Gender

Prior to adolescence (i.e., age 11), males and females have similar rates of depression. However, by age 16, compared to males, females are twice as likely to report symptoms of depression (Angold & Rutter, Reference Angold and Rutter1992), a trend that continues into adulthood (Gonzalez et al., Reference Gonzalez, Berry, McKnight-Eily, Strine, Edwards and Lu2010). Depression and anxiety often co-occur (Sartorius et al., Reference Sartorius, Ustun, Lecrubier and Wittchen1996), and adolescent females consistently report suffering from depression and anxiety at a higher rate than males. In a national survey of 10,123 adolescents, females were twice as likely as males to report suffering from either depression or anxiety (Merikangas et al., Reference Merikangas, He, Burstein, Swanson, Avenevoli and Cui2010), a finding that has been consistently replicated (Derdikman-Eiron et al., Reference Derdikman-Eiron, Indredavik, Bratberg, Taraldsen, Bakken and Colton2011; US Department of Health and Human Services, 2011). Perhaps because of these increased rates of internalizing disorders, compared to males, adolescent females report lower self-esteem (Derdikman-Eiron et al., Reference Derdikman-Eiron, Indredavik, Bratberg, Taraldsen, Bakken and Colton2011; Puskar et al., Reference Puskar, Bernardo, Ren, Haley, Tark and Switala2010). These findings are similar for Latino youth in terms of rates of depression (Cespedes & Huey, Reference Cespedes and Huey2008; McGee, Barber, Joseph, Dudley, & Howell, Reference McGee, Barber, Joseph, Dudley and Howell2005; Robles-Pina, Defrance, & Cox, Reference Robles-Pina, Defrance and Cox2008; Zeiders, Umana-Taylor, & Derlan, Reference Zeiders, Umana-Taylor and Derlan2012), anxiety (McGee et al., Reference McGee, Barber, Joseph, Dudley and Howell2005), and self-esteem (Zeiders et al., Reference Zeiders, Umana-Taylor and Derlan2012).

Age

In general, the risk of suffering from internalizing disorders increases with age. Rates of both anxiety and depression steadily increased in a national sample of adolescents between ages 13 and 18 (Merikangas et al., Reference Merikangas, He, Burstein, Swanson, Avenevoli and Cui2010), a finding that has been replicated (US Department of Health and Human Services, 2011; Twenge & Nolen-Hoeksema, Reference Twenge and Nolen-Hoeksema2002). Self-esteem also appears to increase with age (Wagner, Ludtke, Jonkmann, & Trautwein, Reference Wagner, Ludtke, Jonkmann and Trautwein2013), and in a sample of Latino adolescents, levels of reported self-esteem increased throughout high school (Zeiders, et al., 2013)

SES

Previous research on the influence of SES on adolescent mental health is inconsistent. Findings from a nationally representative sample of 15,112 adolescents yielded that a low SES was associated with an increased risk of suffering from depression (Goodman, Slap, & Huang, Reference Goodman, Slap and Huang2003), a finding that is in line with previous research (Goodman, Huang, Wade, & Kahn, Reference Goodman, Huang, Wade and Kahn2003; Roberts, Roberts, & Chen, Reference Roberts, Roberts and Chen1997). Contrary to these findings, a meta-analysis of 310 samples found that SES was not related to depression (Twenge & Nolen-Hoeksema, Reference Twenge and Nolen-Hoeksema2002). Another group of researchers found that there is a significant, negative relationship between SES and anxiety (Miech, Caspi, Moffitt, Wright, & Silva, Reference Miech, Caspi, Moffitt, Wright and Silva1999), a finding that mirrors national data indicating that SES is more strongly related to anxiety than to other affective disorders like depression (Kessler et al., Reference Kessler, McGonagle, Zhao, Nelson, Hughes and Eshleman1994). Given that individuals identifying as Latino are more likely to live in poverty than are individuals identifying as Caucasian or Asian (DeNavas-Walt, Proctor, & Smith, Reference DeNavas-Walt, Proctor and Smith2012), Latino adolescents may be at risk for suffering from depression and anxiety because of their increased risk of having low SES. There also appears to be a weak, but significant, positive relationship between SES and self-esteem (Mullins, Mullins, & Normandin, Reference Mullins, Mullins and Normandin1992; Twenge & Campbell, Reference Twenge and Campbell2002; Wiltfang & Scarbecz, Reference Wiltfang and Scarbecz1990). Finally, parent education level is often used as a measure of SES, and one study found that father's level of education had a small, positive effect on adolescent self-esteem (Wiltfang & Scarbecz, Reference Wiltfang and Scarbecz1990).

