Examining interactions within the Theory of Planned Behavior in the prediction of intentions to engage in cannabis-related driving behaviors

Abstract

Objective:

As marijuana use becomes more available to college students through increasing legal reform, this paper seeks to examine intentions for driving under the influence of cannabis (DUIC) and riding with a high driver (RWHD) through the lens of the Theory of Planned Behavior (TPB) and assess potential interactions between personal attitudes, subjective norms, perceived behavioral control (PBC), and sex.

Participants:

Undergraduate college students (N=311) completed online surveys in September, 2013.

Method:

Participants self-reported their attitudes towards DUIC, subjective norms, PBC, past DUIC and RWHD, and intentions to DUIC and RWHD.

Results:

Participants’ attitudes towards DUIC, subjective norms, and PBC were strongly associated with intentions to DUIC and RWHD bivariately. In regression models, attitudes and PBC were both positively and significantly related to intentions to DUIC and RWHD.

Conclusions:

DUIC and RWHD are concerns for college populations. Targeting personal attitudes and perceived behavioral control via interventions may reduce these behaviors.

As of 2018, 22 states have decriminalized possession and 8 states along with Washington D.C. have legalized the recreational use of marijuana.¹ Although the U.S. seems to be adopting a more accepting view of cannabis, legalization may pose a heightened risk for college students. A study examining cannabis use among students at Washington State University before vs after legalization found an increase in cannabis use among both underage and legal students. Further, the magnitude of this increase was significantly greater than predicted by national trends.² Recent meta-analyses and longitudinal studies have concluded that, like alcohol, acute cannabis use significantly increases an individual’s risk of being involved in a motor vehicle collision resulting in serious injury or death.^3–5 Recent reports show an increase in the prevalence of cannabis-impaired driving cases since its legalization in Washington and Colorado.^6,7 The same trend has not been observed, however, with respect to alcohol-impaired driving cases.^6,7 Despite this risk, however, college student cannabis users report driving under the influence of cannabis (DUIC) at greater rates than college student drinkers report driving under the influence of alcohol (DUIA).⁵ For example, in a large sample of college students, alcohol users reported DUIA at a rate of 6.8% while cannabis users reported DUIC at a rate of 44% for males and 9% for females.⁸ This is noteworthy given that lifetime cannabis use prevalence ranges from 43–63% and past month prevalence ranges from 16–39%.⁹ Numerous studies have documented the relationship between DUIC and negative outcomes such as risky driving and crash risk.^5,10–12 Although countless studies have investigated the factors influencing college student engagement in DUIA and riding with a drinking driver (RWDD), relatively few have examined predictors of DUIC.^13–15 and riding with a high driver (RWHD).^16,17 As legalization has been related to increased use among college students, these behaviors become a growing concern and research into predictors of DUIC and RWHD becomes increasingly relevant. In order to inform future efforts aimed at preventing such behavior, the current study seeks to use the Theory of Planned Behavior (TPB)¹⁸ as a means for identifying potential predictors of college students’ intentions to DUIC and RWHD.

The Theory of Planned Behavior

The TPB is a theoretical framework for understanding human behavior in which future behavior is said to be directly predicted by an individual’s level of intention to perform or not perform the behavior in question.¹⁸ Intention, in turn, is thought to be determined by a person’s attitudes, defined as positive or negative feelings about the behavior, subjective norms, defined as the extent to which an individual believes close others would approve of the behavior, and perceived behavioral control (PBC), or beliefs about the ease with which the behavior could be performed or avoided. In a meta-analysis of 185 independent studies, the TPB components accounted for significant variance in a wide range of both observed (R² = .20) and self-reported (R² = .31) health behaviors.¹⁹ More recent proposed alterations to the TPB include the addition of past behavior to predictors of intentions and future behavior.²⁰ Past behavior has been shown to often be the strongest predictor of intentions^20,21 and many studies have found past behavior to predict unique variance in intentions above and beyond the other TPB variables.^22–26

The TPB has been successfully applied to the study of many risky driving behaviors including breaking the speed limit,²⁰ dangerous overtaking,²¹ close following,²⁷ disobeying road signs,²⁸ and DUIA.^29–32 In the latter case, some studies have found all three TPB variables to positively predict intentions to DUIA,^29,31 while others have demonstrated a link from attitudes and PBC to intentions but failed to find a significant relationship between subjective norms and intentions.^30,32 Based on the abundance of research linking the TPB with DUIA and other risky driving behaviors, it is expected that TPB could similarly be applied to the study of driving under the influence of cannabis and useful in identifying variables that interventions could manipulate in order to reduce or prevent such behavior.

