Social Support Protects Against Symptoms of Anxiety and Depression: Key Variations in Latinx and Non-Latinx White College Students

Abstract

Social support protects against perceived stress and its harmful effects on psychological well-being. College students in general are at high risk for mental health disorders, and Latinx college students face unique stressors placing them at greater risk of psychological distress. Social support may be a key construct in improving outcomes for college students; however, few studies have empirically tested whether the protective effect of social support is equivalent across racial/ethnic groups. Using a series of regression models, we investigated whether social support moderates the relationship between perceived stress and endorsement of depression and anxiety symptoms in Latinx (n = 265) and non-Latinx White college students (n = 216) and whether this moderating effect varied by group membership. Participants completed a series of questionnaires measuring social support, perceived stress, and depression and anxiety symptoms. The moderating effects of social support varied by group membership and outcomes (i.e., depression and anxiety). Social support moderated the relationship between perceived stress and depression symptoms for both Latinx and non-Latinx White students. However, social support only buffered the effect of perceived stress on anxiety symptom endorsement for Latinx college students. These findings suggest that social support does not function uniformly across racial/ethnic groups or the endorsement of depression and anxiety symptoms. Social support may be particularly important for Latinx students by providing a buffer between perceived stress and symptoms of anxiety.

Social support, or one’s perceptions that other people are available and willing to provide moral, emotional, and/or tangible support, is the “social ‘fund’ from which people may draw” to help manage stressors (Thoits, 1995, p. 64). Social support has been shown to be protective against mental health disorders, including depression and anxiety (Canino et al., 2008; Cruza-Guet et al., 2008; Sumner et al., 2011). The benefits of social support are well established among adults (e.g., Barrera, 1986; Huffman et al., 2015; Rapee et al., 2015); however, the social-support literature in college students and racial/ethnic minority college students in particular remains smaller in comparison (e.g., Baker & Robnett, 2012; Carlson, 2014; Constantine et al., 2002; Menagi et al., 2008; Raffaelli et al., 2013; Romero et al., 2015).

College students face unique stressors including adjusting to college life, pressures to succeed, and issues surrounding the transition from adolescence to young adulthood. Latinx students face many additional stressors, such as discrimination, minority status stress, and acculturative stress, potentially contributing to higher rates of mental illness (Dennis et al., 2010; Finch & Vega, 2003; Torres et al., 2012). As such, identifying the extent to which social support buffers the detrimental effects of stress for college students is critical.

A key limitation in the literature investigating social support among college students is that even when studies have diverse samples that include racial/ethnic groups (Lamis et al., 2016; Raffaelli et al., 2013), they are often not designed to examine group differences in the extent to which social support plays a protective role; thus, limiting inferences around key group differences that can aid in prevention and intervention efforts and guide the future research. We investigated the potential differences in moderation effects of social support on stress and psychological distress in Latinx and non-Latinx White (hereafter referred to as White) college students.

Perceived Stress, Depression, and Anxiety in College Students

Perceived stress refers to the extent to which life events are perceived as unpredictable, uncontrollable, and overwhelming, oftentimes exceeding an individual’ s perceived ability to cope (Nielsen et al., 2016). Perceived stress is a risk factor for various mental health issues including substance use (Dawes et al., 2000; Sinha, 2001), posttraumatic stress disorder (Fincham et al., 2009; Haisch & Meyers, 2004), anxiety (Dyrbye et al., 2006; Wang, 2006), and depression (McGonagle & Kessler, 1990; Rojo-Moreno et al., 2002).

College can be particularly stressful given the emergence of adulthood, newfound independence, and developing and maintaining intimate relationships (American College Health Association, 2012b), as well as academic, financial, and social stressors (Aselton, 2012; Dusselier et al., 2005). According to the American College Health Association (2012a), stress is the primary cause of impaired academic performance. Latinx college students face a unique cluster of stressors related to their minority status (Lee & Turney, 2012) that can negatively influence mental health. Latinx college students often face acculturative stress, discrimination, and questions about belonging on a college campus and their merits (Lopez, 2005).

Approximately 14% of college students are diagnosed with depression, and 23% are diagnosed with anxiety (American College Health Association, 2012a; Furr et al., 2001; Lovell et al., 2015). In general, students of color are more likely to endorse depression symptoms (22%) when compared to White students (15%; Eisenberg et al., 2013), regardless of the racial/ethnic makeup of the college/university (Arbona & Jimenez, 2014). Evidence on rates of anxiety is mixed. Some studies find comparable rates among different groups of college students (Eisenberg et al., 2013), whereas others have found greater risk among students of color. For instance, Latinx students have been found to be more likely to experience anxiety, greater alienation, isolation, and discomfort because of their minority status than other racial–ethnic groups (Saldaña, 1995).

Despite these rates of depression and anxiety, college students do not frequently seek services for mental health problems (Cook, 2007; Eisenberg et al., 2007). The lack of help-seeking behaviors increases the severity of psychological distress, risky behaviors, and college attrition. Social support could mitigate the prevalence of mental health issues and increase the likelihood of help-seeking behavior (Caldeira et al., 2009; Czyz et al., 2013).

