Diverse Family Types and Out-Of-School Learning Time of Young School Age Children

Abstract

=Sources of differentials in out-of-school learning time between children in first marriage biological parent families and children in six nontraditional family types are identified. Analyses of time diaries reveal that children in four of the six nontraditional family types spend fewer minutes learning than do children in first marriage biological parent families. In all four cases, however, the differentials are explained by the presence of siblings age 18+, lower levels of family income, or younger maternal age.

Identifying sources of family structural inequalities in children’s well-being is of major interest to academics and policy makers in advanced industrialized societies (e.g., McLeod, Nonnemaker, & Call, 2004; UNICEF, 2007). The interest is, in part, a response to findings that children in nontraditional families formed by parental union status, such as remarried parent stepfamilies or cohabiting parent families, had lower average levels of well-being, especially on the academic dimension, than children in first marriage biological parent families (e.g., Brown, 2004, 2006; Jeynes, 2007; Magnuson & Berger, 2009). Marriage promotion programs were developed to help close such gaps, under the assumption that characteristics shared by these nontraditional families, regardless of their socioeconomic circumstances, are the source of this inequality (Graefe & Lichter, 2008). Some studies support this perspective (e.g., Brown, 2006; Raley, Frisco, & Wildsmith, 2005). Artis (2007), for instance, found that marriage-cohabitation gaps in academic outcomes of kindergarten-age children were largely explained by family disruption (e.g., parental union instability) separately from economic status of the family, involvement with children, and maternal depression.

Much remains unknown, however, about whether and how family structure influences young children’s own activities, net of sociodemographic effects, in ways that ultimately affect academic performance (see Teachman, 2008). Rather than children’s own activities, explanatory emphases have been placed on the influences of parents’ behaviors, resulting in seemingly contradictory findings. For example, limited parental time involvement in nontraditional families has been reported as the major process underlying lower level of academic performance of children in these families (e.g., Artis, 2007; Hofferth & Anderson, 2003). At the same time, recent results from children’s time diaries found that the level of parental time involvement was not substantially related to children’s academic test scores and did not explain family structural inequality in test scores (Hofferth, 2006). These seemingly contradictory reports are reconciled when accounting for the fact that most academic activities of children, such as doing homework and engaging in lessons, typically do not directly involve parents (Hewison & Tizard, 2004), even though parents support these activities (Jeynes, 2007). Thus, a focus on children’s learning activities without parental involvement should contribute to an improved understanding of the relationship between family structure and children’s academic performances.

This study tested whether family structural influences persist beyond sociodemographic influences in forming variations in younger children’s out-of-school learning activities observed between nontraditional family types and first marriage biological parent families. Differentials in out-of-school learning time of younger children (ages 6 to 12) between a variety of nontraditional family types versus first marriage biological parent family type were investigated. Out-of-school learning activities have been linked to improvements in children’s cognitive and academic performances (e.g., Bartko & Eccles, 2003; Cooper, Robinson, & Patall, 2006), and hence, are of considerable research interest. Family context may more directly influence children’s activities while outside of school when family members have more opportunities to structure children’s activities. Time diaries of children (age 6 to 12) in the Child Development Supplement (CDS) of the Panel Study of Income Dynamics (PSID), which provide high quality and detailed time-use data on durations of children’s activities, were analyzed.

The influence of nontraditional families could persist, or alternatively, could be summarized by the influences of their family members’ sociodemographic positions. The two hypotheses tested here, a family structural hypothesis and a sociodemographic structural hypothesis, correspond to the former and the latter possibilities (see Background). In addition to parental socioeconomic status, which has received much attention (e.g., Artis, 2007; Brown, 2006), sibling composition not supportive of child’s learning activities, a sociodemographic dimension previously not investigated, is examined here. This dimension may help explain the residual differences in academic test scores between first marriage biological parent families and nontraditional families reported in some previous studies (e.g., Artis; Raley et al., 2005).

Using first marriage biological parent families as a reference group to study patterns in family life is common in family research (e.g., Amato, 2004) and we adhered to this practice. In our hypothesis tests, we included six diverse nontraditional family types defined by parental unions and parent-child relationships: (1) remarried biological parent families, or families in which divorced people have remarried and subsequently had biological children; (2) first marriage stepfather families, or families in which the stepfather is married for the first time; (3) remarried stepfather families, or families in which the stepfather has been previously married; (4) cohabiting biological parent families; (5) cohabiting stepfather families; and (6) single-mother families. This set of nontraditional families includes a broader array of, and more refined, family types than those in previous studies examining the impact of family structure on children’s time use. Previous studies combined family types distinct in structure (which should not be assumed as equivalent), such as single-mother families and cohabiting parent families (e.g., Sayer, Bianchi, & Robinson, 2004), and more typically, first marriage parent families with remarried parent families (e.g., Heard, 2007; Teachman, 2008). In some cases, single parent families were excluded from the analysis (e.g., Artis, 2007; Hofferth & Anderson, 2003).

