The Extended Family and Children’s Educational Success

Abstract

Research on family background and educational success focuses almost exclusively on two generations: parents and children. This study argues that the extended family contributes significantly to the total effect of family background on educational success. Analyses using the Wisconsin Longitudinal Study show that, net of family factors shared by siblings from the same immediate family, factors shared by first cousins account for a nontrivial part of the total variance in children’s educational success. Results also show that grandparents’, aunts’, and uncles’ socioeconomic characteristics have few direct effects on educational success. Furthermore, resources in the extended family compensate for lacking resources in low-SES families, which in turn promote children’s educational success. The main conclusion is that the total effect of family background on educational success originates in the immediate family, the extended family, and in interactions between these two family environments.

Keywords

educational success extended family family background first cousins siblings

What is the total effect of family background on children’s educational success? According to conventional models of intergenerational transfers in social stratification research—for example, human capital theory (e.g., Becker and Tomes 1986; Goldberger 1989) and cultural capital theory (e.g., Bourdieu 1977, 1986)—the total effect of family background on educational success originates in parents’ endowments, resources, and propensity to invest in their offspring. These two-generation models of intergenerational transfers focus on parents’ endowments and inputs, and a vast empirical literature documents positive correlations between the educational attainment of parents and their children.

I argue that the conventional two-generation approach to conceptualizing the effect of family background on educational success is too narrow and should be expanded to include the extended family. There are compelling theoretical and empirical reasons for including the extended family. Theoretical models in research on intergenerational family relations propose that, over and above parents, extended family members also affect child outcomes by providing material and affective support to parents and children (e.g., Bengtson 2001; Riley and Riley 1993; Silverstein and Bengtson 1997). Similarly, family demographers argue that two-generation models capture only some of the channels through which intergenerational transmissions occur (e.g., Mare 2011). Consequently, conventional models of intergenerational transfers in social stratification research may be inadequate for understanding the total effect of family background on children’s educational success.

In addition to theory, mounting empirical evidence underscores the extended family’s relevance for children’s educational success. Research shows that most children spend time with extended family members during childhood (Bengtson 2001; Hirshorn 1988). Moreover, scholars have found that the extended family affects child outcomes that are consequential for educational success, such as cognitive development (Modin and Fritzell 2009; Tinsley and Parke 1987), academic achievement (Falbo 1991; Scholl Perry 1996), health (Modin and Fritzell 2009; Osler et al. 2005), and emotional well-being (Dressler 1985; Fergusson, Maughan, and Golding 2008).

This study extends existing social stratification research on family background and educational success in two important regards. The first contribution is proposing a three-generation model of intergenerational transfers. This model builds on the status attainment tradition in social stratification research (e.g., Sewell, Haller, and Portes 1969; Warren and Hauser 1997; Warren, Hauser, and Sheridan 2002), but extends this research by incorporating (1) more extended family members (aunts and uncles as well as grandparents), (2) interdependence between resources in the immediate and extended family, and (3) heterogeneity in the extended family’s effect on children’s educational outcomes.

The second contribution is testing the three-generation model of intergenerational transfers, thereby providing new empirical evidence on the extended family’s effect on children’s educational success. The empirical analysis includes three steps. First, I analyze data from the Wisconsin Longitudinal Study (WLS) and the National Longitudinal Survey of Youth 1979 – Children and Young Adults (NLSY–CYA). Both surveys include respondents who are siblings (related through parents) and first cousins (related through parents’ siblings). I estimate variance components models that decompose the total variance in educational success into factors attributable to the extended family (shared by first cousins), the immediate family (shared by siblings), and individual factors. After controlling for factors shared by siblings from the same immediate family, I find that factors shared by first cousins account for 14 to 28 percent of the total variance in children’s educational success. These estimates provide baseline evidence that the extended family matters.

In the second step, I identify observable socioeconomic status (SES) characteristics of immediate and extended family members that account for similarities in siblings and first cousins in educational success. Previous research, either in the context of three-generation mobility tables (e.g., Biblarz, Bengtson, and Bucur 1996; Erola and Paso 2007; Glass 1954; Svalastoga 1959) or regressions of children’s outcomes on parents and extended family members’ socioeconomic characteristics (e.g., Jencks et al. 1972; Loury 2006; Peters 1992; Ridge 1974; Warren and Hauser 1997), provides some evidence that extended family members affect children’s educational success. I extend previous research by including socioeconomic characteristics of parents, grandparents, and aunts and uncles, but I find little evidence that observed characteristics of extended family members have any direct effects on children’s educational success.

The third step consists of testing for two types of interdependence between effects of the immediate and extended family on children’s educational success. First, research on family relations suggests that extended family members provide extra support to the immediate family when the immediate family has few resources (Eggebeen and Hogan 1990; Hogan, Eggebeen, and Clogg 1993; Wood and Liossis 2007). I test this hypothesis by analyzing whether extended family members’ socioeconomic characteristics have stronger effects on children’s educational success when immediate families have few resources compared to when they have many resources. I find that extended family members’ (and especially grandparents’) socioeconomic characteristics affect children’s educational success in low-SES families but not in high-SES families. This supports the idea that the extended family compensates for lacking resources in low-SES families. Second, research also suggests that the flow of resources within the extended family depends on the quality of family relations (Bengtson 2001; Hirshorn 1988; Sheehan and Petrovic 2008; Silverstein and Ruiz 2006). I test whether effects of extended family members’ characteristics on children’s educational success depend on biological, geographic, or social ties between family members, but I find little evidence to suggest that this is the case.

Theoretical Framework

I propose a theoretical model of intergenerational transfers, illustrated in Figure 1. Conceptually, the model builds on the status attainment tradition in social stratification research and distinguishes three generations of individuals: grandparents (G1), parents (G2), and grandchildren (G3) (Sewell et al. 1969; Warren and Hauser 1997; Warren, Hauser, and Sheridan 2002). I distinguish two types of family effects in the second generation: immediate family G2 effects (G2I; i.e., effects pertaining to parents and siblings) and extended family G2 effects (G2E; i.e., effects pertaining to aunts and uncles). Together, G1 and G2E comprise the types of extended family effects that may exist.¹ Effects of grandparents, aunts and uncles, and parents on G3 educational success are labeled GP, AU, and PR, respectively, and the total effect of family background on G3 educational success is the combined effect of GP, AU, and PR.

Figure 1.

Illustration of Generations and Family Effects

Conventional Models of Intergenerational Transfers

Two-generation models of intergenerational transfers—for example, human capital theory (e.g., Becker and Tomes 1986; Goldberger 1989) and cultural capital theory (e.g., Bourdieu 1977, 1986)—describe endowments and resources that parents transmit to children and mechanisms through which these transmissions occur. Parents transmit endowments such as IQ and health to children through genes (Bouchard and McGue 1981; Plomin et al. 2000; Wardle et al. 2008); and they transmit resources (economic, cultural, and social) by accumulating wealth and shaping the environments in which their children live (Becker and Tomes 1986; Bourdieu 1977).

Incorporating the Extended Family

Most two-generation models assume that intergenerational transmissions follow a Markov process in which endowments and resources are transmitted sequentially from one generation to the next (Lindahl et al. 2011; Mare 2011; Zeng and Xie 2011). Consequently, these models do not accommodate direct effects of the extended family on children’s educational success. In this section I review literature on intergenerational family relations to integrate the extended family into my theoretical framework. I also discuss how interdependence between the immediate and extended family may lead to heterogeneous effects of the extended family on children’s educational success.

