Student participation in gambling: the role of social cognition, past behaviour, and habit

ABSTRACT

The study examined the relationship between Theory of Planned Behaviour constructs, past behaviour, habit, and future behaviour relating to students’ participation in gambling. Using a cross-sectional design, theory constructs, past behaviour, and habit were examined at T1 (N = 250), and gambling behaviour was assessed four weeks later at T2 (N = 180). Results showed attitude and perceived behavioural control (PBC) predicted intention, and intention and PBC predicted behaviour. The inclusion of past behaviour and habit attenuated the effects of attitude and PBC on intention and rendered the impact of intention on behaviour non-significant. The relationship between past behaviour and future behaviour was mediated by habit and PBC. Interventions should focus on attitude and PBC to attend to gambling intentions in addition to automatic processes to attend to student gambling behaviour.

KEYWORDS:

• Gambling behaviour
• Theory of Planned Behaviour
• social cognition
• past behaviour
• habit

Introduction

University students have demonstrated excessive gambling behaviour (Nowak, 2018). Although participation rates compare with young adults more generally (Kessler et al., 2008), university is a setting where behavioural patterns can be established for later life (Sawyer et al., 2012). Gambling can negatively impact academic performance (Dowling et al., 2017) and lead to mental health issues (Rossen et al., 2016). Research is therefore needed to understand the psychology underlying student participation in gambling.

The Theory of Planned Behaviour

The Theory of Planned Behaviour (TPB; Ajzen, 1991) is a prominent social cognition model that enables behaviour to be understood. The theory identifies intention as the proximal determinant of behaviour (see Figure 1). Intention, in turn, is influenced by attitude, subjective norm, and perceived behavioural control (PBC). Research has found theory constructs to explain 40%-45% of the variance in intention and 19%-36% of the variance in behaviour (Armitage & Conner, 2001; Hagger et al., 2002; McEachan et al., 2011).

Figure 1. The Theory of Planned Behaviour (Ajzen, 1991).

Despite the explained variance, research has established past behaviour to play a significant role on future behaviour (Hagger et al., 2018). According to the TPB, past behaviour can influence future behaviour, but this must be mediated through model constructs (Ajzen, 1991). However, research has identified not only indirect effects of this relation, but also direct effects (Brown et al., 2018; Hagger et al., 2016).

A second route to future behaviour has been suggested to represent implicit processes, including habits (Hagger, 2019). Habits have been defined by Gardner (2015) as automatically activated impulses to respond to a stimulus that are triggered by stimulus–response associations formed over repeated experience. Behaviour repeated in the presence of contextual cues can lead to habit development (Lally et al., 2010). Moreover, behaviours performed out of habit are automatic in that they are not reliant on conscious, deliberative processes (Bargh, 1994).

Given gambling is a repetitive behaviour influenced by contextual cues (Ramnerö et al., 2019), it could be that both past behaviour and habit play a role in the behaviour. The purpose of the study was to therefore understand the role of TPB constructs, past behaviour, and habit on student participation in gambling.

Materials and methods

The study was conducted at a university in the UK. A cross-sectional design was used with a four-week follow-up. Participants completed two questionnaires online; the first questionnaire (T1) assessed TPB constructs, past behaviour, and habit, and the second questionnaire (T2) assessed gambling within the previous month. Ethical approval was granted by the school ethics board and all participants consented prior to participation.

TPB constructs were measured at T1 following Ajzen’s (1991) recommendations. Specifically, four items measured attitude, three items measured subjective norm, four items measured PBC, and three items measured intention. Habit was measured using the 4-item behavioural automaticity index (Gardner et al., 2012), and two items assessed past gambling behaviour. Demographics were also measured at T1. Two items assessed behaviour four weeks later at T2. Full measures and psychometric properties can be seen in the supplementary file.

Results

Descriptive statistics and correlation matrix

T1 data was collected on 250 participants (Mage = 19.20, SD = 2.77). The majority of participants were white (n = 235), female (n = 138), in their first year of study (n = 94) and studying full time (n = 241). T2 questionnaires were completed by 180 participants. Tests for systematic differences in sample demographic characteristics and psychological constructs due to attrition revealed no differences between completers and non-completers.

The intercorrelations between study measures can be seen in Table 1. The data showed attitude, PBC, habit, and past behaviour significantly correlated with intention, with habit the strongest correlate of all measures. All constructs significantly correlated with behaviour, with attitude the strongest correlate.

Table 1. Descriptive statistics and intercorrelations between TPB constructs, past behaviour, habit, and behaviour

Variable	2.	3.	4.	5.	6.	7.	M	SD
1. Attitude	.23*	−.62**	.56**	.60**	.53**	.74**	4.53	1.81
2. Subjective norm		−.14*	.09	.08	.14*	.23**	3.64	1.67
3. PBC			−.56**	−.57**	−.56	−.72**	3.48	1.58
4. Intention				.56**	.60**	.58**	3.77	1.70
5. Past behaviour					.51**	.69**	3.44	2.09
6. Habit						.70**	3.39	1.90
7. Behaviour							3.70	2.22

*p < .05, **p < .01; Intention N = 250; Behaviour N = 180.

Predicting gambling intentions and behaviour

A hierarchical linear regression analysis was used to predict intention (see Table 2). At step 1, TPB constructs explained 39% of the variance in intention (F(3, 246) = 52.62, p < .001). Intention was significantly predicted by attitude and PBC, but not subjective norm. At step 2, results showed the predictors explained 49% of the variance in intention (F(5, 244) = 47.91, p < .001). Similar to step 1, attitude and PBC significantly predicted intention whereas subjective norm did not. Past behaviour and habit were significant predictors and significantly increased the variance explained in intention (R² change = .10, F change = 25.279, p < .001).

