Journal of Youth and Adolescence

, Volume 41, Issue 10, pp 1267–1279

Mothers’ Differential Treatment of Adolescent Siblings: Predicting College Attendance of Sisters Versus Brothers

Authors

    • School of PsychologyWalden University
  • Eric Loken
    • Department of Human Development and Family StudiesThe Pennsylvania State University
  • Susan M. McHale
    • Department of Human Development and Family StudiesThe Pennsylvania State University
Empirical Research

DOI: 10.1007/s10964-011-9727-6

Cite this article as:
Bissell-Havran, J.M., Loken, E. & McHale, S.M. J Youth Adolescence (2012) 41: 1267. doi:10.1007/s10964-011-9727-6

Abstract

Current estimates suggest that by 2015, 60% of college students will be women, a change since 1970 when 59% were men. We investigated family dynamics that might explain the growing gender gap in college attendance, focusing on an ethnically diverse sample of 522 mixed sex sibling dyads from the National Longitudinal Study of Adolescent Health. We examined whether the difference between sisters’ and brothers’ reports of their mothers’ expectations for, and involvement in, their education during adolescence predicted their differential odds of college attendance seven years later. Sisters were more likely than brothers to attend college, and this gap was more pronounced among non-Whites and non-Asians. Sisters also had higher grades in school than their brothers. Although there were no gender differences overall in maternal educational expectations or involvement, brothers reported greater maternal involvement than sisters in non-White and non-Asian families. After controlling for family background factors, the average of siblings’ reports of maternal treatment, and differences between siblings’ grades, the results revealed that as sisters reported greater maternal educational expectations than their brothers, it became more likely that only the sister rather than only the brother in the family attended college. The difference between brothers’ and sisters’ reports of their mothers’ educational involvement and their odds of attending college showed the same pattern of association but was not statistically significant. These results suggest that within-family social comparisons may play a role in sisters’ and brothers’ choices about attending college.

Keywords

Academic achievementAdolescenceSiblingsDifferential treatmentGender

Introduction

Gendered patterns in college attendance in the US have changed markedly in recent decades. In 1970, about 59% of undergraduates were men (US Department of Education 2006a). During the 1980s, however, women began to surpass men in college attendance and, by 2007, 57% of students enrolled in post-secondary education were women (US Department of Education 2008). This trend is expected to continue, and by some estimates, women will comprise 60% of the undergraduate population by 2015 (US Department of Education 2006a). From a developmental perspective, experiences during adolescence have implications for choices and opportunities later in life, including those surrounding educational pursuits (Wigfield et al. 2006). As we elaborate below, prior research suggests that adolescents’ experiences with their parents, including their parents’ educational expectations and involvement, are associated with their individual educational attainment (Carter and Wojtkiewicz 2000; Flouri 2006). We draw on this work to test whether differences between brothers’ and sisters’ perceptions of their mothers’ expectations for and involvement in their education during adolescence predict within-family gender differences in siblings’ likelihood of having attended at least one year of college seven years later, in early adulthood.

Studying the mixed sex sibling dyads from the National Longitudinal Study of Adolescent Health (ADD Health), we first tested whether sisters and brothers differed in their likelihood of completing at least the first year of college (which we term college attendance). We then examined whether sisters and brothers differed in their perceptions of their mothers’ educational expectations of or involvement in their education in early and middle adolescence. Finally, we tested whether brother-sister differences in perceptions of mothers’ expectations and involvement (which we term differential maternal treatment) during adolescence explained brother-sister differences in college attendance in early adulthood. In addressing these research goals, we also explored whether there were racial/ethnic group differences in both maternal differential treatment and in sister–brother differences in college attendance.

The Gender Gap in Higher Education

Across all ethnic groups in the US, women now attend college at higher rates than men. Statistics from the 2004–2005 academic year show that the percentage of bachelor’s degrees granted to women was 55% among Asian/Pacific Islanders, 57% among Whites, 61% among Hispanics, and 66% among African Americans (US Department of Education 2006b). These patterns of educational achievement are likely to have important implications for women’s and men’s development across the lifespan. For example, young adults who complete college are less likely to face unemployment and are more likely to find full time work (Cook and Furstenberg 2002; Haggerty 1989). They also have higher incomes: In 1999, individuals who had bachelor’s degrees and worked full time earned an average of $52,000 per year, but those with high school degrees earned only $30,000 (US Census 2002). Success in the labor force, in turn, is linked to higher levels of psychological well-being (e.g. McKee-Ryan et al. 2005). Such findings underscore the significance of identifying experiences in adolescence that set the stage for decisions about education in young adulthood.

The gender gap in education also may have implications for family formation behaviors in young adulthood and beyond. Research on mate selection reveals that individuals tend to choose partners with an educational level that is either comparable to or higher than their own, and further, that the educational level of available mates is related to decisions about whether, and at what age, to marry (Lewis and Oppenheimer 2000). With fewer men than women earning college degrees, women may delay marriage while searching for a mate with a comparable level of education or forgo marriage altogether (Lewis and Oppenheimer 2000). Men also may fail to attract wives, which may have costs later in life; marriage is associated with better mental and physical health, particularly among men (Kiecolt-Glaser and Newton 2001). The gender gap in education also could have implications for divorce, given that spouses who differ in educational level experience more marital instability (Teachman 2002). Overall, previous research suggests that changes in educational attainment patterns by gender could be linked to changes in decisions about marriage and divorce.

