Introduction

Gambling has been an important part of North American Indigenous culture for at least 1000 years prior to European contact (Binde, 2005; Culin, 1907; Williams et al., 2011). Engaging in gambling was believed to activate and promote the gathering of supernatural spirits. Consequently, it was a frequent part of ceremonies associated with ensuring a good harvest or hunt, producing rain, or marking the changing of the seasons. For similar reasons, gambling games were engaged in to help cure sickness, expel demons, aid in fertility, and to facilitate passage to the afterlife after death (Culin, 1907; Salter, 1974, 1980). These games were also an important aspect of inter-tribal interaction as it provided a forum for nonviolent competition between villages, clans, and tribes, as well as an opportunity for socializing and trade (Belanger, 2006; Binde, 2005; Williams et al, 2011). In sum, gambling played an integral role in education, diplomacy, trade, and conflict resolution, and can be considered an important element of Indigenous political economy (Belanger, 2011a).

European colonialism transformed both the nature and types of gambling in North America. In contrast to the more spiritual/ceremonial/social purpose of traditional Indigenous gambling, Western forms have a recreational and commercial orientation. Evidence indicates that North American Indigenous people have now become largely westernized in their understanding and motivations for gambling as well as their pattern of play (Belanger et al., 2017; Williams et al., 2011, 2016). In keeping with this cultural shift, many North American Indigenous groups have become commercial providers of Western forms of gambling (Belanger, 2006, 2011b; Dixon & Moore, 2006; Schaap, 2010).

Some cultural differences still exist. A minority of North American Indigenous people continue to participate in traditional Indigenous gambling games (Belanger et al., 2017; Williams et al., 2011, 2016). Some differences have also been found in Western game preference, with higher Indigenous participation in bingo, electronic gambling machines (EGMs), instant lotteries, and lower participation in financial speculation (Williams et al., 2011, 2016). Overall level and intensity of gambling participation also tends to be higher (Belanger et al., 2017; Williams et al., 2011, 2016). This points to the area where the clearest differences exist, with there being much higher rates of problem gambling, which also appears to be true of Indigenous populations in other countries (Belanger et al., 2017; Breen & Gainsbury, 2013; Hing & Breen, 2014; Hing et al., 2014; Stevens & Young, 2009; Wardman, 2001; Williams et al., 2011, 2016; Young et al., 2007).

While several studies have shed light on the nature and pattern of gambling within North American Indigenous society, most of these have studied specific communities and/or employed relatively small samples. The larger scale investigation by Williams et al. (2011) is limited by the fact that Indigenous people in the study (n = 895) were from both the United States and Canada, despite some potential differences between these countries. This study is also somewhat dated, as the data collection occurred between 2004 and 2007. The other large-scale investigation by Belanger et al. (2017) and Williams et al. (2016) focused on urban Indigenous people (n = 1114) living in Canada’s three Prairie Provinces, which does not necessarily generalize to the rest of Canada.

Thus, one of the purposes of the present study is to provide an updated and comprehensive profile of Canadian Indigenous gambling in 2018. The two datasets utilized in the present study are part of a broad national investigation of gambling and problem gambling in Canada (AGRI National Project, ANP),Footnote 1 with the Indigenous subsamples derived from these datasets being the largest pertaining to Indigenous gambling ever collected in Canada. The second purpose of the present study is to identify strategies to mitigate gambling-related harm among Canadian Indigenous people. It is not sufficient to simply document that gambling-related harms continue to disproportionately impact Indigenous people. Thus, specific focus is given to the predictors of problem gambling within Indigenous people, the variables that differentiate Indigenous from non-Indigenous problem gambling, and how these results can be operationalized into meaningful public health policy.

Method

Canadian Community Health Survey (CCHS)

Sample

The CCHS annually collects information from a target population of 65,000 Canadians aged 12 + who reside in one of Canada’s ten provinces and three territories. The sample excludes full-time members of the Canadian Forces, youth aged 12–17 living in foster homes, the institutionalized population, and people living on First Nation reserves (Statistics Canada, 2019a). The adult (18 +) sample is roughly proportionate to provincial and territorial population size while also ensuring reliable estimates for provincial health regions. Each province is divided into geographic areas consisting of 100 to 600 dwellings (‘clusters’). Households are sampled within each cluster and an individual is randomly selected from each household, with ages 18–35 and 65 + being given a higher probability for selection (Statistics Canada, 2019a).

