costs by calculating the excess costs between gamblers compared to the non-gambler population and then multiplying this excess cost by unit cost of individual harm as following:
Cost of harm = (NG - NP) * C, (1.2) where
NG= estimated number of gamblers with a particular harm
NP= estimated number of gamblers expected to have harm if they had same rate of harm as the general population
C=unit cost of individual with harm.
This Excess cost does not use actual data from real cases. Estimated number of gamblers expected to have experienced a particular harm is calculated by multiplying prevalence rate of a particular harm by the prevalence figures for problem gambling. This gives an estimate of the number of gamblers with specific harms in the general adult population. Multiplying these figures by the estimate of association between gambling and these specific harms (adjusted-OR) produces the number people of experiencing problem gambling who are expected to have these particular harms. The difference of estimated number and expected number of specific harms, (NG − NP), will give an estimate of the number of people with a particular harm associated with problem gambling only.
Method based on Bayes' Theorem
Bayes' Theorem is a mathematical formula used for calculating conditional probabilities (50). In its simplest form, Bayes' Theorem is expressed as:
P(A∣B) is the posterior probability of event A given that event B has occurred.
P(B∣A) is the likelihood of event B occurring given that event A has occurred.
P(A) and P(B) are the prior probabilities of events A and B respectively.
We can use Bayes' Theorem to estimate the proportion of those experiencing long-term work disability who likely had their problem gambling lead to the work disability:
P(gambling led to pension) = [P(past gambling problems | long-term work disability) * P(long-term work disability)] / P(past gambling problems)
Where:
P(past gambling problems | work disability) is the rate of gambling problems before long-term work disability had started among those who are currently on long-term work disability
P(disability pension) is the overall rate of long-term work disability in the population in 2016
P(past gambling) is the rate of gambling problems before long-term work disability had started in the general population
If P(past gambling problems | work disability) is significantly higher than P(past gambling), it suggests that past gambling problems increase the likelihood of long-term work disability. Formula 1.3 provides the proportion of people with long-term work disability whose gambling issues likely occurred before and played a role in their extended work incapacity.
To conclude, during the past decade there has only been limited discussion of the appropriate method for calculating gambling costs. Based on the extant literature, we have selected three methods, including the Causality adjustment factors (with two variations), the Excess costs and the Method based on Bayes' Theorem, and use them in our gambling costs calculations. Our calculations focus on the indirect costs regarding to long-term work disability, in those aged 18–64 years who experienced gambling problems before long-term work disability started. This is a specific context for gambling cost assessment for which high quality data is available, should provide good scope for evaluating alternative costing methods. We do not mean to suggest, however, that these are the only costs – financial or otherwise – that can be due to gambling issues. Rather, the goal of this study is to provoke discussion on methods used in gambling cost studies and hopefully assist the formulation of a consistent approach for cost calculations in the field of gambling studies.