Does ethnic identity moderate the relationship between racial discrimination and cannabis use among US- and non-US born Black Emerging Adults?

Abstract

Race-based experiences of discrimination (EOD) have been documented as a risk factor for substance use among Black individuals, particularly during emerging adulthood, with ethnic identity serving as a protective influence. Our study extends epidemiologic research on EOD and cannabis use by examining this relation in U.S. and non-U.S. born Black emerging adults across immigrant generations (N=466, 30% first-generation immigrants, 49% second-generation immigrants, and 21% non-immigrants). Results from self-reported data indicated EOD were associated with an increased likelihood of lifetime cannabis use, while ethnic identity was not significantly related to any odds of lifetime cannabis use. Odds of lifetime use was lower among first-generation immigrants compared to non-immigrants. Although the interaction between ethnic identity and EOD was not significantly associated with cannabis use, the results indicated that for second-generation immigrants, the probability of lifetime use decreased as ethnic identity increased. These findings underscore the importance of ethnic identity as a protective factor for cannabis use, especially among Black immigrants who have been racialized over generations in the United States, providing implications for future study and intervention.

Introduction

Nationwide trends in the 2019 National Survey on Drug Use and Health: African Americans revealed 26.9% of Black individuals (aged 18–25 years old) reported past month cannabis use and 6.5% reported a cannabis use disorder. Although Black adolescents have lower cannabis usage rates compared with Whites adolescents (Center for Behavioral Health Statistics and Quality, 2015; Johnson et al., 2015; Stock et al., 2011), rates become more similar during emerging adulthood (age 18–29 years old), as Black individuals are at a greater risk of initiating substance use during this time (Doherty et al., 2007; Hurd et al., 2014). The reasons for these race/ethnic differences during emerging adulthood are not fully known.

Race-based experiences of discrimination (EOD) have been implicated as impacting this change among racial/ethnic minority (REM) samples, including Latinx (Cano, 2020; Otinaiano et al., 2014; Viruell-Fuentes, 2007), Asian (Chae et al., 2008), and Black populations. Furthermore, immigrants have been shown to endorse EOD differently from their U.S.-born counterparts (Gee et al., 2006), with reported EOD increasing over time and with increasing generations in the U.S. (Finch et al., 2000; Goto et al., 2002; Read & Emerson, 2005; Waters, 1999). Particularly among Black individuals, while EOD is a significant source of stress across the life span (Gee et al., 2012), it is especially prevalent during emerging adulthood (Arnett & Brody, 2008; Lee et al., 2020). EOD has been associated with depressive and anxiety symptoms (Hurd et al., 2014) and substance use (Bennett et al., 2005; Borrell et al., 2007; Brody et al., 2012; Hurd et al., 2014; Lee et al., 2018; Martin et al., 2003) among Black emerging adults. As adolescents transition into adulthood, they are likely to depart from their familial context into more racially diverse neighborhoods, colleges/universities, and workplaces (Arnett & Brody, 2008). These transitions may increase exposure to EOD and its concomitant adverse consequences.

One factor that may modify this relationship is ethnic identity. In fact, it has been proposed that identity exploration is one of the most distinctive features of emerging adulthood (Arnett & Brody, 2008). Ethnic identity is a multifaceted construct encompassing an individual’s attachment to and knowledge of their ethnic group (Arsenault et al., 2018; Phinney & Ong 2007; Smith & Silva, 2011). It is said to contribute more significance to the overall identity and well-being of REM groups compared to their White counterparts (Phinney, 1996; Smith & Silva, 2011). For these reasons, strong ethnic identity is frequently conceptualized as a protective factor among immigrant and REM populations.

Rejection-Identification Model

Empirical research suggesting the protective influence of ethnic identity has been dominated by the rejection–identification model proposed by Branscombe et al. (1999). This argues that group-based discrimination damages well-being, but ingroup identification has the capacity to counteract these negative effects. In the face of opposition, a strong ethnic identity helps individuals recognize positive virtues about their own ethnic group, minimizing effects of denigrating beliefs perpetuated in society (Brittian et al., 2015; Kiang et al., 2006; Outten et al., 2009; Umaña-Taylor & Guimond, 2010).