Introductory letters explaining the purpose of the survey were first sent to the selected households. CCHS surveys were conducted between January and December 2018 by computer-assisted telephone interviewing using both cell phones and landlines (75%)Footnote 2 and computer-assisted face-to-face interviews (25%). The survey was available in both English and French with interpretative services available for several other languages (Statistics Canada, 2019a). However, the CCHS containing the Gambling Module was only fielded for a six-month period (July–December 2018) and only in the provinces (no territories).Footnote 3 Surveys were obtained from 26,648 individuals in this time period, which represents a 58.4% overall response rate. However, because the section containing the Gambling Module was restricted to ages 15 and older and did not permit proxy respondents, a smaller number of individuals were actually eligible with a total of 23,952 adult (18 +) surveys containing the Gambling Module ultimately being obtained (Statistics Canada, 2019b). Of this group, 1,324 individuals answering affirmatively to the question “Are you an Aboriginal person, that is, First Nations, Métis or Inuk (Inuit)? First Nations includes Status and Non-Status Indians”.Footnote 4

Statistics Canada subsequently created weights for the sample to produce population totals that match the 2016 census population for each age group x gender x health region cell.

Bootstrapping was undertaken in the present study to produce 95% confidence intervals for the gambling and problem gambling prevalence estimates in Tables 1 and 2.

Table 1 Prevalence of past year gambling among indigenous Canadian Adults (18 +) in 2018 compared to Non-Indigenous Canadian Adults (CCHS, weighted)
Full size table
Table 2 PGSI gambling categorizations among Canadian indigenous adults (18 +) in 2018 compared to non-indigenous Canadian Adults (CCHS, weighted)
Full size table

Questionnaire

The CCHS is a survey of health determinants, health status, and health care use. The total length of the average survey is 40 to 45 min (Statistics Canada, 2019a). In cooperation with Statistics Canada a brief assessment of gambling behaviour and problem gambling was developed (Gambling Module) and included in the 2018 administration (Statistics Canada, 2019b). The first part of the Gambling Module was an assessment of past year frequency of engagement in eight different types of gambling using an abbreviated version of the Gambling Participation Instrument (Williams et al., 2017). Respondents were asked about their frequency of in-person or online engagement with: instant lottery tickets, lottery or raffle tickets, electronic gambling machines, casino table games (excluding electronic versions), sports betting (including horse race betting, sports lottery tickets, fantasy sports, bets between friends), bingo, other forms of gambling, and speculative financial market activities (e.g., day trading, penny stocks, shorting, options, currency futures). Time constraints precluded asking questions about gambling expenditure and time spent gambling.

Respondents who gambled once a month or more on at least one type of gambling in the past year were asked the nine questions from the Problem Gambling Severity Index (PGSI), which produces a composite score ranging from 0 to 27 (Ferris & Wynne, 2001).Footnote 5 These composite scores were used to group individuals into one of three categories using the PGSI scoring recommendations of Williams and Volberg (2014), as these provide the best demarcation of problem gambling in the general population: 0 = non-problem gambling, 1–4 = at-risk gambling, 5 +  = problem gambling.Footnote 6 (Note that Statistics Canada data suppression rules prohibit the reporting of both PGSI 5 + and PGSI 8 + values as it would allow someone to subtract the PGSI 8 + rates from the PGSI 5 + rates to identify a group of five or fewer people scoring in the PGSI 6–7 range in some provinces).

Online Panel Survey

Sample

From August 16 to October 10, 2018 a gambling survey was administered to adult (18 +) online panelists from across Canada who were members of Leger Opinion (LEO). Leger Opinion is Canada’s largest online panel, with over 400,000 active members. Eligibility for the present survey was restricted to online panelists who completed a screening question indicating they gambled on one or more types of gambling once a month or more in the past year. Repeated email solicitations were sent out until an achieved sample of at least 10,000 with an equal number from each province or region was achieved (i.e., 1400 each from the provinces of British Columbia, Alberta, Saskatchewan, Manitoba, Ontario, Quebec, and 1400 from the four Atlantic provinces). Of the 10,199 completed surveys, a total of 589 were completed by individuals who self-identified as First Nations, Métis or Inuk (Inuit).

