Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
CHAPTER 8 THE CHILDREN OF TEEN CHILDBEARERS Sandra L. Hofferth INTRODUCTION It is clear that being a child of a teenage mother often entails numerous risks: low birth weight, complications of the mother's preg- nancy and delivery, and health problems associated with poor perinatal outcomes; greater risk of perinatal death; lower IQ and academic achievement later on, including a greater risk of repeating a grade; greater risk of socio-emotional problems; a greater risk of having a fatal accident before age one; and finally, a greater probability of starting one's own family at an early age. Although there are varia- tions from study to study, most studies that survey a representative sample from a population that has had no special interventions and is of diverse socioeconomic makeup, and that do not control for SES or other factors, find that children of teen parents are at greater risk than children of older parents for a host of health, social and economic problems. The critical objective, of course, is to explain why being a child of a teenager entails these risks. This is important because it af- fects the way we plan interventions to prevent undesired outcomes. The implications of an outcome due to physical immaturity (or, in the case of an older mother, the aging process) are different from those that are due to inadequate prenatal care or to inadequate nutrition, to poverty or to ignorance. Explanation is, therefore, the goal of this chapter, which is divided into several sections, each focusing on a specific outcome: health; cognitive development and school achieve- ment; and socioemotional development. The fourth section focuses on intervening factors: e.g., family structure, socioeconomic status, and maternal education. It also looks at the part parenting behaviors play in distinguishing adolescent from older parents and the influence of such behaviors in mediating child outcomes. Finally, the last section focuses on methodological issues and substantive issues that need further research. Two major data sets are used in this chapter, the Collaborative Perinatal Project (CPP) and the Health Examination Survey (HES). The CPP included all patients or a random sample of all patients qualifying 174
175 for prenatal care in the 12 participating medical centers during 6 years of intake (1966-1973~. The total sample size was 53,625. The children of respondents were followed at ages 6 to 8 and a subsample was again followed-up at about 12 years of age. Cycle II of the Health Examination Survey (HES), conducted in 1963- 65, consists of a national random sample of 7110 children age 6-11. The children were given health and psychological exams. Information was also collected from the mother, the school, and from the birth certificate. Cycle III of the HES, conducted in 1966-70, consists of a national random sample of 6768 youth 12-17. The information collected is the same as in Cycle II, with the addition of a questionnaire filled out by the Youth. A small subset of children interviewed in Cycle III had also been interviewed in Cycle II. HEALTH Perinatal Mortality The first outcome of interest is perinatal mortality. A number of studies (see Strobino, this volume; also Makinson, 1985) report a higher incidence of perinatal mortality among teenage mothers. These studies show the relationship between mother's age and perinatal mor- tality as a J-shaped function. That is, it is high at very young ages, declining to a low point in the mid-twenties, and then climbing again among older mothers. The evidence is consistent that perinatal prob- lems increase among mothers above age 30; however, recent evidence from the Danish Perinatal Study and from the Collaborative Perinatal Project in the U.S. show a linear relationship between maternal age and peri- natal mortality with low rates among young women, and increasing rates with maternal age (Mednick and Baker, 1980--or that there is no rela- tionship (Broman, 1981), at least for ages 12-29. There are two major types of explanations for the often found asso- ciation between young age of mother and higher incidence of perinatal problems. First, it is hypothesized that the teenager is physiologi- cally immature; thus her less desirable outcomes (see for example, NCHS, 1984:10~. A second explanation is the differential social characteristics of teenage mothers-lower SES, lack of access to pre- natal care, poor nutrition, poverty and ignorance (see, for example, Baizerman, 1977; Mednick and Baker, 19801. Whatever explanations are used (and different ones may hold for different age groups) should account for the higher levels of perinatal problems among both teen and older mothers. In both the Danish and the U.S. Perinatal studies, relatively high risk groups were overrepresented.2 However, Mednick and Baker (1980: 38) argue convincingly that "In view of the unusually advantageous
