Siblings’ Bidirectional Influence on Adolescents’ Educational Expectations and Postsecondary Enrollment

Siblings’ Bidirectional Influence on Mexican-origin Adolescents’ Educational Expectations and Postsecondary Enrollment

Educational attainment among Latinos has been changing rapidly in recent years, as a result of growth in both K-12 education and postsecondary education. The high school dropout rate for Latinos has reached a record low, dropping from 32% in 2000 to 12% in 2014 (Krogstad, 2016). Even though the Latino high school dropout rate is still higher than Blacks (7%), Whites (5%) and Asians (1%; Krogstad, 2016), more Latino high school graduates have enrolled in postsecondary than in the past. In fact, the percentage of Latinos who enroll in postsecondary education is similar to that of their White peers (i.e., 69% vs. 67%, respectively; Fry & Taylor, 2013).

Latinos are now the largest minority group among the nation’s college student body for both community and four-year colleges, such that Latinos make up 16.5% of the nation’s 18- to 24-year-old college population (Fry & Lopez, 2012). For the first time, Latinos’ representation among the nation’s college student body matched Latinos’ overall population representation, also at 16.5% (Fry & Lopez, 2012). Therefore, some have argued that enrollment in postsecondary education may be becoming a more normative developmental transition after high school for Latino young adults (Lefkowitz, 2005).

From a developmental perspective, the decision to immediately enroll in postsecondary education is influenced by the expectations, goals, and plans that are developed during adolescence (Eccles & Wigfield, 2002; Mello, 2009; Mello et al., 2012), as this is an important period for formulating realistic plans for the future (Markus & Wurf, 1987; Steinberg et al., 2009). Adolescents’ thoughts about the future have been conceptualized in diverse ways: as future orientation (Nurmi, 1994), goals (Massey, Gebhardt, & Garnefski, 2008), possible selves (Markus & Nurius, 1986), hopes and fears (Nurmi, 1993), or aspirations and expectations (Armstrong & Crombie, 2000). Most pertinent to this study is the concept of educational expectations. Educational expectations indicate one’s rational assessment of the educational outcomes that can be achieved given the available resources, socioeconomic background, and academic ability (St. Hilaire, 2002).

Previous literature has demonstrated that Latino adolescents’ educational expectations predict actual postsecondary education enrollment and educational attainment in young adulthood (Mello, 2009; Mello et al., 2012). A strong commitment toward a bachelor’s degree on the part of Latino students expressed as early as the 10th grade played a role in determining enrollment in postsecondary education (Arbona, & Nora, 2007). Latino high school sophomores who held high educational expectations for obtaining a bachelor’s degree were more likely to decide to immediately enroll in a four-year college after graduation than their peers with lower educational expectations (Arbona, & Nora, 2007). Furthermore, data from the National Education Longitudinal Study of 1988 were used to investigate the stability of adolescents’ postsecondary educational expectations from Grade 8 to 2 years after high school. All participants had high early expectations and six years later, approximately 76% of the participants still had high expectations, whereas only 24% of them no longer expected to earn a bachelor’s degree (Trusty, 2000). These findings suggest the importance of examining the influences of adolescents’ educational expectations as they remain stable across adolescence and early adulthood.

On average, Latino adolescents have low educational expectations regarding how much education they will actually complete, despite high educational aspirations that reflect their desire to attain more education (Lopez, 2009). For example, 89% of Latinos (aged 16 to 25) reported that a college degree was important for getting ahead in life; yet, only 48% of Latinos (aged 18 to 25) said they expected to earn a college degree (Lopez, 2009). Therefore, scholars studying Mexican-origin youth have emphasized the importance of identifying culturally relevant factors that promote greater educational expectations and postsecondary enrollment (Gonzales, Germán, Kim, George, Fabrett, Millsap, & Dumka, 2008; Huynh & Fuligni, 2008).

A considerable amount of theoretical and empirical work has demonstrated that sibling relationships function as important socialization contexts and sources of support (Eccles, Early, Fraser, Belansky, & McCarthy, 1997; Tucker, McHale, & Crouter, 2001; Ceja, 2006; Conger & Little, 2010). The vast majority of Mexican-origin adolescents (77%) growing up in the U.S. have at least one sibling (U.S. Census Bureau, 2011), and they spend considerably more time in shared activities with their siblings (> 20 hours per seven days) than with their parents or peers (Updegraff, McHale, Killoren, & Rodríguez, 2010); therefore, siblings may be an important factor that may promote greater educational expectations and postsecondary enrollment. However, relatively few studies have explored the specific role of siblings, relative to the influence of parents, teachers, and peers.

