Discrimination, Marginalization, Belonging and Mental Health among Somali Immigrants in North America

Abstract

In this paper, we examine the relationships among discrimination and mental health for Somali young adults, a group at risk for an unfavorable context of reception, and the way in which individual and community-level factors explain these associations. The present study draws upon data collected during the first wave of the Somali Youth Longitudinal Study, a community-based participatory research project focused on understanding and supporting the healthy development of Somali young adults in four different regions in North America: Boston, MA, Minneapolis, MN, and Portland/Lewiston, ME in the U.S. and Toronto, Canada. Somali men and women, aged 18–30, participated in quantitative interviews that included questions about their health, their neighborhoods, and their thoughts and feelings about their resettlement communities (N=439). Results indicate that discrimination has a direct effect on worse mental health; this effect was mediated through both individual (marginalized acculturation style) and community-level (sense of belonging) factors. These findings suggest that factors associated with a receiving society’s attitudes and behaviors, in addition to its structural supports and constraints, may be particularly important in understanding immigrant mental health.

Health and wellbeing of immigrants in North America is understood to be a function not only of individual characteristics, but also the context of reception, or the receiving societies’ opportunity structures, attitudes and behaviors towards immigrants (Schwartz, Unger, Zamboanga, & Szapocznik, 2010; Viruell-Fuentes, 2007). Receiving societies expose some immigrant groups to high levels of discrimination and marginalization for reasons related to race, religion, ethnicity or simply being an immigrant (Schwartz, Unger, Zamboanga, & Szapocznik, 2010). Current national attention to issues of immigration has heightened the need to examine experiences of discrimination among immigrants as evidence suggests that there has been a recent surge in anti-Muslim hate crimes (Southern Poverty Law Center, 2017). Exclusionary policies aimed at restricting the presence of immigrants also contribute to hostile climates toward newcomers and their children, leading to social isolation, increased discrimination, and worsening health (Perreira and Pedroza 2019).

These experiences of discrimination have robustly been linked to worse health outcomes (American Psychological Association, 2016; Williams & Mohammed, 2009; Williams, Neighbors, & Jackson, 2003; Siddiqi, Shahidi, Ramraj, & Williams, 2017), including poor mental health (Schmitt, Branscombe, Postmes, & Garcia, 2014; Pascoe & Richman, 2009). Additionally, some immigrant groups may be marginalized to neighborhoods that lack safety and collective efficacy (Portes & Rumbaut, 2006), another potent risk factor for poor mental health. In contrast, when the local community provides opportunities for social support, connection, and inclusion, it can facilitate positive adjustment and well-being post resettlement (see Green, Rhodes, Hirsch, Suarez-Orozco, & Camic 2008; Suárez-Orozco, Suárez-Orozco, & Todorova, 2008). Finally, we know that these experiences may be attenuated by individual factors such as gender, acculturation styles, and length of time since migration can also influence how discrimination is experienced and how it relates to mental health (e.g. Ellis et al., 2010; Viruell-Fuentes, 2007). In particular, marginalized acculturation styles often characterized by both low levels of maintenance of heritage culture and identity; along with few connections or relationships with the host culture have been found to worsen PTSD symptoms among Somali adolescents (Lincoln et al., 2016). Increasing our understanding of the relationships among individual, neighborhood, and societal risk and protective factors is needed to inform policies and practices that promote immigrant mental health.

The relationships among discrimination and poor mental health outcomes are well-documented (Umana-Taylor, 2015; Brittian et al., 2015; Williams, Neighbors & Jackson, 2003). In a meta-analysis of 293 studies that examined racial discrimination and health, discrimination was robustly linked to poorer mental and physical health (Paradies et al., 2015). These findings have been replicated in studies examining the impact of discrimination on health among diverse immigrant populations and found consistent associations among discrimination and poor mental health (Viruell-Fuentes,Miranda, & Abdulrahim, 2012; Bernstein et al., 2011; Yip, Gee & Takeuchi, 2008). While the majority of studies have focused on discrimination associated with race, these experiences are also associated with factors including religion and ethnicity (Laird, Amer, Barnett & Barnes, 2007; Johnston & Lordan, 2012; Brittian et al., 2015).

A social ecological framework, informed by models of risk and resilience (Masten & Coatsworth, 1998), suggests that factors at the individual, inter-personal, and community levels may be important levers for interrupting the impact of experiences of discrimination on mental health among immigrants. For example, both aspects of ethnic identity (i.e., an individual’s self-label as a member of an ethnic group) and residing in ethnic enclaves, a geographic area where a large concentration of residents are from a similar ethnic background, can operate as protective factors, buffering the individual from the negative impact of discrimination (Rumbaut, 1994; Phinney, 1990; Mossakowski, 2003; Umana-Taylor, 2015; Brittian et al., 2015).