If results support the family structural hypothesis, then they would imply that in order to reduce family structural inequalities in children’s cognitive and academic well-being through children’s out-of-school learning, programs that support children’s learning activities in nontraditional families would be needed. If results support the sociodemographic structural hypothesis, then they would imply, instead, that supportive programs for children in disadvantageous socioeconomic and demographic positions, such as children in low income families, should be developed.

Family Structural Hypothesis

A family structural hypothesis posits that nontraditional families share characteristics that restrict children’s learning activities, including life disruptions that interrupt children’s learning activities, partly due to caregiving ties that extended beyond the family of residence regardless of their socioeconomic position. An underlying theory - family disruption theory (e.g., Biblarz & Gottainer, 2000; Fomby & Cherlin, 2007)--suggests that children in nontraditional families constructed by parental union dissolution (i.e., remarriage or singlehood), or step-relations (as opposed to biological relations) experience more transitions and instability in their lives than children in first marriage biological parent families (e.g., Raley et al., 2005). These transitions may occur on a regular basis, preventing children from being able to extensively engage in learning activities because such activities may require concentrated efforts by the children and, if organized with other children, complying with a pre-set schedule.

In remarried stepfamilies, for example, children may move between divorced biological parents’ and stepparents’ current residences, partly due to joint custodial arrangements or visitations with nonresident parents, siblings, and extended relatives. They may have to interrupt their activities more frequently in order to do so, or may not be able to participate in organized activities because they need to work around other family members’ schedules. Residential transitions may also interrupt the activities of children with cohabiting parents. Some of these children may live with a part-time cohabiting partner of the mother—the child would then transition between the households of the mother’s and the mother’s partner’s (Boyle, Kulu, Cooke, Gayle, & Mulder, 2006). In single-mother families, rather than living arrangements, the transitions may occur between locations of caregiving. Learning activities may be interrupted by the use of multiple child care arrangements in the mothers’ effort to find quality low-cost care, or any care at all, that matches their work schedule (e.g., Morrissey, 2008).

Some nontraditional families may also share the characteristic of “illegitimate” (i.e., not legally accepted or not socially accepted by many) parenthood (e.g., Ganong & Coleman, 2004). These parents, such as stepparents or cohabiting fathers, may have little authority to set limits, which could constrain the extent to which the family environment could support children’s learning activity. In addition, they may also have incentives to gain children’s affection by being the “easy parent,at least at the onset of the stepparent-child relationship (Mason, Harrison-Jay, Svare, & Wolfinger, 2002)., As a result, they may inadvertently encourage leisure rather than learning.

Thus, a family structural hypothesis leads to the prediction that children in nontraditional families should spend fewer minutes learning than children in first marriage biological parent families, even when holding constant the socioeconomic and demographic positions of the family members. Artis (2007), Raley et al. (2005), and Brown (2004, 2006) provided evidence consistent with this hypothesis.

Sociodemographic Structural Hypothesis

In contrast to a family structural hypothesis, a sociodemographic structural hypothesis posits that the fewer minutes spent learning by children in nontraditional families is ultimately explained by the tendency for parents and children in these families to be placed in disadvantageous socioeconomic and demographic positions. These positions include a sibling composition at risk of resource depletion, fewer siblings, which reduces joint engagement in learning activities among siblings, limited family income, lower level of parental education, and younger parental age. Although socioeconomic status and family structure are interrelated, the sociodemographic structural hypothesis posits that these dimensions parsimoniously summarize the family structural differences in children’s learning time.

Sibship composition

Parents in first marriage biological parent families are typically continuously at risk of having births and have one set of children with one partner. Parents in nontraditional unions, however, can have periods without sexual partners, or can have multiple sets of biological children with different partners. These contrasting childbearing patterns can lead to distinct sibling structures of the children in nontraditional families from those of children in first marriage biological parent families in ways that affect children’s learning time.

Siblings, particularly if they are of similar age and in school, may support a child’s joint participation in learning activities. In contrast, having few or very young siblings, if present, could prevent children from engaging in learning activities. Suggestive evidence that sibship size matters is offered by Hofferth and Sandberg (2001), who found that, net of family income levels, children who lived in families with more children spent more time participating in organized sports. Young siblings may take away parents’ emotional support that would otherwise be available to the school-age child. Previous fragmentary findings show that, compared to first marriage biological parent families, single parent families had fewer children on average (U.S. Census Bureau, 2008). Cohabiting parents also had fewer and very young children, partly because cohabitations tended to form at younger ages and were shorter-lived than first marriages in the United States (Andersson & Phillipov, 2002).

Some remarried parents may sequentially have two sets of children, with a large gap in children’s age between the first and the second sets—one set from the first marriage that ended in divorce, and another born into the remarriage. The parents may wait to have the second set of children until the first set of children gain some social independence, namely when they are teenagers. The second set of children may then experience childrearing resource depletion of the parents. The depletion arises from much of the parents’ financial and psychological childrearing resources being expended on the first set of children who have an earlier birth order (Blake, 1989). Thus, among the young school-age children analyzed here, if fewer minutes of learning time were observed among children in remarried biological parent families, it would be in part explained by them having siblings age 18 or older.