Multiplicative effects of TPB Components

While the TPB posits that PBC should directly influence intentions and behavior, it has been argued that PBC may also moderate the influence of attitudes and subjective norms on intentions.^26,33 For example, attitudes and subjective norms may more strongly influence intentions among those who feel they can competently perform a behavior.²⁶ In line with the idea that TPB components may have multiplicative effects, studies examining predictors of marijuana use have reported a significant PBC × Attitude interaction,^34,35 such that PBC is a stronger predictor of intentions when attitudes toward marijuana are more positive. With respect to driving behaviors, both autonomy (e.g., control over engaging in DUIA) and capacity (e.g., how easy it would be to DUIA) components of PBC have been found to moderate the impact of attitude and subjective norms on intentions to DUIA.²⁸ For example, when capacity was high, both attitudes and subjective norms were stronger predictors of DUIA. Interestingly, perceived autonomy also moderated the relationship between subjective norms and intentions such that peer approval was less important when perceived autonomy was high.

Beyond the interactive effects of PBC, one might also expect subjective norms to moderate the influence of attitudes. Based on the contingency consistency model,³⁶ individuals may be more likely to engage in DUIC if they perceive their peers’ attitudes to be consistent with their own attitude. There has been limited research involving the interaction of attitudes and norms in the context of cannabis use and DUIC. Conner & McMillan³⁴ found that after controlling for habit, self-identity, and moral norms in addition to the core TPB components, attitudes did not moderate the relationship between injunctive norms (i.e., perceived peer approval) and intentions to use cannabis. To date, there has been a dearth of research exploring the contingency consistency model in the context of DUIC and more research is needed to understand moderating effects between the TPB components.

Sex differences and the TPB

Past research suggests that, compared to females, males use marijuana more frequently,³⁷ have more approving attitudes toward DUIC,³⁸ and engage in DUIC significantly more often.^11,13–15 What is not clear is whether sex moderates the influence of TPB components on intentions to DUIC and RWHD. With respect to marijuana use, LaBrie and colleagues³⁷ found that sex moderated the link between norms and cannabis use, with the relationship being stronger among male students. In one of the few TPB studies to examine sex differences in DUIA, Moan and Rise³² found that both normative perceptions and attitudes were more strongly related to intentions to DUIA among men than women.

In summary, no studies have yet applied the TPB to DUIC and RWHD, but extant research on related behaviors like DUIA and general cannabis use suggests that PBC and participant sex may interact with norms and attitude to predict intentions to DUIC and RWHD.

Current Study

As rates of cannabis use are expected to increase amidst recent legal reform, the present study utilizes the Theory of Planned Behavior to identify potential factors that predict college students’ intentions to DUIC and RWHD. In addition, we examine potential interactions between attitude, subjective norms, PBC, and sex. Based on previous research and theory,^26,28,39 we predicted that attitudes and subjective norms would be more strongly and positively related to intentions among male participants and among those reporting greater PBC. Understanding these relationships among TPB variables and intentions could inform the development of interventions and preventative efforts designed to reduce the frequency and negative impacts of DUIC and RWHD.

Methods

Participants and Procedure

Participants were 311 students recruited from the psychology subject pool at a midsized private university. The majority of the students were female (65.6%) and the mean age was 18.7 years (SD = 1.02). The sample was 57.8% Caucasian, 16.6% Asian, 10.0% multiracial, 4.2% Black or African American, and 11.4% “Other.” Sixty-eight participants (21.9%) identified as Hispanic or Latino(a). The sample’s demographics are reflective of the demographics of the institution as a whole. Lifetime marijuana use was reported by 165 participants (53.1%). All procedures and measures received approval from the host university’s Institutional Review Board. Students provided informed consent prior to participating and received class credit for their participation. Data was collected in September, 2013.

Measures

Past DUIC and RWHD behaviors were measured with items asking participants to indicate how many times during the previous three months they had “driven within 2 hours of using marijuana”^13,38 and ridden with a driver who they “knew had been using marijuana.”¹³

Subjective Norms and Attitude toward DUIC and RWHD were assessed using parallel items asking participants to report the extent to which their close friends (subjective norm) and themselves (attitude) approve of “Driving after using marijuana” and “Riding in a car with a driver who has been using marijuana.” The question wording and response options which ranged from 1 (unacceptable) to 7 (acceptable) were adapted from Lewis and colleagues’ injunctive norms measure for drinking behaviors.⁴⁰

Perceived Behavioral Control was measured with an item similar to those previously used to assess PBC for DUIA.^29,30 With regard to DUIC, participants were asked to rate the extent to which they believe that using cannabis “impairs my driving performance”. Responses were given on a scale from 1 (strongly disagree) to 7 (strongly agree) and were reverse-coded. Thus, high PBC scores indicate that participants expect to be able to DUIC without serious impairment. For RWHD the item asked participants to report how much control they feel they have over whether they “ride in a car with a driver who has been using marijuana” on a scale from 0 (no control) to 6 (complete control).

