Growth curve analyses of the relationship between early maternal age and children’s mathematics and reading performance

Social Science Research 50 (2015) 343–366

Contents lists available at ScienceDirect

Social Science Research

journal homepage: www.elsevier .com/locate /ssresearch

Growth curve analyses of the relationship between earlymaternal age and children’s mathematics and readingperformance q

http://dx.doi.org/10.1016/j.ssresearch.2014.09.0010049-089X/� 2014 Elsevier Inc. All rights reserved.

q I would like to thank David J. Harding, Barbara A. Anderson, Yu Xie, Brian A. Jacob, and the anonymous reviewers for Social Science Researchhelpful comments and suggestions for revisions on previous drafts of this paper.

E-mail address: [email protected]

D. Diego TorresHouston Education Research Consortium, Kinder Institute for Urban Research, Rice University, 6100 Main Street, MS-28, Houston, TX 77005, United States

a r t i c l e i n f o

Article history:Received 6 September 2013Revised 17 August 2014Accepted 10 September 2014Available online 21 September 2014

Keywords:Early maternal ageTeen first birthAchievementGrowth curve modeling

a b s t r a c t

Regarding the methods used to examine the early maternal age-child academic outcomesrelationship, the extant literature has tended to examine change using statistical analysesthat fail to appreciate that individuals vary in their rates of growth. Of the one study I havebeen able to find that employs a true growth model to estimate this relationship, theauthors only controlled for characteristics of the maternal household after family forma-tion; confounding background factors of mothers that might select them into early child-bearing, a possible source of bias, were ignored. The authors’ findings nonethelesssuggested an inverse relationship between early maternal age, i.e., a first birth betweenthe ages of 13 and 17, and Canadian adolescents’ mean math performance at age 10. Earlymaternal age was not related to the linear slope of age. To elucidate whether the earlymaternal age-child academic outcomes association, treated in a growth context, is consis-tent with this finding, the present study built on it using US data and explored children’smathematics and reading trajectories from age 5 on. Its unique contribution is that it fur-ther explicitly controlled for maternal background factors and employed a three-levelgrowth model with repeated measures of children nested within their mothers. Thoughthe strength of the relationship varied between mean initial academic performance andmean academic growth, results confirmed that early maternal age was negatively relatedto children’s mathematics and reading achievement, net of post-teen first birth child-specific and maternal household factors. Once maternal background factors were included,there was no statistically significant relationship between early maternal age and eitherchildren’s mean initial mathematics and reading scores or their mean mathematics andreading growth.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

Scholarship on the adverse effects of early maternal age (defined as a first birth in the teen years) on children’s social,academic, behavioral, and health outcomes continues to be mixed. While some researchers find evidence in support of anassociation between early maternal age and many outcomes such as school dropout, delinquency, internalizing andexternalizing behaviors, and lower reading and mathematics scores during childhood and adolescence—either directly or

for their

http://crossmark.crossref.org/dialog/?doi=10.1016/j.ssresearch.2014.09.001&domain=pdf

http://dx.doi.org/10.1016/j.ssresearch.2014.09.001

mailto:[email protected]

http://dx.doi.org/10.1016/j.ssresearch.2014.09.001

http://www.sciencedirect.com/science/journal/0049089X

http://www.elsevier.com/locate/ssresearch

344 D.D. Torres / Social Science Research 50 (2015) 343–366

indirectly via the stunted schooling, low income, and poor parenting skills of young mothers, and often to the exclusion ofmothers’ background factors (Hardy et al., 1997; Hoffman and Maynard, 2008; Moore et al., 1997; Spieker et al., 1999)—others find that any negative relationship between early maternal age and children’s outcomes is difficult to establish, oris spurious, once mothers’ backgrounds are accounted for (Geronimus and Korenman, 1992; Geronimus et al., 1994;Hoffman, 1998; Hotz et al., 1997; Levine et al., 2007; Rosenzweig and Wolpin, 1995; Turley, 2003).

The life course theoretical framework (Neugarten et al., 1965) provides support for some of the earliest studies showingand inverse relationship between early maternal age and children’s outcomes. Specifically, the reason why children ofmothers who began childbearing in the teen years fare poorly is due both to the fact of their mother’s early maternalage and its effect on the health of the maternal household as characterized by their mothers’ completed education andincome and their relationship to the availability or unavailability of social capital-producing resources. Some scholarsargue, on the other hand, that early maternal age is itself an effect of the socioeconomic disadvantage that select youngwomen into early childbearing, and, inasmuch as it is associated with poor outcomes among the children of motherswho had a teen first birth, it is not an additional cause of the difficulties that these young women face later in life(Edin and Kefalas, 2005; Kearney and Levine, 2011, 2012). The later disadvantage of teen mothers is actually part of thetrajectory these women were already on prior to their early maternal age. The failure to account for the background char-acteristics of mothers who began their childbearing early, then, necessarily leads to overestimations of the early maternalage-child outcomes relationship.

Whether early maternal age is truly associated with young women’s later difficulties and, hence, signals negative out-comes for their children, or whether it merely represents part of the normal path some young women travel given their fam-ily backgrounds, it is notable that, with respect specifically to issues of methods, typical longitudinal analyses of change,which often focus on score changes in a two-wave design, have been restricted to explicitly modeling aggregate-levelgrowth, and thus fail to take into account that individuals vary in their rates of growth (Gibbons et al., 1993). Dahintenet al. (2007) provide the only study I could find that examines the effects of early maternal age in a growth context. The out-comes of interest, however, were primarily behavioral; only one outcome, i.e., children’s performance on a math instrument,was academic. Also, the focus was on Canadian adolescents. Finally, because the authors controlled for characteristics of thematernal household after family formation and did not account for differences in the background factors of the women intheir study, it is likely the case that the relationship between early maternal age and children’s math achievement wasoverestimated.

The present study extends the current literature on the early maternal age-child outcomes relationship by replicating theDahinten et al. (2007) study. Its unique contribution is that, in contrast to Dahinten et al. (2007), it (1) uses US data, (2)focuses on children’s achievement on the Peabody Individual Achievement Test (PIAT) for mathematics, reading recognition,and reading comprehension, (3) follows children from age 5, the age at which the majority of US children enter formalschooling, to middle adolescence, (4) explicitly controls for maternal background factors in addition to accounting forpost-teen first birth child-specific and maternal household factors, and (5) employs a three-level growth model to accountfor non-independence of children’s achievement both across time and within families. The specific questions posed are (1) Inthe absence of controls, is early maternal age associated with children’s mean initial and time-dependent mathematics andreading achievement? (2) Granting an association under the preceding question, does early maternal age continue to berelated to children’s mean initial and time-dependent mathematics and reading achievement, net of child-specific andmaternal household controls? and (3) Granting an association under the preceding question, does early maternal age con-tinue to be related to children’s mean initial and time-dependent mathematics and reading achievement when furtheraccounting for the background factors that might select young women into early childbearing? Given the extant literatureon the early maternal age-child outcomes relationship, it is safe to hypothesize that early maternal age is in fact associatedwith children’s mean achievement. Less clear is whether any association will hold once the unobserved heterogeneity inmothers’ backgrounds is explicitly controlled for.

2. Background and significance

2.1. Theoretical framework

The association between early maternal age and children’s outcomes may be explained theoretically by the life courseframework, particularly its emphasis on age-specific expectations or ‘‘on-time transitions’’, and the inability of some, per-haps due to circumstances outside their control, to meet them (Neugarten et al., 1965). Childbearing, which ideally shouldoccur in adulthood, and preferably after formal schooling and marriage, has long-term negative consequences whenachieved in the teen years, both for young mothers and for the children born to them, because it adversely impacts resourceallocation in the home and delays the psychological and emotional maturation young women need to be good parents(Mollborn and Dennis, 2012; Powell et al., 2006). A birth in the teen years is likely to disrupt young women’s completionof education, which, in turn, is associated with penury and its concomitant social disadvantage in adulthood (Coley andLindsay Chase-Lansdale, 1998; Furstenberg et al., 1990). For example, it has long been established that verbal facility, whichis positively associated with reading ability, is highest among children born to and reared by older mothers, who tend also tohave higher developed mental ability (Parcel and Menaghan, 1990).

D.D. Torres / Social Science Research 50 (2015) 343–366 345

Early child-bearers are also more likely to raise a child without the support of the child’s father, which, according toresearch, is detrimental to the educational attainment of the children of these women (Astone, 1993).1 Because they havefew or no resources on which to draw during the childrearing years, young women oftentimes end up as the heads of homescharacterized, in particular, by the lack of fixed routines or regularities, high noise levels, and disparate people coming andgoing, i.e., general disorganization. As Wachs (2005) has made clear, the organization of the household is proximally linked withthe quality of the parenting children receive. A household setup wherein financial and social resources are limited very oftenengenders harsh parenting practices since young mothers may be ill equipped to handle the inexorable demands of childrearing(Scaramella and Conger, 2003). This affects children in a profound way. Older mothers, by contrast, tend to create the most opti-mal home environments for their children, environments particularly amenable to the acquisition of knowledge (Menaghan andParcel, 1991; Parcel and Menaghan, 1994).

In summary, children of mothers who began their childbearing early are at increased risk for certain adverse outcomesbecause their mothers are lacking in the social and human capital needed for success in a postindustrial market economy.Because of family circumstances deriving from their mothers’ early maternal age, which interrupts the expected order of thelife course and leads to educational underachievement, lower income, poor parenting practices, and suboptimal parent–childinteractions, the children of these women begin life with a great disadvantage compared to their peers born to mothers whodelayed childbearing until after the teen years.

Indeed, some of the earliest studies exploring the relationship between early maternal age and child outcomes supportthis directional association. Compared to their peers born to mothers who delay childbearing until after adolescence, chil-dren born to teen mothers are at increased risk for a whole host of poor outcomes (Corcoran, 1998; Hoffman andMaynard, 2008). Spieker et al. (1999) found that preschool and school-age children of teen mothers had a higher rate ofbehavior problems than children of non-teen mothers, while Furstenberg et al. (1987) found that adolescents had a higherrisk of school dropout and delinquency. There is also an inverse association between early maternal age and children’s per-formance on academic tests (Moore et al., 1997). Females born to teen mothers have an elevated likelihood for teen preg-nancy and childbearing, thus likely guaranteeing the transmission of disadvantage to future generations (Botting et al.,1998; McLanahan and Sandefur, 1994). Differences in outcomes by maternal age, across the broad spectrum of outcomemeasures, persist and even widen as children move from childhood into the teen years (Brooks-Gunn and Furstenberg,1986). Perhaps not surprisingly, the children of mothers who began childbearing in the teen years, who are themselves bornafter their mothers leave adolescence, also bear increased risk for suboptimal outcomes compared to later-born children ofmothers who begin childbearing after adolescence (Jutte et al., 2010).

2.2. Maternal background confounders of the early maternal age-child outcomes relationship

Empirical research by a number of scholars in the last half century points to marginalization and despair as largely pre-dictive of behaviors, including an early first birth, that are associated with subsequent poor outcomes and the intergenera-tional transmission of disadvantage (Edin and Kefalas, 2005; Kearney and Levine, 2011, 2012; Lewis, 1969; Wilson, 1987).Specifically, the social inequality present in US society is seen as responsible for creating the environmental conditions inpoor communities that lead young women to prefer the immediate gratification of having a baby to investing in their eco-nomic progress. Because they do not perceive any clear way of escaping the social milieu in which they have grown up, withits myriad problems and seeming obstacles to social advancement, young women drop out of the economic mainstream andengage in behaviors that provide them some immediate, if only brief, relief from those problems (Kearney and Levine, 2011).

An early maternal age under this framework, then, is primarily understood as an effect of socioeconomic disadvantagerather than its cause. For instance, while it might seem intuitive that early maternal age deriving from socioeconomic dis-advantage ought, in turn, lead to greater disadvantage in young mothers’ lives, research by Kearney and Levine (2012) hasactually shown that it is not itself an additional cause of the difficulties young mothers experience later in life. Net of back-ground factors, an early maternal age is not associated with greater economic disadvantage. The social disadvantage expe-rienced by young mothers is a part of the normal trajectory these women were already on prior to their becoming mothers.Therefore, inasmuch as an early maternal age merely reflects the decision of young girls to ‘‘drop out,’’ failing to account forthe background characteristics of young women that might select them into early childbearing, it is argued, necessarily leadsto an overestimation of the real effect of early maternal age on children’s outcomes (Bronars and Grogger, 1994).

A number of studies that have included explicit controls for selection factors deriving from young women’s backgroundshave found that early maternal age is inconsequential for children’s math and reading scores and their fertility timing(Barber, 2001; Levine et al., 2007, 2001). Even after controlling for many previously unobserved background characteristics,however, other studies have found that early maternal age was only reduced in its associational strength, and that it con-tinued to be an important risk factor for grade repetition, early initiation of sexual activity and childbearing, fighting, tru-ancy, male incarceration, educational attainment, economic outcomes, and many other outcomes (Goodman et al., 2004;Grogger, 1997; Haveman et al., 1997; Klepinger et al., 1999; Levine et al., 2007, 2001).

1 In a Bureau of Labor Statistics report by Grogger and Ronan (1995), fatherlessness was not as consequential for blacks after controlling for family-specificunobservables. In fact, blacks raised in single parent households actually acquired more education than they would have if both parents were present in thehome. Whites and Hispanics, on the other hand, suffered with respect to educational attainment when fathers were absent.