More generally, the feminization of higher education may have larger implications for later cohorts of adolescents. Prior research shows that gender stereotypes and norms are linked to youth’s perceptions of their academic capabilities, that youth’s perceptions of their academic abilities are related to their educational expectations, and that youths’ future expectations predict actual educational outcomes (Bleeker and Jacobs 2004; Jacobs et al. 2006; Mello 2008). This line of research was originally inspired by concerns about the effect of stereotypes on girls’ and women’s achievement, particularly in the domains of math and science. An emerging new norm, that college is for women more so than for men, however, could affect the educational aspirations of future generations of boys and men. Germaine to this point, most recent research indicates that US girls today have higher educational expectations than US boys (Mello 2008).

Despite the potentially significant implications of the gendered educational pattern, there is limited research on factors related to the current gender gap in college attendance. The handful of recent studies on the gender gap highlights the role of gendered personal characteristics (Buchmann and DiPrete 2006; Goldin et al. 2006) and documents that, during childhood and adolescence, girls show higher levels of effortful control, are less likely to have behavioral problems, earn higher grades, and have higher rates of high school graduation than boys (Downey and Yuan 2005; Else-Quest et al. 2006; Jacobs 1996). Moving beyond individual characteristics, a body of work targets the family as a key context for gender socialization (McHale et al. 2003), and thus an important step in understanding the gender gap in education is to identify gendered family experiences in adolescence that may set the stage for young women’s and men’s choices about college attendance.

The Role of Parenting in Youth’s Academic Outcomes

Most research on parents’ role in their children’s educational achievement, including studies focused on gender differences, relies on comparisons of boys and girls from different families, an approach we refer to here as between-family comparisons. This work reveals that parents play an important role in their offpsring’s educational outcomes and career aspirations (Grolnick et al. 2000; Jacobs et al. 2006). It also suggests that parents interact differently with girls and boys, and that these differential interactions are related to gender differences in adolescents’ perceptions of their academic capabilities and, ultimately, to differences in young women’s versus young men’s educational and occupational choices and achievements (Jacobs et al. 2006). Two important elements of parenting in adolescence that have been linked to later educational attainment are educational involvement and expectations (Carter and Wojtkiewicz 2000; Flouri 2006).

Results of this work are not entirely consistent, however. With respect to parents’ educational involvement, some work has shown that parents of daughters are more involved with their offspring’s education than are parents of sons (e.g. Carter and Wojtkiewicz 2000). Findings from another study indicated, however, that mothers with sons remained more involved during middle school than did those with daughters (Grolnick et al. 2000). In yet other studies, parents of daughters and sons exhibited similar degrees of educational involvement (e.g. Rogers et al. 2009). A more detailed analysis suggests that the nature of parental involvement differs for girls and boys, with parents of adolescent girls attending more school events and talking more with daughters about school, and parents of boys communicating with school personnel and checking their sons’ homework more often (Bhanot and Jovanovic 2005; Carter and Wojtkiewicz 2000). These patterns may reflect that adolescent girls communicate more with their mothers than do boys (Leaper et al. 1998), and that boys exhibit more behavioral problems and have more difficulties staying on task than girls (Downey and Yuan 2005). In other words, parents may become involved in different ways and for different reasons in girls’ versus boys’ education.

Research on parents’ educational expectations also reveals an inconsistent pattern of findings, possibly because the data on this topic have been collected at different points in time. For example, consistent with national patterns showing that men completed more years of education at the time (US Department of Education 2006a), research in the early 1980s indicated that parents expected sons to complete more years of education than daughters (Eccles and Hoffman 1984). In contrast, data from the 1988 National Education Longitudinal Study revealed that parents expected higher educational attainment in daughters than sons (Carter and Wojtkiewicz 2000). Other data collected in 1988, however, indicated that parents had more money saved for boys’ than for girls’ education (Freese and Powell 1999). The limited body of more contemporary research reveals that adolescent girls describe their parents as placing more importance on educational goals than do boys (Spera 2006). There may not be gender differences, however, in parents’ expectations for their offspring’s actual level of educational attainment (Spera 2006). In short, the available evidence derived from between-family comparisons of contemporary US families does not clearly indicate whether parents favor their adolescent boys or girls with higher educational expectations.

The Implications of Within-Family Dynamics for Youth’s Academic Achievement

In our research, we use a within-family design to study sisters versus brothers—girls and boys from the same families. This design allows for direct comparisons of parents’ treatment of sons versus daughters and allows each family to serve as its own control. This is an important advantage in drawing inferences about the family’s role in adolescent development, because families may differ in unmeasured characteristics (third variables) that affect estimates of constructs of interest such as parents’ values regarding education or parental style (Teachman 1997).

Within-family designs also direct attention beyond the parent–child dyadic relationship to more systemic family processes, including the social comparisons that surround parents’ differential treatment of offspring. In his theory of individual psychology, Alfred Adler proposed that youth’s treatment relative to their siblings was a key dynamic in personality development and adjustment. That is, the social comparisons youth make based on the way they versus their siblings are treated have implications beyond the typically studied measures of the parent–child dyad (Ansbacher and Ansbacher 1956). Consistent with Adler’s ideas, a body of evidence indicates that parents’ differential treatment of their offspring is linked to youth’s outcomes beyond the absolute level of parenting that each sibling receives. For example, in studies based on differences between siblings’ perceptions of their experiences with parents, higher levels of parents’ differential treatment have been linked to larger sibling differences in adjustment, and less favorable treatment in comparison to a sibling has been linked to more internalizing and externalizing symptoms in adolescence, even after controlling for the quality of the dyadic parent–child relationship (Feinberg and Hetherington 2001; Shanahan et al. 2008). Although this work establishes parents’ differential treatment as an important family dynamic, there is almost no research on its role in educational achievement. Instead, as we have suggested, research on parents’ role in offspring educational achievement has focused on dyadic relationship properties and has ignored such larger system dynamics. The limited research that has examined within-family differences in siblings’ achievement has targeted family demographic and structural characteristics such as siblings’ birth order (e.g. Conley et al. 2007), but we know relatively little about the role of parenting processes. Adler’s theory and prior research on parents’ differential treatment, however, suggest that experiencing less maternal educational involvement and lower maternal educational expectations relative to a sibling will have unique implications for youth’s achievement-related cognitions and behaviors, and ultimately their educational achievement outcomes—such as whether they attend college.

