Cross-Cultural Examination of College Marijuana Culture in Five Countries: Measurement Invariance of the Perceived Importance of Marijuana to the College Experience Scale

Marijuana internalized norms, measured by the Perceived Importance of Marijuana to the College Experience Scale (PIMCES; 8 items), has been found to be distinct from marijuana descriptive/injunctive norms and to be a unique robust predictor of marijuana-related outcomes among college students, yet the role of these beliefs has not been studied outside the U.S. Using confirmatory factor analysis, the present work examined the level of measurement invariance (i.e., configural, metric, and scalar) of the PIMCES across five different countries with distinct marijuana-related regulations (i.e., U.S., Argentina, Spain, Uruguay, and the Netherlands), sex, and marijuana user status among college students (n=3,424) recruited between September 2017 and January 2018. To make valid comparisons across groups, metric invariance is needed to compare correlations and scalar invariance is needed to compare latent means. We found strong measurement invariance (i.e., scalar invariance) for the PIMCES across countries, across males and females, and across marijuana users and non-users. College students in the U.S. reported the highest levels of marijuana internalized norms compared to college students from all other countries. As expected, males and marijuana users showed significantly higher scores on the PIMCES than females and non-users, respectively. Bivariate correlations between PIMCES scores and other marijuana-related variables were remarkably similar across males and females, though differences across countries warrant further exploration. Taken together, the degree to which college students view marijuana use to be an integral part of the college experience may be an important target for college student marijuana interventions across various countries/cultures.


Introduction
Normative perceptions have been demonstrated to be robust predictors of substance use outcomes. Although most of this research has focused on alcohol (Borsari & Carey, 2003), a burgeoning literature has revealed that normative perceptions play a similar role in predicting other substance use outcomes (Martens et al., 2006), including marijuana-related outcomes (i.e., marijuana use, marijuana-related consequences ;Napper, Kenney, Hummer, Fiorot, & LaBrie, 2016). In terms of normative perceptions, the literature has predominantly focused on two distinct norms: descriptive norms and injunctive norms (Borsari & Carey, 2001;2003;Cialdini, Kallgren, & Reno, 1991). In the case of marijuana, descriptive norms include estimations of others' frequency and/or quantity of marijuana use, whereas injunctive norms include perceptions regarding others' approval of specific marijuana use patterns/behaviors. Specific to the college population, Osberg et al. (2010) introduced the construct of alcohol internalized norms, or the degree to which college students perceive alcohol to be an important part of the college experience. Compared to descriptive and injunctive norms, alcohol internalized norms (as measured by the College Life Alcohol Salience Scale, CLASS; Osberg et al., 2010) have been shown to be a unique robust predictor of alcoholrelated outcomes among college students even when controlling for descriptive/injunctive norms (Hustad, Pearson, Neighbors, & Borsari, 2014;Osberg, Billingsley, Eggert, & Insana, 2012;Osberg, Insana, Eggert, & Billingsley, 2011;. Modelled after the CLASS, the Perceived Importance of Marijuana to the College Experience Scale (PIMCES) assesses marijuana internalized norms (Pearson, Kholodkov, Gray, & Marijuana Outcomes Study Team, 2017). Similar to what has been found with the CLASS, the PIMCES has been found to be distinct from marijuana descriptive/injunctive norms and to be a unique robust predictor of marijuana-related outcomes (i.e., marijuana user status, frequency of use, marijuana-related consequences) among U.S. college students, even when controlling for descriptive/injunctive norms .
Most research on marijuana norms has been conducted in the United States, and to date, no research has examined marijuana internalized norms outside of the United States.
Considering the wide range of distinct laws and regulations pertaining to marijuana across countries, we expect substantial differences across cultures in normative perceptions about marijuana. For example, marijuana is legal for adult use in Uruguay, but federally illegal throughout the United States. The status of marijuana in many countries (and even states within the United States) occupies a legal middle-ground between licit and illicit with varying levels of decriminalization (i.e., reducing criminal penalties for use and/or possession) and/or level of criminal enforcement (i.e., strict vs. laissez-faire). For example, although Spain, Belgium, and Uruguay all have non-profit "cannabis social clubs," they differ substantially in how they are organized and regulated as well as the methods and limits of distribution (Decorte et al., 2017). Given that prevalence rates do not differ across countries in a strictly linear fashion according to level of legal restrictions (Degenhardt et al., 2008;Simons-Morton, Pickett, Boyce, ter Bogt, & Vollebergh, 2010), it is important to conduct cross-cultural work to examine whether risk/protective factors operate similarly or different in distinct cultural milieu. For example, in both U.S.  and Argentine samples (Pilatti, Read, & Pautassi, 2017), injunctive norms have been shown to be associated with marijuana use frequency showing stronger associations for peers/friends' norms compared to parental norms. Although in separate studies, these norms seem to operate similarly, comparative studies are needed to make stronger conclusions. Similar to cross-cultural work conducted on alcohol internalized norms using the CLASS (Bravo et al., 2017, it is important to conduct cross-cultural research to determine the extent to which findings on marijuana internalized norms are culturally-specific or more universal. It is possible that marijuana internalized norms are a proximal risk factor that can help to explain cross-cultural differences on marijuana-related outcomes. The present study had three primary aims. First, we wanted to determine if the PIMCES can be used to accurately assess marijuana internalized norms across five distinct countries (U.S., Spain, Argentina, Uruguay, and the Netherlands), across males and females, and across marijuana users and non-users. Second, we sought to quantify mean level differences on the PIMCES to determine the degree to which college students perceive marijuana to be an important part of the college experience across these distinct countries, across males and females, and across marijuana users and non-users. Third, we wanted to examine the construct validity of the PIMCES to determine the degree to which the associations between marijuana internalized norms and marijuana-related outcomes are similar/different across countries and across males and females. Based on previous research , we expected the PIMCES to be measurement invariant across sex and user status; we expected males and marijuana users to report higher marijuana internalized norms compared to females and marijuana non-users, respectively; and we expected the associations between marijuana internalized norms and marijuana-related outcomes to be similar for males and females. In terms of cross-cultural comparisons, we did not have a strong basis for hypotheses, thus we considered these analyses to be exploratory.

