Low-income mothers’ entry into postsecondary education during middle childhood: Effects on adolescents

Abstract

This study tests whether young adolescents’ achievement and behavior are associated with their mother’s entry into post-secondary education (PSE) during their middle childhood years. It also examines five family processes that may link maternal PSE to development in middle childhood (income, home learning environment, mother’s educational expectations for child, maternal presence, and family affective climate). The sample selects low-income families from the National Longitudinal Survey of Youth of 1979. Propensity score weighting adjusts for mothers’ self-selection into PSE. We find that adolescents whose mothers entered PSE in their middle childhood scored higher than their peers on math, but similarly on reading, behavior problems, delinquency, and substance use. There were no associations between mothers’ PSE entry and the proposed mediators.

Introduction

Adolescents whose mothers have more education, and post-secondary education (PSE; education beyond high school) in particular, have higher academic achievement and exhibit fewer behavior problems than their peers with less-educated mothers (see Kaushal, 2014 for a review). It is tempting to conclude that this inequality would be remedied if less-educated mothers were to return to school to obtain more education. However, this conclusion assumes that the schooling obtained by women after they become mothers confers similar rewards to the schooling obtained by women before they become mothers. In point of fact, most studies that demonstrate the developmental advantage for adolescents with more educated mothers capture the education their mothers attained before they began childbearing. The few studies that have isolated post-childbearing education have tended to focus on education mothers obtained during their child’s early childhood (ages 0–5; Gardner, Martin, & Petitclerc, 2019; Gennetian, Magnuson, & Morris, 2008; Harding, 2015; Rosenzweig & Wolpin, 1994), and only one of these followed the children through adolescence (Gardner et al., 2019).

Within this small body of research, most studies suggest that mothers’ educational attainment during early childhood benefits their children academically (Gennetian et al., 2008; Harding, 2015; Rosenzweig & Wolpin, 1994), although one study indicates that benefits are limited to children with a consistently coresident father figure (Gardner et al., 2019). The evidence with respect to behavior, on the other hand, is concerning. Harding (2015) found that mothers’ educational attainment in early childhood was associated with small increases in teacher-reported aggression, opposition, and inattention in first grade. The only study to follow a sample through adolescence found that mothers’ entry into PSE in their child’s early childhood was not associated with poorer parent-reported behavior at age 7, but predicted moderate-to-large increases in adolescent-reported delinquency and substance use at age 13 (Gardner et al., 2019).

It is possible that mothers’ diversion of time, attention, and money towards educational activities during early childhood adversely affects children’s social-emotional developmental trajectories. Of interest, then, is whether academic benefits may be retained and behavioral consequences avoided if mothers pursued education during their children’s middle, rather than early, childhood years. There is little research to draw on to evaluate this proposition. Only two studies have examined development during middle childhood as a function of mothers’ concurrent schooling. One found no effects on children (Augustine & Negraia, 2018), and the other found positive effects on achievement for the children of young mothers but did not measure effects on behavior (Magnuson, 2007). It is important to know about the developmental consequences of mothers’ educational activities during their child’s middle childhood years because the majority of mothers who earn an Associate’s (AA) or Bachelor’s (BA) degree do so between 5 and 14 years after their first birth (Augustine, 2016). The current study addresses this knowledge gap by examining the effects of mothers’ entry into PSE during middle childhood on their child’s academic and behavioral development in early adolescence.

We focus exclusively on PSE activities, rather than all educational pursuits (such as finishing high school), and its effects on low-income, as opposed to all, mothers, to inform current policy and program efforts to encourage maternal higher education. Specifically, a recent wave of “two-generation” programs has targeted maternal education as a lever for narrowing gaps in development between low-income children and their peers (Chase-Lansdale & Brooks-Gunn, 2014; Gardner, Brooks-Gunn, & Chase-Lansdale, 2017). These programs supplement early education for low-income children with services promoting educational attainment among their mothers. Some emphasize PSE attainment, in particular, because this level of education confers the greatest benefits to children (Dollaghan et al., 1999; Kahn, Wilson, & Wise, 2005). Yet the wisdom of promoting PSE entry during early childhood has been challenged by recent findings that link mothers’ educational attainment in this period to greater behavior problems among their children (Gardner et al., 2019; Harding, 2015). It is therefore useful to consider the middle childhood years as an alternative period in which to promote low-income mothers’ educational attainment.

Timing of Maternal Educational Attainment

Two-generation programs that promote mothers’ own education along with their children’s are motivated by a fundamental tenet of child development: Interventions are more effective the earlier they occur in the life course because developmental trajectories become increasingly difficult to alter over time (Knudsen, Heckman, Cameron, & Shonkoff, 2006).