The Role of Race/Ethnicity in the Education Gender Gap

The gender gap in attending college is greater among certain racial-ethnic minorities (US Department of Education 2006b), but we know relatively little about the family processes that underlie this pattern. One possibility is that parents’ gendered differential treatment varies by race/ethnicity in ways that have implications for gendered patterns of educational achievement. Because the role of parents’ differential treatment has been largely unexplored among non-European families, it is difficult to make specific predictions. Prior research, however, suggests that parents’ differential treatment may be more pronounced in families facing stressful circumstances (McHale, et al. 2006). Given that some ethnic minority groups in the US face unique challenges, including discrimination and economic disadvantage, it may be that parents treat their daughters and sons more differently in these families. If this is the case, parents’ differential treatment may help to explain the larger gender gap among certain ethnic minorities.

Relevant findings from between-family research are inconsistent, but may shed some light on this issue. For example, between-family comparisons have shown that, among Latinos and African Americans, girls are perceived as having greater potential for academic success and perceive greater family support for academic achievement than boys (Lopez 1995; Sanders and Herting 2000; Wood et al. 2010). Other findings, however, suggest that African American and Latino adolescent boys perceive greater maternal involvement in their education (Roche et al. 2007) and that Latino boys are encouraged more than Latino girls in their post-secondary educational goals (Cauce and Domenech-Rodriguez 2002; Lopez 1995). Again, however, this research is limited to between-family comparisons, and we know little about differential treatment of sisters versus brothers in these ethnic groups.

The Current Study

We built on existing literature to examine the associations between family processes and the odds of sisters versus brothers attending college in an ethnically diverse sample of mixed sex sibling dyads. We focused on mothers to avoid substantially reducing our sample size, particularly among racial-ethnic groups with high rates of single parenthood; about half of our sibling subsample were from family structures other than two parents with biologically-related offspring, and father involvement tends to be limited in such families (e.g. Seltzer 1991).

Our study had three aims. First, given US statistics on gender differences in college attendance, we hypothesized that sisters would be more likely than their brothers to complete the first year of college during young adulthood and that this gender gap would be less pronounced in White and Asian families than in other racial-ethnic groups. Second, in an effort to identify family precursors of these gendered patterns, we tested whether sisters and brothers differed in their reports of maternal educational expectations and involvement during early and middle adolescence. Between-family literature on this topic has not yielded consistent findings and within-family research has been limited. However, given the cohort of youth that we studied –one that was attending middle and high school in 1995–1996, a time when young women were beginning to outnumber young men in college, we predicted that mothers’ educational expectations would favor adolescent girls over their brothers. Further, given that mothers may talk more with adolescent girls than with boys (Leaper et al. 1998), and that our measure of educational involvement focused on conversations about school, we predicted that mothers’ involvement also would favor girls over their brothers. We also hypothesized that these within-family gender differences would be more pronounced in disadvantaged minority families than in White and Asian families, given prior findings that parents in stressful circumstances exhibit more differential treatment.

Finally, extrapolating from prior research on youth adjustment, which has shown that parents’ differential treatment is linked to differential sibling outcomes, we hypothesized that differences between sisters’ and brothers’ reports of maternal involvement in education and/or maternal educational expectations would be associated with the odds of only the sister versus only the brother in the family attending college. In examining this final research question, we controlled for mothers’ average educational expectations and involvement and sibling differences in school grades as well as demographic characteristics. Controlling for the average of siblings’ reports of maternal treatment helps to establish that it is the difference between siblings’ perceptions of maternal treatment that predicts academic outcomes, rather than simply the general level of parenting (Feinberg and Hetherington 2001). Controlling for sibling differences in grades helps to establish that associations between siblings’ differential perceptions of maternal treatment and college attendance were not due simply to differences between siblings’ academic performance.

Method

Participants

The data came from the Add Health pairs sample (Bearman et al. 1997; Slomkowski et al. 2005). Add Health is a longitudinal study of youth in the US. At wave 1 in 1994–1995, participants were in grades 7 through 12; wave 2 was conducted in 1995–1996 and wave 3 data were collected in 2001–2002.

Sample Selection Criteria

The pairs sample included 3,139 sibling dyads at wave 1, and 522 mixed sex dyads from this sample were the focus of the current study. We first removed 245 dyads because: (a) youth did not have data at any of the three study waves, (b) siblings’ ages could not be determined due to missing birth dates, (c) data for some youth were duplicated. From this sample (n = 2,894 dyads), we selected the oldest mixed sex sibling pair from each family who had participated in both waves 1 and 3, choosing the two siblings whose ages most closely approximated consecutive birth orders (n = 719 dyads). (Birth order was randomly assigned for dizygotic, or DZ, twins. Of the 719 families represented, 12% contained more than two study siblings.) Selecting one pair per family allowed us to avoid statistical dependence issues, selecting the oldest pair allowed us to include youth with the greatest opportunity of attending college by wave 3, and using consecutive birth order provided for the greatest similarity between siblings’ family circumstances. We then selected dyads in which both siblings (a) were at least 20 years old at wave 3 (to ensure both members were old enough to have the opportunity to attend college), (b) attended school at wave 1 (given that school grades were important for our analyses), and (c) reported having a mother figure, given our focus on siblings’ perceptions of maternal treatment.