Participants and Procedures
Participants were college students recruited from eight universities across five countries (U.S., Spain, Argentina, Uruguay, and the Netherlands) to participate in a longitudinal online survey study examining antecedents to marijuana outcomes crossculturally (Bravo et al., 2019). Among many cultural differences, these countries also have a wide range of distinct laws and regulations pertaining to marijuana. Participants were recruited for the baseline survey between September 2017 and January 2018. Although 3,482 students were recruited across sites at baseline, only baseline data from students that  Bravo et al., 2019). Study procedures were approved by the institutional review boards (or their international equivalent) at the participating universities.

Measurement Translation
The original English version of the PIMCES  was translated to Spanish and Dutch by native speakers that are also proficient in English. Then, members of the research team compared the original and translated versions and, after thorough discussion, composed a version of the instrument based on consensus (see Appendix A for Spanish/Dutch versions of the items). Several dimensions of marijuana consumption and use patterns (see below) were assessed via self-report questionnaires (mostly single-item questions) and were also translated into Spanish and Dutch in the same manner.

PIMCES.
Marijuana internalized norms were assessed using the 8-item PIMCES ) measured on a 5-point response scale (1=strongly disagree, 5=strongly agree). Previous psychometric testing has shown a single factor solution of the 8-item PIMCES to demonstrate good model fit (e.g., CFI=.967) in a large sample of college students from the U.S.  and excellent internal consistency (α=.892).  (Collins, Parks, & Marlatt, 1985). The MUG asks participants to indicate their frequency and quantity of marijuana use during a typical week. Each day of the week (Monday-Sunday) was broken down into six 4-hour blocks of time (24:00-04:00, 04:00-08:00, 0800-12:00, 12:00-16:00, 16:00-2000, 20:00-24:00), and participants entered the number of grams they consumed during each time block during a typical use week in the past 30 days. From this measure, we calculated both typical frequency (i.e., counting all non-zero values; hypothetical ranges: 0-42) and typical quantity (i.e., summing number of grams) of marijuana use for a typical week. Scores greater than 3 SDs above the mean for typical quantity were Winsorized to address outliers.
To assess subjective intoxication, participants were asked to indicate how high they get on a "typical marijuana use day" (typical subjective intoxication) on a visual analog scale (0=not at all to 100=completely) and reported the number of hours they typically "stay high" on a typical marijuana use day (length of typical intoxication).