Correlations among first cousins’ educational success may arise from shared genes and environments. Like siblings, biologically related first cousins share genes.² With regard to environments, Riley and Riley (1993) describe the extended family as a latent matrix of relationships that may provide support and well-being to family members. Support from the extended family may include direct economic transfers, such as loans, bequests, and inheritances that reduce economic risk and smooth consumption (Altonji, Hayashi, and Kotlikoff 1992; Lacroix, Picot, and Sofer 1998; Mulligan 1997). Support may also be social and emotional, such as practical assistance, childcare, and affection (Bengtson, Biblarz, and Roberts 2002; Coall and Hertwig 2010; Silverstein and Bengtson 1997; Silverstein, Giarrusso, and Bengtson 1998). Finally, support may be indirect, such as through shared social networks and family environments characterized by high educational expectations or cultural capital (Bourdieu 1977, 1986; Picou and Carter 1976; Sandefur, Meier, and Campbell 2006). These types of resources comprise “family capital” (Swartz 2008) that may have direct, non-Markovian effects on G3 educational success, and they are represented by the GP and AU effects in Figure 1.

In addition to explaining why the extended family has a direct effect on G3 educational success, literature on intergenerational family relations also highlights two potential sources of heterogeneity in the extended family’s effect on children’s educational success. These mechanisms, which I label the compensation mechanism and the family quality mechanism, pertain to differences in the propensity of extended family members to provide support to immediate family members. I test both mechanisms in the empirical analysis.

The Compensation Mechanism

My theoretical model of intergenerational transfers assumes that effects of the immediate and extended family on G3 educational success are independent. Moreover, the model assumes these effects influence educational success in a simple, additive fashion. However, research on intergenerational family relations suggests that extended family members (especially grandparents) are particularly likely to support the immediate family in times of need, such as when parents are going through illness, unemployment, or divorce (Eggebeen and Hogan 1990; Hogan et al. 1993; Silverstein, Parrott, and Bengtson 1995; Wood and Liossis 2007). This behavior suggests that a compensation mechanism exists, that is, resources in the family network are mobilized in times of need to offset consequences of lacking resources in the immediate family. The implication of compensating behavior is that GP and AU effects in Figure 1 are not independent of, but rather contingent upon, G2I characteristics; furthermore, this means resources in the extended family are more important for G3 educational success in low-SES families than in high-SES families.

The Family Quality Mechanism

Research on intergenerational family relations also shows that extended family members are more likely to provide support to immediate family members if family relations are high rather than low quality (Bengtson et al. 2002; Silverstein and Bengtson 1997). These results indicate that a family quality mechanism exists, that is, the effect of the extended family on G3 educational success depends on the overall quality of family relations. I consider three dimensions of family quality: biological, geographic, and social (Silverstein and Bengtson 1997; Wood and Liossis 2007). Biological quality concerns whether immediate and extended family members are biologically related; this assumes that extended family members are more likely to support family members with whom they are biologically related compared to non-biological kin (Hamilton, Cheng, and Powell 2007). Geographic quality concerns physical proximity; this assumes that extended family members are more likely to support immediate family members who live close by compared to family members who live further away (Silverstein et al. 1998). Finally, social quality concerns the strength of social ties within families; this assumes that extended family members are particularly likely to support immediate family members with whom they often interact and with whom they share values and a similar outlook on life (Bengtson et al. 2002; Silverstein and Bengtson 1997).

Hypotheses

My theoretical model and the results from previous research indicate three testable hypotheses. The first hypothesis is that genetic and environmental factors in the extended family, which are shared by first cousins, affect children’s educational success. Empirically, this hypothesis entails that over and above the correlation between G3 siblings’ educational attainment (summarizing the total effect of factors pertaining to the immediate family), there should also be a correlation between G3 first cousins’ educational attainment (summarizing the total effect of factors pertaining to the extended family). To test this hypothesis, I use the WLS and NLSY-CYA data and estimate variance components models that decompose the total variance in G3 educational attainment into components attributable to (1) the extended family, (2) the immediate family, and (3) the individual. The key hypothesis to be tested is that the variance component that summarizes the total effect of factors originating in the extended family is statistically significant.

The second hypothesis is that observed characteristics of extended family members account for first cousins’ similarities in educational success. I include observed variables measuring G1 grandparents’ and G2E aunts’/uncles’ socioeconomic characteristics. These variables are proxies for economic, cultural, and social resources in the extended family. I test for direct GP and AU effects on G3 educational success while also controlling for G2I characteristics. Any significant GP and AU effects support the hypothesis that the extended family has a direct effect on G3 educational success.

The third hypothesis pertains to compensation and family quality mechanisms. The compensation mechanism (i.e., the extended family compensates for lacking resources in the immediate family) implies negative interaction effects between G2I and G1/G2E characteristics. In other words, G1 and G2E extended family members’ characteristics will have stronger effects on G3 educational success in low-SES families than in high-SES families. The family quality mechanism (i.e., the extended family’s contribution depends on the overall quality of family relations) implies that the effect of G1/G2E characteristics on G3 educational success depends on the quality of family relations. I expect G3 children will benefit more from resources in the extended family if relations are high rather than low quality. I test this hypothesis by including direct measures of the quality of family relations (biological, geographic, and social) and by testing for interaction effects between these measures and G2E/G1 characteristics. Significant interaction effects provide evidence that effects of G2E/G1 characteristics vary by the quality of family relations.

Data and Variables

Data

I analyzed data from the Wisconsin Longitudinal Study (WLS). The WLS includes a random sample of 10,317 women and men who graduated from Wisconsin high schools in 1957 (and who were born in or around 1939). The WLS collected data on graduates and their parents in 1957, 1964, 1975, 1992/1993, and 2003 to 2005 (WLS 2007, 2008). In addition to WLS graduates, the study also interviewed a random subsample of siblings in 1977, 1993/1994, and 2004 to 2007. The total current population of siblings comprises about 5,600 respondents. Furthermore, the study recently interviewed spouses of both WLS graduates and sibling respondents. Children of WLS graduates and sibling respondents have not been interviewed, but WLS graduates and sibling respondents provided information on all of their children in several WLS waves.

Like Warren and Hauser (1997), I analyzed educational attainment for children of WLS graduates and sibling respondents. In the terminology of Figure 1, these children are G3 respondents; WLS graduates and sibling respondents (and their spouses) are G2 respondents; and parents of WLS graduates and sibling respondents are G1 respondents. In addition to data on three generations, I also exploited the availability of multiple respondents from each WLS family in G2 (WLS graduates and sibling respondents) and G3 (children of WLS graduates and sibling respondents). Figure 2 illustrates this data structure, showing that G3 respondents are nested within G2 families and G2 respondents are nested within G1 families.

Figure 2.

Data Structure in the WLS Analysis Sample

The estimation sample includes 20,747 G3 respondents. These respondents are nested within 7,054 immediate families and 5,652 extended families. Note that because there are more WLS graduates than sibling respondents in the WLS (because only a random subsample of siblings were interviewed), there are also more observations of WLS graduates’ children than of sibling respondents’ children. This means the sample includes 14,115 children of WLS graduates and 6,632 children of sibling respondents. I restricted the analysis sample to G3 children who were at least 25 years old by 2004 and whose parents and aunts/uncles were continuously married throughout the study period (1975 to 2004). The first restriction ensured that children were old enough to have completed their education, and the second restriction ensured that information was available on G2I parents and G2E aunts/uncles. Warren and Hauser (1997) applied similar restrictions.

