Gambling and gambling problem among elite athletes and their professional coaches: findings from a Swedish total population survey of participants in four sports

ABSTRACT

This study assessed the following among elite athletes and their coaches in Sweden: (i) prevalence of gambling and ‘at risk for gambling problems’ (PGSI 3+); and (ii) relations between ‘at risk for gambling problems’ and attitudes towards gambling, experiences of gambling, and individual and demographic factors. A total of 1438 athletes and 401 coaches, in four sports, completed an online survey. Overall, 2% of female athletes and 13% of male athletes were classified as being ‘at risk for gambling problems’. Using an ordinal logistic regression, the results showed associations between ‘at risk for gambling problems’ and eight of the investigated variables: ‘talk about gambling during training’, ‘coaches positive attitude towards gambling’, ‘gambling companies encourage gambling’, ‘I have gambled on own game’, ‘someone I know has gambled on their own game’, ‘gambling is important in the family’, ‘someone in my acquaintance have/had a gambling problem’, ‘alcohol consumption’ among the athletes. Coaches of men’s teams had a higher prevalence (7%) than did coaches of women’s teams (3%). The findings suggest that the sports clubs should have greater knowledge about gambling problems as well as a communication strategy of their acquired knowledge to their athletes and coaches.

KEYWORDS:

• Gambling
• athletes
• sports gambling
• sport club policy
• sport sponsorship
• betting

Introduction

Gambling problems are considered a public health concern in many countries, and are conceptualized as lying along a risk continuum (Lancet, 2017; Shaffer & Korn, 2002). Worldwide, the prevalence of problem gambling is estimated to 0.5–8% in the adult population, with the Swedish prevalence being about 0.4% (0.7% for men and 0.2% for women). However, a commonly reported prevalence rate is 2%, a figure which is derived from merging groups with ‘moderate risk gambling’ and ‘problem gambling’ (Public Health Agency of Sweden, 2016b; Williams, Volberg, & Stevens, 2012). The latest population survey in Sweden showed that 58% of those aged 16–84 years had gambled at some point during the previous year (Public Health Agency of Sweden, 2016b). Although this represents over half of the population, the figures have decreased since the 2008–2009 survey, in which over 70% of respondents reported gambling at some point in the previous twelve months (Public Health Agency of Sweden, 2016b). Despite this decrease, the turnover on the Swedish gambling market increased during this period (Swedish Gambling Authority, 2017).

Overall, gambling problems are unevenly distributed in society, being more common among individuals of low socioeconomic status and those with easy access to gambling in their vicinity (Johansson, Grant, Kim, Odlaug, & Götestam, 2009; Public Health Agency of Sweden, 2013; Welte, Barnes, Wieczorek, & Tidwell, 2009).

One way to improve our understanding of gambling problems in society is to study groups in environments where gambling behaviours have increased. The occurrence of gambling in workplaces is relatively unexplored. However, some studies point to the importance of social networks for the occurrence of problem gambling (Reith & Dobbie, 2011). For example, having friends, family members, or co-workers who gamble is associated with problem gambling (Mazar, Williams, Stanek, Zorn, & Volberg, 2018; Public Health Agency of Sweden, 2013). Gambling industry employees are one group with a distinct presence of gambling in their workplaces. Hing and Gainsbury (2013) have previously shown that this group has a higher proportion of individuals with gambling problems than in society in general. They found five risk factors relating to problem gambling: workplace motivators, workplace triggers, the influence of colleagues, limited social opportunities, and familiarity with and interest in gambling. It remains unclear, however, which workplace factors are most influential. The study also identified two factors that are protective against problem gambling: experiencing other people’s losses, and having work colleagues who, through their discouraging actions, inhibit others from gambling. The importance of attitudes among colleagues, both as protective and risk factors, has also been highlighted in studies of alcohol use (Duke, Ames, Moore, & Cunradi, 2013; Hodgins, Williams, & Munro, 2009). Even so, the factors that have the greatest impact on the likelihood of gambling likely vary between professions.

Sport clubs as workplaces and their relation to the gambling industry

Sport clubs are not inherently related to gambling, but have come to be increasingly associated with sports betting. The most obvious association between elite sport and gambling is seen in gambling advertising connected with sport events. The advertising revenues from regulated gambling companies have become a significant source of funding for the sports sector (Lopez-Gonzalez & Griffiths, 2018; Lopez-Gonzalez & Tulloch, 2015; Milnes, 2018). In the UK, the dominant position of the gambling industry as sponsors, especially of football, has been critically discussed in recent years (Bunn et al., 2018; Lewanczik, 2019). Illustrating this relationship, nine of the 20 English Premier League clubs and 17 of the 24 Championship teams have the logos of betting companies on their shirtfronts (Davies, 2019). The pattern is the same in Sweden: for many years, various sports have been primarily financed by revenues from the state’s former gambling monopoly as well as sponsored directly by gambling companies (Norberg, 2015). The ‘gamblification’ of sports has been highlighted in this context, in which gambling, particularly sports betting, has been actively promoted as a natural part of sport (Lopez-Gonzalez & Griffiths, 2018). Part of this is the athletes’ marketing of gambling products. This has taken a new path in Sweden, where an athlete co-owns and markets a gambling company (Bethard, 2019).

Match-fixing is another issue that has attracted attention to gambling within sport. The Swedish Sports Confederation has prohibited betting on one’s own game, and bans are also in place in many other countries, although they are not always heeded (Huggins, 2018; Husting, Iglesias, Kern, & Buinickaite, 2015; Swedish Sports Confederation, 2015). Match-fixing has been called a threat to the sport, and this concern has put the question of athletes’ own betting and other gambling activities on the agenda (Huggins, 2018; McNamee, 2013). In a previous European study, 37% of the included athletes had bet on their own games (Grall-Bronnec et al., 2016). Betting on one’s own game was also associated with gambling problems. The association between gambling and match-fixing in sports has not been substantially explored in research, though it is generating interest from both the gambling industry and the sport environment.

