Introduction

Adolescence is a time of considerable psychological and social changes characterized by an increase in peer interpersonal relationships, school demands, and a high risk for depressive symptoms. Peer relationships and the school environment are important developmental context associated with stressors and demands that have been linked with depressive symptoms among adolescents (Hankin et al. 2007). At the same time, the prevalence of depressive symptoms increases across adolescence, and females of Mexican-origin show a higher risk for depressive symptoms compared to females from other ethnic-racial groups (Joiner et al. 2001). During adolescence, parental support can serve a protective role that ameliorates the association between stressors and depressive symptoms and promotes adolescents’ healthy development (Rueger and Malecki 2011). The stress exposure (Rudolph and Hammen 1999) and stress generation model (Hammen 1999) highlight the importance of considering the reciprocal link between stressors and depressive symptoms to better understand the direction of association. Differences in reports of peer and school stressors between early and middle adolescents also point to the importance of examining when stressors within these domains may be more prevalent (Hankin et al. 2007). Extant literature, however, has primarily focused on cross-sectional relations between stressors and depressive symptoms, and less is known about the changes in these relations across adolescence and directions of the effects. The current study builds on prior work by examining (1) the longitudinal and reciprocal associations at three time points across four years between peer stressors, school stressors, and depressive symptoms among a sample of Mexican-origin adolescent females (2) the protective role of parental support and (3) potential cohort differences between early-mid and middle-late adolescence.

Domain Specific Stressors and Depressive Symptoms During Adolescence

Peer and school stressors are important contexts for adolescent development and mental health (Morales and Guerra 2006). Research on Latino adolescents highlights the role of domain specific stressors (i.e., peer and school stressors), which grants the opportunity to investigate the unique associations between specific stressors and adolescents’ adjustment outcomes. For example, in a cross-sectional study of recently immigrated adolescents, Latinos reported more peer stressors compared to their Asian and Caribbean counterparts (Patel et al. 2016). In another cross-sectional study where over half of the sample was of Mexican origin, family and peer stressors were the salient stress domains that were significantly and positively associated with more depressive symptoms (Deardorff et al. 2003). A cross-sectional study with an ethnically diverse sample of youth found that family stressors and peer stressors were positively associated with depressive symptoms (Gonzales et al. 2001). In addition, a longitudinal study on a diverse sample of elementary school children found that school stressors, which included measures of school and peer problems, were associated with lower academic outcomes, aggression, and depressive symptoms (Morales and Guerra 2006). Altogether, past work, mostly cross-sectional, provides evidence for considering domain specific stressors such as peer and school stressors and their contribution to Mexican-origin females’ adjustment.

Additional longitudinal research is needed to unpack the potential reciprocal associations between domain specific stressors and depressive symptoms that are postulated in seminal theoretical work. Specifically, the stress exposure model (Rudolph and Hammen 1999) indicates that stressors and negative life events increase the risk for depressive symptoms. Alternatively, the stress generation model (Hammen 1991) suggests that depressive symptoms predict later stressors such that individuals who report depressive symptoms are active contributors of their environment and, in turn, actively generate stressful experiences through their behaviors and characteristics (Hammen 2006). The current study combines the stress exposure and stress generation model to test more rigorously the direction of association between stressors and depressive symptoms across time. Consistent with both theoretical models, there is evidence with ethnically diverse samples (Hankin et al. 2007) that early peer stressors are associated with later depressive symptoms (i.e., stress exposure) and that adolescents’ depressive symptoms are related to more stressors, particularly romantic stressors (i.e., stress generation). These complex associations between stressors and depressive symptoms, however, have not been examined with Mexican-origin adolescent samples. Given that Mexican-origin adolescent females are at high risk for depressive symptoms, it is important to consider how stressors experienced within key developmental contexts may be related to future depressive symptoms and how depressive symptoms are associated with future reports of peer and school stressors among this population. Accordingly, using longitudinal data across adolescence, the reciprocal associations between stressors and depressive symptoms among Mexican-origin females were examined to reveal the directionality of these associations and identify when associations may be most pertinent for this population group who are at high risk for depressive symptoms.

Peer Stressors and Depressive Symptoms

Given that youth spend a substantial amount of time with peers, examining the reciprocal process between peer stressors and depressive symptoms lets us delve into an important developmental context (Parker et al. 2006). The peer context allows adolescents to develop their social networks and build closer relationships, but experiences with peers differ between males and females, with females reporting more intimate friendships compared to males (Parker et al. 2006; Zarbatany et al. 2000). The different experiences that males and females report on interpersonal relationships may, in turn, influence perceptions of stress within peer contexts. Among non-Latino samples, females, for example, reported more friendship stress than males, whereas males reported more peer group stress than females (Rudolph 2002). For early adolescent girls, friendship stress has been associated with higher levels of depressive and anxiety symptoms (Rudolph 2002). Peer-related relationship losses among females (i.e., close friend and romantic breakup) have also been associated with more internalizing (i.e., anxiety/depressive symptoms) and externalizing (i.e., rule-breaking behavior) problems compared to males (Bakker et al. 2010). These findings indicate that females may be at a higher risk for suffering negative consequences from experiencing peer stressors compared to males.