Sample Description

The sisters (n = 522) were an average of 15.8 years old (SD = 1.60) at wave 1 and 22.1 (SD = 1.62) at wave 3. The brothers (n = 522) were an average of 16.0 (SD = 1.67) years of age at wave 1 and 22.4 (SD = 1.67) at wave 3. The dyads had an average age-spacing of 1.6 years (SD = 1.2), and brothers were the older sibling in 53% of the dyads. The median 1994 family income was $37,750, 24% of the primary parents (see “Measures”) reported being a college graduate or having some professional training beyond college, and 49% of the sibling dyads were in two-parent biological families at wave 1. Approximately 50% of the sibling dyads were White, 24% were African American, and 12% were Hispanic. The remaining 14% of dyads were Asian/Pacific Islander, American Indian/Alaskan Native, another less common race/ethnicity, or the siblings in the dyad were of different racial-ethnic groups. The sample contained 108 dizygotic twin dyads, 259 full sibling dyads, 75 half sibling dyads, 23 cousin dyads, and 57 non-related sibling dyads (e.g. step-siblings).

Measures

Parental Education

In the ADD Health study, one parent in the household, referred to as the “primary parent”, was chosen to be interviewed. In our sub- sample, the primary parent was the biological mother in about 86% of cases. Primary parents reported their educational level at wave 1. The original ADD Health response scale was altered for this study so that the responses were ordered on a 0 (lowest) to 7 (highest level of education) scale (0 = “never went to school”, 1 = “eighth grade or less”, 2 = “more than eighth grade but not high school graduate”, 3 = “went to business, trade, or vocational school instead of high school”, 4 = “high school diploma”, 5 = “some post high school education”, 6 = “college graduate”, 7 = “some professional training beyond college”).

Family Income

At wave 1, primary parents reported their total family income in thousands of dollars for the year 1994. Because this variable was positively skewed, we used the natural log.

Family Structure

At wave 1, adolescents reported who currently lived in their household and how each person was related to them. We grouped families into two categories: two-parent biological (both siblings lived with two parents who were biologically related to them) versus other.

Sibling Age Difference

This was calculated by subtracting brothers’ from sisters’ age (in years).

Dyad Race/Ethnicity

Two groups were formed based on adolescents’ wave 1 reports of their race and ethnicity: (1) “White/Asian” (both siblings were White (non-Hispanic), both siblings were Asian/Pacific Islander, or one sibling was White and one sibling was Asian/Pacific Islander), and (2) “Other” (all other dyads). We decided on two racial-ethnic groups to minimize small cell sizes, and these two groups were chosen based on the percentages of youth in our subsample, by race/ethnicity, who attended at least one year of college (Whites: 57%, Asian/Pacific Islanders: 66%, African Americans: 48%, and Hispanics: 41%).

Maternal Involvement in Education

At wave 1, adolescents reported on whether they had been involved in two activities with their mother over the past four weeks: talked about school work or grades and talked about other things they were doing in school. Ratings were summed, and scores ranged from zero to two, with “two” indicating that adolescents reported being involved in both activities with their mother (α = .67).

Maternal Educational Expectations

At wave 1, adolescents responded to the following item: “On a scale from 1 to 5, where 1 is low and 5 is high, how disappointed would [your mother] be if you did not graduate from college?”

Adolescents’ School Grades

Adolescents self-reported their grades (4 = “A”; 3 = “B”; 2 = “C”; 1 = “D or lower”) for the most recent quarter for each of four academic subjects: history/social studies, science, math and English/language arts. Grades were averaged to create an overall GPA if youth reported grades for at least two subjects.

College Attendance

At wave 3, youth were asked: “What is the highest grade or year of regular school that you have completed?” Answer options ranged from 6 (“sixth grade”) to 22 (“5 or more years of graduate school”). We operationalized college attendance as completing versus not completing one year of college. Siblings whose ratings were 13 (one year of college) or higher were assigned a score of 1 to indicate that they had attended at least one year of college; otherwise, they were assigned a score of 0.

Sibling Differences

Maternal differential treatment scores (involvement in education and educational expectations) and an index of sibling differences in school grades were formed by subtracting brothers’ from sisters’ reports.

Missing Data

The variable with the highest proportion of missing data was family income (90 cases missing; 17% of the sample), followed by primary parental education (53 cases; 10%), and the difference between siblings’ grades (19 cases; 4%). Because these missing data created a loss of up to 116 complete cases in the multinomial logistic regression models (see “Results”), we used PROC MI in SAS to impute five data sets that were used for these analyses in this study. A Markov chain Monte Carlo (MCMC) method was used with multiple chains (see Schafer 1997 for details) and full-data imputation. PROC MIANALYZE in SAS was used to combine the five sets of results.

Results

The results are organized in terms of our research goals. We first examined differences between sisters’ and brothers’ reports of college attendance by wave 3 (age range: 20 to 27), both overall and by race/ethnicity. Next, we examined differences between sisters’ and brothers’ reports of their mothers’ educational expectations and involvement and their school grades. Finally, we tested our hypothesis that differences between sisters’ and brothers’ reports of maternal treatment during adolescence would predict the odds of sisters only versus brothers only attending college, after controlling for mothers’ average educational expectations and involvement, as well as sibling differences in school grades.

