Self-Esteem, Cultural Identification and Substance Use Among American Indian Youth

Abstract

Objectives:

Determine potential mediating and moderating effects of self-esteem on the relationship between cultural identification and substance use among American Indian (AI) youth.

Methods:

Anonymous surveys were administered to middle and high school AI students assessing levels of last month alcohol and marijuana use, cultural identification, and self-esteem. Structural equation modeling assessed potential mediating and moderating effects of self-esteem on the relationship between AI and white identification on alcohol and marijuana use.

Results:

No mediating effects were found for middle school students. Among high school students, no mediating effects were found for alcohol. Among female high school students, a small mediating effect was found between AI identification and marijuana use, with AI identification protecting against use. Across all students, no moderating effects of self-esteem were found.

Conclusions:

The literature on cultural identification and substance use among minority youth is unclear. It was anticipated that consideration of self-esteem as a potential mediator or moderator might help clarify some of the confusion. However, no moderating effects were found and mediating effects were limited to female high school students for marijuana. Consideration should be given to assessing other measures of cultural identification and subscales of self-esteem to confirm that these findings are robust.

The role of cultural identification as a protective factor for substance use among minority populations is a confusing one at best. On the surface, it would appear that being more connected to one’s cultural heritage would reduce risk for problem behaviors such as substance use by enhancing focus on prosocial behaviors and lowering attention to negative behaviors. A number of authors have proposed such a relationship (Jones & Galliher, 2015). LaFromboise, Hoyt, Oliver, and Whitbeck (2006) suggested that cultural identification can serve as a key resilience factor, protecting American Indian (AI) youth from negative outcomes, and they found support for this hypothesis among upper Midwest middle school AI youth. Further support for the protective effects of cultural identification was found in a meta-analysis across various groups of racial minority youth (Rivas-Drake et al, 2014a, 2014b). Results indicated that higher levels of ethnic identification resulted in more favorable psychosocial adjustment and reduced health risks, including substance use. These findings are consistent with both social identity theory (Tajfel & Turner, 1986) and Phinney’s model of ethnic identity (1990) that link identification with one’s group to positive psychosocial outcomes. Tajfel and Turner (1986) argue that identification with one’s social group results in an increased sense of self-worth. Phinney (1990) supports this hypothesis, but also raises the question as to whether identification with a low status culture may diminish self-esteem. These theoretical perspectives suggest that self-esteem may be an important outcome of identity development.

There is also evidence that for minority youth, identification with the majority culture may provide benefits. As noted by Jones and Galliher (2007), AI youth are exposed not only to their own culture, but to the majority culture as well, and they may identify with both. In the original formulation of the Orthogonal Cultural Identification Scale (OCIS), the measure of ethnic identity used in this study, it was proposed that identification with either the minority or majority culture would provide benefits and protection (Oetting & Beauvais, 1990–1991). Support was found for this proposition among AI youth, for whom identification with both their own and white culture was associated with more perceived family caring, higher self-esteem, and better school adjustment (Oetting & Beauvais, 1990–1991). Galliher, Jones, and Dahl (2011) reported similar findings with the OCIS in which identification with the white majority culture among Navajo adolescents served to protect against the negative effects of discrimination.

While some studies support a negative relationship between identification with AI culture and substance use, other studies cast doubt on such a relationship. In a review, Beauvais (1998) concluded that the relationship between the two is not well supported. A number of empirical studies have reached a similar conclusion. Trimble (1996) found little relationship between cultural identification and alcohol use among a sample of AI adolescents. Baldwin, Brown, Wayment, Nez, & Brelsford (2011) reported no direct relationship between cultural identification and substance use among AI youth, but did find relationships of cultural identity to social support and protective family and peer influences. Galliher, et al. (2011) found differing relationships over a two year period among Navajo adolescents. At Time 1, higher levels of cultural identification were associated with better adaptive functioning including lower levels of substance use. But with control for Time 1 psychosocial functioning, the protective effects of cultural identification two years later were not significant.