176 treatment conditions prevailing in the university hospital samples as a group, the teenage mothers in these samples probably received con- siderably more intensive and higher quality treatment than teenagers in the population at large. n Because a clear relationship does exist between quality of medical care and perinatal mortality rates (Mednick and Baker, 1980:39-40), the latter argue that The relatively lower mortality rates observed among the teenage subjects, compared with the rates in older age groups in the American and Danish Perinatal samples, are due to the provision of adequate pre- and perinatal medical treatment . . . the previously reported higher mortality rate associated with teenage deliveries was not caused by physiological characteristics of the teenage organism but rather by social factors that have the ultimate effect of lowering the quality of medical treatment received by teenage mothers in the general population. In contrast, constitutional changes do appear to determine the in- creased risk of perinatal mortality with increasing age of mother at birth. The results from studies of representative samples as well as from special hospital samples show a similar relationship for mothers over 30; increased age appears to be associated with increased rate of death (Mednick and Baker, 1980~. Neonatal Health Vital statistics data (NCHS, 1984) show that children of teen mothers are more likely to be below 2500 grams at birth than children of mothers 20 to 39, and the younger the age of the mother the higher the proportion of infants of low birth weight. In 1982, twice as many infants of 10-14 year olds (13.8 percent) were low birth weight as in- fants of 20-24 year olds (6.9 percent). In that year 9.3 percent of the infants of 15-19 year olds were low birth weight. Low birth weight babies are subject to higher risks of death, mental retardation, and other health problems (Williams and Chen, 1982~. Low birth weight has also been implicated in poor intelligence and achievement test scores in childhood (see, for example, Edwards and Grossman, 1979; Mednick and Baker, 1980~. A second measure of neonatal health is the Apgar score. The Apgar score is a summary measure used to evaluate the neonate's overall physical condition at birth. It is a composite evaluation of five factors--heart rate, respiratory effort, muscle tone, irritability, and color-each of which is assigned a value from O to 2. The overall score is the sum of the five values, with a score of 10 being optimal (NCHS, 1984:12~. Infants of teen childbearers are more likely to score under 7 at either one or five minutes after birth than are infants of mothers 20 to 39. These results hold for both blacks and whites, though the proportion of low birth weight infants and the percent with low Apgar scores are consistently higher among blacks than among whites.
177 Although these relationships appear to hold in the population as a whole, there appears to be little difference between children of adolescent and non-adolescent mothers in special samples where prenatal and postnatal care are good. Sandier et al. (1981) evaluated the rela- tionship between the age of mother and two measures of newborn be- havior: 1) the Neonatal Behavioral Assessment Scale (Brazelton) and 2) a measure of infant temperament (Carey "My Baby" scale). No differ- ences were found on the Brazelton Scale or Carey scale between children of adolescents and post adolescents (age not defined) within the first few days after birth. Lester et al. (1982, 1983) used the Brazelton Scale on the second day after birth of a sample of Puerto Rican and American infants of teen mothers. In addition they obtained information on a number of health measures from medical records. In a regression analysis con- trolling for ponderal index, gestational age, marital status, drug score, 1 minute Apgar and the number of maternal parturitional and fetal nonoptimal conditions, none of the associations between maternal age and Brazelton scale cluster scores were significant. There did appear to be an interaction in the Puerto Rican sample between a com- plications index and age. Infants of young mothers with few complica- tions had a wider range of states of arousal than infants of older mothers with few complications. In both these studies (Sandier et al. and Lester et al.), mothers received excellent prenatal and postnatal medical care through a special program for low income families. A number of recent studies failed to find any difference by age of mother in health status of neo- nates at birth (Apgar score, birth weight, prematurity, birth trauma, etc.) once initial differences such as differences in SES between adolescents and non-adolescents were controlled (Zuckerman et al. 1983; Rothenberg et al., 19811. Net of SES, Broman (1981) found older women to have higher birthweights among blacks, but not whites. Also net of SES, Broman (1981) found the youngest adolescents (12-15) to have lower Apgar scores than older adolescents among whites and blacks. The dif- ferences were very small, however. Infant Health Status The medical risk to neonates of adolescent childbearers does not appear to be biological, but, rather, due to differential access to adequate medical care (Mednick and Baker, 1980~. Less research has focused on the effect of age of mother on the health status of in- fants, that is from the first 28 days to one year of age. Two studies have addressed maternal age differences and infant health status (Hardy, 1978; and Mednick and Baker, 1980~. Hardy pre- sents one figure which shows that the risk of infant death after the neonatal period is higher for the infants of black teen mothers than for the infants of black older mothers. However, no differences among