While ‘opt-in’ online panels such as Leger Opinion are demographically representative of the population, they sometimes have some important behavioural differences (Brüggen et al., 2016). In particular, they often contain more heavy gamblers and people with gambling problems than in the general population (Lee et al., 2015; Williams & Volberg, 2012). Thus, to better ensure representativeness, the gambling behaviour and demographic profile of the online panel sample was weighted to match the gambling behaviour and demographic profile for the same sub-sample in the CCHS 2018 Gambling Module (i.e., everyone 18 + who gambled once a month or more in the past year). Weights were created via iterative raking using the following six variables: number of types of gambling engaged in, PGSI category, provincial population size, gender, educational attainment, and age group. Weights were then winsorized to a maximum value of six times the average weight so as to minimize mean squared error.

Questionnaire

The main utility of the online panel survey was that it more comprehensively assessed mental health and substance use comorbidities as well as various aspects of gambling compared to the CCHS survey. The online panel survey took an average of 19.5 min to complete, ranging from 14 to 28 min. In addition to collecting demographic information, the questionnaire collected information in the following areas relevant to the present investigation:

Comorbidities

Past year use of tobacco, alcohol, cannabis, and non-medical use of other drugs; past year substance use disorder (assessed using DSM-5 criteria; American Psychiatric Association, 2013); lifetime and family history of substance abuse; past year and lifetime history of behavioural addictions; history of child abuse or neglect; number of significant negative past year life events (adaption of the Life Events Questionnaire; Vuchinich et al., 1986); past year post-traumatic stress, generalized anxiety, panic disorder, and major depression (all assessed using DSM-5 criteria with a composite variable of any past year DSM-5 mental disorder also being created); and lifetime and family history of mental health problems.

Impulsivity

Impulsivity was assessed with the impulsivity subdomain from the NEO Personality Inventory—Revised (NEO PI-R) (Costa & McCrae, 1992). Internal reliability of the NEO-PI-R domain scores are high, ranging from 0.86 to 0.92, and the internal reliabilities of the subdomains range from 0.58 to 0.82 (Costa & McCrae, 1992). The concurrent and discriminant validity of the NEO has been well established in both normal and clinical populations (Costa & McCrae, 1992).

Past Year Gambling Participation

Assessed with the full Gambling Participation Instrument (Williams et al., 2017) which assesses gambling participation in terms of types engaged in, means of access (remote or in-person), frequency, time, and expenditure. Depending on the domain, the test–retest reliability coefficients of this instrument are fair (0.46) to excellent (0.84), and the validity coefficients are good to excellent, ranging from 0.60 to 0.91.

Motivation for Gambling

Assessed with a single question having eight different response options, with these response options derived from extensive unpublished analysis of both open-ended and closed-ended responses obtained in several prior population surveys involving several thousand people. Multiple responses were permitted.

Context for Gambling

Five face valid questions concerning whether the person typically gambles alone or with friends/family, and the use of tobacco, alcohol, cannabis, and other drugs while gambling.

Gambling Social Exposure

Four face valid questions concerning being exposed to gambling prior to age 18, the prevalence of regular gambling and problem gambling among the person’s current social group, and the availability of gambling at the person’s place of work or school.

Gambling Fallacies

Assessed with the 10 item Gambling Fallacies Measure (GFM) (Leonard et al., 2015). Factor analysis across multiple datasets has found the GFM to consist of two factors: a failure to understand the random and uncontrollable nature of most gambling games and a failure to take statistical probabilities into account. The hierarchical coefficient omega shows adequate (0.61) internal consistency. The overall one-month test–retest reliability of the instrument is good (0.70). Depending on the dataset, GFM scores have been found to be consistently and significantly associated with intelligence, educational attainment, paranormal beliefs, and gambling ‘to win money’ as a primary motivation. Higher GFM scores denote greater resistance to fallacies.

Family History of Problem Gambling

A single question asking whether anyone in the person’s immediate family has ever had a gambling problem.