Individual level factors including gender, length of time in resettlement, generational status, ethnic identity and acculturation processes (i.e., the processes which one’s values and customs were changed based on the influence of another culture), can shape the way in which discrimination relates to health. Although some studies have found gender to moderate the relationship between discrimination and anxiety (Banks, Kohn-Wood, & Spencer, 2006), others have found that the intersection of race and gender in particular contributes to poor mental health outcomes, including depression and anxiety (Perry, Harp, & Oser, 2013; Borrell, Kiefe, Williams, Diez-Roux, & Gordon-Larsen, 2006). Assari and Lankarani (2017) note in their findings a stronger association between perceived discrimination and psychological distress for Arab American males than females. In a study of Somali immigrant adolescents, both acculturation and gender moderated the effects of discrimination on mental health outcomes (Ellis et al., 2010); the moderation was evident by the fact that Somali girls who more actively participated in Somali cultural activities reported less symptomology than those who endorsed active participation in American cultural activities. In contrast, for boys, greater participation in American culture was associated with better mental health status.

Length of time in resettlement as well as generational status, may further shape the experiences of discrimination and mental health. In recent years, research has demonstrated that an increasing length of stay in the U.S. may have a detrimental impact on the mental health of immigrants, increasing the risk of depression, anxiety, and PTSD (Breslau et al., 2007; Guajardo, Slewa-Younan, Smith, Eagar, & Stone, 2016; Gee, Ro, Shariff-Marco, & Chae, 2009; Marshall, Schell, Elliott, Berthold, & Chun, 2005). This decline in mental health status appears to be related to experiences of perceived discrimination and “othering” in host countries (de Maio & Kemp 2010; Viruell-Fuentes, 2007). There are also important distinctions between how first and second generation immigrants form ethnic identity and perceive discrimination. A study of 40 first and second-generation Mexican immigrant women in Detroit found that while first-generation immigrants reported very little discrimination, second generation immigrants reported experiencing the cumulative effects of being an outsider and of witnessing their families and peers experience discrimination, as well as a heightened awareness of “othering” messages (Viruell-Fuentes, 2007).

Researchers have also explored the influence of neighborhood characteristics on the relationship between discrimination and mental health. For instance, in a study of newly arrived immigrants, longitudinal analyses revealed the buffering effects of neighborhood collective efficacy on the relationship between perceived discrimination and depressive symptoms (Chou, 2012). Neighborhood characteristics have also been found to mediate the association of discrimination on mental health such that perceived discrimination negatively influences neighborhood social capital, which in turn hinders self-reported mental health (Chen & Yang, 2014).

Somali Immigrants: Multiple Marginalized Identities

The experience of Somali immigrants provides one important example of the complex interplay between immigration, experiences of discrimination, acculturation, neighborhood characteristics, and well-being. Significant numbers of Somalis first arrived in the US in the early 1990s. As Black Muslim immigrants, Somalis face ‘triple jeopardy’ of being marginalized because of race, religion and immigration status (Ellis et al., 2010). As one of our study participants stated,

“But when you go outside, that’s when the racism hits you. Boom. You get slapped right in the face. They’re not racist like back in the day racist, like we’re going to lynch you or whatever. It’s more of a systematic racism, you know what I’m saying?”

26 year old, male.

Somalis are Black and predominantly Muslim; as immigrants of minority racial and religious groups, they identify with multiply marginalized identities and experience the discrimination related to these intersecting identities. Compared to other immigrant groups, Somali Americans report higher levels of perceived housing discrimination (Dion, 2001). Somali Americans also have the highest unemployment rate among East African immigrants in the U.S. and the lowest rate of college graduation (Diamant, 2017).

The context of reception (i.e., the situation or circumstances in which immigrants are received by their resettlement communities) experienced by Somalis in North America is further complicated by societal hostility towards Muslims. Experiences of exclusion and prejudice are unfortunately becoming characteristic of the Somali immigrant experience in North America, in contrast to earlier Muslim immigrants who migrated to the region (Pew Research Center, 2011). Within the United States, anti-Muslim hate crimes have been increasing (Kishi, 2017), and within Canada, the number of hate crimes targeted at Muslims increased by more than 10-fold the year after September 11^th, 2001 and remained elevated more than 10 years later (Allen, 2015; Helly, 2004; Perry, 2014).

Young Immigrants

Adolescence and early adulthood represent a unique developmental period in which the detrimental impacts of discrimination may be magnified by identity formation processes occurring during this time period (Brenner, Crosnoe, & Eccles., 2015; Resnick et al., 1997). Among young adult Blacks and Latinos, recent studies examining ethnic discrimination have demonstrated associations with increased rates of depressive symptoms (Brittian et al., 2015). For immigrant adolescents and young adults, the toll taken by the chronic stress of exposure to discrimination interacts with the formation of ethnic identity within a new culture as seen among Mossakowski’s studies of 1^st and 2^nd generation Filipino Americans (Mossakowski, 2003). Several studies have noted that a strong ethnic identity may heighten awareness in young people in particular to subtle forms of discrimination (Operario & Fiske, 2001; Yip, Gee & Takeuchi, 2008; Shelton & Sellers, 2000). Finally, there is evidence of the differential impact of sources of discrimination on the mental health of adolescents. In one study of Latino adolescents, differences in the impact of discrimination within schools, from adults outside of schools, and within on-line communities on mental health were found (Umana-Taylor, 2015). In our own work (Ellis et al., 2010), with Somali adolescents we found that experiences of discrimination were highly prevalent and that both acculturation and gender moderated the strong effects of discrimination on mental health outcomes.