Gambling and alcohol consumption among athletes

To date, the occurrence of gambling within sport environments has been little examined. The research focus has been on athletes, and little is known about the gambling habits and attitudes of sports employees such as coaches and sports managers. Prevalence rates of gambling problems have been measured, with a range of values in different populations. American college athletes have been in focus, with one American study finding prevalence rates of 4.3% among male and 0.4% among female college athletes (Huang, Jacobs, Derevensky, Gupta, & Paskus, 2007). Another study found no significant difference in problem gambling between college athletes and their college peers, with a prevalence rate of under 1% among the college athletes (Nelson et al., 2007). A pattern identified in other studies is that male college athletes gamble more and have a higher incidence of problem gambling than do women college athletes, paralleling the situation in the general population (Ellenbogen, Jacobs, Derevensky, Gupta, & Paskus, 2008; Huang et al., 2007; Nelson et al., 2007; Public Health Agency of Sweden, 2017; Wardle et al., 2011).

In Europe, two published articles and one report have shown problem gambling prevalence rates of 6–8% (Grall-Bronnec et al., 2016; Håkansson, Kenttä, & Åkesdotter, 2018; Wardle & Gibbons, 2014), though these studies used different measurements and study settings. The British report by Wardle and Gibbons (2014) included male elite cricket and football athletes. The 6% prevalence of problem gambling estimated using the Problem Gambling Severity Index (PGSI) and the cut-off of 8 points or more (out of 27 points) evoked concern within the Professional Players Federation in the UK. A Swedish study by Håkansson et al. (2018) explored problem gambling among national athletes in university studies; using a lifetime measurement, they estimated a prevalence of problem gambling of 14% in males and 1% in females. This could be considered an elevated prevalence of problem gambling compared with similar age groups in Swedish society in general (Public Health Agency of Sweden, 2016b).

Elite sport is a demanding occupation with insecure working conditions and high expectations of consistent performance. Gambling has not been a prime concern so far, but there is rising concern within the sports community about mental health problems among athletes, and strategies for approaching such problems have recently been formulated (Foskett & Longstaff, 2018; Moesch et al., 2018; Rice et al., 2016). These strategies are intended to break the silence around the subject and reduce the stigma of seeking help. Alcohol use and risky or excessive alcohol consumption have also been a recurring issue in sports, and various ways of coming to terms with this have emerged over time (Kingsland et al., 2013; O’Brien & Lyons, 2000). The association between alcohol and gambling is also of interest. In the latest population survey in Sweden, almost 6% of the male population (aged 16–84 years) displayed risky alcohol consumption and concurrent problem gambling (Public Health Agency of Sweden, 2016b). So far, only one study of college athletes has shown a connection between alcohol use and problem gambling (Huang et al., 2007), while Håkansson et al. (2018) found no such association.

If the prevalence of problem gambling in the sport environment is elevated relative to that in the general population, the question arises as to what factors in that environment contribute to the problem. In football, problem gambling has been associated with factors such as gambling as part of the social networks of young players, high salaries, spare time, gambling as a shared leisure pursuit, and the competitive and emotional challenges of the game (Lim et al., 2016). Wardle and Gibbons (2014) reported that 25% of respondents thought they were encouraged to gamble by teammates and 31% that they were encouraged by gambling companies. Moreover, social factors including the environment’s expectation of gambling participation and recurring discussions of gambling could contribute to higher prevalence (Hing, Russell, Vitartas, & Lamont, 2016; Lim et al., 2016; Lopez-Gonzalez, Guerrero-Sole, Estevez, & Griffiths, 2018). Exploring the extent to which personal characteristics or environmental factors (e.g. high salaries or a gambling-encouraging environment) predict problem gambling could provide information to be used in future work to prevent problem gambling.

The present study estimates the prevalence of gambling and ‘at risk for gambling problems’ and examines the association between ‘at risk for gambling problems’ and demographic factors as well as attitudes towards and experiences of gambling among elite athletes and coaches within the elite sport environment.

Methods

Design and participants

In this cross-sectional study, members of all Swedish teams competing in the highest divisions of football, ice hockey, floorball, and basketball were invited to participate in an online survey, meaning that the total population was targeted. These sports were selected for three reasons: they are team sports; they are popular sports with different financial conditions; and their team members represent a range of sociodemographic groups in Sweden. In the two sports with more than three divisions each (i.e. men’s football and ice hockey), all teams in the third highest division were also invited to participate, to explore whether full-time employment in the sport was a risk or protective factor for gambling problem. Overall, the survey reached 184 sport clubs, and an estimated 3717 athletes and 813 coaches were invited to participate. Overall, the response rates were 39% for the athletes and 49% for their coaches. Group-specific response rates differed greatly, with the highest response rate for athletes being in men’s basketball (93%) and the lowest in men’s floorball (25%). Sociodemographic characteristics of the participating athletes and coaches are presented in Table 1.

Table 1. Sociodemographic characteristics of the athletes and their coaches.