These conflicting findings raise questions regarding our level of understanding of the role of cultural identification as protecting against substance use behaviors among AI youth. One of the assumptions underlying the relationship between cultural identification and positive outcomes is that greater connection to one’s culture will exert its protective effects, at least partially, through an enhanced sense of well-being such as improved self-esteem. Phinney (1990) was one of the first to identify this relationship, but she found it was more strongly present among those youth who were bicultural. Umaña-Taylor (2004), along with Lorenzo-Hernandez and Ouellette (1998), further supported the hypothesis that if ethnic group members evaluate their group positively, the result will be enhanced self-esteem. These studies suggest a possible mediating effect of cultural identification on substance use through self-esteem. Two studies have specifically evaluated the mediating role of self-esteem between ethnic identity and problem behaviors. Smokowski, Evans, Cotter, and Webber (2014) found in a sample of Lumbee, Caucasian, African American, and Latino adolescents, that self-esteem mediated the relationship between ethnic identity and anxiety, depressive, and externalizing behaviors. While their measure of externalizing behaviors did not include substance abuse, we suspect that the same mediating effect of self-esteem may operate on this variable as well. Additionally, Fisher, Zapolski, Sheehan, and Barnes-Najor (2017) found that self-esteem partially mediated between ethnic identity and substance use among multiracial youth from a large Midwestern city.

It is also possible that self-esteem serves a moderating role. That is, the effects of cultural identification on substance use will exert stronger effects when self-esteem is high. Higher self-esteem may be a necessary condition for cultural identification to provide its protective effects. But as noted, the literature remains unclear as to whether or not cultural identification reduces substance use risk. Examining the potential mediating and moderating roles of self-esteem in a study of cultural identification and substance use may help explain the conflicting results in the adolescent substance use literature. No study to date has examined this relationship in a population-based sample of reservation-based American Indian youth.

We conducted two studies, one evaluating whether self-esteem serves as a mediating effect between cultural identification and alcohol and marijuana use, and the other assessing whether self-esteem moderates the relationship between these variables. Alcohol and marijuana are the two substances used most frequently by adolescents (Miech, Johnston, O’Malley, Bachman, & Schulenberg, 2016), and among reservation AI adolescents, rates of these two substances are exceedingly high in comparison to national samples of adolescents as reported in Monitoring the Future (Stanley, Harness, Swaim, & Beauvais, 2014; Swaim & Stanley, 2018). Figure 1 presents the hypothesized model, with two measures of cultural identification (American Indian and white) related to last month alcohol and marijuana use through either mediating or moderating effects of self-esteem. In light of Phinney’s (1991) finding that self-esteem exerted its effects on positive outcomes more strongly among bi-cultural youth, measures of both AI and white identification were examined. We used the OCIS (Oetting, Swaim, & Chiarella, 1998) that, in contrast to bi-cultural measures, allowed us to consider the roles of white and AI identification independently. This measure asks students to report their identification, not only with their own cultural heritage, but also identification with the majority white culture. For both types of identification, the same item stems are used, but with reference either to American Indian or white culture.

Figure 1.

Hypothesized mediational model

Differing relationships between cultural identification, measured using the OCIS and self-esteem, have been found for males and females. For example, among a sample of Navajo adolescents, Jones and Galliher (2007) found that for males, AI identification was negatively related to self-esteem while white identification was positively related. Among females, both types of identification were positively related to self-esteem. Among Lumbee adolescents, Bryant and LaFromboise (2005) found no differences by gender, but all youth reported higher identification with AI as opposed to white culture. Phinney (1990) noted that females may identify more strongly with their culture than do males. Whitesell, Mitchell, Kaufman, Spicer, & Voices of Indian Teens Project Team (2006) also reported that among Northern Plains, Pueblo, and Southwest American Indian students, females reported higher levels of cultural identification.

Not only is ethnic identity likely to differ by gender, developmental level is also likely to alter its emergence and expression. Development of an ethnic identity falls within the larger context of development of the self-concept, key tasks that occur during childhood and adolescence. Concept of the self and identification with one’s own ethnic group in relation to other groups emerge over time, with changes in understanding and complexity occurring over the course of development. Gfellner and Armstrong (2011, 2012) reported that important differences in ethnic identity occur among Indigenous adolescents at varying developmental stages. Because of the potential variation in ethnic identity across gender and age, we conducted separate analyses by gender and middle school vs. high school students for tests of mediation and moderation.

This study adds to our understanding of cultural identity and adolescent substance use in several ways. First, it examines relationships between cultural identification and the use of alcohol and marijuana, the most commonly used substances among AI adolescents, in a large, national sample of AI youth. In addition, it investigates the potential mediating and moderating effects of self-esteem on these relationships, with separate tests for gender and grade group. No prior studies have included these factors in similar investigations.