It is important to point out that because the WLS is not representative of similarly aged U.S. cohorts, the children of WLS graduates and their siblings do not comprise a nationally representative sample. WLS graduates (and their siblings) all come from one geographic area in the United States, most graduated from high school, and almost all are white. Consequently, although the WLS is broadly representative of white, non-Hispanic Americans who completed at least high school (WLS 2007), the empirical results do not generalize to all Americans of similar age. I further discuss limitations of using a selective sample below.

Very few datasets include sibling respondents in more than one generation. The only other dataset I am aware of is the National Longitudinal Survey of Youth 1979 – Children and Young Adults (NLSY-CYA, see CHRR 2006a). The NLSY-CYA includes all children born to women who participated in the National Longitudinal Survey of Youth 1979 (NLSY79; see CHRR 2006b). Because the NLSY79 includes (G2) sisters who eventually become mothers, the NLSY-CYA includes (G3) children who are each others’ siblings and first cousins. Unfortunately, the NLSY-CYA cannot systematically address the current research question because it only includes extended family members on the mother’s side, it contains limited information on extended family members, and sample sizes for G3 respondents are small because the NLSY79 respondents were quite young when they were first interviewed. However, I used the NLSY-CYA as a supplementary data source and calculated baseline estimates of the total effect of the immediate and extended family on various measures of educational success (see Table 3).

Variables

Table 1 shows descriptive statistics for all variables included in the analysis.

Table 1.

Descriptive Statistics for WLS Analysis Sample; Means, Standard Deviations, and Available Cases

	Mean	SD	N
Child Characteristics (G3)
Years of schooling	14.285	2.094	20,747
Sex (1 = female; 0 = male)	.492	.500	20,747
Age in 2004	38.228	5.399	20,747
Immediate Family: Parents (G2I)
Father’s education	13.658	2.723	19,266
Mother’s education	12.957	1.819	19,794
Father’s SES	49.051	25.460	20,666
Family income^a	19.160	11.639	20,747
Family size	3.701	1.676	20,739
Extended Family: Aunt/Uncle (G2E)
Uncle’s education	13.666	2.718	13,349
Aunt’s education	13.170	2.066	13,752
Uncle’s SES	48.492	25.317	14,399
Aunt’s/uncle’s family income^a	18.000	10.892	15,429
Aunt’s/uncle’s family size	3.017	1.834	16,032
Extended Family: Grandparents (G1)
Grandfather’s education	9.625	3.381	20,566
Grandmother’s education	10.367	2.820	20,027
Grandfather’s SES	32.007	21.039	20,712
Grandparents’ family income	6.260	5.942	19,729
Grandparents’ family size	4.469	2.611	16,032
Family Quality: G2I/G2E
WLS grad/sib respondent: not biological siblings	.040	.195	20,747
WLS grad/sib respondent: twins	.016	.124	20,747
WLS grad/sib respondent: both live in WI 1975	.163	.370	20,747
WLS grad/sib respondent: both live in WI 1993	.426	.495	20,747
WLS grad/sib respondent: frequency of contact	8.897	4.664	18,151
WLS grad/sib respondent: feel close to each other	3.088	.801	18,173
Family Quality: G2I/G1
WLS grad/sib respondent: lived with both parents up to age 16 years	.909	.288	16,325
WLS grad/sib respondent: felt taken care of/protected	1.515	.915	17,811
WLS grad/sib respondent: father swore or insulted	1.345	.690	17,732
WLS grad/sib respondent: mother swore or insulted	1.187	.528	17,432
Extended Family Characteristics (G2)
G2 mean cognitive ability^a	6.369	1.247	20,747
G2 mean health	.821	.156	20,740

Some missing data imputed. See text for description.

Dependent variable

The dependent variable is years of completed schooling for G3 respondents. WLS graduates and sibling respondents provided information on the educational attainment of all their children.

Explanatory variables

The principal explanatory variables pertain to G1 grandparents, G2E aunt/uncle, and G2I parents. These variables, which are intended as proxies for economic, cultural, and social resources in the extended and immediate family, capture some of the GP, AU, and PR effects shown in Figure 1. I included the same variables in G1, G2E, and G2I to the extent possible.

First, I included father’s/mother’s, aunt’s/uncle’s, and grandfather’s/grandmother’s level of education measured by years of completed schooling. Note that variables pertaining to G1 grandparents always refer to parents of WLS graduates and sibling respondents (not to parents of the spouses of these respondents). Given that I also controlled for socioeconomic status and family income, variables measuring educational attainment were intended as proxies for the intellectual climate in the home (the WLS does not include any direct measures of cultural capital or learning climate). Second, I included father’s/uncle’s/grandfather’s SES measured by Duncan’s (1961) SEI scale. Third, I included total annual family income from earnings for grandparents, parents, and aunt/uncle. Total family income for G1 grandparents was measured in 1957 dollars, and total family income for G2I parents and G2E aunt/uncle were measured in 1976 dollars. As is often the case, there were nontrivial missing data on parents’ and aunt’s/uncle’s income. I imputed some of the missing data on parents’ and aunt’s/uncle’s income using Stata’s impute procedure and additional income information from the WLS. Specifically, I used information on WLS graduates’ and sibling respondents’ total earnings in 1974 and their total family income in 1992 to 1993 to impute missing information on the family income variables. For parents’ total family earnings, I imputed 4,952 observations (bringing the total number of valid observations to 20,747); for aunt’s/uncle’s total family earnings, I imputed 5,469 observations (bringing the total number of valid observations to 15,429).³ Fourth, I included family size, which measures the total number of children in each of the G1/G2E/G2I families. Finally, because the WLS includes more WLS graduates than sibling respondents (as described earlier, the WLS includes only a random subsample of siblings) and, as a consequence, there are more children of WLS graduates than children of sibling respondents in the sample, I included dummy variables for missing observations of aunt’s and uncle’s education, uncle’s SES, and aunt’s/uncle’s family income and family size.⁴ For G3 children, I controlled for sex (dummy variable for female) and age in years in 2004.

I also included two additional sets of explanatory variables. The first set includes 10 variables intended to capture the biological, geographic, and social quality of the relationship between WLS graduates and their sibling respondents (i.e., G2I/G2E relationships) and between WLS graduates/sibling respondents and G1 grandparents (i.e., G2I/G1 relationships). I have indicators of all three dimensions of family quality (biological, geographic, and social) for G2I/G2E relationships, but only indicators of social quality for G2I/G1 relationships. To capture the biological quality of G2I/G2E family relationships, I created two dummy variables indicating (1) if the WLS graduate and sibling respondent are not biological siblings and (2) if they are twins. To capture geographic quality, I included two dummy variables indicating whether the WLS graduate and sibling respondent both lived in Wisconsin (3) in 1975 and (4) 1993. To capture social quality, I created two variables from the 1993 wave measuring (5) frequency of contact between the WLS graduate and sibling respondent over the past 12 months (0 to 14 scale ranging from “no contact” to “contact more than once per week”) and (6) the extent to which the WLS graduate and sibling respondent feel close to each other (1 to 4 scale with response categories: 1 = “not at all close,” 2 = “not very close,” 3 = “somewhat close,” and 4 = “very close”). To capture the social quality of G2I/G1 family relationships, I created four variables from the 2004 to 2005 wave. These variables measure (7) whether the WLS graduate and sibling respondent lived with both parents up to age 16 years, (8) how often until age 18 the WLS graduate and sibling respondent knew there was somebody to take care of and protect them (1 to 5 scale with response categories: 1 = “never,” 2 = “rarely,” 3 = “sometimes,” 4 = “often,” and 5 = “very often”), (9) to what extent the WLS graduate and sibling respondent’s father insulted or swore at them (1 to 4 scale with response categories: 1 = “not at all,” 2 = “a little,” 3 = “some,” and 4 = “a lot”), and (10) to what extent their mother insulted or swore at them (same response scale as for father).