College Attendance by Sisters Versus Brothers

Findings from McNemar’s test revealed that sisters and brothers differed in their frequency of attending college by wave 3, Z2 = 12.27, df = 1, p < .01, with sisters showing an 8.62% higher probability than their brothers of completing at least one year of college (see Table 1). McNemar’s test allows researchers to examine, in a 2 × 2 contingency table, whether members of matched pairs differ in their frequency of displaying a trait, or outcome. When only one study sibling in the family attended college, it was 1.8 times more likely to be the sister, rather than the brother. When we examined the gender gap as a function of race/ethnicity, we found no statistically significant differences between brothers’ and sisters’ frequency of attending college in “White/Asian” dyads; the single largest group was both brothers and sisters attend college. In “other ethnicity” dyads, however, the probability of sisters attending college was 13.55% higher than the probability of brothers attending, Z2 = 12.55, df = 1, p < .01. Here, the single largest group was neither brother nor sister attends college (see Table 1).
Table 1

Frequencies (and proportions) of dyads with sisters and brothers who did and did not attend one year of college

 

Sister does not attend

Sister attends

Total

All dyads

 Brother does not attend

162 (31%)

105 (20%)

267 (51%)

 Brother attends

60 (12%)

195 (37%)

255 (49%)

 Total

222 (43%)

300 (57%)

522

White/Asian

 Brother does not attend

77 (25%)

57 (19%)

134 (44%)

 Brother attends

41 (13%)

133 (43%)

174 (56%)

 Total

118 (38%)

190 (62%)

308

Other race/ethnicity

 Brother does not attend

85 (40%)

48 (22%)

133 (62%)

 Brother attends

19 (9%)

62 (29%)

81 (38%)

 Total

104 (49%)

110 (51%)

214

We next estimated a multinomial logistic regression model to examine whether the difference between sisters’ and brothers’ odds of attending college varied by race/ethnicity after controlling for family income, education, and structure, to account for the fact that these variables are typically confounded with race/ethnicity in the US. We also included a control for the age difference between brothers and sisters (see Table 2). Multinomial logistic regression allows researchers to use multiple independent variables to predict an outcome with more than two categories. In the model, the odds of being in each of three family outcome categories (neither sibling attended college, only the sister attended, both siblings attended) were compared to the odds of being in the baseline category (only the brother attended). Of the three sets of odds ratios in the model, the set pertaining to the odds of sisters only versus brothers only attending college was of primary interest. After controlling for other family background characteristics, the odds were 2.10 times higher in “other” families, than they were among Whites/Asians, that only the sister, rather than only the brother, completed at least one year of college. The results also showed, however, that the odds were significantly higher that both siblings rather than only the brother completed one year of college in families with higher parental education, in non-White/non-Asian sibling dyads, and in two parent biological families.
Table 2

Multinomial logistic regression model predicting siblings’ odds of college attendance with family background characteristics

 

Comparison: neither sibling

Comparison: sister only

Comparison: both siblings

B

SE

OR

B

SE

OR

B

SE

OR

Intercept

2.48

.90

11.94**

.15

.91

1.16

−2.64

1.00

.07**

Age differencea

.01

.08

1.01

.13

.08

1.14

.15

.08

1.16

Family income (logged)

−.39

.23

0.68

−.26

.24

.77

.21

.26

1.23

Parental education

−.10

.12

0.90

.19

.13

1.21

.49

.13

1.63**

Race/ethnicityb

.51

.36

1.67

.74

.38

2.10*

.73

.37

2.08*

Family structurec

−.03

.33

.97

.59

.35

1.80

1.08

.33

2.94**

Baseline = only the brother attends college

p < .05; ** p < .01; G2 = 1190.04; df = 1509

aSister–brother difference

bRace/ethnicity: 0 = Asian or White family, 1 = dyad race/ethnicity other than Asian or White

cFamily structure: 0 = non two-parent biological family, 1 = two-parent biological family

Sister–Brother Differences in Adolescence

Results from paired t-tests indicated that there were no mean differences between sisters’, M = 1.20, SD = .85, and brothers’ reports, M = 1.20, SD = .84, of maternal involvement in education or maternal educational expectations (M = 3.94, SD = 1.30; M = 3.86, SD = 1.32, for sisters and for brothers, respectively). A paired t-test did show, however, that sisters, M = 2.84, SD = .77, reported higher grades than brothers, M = 2.59, SD = .75, t = 6.18, p < .01. Overall, the average difference score for maternal expectations was .06, SD = 1.64, and the average difference score for maternal involvement was 0, SD = 1.09. Differential expectations and involvement were positively correlated, r = .10, p < .05, but not highly so, indicating that they reflected distinct socialization processes.

We next used a 4 college attendance group (neither sibling, only sisters, only brothers, both siblings attend college) × 2 sibling (sisters, brothers) mixed model ANOVA, with sibling as a within-group factor, to examine whether differences between sisters’ and brothers’ reports of maternal treatment or grades predicted the dyadic pattern of college attendance. The results for maternal educational expectations (Table 3), revealed a main effect for college attendance group, F (3, 511) = 8.82, p < .05, and a significant interaction between sibling and college attendance group, F (3, 511) = 3.70, p < .05. Follow-up paired t-tests showed that in families in which only brothers attended, brothers had reported higher maternal expectations during adolescence, M = 4.20, SD = .99, than their sisters, M = 3.77, SD = 1.49, t = 2.17, p < .05, and in families in which only sisters attended, sisters had reported higher expectations, M = 4.09, SD = 1.17, than their brothers, M = 3.65, SD = 1.36, t = 2.87, p < .05. There were no mean differences in perceptions of maternal treatment between sisters and brothers in families in which siblings had the same college attendance outcomes. For maternal involvement in education, neither the main effects nor the interaction were significant (Table 3). Finally, for grades, the main effects for both sibling, F (1, 499) = 21.74, p < .05, and college attendance group, F (3, 499) = 41.39, p < .05, were significant, but were qualified by a significant sibling × attendance group interaction, F (3, 499) = 9.16, p < .05. Follow-up paired t-tests showed that sisters reported higher grades than brothers in three of the attendance groups; there was no difference between siblings’ grades in the “brother only” group.
Table 3