However, another fundamental tenet of development, advanced by life course theory (Elder, 1988), is that events in children’s lives vary in impact according to their developmental stage. Accordingly, mothers’ educational activities may affect children differently across distinct life stages. The literature to date suggests that mothers’ educational activities during early childhood may boost their children’s achievement (Gennetian et al., 2008; Harding, 2015; Rosenzweig & Wolpin, 1994), but also predict poorer behavior (Gardner et al., 2018; Harding, 2015). It behooves interventionists to consider the possibility that middle childhood may provide as or more advantageous a window as early childhood in which to advance child development via mothers’ education. As children age, mothers spend less time on caregiving (Kalil, Ryan, & Corey, 2012), and the diversion of their time and energy towards educational activities may have fewer adverse consequences for children’s behavior once they enter school.

Of the two extant studies that examined maternal educational attainment during middle childhood, one found no effects on achievement and behavior between ages 5 and 14 (Augustine & Negraia, 2018). The other found positive effects on achievement between ages 6 and 10, but not between ages 10 and 12, and only for children of the least advantaged mothers; moreover, this study did not evaluate behavior (Magnuson, 2007). Interestingly, both studies analyzed the NLSY79, the data set used here, but neither was restricted to mothers pursuing PSE in particular. Additionally, Augustine and Negraia (2018) included all mothers in their sample, whereas Magnuson (2007) included only young mothers with low cognitive scores. Neither sample corresponds to the mothers targeted by two-generation programs, who are low-income but unlimited by age and cognitive score. Therefore, further research is needed to understand the implications of PSE entry in middle childhood among that population of interest. Ideally, mothers who are considering pursuing more education should be armed with knowledge about the likely consequences for their children at different periods of development. There may, for example, be trade-offs by developmental stage, such that educational activities undertaken in early childhood confer greater academic rewards for children but also greater behavioral risk relative to educational activities undertaken in middle childhood.

Potential Mechanisms Linking Maternal Educational Attainment and Middle Childhood Development

To the extent that mothers’ PSE entry in middle childhood affects children, it should be informative to identify the mechanisms responsible. Theory and research suggest five family-level processes through which mothers’ educational activities may translate into advantages or disadvantages for offspring: income, the home learning environment, mothers’ educational expectations for children, maternal presence at home, and family affective climate. We now briefly review each of these in turn.

First, theories of capital posit, and empirical evidence confirms, that a mother’s educational attainment boosts her financial capital by resulting in higher earnings (Becker, 1975; U.S. Bureau of Labor Statistics, 2018), and income is positively associated with child and adolescent development (Duncan & Brooks-Gunn, 1997). However, in the short term, schooling may result in foregone income if it displaces paid employment. Gardner et al. (2019) found that low-income mothers entering PSE in their child’s first five years reported similar income as other mothers at child age 7, but higher income by child age 13. There were no contemporaneous links between mothers’ educational attainment and income in middle childhood in Magnuson (2007), but mothers in another study who attained more education between their child’s Head Start and fifth grade years had higher household incomes by fifth grade than mothers who did not (Pressler, Raver, & Masucci, 2016). Yet the education attained in both of these studies was not necessarily PSE. PSE should take longer than other forms of education such as high school completion, but it should also net larger income returns (U.S. Bureau of Labor Statistics, 2018).

Maternal PSE in middle childhood may also influence development by improving the quality of the home learning environment. Educational activities are thought to boost a mother’s human capital by imparting skills and knowledge that promote children’s cognitive development (Becker, 1975; Harding, Morris, & Hughes, 2015). Educational activities also increase mothers’ social capital by extending social networks to more educated peers, whose attitudes, connections, and behaviors are associated with more favorable child development (Bourdieu, 1986; Harding et al., 2015; Lareau, 1987). Ample research shows that mothers who are more educated provide more stimulating home environments. They spend more time with their children on activities that promote development (Kalil et al., 2012). They read and talk more to their children, and use more complex speech (Hoff, 2003; Hoff-Ginsburg, 1991, 1998; Kuo, Franke, Regaldo, & Halfon, 2004). Several studies have found that mothers who pursue education provide higher-quality home environments for their children (Attewell & Lavin, 2007; Harding, Morris, & Hill, 2017; Magnuson, 2007; Magnuson, Sexton, Davis-Kean & Huston, 2009), but only one (Magnuson, 2007) examined maternal education in middle childhood per se, and its measure of the home environment combined cognitive stimulation and emotional responsiveness. Further, the one study examining maternal PSE in particular found no improvements in the home learning environment (Gardner et al., 2019), albeit during early childhood.

Greater exposure to, and success with, educational pursuits may also lead mothers to form higher expectations for their child’s educational attainment. More educated parents hold higher educational expectations for their child (Davis-Kean, 2005). Higher parental expectations predict children’s higher expectations for themselves (Sewell, Haller, & Ohlendorf, 1970), which in turn predict greater attainment (Martin & Gardner, 2015). A previous study found that mothers who entered school during their child’s early childhood years did not hold higher educational expectations for their child at age 13 than other mothers (Gardner et al., 2019), but the effects of mothers’ schooling may have stronger effects on age 13 expectations if the schooling occurred more recently.