We hypothesize that higher levels of AI and white identification will be associated with lower levels of each type of substance use. We further hypothesize that measures of cultural identification will be negatively related to alcohol and marijuana use through the mediating effects of self-esteem. Due to the lack of prior empirical findings, no specific hypotheses are offered based on mediating effects by gender or grade. These factors are investigated on an exploratory basis. However, among both genders, the protective effects of ethnic identification is hypothesized to be strengthened for those students with higher levels of self-esteem.

Method

This study is part of a larger ongoing investigation of the epidemiology and etiology of substance use among AI youth who live on or near reservations. Anonymous surveys of substance use and related factors were obtained from students in 46 schools during the 2009 to 2012 school years. The sample is a national sample of AI youth, with sampling conducted across five geographic regions where the majority of reservations are located (Northwest, Northern Plains, Upper Great Lakes, Southeast, and Southwest).

Participants

Seventh through twelfth grade students who self-identified as AI and who did not self-identify as white or other ethno-racial groups comprised the sample. All participants attended schools on or within 25 miles of a recognized AI reservation. According to our agreement with participants and schools, tribal affiliation and locations of the schools are kept confidential. The geographic distribution, based on cultural and other similarities of the AI population (Snipp, 2005) were Northwest (3.2%), Northern Plains (51.5%), Upper Great Lakes (11.2%), Southeast (3.6%), and Southwest (30.4%). Sample sizes consisted of 697 female and 767 male middle school (MS) students (7^th-8^th grades), and 975 female and 936 male high school (HS) students (9^th-12^th grades). For tests of both mediating and moderating effects, the sample was divided into MS and HS grade groups and by gender. Across all students, the mean age of students was 14.76.

Procedures

All data collection procedures were approved by the university IRB. Prior to all data collection, resolutions of support were obtained from participating tribal authorities or local school boards as required by each location. Parents/guardians were notified of the school survey through a parent notification letter and a broad media release, and were given the opportunity to remove their child from the project. Less than one percent of parents declined their child’s participation. Teachers or other school personnel trained in human subjects procedures through online or telephonic IRB training administered the surveys.

Prior to survey administration, students were informed that their participation was voluntary and that they could discontinue taking the survey at any time or could leave any item blank that they did not want to answer. All surveys were conducted anonymously and were completed during one class period. Those students whose parents opted them out of the study were given other school activities during survey administration. After surveys were completed, students placed their completed survey in a large envelope in random order. Once all surveys were placed in the envelope, the survey administrator sealed it for return to our offices.

Measures

Two measures of substance use were used as dependent variables – frequency of drinking in the last month and frequency of marijuana use in the last month, with 5 response categories for each measure (none (1), 1–2 times (2), 3–9 times (3), 10–19 times (4), and 20 or more times (5)).

Cultural identification was measured using the OCIS (Beauvais, 2006). In contrast to many bi-cultural measures that assume higher identification with the majority culture results in lower identification with a minority culture and vice versa, this measure allows for the independent assessment of both minority and majority identification. This offers multiple possibilities in the way cultural identification is measured including high identification with both AI and white culture, low identification on both, or high on one and low on the other. Measurement invariance has been established among AI students across middle and high school youth for this instrument (Oetting, et al., 1998). Six questions were asked that include how one’s family lives, how the student lives, family and personal success, and involvement in special activities and traditions. For each question, students responded about the degree to which they agreed (a lot, some, not much, not at all) according to an American Indian way of life or culture or a white American way of life or culture (AI Cronbach alpha = .91; white Cronbach alpha = .91).

Self-esteem was measured with an 8-item scale that has demonstrated good psychometric properties and association with substance use among adolescents (Oetting, Swaim, Edwards, & Beauvais, 1989; Swaim, Oetting, Edwards, & Beauvais, 1989; Swaim & Wayman, 2004; Cronbach alpha = .89). Survey items (e.g., I like myself, peers like me, I am able to do things well) were answered on a scale of 1 to 4 with response alternatives of “not at all,” “not much,” “some,” and “a lot.”

Analytic Approach

Descriptive statistics in the form of mean levels of measures and percentages of alcohol and marijuana last month are presented in Table 1. Two-factor (gender by age group) analyses of variance were run to determine mean differences in measures of identification, self-esteem, and last month alcohol and marijuana use.