Explicitly controlling for maternal background factors alone, while it may reduce, or explain away entirely, any associa-tion between early maternal age and children’s outcomes, is still subject to bias deriving from the differential selection intoteen childbearing in the population (Harding, 2003). Given that early maternal age is non-randomly concentrated among thesubpopulation of blacks and non-white Hispanics (Hamilton et al., 2009), establishing causality between it and child out-comes requires finding a reliable comparison group of young women who are similar on preexisting measures such thatany differences in the outcomes between their children must be attributable to early maternal age (Harding, 2003). A fewprimary techniques have been employed to address issues of selection in the early maternal age-child outcomes relationship,including quasi-natural experiments, instrumental variables (IV) estimation, and sibling, or within-family, fixed effects mod-eling. Research that has employed the first two of these three methods typically focus on early maternal age’s associationwith women’s later socioeconomic outcomes rather than with the outcomes of their children.2 Since I am primarily con-cerned with early maternal age effects on child outcomes, I highlight sibling fixed effects models.

Sibling, or within-family, fixed effects models compare the long-term outcomes of first-born children of sisters who dif-fered in the timing of their initiation of childbearing. Again, the goal is to find a comparison group of women who are similarto early childbearing women on pre-early maternal age characteristics such that differences in subsequent outcomes formothers and their children may be attributed to early maternal age. The results, consistent across several studies, reveal thatthere is neither a disadvantage to being born to a mother whose first birth was in her teen years, nor does it negativelyimpact young women’s later socioeconomic well-being. Geronimus et al. (1994), comparing various outcomes among firstcousins, one set born to the sister who began childbearing early and the other set born to the sister who postponed child-bearing, found that there was no statistically significant negative effect of a teen first birth on verbal memory, behavior prob-lems, or reading recognition scores after controlling for young mothers’ family background and current socioeconomicstatus. Interestingly, an early maternal age was positively associated with children’s picture vocabulary, mathematics,and reading comprehension scores. Because Geronimus et al.’s (1994) data were drawn from what is now referred to asthe National Longitudinal Survey of Youth, 1979 – Children and Young Adults (NLSY79-CYA), which began administeringcognitive and behavioral assessments biennially in 1986, the sample size in 1994 was limited, and, thus, so were the gen-eralizability of the results. Taking advantage of four additional rounds of data, and thereby increasing the sample size,Turley’s (2003) replication of the Geronimus et al. (1994) study also confirmed (1) once family background factors were con-trolled, the cousins born to the older mothers did not perform significantly better on achievement measures than did theircousins born to younger mothers, and (2) even before controlling for family background factors, there was no difference inthe trajectory of the cousins’ scores. The takeaway is that the negative association observed between early maternal age andchild outcomes primarily reflects young mothers’ origins (i.e., family structure, family income, etc.) and not early childbear-ing per se, nor its accompanying indirect effects via the home environment, parenting skills, or income.

Problematic with cousin comparisons, or within-family fixed effects models, generally, is the fact that they often producelarge standard errors, possibly leading to a showing of no effect where an effect might exist. More important, though, is themuted assumption that the cousins’ mothers (i.e., sisters) are more alike than they are dissimilar and that they remain thusacross the life course. Choosing fixed effects models over explicit control models may alleviate concerns over selection bias,but such models pose their own shortcomings in that they can only control for those background traits sisters share; they failto capture the differential effects of traits that vary across time.

3. The current study

Given that early maternal age is related to some child outcomes but not others, even after either explicitly controlling orusing sophisticated analytic methods to account for those background factors that select young women into early childbear-ing, it is likely the case that it will remain an important area of research in the near term. Despite the number of existing

2 While much of the early maternal age literature employing quasi-natural experiments and the instrumental variables approach focus on later outcomes ofmothers who began their childbearing in the teen years, which is not the purpose of the present study, it is nonetheless instructive inasmuch as these lateroutcomes are associated with child outcomes. As the name suggests, quasi-natural experiments involve exploiting randomly occurring natural phenomenasuch as twin births and miscarriages to achieve the optimal comparison group for teen mothers. Twin studies, which compare the socioeconomic effects of anunplanned birth for mothers whose first birth was to a singleton and those whose first birth was to twins, suggest that an early maternal age has small, butstatistically significant, adverse effects on women’s later labor force participation, educational attainment, welfare recipiency, poverty status (Grogger andBronars, 1993), all of which are doubtless to be associated with poor outcomes for the children of teen mothers in a host of domains. Miscarriage studies, on theother hand, show that the early disadvantage to young women deriving from their early maternal age is reversed as they enter into young adulthood (Hotzet al., 1997, 2005). See Ashcraft et al. (2013), Fletcher and Wolfe (2009), Grogger and Bronars (1993), and Hoffman (1998) for limitations to using this approach.The purpose of the IV approach is to find a variable, or instrument, that, while it is correlated with teen motherhood, is not predictive of factors that teenmotherhood is believed to predict except insofar as it does so indirectly through teen motherhood. Instruments that have been used in the literature areabortion, age at menarche, and miscarriage, and the findings have been mixed. Using age at menarche as an instrument, Chevalier and Viitanen (2003) foundthat an early maternal age decreased additional educational attainment after age 16 among young women and led to both a reduction in their employmentexperiences and modest to moderately large differences in employment remuneration later in the life course. Miller (2009), instrumenting for whether awoman’s first pregnancy ended in a miscarriage, whether the conception of her first child occurred while taking contraceptives, and elapsed time from firstconception attempt to first birth, found that the failure to delay motherhood was associated with lower test scores on the Peabody Individual Achievement Test(PIAT). Others, though, have found any effect of teen motherhood on maternal household factors that might impinge on children’s outcomes to be nonexistentonce unobserved heterogeneity was taken into account (Goodman et al., 2004). See Ashcraft and Lang (2006), Harding (2003), and Imbens and Angrist (1994)for limitations to using the IV approach.


studies available, however, it is notable that few have truly considered the early maternal age-child outcomes relationship ina growth context, particularly as it relates to child academic outcomes. This is an important point because, whereas typicalstudies of change focus on aggregate-level, or population, growth only, individual growth models, an outgrowth of hierar-chical linear modeling (HLM), afford researchers the opportunity to study change over time for outcomes of interest atthe aggregate- and individual-levels (DeLucia and Pitts, 2006). This technique represents an advantage over methods likechange-score designs used in the past, foremost among them the ability to ascertain the extent to which theoretically mean-ingful variables, in this case those related to maternal background and maternal household factors, contribute to individualdeviations from the aggregate-level, or population, slope or curve. That is, it can distinguish between within- and between-individual variance in estimating outcomes of interest. Individual growth models also allow for the treatment of the timevariable as continuous rather than as incremental (Raudenbush and Bryk, 2002; Gibbons et al., 1993). Finally, because theyare so flexible, it is not necessary that all individuals be measured at each occasion in growth models.

Dahinten et al. (2007) provide the only study employing a growth model I have been able to find that included an aca-demic outcome. While their primary aim was to highlight how an early maternal age was related to Canadian adolescents’psychological and behavioral outcomes, they also considered its association with children’s math performance. Afteraccounting for maternal household factors such as income and maternal education, as well as controlling for maternaldepression and parenting behaviors, the authors found a statistically significant negative relationship between an early firstbirth—i.e., a first birth between the ages of 13 and 17—and children’s mathematics scores at age 10, the initial status in theirstudy. Early maternal age was not, however, related to the linear slope in mathematics scores. Whether additional controlsfor maternal background, or selection, factors might have changed the relationship between early maternal age and eitherinitial status or the linear slope remains unclear.

The current study extends the current literature by examining how, using individual growth curve modeling, children’smathematics, reading recognition, and reading comprehension scores are influenced by early maternal age prior to and aftercontrolling for both maternal background and maternal household factors. Adding specifically to Dahinten et al.’s (2007)contribution to the literature, in which the authors use Canadian data, I instead utilize United States data and focus and chil-dren’s academic growth curve trajectories from age 5, the time at which the majority of US children enter formal schooling,to age 14. Also, whereas Dahinten et al.’s approach only accounted for repeated measures of children, even though therewere multiple children in some families, I assume no degree of independence of scores of siblings reared by the samemother. By employing a three-level hierarchical model to account for non-independence both across time and within fam-ilies, I avoid bias in the estimate of the standard errors of my chosen predictors.

Given the results from most of the available research examining the relationship between early maternal age and chil-dren’s academic achievement, I propose the following hypotheses:

H1. Absent controls, early maternal age will be significantly related both to children’s initial and their time-dependentmathematics, reading recognition, and reading comprehension scores.

H2. Conditional on post-birth maternal household factors, early maternal age will continue to have a statistically significantassociation with children’s academic achievement at age 5 and across the elementary and secondary years.

That the literature remains mixed on whether there is a statistical significance relationship between early maternal ageand children’s academic achievement, even after taking into account unobserved heterogeneity in background characteris-tics in addition to post-birth maternal household factors, makes it difficult to hypothesize whether explicitly controlling forthese factors will prove the estimates deriving from models under the second hypothesis to not only be overestimated butspurious. Answering this question is one of the aims of the present study.

R1. Granting a statistically significant early maternal age association with children’s academic outcomes net of child-specificand maternal household factors, does it hold when explicit controls are included for maternal background factors?

4. Methods

4.1. Sample

To examine empirically the hypotheses, and to answer the research questions, put forward, I analyzed data from theNational Longitudinal Survey of Youth, 1979 – Children and Young Adults (NLSY79-CYA). The respondents in the NLSY79-CYA are the biological offspring of the female respondents of the National Longitudinal Survey of Youth, 1979 (NLSY79), anationally representative sample of 12,686 adolescents and young adults who were aged 14–22 at the NLSY79’s initiationin 1979. With the accumulation of time, and the attendant growth in family size among NLSY79 women, the weightedmother–child data, excluding women who were in the military or who were part of the economically disadvantaged whiteoversample (N = 1352 out of 6283 total NLSY79 women) are increasingly representative of a cross-section of women whowere between 14 and 22 in 1979. Because some of the respondents in the NLSY79-CYA are still being born toward the


end of the panel and are therefore not observed for their full childhood on the relevant measures, I excluded these individ-uals and focused only on those who were age 5–10 in the 1986 wave, and followed them through the 2000 wave. This con-straint on the data not only allowed for more sensible analysis of those respondents with more complete score trajectories,but it also proved useful in alleviating concerns regarding the imbalance between the number of children born to early andlate child-bearers present when all years are included. While there are achievement scores for 5-year olds born to post-ado-lescent mothers in 1990, for 6-year olds born to post-adolescent mothers in 1991, for 7-year olds born to post-adolescentmothers in 1992, and so on, there are none for 5-year olds born to adolescent mothers in 1990, for 6-year olds born to ado-lescent mothers in 1991, for 7-year olds born to adolescent mothers in 1992, and so on. By focusing on those with more com-plete score trajectories, I have removed a potential confound between period and the birth age of the mother. The final dataconsisted of 13,711 person-years representing 3174 children from 1899 families.

Biennially, from 1986 to 2006, NLSY79-CYA respondents were asked survey questions similar to those asked of theirmothers. While many questions touched on topics relating to employment, schooling, interactions with parents and peers,dating habits, and delinquent behavior, respondents were also administered, when they met the requisite developmentalstandards, several cognitive assessments. I focused on three of those assessments, brief descriptions of which follow.

4.2. Dependent measures

The outcome measures were raw scores on the mathematics, reading recognition, and reading comprehension portions ofthe Peabody Individual Achievement Test (PIAT), which has been administered to NLSY79-CYA children age 5–14 in theeven-numbered years from 1986 to the present. The PIAT mathematics test consists of 84 multiple-choice questions ofincreasing difficulty, which assess children’s ability to, at the most basic level, recognize numbers, and to, at a more advancedlevel, handle trigonometric concepts. The PIAT reading recognition test also consists of 84 multiple-choice questions ofincreasing difficulty and is designed to assess children’s ability to match letters, name names, and read single words aloud.Finally, the PIAT reading comprehension test consists of 66 sentence items of increasing difficulty in which a child choosesfrom an array of pictures the one that best represents a sentence’s meaning.3

4.3. Independent and control measures

Early maternal age, my primary predictor of interest in this study, has been treated differently in the existing literature,with some researchers treating all teen births prior to age 20 the same. Work by others, however, has attempted to distin-guish between early (i.e., prior to age 16, 17, or 18) and late (i.e., after age 16, 17, or 18) teen child-bearers on the assumptionthat there may be different early maternal age effects between these two teen groups (Coyne et al., 2013; Phipps and Sowers,2002). There are several factors that justify opting for the latter strategy over the former strategy. First, the extant literatureshows that a young teen first birth is associated with a greater likelihood of school non-completion than is an older teen firstbirth (Hoffman and Maynard, 2008). Second, as a consequence of the preceding point, the daughters of mothers whose teenfirst birth was at a younger rather than older age are also more likely to be teen mothers themselves, and all their childrenhave a greater likelihood of experiencing other social and economic disadvantage in their lives (Wildsmith et al., 2012).Finally, relative to their older peers, younger teen mothers are at greater risk of having a second teen birth and they tendto also experience more closely-spaced subsequent births, factors that put a strain on the allocation of the scarce resourcesamong their children (Manlove et al., 2000). Consistent with Dahinten et al. (2007), then, I created an ordinal variable todenote three categories for maternal age. A mother was coded 1 if she had a post-adolescent first birth (i.e., a first birthat age 20 or older), 2 if she had a first birth during adolescence at age 18 or 19, and 3 if she had an adolescent first birthduring adolescence prior to age 18. From this ordinal variable I then created dummy variables for which the first categorywas the reference category in my analyses.