Means (SD) for Sisters’ and Brothers’ Reports of Maternal Treatment and Grades as a Function of College Attendance and Race/Ethnicity

 

Maternal involvement in education

Maternal educational expectations

Grades

Sisters

Brothers

Sisters

Brothers

Sisters

Brothers

College attendance

 Sister only

1.28 (.87)

1.23 (.87)

4.10 (1.17)c

3.65 (1.36)d

3.04 (.70)e

2.50 (.66)f

 Brother only

1.05 (.85)a

1.37 (.84)b

3.77 (1.49)c

4.20 (.99)d

2.44 (.75)

2.66 (.71)

 Both siblings

1.30 (.81)

1.25 (.81)

4.17 (1.04)

4.12 (1.16)

3.15 (.66)e

2.89 (.73)f

 Neither sibling

1.09 (.87)

1.08 (.86)

3.61 (1.51)

3.56 (1.50)

2.47 (.74)e

2.26 (.71)f

Dyad race/ethnicity

 White/Asian

1.23 (.83)

1.13 (.87)

3.86 (1.32)

3.85 (1.29)

2.95 (.77)e

2.76 (.75)f

 Other

1.16 (.87)a

1.31 (.79)b

4.05 (1.26)

3.88 (1.36)

2.67 (.76)e

2.35 (.69)f

a, bSister–brother paired means for maternal involvement with different superscripts are significantly different (p < .05)

c, dSister–brother paired means for maternal expectations with different superscripts are significantly different (p < .05)

e, fSister–brother paired means for grades with different superscripts are significantly different (p < .05)

We next examined whether differences between sisters’ and brothers’ reports of maternal treatment or grades varied by dyad race/ethnicity. Results for maternal educational expectations from a 2 race/ethnicity group (White/Asian, Other) × 2 sibling (sisters, brothers) mixed model ANOVA, with sibling as a within-group factor, revealed that there were no significant main or interaction effects. For grades, the main effects for race/ethnicity, F (1, 501) = 40.56, p < .01, and sibling, F (1, 501) = 40.43, p < .01, were significant (Table 3), with grades being higher for White/Asian than for “Other” dyads and grades being higher for sisters than for brothers. For maternal involvement, there were no significant main effects, but there was a significant interaction between sibling and race/ethnicity, F (1, 520) = 7.04, p < .01. Follow-up paired t-tests showed that in White/Asian dyads, there were no mean differences between sisters’ and brothers’ reports, but that in “other ethnicity” families, brothers had reported higher maternal involvement, M = 1.31, SD = .79, than sisters, M = 1.16, SD = .87, t = 2.07, p < .05 earlier in adolescence.

Sister–Brother Differences in Perceptions of Maternal Treatment as a Predictor of Sisters’ Versus Brothers’ Odds of Attending College

As a first step toward illuminating predictors of the odds of only the sisters versus only the brothers attending college, the following measures were entered into separate multinomial logistic regression models: the average of sisters’ and brothers’ grades, the difference between sisters’ and brothers’ grades, the average of sisters’ and brothers’ reports of mothers’ educational expectations, the difference between their reports of mothers’ educational expectations, the average of sisters’ and brothers’ reports of mothers’ involvement in education, and the difference between their reports of mothers’ involvement in education (see Table 4). “Only the brother attends college” was again used as the baseline outcome, with the same three comparison outcomes discussed above. These analyses revealed that the odds increased that only sisters, rather than only brothers, completed at least one year of college as a function of increases in: differential perceptions of mothers’ expectations (favoring sisters), differential perceptions of mothers’ involvement (favoring sisters), and both the sibling average and sibling differences (favoring sisters) in school grades. For example, for each unit increase in differential expectations that favored sisters, the odds were 1.39 times higher that only the sister, rather than only the brother, completed at least one year of college.
Table 4

Preliminary multinomial logistic regression models predicting siblings’ odds of college attendance

Modela

 

Comparison: neither sibling

Comparison: sister only

Comparison: both siblings

B

SE

OR

B

SE

OR

B

SE

OR

1

Average gradesb

−.63

.29

.53*

.72

.30

2.05*

1.49

.29

4.44**

2

Difference in gradesc

.55

.17

1.73**

.97

.19

2.65**

.56

.17

1.75**

3

Average maternal expectationsb

−.35

.15

.70*

−.11

.17

.90

.20

.16

1.22

4

Mothers’ differential expectationsc

.17

.09

1.19

.33

.10

1.39**

.19

.09

1.21*

5

Average maternal involvementb

−.29

.23

.75

.11

.25

1.11

.16

.23

1.18

6

Mothers’ differential involvementc

.28

.14

1.32*

.31

.15

1.37*

.32

.14

1.38*

Baseline = only the brother attends college

* p < .05; ** p < .01

aEach predictor was entered into a separate model

bAverage of sisters’ and brothers’ reports

cSister report–brother report

We next estimated a multinomial logistic regression model to examine whether sister–brother differences in perceptions of maternal treatment were associated with the odds of only sisters versus only brothers attending college, this time including the following controls: family demographic variables, the difference between sisters’ and brothers’ grades, the average of sisters’ and brothers’ grades, the average of sisters’ and brothers’ reports of mothers’ involvement in education, and the average of their reports of mothers’ educational expectations (see Table 5). As a preliminary step, we ran this model separately for each of four biological relatedness groups (dizygotic twins, full siblings, half siblings, and non-related siblings), but because the differential perceptions of maternal treatment coefficients for the odds of only the sister versus only the brother attending college did not vary across these groups, we report results for the sample as a whole. The results revealed that, as the average of sisters’ and brothers’ grades increased, OR = 2.46, p < .05, and as the difference between their grades (favoring sisters) increased, OR = 2.51, p < .05, it became more likely that only the sisters, rather than only the brothers, attended college. Further, as differential perceptions of mothers’ educational expectations (favoring sisters) increased, only sisters attending college became more likely relative to only brothers attending, OR = 1.40, p < .05. The effect for differential perceptions of mothers’ involvement, although in the predicted direction, was not significant, OR = 1.27, p = .15.
Table 5

