Journal of Gambling Studies

, Volume 30, Issue 1, pp 47–60

Predicting Gambling Problems from Gambling Outcome Expectancies in College Student-Athletes


    • International Centre for Youth Gambling Problems and High-Risk Behaviors, School/Applied Child PsychologyMcGill University
  • Caroline E. Temcheff
    • Département de psychoéducationUniversité de Sherbrooke
  • Rina Gupta
    • International Centre for Youth Gambling Problems and High-Risk Behaviors, School/Applied Child PsychologyMcGill University
  • Jeffrey Derevensky
    • International Centre for Youth Gambling Problems and High-Risk Behaviors, School/Applied Child PsychologyMcGill University
  • Thomas S. Paskus
    • The National Collegiate Athletic Association
Original Paper

DOI: 10.1007/s10899-012-9355-4

Cite this article as:
St-Pierre, R.A., Temcheff, C.E., Gupta, R. et al. J Gambl Stud (2014) 30: 47. doi:10.1007/s10899-012-9355-4


While previous research has suggested the potential importance of gambling outcome expectancies in determining gambling behaviour among adolescents, the predictive ability of gambling outcome expectancies has not yet been clearly delineated for college-aged youth. The current study aims to explore the relationships between gender and outcome expectancies in the prediction of gambling severity among college student-athletes. Data from the National Collegiate Athletic Association (NCAA) study assessing gambling behaviours and problems among U.S. college student-athletes were utilized. Complete data was available for 7,517 student-athletes. As expected, male college student-athletes reported more gambling participation as well as greater gambling problems than their female counterparts. Findings showed positive relationships between the outcome expectancies of financial gain, and negative emotional impacts and gambling problems. That is, those who endorsed more items on the outcome expectancy scales for financial gain and negative emotional impacts also tended to endorse more gambling-related problems. Findings also showed a negative relationship between outcome expectancies of fun and enjoyment, and gambling problems over and above the variance accounted for by gender. Those with gambling problems were less likely to have the expectation that gambling would be fun than those without gambling problems. Despite NCAA efforts to curb gambling activity, the results suggest that college student-athletes are at risk for over-involvement in gambling. Therefore, it is important to explore gambling outcome expectancies within this group since the motivations and reasons for gambling might be able to inform treatment initiatives.


GamblingYouthsCollegeStudentsAthletesOutcome expectancies


College students encompass the largest group of young adults in the United States and other Western countries (Rigotti et al. 2000). As a means of exploring their newly-acquired independence and of signalling their maturity, many college students engage in high-risk activities such as cigarette smoking, frequent alcohol use or heavy episodic drinking (HED), and illicit drug use (Laska et al. 2009; Mohler-Kuo et al. 2003; Wechsler et al. 2002). Evidence from prevalence studies reveals that many of these youth are also participating in gambling, another potential risk behaviour (LaBrie et al. 2003; Weinstock et al. 2007).

Youth gambling behaviour can be considered on a continuum ranging from non-gambling at one extreme to problem or pathological gambling at the other extreme, with social or recreational gambling lying in between. As is the case with other potentially risky behaviours such as alcohol consumption and drug use, for most college students gambling is an occasional recreational activity with few negative consequences. The majority of college students who gamble do so for enjoyment and excitement, and are generally able to set and maintain limits for gambling participation (Gupta and Derevensky 2008). However, a small but meaningful proportion of students become over-involved in gambling activities and experience serious gambling-related problems. From existing published studies in the United States and Canada, Blinn-Pike et al. (2007) employed a meta-analytic strategy to synthesize the prevalence estimates of disordered or problem gambling among college students. They concluded that approximately 7.9 % of college students met the criteria for problem gambling. In addition, national surveys of college student gambling reveal that between 3 and 18 % gamble at least once a week (Barnes et al. 2010; Labrie et al. 2003; Nelson et al. 2007). Of particular concern is that estimates of problem gambling among college students exceed prevalence rates in the general adult population (which are established to be between 0.4 and 4.7 %; Stucki and Rihs-Middel 2007), and that gambling more than once per month has been recently identified as a behavioural marker of pathological gambling among college students (Weinstock et al. 2008). As such, problem gambling is increasingly being recognized as a serious public health issue on college and university campuses (Stuhldreher et al. 2007).