Strong cultural networks have also been identified as fostering resilience among young adult immigrants (Akinsulure-Smith 2017; Ellis et al. 2010). Ethnic/racial social support has been found to offset the effects of discrimination on mental health and psychological distress (Ajrouch, Reisine, Lim, Sohn, & Ismail, 2010; Elligan & Utsey, 1999; Finch & Vega, 2003). Other studies, however, have found limited evidence for this buffering effect, suggesting inconsistencies in the measurement of social support and in quantitative vs. qualitative report of this relationship (see Brondolo, Gallo, & Myers, 2009). Attention to the severity of environmental stressors is of further importance; specifically, although social support is a significant buffer among those experiencing moderately high levels of discrimination, it does not seem to counteract the impact of acute stress caused by frequent discriminatory incidences (Ajrouch, Reisine, Lim, Sohn, & Ismail, 2010; Finch & Vega, 2003).

Finally, the role of neighborhood factors in the associations amongst discrimination and mental health for young immigrants has been less explored. However, research demonstrating the positive impact of neighborhood collective efficacy on youth mental health—a relationship that might in part be understood through residents’ use of community resources and participation in neighborhood organizations (Xue, Leventhal, Brooks-Gunn, & Earls, 2005) is informative here. Community involvement and sense of belonging may help to protect against adversity in resettlement, including discrimination and marginalization. For young immigrants facing uncertainty and isolation in the process of integration, institutional supports through community organizations and strong social networks may facilitate access to resources and information, strengthen their sense of belonging at the community level, and enhance their psychological well-being (Arévalo, Tucker, & Falcón, 2015; Stewart, 2014; Stewart et al., 2008).

In this paper, we examine how discrimination relates to mental health among one young adult immigrant group (Somalis) at risk for an unfavorable context of reception in North America, and whether individual acculturation styles and community-level context of reception factors explain this association. We explore the ways in which individual and community level factors help us to understand these relationships and identify potential levers for intervention to disrupt them. First, we explore the multiple types of discrimination experienced by young Somalis and the reasons attributed to these experiences. Next, the impacts of experiences of discrimination on mental health outcomes and the way in which these associations are mediated or moderated by individual (marginalized acculturation, gender, and years residing in North America) and community level (collective efficacy, ethnic community belonging, and receiving community belonging) factors are examined.

Methods

Design and Setting

Data from the first wave of the Somali Youth Longitudinal Study (SYLS; Wave 1: May 2013 – January 2014) were analyzed to examine the associations among experiences of discrimination and mental health outcomes. SYLS is a community-based participatory research (CBPR; Israel, Schulz, Parker, & Becker, 1998) project focused on understanding and supporting the mental health of Somali young adults in four different regions in North America: Boston, MA, Minneapolis, MN, and Portland/Lewiston, ME in the U.S. and Toronto, Canada.

In line with CBPR methodology, academic and Somali community stakeholders (community advisors and staff) worked together on all aspects of SYLS from design (study goals) and data collection (protocols and procedures) to interpretation of the results and dissemination of findings. Additional Somali community partnerships with a variety of organizations were formed in each site location (e.g., religious institutions, social services agencies, and a refugee resettlement agency). These partnerships enabled cultural and linguistic issues to be considered in every stage of the research process and also allowed the development of a scientifically rigorous and community acceptable research protocol. For more detailed information regarding SYLS’ CPBR approach see Ellis et al., 2016.

Sample

Participants were recruited through snowball sampling (as described in Ellis et al., 2015). Steps were taken to overcome the known difficulty of obtaining a random sample in immigrant and refugee communities (Spring et al., 2003) and to reach as many Somali community members as possible. This included building trust with Somali community leaders in each of the study’s sites, hiring Somali research staff, attending community meetings, using word of mouth to spread information about the study, and receiving referrals from previous participants (Ellis et al., 2007). Steps were taken to ensure participants represented a range of diverse experiences and different sectors of the Somali community (e.g., religiosity, clan, socioeconomic status), including recruitment in diverse gathering spaces. Participants were eligible for inclusion in the study if they identified as Somali, were between the ages of 18- and 30 years, and had lived in the United States or Canada for at least one year. Twenty-three percent of participants were born in the U.S. or Canada; as such, the term immigrant (i.e., a person who comes to live permanently in a foreign country) is used as a broad descriptor of the sample and includes first and second generation young adults who continue to identify as Somali. The sample includes 439 Somali young men and women who participated in SYLS’ Wave 1. Demographics of participants are presented in Table 1.

Table 1.

Demographic Characteristics of the Study Sample Characteristic (N = 439)

Characteristics	n (%)	Mean ± SD (Range)
Gender
Female	165 (37.6 %)
Male	274 (62.4 %)
Years in the United States/Canada		13.08± 5.82 (1–24)
Age		21.36± 2.98 (18–33)
Location of Interview
Boston, MA	116 (26.4 %)
Toronto, CA	115 (26.2 %)
Minneapolis, MN	107 (24.4 %)
Lewiston/Portland, ME	101 (23 %)
Currently In School or Employed
Yes	354 (80.6 %)
No	85 (19.4 %)

Data Collection

Somali research staff obtained informed consent and interviews were conducted by a non-Somali research assistant. Each interview lasted approximately 1.5 hours and consisted of verbally administered standardized instruments in English with a Somali cultural broker available to provide concept clarification or translation as desired.¹ For more detailed information regarding specific SYLS study protocols see Ellis et al., 2016. The Institutional Review Boards of Boston Children’s Hospital and Northeastern University approved this study. All participants were paid $40 as a thank you for their time.