Other family processes potentially affected by mothers’ educational activities are suggested by family systems theory, which stresses that any significant change undertaken by one family member has ramifications for family systems that impact other family members (Minuchin, 1985). We hypothesize that mothers’ entry into PSE should have implications for their presence in the home and their family’s affective climate (expressions of positive and/or negative emotion). Mothers who enroll in PSE have less time to spend on caregiving or leisure with their children (Augustine, Prickett, & Negraia, 2018), and unsupervised time in middle childhood is linked to behavior problems (Dishion & McMahon, 1998). Mothers in school may also encounter stress or fatigue while striving to complete assignments or pay tuition (Augustine et al., 2018), and maternal mental distress is linked to discord with fathers, as well as less warmth and greater hostility with children (Conger, Patterson, & Ge, 1995; Crnic & Low, 2002). A recent study of low-income mothers’ entry into PSE in early childhood did not find links with mothers’ presence at home or the family’s affective climate (Gardner et al., 2019), but no such test has been performed for middle childhood.

The Present Study

The present study addresses several gaps in our knowledge about whether and how mothers’ post-childbearing education shapes child development. First, it examines the middle childhood years in particular because little is known about the effects of mothers’ educational activities during this developmental stage. Second, it focuses on low-income mothers who enter PSE to shed light on the wisdom of current two-generation programs that promote PSE entry for low-income mothers. These programs often target the mothers of young children, but recent research indicates that young adolescents may be at a behavioral disadvantage as a result of their mothers’ entry into PSE during their early childhood (Gardner et al., 2019). This study examines whether mothers’ entry into PSE during middle childhood conveys a similar behavioral disadvantage. Third, this study examines five family processes that may serve as mediating mechanisms for the effects of mothers’ PSE entry during middle childhood.

We draw on a national sample of low-income mothers and children in the United States from the National Longitudinal Survey of Youth 1979 Cohort (NLSY79) using an analytic approach designed to minimize selection bias. Among mothers with low education, those who pursue education after childbearing are generally more advantaged than those who do not (Augustine & Negraia, 2018; Domina & Roksa, 2012; Magnuson, 2007; Pressler et al., 2016; but see Augustine, 2016), which predisposes their children to favorable developmental outcomes. For this reason, it is difficult to isolate the causal effects of maternal education on child development from the effects of characteristics predisposing mothers to enter education. The present study makes use of propensity score weighting, a technique that allows children to be compared according to whether their mothers entered PSE in middle childhood while accounting for the observed characteristics that predisposed their mothers to do so.

Methods

Sample

The NLSY79 is a cohort study of 12,686 American 14-to 21-year-olds that began in 1979. Two subsamples have been followed through to the present day: a nationally representative cohort of 6,111 non-institutionalized, civilian youth and a supplemental oversample of 3,652 Black and Latino youth. In 1986, female participants’ children began to be followed in a substudy now called the NLSY79 Child and Young Adult Cohort. NLSY79 participants were assessed annually through 1994 and biennially thereafter. From the women in this sample who became biological mothers (n = 4,928), we selected those who gave birth to at least one child between 1980 and 1994 (n = 4,309), were classified as low-income (≤ 200% of the federal poverty level) in the year prior to their first birth (n = 1,865), and had not entered a PSE program as of that birth (n = 1,265). Women who gave birth to all of their children before 1980 were excluded because their pre-birth income data were not available, and women who gave birth to all of their children after 1994 were excluded because, at the time of analysis, age 13 follow-up data were not available for their children. Twenty-two mothers with extreme values on control variables (e.g., very late age at first birth) were also excluded. The remaining 1,243 mothers gave birth to a total of 3,696 children. Of those, 86 children were excluded because their mother entered PSE before they were born, and an additional 157 were excluded because their mother entered PSE during their early childhood (between birth and age 5). Children who were observed at age 13 (when they had exited middle childhood and entered early adolescence) were selected for the final analytic sample, which consisted of 1,766 young adolescents belonging to 1,109 mothers. The characteristics of these adolescents are presented in Table 1.

Table 1.

Description of adolescents in sample (n = 1,766)

Characteristic	M (SE)	%
Mother
Entered PSE age 6–11		8%
Age at first birth	19.69 (.15)
Date of birth	June 1961
AFQT percentile	25.76 (1.24)
Self-esteem	20.89 (0.19)
Locus of control	9.23 (0.14)
White		42%
Black		27%
Latina		29%
Other race/ethnicity		2%
Foreign born		6%
Grandmother HS graduate		27%
HS graduate at child’s birth		69%
Married at child’s birth		54%
Employed at child’s birth		46%
Household income at child’s birth (1994 US $)	19,714 (940)
Adolescent
Female		43%
Birth order	1.88 (0.05)
Math	96.37 (0.69)
Reading	98.59 (0.77)
Behavior problems	108.30 (0.76)
Delinquency	0.58 (0.03)
Substance use		35%
Family processes in middle childhood
Income (1994 US $)	26,804 (1,032)
Home learning environment	0.03 (0.04)
Maternal educational expectations	2.31 (0.09)
Maternal presence	−0.03 (0.02)
Family affective climate	−0.10 (0.03)

Note. All statistics are weighted. Standard errors rather than standard deviations are reported because estimates are based on multiply imputed data. Robust standard errors adjust for siblings. PSE= post-secondary education; AFQT = Armed Forces Qualification Test; HS = high school or graduate equivalency diploma.