Parent–child conflict

Adolescence marks a period of exploration when youth seek to obtain independence from their parents. However, because of the centrality of family in Latino culture (Leidy, Guerra, & Toro, Reference Leidy, Guerra and Toro2012), parent and family factors continue to impact adolescent mental health functioning. For example, parent–child conflict was associated with increased levels of internalizing problems and decreased levels of self-esteem in Latino adolescents in several investigations (Kuhlberg Pena, & Zayas, Reference Kuhlberg, Pena and Zayas2010; Smokowski & Bacallao, Reference Smokowski and Bacallao2006; Smokowski, Rose, & Bacallao, Reference Smokowski, Rose and Bacallao2010). Parent–child conflicts about culture may be particularly harmful: one study found that cultural conflicts with mothers and fathers were significantly associated with decreased self-esteem, and conflicts with fathers were associated with increased depressive symptoms (Behnke et al., Reference Behnke, Plunkett, Sands and Bamaca-Colbert2011). In addition, normative parent–child conflicts are often exacerbated by acculturation stress (Szapocznik & Kurtines, Reference Szapocznik, Kurtines and Padilla1980; Szapocznik, Kurtines, & Fernandez, Reference Szapocznik, Kurtines and Fernandez1980). In contrast, supportive parenting behavior was associated with increased self-esteem in a sample of 324 9th- and 10th-grade Latino students (Bamaca, Umana-Taylor, Shin, & Alfaro, Reference Bamaca, Umana-Taylor, Shin and Alfaro2005).

Parent mental health

Parent–child interactions can be negatively impacted by poor parent mental health (for a review, see Lovejoy, Gracyk, O'Hare, & Neuman, Reference Lovejoy, Graczyk, O'Hare and Neuman2000), suggesting that adolescent mental health is also affected by parental mental health. Adolescent and parental mental health functioning are inextricably linked. A depressed, anxious, or fearful parent might be unable to adequately care for a child's physical and emotional needs, resulting in distress for that child. An extensive body of literature documents that higher levels of current depressive symptoms in parents are associated with higher levels of internalizing symptoms (e.g., depression and anxiety; for a review, see Beardslee, Gladstone, & O'Connor, Reference Beardslee, Gladstone and O'Connor2011; England & Sim, Reference England and Sim2009; Fear et al., Reference Fear, Champion, Forehand, Colletti, Reeslund and Roberts2009) and lower self-esteem (Hirsch, Moos, & Reischl, Reference Hirsch, Moos and Reischl1985) in children and adolescents. In a 20-year longitudinal study of 151 children, those who had severely depressed parents were three times more likely to suffer from anxiety, depression, and substance abuse at all time points compared to their counterparts who did not have a depressed parent (Weissman et al., Reference Weissman, Wickramaratne, Nomura, Warner, Pilowksy and Verdeli2006). However, less is known about this link in Latino immigrant families.

Genetic heritability is one plausible explanation for these results (O'Connor, McGuire, Reiss, Hetherington, & Plomin, Reference O'Connor, McGuire, Reiss, Hetherington and Plomin1998; Thapar & McGuffin, Reference Thapar and McGuffin1996). However, parenting styles, marital relationships, and parent–child interactions are negatively affected by poor parent mental health (see Lovejoy et al., Reference Lovejoy, Graczyk, O'Hare and Neuman2000, for a review) and may be as influential as genetics because of social learning. For example, in one study of 527 mothers and fathers of 12-year-old children, increased parental anxiety, depression, and social dysfunction were associated with higher levels of punitive parenting, lower levels of authoritative parenting, and less parent involvement (Leinonen, Solantaus, & Punamaki, Reference Leinonen, Solantaus and Punamaki2003). The increased stress caused by conflict ridden parenting styles, poor marital relationships, and dysfunctional parent–child interactions contribute to the decreased mental health functioning of the children of depressed and anxious parents.

Further, one study compared 124 families with a mentally ill parent to 127 families without a mentally ill parent and examined parenting behaviors as well as internalizing disorders in children ages 11–16. Results demonstrated that children from the families with a mentally ill parent had higher rates of internalizing problems. Families with a mentally ill parent had a more negative family environment, and mentally ill parents monitored their children less (Loon, Ven, Doesum, Witteman, & Hosman, Reference Loon, Ven, Doesum, Witteman and Hosman2013). In addition, parental anxiety, fear, depression, worry, or hopelessness may incapacitate parents to such a degree that their child(ren) feel the need to take care of them. Children of mentally ill parents often become caretakers and give up age-appropriate activities, such as spending time with friends or playing sports, in order to care for their parent (Aldridge & Becker, Reference Aldridge and Becker2003). Taking on the adult role of caretaker may lead to poor mental health functioning for the child and cut the child off from other healthy environmental relationships. The majority of the extant literature examining the impact of parent mental health on adolescent mental health examines rates of depression. The authors were unable to find studies of how parent levels of hopelessness, fear, and humiliation influence their offspring, areas that this study addresses. These factors take on particular importance within the acculturation process for immigrant families (Smokowski & Bacallao, Reference Smokowski and Bacallao2010).