Measurement

Three mental health outcomes were examined: anxiety, depression, and posttraumatic stress symptoms. Independent variables included: perceived experience of everyday discrimination, major experiences of discrimination, marginalized acculturation style, and societal factors relevant to the context of reception (neighborhood cohesion, receiving community belonging, and Somali community belonging). In addition, demographic data, including gender and how long (in years) they had lived in the U.S./Canada were included.

Mental health

Symptoms of anxiety and depression were assessed using the associated subscales of the Hopkins Symptom Checklist-25 (HSCL-25; Parloff, Kelman, & Frank, 1954). Participants were asked about symptoms of anxiety (e.g., feeling fearful, tense or keyed up, nervous or shaky) and depression (e.g., self-blame, crying easily, poor appetite) they had experienced over the past four weeks. Interviewees were asked to rank these symptoms on a 4-point scale from 1 (not at all) to 4 (extremely). In the present study, Cronbach’s alphas for the anxiety and depression subscales of the HSCL were 0.87 and 0.91, respectively.

Participants were also administered the Harvard Trauma Questionnaire (HTQ; Mollica et al., 1996), which asked participants to rate how much they had been bothered by symptoms of post-traumatic stress on a 4-point scale (not at all, a little, quite a bit, or extremely) in the past four weeks. (α = 0.90). Both the HSCL-25 and HTQ have been validated for use with refugee and immigrant populations cross-culturally (Ellis et al., 2015; Mollica et al., 1996; Kleijn, Hovens, & Rodenburg, 2001), including Somalis (Fox & Tang, 2000; Stutters & Ligon, 2001). The construct validity of the measures was evaluated by use of Confirmatory Factor Analysis (CFA). Results for the HSCL indicated excellent model fit as all factor loadings were significant. Furthermore, the RMSEA was equal to 0.04, and both the Comparative Fit Index (CF) and Tucker-Lewis indices were acceptable, i.e, 0.980 and 0.978, respectively. Internal consistency by use of omega reliability was equal to 0.949. For the HTQ measure the CFA also confirmed the excellent fit of the data to the hypothetical structure (RMSEA=0.054, CFI=0.971, TLI=0.967) with all measurement paths being significantly different from zero. Omega reliability of the HTQ was equal to 0.951). The means of these mental health scales were used in analyses.

Discrimination

Perceived experience of discrimination was assessed using the nine-item Every Day Discrimination Scale (EDD; Williams, Yu, Jackson, & Anderson, 1997). Participants were asked about experiences of discrimination in their day-to-day lives. A sample item is, “In your day-to-day life, how often are…you threatened or harassed?” They were asked to indicate “never,” “once every few years,” “a few times a year,” “a few times a month,” “at least once a week,” or “every day.” In the present study, Cronbach’s Alpha for the EDD was 0.87.

The Major Experiences of Discrimination (MED) Scale (Kessler, Mickelson, & Williams, 1999) was also administered. Participants were asked “yes” or “no” whether at any time in their life they had experienced any of six types of discrimination (see Table 1 for types). If an experienced was endorsed, participants were then asked how many times that event has occurred over the course of their lifetime. Participants were also asked to provide an open-ended response describing what they believed to be the main reason(s) for these discriminatory experiences. Multiple coders then developed a coding scheme through an iterative process, including tagging responses and generating codes based on tag clusters. Seven major categories emerged, coded as: ethnicity, bad people/bad world, race, ignorance/fear of others, religion, appearance, assumptions made by others, neighborhood, and other. As many respondents provided multiple answers their first response was chosen for these analyses.

Both the EDD and MED are well validated in the literature (EDD; Williams, 2012; Clark, Coleman & Novak, 2004, Peek, Nunez-Smith, Drum, & Lewis, 2011, MED; Kessler, Mickelson, & Williams, 1999), and the EDD has demonstrated reliability with Somali immigrants (Ellis et al., 2008). Using the current data, the construct validity of EDD was confirmed using CFA (RMSEA=0.070, CFI=0.977, TLI=0.969). Omega reliability was equal to 0.933. For the MED, construct validity was demonstrated as the chi-square test of exact fit was not significant [χ2(9)=6.096, p=0.730]. Further indices pointed to the excellent fit of the data to the model (RMSEA=0.001, CFI/TLI=1.00). Omega reliability of the MED was equal to .779. For both the MED and EDD, frequencies were calculated for each item to assess how common each experience of discrimination was within the present sample. Additionally, a mean score was computed based on the frequency in which each type of mistreatment (experience of discrimination) was endorsed; specifically, item responses were summed and then averaged across the items.

Marginalized acculturation

The presence of a marginalized acculturation style, a rejection of both one’s culture of origin and the cultural norm, was evaluated using the 9-item marginalization subscale of the East Asian Acculturation Measure (Barry, 2001). Sample items include “Generally, I find it difficult to socialize with anybody, American or Somali” and “Sometimes I feel that Somalis and Americans do not accept me.” Response choices ranged from 1 (completely disagree) to 7 (completely agree); a mean of the 9 items was calculated and used in analyses. This subscale has demonstrated good internal consistency with Somali immigrants (α = .82) (Ellis et al., 2015). In the present study, the unidimensional simple structure was confirmed using CFA (RMSEA=0.07, CFI=0.977, TLI=0.969). Furthermore, omega reliability was equal to 0.903), and did in the present study as well (α = .81).