Table 1.

Means (SD) for Total Sample and by Gender and Grade Levels

	AI Ident	White Ident	Self Esteem	Alc Last Month	Marj Last Month
Group
Total Sample	3.01(.87)	2.09(.93)	2.88(.53)	1.38(.74)	1.88(1.39)
MS Females	3.04(.81)	1.95(.86)	2.79(.52)	1.38(.69)	1.67(1.12)
MS Males	2.93(.95)	1.93(.93)	2.85(.60)	1.22(.59)	1.74(1.29)
HS Females	3.05(.84	2.23(.94)	2.91(.47)	1.46(.79)	1.96(1.43)
HS Males	3.02(.86)	2.17(.93)	2.93(.54)	1.44(.82)	2.07(1.57)

Note: AI Ident = American Indian Identification (mean of six items/6); White Ident = White Identification (mean of six items/6); Self Esteem (sum of 8 items/8); Alc = Alcohol; Marij = Marijuana; MS = grades 7 – 8; HS = grades 9 – 12.

Relationships between measures were assessed within an SEM format, with all analyses conducted using Mplus 8.0. The percentage of missing data ranged from 0 to 17.3%. Missing data were handled using the FIML feature in Mplus which permitted us to make use of all observations. Measurement models were evaluated prior to tests of mediating and moderating effects to assess the measurement quality of the two latent measures of cultural identification and the latent measure of self-esteem. Variances of the three latent variables were fixed to unity, with all latent factors freely correlated. In order to account for non-normality of measures, model parameters were estimated with the robust maximum likelihood estimator. Model fit was evaluated with RMSEA, CFI, and SRMR. Configural, scalar, and metric invariance (model form, factor loading, and item intercepts respectively) were tested across four groups (MS males, MS females, HS males, HS females) using the “Model Configural Scalar Metric” analysis command in Mplus which uses a Chi-square difference test using scaling correction factors. In addition, recommendations by Chen (2007) for changes in approximate fit statistics were used to supplement this information as Chi-square tests can be overly sensitive to measurement invariance in large samples. Next, structural models were assessed separately for mediating and moderating effects of self-esteem. For mediating effects, 95% bias-corrected bootstrap confidence intervals (MacKinnon et al., 2004; Shrout & Bolger, 2002) were estimated. In order to test for bicultural relationships with self-esteem and substance use, for each group, we computed a likelihood ratio test statistic to compare the model in Figure 1 with a model that also included the interaction of white identification and American Indian identification as a predictor of self-esteem and substance use. Because these tests showed no significant improvement in the model by adding the cultural identification interaction term into the model, we do not report these findings. For moderation, the XWITH option with TYPE=RANDOM was used to define the interactions between the self-esteem latent measure and the two latent measures of cultural identification. Moderation was tested using the likelihood ratio test and by examining the significance of the coefficients on the interaction terms.

Results

Table 1 presents means and standard deviations of the study variables for the entire sample and by gender/grade group along with percentages of alcohol and marijuana use in the last month. The two cultural identification scales and the self-esteem scale are summative measures divided by the number of items in the scale, yielding a range of 1 to 4. The alcohol and marijuana last month measures were single items that ranged from 1 to 5.

For AI identification there were significant effects for gender, F(1,3627) = 24.51, p < .001, with female students reporting higher AI identification than male students, and for age group, F(1, 3627) = 9.77, p = .002, with MS students reporting lower levels of AI identification than HS students. The interaction was not significant. For white identification a similar pattern of results was found with a significant effect for gender, F(1,3627) = 32.39, p <.001, with female students reporting higher white identification than male students, and a significant effect for grade level, F(1,3627) = 12.71, p <.001, with MS students reporting lower white identification than HS students. As for AI identification, the interaction was not significant. For self-esteem, the only significant effect was for gender, F(1,3627) = 28.50, p<.001, with female students reporting lower self-esteem than male students. Age group was the only significant effect for alcohol use in the last month, F(1,3627) = 14.38, p<.001, with HS students reporting more use than MS students. Similar results were found for marijuana use in the last month, with only age group significant, F(1,3627) = 5.41, p = .02. HS students reported more frequent use than MS students.