Past Year Problem Gambling

Two instruments were employed. The first was the Problem Gambling Severity Index (PGSI) (Ferris & Wynne, 2001). The second was the Problem and Pathological Gambling Measure (PPGM) (Williams & Volberg, 2014). The PGSI was included to match the PGSI categorization pattern in the online panel data to the CCHS PGSI categorization pattern. The PPGM was included as it is the best validated measure for assessing problem gambling in the general population (Williams & Volberg, 2014) and most comprehensively captures the multidimensional nature of this construct (Christensen et al., 2019).

Results

Gambling Participation

Table 1 presents the past year prevalence of gambling among Indigenous Canadian adults (18 +) in 2018 as compared to non-Indigenous Canadian adults as determined by the CCHS. Altogether, 75.3% of the Indigenous sample reported engaging in one or more types of gambling in 2018 compared to 63.9% for non-Indigenous Canadians. Significantly higher rates of Indigenous participation were observed for bingo, electronic gambling machines, and instant lotteries.

Gambling Categorizations

Table 2 presents the PGSI gambling categorizations among Indigenous Canadian adults (18 +) in 2018 as compared to non-Indigenous Canadians as determined by the CCHS. As seen, Indigenous Canadians have significantly higher rates of non-problem gambling and at-risk gambling. They also have higher rates of problem gambling, but these differences are not significant as the 95% confidence intervals overlap.

Gambling Motivations, Fallacies, Context, Social Exposure, Family History, and Comorbidities

Table 3 presents online panel descriptive statistics pertaining to motivations for gambling, level of gambling fallacies, context for gambling, social exposure to gambling, family history, and comorbidities among Indigenous and non-Indigenous regular gamblers. It excludes the 34.1% of Indigenous and 33.8% of non-Indigenous people who gambled less than once a month in the past year as well as the 24.7% of Indigenous and 33.8% of non-Indigenous people who did not gamble at all. Motivations for gambling appear to be more similar than dissimilar between the Indigenous and non-Indigenous samples, although gambling to socialize is more prominent among Indigenous gamblers. There are no meaningful differences in the level of gambling fallacies between the Indigenous and non-Indigenous sample. Substance use while gambling is more common in the Indigenous sample. Clear group differences occur with higher Indigenous social exposure to gambling and having someone in the immediate family with a gambling problem. Similarly, there are noticeably higher levels of DSM substance use disorders and mental disorders among the Indigenous sample. There are statistically significant differences between the groups on all variables due to the very large sample sizes and the population weights.

Table 3 Gambling motivations, fallacies, context, social exposure, family history and comorbidities among Canadian indigenous adult (18 +) Gamblers in 2018 compared to non-indigenous Canadian gamblers (online panel, weighted)
Full size table

Prediction of Problem versus Non-Problem Gambling within the Indigenous Sample

The online panel data was used for the multivariate prediction of problem versus non-problem gambling as it includes a much more comprehensive range of potential predictors than the CCHS. Because of the large number of potential predictors, variables were not included in the multivariate analysis unless they demonstrated a significant univariate association (p < 0.05) with problem gambling. This reduced the list of independent variables from 53 to 43. A forward stepwise binary logistic regression was then conducted to identify the variables that best differentiated the 130 Indigenous PPGM-classified Problem Gamblers from the 455 Indigenous PPGM Recreational and At-Risk Gamblers. The prevalence of PPGM identified problem gambling within the sample was used as the group classification cut point. Variable entry order was determined by the size of the Wald statistic, with a minimum entry level of p = 0.05 and a removal level of p = 0.01. The series mean was used to replace missing values for age group, educational attainment, and income group.

Model fit was maximized with a constant and nine predictors. A test of the full model with the nine predictors against a constant-only model was statistically significant, χ2 (9, n = 585) = 191.5, p < 0.001, indicating that the nine predictors alone could reliably distinguish between PPGM problem and non-problem gamblers. The variance accounted for was modest, with Nagelkerke R2 = 27.9%. The final model correctly classified 78.7% of Non-Problem Gamblers and 76.9% of Problem Gamblers.

Table 4 shows the regression coefficients, Wald statistics, and odds ratios for each of the significant predictors. What this table illustrates is that EGM participation is the most important predictor of problem gambling, with there being additive predictive power with more gambling fallacies; having a past year mental disorder; sports betting participation; past year substance use disorder; gambling to escape, relax or relieve stress; participation in ‘other’ types of gambling; non-medical use of drugs (other than alcohol, tobacco, cannabis); and male gender.