Social Support as a Protective Factor

Social support helps to manage stressors and protect against mental health disorders, including depression and anxiety (Canino et al., 2008; Cruza-Guet et al., 2008; Sumner et al., 2011). When students perceive that their family and friends support them, they are able to more effectively adjust to college demands (Alvan et al., 1996; Friedlander et al., 2007) and their risk for depression and psychological distress decreases (Brandy et al., 2015; Rodriguez et al., 2003). Among diverse college students, including Latinx students, the quality of peer social support has been associated with a decrease in self-injurious behavior (Whitlock et al., 2015). Furthermore, perceived social support, specifically the extent to which an individual feels supported by family and friends, is associated with fewer feelings of loneliness (Nicpon et al., 2006) and depressive symptoms (Raffaelli et al., 2013).

Social support offers protective benefits by providing a sense of connection with others (Berkman et al., 2000), and by contributing to increased self-esteem (Uchino et al., 2012). It is also associated with greater ability to cope with stressors and seek assistance from others in times of need (Berkman et al., 2000; Smith & Christakis, 2008; Uchino et al., 2012). Support in the form of emotional, instrumental, or other tangible assistance, in addition to providing individuals with a sense of belonging, offers helpful ways to appraise and cope with stressful situations (Cohen & Wills, 1985). According to the stress-buffering hypothesis (Cohen et al., 1985), social support attenuates both the impact of and the reaction to stress on well-being to the degree that individuals can better cope with the stressor(s). This buffering (i.e., moderating) effect of social support has been supported by the existing literature. The presence of high levels of social support has been found to lessen the impact of stress on physical (Uchino et al., 2012) and mental health outcomes (Anderson, 2004; Corona et al., 2017; Kawachi & Berkman, 2001). In a recent study investigating the moderating effects of social support on depression, undergraduate college students who reported high levels of perceived stress also reported higher scores on a depression measure than those with lower levels of perceived stress. However, depression symptom endorsement was lower among those with high levels of stress who perceived high social support than those who perceived low social support (Wang et al., 2014).

Key Contributions of Our Study Design

We compared two racial/ethnic groups (Latinx and White college students) in order to assess potential differences in the buffering effects of social support across race/ethnicity in the presence of perceived stress. The benefits of social support may be culture specific (Chen et al., 2012; Kim et al., 2008; Park et al., 2013), potentially leading to group differences in the buffering effects of social support. In more collectivistic cultures, for example, social support is more readily available and often provided without the explicit request of the individual (Kim et al., 2008). Furthermore, although the Latinx community is heterogeneous, there is a strong cultural value of familismo (familism), which refers to strong identification and attachment to family, as well as strong feelings of “loyalty, reciprocity, and solidarity” among members of the same family (Sabogal et al., 1987). In a racially/ethnically diverse sample of college students, Campos et al. (2014) found that familism was associated with better psychological health by way of greater closeness to family members and greater perceived social support. However, Latinx students reported the highest level of familism relative to the other racial/ethnic groups.

Furthermore, we evaluate whether the effects of social support differ across multiple mental health outcomes. Depression and anxiety, although often highly correlated, remain different constructs with differences in their etiology and treatment (Corona et al., 2017). Both familial and peer support have been found to buffer the effects of acculturative stress (e.g., challenges related to language skills, intercultural relations, discrimination, and pressure to assimilate) on depression but not anxiety (Corona et al., 2017; Crockett et al., 2007). Thus, research on the differences between the benefits of social support for depression versus anxiety requires further exploration.

Finally, our study assesses social support as a larger construct rather than concentrating solely on sources of support, such as family or friends. In addition, we account for common subtypes of support, including appraisal and perceived availability of support (Finch & Vega, 2003), as well as the perception of belonging to a network. Belonging may be a crucial aspect of social support, as it relates to cultural constructs such as familism and collectivism, known to be particularly relevant for Latinxs (Sabogal et al., 1987).

Purpose of Study

This study addresses the above gaps in the literature by exploring the moderating effects that perceived social support may have in buffering Latinx and White undergraduate college students against the negative effects of stress. Specifically, we assess whether the protective effect of social support varies across these two racial/ethnic groups. A series of regression analyses were used to investigate the role of social support on psychological health, specifically on the symptoms of depression and anxiety. Our hypotheses were as follows: (a) the moderating role of social support would vary by group and (b) the moderating role of social support would vary by outcome (e.g., depression vs. anxiety symptomatology). Specifically, we hypothesized that social support would have a greater moderating effect for Latinx students than their White counterparts.

Method

Participants

The study included self-identified Latinx (n = 265) and White (n = 216) undergraduate college students from a designated Hispanic-serving research university in a Southwestern city (Table 1). The Latinx sample averaged approximately 20 years of age (SD = 3.8) and was comprised of mostly women (65.3%, n = 177). The White sample averaged approximately 22 years of age (SD = 7.3) and was also comprised of mostly women (67.4%, n = 149).