Certain concomitant demographic correlates of gambling among college students have been identified in the extant literature. Both cross-sectional and longitudinal research has consistently shown that being male is a risk factor for gambling involvement and problem gambling. Compared to women, college-aged men are more likely to gamble frequently and to experience gambling-related problems (LaBrie et al. 2003; Slutske et al. 2003). It is nevertheless unclear whether the observed gender differences between college students relate to male–female differences in other areas, such as general motivations to gamble (Stinchfield et al. 2006). In addition to demographic correlates, several studies have identified substance use/abuse and parental history for gambling problems as salient risk factors for problem gambling among college students (Goudriaan et al. 2009; King et al. 2010; LaBrie et al. 2003).

Although useful, these identified risk factors cannot fully explain why some college students choose to gamble or why they gamble excessively. Social cognitive models of health behaviour (e.g., Theory of Planned Behavior, Ajzen 2002) have been proposed to be potentially useful for explaining gambling behaviour, including the decision to gamble or the selection of a specific gambling activity (Evans 2003). These social cognition models place importance on a number of proximal determinants of behaviour that are believed to mediate the effects of other influential factors (e.g., gender, age, parental models) (Conner and Norman 2005). With respect to gambling, a growing body of research has identified attitudes towards the behaviour, perceptions of significant others’ evaluations of the behaviour, and appraisals of control over refusing the behaviour as significant predictors of intentions to initiate or continue gambling, and of gambling involvement and problem gambling (Martin et al. 2010; Wu and Tang 2011). College students’ expectations concerning the potential benefits associated with gambling behaviour are also reported to be positively correlated with gambling frequency and gambling problems, whereas their perceptions of the possible negative consequences are found to be negatively related to gambling frequency (Wickwire et al. 2007). These perceived costs and benefits of engaging in gambling behaviour (i.e., outcome expectancies) are posited to be predictive of an individual’s overall gambling attitudes (Moore and Ohtsuka 1999a, b).

Within the college environment, student-athletes are proposed to be a special sub-population at particular risk for problematic health-related behaviours. Owing to their dual status, college student-athletes are required to balance both academic and athletic demands. There is speculation that the risk for maladaptive health-related behaviours is likely exacerbated by the heightened physical and psychological stress, as well as the greater time constraints resulting from this dual status (Yusko et al. 2008a, b). Given that the experiences of college student-athletes may be characteristically different from non-athletes, it is possible that preventive interventions and treatment initiatives geared towards the general student body are not appropriate for student-athletes. Research attention focused specifically on student-athlete participation in high-risk activities is therefore warranted.

Investigations examining the substance use behaviours of college student-athletes and non-athletes have revealed that student-athletes engage in heavy episodic drinking and use smokeless tobacco at greater frequency than their non-athlete peers (Yusko et al. 2008a, b). There is also emerging evidence suggesting that gambling problems are prevalent among students involved in college athletics (Huang et al. 2007), and that rates of problem gambling among current and former student-athletes are comparable to or higher than those observed in non-athletes (Nelson et al. 2007; Weinstock et al. 2007).