178 whites by age of mother at birth were found. This study did not con- trol for the SES of the mothers, however e The Mednick and Baker (1980) study, using Danish data, looked at the physical health status of the infant at one year as an outcome measure (see Makinson, 1985, for results of other non-U.S. studies). They found that the relationship between mother's age and infant's first year physical health status was curvilinear. That is, infants of mothers under 20 and over 35 were the healthiest; those of mothers in their twenties had the most health problems. Comparing neonatal and one year outcomes, children of the youngest mothers were the best off at both points. In contrast children of older mothers were less well off at birth, but very well off at one year. This suggests dif- ferent mechanisms influencing the different outcomes at two points in time: biological factors at birth, environmental factors at one year. Older mothers may have the most biological problems but the best en- vironment. Age-related social variables may be enough to compensate for the negative biological effects seen at birth. Mednick and Baker show that the most important predictors of health status at one year were birthweight and being female. After controlling for these impor- tant factors, a number of environmental factors were associated with better child health, including an older mother, fewer previous preg- nancies, and less exposure to institutional day care. Why the infants of Danish mothers under 20 were healthiest at one year also needs explanation. Mednick and Baker hypothesized that teen mothers may have older adults to rely on for support. They found that infants living with their grandmothers had the best mean health score; infants living with both biological parents a mid-range score, while infants who lived with their unmarried mother or in an institution or foster home showed the worst scores at one year. In one analysis, after controlling for birthweight and pregnancy complications, number of nurturing adults was strongly related to a positive one year health status among children of teen mothers. Mothers in their twenties may lack the parental support of the young mothers as well as the maturity and experience that come with age. In conclusion, it appears that once the birth occurs and survival is assured, health status varies strongly with social and environmental variables. In the case of the older mother, age implies a number of positive psycho-social and environmental aspects. In the case of the young mother, it may imply the availability of alternative caregivers to help out. The worst one-year outcomes occurred among children of 18-29 year olds. Once infant survival is assured, environmental and social variables begin to emerge as important to the continued physical growth and development of the child" (Mednick and Baker, 1980:65~.
179 Path Analysis of Infant Health Status at One Year The previous analysis of health status at one year did not control for a number of other factors that might affect health: health status at birth or complications of pregnancy and delivery. The question is whether there are residual effects of non-medical variables that may impact on one year infant status. A number of studies (e.g., Sameroff, 1979) have shown that environmental factors do not have major effects on cognitive and neurological measures within the first 12 months of life. Measures of infant physical health and motor development have been shown to be sensitive to variations in prenatal environment. Good perinatal care can insure good perinatal outcomes even when environ- mental conditions are less than adequate. During the first year of life, environmental influences may increase in importance as the posi- tive effects of good prenatal care wear off. Intervention postnatally is less common than prenatal medical intervention. During the year after birth, the Danish cohort studied by Mednick was more similar in medical care to the general population. Thus effects of environmental factors could be expected to show at one year. Mednick and Baker (1980) developed a path model to trace the causal connections between background, mother's age, and intervening medical and health factors on one year infant outcomes. Background factors (spacing, mother's age, previous health, data on previous pregnancy, wontedness, use of institutional day care, SES, mother's employment and family size) were assumed to predict one year infant outcome through the following health and medical variables: complications of pregnancy and delivery, multiple births, birth weight, and neonatal physical and neurological status. Two random samples were pulled from the full sample and models were tested separately on each sample. Unfortunately the results differed substantially between the two samples. Mother's age did not have a consistent direct or indirect effect on one year physical or neurological status or one year motor development. In sample 1, older mothers had children with poorer one year physical status. In sample 2, older mothers had children with better one year neurological status (direct effect) and better one year motor devel- opment through improved