Table 1.

Demographic and Measurement Characteristics for the Latinx and Non-Latinx White Students

Variable	Latinx M (SD)	Non-Latinx White M (SD)	Statistic
N	265	216
Age (in years)	20.3 (3.80)	22.01 (7.30)	t(480) = 3.26**
Sex (female, %)	177 (65.30)	149 (67.40)	χ2(1, N = 481) = 0.26
Father education†	4.92 (1.73)	6.01 (1.56)	t(479) = −7.21**
Mother education†	5.14 (1.75)	5.95 (1.41)	t(489) = −5.54**
Generational status (n, %)
1st generation	21 (7.7)	12 (5.4)	χ2(1, N = 492) = 1.05
2nd generation	49 (18.1)	21 (9.5)	χ2(1, N = 492) = 7.34**
3rd generation	14 (5.2)	14 (5.9)	χ2(1, N = 492) = 0.12
4th generation	52 (19.2)	54 (24.4)	χ2(1, N = 492) = 1.98
5th generation	134 (49.4)	118 (53.4)	χ2(1, N = 492) = 0.76
Perceived stress	2.79 (0.53)	2.74 (0.45)	t(490) = −0.89
Social support
Appraisal	3.52 (0.33)	3.43 (0.41)	t(490) = −1.60
Belonging	3.43 (0.33)	3.30 (0.38)	t(489) = −2.30*
Tangible	3.50 (0.30)	3.43 (0.29)	t(490) = −1.33
Social support total	3.48 (0.47)	3.39 (0.52)	t(490) = −2.03*
BDI-II	9.19 (9.39)	9.60 (9.53)	t(489) = 0.29
BAI	8.17 (7.62)	7.96 (7.79)	t(489) = −0.60

Note. BDI-II = Beck Depression Inventory-II; BAI = Beck Anxiety Inventory.

^†Parental education was calculated using each parent’s level of education measured on a 1–8 scale (1 = below an 8^th grade education, 8 = doctoral level graduate degree).

^*p < .05.

^**p < .01.

Procedures

Participants were part of a larger, Internal Review Board approved study investigating psychological factors and mental health in a racially/ethnically diverse sample. Constructs investigated included somatization, alexithymia, resilience, social support, interpersonal relationships, personality traits, coping skills, acculturation, and spirituality and religion. The study was advertised in introductory psychology courses and in an online research enrollment system. Participants were specifically recruited from introductory psychology courses because they could receive class credit in return for their study participation. After they consented and enrolled in the study, participants completed a series of questionnaires measuring multiple psychological constructs associated with mental health outcomes, all of which were in English only. The study was conducted in person between 2008 and 2011.

Measures

Demographic Information

Ethnicity was determined through self-report on a demographic questionnaire as was age, gender, and generational status. Because college students are not likely to be financially independent, direct reporting of income was difficult to interpret; thus, we used parental education variables as a measure of socioeconomic status (SES). Both mothers’ and fathers’ educational attainments were assessed on an eight-point scale ranging from below an 8th grade education through doctoral level graduate degree attainment (e.g., MD, PhD, and PsyD).

Interpersonal Support Evaluation List-12

The Interpersonal Support Evaluation List-12 (ISEL-12) is a 12-item, self-reported, social-support measure adapted from a larger 40-item scale assessing perceived availability of potential social-support sources (Cohen et al., 1985). The abbreviated instrument assesses three constructs: appraisal (advice or guidance [e.g., “I feel that there is no one I can share my most private worries and fears with.”]), belonging (empathy, acceptance, and concern [e.g., “If I wanted to have lunch with someone, I could easily find someone to join me.”]), and tangible (help or assistance and financial/material support [e.g., “If I were sick, I could easily find someone to help me with my daily chores.”]) support. Each construct is measured using four response options ranging from 0 (definitely false) to 4 (definitely true). All items were averaged to obtain a total score per participant. The ISEL-12 has been validated for use with White and Latinx populations (Merz et al., 2014). The overall Cronbach’s alpha was .87 for our sample, indicating high internal consistency, and thus, good reliability. Cronbach’s alpha was .85 for the Latinx sample and .89 for the White sample.

Perceived Stress Scale

The Perceived Stress Scale (PSS) is a brief 10-item, self-report instrument that assesses the degree to which an individual perceives life situations as stressful (Cohen & Williamson, 1988). Respondents are asked to consider how often they have experienced various thoughts/feelings related to stress in the last month (e.g., “How often have you been able to control irritations in your life,” How often have you felt that you were on top of things?”). It has demonstrated adequate internal and test–retest reliability and is correlated with symptom severity measures, including those assessing depressive symptoms. Response options range from 0 (never) to 4 (very often). After accounting for reverse items, all items were averaged to obtain a total score per participant. The PSS is commonly used in studies of Latinx health (Chavez-Korell & Torres, 2013; Flores et al., 2008; Perez et al., 2015). Cronbach’s alpha was .89 for our sample, indicating good reliability. Cronbach’s alpha was .89 for the Latinx sample and .88 for the White sample.