Despite emerging evidence about its prevalence, gambling among college student-athletes has received considerably less research attention. Additionally, no studies have systematically evaluated the relationship between gambling-related outcome expectancies (i.e., the perceived positive and negative consequences of engaging in gambling) and college student-athletes’ involvement in gambling or the severity of their gambling problems. While previous studies have identified the potential importance outcome expectancies in the prediction of gambling participation and gambling problems among adolescents and college students (Gillespie et al. 2007b; Wickwire et al. 2010, 2007), the predictive ability of positive and negative outcome expectancies has not yet been clearly delineated for college student-athletes. Moreover, preliminary evidence suggests that the value of gambling outcome expectancies in the prediction of gambling severity may differ for adolescent males and females (Gillespie 2010; Gillespie et al. 2007b), although it was noted that this may be an artefact of the small sample of female problem gamblers. Further exploration of the relationships between gender and outcome expectancies in the prediction of gambling severity among youth is therefore needed. While identification of gambling outcome expectancies and the relationships between gender and outcome expectancies is only one piece of the much larger puzzle of predicting gambling severity among college student-athletes, its exploration as a line of inquiry may nevertheless have the potential to inform future treatment initiatives targeting this particular population.

The principle objective of the current study was to evaluate the predictive ability of gambling outcome expectancies while controlling for the effects of gender in the prediction of gambling severity among college student-athletes. In addition, this study sought to ascertain if gambling outcome expectancies differentially predicted gambling severity for women and men involved in college athletics.



The current study utilized data from the 2008 National Collegiate Athletic Association (NCAA) study designed to assess gambling behaviours and problems among U.S. college student-athletes. This self-reported, voluntary, and anonymous survey was used to obtain a comprehensive assessment of gambling behaviour and experiences among U.S. college student-athletes.

In order to obtain sufficiently large and representative samples of NCAA student-athletes from all sports in each of the three divisions, a random stratified sampling procedure was used to select teams for participation. All member colleges of the NCAA were invited to participate, but no school was asked to collect responses from more than three of its athletics teams. All student-athletes from sampled teams were given the opportunity to take the survey. The school-level response rate was estimated to exceed 60 % based on previous surveys conducted using a similar procedure and given the total number of returned surveys. However, exact institutional response rates could not be established as survey responses were submitted anonymously.

A total of 19,942 surveys were administered to student-athletes in 2008. Males (approximately 62 % of participants) were marginally overrepresented in the sample, compared to gender proportions in the student-athlete population reported by the NCAA (57.2 % males; NCAA 2010).


Gambling Activities Questionnaire—Adapted (GAQ; Gupta and Derevensky 1996). The GAQ is designed to assess four general domains related to gambling behaviour, including descriptive information, cognitive perceptions, familial gambling, and comorbidity with other high-risk and delinquent behaviours. For this study, a modified version of the GAQ was administered to collect descriptive information regarding the frequency of participation in 14 common gambling activities over the previous 12 months (“not at all”, “less than once a month”, “at least once a month”, “at least once a week”, and “daily”). The frequency of gambling behaviour across various types of activities was used to categorize participants as non-gamblers (did not engage in any gambling activities in the previous 12 months) and gamblers (endorsed gambling at least once in any activity in the past 12 months).

Diagnostic Statistical Manual–IV–Text Revision (DSM–IV–TR; American Psychiatric Association 2000). The DSM-IV-TR provides a list of 10 diagnostic criteria for pathological gambling, including: preoccupation with gambling; loss of control; need to increase wagers to achieve same level of excitement (tolerance); withdrawal symptoms; escape; chasing of losses; lying to family; illegal activities to pay for gambling; disruptions to family or employment; and borrowing money to pay for gambling debts. The questionnaire contained one item corresponding to each of the diagnostic criteria, and responses to the 10 items were summed to create a DSM Gambling Screen score. Scores ranged from 0 to 10, with higher scores indicating endorsement of more criteria. Standard cut-off scores for problem gambling classification were used to categorize past-year gamblers as “social gamblers” (DSM score of 0–2), “at-risk gamblers” (DSM score of 3–4), and “probable pathological gamblers (PPG)” (DSM score of 5 or more). This questionnaire format has been found to exhibit satisfactory reliability (e.g., Cronbach’s alpha = .92), validity and classification accuracy (e.g., specificity = .98) (Stinchfield et al. 2005). Internal consistency of the 10-item scale in this sample was acceptable with a Cronbach’s coefficient alpha = .85, and with an average correlation between the items of r = .40.