The second set includes two variables measured at the level of G2 families. These variables are intended as proxies for families’ mean cognitive ability and health, and their relevance is described below. The first variable measures mean cognitive ability of all observed G2 family members (i.e., father, mother, aunt, and uncle). In the 2004 to 2007 WLS waves, the WLS graduate, the sibling respondent, and their spouses all completed the Cognition Similarities Module, which includes a subset of items from the Weschler Adult Intelligence Scale (WAIS). I calculated the mean response on this variable for all observed G2 family members.⁵ The second variable measures mean health of all observed G2 family members. The WLS includes the Health Utilities Index-Mark 3 (HUI), which summarizes respondents’ health status based on eight dimensions: vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain (Furlong et al. 2001). The HUI takes values between 0 and 1, with higher values indicating better health.

Empirical Approach

The WLS sample consists of G3 respondents nested within immediate families (via G3 siblings) and, in some cases, also nested within extended families (via related G2 family members, see Figure 2). I used variance components models to estimate the total effect of the immediate and extended family on G3 educational success (Skrondal and Rabe-Hesketh 2004). I began with a simple variance components model and then extended this model to incorporate explanatory variables pertaining to G3, G2E, G2I, and G1 characteristics. In the simple variance components model, the total variance in G3 educational attainment can be written as follows:

\begin{matrix} y_{isf} = α + v_{f} + u_{sf} + ε_{isf}, \\ v \sim N (0, σ_{v}^{2}), u \sim N (0, σ_{u}^{2}), ε \sim N (0, σ_{ε}^{2}), \end{matrix}

where y_isf denotes years of completed schooling for G3 respondent i (i=1,…,I) belonging to immediate family s (s=1,…,S) and to extended family f (f=1,…,F). In addition to the overall intercept α, Equation 1 includes two random intercepts that capture deviations from the mean level of schooling arising from nesting of respondents within extended families (u_sf, i.e., siblings and first cousins belonging to the same extended family f) and immediate families (ε, i.e., siblings belonging to the same immediate family s). The error term ϵ _isf captures the effect of individual factors. In this model, the total variance in years of completed schooling can be separated into variance components pertaining to the extended family, the immediate family, and individual factors, respectively:

var (y_{isf}) = σ_{v}^{2} + σ_{u}^{2} + σ_{ε}^{2}

where $σ_{v}^{2}$ is the contribution from the extended family and $σ_{u}^{2}$ is the contribution of the immediate family, and the total effect of family background on educational success is $σ_{v}^{2} + σ_{u}^{2}$ . The relative importance of each family environment can be expressed via the intra-class correlation (ICC) coefficient ρ, which measures the proportion of the total variance in G3 educational attainment attributable to each family environment. Below, I provide estimates of $ρ_{v}$ and $ρ_{u}$ .

Two aspects should be kept in mind when interpreting the ICC. First, the ICC is a summary measure that expresses the influence of all factors in the immediate and extended family that affect educational success. Consequently, rather than directly measuring family factors that lead to siblings and first cousins resembling each other with regard to educational success, the ICC summarizes the total, combined effect of these factors on educational success. Second, the ICC captures combined effects of genes and environments.⁶ It is straightforward to extend the simple variance components model in Equation 1 to include explanatory variables pertaining to different family environments. I used a linear model specification and extended the simple variance components model in the following way:

y_{isf} = β_{1} G 3 + β_{2} G 2 I + β_{3} G 2 E + β_{4} G 1 + v_{f} + u_{sf} + ε_{isf},

where, using the terminology from Figure 1, vectors of explanatory variables G3 refer to children, G2I to parents, G2E to aunt/uncle, and G1 to grandparents. The G2I, G2E, and G1 vectors include variables capturing parents’, aunt’s/uncle’s, and grandparents’ socioeconomic characteristics. The G3 vector includes children’s sex and age. The β’s (β = 1,2,3,4) are vectors of regression coefficients, and the random intercepts v and u capture residual first cousin and sibling similarity in educational attainment not accounted for by the observed explanatory variables. The regression model in Equation 3 replicates the basic features of the theoretical model in Figure 1 and tests for direct effects of extended family members’ socioeconomic characteristics on G3 educational success.

I extended the model in Equation 3 to test the compensation and family quality mechanisms described above. I tested the compensation mechanism by adding interaction effects between the G2I and G2E variables and between the G2I and G1 variables to the model. The compensation mechanism predicts negative interaction effects between the G2E/G1 and G2I variables, thus suggesting that extended family members’ characteristics may have a stronger effect on children’s educational success in low-SES families than in high-SES families. In a three-generation model, however, negative interaction effects could also pick up situations in which a G2I family member deviates negatively from her siblings or parents with respect to educational attainment and SES (perhaps due to poor choices, ill health, or bad luck); in this case, instead of a compensation effect, negative interaction effects could capture G3 respondents returning to the “natural” education level in their extended family. I tested the family quality mechanism by including family quality variables described above and by adding interaction effects between these variables and the G2E/G1 variables. Models with interaction effects also include two variables measuring mean cognitive ability and health across all G2 family members. I included these variables to address the possibility that the interaction effects indirectly pick up “regression towards the mean” effects. In addition, by adding mean cognitive ability and health to the model, I hoped to control for the effect of some characteristics in the extended family (e.g., “good” or “bad” genes or environments) that are shared by G2 family members and transmitted to G3 children (Mundlak 1978).

Results

Presentation of the empirical results is divided into three parts.⁷ First, I present results from simple variance components models. These models provide baseline estimates of the total effect of the immediate and extended family on G3 educational success in the WLS and NLSY-CYA. Second, I present results from regression models in which I included observed socioeconomic characteristics of grandparents, aunts/uncles, parents, and children themselves. These results are informative about the strength of direct GP, AU, and PR effects. Third, I test whether the WLS data support the hypothesized compensation and family quality mechanisms described above.

Total Effect of the Immediate and Extended Family on Educational Success

Table 2 presents results from simple variance components models that decompose the total variance in G3 educational success into components attributable to the extended ( $σ_{v}^{2}$ ) and immediate ( $σ_{u}^{2}$ ) family. Measures of educational success in the WLS include years of completed schooling and college completion and, in the NLSY-CYA, test scores in math and reading ability during childhood, high school graduation, years of completed schooling, and college completion.

Table 2.