Context of reception

At the neighborhood level, context of reception was assessed using two 5-item subscales of the Collective Efficacy Scale (Sampson, Raudenbush, & Earls, 1997; Sampson, Morenoff, & Earls, 1999) that measure neighborhood social cohesion and intergenerational closure. These subscales have demonstrated reliability and validity (Sampson, Raudenbush, & Earls, 1997; Wickes, Hipp, Sargeant, & Homel, 2013; Sampson, Morenoff, & Earls, 1999), including with Somali immigrants (Ellis et al., 2015). In the present study, Cronbach’s alphas for the social cohesion and intergenerational closure subscales were 0.75 and 0.74, respectively. Sample items include “people in my neighborhood can be trusted (social cohesion) and “adults in my neighborhood know who the local children are” (intergenerational closure). Respondents were asked to think about their neighborhood and rate how much they agree or disagree with each statement on a 4-point Likert scale response, ranging from 1 (strongly disagree) to 4 (strongly agree). The factor structure of the measure was partly supported by the CFA model (RMSEA=0.12, CFI=0.89, TLI=0.86). Omega reliability, however, was excellent at 0.867 suggesting the presence of a well-defined unidimensional construct. A mean of the 10 items was computed to summarize the variable.

At the community level, context of reception was assessed via the Psychological Sense of Community Membership scale (PSCM) adapted from Goodenow (1993) and Hagborg, (1998). This instrument was administered twice, once asking about Somali community belonging and once asking about American/Canadian community belonging. Participants were asked how true 18 statements were for them in reference to their Somali community on a scale of 1 (not at all true) to 5 (completely true); these questions were then repeated in reference to participants’ American/Canadian community. The PSCM has demonstrated excellent internal consistency with Somali young adults (α = .91) for each of the scales (Ellis et al., 2015). In the present study, Cronbach’s alphas for the Somali community belonging and the American/Canadian community belonging subscales were 0.72 and 0.77, respectively. A sample item is, “There’s at least one person in the Somali (American/Canadian) community I can talk to if I have a problem.” With the present data, the unidimensional structure was borderline using CFA for both the Somali and the non-Somali samples (RMSEA_Somali=0.09; _{RMSEAnon-Somali}=0.09; CFI_Somali=0.820, _{CFInon-Somali}=0.863, TLI_Somali=0.80, _{TLInon-Somali}= 0.845). However, internal consistency reliability using omega was equal to 0.897 and 0.918, respectively, suggesting a strong unidimensional structure and a small amount of measurement error. A mean-based term for each 18-item subscales was calculated and used in analyses.

Data Analysis

First, we examined frequencies among experiences of discrimination, potential co-variates and confounders, and our mental health outcomes. Particular attention was paid to the frequency and patterning of experiences of discrimination incurred by our participants so as to better understand the ways in which these experiences might relate to mental health outcomes.

Structural equation modeling (SEM)

Direct and indirect relationships within our path models were evaluated by use of Structural Equation Modeling (SEM). The methodology involves utilizing latent or measured variables (Bentler, 2007; Joreskog, 1973) and in the present study we opted for the latter. The SEM analytical methodology employs a simultaneous equation approach with the goal of minimizing discrepancies between a population covariance matrix (implied variance-covariance matrix Σ(θ)) and the sample-based variance-covariance matrix [Σ]. Model fit is evaluated by various means such as the omnibus Chi-square statistic, which is a test of exact fit (MacCallum, Browne, & Sugawara, 1996). As such, little value is placed on the statistic as minimal discrepancies are associated with significant findings. To the aid of the chi-square statistic there is a range of descriptive fit indices that test the fit of the model in relation to other competing models. Amongst them, most prominent are the Comparative Fit Index (CFI) and the Tucker-Lewis index (TLI) (e.g., Bentler, 1990; 2007). For both, values greater than .95 are indicative of good model fit (Hu & Bentler, 1999). Last, a recommended choice is the use of unstandardized residuals in the form of the Root Mean Square Error of Approximation (RMSEA, Steiger & Lind, 1980). Acceptable values of the residuals are in the range of 5%−8%. As pointed by MacCallum, Browne, & Sugawara (1996), evaluation of the residuals provides one of the most unbiased estimate of model fit. The level of significance of the structural coefficients in the path models was set to 5% for a two-tailed test.

Power analysis of SEM models

Power for path analytic models involved testing the hypothesis that any slope coefficient greater than .50 standard deviations should be associated with a 5% or lower probability using a two-tailed test (Cohen, 1992). Consequently, a Monte Carlo simulation involved generating continuous indicators as per our path models using zero mean and variance equal to 1. Path coefficients were posited to .5 and 1,000 replicated data sets were generated using Mplus, 8.2 (Muthen & Muthen, 2006). Using an estimated sample size of 439 valid cases, results indicated that power levels to identify as significant effects that exceed .5 standard deviations was more than 99%.