Multinomial logistic regression model predicting siblings’ odds of college attendance with differential perceptions of maternal treatment

 

Comparison: neither sibling

Comparison: sister only

Comparison: both siblings

B

SE

OR

B

SE

OR

B

SE

OR

Intercept

5.09

1.28

162.39**

−1.71

1.39

.18

−5.93

1.43

.00**

Age differencea

.06

.08

1.06

.23

.09

1.26*

.22

.08

1.25**

Family income (logged)

−.36

.24

.70

−.25

.26

.78

.17

.28

1.19

Parental education

−.02

.13

.98

.19

.14

1.21

.44

.14

1.55**

Race/ethnicityb

.73

.38

2.08

.98

.42

2.66*

1.11

.41

3.03**

Family structurec

.03

.35

1.03

.62

.38

1.86

1.04

.35

2.83**

Average gradesd

−.44

.32

.64

.90

.34

2.46**

1.32

.32

3.74**

Difference in gradesa

.54

.19

1.72**

.92

.21

2.51**

.58

.19

1.79**

Mothers’ educational expectationsd

−.41

.16

.66*

−.21

.18

.81

−.03

.18

.97

Mothers’ differential expectationsa

.14

.10

1.15

.34

.11

1.40**

.25

.11

1.28*

Mothers’ educational involvementd

−.37

.26

.69

.04

.28

1.04

.07

.26

1.07

Mothers’ differential involvementa

.26

.15

1.30

.24

.17

1.27

.25

.16

1.28

Baseline = only the brother attends college

* p < .05; ** p < .01; G2 = 1087.97; df = 1530

aSister report—brother report

bRace/ethnicity: 0 = Asian or White family, 1 = dyad race/ethnicity other than Asian or White

cFamily structure: 0 = non two-parent biological family, 1 = two-parent biological family

dAverage of sisters’ and brothers’ reports

Discussion

The gender gap in undergraduate education has reversed in recent decades, and young women now outnumber men, a trend that may have implications for subsequent cohorts of adolescents and lifespan implications for both women and men. Although recent research has highlighted how gendered personal characteristics may explain this growing gender gap, we know little about how within-family experiences during adolescence may set the stage for the choices young women and men make about college attendance. Furthermore, much of what we know about how parenting factors are related to offspring’s educational outcomes is based on between-family designs that examine boys and girls from different families and do not assess important sibling and family system dynamics, including the implications of comparisons siblings make between the ways they versus their siblings are treated by parents. Our analyses extend previous work by documenting that the gender gap prevails within families and by showing that differences between sisters’ and brothers’ perceptions of what their mother expects of them with regard to their educational attainment predict their differential success of completing at least a year of college seven years later. Importantly, mothers’ differential expectations emerged as a significant predictor beyond the mean level of mothers’ expectations, consistent with the idea that the social comparisons youth make between themselves and their siblings are a key dynamic in family socialization (Ansbacher and Ansbacher 1956).

Sister–Brother Differences in College Attendance

Our first research goal was to examine differences between sisters’ and brothers’ college attendance. Despite being raised in the same family, and, in most cases, being genetically related, siblings in mixed sex dyads had markedly different odds of attending college. Our analyses revealed that about 49% of boys attended college as compared to 57% of girls. Furthermore, in about a third of the sample, sisters and brothers differed in whether or not they attended college: in 20% of families, sisters but not brothers attended college, and in another 12%, brothers but not sisters attended college. Although our sample is not nationally representative, these gender differences are consistent with national data showing that, in recent years, US women outnumber men in higher education.

In addressing our first research goal, we also expanded on prior research to test whether gendered patterns in siblings’ differential achievement varied across racial/ethnic group. Given prior findings based on between-family comparisons (US Department of Education 2006b), we expected that the group comprised primarily of African American and Hispanic dyads would exhibit a larger gender gap. Consistent with this expectation, 51% of girls but only 38% of boys in this group attended college. Within-family comparisons again revealed that, in almost a third of this group, sisters and brothers had different educational outcomes: in 22% of families, sisters but not brothers attended college, whereas in 9% of families, brothers but not sisters attended college. Prior research using between-family analyses has found larger gender gaps in college attendance among lower SES families (e.g. Goldin et al. 2006) and other within-family comparisons have revealed larger sibling differences in achievement in the context of lower economic resources (Conley 2004). In our analyses, the gender gap among “other” racial-ethnic dyads was evident even after controlling for income and education, suggesting that the wider gap is not simply due to African Americans and Hispanics being overrepresented among low SES families.