Beck Depression Inventory-II

The Beck Depression Inventory-II (BDI-II) (Becket al., 1996) is a21-item, self-administered screening measure used to assess severity of depression symptoms in clinical and nonclinical samples. Items pertain to cognitive and affective symptoms (e.g., sadness and guilt) and somatic symptoms (e.g., changes in appetite and sleep). Respondents rate the degree to which these symptoms have bothered them in the past month using a 4-point scale of 0 (symptom absent) to 3 (severe symptom). The BDI-II total score ranges from 0 to 63, with high scores indicating greater severity of depression. The BDI-II has been shown to have adequate reliability and construct validity in general clinical samples (Beck et al., 1996) and in ethnically diverse college samples (Carmody, 2005; Storch et al., 2004). Cronbach’s alpha was .93 for our sample for both Latinx and White participants, indicating excellent reliability.

Beck Anxiety Inventory

The Beck Anxiety Inventory (BAI; Beck & Steer, 1993) is a 21-item, self-administered measure that discriminates between anxiety and depression while maintaining convergent validity (Beck et al., 1988). This measure has been found valid in differentiating between anxiety and other psychiatric disorders and for use with Latinx and White college samples (Contreras et al., 2004). Response options range from 0 (not at all) to 3 (severely). Items are added together to obtain a BAI total score that ranges from 0 to 63 with high scores indicating greater severity of anxiety symptoms. Cronbach’s alpha was .92 for our sample, indicating excellent reliability. Cronbach’s alpha was .93 for the Latinx sample and .89 for the White sample.

Analytic Strategy

Preliminary Analyses

Using independent sample t tests and chi-squared tests, we first examined mean values and standard deviations for sociodemographic, stress, social-support, depression, and anxiety measures by evaluating the statistical differences in these variables according to ethnic group status. We also examined the bivariate associations among social support, perceived stress, depression, and anxiety separately by racial/ethnic group using Pearson’s correlations.

Main Outcome Analyses

With a series of regression models, we explored the relationship among perceived stress, social support, and the mental health outcome variables of depression and anxiety. Social support was conceptualized as a moderator between perceived stress and the mental health outcomes. Following a step-by-step approach, we first tested a regression model for each outcome variable independently for each racial/ethnic group controlling for gender, age, parental educational level, and generational status. We explored main effects and interactions separately; however, as results are consistent across models, we report them in combination for simplicity. Both our moderator and main independent variable (i.e., social support and perceived stress) were grand mean centered in these models in order to aid interpretation in the event of a significant interaction.

Given the co-occurrence of depression and anxiety in the general population, we employed seemingly unrelated regression models (Zellner, 1962) to allow for correlated error terms between regression models in our final models. These regression models provided a generalization of more traditional MANCOVA models and, in the same way, allowed for more efficient estimation in cases in which outcomes are correlated. To allow for ease of access to within-group results, we calculated three seemingly unrelated regression models: Anxiety and depression for Latinxs and for Whites and a third model for the full sample testing a three-way interaction between perceived stress, social support, and ethnicity. Technically, results for the within-group models can be recovered from parameters estimated during the course of fitting the model for the full sample. R-squared statistics, representing proportion of explained variance, were calculated for each of these models. Stata software version 12.0 (StataCorp, 2011) was used for all analyses, and p ≤ .05 was used as the threshold for statistical significance. We graphically probed significant interactions using simple slopes graphs, where the significance of the relationship between x and y is plotted across several conditional levels of the moderator (Aiken & West, 1991; Preacher et al., 2006).

Diagnostics

We used several techniques to explore the properties of these linear regression models. Cameron and Trivedi’s (1990) information matrix test was used to detect indications of heteroscedasticity, skewness, and kurtosis in model residuals. Consistent with the findings of Long and Ervin (2000), we used Huber and White’s “sandwich” estimator to calculate heteroscedasticity-consistent standard error estimates in order to quantify possible impacts of heteroscedasticity on model characteristics. The link test from Pregibon (1980), which assesses the significance of squared predicted values alongside predicted values as model-based predictors of observed values, was used as a broad test of model specification. We also detected possibly unduly influential observations using Cook’s distance (Cook, 1977) and fit models with and without such values to assess their impact on model characteristics. Variance inflation factors were used to screen for problematic multicollinearity. There was no missing data for our key independent variables and only one case had missing data for the dependent variables.

Regression diagnostics showed generally robust characteristics. All variance inflation factors were less than <1.5, indicating a lack of problematic multicollinearity. Although models for depression in both Latinxs and Whites did show evidence of significant heteroscedasticity, heteroscedasticity-consistent standard error estimates inflated standard errors only for regression coefficients for perceived stress and its interaction with social support and by a maximum of 30%, too little to substantively impact statistical significance. Several cases were found across models with Cook’s distance values exceeding the conventional cutoff of 4/n, but no more than 3 values larger than .1 were found in any model. Although excluding observations with Cook’s distance values >.1 did shift statistically significant regression coefficients by up to 20%, such exclusions did not substantively impact statistical significance for any statistically significant regression coefficients or bring any nonsignificant results into statistical significance. Link tests for model specification were nonsignificant for both the anxiety and depression models for Whites but were significant for Latinxs. However, the magnitude of the misspecifications, as measured by the coefficients of the square predicted values, was small [b = .01, t(261) = 2.69, p = .02 for anxiety; b = .01, t(261) = 2.43, p = .02 for depression].