neonatal physical status (indirect effect). However, given that the same findings don't hold up in both samples, there appears to be no consistent direct or indirect effect of mother's age on infant status at age one. Infant status at age one was influenced directly by birthweight and neonatal neurological status and indirectly by neonatal physical status. In addition, exposure to institutional day care significantly reduced rating of health status at one year. Thus mother's employment showed an indirect effect via daycare on one year health status. Higher birthweight was associated with improved one year motor development. The analysis supports the conclusion that neonatal status is strongly influenced by factors subject to medical intervention. Maternal age, SES, and even previous pregnancy history effects on neo
180 natal health are weak in a sample which received excellent medical care. By age one, neonatal status exerts the strongest influence on physical and motor status. Although none of the expected background factors has an impact at age one, environmental influence on physical status can be seen through the direct (negative) impact of institu- tional day care, and the indirect (and also negative) influence of maternal employment. This analysis assumed a linear relationship between maternal age and outcomes. In fact, other analyses by the same researchers have shown a non-linear relationship. The weakness of maternal age effects may be due to differential influences across the life cycle. Finally, this analysis did not and could not include the potential ameliorating influence of other adults in the home for the very young mother. Thus although high quality medical care appears to have reduced the environmental influence on children's health over the first year, there is evidence that social conditions, which did not have an impact during that first year, at one year do have an impact. Fatal Infant Accidents Further evidence for the importance of environmental factors is found in a study using linked birth and death records from North Carolina and Washington State for 1968 through 1980. Wicklund et al. (1984) found a strong inverse relationship between maternal age and mortality rates from accidents for children under one, net of parity and educational level of mother (a proxy for SES). The actual mortal- ity rate from accidents during the first year of life is actually quite low--in 1980 in North Carolina about 3 out of 10,000 live births died from accidents in the first year in Washington state the rate was 1.47 per 10,000 live births. There were substantial differences by race, maternal education and age of mother, however. Children of mothers under 20 who had 9 or more years of schooling were substantially more likely to die from accidents in the first year of life than children of mothers 20 and over with the same amount of schooling. Among chil- dren of mothers with very low levels of schooling, those with mothers 24 and younger were more likely to die than those with mothers 25 and older. Education was also strongly inversely related to infant mor- tality from accidents and parity was directly related. That is, mortality rates were lower for children with a more educated mother and one with fewer children. Black children had almost twice the rate of deaths from accidents in the first year as white children. The leading causes of infant accident mortality in North Carolina were suffocation by inhalation and/or ingestion of food, and suffoca- tion by mechanical means (e.g., in bed or cradle, by plastic bag, etc.), with transport accidents coming third. In Washington State, transport accidents were the leading cause, with mechanical suffoca- tion and food suffocation next. Parental care is crucial for the
181 safety and well-being of children; and such parental care appears to be less dependable among families in which the mother is young, black, less well-educated and has more children. Neuropsychological Status/Motor Development The Collaborative Perinatal Project was originally designed by the National Institute of Neurological and Communicative Diseases and Stroke (NINCDS) as a study of the neurological problems of children. Children were assessed at age four using the Graham-Ernhart Block-Sort, a battery of fine-motor development tests and a battery of gross-motor development tests. The Bender-Gestalt test was the main measure of functioning for seven year olds. These tests measure motor functioning and development, an indicator of brain damage. Three studies using the CPP have looked at the association between age at birth of the child and motor development. Marecek (1979) found no consistent evidence for a relationship between age at first birth of the mother and the child's motor development at age 4. At age 7, Marecek found a slight curvilinear relationship such that children of both older and younger mothers do slightly less well on the Bender- Gestalt test than children of mothers in their late teens and early twenties. Hardy et al. (1978), in contrast, using the Baltimore sub- sample of the CPP, found a significant difference in scores on the Bender-Gestalt test at age 7, favoring the children of older mothers. Neither of these studies controlled, however, for differences in socioeconomic status of the family. Controlling for SES, Broman (1981) found that both gross and fine motor scores of 4 year olds were lower among children of black adolescent mothers than black older mothers. Only the gross motor scores were lower among the children of white adolescent mothers compared with the children of white older mothers. No analysis of motor development was reported by Broman for children age 7. Because the results appear to be inconsistent it is reasonable to conclude, as did Marecek, that there is no evidence of a real or sub- stantial difference in motor development/brain damage by age of mother at first birth (Marecek) or age of mother at birth of index child (Broman and Hardy et al.