4.3.1. Maternal background controlsThe maternal background factors included as controls were the maximum Duncan Socioeconomic Index value and the

highest grade of the young girls’ mother or father when the girls were 14, the natural log of net family income in the yearprior to NLSY79 study inception, race (dichotomized to distinguish between Black and Hispanic females [coded 1] and allothers, who were predominantly white [coded 0]), the number of siblings in the household at NLSY79 study inception,whether the family was headed by both biological parents at NLSY79 study inception, and urban versus rural residence atNLSY79 study inception.

Kessler et al. (1997) found that young women who enter motherhood prematurely tended to have a long history of con-duct disorder, including aggression, and Woodward and Fergusson (1999) have highlighted the fact that a history of aggres-sion is proximally related to an increased risk for early childbearing. If teen pregnancy and early maternal age are simplyillustrative examples of an entire universe of problem behaviors in which some young women engage, from premaritalsex to alliances with antisocial peer groups to illicit drug use (Miller-Johnson et al., 1999), their effects on children’s

3 The reading comprehension test was only administered to those who scored at least a 19 on the reading recognition portion of the PIAT; those children whoscored less than 19 on the reading recognition test had that score recorded as their reading comprehension score.


outcomes should, to some extent, be accounted for by such behavior. Therefore, in considering the degree to which earlymaternal age effects on children’s mathematics, reading recognition, and reading comprehension scores are explained bymaternal background factors, I also focused on those factors that denoted women’s aggressive and antisocial behaviors priorto beginning childbearing.

In 1980, the NLSY79 gathered information from respondents on their delinquent behaviors in the previous year. Origi-nally entered on an ordinal scale from 0 (NEVER) to 6 (MORE THAN 50 TIMES), the relevant items asked, for example, thenumber of times in the past year a respondent (1) ‘‘purposely damaged property that did not belong to you?’’, (2) ‘‘gotteninto a physical fight at school or work?’’ or (3) ‘‘used force or strong arm methods to get money or things from a person?’’For purposes of the current study, I constructed from 18 of 21 items a scale of delinquent behavior at NLSY79 study incep-tion.4 Each item was standardized, and then all of the z-transformed items were summed before being standardized again. Thescale reliability coefficient, or Cronbach’s alpha, was about .78.

Finally, I also entered as a maternal background control young women’s scores on the Armed Forces Qualification Test(AFQT)—a test of cognitive ability or aptitude that is drawn from the arithmetic reasoning, word knowledge, paragraph com-prehension, and numerical operations portions of the Armed Services Vocational Aptitude Battery (ASVAB; National Longi-tudinal Surveys, 1999), and which is highly correlated with formal tests yielding an intelligence quotient (IQ).

4.3.2. Child-specific and maternal household controlsThe child-specific variables entered as controls were sex (coded 0 for females, 1 for males), whether the biological father

resided in the home at the first measurement occasion, region of the country at first measurement occasion, weeks in ges-tation, and birth weight, measured in ounces. For each of the three data sets, these latter two variables were centered at thegrand mean.

Sullivan et al. (2011), using a cross-section of the Panel Study of Income Dynamics (PSID) data, have shown that motherswhose educations were interrupted due to an adolescent first birth, but who later resumed their studies, provided betterhome environments for their children than their peers whose educations were also interrupted due to an adolescent firstbirth but who did not resume their studies. In addition to increased education having a direct positive effect on the homeenvironment, it is also indirectly associated with improvements to the home environment via increased family income.And all of these are positively associated with children’s achievement. I also included as controls, therefore, the grandmean-centered time-varying values for the children’s mothers’ highest grade completed and the natural log of net familyincome.

Since the NLSY79-CYA includes multiple children in the same family, and because the social influence hypothesis sug-gests treating the children born to women in post-adolescence differently than their siblings born when the women werestill in their teen years (Jaffee et al., 2001), I also included an additional control denoting whether a specific child was bornduring her mother’s adolescence (coded 0) or thereafter (coded 1). This unique control allows one to explicate the degree towhich early maternal age is differentially associated with children’s achievement by whether they were born during theirmothers’ adolescent years or during her post-adolescent years.

4.4. Data analysis plan

Missing values of the independent variables were handled via multivariate imputation by fully conditional specification(Lee and Carlin, 2010; Raghunathan et al., 2001) using the -mi impute- command in Stata 13. In accordance with Schafer andOlsen (1998) and Rubin (1987), who suggest no more than 3–5 imputations as the gains to efficiency tend diminish after thispoint, I created 3 imputed data sets and then made use of Stata’s -mixed- command to carry out a growth curve analyses onchildren’s PIAT mathematics, reading recognition, and reading comprehension performance.

The growth curve models for the three outcomes were specified such that, at level-one, individual growth in raw scoreswas related to the linear and quadratic functional forms of age. In equation form, this was quantified as:

4 Thedrank areliabili

Ytij ¼ p0ij þ p1ijðatij � LÞ þ p2ijðatij � LÞ2 þ etij; ð1Þ

where the centering constant, L, was set at chronological age 5, the age at which most children in the US enter formal school-ing, the moment when educational gaps resulting from disadvantage prior to compulsory schooling begin to widen, and theearliest age at which the relevant PIAT measures are administered. The parameter denoting the intercept, p0ij, represents theinitial test score of child i in family j at age 5. The linear parameter, p1ij, is the instantaneous growth rate for child i in family jat age 5, and the nonlinear component, p2ij, represents the mean acceleration or deceleration growth rate over time for child iin family j. Finally, the random error term, etij, is the unique effect of child i in family j at age t on achieved raw score. Thisrandom effect is believed to have a multivariate normal distribution with a mean vector of 0 and a diagonal covariancematrix with variance r2.

excluded delinquency items were (1) number of times ran away in past year, (2) number of times skipped school in past year, and (3) number of timeslcohol in past year. Their exclusion from the bevy of other delinquency variables available, which often focused on much riskier behavior, led to a higherty estimate of the created measure; hence, their absence from final analyses.


The level-2 model, which treated the level-1 covariates denoting the intercept and the linear and quadratic terms for ageas outcomes, was be specified as:

5 Whevaluatfixed-q

ppij ¼ bp0j þXQp

q¼1

bpqjWqij þ rpij; ð2Þ

where bp0j is the child-level intercept, Wqij is the vector of q = 1,2,3, . . . ,Qp level-2 predictors of child effect ppij, bpqj is thecorresponding coefficient for each level-2 predictor, and rpij is the child-level random effect.

Finally, to predict the effect of family- or mother-level characteristics on children’s mathematics and reading achieve-ment, particularly the effect of being born to a mother who began childbearing at (1) age 18 or 19 versus age 20 or olderor (2) age 17 or younger versus age 20 or older, I quantified the level-3 equation as:

bpqj ¼ bpq0 þXSpq

s¼1

cpqsZsk þ upqj; ð3Þ

where bpq0 is the mother-level intercept, Zsk is the vector of s = 1,2,3, . . . ,Spq predictors of the mother effect, bpqj, cpqs is thecorresponding coefficient for each level-3 predictor, and upqj is the level-3 random effect.5

For each PIAT outcome, I estimated four models, the first of which was an unconditional model. The second model esti-mated the degree to which early maternal age conditioned the intercept, the slope of age, and the slope of the quadraticterm, age-squared. That is, the intercept and slopes denoted by the coefficients in equation 1 were treated as outcomes ofthe level-three dummy predictors, early maternal age at 18 or 19 (a dummy variable denoting whether a woman had herfirst child after the teen years [coded 0] or when she was 18 or 19 years old [coded 1]) and early maternal age at age 17or younger (a dummy variable denoting whether a woman had her first child after the teen years [coded 0] or when shewas 17 years old or younger [coded 1]). The third and fourth analytic model examined the effects of early maternal ageon the intercept, the instantaneous growth rate, and acceleration or deceleration in the growth trajectory, controlling for,first, the child-specific and maternal household factors (Model 3), and second, the family-specific maternal background fac-tors (Model 4).

Because they are likely associated with or influenced by early maternal age, including the time-invariant factor denotingwhether a child’s father lived in the home at her birth, or the time-varying factors denoting family income and mother’shighest grade completed, could lead to an underestimation of the early maternal age effect. Too allay concerns of over-controlling for the early maternal age effect I carried out robustness checks that excluded these variables. The resultant mod-els, which did not differ from the results presented here, are included in an Appendix. All of the models included randomintercepts at all levels and random slopes in age at level two. Full maximum likelihood estimation was used for all models.

5. Results

5.1. Descriptive statistics

A comparison of the means and standard deviations of the maternal background and household factors by early maternalage status is shown for the pooled data in Table 1. Women who began childbearing in their teen years came from back-grounds that were, on average, more disadvantaged than the backgrounds of their peers who postponed childbearing untiladulthood, and they also headed households that tended to perpetuate this disadvantage to the next generation. Comparedto the parents of adult child-bearers, the parents of early child-bearers had lower levels of educational attainment, worked incareers that were less prestigious, and earned between 26 (in the case of those whose first birth was at age 18 or 19) and 46(in the case of those whose first birth was at age 17 or younger) percent less in income. An early first birth was also asso-ciated with larger sib-ship size, greater likelihood of broken home, and depressed cognitive ability. For all of the foregoingfactors, the differences were largest between post-adolescent child-bearers and those who first gave birth at age 17 oryounger. The differences in family background by maternal age status were consistent across the generations as earlychild-bearers in the age 18 or 19 category and the age 17 or younger category, not unlike their parents, had, on average,lower levels of educational attainment and earned, respectively, 26 and 39 percent less income than did their postponingpeers. Again, the largest differences in means across the various child-specific and maternal household factors were evidentbetween post-adolescent first time child-bearers and their pre-18 year old teen first time child-bearer counterparts. Not sur-prising is the fact that minorities such as blacks and non-white Hispanics were more likely than whites to have an early firstbirth. Also, relative to the other maternal age groups, pre-18 teen first births tended to be concentrated at a higher level inthe south than in other regions of the country.

Using ANOVA to test the equality of unadjusted means for children’s PIAT mathematics, reading recognition, and readingcomprehension raw scores by early maternal age status, Table 2 shows that, compared to children whose mothers delayed

ile diagnostics of various models suggested that growth in mathematics and reading achievement be specified as quadratic, reliability estimates andions of model fit called for estimating the variance for the linear parameter only and not for the quadratic parameter. The resulting model, then, is auadratic model.

Table 1Family descriptive statistics by early maternal age status with one-way ANOVA comparing means.

Variables Post-teen First Birth First birth at age 18 or 19 First birth at age 17 or younger F-sig

n Mean SD n Mean SD n Mean SD

Maternal family backgroundDuncan’s Socioeconomic Index 792 35.1 22.6 352 29.1 19.2 339 24.0 17.4 36.82***

Parent’s Highest Grade Completed 936 11.2 3.4 445 10.5 3.0 436 10.0 3.1 25.67***

Family Income (log 1979 $) 742 9.3 1.1 349 9.0 1.0 351 8.7 1.5 34.07***

Black or Hispanic 961 0.5 0.5 473 0.6 0.5 465 0.7 0.4 39.34***

Number of Siblings 960 4.2 2.8 472 4.6 2.9 463 5.0 3.1 10.46***

Delinquency 961 5.5 6.6 473 5.5 6.4 465 5.1 6.3 0.62Both Biological Parents 960 0.7 0.5 472 0.5 0.5 463 0.5 0.5 30.97***

Urban Residence 956 0.8 0.4 472 0.8 0.4 462 0.8 0.4 0.04AFQT Percentile Score 922 38.0 27.2 458 26.8 22.5 444 20.3 19.0 88.33***

Child-specific and maternal householdMale 1451 0.5 0.5 828 0.5 0.5 895 0.5 0.5 1.08Gestation Age 1401 38.8 2.2 807 38.8 2.1 858 38.6 2.4 3.07*

Birth Weight (in ounces) 1425 117.3 21.0 820 115.8 19.9 881 111.4 22.2 22.04***

Net Family Income (log $) 1255 9.8 1.5 726 9.5 1.2 746 9.3 1.3 34.15***

Mother’s Highest Grade 1448 12.6 2.1 824 11.4 1.6 890 10.5 3.5 190.95***

Father Lives in the Home 1426 0.6 0.5 799 0.5 0.5 862 0.3 0.5 85.26***

RegionNortheast 1442 0.2 0.4 825 0.1 0.3 890 0.1 0.3 6.57**

Midwest 1442 0.2 0.4 825 0.2 0.4 890 0.2 0.4 3.42*

South 1442 0.4 0.5 825 0.4 0.5 890 0.5 0.5 18.95***

West 1442 0.2 0.4 825 0.2 0.4 890 0.2 0.4 6.19**

Note: While the child-specific and maternal household variables denoting net family income and mother’s highest grade vary across time, only the firstreported values for each child are included in the calculation of the means and standard deviations.

* p < .05.** p < .01.