Sister–Brother Differences in Reports of Maternal Treatment

Our second research goal was to assess sister–brother differences in reports of maternal educational involvement and expectations—two family factors we expected to explain sibling differences in college attendance. We also examined sister–brother differences in school grades, given that grades may be both a cause and a consequence of maternal differential treatment. Within-family comparisons revealed that sisters reported higher grades than brothers, a finding that is consistent with between-family analyses on gender differences during adolescence (Downey and Yuan 2005). There were no overall mean differences between sisters’ and brothers’ reports of maternal treatment, however: in the face of sisters’ significantly higher grades overall, they did not experience “preferential treatment” by their mothers. Further analyses revealed, however, that reports of mothers’ educational expectations varied as a function of sister–brother differences in college attendance in expected ways: in families in which sisters but not brothers attended college, perceptions of maternal expectations favored girls, in families in which brothers but not sisters attended college, perceptions of maternal expectations favored boys, and in families in which sibling attendance outcomes did not differ, neither did maternal expectations.

Although perceptions of mothers’ educational involvement did not covary with gendered patterns of college attendance for the sample as a whole, in the “other race/ethnicity” group mothers were more involved with their sons than with their daughters. This finding was contrary to our hypothesis in one respect, given that we expected sisters to be “favored”, but notably, it was consistent with earlier results that differential treatment is more likely in families facing stressful circumstances (McHale et al. 2006) and data documenting that Hispanic and African American families face more economic hardship and discrimination in contemporary US society than Whites (Fisher et al. 2000; Teachman 2000). Although mothers may involve themselves in their children’s education when youth are performing well, they also may become more involved when youth are performing poorly and need more coaching or supervision. African American and Hispanic boys lag behind girls to a greater extent than do White boys when it comes to academic achievement in adolescence (e.g., Garibaldi 1992), and compared to girls, minority boys who are disadvantaged in the educational system may place less value on the role of education in future achievement, as well as being more frequent victims of discrimination (Colon and Sanchez 2010; Noguera 2003). Mothers may be responding to these types of disadvantages by making concerted efforts to monitor their sons’ academic performance. Different types of and motivations for, maternal educational involvement may obscure the links between mothers’ differential involvement and their children’s educational achievement. In future work, more fine-grained measures of maternal involvement should separate out “remedial” from “supportive” types of maternal educational involvement toward illuminating the role of maternal differential treatment and sibling social comparisons in youth’s educational outcomes.

The Role of Differential Perceptions of Maternal Treatment in College Attendance

Analyses addressing our third research goal yielded results consistent with our hypothesis that sibling differences in reports of mothers’ treatment would predict sister–brother differences in college attendance: beyond the effects of mothers’ mean level of educational expectations, differential perceptions of mothers’ expectations predicted the odds of only sisters versus only brothers attending college. This finding is consistent with previous literature on the role of differential treatment in adolescent siblings’ differential outcomes (McHale et al. 2006), but is novel in its focus on gendered differential treatment and its extension to the domain of differences in siblings’ educational outcomes. Our work also builds on prior findings on how family demographics are related to within-family differences in achievement (Conley et al. 2007) to illuminate the family processes through which these differences may emerge during adolescence. It is noteworthy that the association between differential maternal expectations and sisters’ versus brothers’ odds of attending college emerged after controlling for sister–brother differences in grades. In this way our findings are consistent with ideas advanced by Adler on the importance of social comparisons within the family for the development of children’s sense of self (Ansbacher and Ansbacher 1956). Youth’s self concepts—including their goals and aspirations—may be based not only on their individual achievements but also on how they compare with their siblings. When youth perceive that their sibling is treated by parents as the “smart” child in the family, adolescents may come to think of themselves as less academically capable, and in an effort to minimize rivalry, they also may search for an alternative niche within the family, leaving the role of “student” to their sibling.

Limitations and Future Research

A limitation of this study was the reliance on youth self reports of maternal treatment. Importantly, however, perceptions can have real implications (Graham et al. 1991), and the difference score approach, which takes into account both siblings’ perceptions and is a commonly used measure of parents’ differential treatment in adolescence, avoids the problem of mono-reporter bias. Our study also was limited in its focus on the parenting of mothers in the US. Fathers, as well as mothers, have an impact on their children’s educational and occupational choices (Hellerstein and Morrill 2011). Future studies on the role of family dynamics in the gender gap should incorporate information on family gender socialization and fathers’ role in their sons’ and daughters’ academic achievement. The ethnic group differences we uncovered may signify the role of socio-cultural processes in sister–brother differences and underscore the importance of studying gender differences in educational achievement around the world. Finally, our analyses focused on one mixed sex dyad from each family. Siblings’ experiences within same-sex dyads also may have an impact on patterns of college attendance and, in families with more than two offspring, social comparisons with multiple siblings are possible. It remains for future studies to delve into the complexities of (differences in) differential treatment across multiple sibling dyads in the family. These limitations aside, our findings underscore the significance of siblings’ differential perceptions of maternal expectations during adolescence and suggest that a focus on siblings is central to understanding the family’s role in adolescent academic development. More generally, siblings have been neglected in the study of family socialization, with most prior research on family socialization processes based on the seeming assumption that studying one child in a family is sufficient for understanding how families work to foster adolescent development and achievement. Our findings suggest that future research should move beyond a focus on the parent-adolescent dyad to examine the role of larger family system dynamics in the widening gender gap in educational achievement. Finally, the results of this study underscore the need for further research on other factors related to, and potential consequences of, the current gender gap in higher education. A body of work, inspired by concerns about the under-representation of women in math and science, examined how girls’ experiences were linked to their later educational aspirations and outcomes. Researchers now also need to study how boys’ experiences have implications for the current under-representation of young men in higher education.

Acknowledgments

This research was supported by a grant from NICHD (HD045309), Nancy Landale, PI, and uses data from the National Longitudinal Study of Adolescent Health (Add Health) a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/addhealth). No direct support was received from grant P01-HD31921 for this analysis. We thank Mona Ostrowksi for her programming assistance during this project.

Copyright information

© Springer Science+Business Media, LLC 2011