Intention to DUIC and RWHD was measured using items asking participants the extent to which they agree with the statements “I do not intend to drive after using marijuana” and “I do not intend to ride in a car with a driver who has been using marijuana”. Response options ranged from 1 (strongly disagree) to 7 (strongly agree) and scores were reverse-coded to create an intention score. Thus, a high intention score indicates that a participant did intend to DUIC and RWHD, while a low score indicates that he or she did not.

Data Analytic Plan

First, bivariate relationships among study variables were examined and sex differences were explored using t-tests. Next, the efficacy of the TPB in predicting intentions to DUIC and RWHD was assessed using two hierarchical linear regression models. At Step 1 in both models, student sex (female = 0, male = 1) and past DUIC or RWHD behavior were entered as covariates, followed by attitude, subjective norms, and PBC in Step 2. The 2-way interaction terms involving the TPB variables and sex were entered in Step 3. All continuous variables were mean centered before creating interaction terms. Simple slopes were plotted at the 25^th and 75^th percentile.

Results

Descriptives and Correlations

Descriptive statistics for study variables are presented in Table 1 with bivariate correlations in Table 2. T-tests revealed that males and females differed significantly on all study variables except for PBC for RWHD. Specifically, males reported committing more DUIC and RWHD in the past, being more approving of DUIC and RWHD, and perceiving their friends as more approving. In addition, males in the present study were less likely to feel that cannabis impairs their control of a vehicle and more likely to report that they intend to DUIC and RWHD in the future. Further, all DUIC variables were significantly associated with one another, as were almost all RWHD variables (PBC was not associated with past RWHD behavior).

Table 1.

Means and (SD’s) for all study variables.

	Females (N=204) Mean (SD)	Males (N=107) Mean (SD)	Overall Mean (SD)
1. Past DUIC Behavior	.99(3.71)	2.14(5.09)	1.39(4.26)*
2. PBC for DUIC	2.42(1.41)	2.83(1.58)	2.56(1.48)*
3. Attitude toward DUIC	1.47(1.65)	2.11(1.88)	1.69(1.76)**
4. DUIC Subjective Norms	2.28(2.03)	3.14(2.04)	2.58(2.07)***
5. Intentions to DUIC	2.28(1.68)	2.95(1.88)	2.51(1.78)**
6. Past RWHD Behavior	2.38(5.72)	5.08(9.07)	3.30(7.15)**
7. PBC for RWHD	5.59(1.35)	5.75(1.26)	5.65(1.32)
8. Attitude toward RWHD	1.58(1.73)	2.27(1.91)	1.82(1.82)***
9. RWHD Subjective Norms	2.31(2.05)	3.19(2.05)	2.61(2.09)***
10. Intentions to RWHD	2.56(1.83)	3.22(2.04)	2.79(1.93)**

Note. Degree of significance in the differences between males and females is noted in the “overall” column.

^*p < .05.

^**p < .01.

^***p < .001.

Table 2.

Bivariate correlations for all study variables.

	1	2	3	4	5	6	7	8	9	10	11
1. Sex	--
2. Past DUIC Behavior	.10	--
3. PBC for DUIC	.11	.43***	--
4. Attitude toward DUIC	.15**	.49***	.78***	--
5. DUIC Subjective Norms	.18**	.35***	.56***	.66***	--
6. Intentions to DUIC	.16**	.54***	.65***	.69***	.50***	--
7. Past RWHD Behavior	.19***	.64***	.49***	.50***	.49***	.55***	--
8. PBC for RWHD	.03	−.01	−.10	−.07	−.12*	−.10	−.06	--
9. Attitude toward RWHD	.16**	.43***	.76***	.92***	.69***	.64***	.53***	−.14*	--
10. RWHD Subjective Norms	.19***	.33***	.54***	.64***	.96***	.47***	.47***	−.14*	.69***	--
11. Intentions to RWHD	.16**	.41***	.66***	.68***	.58***	.79***	.53***	−.23***	.73***	.57***	--

^*p < .05.