Peer relationships become increasingly more salient across adolescence (Brown and Larson 2009), which may influence how early versus middle adolescent girls are affected by peer stressors. A longitudinal study examining differences between two ethnically diverse cohorts, 8th graders and 10th graders, found that peer stressors decreased across time but, on average, 10th graders encountered more peer stressors compared to 8th graders (Hankin et al. 2007). In addition, for both cohorts, peer stressors were associated with depressive symptoms, whereas depressive symptoms were not related to peer stressors (Hankin et al. 2007). Altogether, past work suggests that peer relationships become an increasingly important socializing context for adolescents that can have detrimental health consequences, which may be more pertinent at certain stages of adolescence such that peer stressors may increase from early to middle adolescence as youth become more embedded within the peer context (Hankin et al. 2007). In addition, the burgeoning literature has provided support for the stress exposure model where stressors are positively associated with depressive symptoms but not the other direction for the stress generation model, suggesting that during adolescence peer stressors may take the dominant role in leading to more depressive symptoms. Although past work has not found that depressive symptoms are associated with more peer stressors, research examining bidirectional associations among Latino adolescents is non-existent. Accordingly, to better understand the exchanges between the peer context and depressive symptoms among Mexican-origin adolescent females—a group understudied for these associations but at high risk for depressive symptoms, both directions in the longitudinal associations between peer stressors and depressive symptoms were examined. Further, given the developmental changes in the encounters of peer relationships during adolescence, potential cohort differences in these associations between early versus middle adolescent females at three time points across four years were also examined.

School Stressors and Depressive Symptoms

Schools are an important context that can present developmentally-related stressors and demands for adolescents (Eccles and Roeser 2011). Qualitative accounts, not specific to Latino youth, suggest that sources of school stressors may include students’ own high standards for themselves and feelings of being pressured by parental expectations, both of which were more likely to be reported by females than males (Låftman et al. 2013). Considering that adolescents spend a significant amount of time in the school context, it is important to understand the potential implications of school stressors on depressive symptoms and whether depressive symptoms may be associated with future school stressors.

Research examining associations between school-related stressors and depressive symptoms so far has found support for the stress exposure model, such that school stressors are associated with more depressive symptoms. A study with non-Latino samples suggested that school stressors, such as schoolwork pressures and being treated poorly by teachers, were associated with psychosomatic pain and psychological outcomes such as sadness, irritability, and feeling unsafe and nervous (Hjern et al. 2008). Additional school-related stressors such as school performance (e.g., difficulties with school subjects) have been associated with depressive symptoms across studies (Moksnes et al. 2014; Moksnes et al. 2016). A cross-sectional study with an ethnically diverse sample of adolescents, however, did not find a significant link between school stressors and depressive symptoms (Deardorff et al. 2003). Past work examining whether depressive symptoms predict later school stressors is non-existent. Therefore, this study examined potential reciprocal associations for a better understanding of whether depressive symptoms play a role in stressors that adolescents experience within the school context.

School stressors may change across adolescence and differ for girls, compared to boys, which in turn may have different implications for depressive symptoms at different time points in girls. In one study with an ethnically diverse sample of adolescents, described above, school stressors declined across time for the 8th grade cohort whereas, for 10th graders, school-related stressors did not change across time (Hankin et al. 2007). These results provide evidence on the importance of examining the longitudinal association between school stressors and depressive symptoms across adolescence. Finally, as mentioned above, the influence of certain domain specific stressors on adjustment may be more prevalent at certain developmental stages, such that reports on school stressors may be different among early and middle adolescents (Hankin et al. 2007). Indeed, compared to elementary schools, secondary schools are an important developmental context that presents unique demands and expectations for adolescents, such as more competitive environments and increased interpersonal relationships with more teachers. Thus, beyond testing the reciprocal associations between school stressors and depressive symptoms, in this study potential differences in these associations among a cohort of early and middle adolescent females were examined.

Potential Protective Factors

In the presence of stressors and their potential detrimental contribution to adolescent depressive symptoms, identifying protective mechanisms is of utmost importance. According to the stress-buffering model, emotional resources can minimize the effect of stressors on health outcomes (Cohen and Wills 1995), suggesting a potential buffering effect of emotional support on the associations between stressful experiences and depressive symptoms. From among multiple sources of emotional support, the current study examined parental support as a moderator in the relation between peer and school stressors and depressive symptoms across time. Research has shown the protective role of parental support in adolescents’ stress experiences. For example, adolescents who reported low levels of parent support experienced more interpersonal stressors (Auerbach et al. 2011). Further, a study on an ethnically diverse sample of adolescents found a significant association between peer stress and depressive symptoms only when parent-child relationship quality was low, not high (Hazel et al. 2014). Less is known about the role of parental support in the association between peer and school stressors and depressive symptoms among Latino adolescents over time. Previous work has mixed findings on whether parental support could exacerbate or protect Latino adolescents from stressors and depressive symptoms. A cross-sectional study with the same sample of Mexican-origin adolescent females from the current study found that for middle adolescent females who reported high levels of maternal support, experiences of discrimination from teachers were associated with more depressive symptoms—an exacerbating effect (Kam and Bámaca-Colbert 2013). In contrast, another study, using a sample of Mexican adolescents and young adults, found that family support buffered the effect of stress on depressive symptoms (Raffaelli et al. 2013). To illuminate the role of parental support in adolescents’ experiences of stressors and depressive symptoms over time, the current study examined parental support as a potential moderator in the longitudinal associations between peer and school stressors and depressive symptoms among early and middle adolescents.

Covariates

In examining the longitudinal stressor-depressive symptoms association and the moderating role of parental support, the potential confounding effects of adolescents’ nativity and their family structure were taken into account. Past work on Latino adolescents suggests that birth country has been associated with youth stress-related experiences and adjustment outcomes. Study findings among young adults suggest that foreign-born adults reported less stressful events compared to U.S. born adults and foreign-born women experienced a decline in depressive symptoms compared to U.S. born women (Tillman and Weiss 2009). Also, among Latino adolescents, U.S. born youth experienced more family conflict and less family support compared to foreign-born youth (Barajas-Gonzalez and Brooks-Gunn 2014). Further, adolescents’ family structure has been associated with youth’s adjustment outcomes. In particular, adolescents in single-parent families were more likely to report poor adjustment outcomes such as more depressive symptoms and more family conflict compared to adolescents in two-parent families (Zeiders et al. 2011). Therefore, in the present study adolescents’ birth country and family structure were included as covariates.