Measures

Mothers’ PSE entry.

PSE was defined as any vocational program (a non-degree granting program offered by a vocational/technical institute, apprenticeship program, etc.) or academic program (an AA, BA, or graduate degree program) that followed high school or GED completion. Mothers’ high school graduation and postsecondary entry and graduation dates were collected at all survey waves. These dates were used to construct an indicator of initial entry into PSE for all calendar years. Based on these indicators, each adolescent was classified according to whether their mother entered PSE during their middle childhood (ages 6–11) or never entered PSE.

Of the 1,766 adolescents in our sample, 136 (8%) had a mother who entered PSE during their middle childhood. Slightly more adolescents had mothers who entered vocational programs (n = 76) than college programs (n = 60; note that when mothers entered both program types they were classified as entering college). Approximately two-thirds (n = 48, or 63%) of mothers entering vocational programs completed them during middle childhood. Less than one-tenth (n = 5, or 8%) of mothers entering college earned a degree (AA or BA) during middle childhood, although 17% (n = 10) ended up completing a vocational program during this period.

Predictors of mothers’ PSE entry.

The first stage of our analysis predicted mothers’ PSE entry in order to generate propensity score weights. Our predictor variables consisted of adolescents’ sex and birth order as well as characteristics of adolescents’ mothers. Mothers reported annually (or sometimes bienially) on their marital status, employment status, high school graduation or GED completion status, and household income. These in turn were used to create corresponding predictors measured as of the year the adolescent was born.

Self-reports at baseline (1979) provided measures of the mother’s race/ethnicity (Black, White, Latina, other), foreign-born status, date of birth, age at first birth, locus of control (Rotter, 1966), and whether her mother (i.e., the adolescent’s grandmother) had earned a high school diploma. Measures of mothers’ self-esteem (Rosenberg, 1965) and academic aptitude (the Armed Forces Qualification Test, or AFQT, percentile score), administered in 1980, were also used as predictors.

Adolescent outcomes.

Adolescent outcomes were assessed at age 13, two years after the end of middle childhood. If a child’s 13^th year fell between biennial data collection waves, we used data collected at age 14.

Academic outcomes.

Adolescents’ academic outcomes were measured using the Peabody Individual Achievement Test (PIAT; Dunn & Markwardt, 1970) math and reading tests. Scores were standardized within the NLSY Child and Young Adult sample (M = 100, SD = 15).

Socioemotional outcomes.

Adolescents’ internalizing and externalizing problems were measured via maternal report on 28 items from the Behavior Problems Index (BPI; Peterson & Zill, 1986). The total standardized score (M = 100, SD = 15) was used (α = .88; Baker, Keck, Mott, & Quinlan, 1993). A measure of delinquency was created based on adolescents’ self-reports of past-year participation in nine forms of delinquent behavior (e.g., “stayed out later than your parents said you should”). Responses were coded 0 (never) to 3 (more than twice) and averaged (α = .74). A binary measure of substance use was coded affirmatively if the adolescent reported having ever used alcohol, marijuana, or inhalants.

Family process mediators.

Potential mediators were measured at the end of the middle childhood period, namely age 11 (or, if not assessed at that age, age 12).

Income.

We summed mothers’ reports of their prior year family income and income earned by unmarried resident partners to create a measure of total household income. It was expressed in 1994 constant dollars.

Home learning environment.

The standardized score on the Home Observation Measurement of the Environment-Short Form (HOME-SF; Caldwell & Bradley, 1984) Cognitive Stimulation scale captures learning materials and activities at home (e.g., books, trips to museums; α = .62; Baker et al., 1993).

Maternal educational expectations.

Mothers reported how far they thought their child would go in school on a scale from 0 = leave high school before graduation to 4 = get more than 4 years of college.

Maternal presence.

The first and second items were, respectively, the mother’s and child’s reports of how often the mother knew whom her child was with when the child was not home (mother: 1 = rarely to 4 = all of the time; child: 1 = hardly ever to 3 = often). The third and fourth items were the child’s reports of how much time the mother spent with the child (1 = not enough to 3 = too much) and how often the mother missed events or activities that were important to the child (1 = a lot to 3 = almost never). Responses were standardized and averaged. Reliability across items was low (α = .36) because there were two reporters and items did not highly covary; for example, the mother’s knowledge of whom her child was with and the child’s opinion on whether the mother spent enough time with him/her were only modestly correlated. Nevertheless, all items were included to provide as rich a measure as possible. Sensitivity checks showed that when models were run separately for each of the four items, results did not change substantively.