Results

Experiences of Discrimination

See Table 2 for the prevalence of each discrimination type. 91.6% of participants reported at least one experience of day-to-day discrimination (EDD), while 64.2% of participants reported at least one experience of major discrimination (MED). Endorsement rates for the types of EDD were quite high; the highest being for “people acting like they’re better than you are” (79%) and the least frequently endorsed being, “threatened or harassed” (36%). Approximately 19% of the sample (N=81) reported having experienced all nine instances of everyday discrimination. Approximately 28% (N=121) of the sample reported experiencing 1 incident of major discrimination over the course of their lifetime; 21% reported experiencing 2 of the 6 items; 11% reported experiencing three.

Table 2.

Experiences of Discrimination Among Somali Refugees

Everyday discrimination (EDD)	N (%)
Treated with less courtesy than others	318 (72.4%)
Treated with less respect than others	287 (65.4%)
Receive poorer service than others	293 (66.7%)
People act as if they think you’re not smart	302 (68.8%)
People act as if they are afraid of you	246 (56.0%)
People act as if they think you are dishonest	226 (51.5%)
People act as if they’re better than you are	348 (79.3%)
Called names or insulted	250 (56.9%)
Threatened or harassed	157 (35.8%)
Lifetime discrimination (MED)	N (%)
Fired from a job or denied a promotion	79 (17.8%)
Not hired for a job	122 (27.8%)
Stopped, searched, questioned, physically threatened, or abused by the police	184 (41.9%)
Discouraged by a teacher from continuing education	124 (28.2%)
Prevented from moving into a neighborhood/apartment	28 (6.4%)
Denied bank loan	8 (1.8%)

Lastly, our participants reported diverse reasons (attributions) for experiencing these events (Table 3). Race was the most common perceived source of discrimination (EDD: 13.6%; MED: 23%); ethnicity (EDD: 13.9%; MED: 14.9%), religion (EDD: 13%; MED: 9.2%), assumptions made by others (EDD: 10.2%; MED: 15.7%), and “bad people/bad world” (EDD: MED; lifetime: 2.4%) were also frequently reported.

Table 3.

Reasons Ascribed to Experiences of Discrimination Among Somali Refugees

	Everyday discrimination – N (%)1	Lifetime discrimination – N (%)2
Ethnicity	47 (13.3%)	37 (14.9%)
Bad people / bad world	47 (13.3%)	6 (2.4%)
Race	48 (13.6%)	56 (22.5%)
Ignorance/fear of others	29 (8.2%)	5 (2.0%)
Religion	46 (13.0%)	23 (9.2%)
Appearance	31 (8.8%)	19 (7.6%)
Assumptions made by others	36 (10.2%)	39 (15.7%)
Neighborhood/social associations	17 (4.8%)	25 (10.0%)
Abuse of power/authority	-	5 (2.0%)
Other	53 (15.0%)	34 (13.7%)

¹N = 354

²N = 249

Path Models: In Search of the Optimal Model Fit

An iterative procedure was utilized to identify an optimal statistically and theoretically viable model. The original model (see Figure 1) posited that our dependent variable(s) (anxiety, symptoms of depression, and posttraumatic stress symptoms) would be predicted directly and indirectly from individual and community-level context of reception factors. Table 4 shows a series of nested models that lead to optimal model fit. As shown in the table, the model in Figure 1 did not provide a good match to the sample-based variance-covariance matrix in relation to the hypothesized (population) matrix. The model fit was not acceptable by any standards. Inspection of the parameters of the model suggested that the relationship between anxiety symptoms and collective efficacy should be fixed to zero as it was almost zero. This constraint improved model fit significantly, and there was also the gain of 1df (from 6 to 7). Further inspection of Model 2 by use of the Lagrange Multiplier suggested that the model would be significantly improved through adding residual covariations between the marginalization variables, marginalized acculturation and the two community belonging variables. Models 3 and 4 tested this proposition, resulting sequentially in improved model fit. Residual covariances reflect at time statistical nuisances and may be hard to interpret. In the present case, we fitted a one-factor model to the three components of marginalization and that linear combination resulted in the explanation of 61.125% of the items’ variances. Apparently, the remaining variance likely represented a common third variable that was accountable for the relationship among residual variances. There is a plethora of social processes that govern how a person or a group are treated, but at present there were no relevant measures to try and tap that part of the unexplained item variability. Model 4 included a further improvement by dropping the non-significant path between discrimination and collective efficacy, with the model reaching borderline model fit by use of descriptive fit indices. The last modification involved a direct path linking Somali community belongingness to collective efficacy. The final model was acceptable by all statistical means. First, the chi-square test was non-significant suggesting exact fit (no significant differences between sample-based model fit and an errorless model). Both descriptive fit indices were at 1.0 and the RMSEA values were less than .001. Consequently, all findings pointed to the presence of excellent model fit for anxiety symptoms, but also for the remaining dependent variables (see note in Table 4).

Figure 1.

Theoretical model for the prediction of mental health from discrimination.

Table 4.