	Athletes, N = 1438		Coaches, N = 401
Characteristics	n	%	n	%
Sex
Males	934	64.9	317	79.1
Females	502	34.9	79	19.7
Don’t know	1	.1	0	.0
Missing	1	.1	5	1.2
Age, years
<18	89	6.2	-	-
18–24	826	57.4	65	16.2
25–29	351	24.4	42	10.5
30–35	115	8.0	56	14.0
>35	53	3.7	235	58.6
Missing	4	.3	3	.7
Brought up outside Sweden
No	1227	85.3	378	94.3
Yes	162	11.3	15	3.7
Missing	49	3.4	8	2.0
Employment within the sport
Full-time employment	334	23.2	124	30.9
Half-time employment	507	35.3	86	21.5
Other arrangements	296	20.6	80	20.0
No income from the sport	235	16.3	100	24.9
Missing	66	4.6	11	2.7
Living conditions
Living alone	405	28.2	87	21.7
Living with partner	436	30.3	102	25.4
Living with partner and children	145	10.1	164	40.9
Living with parents/siblings	335	23.3	26	6.5
Living with other person/persons	68	4.7	13	3.2
Missing	49	3.4	9	2.3
Gambling is important among my current family members
1 – Not correct at all	1086	75.5	329	82.0
2	124	8.6	31	7.7
3	62	4.3	10	2.5
4	12	0.8	3	0.8
5 – Completely correct	8	0.6	5	1.3
6 – Don’t know	19	1.3	2	.5
Missing	127	8.9	21	5.2

The coaches could be female or male, regardless of whether their teams were for women or men. In analysing the coaches, we therefore included a variable that indicated whether the coaches were active in women’s or men’s teams.

Measures

As part of the study, two similar web-based questionnaires – one for the athletes and one for their coaches – were designed in collaboration with the Public Health Agency of Sweden. The questionnaires covered information about the study, such as the principles of confidentiality and voluntary participation, as well as 43 questions for the athletes and 40 for the coaches. The surveys were designed using the web-based programme Easyresearch, which generated a web link to the questionnaires to be completed online by the participants (Questback, 2019). All items included in the questionnaires are presented in Tables 2 and 3, including the numbers of missing responses and non-analysable responses (i.e. ‘Don’t know’ and ‘Someone else’) for each item.

Table 2. Item missing values for all variables except sociodemographic characteristics among the athletes (N = 1438).

Item	Valid responses	Non-analysable responses	Missing values
I have not been able to exercise because of injury/illness	1350	-	88
I have limited socialization with people outside the sport	1343	-	95
Gambling is important to my teammates	1261	44	133
We often talk about gambling during training	1261	33	144
Our coaches have a positive attitude towards gambling	806	486	146
Permissive/non-permissive attitudes towards gambling in my sports club	778	513	147
Gambling companies encourage gambling among athletes	1279	-	159
I believe that my own games should be subjected to gambling by gambling companies	1266	-	172
I have gambled on my own game during the last 12 months	1278	-	160
Someone I know has gambled on their own game during the last 12 months	1221	-	217
Thinking over the last 12 months, have you bet more than you could really afford to lose? (PGSI, item 1)	1246	-	192
Thinking over the last 12 months, have you needed to gamble with larger amounts of money to get the same feeling of excitement? (PGSI, item 2)	1231	-	207
Thinking over the last 12 months, have you gone back another day to try to win back the money you lost? (PGSI, item 3)	1232	-	206
Thinking over the last 12 months, have you borrowed money or sold anything to gamble? (PGSI, item 4)	1234	-	204
Thinking over the last 12 months, have you felt that you might have a problem with gambling? (PGSI, item 5)	1232	-	206
Thinking over the last 12 months, have people criticized your betting or told you that you had a gambling problem, whether or not you thought it was true? (PGSI, item 6)	1231	-	207
Thinking over the last 12 months, have you felt guilty about the way you gamble or what happens when you gamble? (PGSI, item 7)	1232	-	206
Thinking over the last 12 months, has gambling caused you any health problems, including stress or anxiety? (PGSI, item 8)	1232	-	206
Thinking over the last 12 months, has your gambling caused any financial problems for you or your household? (PGSI, item 9)	1229	-	209
Someone in my acquaintance has or has had a gambling problem	1244	-	194
If so, someone in my sports club	1234	-	204
If so, someone in another sports club	1234	-	204
If so, someone in my acquaintance, outside the sport	1238	-	200
If so, someone in my acquaintance, in the sport	1236	-	202
Other	1205	30	203
How often do you have a drink containing alcohol? (AUDIT-C, item 1)	1267	-	171
How many drinks containing alcohol do you have on a typical day when you are drinking? (AUDIT-C, item 2)	1259	-	179
How often do you have six or more drinks on one occasion? (AUDIT-C, item 3)	1252	-	186

Table 3. Item missing values for all variables except sociodemographic characteristics among the coaches (N = 401).

Item	Valid responses	Non-analysable responses	Missing values
Active in women’s sports	390	-	11
I have limited socialization with people outside the sport	385	-	16
Gambling is important to my colleagues	362	15	24
We often talk about gambling during training	365	13	23
Permissive/non-permissive attitude towards gambling	259	121	21
Gambling companies encourage gambling among athletes	377	-	24
I believe that my own games should be subjected to gambling by gambling companies	372	-	29
I have gambled on my own game during the last 12 months	375	-	26
Someone I know has gambled on their own game during the last 12 months	376	-	25
Thinking over the last 12 months, have you bet more than you could really afford to lose? (PGSI, item 1)	370	-	31
Thinking over the last 12 months, have you needed to gamble with larger amounts of money to get the same feeling of excitement? (PGSI, item 2)	370	-	31
Thinking over the last 12 months, have you gone back another day to try to win back the money you lost? (PGSI, item 3)	371	-	30
Thinking over the last 12 months, have you borrowed money or sold anything to gamble? (PGSI, item 4)	368	-	33
Thinking over the last 12 months, have you felt that you might have a problem with gambling? (PGSI, item 5)	369	-	32
Thinking over the last 12 months, have people criticized your betting or told you that you had a gambling problem, whether or not you thought it was true? (PGSI, item 6)	371	-	30
Thinking over the last 12 months, have you felt guilty about the way you gamble or what happens when you gamble? (PGSI, item 7)	371	-	30
Thinking over the last 12 months, has gambling caused you any health problems, including stress or anxiety? (PGSI, item 8)	368	-	33
Thinking over the last 12 months, has your gambling caused any financial problems for you or your household? (PGSI, item 9)	371	-	30
Someone in my acquaintance has or has had a gambling problem	371	-	30
f so, someone in my sports club	371	-	30
If so, someone in another sports club	371	-	30
If so, someone in my acquaintance, outside the sport	371	-	30
If so, someone in my acquaintance, in the sport	371	-	30
Other	366	5	30
How often do you have a drink containing alcohol? (AUDIT-C, item 1)	374	-	27
How many drinks containing alcohol do you have on a typical day when you are drinking? (AUDIT-C, item 2)	374	-	27
How often do you have six or more drinks on one occasion? (AUDIT-C, item 3)	370	-	31