Observational learning and role modeling are key socialization processes through which adolescents acquire beliefs and behaviors (Bandura, 1977). During adolescence and young adulthood, siblings may be important role models and sources of information when adolescents are constructing their educational expectations and decisions to enroll in postsecondary education (Ali, Hawley McWhirter, Chronister, 2005; Ceja, 2006; Conger & Little, 2010). This may be particularly true among predominantly immigrant families as parents may be less familiar with the U.S. educational system (Ceja, 2006; Sanchez, Reyes, & Singh, 2006).

Most empirical research has examined top-down models of socialization such that they have focused on how older siblings serve as models for their younger brothers and sisters. For example, among Mexican-origin students, older siblings’ influence is associated with educational expectations and school attendance. Specifically, Hess and D’Amato (1996) found that Mexican-origin third- to fifth-grade students who had older siblings with a high school diploma or who were pursuing a high school diploma reported higher expectations of completing high school and had fewer absences than did students with siblings who had dropped out of school. However, little is known on how younger siblings influence their older brothers and sisters (Whiteman, Becerra, & Killoren, 2009). This study is one of the first to utilize a theory of relational interdependency such as the actor-partner interdependence model, which purports multilayered, reciprocal, and causal pathways that connect both older and young siblings’ influence on each other’s beliefs and behaviors (Kenny, Kashy, & Cook, 2006).

Research and theory on sibling influence highlight two opposing socialization processes. Sibling deidentification operates to make siblings’ different and social modeling, which operates to make siblings alike (Whiteman et al., 2007). These theoretical frameworks in conjunction with one another help explain why some siblings are more alike and others more different in their educational outcomes (Whiteman et al., 2007). Although these two theories are not alone in explaining why sibling similarities (i.e., shared genetics and shared environments, including shared parenting) and differences (i.e., nonshared genetics and nonshared environments, including parents’ differential treatment) may develop, they have been invoked as explanations across a wide range of studies (Whiteman et al., 2007).

Sibling Deidentification

Theorists have offered many explanations for sibling deidentification. Deidentification is believed to transpire to help protect siblings from social comparison, rivalry, envy, and possible resentment (Feinberg & Hetherington, 2000; Festinger, 1954; Schachter et al., 1976; Tesser, 1980; Sulloway, 1996). Siblings may be highly motivated to construct independent identities, unique from that of their siblings, by carving out their own niches through excelling in different domains (Schachter et al., 1976). From this perspective, if an older sibling excels in school, deidentification theory would predict that the younger sibling will consciously or unconsciously choose to specialize in a different domain, such as music or sports (Schachter et al., 1976). Psychodynamic theories suggest that identifying with or imitating siblings serves to exacerbate sibling rivalry as siblings seek the same goals, achievements, and gratifications. Feinberg and Hetherington (2000) reported that those adolescents who were most like their older siblings were most apt to experience deidentification.

Taken together, this small group of studies suggests that sibling deidentification dynamics may operate to make siblings different from one another. Unfortunately, these studies fail to measure these processes directly. Instead, deidentification dynamics are posited as post hoc explanations for patterns of differences between siblings (Whiteman et al., 2007). However, recent studies have demonstrated that sibling differentiation processes can be measured in a variety of ways (Whiteman & Christiansen, 2008; Whiteman, McHale, & Crouter, 2007). For example, a study measuring sibling deidentification should use a design that includes both siblings because much of the earlier work on sibling deidentification was limited to one respondent per family. Using this method, Whiteman and Christiansen (2008) found that differentiation dynamics were relevant for older as well as younger siblings, suggesting that siblings mutually view each other as having distinct characteristics and behaviors. This finding is notable because it challenges traditional deidentification assumptions that younger siblings differentiate from their older brothers and sisters, but not vice versa. Therefore, it is important that future research include the perspectives of both siblings, so that hypotheses about reciprocal influences, as well as the contributions of birth order and sex constellation, can be tested directly (Whiteman et al., 2007).