Table 4 Logistic regression of characteristics differentiating indigenous PPGM problem gamblers (n = 130) from Indigenous non-problem gamblers (n = 455) (online panel, unweighted)
Full size table

There are a few other observations that also potentially implicate EGMs as an important determinant of Indigenous problem gambling. As seen in Table 5 the Indigenous At-Risk + Problem Gambling rate in the fourth column varies considerably between provinces/regions and is strongly associated with the density of EGMs (slot machines or video lottery terminals)Footnote 7 in that province/region (r = 0.83, p = 0.020, n = 7) as well as the provincial/regional rates of Indigenous EGM participation in 2018 (r = 0.85, p = 0.016, n = 7). It is also the case that Ontario and British Columbia are the only two provinces that do not permit EGMs outside of dedicated gambling venues (i.e., casinos, horse racetracks, bingo halls), and these are the two provinces with the lowest rates of Indigenous At-Risk + Problem Gambling. Perhaps most concerning is the fact that EGM density has the strongest association with percentage of the provincial/regional population that is Indigenous (r = 0.95, p = 0.001, n = 7).

Table 5 Relationship between provincial/regional EGM density, indigenous EGM participation, indigenous at-risk + problem gambling rates, and indigenous population as a percentage of total population
Full size table

Prediction of Problem versus Non-Problem Gambling within the Non-Indigenous Sample

For comparison purposes, the same logistic regression procedure described above was used to determine the variables that best differentiated the 1216 Non-Indigenous Problem Gamblers from the 8253 Non-Indigenous Recreational and At-Risk Gamblers. Model fit was maximized with a constant and 21 predictors. A test of the full model with the 21 predictors against a constant-only model was statistically significant, χ2 (21, n = 9469) = 3886.72 p < 0.0001, indicating that the 21 predictors alone could reliably distinguish between problem and non-problem gamblers. The variance accounted for was quite good, with Nagelkerke R2 = 52.1%. The final model correctly classified 85.7% of Non-Problem Gamblers and 82.1% of Problem Gamblers.

Table 6 shows the regression coefficients, Wald statistics, and odds ratios for each of the significant predictors. What this table illustrates is that EGM participation is also the most important predictor of problem gambling for non-Indigenous problem gamblers, with there being additive predictive power with more gambling fallacies; impulsivity; family history of problem gambling; male gender; gambling to escape, relax or relieve stress; past year ‘other’ type of gambling participation; regular socialization with problem gamblers; past year sports betting; past year mental disorder; lifetime history of drug or alcohol problems; past year bingo participation; lifetime history of a behavioural addiction; past year tobacco or e-cigarette use; gambling with family when growing up; greater number of negative life events in the past 12 months; a lifetime history of mental health problems; being in a lower household income group; past year financial speculation participation; and younger age group.

Table 6 Logistic regression of characteristics differentiating non-indigenous PPGM problem gamblers (n = 1216) from non-indigenous non-problem gamblers (n = 8253) (online panel, unweighted)
Full size table

Discussion

Summary

One of the purposes of this study was to provide an updated profile of gambling and problem gambling among Indigenous Canadians. In general, the relative popularity and overall pattern of gambling participation among Indigenous Canadians in 2018 is fairly similar to non-Indigenous Canadians. Some important differences do exist, with higher levels of overall gambling, as well as greater participation in EGMs, bingo, and instant lotteries. These results are similar to what has been found in previous North American research (Belanger et al., 2017; Williams et al., 2011). Also consistent with prior research is the finding that the rate of problem gambling is much higher among Indigenous Canadians (2.0%) compared to non-Indigenous Canadians (0.5%). This rate of problem gambling is the highest of any racial/ethnic group in Canada in 2018 (Williams et al., 2021). That said, it is important to note that the overall level of problem gambling is lower in both the Indigenous and non-Indigenous samples compared to historical levels (Williams et al., 2020).

Motivations for gambling are also very similar between Indigenous and non-Indigenous gamblers, although gambling to socialize tends to be more common among Indigenous gamblers and gambling to compete or for the challenge tends to be more common among non-Indigenous. There are no meaningful differences between level of gambling fallacies between the two groups. Substance use while gambling does appear to be somewhat more common among Indigenous gamblers. However, the clearest group differences concern higher Indigenous social exposure to gambling, having someone in the immediate family with a gambling problem, and higher levels of DSM substance use and mental health disorders.