Current Study

Adolescence is a time of considerable changes characterized by increased peer relationships, school demands, and a high risk for depressive symptoms. Guided by the stress exposure model and the stress generation model and past empirical work, the first aim of the current study was to examine the concurrent and long-term reciprocal associations between domain specific stressors (i.e., peer and school) and depressive symptoms across three time points. Further, guided by the stress-buffering model, the second aim was to examine whether parental support would moderate the link between stressors and depressive symptoms across time. This study used a cross-lagged panel model to test the reciprocal and longitudinal linkages between stressors and depressive symptoms and whether parental support moderated the association between peer or school stressors and depressive symptoms. Finally, in the face of developmental changes in levels of stress experiences among adolescents, multiple group analysis was conducted to test whether associations differed among a cohort of Mexican-origin early-to-middle and middle-to-late adolescent females.

Past work suggests that peer stressors are associated with more depressive symptoms (Deardorff et al. 2003), therefore, it was hypothesized that peer stressors would be associated with future depressive symptoms among this sample. Studies have also shown that school-related stressors are associated with more depressive symptoms in adolescence (Moksnes et al. 2016). Thus, it was hypothesized that school stressors would be associated with later depressive symptoms. Given the limited empirical research examining whether depressive symptoms are associated with later peer or school stressors, these lagged associations were exploratory. Based on previous empirical findings showing that support moderates the link between stressors and depressive symptoms (Raffaelli et al. 2013), it was hypothesized that parental support would moderate the reciprocal associations between stressors and depressive symptoms across time. Peer stressors may become more salient among middle adolescents compared to early adolescents, suggesting that reports on peer-related stressors differ across adolescence and may play a bigger role among middle adolescents (Hankin et al. 2007). Therefore, it was hypothesized that peer stressors at an earlier time would be associated with later depressive symptoms and these associations may be particularly salient for adolescents who are in middle adolescence. Finally, no hypothesis was made about whether differences would emerge in the association between school stressors and depressive symptoms among early and middle adolescents.

Method

Participants

Data were from a longitudinal study focused on cultural, relational, and developmental influences on the well-being of Mexican-origin adolescent females (Bámaca-Colbert et al. 2014). Adolescent daughters and their mothers participated at baseline (i.e., T1) but only daughters participated at subsequent assessments (i.e., Time 2 (T2) and Time 3 (T3)). Thus, the current study included adolescent data across the three assessments. A total of 338 girls participated at baseline, 155 participated at T2, and 146 participated at T3. At baseline, 59.8% of the adolescents lived with both birth mother and birth father, 21.3% lived with birth mother only, 16.3% lived with birth mother and stepfather, 1.5% lived with birth father and stepmother, and 1.2% lived with other. A majority of the adolescents (66%) were born in the U.S. Early adolescents were in 7th grade at T1 (n = 170, Mage= 12.27, SD = 0.48), in 9th grade at T2 (n = 78, Mage=14.81, SD = 0.49), and in 10th grade at T3 (n= 76, Mage = 15.84, SD = 0.52). Middle adolescents were in 10th grade at T1 (n = 168, Mage = 15.21, SD = 0.46), in 12th grade at T2 (n= 75, Mage=17.76, SD = 0.54), and the majority were post-high school at T3 (n = 70, Mage= 18.69, SD = 0.58).

Design and Procedures

Participants were recruited from middle schools and high schools in a southwestern metropolitan area (for a detailed description of the recruitment procedure, see Bámaca-Colbert and Gayles 2010). School district and principal approval were received from all participating schools. School personnel distributed in classrooms a recruitment letter that explained the purpose of the study and eligibility criteria: adolescents had to be in 7th or 10th grade and both parents had to be of Mexican descent. Regardless of mothers’ interest in participating, all students who returned the forms were entered into a drawing to win a gift card. At T1, adolescents who returned signed assent and consent forms completed a self-administered questionnaire at their school and research assistants were available for assistance. The questionnaire was available in English or Spanish and took approximately one hour to complete. Adolescent participants were invited for a follow-up survey 2.5 years later for T2 and 3.5 years later for T3. Participants were compensated for their participation. Reasons for attrition in the subsequent waves were due to inability to connect with participants (e.g., change in residence, unanswered phone calls, disconnected or incorrect number). To handle missing data, Full Information Maximum Likelihood was used to estimate the models. This approach uses all of the available information to model parameter estimates. In addition, sensitivity analysis was conducted to check the robustness of the findings.

Measures

Peer stressors

Peer stressors were measured using the eight-items subscale from the Multicultural Events Scale for Adolescents (MESA; Gonzales et al. 2001). The peer stressors subscale measures adolescents’ interpersonal conflict with peers at school (e.g., “Other teens said mean or bad things to you.”, “You had a disagreement or fight with a close friend.”) and this subscale has been used with Latino samples (Santiago et al. 2017). Adolescents reported whether certain events happened to them in the past three months (Yes = 1, No = 0) and scores were summed to create a count of the number of stressors each participant experienced.

School stressors

School stressors were measured using the seven-item subscale from the Multicultural Events Scale for Adolescents (MESA; Gonzales et al. 2001). The school stressors subscale measures adolescents’ interpersonal conflict with school personnel (e.g., “You had a disagreement with a teacher or principal”) and stressors related to doing well in school (e.g., “You did poorly on an exam or school assignment”). This subscale has been used with Latino samples (Santiago et al. 2017). Adolescents reported whether certain events happened to them in the past three months (Yes = 1, No = 0) and scores were summed to create a count of the number of stressors each participant experienced.