^**p < .01.

^***p < .001.

Linear Regression Models

Results of the hierarchical linear regressions are presented in Table 2. Past behavior was a strong predictor of intentions in both models, with the first regression step explaining 31.0% and 28.3% of the variance in intentions to DUIC and RWHD respectively. The addition of the TPB variables in Step 2 significantly improved both models, explaining an additional 24.7% and 29.5% of the variance in intentions to DUIC and RWHD respectively. Importantly, PBC and attitudes were significantly and positively related to intentions in both models but subjective norms failed to reach statistical significance in either.

The addition of 2-way interactions in Step 3 significantly improved the DUIC model. A significant interaction revealed that PBC moderated the relationship between attitude and intentions (Figure 1), with attitude more strongly positively associated with intentions when PBC was high (b = .53, p < .001) compared to low (b = .22, p = .057).

Figure 1.

Relationship between attitudes toward DUIC and intentions to DUIC at high (+1 SD) and low (−1 SD) PBC.

With respect to RWHD, none of the 2-way interactions emerged as significant.

Comment

Results from this study indicate that cannabis use is still prevalent among college students; 53% of the sample reported lifetime use, 18.6% reported DUIC at least once in the past 3 months, and 43.9% reported RWHD at least once in the past 3 months. In a previous unpublished sample of nearly 1000 students from the same university we found a similar lifetime prevalence rate of 53.7%. Another study found an average lifetime prevalence of cannabis use of 53.3% across 11 different colleges.⁹ While there are well-established social sanctions against driving under the influence of alcohol, with one study finding rates of 4.4% for DUIA⁸, the current findings that nearly 1/5 of the participants engaged in DUIC in the past three months suggests that cannabis-related driving behaviors may not yet have a similar social sanction and appear to be a loci for intervention.

Regression analyses suggest that after controlling for past behavior, participants’ own attitudes toward driving under the influence of cannabis (DUIC) and riding with a high driver (RWHD) as well as their perceived behavioral control (PBC), or beliefs about the extent to which they maintain control of a vehicle while high or are able to refuse to ride with a high driver, significantly predicted their future intentions to DUIC and RWHD. Further, significant interactions suggested that the relationship between attitudes and intentions to DUIC was stronger for students higher in PBC. Additionally, and consistent with previous DUIA findings,^30,32 subjective norms were not a significant predictor of intentions to DUIC or RWHD after controlling for other TPB items in the models. While the measure we used for subjective norms is standard, it could be that there is a more sensitive measure that would have better captured the effect of normative beliefs.

Though this is the first TPB study to employ cannabis-related driving intentions as an outcome, some comparisons to previous findings are possible. For instance, Castanier and colleagues²⁸ reported that attitudes and subjective norms were stronger predictors of intentions to DUIA when PBC was high. The current findings suggest a similar trend for intentions to DUIC; among those with high PBC, attitudes were more strongly related to intentions (particularly in the case of females) and, among males, perceived norms were more strongly positively related to intentions for those higher in PBC. Further research is needed to explore the reasons for sex differences in these TPB relationships. The findings provide some evidence that individuals who believe they cannot control their driving ability while under the influence of cannabis are less likely to engage in this behavior—even if they approve of DUIC or believe others view it as acceptable.

Implications

The findings regarding participants’ perceived behavioral control offer researchers and universities with a promising area in which to intervene. That is, although several studies have documented that cannabis use can significantly reduce performance in driving tests⁴¹ and impair numerous driving-related abilities,^42,43 the majority of college students still believe their driving is, at most, only slightly impaired by cannabis.^44–46 Further, students report that they would not be deterred by TV campaigns about the effects of cannabis on driving because they feel they can compensate for these detrimental effects.⁴⁴ In fact, the majority of students report that even if they could be convinced that cannabis increases crash risk they would still continue to DUIC.¹⁴ The current findings, however, suggest students may be mistaken in their estimates of how they would behave if they could be convinced that cannabis use impairs their driving. Simple slope analysis revealed that when a student believes cannabis significantly impairs control of a vehicle and increases crash risk (PBC), then his or her attitudes toward DUIC may be largely irrelevant in predicting intentions. Thus, developing more effective ways to communicate how cannabis impairs one’s ability to drive (i.e., more effective ways to correct students’ PBC) appears to represent an optimal topic of focus for future prevention and intervention efforts.