Family income, parental education, and parental age

Lower family income levels, which in part measure limitations in family financial resources, may prevent provision of goods for a home environment conducive to learning (Brooks-Gunn, Klebanov, Smith, & Lee, 2001). Lower level of family income also measures lower social class; lower class parents may place relatively limited emphasis on structured learning activities (Laureau, 2003). Thus, low family income levels should be related to fewer minutes of children’s learning. Evidence supports this predicted relationship (e.g., Evans & Rosenbaum, 2008). Parents in informal unions, namely first marriage stepfamilies, cohabitations, or single parenthood, on average, had lower income than first marriage biological parents (e.g., Manning & Brown, 2006). Exceptions may apply to remarried biological parent and remarried stepparent families whose income levels are similar to that of first marriage biological parents (Sayer, 2006). Thus, other than in remarried parent families, lower family income should at least partly explain the fewer minutes of learning time of children in nontraditional families than children in first marriage biological parent families.

Mothers with higher levels of education are theorized to value quality children rather than a large number of children, and hence have incentives to support the learning activities of their children (e.g., DeLeire & Kalil, 2005). Thus, children whose parents have lower levels of education should spend fewer minutes learning than children whose parents have higher levels of education. Consistent with this prediction, children of better educated parents spent more time studying (Hofferth & Sandberg, 2001) and reading (Bianchi & Robinson, 1997). Those in informal unions and stepfamilies were noted to have parents with lower educational levels on average (Hofferth, 2006). Therefore, parents’ lower levels of education should partly explain the restricted learning time of children in nontraditional families.

Parental immaturity is another dimension that could limit learning time. Older mothers may be more mature, a characteristic that facilitates learning in children (e.g., McLanahan, 2004). Mothers in cohabiting or single parent families tend to be younger than mothers in first marriage (DeKlyen, Brooks-Gunn, McLanahan, & Knab, 2006) but should tend to be older if they are remarried. Thus, the fewer minutes of learning of children in first marriage stepfamilies, cohabiting biological parent families, cohabiting stepparent families, and single-mother families may be in part explained by the younger ages of mothers.

Relationships Between Learning Time and Academic/Cognitive Well-Being

Previous findings demonstrate that a wide array of learning activities is related to higher cognitive and academic scores of children. Activities such as academic work (e.g., homework, studying for exams, or working on school projects), extracurricular activities (e.g., playing musical instruments), and participation in organized activities (e.g., clubs and sports teams) have been linked to children’s improved academic and cognitive test scores (e.g., Bohnert, Aikins, & Edidin, 2007; Cooper et al., 2006). Though counter-intuitive, learning activities also include hobbies, such as crafts and pet care (McHale, Crouter, & Tucker, 2001), and games, including board games and video games. Engagement in hobbies and games has been linked to improved quantitative reasoning (Ramani & Siegler, 2008) and recognition and systematic skills (Shaffer, Squire, Halverson, & Gee, 2005). Computer use, an emerging activity among 6–12 year olds (Ono & Tsai, 2008), was noted as a form of learning related to improved cognitive outcomes, even when playing games (Lieberman, Bates, & So, 2009). As a note, TV watching is not treated as a learning activity here. This is because it is a passive activity (Holder, Coleman, & Shen, 2009) that has been shown to be negatively related to academic achievement (Smith, 1992).

Method

Data

The data we analyze were drawn from the 1997 and 2003 Child Development Supplements (CDS) of the Panel Study of Income Dynamics (PSID). The PSID, which began in 1968 with 4,800 households, was designed to examine economic, social, and demographic changes in the family over time, but was not intended for psychological analyses (see Hill, 1991, for a detailed description of the PSID), and hence lacked depression measures until 2007. Both the 1997 and 2002/3 CDS time diaries were collected in paper format and mailed to respondents with instructions. Respondents were then contacted for an in-house or telephone interview and asked to provide time diary information to the interviewer. Among 6 to 12 year olds, 8% of the diaries were filled out by the child alone (average age = 11). Other diaries were filled out by the primary caregiver (65%), another caregiver (7%), or a child together with the primary caregiver (20%). Time diary data in both waves were collected for one weekday and for one weekend day, and a 24 hour chronology of activities was also obtained. Time use information collected by the diary method is a significant improvement over information collected with stylized methods (Juster, Ono, & Stafford, 2003).

Children who lived with a biological mother in a first marriage, remarriage, cohabitation or single-mother arrangement, and with a biological father or a stepfather, served as the unit of analysis. There were 2007 children in such families, a subsample of the approximately 2,450 CDS children in this age range. Unfortunately, sample size for stepmother families was too small to obtain stable estimates (n = 27). Single biological father families were also too small for analysis (n = 9). Thus, the analyses presented here only included children living with biological mothers. Approximately 400 children among the CDS 6–12 year olds were not living with a biological parent. and lived primarily with their grandparents, and infrequently with other relatives, foster parents, adoptive parents, or in institutions.