Immigrants present an interesting case for rejection-identification as they are in a unique position of having to navigate two cultural milieus, the culture of their society of origin and the culture of their host society. Non-U.S. born immigrants’ ethnic identities are established in the country of origin and are likely to remain tied to that common history and birthplace. (Wiley, 2013; Wiley et al., 2013). In fact, several studies indicate that immigrants who perceive rejection from the host culture are less likely to develop positive relationships with the host society, instead turning increasingly toward ethnic identification (Badea et al., 2011; Jasinskaja-Lahti et al., 2009; Verkuyten & Yildiz, 2007).

This contrasts with the situation for non-immigrants, whose ethnic and national identities are established around the same time and carry similar investments. Without the country of origin as a frame of reference, ethnic identities are more likely to be influenced by what happens in the U.S. and how one is viewed (Branscombe et al., 1999; Postmes & Branscombe, 2002; Wiley et al., 2013). Thus, non-immigrants’ ethnic identities may be more susceptible to acceptance or rejection (e.g., via EOD) by majority group members of the society (Wiley, 2013).

Substance Use Patterns

Some studies have suggested that strong ethnic identity reduces the correlation between EOD and substance use (Brook et al., 2010; Love et al., 2006; Marsiglia et al., 2004; Pugh et al., 2007; Richman et al., 2013; Rowley et al., 1998). However, other studies have found opposite (Zamboanga et al., 2009) or inconsistent (Kulis et al., 2012) associations, illustrating the relationship’s complexity. Most of these studies do not factor in the importance of immigration status in these relationships. Research findings often indicate lower risk of substance use among first-generation immigrants relative to those of other immigrant generations (Blake et al., 2001; Di Cosmo et al., 2011; Gfroerer & Tan, 2003; Hamilton et al. 2009a; Hamilton et al. 2012; Hjern 2004; Pena et al., 2008; Vega et al., 1993; Tran et al., 2010).

The present study

The present study will address gaps in the literature by examining the relationships between EOD, ethnic identity, and both lifetime and recent cannabis use among an ethnically diverse population of first- and second-generation Black emerging adult immigrants and non-immigrants. The primary aim is to test the hypothesis that EOD will be positively associated with cannabis use and that this relation will be moderated by stronger ethnic identity and immigrant generation status. The secondary aim is to examine whether immigrant generation status is directly related to these variables (EOD, ethnic identity, cannabis use) and to test whether immigration generation status moderates the relation between ethnic identity and cannabis use.

Methods

Sample

The present sample was drawn from a larger study composed of 1799 individuals recruited from an urban public university system in the Northeast U.S. Inclusion criteria for the present study included age 18–29 years old, self-identifying as Black/African American/of African descent, and identifying within one of the following categories: (a) 1^st generation immigrant (participant is non-U.S. born); (b) 2^nd generation immigrant (one or more parent is non-U.S. born but participant is U.S. born); or (c) Non-immigrant (both parents and participant U.S. born). This resulted in a final sample of 466 Black undergraduate college students with 30.5% first-generation immigrants, 48.5% second-generation immigrants, and 21.0% non-immigrants. The sample was predominantly female (61.5%) with a mean age of 20.1 years (Table 1). Participants completed a computerized battery of self-report questionnaires in a research lab. The protocol was approved by the institutional review board from which students were recruited and written informed consent was obtained from all participants. All participants received course credit for their participation in the study.

Table 1.

