Confidence in one’s ability to refuse a bet may be among the most revealing psychological signatures of problem gambling, and after decades of fragmented findings, researchers have finally quantified just how strong that signal is. In a preregistered systematic review and meta-analysis published in the International Journal of Mental Health and Addiction, a team led by Natasa Nagy and Zsolt Horváth synthesized evidence from 21 studies encompassing 24 independent samples and uncovered a significant, negative, moderate correlation between problem gambling severity and gambling refusal self-efficacy — the confidence a person has in their capacity to resist gambling in high-risk situations. In plain terms: the lower a gambler’s belief in their ability to say no, the more severe their gambling problems tend to be. The association survived sensitivity analyses and publication bias assessments, yet it concealed substantial heterogeneity, a statistical fingerprint the authors say reveals as much about how science measures self-belief as it does about gambling itself.

The stakes of understanding this relationship are considerable. Gambling disorder is classified as an addictive disorder in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, defined by a cluster of symptoms that includes tolerance, withdrawal-like restlessness, impaired control, preoccupation with gambling, gambling to escape distress, chasing losses, lying about the extent of one’s gambling, and jeopardizing relationships or careers. The broader term problem gambling, widely used in the research literature, covers both sub-clinical impairments and the full range of harmful consequences gambling can produce. A recent systematic review and meta-analysis of representative population-based studies — using standardized screening instruments such as the Problem Gambling Severity Index, DSM-IV/DSM-5 diagnostic criteria, or the South Oaks Gambling Screen — estimated that 8.7 percent of adults worldwide experienced at least one adverse consequence of gambling in the past 12 months, while 1.4 percent met criteria for problem gambling. And the damage rarely stays contained: studies document ripple effects extending to families and wider communities, from financial strain and relationship breakdown to psychological and physical health problems, impaired school and work performance, criminal activity, and cultural and community harm.

The theoretical engine behind the new analysis is social cognitive theory, the influential framework advanced by psychologist Albert Bandura in 1977, which holds that people’s beliefs about their own capabilities are central drivers of behavior. Within this tradition, refusal self-efficacy describes an individual’s belief in their ability to resist engaging in a specific behavior in specific situations, and it has been operationalized separately for each potentially addictive behavior. Gambling refusal self-efficacy, or GRSE, therefore captures a gambler’s confidence in their ability to refuse gambling across a range of tempting contexts — after a bad day, during a winning streak, when friends are betting, or when other substances lower inhibitions. The construct has already proven its worth elsewhere in addiction science: a meta-analysis of drinking refusal self-efficacy found that people with lower confidence in their ability to turn down alcohol drank more heavily and experienced more alcohol-related problems, establishing a pattern gambling researchers have long suspected would replicate.

Individual gambling studies have consistently linked lower GRSE to problem gambling, but the direction of causation has remained stubbornly unclear — a limitation the new paper confronts head-on by laying out three competing pathways. In the first, low refusal self-efficacy acts as a vulnerability factor: shaped by traits such as heightened reward sensitivity or low self-control, by adverse learning experiences, stressful environments, or susceptibility to social pressure, it predisposes people to gamble more frequently, gradually raising the risk of developing problems. In the second, the arrow reverses: gambling itself, through repeated experiences of lost control, erodes refusal self-efficacy over time and spreads that erosion across multiple high-risk situations, a decline reinforced by effects on the central nervous system that impair executive functioning and self-regulatory capacity. In the third, the association is manufactured by underlying third variables such as impulsivity, stress, or emotion regulation difficulties, which simultaneously worsen gambling and undermine self-belief. Which pathway dominates matters enormously: if self-efficacy is a cause, it is a treatment target; if it is a casualty, it is an early-warning symptom; and if it is merely a correlate, both may require attention.

To assemble the evidence base, the research team — which also included Róbert Urbán, Andrea Czakó, and Zsolt Demetrovics — conducted the review in accordance with the PRISMA guidelines and searched five databases: PubMed, ScienceDirect, Academic Search Complete, Web of Science, and PsycINFO, beginning on 25 March 2024, using a keyword strategy that paired gambling terms with the full family of self-efficacy variants: refusal, abstinence, regulatory, and resistance self-efficacy. Studies were eligible only if they were peer-reviewed, written in English, quantitative and observational or correlational in design, measured both problem gambling and GRSE with standardized instruments, and reported a correlation between the two — or sufficient statistical information to compute one. The team excluded studies whose effect sizes were confounded with intervention-related change, those that operationalized problem gambling purely through behavioral involvement such as gambling frequency or money spent, those relying on general or non-gambling-specific self-efficacy scales, and studies using single-item or unvalidated measures. Where statistics needed for effect size calculation or moderation analyses were missing, corresponding authors were contacted at least three times; when the data could not be obtained, the studies were dropped. Two reviewers screened and coded the literature independently, achieving 84.69 percent raw agreement and a Cohen’s kappa of 0.68 at the title-and-abstract stage, with disagreements resolved through consensus and, where necessary, consultation with a third author — a process that ultimately locked in 21 studies containing 24 independent samples.