*** p < .001.


childbearing until after the teen years, children born to teen mothers in both of the early maternal age categories faredpoorer throughout childhood and adolescence. Depending on the specific outcome under scrutiny, and particularly intothe middle of the teen years, those with a mother who began childbearing early were one (in the case of those whose motherhad a first birth at age 18 or 19) to two (in the case of those whose mother had a first birth at age 17 or younger) years behindin ability relative to their peers of older mothers. Indeed, these differences were already present and large at the initial age inthe study and are greatest between children of post-adolescent first time child-bearers and children of pre-18 year old teenfirst time child-bearers.

5.2. Mathematics growth

Table 3 shows the results of the fixed quadratic growth curve models for the PIAT mathematics data. In agreement withresearch produced by Dahinten et al. (2007), the random effects parameters in the unconditional model (Model 1) showedthat there was significant variation in individual raw mathematics scores at age 5 and across age, allowing for nonlinearity.The mean mathematics achievement at age 5 was 7.269 points, with yearly gains of about 9 points accompanied by slightdeceleration in gains over time. Conditional on early maternal age, Model 2 revealed a mean initial mathematics score of7.946, also with decelerating gains of about 9 raw points per year thereafter. In partial confirmation of my hypothesis ofa negative association between early maternal age and all level-1 age factors in the absence of either child-specific andmaternal household or maternal background controls, Model 2 also revealed that a pre-18 year old teen first birth wasrelated both to children’s initial mathematics scores (c002 = �1.929, se = 0.485, p < .001) and to their linear growth trajecto-ries (c102 = �0.572, se = 0.177, p < .01). There was no association, however, between an 18 or 19 year old first birth and chil-dren’s mean mathematics achievement at age 5 (c001 = �0.780, se = 0.466, p > .05) or their mean growth trajectories(c101 = �0.193, se = 0.179, p > .05).

The inclusion of the child-specific and maternal household factors in Model 3 reduced the large statistically significantnegative association between early maternal age and mean initial mathematics score by greater than .75 points to a smallerthough still statistically significant �1.135 (se = 0.541, p < .05) for children born to mothers who began childbearing prior toage 18. A pre-18 year old first birth continued to be related to the linear growth in mathematics achievement as well(c102 = �0.558, se = 0.217, p < .05). Children born to mothers who began childbearing at age 18 or 19 were not at a disadvan-tage relative to their peers born to postponing mothers in either their mean mathematics achievement assessed at age 5 orany point thereafter, net of controls. Again, this lent partial confirmation to my second hypothesis, which argued that therewould be a statistically significant inverse relationship between early maternal age and children’s mean initial and time-dependent mathematics achievement, controlling only for child-specific and maternal household factors.

Table 2Mean PIAT mathematics, reading recognition, and reading comprehension raw scores by child age by early maternal age status (with one-way ANOVAcomparing means).

Age Post-teen First Birth First birth at age 18 or 19 First birth at age 17 or younger F-sig

n Mean n Mean n Mean

Mathematics data5 338 12.2 155 10.9 105 9.5 16.90***

6 606 15.2 285 14.6 249 12.9 15.77***

7 663 23.5 318 21.1 297 19.1 32.27***

8 672 30.1 337 28.9 335 27.4 9.87***

9 700 38.3 391 36.1 365 34.1 23.81***

10 678 43.3 371 42.5 407 39.3 24.02***

11 719 48.4 402 45.8 400 43.1 41.00***

12 612 50.8 343 49.4 372 46.6 23.64***

13 767 53.7 448 51.8 466 48.1 42.74***

14 677 54.9 360 52.6 412 50.7 19.84***

Reading recognition data5 320 13.0 151 11.6 103 10.4 11.43***

6 583 16.9 274 16.4 248 15.1 11.20***

7 656 25.4 318 23.6 296 22.3 16.42***

8 672 33.1 341 31.9 334 30.6 6.80**

9 694 41.3 390 38.6 366 35.7 33.62***

10 676 46.0 367 44.4 406 42.2 13.45***

11 718 51.6 396 48.6 398 44.4 42.05***

12 609 55.3 345 53.0 369 50.1 17.53***

13 766 59.5 450 57.2 459 51.8 45.56***

14 673 61.5 363 59.0 413 55.9 21.76***

Reading comprehension data5 313 12.8 150 11.5 101 10.3 10.89***

6 547 16.5 261 15.9 237 14.9 8.39***

7 604 24.1 283 22.5 269 21.5 11.69***

8 643 31.0 319 29.7 312 28.7 6.57**

9 687 38.5 381 35.9 350 33.2 33.57***

10 670 42.4 362 40.9 397 39.2 11.68***

11 710 46.5 394 44.3 400 40.7 36.92***

12 602 49.3 343 47.3 364 46.0 11.16***

13 762 52.3 446 49.9 459 46.3 37.75***

14 672 53.8 360 51.2 404 49.4 17.81***

⁄ p < .05.** p < .01.

*** p < .001.


Model 4 included the maternal background factors with the aim of answering whether the statistically significant nega-tive association of early maternal age and mean mathematics achievement under Model 3 held. Neither early maternal agecategory was significantly associated with adverse mathematics achievement, either at school entry or across time. Moreconsequential to children’s mean initial mathematics score and score trajectories than their mothers having started child-bearing during her adolescent years was their mothers’ cognitive ability, urban residence, family income, and educationalattainment. These findings were robust to the exclusion of the time-invariant factor denoting whether a child’s father livedin the home at the child’s birth and the time-varying factors denoting the family income of the maternal household andmother’s completed education, factors that are associated with or likely negatively impacted by an early first birth (seeTable A1).

5.3. Reading recognition growth

As was the case with the mathematics data, the random effects parameters under Model 1 of Table 4 revealed that therewas considerable variation in raw PIAT reading recognition scores at age 5 and throughout the primary and secondary years.The fixed effects showed a mean initial reading recognition achievement score of 9.264 (se = 0.194) and a yearly learning rateof about 9 points at age 6, with slight deceleration in gains thereafter. Controlling for early maternal age effects only, Model 2of Table 4 revealed a mean initial reading recognition score of 9.619 (se = 0.268). Again, mean yearly gains reading recogni-tion achievement was on the order of about 9 points, holding constant the deceleration in gains from age 6 on. Model 2 onlyprovided partial confirmation of an early maternal age effect on children’s mean achievement as the association was limitedto the linear trajectory in mean achievement for children born to mothers in both the 18 and 19 year old and pre-18 year oldteen first birth categories (c101 = �0.513, se = 0.184, p < .01 and c102 = �0.959, se = 0.183, p < .001, respectively) and to thenonlinear trajectory for children born only to mothers in the pre-18 year old teen first birth category (c202 = 0.035,se = 0.017, p < .05). Early maternal age was not significantly related to children’s mean initial achievement.

Table 3Fixed quadratic growth curve model for PIAT mathematics data.

Model 1 Model 2 Model 3 Model 4

Fixed effectsInitial Status at age 5, p0ij

Intercept, b00j 7.269*** (0.194) 7.946*** (0.267) 8.583*** (0.752) 9.127*** (0.951)Age at first birth 18/19, c001 �0.780 (0.466) �0.340 (0.510) �0.070 (0.499)Age at first birth < 18, c002 �1.929*** (0.485) �1.135* (0.541) �0.677 (0.541)

Maternal family backgroundFamily Income (log 1979 $), c003 0.040 (0.199)Duncan’s SEI, c004 �0.110 (0.219)Parent Highest Grade, c005 0.198 (0.222)Both Biological Parents, c006 0.466 (0.394)Black or Hispanic, c007 �0.730 (0.462)Number of Siblings, c008 0.022 (0.067)Delinquency, c009 0.001 (0.020)AFQT Percentile Score, c0010 0.038*** (0.009)Urban Residence, c0011 �0.787 (0.452)

Child-specific and maternal household factorsMale, b01j �0.626 (0.345) �0.649 (0.341)Gestation Age, b02j 0.097 (0.090) 0.097 (0.089)Birth Weight (in ounces), b03j 0.009 (0.010) 0.006 (0.009)Post-Adolescent Birth, b04j �0.120 (0.497) �0.194 (0.493)Father Lives in Household, b05j �0.014 (0.359) 0.126 (0.368)Region (Ref. = Northeast)

Midwest, b06j �0.209 (0.622) �0.565 (0.609)South, b07j �0.932 (0.568) �0.813 (0.559)West, b08j �0.248 (0.628) �0.334 (0.611)

Linear slope (Age), p1ij

Intercept, b10j 8.826*** (0.073) 9.024*** (0.107) 9.300*** (0.306) 8.993*** (0.388)Age at first birth 18/19, c101 �0.193 (0.179) �0.232 (0.206) �0.026 (0.208)Age at first birth < 18, c102 �0.572** (0.177) �0.558* (0.217) �0.181 (0.223)

Maternal family backgroundFamily Income (log 1979 $), c103 0.119 (0.080)Duncan’s SEI, c104 0.201* (0.088)Parent Highest Grade, c105 �0.084 (0.088)Both Biological Parents, c106 �0.235 (0.157)Black or Hispanic, c107 �0.184 (0.185)Number of Siblings, c108 �0.032 (0.026)Delinquency, c109 �0.001 (0.008)AFQT Percentile Score, c1010 0.018*** (0.004)Urban Residence, c1011 0.558** (0.180)

Child-specific and maternal household factorsMale, b11j 0.063 (0.146) 0.088 (0.145)Gestation Age, b12j �0.014 (0.037) �0.004 (0.037)Birth Weight (in ounces), b13j 0.008* (0.004) 0.005 (0.004)Post-Adolescent Birth, b14j �0.159 (0.207) �0.141 (0.206)Father Lives in Household, b15j �0.070 (0.153) �0.267 (0.159)Region (Ref. = northeast)

Midwest, b16j �0.105 (0.245) �0.106 (0.247)South, b17j �0.392 (0.223) �0.069 (0.226)West, b18j �0.031 (0.248) 0.117 (0.248)

Quadratic slope (Age2), p2ij

Intercept, b20j �0.404*** (0.007) �0.413*** (0.011) �0.456*** (0.029) �0.417*** (0.037)Age at first birth 18/19, c201 0.007 (0.017) 0.011 (0.020) 0.000 (0.021)Age at first birth < 18, c202 0.030 (0.017) 0.032 (0.021) 0.012 (0.022)

Maternal family backgroundFamily Income (log 1979 $), c203 �0.009 (0.008)Duncan’s SEI, c204 �0.016 (0.008)Parent Highest Grade, c205 0.008 (0.008)Both Biological Parents, c206 0.017 (0.015)Black or Hispanic, c207 0.002 (0.018)Number of Siblings, c208 0.003 (0.003)Delinquency, c209 �0.000 (0.001)AFQT Percentile Score, c2010 �0.001* (0.000)Urban Residence, c2011 �0.056** (0.017)

(continued on next page)


Table 3 (continued)


Child-specific and maternal household factorsMale, b21j 0.007 (0.014) 0.005 (0.014)Gestation Age, b22j 0.002 (0.003) 0.001 (0.003)Birth Weight (in ounces), b23j �0.000 (0.000) �0.000 (0.000)Post-Adolescent Birth, b24j 0.012 (0.020) 0.010 (0.020)Father Lives in Household, b25j 0.023 (0.015) 0.033* (0.015)Region (Ref. = Northeast)

Midwest, b26j 0.031 (0.024) 0.026 (0.024)South, b27j 0.041 (0.022) 0.018 (0.022)West, b28j �0.004 (0.024) �0.014 (0.024)

Family Income (log $), b30j 0.324*** (0.068) 0.199** (0.068)Mother’s Highest Grade, b40j 0.914*** (0.107) 0.396*** (0.110)

Random effectsLevel 1

Temporal Variation, etij 34.078 34.112 34.125 34.127Level 2 (children within families)

Individual Initial Achievement, r0ij 2.290*** (0.492) 2.418*** (0.508) 2.628*** (0.527) 2.895*** (0.553)Individual Learning Rate, r1ij 0.532*** (0.035) 0.515*** (0.035) 0.483*** (0.034) 0.395*** (0.031)

Level 3 (between families)Family Mean Achievement, u00j 20.058*** (1.357) 18.132*** (1.303) 15.091*** (1.212) 10.922*** (1.021)

AIC 92954.326 92850.227 92676.187 92218.154BIC 93014.454 92955.451 92976.830 92721.730

Note: The total number of level-1, level-2, and level-3 units was, respectively, 13,576, 3151, and 1895. Standard errors in parentheses.* p < 0.05.

** p < 0.01.*** p < 0.001.


Model 3 of Table 4, which included the child-specific and maternal household controls, revealed a mean reading recog-nition score at age 5 of 11.662 points (se = 0.766). Accounting for a slowing of gains over time, the yearly learning rate was9.808 points (se = 0.316). While neither of the early maternal age categories was significantly related to children’s mean ini-tial mean reading recognition achievement or their mean nonlinear trajectory, they were negatively related to children’smean linear score trajectory (c101 = �0.571, se = 0.211, p < .01 and c102 = �0.992, se = 0.223, p < .001, respectively), providingpartial confirmation to my second hypothesis. Both the coefficients and their statistical significance were slightly larger inModel 3 than in Model 2, but there was no substantive difference in the interpretation.