~. COGNITIVE DEVELOPMENT AND SCHOOL ACHIEVEMENT The major source of data on infants and young children is the Collaborative Perinatal Project (CPP). Although a number of different researchers have utilized data from this study in their analyses, the subsamples they have used have been slightly different. As a result, the results should not be expected to be identical. Marecek (1981) used the Philadelphia subsample of the survey in her study, Hardy
182 (1980) used the Baltimore subsample, while Belmont et al. (1981) and Cohen et al. (1980) and Broman (1981) used the entire sample of respon- dents and their children. Although blacks were overrepresented in the entire sample, they were especially predominant in both the Baltimore and the Philadelphia samples. These sample hospitals served a primar- ily black, low SES area of their respective cities. Thus the results from Marecek and Hardy may differ from those of the other studies. Infants The Collaborative Perinatal Project assessed the developmental status of infants at 8 months using the Bayley Scales of Infant Devel- opment and the Infant Behavior Profile. The Bayley Test consists of two separate scales: the Mental Scale and the Motor Scale. The Infant Behavior Profile was designed to evaluate qualitative aspects of chil- dren's behavior. Finally, the 8 month exam includes summary ratings of general development based on evaluations of the examiners Marecek found that, among blacks, first born children of mothers under 20, as a group, scored lower on the average than first born children of older mothers on the Bayley mental scale. Among whites, in contrast, first born children of mothers 20 to 25 scored lower on the average than children of mothers 18 to 19; children of mothers under 18 did not differ from the other groups. Differences are very small, however. The author then looked at the individual components of the scale. Black children of women under 20 scored lower on three components than black children of older mothers--incidental spontaneous exploration, social interaction, and awareness of object constancy. In the white sample, children of mothers under 18 scored lower on 2 come portents--incidental spontaneous exploration and ability to sustain attention than children of older mothers. On the Bayley motor scale there were no differences by mother's age at first birth for blacks or whites. Ratings on the Infant Behavior Profile ranged from 1 to 5, with an extreme underresponse to physical stimuli rated 1 and an overresponse rated 5. Among whites there were no differences by age of mother. Among blacks, age of mother at first birth had an effect on three ratings. Black children of childbearers under 20 were likely to be rated weak in their responses to physical stimuli, with black sons of adolescent childbearers more likely than black sons of older child- bearers to be slow in their responses. Children of older childbearers were more likely to be overly apprehensive in response to the examiner relative to children of younger childbearers. However, only a very small proportion of the sample was in either of the extreme cate- gories--1 or 5. On the summary ratings of black children's development (as rated by a pediatrician on a three-point scale--normal, suspect or abnormal) age was not related to development for daughters. For sons, age was
183 related to development. Twice as many sons of childbearers under 18 at first birth were rated suspect compared with sons of mothers 18 and older at first birth (9.2 compared with 4.1 percent). There was no relationship for whites. Marecek draws two conclusions: 1) The effects of maternal age on infant mental development are small, and 2) Maternal age is more likely to affect boys' development than girls'. Thus on these measures little difference was found between children of older and younger mothers. Marecek did not control for SES. In her study using the entire CPP, Broman examined the relationship between age and Bayley scale score within SES categories. Broman also found differences by age net of SES, with Bayley mental and motor scale scores higher among infants of younger (13-15 and 16-17) than older mothers (20-29~. Since the differences in both studies are small, and they operate in different directions, the conclusion would appear to be that there is little difference on these measures between children of older and younger mothers. Early Childhood The Collaborative Perinatal Project (CPP) used the Stanford-Binet Intelligence Scale to measure the intelligence of children at age 4. Using this measure, Hardy et al. (1978) found a significant difference