Gambling Expectancy Questionnaire (GEQ; Gillespie et al. 2007a). The GEQ is a 23 item self-report measure that comprises four discrete subscales representing three positive outcome expectancies (Enjoyment/Arousal; Self-Enhancement; Money) and one negative outcome expectancy (Over-involvement/Negative Emotional Impact). For each scale, items are scored on a 7-point Likert scale ranging from 0 (no chance) to 7 (certain to happen). The Enjoyment/Arousal scale consists of seven items denoting enjoyment, excitement, boredom, escape/tension reduction, and social interaction (Cronbach’s alpha = .94 in this sample). The Self-Enhancement scale is comprised of four items reflecting themes of social acceptance and independence (Cronbach’s alpha = .93). The Money subscale contains five items denoting the theme of gambling for financial gain (Cronbach’s alpha = .92). Finally, the only negative outcome expectancies scale, the Over-involvement/Negative Emotional Impact subscale consists of 10 items reflecting negative themes of preoccupation, relational disruptions, and negative emotions resulting from gambling (Cronbach’s alpha = .93).


Ethical approval for the protocol of the current study was obtained from the institutional review board at NCAA and from the research ethics committees of respective institutions where the surveys were administered. Once institutions were identified and sports were selected for those schools, the faculty athletics representatives (FARs) at each member institution were contacted to assist with survey administration. Each FAR was provided with a specific protocol to follow and a script to read, which emphasized that the study was completely voluntary, that each participant’s responses were anonymous, and that voluntary completion of the study constituted informed consent to participate. The surveys were group-administered by FARs to all adult student-athletes of a sampled team without coaches or other team personnel present. One student-athlete was assigned the responsibility of collecting the completed surveys, placing them in a sealed envelope, and mailing the pre-addressed envelopes to an independent third-party vendor that compiled and entered the data.

Data Preparation

Prior to the present study, rigorous data cleaning procedures were implemented in order to eliminate as much invalid data resulting from insincere or dubious survey responses as possible. Included in these cleaning procedures were a series of validity checks and Item Response Theory techniques to identify questionable patterns of responding. Cases revealing strong evidence of insincere responding (e.g., statistically unlikely combination of responses, inconsistent responses, responses in certain portions of the survey that contradicted responses elsewhere) were excluded. Application of the data cleaning procedures resulted in a sample of 17,675 student-athletes with valid data, with excluded participants representing 11.4 % of the total surveys. Details on the data cleaning procedures have been provided elsewhere (Shead et al. in press).

Despite the application of data cleaning procedures, 8.7 % of the sample with valid data did not respond to any of the GEQ items, while an additional 8.9 % of the sample did not respond to one or more items of the measure. The “available-case method” of excluding incomplete cases from the analyses was selected as a method for dealing with missing observations. Given the large size of the sample and the relatively marginal proportion of missing values for each GEQ item, this method was estimated to be appropriate as non-random patterns of missing values will have little influence on the results when the pattern concerns a small number of cases (Tabachnick and Fidell 2001). Application of this exclusion criteria resulted in a sample of 14,559 student-athletes with complete and valid data. No differences in the gender distribution, rates of gambling participation, and rates of gambling severity were observed between this sample (N = 14,599) and the original cleaned sample (N = 17,675).

Since the present study examined the relationships between gender and outcome expectancies in the prediction of gambling severity, an additional set of filters was implemented to ensure problem gambling severity rates were valid. The DSM-IV-TR section of the questionnaire, which follows the GAQ, contains the instructions, “If you have not gambled, bet or wagered in any way during the past 12 months, please skip [this section]”. Despite the provision of an explicit instruction, a number of participants who reported participation in gambling activities on the GAQ skipped the DSM-IV-TR. Accordingly, the following three guidelines were used to filter out and categorize respondents: (1) those who missed the GAQ and DSM-IV-TR were classified as “missing” (N = 2,084); (2) those who indicated “no gambling” in the past year on the GAQ were categorized as “non-gamblers” irrespective of response or non-response for the items of the DSM-IV-TR (N = 3,827); and (3) those who indicated gambling participation in the past year on the GAQ were but skipped the DSM-IV-TR were classified as “missing” (N = 1,131). All three groups of participants were omitted from all analyses. A final analytic sample of 7,517 participants was retained following application of the data filters.