Child related measures in the CDS for 6–12 year olds are available at two time points, 1997 and 2003. The second wave of the CDS was collected six years after the first wave, producing a substantial information gap in children’s time use between waves, and thus, reducing the dataset’s usefulness for longitudinal analyses. But, the two waves of data are useful for increasing the sample size. Children in the sample were ages 0–12 in 1997 and approximately ages 6–18 during the 2002-3 school year. Virtually none of the children who were 6–12 in 1997 remained in this age group during the 2002-3 survey, making the 6–12 year olds in 2003 a group distinct from those in the 1997 group. By combining the uniquely identified 6–12 year olds from the two waves of data, the sample size of CDS children in this age group was doubled.

Variables

Dependent variable

The dependent variable was the time (in minutes) children spent in learning activities per week, calculated as five times the week day diary minutes plus twice the weekend day diary minutes. To reflect the theoretical perspective that engagement in active learning during childhood is also important for later life attainment in middle class America (e.g., Laureau, 2003), active forms of learning among the 6–12 year olds such as sports as well as sedentary forms of learning such as reading and studying were included in this measure. As a result, our specification was broader than those of previous studies that focused on sedentary learning activities (e.g., Hofferth & Sandberg, 2001; Yeung, Sandberg, Davis-Kean, & Hofferth, 2001). The reliability of the time-use measure employed here was enhanced by the broader categorization of activities (Juster, Ono, & Stafford, 2003). Specifically, included activity types were: academics (e.g., homework, studying for exams, completing school projects, tutoring and using the computer or visiting the library for school-related reasons), extracurricular activities (e.g., school clubs, band, choir, math club, and lessons), sports (e.g., gymnastics and basketball), organized activities (e.g., volunteering), hobbies (e.g., crafts, pet care, photography, sewing, and reading), non-computer based games (e.g., board games and puzzles), and computer use (e.g., reading on the computer and computer games). The three detailed activities in which children in the sample spent the most time per week were: team sports (230 minutes), computer games (59 minutes), and reading (69 minutes). In the CDS, boys spent approximately 1 more hour playing sports and 10 minutes more using computers than girls, whereas girls spent 15 more minutes reading. For both boys and girls, homework ranked only after these top three activities (32 minutes).

In light of the unanticipated absence of a relationship between parental time involvement and children’s academic outcomes (Hofferth, 2006), we tested whether or not the learning time measure developed here was actually related to children’s academic and cognitive outcomes (see Appendix). Specifically, the Woodcock-Johnson (W–J) problem solving and letter-word scores were used as outcomes. We were unable to use the calculation score because it was not available in the 2002/3 CDS. Digit span memory test score was used as a measure of cognitive outcomes. The results showed that out-of-school learning time was statistically significantly related to the higher achievement and cognitive test scores, except for the W-J verbal component. The conclusion was unchanged when further adding family type to the model (not shown).

Covariates

The covariate of central interest was family type. Categories were created from the cross-classification of parental union type (i.e., first marriage, remarriage, cohabitation, and single) and relationship type (i.e., biological versus step-relations). The variable consisted of seven categories: first marriage biological father (omitted), remarried biological father, cohabiting biological father, first marriage stepfather, remarried stepfather, cohabiting stepfather, and single mother.

To account for sibship composition, we included two covariates. First, age of an older or oldest sibling if present captured the resource depletion that arises from the child being part of the “second set” of children. It also captured the presence of an older sibling of similar age who may support the CDS child’s learning activity and had five categories: no sibling (omitted), sibling present but is younger, older/oldest sibling present and is age 18+, older/oldest sibling present and is age 13–17, and older/oldest sibling present and is age 6–12. This variable was originally differentiated by the age groups of the younger sibling, but this differentiation was later eliminated due to the absence of statistically significant differences. Second, having 2 or more siblings was included in the analysis to capture the presence of more siblings who will support the learning activities of the CDS child. The CDS child was not included in the count. We originally used more detailed categories (i.e., 2, 3, and 4+), but the detailed categories were later collapsed into 2+ due to nonsignificant differences.

To account for parents’ sociodemographic status, we included family income, education of the mother, and age of the mother. Family income had four categories: $0–$39,999, $40,000–$59,999, $60,000+, and no information. A more detailed six category classification of family income was used in earlier regressions. However, additional analyses indicated that a three category specification fits as well as the six category specification and was also more parsimonious. Education had five categories: some high school (omitted), high school graduate, some college, college graduate, and no information. On the basis of preliminary analysis, maternal age was categorized into: younger than 35 (omitted), 35–39, and 40+.