Sample Characteristics and Descriptive Statistics

	Total (n=466, 100%)	1st generation immigrant (n=142, 30.5%)	2nd generation immigrant (n=226, 48.5%)	Non-immigrant (n=98, 21.0%)	F/x2
Age in years
mean (SD)	20.1 (2.3)	20.4 (2.7)	19.7 (2.1)	20.4 (2.5)	4.83*
Female gender
n (%)	318 (68.5%)	92 (64.8%)	158 (69.9%)	68 (70.8%)	1.36
Household Income n (%)
At or above $20,000 annually	351 (77.1%)	110 (79.1%)	176 (80.0%)	65 (67.7%)	6.18*
Lifetime Marijuana Usea
n (%)	153 (33.0%)	28 (19.9%)	83 (36.9%)	42 (43.3%)	17.19**
Recent Marijuana Use (past three months)a					6.93
No Use within past 3 months	82 (53.6%)	18 (64.3%)	46 (55.4%)	18 (42.9%)
1–2 times	37 (24.2%)	6 (21.4%)	20 (24.1%)	11 (26.2%)
3–5 times	23 (15.0%)	4 (14.3%)	12 (14.5%)	7 (16.7%)
More than 5 times	11 (7.2%)	0 (0%)	5 (6.0%)	6 (14.3%)
Dichotomization of Recent Marijuana Use (past three months)a					3.02
No Use or Little Use	119 (77.8%)	24 (85.7%)	66 (79.5%)	29 (69.0%)
Substantial Use	34 (22.2%)	4 (14.3%)	17 (20.5%)	13 (31.0%)
Experiences of Discrimination Domains a
mean (SD)	2.1 (1.9)	1.7 (1.9)	2.1 (1.8)	2.7 (1.9)	9.36**
0 domains	113 (24.2%)	53 (37.3%)	48 (21.2%)	12 (12.2%)
1 domain	90 (19.3%)	28 (19.7%)	49 (21.7%)	13 (13.3%)
2 domains	81 (17.4%)	22 (15.5%)	35 (15.5%)	24 (24.5%)
3 domains	73 (15.7%)	11 (7.7%)	45 (19.9%)	17 (17.3%)
4 or more domains	109 (23.4%)	28 (19.7%)	49 (21.7%)	32 (32.7%)
Ethnic Identity Scorea
mean (SD)	3.6 (0.84)	3.5 (0.88)	3.7 (0.79)	3.5 (0.86)	3.28*

^†p<0.1

^*p < .05

^**p<.001

^aAdjusted for age, gender, household income with non-immigrants as the reference group

Chi-squared tests used to calculate p values for categorical variables, ANOVA used to calculate p values for continuous variables.

Measures

Lifetime cannabis use.

To capture lifetime cannabis use, participants were asked “Have you ever used (e.g., smoked, eaten) marijuana/weed/pot” to which they replied “yes” or “no.” This item was derived from The Marijuana/Cannabis Use Questionnaire (MUQ; Brunswick, 1992), a 5-item self-report questionnaire for determining a respondent’s cannabis use. One item was used to capture lifetime cannabis use.

Frequency of recent cannabis use.

To assess frequency of recent use we used one item from the Drug Use Frequency questionnaire (DUF; O’Farrell et al., 2003): “How often did you use the following drugs or medications during the past 3 months?”. All participants, regardless of lifetime cannabis use, responded to this item using a 5-point Likert scale with values of (1) never, (2) once or twice, (3) several times a month, (4) several times a week, and (5) daily. Responses were dichotomized into none-to-little use (responses (1) never and (2) once or twice) vs. substantial use (responses (3) several times a month to (5) daily). This dichotomization was determined based on a similar process used in previous studies in non-clinical populations meant to capture meaningful degrees of usage (Green et al., 2006; Lynskey et al., 2003; Reeves et al., 2014).

Race-Based Experiences of Discrimination (EOD).

To assess experiences of racial discrimination, participants completed the Experiences of Discrimination scale (EOD; Krieger et al., 2005). This is a self-report questionnaire comprised of nine items that inquire about EOD due to race, ethnicity, or color across a variety of settings such as school, employment, and housing, as well as treatment from figures of authority such as police, medical professionals, and bank officials. The nine situations are tallied into a total racial discriminatory domain score (range 0–9) and frequency of occurrences for all nine situations is summed into a total score (0–27). This measure has been used to capture experiences in racially/ethnically diverse college samples (Rouhakhtar, et al., 2019; Sanchez & Awad; 2016) and has demonstrated strong psychometric properties including good internal consistency reliability (Cronbach’s alpha 0.74) and test–retest reliability (r=0.70) in a racially/ethnically diverse non-clinical sample of adults (Krieger et al., 2005). For the present analyses we used total domain score.

Ethnic identity.