Intercorrelations between the scales are shown in Table 1 (n = 7,517). Estimation of multicollinearity was conducted via collinearity diagnostics for linear regression analyses (Field 2005). A linear regression analysis was conducted to obtain estimates of tolerance and variance inflation factor (VIF) for the predictors using gambling severity as the outcome variable. None of the tolerance and VIF values indicated collinearity problems. Therefore, no statistical methods were implemented to deal with multicollinearity.
Table 1

Summary of intercorrelations between outcome expectancies, gender, and dependent variable of gambling severity among gamblers (N = 7,517)


Gambling severitya




Negative impact


Gambling severitya






Enjoyment / arousal















Negative Impact





“Negative Impact” refers to the over-involvement/emotional impact expectancies

aCorrelation coefficients are point-biseral, rpb

Data Analysis

Due to insufficient numbers of female PPG gamblers, the severity of participants’ gambling behaviours was dichotomized as whether or not they were problem gamblers in the past year. All at-risk gamblers and PPG gamblers were classified as “problem gamblers”, whereas social gamblers were categorized as “non-problem gamblers”.

To address the research goal of the current study, logistic regression was selected as the data analytic method given that the distributions of the independent variables (gender, four scales of the GEQ) were not likely to satisfy the assumptions of normality (Tabachnick and Fidell 2001). The logistic regression was undertaken to evaluate the contribution of each predictor while controlling for the effects of the other predictors (Tabachnick and Fidell 2001). Once all main effects were examined, further regression analyses were carried out examining interactions between gender and outcome expectancies. Since no interactions were significant, they were subsequently dropped from the final model.


Gambling Severity Rates

Based on DSM-IV-TR criteria, overall, 95.1 % of student-athletes were considered social gamblers, 2.4 % were considered at-risk for pathological gambling, and 2.5 % met criteria for pathological gambling. Gender differences in the rates of gambling participation and gambling severity were observed, and are reported in Table 2.
Table 2

Gambling severity rates (past year) grouped by gender (N = 7,517)


Social (%)

At-risk (%)

PPG (%)

Male (n = 5,424)




Female (n = 2,093)




Logistic Regression Analysis Predicting Problem Gambling

A logistic regression analysis was performed to determine whether gender and gambling outcome expectancies differentiated student-athletes who gamble but have no associated difficulties from those who are might be developing or who have gambling problems. None of the two-way interactions included in the model were found to be statistically significant, and were subsequently dropped in the logistic regression analysis. A test of the model against the intercept-only model was statistically significant [χ² (5, n = 7,517) = 381.35, P = .00], indicating that the predictors, as a set, reliably distinguished between non-problem gamblers and problem gamblers. A moderate improvement in fit was observed for the full model over the intercept model, with McFadden’s ρ² = .13. Prediction success was adequate, with 69.4 % of non-problem gamblers and 74.4 % of the problem gamblers correctly predicted, for an overall success rate of 69.6 % (see Table 3).
Table 3

Classification table for logistic regression model predicting gambling severity (n = 7,517)



% correct

Non-problem gambler

Problem gambler

Non-problem gambler




Problem gambler




Overall % correct



Base rate = .05, hit rate = .70, sensitivity = .74, specificity = .69, false positive rate = .02, false negative rate = .89, positive predictive value = .11, negative predictive value = .98