Contrasting first marriage biological parent families and nontraditional family types in their sociodemographic differentials facilitated understanding of how sociodemographic effects could summarize the effects of nontraditional family types in our regressions of interest, which predicted learning time. Thus, we applied simple regressions with family type as the main covariate and sociodemographic characteristics as the outcomes, holding constant only the four child characteristics in all models (see below, not shown but available by the authors upon request). To describe briefly, results showed that all six nontraditional family types have statistically significantly (p ≤ .05) lower levels of maternal education than first marriage biological parent families. Children in all six nontraditional family types except for remarried biological father family type had significantly lower family income and younger mothers and siblings relative to children in first marriage biological parent families. Children in remarried biological parent families, on average, had significantly older mothers, fewer siblings, were both more likely to be an only child and have siblings older than age 17, but were less likely to have older siblings in the same age range of 6 to 12. Remarried biological parents with children were most likely a mix of: (a) parents who divorced without children and had their first biological child in their remarriage; and (b) parents who had children in their former marriage and started their second biological family in their remarriage.

We included four controls for child-sample characteristics: age, gender, race, and year of the interview. Older children not only spend more time in academic activities such as studying, but other structured learning activities as well (Hofferth & Sandberg, 2001; Bianchi & Robinson, 1997). Based on preliminary findings, a linear constraint was imposed. Gender was coded: 0 = boy and 1 = girl. Race was coded in three categories: White (omitted), Black, and Others. All regressions also included the year in which the child was interviewed (0 = 1997, 1 = 2003) to capture differences in the sample structure and survey design between the two waves.

Results

Sample Description

Table 1 contains the sample means and standard deviations of the variables included in the analyses. The means indicate that children in the sample spent an average of 11 hours (652 minutes) per week engaged in out-of-school learning activities. The average age of children in the sample was 9, with similar number of boys and girls. About half (49%) of the children in the sample were Non-Hispanic White, and the remaining half was 39.4% Black, and 11.6 % Others. Just over half of the children lived in first-marriage biological parent families (51.5%), and the other half lived in one of the nontraditional family types. Of those children living in nontraditional family types, most of them (30.8% of the total sample) lived in single-mother families.

Table 1.

Sample Means and Standard Deviations (SD): CDS Children Ages 6 – 12 (N = 2,007).

Variable	Means	SD
Dependent Variable
Weekly out-of-school learning time in minutes	651.910	522.340
Covariates
Family types
First marriage biological parents	.515	--
First marriage stepfather	.038	--
Remarried biological parents	.052	--
Remarried stepfather	.038	--
Cohabiting biological parents	.023	--
Cohabiting stepfather	.026	--
Single-mother	.308	--
Selection
Sibship compositions
Older/oldest sibling’s age if present
Only child	.113	--
No older sibling (younger sibling only)	.128	--
Older/oldest sibling age 18+	.214	--
Older/oldest sibling age 13- 17	.200	--
Older/oldest sibling age 6–12	.347	--
2+ children siblings of the CDS child	.391	--
Parental status
Family Income
No information	.034	--
$0–$39,999	.471	--
$40,000–$59,999	.199	--
$60,000+	.296	--
Maternal education
Some high school	.212	--
High school graduate	.293	--
Some college	.260	--
College graduate	.180	--
No information	.055	--
Maternal age
−34	.415	--
35–39	.284	--
40+	.301	--
Child characteristics
Age	9.042	1.751
Gender (0 = boy, 1 = girl)	.491	--
Race
Non-Hispanic White	.490	--
African American	.394	--
Others	.116	--
Interview year is 2003	.529	--

With respect to sibship composition, more than a third of the children (35%) had an older sibling in the same age range of 6 to 12. Approximately 13% of the children had siblings but younger siblings only. About 20% of the children had much older siblings who were age 18+, and another 20% had somewhat older teenage siblings between the ages of 13 and 17. About 10% of the children were an only child. Parental socioeconomic and demographic status measures showed that slightly less than half (47.1%) of the children were in families with incomes below $40,000. The children’s mothers were fairly evenly spread across maternal educational levels. About 40% of the mothers were in their early 30s, 30% in their late 30s, and another 30% in their 40s and above.

Regression Analysis

Three Ordinary Least Square (OLS) regression models were applied to the sample. Weighted OLS also yielded similar estimates; under this condition, OLS results are preferred, and hence are presented here (see Winship & Radbill, 1994). The first model is the baseline model, holding constant only children’s age, sex, race, and year of interview. The second model tested the role of sibship composition by adding two measures, namely age of the oldest sibling if present and having two or more siblings, to Model 1. The third model tested the role of parental status by adding the three measures of parental sociodemographic status, namely family income, maternal education, and maternal age, to Model 2. If results showed that, even when holding constant sibship composition and parental status (Model 3), significantly lower levels of time were found in all six nontraditional family types, then they would be consistent with the family structural hypothesis. If no statistically significant differences in learning time were found in any of the six nontraditional family types when holding constant sibship composition and parental status, then the results would support the sociodemographic structural hypothesis.