Sibling Modeling

Most social learning theories suggest that in addition to learning through their own behaviors and actions, individuals form ideas about and learn new behaviors through the observation of others. Bandura (1977) posited several necessary conditions for observational learning to occur. First, a model must possess salient qualities that attract attention in order for imitation to occur. One of the most important determinants of whether a model will attract another person’s attention is the frequency of contact with the model. Individuals who are regularly encountered provide more opportunities for observation, and hence their behaviors may be learned more thoroughly. Because adolescents spend so much time with their siblings (Updegraff, McHale, Whiteman, Thayer, & Delgado, 2005), they are potentially very salient models. Bandura (1977) also suggested that models will attract an individual’s attention if they possess attractive qualities, such as power, mastery, and nurturance. Older siblings may well possess power and mastery in more domains given that they are older and more advanced developmentally making them more likely models for younger siblings to observe and imitate (Bandura, 1977; Whiteman et al., 2011). For instance, older siblings may be able to provide essential information to their younger siblings regarding the steps they should follow to enter college (e.g., entrance exams, application process, financial aid; Buriel & De Ment, 1997; Ceja, 2006; Hurtado-Ortiz & Gauvain, 2007). Thus, older siblings have the potential to inspire their younger siblings (Ceja, 2006; Sanchez et al., 2006) and serve as positive role models through their educational expectations and attainments. These predictions are in line with a social learning perspective, which emphasizes the significance of individuals of higher status as role models (Whiteman et al., 2011), such as older siblings for their younger siblings. However, the ways in which younger siblings influence their older brothers and sisters are largely unknown and represent an important area for future research.

The Conjunction of Sibling Deidentification and Sibling Modeling Theories

The period from adolescence to young adulthood is hypothesized to be a time of developmental change in the structure of sibling relationships (Buhrmester & Furman, 1990; Conger & Little, 2010). In adolescence, sibling relationships are characterized as more hierarchical in that hypothesized influences are stronger from older to younger siblings than from younger to older siblings. Accordingly, extent data documents the role of older siblings as models for younger siblings’ behaviors and attitudes (e.g., East, 1998; Slomkowski et al., 2001; Whiteman et al., 2007). As individuals transition through adolescence and into adulthood, sibling relationships are expected to become more egalitarian (Buhrmester & Furman, 1990). These shifting dynamics from a hierarchical to a more egalitarian relationship structure may mean that sibling influences will become more reciprocal as the structure of the relationship becomes more balanced. Although the influence of younger siblings on older siblings is tested much less often than the influence of older siblings on younger siblings (East, 1998; Slomkowski et al., 2001; Whiteman et al., 2007b), this may be an important oversight as siblings transition from adolescence to adulthood when more reciprocal sibling influences may emerge.

Furthermore, the literature on sibling similarities and differences suggests that siblings are indeed influential in each other’s lives; however, the magnitude and direction of that influence is not always clear. In order to disentangle how siblings influence one another, mechanisms of similarity (i.e., social learning) need to be considered concurrently with mechanisms related to sibling differences (i.e., deidentification). Although these forces are posed as independent processes, whether they operate independently or in concert with one another is largely unknown.

The Moderating Role of Gender Constellation

Sibling relationships vary as function of their structural characteristics (e.g., sibling dyad gender constellation, sibling age spacing; Buhrmester & Furman, 1990; McHale, Bissell, & Kim, 2009; Whiteman, McHale, & Soli, 2011). Prior work on sibling relationships demonstrated that sibling characteristics merit attention as they have implications for siblings’ influences on one another (Buhrmester & Furman, 1990; McHale, Bissell, & Kim, 2009). Both classic and recent empirical evidence indicate that the sex constellation of sibling dyads is associated with sibling relationship quality during childhood and adolescence. Research has established that the warmest, most affectionate and supportive sibling dyads are typically those between sisters (Buhrmester & Furman, 1990; Furman & Buhrmester, 1992). The extant literature also suggests that siblings in same-sex (i.e., sister–sister, brother– brother) dyads tend to be closer than those in mixed-sex dyads (Furman & Buhrmester, 1985a, 1985b), although some longitudinal research suggests that siblings in mixed-sex dyads experience an initial decline in intimacy from childhood through early adolescence, followed by an increase (Kim, McHale, Crouter, & Osgood, 2007; Updegraff, McHale, & Crouter, 2002), whereas same-sex siblings evidence little change in intimacy over time. Therefore, sibling deidentification and modeling process are likely to differ due to gender constellation. For instance, McHale, and colleagues (2001) suggested that sibling dyads composed of older and younger siblings of different sex are most likely to demonstrate deidentification processes, both concurrently and over time. Whereas, previous research suggests that models that are similar to the self are more likely to be imitated. As such, same-sex siblings are thought to be more powerful models than opposite-sex siblings (Rowe & Gulley, 1992). Based on prior work (Buhrmester & Furman, 1990; McHale, et. al., 2009; Whiteman et al., 2011), stronger associations are expected between siblings’ educational expectations for same-sex as compared to mixed-sex dyads, given that individuals are more likely to identify themselves with others that are similar to them (Whiteman et al., 2011). Due to these findings, more work needs to be conducted examining deidentification and the effect of gender constellation.