in IQ score at age 4 between children of black mothers who bore that child at 17 or younger and those who were 20 to 24 at that birth. The difference is about 4 IQ points on the average. There was no differ- ence for whites. Marecek also failed to find a difference on the Stanford-Binet by age at first birth among whites. Among blacks, age of mother at first birth had no significant effect on girls' IQ but had a marginally sig- nificant effect for boys (probability less than .081. The sons of mothers under 18 scored lower on the average than those of mothers 20 to 25 at first birth, with those of mothers 18 to 19 intermediate. Marecek finds that among children of childbearers under 18 and 18-19, boys scored lower on the average than girls, while there was no sex difference among children of mothers 20 to 25 at first birth. As a result, she concludes that boys tend to be affected more strongly by mother's first birth age than girls. Controlling for SES, Broman (1981) found a 5-6 point difference in IQ at age 4 between both black and white children of older and younger mothers, favoring the former. SES effects were larger than age effects, however. Furstenberg (1976) also found a difference in cognitive performance between black children of adolescent parents and black children of classmates who delayed childbearing until age 18, even when differen- tial school attendance was controlled. Cognitive performance (as mea- sured by the Preschool Inventory) was higher among comparison group children.
184 Middle Childhood Intelligence The CPP measure of intelligence used for children age 7 was the Wechsler Intelligence Scale for Children (WISC). The WISC consists of two major scales: the verbal and the performance scales, each with 6 subtests. A subset of 7 subtests of the WISC was used on the Collab- orative Perinatal Study, three verbal (information, comprehension, vocabulary) and four performance (digit span, picture arrangement, block design, and coding) subtests (Marecek, 1979~. In the Marecek study, no difference in intelligence by mother's age at first birth was found for whites. Among blacks, both Hardy and Marecek found differences by mother's age at first birth. As a group, children of childbearers under 18 tended to do less well than children of later childbearers. However, Marecek found a sex difference here. The relationship was curvilinear for daughters, linear for sons. As a group, daughters of mothers 18-19 tended to do best; sons of mothers 20-25 tended to do best. In addition, sons of childbearers under 18 tended to do less well than daughters of childbearers under 18. How- ever, there was no sex difference among children of later childbearers. This again suggests a stronger impact of maternal age on boys on the average than on girls. There were several differences among blacks by type of scale. On the performance scale, sons of 20-25 year old mothers scored highest; daughters of 18-19 year old mothers scored highest. On the verbal scale, sons of childbearers 20-25 scored highest. On the same scale daughters of childbearers 18-19 scored higher than daughters of child- bearers under 18 with little difference between daughters of 20-25 year olds and 18-19 year olds. Marecek (1979) estimated path models of the direct and indirect effects of adolescent childbearing on WISC scores. These models were developed only for blacks, since the white samples were too small for meaningful analyses. The variables included were age of mother at first birth, mother's education, mother's marital status, number of parents in the household, per capita income, and child's behavior con- trol. The results showed no direct effect of mother's age at first birth on child's IQ score. There were small indirect effects through mother's marital status, number of parents in the household and per capita household income which were stronger for males than for females. The models were, unfortunately, unable to explain much of the covari- ance in terms of other variables in the model, either because 1) some important intervening factors may have been left out or 2) there may be some direct effects not captured by the variables in the model. The total correlation between age of mother at first birth (l=under 18; 0=18-2S) and IQ score was -.042 for males on the verbal scale, -.102 for males on the performance scale, -.105 for females on the verbal scale and -.017 for females on the performance scale. The
185 total indirect effects identified were -.038, -.039, -.024 and -.014 respectively. The results suggest that age of mother is related to both verbal and performance IQs among black male children but only to verbal IQ among black female children. However, the relationship is very small, accounting for less than one percent of the variance in children's IQ scores, and, furthermore, is indirect (Marecek, 19797. Of course, the sample was relatively homogeneous with regard to SES, and mother's age was restricted to a maximum of 25 years. Broman (1981) found, using the whole CPP, that scores on the WISC for seven year aids were only marginally associated with maternal age among whites, and unrelated to maternal age among blacks, controlling for the effects of SES. SES effects were very large, in contrast. The second major national study which collected