Structural Equation Modeling (SEM) Comparison between Nested Models for the Prediction of Mental Health Problems

Model Comparison	Chi-square	D.F	Δ-Chi-square	p-value	Δ-D.F	CFI	Δ-CFI	TLI	Δ-TLI	RMSEA	Δ-RMSEA
1. M1: Theoretical Model, Original conceptualization	149.033***	6	-	-	-	0.579	-	0.000	-	0.233	-
2. M2: No correlation between Anxiety and collective efficacy	153.646***	7	4.613*	0.019	1	0.568	0.011	0.074	0.074	0.218	0.015
3. M3: Correlation between belonging variables	101.853***	6	51.793***	<0.001	1	0.718	0.150	0.294	0.220	0.191	0.027
4. M4: Correlation between belonging and marginalization and dropping of non-sig EDD to collective efficacy path	39.868**	5	61.985***	<0.001	1	0.897	0.179	0.692	0.398	0.138	0.053
5. M5: Collective efficacy predicted by belonging- Somali	2.269	4	37.599***	<0.001	1	1.000	0.103	1.000	0.308	<0.001	0.138

Note: Conventions for Delta fit index values are .20 as per Fan and Sivo (2007). All difference chi-square values were significant suggesting model fit improvement when moving from one model to the next. a test of “exact fit” as per MacCallum, Browne and Sugawara (1996). Results reflect the model in which anxiety was the dependent variable. As expected, very similar findings and excellent model fit were observed for the models with depression and posttraumatic stress as the dependent variables [Depression: χ²(4) = 2.523, p=.989; CFI=1.0, TLI=1.0, RMSEA=<.001] [posttraumatic stress: χ²(4) = 2.035, p=.968; CFI=1.0, TLI=1.0, RMSEA = <.001].

^*p<.05.

^**p<.01.

^***p<.001.

Figure 2 shows the structural models predicting, anxiety symptoms (upper panel), symptoms of depression (middle panel) and posttraumatic stress symptoms (lower panel) from marginalization and discrimination. As shown in Figure 2, upper panel, discrimination exerted both direct and indirect effects over anxiety symptoms. The direct effect was equal to .200 standard deviations and the indirect .096 for a total effect of .296. Marginalized acculturation was predicted positively from discrimination (b=.307) in that for one standardized unit of change of discrimination acculturation changed by 0.30 standardized units. On the other hand, Somali community belongingness and receiving community belongingness were predicted negatively from discrimination (b_Som=−.291, b_Re=−.316). Last, anxiety symptoms were predicted positively from marginalized acculturation (b=.312) and receiving community belongingness (b=.190) and negatively from Somali community belongingness (b=−.247). All structural coefficients were significant at p<.05 using a two-tailed test.

Figure 2.

Structural equation model predicting mental health from marginalization components and discrimination. Estimates are standardized. Parameters next to circled variables represent residual variances.

Testing Gender and Years in Resettlement

Effects of gender

The invariance of the parameters across gender were tested through estimating the coefficients (slopes and intercepts) for one group (i.e., males) and then constraining them to be equivalent for females in order to test the hypothesis that the observed findings (slopes) were of the same magnitude across males and females. Evidence in favor of gender invariance would be manifested with a non-significant Chi-square value pointing to non-significant misfit in the presence of the constrained relationships. Results for anxiety symptoms as the dependent variable indicated that the model fitted the data well with inclusion of that constraint [χ²(19) = 26.269, p=.123]. Results were replicated with depression being the outcome variable [χ²(19) = 21.364, p=.317] and the same held for posttraumatic stress [χ²(19) = 20.122, p=.387] with the chi-square tests being non-significant pointing to the presence of “exact fit” and, thus, the presence of invariance across gender. Consequently, the behavior across gender was consistent with regard to the posited relationships and models were analyzed in full, not splitting the sample by gender.

The role of years in the U.S.

To evaluate the invariance by amount of years residing in the U.S., a series of interaction terms were created for time in the U.S. with discrimination, the main independent variable. A multiplicative term was created and models included both main effects and interaction term. Results indicated that the interaction of discrimination with years in the U.S. was not significant across anxiety symptoms (b=.005, p=.257), symptoms of depression (b=.001, p=.868) and posttraumatic stress symptoms (b=.006, p=.114). Consequently, time in the U.S. was not further considered as an influential variable in our path models.

In summary, experiences of discrimination, had a direct effect on worsening symptoms of depression, anxiety and post-traumatic stress, and this effect was also mediated through both individual and community-level factors. Higher mean discrimination scores were associated with a more marginalized acculturation style, which in turn was associated with worse mental health outcomes. Higher mean discrimination scores were associated with lower levels of both American/Canadian community belonging and Somali Community belonging. The effect of community belonging on mental health outcomes, however, was complex. While ethnic belonging functioned as hypothesized, with more belonging mitigating the adverse effect of discrimination on mental health, American/Canadian belonging functioned in the opposite way. Higher levels of American/Canadian belonging were associated with worse mental health. This finding held true across path models for all three mental health symptom types. Lastly, neither gender nor years in the U.S. were influential variables in the final path models.

Discussion

Our Somali young adult participants experience high rates of discrimination, both every day and through major events across their lifetime. Importantly, they attribute these experiences to multiple aspects of their identity, reinforcing the need for an intersectional understanding of the associations among experiences of discrimination and mental health outcomes. Discrimination experienced by Somali young adults has a powerful impact on mental health status, including measures of anxiety, depressive and posttraumatic stress symptoms. These results provide further evidence of the importance of the context of reception in promoting the well-being of immigrant communities as well as the interplay amongst multi-level factors in promoting health and well-being. Across three different manifestations of psychological distress (anxiety, depressive, and posttraumatic stress symptoms) findings were similar: discrimination had a direct effect and was associated with worsening mental health, and this effect was also mediated through both individual and community-level factors. Specifically, marginalized acculturation style, Somali community belonging, and receiving community belonging all mediated the association between discrimination and mental health, though in different directions.