Two scales can be constructed from nine and three items, respectively, from the questionnaires: the Problem Gambling Severity Index (PGSI) (Ferris & Wynne, 2001) and the Alcohol Use Disorders Identification Test – Consumption (AUDIT-C) (Bush, Kivlahan, McDonell, Fihn, & Bradley, 1998).

The PGSI measures the severity of gambling problems during the past 12 months. It comprises nine items, each of which is coded 0–3, with the total score ranging between 0 and 27 points. Previous research has established that the PGSI has adequate psychometric properties (Miller, Currie, Hodgins, & Casey, 2013). Our questionnaire also included the question ‘Have you gambled during the last 12 months?’ as a filter question; for respondents who answered no, a zero was imputed for each of the nine items in the PGSI scale. The PGSI scale is divided into five levels: Non-gambling, 0 = Non-problem gambling, 1–2 = Low risk gambling, 3–7 = Moderate risk gambling, and 8 and higher = Problem gambling (Ferris & Wynne, 2001). In order to compare our results to the Swedish longitudinal population study Swelogs (Public Health Agency of Sweden, 2016a), we used the same cut-off in the present study. The categories are ’non-gambling’, ‘non-problem gambling’ (score 0), ‘low risk gambling’ (score 1–2), ’at risk for gambling problems’ (score 3+). In the analyses of the associations between ‘at risk for gambling problems’ and criterion variables, three levels of PGSI constitutes the outcome variable e.g. ‘non-problem gambling’ (including non-gambling), ‘low risk gambling’, ‘at risk for gambling problems’. The remaining items of the questionnaire are included in the analyses as independent variables. In the analyses of the associations between ‘at risk for gambling problems’ and criterion variables, the independent variables are grouped into two blocks: the first, ‘Experiences of and attitudes to gambling in sports’, comprises variables capturing attitudes towards gambling within the sport club; the second, ‘Individual and sociodemographic factors’, comprises variables such as living conditions, social life, employment in the sport, and alcohol consumption.

The three-item AUDIT-C measures risky alcohol consumption. The items are coded 0–4, with the total score ranging from 0 to 12 points (Bush et al., 1998). The AUDIT-C has displayed good validity and reliability in previous research (Campbell & Maisto, 2018).

Data collection

The data were collected between November 2016 and March 2017. The first author and the project assistant contacted each club via email and/or phone. After initial contact, the project assistant sent the web link to the sports manager or chairperson of each selected club, who then forwarded the link to the athletes and coaches via email or through a Facebook group for team members. Each club received three participation reminders from the first author or the research assistant. The athletes and coaches could choose the time and mode (e.g. via computer or mobile phone) of completing the questionnaire, which was estimated to require about 20 minutes, and the web link remained open to the participants for one month.

Ethical considerations

All participants were provided with written information about study participation. Given the importance of maintaining confidentiality, completing the questionnaire was considered as providing consent. Ethical permission was granted by the Regional Ethical Review Board in Stockholm (2016/2049-31).

Data analysis

Due to the high number of missing data for some items, multiple imputation (20 imputations) was performed using Blimp software before any analyses were conducted (Enders, Keller, & Levy, 2018). A great advantage of this software is that values are imputed based on their level of measurement (i.e. continuous, ordinal, or nominal). The PGSI scores were first categorized into four levels (i.e. non-gambling, non-problem gambling, low risk gambling, and ‘at risk for gambling problems) and cross-tabulated against sports, stratified by sex for the athletes and their coaches. The proportions of participants experiencing ‘at risk for gambling problems’ in all sports taken together for each of the four groups (i.e. male athletes, female athletes, male coaches, and female coaches) were then compared with the proportions for their counterpart sex and age groups in a stratified random sample of the Swedish general population using a two-sample proportion test. In the analyses of the association with criterion variables, our outcome was the PGSI score divided into three categories (i.e. non-problem gambling (including non-gambling), low risk gambling, and ‘at risk for gambling problems’) that constitute an ordinal scaled variable coded 0–2, consequently analysed using ordinal logistic regression (Liu, 2016). The proportional odds assumption – that is, the assumption that the odds ratio for being in category 1 or 2 versus category 0 is the same as the odds ratio for being in category 2 versus category 1 or lower – was tested using the Brant test (Brant, 1990). Moreover, we also included the independent variables in two blocks, ‘Experiences of and attitudes to gambling in sports’ and ‘Individual and sociodemographic factors’, to investigate the degree to which these two blocks explained the variance of ‘at risk for gambling problems’. The proportion of variance explained by the two blocks of variables was tested using the Wald chi-squared test.