Turning to Model 4 of Table 4, which built on Model 3 by adding the maternal background controls, the mean initial read-ing recognition achievement was 10.472 points (se = 0.975). Children’s mean annual gains in reading recognition knowledgewere just less than 10 points, accounting for nonlinearity in children’s score trajectories. The early maternal age effect wasreduced to statistical nonsignificance. The implication of this was that the significant negative relationship found underModel 3, which only accounted for child-specific and maternal household factors, did not hold once explicit controls forselection factors were included. Having a mother who began childbearing either at 18 or 19 years of age or prior to age18 was not as consequential to children’s mean initial reading recognition achievement or their mean reading recognitiontrajectory as was their mothers’ cognitive ability, whether their mothers came from an intact family (i.e., a family headedby both biological parents), their grandparents’ and mothers’ income, and their mothers’ educational attainment. Thiswas not the case for all children within the same family whose mother began childbearing at an early age. Holding constantwhether a woman had an early maternal age, her children born to her in post-adolescence had significantly lower mean ini-tial achievement than a sibling born during her adolescent, or teen, years. Consistent with the results of the mathematicsdata, the findings of no effect of an early maternal age in Model 4 of Table 4 was robust to the exclusion of whether a child’sfather resided in the home at her birth, the family income of the maternal household, and mother’s completed education.

5.4. Reading comprehension growth

Allowing for curvature in the trajectory of scores, the random effects under Model 1 of Table 5 suggested a good deal ofvariation in children’s PIAT reading comprehension scores at initial status and over time. The fixed effects showed a meaninitial reading comprehension score of 9.741 points (se = 0.206), with gains of about 8 points made each year thereafter,holding constant a slight deceleration in annual gains. Accounting for the effects of early maternal age in Model 2 did notchange these values very much. The mean initial reading comprehension score was 10.056 (se = 0.285), while yearly gainswere 8.265 (se = 0.119). Consistent with the results of the PIAT reading recognition data, and in partial confirmation of myfirst hypothesis, the effect of having a mom who either began childbearing at 18 or 19 or prior to age 18 was significantlyrelated to the linear slope of age, but not to the mean initial reading comprehension achievement. That is, absent controls,while the mean initial reading comprehension score of children of early child-bearers was not significantly lower than themean achievement of their counterparts born to postponing mothers, it was lower as a consequence of age, with these

Table 4Fixed quadratic growth curve model for PIAT reading recognition data.



Intercept, b00j 9.264*** (0.194) 9.619*** (0.268) 11.662*** (0.766) 10.472*** (0.975)Age at first birth 18/19, c001 �0.370 (0.467) �0.447 (0.514) �0.450 (0.509)Age at first birth < 18, c002 �0.914 (0.486) �0.689 (0.547) �0.594 (0.553)

Maternal family backgroundFamily Income (log 1979 $), c003 0.122 (0.205)Duncan’s SEI, c004 0.369 (0.223)Parent Highest Grade, c005 0.316 (0.226)Both Biological Parents, c006 0.107 (0.404)Black or Hispanic, c007 1.107* (0.471)Number of Siblings, c008 0.069 (0.069)Delinquency, c009 �0.020 (0.021)AFQT Percentile Score, c0010 0.027** (0.010)Urban Residence, c0011 �0.042 (0.463)

Child-specific and maternal household factorsMale, b01j �0.309 (0.354) �0.382 (0.352)Gestation Age, b02j 0.138 (0.092) 0.151 (0.091)Birth Weight (in ounces), b03j 0.007 (0.010) 0.008 (0.010)Post-Adolescent Birth, b04j �1.083* (0.507) �1.260* (0.505)Father Lives in Household, b05j �0.096 (0.366) 0.298 (0.377)Region (Ref. = Northeast)

Midwest, b06j �1.311* (0.629) �1.266* (0.625)South, b07j �1.042 (0.575) �0.844 (0.574)West, b08j �1.356* (0.634) �1.307* (0.626)


Intercept, b10j 8.684*** (0.075) 9.065*** (0.110) 9.808*** (0.316) 9.757*** (0.399)Age at first birth 18/19, c101 �0.513** (0.184) �0.571** (0.211) �0.217 (0.212)Age at first birth < 18, c102 �0.959*** (0.183) �0.992*** (0.223) �0.405 (0.228)

Maternal family backgroundFamily Income (log 1979 $), c103 0.188* (0.083)Duncan’s SEI, c104 �0.127 (0.090)Parent Highest Grade, c105 �0.004 (0.090)Both Biological Parents, c106 �0.055 (0.162)Black or Hispanic, c107 �0.409* (0.190)Number of Siblings, c108 �0.026 (0.027)Delinquency, c109 0.013 (0.008)AFQT Percentile Score, c1010 0.031*** (0.004)Urban Residence, c1011 0.154 (0.186)Child-Specific and

Maternal household factorsMale, b11j �0.668*** (0.150) �0.630*** (0.149)Gestation Age, b12j 0.003 (0.038) 0.012 (0.038)Birth Weight (in ounces), b13j 0.007 (0.004) 0.002 (0.004)Post-Adolescent Birth, b14j �0.176 (0.212) �0.116 (0.211)Father Lives in Household, b15j 0.074 (0.154) �0.174 (0.160)Region (Ref. = Northeast)

Midwest, b16j �0.096 (0.253) �0.249 (0.255)South, b17j �0.554* (0.231) �0.239 (0.233)West, b18j �0.452 (0.256) �0.312 (0.255)


Intercept, b20j �0.341*** (0.007) �0.356*** (0.011) �0.421*** (0.029) �0.400*** (0.037)Age at first birth 18/19, c201 0.030 (0.017) 0.035 (0.020) 0.014 (0.020)Age at first birth < 18, c202 0.035* (0.017) 0.039 (0.021) 0.008 (0.022)

Maternal family backgroundFamily Income (log 1979 $), c203 �0.017* (0.008)Duncan’s SEI, c204 0.018* (0.008)Parent Highest Grade, c205 �0.006 (0.008)Both Biological Parents, c206 �0.001 (0.015)Black or Hispanic, c207 0.029 (0.018)Number of Siblings, c208 �0.001 (0.003)Delinquency, c209 �0.002* (0.001)AFQT Percentile Score, c2010 �0.002*** (0.000)Urban Residence, c2011 �0.015 (0.017)



Table 4 (continued)


Child-specific and maternal household factorsMale, b21j 0.046** (0.014) 0.043** (0.014)Gestation Age, b22j �0.002 (0.003) �0.002 (0.003)Birth Weight (in ounces), b23j �0.000 (0.000) 0.000 (0.000)Post-Adolescent Birth, b24j 0.007 (0.019) �0.000 (0.019)Father Lives in Household, b25j 0.014 (0.015) 0.029 (0.015)Region (Ref. = Northeast)

Midwest, b26j 0.021 (0.024) 0.032 (0.024)South, b27j 0.047* (0.022) 0.028 (0.022)West, b28j 0.044 (0.024) 0.037 (0.024)

Family Income (log $), b30j 0.260*** (0.069) 0.166* (0.069)Mother’s Highest Grade, b40j 0.809*** (0.111) 0.308** (0.115)





AIC 93795.813 93683.123 93504.981 93093.211BIC 93855.875 93788.232 93805.292 93596.232


** p < 0.01.*** p < 0.001.


children growing at a slower rate (c101 = �0.431 points [se = 0.197, p < .05] slower growth for children of mothers who beganchildbearing at age 18 or 19 and c102 = �0.755 points [se = 0.196, p < .001] slower growth for children born to mothers whobegan childbearing prior to age 18).

Model 3 of Table 5 built on Model 2 by further conditioning on the child-specific and maternal household covariates. Themean initial achievement on the reading comprehension measure was about 13 points, with per year gains of about 8.5points. This was net of a slight deceleration in gains over time. Early maternal age was not related to reading comprehensionachievement either at age 5 or across age among children born to the mothers whose childbearing years began at age 18 or19. Although there was also no association between children’s mean initial reading comprehension achievement and beingborn to a mother who began childbearing prior to age 18, there was an association between children’s mean trajectory inreading comprehension achievement and their mothers’ early maternal age, which partially confirmed my second hypoth-esis. Relative to the mean annual growth made by children of mothers who postponed childbearing until adulthood, themean annual growth made by children of mothers who began childbearing prior to age 18 was c102 = �0.755 points(se = 0.196, p < .001) lower.

In addition to the child-specific and maternal household factors, Model 4 of Table 5 included the maternal backgroundfactors as controls. The mean initial reading comprehension score was 11.454 (se = 1.029). Accounting for slight decelerationin gains over time, the mean years learning rate was 8.618 (se = 0.429). In response to the research question of whether thenegative association found to exist between early maternal age and reading comprehension under Model 3 held once con-trols for maternal background were accounted for, Model 4 revealed that neither early maternal age category was signifi-cantly related to children’s mean reading comprehension achievement, either at age 5 or across the years of formalschooling. Consistent with the other two academic outcomes studied here, mother’s cognitive ability and the family incomeof the maternal household were more important to children’s age-5 reading comprehension score and score trajectory thanthe age at which their mothers began childbearing. Also more important was the grandparents’ highest grade completed,children’s sex, parents’ number of siblings, region, and whether a specific child’s birth was after the mother’s adolescentyears. The results shown in Model 4 of Table 5 were also robust to the exclusion of post-first birth maternal household fac-tors possibly negatively associated with an early maternal age, i.e., whether father resided in the home at the child’s birth,the family income of the maternal household, and mother’s completed education.

6. Summary and discussion

There continues to be an intense debate surrounding the issue of whether there is a significant negative effect of an earlyfirst birth on various children’s outcomes (Geronimus et al., 1994; Hoffman and Maynard, 2008; Jutte et al., 2010; Turley,2003). Whatever the reported results have been, however, there has been little that has contributed to understanding theearly maternal age-child outcomes relationship in a developmental context. Dahinten et al. (2007) provide the only research

Table 5Fixed quadratic growth curve model for PIAT reading comprehension data.



Intercept, b00j 9.741*** (0.206) 10.056*** (0.285) 12.831*** (0.819) 11.454*** (1.029)Age at first birth 18/19, c001 �0.285 (0.495) �0.373 (0.540) �0.431 (0.533)Age at first birth < 18, c002 �0.889 (0.516) �0.813 (0.577) �0.821 (0.582)

Maternal family backgroundFamily Income (log 1979 $), c003 0.096 (0.216)Duncan’s SEI, c004 0.109 (0.235)Parent Highest Grade, c005 0.629** (0.237)Both Biological Parents, c006 0.460 (0.424)Black or Hispanic, c007 0.800 (0.495)Number of Siblings, c008 0.106 (0.072)Delinquency, c009 �0.027 (0.022)AFQT Percentile Score, c0010 0.028** (0.010)Urban Residence, c0011 0.017 (0.483)

Child-specific and maternal household factorsMale, b01j �0.875* (0.374) �0.922* (0.371)Gestation Age, b02j 0.112 (0.097) 0.116 (0.096)Birth Weight (in ounces), b03j 0.006 (0.010) 0.008 (0.010)Post-Adolescent Birth, b04j �1.028 (0.540) �1.230* (0.536)Father Lives in Household, b05j �0.564 (0.393) �0.193 (0.402)Region (Ref. = Northeast)

Midwest, b06j �1.415* (0.670) �1.421* (0.662)South, b07j �1.281* (0.613) �1.074 (0.609)West, b08j �1.879** (0.674) �1.787** (0.662)


Intercept, b10j 7.971*** (0.081) 8.265*** (0.119) 8.529*** (0.341) 8.618*** (0.429)Age at first birth 18/19, c101 �0.431* (0.197) �0.404 (0.227) �0.115 (0.227)Age at first birth < 18, c102 �0.755*** (0.196) �0.622** (0.240) �0.144 (0.245)

Maternal family backgroundFamily Income (log 1979 $), c103 0.149 (0.089)Duncan’s SEI, c104 �0.004 (0.097)Parent Highest Grade, c105 �0.156 (0.097)Both Biological Parents, c106 �0.161 (0.173)Black or Hispanic, c107 �0.088 (0.203)Number of Siblings, c108 �0.073* (0.029)Delinquency, c109 0.014 (0.009)AFQT Percentile Score, c1010 0.026*** (0.004)Urban Residence, c1011 0.132 (0.198)

Child-specific and maternal household factorsMale, b11j �0.396* (0.161) �0.388* (0.160)Gestation Age, b12j �0.014 (0.041) �0.003 (0.040)Birth Weight (in ounces), b13j 0.008 (0.004) 0.004 (0.004)Post-Adolescent Birth, b14j �0.146 (0.228) �0.074 (0.226)Father Lives in Household, b15j 0.448** (0.168) 0.262 (0.174)Region (Ref. = Northeast)



Intercept, b20j �0.359*** (0.008) �0.372*** (0.012) �0.395*** (0.032) �0.374*** (0.041)Age at first birth 18/19, c201 0.022 (0.019) 0.009 (0.022) �0.004 (0.022)Age at first birth < 18, c202 0.033 (0.018) 0.011 (0.023) �0.009 (0.024)Maternal family backgroundFamily Income (log 1979 $), c203 �0.014 (0.008)Duncan’s SEI, c204 0.006 (0.009)Parent Highest Grade, c205 0.007 (0.009)Both Biological Parents, c206 0.012 (0.016)Black or Hispanic, c207 �0.018 (0.020)Number of Siblings, c208 0.004 (0.003)Delinquency, c209 �0.002* (0.001)AFQT Percentile Score, c2010 �0.001** (0.000)Urban Residence, c2011 �0.022 (0.019)



Table 5 (continued)


Child-specific and maternal household factorsMale, b21j 0.045** (0.015) 0.046** (0.015)Gestation Age, b22j 0.001 (0.004) 0.000 (0.004)Birth Weight (in ounces), b23j �0.000 (0.000) �0.000 (0.000)Post-Adolescent Birth, b24j �0.010 (0.021) �0.018 (0.021)Father Lives in Household, b25j �0.026 (0.016) �0.025 (0.017)Region (Ref. = Northeast)

Midwest, b26j 0.025 (0.026) 0.025 (0.026)South, b27j 0.044 (0.024) 0.030 (0.024)West, b28j 0.044 (0.026) 0.043 (0.027)

Family Income (log $), b30j 0.330*** (0.075) 0.189* (0.074)Mother’s Highest Grade, b40j 0.852*** (0.114) 0.195 (0.118)





AIC 90828.191 90736.052 90513.076 89979.757BIC 90887.961 90840.649 90811.926 90480.330


** p < 0.01.*** p < 0.001.


I know of that makes use of longitudinal data to carry out growth curve analyses testing the early maternal age-child out-comes relationship. Their study, though, focused on adolescents, and most of the outcomes, save one, were behavioral. Thepresent study makes a unique contribution to the literature in that it utilized a three-level growth model approach to assessthe mathematics and reading score trajectories from entry into formal schooling through middle adolescence. Like somecross-sectional and family fixed-effect studies, but unlike Dahinten et al. (2007), the present study, after controlling forchild-specific and maternal household factors, also accounted for maternal background factors that might select youngwomen into early childbearing. Specific hypotheses proposed were that early maternal age would have a statistically signif-icant association with children’s mean initial mathematics and reading achievement and mean achievement trajectory, bothin the absence of controls and net of child-specific and maternal household factors. I also sought to explicate whether thefinding of a significant early maternal age relationship held after explicitly controlling for maternal background factors.