At the individual level, higher mean discrimination scores were associated with a more marginalized acculturation style, which in turn was associated with worse mental health outcomes. The finding that marginalized acculturation is associated with poor mental health is in keeping with past research. In a study of more than 3,000 immigrant-origin Canadians, a marginalized acculturation style was found to exacerbate the effect of discrimination on mental health (Berry & Hou, 2017). Acculturation styles need to be understood in the context of reception; it is possible that higher levels of discrimination push immigrants towards a more marginalized acculturation style, as these experiences may simultaneously undermine one’s sense of identity with the host society (here American or Canadian) while also reinforcing a distancing of oneself from one’s devalued ethnic community. Under this frame, acculturation style is not necessarily a ‘choice’ made by immigrants but can be understood to describe a dynamic process shaped in part by the context of reception.

As noted, the effect of community belonging on mental health outcomes among our participants was complex. While ethnic belonging functioned as hypothesized, with more belonging mitigating the adverse effect of discrimination on mental health, American/Canadian belonging functioned in the opposite way. Higher levels of American/Canadian belonging were associated with worse mental health. This may reflect processes by which those who experience both a high level of discrimination and feel a strong sense of belonging to the receiving society may be particularly affected by the messages of rejection communicated via discrimination. Essentially, those who value their connection to American/Canadian community may feel they have “more to lose” by being rejected, or discriminated against, by society. Given this, the potential for ethnic community belonging to serve as a protective buffer against the adverse effects of discrimination may be important to understand.

Contrary to expectations, neighborhood collective efficacy was not related to either discrimination or mental health. Prior research has demonstrated how the context of reception for immigrants is shaped in part by the neighborhoods in which they settle and the communities of which they may feel a part. Immigrants are often marginalized into poor or unsafe neighborhoods as a result of financial stressors in resettlement (Carter & Osborne, 2009). Neighborhood social context has been shown to affect the health of immigrants in a number of ways. For example, social capital and social integration have been implicated as predictors of physical and emotional health (Chen & Yang, 2014; Rose, 2000), including among immigrant populations (Lorant, Van Oyen, & Thomas, 2008). Additionally, social cohesion has been shown to fully mediate the relationship between neighborhood poverty and mental health among Latino immigrants (Hong, Zhang & Walton, 2014). In contrast, when the local community provides opportunities for social support, connection, and inclusion, it can facilitate positive adjustment and well-being post resettlement (Green, Rhodes, Hirsch, Suarez-Orozco, & Camic 2008; Suárez-Orozco, Suárez-Orozco, & Todorova, 2008). More work is needed to understand the relationships among experiences of discrimination, neighborhood collective efficacy, other neighborhood characteristics, and mental health among immigrants. These efforts should aim to increase our understanding of the meaning and role of neighborhood, as well as measure neighborhood effects, in the lives of immigrants; and in particular, young immigrants. Lastly, these data speak to the impact of experiences of discrimination on mental health during a specific developmental period, as our sample is restricted to young Somali adults aged 18–30. Youth transitioning from adolescence to young adulthood are faced with fundamental developmental tasks of defining their identity and their relationship to society (Flanagan, Levine, & Settersen, 2009; Flanagan, Syversten, Gill, Gallay, & Cumsille, 2009; Sherrod & Lauchardt, 2009); experiences of discrimination and belonging may thus be particularly salient to mental health during this time. Longitudinal approaches, which examine experiences of discrimination and mental health across the life-course are needed to better understand the impact of multiple experiences of discrimination across different developmental stages.

This study of Somali young adults increases our understanding of the importance of the context of reception in the mental health and well-being of young adult immigrants. As noted, our current national immigration policy and anti-immigrant rhetoric have contributed to increasingly hostile environments, including rising rates of hate crime and discrimination, among immigrants and their children which may have long term impacts on health (Perreira & Pedroza, 2019). These data suggest that factors associated with the context of reception may be particularly important to assess in working with young adult immigrants to support their mental health. Furthermore, a better understanding of the ways in which individual and societal factors work together or through each other to shape immigrant mental health is needed. Understanding the ways in which these instances of discrimination and exclusion contribute to disparities in mental health, in addition to mechanisms of resilience in the face of these discriminatory experiences, will inform our understanding of these factors among immigrant groups and may provide knowledge to inform our understandings of the experiences of other people that similarly contend with multiple marginalized statuses. Ultimately, attention to such approaches is much needed if we are to develop clinical strategies, public health practice, and policy agendas that can shape more favorable contexts of reception to promote mental health and well-being for young adult immigrants.

Public Policy Relevance Statement:

Understanding the ways in which Somali young adults’ experiences of discrimination and exclusion contribute to worsening mental health, and mechanisms of resilience in the face of these discriminatory experiences, will inform our understandings of these factors among immigrant groups and other groups of people that similarly contend with multiple marginalized statuses. These findings are needed to inform policy and practice to promote the mental health and well-being of immigrants.