Family affective climate.

Four variables were drawn on: (1) the mother’s report of happiness in her relationship with her romantic partner (1 = not too happy to 3 = very happy); (2) the mother’s report of the frequency of three types of positive exchanges with her partner (e.g., tell each other about their day; 1 = less than once a month to 4 = almost every day); (3) the mother’s report of the frequency of arguments with her partner about nine topics (1 = often to 4 = never); and (4) the HOME-SF Emotional Support scale, which captures mother-reported and interviewer-observed indicators of maternal warmth and lack of hostility (e.g., mother’s display of affection towards child; α = .58; Baker et al., 1993). For families with partnered mothers, scores across all four variables were standardized and averaged (α = .57). The alpha was low because the observed measure of the mother-child relationship did not correlate highly with the mother’s reports of her relationship quality, but all measures were included to achieve as complete a portrait as possible of the family’s affective climate. For families with single mothers, family affective climate was captured only by the HOME-SF Emotional Support scale, which was standardized. Sensitivity checks showed that when models were run separately for the Emotional Support scale and a composite of the relationship variables, results did not change substantively.

Controls.

All variables used as predictors of mothers’ propensity to enter PSE were also used as controls in models of adolescent outcomes and family process mediators. The age of the adolescent at measurement of the adolescent outcome (13 or 14) was also used as a control variable.

Missing Data

Missing data were addressed using all cases prior to selection of our analytic sample. When annual data were needed but unavailable due to biennial data collection patterns, data were carried forward from the previous data collection wave. Rates of missing data due to attrition, item non-response, and irregular item administration patterns over time ranged from 0% to 36% for covariates, from 9% to 16% for outcomes, and from 0% to 55% for family process mediators. Missingness was highest on the measure of family affective climate because the HOME Emotional Support was not completed when the data collector was unable to observe the child and mother together. Data were multiply imputed using mi ice in Stata 13.0 (StataCorp, 2013). We generated 25 imputed data sets.

Analytic Strategy

We used Stata 13.0 for all analyses. Propensity score weighting was used to minimize selection bias (Austin, 2011; Austin & Stuart, 2017; Rosenbaum & Rubin, 1983). This technique is applied to observational data when exposure to a condition of interest, called the “treatment,” is not distributed randomly across subjects. The weighting serves to impose balance between subjects that were exposed to the treatment and those who were not (“controls”). This approach was advisable for the present analyses given pre-existing differences between women who return to school after childbearing and those who do not (Augustine & Negraia, 2018; Domina & Roksa, 2012; Gardner et al., 2019; Magnuson, 2007).

In the present study, we were concerned with pre-existing differences between young adolescents whose mothers entered PSE during their middle childhood (the treatment group) and those whose mothers never entered PSE at all (the control group). As a first step, a logistic regression model was used to predict whether adolescents had a mother who entered PSE in their middle childhood as opposed to not at all. Using this model, we then estimated each adolescent’s propensity to have a mother who entered PSE in middle childhood (i.e., their predicted probability based on the logistic regression model). All models used robust standard errors to adjust for siblings sharing the same mother. Following Stuart (2010), propensity scores were generated separately within each multiply imputed data set, and then combined across data sets.

As the next step, inverse probability of treatment weights (IPTW) were created based on propensity scores (Austin, 2011; Austin & Stuart, 2015; Rosenbaum & Rubin, 1983). The IPTW is equal to the inverse of the probability that a given adolescent experienced the condition (treatment or control) the adolescent actually received. ITPW were computed to yield the average treatment effect in the treated (ATT; Austin, 2011; Austin & Stuart, 2017). For adolescents in the treatment group, the ITPW was set equal to 1. For adolescents in the control group, this weight was computed as their propensity score divided by 1 minus their propensity score. The ATT describes the average effect of the treatment only on the adolescents who experienced the treatment.

Ideally, application of an ITPW should render the treatment and control groups balanced in the distribution of observed baseline characteristics. To diagnose the extent to which balance was achieved, the standardized difference (M_exposed – M_unexposed) / SD_average) was computed for each baseline characteristic before and after weighting with the ITPW (Stuart, 2010). Scholars suggest that standardized differences of less than 0.03 to 0.05 (i.e., 3–5% of a SD) indicate balanced groups (Caliendo & Kopeinig, 2008). A significant improvement in balance across baseline characteristics was achieved by weighting (Table 2). Before weighting, the average standardized difference between the treatment and control groups was .09. After weights were applied, the average standardized difference fell to .02.

Table 2.