Absent controls, the results lend partial confirmation to the hypothesis of a statistically significant early maternal ageeffect on children academic achievement. Early maternal age is negatively associated with children’s mean mathematicsscore at age 5 and the linear slope, or yearly gains, from age 6 on. This only applied to children of mothers who teen child-bearing began prior to age 18, however. The children of mothers who had an adolescent first birth at age 18 or 19 fared noworse than their peers born to mothers whose first birth was in adulthood. Relative to children whose mothers had a firstbirth at 20 or older, children in both early maternal age groups are disadvantaged with respect to reading recognition andreading comprehension achievement, though only as a consequence of age; children in all three maternal age groups havesimilar mean initial reading performance.

Accounting for child-specific and maternal household factors, and again leading to partial confirmation of the hypothesisput forward, the preceding trends are unchanged despite the fact that the estimates of a negative association are somewhatsmaller. That is, consistent with much of the early literature (Brooks-Gunn and Furstenberg, 1986), there does appear to besome relationship between an early maternal age and children’s academic outcomes, even if that relationship varies amongmean initial scores and the linear and/or nonlinear trajectories. The estimate of a negative effect of early maternal age tendedto be larger among children born to mothers whose teen first birth was at younger (prior to age 18) rather than older age(age 18 or 19).

Further accounting for maternal background factors suggests that the negative association between early maternal ageand children’s mathematics and reading achievement is spurious. Children born to mothers in each of the maternal age cat-egories perform equally on the PIAT mathematics, reading recognition, and reading comprehension measures, both at schoolentry, or age 5, and across early childhood into middle adolescence. This is consistent with findings from more recent liter-ature showing no negative relationship between early maternal age and children’s academic outcomes (Geronimus et al.,1994; Levine et al., 2001, 2007; Turley, 2003). Despite concerns of possibly over-controlling the models in this study, thisfinding of no association was robust to the exclusion of covariates possibly affected by an early first birth (i.e., whether afather was in the home at his child’s birth, the income of the maternal household, and mother’s completed education).

Table A1Fixed quadratic growth curve model for PIAT mathematics data. (As a check on robustness, these models exclude variables denoting whether a child’s fatherresided in the home at the time of birth, family income of the maternal household, and mother’s completed years of education.)

Model 3 Model 4

Fixed EffectsInitial Status at age 5, p0ij

Intercept, b00j 8.909*** (0.736) 9.227*** (0.934)Age at first birth 18/19, c001 �0.798 (0.514) �0.203 (0.499)Age at first birth < 18, c002 �1.883*** (0.538) �0.904 (0.538)

Maternal family backgroundFamily Income (log 1979 $), c003 0.061 (0.200)Duncan’s SEI, c004 �0.098 (0.220)Parent Highest Grade, c005 0.275 (0.220)Both Biological Parents, c006 0.495 (0.395)Black or Hispanic, c007 �0.636 (0.457)Number of Siblings, c008 0.015 (0.067)Delinquency, c009 �0.002 (0.020)AFQT Percentile Score, c0010 0.046*** (0.009)Urban Residence, c0011 �0.802 (0.453)

Child-specific and maternal household factorsMale, b01j �0.674 (0.346) �0.651 (0.341)Gestation Age, b02j 0.105 (0.091) 0.100 (0.089)Birth Weight (in ounces), b03j 0.011 (0.010) 0.007 (0.009)Post-Adolescent Birth, b04j �0.199 (0.498) �0.230 (0.494)Region (Ref. = Northeast)



Intercept, b10j 9.239*** (0.298) 8.825*** (0.379)Age at first birth 18/19, c101 �0.210 (0.206) �0.012 (0.208)Age at first birth < 18, c102 �0.558** (0.215) �0.155 (0.222)

Maternal family backgroundFamily Income (log 1979 $), c103 0.122 (0.080)Duncan’s SEI, c104 0.194* (0.088)Parent Highest Grade, c105 �0.067 (0.087)Both Biological Parents, c106 �0.245 (0.157)Black or Hispanic, c107 �0.115 (0.183)Number of Siblings, c108 �0.032 (0.026)Delinquency, c109 �0.000 (0.008)AFQT Percentile Score, c1010 0.017*** (0.004)Urban Residence, c1011 0.565** (0.180)

Child-specific and maternal household factorsMale, b11j 0.078 (0.146) 0.079 (0.145)Gestation Age, b12j �0.016 (0.037) �0.003 (0.037)Birth Weight (in ounces), b13j 0.008* (0.004) 0.004 (0.004)Post-Adolescent Birth, b14j �0.109 (0.207) �0.135 (0.206)Region (Ref. = Northeast)

Midwest, b15j �0.134 (0.245) �0.119 (0.247)South, b16j �0.370 (0.223) �0.059 (0.226)West, b17j �0.024 (0.248) 0.106 (0.248)


Intercept, b20j �0.444*** (0.029) �0.398*** (0.036)Age at first birth 18/19, c201 0.008 (0.020) �0.001 (0.021)Age at first birth < 18, c202 0.030 (0.021) 0.009 (0.022)

Maternal family backgroundFamily Income (log 1979 $), c203 �0.009 (0.008)Duncan’s SEI, c204 �0.015 (0.008)Parent Highest Grade, c205 0.006 (0.008)Both Biological Parents, c206 0.018 (0.015)Black or Hispanic, c207 �0.005 (0.018)Number of Siblings, c208 0.003 (0.003)Delinquency, c209 �0.000 (0.001)AFQT Percentile Score, c2010 �0.001* (0.000)Urban Residence, c2011 �0.057** (0.017)

Child-specific and maternal household factorsMale, b21j 0.006 (0.014) 0.006 (0.014)Gestation Age, b22j 0.002 (0.003) 0.001 (0.003)



Table A1 (continued)

Model 3 Model 4

Birth Weight (in ounces), b23j �0.000 (0.000) �0.000 (0.000)Post-Adolescent Birth, b24j 0.008 (0.020) 0.010 (0.019)Region (Ref. = Northeast)

Midwest, b25j 0.035 (0.024) 0.028 (0.024)South, b26j 0.039 (0.022) 0.017 (0.022)West, b27j �0.003 (0.024) �0.012 (0.024)


Temporal Variation, etij 34.019 34.142Level 2 (children within families)

Individual Initial Achievement, r0ij 2.522*** (0.517) 2.978*** (1.314)Individual Learning Rate, r1ij 0.498*** (0.034) 0.399*** (0.039)

Level 3 (between families)Family Mean Achievement, u00j 17.171*** (1.268) 11.174*** (1.036)

AIC 92778.564 92236.063BIC 93041.626 92702.058


** p < 0.01.*** p < 0.001.


Ascertaining with certainty whether an early first birth does or does not lead to adverse outcomes for the children ofearly-starting mothers is important from a policy perspective. If it does lead to poorer outcomes, then the host of programsaimed at the prevention of teen pregnancy and teen motherhood, from encouraging abstinence to providing comprehensiveeducation about birth control to equipping young women to resist negative pressure from peers (Bennett and Assefi, 2005;Kohler et al., 2008), inasmuch as they are effective, are justified. Whenever and wherever such programs are implemented,whether in the communities where at-risk young women live or in their schools, the result would necessarily be a closing ofracial gaps in intergenerational mobility given the prevalence of teen pregnancy and early maternal age among racial andethnic minorities (Hamilton et al., 2009). If, on the other hand, an early maternal age is not truly associated with children’soutcomes, if its effect is merely a statistical artifact deriving from the failure to address selection issues, then programsaimed at prevention are unjustified to the extent that they misallocate limited resources away from other problems thatyoung people from disadvantaged backgrounds confront and which are more likely to be harmful to their children, from druguse to juvenile criminal behavior to school noncompletion, among other social ills. Moreover, the focus on teen pregnancy orearly maternal age as a social problem when it is not could have its own indirect negative impact on children’s outcomesinsofar as it stigmatizes young mothers (Lewis et al., 2007; SmithBattle, 2013). One mechanism through which the stigmaof an early maternal age might influence children’s outcomes is via depression (Wiemann et al., 2005). Knowledge that theyhave engaged in behavior for which the larger society views them with contempt leads to an increased risk of both socialisolation and abuse among young mothers, which in turn leads to depression, which is linked to poor child outcomes inthe domains of health, development, and behavior (Augustine and Crosnoe, 2010; Wachs et al., 2009).

In the present study, those factors that appear to be of greater consequence than an early maternal age to both children’smathematics and reading development are the time-invariant cognitive ability and time-varying completed education oftheir mothers, as well as the income of the maternal household. Region of residence, urbanicity, child’s sex, and maternalgrandparents’ completed education are also more important to children’s reading outcomes, though not to their mathemat-ics outcomes, than is having a mother who began childbearing early. In line with research by Jaffee et al. (2001), which sug-gests treating children born during their mother’s teen years differently than their siblings born after the teen years, thepresent study included an explicit control for the effect of child-specific maternal age. I note that, while there is no impactof a post-adolescent birth on children’s mean mathematics achievement at age 5 or on their mean mathematics and readinggrowth, there is a statistically significant impact on their mean initial reading achievement on both PIAT reading measures.Regarding these differences, the results suggest that the children born during their mother’s adolescence fared better by lar-ger than a full raw point than their siblings born during their mother’s adulthood, net of the effect of an early maternal ageand the child-specific and maternal household and maternal background controls. Why children born during their mother’steen years might perform better than later-born siblings born during their mother’s adulthood could be attributable to sev-eral related factors. First, while research on the effects of birth order on children’s academic performance has been limiteddue to a lack of representative data to control for family size and address issues of cohort effects of both children and parents,there is some evidence that suggests that higher birth order has a significant and large impact on children’s achievement(Black et al., 2005).

Second, the availability of supplemental resources may be greater for mothers when they are young, but may be severelyreduced as they age. For instance, if the children of teenage mothers are the recipients of a decent quantity and quality ofnon-parental, center-based care, or if they receive additional support from state and federal agencies because of their youngage, the negative impact associated with their mothers’ teenage birth and its concomitant social disadvantage may be

Table A2Fixed quadratic growth curve model for PIAT reading recognition data. (As a check on robustness, these models exclude variables denoting whether a child’sfather resided in the home at the time of birth, family income of the maternal household, and mother’s completed years of education.)