Parents’ marital relationship

Research suggests that marital conflict hinders the formation of positive parent–child relationships (see Erel & Burman, Reference Erel and Burman1995, for a review), and a poor parent–child relationship is a risk factor for poor child mental health (e.g., depression; Hair, Moor, Hadley, Kaye, & Orthner, Reference Hair, Moor, Hadley, Kaye and Orthner2009; Stein, Gonzalez, & Huq, Reference Stein, Gonzalez and Huq2012). In addition, children whose parents have conflict-ridden marriages are at risk for poor mental health outcomes, such as depression and anxiety (Cummings & Davies, Reference Cummings and Davies2010; for a review, see Grych & Fincham, Reference Grych and Fincham1990). It follows that less marital conflict is associated with improved mental health outcomes for children and adolescents (Hair et al., Reference Hair, Moor, Hadley, Kaye and Orthner2009). One way of gauging the level of conflict in marriage is to assess marital satisfaction, also termed dyadic adjustment. In a nationally representative sample of 3,316 adolescents ages 12–14, those whose parents reported poor marital quality had worse physical, mental health, and substance use outcomes (Hair et al., Reference Hair, Moor, Hadley, Kaye and Orthner2009). Given that marital conflict is a risk factor for poor adolescent mental health outcomes, positive family environments are commonly associated with healthy adolescent mental functioning (Bacallao & Smokowski, Reference Bacallao and Smokowski2007).

Familism

Familism is defined as close-knit attitudes, behaviors, and family structures within an extended family system and is believed to be one of the most important factors influencing the lives of Latinos (Coohey, Reference Coohey2001, p. 30). Familism is the belief in the importance and centrality of one's family and can be a protective factor that functions to strengthen Latino families. For example, familism was a deterrent to child maltreatment in both Latino and non-Latino families (Coohey, Reference Coohey2001) and was negatively associated with acculturation stress (Gil et al., Reference Gil, Wagner and Vega2000). High levels of familism are also associated with decreased parent–child conflict (Kuhlberg et al., Reference Kuhlberg, Pena and Zayas2010; Smokowski, Rose, & Bacallao, Reference Smokowski, Rose and Bacallao2008). In addition, familism buffered against mental health problems in adolescents; high levels of familism were associated with fewer internalizing problems and higher self-esteem (Smokowski & Bacallao, Reference Smokowski and Bacallao2006). However, the pressure of being part of a tight-knit, unified family that values familism might also cause stress for adolescents. In one study of 226 Latina females, high levels of familism were associated with increased internalizing disorders (Kuhlberg et al., Reference Kuhlberg, Pena and Zayas2010).

Family cohesion and adaptability

A positive family climate is one in which family members are united and can adapt to developmental or contextual changes. These family characteristics result in increased family satisfaction, a protective factor that results in positive youth developmental outcomes. Family satisfaction was associated with higher levels of self-esteem and optimism and lower levels of hopelessness in a sample of 278 middle and high school students. In contrast, high levels of negative affect in the family were associated with lower self-esteem and optimism scores and higher reported hopelessness (Phillips, Reference Phillips2012). Cohesive families (families with high levels of emotional closeness, dependability, support, and clear communication) likely have a positive family environment that serves to bolster adolescent mental health. In support of this assertion, one study of 282 parent and child dyads (ages 9–12) reported that increases in family cohesion predicted increased child problem-solving skills and social self-efficacy (Leidy et al., Reference Leidy, Guerra and Toro2012). In another sample of 1,876 Latino youth, family cohesion was significantly associated with decreased depressive symptoms (Lorenzo-Blanco, Unger, Baezconde-Garbanati, Ritt-Olson, & Soto, Reference Lorenzo-Blanco, Unger, Baezconde-Garbanati, Ritt-Olson and Soto2012), a finding that has been replicated in other samples of Latino youth (Williams, Reference Williams2010). Family cohesion is also negatively associated with anxiety (Varela, Sanchez-Sosa, Biggs, & Luis, Reference Varela, Sanchez-Sosa, Biggs and Luis2009). However, another study found that family cohesion and adaptability made little impact on Latino adolescents’ levels of depression (Copeland, Reference Copeland1998) indicating the need for additional research.

Time living in the United States and assimilation

Minimal research investigates how the amount of time living in the United States impacts adolescent internalizing symptoms. An early study of Asian refugees resettling in Canada found that more time in the new country was associated with better mental health functioning (Beiser, Reference Beiser1988). This finding was replicated in a sample of 281 Latino adolescents ages 12–19 who had a reduced risk of reporting symptoms of anxiety and depression with increased time spent in the United States (Potochnick & Perreira, Reference Potochnick and Perreira2010). Given the initial shock of immigrating to a new country and being surrounded by a different language, strange customs, and an unfamiliar culture, it follows that this experience could negatively impact mental health.