Participants answered items from the Multigroup Ethnic Identity Measure-Revised (MEIM-R; Phinney & Ong, 2007), which is a widely used 6-item self-report measure identifying the individual’s level of exploration and commitment to their ethnic identity. Items are presented on a Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The measure is scored by averaging all six items with higher values representing a stronger ethnic identity. MEIM has previously been coded continuously to capture degrees of ethnic identity (Lin et al., 2019; Loyd et al., 2019; Williams et al., 2018). The psychometric properties of this measure have been confirmed in multiple studies, which have documented strong construct validity (r = 0.67–0.87; Homma et al., 2014), internal consistency estimates between 0.81–0.88 (Herrington et al., 2016) as well as measurement invariance across diverse groups (Brown, et al., 2013; Herrington et al., 2016), including REM (Phinney et al., 2007) and immigrant student samples (Homma et al., 2014).

Demographics.

Information on gender (binary, male=0, female =1), age (years), race, ethnicity, generational immigrant status (first- or second- generation, non-immigrant); and family household income poverty index (<$20,000) (United States Department of Health and Human Services, 2017) was collected for each participant.

Statistical Analysis

We verified the skewness and kurtosis coefficients were within the bounds suggested in the literature (West et al., 1995). Bivariate tests were used to examine relationships between demographic variables, EOD, ethnic identity, and cannabis usage. Chi-square tests of independence were used to determine relations between categorical variables. Pearson correlations assessed associations between continuous variables. Independent samples t-tests or analyses of variance (ANOVAs) with post hoc Bonferroni corrected t-tests were used to assess mean differences in EOD and ethnic identity between cannabis groups and across immigrant generation status.

Logistic regression models were used to determine the relation between EOD and lifetime cannabis use in analyses adjusted for demographics. To determine whether immigrant status and ethnic identity moderated relations between EOD and lifetime cannabis use, we created interaction terms for logistic regression models that were tested separately for EOD and immigrant status and EOD and ethnic identity. We also tested the interaction between ethnic identity and immigrant status in a separate model. Identical analyses were carried out with recent cannabis use frequency as the focal outcome. In all logistic regressions, no lifetime use, none-to-little cannabis use, and non-immigrants were reference groups.

Results

Descriptive and Bivariate Analyses

As shown in Table 1, ANOVA revealed a statistically significant age difference across immigrant generation (F[2,460]=4.83, p<0.01), with first-generation s significantly older than second-generation (mean difference=0.67, p<0.05). There were no statistically significant gender differences (X²[2, N = 464]=1.36, p=0.51). Chi-square analysis revealed a statistically significant difference in likelihood of family income below poverty (X²[2, N = 455]=6.18, p<0.05). A higher proportion of non-immigrants (32.6%) reported family income below poverty than second-generation (19.9%) (Table 1).

Bivariate analyses revealed gender differences in ethnic identity (F[1,464]=2.35, p=0.07) approached significance, while gender differences in EOD were not significant (F[1,464]=5.05, p=0.23). EOD was significantly different by poverty level (F[1,455]=5.29, p<0.05), with those below the poverty level reporting greater EOD domains than those above the poverty level (M=2.50, SD=1.81 vs. M=2.02, SD=1.85, respectively). Ethnic identity was also significantly different by poverty index (F[1,455]=5.52, p<0.05). Individuals above the poverty level reported stronger mean ethnic identity (M=3.6, SD=0.80) than those below the poverty level (M=3.4, SD=0.94).

Cannabis Use

One third of the total sample reported lifetime cannabis use (Table 1). There was a significant difference in lifetime cannabis use across immigrant status (X²[2,N=463]=17.19, p<0.001). Specifically, a lower proportion of first-generation immigrants (19.9%) reported lifetime use compared to both second- (36.9%) and non-immigrants (43.3%). Among the 153 lifetime users, there were no significant differences across immigrant status in frequency of recent use (X²[6,N=153]=6.93, p=0.33).

Experiences of Racial Discrimination (EOD)

A substantial proportion of participants reported EOD with approximately 75% of the total sample endorsing at least one domain (Table 1). Non-immigrants reported EOD in the greatest number of domains (M=2.7, SD=1.9) compared to first- (mean difference score 1.03, p<0.001) and second-generation immigrants (mean difference score 0.57, p<0.05). First-generation immigrants were more likely to report no EOD domains (37.3%), when compared to 21.2% and 12.2% of second-generation and non-immigrants, respectively.