The pooled result was clear: a significant, negative, moderate association between problem gambling and gambling refusal self-efficacy, accompanied by substantial heterogeneity — a statistical signature indicating that individual studies varied widely in the strength of the correlation they reported. In a meta-analysis, each study’s correlation coefficient is pooled into a single weighted estimate, with larger and more precise studies carrying more statistical weight, and here sensitivity analyses and publication bias assessments supported the robustness of the finding. Moderation analyses, which the authors framed as explicitly exploratory, revealed that two study characteristics significantly shaped the size of the effect: the number of participants and, critically, the type of instrument used to measure GRSE. Strikingly, none of the demographic or clinical moderators tested — gender, age, and clinical status — showed significant effects, despite earlier literature reporting lower GRSE among men, among high school students relative to university students, and among gamblers in clinical samples. The entire analysis was preregistered on the Open Science Framework, with predefined aims, search strategy, inclusion and exclusion criteria, data collection and coding procedures, and an analytical plan specified in advance, and any deviations from that protocol were transparently flagged.

That instrument effect is no statistical footnote; it goes to the heart of how self-efficacy is conceptualized. The Gamblers Self-Efficacy Questionnaire, developed by May and colleagues in 2003, measures a person’s belief in their ability to control gambling behavior in high-risk situations, spanning unpleasant and pleasant psychological and physical states, interpersonal situations and relationships, and control over gambling urges. The Gambling Abstinence Self-Efficacy Scale, introduced by Hodgins and colleagues in 2004, instead assesses a gambler’s confidence in abstaining from gambling in relapse-specific situations, including winning or external triggers, negative emotions, positive mood and testing of urges, and social factors. The Gambling Refusal Self-Efficacy Questionnaire, published by Casey and colleagues in 2008, evaluates confidence in refusing gambling across contexts that include psychoactive substance use, positive and negative emotional states, and gambling-related situations and thoughts. Control, abstinence, refusal — three related but non-identical constructs — may explain why studies relying on different scales arrived at correlations of different magnitudes, and why the meta-analysts found the measurement tool itself to be a significant moderator.

The authors also confronted a persistent blind spot in gambling research: social inequality. Previous reviews have reported elevated rates of problem gambling and gambling-related harm among certain ethnic minority groups and among people experiencing socioeconomic disadvantage, but the new analysis deliberately declined to test ethnicity or socioeconomic status as moderators. The reason is methodological. Ethnicity is operationalized in strikingly different ways across studies — country of birth, language spoken at home, or self-identified cultural background — each capturing different facets of cultural identity and introducing measurement inconsistencies. Socioeconomic status is similarly multidimensional, invoked variously through income, education, occupation, employment status, or composite deprivation indices. Combining such non-equivalent indicators in a meta-regression, the authors reasoned, could distort conclusions rather than clarify them, so they restricted their moderator analyses to gender, age, and clinical status, where reporting was sufficiently consistent for meaningful cross-study comparison.

The findings arrive with candid caveats. Risk of bias assessment suggested moderate or high risk in most of the included studies, and the evidence base remains dominated by observational designs that cannot untangle whether low self-efficacy precedes gambling problems, results from them, or both in a self-reinforcing loop. The authors conclude that lower GRSE is reliably associated with more severe problem gambling, but they argue that the field needs improved designs and measurement — such as longitudinal cohorts tracking refusal self-efficacy and gambling severity over time, and greater standardization of instruments — before the correlational signature documented here can be converted into confident causal claims. Even so, the practical stakes are difficult to ignore. If low refusal self-efficacy genuinely predisposes people to escalating gambling, interventions that build situation-specific confidence could act as prophylactics; if problem gambling instead corrodes self-efficacy, the construct becomes an early-warning marker for the erosion of control, detectable before financial and relational damage compounds.

For now, the meta-analysis delivers the field its first pooled, weighted estimate of a relationship clinicians and researchers have long observed anecdotally: people who doubt their ability to say no to a bet are, in aggregate, the people most entangled in gambling’s harms. As legalized online gambling expands across jurisdictions and mobile platforms compress the distance between impulse and wager into a single tap, the psychology of resistance — and the measurement of its failures — is shifting from academic curiosity to public health priority. What the new synthesis makes unmistakably clear is that the question is no longer whether gambling refusal self-efficacy relates to problem gambling, but how to measure it consistently, when it breaks down, and whether rebuilding it can pull people back from the brink.