Measurement Models

Results of configural invariance were supported across the four groups based on a well-fitting configural model (CFI = .967; RSMEA = .043; SRMR = .040). The Chi-square difference test comparing the metric against the configural model was non-significant (Chi-square (45) = 47.52, p. = .37), indicating invariance of factor loadings across the four groups. Standardized factor loadings for the entire sample and for each group are presented in Table 2. The Chi-square difference test comparing the scalar against the metric model was significant (Chi-square (45) = 261.24, p < .001), indicating variance of item intercepts across the four groups. Examination of the item intercepts indicated differences across MS and HS students on four of the five indicator items for the white identification latent factor. Chen’s (2007) standards for assessing fit statistic differences for measurement invariance, reaching or falling below the following levels (ΔCFI =.002, ΔRMSEA = 0.006, and ΔSRMR = 0.002) were approached but not fully achieved for scalar invariance (CFI = .010; RMSEA = .004; SRMR = .004). These results suggest that partial measurement invariance was supported.

Table 2.

Standardized Factor Loadings (SE) for Total Sample and by Grade Level/Gender

	Total Sample	MS Males	HS Males	MS Females	HS Females
Factor
AI Identification
AI traditions	64(.02)	.65(.04)	.64(.02)	.60(.04)	.67(.04)
Family WOL	87(.01)	.88(.02)	.85(.02)	.80(.03)	.89(.01)
You WOL	.85(.01)	.87(.02)	.85(.01)	.82(.03)	.87(.01)
Family Success WOL	.84(.01)	.88(.01)	.85(.01)	.81(.02)	.82(.01)
You Success WOL	.85(.01)	.86(.02)	.88(.01)	.80(.02)	.85(.01)
White Identification
White traditions	.68(.02)	.66(.03)	.69(.03)	.67(.04)	.67(.02)
Family WOL	.86(.01)	.87(.02)	.84(.01)	.81(.02)	.90(.01)
You WOL	.87(.01)	.87(.02)	.88(.01)	.82(.03)	.90(.01)
Family Success WOL	.83(.01)	.83(.02)	.83(.02)	.81(.02)	.82(.02)
You Success WOL	.83(.01)	.86(.02)	.83(.02)	.79(.03)	.84(.01)
Self-Esteem
Proud of myself	.50(.03)	.57(.05)	.54(.06)	.48(.04)	.45(.03)
Able to do things well	.59(.02)	.67(.04)	.61(.05)	.45(.04)	.57(.02)
Peers like me	.77(.02)	77(.03)	.80(.02)	.78(.03)	.72(.05)
I am lucky	.59(.02)	.63(.04)	.60(.03)	.56(.04)	.55(.02)
Peers like to be with me	.70(.03)	.67(.04)	.74(.03)	.73(.05)	.64(.07)
I am good looking	.65(.01)	.68(.02)	.67(.02)	.63(.03)	.60(.03)
I am good at games	.50(.02)	.54(.05)	.57(.05)	.49(.04)	.43(.03)
Peers ask to do things with me	.52(.03)	.51(.05)	.60(.04)	.43(.05)	.50(.06)

Note. MS = Middle School; HS = High School; WOL = Way of Life

Structural Models

Standardized solution structural parameters are presented in Table 3 by gender/grade level. With the exception of HS female students, self-esteem was significantly and positively related to both AI and white identification. Last month alcohol use was not directly related to AI or white identification, nor to self-esteem. For marijuana use last month, white identification was associated with lower levels of last month marijuana use among both HS male and female students. The only direct relationship from measures of substance use to self-esteem was found among HS female students, for whom higher levels of marijuana use were associated with higher self-esteem.

Table 3.

Standardized Model Parameter Estimates (SE) for Total Sample and by Grade Level/Gender

	Total Sample	Middle School Male Students	High School Male Students	Middle School Female Students	High School Female Students
Parameter
Self-esteem on
Indian ID	.17***(.04)	.31***(.05)	.16**(.05)	.14**(.06)	.18***(.04)
White ID	.08*(.04)	.10*(.05)	.15***(.04)	.13**(.05)	.06(.04)
Alcohol on
Indian ID	−.03(.03)	−.03(.04)	.03(.04)	.02(.05)	.04(.04)
White ID	.01(.02)	.02(.04)	−.01(.04)	−.06(.05)	−.03(.04)
Self-esteem	.07(.05)	.01(.05)	.02(.05)	.06(.06)	.08(.04)
Marijuana on
Indian ID	.07(.04)	.07(.05)	.05(.03)	.05(.05)	.06(.04)
White ID	−.09(.05)	.03(.05)	−.11**(.04)	−.06(.04)	−.09*(.04)
Self-esteem	.08(.07)	−.05(.05)	.01(.04)	.06(.05)	.08*(.04)

Note. Indian ID = American Indian identification; White ID = white identification.