Bias estimates before and after weighting

	Middle childhood PSE vs no PSE at all
	Before weighting	After weighting	| |Before|−|After| |
Female	−0.13	−0.01	0.12
Birth order	−0.10	−0.00	0.10
Mother’s age at first birth	−0.17	−0.00	0.17
Mother’s date of birth	0.07	−0.00	0.07
Mother’s AFQT percentile	0.09	0.01	0.08
Mother’s self-esteem	0.01	0.01	0.00
Mother’s locus of control	−0.05	−0.01	0.04
Mother white	0.27	0.04	0.23
Mother black	−0.25	−0.01	0.24
Mother Latina	0.00	−0.02	0.02
Mother other race/ethnicity	−0.04	−0.07	0.03
Mother foreign-born	−0.11	0.06	0.05
Grandmother HS graduate	−0.13	−0.01	0.12
Mother HS graduate at adolescent’s birth	0.04	0.01	0.03
Mother married at adolescent’s birth	−0.06	−0.00	0.06
Household income at adolescent’s birth	−0.09	−0.03	0.06
Mother employed at adolescent’s birth	0.00	−0.00	0.00
Average of absolute values	0.09	0.02	0.07

Note. PSE= post-secondary education; AFQT = Armed Forces Qualification Test; HS = high school or graduate equivalency diploma.

The ITPW were used as probability weights (pweights in Stata; StataCorp, 2013) in OLS regression models of both early adolescent outcomes and middle childhood family processes. Additionally, all baseline characteristics used to predict propensity scores were included as covariates because the weights did not completely eliminate bias (Rubin & Thomas, 2000). Robust standard errors accounted for siblings nested within mothers.

Results

Adolescent Outcomes

Adolescents whose mothers had entered PSE during their middle childhood scored 2.92 points higher on math at age 13 than adolescents whose mothers had never entered PSE (SE = 1.16, p < .05; Table 3). This differential corresponds to an effect size of d = .19. However, adolescents whose mothers had entered PSE during their middle childhood did not score significantly higher on reading than their peers. Furthermore, behavior problems, delinquency, and substance use did not differ among adolescents according to whether their mother had entered PSE in middle childhood.

Table 3.

Adolescent outcomes as a function of maternal PSE entry in middle childhood

	Academic				Socioemotional
	Math		Reading		Behavior problems		Delinquency		Substance use
	b	se	b	se	b	se	b	se	OR	se
Maternal PSE entry in middle	2.92*	1.16	1.37	1.35	0.95	1.40	−0.01	0.05	1.25	0.25
childhood
Female	−1.07	1.67	3.13*	1.39	0.24	1.44	−0.16**	0.05	1.05	0.20
Birth order	−0.93	0.53	0.32	0.62	0.47	0.66	0.02	0.02	0.87	0.08
Age at outcome	0.15	1.30	1.67	1.45	0.82	1.48	0.13*	0.06	1.52	0.35
Mother’s age at first birth	0.58**	0.19	0.37	0.20	−0.50	0.26	−0.01	0.01	0.96	0.04
Mother’s date of birth	−0.04	0.02	0.03	0.03	−0.00	0.03	0.00	0.00	1.00	0.00
Mother’s AFQT score	0.15***	0.03	0.16***	0.04	0.00	0.04	−0.00	0.00	1.00	0.01
Mother’s self-esteem	0.42**	0.16	0.58*	0.18	−0.30	0.19	−0.00	0.01	1.02	0.03
Mother’s locus of control	0.06	0.27	−0.15	0.32	0.23	0.33	−0.00	0.01	0.99	0.05
Mother Black	−1.96	1.40	−3.02	1.73	−0.03	2.00	0.01	0.07	0.91	0.25
Mother Latina	0.16	1.67	0.93	2.07	−1.00	1.83	0.04	0.07	1.42	0.40
Mother other race/ethnicity	0.51	3.34	−0.24	4.41	4.17	2.47	0.24	0.18	1.07	0.86
Mother foreign-born	4.92	2.52	6.10	3.31	0.64	2.55	0.04	0.11	1.31	0.52
Grandmother HS graduate	−0.22	1.59	−0.12	1.77	4.36**	1.65	0.09	0.06	1.68*	0.42
Mother married at adol’s birth	0.15	1.16	0.37	1.44	−0.46	1.47	−0.10	0.06	0.78	0.17
Mother HS graduate at adol’s birth	1.09	1.36	1.87	1.56	−4.48*	1.85	−0.03	0.06	0.77	0.19
Mother employed at adol’s birth	2.83*	1.23	3.25*	1.46	0.19	1.43	−0.02	0.06	1.27	0.27
Household income at adol’s birth	−0.00	0.00	0.00	0.00	−0.00	0.00	0.00	0.00	1.00	0.00

Notes. All models are weighted by the mother’s propensity to enter PSE in middle childhood versus not at all, and are based on 25 imputed data sets. PSE = post-secondary education; AFQT = Armed Forces Qualification Test; HS = high school or graduate equivalency diploma; adol = adolescent.

^*p < .05;

^**p < .01;

^***p < .001.