As individuals become accustomed to their surroundings and the new culture, mental health may improve. Smokowski, Rose, and Bacallao (Reference Smokowski, Rose and Bacallao2010) reported that, over 3 years, Latino adolescents’ internalizing problems decreased and their self-esteem increased. However, this study of 349 Latino adolescents also found that increased time living in the United States was significantly associated with lower self-esteem (Smokowski, Rose, & Bacallao, Reference Smokowski, Rose and Bacallao2010). Perhaps increased time in a new country results in additional levels of acculturation stress, which is a known predictor of poor adolescent mental health (Hovey & King, Reference Hovey and King1996; Katragadda & Tidwell, Reference Katragadda and Tidwell1998; Sirin, Ryce, Gupta, & Rogers-Sirin, Reference Sirin, Ryce, Gupta and Rogers-Sirin2013; Stein et al., Reference Stein, Gonzalez and Huq2012). A third study found that time in the United States was not significantly associated with self-esteem (Cavazos-Rehg & DeLucia-Waack, Reference Cavazos-Rehg and DeLucia-Waack2009).

English fluency in the home is another indicator of assimilation in adolescents that has been linked to internalizing symptoms. In a sample of junior and senior high school students in Texas, the relative impact of SES, family composition, and linguistic fluency on anxiety was greater than any other factors (Glover, Pumariega, Holzer, Wise, & Rodriguez, Reference Glover, Pumariega, Holzer, III Wise and Rodriguez1999). Similarly, in a sample of multiethnic Latino high school students, English-language competence was linked with lower reports of depression (Rumbaut, Reference Rumbaut, Rumbaut and Cornelius1995). These inconsistencies in assimilation indicators and effects point to the need for additional research.

Acculturation conflict

The process of acculturation can be stressful and may result in negative mental health outcomes (e.g., depression, anxiety, and low self-esteem), especially if an individual's culture of origin and the dominant culture are conflicting. In one sample of 190 Latino adolescents in Grades 7 through 10, 75% reported acculturative stress (e.g., worrying about immigration, or discrepancies between culture of origin and host culture values; Stein et al., Reference Stein, Gonzalez and Huq2012). Acculturation stress has been associated with both anxiety and depression (e.g., Hovey & King, Reference Hovey and King1996; Katragadda & Tidwell, Reference Katragadda and Tidwell1998; Sirin et al., Reference Sirin, Ryce, Gupta and Rogers-Sirin2013, Stein et al., Reference Stein, Gonzalez and Huq2012). One study found that in a sample of 1,119 high school students, Latinos in the marginalized acculturation group (e.g., those with fewer attachments and adaptions to Latino and other cultures) reported higher rates of depression and lower levels of optimism compared to the bicultural group (Carvajal, Hanson, Romero, & Coyle Reference Carvajal, Hanson, Romero and Coyle2002). However, in a study of 1,922 Latino adolescents, acculturation stress was indirectly associated with increased depression through family conflict (Lorenzo-Blanco et al., Reference Lorenzo-Blanco, Unger, Baezconde-Garbanati, Ritt-Olson and Soto2012).

Acculturation conflict also appears to influence self-esteem. In a sample of Cuban adolescents, there was a significant correlation between acculturation conflicts and self-derogation (Vega, Gil, Warheit, Zimmerman, & Apospori, Reference Vega, Gil, Warheit, Zimmerman and Apospori1993). Another group of researchers found that acculturation stress was significantly associated with low self-esteem and that self-esteem mediated the effect of ethnic identity on externalizing problems (Schwartz et al., Reference Schwartz, Zamboanga and Hemandez Jarvis2007). This indicates that acculturation stress is associated with decreased self-esteem, which is a significant mediator between acculturation and behavioral outcomes.

Discrimination experiences

Discrimination experiences are another risk factor for poor adolescent mental health. Latino adolescents who did not primarily speak English at home were at an increased risk of being alienated and bullied compared to youth who spoke only English at home (Yu, Huang, Schwalberg, Overpeck, & Kogan, Reference Yu, Huang, Schwalberg, Overpeck and Kogan2003). Victimization experiences, especially perceived discrimination, predicted internalizing problems (e.g., depression and anxiety) and low self-esteem in a sample of Latino adolescents (Smokowski & Bacallao, Reference Smokowski and Bacallao2006, Reference Smokowski and Bacallao2007). In a sample of middle school students of Mexican descent, a higher number of stressors, which included in-group and out-group discrimination, were associated with depressive symptoms for both those born in the United States and those born in Mexico (Romero & Roberts, Reference Romero and Roberts2003). This finding was replicated in a sample of 442 middle and high school Latino students (Stein et al., Reference Stein, Gonzalez and Huq2012) and a sample of 1,124 Latino adolescents (Lorenzo-Blanco et al., Reference Lorenzo-Blanco, Unger, Ritt-Olson, Soto and Baezconde-Garbanati2011).