^*p<.05

^**p<.01

^***p<.001.

Mediating Effects.

Table 4 presents the direct, indirect, and total effects of measures of cultural identification (AI and white) on alcohol and marijuana use in the last month, for female and male students respectively. N mediating effects were found for either alcohol or marijuana among female MS students. For HS female students, no mediating effects were observed for alcohol use. However, AI identification and last month marijuana use were indirectly related through self-esteem (β = .014, bootstrap CI = .001, .03). The direct effect of AI identification on last month marijuana use was not significant, but the total effect was (β = .08, bootstrap CI = .01, .14). There was no mediating effect from white identification to last month marijuana use; however, the direct effect was significant (β = −.08, bootstrap CI = −.15, −.01).

Table 4.

Direct, Indirect, Total, and Bootstrap Confidence Interval Estimates for Mediation of Self-Esteem between Cultural Identification and Substance Use

	Middle School Female Students		High School Female Students
	β (BS 95% CI)	β (BS 95% CI)	β (BS 95% CI)	β (BS 95% CI)
Model	Alcohol	Marijuana	Alcohol	Marijuana
Direct
AI ID	.016 (−.073, .101)	.047 (−.048, .128)	.044 (−.026, .118)	.062 (−.010, .128)
White ID	−.062 (−.162, .028)	−.059 (−.138, .023)	−.033 (−.105, .037)	−.083 (−.152, −.009)*
Indirect
AI ID	.009 (−.006, .034)	.008 (−.004, .026)	.013 (.000, .031)	.014 (.001, .032)*
White ID	.008 (−.005, .034)	.007 (−.003, .030)	.005 (−.001, .016)	.005 (−.001, .018)
Total
AI ID	.024 (−.062, .107)	.055 (−.038, .136)	.058 (−.006, .131)	.077 (.009, .144)*
White ID	−.054 (−.140, .032)	−.052 (−.130, .026)	−.029 (−.096, .040)	−.078 (−.147, −.006)*
	Middle School Male Students		High School Male Students
	β (BS 95% CI)	β (BS 95% CI)	β (BS 95% CI)	β (BS 95% CI)
Model	Alcohol	Marijuana	Alcohol	Marijuana
Direct
AI ID	−.027 (−.111, .061)	.070 (−.020, .160)	.030 (−.045, .099)	.054 (−.011, .118)
White ID	.021 (−.058, .114)	.030 (−.059, .119)	−.003 (−.091, .083)	−.105 (−.183, −.039)*
Indirect
AI ID	.003 (−.026, .032)	−.016 (−.053, .012)	.003 (−.015, .020)	.000 (−.015, .015)
White ID	.001 (−.011, .015)	−.005 (−.025, .003)	.003 (−.016, .021)	.000 (−.014, .013)
Total
AI ID	.024 (−.102, .067)	.054 (−.031, .137)	.033 (−.044, .100)	.054 (−.009, .116)
White ID	.022 (−.059, .117)	.024 (−.063, .116)	.000 (−.083, .079)	−.105 (−.179, −.039)*

Note. BS CI = Bootstrap Confidence Interval; AI ID =American Indian identification; White ID = White identification

^*=p < .05.

There were no mediating effects for MS males, and no mediating effects for HS males. The only significant effect for HS males was a negative direct effect from white identification to last month marijuana use, (β = −.11, bootstrap CI = −.18, −.04).

Moderating Effects.

For all groups, the models with moderation did not differ significantly from the non-moderation models according to the likelihood ratio tests, and examination of estimates of interaction terms confirmed the lack of moderation for all groups. Results of Chi-square tests were as follows: MS females (Χ²(4) = 4.54, p = .34); MS males (Χ²(4) = 2.02, p = .73); HS females (Χ²(4) = 1.28, p = .86); HS males (Χ²(4) = 1.95, p = .74).