Results

Demographics

Table 1 shows the demographic information about participants. White students, older by almost two years, had significantly higher average parental education (i.e., more bachelor’s degrees and graduate school experience and degrees) than Latinx students. More Latinx were second-generation college students than were White students. Mean scores and comparisons on the ISEL-12, PSS, BDI-II, and BAI for the two groups are also shown in Table 1. Latinx students reported significantly higher levels of belonging than their White counterparts [3.43 (Latinx) vs. 3.30 (White) on a scale of 1–4]. Overall, there was low variability in both age and immigration status in this sample. Fewer than 10% of both Latinx and White students reported being first generation, and only 8% of Whites and 3% of Latinxs reported being over age 30.

Buffering Effect of Social Support on Risk of Anxiety and Depression

Social support, perceived stress, depression, and anxiety are significantly intercorrelated for both Latinxs and Whites with perceived stress having a somewhat stronger correlation with social support among Whites (Table 2). In initial regression models, demographic variables were not found to be significant predictors of anxiety or depression for either Latinx or White students. Nonetheless, due to a lack of variability in this sample of some demographics, our results should not be taken as an indication of their general importance to populations with greater variability. In these initial regression models, without any interaction terms, we examined the unique contributions of focal variables (i.e., perceived stress and social support) to each outcome while controlling for each other and demographics. For Latinxs, both perceived stress and social support are important contributors to anxiety (β = .53, p < .001; β = −.17, p < .001, respectively) and depression symptoms (β = .62, p < .001; β = −.21, p < .001, respectively). For White students, only perceived stress was a significant contributor to anxiety (β = .54, p < .001), with the association between anxiety and social support not reaching statistical significance (β = −.07, p = .27). In terms of depression, findings for White students are consistent with those of Latinx students, with both perceived stress and social support contributing to depression symptomatology (β = .58, p < .001; β = −.27, p < .001, respectively). Thus, for depression, Latinx and White students had similar findings with both perceived stress and social support having significant main effects. However, for anxiety, results start to differ across racial/ethnic groups with social support only being significantly associated with this outcome for Latinx students.

Table 2.

Bivariate Correlations Among Main Variables of Interest

	1	2	3	4
1. Social support	—	−.28	−.38	−.30
2. Perceived stress	−.46	—	.69	.59
3. Depression	−.50	.69	—	.76
4. Anxiety	−.29	.59	.72	—

Note. Correlations for Latinx students shown above the diagonal (n = 270). Correlations for Non-Latinx White students shown below the diagonal (n = 221). All correlations are significant at p < .001.

Overall Model

The overall model included both groups (N = 480) with Latinxs coded as the reference category for ethnicity (Table 3). We found that the main effect of social support on anxiety and the effect of gender on anxiety did not vary by ethnic group. The three-way interaction term shown in Table 3 revealed a significant difference in social-support stress-buffering effects by ethnic group and by outcome. For anxiety, the buffering effects of social support are significantly different for Latinx compared to White students (B = 5.46, p < .01). However, moderating effects do not differ significantly by racial/ethnic group for depression (B = 1.46, p = .33). Thus, stress-buffering dynamics appear to operate similarly for both Latinx and White students when considering depressive symptomatology as the outcome but not when considering anxiety as the key outcome. In terms of main effects, both perceived stress and social support appear to be key variables associated with anxiety and depression for both groups.

Table 3.

Results of Three-Way Interaction Model Examining the Stress-Buffering Hypothesis for Latinx and Non-Latinx White Students

	Full sample
Effects	Coefficient (SE)	Significance
Anxiety
Gender	.78 (0.70)	.264
Age	−0.01 (0.02)	.570
Parental education	−0.08 (0.22)	.723
Immigranta	−2.53 (1.27)	.047
Perceived stress	7.41 (0.62)	.001
Social support	−2.71 (0.97)	.005
PS * SS	−6.91 (1.22)	.001
Ethnic group	−0.29 (0.70)	.680
PS * EG	−1.12 (1.01)	.272
SS * EG	2.29 (1.44)	.113
PS * SS * EG	5.46 (1.65)	.001
Constant	8.09 (0.66)	.001
Depression
Gender	0.46 (0.64)	.471
Age	0.01 (0.02)	.740
Parental education	0.05 (0.20)	.786
Immigranta	−1.21 (1.17)	.300
Perceived stress	8.30 (0.57)	.001
Social support	−3.55 (0.89)	.001
PS * SS	−4.98 (1.13)	.001
Ethnic group	0.42 (0.64)	.516
PS * EG	−0.47 (0.93)	.611
SS * EG	−0.33 (1.32)	.802
PS * SS * EG	1.46 (1.51)	.333
Constant	8.30 (1.25)	.001

Note.