Culture-of-origin involvement

Culture-of-origin involvement refers to an individual's sense of belonging to his or her ethnic group and the influence that ethnic group membership has on individual perceptions, feelings, and behavior. Culture-of-origin involvement can be seen as a counterpart to assimilation on the spectrum of acculturation and describes an individual's connection to his or her native culture, whereas assimilation denotes the connection with the adopted culture. Some research suggests that Latino enculturation, maintaining aspects of traditional Latino culture, may be protective. McDonald et al. (Reference McDonald, McCabe, Yeh, Lau, Garland and Hough2005) found that, for adolescent Latinas, high affiliation with Latino culture protected those with low self-esteem from developing later internalizing symptoms. It seems that a degree of culture-of-origin involvement leads to positive mental health outcomes, such as ethnic identity and self-esteem (Gonzales, Knight, Birman, & Sirolli, Reference Gonzales, Knight, Birman, Sirolli, Maton, Shellenbach, Leadbetter and Solarz2004; Martinez & Dukes, Reference Martinez and Dukes1997; Phinney, Cantu, & Kurtz, Reference Phinney, Cantu and Kurtz1997; Phinney & Chavira, Reference Phinney and Chavira1995). In one study of 669 Latino, African American, and Caucasian US-born high school students, researchers found that ethnic identity was associated with higher self-esteem in all three ethnic groups. In contrast, assimilation (i.e., US culture identity) predicted self-esteem only for Caucasian adolescents (Phinney et al., Reference Phinney, Cantu and Kurtz1997).

The ability to maintain culture-of-origin identity and simultaneously adopt elements of the host culture (i.e., biculturalism) is associated with higher self-esteem, stronger social skills, and heightened psychological well-being (Bautista de Domanico, Crawford, & Wolfe, Reference Bautista de Domanico, Crawford and Wolfe1994; LaFromboise et al., Reference LaFromboise, Coleman and Gerton1993; Smokowski & Bacallao, Reference Smokowski and Bacallao2010). In addition, bicultural adolescents had the lowest levels of acculturation stress and were less likely to report low levels of family pride compared to low- and high-assimilated Latino adolescents (Gil, Vega, & Dimas, Reference Gil, Vega and Dimas1994). The positive outcomes of biculturalism extend to families as well. One study found that compared to low- and high-assimilated families, bicultural families displayed significantly lower levels of conflict and were more helpful, committed to, and supportive of family members (Miranda, Estrada, & Firpo-Jimenez, Reference Miranda, Estrada and Firpo-Jimenez2000).

Time

Because exact dates for each interview were obtained, the multilevel model (Raudenbush & Bryk, Reference Raudenbush and Bryk2002) used a variable occasions design for time in which time was measured continuously rather than with a fixed occasions design using the wave indicators (W = 0, 1, 2, 3, pertaining to baseline and 6-, 12-, and 18-month follow-ups, respectively). Wave indicators occur at arbitrary times for each respondent, making the variable occasions design preferable. In the context of our objective to examine the change in internalizing symptoms among Latino adolescents, years living in the United States anchored time to a common experience for all adolescents in the sample. We were thus able to more intuitively interpret change in internalizing symptoms relative to this common experience rather than to some arbitrary study time point. For adolescents born in the United States, years living in the United States was equivalent to their age. Time was centered on the mean years living in the United States for each adolescent over the four measurement occasions, a center that may have occurred between measurement occasions.

Because there were four measurement occasions, we included both linear (T) and quadratic (T ²) variables. In the presence of the quadratic term, the actual value of linear change depends upon the value of time at which it is assessed and is interpreted as the instantaneous rate of change in internalizing symptoms per month that the adolescent lived in the United States. With time mean centered, it is also possible to interpret the linear time coefficient as the rate of change at time zero or the adolescents’ mean time living in the United States. The quadratic component is the change in this rate at all time points, with positive values indicating that the slope increases or gets steeper over time and negative values indicating that the slope decreases or gets flatter over time (regardless of whether the slope is upward or downward). With the exact amount of time measured in years, we have greater precision in the time measure, which is essential given the use of quadratic time.

Child characteristics

Child characteristics included gender (1 = female, 0 = male), mean centered age (in years), parent education level (1 = elementary or no schooling, 0 = high school or higher), and sample mean-centered household annual income (in thousands of dollars). Child characteristics and all other independent variables were treated as time invariant and measured at baseline data collection at the beginning of the internalizing trajectory.

Parent and adolescent culture-of-origin involvement

Culture-of-origin involvement, the process of maintaining ethnic identity through use of native language, media use, and enacting traditions from one's native culture, was assessed using a 20-item subscale from the Bicultural Involvement Questionnaire (BIQ; Szapocznik et al., Reference Szapocznik, Kurtines and Fernandez1980). Items were rated on a 5-point Likert scale ranging from not at all to very much. Example items included “How comfortable do you feel speaking Spanish (at home, with friends, in general)” and “How much do you enjoy . . . music, television programs, books, and magazines from your native country?” Internal consistency reliability was 0.89 for adolescents’ and 0.90 for parents’ culture-of-origin involvement in the current sample.

Parent and adolescent US cultural involvement

US cultural involvement, defined as assimilation of host culture language, media, norms, and traditions, was also measured using a subscale of the BIQ (Szapocznik et al., Reference Szapocznik, Kurtines and Fernandez1980). This subscale exactly parallels the culture-of-origin involvement subscale and has 20 items that measure participants’ use of English language, non-Latino US food, recreation, and media. Items were rated on a 5-point Likert scale with anchors labeled not at all to very much. In this sample, internal consistency reliability was 0.90 for adolescents and 0.93 for parents.