For the second outcome, intentions to RWHD, the same direct relationships that were observed with DUIC (involving attitudes and PBC) also emerged. However, similar interactions were not found. One potential explanation for this may be the operationalization of the PBC variables used in our study. With respect to DUIC we utilized an item assessing participants’ efficacy for performing the behavior safely, which is thought to be a better predictor of behavioral intentions.¹⁹ For RWHD, on the other hand, this definition is not applicable and PBC was operationalized as perceived control over whether to perform or not perform the behavior in question, another widely-utilized definition. This variable was still found to predict unique variance in intentions to RWHD above and beyond other TPB and control variables. However, the interactions computed using this PBC variable did not significantly improve the model. This suggests that attitudes and PBC may both exert their influence on behavioral intentions individually in the context of RWHD and, therefore, may both be important intervention targets. However, these findings are preliminary and more research is clearly needed.

In addition to the prevention-related applications of the current findings, these results also have implications with regard to future TPB studies in various domains. Few studies examine the full complement of interactions involving TPB variables and participant sex as we did in the present study. This may mean that potential moderating effects of PBC variables are being overlooked.

Limitations and Directions for Future Research

First, it is important to note that this study used a convenient sample of 311 students from a subject pool. Therefore, the generalizability of our results may be limited. The current study did not explore whether the TPB variables predicted future cannabis-related driving behaviors. While much evidence suggests that intentions are a strong predictor of behavior,¹⁹ future researchers may wish to extend the current findings by adding a measure of future behavior to the models. This would also allow for the assessment of a direct path from PBC to behavior, which is predicted by the TPB¹⁸ and has been found in some studies.⁴⁷ Another limitation is that the intention measure may be conservative because participants may have felt uncomfortable disclosing their true intentions to DUIC and RWHD. Additionally, because data was collected prior to legalization of recreational cannabis use in California, it is possible that cannabis use and associated behaviors may now be greater, as observed in Washington and Colorado post-legalization.^2,6,7,48 More research is necessary to assess this relationship now that cannabis is legal recreationally in California. Lastly, future researchers may wish to apply other theoretical models to the prediction of cannabis-impaired driving outcomes. For instance, the Prototype Willingness Model⁴⁹ could be used to explore predictors of willingness to engage in unplanned DUIC / RWHD behaviors in addition to predictors of more deliberative intentions to engage in these behaviors. Because this model was specifically created to address the seemingly non-rational decision-making of adolescent health-risk behavior, it may be especially well-suited to the prediction of DUIC and RWHD.

Conclusion

Though DUIC has received less media and research attention than DUIA, there is evidence that cannabis users are more likely to drive under the influence than are alcohol users.⁸ As more U.S. states legalize and decriminalize recreational cannabis use, making cannabis more accessible to college students, it becomes increasingly important to investigate its potential risks and the factors that motivate behavioral decisions surrounding cannabis-related driving and riding. In the service of this interest, the current research used the TPB as an initial means for identifying predictors of these behaviors. Attitudes and PBC emerged as significant predictors of intentions and a significant interaction was found. These results carry important implications for researchers and universities working to both minimize cannabis use and prevent its associated risky driving-related behaviors.

Table 3.

Hierarchical linear regressions predicting intentions.

	Intentions to DUIC
	B	SE	p	R2	ΔF
Covariates				.310	68.683***
Sex	.101	.186	.035
Past DUIC Behavior	.534	.021	.000
Main Effects				.557	56.180***
PBC for DUIC	.284	.077	.000
Attitude toward DUIC	.362	.072	.000
DUIC Subjective Norms	.046	.045	.306
2-Way Interactions				.580	2.804*
PBC X Attitude	.105	.038	.006
Norms X PBC	.000	.048	.998
Attitude X Norms	.004	.038	.911
Sex X Attitude	−.167	.146	.254
Sex X PBC	.019	.163	.906
Sex X Norms	.080	.092	.382
	Intentions to RWHD
	B	SE	p	R2	ΔF
Covariates				.283	59.301***
Sex	.263	.203	.195
Past RWHD Behavior	.140	.013	.000
Main Effects				.578	69.701***
PBC for RWHD	−.184	.056	.001
Attitude toward RWHD	.597	.059	.000
RWHD Subjective Norms	.071	.049	.148
2-Way Interactions				.585	.792
PBC X Attitude	−.026	.038	.505
Norms X PBC	.052	.033	.119
Attitude X Norms	−.001	.024	.953
Sex X Attitude	−.034	.120	.774
Sex X PBC	−.123	.136	.367
Sex X Norms	−.092	.109	.400

^*p < .05.

^***p < .001.