^aImmigrant controls for generational status and compares immigrant and U.S.-born participants (reference group). None of the interactions including demographic variables (i.e., gender, age, or immigrant status were significant). Final parsimonious model presented with control variables included.

PS = Perceived Stress; SS = Social Support; EG = Ethnic Group)

Main Effects by Racial/Ethnic Group

Given the significant three-way interaction, we further explored main and moderating effects of social support by racial/ethnic group. Results by group are shown in Table 4. For Latinxs, we found a significant main effect of perceived stress on both anxiety (B = 7.51, p < .001) and depression (B = 8.31, p < .001), with higher levels of perceived stress being associated with more symptomatology. Social support was a significant protective factor for anxiety (B = −2.61, p < .05) and depression (B = −3.57, p < .001). For Whites, we also found a significant, and deleterious, main effect of perceived stress on both anxiety (B = 6.14, p < .001) and depression (B = 7.83, p < .001). Social support was not a significant predictor of anxiety (B = −0.67, p = .50); however, the main effect for depression (B = −3.97, p < .001) was significant, and protective, for this group. In summary, for anxiety, we found indications of a significant difference in the main effect of social support for Latinxs and Whites, while the impact of perceived stress did not seem to differ. For depression, the main effects of perceived stress and social support did not differ between Latinxs and White college students. Simple slope graphs depicting effects by racial/ethnic group and by outcome of interest are shown in Figures 1 and 2.

Table 4.

Results of Two-Way Interaction Regression Model Examining the Stress-Buffering Hypothesis for Latinx and Non-Latinx White Students

	Latinx		Non-Latinx White
Effects	Coefficient (SE)	Significance	Coefficient (SE)	Significance
Anxiety
Gender	0.57 (0.98)	.561	1.22 (0.96)	.205
Age	0.02 (0.03)	.517	−0.04 (.02)	.145
Parental education	−0.22 (.29)	.448	0.22 (.34)	.508
Immigrana	−2.45 (1.76)	.164	−3.04 (1.84)	.098
Perceived stress	7.51 (0.67)	.001	6.14 (0.77)	.001
Social support	−2.61 (1.04)	.012	−0.67 (0.99)	.503
PS * SS	−6.99 (1.31)	.001	−1.43 (0.99)	.148
Constant	8.29 (1.86)	.001	6.75 (2.25)	.003
Depression
Gender	0.76 (0.86)	.372	0.19 (0.96)	.840
Age	0.03 (.03)	.245	−0.01 (0.02)	.611
Parental education	0.02 (0.24)	.935	0.23 (0.34)	.501
Immigranta	−1.01 (1.53)	.509	−1.72 (1.83)	.348
Perceived stress	8.31 (0.58)	.001	7.83 (0.76)	.001
Social support	−3.57 (0.91)	.001	−3.97 (0.99)	.001
PS * SS	−4.89 (1.14)	.001	−3.42 (0.99)	.001
Constant	7.93 (1.62)	.001	8.27 (2.24)	.001

Note. Perceived stress and social support were grand mean centered. Given well-known methodological issues calculating and interpreting standardized regression coefficients when interactions are examined, unstandardized coefficients are reported. Demographic variables included as controls with parental education as proxy for income. PS = Perceived Stress; SS = Social Support.

^aImmigrant controls for generational status and compares immigrant and U.S.-born participants (reference group).

Figure 1. Depression Simple Slopes Interaction by Ethnicity.

Note. Figure showing relation between y and x as a function of z. CV_z = conditional values of z, which represents various levels of social support (${\text{cv}}_1=\overline{z}+1s{d}_z$, ${\text{cv}}_2=\overline{z}$, and ${\text{cv}}_3=\overline{z}-1s{d}_z$). See the online article for the color version of this figure.

Figure 2. Anxiety Simple Slopes Interaction by Ethnicity.

Note. Figure showing relation between y and x as a function of z. CV_z = conditional values of z, which represents various levels of social support (${\text{cv}}_1=\overline{z}+1s{d}_z$, ${\text{cv}}_2=\overline{z}$, and ${\text{cv}}_3=\overline{z}-1s{d}_z$). See the online article for the color version of this figure.

Buffering Effects by Ethnicity

No indications of group differences were found in social-support stress buffering when analyzing depression as an outcome. As seen in Table 4, social support interacts with perceived stress to lessen the impact of perceived stress on depression significantly for both Latinxs (B = −4.89, p < .001) and Whites (B = −3.42, p < .001). In this case, the negative effects of perceived stress on depression diminish at higher levels of social support for both groups (Figure 1). Nonetheless, for anxiety, social support significantly buffered the impact of perceived stress on anxiety for Latinxs (B = −6.99, p < .001) but not for Whites (B = −1.43, p = .15). In particular, at low levels of social support, stress has a greater impact on anxiety for Latinxs than for Whites, and this disparity narrows significantly as levels of social support increase (Figure 2). In other words, the negative effects of perceived stress on anxiety diminish at higher levels of social support for Latinx students, with increasing values of social support reducing the magnitude of the association between perceived stress and anxiety (i.e., significantly flattening the positive slope).