The Moderating Role of Familism

A cultural-ecological framework informed our consideration of youths’ familism values as a moderator. From this perspective, cultural processes play a role in how one interprets and makes meaning of these family dynamics (Garcia Coll et al., 1996; Spencer, 2006). A small body of research on siblings in ethnic minority families reveals that the linkages between siblings’ influence and youths’ adjustment vary depending on youths’ cultural values (Solmeyer & McHale, 2015). Specifically, children from cultures in which families are highly regarded may be more influenced by their siblings than children from cultures in which family is not valued as much (Baham, 2009). Thus, sibling relationships may play a more important role in youth’s educational expectations and decisions to enroll in postsecondary education for children with strong family values (Baham, 2009).

One way that a child’s beliefs and values about family can be measured is through the construct of familism. Familism refers to a person’s family values and the importance one places on the family, and is defined as “a cultural value that involves individuals’ strong identification with and attachment to their nuclear and extended families, and strong feelings of loyalty, reciprocity, and solidarity among members of the same family” (Marin & Marin, 1991, p. 13). Familism is a critical aspect of the Latino culture (Gaines, Rios, & Buriel, 1997; Updegraff, McHale, et al., 2005) and thus should be included when examining interpersonal relationships within that culture. Research has illustrated that values placed in familism and family obligations can and do relate to academic performance; for example, during middle to late adolescence, support of the family and family respect was significantly positively related to better study habits and higher academic expectations in Asian-, Latino-, and European-American adolescents (Fuligni, Tseng, & Lam, 1999). This highlights that familism can be an important predictor of educational outcomes and should be examined in further detail.

The Present Study

Given that family relationships are salient and siblings may be particularly important due to their shared experiences and involvement in the U.S. educational system, the first goal was to examine the role of sibling influence in two aspects of Mexican-origin adolescents’ education: expectations and enrollment in post-secondary education (see Figure 1 and 2). The second goal was to test sibling gender constellation as a potential moderator of the associations between younger and older siblings’ influence and educational expectations and enrollment in postsecondary education (see Figure 3 and 4). The third goal was to test youths’ familism values as potential moderators of the associations between younger and older siblings’ influence and educational expectations and enrollment in postsecondary education (see Figure 5 and 6). Demographic background factors (i.e., parents’ education, parents’ nativity, and sibling age gap) were controlled for in all analyses.

Method

Participants

Data were drawn from a larger longitudinal study of adolescent development and family socialization including 246 Mexican American adolescents and their families (Updegraff et. al., 2005). Participants were recruited through schools in and around a southwest metropolitan area. Based on the larger study goals, criteria for participation were as follows: (1) 7th graders and an older sibling were living at home and not learning disabled, (2) biological mothers and biological or long-term adoptive fathers (i.e., 10 or more years) were living at home, (3) mothers were of Mexican-origin and (4) fathers worked at least 20 h per week. Although not required for participation, 93% of fathers also were of Mexican descent. We focused on two-parent families, who represent the predominant arrangement in Mexican American families in the U.S. (65 %; U.S. Census Bureau, 2014) and in the county from which the sample was drawn (U.S. Census Bureau, 2000).