The other purpose of this study was to identify strategies to mitigate gambling-related harm among Indigenous Canadians. Multivariate analysis established that EGM participation is the most important predictor of Indigenous problem gambling, although there is additive predictive power with a higher level of gambling fallacies; presence of a mental disorder; sports betting participation; presence of a substance use disorder; gambling to escape, relax or relieve stress; participation in ‘other’ types of gambling; non-medical use of drugs (other than alcohol, tobacco, cannabis); and male gender. Multivariate analysis established that non-Indigenous problem gamblers had a very similar set of predictors, with seven of the nine Indigenous predictors being in the top 10 non-Indigenous predictors and with past year EGM participation and a higher level of gambling fallacies being the top two predictors in both analyses. Some differences did exist, with the main ones being impulsivity, family history of problem gambling, and socializing with problem gamblers being uniquely predictive for non-Indigenous problem gamblers and past year substance use disorder and past year non-medical use of drugs being uniquely predictive for Indigenous problem gamblers.

Conclusions

One of the most important findings is that the variables predictive of Indigenous problem gambling in this study are the same variables that have been identified as etiologically related to problem gambling in most populations (Dowling et al., 2017; Johansson et al., 2009; Sharman et al., 2019; Welte et al., 2017; Williams et al., 2012). For the most part, the elevated rate of Indigenous problem gambling appears to be due to the elevated prevalence of these common risk factors. To further underscore this point, multivariate analysis of the 2018 CCHS data established that Indigenous identify is not a risk factor for problem gambling when taking these risk factors into account (Williams et al., 2021). Thus, while culturally appropriate prevention and treatment initiatives likely facilitates better engagement and treatment retention, it is not clear that the nature of these interventions needs to be culturally specific.

The higher rate of Indigenous EGM participation (20.4% versus 12.8% for non-Indigenous) is a particularly important risk factor as EGM participation was the strongest predictor of Indigenous problem gambling in the present study. Worldwide, EGM participation is one of the most powerful predictors of problem gambling due to its continuous nature (Binde et al., 2017; MacLaren, 2016; Storer et al., 2009; Williams et al., 2012). In Canada, both historically (Cox et al., 2005) and as seen in the present study (see also Williams et al. 2021), EGM participation has also been documented to be among the most important predictors of problem gambling.

Indigenous EGM participation, in turn, is associated with provincial EGM availability/density. EGMs are typically placed where they have good potential for revenue generation, which usually results in their location in areas with higher concentrations of vulnerable populations (Raisamo, 2019; Rintoul et al., 2013; Wardle et al., 2014). The present findings show this to also be true in Canada, with Indigenous Canadians having both the highest rates of problem gambling of any racial/ethnic group (Williams et al., 2021) and residing disproportionately in the provinces with the highest density of EGMs. However, the higher EGM density in certain provinces is also because First Nations are providing a large number of EGMsFootnote 8 in these provinces in addition to provincial government provision (Marshall, 2019). In 2017/2018 First Nations provided 31.4% of the EGMs in Saskatchewan, 28.7% in Manitoba, 15.1% in Alberta, 12.2% in Ontario, 12.1% in the Atlantic provinces, 5.4% in Quebec, and 1.7% in British Columbia.Footnote 9 Having two largely independent providers has resulted in a higher overall number and density of EGMs in some provinces. (Note that First Nation VLT provision in MB and NB (not NS) is done in conjunction with the provincial government operator rather than independently and that First Nation casinos in most provinces are still provincially regulated).

The reality is that EGM revenue serves as an important source of funding for many First Nations bands. This illustrates a difficult tension between the need for economic vitality and the need to protect the health of the community. Provincial governments and First Nation bands need to be fully aware of this trade-off and, at a minimum, better coordinate provision to reduce the EGM saturation that currently exists, as EGM density in Manitoba, Saskatchewan and Alberta is very high by international standards (Ziolkowski, 2018). Another option to reduce EGM density is to restrict EGM provision to a single provider but have a revenue sharing agreement. This is the primary model in British Columbia where the provincial government provides almost all the EGMs, but 7% of all net gambling revenue goes to First Nations. Similar arrangements exist in other provinces (In Alberta the First Nations Development Fund receives revenue from the Alberta Lottery fund; In Ontario 1.7% of gross gambling revenue goes to First Nations; In Nova Scotia the revenue from Sydney Casino is shared with First Nations).