Family Processes

Mothers’ entry into PSE in middle childhood did not predict any of the five contemporaneous family processes hypothesized to be potential mediators of adolescent developmental outcomes (Table 4). Specifically, children whose mothers entered PSE in their middle childhood did not score differently than other children on family income, quality of the home learning environment, maternal educational expectations, maternal presence in the home, or family affective climate during middle childhood. Because maternal PSE did not predict these family processes, they did not qualify as potential mediators of the effects of maternal PSE entry on the only adolescent outcome predicted by maternal PSE, math (Barron & Kenney, 1986; Kenny, Kashy, & Bolger, 1998). Therefore, no formal tests for mediation were run.

Table 4.

Middle childhood family processes as a function of maternal PSE entry in middle childhood

	Income		Home learning environment		Educational expectations		Maternal presence		Family affective climate
	b	SE	b	SE	b	SE	b	SE	b	SE
Maternal PSE entry in middle	477	2,467	0.06	0.09	−0.01	0.11	−0.03	0.05	0.04	0.09
Childhood Female	−489	1,945	0.10	0.09	0.13	0.11	0.08	0.05	−0.02	0.09
Birth order	−1,251	1,040	−0.05	0.04	−0.04	0.05	−0.05*	0.02	−0.05	0.05
Age at outcome	75	2,361	0.02	0.09	0.10	0.58	−0.01	0.05	0.02	0.08
Mother’s age at first birth	−381	395	−0.00	0.01	0.04	0.02	0.02*	0.01	0.00	0.01
Mother’s date of birth	44	50	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Mother’s AFQT score	141*	68	0.00	0.00	−0.00	0.00	0.00	0.00	0.00	0.00
Mother’s self-esteem	51	320	0.03**	0.01	0.04*	0.01	0.00	0.01	0.02	0.01
Mother’s locus of control	−247	623	0.00	0.02	0.04	0.03	0.00	0.01	−0.02	0.02
Mother Black	−2,007	2,603	−0.14	0.12	0.19	0.21	0.06	0.07	−0.17	0.10
Mother Latina	1,275	3,679	−0.16	0.11	−0.03	0.19	0.08	0.07	0.11	0.13
Mother other race/ethnicity	−10,461*	4,018	−0.16	0.19	0.13	0.37	−0.12	0.14	0.24	0.28
Mother foreign-born	1,117	6,780	0.24	0.18	0.65*	0.27	0.06	0.09	−0.01	0.15
Grandmother HS graduate	4,087	3,215	0.07	0.12	0.13	0.13	0.03	0.06	−0.02	0.10
Mother married at adol’s birth	7,524**	2,271	0.05	0.10	0.03	0.13	0.07	0.05	0.15	0.10
Mother HS graduate at adol’s birth	3,895	2,341	0.04	0.11	0.32	0.21	0.17*	0.07	−0.04	0.10
Mother employed at adol’s birth	2,971	2,183	0.15	0.09	0.11	0.13	-0.01	0.05	0.04	0.08
Household income at adol’s birth	0*	0	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00

Notes. All models are weighted by the mother’s propensity to enter PSE in middle childhood versus not at all, and are based on 25 imputed data sets. PSE = post-secondary education; AFQT = Armed Forces Qualification Test; HS = high school or graduate equivalency diploma; adol = adolescent.

^*p < .05;

^**p < .01;

^***p < .001.

Sensitivity Tests

An earlier study of maternal entry into PSE in early childhood found that adolescent academic outcomes were moderated by the continuous coresidence of a father figure in the home throughout early childhood (Gardner et al., 2019). Specifically, children whose mothers entered PSE in early childhood experienced gains in achievement at age 13 if they had a continually coresident father figure in early childhood. In the present study, we ran interaction models (PSE entry in middle childhood X continuous coresident father in middle childhood) to test the possibility that mothers’ entry into PSE during middle childhood was similarly moderated by continuous father figure coresidence during that period. There was no evidence of such moderation (results available upon request) in models of adolescent outcomes.

Tests were also performed to explore the possibility that associations between maternal PSE entry in middle childhood and adolescent outcomes were limited to children whose mothers actually completed their program – that is, finished a vocational program or earned an AA or BA. Results suggested that even when mothers who entered PSE during their middle childhood finished their program, their children’s adolescent outcomes did not differ from those of adolescents whose mothers did not enter PSE at all (results available upon request).

Discussion

Drawing on a national low-income sample from the United States, this study finds that adolescents whose mother entered PSE during their middle childhood years (ages 6–11) scored higher at age 13 on math than their peers whose mothers never entered PSE. The effect size (d = .19) was small (d < .20; Cohen, 1988). However, they scored similarly on reading and all measures of problem behavior (parent-reported behavior problems and self-reported delinquency and substance use). They also scored similarly on all family processes hypothesized to serve as mediating mechanisms for the effects of maternal PSE entry on adolescent outcomes (family income, home learning environment, educational expectations, maternal presence, and family affective climate).