Parent and adolescent biculturalism

Parent and adolescent biculturalism were calculated by adding the culture-of-origin involvement and U.S. cultural involvement subscales of the BIQ (Szapocznik et al., Reference Szapocznik, Kurtines and Fernandez1980). This rendered an overall continuous measure of the parent's and adolescent's involvement in both their culture of origin and the US culture (i.e., biculturalism). In this sample, internal consistency reliability was 0.91 for adolescents and 0.92 parents.

Parent and adolescent acculturation conflict

Acculturation conflict, the degree to which foreign-born individuals experience problems as a result of adopting US culture, was measured with a four-item scale created by Vega and colleagues (Reference Vega, Kolody, Aguilar-Gaxiola, Alderete, Catalano and Caraveo-Anduaga1998) and response options were measured on a 5-point Likert scale with anchors ranging from not at all to frequently. Items included the following: (a) How often have you had problems with your family because you prefer American customs? (b) How often do you think that you would rather be more American if you had a chance? (c) How often do you get upset at your parents because they don't know American ways [not in parent version]? (d) How often do you feel uncomfortable having to choose between non-Latin and Latin ways of doing things? In the current sample, internal consistency reliability was 0.76 for adolescents and 0.87 for parents.

Parent and adolescent perceived discrimination

Perceived discrimination, the degree to which individuals feel that they are treated unfairly because of their race, was measured using three items and a 5-point Likert scale with anchors ranging from not at all to frequently (Vega, Khoury, Zimmerman, Gil, & Warheit, Reference Vega, Khoury, Zimmerman, Gil and Warheit1995). One item asked, “How often do people dislike you because you are Latino?” The other questions asked about unfair treatment because of being Latino. Higher scores indicated increased perceptions of discrimination. Internal consistency reliability was 0.75 for both adolescents and parents in the current sample.

Adolescent report on parent–adolescent conflict

Parent–adolescent conflict was assessed using the Conflict Behavior Questionnaire—20 (Robin & Foster, Reference Robin and Foster1989). The 20-item questionnaire assesses positive and negative parent–adolescent interactions that occur in nonconflictual and argumentative exchanges. Items are rated as yes or no. Example items included “My parent(s) don't understand me” and “My parent(s) put me down.” The reliability was 0.89 for adolescents in this sample.

Adolescent report of familism

Familism refers to the degree of family unity and trust present in the family and was assessed using seven items rated on a 4-point Likert scale with anchors labeled as strongly agree to strongly disagree (Gil et al., Reference Gil, Wagner and Vega2000). Example items included “We share similar values and beliefs as a family” and “Family members respect one another.” Internal consistency reliability was 0.87 in the current sample.

Parent report of family cohesion

Family cohesion, the degree to which family members support one another and spend time together, was measured using 8 items from Olson's (Reference Olson1992) 16-item family cohesion subscale from the Family Adaptability and Cohesion Evaluation Scale II survey. Six items were not included because of translation difficulties and 2 items were removed because they were repetitive. Example items included “Family members are supportive of each other during difficult times” and “Our family does things together.” Items were rated on a 5-point Likert scale with not at all and all the time serving as anchors. Internal consistency reliability was 0.77 in the current sample.

Parent report of family adaptability

Family adaptability refers to the level of flexibility a family has regarding rules, roles, discipline, and negotiations. Twelve items from Olson's (Reference Olson1992) 14-item family adaptability subscale from the Family Adaptability and Cohesion Evaluation Scale II survey were used to assess levels of family adaptability. Two items were removed because they were repetitive. Example items included “Discipline is fair in our family” and “Our family tries new ways of dealing with problems.” Internal consistency reliability was 0.72 in the current sample.

Parent report of parent depression

Parent depression was measured using 12 items from the 20-item Center for Epidemiological Studies Depression Scale (Radloff, Reference Radloff1977). Eight items were removed because they did not assess constructs of depression (e.g., I felt fearful) or would not have been accurately translated into Spanish (e.g., I could not get going). Participants were asked to indicate how often over the past 7 days “I felt depressed” or “I had trouble keeping my mind on what I was doing,” and items were rated on a 4-point Likert scale (rarely or none of the time/less than once a day, some or a little of the time/1–2 days, occasionally/3–4 days, or most or all of the time/5–7 days). Internal consistency reliability was 0.81 in this sample.

Parent report of parent humiliation

Parent's fear of humiliation was assessed with Hartling's (Reference Hartling1995) seven-item fear of humiliation scale. Items were rated on a 5-point Likert scale with anchors of not at all and extremely. Example items included “At this point in your life, how concerned are you about being . . . Teased? Treated as invisible? Called names or referred to in derogatory terms?” Internal consistency reliability was 0.87 in the current sample.

Parent report of parent hopelessness

Parents’ level of hopelessness was assessed using 13 items from Beck's 20-item hopelessness scale (Beck, Weissman, Lester, & Trexler, Reference Beck, Weissman, Lester and Trexler1974). Seven items were removed for brevity on a long interview. Example items included “I look forward to the future with enthusiasm” and “My future seems dark to me.” Response options included true or false. Internal consistency reliability was 0.73 in the current sample.