Results

Prevalence of gambling and ‘at risk for gambling problems’ among athletes and coaches

The results indicated that 46% of the 490 female athletes and 75% of the 903 male athletes had gambled during the previous 12 months (Table 4). ‘at risk for gambling problems’ was low among female athletes (under 2%) and higher among male athletes (13%). The highest prevalence of ‘at risk for gambling problems’ was found in men’s floorball (15%) and the lowest prevalence in women’s football (less than 1%). In men’s football and ice hockey, we also compared the prevalence of athletes’ ‘at risk for gambling problems’ between two sport divisions. In football, the prevalence was 14% in the highest division (Allsvenskan; 148 athletes) and 13% in the third highest division (Division 1; 214 athletes). In ice hockey, the prevalence was approximately 13% in both the highest division (SHL; 143 athletes) and the third highest division (Hockeyettan; 222 athletes). In the comparison between male athletes and a stratified random sample of males in the Swedish general population in approximately the same age group (16–39 years), the athletes had a higher prevalence of ‘at risk for gambling problems’ (13% vs. 4%). The difference between the two groups was highly significant: z = – 8.585, P (|Z| > |z|) < 0.001. Among females, the difference in ‘at risk for gambling problems’ between the athletes and the previously mentioned community sample in Sweden was much smaller (1.7 vs. 1.1%) and not statistically significant: z = – 1.058, P (|Z| > |z|) = 0.290.

Table 4. Pooled proportions of gambling habits among Swedish athletes in four sports compared with a stratified random sample of the Swedish population in the age group of 16–39 years.

Gambling habits (PGSI)	Football %	Ice hockey %	Floorball %	Basketball %	Total %	Swedish population* %
Males, n	362	365	90	86	903	1895
Non-gambling	25.2	21.8	18.0	42.1	24.7	42.0
Non-problem gambling	50.7	52.6	52.7	39.7	50.6	45.2
Low risk gambling	11.0	12.2	14.7	8.0	11.6	8.6
At risk for gambling problems	13.1	12.2	14.6	10.2	13.1	4.2
Females, n	179	108	138	65	490	1678
Non-gambling	58.4	46.5	49.9	63.1	54.0	52.7
Non problem gambling	39.4	49.5	45.6	30.8	42.2	43.7
Low risk gambling	1.3	0.4	3.0	4.8	2.1	2.5
At risk for gambling problems	0.9	3.6	1.5	1.3	1.7	1.1

*Personal communication with the Public Health Agency of Sweden.

Of the 315 male coaches, 76% had gambled during the previous 12 months and 6% displayed ‘at risk for gambling problems’ (Table 5). In comparison with a stratified random sample of the Swedish male population in approximately the same age group (20–49 years), the male coaches had a slightly higher prevalence of ‘at risk for gambling problems’ (6% vs. 4%). The difference between the two groups was statistically significant: z = – 2.059, P (|Z| > |z|) = 0.040. The prevalence of ‘at risk for gambling problems’ was 2% among female coaches, versus 1% in the previously mentioned sample of the Swedish general population. However, the difference was too small to be statistically significant: z = – 0.859, P (|Z| > |z|) = 0.390. When we compared coaches who were active in teams of the opposite sex with coaches who were active in teams of the same sex, we found differences for both female and male coaches. Among the 65 female coaches active in women’s teams, the proportions of gambling during the previous 12 months and of ‘at risk for gambling problems’ were 52% and 1%, respectively, while the proportions were 70% and 8%, respectively, among the 12 female coaches active in men’s teams. Among the 221 male coaches active in men’s teams, the proportions of gambling during the previous 12 months and of ‘at risk for gambling problems’ were 75% and 7%, respectively, while the proportions were 78% and 5%, respectively, among the 94 male coaches active in women’s teams. The differences in ‘at risk for gambling problems’ between coaches active in teams of the same sex versus the opposite sex were not statistically significant for either female or male coaches: z = – 1.298, P (|Z| > |z|) = 0.194 and z = 0.723, P (|Z| > |z|) = 0.470, respectively.

Table 5. Pooled proportions of gambling habits among Swedish coaches in four sports compared with a stratified random sample of the Swedish population in the age group of 20–49 years.

Gambling habits (PGSI)	Football %	Ice hockey %	Floorball %	Basketball %	Total %	Swedish population* %
Males, n	111	113	65	26	315	2404
Non-gambling	23.5	22.0	16.8	49.0	23.7	36.1
Non- problem gambling	63.6	65.3	65.5	50.0	63.5	52.8
Low risk gambling	5.6	7.4	8.0	0.2	6.3	7.1
At risk for gambling problems	7.3	5.3	9.7	0.8	6.5	4.0
Females, n	12	44	17	4	77	2255
Non-gambling	52.1	45.3	25.9	100.0	45.0	46.5
Non- problem gambling	47.1	48.0	74.1	0.0	51.1	49.6
Low risk gambling	0.0	2.8	0.0	0.0	1.6	2.7
At risk for gambling problems	0.8	3.9	0.0	0.0	2.3	1.2

*Personal communication with the Public Health Agency of Sweden.

Relations between ‘at risk for gambling problems’ and demographic factors, attitudes, and experiences of gambling

Results among athletes

In the ordinal logistic regression, we first included all independent variables from the first block, ‘Experiences of and attitudes to gambling in sports’ (Model I, Table 6). In the next step of the analysis, we added all independent variables from the second block, ‘Individual and sociodemographic factors’ (Model II, Table 6). The estimated variables from the first block differed slightly between Models II and I, since those in Model II were adjusted for the variables in the second block. The Wald test of explained variance including all variables from the first block gave Chi²₍₁₀₎ = 45.96, P > F < 0.001, and adding all variables from the second block gave Chi²₍₁₀₎ = 27.57, P > F = 0.002. The Brant test of the proportional odds assumption gave Chi²₍₁₀₎ = 5.63, P > Chi² = 0.845 for the first block and Chi²₍₁₀₎ = 5.60, P > Chi² = 0.848 for the second block, indicating that the final Model II did not violate the proportionality assumption.

Table 6. Associations among the athletes (N = 1393) between problem gambling and 18 predictor variables divided into two blocks; results of two proportional odds models: single-block model and full model.