Effect Size

Effect sizes were calculated for each model in terms of proportion of explained variance in the form of R-squared values. In terms of anxiety, all models accounted for a significant proportion of the variance: R² = .42 for Latinxs, R² = .38 for White students, and R² = .40 for the combined model. Similarly, all depression models accounted for a significant proportion of the variance: R² = .53 for Latinxs, R² = .58 for White students, and R² = .54 for the combined model.

Discussion

Consistent with the literature, social support buffered the effects of perceived stress on depression symptoms similarly for both groups. However, for anxiety symptoms, group membership moderated this effect. Social support was a buffer against symptoms of anxiety in the context of perceived stress for Latinx but not for White students. These findings highlight potential key differences between Latinx and White students in the protective role of social support against different forms of psychological distress.

Degree of endorsement of depression and anxiety symptoms did not differ by group. That depression did not vary by group was unexpected in light of findings that suggest that students of color may be more likely to endorse depression (Eisenberg et al., 2013), whereas findings about differences in anxiety have been mixed. This study was conducted at a Hispanic Serving Institution (HSI), which may provide important context. Vaughan et al. (2015) reported that Latinx students enrolled at an HSI consumed less alcohol than Latinx students at non-HSIs. Although this study focused on alcohol use, it does speak to the possibility that attending an HSI as a Latinx student may be protective. To our knowledge, research investigating the extent to which enrollment at an HSI is protective is limited and definitely warrants further investigation.

Social Support as a Protective Factor

Consistent with other studies (Brausch & Decker, 2014; Uebelacker et al., 2013), social support was protective against depression symptoms in both study groups. Among college students, greater social support has been associated with lower levels of depressive symptoms following stressful life events (Ling et al., 2010). Social support has also been shown to mitigate the effects of stressful experiences, such as discrimination, on depressive symptoms (Kim, 2014).

Social support was protective against anxiety symptoms for Latinx students but not White students, suggesting that social support may be more critical for Latinxs in the context of anxiety. Particularly at low levels of social support, perceived stress had a stronger association with anxiety symptoms for Latinxs than it did at high levels of social support. No significant buffering effects of social support for anxiety were found for White students. Several different mechanisms may explain this group difference. It is possible that while overall levels of perceived social support are similar for both groups, there might be variation in how students use and receive that support. Previous studies have found that in more collectivistic cultures, social support is not only more readily available but also individuals have to make fewer explicit requests in order to receive support to cope with stressors including academics, relationships, and job concerns (Kim et al., 2008). Indeed, Latinx undergraduate students are more likely to live at home with their families during their college attendance than their White counterparts (Mae, 2014). For Latinxs, living at home potentially increases opportunities to receive direct help, support, and social interactions that can reduce social isolation and increase well-being. Conversely, the absence of social support may be more detrimental for Latinx students on mental health outcomes, especially in the context of stress.

Family respeto and religiosity have been found to moderate the relationship between discrimination related stress and anxiety among Latinx college students (Corona et al., 2017). These same cultural values were not a significant moderator when examining depression in the same study (Corona et al., 2017). Thus, it is possible that social support for Latinxs is closely tied to other cultural values such as familismo, respeto, and religiosity, which in turn have different associations with anxiety and depression. In alignment with this hypothesis, we found significant group differences in the belonging subscale of social support. Belonging, or the need to belong to a network of mutual support and connection, may reflect values and desires like collectivism and familismo.

Furthermore, Latinx students may rely more on social support to cope with additional stressors they encounter based on their minority status. Latinx students whose parents have low educational attainment may experience difficulty navigating higher education (Zalaquett, 1999). Latinxs are almost three times as likely as non-Latinxs to be first-generation college students (Saenz et al., 2007). First-generation Latinx college students reported facing greater perceptions of barriers to remain in school than second-generation students (Gloria et al., 2009). In addition, for many racial/ethnic minority students, including Latinxs, entering college may be their first exposure to being a minority in a predominantly White environment (Alvarez et al., 2009). Latinx college students most commonly cope with stressors by relying on social-support networks (Chiang et al., 2004).

Finally, as previously indicated, anxiety is more closely associated with the future-related concerns, while depression better aligns with the past events (Eysenck et al., 2010). In this regard, social support may be a stronger buffer of the effects of stress on anxiety for Latinxs who may have greater concerns about future postgraduation transitions and career-oriented concerns. Given their minority status, and often lower educational attainment among their family members, Latinxs potentially have fewer collective experiences and support during this transition than Whites. Moreover, for those with little social support, facing this transitional challenge alone can lead to added stress, which could subsequently have a greater negative impact on anxiety. In contrast, for those with high levels of social support, the effects of stress on anxiety may be like that of Whites; participants with high levels of social support may know that others will support them through their transitions and will assist them in times of need.