Depressive symptoms

Depressive symptoms were measured using the 20-item Center for Epidemiologic Studies Depression Scale (CES-D; Radloff 1977). The CES-D measures the prevalence of depressive symptoms and includes subscales of somatic symptoms, interpersonal problems, and positive and negative affect (e.g., “I was bothered by things that usually don’t bother me.”, “I had crying spells”). This measure has been used with Mexican-origin samples (Bámaca-Colbert et al. 2012). Participants responded to statements about how often they felt depressive symptoms within the last week on a 4-point Likert scale (0 = rarely or none of the time (less than 1day) to 3 = mostly or almost all the time (5–7 days). Positively worded items were reverse coded, and all items were averaged into an overall depressive symptoms score. The alpha coefficients for the current sample ranged from 0.90 to 0.92 for T1 to T3.

Parental support

Parental support was measured using a nine-item adapted version of the Inventory of Parent and Peer Attachment Scale (Armsden and Greenberg 1987). This measure assesses adolescent’s positive bonds to mothers and fathers. Adolescents responded to questions about mothers and fathers, separately (e.g., “I tell my parent about my worries and problems”, “This parent shows me that he/she understands me”). This measure has been used with Mexican-origin samples (Bámaca-Colbert et al. 2012). Items were rated on a 4-point Likert scale (1 =almost never or never to 4 =almost always or always). Past research has calculated the average score of adolescent reported scores of relationship quality with mothers and fathers to capture the positive relationship that adolescents experienced with both parents (Hazel et al. 2014). In the current study, items for mothers and fathers (18-items) were averaged into an overall parental support score separately at T1 and T2. Alpha coefficients for the current sample were 0.94 for T1 and 0.94 for T2.

Covariates

In the current study adolescent’s family structure (intact family = 1, other = 0) and adolescent birth country (U.S. = 1, Mexico = 0), were included as covariates.

Data Analysis Strategy

Correlations were estimated to examine the association among all variables of interest. Table 1 includes the correlations and descriptive statistics for the early- and middle-adolescent cohorts. The skewness depicted in Table 1 indicated that the data met the normality assumption (Kline 2015). Next, a Little’s missing completely at random test (Little 1988) indicated that the data were missing completely at random for the peer stress model \(\chi ^2\left( {{\mathrm{53}}} \right) = {\mathrm{63}}{\mathrm{.376}}\), p = 0.16 and the school stress model \({\mathrm{\chi }}^2\left( {{\mathrm{53}}} \right) = {\mathrm{63}}{\mathrm{.409}}\), p = 0.16.

Table 1 Correlations, means, and standard deviations among variables of interest for early and middle adolescent cohort
Full size table

The structural equation model was estimated using Mplus version 8 (Muthén and Muthén 1998–2017). Because the data for the current study met the assumptions of normality and the missing data was missing completely at random, to handle missing data across variables which ranged from as low as less than 1% for peer stressors T1 and as high as 57.10% for depressive symptoms T3, Full Information Maximum Likelihood (FIML) was used for the model estimation (Widaman 2006). Model fit was evaluated using the standard chi-square statistical fit index and degrees of freedom, root mean square error of approximation (RMSEA; Browne and Cudeck 1992), standardized root mean residual (SRMR; Bentler 1995), and the comparative fit index (CFI; Bentler 1990) as measures of overall fit. Model fit was considered good when the RMSEA value was less than 0.05, SRMR was less than 0.08, and CFI value was greater than 0.95. Multiple group analysis was used to test the hypothesized model with the developmental period as a grouping variable (early adolescent cohort and middle adolescent cohort at Time 1).

To test the hypothesis of reciprocal influences between domain specific stressors and depressive symptoms, one model was estimated for peer stressors (PS) and depressive symptoms (DS) and a separate model was estimated for school stressors (SS) and depressive symptoms (DS). Specifically, a cross-lagged analytic approach was employed using three waves of data for each model. Cross-lagged models are a useful strategy for estimating reciprocal associations (Selig and Little 2012). Autoregressive paths were also estimated to account for the stability of the construct from one time point to the next and the correlations of the variables within each wave were estimated. PST1, SST1, and DST1, were treated as the exogenous variables and PST2, PST3, SST2, SST3, DST2, and DST3 were treated as the endogenous variables. To test for moderation, all variables in the model were centered, and parental support T1 was included as a moderator between (1) PST1 and DST2, (2) DST1 and PST2, (3) SST1 and DST2, and (4) DST1 and SST2 and parental support T2 as a moderator between (1) PST2 and DST3, (2) DST2 and PST3, (3) SST2 and DST3, and (4) DST2 and SST3. Control variables were also included in the model and adolescents’ birth country and family structure were regressed on PS, SS, and DS at Time 2 and Time 3.

Results

Descriptives

Independent samples t-test were conducted to test for potential cohort differences for early and middle adolescents. No significant differences were found in depressive symptoms (DS). There was a significant difference in peer stressors T1 (PST1), early adolescents (M= 2.40, SD= 2.26) reported more PST1 than middle adolescents (M = 1.92, SD = 2.03), t(329.50) = 2.05, p = 0.04 and for peer stressors T3 (PST3), early adolescents (M = 2.16, SD = 1.97) reported more PST3 than middle adolescents (M = 1.39, SD = 2.00), t(144) = 2.35, p = 0.02. For school stressors T3 (SST3), early adolescents (M = 1.83, SD= 1.37) reported significantly more SST3 than middle adolescents (M= 0.87, SD= 1.18), t(144) = 4.51, p= 0.00. For parental supportT1, early adolescents (M = 3.18, SD= 0.71) reported more parental support T1 than middle adolescents (M = 2.98, SD = 0.74), t(336) = 2.53, p= 0.01, and for parental supportT2, early adolescents (M = 2.56, SD= 0.76) reported less parental support T2 than middle adolescents (M = 2.85, SD = 0.76), t(149) = −2.35, p= 0.02. No other significant cohort differences were found.