Marijuana norms. All participants (regardless if they have ever used marijuana)
answered questions about marijuana norms. Marijuana descriptive norms were assessed using the same marijuana use frequency/quantity measure to assess one's own marijuana use (see MUG above); however, participants also estimated the frequency/quantity of use of their "close friends." Using a similar grid-based measure, marijuana injunctive norms were assessed by asking the amount of marijuana use that "close friends" would approve of in a typical week. Similar to the MUG, indicators of both typical quantity and frequency were calculated for each norm. Moreover, nine additional items (Pearson, Liese, Dvorak, & Marijuana Outcomes Study Team, 2017) were used to assess injunctive norms by asking one's perceptions about others' approval of specific marijuana use behaviors (using marijuana, using marijuana to get high, using marijuana daily) on a 7-point response scale (1=strongly disapproving to 7=strongly approving) for three reference groups (best friends, college students, parents). We averaged across the three behaviors to create composite scores for each reference group.
Negative marijuana-related consequences. Negative marijuana-related consequences were assessed using the 21-item Brief-MACQ (Simons et al., 2012) measured on a dichotomous response scale (0=no, 1=yes). By summing these responses, the total score reflects the total number of consequences that the individual had experienced in the past month. Previous psychometric testing has shown a single-factor solution to the B-MACQ to demonstrate good model fit (e.g., CFI=.96) and strong test-retest reliability (intra-class correlation=.80) in a sample of college student marijuana users in the U.S.
Further, the B-MACQ has been shown to be weakly correlated with alcohol use, moderately correlated with alcohol consequences and indicators of marijuana use, and strongly correlated with another measure of marijuana-related consequences (Simons et al., 2012). We have found the MACQ to be scalar invariant across sex and countries using the present sample (Bravo et al., 2019).

Statistical Analysis
To examine the internal structure of the PIMCES across sites and in the total sample, we conducted confirmatory factor analyses (CFA) using a diagonally weighted least squares (WLSMV) estimator in Mplus 7.4 (Muthén & Muthén, 1998-2018. To evaluate overall model fit, we used model fit criteria suggested by Marsh, Hau, and Wen (2004). Specifically, a Comparative Fit Index (CFI)/Tucker-Lewis Index (TLI)>.90 is considered acceptable and CFI/TLI>.95 is considered optimal; Root Mean Square Error of Approximation (RMSEA)<.06 is considered acceptable. Upon deciding on the best fitting model across all countries, we calculated Cronbach's alpha to test the internal consistency of the measure across sites. We conducted multi-group CFA to examine factorial invariance (configural, metric, and scalar) of the PIMCES across sex (males vs. females), lifetime user status (yes vs. no) and countries (i.e., U.S., Argentina, Uruguay, Spain, and the Netherlands). Metric invariance (equivalence of item-factor loadings) is necessary to compare associations between the PIMCES and other constructs across different groups, and scalar invariance (equivalence of item thresholds) is necessary to compare mean levels across groups. Since the χ 2 test statistic is sensitive to sample size (Brown, 2015), we used model comparison criteria of ΔCFI/ΔTFI≥.01 (Cheung & Rensvold, 2002) and ΔRMSEA≥.015 (Chen, 2007) to indicate significant decrement in fit when testing for measurement invariance. Finally, evidence of construct validity was assessed using correlation analyses among the PIMCES latent factor with marijuana use indicators and perceived norms.

CFA and Measurement Invariance of the PIMCES
The 8-item PIMCES showed adequate-to-excellent fit to the data on most indices for the total sample: χ 2 (20)=424.28, p<.001, CFI=.984, TLI=.977, RMSEA=.077 (90% CI [.071, .083]) and country subsamples (see Table 1). The standardized loadings (available from the authors upon request) were all salient (i.e., ≥.30; Brown, 2015). Internal consistency for the PIMCES was .87 in the total sample and ranged between .81-.88 across countries. Based on the findings reported above, measurement invariance testing was conducted and the PIMCES was found to be invariant across sex, the five countries, and lifetime user status (i.e., configural, metric, and scalar invariance was met; see Table 1). 1