Outcome variable
Problem gambling according to the PGSI^1 divided into non-problem gambling, low risk gambling problem, at risk for gambling problems
	Model I		Model II
Predictor variables	Coefficient	Standard error	Coefficient	Standard error	OR
Gambling is important to my teammates	0.243**	0.093	0.145	0.100	1.16
We often talk about gambling during training	0.532***	0.105	0.502***	0.108	1.65
Our coaches have a positive attitude towards gambling	–0.195*	0.089	–0.220*	0.099	0.80
My sports club has the following attitude towards gambling
Non-permissive attitude (ref)
No attitude (no talk about gambling in the club)	0.164	0.211	0.137	0.220	1.15
Permissive attitude	0.318	0.372	0.343	0.378	1.41
Gambling companies encourage gambling among athletes	0.579**	0.212	0.457*	0.215	1.58
I believe that my own games should be subjected to gambling by gambling companies	–0.108	0.198	–0.120	0.209	0.89
I have gambled on my own game during the last 12 months	1.355**	0.395	1.209**	0.397	3.35
Someone I know has gambled on their own game during the last 12 months	0.539*	0.198	0.514*	0.205	1.67
Someone in my acquaintance (inside my sport) has or has had a gambling problem	0.336	0.233	0.176	0.236	1.19
Age in years			–0.106	0.138	0.90
Brought up outside Sweden	-	-	–0.265	0.536	0.77
I live alone	-	-	–0.155	0.219	0.86
I am being paid by the sports club
Nothing (ref)
Part-time	-	-	0.064	0.294	1.07
Full-time	-	-	0.421	0.335	1.52
I have not been able to exercise because of injury/illness	-	-	–0.207	0.213	0.81
I have limited socialization with people outside the sport	-	-	–0.238	0.190	0.79
Gambling is important among my current family members	-	-	0.506***	0.131	1.66
Someone in my acquaintance (outside my sport) has or has had a gambling problem	-	-	0.595**	0.197	1.81
Alcohol consumption according to the AUDIT-C^2	-	-	0.140**	0.053	1.15

***p < .001

**p < .01

*p < .05

^aPGSI = Problem Gambling Severity Index

^bAUDIT-C = Alcohol Use Disorders Identification Test – Consumption.

Five of the variables in the first block were significantly associated with ‘at risk for gambling problems’ in the final Model II: ‘We often talk about gambling during training’ (OR = 1.65, t = 4.63, P > |t| < 0.001); ‘Our coaches have a positive attitude towards gambling’ (OR = 0.80, t = – 2.23, P > |t| = 0.027); ‘Gambling companies encourage gambling among athletes’ (OR = 1.58, t = 2.12, P > |t| = 0.036); ‘I have bet on my own game during the last 12 months’ (OR = 3.35, t = 3.05, P > |t| = 0.003); and ‘Someone I know has bet on their own game during the last 12 months’ (OR = 1.67, t = 2.50, P > |t| = 0.013). After adjusting for variables in the second block, one of the variables that was statistically significant in Model I (‘Gambling is important to my teammates’) ceased to be statistically significant in Model II.

Three of the variables in the second block were statistically related to ‘at risk for gambling problems’: ‘Gambling is important among my current family members’ (OR = 1.66, t = 3.86, P > |t| < 0.001); ‘Someone in my acquaintance (outside my sport) has or has had a gambling problem’ (OR = 1.81, t = 3.03, P > |t| = 0.003); and ‘Alcohol consumption’ (OR = 1.15, t = 2.63, P > |t| = 0.010).

Results among coaches

The analysis for the coaches was performed in the same way as for the athletes. The Wald test of explained variance including all variables from the first block, ‘Experiences of and attitudes to gambling in sports’, gave Chi²₍₈₎ = 28.93, P > F < 0.001, and adding all variables in the second block, ‘Individual and sociodemographic factors’, gave Chi²₍₉₎ = 11.40, P > F = 0.2493. The Brant test of the proportional odds assumption gave Chi²₍₁₅₎ = 72.78, P > Chi² < 0.001, indicating that Model II violated the proportionality assumption. We therefore performed the Brant test on Model I as well, which gave Chi²₍₇₎ = 18.19, P > Chi² = 0.011, indicating that Model I also violated the proportional odds assumption. The second block of variables did not contribute to explaining the variance in gambling problems among the coaches, reducing the number of variables included in the final model (i.e. only Model I). However, since the proportionality assumption was violated, making the presentation in tables more tedious, and only one statistically significant result emerged from Model I, we present this significant result for the coaches in text form only.

The only variable of statistical significance in the final model (i.e. Model I) was ‘We often talk about gambling during training’, which was associated with being in the two higher categories of the scale versus being in the lowest category: OR = 1.82, t = 2.95, P > |t| = 0.003. However, this variable was not statistically significant in the comparison of the odds of being in the highest versus the two lower categories combined.

Discussion

This study explored the prevalence of ‘at risk for gambling problems’, gambling experiences, and attitudes to gambling within the sport environment. Both gambling and ‘at risk for gambling problems’ were more common in the sample than in the equivalent age groups in the Swedish general population (Public Health Agency of Sweden, 2016b). There was a marked difference, however, in the prevalence of ‘at risk for gambling problems’ between the male athletes (13%) and female athletes (2%), confirming the results of Håkansson et al. (2018). The prevalence of ‘at risk for gambling problems’ was three times higher in male athletes than in males in the same age group in the Swedish general population (Public Health Agency of Sweden, 2016b). Merging ’moderate risk gamblers’ and ‘problem gamblers’ in the British report by Wardle and Gibbons (2014), and thereby using the same cut-off as used in the present study (PGSI 3+), the prevalence among the male British football and cricket athletes would have been 20%. This is higher than the proportion found in the present study and is an interesting result, as the prevalence of problem gambling in the British population is 2.8% (Conolly, Davies, Fuller, Heinze, & Wardle, 2018).