Results from Variance Components Models Predicting Educational Success

	WLS		NLSY-CYA
Outcome	Years of Schooling^a	College^a	PIAT Math^b	PIAT Reading Recognition^b	High School^c	Years of Schooling^a	College^a
Mean	14.285	.465	51.566	57.080	.819	12.781	.304
Variance Components
ICC cousins ( $ρ_{v}$ )	.144***	.208***	.234***	.187***	.138*	.262***	.286*
ICC siblings ( $ρ_{u}$ )	.264***	.349***	.143***	.185***	.229***	.109***	.160***
ICC individuals^d			.255***	.352***
N Observations	20,747	20,747	26,599	26,537	4,679	2,250	2,250

Note: Significance of variance components tested by means of likelihood ratio tests. Models for high school and college completion are logit models.

Includes respondents age 25 years or older.

Peabody Individual Assessment Test (PIAT) percentile scores.

Includes respondents age 19 years or older.

NLSY-CYA includes repeated observations of math and reading ability test scores for each respondent. Variance components model adds extra variance component to capture within-individual variation in test scores.

p < .05; ** p < .01; *** p < .001 (two-tailed tests).

Table 2 shows that in the WLS, family factors (genetic and environmental) shared by first cousins account for 14.4 and 20.8 percent, respectively, of the total variance in years of completed schooling and the likelihood of graduating from college. By contrast, family factors shared by siblings account for 26.4 and 34.9 percent, respectively. These results suggest that factors in the extended family contribute to the total effect of family background on G3 educational success; furthermore, a two-generation Markov process does not represent the total effect of family background on educational success. Supplementary estimates from the NLSY-CYA provide a similar impression. Here, the extended family accounts for 26.2 and 28.6 percent of the total variance in years of completed schooling and likelihood of graduating from college, respectively. Furthermore, factors in the extended family account for around 20 percent of the total variance in early math and reading ability test scores. In summary, although the estimated ICCs do not provide any information on why siblings and first cousins resemble each other with regard to educational success, they provide strong evidence that factors in the immediate and extended families do contribute to educational success.

Direct Effects of the Extended Family

I now test for direct effects of extended family members’ socioeconomic characteristics on G3 educational success. These characteristics are proxies for economic, cultural, and social resources in the extended family and are part of the direct GP and AU effects in Figure 1.

Table 3 shows results from regressions of years of completed schooling by G3 respondents on different types of family background characteristics. I incrementally added more variables to the models, accounting for direct PR, AU, and GP effects. Table 3 shows two major findings. First, net of PR effects, there is little evidence that extended family members’ socioeconomic characteristics have any direct effect on G3 educational success. Second, the random intercept that captures residual first cousin similarity in educational attainment becomes insignificant when controlling for observed G2I, G2E, and G1 characteristics. Together, these results suggest extended family members’ characteristics have few direct effects on G3 educational success (cf. Loury 2006; Peters 1992; Warren and Hauser 1997). Furthermore, the observed variables completely account for the baseline first cousin similarities in educational success reported in Table 2. Below, I present empirical evidence suggesting that extended family affects children’s educational success in a manner not explained by the baseline models in Table 3.⁸

Table 3.

Effects of Immediate and Extended Family Characteristics on Years of Completed Schooling in the WLS; Parameter Estimates with Standard Errors in Parentheses

	Model 1	Model 2	Model 3	Model 4
Sex (1 = female; 0 = male)	.168*** (.028)	.161*** (.027)	.162*** (.029)	.162*** (.029)
Age in 2004	−.016*** (.003)	.003 (.003)	.003 (.003)	.003 (.003)
Immediate Family (G2I)
Father’s education		.174*** (.009)	.169*** (.009)	.167*** (.010)
Mother’s education		.164*** (.012)	.156*** (.012)	.152*** (.012)
Father’s SES		.010*** (.001)	.010*** (.001)	.009*** (.001)
Family income		.008*** (.002)	.008*** (.002)	.008*** (.002)
Family size		−.070*** (.014)	−.070*** (.012)	−.069*** (.012)
Extended Family: Aunt/Uncle (G2E)
Uncle’s education			.020 (.010)	.019 (.010)
Aunt’s education			.040** (.012)	.037** (.012)
Uncle’s SES			−.001 (.001)	−.001 (.001)
Aunt’s/Uncle’s family income			−.001 (.002)	−.001 (.002)
Aunt’s/Uncle’s family size			.009 (.011)	.009 (.011)
Extended Family: Grandparents (G1)
Grandfather’s education				.003 (.007)
Grandmother’s education				.009 (.008)
Grandfather’s SES				.0008 (.001)
Grandparents’ family income				−.0008 (.003)
Grandparents’ family size				.005 (.008)
Variance Components
ICC cousins	.127***	.013	.014	.012
ICC siblings	.273***	.253***	.250***	.251***
−2 Log-Likelihood	−34,968	−33,963	−33,952	−33,950
Number of Observations	16,858	16,858	16,858	16,858

Note: Models estimated by maximum likelihood. Significance of variance components tested by means of likelihood ratio tests. Models also include dummy variables for missing values on aunt’s/uncle’s education, uncle’s socioeconomic status, aunt’s/uncle’s family income, and aunt’s/uncle’s family size.

p < .05; ** p < .01; *** p < .001 (two-tailed tests).

Compensation and Family Quality Effects

Three reasons might explain why I found no direct effects of extended family members’ socioeconomic characteristics on G3 educational success. First, the G2E and G1 variables might not capture the relevant G2E and G1 characteristics that affect G3 educational success. Second, the WLS, which includes only one set of aunts and uncles (and one set of first cousins) rather than the entire extended family, might provide insufficient variation to identify statistically significant direct effects of extended family members’ characteristics on G3 educational success. Neither of these possibilities can be addressed due to data limitations in the WLS. Third, the three-generation model and its empirical implementation, which assumes independent and homogenous effects of different family environments, might not adequately capture how the extended family actually contributes to children’s educational success. I now turn to this possibility.

The compensation and family quality mechanisms hypothesize interdependence between immediate and extended family members’ characteristics, as well as heterogeneity in the extended family’s effect on G3 educational success. The compensation mechanism predicts that extended family members provide extra resources to immediate family members when the immediate family lacks resources (and fewer resources when the immediate family has plentiful resources). The family quality mechanism predicts that extended family members’ characteristics are more important when family members have strong biological, geographic, or social ties. I test both hypotheses by extending Model 4 in Table 3 to include interaction effects between immediate and extended family members’ characteristics (compensation mechanism) and measures of family quality (family quality mechanism). This model specification departs from Model 4 in two regards. First, I included the two variables measuring cognitive ability and health across all G2 family members. I included these variables to reduce the risk that the results capture the type of regression to the mean described previously. Second, I omitted the random intercept for first cousin similarity in educational success, which was found to be insignificant in Model 4, thus leaving only one random intercept for sibling similarity in the models.

Table 4 summarizes results from models that include interaction effects between G2E and G2I characteristics and G1 and G2I characteristics. The table shows estimated coefficients for each of the G2I and G2E main effects and the interaction effects. We see a clear pattern of positive main effects of the G2E/G1 and G2I variables and negative interaction effects between these variables (interaction effects in Table 4 that do not include stars to denote significance levels are significant at the 10 percent level). This pattern, which is especially evident for the G1xG2I interactions, is consistent with the compensation mechanism stating that extended family members compensate for a lack of resources in immediate families, which in turn affects G3 educational success.

Table 4.