Latent Mean Comparisons
Based on the results of our measurement invariance analyses, we conducted latent mean comparisons to test for latent score mean differences by country, lifetime user status, and sex (controlling for the effects of the other factors). We used dummy-coded indicators for country, lifetime user status (0=no, 1=yes) and sex (0= male, 1=female) as predictors of a latent factor of the PIMCES. A statistically significant result indicates a significant mean difference in the latent factor between the reference group and the predictor group (these results are available upon request).
Independent of country and lifetime user status, females reported significantly lower scores than males on the PIMCES (b=-.227, p<.001). Independent of country and sex, lifetime users reported significantly higher scores than non-users on the PIMCES (b=.510, p<.001). Using the U.S. as the reference group and controlling for sex and lifetime user 1 *There were no endorsement of all response options for items 2,5,7, and 13 in the Uruguay sample; thus, measurement invariance of the 8-item version was tested across the U.S., Argentina, Spain, and Netherlands. A 4-item version (no items 2,5,7, and 13) was tested across all five countries. Each of these analyses supported measurement invariance of the measure (see Table 1). status, we found that college students from all other countries reported significantly lower scores than U.S. college students on the PIMCES: Argentina (b=-.250, p<.001), Uruguay (b=-.220, p=.012), Spain (b=-.330, p<.001), and the Netherlands (b=-.394, p<.001). Of all possible mean differences across the other countries, there was only one significant latent mean difference such that Dutch students reported significantly higher scores on the PIMCES than Argentinian students (b=.144, p=.015).

Construct Validity
Bivariate correlations were conducted between the latent factor of the PIMCES and other marijuana-related variables among past month marijuana users (n=1117) across sex and across countries. In the analyses conducted across sex (see Table 2), the associations between PIMCES scores and marijuana-related variables were small-to-moderate across both males and females, though they tended to be slightly larger among females than males (average r diff =.03). The association between PIMCES and best friend injunctive norms was stronger among females (r=.30) than males (r=.14) and the association between PIMCES and typical frequency was stronger among females (r=.27) than males (r=.16). All other differences were |.08| or smaller (average |r diff |=.05, SD=.05). In the analyses conducted by country (see Table 3), larger differences emerged in the associations between PIMCES and other marijuana-related variables. Across 130 possible comparisons, we found that the average difference in correlations was |.15| (SD=.12). The largest correlation differences were between PIMCES and typical quantity injunctive norms in the Netherlands (r=.41) vs.

Discussion
In the first study of its kind, we examined marijuana internalized norms, or the degree to which college students view marijuana use to be an integral part of the college experience, among college students across five distinct countries (United States, Uruguay, Argentina, Spain, and the Netherlands) spanning three continents (North America, South America, and Europe) with three different languages (English, Spanish, and Dutch). We found strong measurement invariance (i.e., scalar invariance) for the PIMCES across these countries, suggesting that this measure taps into the same construct across these distinct cultures. Consistent with our hypotheses and with previous research , we also found the PIMCES to be scalar invariant across sex and user status, suggesting that this measure taps into the same construct across males and females as well as across marijuana users and non-users.
Also consistent with our hypotheses and previous research , we found that males reported higher marijuana internalized norms than females, and marijuana users reported higher marijuana internalized norms than non-users. The legal regulatory status of marijuana across these countries are complicated. In Uruguay, marijuana is legal for adult use, but frequent users have been found to circumvent the legal market (Boidi, Queirolo, & Cruz, 2016). In the Netherlands, Spain, and Argentina, marijuana is decriminalized though specific policies, availability, and prevalence rates differ across these countries (Decorte et al., 2017). Despite more legal restrictions regarding marijuana use in the United States compared to these other countries, college students in the United States reported the highest levels of marijuana internalized norms compared to college students in any of the other countries. Just like prevalence rates (Degenhardt et al., 2008), it is important to note that normative perceptions are not strictly (or even strongly) determined by legal restrictions.
In terms of the relationships between marijuana internalized norms and other marijuana-related constructs across males and females, we observed very similar associations, although these relationships tended to be slightly stronger among females suggesting that PIMCES has similar construct validity among males and females. In terms of the relationships between marijuana internalized norms and other marijuana-related constructs across countries, we observed a rather wide range of differences with an average correlation difference of .15. For example, the associations between marijuana internalized norms and typical quantity descriptive norms ranged from .04 to .14, the associations with past month marijuana use frequency ranged from .10 to .38, and the associations with negative consequences ranged from .18 to .41. The associations with many of these constructs included a mix of positive, negative, and/or non-significant associations. These findings could reflect real cultural differences in how internalized marijuana norms exert effects on outcomes, measurement non-invariance on the outcomes examined, or samplespecific differences. For example, it is plausible that college students in distinct countries differ in their ability to estimate marijuana use in grams based on varied formulations of marijuana products obtained across these countries, leading to large differences across countries in these associations. It must also be noted that the sample sizes of marijuana users in some of these countries were quite small (e.g., n=43 in Uruguay), suggesting that some of these differences could be spurious.