To recruit participants, letters and brochures describing the study goals (in English and Spanish) were sent to 1,856 families with Latino 7th graders in five public schools districts and five parochial schools. Follow-up telephone calls were conducted by trained bilingual staff to determine each family’s eligibility and interest in participating in the project. The contact information of 396 families (21%) was incorrect and attempts to find updated information were unsuccessful and 146 families (10%) refused to be screened for eligibility. Eligible participants included 421 adolescents and their families (i.e., 32 % of those who were contacted and screened). Of those who were eligible, 284 families (67 %) agreed to participate, 95 (23 %) refused, and 42 families (10 %) moved before the recruitment process was completed. Interviews were completed with 246 adolescents and their families. Those who agreed but did not participate in the final sample (n = 38) were families that we were unable to locate or with whom we were unable to complete a home interview after repeated attempts.

At Time 1 (T1), mothers and fathers averaged 39 years (SD = 4.63) and 42 years of age (SD = 5.80), respectively. Most parents were born in Mexico (71% of mothers and 69% of fathers) and preferred to complete the interview in Spanish (66% of mothers, and 67% of fathers). Parents reported an average of 10 years of education (M = 10.34; SD = 3.74 for mothers, and M = 9.88, SD = 4.37 for fathers). The majority of foreign-born parents completed their education outside the U.S. (88% of mothers and 93% of fathers, respectively). Parents came from a range of socioeconomic levels, with the percentage of families meeting federal poverty guidelines (18.3%) being similar to two-parent Mexican American families in poverty in the county where the sample was drawn (i.e., 18.6%; U.S. Census Bureau, 2000). Median family income was $40,000 (range from $3,000 to over $250,000). Younger siblings were 12.51 (SD = 0.58) and older siblings were 15.48 (SD = 1.58) years of age. Over 51% of younger siblings (n = 125) and 50% of older siblings (n = 123) were female. Younger siblings were most likely to be born in the US (62%; n = 153), whereas older siblings were more likely to be born in Mexico (54%; n = 132). The majority of participants preferred to complete the interview in English (83%).

At Time 2 (T2), five years after the initial wave of data collection, over 75% of the families participated (n = 185). Younger siblings were 17.72 (SD = .57) and older siblings were 20.65 (SD = 1.57) years of age at T2.Those who did not participate could not be located (n = 43), had moved to Mexico (n = 2), could not presently participate or were difficult to contact (n = 8), or refused to participate (n = 8). When compared to the participant families (n = 185), non-participant families at T2 (n = 61) reported significantly lower income at Time 1 (M = $37,632; SD =$28,606 for non-participant families and M = $59,517; SD = $48,395 for participant families) and lower maternal education (M = 9.48; SD = 3.45 for non-participant families and M = 10.62; SD = 3.79 for participant families) and paternal education (M = 9.06; SD = 4.13 for non-participant families and M = 10.16; SD = 4.43).

At Time 3 (T3), seven years after the initial wave of data collection and two years after T2, over 70% of the families participated (n = 173). Younger siblings were 19.60 (SD = .66) and older siblings were 22.57 (SD = 1.57) years of age. Those who did not participate could not be located (n = 45), had moved to Mexico (n = 4), could not presently participate or were difficult to contact (n = 4), or refused to participate (n = 8). The 12 remaining non-participant families were classified as mixed-status as family members within these families did not participate for different reasons (e.g., in one family the father refused to participate and we were unable to locate the mother, younger sibling, and older sibling). When compared to the participant families (n = 173), non-participant families at T3 (n = 73) reported significantly lower income at T1 (M = $41,636; SD =$39,095 for non-participant families and M = $59,137; SD = $46,674 for participant families), lower maternal education (M = 9.35; SD = 3.53 for non-participant families and M = 10.75; SD = 3.75 for participant families), and lower paternal education (M = 8.49; SD = 4.08 for non-participant families and M = 10.46; SD = 4.37 for participant families).

Procedures

The same procedures were used at each wave of data collection. Trained bilingual interviewers collected data in separate home interviews in family members’ preferred language (either English or Spanish). At the beginning of the interview, interviewers obtained informed consent at T1 and at T2 (for T2 and T3). Due to variability in reading abilities, interviewers read questions aloud and entered responses into a laptop computer. Home interviews averaged between 2 to 3 hours in duration. Families were given a $100 honorarium for the interviews at T1, $125 at T2, and each family member was paid separately $75 at T3. The Institutional review board approved all procedures.

Measures

All measures were forward and back-translated into Spanish for local Mexican dialect (Foster & Martinez, 1995). All final translations were reviewed by a third native Mexican American translator and discrepancies were resolved by the research team. Cronbach’s alphas for all measures were acceptable for English- and Spanish-speaking participants; thus for efficiency, all alphas are reported for the overall sample rather than separately by language.