Results from Model 1 are presented in Table 2. They show that, when not holding constant sibship composition and parental status, children in four but not two nontraditional families—remarried stepfather families and cohabiting biological parent families—engaged in anywhere from 121 to 180 fewer minutes of learning time per week than did children in first marriage biological parent families. This result is not consistent with the family structural hypothesis, which predicts that children in all nontraditional families spend fewer minutes engaging in learning activities than children in first marriage biological parent families.

Table 2.

Regression Results from Applying Three Models: CDS children, Ages 6 – 12 (N = 2,007)

	Model 1: Baseline		Model 2: +Sibship Composition		Model 3: +Parental Status
Variables	Coefficient	Std. Error	Coefficient	Std. Error	Coefficient	Std. Error
Family types
First marriage biological parent	--	--	--	--	--	--
First marriage stepfather	−120.645*	60.224	−125.564*	59.667	60.968	59.667
Remarried biological parent	−121.706*	51.267	−86.201	51.170	51.392	51.170
Remarried stepfather	−71.543	58.957	−72.046	58.222	58.421	58.223
Cohabiting biological parent	−128.793	75.821	−115.897	74.910	75.464	74.910
Cohabiting stepfather	−180.389**	70.916	−175.043**	69.984	70.776	69.984
Single-mother	−121.750***	29.013	−111.960***	28.744	32.178	28.744
Selection
Sibship compositions
Older/oldest sibling’s age if present
Only child	--	--	--	--	--	--
No older sibling (younger sibling only)	--	--	28.065	47.690	86.146	48.506
Older/oldest sibling(s), age 18+	--	--	−188.692***	42.356	−194.871***	42.166
Older/oldest sibling(s), age 13- 17	--	--	39.704	43.787	50.089	43.643
Older/oldest sibling age 6–12	--	--	19.470	40.500	53.242	40.791
2+ children siblings of the CDS child			11.840	23.978	30.022	24.217
Parental status
Family income
$0–$39,999	--	--	--	--	--	--
$40,000–$59,999	--	--	--	--	47.707	32.519
$60,000+	--	--	--	--	87.222**	34.000
No information	--	--	--	--	−37.403	69.850
Maternal education
Some high school	--	--	--	--	--	--
High school graduate	--	--	--	--	10.580	33.717
Some college	--	--	--	--	11.931	35.634
College graduate	--	--	--	--	63.024	42.923
No information	--	--	--	--	−57.147	52.854
Maternal age
−34	--	--	--	--	--	--
35–39	--	--	--	--	58.076*	28.903
40+	--	--	--	--	134.881***	31.712
Child characteristics
Age	39.644***	6.388	42.351***	6.472	36.369***	6.535
Gender (0 = boy, 1 = girl)	−58.672**	22.277	−64.740**	22.001	−60.567**	21.839
Race
Non-Hispanic White	--	--	--	--	--	--
African American	−184.827***	27.734	−154.474***	27.869	28.491	27.870
Others	−214.998***	36.529	−194.624***	36.367	41.849	36.367
Interview year is 2003	−103.728***	22.419	−99.604***	23.779	−130.201***	24.735
Intercept	535.379	60.999	509.655	71.740	391.762	391.762
R2	0.089		0.111		0.133

Note. *p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

Two-tailed test.

When controlling only for sibship composition in Model 2, results consistent with the sociodemographic structural hypothesis appeared. The significantly lower level of learning time among children in remarried biological parent families observed in Model 1 was absent in Model 2. The results of Model 2 showed that sibship composition, in particular the presence of older siblings age 18 or older, explained the lower level of reported learning time of children in remarried biological parent families relative to children in first marriage biological parent families. In Model 2, the effect of older/oldest sibling age 18+ was statistically significant, and indicated that children with older siblings age 18+ spent 189 fewer minutes engaged in learning time than only children. These results are consistent with the resource depletion perspective for young children who have substantially older siblings. The presence of older siblings around the same age and two or more siblings did not reach statistical significance, although both have positive coefficients as predicted.

Results from Model 3, which added parental status measures to Model 2, provided full support for the sociodemographic structural hypothesis. When controlling for parental status in addition to sibship composition, no children in nontraditional family types had statistically significantly lower levels of learning time relative to the learning time of children in first marriage biological parent families. The parental status covariates added to Model 3 indicated that children who lived with parents in the lowest income category (below $40,000) as opposed to the highest income category ($60,000+) or who had younger mothers spent fewer minutes engaging in learning activities. Compared to children of mothers in the lowest income category ($0–$39,999), children of mothers in the highest income category spent 87 more minutes of learning time per week. Compared to mothers who were younger than age 35, children with mothers who were between age 35–39 spent approximately 58 more minutes, and those whose mothers were age 40+ spent 135 more minutes per week in learning activities. Little difference in learning time was observed across maternal educational levels. Thus, the fewer minutes of learning time of children in cohabiting stepfather families, first marriage stepfather families, and single-mother families (See Models 1 and 2) were explained by lower family income and younger maternal age.