Ethnic Identity

There were significant differences across immigrant status on ethnic identity scores (F[2,463]=3.28, p<0.05). Second-generation immigrants had strongest ethnic identity compared to first-generation (mean difference 0.19, p<0.1) and non-immigrants (mean difference 0.20, p=0.14). Pearson correlations revealed a significant positive relation between EOD and ethnic identity (r= 0.21, p< 0.001).

Discrimination, Ethnic Identity and Cannabis Use

Independent samples t-tests revealed a significant difference in mean number of EOD domains between lifetime cannabis users (M=2.75, SD=1.91) and non-users (M=1.82, SD=1.76), t(461)=−5.19, p<0.001. However, there was no significant difference in ethnic identity between lifetime cannabis users (M=3.6, SD=0.83) and non-users (M=3.6, SD=0.84). Independent samples t-tests revealed the difference between none-to-little cannabis use (M=2.70, SD=1.83) and substantial use (M=2.94, SD=2.19) was not significant, p=0.50. Similarly, the difference in ethnic identity between these groups was not significant, p=0.14.

Logistic Regression Analyses

The logistic regression models depicting lifetime cannabis use and frequency of recent use are shown in Tables 2 and 3, respectively. Odds of lifetime use was lower among first-generation immigrants (AdjOR of 0.38, 95% CI: 0.21–0.70) compared to non-immigrants. There was no significant difference in odds of use for second-generation and non-immigrants. EOD were associated with an increased likelihood of lifetime cannabis use (AdjOR of 1.32, 95% CI:1.17–1.50). This did not explain the relationship between immigrant status and lifetime cannabis use. Ethnic identity was not significantly related to any odds of lifetime cannabis use. Test of the two-way interaction between EOD and ethnic identity in association with lifetime cannabis use was not significant (OR of 0.98, 95% CI: 0.85–1.14). Tests of the two-way interaction between EOD and immigrant status were also not significant. Tests of the two-way interaction between immigrant status and ethnic identity on lifetime cannabis use approached significance (p=0.07) for the comparison between second-generation and non-immigrants. For second-generation immigrants, the probability of cannabis use decreased as ethnic identity increased. At higher degrees of ethnic identity, the difference in probability of lifetime cannabis use between second-generation and non-immigrants became more pronounced (Figure 1). The logistic regression models for frequency of recent use were not significant (Table 3).

Table 2.

Results of Logistic Regression Model for Lifetime Marijuana Use

	Model 1		Model 2		Model 3		Model 4		Model 5
Variables	B (SE)	p value	B (SE)	p value	B (SE)	p value	B (SE)	p value	B (SE)	p value
Dependent Variable = lifetime marijuana use
Step 1: Demographic Variables
Age	0.07 (0.04)	0.80
Gender	−0.36 (0.21)	0.09†
Income	0.10 (0.24)	0.67
Step 2: Immigrant Generation
Age			0.09 (0.04)	0.03*
Gender			−0.43 (0.22)	0.05*
Income			0.17 (0.25)	0.48
1st generation immigrant			−1.20 (0.30)	<0.001**
2nd generation immigrant			−0.26 (0.26)	0.30
Step 3: Experiences of Discrimination Domain
Age					0.04 (0.04)	0.35
Gender					−0.37 (0.22)	0.10†
Income					0.28 (0.25)	0.28
1st generation immigrant					−1.00 (0.31)	<0.001**
2nd generation immigrant					−0.15 (0.26)	0.57
Discrimination					0.25 (0.06)	<0.001**
Step 4: Ethnic Identity
Age							0.04 (0.04)	0.35
Gender							−0.34 (0.22)	0.13
Income							0.32 (0.26)	0.22
1st generation immigrant							−0.99 (0.31)	<0.01*
2nd generation immigrant							−0.11 (0.26)	0.67
Discrimination							0.27 (0.06)	<0.001**
Ethnic Identity							−0.17 (0.13)	0.20
Step 5: Moderation
Age									0.04 (0.04)	0.40
Gender									−0.33 (0.22)	0.14
Income									0.31 (0.26)	0.23
1st generation immigrant									−1.00 (0.31)	<0.001**
2nd generation immigrant									−0.14 (0.27)	0.61
Discrimination									0.27 (0.06)	<0.001**
Ethnic Identity									0.19 (0.26)	0.47
1st generation x Ethnic Identity									−0.31 (0.36)	0.39
2nd generation x Ethnic Identity									−0.56 (0.31)	0.07†
Constant	−0.51*		−0.06		−0.85*		−0.97*		0.93*
-2 Log-likelihood	574.408		554.119		536.017		534.373		531.049
X 2	6.34†		20.30**		18.09**		1.64		3.32