Peer Stress Models

Initial model

For the overall sample, an unconstrained model was fit to the data first. Autoregressive paths were estimated for each construct (e.g., PST1→ PST2, PST2→ PST3). Cross-lagged paths (e.g., PST1→ DST2, DST1→ PST2) and within time associations were included (e.g., PST1 ←→ DST1). Parental support T1 and T2 were included as moderators on the cross-lagged paths. All exogenous variables were correlated with the covariates (i.e., adolescent birth country, family structure) and all endogenous variables were regressed on the covariates. Fit indices indicated that the model fit the data well \(\chi ^2\left( {{\mathrm{14}}} \right)\) = 6.012, p = 0.97; CFI = 1.00; RMSEA = 0.000; SRMR = 0.015.

Autoregressive paths indicated that PST1 was significantly associated with PST2 (β = 0.48, p= 0.00) and PST2 was significantly associated with PST3 (β = 0.50, p= 0.00). DST1 was significantly associated with DST2 (β = 0.30, p= 0.00) and DST2 was significantly associated with DST3 (β = 0.53, p= 0.00). Cross-lagged paths indicated that PST2 was significantly associated with more DST3 (β = 0.15, p= 0.04). The interaction between parental support T1 and PST1 was significantly associated with DST2 (β = −0.20, p= 0.00), and the interaction between parental support T2 and DST2 was significantly associated with PST3 (β = −0.19, p= 0.02). There were no significant findings for the covariates.

To examine the interaction between parental support, PS, and DS, the moderator was centered (parental support T1) and multiplied by the centered PS and DS variables (Aiken and West 1991). Simple slopes were plotted and tested for significance. First, simple slopes were calculated with values corresponding to ± 1 standard deviation above and below the mean. Among adolescents who reported low levels of parental support T1, PST1 was significantly related with more DST2 (B = 0.09, p= 0.00) but not among adolescents who reported high levels of parental support (B = −.05, p = .12). Among adolescents who reported low levels of parental support T2, DST2 was significantly related to more PST3 (B = 0.93, p= 0.03) but not among adolescents who reported high parental support (B = −0.85, p = 0.16).

Multiple group analysis

Next, the equivalence of paths across early and middle adolescent groups was tested using multiple group analysis. A fully constrained model was compared to a freely estimated model. The chi-square difference test results indicated that the fully constrained model where all paths were equal for early and middle adolescents was significantly different from the freely estimated model \(\chi ^2\)(77) = 152.487, p < 0.05. The model fit indices for the constrained model indicated that the model did not fit the data well \(\chi ^2\left( {{\mathrm{105}}} \right)\) = 170.752, p = 0.00; CFI = 0.701; RMSEA = 0.061; SRMR = 0.074. This suggested that the paths for early adolescents and middle adolescents should not be set to be equal across groups. The model with the unconstrained paths was retained as the final model as this model fit the data well \(\chi ^2\left( {{\mathrm{28}}} \right)\) = 18.265, p = 0.92; CFI = 1.00; RMSEA = 0.000; SRMR = 0.029. See Fig. 1 and Table 2 for path coefficient values.

Fig. 1
figure 1

Model fit: χ2(28) = 18.265, p > 0.05; CFI = 1.00; RMSEA = 0.00; SRMR = 0.029. Standardized regression coefficients for the longitudinal association between peer stressors (PS), depressive symptoms (DS), and the interaction effect of parental support for early adolescents and middle adolescents presented after the forward slash. Correlations among the exogenous variables, within wave correlations, and covariates regressed on the endogenous variables were estimated but not depicted for clarity of presentation. Adolescents’ grade in school at each time point is depicted above. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image
Table 2 Multiple group reciprocal and longitudinal associations among stressors, depressive symptoms, and parental support for early and middle adolescents
Full size table

For early adolescents, findings indicted that all autoregressive paths were significant (i.e., PST1 to PST2; PST2 to PST3; DST1 to DST2; DST2 to DST3. Cross-lagged paths indicated that PST2 was associated with more DST3. There were no other significant findings. The proportion of the variance explained by the model for PST2 was R2 = 0.21, PST3R2 = 0.23, DST2R2 = 0.27, and DST3 was R2 = 0.46.

For middle adolescents, findings indicted that all autoregressive paths were significant (i.e., PST1 to PST2; PST2 to PST3; DST1 to DST2; DST2 to DST3). Cross-lagged paths indicated that DST2 was associated with more PST3. Results indicated a significant interaction between PST1 and parental support T1 on DST2. The interactions were plotted using simple slope calculations with values corresponding to ± standard deviation above and below the mean. Among adolescents who reported low parental support, PST1 was associated with more DST2 (B = 0.15, p= 0.00) but not among middle adolescents who reported high levels of parental support (B= −0.08, p = 0.16), see Fig. 2.

Fig. 2
figure 2

Among middle-late adolescents who reported low parental support, peer stressors T1 was associated with more depressive symptoms T2. Unstandardized coefficients are presented. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

Adolescents’ birth country was significantly associated with DST2 (β = −0.25, p = 0.02) such that adolescents who were born in the U.S. reported less DST2. Adolescents’ family structure was significantly associated with PST3 (β = 0.21, p = 0.04) such that adolescents who lived in intact families reported more PST3. There were no other significant associations. For middle adolescents, the proportion of variance explained by the model for PST2 was R2 = 0.28, PST3R2 = 0.42, DST2R2 = 0.31, and DST3 was R2 = 0.39.