Covariates measuring child characteristics showed that, when not controlling for parental status, racial minority children exhibited lower levels of learning time than non-Hispanic White children (Models 1 and 2). Once parental status measures were held constant in Model 3, however, no evidence that minority children spent fewer minutes learning than White children was present. All models showed that girls spent about 60 fewer minutes learning than boys. As child age increased by a year, learning time increased by approximately 40 minutes on average.

As a note, we also conducted multiple imputations analyses, in which the no information categories in family income and mother’s education were imputed with father’s education, parents’ age, parents’ work hours, race, and family type (not shown but available from the authors upon request). The conclusions remained unchanged. Because multiple imputations has its own limitations (e.g., Allison, 2002), the analyses using the no information categories are presented here. The R-squared of Model 3 is comparable to models of CDS time diary data in previously published analyses of children’s time use (e.g., Hofferth & Anderson, 2003). The unexplained proportion of variance could be attributable to individual and school characteristics not accounted for in the model or time diary day variations.

Discussion

Sources of variation in children’s learning time between first marriage biological parent families and diverse nontraditional family types are investigated. Results showed that children spend as much as two fewer hours per week of learning time in four of the six nontraditional family types. However, in support of the sociodemographic structural hypothesis, no substantial differences in children’s learning time across family types persist when comparing children in similar sociodemographic positions.

The family structural hypothesis posits that nontraditional families are characterized in part by life disruptions stemming from frequent shifts in caregiving locations and “illegitimate” parenthood, both of which interfere with children’s learning activities, regardless of their family’s sociodemographic position. In contrast, the sociodemographic structural hypothesis posits that the relative disadvantage of children in nontraditional families can be summarized by their contrasting sibship composition, lower parental socioeconomic status, or youth of the parents compared to those of first marriage biological parent families. Empirically, lower (average) parental socioeconomic status and younger (average) maternal age are observed in all nontraditional families except in remarried biological parent families. Remarried biological parents, many of whom have a second set of biological children later in life, may be a select group that is both older and as economically well off as first marriage biological parents. The substantially lower socioeconomic status and younger age of parents in remarried stepfamilies compared to those in first marriage biological parent families were not necessarily expected a priori. But given that these factors are also major determinants of divorce (e.g., Ono, 1998), the results are not surprising.

In support of the sociodemographic structural hypothesis, children in four of the six nontraditional family types (i.e., remarried biological, first marriage stepfather, cohabiting stepfather, and single mother) reported statistically significantly lower levels of learning time. Lower levels of learning time in three of the four family types (i.e., first marriage stepfamilies, cohabiting stepfamilies, and single-mother families) are explained by lower levels of family income, a measure of resource limitation, and mothers’ younger age. The positive influences of higher income and maternal maturity are consistent with those reported in previous studies (e.g., McLanahan, 2004). Children’s fewer minutes of learning time in remarried biological parent families is explained by the presence of an oldest child who is age 18+, a proxy for resource depletion after the parents raised another set of children from the prior marriage before the CDS child was born in the remarriage. Whether the inclusion of other sociodemographic structural characteristics, such as the presence of non-parental adults in the home, maternal employment, and maternal age at first birth alters the conclusion should be investigated in future studies.

The findings reported in this study are in contrast to findings of recent studies that linked family structure and children’s academic test scores, and thus, supported the family structural hypothesis (e.g., Artis, 2007; Brown, 2004; Raley et al., 2005). These contrasting conclusions may be in part due to the fact that the current study uses children’s out-of-school learning time rather than their test scores as the outcome. The current study’s findings suggest that children’s out-of-school learning activities may not be a pathway through which family structure ultimately limits children’s cognitive and academic test scores. Family structural effects not summarized by sociodemographic effects could alternatively affect children’s academic performance through in-school learning (Teachman, 2008), for example, by limiting children’s study network. The possible alternative ways in which family disruptions may affect children’s learning activities should be investigated in future studies. Stronger conclusions would also necessitate annual multi-wave longitudinal analyses. Such data would enable a closer look at how learning time mediates family structural and sociodemographic effects on academic or cognitive outcomes. Nevertheless, our results serve as a baseline for future research.

One of the major findings of this study is that sibship composition helps explain the gap in children’s out-of-school learning time between remarried biological parent families and first marriage biological parent families. Results are consistent with the view that some children in remarried biological parent families who have substantially older siblings are disadvantaged by resource depletion. Further elaborations of the resource depletion process would be useful.

This study focuses on the more common types of nontraditional families formed by parental unions. The range of diversity of family types was somewhat limited by small sample size, and did not enable us to include single father families and stepmother families. The range provided here, however, is broader and more refined than the array of nontraditional families used in previous studies (e.g., Hofferth, 2006). The results imply that first marriage and remarried parent families are not similar family structures and need to be separated in analyses. For example, children in remarried biological parent families have distinct sibship compositions from children in first marriage biological parent families in that they are more likely to either have much older siblings, or are more likely to be an only child. When very large time diary samples of young children become available, future studies could also include same-sex parent families (see Gates, Badgett, Macomber, Ehrle, & Chambers, 2007) or two-adoptive parent families.