Sibling Intimacy (T1). The intimacy measure was originally developed by Blyth and colleagues (Blyth & Foster-Clark, 1987). The purpose of this measure is to assess and compare individuals’ perceptions of emotional closeness in different interpersonal relationships, including those with parents, siblings, friends, extended family members, and other adults. A sample item was “”How much do you go to (sibling’s name) for advice or support?” Both older and younger siblings reported on a 5-point Likert scale ranging from (1) not at all to (5) very much. Higher scores indicated more sibling intimacy. The average was 3.30 (SD = .78) for older siblings, in reference to intimacy with their younger sibling and the average was 3.36 (SD = .73) for younger siblings, in reference to intimacy with their older sibling. Cronbach’s alpha was .81 for younger siblings and .83 for older siblings.

Sibling Negativity (T1). Negativity was assessed with items from Furman and Buhrmester’s (1985) Network Relationship Inventory. This subscale includes a series of five questions about the extent to which the adolescent and his/her friend disagree, argue, and feel angry with each other. A sample item was “How much do you and (sibling’s name) get upset or mad at each other?” Both older and younger siblings reported on a 5-point Likert scale ranging from (1) not at all to (5) very much. Higher scores indicated more sibling negativity. The average was 3.15 (SD = .94) for older siblings, in reference to negativity with their younger sibling and the average was 3.09 (SD = .90) for younger siblings, in reference to negativity with their older sibling. Cronbach’s alpha was .89 for younger siblings and .92 for older siblings.

Sibling Relationship Quality (T1). Sibling relationship quality has been described as multidimensional demonstrating a ‘‘love–hate’’ relationship (Dunn, 1993). A difference score was created between the sibling intimacy measure and the sibling conflict measure, in order to be representative of the overall sibling relationship quality. Scores above zero indicated a more positive sibling relationship; whereas, scores below zero indicated a more negative sibling relationship. The average was 0.14 (SD = 1.42) for older siblings, in reference to overall sibling relationship quality with their younger sibling and the average was 0.27 (SD = 1.38) for younger siblings, in reference to overall sibling relationship quality with their older sibling.

Sibling Deidentification (T1). Both older and younger siblings reported how often they tried to be different from their sibling by completing an 8-item scale measuring sibling deidentification developed by Whiteman, McHale, and Crouter (2007). A sample item was “I have learned from watching (SIB’S NAME) what NOT to do”. Siblings responded using a Likert-type scale, ranging from (1) never to (5) very often. Items were averaged such that higher scores indicated greater deidentification. The average was 3.09 (SD = .92) for older siblings, in reference to deidentifying with their younger sibling and the average was 3.37 (SD = .83) for younger siblings, in reference to deidentifying with their older sibling. Cronbach’s alpha was .83 for younger siblings and .82 for older siblings.

Sibling Modeling (T1). Both older and younger siblings reported how often they tried to be like their sibling, the degree to which their sibling set a positive example for them, and the extent to which their sibling encouraged them to participate in particular activities by completing an 8-item scale measuring sibling modeling developed by Whiteman, McHale, and Crouter (2007). A sample item was “(Sibling’s name) sets an example for how I should behave.” Siblings responded using a Likert-type scale, ranging from (1) never to (5) very often. Items were averaged such that higher scores indicated greater modeling. The average was 2.33 (SD = .82) for older siblings, in reference to modeling after their younger sibling and the average was 2.86 (SD = .82) for younger siblings, in reference to modeling after their older sibling. Cronbach’s alpha was .87 for younger siblings and .84 for older siblings.

Siblings’ Gender Constellation (Moderator; T1). Older and younger siblings reported on their own gender (0 = females; 1 = males) and sibling dyad gender constellation was calculated based on participants’ responses (opposite-sex dyads= 0; same-sex dyads= 1). Approximately, 55% of siblings were same sex dyads (n = 134).

Familism (Moderator; T1). Siblings completed the 17-item Familism subscaleof the Mexican American Cultural Values Scale(Knight, Gonzales, Saenz, Bonds, German, Deardorff, & Updegraff, 2010). Siblings rated each item (e.g., “Parents should teach their children that the family always comes first?”) on a 5-point scale from strongly disagree to strongly agree. Cronbach’s alphas were .83 for younger siblings and .85 for older siblings.