School Stress Models

Initial model

For the overall sample, an unconstrained model was fit to the data first. Autoregressive paths were estimated for each construct (e.g., SST1 → SST2, SST2 → SST3). Cross-lagged paths (e.g., SST1 → DST2, DST1 → SST2) and within time association were included (e.g., SST1 ←→ DST1). Parental support was included as a moderator on the cross-lagged paths. All exogenous variables were correlated with the covariates (i.e., adolescent birth country, family structure) and all endogenous variables were regressed on the covariates. Fit indices indicated that the model fit the data well \(\chi ^2\left( {{\mathrm{14}}} \right)\) = 19.061, p = 0.16; CFI = 0.975; RMSEA = 0.033; SRMR = 0.030.

Autoregressive paths indicated that SST1 was significantly associated with SST2 (β = 0.32, p = 0.00) and SST2 was significantly associated with SST3 (β = 0.40, p = 0.00). DST1 was significantly associated with DST2 (β = 0.32, p= 0.00) and DST2 was significantly associated with DST3 (β = 0.55, p = 0.00). Cross-lagged paths indicated that SST1 was significantly associated with DST2 (β = 0.19, p = 0.00). Adolescents’ family structure was significantly associated with SST3 (β = 0.17, p= 0.02) such that adolescents in an intact family structure reported more SST3. There were no other significant associations.

Multiple group analysis

Next, the equivalence of paths across early and middle adolescent groups was tested using multiple group analysis. A fully constrained model was compared to a freely estimated model. The chi-square difference test results indicated that the fully constrained model where all paths were equal for early and middle adolescents was significantly different from the freely estimated model \(\chi ^2\)(77) = 104.932, p< 0.05. The model fit indices for the constrained model indicated that the model did not fit the data well \(\chi ^2\left( {{\mathrm{105}}} \right)\) = 141.338, p = 0.01; CFI = 0.827; RMSEA = 0.045; SRMR = 0.087. This suggested that the paths for early adolescents and middle adolescents should not be set to be equal across groups. The fully unconstrained model had a good model fit \(\chi ^2\left( {{\mathrm{28}}} \right)\) = 36.406, p= 0.13; CFI = 0.96; RMSEA = 0.042; SRMR = 0.040 and was retained as the final model. See Fig. 3 and Table 2 for path coefficient values.

Fig. 3
figure 3

Model fit: χ2(28) = 36.406, p > 0.05; CFI = 0.960; RMSEA = 0.042; SRMR = 0.040. Standardized regression coefficients for the longitudinal association between school stressors (SS), depressive symptoms (DS), and the interaction effect of parental support for early adolescents and middle adolescents presented after the forward slash. Correlations among the exogenous variables, within wave correlations, and covariates regressed on the endogenous variables were estimated but not depicted for clarity of presentation. Adolescents’ grade in school at each time point is depicted above. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

For early adolescents, the autoregressive paths were significant (i.e., SST1 to SST2; SST2 to SST3; DST1 to DST2; DST2 to DST3). There were no significant cross-lagged associations and no significant interactions. Finally, family structure was significantly associated with SST3 (β = 0.21, p = 0.04) such that adolescents in an intact family structure reported more SST3. For early adolescents, the proportion of the variance explained for SST2 was R2 = 0.15, SST3 was R2= 0.41, DST2 was R2= 0.28, and DST3 was R2 = 0.38.

For middle adolescents, the autoregressive paths were significant (i.e., SST1 to SST2; SST2 to SST3; DST1 to DST2; DST2 to DST3). Cross-lagged paths indicated that DST2 was significantly associated with more SST3 (β = 0.34, p = 0.02). Findings indicated a significant interaction between parental support T2 and DST2 on SST3. Simple slopes were calculated with values corresponding to ± 1 standard deviation above and below the mean. Among adolescents who reported low parental support, DST2 was associated with more SST3 (\(B = {\mathrm{1}}{\mathrm{.32}}\), p = 0.000) but not among middle adolescents who reported high levels of parental support at T1 (\(B = {\mathrm{ - 0}}{\mathrm{.01}}\), p = .983), see Fig. 4. Finally, adolescents birth country was significantly associated with DST2 (β = −0.25, p= 0.02) such that adolescents who were born in the U.S. reported fewer DST2. Adolescents birth country was also significantly associated with SST3 (β = 0.24, p = 0.03) such that adolescent born in the U.S. reported more SST3. Lastly, adolescents living in an intact family structure reported more SST3 (β = 0.21, p = 0.05). For middle adolescents, the proportion of variance explained for SST2 was R2 = 0.15, SST3R2 = 0.41, DST2R2 = 0.28, and DST3 was R2 = 0.38. The current study findings provide support for the reciprocal associations between stressors and depressive symptoms and the moderating role of parental support among an early and middle adolescent cohort at three time points across four years.