The prevalence of ‘at risk for gambling problems’ among female athletes (2%) presents another picture. Among the female athletes, only ice hockey players had a distinctly higher prevalence of ‘risk for problem gambling’ (3.6%) than that of females in the Swedish general population. It is also interesting to keep in mind that the female coaches in women’s ice hockey had the highest prevalence (3.9%) of all studied female groups. This could indicate an accepting gambling culture slightly more similar to that of the male athletes and coaches. One explanation as to why gambling and gambling problems within the sport environment seem to be common among male athletes is the high salaries in some sports (Lim et al., 2016). This could be a factor contributing to the difference in prevalence between female and male athletes since there is a big difference in salaries between female and male athletes. Relatively few female elite football players or ice hockey players live solely on their pay from their sport. On the other hand, floorball and the third divisions in men’s football and hockey do not have substantially higher salaries than those of female athletes in the highest division in ice hockey and football, but the prevalence of ‘risk for problem gambling’ in all these groups was similar, i.e. about 13%. High salaries are probably not the most prominent explanation for ‘at risk for gambling problems’, but can contribute to more frequent gambling.

There were no associations between ‘at risk for gambling problems’ and employment aspects (i.e. full-time paid employment or no payment at all for the sport). However, being paid by the sport club is linked to the amount of leisure time the athletes have. In floorball and basketball, most athletes have other occupations besides the sports, which reduces their leisure time. The similarity of ‘at risk for gambling problems’ prevalence between (male) sports and divisions does not seem to support the argument that extensive leisure time is a prominent explanation for the prevalence figures (Lim et al., 2016).

One reason for the discrepancy in ‘at risk for gambling problems’ between males and females might be found in the way they are depicted in marketing strategies. The lack of women’s representation in gambling marketing and active strategies to align sports betting with masculinity might not encourage female athletes to associate gambling with a successful sporting life (Deans, Thomas, Daube, Derevensky, & Gordon, 2016; Hunt & Gonsalkorale, 2018). Further research is needed to understand how gambling is contextualized in a gender-segregated environment like that of sports.

In the analysis of factors associated with ‘at risk for gambling problems’, we found that five variables in the block ‘Experiences of and attitudes to gambling within sports’ and three in the block ‘Individual and sociodemographic factors outside of the sport’ were statistically significantly associated with ’at risk for gambling problems’ among the athletes. Most of these factors are found within the sport environment, and reflect an environment where athletes regularly talk about gambling and where gambling is important among teammates. However, the variables ‘Gambling is important among my current family members’ and ‘Someone in my acquaintance (outside my sport) has or has had a gambling problem’ were also associated with ‘at risk for gambling problems’. It is conceivable that these risk factors taken together are a manifestation of an environment in which both family and other relatives are interested in sports, in turn creating an interest in betting, although not necessarily gambling in general. The ‘gamblification’ of sport has meant that betting is now seen as part of sporting as an interest, giving rise to, on one hand, conversation topics and a sense of connection, and, on the other hand, the risk that more people will engage in gambling (Hing et al., 2016; Lim et al., 2016; Lopez-Gonzalez & Griffiths, 2018; Lopez-Gonzalez et al., 2018). Our results suggest that recurrent talk about gambling or betting is in itself associated with ‘at risk for gambling problems’, and may stem from a culture in which betting is normalized through constant marketing in the sport context (Deans et al., 2016; Lopez-Gonzalez, Estévez, & Griffiths, 2019). On the other hand, research has also shown that the network of family and colleagues can be a protective factor (Dowling et al., 2017; Hing & Gainsbury, 2013). This may be the case in women’s sports, where a smaller proportion of participants reported talking about gambling with their teammates. This could be a protective factor against starting to gamble when a young female athlete joins a team; conversely, a young male athlete joining a team in which gambling is part of the daily conversation is given a social incentive to gamble himself.

Both elite athletes and coaches work in an environment where it is common to watch sport competitions in the media, meaning that they are also regularly exposed to the marketing of gambling. Well-known sports profiles used in marketing gambling help normalize betting, probably not only among the sport’s audience but also among its athletes (Deans et al., 2016; Lamont, Hing, & Vitartas, 2016). In addition to the marketing of gambling, there is also the sponsorship from the gambling industry. Among the studied athletes, the attitude that gambling companies encourage gambling among athletes was associated with gambling problems. This encouragement might be the result of gambling industry sponsorship of athletes, which involves invitations to events such as horse racing and events at casinos (Lim et al., 2016). It is not possible to comment on what the encouragement consisted of among the studied athletes, but athletes in other countries also reportedly experience some kind of encouragement to gamble (Lim et al., 2016; Wardle & Gibbons, 2014). Considering the sponsorship from gambling companies, this encouragement and associated dependence could result in the absence of constructive criticism of the relationship between the gambling industry and sport clubs (Statista, 2018; Swedish Ice Hockey Association, 2019).

If we assume that gambling is normalized and sometimes encouraged by the gambling companies, this may also contribute to betting on one’s own game, even though this is prohibited. Betting on one’s own game has previously been found to be a risk factor for problem gambling (Grall-Bronnec et al., 2016). Our results are in line with this, and in addition reveal that knowing someone who has bet on their own game is another factor associated with ‘at risk for gambling problems’. It could be asked whether betting on one’s own game is a logical reaction to the ‘gamblification’ of the sport environment. One reason given by athletes for betting on one’s own game despite the rules is the feeling that ‘our team will win’, so refraining from betting on a ‘certain win’ would be a departure from one’s ‘normal’ gambling habits. As long as athletes bet on a ‘winning situation’, the risk of being condemned for unacceptable behaviour by peers might be unlikely. However, in terms of the integrity of the game, this association is problematic and well worth some future attention.