Table 2. A hierarchical linear regression analyses of attitude, subjective norm, PBC (step 1), habit and past behaviour (step 2) on intention

Predictor	Step 1 β	Step 2 β
Attitude	.36***	.18**
Subjective norm	−.04	−.03
PBC	−.34***	−.15*
Habit		.31***
Past behaviour		.21***
R^2	.39	.49

*p < .05, **p < .01, ***p < .001; Intention N = 250.

A hierarchical linear regression analysis was used to predict T2 behaviour (see Table 3). At step 1, intention and PBC both significantly predicted behaviour and accounted for 57% of the variance (F(2, 177) = 119.54, p < .001). At step 2, the predictors explained 70% of the variance in behaviour (F(4, 175) = 104.28, p < .001) and the explained variance significantly increased with the inclusion of past behaviour and habit (R² change = .13, F change = 38.442, p < .001). Behaviour was significantly predicted by PBC, past behaviour, and habit, but not intention.

Table 3. A hierarchical linear regression analyses of intention, PBC (step 1), habit and past behaviour (step 2) on behaviour

Predictor	Step 1 β	Step 2 β
Intention	.26***	.06
PBC	−.57***	−.31***
Habit		.33***
Past behaviour		.28***
R^2	.57	.70

***p < .001; Behaviour N = 180.

Mediation analyses

Using Hayes’ PROCESS macro (Hayes, 2017), the serial multiple mediator model (model 6) examined relationships between the key predictors of gambling behaviour: past behaviour, habit, and PBC. Results showed past behaviour significantly predicted habit (a₁), PBC (a₂) and behaviour (c). Habit significantly predicted PBC (d₂₁) and behaviour (b₁), and PBC significantly predicted behaviour (b₂). Past behaviour significantly predicted behaviour when controlling for habit and PBC (c’₁). The mediation analyses showed the indirect effects of habit (a₁b₁ = 0.1870, CI = 0.1140 to 0.2826), PBC (a₂b₂ = 0.1498, CI = 0.0796 to 0.2304), and habit and PBC (a₁d₂₁b₂ = 0.0628, CI = 0.0326 to 0.0966) were significantly positive. These effects can be seen in Figure 2.

Figure 2. A statistical diagram of the serial multiple mediator model for the impact of past behaviour on gambling behaviour through habit and PBC.

**p < .01, ***p < .001.

Discussion

In accordance with the TPB, attitude and PBC were significant predictors of intention. These findings broadly share similarities with previous research adopting the TPB (Martin et al., 2010; St-Pierre et al., 2015; Wu & Tang, 2012). In relation to subjective norm, research adopting general measures have found similar null effects (St-Pierre et al., 2015). However, some influence has been identified in studies examining specific referents (Martin et al., 2010; Neighbors et al., 2007). For example, Martin et al. (2010) found no impact of peer norms, but friend and family norms were influential. Therefore, this lack of effect could either be a consequence of the measures used or it could be that students’ gambling intentions are influenced less by the perceptions of significant others and more by evaluations of the behaviour and issues of control.

In relation to behaviour, the significant influence of intention and PBC corroborates previous findings (Martin et al., 2010; St-Pierre et al., 2015; Wu & Tang, 2012). However, the inclusion of past behaviour and habit rendered the influence of intention non-significant. Mediation established the influence of past behaviour on future behaviour partially operated through habit and PBC, independent of intention. With regards to habit, the suggestion is that repetitive behaviour can lead to habit development and once developed, behaviour is less likely to be reliant on intention because habits are automatically activated (Ouellette & Wood, 1998). The findings partially supported this assumption. However, in line with the TPB, PBC partially mediated the past behaviour-future behaviour relationship. This suggests repeatedly gambling weakens perceptions of control and these weakened perceptions result in gambling behaviour. The negative association between PBC and behaviour has been observed in other health risk behaviours, such as alcohol consumption (Norman & Conner, 2006). Finally, the results suggest the path from past behaviour to future behaviour operates through both habit and PBC, with the former influencing the latter. Frequently performing behaviour could lead to the development of habits and when certain cues are encountered, habits weaken perceptions of control. These weakened control beliefs then result in gambling behaviour.

The findings suggest gambling interventions should target positive attitudes and weak perceptions of control. Different behaviour change techniques (Michie et al., 2013), such as providing information about emotional consequences and self-monitoring, could be adopted to modify these determinants. Indeed, self-monitoring was identified by Humphreys et al. (2021) as an effective technique within gambling interventions. However, in relation to behaviour, attempting to manipulate deliberative, reasoned processes may prove limited. Rather, interventions should target automatic processes associated with past behaviour and habit. A number of suggestions have been made to attend to automatic processes (see Sheeran et al., 2013). For example, the development of specific plans may protect against unwanted gambling cues.

Strengths of the study include the examination of past behaviour and habit, and the focus on a detrimental health risk behaviour often performed by students (Nowak, 2018). Limitations concern the correlational design, the use of self-report, and the potential lack of generalizability.

Conclusion

The study identified a number of psychological influences on student gambling behaviour. To reduce intentions to gamble, interventions should focus on weakening attitudes and increasing perceptions of control. In relation to gambling behaviour, focus should be given to the automatic processes associated with past behaviour and habit. Modifying these psychological mechanisms could lead to a reduction in students’ gambling behaviour.

Ethics approval statement

The study received ethical approval from the school ethics board.

Disclosure statement

The author declares no conflict of interest.

Additional information

Funding

There is no funding sources associated with the manuscript.