^†p<0.1

^*p < .05

^**p<.001

Note. For immigrant generation, non-immigrants were the reference group. For lifetime marijuana use, No use was the reference group. For frequency of recent marijuana use, None/Little Use was the reference group.

Table 3.

Results of Logistic Regression Model for Dichotomized Recent Marijuana Use

	Model 1		Model 2		Model 3		Model 4		Model 5
Variables	B (SE)	p value	B (SE)	p value	B (SE)	p value	B (SE)	p value	B (SE)	p value
Dependent Variable = dichotomization of recent marijuana use
Step 1: Demographic Variables
Age	0.06 (0.06)	0.30
Gender	−0.39 (0.40)	0.33
Income	−0.25 (0.46)	0.58
Step 2: Immigrant Generation
Age			0.07 (0.06)	0.25
Gender			−0.40 (0.40)	0.31
Income			−0.25 (0.46)	0.59
1st generation immigrant			−0.86 (0.60)	0.16
2nd generation immigrant			−0.47 (0.43)	0.27
Step 3: Experiences of Discrimination Domains
Age					0.07 (0.06)	0.29
Gender					−0.39 (0.40)	0.33
Income					−0.23 (0.46)	0.62
1st generation immigrant					−0.85 (0.61)	0.16
2nd generation immigrant					−0.46 (0.44)	0.29
Discrimination					0.03 (0.10)	0.79
Step 4: Ethnic Identity
Age							0.08 (0.07)	0.19
Gender							−0.25 (0.41)	0.55
Income							−0.24 (0.47)	0.60
1st generation immigrant							−0.92 (0.62)	0.14
2nd generation immigrant							−0.47 (0.44)	0.29
Discrimination							0.07 (0.11)	0.54
Ethnic Identity							−0.41 (0.25)	0.10
Step 5: Moderation
Age									0.09 (0.07)	0.18
Gender									−0.25 (0.41)	0.55
Income									−0.24 (0.47)	0.60
1st generation immigrant									−0.86 (0.62)	0.16
2nd generation immigrant									−0.46 (0.45)	0.30
Discrimination									0.07 (0.11)	0.55
Ethnic Identity									−0.53 (0.40)	0.19
1st generation x Ethnic Identity									0.34 (0.61)	0.58
2nd generation x Ethnic Identity									0.09 (0.55)	0.87
Constant	−0.79*		−0.39		−0.50		−.71		0.32
-2 Log-likelihood	164.969		162.632		162.560		159.871		159.553
X 2	2.51		2.34		0.07		2.69		−0.71

^†p<0.1

^*p< .05

^**p<.001

Note. For immigrant generation, non-immigrants were the reference group. For lifetime marijuana use, No use was the reference group. For frequency of recent marijuana use, None/Little Use was the reference group.

Figure 1.

Plot of logistic regression model for the interaction between ethnic identity and second generation immigrants vs. nonimmigrants.

Discussion

The current investigation explores the relationship between cannabis use and EOD, as well as the protective influence of ethnic identity, among U.S.-born and non-U.S. born Black emerging adults. Given the paucity of adequate research regarding the impact of EOD on cannabis use patterns of diverse REM groups (Gee, 2002; Williams et al., 2003), this study serves in describing this unique experience in a sample of Black immigrants and non-immigrants in which a substantial proportion endorse EOD.

Results indicate that first- and second-generation Black immigrants report less frequent EOD, stronger ethnic identity, and lower odds of lifetime cannabis use when compared to non-immigrants. There was no difference with frequency of recent cannabis use. However, there was a positive link between EOD and lifetime cannabis use, which has similarly been documented in prior studies (Finch et al., 2000; Noh et al., 1999; Williams et al., 2003). Furthermore, in line with the rejection-identification model, EOD were associated with stronger ethnic identity. Logistic regression indicated self-reported EOD were related to increased lifetime cannabis use, but these associations were not significantly moderated by ethnic identity.