While reducing EGM density would have a positive public health impact, it also would not be a panacea considering that the highest rates of problem gambling have actually been found in the territory of Nunavut (which is 85.9% Indigenous), where the combined results of the CCHS 2015 and CCHS 2016 survey established that 4.5% of the Nunavut adult (18 +) population had gambling problems (PGSI 5 +)) (Statistics Canada, 2015, 2016a).Footnote 10 This is notable because Nunavut has no EGMs as well as considerably fewer legal gambling opportunities compared to the provinces (i.e., no casinos, horse race tracks, or legal online gambling except for the Canada-wide availability of online horse race betting (HorsePlayer Interactive)). It is also important to remember that people with gambling problems engage in many different types of gambling, all of which contribute to their problems. In the present study only 39.6% of Indigenous problem gamblers indicated that there was a particular type of gambling that contributed to their problems more than others (with EGMs being the most commonly identified problematic type followed by sports betting).

Reducing gambling-related fallacies is another important modifiable risk factor that merits attention. Gambling fallacies are robustly associated with problem gambling across different populations (Goodie & Fortune, 2013). That said, similar to EGM reduction, an exclusive focus on modifying gambling fallacies only produces modest treatment outcomes (e.g., Fortune & Goodie, 2012), as this strong association is due to problem gambling also causing gambling fallacies, and because gambling fallacies are only one of many different variables contributing to problem gambling (Leonard et al., 2021; Leonard & Williams, 2016; Nicholson et al., 2016).

Substance use/abuse and having a mental disorder are the final group of etiologically important modifiable risk factors. Here again, these variables are robustly associated with problem gambling in all populations (Dowling et al., 2017; Johansson et al., 2009; Sharman et al., 2019; Welte et al., 2017; Williams et al., 2012). The present study found these to be especially important as they are more prevalent among Indigenous gamblers and stronger predictors of Indigenous compared to non-Indigenous problem gambling (whereas impulsivity tends to be a stronger risk factor for non-Indigenous problem gambling). Substance abuse and mental health problems in turn, are linked to a range of disadvantageous social conditions as antecedents (e.g., poverty, unemployment, lower educational attainment, discrimination, adverse childhood events) (Hudson, 2005; Hughes et al., 2017; Kivimaki et al., 2020; Whitesell et al., 2012). Rectifying these antecedents would likely have the strongest and most enduring impact on reducing these comorbidities and the risk for problem gambling.

Limitations

There are four important limitations to this study. The first is that the identified predictors of problem gambling are cross-sectional rather than longitudinal, which weakens causal attributions. The second is that the provincial association between province/region and EGM density as well as Indigenous gambling and problem gambling rates is based on a very small number of data points. A finer-grained geographic analysis of these relationships is needed to confirm these results. The third is that that the CCHS sample does not include Indigenous adults living on First Nation reserves or residing in the three Canadian territories. A total of 37.6% of First Nations people live on reserves. With First Nations representing 58.4% of all Canadian Indigenous people this means that 21.9% of all Canadian Indigenous adults were excluded from the CCHS sampling approach plus the 3.6% of the Canadian Indigenous population living in the Canadian territories. On-reserve First Nations individuals tend to have somewhat poorer physical and mental health compared to off-reserve individuals (Carrière et al., 2018; Curtis, 2007), which means that the prevalence of problem gambling would likely have been even higher if this group had been included. The prevalence of problem gambling among Indigenous Canadians living in the territories is also expected to be higher than average based on what is known about Nunavut. The final limitation concerns the fact that the online panel sample excludes the 9% of Canadians who did not use the Internet in 2018 (Statistics Canada, 2019c). The impact of this is uncertain, as non-Internet users are more likely to be elderly and/or with lower educational attainment. While older Canadians tend to have lower rates of gambling-related problems, people with lower educational attainment tend to have higher rates of gambling-related problems (Williams et al., 2021).