Fig. 4
figure 4

Among late adolescents who reported low parental support, depressive symptoms T2 was associated with more school stressors T3. Unstandardized coefficients are presented. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

Sensitivity Analysis

The robustness of the above findings was checked by including a set of auxiliary variables. The auxiliary variables, not part of the analysis, were used to help with FIML estimation (Collins et al. 2001). The inclusion of auxiliary variables helps the precision and efficiency of the model and does not compromise the model of substantial interest (Graham 2003). Auxiliary variables included in these models were adolescents reports on anxiety symptoms (20-item State-Trait Anxiety Inventory for Children-Trait Version (STAIC-T); Spielberger et al. 1973) and family economic stressors, both from Time 1 (10-item, Multicultural Events Scale for Adolescents (MESA); Gonzales et al. 2001). Most of the substantive findings held in the models with auxiliary variables but more nuances emerged. For example, for the peer stress model, the findings for the early adolescent cohort held with the exception that the cross-lagged association PST2→DST3 was no longer significant. For the middle adolescent cohort, the findings held, and one new finding emerged such that among adolescents who reported low parental support, DST2 was significantly related to more PST3 (B = 1.785, p = 0.000). For the school stress model, the findings for the early adolescent cohort held and one new finding emerged such that DST1 was significantly related to SST2. For the middle adolescent cohort, the main findings held, and one new finding emerged such that among adolescents who reported low parental support, SST1 was significantly related to DST2 (B = 0.179, p = 0.002).

Discussion

In adolescence, peer relationships and the school environment are important developmental contexts that present new experiences and demands. This is a period when there is increased interpersonal stressors in the peer and school contexts coupled with depressive symptoms. Alongside, parental support continues to be an important buffer in adolescent adjustment. However, there is a scarcity of research addressing how peer and school stressors and depressive symptoms are associated across time and how parental support moderates this link among Mexican-origin samples. Furthermore, it is unclear whether these associations may be different across early-middle and middle-late adolescence. The current study was guided by three theoretical frameworks: the stress exposure model, which posits that stressors contribute to depressive symptoms, the stress generation model, which suggests that depressive symptoms may play a role in reports of future stressors, and the stress-buffering model, which posits that support may buffer against the positive link between stressors and depressive symptoms. Using data from a longitudinal study of Mexican-origin female adolescents, the current study aimed to addresses existing gaps in the literature by examining the reciprocal and longitudinal associations between stressors and depressive symptoms in early and middle adolescence. Furthermore, parental support was examined as a moderator across time between stressors and depressive symptoms. Overall, the current study findings were in line with the stress generation, stress exposure, and stress-buffering models such that prior levels of peer stressors were associated with subsequent depressive symptoms and prior levels of depressive symptoms were associated with later peer and school stressors. In addition, parental support moderated the link between stressors and depressive symptoms and differences emerged among the early and middle adolescent cohorts.

This study’s findings provide support for the longitudinal associations among peer and school stressors and depressive symptoms and for differences across adolescence (i.e., between early and middle adolescent cohorts) in relational patterns. In line with the stress exposure model and past work that found that peer stressors were associated with depressive symptoms (Hazel et al. 2014) and that peer salience increases across adolescence (Brown and Larson 2009), it was hypothesized that prior peer stressors would be associated with later depressive symptoms particularly among middle adolescents. This study’s findings supported the hypothesis, such that for early adolescents who were in middle adolescence at Time 2, peer stressors T2 were associated with more depressive symptoms T3. As adolescents begin to spend more time with peers, adolescent females may be more likely to experience stressors and challenges within this key developmental context. The stressors experienced early on may contribute to later feelings of depressive symptoms especially during this developmental stage when adolescent females may be more likely to report peer-related stressors within a significant context for emotional experiences (Hankin et al. 2007). Early adolescents who were in mid-adolescence at Time 2 may be more vulnerable and more likely to report challenges within the peer context given the importance that adolescent females place on peers and the instability of peer relationship that youth may experience during this developmental stage (Brown and Larson 2009; Rudolph 2002). Furthermore, it is possible that among Mexican-origin females, the salience of peer stressors emerges at a later time such as middle-late adolescence due to socializing experiences in the family that may keep younger females closer to the family context, resulting in relationships outside of the family to take on greater importance later in adolescence. As adolescents move into high school and begin to navigate new peer contexts, these experiences may be associated with stressors and challenges within new interpersonal relationships. These findings provide support for the salience of peer-related stressors at certain developmental stages that are associated with later depressive symptoms. These results may inform schools, teachers, staff, and parents on the importance of the peer contexts, such that interpersonal stressors occurring within this domain may negatively influence the well-being of Mexican-origin females in middle adolescence.

Guided by the stress generation model (Hammen 1991), the prospective association between prior levels of depressive symptoms and future stressors was exploratory. This study’s findings provided partial support and suggest that among the middle adolescent cohort, who were in late adolescence at Time 2, depressive symptoms T2 were associated with more peer stressors T3. Thus, these results provide support for the idea that earlier levels of depressive symptoms place individuals at a vulnerability that may influence perceived stressors (Liu and Alloy 2010). Further, these results support the importance of considering how depressive symptoms may spillover and play a role in other areas of adolescents’ lives including perceptions of stress within the peer context. These results also suggest the importance for future research to consider the reciprocal associations between depressive symptoms and reports of domain specific stressors. Past work suggests that depressive symptoms may peak during middle to late adolescence (Adkins et al. 2009) which highlight the importance of examining how symptoms of depression may contribute to experiences within key developmental domains (e.g., interpersonal) during adolescence.

In line with the stress-buffering hypothesis, it was hypothesized that parental support would moderate the link between peer stressors and depressive symptoms. Findings supported this premise but only among the middle adolescent cohort at Time 1. Specifically, for middle adolescents who reported low levels of parental support at Time 1, peer stressors T1 were associated with more depressive symptoms T2 (2.5 years later from the initial assessment). This is consistent with past research that has found that stressors are associated with more depressive symptoms under conditions of low support (Raffaelli et al. 2013). Based on this finding, the link between peer stress and depressive symptoms, in the context of parental support, appears to vary across developmental stages (Hankin et al. 2007) and suggest that low parental support may place adolescent females at a higher risk in middle-to-late adolescence. Thus, specific stressors appear to be more detrimental under conditions of low parental support, underscoring the salience of parental support in protecting youth from negative encounters within the peer context from middle to late adolescence. Among adolescent females both family and peer interpersonal relationships are especially important (Rudolph 2002). Therefore, perceptions of low parental support, compounded with stressors may be especially detrimental and may influence females’ responses to stressors within specific domains such as the peer context. These findings underscore the importance for school programs and interventions to continually engage and make parents aware that the emotional support that they provide to their children may ameliorate the association between stressors their youth experience and later mental health problems.