Model 3 Model 4


Intercept, b00j 11.942*** (0.749) 10.668*** (0.957)Age at first birth 18/19, c001 �0.871 (0.516) �0.576 (0.509)Age at first birth < 18, c002 �1.361* (0.542) �0.806 (0.550)

Maternal family backgroundFamily Income (log 1979 $), c003 0.133 (0.205)Duncan’s SEI, c004 0.381 (0.224)Parent Highest Grade, c005 0.367 (0.225)Both Biological Parents, c006 0.141 (0.405)Black or Hispanic, c007 1.137* (0.465)Number of Siblings, c008 0.065 (0.069)Delinquency, c009 �0.023 (0.021)AFQT Percentile Score, c0010 0.033*** (0.009)Urban Residence, c0011 �0.047 (0.464)

Child-specific and maternal household factorsMale, b01j �0.363 (0.355) �0.379 (0.351)Gestation Age, b02j 0.145 (0.092) 0.152 (0.091)Birth Weight (in ounces), b03j 0.008 (0.010) 0.009 (0.010)Post-Adolescent Birth, b04j �1.175* (0.507) �1.281* (0.505)Region (Ref. = Northeast)

Midwest, b05j �1.161 (0.637) �1.229* (0.626)South, b06j �1.024 (0.582) �0.820 (0.575)West, b07j �1.451* (0.642) �1.294* (0.627)


Intercept, b10j 9.816*** (0.307) 9.635*** (0.390)Age at first birth 18/19, c101 �0.559** (0.211) �0.203 (0.212)Age at first birth < 18, c102 �1.010*** (0.221) �0.383 (0.228)

Maternal family backgroundFamily Income (log 1979 $), c103 0.192* (0.082)Duncan’s SEI, c104 �0.130 (0.090)Parent Highest Grade, c105 0.007 (0.090)Both Biological Parents, c106 �0.065 (0.162)Black or Hispanic, c107 �0.358 (0.188)Number of Siblings, c108 �0.025 (0.027)Delinquency, c109 0.013 (0.008)AFQT Percentile Score, c1010 0.031*** (0.004)Urban Residence, c1011 0.162 (0.186)

Child-specific and maternal household factorsMale, b11j �0.649*** (0.150) �0.635*** (0.149)Gestation Age, b12j �0.000 (0.038) 0.013 (0.038)Birth Weight (in ounces), b13j 0.008 (0.004) 0.002 (0.004)Post-Adolescent Birth, b14j �0.120 (0.212) �0.106 (0.210)Region (Ref. = Northeast)

Midwest, b15j �0.120 (0.253) �0.258 (0.255)South, b16j �0.547* (0.231) �0.231 (0.233)West, b17j �0.442 (0.256) �0.318 (0.255)


Intercept, b20j �0.413*** (0.028) �0.382*** (0.036)Age at first birth 18/19, c201 0.031 (0.020) 0.012 (0.020)Age at first birth < 18, c202 0.037 (0.021) 0.004 (0.022)

Maternal family backgroundFamily Income (log 1979 $), c203 �0.018* (0.008)Duncan’s SEI, c204 0.019* (0.008)Parent Highest Grade, c205 �0.008 (0.008)Both Biological Parents, c206 �0.000 (0.015)Black or Hispanic, c207 0.023 (0.018)Number of Siblings, c208 �0.001 (0.003)Delinquency, c209 �0.002* (0.001)AFQT Percentile Score, c2010 �0.002*** (0.000)Urban Residence, c2011 �0.016 (0.017)




Model 3 Model 4

Child-specific and maternal household factorsMale, b21j 0.044** (0.014) 0.044** (0.014)Gestation Age, b22j �0.002 (0.003) �0.003 (0.003)Birth Weight (in ounces), b23j �0.000 (0.000) 0.000 (0.000)Post-Adolescent Birth, b24j 0.003 (0.019) �0.000 (0.019)Region (Ref. = Northeast)

Midwest, b25j 0.024 (0.024) 0.034 (0.024)South, b26j 0.046* (0.022) 0.028 (0.022)West, b27j 0.045 (0.024) 0.038 (0.024)





AIC 93584.938 93104.949BIC 93847.710 93570.431


** p < 0.01.*** p < 0.001.


ameliorated to so great a degree that these children experience an early advantage over their siblings born to adult mothers.There is some evidence to support this as De Haan (2010) has found that earlier-born children are more likely to benefit fromin-kind transfers than are their later-born siblings, a fact that doubtless leads to fewer resources available to higher birthorder siblings born to once teen mothers.

Third, with respect specifically to the outcomes of children born to early-starting mothers, threats to resource allocationmay be exacerbated by young mothers’ propensity to have closely-spaced subsequent birth (Manlove et al., 2000).

Despite its use of a method not normally used to examine the early maternal age-child academic outcomes relationship,the present study is not without its limitations. First, because it is not an experiment, the associations that were foundbetween the two early maternal age categories and children mathematics and reading performance cannot be viewed as cau-sal. Future research attempting to employ a growth curve model will perhaps benefit from utilizing an analytic techniquedesigned to correct for bias deriving from selection issues. It is important to note that in preliminary analyses of the dataused here I attempted to instrument for abortion, miscarriage, and a combination of the two. The result was large negativeestimates for each of the three outcomes. A major drawback of the IV approach, however, is that it leads to large standarderrors (Harding, 2003), which, in this case, it did, leading to the inference of no effect of early maternal age despite unrea-sonably large estimates.

Second, given the age of the data (I follow 5–10 year olds from 1986 to 2000), questions may arise to the external gen-eralizability of the findings over time. If the resources available to children vary significantly across time such that the con-text of one cohort differs dramatically from another, then the findings may be viewed only as time-specific. Similarly, if thereare significant differences in the rearing practices of mothers from one cohort to another, the results presented here cannotbe said to be applicable to all children. Future replications of this or the Dahinten et al. (2007) study, then, will benefit fromusing representative data large enough to control for a cohort effect of both children and mothers.

Third, while one of the advantages of growth curve modeling is that it allows for the ability to estimate the variation inintercepts and slopes across individuals, this advantage is obtained at the expense of possibly biased estimates deriving fromthe omission of fixed effects. A suggested remedy to this drawback that future research might employ is to incorporate intothe multilevel framework of the growth curve model, perhaps through variable manipulation or some other strategy, theadvantages of the fixed effects approach.

The fourth limitation relates to the consideration of early maternal age effects due both to other child-specific factors andhigher levels of nesting. Accepting the proposition, for instance, that whites and minorities, and boys and girls, occupy dif-ferent social milieus, there may be different mediating impacts of the early maternal age-child academic outcomes relation-ship due to (1) neighborhoods, (2) schools, (3) externalizing behaviors such as fighting or drug use, and (4) peer associations,among other things. To the degree that extreme social disadvantage, coupled with its nonrandom distribution, leads to emo-tional and mental instability or sickness at a higher rate than would normally be uncovered in the population, the negativefeedback makes social disadvantage that much worse. Future research must somehow account for the cumulative disadvan-tage resulting from these realities.

Finally, in many ways the present study was a replication of the Dahinten et al. (2007) study, which sought to elucidateearly maternal age effects in a growth curve contexts for primarily behavioral outcomes among Canadian adolescence, net of

Table A3Fixed quadratic growth curve model for PIAT reading comprehension data. (As a check on robustness, these models exclude variables denoting whether achild’s father resided in the home at the time of birth, family income of the maternal household, and mother’s completed years of education.)

Model 3 Model 4


Intercept, b00j 12.890*** (0.801) 11.368*** (1.009)Age at first birth 18/19, c001 �0.809 (0.543) �0.513 (0.532)Age at first birth < 18, c002 �1.470* (0.573) �0.918 (0.578)

Maternal family backgroundFamily Income (log 1979 $), c003 0.119 (0.216)Duncan’s SEI, c004 0.107 (0.236)Parent Highest Grade, c005 0.683** (0.235)Both Biological Parents, c006 0.478 (0.425)Black or Hispanic, c007 0.887 (0.488)Number of Siblings, c008 0.107 (0.072)Delinquency, c009 �0.027 (0.022)AFQT Percentile Score, c0010 0.032** (0.010)Urban Residence, c0011 0.019 (0.484)

Child-specific and maternal household factorsMale, b01j �0.941* (0.376) �0.938* (0.370)Gestation Age, b02j 0.121 (0.098) 0.119 (0.096)Birth Weight (in ounces), b03j 0.007 (0.010) 0.008 (0.010)Post-Adolescent Birth, b04j �1.141* (0.541) �1.278* (0.536)Region (Ref. = Northeast)

Midwest, b05j �1.280 (0.679) �1.407* (0.663)South, b06j �1.226* (0.621) �1.044 (0.610)West, b07j �2.010** (0.683) �1.800** (0.663)


Intercept, b10j 8.717*** (0.331) 8.743*** (0.419)Age at first birth 18/19, c101 �0.420 (0.226) �0.117 (0.227)Age at first birth < 18, c102 �0.706** (0.238) �0.175 (0.244)

Maternal family backgroundFamily Income (log 1979 $), c103 0.143 (0.089)Duncan’s SEI, c104 0.005 (0.097)Parent Highest Grade, c105 �0.166 (0.096)Both Biological Parents, c106 �0.163 (0.173)Black or Hispanic, c107 �0.119 (0.201)Number of Siblings, c108 �0.074* (0.029)Delinquency, c109 0.013 (0.009)AFQT Percentile Score, c1010 0.027*** (0.004)Urban Residence, c1011 0.129 (0.198)

Child-specific and maternal household factorsMale, b11j �0.365* (0.161) �0.377* (0.160)Gestation Age, b12j �0.020 (0.041) �0.006 (0.040)Birth Weight (in ounces), b13j 0.009* (0.004) 0.004 (0.004)Post-Adolescent Birth, b14j �0.065 (0.228) �0.039 (0.226)Region (Ref. = Northeast)



Intercept, b20j �0.406*** (0.031) �0.387*** (0.040)Age at first birth 18/19, c201 0.009 (0.022) �0.003 (0.022)Age at first birth < 18, c202 0.016 (0.023) �0.006 (0.024)

Maternal family backgroundFamily Income (log 1979 $), c203 �0.013 (0.008)Duncan’s SEI, c204 0.005 (0.009)Parent Highest Grade, c205 0.008 (0.009)Both Biological Parents, c206 0.012 (0.016)Black or Hispanic, c207 �0.015 (0.019)Number of Siblings, c208 0.004 (0.003)Delinquency, c209 �0.002* (0.001)AFQT Percentile Score, c2010 �0.001** (0.000)Urban Residence, c2011 �0.022 (0.019)

Child-specific and maternal household factorsMale, b21j 0.043** (0.015) 0.045** (0.015)Gestation Age, b22j 0.001 (0.004) 0.000 (0.004)




Model 3 Model 4

Birth Weight (in ounces), b23j �0.000 (0.000) �0.000 (0.000)Post-Adolescent Birth, b24j �0.017 (0.021) �0.021 (0.021)Region (Ref. = Northeast)

Midwest, b25j 0.027 (0.026) 0.026 (0.026)South, b26j 0.045 (0.024) 0.030 (0.024)West, b27j 0.043 (0.026) 0.041 (0.026)





AIC 90607.038 89984.822BIC 90868.532 90448.039


** p < 0.01.*** p < 0.001.


child-specific and maternal household factors. The difference here is that I focused primarily on academic outcomes of USchildren from early childhood to middle adolescence while additionally controlling for maternal background factors. Futurereplications of either study will want to address both behavioral and academic outcomes to test the hypothesis than an earlymaternal age is truly detrimental to children’s life outcomes, net of both child-specific and maternal household and maternalbackground factors. The failure of the present study to assess the relationship of early maternal age with children’s behav-ioral outcomes in addition to assessing its relationship with their academic outcomes, since the findings can only be under-stood as subject-specific, may therefore be viewed as a limitation.

Appendix A

See Tables A1–A3.

References

Ashcraft Adam, Kevin Lang, 2006. The consequences of teenage childbearing. NBER Working Paper 12485, National Bureau of Economic Research;Cambridge, MA.

Ashcraft, Adam, Fernández-Val, Iván, Lang, Kevin, 2013. The consequences of teenage childbearing: consistent estimates when abortion makes miscarriagenon-random. Econ. J. 123 (571), 875–905.

Astone, Nan M., 1993. Are adolescent mothers just single mothers?. J. Res. Adolescence 3, 353–371.Augustine, Jennifer M., Crosnoe, Robert, 2010. Mothers’ depression and educational attainment and their children’s academic trajectories. J. Health Soc.

Behav. 51 (3), 274–290.Barber, Jennifer S., 2001. The intergenerational transmission of age at first birth among married and unmarried men and women. Soc. Sci. Res. 30 (2), 219–

247.Bennett, Sylvana E., Assefi, Nassim P., 2005. School-based teenage pregnancy prevention programs: a systematic review of randomized controlled trials. J.

Adolesc. Health 36 (1), 72–81.Black, Sandra E., Devereux, Paul J., Salvanes, Kjell G., 2005. The more the merrier? The effect of family size and birth order on children’s education. Q. J. Econ.

120 (2), 669–700.Botting, Beverly, Rosato, Michael, Wood, Rebecca, 1998. Teenage mothers and the health of their children. Popul. Trends 93, 19–28.Bronars, Stephen G., Grogger, Jeff, 1994. The economic consequences of unwed motherhood: using twin births as a natural experiment. Am. Econ. Rev. 84

(5), 1141–1156.Brooks-Gunn, Jeanne, Furstenberg Jr., Frank F., 1986. The children of adolescent mothers: physical, academic, and psychological outcomes. Dev. Rev. 6 (3),

224–251.Chevalier, Arnaud, Viitanen, Tarja K., 2003. The long-run labour market consequences of teenage motherhood in Britain. J. Popul. Econ. 16 (2), 323–343.Coley, Rebekah L., Lindsay Chase-Lansdale, P., 1998. Adolescent pregnancy and parenthood: recent evidence and future directions. Am. Psychol. 53 (2), 152–

166.Corcoran, Jacqueline, 1998. Consequences of adolescent pregnancy/parenting: a review of the literature. Soc. Work Health Care 27 (2), 49–67.Coyne, Claire A., Långström, Niklas., Rickert, Martin E., Lichtenstein, Paul., D’Onofrio, Brian M., 2013. Maternal age at first birth and offspring criminality:

using the children of twins design to test causal hypotheses. Dev. Psychopathol. 25 (1), 17–35.Dahinten, V. Susan, Shapka, Jennifer D., Douglas Willms, J., 2007. Adolescent children of adolescent mothers: the impact of family functioning on trajectories

of development. J. Youth Adolesc. 36, 195–212.De Haan, Monique, 2010. Birth order, family size and educational attainment. Econ. Educ. Rev. 29 (4), 576–588.DeLucia, Christian, Pitts, Steven C., 2006. Applications of individual growth curve modeling for pediatric psychology research. J. Pediatr. Psychol. 31 (10),

1002–1023.Edin, Kathryn, Kefalas, Maria, 2005. Promises I Can Keep: Why Poor Women Put Motherhood before Marriage. University of California Press, Berkeley, CA.Fletcher, Jason M., Wolfe, Barbara L., 2009. Education and labor market consequences of teenage childbearing: evidence using the timing of pregnancy

outcomes and community fixed-effects. J. Human Resour. 44 (2), 303–325.

http://refhub.elsevier.com/S0049-089X(14)00174-4/h0005































Furstenberg Jr., Frank F., Brooks-Gunn, Jeanne, Philip Morgan, S., 1987. Adolescent Mothers in Later Life. Cambridge University Press, New York.Furstenberg Jr., Frank F., Levine, Judith A., Brooks-Gunn, Jeanne, 1990. The children of teenage mothers: patterns of early childbearing in two generations.