The finding of lower cannabis use among first-generation immigrants was consistent with the ‘immigrant paradox,’ a well-documented phenomenon referring to findings that immigrants show better outcomes (e.g., psychological wellbeing, healthier behaviors) than their non-immigrant peers (Di Cosmo et al., 2011; Prado et al., 2012). However, there was no significant difference in odds of lifetime cannabis use between second-generation and non-immigrants. Studies have shown that the immigrant advantage deteriorates over time and across generations, suggesting a link to increasing contact with the majority population during the process of acculturation (O’Loughlin et al., 2010; Vega et al., 1993).

Our results suggest some advantages of the immigrant paradox seen in first-generation immigrants may be preserved in the second-generation through a strong connection with one’s ethnic group and culture. Among second-generation immigrants, the probability of cannabis use decreased as degree of ethnic identity increased. This suggests that ethnic identity may be protective against lifetime cannabis use, especially for second-generation immigrants. Although these findings are consistent with existing research depicting the protective influence of ethnic identity on substance use (Brook et al., 1998; Herd & Grube, 1996; Pugh & Bry, 2007), this study indicates that associations are not uniform across immigrant groups, which may reflect the dynamics of being racialized as an immigrant over generations in the U.S.

Furthermore, the study findings that EOD were associated with stronger ethnic identity lends support to the rejection-identification model (Branscombe et al., 1999). In accordance with this model, identification with one’s ingroup is an adaptive strategy when faced with rejection (Flores, 2000; Itzigsohn & Dore-Cabral, 2000; Outten et al., 2009; Waters, 1999). However, as suggested earlier, differences in immigrant status provides an interesting case for rejection-identification. Our results indicated that although second-generation immigrants, who have grown up in the U.S. but also often navigate a different home culture, had the strongest ethnic identity, they did not endorse the most EOD. Perhaps the factors driving a stronger commitment to one’s ethnic group for second-generation immigrants includes both an adaptive strategy to deal with the rejection of U.S.-based EOD and having some exposure to their parents’ home ethnic culture (e.g., spending summers in the country of origin with other extended family) (Benson, 2006; Waters, 1994; Wiley et al., 2012), making understanding their experiences a promising avenue for future research.

Lastly, findings from models in the study indicated that the interaction between ethnic identity and EOD was not significantly associated with lifetime cannabis use. Contrary to our hypotheses and prior studies (Brook et al., 2010; Chae et al., 2008; Love et al., 2006; Outten et al., 2009; Pugh et al., 2007; Richman et al., 2013; Rowley et al., 1998), ethnic identity was not found to be a significant moderator of the relationship between EOD and cannabis use. One potential explanation may be the lower prevalence and frequency of cannabis use in this sample, limiting our statistical power to capture this nuanced relationship. Another potential explanation may be the study location, a minority-majority university in New York City. Research shows that ethnic identity may be more salient in contexts where REM groups constitute a minority relative to Whites (Umaña-Taylor & Shin, 2007). Participants in this study were recruited from a university where Whites constitute less than 20% of the student body, and they mostly live in their own ethnic group neighborhoods. In such contexts, ethnic identity may not as strongly moderate associations between EOD and substance use patterns.

Strengths, Limitations, and Future Directions

Despite the novel nature of the study, there are limitations that should be considered. First, due to the cross-sectional design, temporality cannot be established. In addition, the process of forming an ethnic identity may be cyclical, meaning that new experiences can prompt individuals to further explore their ethnicity and make new commitments to their ethnic group (Brittian et al., 2015). In this case, prospective longitudinal research will be important for future studies. Another potential limitation is the reliance on self-report measures. Although personal perception is vital to characterizing these individual experiences, we must factor the possibility of underreporting and recall bias. Nonetheless, there are noteworthy strengths to the current study. Our findings reinforce the notion that patterns of cannabis use are not universal and warrant further investigation across ethnic groups, as cultural factors may shape responses to stressors such as EOD. Further longitudinal research utilizing a larger size is needed to separate the potentially complex ways in which societal and personal factors influence substance use behaviors.