Summary of Interaction Effects Testing Compensation Mechanism; Unstandardized Parameter Estimates; Dependent Variable Is Years of Completed Schooling

	G2I Variable
	Father’s Education	Mother’s Education	Father’s SES	Family Income
G2ExG2I interactions
G2E variable
Uncle’s education
βG2ExG2I		−.0035*	−.0002
βG2E		.064**	.023
βG2I		.179***	.010***
Aunt’s education
βG2ExG2I	−.002*	−.003	−.0002*
βG2E	.064**	.071**	.043**
βG2I	.178***	.172***	.011***
Uncle’s SES
βG2ExG2I		−.0005
βG2E		.006
βG2I		.165***
Aunt’s/uncle’s family income
G1xG2I interactions
G2E variable
Grandfather’s education
βG2ExG2I	−.003	−.007*		−.001**
βG2E	.047	.093*		.024*
βG2I	.190***	.216***		.020***
Grandmother’s education
βG2ExG2I	−.005*	−.010**	−.0005	−.002***
βG2E	.078*	.140**	.031*	.040**
βG2I	.214***	.257***	.014***	.027***
Grandfather’s SES
βG2ExG2I		−.0008		−.0001
βG2E		.012*		.003
βG2I		.175***		.013***
Grandparents’ family income
βG2ExG2I	−.002*	−.003*	−.0002	−.0003*
βG2E	.032*	.040*	.016	.007
βG2I	.172***	.164***	.010***	.010***

Note: N = 16,858. Interaction terms entered individually in each model. Models also include all other variables in Model 4 in Table 3, the variables measuring mean cognitive ability and health across G2 family members, and a random intercept for sibling similarity in educational success.

p < .05; ** p < .01; *** p < .001 (two-tailed tests).

Figure 3 plots the interaction effect between grandfather’s education and family income (G1xG2I) to illustrate the compensation mechanism. The figure shows predicted years of G3 schooling as a function of grandfather’s education and evaluated at different points on the distribution of family income. Instead of using the marginal distribution of family income in the calculations, I used actual income quintiles at different typical levels of grandfather’s education (0 to 7, 8, 9 to 11, 12, and more than 12 years of schooling). This approach provides a realistic depiction because it uses the observed distribution of family income in the WLS. In line with the compensation mechanism, Figure 3 shows that the effect of grandfather’s education on G3 educational attainment is stronger at the lower end of the distribution of family income than at the top. Among G2I families below the 10th income quintile, the predicted difference in educational attainment between having a grandfather with 0 to 7 years of schooling and a grandfather with more than 12 years of schooling is .16 years of schooling. Among families below the 25th income quintile, the estimated difference is reduced to .11 years of schooling, and below the 50th quintile the difference is .07 years. From the 75th percentile onward there is no effect of grandfather’s education on G3 educational success.

Figure 3.

Effect of Grandfather’s Education at Different Quintiles of Family Income

Similar figures could be produced for other interaction effects in Table 4, and the general pattern is that extended family members’ socioeconomic characteristics matter more in low-SES families than in high-SES families. Although statistically significant, it is important to point out that the observed compensation effects are not large in substantive terms. For example, in Figure 3 the difference in the effect of grandfather’s education on G3 educational success at the bottom and the top of the distribution of family income (.16 years of schooling versus no difference) is equivalent to the effect of about one year of parental education. This is not a large effect. However, the main goal here is to demonstrate a consistent pattern of interaction effects and, given the large set of G1, G2E, G2I, and G3 variables included in the models, I did not expect large compensation effects. Finally, although I included a large set of socioeconomic variables and controlled for mean cognitive ability and health across all observed G2 family members, I cannot rule out bias from regression to the mean. For example, rather than capturing a compensation effect, the interaction effects may, to some extent, pick up G3 respondents returning to the mean education level in their extended family. Unfortunately, due to limitations in the WLS, I cannot effectively rule out this possibility.

In addition to the compensation mechanism, I tested the family quality mechanism, that is, the effect of the extended family on children’s educational success depends on the quality of biological, geographic, and social relations between immediate and extended family members. I tested the family quality mechanism by introducing the measures of family quality listed in Table 1 as main effects in the regression models, as well as interaction effects between these measures and the G2E and G1 variables. Again, I entered each main and interaction effect individually into the model. This setup allowed me to test if, for example, the effect of uncle’s education on G3 educational success depends on whether this uncle is biologically related to one of the G3 respondents’ parents (the WLS graduate or sibling respondent), whether the uncle lives close by or far away, or how close the uncle and the parent feel to each other. I ran the models with interaction effects and, surprisingly, found few statistically significant interaction effects.⁹ This result is unexpected but suggests that the extended family’s effect on G3 educational success is not mediated by the quality of family relations (at least not by the types of family quality measured in the WLS).

Conclusions

I argued that the total effect of family background on educational success is more complex than previously assumed because it includes effects from the immediate family, the extended family, and interactions between these two environments. Two-generation models of intergenerational transfers, which are common in social stratification research, do not take the extended family into account, and most empirical research only analyzes two-generation relationships.

Building on the status attainment tradition in social stratification research, I proposed and tested key features of a three-generation model of intergenerational transfers that, in addition to parents, also included effects from extended family members (i.e., grandparents and aunts and uncles) and interactions between different family environments. I analyzed data from the Wisconsin Longitudinal Study and the National Longitudinal Survey of Youth 1979 – Children and Young Adults and found that, net of family factors shared by siblings, factors shared by first cousins account for a nontrivial part of the total variance in educational success. These results suggest that the extended family contributes to educational success through a combination of shared genes, environments, and resources.

I tested for direct effects of extended family members’ (i.e., grandparents and aunts and uncles) socioeconomic characteristics on educational success and found little evidence of direct effects. This result is similar to previous findings that extended family members’ characteristics have little effect on educational success on average (e.g., Warren and Hauser 1997). However, research on family relations suggests that the extended family’s effect need not materialize at the mean. Rather, the effect likely varies across the distribution of family SES and by the quality of family relations. The analysis presented suggests that extended family members’, especially grandparents’, socioeconomic characteristics matter more for children’s educational success in low-SES families than in high-SES families. This result is consistent with the hypothesis that resources in the extended family compensate for negative consequences of growing up in a low-SES family. By contrast, my results provide no evidence that the extended family’s effect on educational success depends on the quality of biological, geographic, or social ties within families.

This study’s key contribution is to expand how social stratification researchers measure the effect of family background on educational success. Conventional two-generation models, which assume that processes of intergenerational transfers follow a Markov process, are myopic and do not take into account, first, all the relevant channels through which family background may affect educational success and, second, that interdependencies and complementarities may exist between different family environments. I proposed a three-generation model of intergenerational transfers, but research on family relations provides building blocks for further incorporating the extended family into a comprehensive theoretical model of intergenerational transfers (e.g., Mare 2011; Silverstein and Bengtson 1997).

Finally, it is important to note several limitations in the present analysis and clarify their implications. First, neither the WLS nor the NLSY-CYA includes information about all members of the extended family. This limitation implies that I observed only parts of the extended family’s effects. Unfortunately, no data presently exist that include information on the entire extended family.

Second, the WLS sample of G3 respondents is not representative of similarly aged U.S. cohorts because almost all of their parents are high school graduates, white, grew up in intact families, and originate from the same geographic area. This limitation means conclusions do not generalize to the U.S. population. Furthermore, this limitation means that compared to a nationally representative sample, I probably underestimated the total variance in G3 educational attainment and the total effect of the immediate and extended family on educational success.