Parent report of parent self-esteem

Self-esteem was measured using 7 items from Rosenberg's (Reference Rosenberg1965) 10-item self-esteem scale. Three items were removed to limit the length of the interviews. Items were rated on a 4-point Likert scale (strongly agree, agree, disagree, or strongly disagree), and example items included “I feel positive about myself” and “All in all, I feel that I am a failure.” Internal consistency reliability was 0.88 in the current sample.

Parent report of parent worry

Parents’ level of worry was assessed using 5 items from the 16-item Penn State Worry Questionnaire (Meyer, Miller, Metzger, & Borkovec, Reference Meyer, Miller, Metzger and Borkovec1990). A subset of items was chosen for brevity on a long interview. Items were rated on a 5-point Likert scale with anchors including strongly disagree and strongly agree. Example items included “My worries overwhelm me” and “I am always worrying about something.” Internal consistency reliability was 0.90 in the current sample.

Parent report of parent fear and avoidance

Parent fear and avoidance were measured with a 13-item modified version of the Fear Questionnaire (Marks & Mathews, Reference Marks and Mathews1979). Participants were asked, “How much do you fear and avoid the following situations?” and were then presented with potentially anxiety-producing situations (e.g., speaking in public, being embarrassed or humiliated, traveling alone by bus). Eleven of the original 15 items were used, and 2 additional items were added that specifically addressed fears of an immigrant population (i.e., Speaking in English and Speaking to Americans). Items were rated on a 3-point Likert scale (never, sometimes, or often), and internal consistency reliability was 0.84 in this sample.

Parent report of Dyadic Adjustment Scale

Dyadic adjustment assesses the quality of a marital relationship. Spanier's (Reference Spanier1976) 19-item Dyadic Adjustment Scale was used to assess marital quality in the current sample. Items were rated on a 5-point scale (all the time, most of the time, sometimes, hardly ever, or never). Example items included “How often have you considered divorce, separation, or terminating the relationship?” and “How often do you confide in your partner?” Internal consistency reliability was 0.98 in the current sample.

Parent report of Conflict Tactics Scale

Conflict tactics refer to how parent(s) handle conflict with their partner. Eight items from the 14-item Conflict Tactics Scale (Straus, Reference Straus1979) were used in the current study. Items were rated on a 5-point Likert scale (all the time, most of the time, sometimes, hardly ever, or never). Participants responded to the prompt “When you and your partner disagree, how often do you do the following?” Eight possible behaviors were listed, such as “I discuss things calmly” and “I hit my partner.” Participants then rated these eight behaviors for their partner. Internal consistency reliability was 0.95 in the current sample.

Adolescent internalizing symptoms

Adolescents reported on their internalizing symptoms using the Youth Self-Report (YSR; Achenbach, Reference Achenbach1991). This subscale consists of 24 items using a 3-point Likert scale (not true, sometimes true, or often true), and example items included “There is very little that I enjoy” and “I cry a lot.” Prior studies indicate that the YSR has internal consistency reliability over 0.90 and a test–retest reliability of 0.80 (Achenbach, Reference Achenbach1991). The internal consistency reliability was 0.85 for adolescents in this sample.

Parents reported on adolescent internalizing symptoms using the 25-item internalizing problems subscale of the Child Behavior Check List (CBCL/4–18; Achenbach, Reference Achenbach1991). The CBCL/4–18 contains similar questions as the YSR, but they are answered by the parent about the adolescent. The CBCL has been widely used with Latino children (see Vega et al., Reference Vega, Khoury, Zimmerman, Gil and Warheit1995), is considered a “gold standard” of child mental health measures, and has been used in thousands of studies (Rush & The Task Force for the Handbook of Psychiatric Measures, Reference Rush2000). Parents rated their children on a number of internalizing behaviors, such as “cries a lot,” “feels worthless or inferior,” and “is nervous, high strung, or tense.” The internal consistency reliability was 0.89 in the current sample.

Hierarchical linear model specifications

The unconditional intraclass correlation, which measures the proportion of variation (excluding variation over time) attributed to raters and adolescents, was shown to be 75% at the rater level and 25% at the participant level (with time included, 48% and 16%, respectively). These levels represent conclusive evidence that the data lacked the independence required of conventional linear modeling methods (such as ordinary least squares). Consequently, we used a hierarchical linear model (Raudenbush & Bryk, Reference Raudenbush and Bryk2002). The first level of the multilevel model consisted of parent and adolescent raters’ assessment of adolescent internalizing symptoms (INTS) taken at four time points with the time (T; instantaneous rate of change) and quadratic (T ²; change in change) variables.