data on children is the Health Examination Survey (HES), Cycles II and II. Cycle TI collected data on children 6 to 11 in 1963-65 and Cycle III collected data on children 12-17 in 1966-70. There is some overlap in the samples of Cycles II and III, that is, some of the same children were interviewed in cycles II and III. Belmont et al. and Cohen et al. both used Cycles II and III as well as the Collaborative Perinatal Study. Levin (1983) used cycles II and III of the HES only. Davis and Grossbard-Schectman {1980) used Cycle II of the HES. Two subtests of the-WISC, vocabulary and block design, were included on the HES. Belmont et al. attempted to address the issue of whether there was a unique disadvantage for children's intelligence of having a mother who was a teenager. Thus they used a measure both of mother's age (in years) and a dummy variable indicating whether or not the mother was a teenager at the birth of the study child. They found a linear rela- tionship between maternal age and the IQ score of the child (WISC) in all three surveys; howeve r, they found no evidence of excess disad- vantage to the offspring of mothers under 20 above and beyond the linear relationship. (For example, children of mothers age 20 do less well than children of even older mothers). Belmont el al. found the contribution of maternal age in years to be very small, contributing less than 1 percent of the variance ex- plained in intelligence. The contribution of teenage mother status was even smaller. The most important factor explaining variance in child's intelligence was education of the parents. Family size, age and sex of the child were also important. They found some interesting subgroup differences and interactions. In particular, they found that the impact of mother's age at birth of the study child was stronger among 6 year old than among 11 year old children. It was stronger among urban than rural children, among blacks than whites, and among children in larger families. IQ scores were lower for blacks than whites overall. Finally, scores of children in the CPP, which is from a lower SES population, were lower than those of children in the HES Cycles II and III.
186 In general, the results from the three data sets were very similar. The authors' (Belmont et al.) conclusion was that the offspring of teen mothers suffer TQ depression only because of associated social disad- vantages and not because of any "immaturity of the mother. n Cohen et al. (1980) also conducted a path analysis on these data to attempt to tease out some of the causal sequences leading to lower intelligence scores of the children of teen mothers. Here again, the authors looked at the influence of maternal age in years as well as whether the child's mother was under 20 or 20 and over. There was no direct effect of teen maternity on the child's IQ. The results showed that the effect of linear maternal age was significant in five of the six samples (three surveys and two race groups) after controlling for a number of other factors. The exception was blacks in Cycle III of the HES. The size of the effect was rather small, however; approxi- mately .09 to .21 IQ points per year of maternal age. Again, the re- sults from the three surveys were similar. There was no apparent difference either by age of child (6-11 versus 12-17) or by type of population (low income versus all income levels). The effect of maternal education is much larger--approximately one IQ point per year of maternal age among whites, somewhat less (.4 to .9) in the black sample (Cohen et alp. The effects of paternal edu- cation were smaller. Maternal employment had inconsistent effects--a negative impact in HES Cycle II and a positive impact in the CPP. Father absence and large family size both had negative effects on IQ. Even though teen maternity (being a child of a teen mother under 20 compared with being a child of a mother 20 or older) had no direct ef- fect on TQ, it had several indirect effects (Cohen et al.~. The largest of these was through maternal and paternal education. The total mean TQ difference for whites ranged from -2.45 in the HES Cycle III to -2.71 in the CPP to .4.22 in HES Cycle II. Among blacks the in- direct effect was only significant in the CPP, 1.00 IQ point. Teen mothers were more likely to have had less schooling, which reduces their children's IQ. Family structure had a smaller, but significant impact. Children of teen mothers were more likely to live in homes not headed by both biological parents, and this was associated with lower IQ scores. As a result the total mean difference between the TQ scores of white children of teen mothers compared with those of white children of older mothers ranged from .2.45 in the HES Cycle II to -2.71 in the CPP to -4.22 in HES Cycle II. The IQ scores of black children of teen mothers were significantly lower by one IQ point only in the CPP. Linear maternal age had both direct and indirect effects. Here the direct effects dominated (Cohen et al. ~ . The major indirect effects operated through family size, in particular the number of births sub- sequent to the study child (since birth order is also controlled). Delaying a first birth one year was associated with a rise in IQ of from .09 IQ points (HES, III) to .25 IQ points (HES,IT) for whites.