The contribution of each of the individual predictors on gambling severity, while controlling for the other variables in the model, is summarized in Table 4. As expected, gender emerged as a statistically significant predictor in the logistic regression model, with the odds of a male student-athlete having a gambling problem in the past 12 months approximately 5 times greater than the odds for a female student-athlete, even after controlling for the other model variables. Expectancies of Enjoyment/Arousal and Money were found to be statistically significant predictors of problem gambling. For each one-unit increase on the Enjoyment/Arousal subscale of the GEQ, the odds of having a gambling problem in the past 12 months decreased by a factor of .80. Conversely, a one-unit increase on the Money subscale of the GEQ was associated with college-student athletes being 1.4 times more likely to have a gambling problem. Further, high scores on the expectancy scale of Over-Involvement and Negative Emotional Impact (referred to as “Negative Impact” in Table 4) were found to be a statistically significant predictor of problem gambling; the odds of having a gambling problem over the past 12 months increased by 2.3 times for each one point increment on the subscale. No statistically significant association of Self-Enhancement outcome expectancies and gambling severity was found.
Table 4

Logistic regression analysis predicting gambling severity (n = 7,517)




Wald test (z-ratio)




















[3.33, 7.56]








[.67, .95]








[.67, 1.11]








[1.11, 1.87]

Negative Impact







[1.93, 2.76]

Nagelkerke R2 = .15

β is the parameter estimate. SE standard error, OR odds ratio. CI.95 95% confidence interval, NA not applicable. “Negative Impact” refers to the outcome expectancy of over-involvement/negative emotional impact

* Indicates a statistically significant result

aFemale is the reference for gender


College students in general, and college student-athletes in particular, have been found to be an at-risk group for the development of gambling problems (Huang et al. 2007; Nelson et al. 2007; Weinstock et al. 2007). Within this sample, 3.2 % of male student-athletes and 0.7 % of female student-athletes met criteria for probable pathological gambling and an additional 3.2 % of male and 0.5 % of female student-athletes met criteria for at-risk gambling. This finding suggests that an identifiable proportion of college student-athletes are experiencing a number of serious gambling-related issues. Given changing trends in gambling (e.g., wider availability of Internet and mobile gambling opportunities, expansion of land-based gaming, increased media exposure), it is not surprising that gambling has become a popular activity among student-athletes and that, for some, gambling can transition from a form of entertainment to a more serious problem. Understanding the motivations for gambling among college students is therefore of crucial importance for the development of effective prevention programs among this at-risk population.

Gender and Outcome Expectancies in the Prediction of Gambling Severity

The findings identify specific cross-sectional predictors of problem gambling for college student-athletes. Consistent with previous research, gender emerged as a robust predictor of gambling-related problems (King et al. 2010; Nelson et al. 2007): college-aged men were nearly five times more likely to report experiencing gambling-related problems compared to their female counterparts. However, no gender differences in positive and negative outcome expectancies in the prediction model for gambling severity were revealed. A possible explanation for this is that an insufficient number of female problem gamblers (n = 26) were available in the sample to detect any significant interactive effects.

Overall, the results demonstrated that gambling-related expectancies of young adult student-athletes are predictive of gambling problems. Negative outcome expectancies of over-involvement and adverse emotional effects demonstrated the strongest relationship with gambling severity for both male and female samples. Specifically, although the majority of student-athletes who gamble say that they do so “for fun” (Nower and Blaszczynski 2010; Paskus et al. in press), student-athletes who expected gambling to make them feel badly were twice as likely to report experiencing gambling problems than student-athletes who did not have this expectation. This mirrors similar findings from the extant gambling expectancy literature (Gillespie 2010; Gillespie et al. 2007b; Wickwire et al. 2010). It is plausible that the positive relationship between negative outcome expectancies and gambling severity is recursive in nature: as gambling behaviour intensifies so do negative consequences, which may in turn strengthen negative outcome expectancies. Only longitudinal, repeated measures methodologies would be able to decisively tease apart the developmental links and sequences between gambling problems and negative gambling expectancies. Regardless, this finding suggests that for college student-athletes who gamble, current negative outcome expectancies are not a protective factor against the presence of gambling problems, but rather are characteristic of those at greatest risk for problem gambling.