Practical and Clinical Implications
The PIMCES appears to have merit from a psychometric perspective as it demonstrates strong psychometric properties for males and females as well as marijuana users and non-users sampled from diverse countries/cultures. The patterns of associations between marijuana internalized norms with other marijuana norms measures, indicators of marijuana use, and marijuana-related negative consequences continues to support that these internalized norms reflect a construct that is unique from other marijuana norms and is one that plays a unique role in college students' decision whether and how to use marijuana.
Clearly, additional research is needed to examine how marijuana internalized norms develop over time and to determine what factors influence these perceptions. For example, alcohol internalized norms have been shown to be influenced by exposure to alcoholpromoting movies (Osberg et al., 2012).
Despite some differences across countries in terms of correlations with other measures, we found that marijuana internalized norms were robustly related to marijuanarelated outcomes across males and females and across the five countries included in the present sample. From a clinical perspective, these findings provide preliminary support for marijuana internalized norms as an intervention target. Although our findings provide limited information regarding specific strategies to modify these beliefs, it is possible that the efficacy of norm-based interventions targeting college student marijuana use (Lee et al., 2010) could be improved through targeting marijuana internalized norms. Perhaps such approaches would be efficacious across distinct cultural contexts, despite large differences in college/campus cultures.

Limitations
The present study had notable limitations that prevent our ability to make stronger inferences. Due to the cross-sectional/non-experimental design, we cannot determine whether marijuana internalized norms are causally related to the marijuana-related outcomes that we examined. Given our convenience sampling approach, we cannot be sure that our subsamples are representative of the college student populations in each country.
Thus, we do not wish to overinterpret differences found between countries as they could have resulted from small differences in sampling strategies (e.g., differential representativeness of college students in each country). Although our overall sample size was large, the sample sizes in some of our subsamples were rather small. Further, our selection of countries was largely opportunistic and did not include any countries in Africa, Asia, or Oceania. Taken together, longitudinal and experimental research that examines these associations prospectively in large, representative samples of college students in a wider selection of countries reflecting the full range of variability in marijuana-related policies, use, and norms would be ideal.

Conclusion
Our results are promising in that the PIMCES appears to assess marijuana internalized norms equally well among males and females, marijuana users and non-users, and across college students in the five countries examined. Now that we have Spanish and Dutch versions of the instrument, we can examine the psychometric properties and construct validity of the PIMCES in English-, Spanish-, and/or Dutch-speaking countries.
Given the limitations of the present study, our findings should be considered preliminary; however, our results support further examination of normative perceptions across cultures.
Characterizing cultural differences in normative perceptions and their influence on marijuana-related outcomes can inform the etiology of marijuana use as well as development and tailoring of norms-based interventions.  Chen, 2007) and ΔCFI/ΔTFI ≤.01 (decrease indicates worse fit; Cheung & Rensvold, 2002) to test for measurement invariance. *In the Uruguay sample, items 2, 5, 7 and 13 did not contain all values and thus measurement invariance was tested without these items. .26 Note. USA =United States (n = 696); AR=Argentina (n = 153), UY = Uruguay (n = 43), ES=Spain (n = 160), NL=Netherlands (n = 65). Due to no convergence in the Uruguay sample, all correlations were conducted with a MLR estimator. Medium correlation differences are italicized (.15 < r diff < .27), large differences are bolded (.28 < r diff < .39), and substantial difference are bolded and underlined (r diff > .40).