Discussion

A preliminary hypothesis of this study was that AI and white identification would be negatively related to last month use of alcohol and marijuana. We obtained only partial support for this hypothesis. First, no relationship was found between either measure of identification and substance use among MS students. This does not necessarily mean that culture is unimportant for middle school or younger youth. Albright and LaFromboise (2010) obtained similar findings among a sample of MS reservation students in which AI ethnic identification did not provide protective effects. While they found a significant negative relationship between white identification and hopelessness, no significant relationship was found for AI identification. We found no relationship to substance use for either AI or white identification among the younger students in our sample. Albright and LaFromboise (2010) interpreted the lack of relationship for AI identification to developmental level, suggesting that due to emergent cognitive and social development processes, the concept of cultural identity may not be fully active among MS youth. This does not imply that cultural processes are unimportant at this developmental stage, but that their influence is not yet fully available due to continuing cognitive maturation.

Another factor that may have suppressed any cultural effects among these younger students is social context. Quintana (2007) argues that exposure to racism may accelerate identity exploration. While all American Indians are at risk for being discriminated against, such exposure may be less likely among middle school reservation AI students who are more likely to attend tribal schools comprised largely of other AI youth, and may be less prevalent among reservation-based AI youth in general, compared to those living in urban locations who reside alongside many non-AI peers.

Among HS students, no relationships were found between alcohol use and either measure of cultural identification. For marijuana use, AI identification was not related, but white identification was negatively related to use among both male and female students. While we predicted that both AI and white identification would buffer against substance use, this was found only for white identification among the older students. The absolute value of coefficients for measures of substance use on AI identification was positive, but non-significant, suggesting that identification with their own culture could relate to higher marijuana use. This is not altogether surprising. Oetting and Beauvais (1990–1991) reported a similar finding in the original paper on Orthogonal Cultural Identification. Similarly, in regard to smoking among urban AI adults, Angstman, Harris, Golbeck, & Swaney (2009) found that AI identification increased risk while white identification decreased risk. While most authors have posited a beneficial influence of both in-group and out-group identification, in the case of American Indian youth who are exposed to very high rates of substance use by peers, identification with their own group may actually increase risk for substance use. The other finding of interest here is that the beneficial effects of white identification were found for marijuana, but not for alcohol. This may be due to the very high rate of marijuana use among reservation-based AI youth compared to white youth in the same locations (Stanley, Swaim, & Dieterich, 2017). If AI students are identifying with white culture, this identification may protect against marijuana use due to the much lower rates of use among white students. By comparison, rates of alcohol use are more similar between AI and white HS students (Swaim & Stanley, 2018). Thus, identification with white culture would not offer the same protection against alcohol as it may provide for marijuana. Additionally, Whitesell et al., (2014) found among Northern Plains AI adolescents, that cultural engagement was not associated with either alcohol or marijuana initiation or use trajectories over a one-year period.

Identification with either AI or white culture related positively to self-esteem, except among HS female students. In general, AI identification appeared to be more strongly related to self-esteem than white identification, and perhaps the strongest among MS male students. These findings are consistent with prior findings reported by both Umaña-Taylor (2004) and Lorenzo-Hernandez and Ouellette (1998) who found a positive relationship between self-esteem and ethnic identification in other minority groups. The primary purpose of this study was to assess potential mediating and moderating effects of self-esteem on the relationship between cultural identification and two measures of substance use (alcohol and marijuana use) among AI students. Given the conflicting nature of the literature regarding cultural identification and substance use, we anticipated that these analyses might help clarify some of disparate findings reported by others.

Among MS students, this was not supported for either female or male students. This is in contrast to findings reported by Zamboanga, Schwartz, Jarvis, and Van Tyne (2009). Among a similarly aged sample of Hispanic youth, they found that self-esteem mediated between cultural orientation and alcohol use. The difference between our sample of AI youth and other studies with Hispanic youth raises questions as whether these relationships may differ across other cultural groups.

Another important consideration is the degree to which self-esteem may be perceived differently among American Indian youth. In a meta-analytic study of race and self-esteem across various cultural groups, Twenge and Crocker (2002) noted that the self-esteem construct may differ among American Indians who tend to be more collectivistic than individualistic.