The current study’s findings also pointed to another significant interaction, such that among adolescents who reported low levels of parental support at Time 2, depressive symptoms T2 was associated with more school stressors T3. As stated before, findings such as this underscore the important role that parental support continues to play into late adolescence as it can help buffer the negative role that prior depressive symptoms may have on later school related stressors. These findings also point to the significance of school stressors at the developmental period of late adolescence. Mexican-origin adolescents in the middle adolescent cohort were likely to be preparing to transition away from high school between Times 2 and 3. It is possible that school-related stressors become more salient at this period of development such that these youth are beginning to think about the next stage such as graduating and continuing school, going to work, or both. Therefore, prior levels of depressive symptoms may play a role in future perceptions of challenges and stressors within the school domain such as doing poorly in school or experiencing negative interpersonal relationships with teachers. Overall, these findings indicate that parental support continues to prove an important source of support that may help offset the link between depressive symptoms and school stressors even when these youth are moving into late adolescence.

Although not the main focus of the study, it is important to note that some significant patterns emerged between covariates and variables of interest. For instance, middle adolescents born in the U.S. reported fewer depressive symptoms at Time 2. Interestingly, adolescents living in intact families reported more school stressors at Time 3 (both early and middle adolescent cohorts), and peer stressors at Time 3 (only middle adolescent cohort) compared to adolescents living in other family structures. Finally, middle adolescents born in the U.S reported more school stressors at Time 3 compared to middle adolescents born in Mexico.

The current study adds to existing knowledge on the longitudinal and reciprocal relations between stressors and depressive symptoms among Mexican-origin females, but results must be considered in light of some limitations. Specifically, the sample was comprised of Mexican-origin adolescents who attended schools in a southwestern metropolitan area in the U.S., therefore, findings cannot be generalized to females of Mexican-origin residing in other geographical locations. In addition, the larger project focused on females and, therefore, not examining these patterns among males of Mexican origin is another limitation. Further, from Time 1 to Time 2, there was a 2.5-year gap which may have contributed to the null cross-lagged findings in the current study from Time 1 to Time 2. Moreover, the current study focused on parental support, but it is as important to examine the role of other sources of support (e.g., teachers, friends, romantic partner) that become increasingly important in adolescence and may also protect adolescents against the positive association between stressors and depressive symptoms.

Despite the aforementioned caveats, the current study’s findings add to extant longitudinal research focused on Latino adolescents and underscore the peer and school contexts for Mexican-origin female adolescents’ depressive symptoms. Findings also draw attention to the salience of earlier parental support influences across adolescence and how supportive parenting moderated associations between different stressors and depressive symptoms. Findings call for prevention programs to continue to focus on parenting practices and highlight to parents and families the important implication that supportive parenting can have for adolescent long-term adaptation to external stressors. Programs should also continue to emphasize the importance of parental support across adolescence and discuss with parents the multiple stressors that their daughters can experience within the peer and school context, as parents, especially immigrant parents, may be unaware of these challenging experiences for their children. These findings are also important for school personnel and teachers. For example, school teachers and staff should consider how peer relationships and school experiences are important developmental contexts for adolescents that may contribute to later adjustment outcomes particularly for female adolescents who are Mexican-origin. Sources of support from teachers and staff may also help ameliorate the positive link between peer and school stressors and depressive symptoms. The peer and school environments are important developmental contexts for all adolescents and schools should consider working closely with parents to help support and ameliorate the stressful experiences that can play a role in adolescents’ depressive symptoms.

Conclusion

Past work underscores the role that stressors can have on adolescent depressive symptoms, but there is scant knowledge about how depressive symptoms are associated with reports of peer and school stressors across time. At the same time, parental support continues to play an important role across adolescence that may help offset the positive link between stressors and depressive symptoms. Given that Mexican-origin adolescent females show a higher risk for depressive symptoms during adolescence compared to adolescents from other ethnic-racial groups (Twenge and Nolen-Hoeksema 2002), the current study findings add to extant research focused on this population on how interpersonal stressors outside the family are implicated in their depressive symptoms. First, the results highlight the importance of considering the longitudinal and reciprocal association among stressors and depressive symptoms such that peer stressors were associated with more depressive symptoms and depressive symptoms were associated with more peer stressors for middle adolescents and late adolescents, respectively. Further, parental support emerged as a key source of support and as indicated by these findings, this source of support continues to have important implications in different areas for Mexican-origin females across adolescence. Specifically, among adolescents who reported low parental support, peer stressors were linked with more depressive symptoms and depressive symptoms were linked with more school stressors among middle-late adolescents. The implications of this study include the importance of considering the reciprocal and longitudinal associations among peer stressors, school stressors, depressive symptoms, and the moderating role of parental support. The study’s findings also point to the importance of examining differences across time in order to better understand how these important factors are associated with the well-being and adjustment outcomes across adolescence among Mexican-origin females. At a time when peers and school contexts are increasingly important, having a better understanding of how these contexts contribute to Mexican-origin females’ long-term adjustment is critical and the current study’s findings is an important step in this direction given the few existing studies focused on these contextual realities among this population.