(1) $$\matrix{{\hbox {INTS}}_{tij} &= {\rm \beta}_{0ij} + {\rm \beta}_{1\iota} T_{tij} + {\rm \beta}_{2ij} T^2_{tij} \cr &\quad {+}\, r_{tij} \, \lsqb r_{tij} \sim N \lpar 0\comma \; {\rm \sigma}\,^2\rpar \rsqb.}$$

The second level regressed the parameters of the first level model on population averages with accommodations made for variation in the intercept (β_0ij ) and instantaneous rate of change parameters (β_1ij ; no random effect was proposed for the quadratic term). This level contained only random effects pertaining to rater that adjusted for variation specific to how adolescents and parents’ perceived the adolescents' internalizing symptoms.

(2a) $${\rm \beta}_{0ij} = {\rm \pi}_{00j} + u_{0ij} \quad \lsqb u_{0ij} \sim N \lpar 0\comma \; {\rm \tau}_0\rpar \rsqb \comma \;$$

(2b) $${\rm \beta}_{1ij} = {\rm \pi}_{10j} + u_{1ij} \quad \lsqb u_{1ij} \sim N \lpar 0\comma \; {\rm \tau}_1\rpar \rsqb \comma \;$$

(2c) $${\rm\beta}_{2ij} = {\rm \pi}_{20j}\comma \;$$

In the third level, the random intercept of Equation 2a was regressed on adolescent and family characteristics and random effects specific to variation between adolescents Equation 3a. In addition, the rater-level means of instantaneous rate of change (from Equation 2b) and acceleration (from Equation 2b) were then regressed on adolescent-level constants. The coefficients reported in the results are the γ coefficients, representing participant-level means of all effects.

(3a) $${\rm \pi}_{00j} = {\rm \gamma}_{000} + X {\rm \gamma}_{001} + e_{00j} \lsqb e_{00j} \sim N \lpar 0\comma \; \rm\phi _{0}\rpar \rsqb \comma \;$$

(3b) $${\rm \pi}_{10j} = {\rm \gamma}_{100} + e_{10j}\lsqb e_{10j} \sim N \lpar 0\comma \; \rm\phi _{1}\rpar \rsqb \comma \;$$

(3c) $${\rm \pi}_{20j} = {\rm \gamma}_{200} \comma \;$$

As shown in Equation 3b, a random term associated with the linear time parameter was also tested.

Missing data

There were missing values throughout the data, including the dependent and independent variables. The dependent variable nonresponse rate was 10% at Wave 1, 14% at Wave 2, 23% at Wave 3, and 27% at Wave 4. Independent variables had nonresponse rates that varied from less than 1% (for parent education, gender, age, and time living in the Unites States) to 55% (for Conflict Tactics Scale; see Table 1). Both parent marriage variables had higher rates of nonresponse (with dyadic adjustment at 50% missing) than the other scales. Given the potential for bias from nonresponse, analyses were conducted on data subjected to multiple imputation (Schafer, Reference Schafer1997). After demonstrating that the data were not missing completely at random according to a chi-square test developed by Little (Reference Little1988), we conducted an extensive analysis to develop an informed imputation model that would help reduce the bias from the missing values. These analyses demonstrated that 10 imputations were sufficient, and that a model consisting of all 15 analysis variables as well as 36 additional “auxiliary” variables that had the potential to explain the missing values could be supported by the data without a high level of collinearity. All analysis variables were included in the model to ensure that the covariance structure between the analysis variables was represented in the imputed data (without it, the relationships in the observed data would be attenuated and the imputed values would reflect this attenuation, biasing the parameter estimates). The auxiliary variables consisted of unused demographic variables and scales, including several measures that have been used and reported in other studies (Smokowski & Bacallao, Reference Smokowski and Bacallao2010). Ten data sets were simulated in the imputation process. These procedures were completed using SAS Proc MI, and results from running the models on each imputed data set were summarized using SAS Proc MIAnalyze.

Sensitivity, robustness, and diagnostics

The model described in Equations 1a–3c was tested exactly as specified. Variations on this model were also tested. These alternative models included (a) variations that excluded random terms u _1ij (Equation 2b) and e _10j (Equation 3b); (b) variations that excluded the quadratic term; (c) variations that used the arbitrary wave-based time variable (with values 0, 1, 2, and 3 for all participants); and (d) models without the two variables measuring parent marriage quality, which as noted had very high rates of missingness. Variations a–d were all tested using model fit. Fit was assessed using a chi-square test of the difference in likelihood ratios, the Akaike information criterion (AIC), and the Bayesian information criterion (BIC). For the AIC and the BIC, values closer to zero indicated better fit. The final model was selected from the full model and variations a–c. For variation d, we focused on how the coefficients of the model changed after the parent marriage quality variables were removed. Collinearity of the models was assessed using a variance inflation factor, with values approaching 10 indicating a high level of collinearity. Particular attention was paid to the relationship between age and time living in the United States, because these are the same for the adolescents born in the United Sttes. Influence and normality diagnostics were assessed on the final model as well.

Model details

SAS version 9.1 Proc Mixed was used. Restricted maximum likelihood was used to estimate the reported models (a maximum likelihood version of each model was run to facilitate the comparisons between models with different fixed effects as described by model variations b, c, and d; refer to Snijders & Bosker, Reference Snijders and Bosker1999).