Among HS female students, self-esteem provided a small mediating effect between AI identification and marijuana use. AI identification increased the likelihood of marijuana use, and self-esteem enhanced this relationship to a small degree. White identification provided a direct protective effect against marijuana use among HS females, with no mediating effect for self-esteem. A similar relationship was found among HS males, with white identification protecting against marijuana use, but no mediating effect for self-esteem. For all groups, no case can be made for the moderating effects of self-esteem. We noted at the beginning of this paper that the role of cultural identification in relation to adolescent substance use is unclear and marked by inconsistency in the literature. The same is true for self-esteem (Swaim & Wayman, 2004), with some studies supporting a negative relationship, and others finding a positive relationship to substance use. Some studies have attempted to clarify this inconsistency, noting that the direction of effects may depend on the form of self-esteem that is measured. For example, among a sample of both MS and HS students, Wild, Flisher, Bhana, and Lombard (2004) found that positive global and body image self-esteem predicted protection from risk behaviors in female students while positive peer self-esteem was associated with increased risk for both male and female students. Our study measured self-esteem as a global construct, but previous research among Mexican American adolescents has demonstrated differential effects by separate self-esteem dimensions (Swaim & Wayman, 2004). Among Mexican American female students, the authors found that one component of self-esteem, self-confidence, protected against alcohol use, while other components (perceived competence and social acceptance) did not. Among Mexican-American male students, no relationship between self-esteem dimensions and alcohol use was observed. In this paper, we sought to first examine potential relationships for global self-esteem. Follow-up study with AI youth, considering various components of self-esteem within a prospective model may produce differing findings, both for mediating and moderating relationships.

Overall, the results reported here indicate that AI cultural identification did not relate to alcohol or marijuana use at the developmental stage of MS. Some protection against marijuana use was found among HS females and males who identified with white culture, but level of self-esteem did not alter this relationship. The lack of relationship for MS students may be due to this construct not reaching full development by this age level. Among both MS and HS students, no relationship was found between cultural identification and alcohol use.

Therefore, the roles of cultural identification and self-esteem, as they relate to AI adolescents substance use, remain somewhat elusive. While the measure of cultural identification utilized here (Orthogonal Cultural Identification; Oetting & Beauvais, 1990–1991) has been widely used in the literature, a number of other measurement approaches to cultural identification are available (Kulis, Robbins, Baker, Denetsosie, & Parkhurst, 2015; Schwartz, Zamboanga, & Weisskirch, 2008) that may capture different components of this construct and may relate differently to substance use. Examination of the role of self-esteem, cultural identification, and substance use with these measures should be considered in future research.

Limitations

This study has several limitations. First, although it used a large sample of AI adolescents living on or near reservations, it does not reflect a random sample of all schools on or near reservations due to voluntary participation, and thus, the results may differ from those obtained from the entire population of AI youth living in these locations. However, this sample does represent the largest sample of AI adolescents living on or near reservations. Second, because the survey took place in schools, the sample did not include adolescents who have dropped out of school. With as many as 28 percent of AI students dropping out of school (National Center for Education Statistics, 2018), the study may underestimate relationships between substance use and cultural identification due to less variance in the sample’s measures. Third, because the sample is cross-sectional in nature, we were not able to assess temporal ordering in the tests of mediation, and could not infer causality between variables where significant relationships were found. The results suggest possible mediating effects of self-esteem, but will require a prospective design in which temporal sequencing is present. Fourth, the data were based on self-report. However, self-report is the best method for measuring cultural identification and self-esteem, and for the substance use variables, self-report is the most feasible method for collecting substance use rates for a large sample of adolescents where physical markers for substance use are infeasible, and parents, teachers, and others may have inaccurate knowledge about the substance use of the adolescents. However, responses from some students may still include socially desirable and other response biases. Finally, scalar invariance was not obtained with the latent factor for white identification contributing to differences in item intercepts for this factor. It is important to note that scalar invariance is rarely obtained (Vandenberg & Lance, 2000), and has been seen as “an unachievable ideal that in practice can only be approximated (Marsh et al., 2017). It is likely that the failure to achieve scalar invariance for white identification among MS students was due to their limited exposure to white culture at this age level. Most AI students from reservations attend tribal middle schools where nearly all students are native. As they move into high school, many attend public schools where they are now exposed to other white students and white culture. This would help explain the differences in intercepts for white culture between these two age groups. Future research should investigate how mean levels of identification with the majority culture may change over the developmental period of adolescence among reservation-based AI students.