While the gambling industry is a substantial sponsor of and advertiser at sport events in Europe, so is the alcohol industry (Institute of Alcohol Studies, 2017; Mongan, 2016). Given the relation between risky alcohol consumption and problem gambling, the high profile of the alcohol industry highlights the need to monitor alcohol use in the sport environment (Lorains, Cowlishaw, & Thomas, 2011; Welte et al., 2009). In contrast to Håkansson et al. (2018), we found a significant association between athletes’ alcohol consumption and ‘at risk for gambling problems’, supporting earlier findings from American studies (Huang et al., 2007). The association between gambling and risky alcohol consumption among young men is well documented (Lorains et al., 2011; Public Health Agency of Sweden, 2013). Nevertheless, it is not obvious in the context studied here, as alcohol is strictly monitored during the athletes’ competition season, so it would be of interest to further explore how alcohol and gambling interact in the sporting environment.

‘At risk for gambling problems’ among the male coaches seemed to be slightly elevated (6%) compared with the level in the general population, although it was substantially lower than among the male athletes. The female coaches reported a low prevalence of ‘at risk for gambling problems’, though the difference between male and female coaches in the prevalence of problem gambling was less than the difference between the male and female athletes. We do not know whether this difference in prevalence between athletes and coaches is an effect of age or other factors. One factor that may have contributed is the preferred gambling form. Accessibility has increased significantly since mobile phones became a platform for gambling. It has also been shown that the younger generation engages in mobile-phone gambling more often than does the older generation (Barnfield-Tubb & Harris, 2018; Edgren, Castrén, Alho, & Salonen, 2017). The younger generation in this study, the athletes, started gambling at a time when gambling access was on-line, which might have created other habits from the beginning.

Our results also indicate the importance of the workplace for gambling, though only two workplace factors were associated with ‘at risk for gambling problems’: ‘Gambling is important to my teammates’ and ‘We often talk about gambling during training’. Being a coach with an interest in betting and gambling might not be controversial, as most athletes and coaches reported an absence of gambling policy or rules regarding gambling. Also, earlier studies have shown that sports betting seems to have few or no stigmatizing connotations compared with other forms of gambling, which could indicate that there are no inherent inhibitory factors (Hing, Nuske, Gainsbury, & Russell, 2015; Lopez-Gonzalez et al., 2019). More research is needed to find out how coaches and others in leading positions within sport relate to their own gambling and its possible consequences.

Strengths and limitations

This is the first European study to explore gambling and gambling problem among both professional athletes and coaches. The survey targeted all athletes and coaches in defined elite groups and the study population large. Although the study had a very low response rate for some of the sports, the similarity in problem gambling prevalence between the sports as well as between the different levels of sport strengthens the likelihood of a valid result. Moreover, a lower response rate is to be expected in a total population survey, which is one of its shortcomings compared with a sample survey. Partially missing data (i.e. item nonresponse) were handled by multiple imputation, which is a powerful way of handling missing values before any analysis is performed (Royston, 2004). Another shortcoming is that we designed the web survey ourselves and used it without initial validation. However, no other options were available, and to ensure relevance we constructed the survey in cooperation with experts in the field. Even so, self-report data generally imply a level of uncertainty. It is difficult to know the degree of honesty among the participants due to social desirability bias. One final limitation is the potential correlation between data from the athletes and coaches in the same team. This possibility was not taken into account in the analyses due to a strongly expressed requirement for anonymity from our sources within the sport environment; consequently, we did not ask the athletes or coaches what team they belonged to, as this would have threatened their anonymity.

Conclusion

This study builds our knowledge of gambling and gambling problem among athletes and coaches in four sports. The difference in the prevalence of ‘at risk for gambling problems’ between female and male athletes raises the question of how the sport environment affects the degree of gambling. Gambling problem seems to be related to environmental factors such as talking about gambling and other signs of gambling interest. The responsibility for addressing and resolving gambling problem in sport teams, rests with the coaches and their managers. However, with a 6% prevalence of ‘at risk for gambling problems’ among the coaches, carefully monitoring the implementation of prevention efforts and who should be responsible for them may be justified. Our findings suggest that the sports clubs need to develop a policy or guidelines in order to guide the athletes and coaches on how to relate to gambling at work, as well as inform were to access help if needed. Additional research is needed to understand how gambling problem is contextualized within sports, and the extent to which preventive efforts can reduce gambling problem.

Conflict of interest Funding sources

This study was financed by the Public Health Agency of Sweden (ref: 1592-20216-3.4.4).

Competing interests

Ingvar Rosendahl and Natalie Durbeej have no competing interests. Maria Vinberg has received unrestricted research grants from Svenska Spels Forskningsråd (2018) to explore gambling problem among grassroots athletes in four sports.

Constraints of publishing

The authors declare that there are no constraints on publishing.

Acknowledgements

We would like to thank our research assistant Anders Axelsson for his work in contacting the sport clubs. Another person we want to thank is Ulla Romild at Public Health Agency of Sweden for providing us figures on the prevalence of gambling habits in the general population of Sweden. We also thank the Swedish Sport Federation for their support and the Swedish Ice Hockey Central Organization for helping us contact the clubs. Finally, we are grateful to all the participants who completed the survey.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Maria Vinberg

Maria Vinberg is a PhD student at the Centre for Psychiatry Research, Karolinska Institutet, Stockholm County Council, Sweden. Her research focuses on gambling within sport from a public health perspective.

Natalie Durbeej

Natalie Durbeej, PhD, is a researcher at Uppsala University, Sweden. Her research focuses on mental health problems, substance use problems, and interventions among children and young people.

Ingvar Rosendahl

Ingvar Rosendahl, PhD, is a statistician and epidemiologist at the Centre for Psychiatry Research, Karolinska Institutet, Stockholm County Council, Sweden. For the last ten years, his research has focused on the aetiology and treatment of addictive behaviours.