For both male and female student-athletes, expectancies of financial gain were also shown to be positively related to gambling problems, a finding that has been previously reported for adolescent samples (Gillespie et al. 2007b; Wickwire et al. 2010). In a recent study, monetary expectancies were observed to contribute significantly to increases in physical arousal and subjective excitement during gambling, and that these reactions persisted over time even in the face of losses (Wulfert et al. 2008). It is therefore plausible that the physiological and psychological changes in response to expectations of financial gain may function as powerful reinforcers that sustain excessive gambling behaviour in spite of persistent losses.

An unexpected finding was that for both male and female student-athletes, expectancies of enjoyment were negatively related to gambling severity, with stronger expectations of “fun” and “enjoyment” from gambling being inversely related to the likelihood of having a gambling problem. While previous studies using adolescent samples have reported a positive relationship between expectancies of enjoyment and stimulation and problem gambling (Gillespie et al. 2007b), it may be that for college student-athletes who gamble excessively, the long-term negative consequences of gambling become much more salient than the immediate heightened arousal or enjoyment benefits. Therefore, as student-athletes’ experiences with the negative consequences of gambling intensify with excessive play, their expectations of “fun” and “enjoyable” outcomes from gambling activities may become less evident or relevant.

Treatment Implications

The fact that students who indicated holding the expectation that gambling will make them feel badly were twice as likely to report having a gambling problem is not surprising, since their level of gambling severity is associated with negative consequences. Thus, it is likely that student-athletes who have experienced harm from gambling are more aware of its negative consequences. Directly in line with this is the finding that problem gamblers endorsed expectancies of “fun” less often than those who gamble without having gambling problems. These findings suggest that once a gambling problem has developed, gambling behaviour is not driven by positive expectations. On the contrary, young adults appear to be keenly aware of negative outcomes that result from continuing their gambling behaviours. That student-athletes choose to continue gambling despite negative expectations suggests a loss of control over their gambling participation. Therefore, focusing interventions on expectancies might not be effective targets for cognitive interventions with college-aged problem gamblers, but rather a focus on self-control strategies and time management is called for. Furthermore, consistent with past research (Gillespie 2010; Gillespie et al. 2007b; Wickwire et al. 2010), the fact that expectancies of financial gain were predictive of gambling problems suggests that young problem gamblers may hold cognitive distortions in the form of unrealistic expectancies of financial gain, highlighting another salient area for cognitive work with young adult problem gamblers.

That young problem gamblers continue to gamble despite having more negative expectancies and fewer positive expectancies of enjoyment might also be an indication that they use gambling as a way of dealing with difficult emotions, often referred to as “escaping”. It is possible that although college-aged students experiencing gambling problems report expecting negative impacts from their gambling, they may in fact simultaneously be experiencing salient and immediate positive experiences, thus allowing them to turn a blind eye to the more delayed negative consequences of gambling. One possible outcome expectancy that was not tested in this study is the expectancy that gambling would help cope with negative emotions or stressful situations. However, it is possible that problem gamblers may have greater expectations for gambling in terms of helping them cope with difficult situations than non-problem gamblers. Taken together, these results suggest that treatment for young problem gamblers should incorporate a variety of elements including cognitive therapy and affective regulation in addition to providing alternatives to gambling, encouraging healthy coping skills, and teaching self-control mechanisms. In fact, providing personalized feedback on reported outcome expectancies is a central component of an empirically-supported, motivationally-enhanced cognitive-behaviour treatment for adult pathological gamblers (Wickwire et al. 2010). The current findings may therefore be useful in the development of treatment programs intended to reduce gambling among youth gamblers.