Third, the WLS (and NLSY-CYA) includes information about mostly “hard” socioeconomic characteristics of extended family members. Consequently, I am unable to assess indicators of “soft” characteristics, such as educational expectations, that might have a direct effect on educational success. These characteristics likely transmit more easily across generations in a non-Markovian way than do hard socioeconomic resources and are likely to manifest in first cousin correlations in educational success. Their impact on educational success should be analyzed in future research.

Fourth, observed similarity in educational success among siblings and first cousins summarizes the influence of both genetic and environmental factors (Plomin et al. 2000). It would be informative to separate sibling and first cousin similarity in educational success into components attributable to shared genes versus shared environments. This type of variance separation is not possible in the WLS or NLSY-CYA but would make an important extension of the results presented here.

Fifth, although I control for a comprehensive set of family background characteristics, I cannot rule out the possibility that regression to the mean affects estimates of the compensation effect reported here. Unfortunately, given the design of the WLS, I am unable to address this possibility.

Despite these limitations, this study should stimulate additional research that seeks to estimate the total effect of family background on educational success. The effect of family background on children’s outcomes is a core topic in social stratification research, and understanding how family background affects educational success is important for designing policies to address social inequality. Future social stratification research should undertake the important task of identifying the total effect of family background on educational success, the relative contributions from different family environments to this effect, factors in different family environments that lead to sibling and first cousin similarity in educational success, and interactions between family environments that affect children’s outcomes.

Footnotes

Acknowledgements

I thank the anonymous reviewers and the ASR editors for excellent comments and suggestions. This work was presented at the fall 2009 CIQLE Workshop at Yale University, the 2010 RC28 Spring Meeting in Haifa, Israel, and at various seminars in Denmark. I thank participants at these events and, in particular, Richard Breen, Anette Fasang, Karl Ulrich Mayer, Anders Holm, Kristian Karlson, Signe Hald Andersen, and David Reimer for comments.

Funding

I gratefully acknowledge funding from the Danish Social Science Research Council (grant 275-07-0046). This research used data from the Wisconsin Longitudinal Study (WLS) of the University of Wisconsin-Madison. Since 1991, the WLS has been supported principally by the National Institute on Aging (AG-9775 and AG-21079), with additional support from the Vilas Estate Trust, the National Science Foundation, the Spencer Foundation, and the Graduate School of the University of Wisconsin-Madison.

Data Note

A public-use file of data from the Wisconsin Longitudinal Study is available from the Wisconsin Longitudinal Study, University of Wisconsin-Madison, 1180 Observatory Drive, Madison, Wisconsin 53706 and at .

Notes

Mads Meier Jæger is Professor at the Department of Education, Aarhus University. His research focuses on social mobility and on social and cultural stratification. His recent work includes “Does Cultural Capital Really Affect Academic Achievement? New Evidence from Combined Sibling and Panel Data” (Sociology of Education 2011), “‘A Thing of Beauty is a Joy Forever’? Returns to Physical Attractiveness over the Life Course” (Social Forces 2011) and “Equal Access but Unequal Outcomes: Cultural Capital and Educational Choice in a Meritocratic Society” (Social Forces 2009).

References

Altonji

Joseph G

Hayashi

Fumio

Kotlikoff

Laurence J

. 1992. “Is the Extended Family Linked? Direct Tests Using Micro Data.” American Economic Review 82:1177–98.

Becker

Gary S

Tomes

Nigel

. 1986. “Human Capital and the Rise and Fall of Families.” Journal of Labor Economics 4:S1–S39.

Bengtson

Vern L

2001. “Beyond the Nuclear Family: The Increasing Importance of Multigenerational Ties.” Journal of Marriage and Family 63:1–16.

Bengtson

Vern L

Biblarz

Timothy J

Roberts

Robert E. L

. 2002. How Families Still Matter: A Longitudinal Study of Youth in Two Generations. New York: Cambridge University Press.

Biblarz

Timothy J

Bengtson

Vern L

Bucur

Alexander

. 1996. “Social Mobility Across Three Generations.” Journal of Marriage and the Family 58:188–200.

Bouchard

Thomas J

McGue

Matthew

. 1981. “Familial Studies of Intelligence: A Review.” Science 212:1055–1059.

Bourdieu

Pierre

. 1977. “Cultural Reproduction and Social Reproduction.” Pp. 487–511 in Power and Ideology in Education, edited by Karabel

Halsey

A. H

. New York: Oxford University Press.

Bourdieu

Pierre

. 1986. “The Forms of Capital.” Pp. 241–58 in Handbook of Theory and Research in the Sociology of Education, edited by Richardson

J. G

. New York: Greenwood Press.

CHRR (Center for Human Resource Research). 2006a. NLSY79 Child and Young Adult Data Users Guide. Columbus, OH: Center for Human Resource Research, The Ohio State University.

10.

CHRR (Center for Human Resource Research). 2006b. NLSY79 User’s Guide. Columbus, OH: Center for Human Resource Research, The Ohio State University.

11.

Coall

David A

Hertwig

Ralph

. 2010. “Grandparental Investment: Past, Present and Future.” Behavioral and Brain Sciences 33:1–59.

12.

Dressler

William W

1985. “Extended-Family Relationships, Social Support, and Mental Health in a Southern Black Community.” Journal of Health and Social Behavior 26:39–48.

13.

Duncan

Otis Dudley

. 1961. “A Socioeconomic Index for All Occupations.” Pp. 109–138 in Occupations and Social Status, edited by Reiss

A. J

. New York: Free Press of Glencoe.

14.

Eggebeen

David J

Hogan

Dennis P

. 1990. “Giving between Generations in American Families.” Human Nature 1:211–32.

15.

Erola

Jani

Paso

Moisio

. 2007. “Social Mobility over Three Generations in Finland, 1950–2000.” European Sociological Review 23:169–82.

16.

Falbo

Toni

. 1991. “The Impact of Grandparents on Children’s Outcomes in China.” Marriage and Family Review 16:369–76.

17.

Fergusson

Emma

Maughan

Barbara

Golding

Jean

. 2008. “Which Children Receive Grandparental Care and What Effect Does it Have?” Journal of Child Psychology and Psychiatry 49:161–69.

18.

Furlong

William J

Feeny

David H

Torrance

George W

Barr

Ronald D

. 2001. “The Health Utilities (HIU) System for Assessing Health-Related Quality of Life in Clinical Studies.” Annals of Medicine 33:375–84.

19.

Glass

David V

1954. Social Mobility in Britain. London: Routledge and Kegan Paul.

20.

Goldberger

Arthur S

1989. “Economic and Mechanical Models of Intergenerational Transmission.” American Economic Review 79:504–513.

21.

Hamilton

Laura

Cheng

Simon

Powell

Brian

. 2007. “Adoptive Parents, Adaptive Parents: Evaluating the Importance of Biological Ties for Parental Investment.” American Sociological Review 72:95–116.

22.

Hirshorn

Barbara

. 1988. “Grandparents as Caregivers.” Pp. 200–214 in Handbook on Granparenthood, edited by Szinovâacz

. Westport, CT: Greenwood Press.

23.

Hogan

Dennis P

Eggebeen

David J

Clogg

Clifford C

. 1993. “The Structure of Intergenerational Exchanges in American Families.” American Journal of Sociology 98:1428–58.

24.