Educational Expectations (T2). Participants’ reported on their educational expectations by responding to the following item: “How far do you really think you will go in school?” Response choices for both questions were on a continuous scale representing the total number of years of education (e.g., 12 = high school diploma, 21 = MD, JD, DO, DDS, OR Ph.D.). The average for older siblings was 15.59 (SD = 2.39) and the average for younger siblings was 15.40 (SD = 2.23), which is equivalent to expecting to complete about 3 years of post-secondary education.

Enrollment in Postsecondary Education (T3). Participants’ reported their highest grade completed on a continuous scale (e.g., 12 = high school diploma, 21 = MD, JD, DO, DDS, OR Ph.D.). Researchers later coded participants’ enrollment in post-secondary education. Participants’ who reported their highest grade completed was a high school diploma or lower were coded as 0 = not enrolled in post-secondary education Participants’ who reported their highest grade completed was greater than a high school diploma were coded as 1 = enrolled in post-secondary education. Approximately, 48% of older siblings (n = 119) and 53% of younger siblings (n = 131) enrolled in post-secondary education after high school.

Family and Demographic Characteristics (Covariates; T1). Mothers and fathers reported on their highest level of education ranging from less than a high school degree (e.g., a score of 10 signified 10th grade) to graduate or professional degree (a score of 21 signified PhD, JD, or MD). Parents also reported on their country of birth, which was coded as 1 = both parents US-born (N=62), 2 = one parent U.S. born (N=21), and 3 = both parents foreign-born (N=163). Mothers reported on each sibling’s age and age gap between siblings was calculated as the difference score between older and younger siblings’ age (M= 2.96, SD=1.63). All models include covariates for parents’ education, parents’ nativity and age gap between siblings.

Plan of Analysis

An Actor-Partner Interdependence Model (APIM) framework can be a great tool to capture the essence of sibling relationships (Kenny, Kashy, & Cook, 2006). Actor effects are intrapersonal effects. In APIM distinguishable regression models, there are two actor effects. Partner effects are interpersonal effects. In APIM distinguishable regression models, there are two partner effects. An APIM framework was used to estimate the associations between sibling influence and youths’ educational outcomes (see Figure 1 and 2).

There are multiple ways to analyze distinguishable dyads using APIM such as pooled regressions, structural equation modeling and multilevel modeling. A multilevel modeling approach was used because it is better able to estimate interaction effects (i.e., moderation) and handle missing data. An APIM was conducted to account for siblings’ interdependence using PROC MIXED in SAS 9.3 (Kenny, Kashy, & Cook, 2006). These models require a pairwise dataset; so, data was transformed from an individual dataset to a dyad dataset to a pairwise dataset. Furthermore, a variable is required to distinguish one sibling from another; a simple older versus younger sibling dummy variable was sufficient for these purposes. (Kenny, Kashy, & Cook, 2006). Missing data was imputed using maximum likelihood estimation (Enders, 2010). To test for moderation, familism values are multiplied by both the “actor” and “partner” terms of the APIM distinguishable regression model (see Figure 3, 4, 5 and 6).

Fit of models that are not saturated (i.e., models have remaining degrees of freedom) were assessed with the chi-square statistic, root mean square error of approximation (RMSEA ≤ .05), the comparative fit index (CFI ≥ .95), and standardized root mean square residual (SRMR ≤ .08). These particular fit indices are suggested as a good combination to assess the fit of models with small sample sizes (e.g., N < 250; Hu & Bentler, 1999).

Figure 1. The actor and partner effects of sibling modeling as predictors of educational expectations and enrollment in postsecondary education.

Figure 2. The actor and partner effects of sibling deidentification as predictors of educational expectations and enrollment in postsecondary education.

Figure 3. The actor, partner, and interaction effects of sibling modeling and gender constellation as predictors of educational expectations and enrollment in postsecondary education.

Figure 4. The actor, partner, and interaction effects of sibling modeling and gender constellation as predictors of educational expectations and enrollment in postsecondary education.

Figure 5. The actor, partner, and interaction effects of sibling modeling and familism as predictors of educational expectations and enrollment in postsecondary education.

Figure 6. The actor, partner, and interaction effects of sibling modeling and familism as predictors of educational expectations and enrollment in postsecondary education.

References