The two earlier studies that previously examined maternal educational attainment in middle childhood yielded mixed findings. Augustine and Negraia (2018) found null associations between maternal educational attainment and achievement and behavior between ages 5 and 14. Magnuson (2007) found a positive association for achievement but in a circumscribed population. Among the children of young mothers with a high school degree or less, maternal educational advances between ages 6 and 10 were accompanied by increases in reading and math scores, with effect sizes in the small-to-moderate range. No such effects were obtained among the children of older mothers who returned to school. Notably, child behavior was not assessed.

Interestingly, both of these prior studies used the NLSY79, the same data set used in the present study. Magnuson (2007), Augustine and Negraia (2018) and the present study vary in sample eligibility, measures of maternal educational attainment, and methodological approach. Taken together, however, they suggest that mothers’ educational attainment during the middle childhood years may have small but positive associations with achievement and no associations with behavior by early adolescence. Further, the combined findings of this study and Gardner et al. (2019) suggest that there may be a trade-off in adolescent outcomes according to whether maternal PSE entry occurs in middle versus early childhood. The 13-year-olds in Gardner et al. (2019) whose mothers had entered PSE in their early childhood did not benefit academically unless they had coresident fathers, but the size of their advantage was large. By comparison, all 13-year-olds in the present study whose mothers had entered PSE in their middle childhood benefited academically (on average), but only modestly so. On the other hand, the youth in Gardner et al. (2019) whose mothers entered PSE in early childhood had substantively more behavior problems than their peers, whereas the youth in this study whose mothers entered PSE in middle childhood fared no worse on behavior than their peers.

Interestingly, Gardner et al. (2019) found that the implications of maternal PSE in early childhood grew over time, such that unfavorable associations with behavior were visible not at age 7 but at age 13. Relations with achievement also grew over time, with no associations at age 7 but moderate-to-large associations with math and reading at age 13 in families with coresident fathers. The delay in associations may in part be attributable to the fact that many of the mothers did not complete their PSE program until after child age 7. Future research should test whether associations between maternal PSE entry in middle childhood and child development similarly grow over time, for better or for worse, as this sample ages into late adolescence. Some of the 13-year-olds in Gardner et al. (2019) were followed up for as long as 13 years, whereas the longest follow-up in the present sample was 7 years. It will be helpful to reassess this sample in several years after more mothers have had a chance to complete their program. Recall that only about half of the mothers who entered a PSE program in our sample actually completed it while their child was still in middle childhood.

Further, most mothers who completed a PSE program entered a vocational program, not a college program. Only five of the 136 mothers in our sample who entered PSE in middle childhood actually earned an AA or BA in middle childhood. Studies show that college students who are parents take longer than other students to finish their degree, in part because they are more likely to enroll part-time so that they can continue to work part- or full-time (Goldrick-Rab & Sorensen, 2014). Low-income students are also more likely than other students to enter college with insufficient academic preparation and in need of remedial services that stall academic advancement (Deil-Amen & Rosenbaum, 2002). It is possible that this study would have identified larger associations between PSE entry and family processes and child outcomes had more mothers earned college degrees. Therefore, efforts to promote PSE among low-income mothers must try to overcome not only barriers to program entry, but also barriers to program completion. Then again, the investments of time and money needed to enter and complete a college program within the window of middle childhood may well lie beyond the reach of most low-income mothers.

Several limitations of this study should be acknowledged. First, propensity score weighting serves to improve balance between the treatment and control groups on measured but not unmeasured, variables. There may be unmeasured factors, such as maternal mental health, which could have influenced both mothers’ predisposition to enter PSE as well as their children’s development (Pressler et al., 2016). Second, as discussed above, our ability to detect associations between maternal PSE entry and family processes and adolescent outcomes would have been stronger with a larger number of mothers who entered PSE, particularly college. Third, some measures of family processes may have been weak. For example, our measure of maternal presence did not assess the amount of time children spent unsupervised, a risk factor for delinquent behavior (Dishion & McMahon, 1998). Notably, our measure of family affective climate did not capture maternal distress, although it captured the marital relations and parenting behaviors that are likely influenced by maternal distress. Fourth, the age of the data may limit their interpretability due to changes in the landscape of PSE. For example, online programs that are now available were likely uncommon in our sample. Nevertheless, the unusual advantages of our data set – its comparatively large sample size and representativeness, as well as its longitudinal design – render it worthy of consideration. Last, these data are observational, not experimental. The mothers in our sample who entered PSE in their child’s middle childhood did so (presumably in most if not all cases) without assistance from an intervention. It remains to be seen whether mothers whose PSE entry is facilitated by an intervention produce the same effects on children as mothers whose PSE entry is unassisted. For research purposes, it would be ideal for a program to randomly assign participants to treatment and control conditions, and then compare the two groups’ family processes and child outcomes over time.

Highlights.

• Mothers who return to school may boost child achievement but impair behavior

• This study examined low-income families in a national, longitudinal sample using propensity score weighting

• We examined adolescents whose mother entered post-secondary education in their middle childhood years

• They scored slightly higher than matched peers on math but the same on reading and behavior