In contrast to previous studies (e.g., Hofferth & Sandberg, 2001), this study used a measure of out-of-school learning time that is generally related to cognitive/academic development (with the exception of verbal skills). Whether it is possible to develop a quality measure of out-of-school learning time that is simultaneously related to all components of academic and cognitive outcomes—verbal, nonverbal and cognitive components—would be useful to investigate in future studies. Although a broader measure of learning time has the benefits of reliability and larger variations in learning time, it may have a weak relationship to specific child outcomes. Thus, children’s learning time constructed from time diaries may have tradeoffs between the reliability of the learning time measure and its links to child outcomes. A consequence of the broader definition of learning activity is that it more heavily reflects active learning. Our findings that boys are reported to spend more out-of-school learning time than girls and that the effect of child-race is absent when holding constant parental status may partly be products of this definition (see Jacobs, Vernon & Eccles, 2005; Larson, Richards, Sims, & Dworkin, 2001; Vilhjalmsson & Kristjansdottir, 2003).

There were two nontraditional family types in which substantially fewer minutes of learning were not found, namely cohabiting biological father families and remarried stepfather families. Given the sizable negative coefficient of cohabiting biological parent families in Model 1, part of the reason that the difference in learning time between cohabiting biological parent families and first marriage biological parent families does not reach statistical significance may be its small sample size. Contrasts between these family types may be reexamined when larger samples of children in cohabiting biological parent families become available. In the case of remarried stepfather families, the results may be reflective of stepparent compensation, an adaptive process noted by Hamilton, Cheng, and Powell (2007).

No evidence is available in this study that shifts in caregiving locations or “illegitimate” parenthood substantially disadvantage children’s out-of-school learning activities in nontraditional families. For the academic well-being of young school-age children, the implication of family disruptions or illegitimate parenthood may not be as substantial as theories suggest (Teachman, 2008). Similar conclusions may not apply to other dimensions of well-being, however, such as socioemotional or psychological well-being, which may indirectly affect children’s in-school learning activities. For example, disruptions in daily activities may increase children’s sociobehavioral problems, narrowing their friendship network in ways that limit their participation in in-school learning activities. Replication of the analysis with other dimensions of well-being and in-school learning activities would be a useful next step.

The results have policy implications. They suggest that family structure itself has a minor relationship to children’s learning activities. It is the economic and demographic positions in which nontraditional families are situated that limit these children’s level of learning activity. Thus, to reduce inequality in academic and cognitive well-being across family structures, programs that support the learning activities of children in economically disadvantaged families with young mothers could be developed. For example, TANF could include learning activity programs for children of young low income mothers. Programs that supplement the learning resources available to children who have much older siblings may also be useful for biological children of remarried parents.

Appendix

Regression of Academic and Cognitive Scores on Learning Time, Holding Constant Race, Age, and Gender: CDS children, Ages 6 – 12

	Woodcock-Johnson Problem Solving		Digit Span Memory		Woodcock-Johnson Letter Word
Variable	Coefficient	Std. Error	Coefficient	Std. Error	Coefficient	Std. Error
Learning time (X10)	0.021*	0.009	0.016***	0.002	−0.077	0.091
Family income
$0–$39,999	--	--	--	--	--	--
$40,000–$59,999	3.725***	1.311	0.382	0.414	49.824***	13.077
$60,000+	8.861***	1.292	0.522	0.380	141.334***	12.074
Maternal education
Some high school
High school graduate	0.846	4.112	0.272	1.182	13.400	40.369
Some college	5.942	4.115	2.023	1.185	−51.800	40.450
College graduate	9.311*	4.204	2.806*	1.216	−54.670	41.230
No information	0.001	4.534	−0.145	4.539	37.263	42.530
Maternal age
−34	--	--	--	--	--	--
35–39	0.947	1.207	−0.209	0.374	−4.231	11.834
40+	1.396	1.241	0.084	0.388	30.564*	12.216
Children’s race
Non-Hispanic White	--	--	--	--	--	--
African American	−7.168***	1.183	0.088	0.376	−1.527	11.920
Others	−3.775*	1.629	−0.061	0.453	208.750***	14.327
Children’s gender (0 = boy,     1= girl)	−0.728	0.928	0.893***	0.290	0.708	9.175
Children’s age	0.774***	0.276	0.808***	0.277	−1.022	2.722
Interview year is 2003	−0.251	2.472	1.483*	0.604	−319.051***	19.355
Intercept	95.190	4.627	.637	1.353	180.096	28.611
R2	0.202		0.111		0.414

^*p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

Two-tailed test.

Notes: n = 1071 for problem solving scores; n = 1241 for digit span memory score; n = 1131 for letter-word score. All of these cases had known level of family income and hence the missing value category for family income is absent from these regressions.