Fam. Plann. Perspect. 22 (2), 54–61.Geronimus, Arline T., Korenman, Sanders, 1992. The socioeconomic consequences of teen childbearing reconsidered. Q. J. Econ. 107 (4), 1187–1214.Geronimus, Arline T., Korenman, Sanders, Hillemeier, Marianne M., 1994. Does young maternal age adversely affect child development? Evidence from

cousin comparisons in the united states. Popul. Dev. Rev. 20 (3), 585–609.Gibbons, Robert D., Hedeker, Donald., Elkin, Irene., Watermaux, Christine, Kraemer, Helena C., Greenhouse, Joel B., Tracie Shea, M., Imber, Stanley D., Sotsky,

Stuart M., Watkins, John T., 1993. Some conceptual and statistical issues in analysis of longitudinal psychiatric data. Arch. Gen. Psychiatry 50, 739–750.Goodman Alissa, Kaplan Greg, Walker Ian, 2004. Understanding the effects of early motherhood in Britain: the effects on mothers, IFS Working Paper 04/20,

Institute for Fiscal Studies; London.Grogger, Jeff, 1997. Incarceration-related costs of early childbearing. In: Maynard, R.A. (Ed.), Kids Having Kids: Economic Costs and Social Consequences of

Teen Pregnancy. Urban Institute Press, Washington, DC, pp. 231–256.Grogger, Jeff, Bronars, Stephen, 1993. The socioeconomic consequences of teenage childbearing: findings from a natural experiment. Fam. Plann. Perspect.

25 (4), 156–164.Grogger Jeff, Ronan Nick, 1995. The Intergenerational Effects of Fatherlessness on Educational Attainment and Entry-Level Wages. National Longitudinal

Surveys Discussion Paper, 96–30. Bureau of Labor Statistics, Washington, DC.Hamilton, Brady E., Martin, Joyce A., Ventura, Stephanie J., 2009. Births: preliminary data for 2009. National Vital Stat. Rep. 59 (3), 1–29.Harding, David J., 2003. Counterfactual models of neighborhood effects: the effect of neighborhood poverty on dropping out and teenage pregnancy. Am. J.

Sociol. 109 (3), 676–719.Hardy, Janet B., Shapiro, Sam., Astone, Nan M., Miller, Therese L., Brooks-Gunn, Jeanne, 1997. Adolescent childbearing revisited: the age of inner-city

mothers at delivery is a determinant of their children’s self-sufficiency at age 27 to 33. Pediatrics 100, 802–809.Haveman, R.H., Wolfe, B., Peterson, E., 1997. Children of early childbearers as young adults. In: Maynard, R.A. (Ed.), Kids Having Kids: The Economic and

Social Consequences of Teenage Childbearing. Urban Institute Press, Washington, DC, pp. 257–284.Hoffman, Saul D., 1998. Teenage childbearing is not so bad after all... or is it? a review of the new literature. Fam. Plann. Perspect. 30 (5), 236–243.Hoffman, Saul D., Maynard, Rebecca A. (Eds.), 2008. Kids Having Kids: The Economic and Social Consequences of Teenage Childbearing, second ed. Urban

Institute Press, Washington, DC.Hotz, V. Joseph, McElroy, Susan Williams, Sanders, Seth G., 1997. The impacts of teenage childbearing on the mothers and the consequences of those

impacts for government. In: Maynard, R.A. (Ed.), Kids Having Kids: Economic Costs and Social Consequences of Teen Pregnancy. Urban Institute Press,Washington, DC, pp. 55–94.

Hotz, V. Joseph, McElroy, Susan Williams, Sanders, Seth G., 2005. Teenage childbearing and its life cycle consequences: exploiting a natural experiment. J.Human Resour. 40 (3), 683–715.

Imbens, Guido W., Angrist, Joshua D., 1994. Identification and estimation of local average treatment effects. Econometrica 62 (2), 467–475.Jaffee, Sara, Caspi, Avshalom, Moffitt, Terrie E., Belsky, Jay, Silva, Phil, 2001. Why are children born to teen mothers at risk for adverse outcomes in young

adulthood? Results from a 20-year longitudinal study. Dev. Psychopathol. 13 (2), 377–397.Jutte, Douglas P., Roos, Noralou P., Brownell, Marni D., Briggs, Gemma., MacWilliam, Leonard, Roos, Leslie L., 2010. The ripples of adolescent motherhood:

social, educational, and medical outcomes for children of teen and prior teen mothers. Acad. Pediatr. 10 (5), 293–301.Kearney, Melissa S., Levine, Phillip B., 2011. Income inequality and early non-marital childbearing: an economic exploration of the ‘culture of despair’, NBER

Working Paper 17157, National Bureau of Economic Research; Cambridge, MA.Kearney, Melissa S., Levine, Phillip B., 2012. Why is the teen birth rate in the United States so high and why does it matter?, NBER Working Paper 17965,

National Bureau of Economic Research; Cambridge, MA.Kessler, Ronald C., Berglund, Patricia A., Foster, Cindy L., Saunders, William B., Stang, Paul E., Walters, Ellen E., 1997. Social consequences of psychiatric

disorders, ii: teenage parenthood. Am. J. Psychiatry 154 (10), 1405–1412.Klepinger, Daniel, Lundberg, Shelly, Plotnick, Robert, 1999. How does adolescent fertility affect the human capital and wages of young women?. J. Human

Resour. 24 (2), 421–448.Kohler, Pamela K., Manhart, Lisa E., Lafferty, William E., 2008. Abstinence-only and comprehensive sex education and the initiation of sexual activity and

teen pregnancy. J. Adolesc. Health 42 (4), 344–351.Lee, Katherine J., Carlin, John B., 2010. Multiple imputation for missing data: fully conditional specification versus multivariate normal imputation. Am. J.

Epidemiol. 171 (5), 624–632.Levine, Judith A., Pollack, Harold, Comfort, M.E., 2001. Academic and behavioral outcomes among the children of young mothers. J. Marriage Fam. 63, 355–

369.Levine, Judith A., Emery, Clifton R., Pollack, Harold, 2007. The well-being of children born to teen mothers. J. Marriage Fam. 69, 105–122.Lewis, Oscar, 1969. The culture of poverty. In: Moynihan, D.P. (Ed.), On Understanding Poverty: Perspectives from the Social Sciences. Basic Books, New

York, pp. 187–200.Lewis, Carol M., Scarborough, Megan, Rose, Charlie, Quirin, Katherine B., 2007. Fighting stigma: an adolescent mother takes action. J. Women Social Work 22

(3), 302–306.Manlove, Jennifer, Mariner, Carrie, Papillo, Angela Romano, 2000. Subsequent fertility among teen mothers: longitudinal analyses of recent national data. J.

Marriage Fam. 62 (2), 430–448.McLanahan, Sara, Sandefur, Gary, 1994. Growing Up with a Single Parent: What Hurts, What Helps. Harvard University Press, Cambridge, MA.Menaghan, Elizabeth G., Parcel, Toby L., 1991. Determining children’s home environments: the impact of maternal characteristics and current occupational

and family conditions. J. Marriage Fam. 53 (2), 417–431.Miller, Amalia R., 2009. Motherhood delay and the human capital of the next generation. Am. Econ. Rev. 99 (2), 154–158.Miller-Johnson, Shari, Winn, Donna -Marie, Coie, John, Maumary-Gremaud, Anne, Hyman, Claire, Terry, Robert, Lochman, John, 1999. Motherhood during

the teen years: a developmental perspective on risk factors for childbearing. Dev. Psychopathol. 11 (1), 85–100.Mollborn, Stefanie, Dennis, Jeff A., 2012. Investigating the life situations and development of teenage mothers’ children: evidence from the ECLS-B. Popul.

Res. Policy Rev. 31 (1), 31–66.Moore, Kristin A., Morrison, Donna Ruane, Greene, Anglea Dungee, 1997. Effects on the children born to adolescent mothers. In: Maynard, R.A. (Ed.), Kids

having Kids: Economic Costs and Social Consequences of Teen Pregnancy. Urban Institute Press, Washington, DC, pp. 145–180.Neugarten, Bernice L., Moore, Joan W., Lowe, John C., 1965. Age norms, age constraints, and adult socialization. Am. J. Sociol. 70 (6), 710–717.Parcel, Toby L., Menaghan, Elizabeth G., 1990. Maternal working conditions and children’s verbal facility: studying the intergenerational transmission of

inequality from mothers to young children. Soc. Psychol. Q. 53, 132–147.Parcel, Toby L., Menaghan, Elizabeth G., 1994. Parents’ Jobs and Children’s Lives. Aldine de Gruyter, Hawthorne, NY.Phipps, Maureen G., Sowers, Mary Fran, 2002. Defining early adolescent childbearing. Am. J. Public Health 92 (1), 125–128.Powell, Brian, Steelman, Lala Carr., Carini, Robert M., 2006. Advancing age, advantaged youth: parental age and the transmission of resources to children.

Soc. Forces 84 (3), 1359–1390.Raghunathan, Trivellore E., Lepkowski, James M., Van Hoewyk, John, Solenberger, Peter, 2001. A multivariate technique for multiply imputing missing

values using a sequence of regression models. Surv. Meth. 27 (1), 85–96.Raudenbush, Stephen W., Bryk, Anthony S., 2002. Hierarchical Linear Models: Applications and Data Analysis Methods, second ed. Sage Publications,

Thousand Oaks, CA.








































































Rosenzweig, Mark R., Wolpin, Kenneth I., 1995. Sisters, siblings, and mothers: the effect of teen-age childbearing on birth outcomes in a dynamic familycontext. Econometrica 63 (2), 303–326.

Rubin, Donald B., 1987. Multiple Imputation for Nonresponse in Surveys. Wiley, New York.Scaramella, Laura V., Conger, Rand D., 2003. Intergenerational continuity of hostile parenting and its consequences: the moderating influence of children’s

negative emotional reactivity. Soc. Dev. 12 (3), 420–439.Schafer, Joseph L., Olsen, Maren K., 1998. Multiple imputation for multivariate missing-data problems: a data analyst’s perspective. Multivar. Behav. Res. 33

(4), 545–571.SmithBattle, Lee I., 2013. Reducing the stigmatization of teen mothers. MCN: The American Journal of Maternal/Child Nursing 38 (4), 235–241.Spieker, Susan J., Larson, Nancy C., Lewis, Steven M., Keller, Thomas E., Gilchrist, Lewayne, 1999. Developmental trajectories of disruptive behavior problems

in preschool children of adolescent mothers. Child Dev. 70 (2), 443–458.Sullivan, Kate, Clark, Jamie, Castrucci, Brian, Samsel, Rachel, Fonseca, Vincent, Garcia, Imelda, 2011. Continuing education mitigates the negative

consequences of adolescent childbearing. Matern. Child Health J. 15 (3), 360–366.Turley, Ruth N.L., 2003. Are children of young mothers disadvantaged because of their mother’s age or family background?. Child Dev. 74 (2), 465–474.Wachs, Theodore D., 2005. Person-environment ‘Fit’ and individual development. In: Malden, D.M. Teti (Ed.), Handbook of Research Methods in

Developmental Science. Wiley-Blackwell, MA, pp. 443–466.Wachs, Theodore D., Black, Maureen M., Engle, Patrice L., 2009. Maternal depression: a global threat to children’s health, development, and behavior and to

human rights. Child Dev. Perspect. 3 (1), 51–59.Wiemann, Constance M., Rickert, Vaughn I., Berenson, Abbey B., Volk, Robert J., 2005. Are pregnant adolescents stigmatized by pregnancy?. J. Adolesc.

Health 36 (4), 352.e1–352.e7.Wildsmith, Elizabeth, Manlove, Jennifer, Jekielek, Susan, Moore, Kristin Anderson, Mincieli, Lisa, 2012. Teenage childbearing among youth born to teenage

mothers. Youth Soc. 44 (2), 258–283.Wilson, William Julius, 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. University of Chicago Press, Chicago, IL.Woodward, Lianne J., Fergusson, David M., 1999. Early conduct problems and later risk of teenage pregnancy in girls. Dev. Psychopathol. 11 (1), 127–141.
























Documents

Growth curve analyses of the relationship between early maternal age and children’s mathematics and reading performance