Limitations and Strengths

This study has several limitations. First, the present study investigated perceived social support in general and did not specify the source (e.g., family or friend) nor did it investigate the source(s) of stress. Notably, we did not use a measure of acculturative stress. In addition, investigating the relationship between class level (i.e. freshman, sophomore, junior, and senior) and perceived stress could have provided important context, as levels of stress may vary. In addition, the cross-sectional nature of our study limits the extent to which cause–effect relationships may be inferred. However, theoretical and empirical evidences based on stress-coping models and other frameworks suggest that stress can be conceptualized as the initial temporal variable, while health outcomes are thought of as consequences of the stress (Clark et al., 1999; Leblanc et al., 2004; Pearlin et al., 1990). Furthermore, our study focused on the undergraduate student population recruited from introductory psychology courses; thus, the findings may not generalize to the general undergraduate population or other age or racial/ethnic groups. Moreover, an index of SES was based on mother’s and father’s education level, which likely does not capture the complexities of the SES construct. Future studies with larger samples and different demographics should investigate other potential SES variables. Our study is also limited by relying on an all English-speaking, primarily nontreatment seeking sample. It is important to consider how our model would apply to immigrant Latinx college students with limited English proficiency given additional stressors, including acculturative stress and the current political climate and anti-immigrant sentiment. Finally, we investigated the moderating role of group membership (Latinx or White) on the relationship between stress and anxiety and depression. Future studies should further explore the unique protective role of social support and differences based on racial/ethnic group membership and detangle potential mechanisms and differences between sources of social support, as well as sources of stress.

The present study offers several important contributions. First, we tested a theoretical model exploring differences among Latinxs and White college students for the stress-buffering hypothesis across two mental health outcomes. Second, we included more complex modeling to investigate the buffering effect of social support on anxiety than in previous studies. Regression approaches that have neglected considering interaction terms fail to detect subtle but important group differences. Our results showed a moderated buffering effect of social support on anxiety. Thus, our study revealed a potential difference in the role of social support as a protective factor that can serve to uncover and explore other group differences and constructs relevant to Latinxs and other minority groups. Furthermore, social-support measures do not always include various aspects of social support (e.g., emotional, tangible, and belonging), and oftentimes concepts such as social networks or social integration are conflated with social support (Gottlieb & Bergen, 2010). We assessed social support as a larger construct rather than concentrating on the sources such as family or friends’ support.

Implications

Postsecondary education harbors unique and potentially impactful stressors that may contribute to negative mental health outcomes. In the context of anxiety, social support appears to be more useful and relevant for Latinxs than for Whites. Interventions that specifically target social support may be more valuable for Latinxs presenting anxiety symptoms than for their White counterparts. For instance, clinicians could consider employing behavioral activation in an effort to increase engagement in pleasant and meaningful activity that would increase contact with Latinx students’ support network. Similarly, problem-solving approaches could be utilized to assist in the identification of social support and address barriers to accessing support. Acceptance-based interventions could focus on increasing values congruent activity, which might include reaching out to family and/or friends, despite anxiety. Moreover, policies that address informal and formal sources of support on college campuses may contribute to decreased stress and improved mental health outcomes. Such interventions could take many forms, including supporting student organizations and centers that provide guidance and gathering opportunities for Latinx students.

Further research is necessary to identify trends that might have been too nuanced in our study (e.g., type of social support, and source of support) to understand differences between Latinxs and Whites. For instance, perceived support has been found to be associated with a reduction on subsequent depression, whereas instrumental support has been shown to predict depression (Cheng, 1998). Unpacking the construct of social support and identifying how type and source of social support might differentially impact depression and anxiety can further our understanding of between-group differences. Another consideration is the extent to which isolation might account for group differences in this study. Social isolation is a common feature of depression for which social support might be particularly beneficial. However, for anxiety, this may not be the case as it is not characterized by feelings of isolation to the same extent that depression is. The possibility that social support is not as strong as a buffer against anxiety should be considered further.

Conclusions

Our findings suggest that social support is protective against the adverse effects of perceived stress on depression for both Latinx and White college students. However, social support moderated the relationship between perceived stress and symptoms of anxiety only for Latinx college students, suggesting that social support may be particularly important to Latinx students in the context of stress to prevent anxiety symptoms. This study adds to the growing literature providing evidence on the differential role of social support on psychological distress between Latinxs and White college students. These findings suggest that social support does not function in the same way across ethnic groups and across types of distress. These results are important to consider when developing preventive and intervention programs for ethnically diverse college students.

Public Significance Statement.

Pursuing a college degree can be a stressful experience, especially for Latinx students who experience additional stressors including acculturative stress and perceived discrimination that may place them at high risk of developing depression and anxiety. This study suggests that social support is protective against the adverse effects of perceived stress on depression for both Latinx and White college students; however, social support may be particularly important for Latinx students in the context of stress to prevent anxiety symptoms. The differing effects of social support across ethnic groups and across types of distress are important to consider when developing preventive and intervention programs for ethnically diverse college students.