Jencks

Christopher

Smith

Michael

Ackland

Henry

Bane

Mary Jo

Cohen

David

Gintis

Herbert

Heyns

Barbara

Michelson

Stephen

. 1972. Inequality: A Reassessment of the Effect of Family and Schooling in America. New York: Harper and Row.

25.

Lacroix

Guy

Picot

Michel

Sofer

Catherine

. 1998. “The Extent of Labour Specialization in the Extended Family: A Theoretical and Empirical Analysis.” Journal of Population Economics 11:223–37.

26.

Lindahl

Mikael

Palme

Mårten

Massih

Sofia Sandgren

Sjögren

Anna

. 2011. “Transmission of Human Capital across Four Generations: Intergenerational Correlations and a Test of the Becker-Tomes Model.” Working Paper 2011-21, Uppsala Center for Labor Studies, Uppsala University, Uppsala, Sweden.

27.

Little

Roderick J. A

Rubin

Donald B

. 2002. Statistical Analysis with Missing Data, 2nd ed. New York: Wiley

28.

Loury

Linda Datcher

. 2006. “All in the Extended Family: Effects of Grandparents, Aunts, and Uncles on Educational Attainment.” American Economic Review 96:275–78.

29.

Mare

Robert D

2011. “A Multigenerational View of Inequality.” Demography 48:1–23.

30.

Modin

Bitte

Fritzell

Johan

. 2009. “The Long Arm of the Family: Are Parental and Grandparental Earnings Related to Young Men’s Body Mass Index and Cognitive Ability?” International Journal of Epidemiology 38:724–32.

31.

Mulligan

Casey B

1997. Parental Priorities and Economic Inequality. Chicago: University of Chicago Press.

32.

Mundlak

Yair

. 1978. “On the Pooling of Time Series and Cross Section Data.” Econometrica 46:69–85.

33.

Osler

Merete

Andersen

Anne-Marie N

Lund

Rikke

Holstein

Bjørn

. 2005. “Effect of Grandparent’s and Parent’s Socioeconomic Position on Mortality among Danish Men Born in 1953.” European Journal of Public Health 15:647–51.

34.

Peters

H. Elizabeth

. 1992. “Patterns of Intergenerational Mobility in Income and Earnings.” Review of Economics and Statistics 74:456–66.

35.

Picou

J. Steven

Carter

T. Michael

. 1976. “Significant-Other Influence and Aspirations.” Sociology of Education 49:12–22.

36.

Plomin

Robert

DeFries

John C

McLearn

Gerald E

McGuffin

Peter

. 2000. Behavioral Genetics. New York: Worth Publishers

37.

Ridge

John M

1974. “Social Mobility in Britain Revisited.” Oxford Studies in Social Mobility, Working Paper 2. Oxford, UK: Clarendon Press.

38.

Riley

Matilda W

Riley

John W

. 1993. “Connections: Kin and Cohort.” Pp. 169–89 in The Changing Contract across Generations, edited by Bengtson

V. L

Achenbaum

W. A

. New York: Aldine de Gruyter.

39.

Sandefur

Gary

Meier

Ann

Campbell

Mary E

. 2006. “Family Resources, Social Capital, and College Attendance.” Social Science Research 35:525–53.

40.

Scholl Perry

Kristin

. 1996. “Relationships among Adolescents’ Ego Development, Their Academic Achievement, and the Amount of Their Contact with and Social Distance from Grandparents.” PhD dissertation, Department of Psychology, Pace University, New York.

41.

Sewell

William H

Haller

Archibald O

Portes

Alejandro

. 1969. “The Educational and Early Occupational Attainment Process.” American Sociological Review 34:82–92.

42.

Sheehan

Nancy W

Petrovic

Kimberly

. 2008. “Grandparents and Their Adult Grandchildren: Recurring Themes from the Literature.” Marriage & Family Review 44:99–124.

43.

Silverstein

Merril

Bengtson

Vern L

. 1997. “Intergenerational Solidarity and the Structure of Adult Child–Parent Relationships in American Families.” American Journal of Sociology 103:429–60.

44.

Silverstein

Merril

Giarrusso

Roseann

Bengtson

Vern L

. 1998. “Intergenerational Solidarity and the Grandparent Role.” Pp. 170–83 in Handbook on Grandparenthood, edited by Szinovacz

M. E

. Westport, CT: Greenwood.

45.

Silverstein

Merril

Parrott

Tonya M

Bengtson

Vern L

. 1995. “Factors that Predispose Middle-Aged Sons and Daughters to Provide Social Support to Older Parents.” Journal of Marriage and the Family 57:465–76.

46.

Silverstein

Merril

Ruiz

Sarah

. 2006. “Breaking the Chain: How Grandparents Moderate the Transmission of Maternal Depression to their Grandchildren.” Family Relations 55:601–612.

47.

Skrondal

Anders

Rabe-Hesketh

Sophia

. 2004. Generalized Latent Variable Modeling: Multilevel, Longitudinal, and Structural Equation Models. Baton Rouge, FL: Chapman and Hall/CRC Press.

48.

Svalastoga

Kaare

. 1959. Prestige, Class, and Mobility. Copenhagen: Gyldendal.

49.

Swartz

Teresa Toguchi

. 2008. “Family Capital and the Invisible Transfer of Privilege: Intergenerational Support and Social Class in Early Adulthood.” New Directions for Child and Adolescent Development 219:11–24.

50.

Tinsley

Barbara J

Parke

Ross D

. 1987. “Grandparents as Interactive and Social Support Agents for Families with Young Infants.” International Journal of Aging and Human Development 25:259–77.

51.

Wardle

Jane

Carnell

Susan

Haworth

Claire M. A

Plomin

Robert

. 2008. “Evidence for a Strong Genetic Influence on Childhood Adiposity Despite the Force of the Obesogenic Environment.” American Journal of Clinical Nutrition 87:398–404.

52.

Warren

John R

Hauser

Robert M

. 1997. “Social Stratification Across Three Generations: New Evidence from the Wisconsin Longitudinal Study.” American Sociological Review 62:561–72.

53.

Warren

John R

Hauser

Robert M

Sheridan

Jennifer T

. 2002. “Occupational Stratification Across the Life Course: Evidence from the Wisconsin Longitudinal Study.” American Sociological Review 67:432–55.

54.

Wisconsin Longitudinal Study (WLS). 2007. Wisconsin Longitudinal Study Handbook, edited by Wollmering

with contributions by Wisconsin Longitudinal Study staff. Madison, WI: University of Wisconsin-Madison.

55.

Wisconsin Longitudinal Study (WLS). 2008. Wisconsin Longitudinal Study [Graduates, Siblings, and Spouses]: 1957–2005, Version 12.24 [machine-readable data file] / Robert M

Hauser

Sewell

William H

[principal investigators]. Madison, WI: University of Wisconsin-Madison, WLS [distributor] (http://www.ssc.wisc.edu/wlsresearch/documentation/).

56.

Wood

Suzanne

Liossis

Poppy

. 2007. “Potentially Stressful Life Events and Emotional Closeness between Grandparents and Adult Grandchildren.” Journal of Family Issues 28:380–89.

57.

Zeng

Zhen

Xie

. 2011. “The Effects of Grandparents on Children’s Schooling: Evidence from Rural China.” Working Paper No. 2011-11, Center for Demography and Ecology, University of Wisconsin-Madison, Madison, WI.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.09 MB

8.34 MB