It is important to acknowledge that the NCAA has specific statutes and regulations prohibiting sports wagering among their college student-athletes. In addition, the NCAA has established comprehensive as well as sports-specific programs aimed at educating athletes about these regulations. Despite these efforts, results of this study demonstrate that many athletes continue to gamble, and a small but significant proportion of them are experiencing problems relating to their gambling behaviour. Further, it is possible that some college student-athletes are reluctant to seek help for their gambling out of fear of losing their eligibility to play on school teams. It may be important to conduct a review of the procedures and possibly the values underlying the regulations in order to allow those with gambling problems to seek therapy without jeopardizing their eligibility.

Limitations and Future Directions

This research allows a greater appreciation for the significance of outcome expectancies in the prediction of gambling problems among students involved in college athletics. However, this study represents only an initial step toward understanding college student-athletes’ motivations and reasons for gambling, and the findings must be interpreted in light of their preliminary nature. Several additional limitations to this study must be noted. First, no inferences regarding causality can be made due to the correlational nature of the data. This is particularly important when outcome expectancy constructs are considered, as it is impossible to discern whether current expectations drive current gambling behaviour or whether past gambling experiences determine current outcome expectancies. One could easily assume that as an individual travels across the gambling severity spectrum, the motivations or expected gains driving their behaviors would change over time. It is highly plausible that at the onset of gambling participation, one would hold positive expectancies, but as the loss of control sets in, those expectations change as a result of having experienced negative consequences. Understanding expectancies driving gambling behavior along the gambling spectrum has meaningful implications for prevention, thus highlighting the benefits of a longitudinal investigation of motivations and expectancies held by people who gamble. Although this research is necessarily correlational, future research investigating longitudinal links between outcome expectancies and development of gambling behaviours and problems would likely add to our clinical understanding of young adult problem gamblers.

Second, the data in the present study were based on self-report questionnaires. A series of checks, Item Response Theory techniques, and filters were used to ensure the validity of the data. However, it is impossible to evaluate each respondent’s true engagement in the questionnaire and/or the seriousness with which they completed each section. Also, given the implications of some of the questions asked in the survey (e.g., items pertaining to the serious violations of NCAA rules that could result in loss of eligibility or possible criminal charges), it is possible that a number of participants were cautious in reporting their actual gambling activities or related problems. Nevertheless, the prevalence rates of gambling behaviour and gambling-related problems suggest a representative sample that is consistent with previous research.

Despite efforts to examine how gambling outcome expectancies differentially predict gambling severity for women and men, the insufficient number of female problem gamblers (n = 26) available in the sample does not allow for firm conclusions to be drawn about the relationships between gender and outcome expectancies in the prediction of gambling problems among students involved in college athletics. To improve the results’ generalizability, future studies are needed to replicate these findings with a larger sample of female problem gamblers. Further, youth problem gambling may be influenced by numerous other biological, psychological and social-cognitive factors. Although the findings and treatment implications of this study warrant definite consideration, future research that aims to develop a more comprehensive model delineating the direct and meditational relationships between risk factors, outcome expectancies, and gambling involvement/problem gambling is needed.

Finally, while the results of the current study are largely consistent with the extant clinical literature suggesting that young adults problem gamblers might be motivated by financial gain and many among them may be keenly aware of potential long-term negative impacts of their gambling (Lesieur 1984; Toce-Gerstein and Gerstein 2004), this study does not allow us to ascertain how well these results would generalize to the population of college students and young adults in general given that we do not have a non-athletic comparison group.


The findings of this study showed that among college student-athletes, positive outcome expectancies of fun, enjoyment, and financial gain, as well as negative outcome expectancies of over-involvement and negative emotional impacts of gambling were significant predictors of gambling problems over and above the variance accounted for by gender. Since college student-athletes are another population at risk for over-involvement in gambling, it is important to explore gambling outcome expectancies within this population as the motivations and reasons for gambling might be able to inform treatment initiatives.


The financial support for this study was provided by the National Collegiate Athletic Association (NCAA). The authors wish to thank the nearly 20,000 student-athletes for their close attention and candor in responding to the extensive questionnaire, representing NCAA’s second national study on collegiate sports wagering and gambling behaviors.

Copyright information

© Springer Science+Business Media New York 2013