187 Levin (1983) also used the HES survey, Cycles II and III. He found significant effects of mother's age at birth of the study child on WISC vocabulary and block design scores, as well as the two subtests to- gether, controlling for sex and age of child. These relationships did not disappear when controls were introduced for race, birth order, in- come, education, household structure, household size and ecological factors. However, as the author pointed out, the sizes of the effects are small mother's age at birth explains less than half of one percent of the variance in cognitive variables controlling for other factors. Total variance explained ranges from .18 to .33. Davis and Grossbard-Schechtman analyzed Cycle II of the HES. Their study focused on 10 to 11 year olds. They explored the impact of two different variables: mother's age at birth of index child and whether or not the mother was an adolescent (under 18) at birth of the study child. Children of mothers 40 and over were excluded from the study. The authors concluded that, net of other factors, having a mother who was an adolescent did reduce the scores on the WISC. However, the dif- ferences were only marginal. Having an adolescent mother was asso- ciated, on the average, with an IQ score lower by 2.14 points on the vocabulary scale (marginally significant at p .10) and by 2.15 points on the block design scale (not significant). Thus the effects of these variables were relatively small. Having a mother with a low level of education was more harmful--approximately one IQ point for each year of schooling. According to Davis and Grossbard-Schechtman, the age of mother (in years) appeared to significantly affect WISC vocabulary scores, but not block design scores. The effect also appeared to be non-linear. At higher levels of age, the scores appeared to turn down slightly. The effect of maternal age on vocabulary scores was small--it in- creased about one half of one IQ point for each year of maternal age. This effect held net of a large number of other factors, including health factors. Maternal age had a significant but very small impact on grade retention. Again this effect was slightly non-linear. The probability of repeating a grade dropped as age of mother rose, but rose again slightly among children of older mothers. There was no association of age of mother with child's reading score. Finally, boys' vocabulary scores appeared to be affected more strongly than girls' by mother's age at birth. A boy's score increased with mother's age until the mother reached age 35; afterwards, the older the mother, the lower the boy's score on the average. A third nationally representative survey of children has also been used to study the relationship between parental age at birth and the child's intelligence (Moore et al., 1985~. This survey, the National Survey of Children, collected data in 1976 on 2,301 children aged 7-11 in 1,747 households. Interviews were conducted with the eligible child (or two eligible children if there were two or more children in the household), and with the parent most capable of providing information about the child, usually the mother. In addition, school information
188 was obtained on 1,682 of the children. Not all the children are first born, although the analyses examine the impacts of the age of mother at first birth. The measure of intelligence used in this data set is the Peabody Picture Vocabulary Test, a test that measures both innate ability as well as stimulation in the home environment. The results show that children of teen mothers are less likely than children of older mothers to do well on this test, a result which is statistically significant for whites but not for blacks (Moore et al., 1985~. When the data are adjusted for the effects of mother's education, family configuration, sex of child, family income, number of siblings, and number of moves in last five years, the differences between children of mothers younger and older at first birth decline sharply. A1- though children of youngest mothers generally have the lowest scores, among whites and blacks the children of the very youngest mothers at first birth {those less than or equal to 15), in fact, have high scores relative to children of older mothers. Additional analyses conducted by these researchers suggest that what differentiates the scores of children is whether the mother dropped out of school at an early age and didn't return or whether she continued in school/ returned and completed more schooling later on. The very earliest childbearers may be the most likely to continue/return to school and, as a result, their children may not suffer as much. However, this is still speculation, as no research yet shows this to be true. Furs- tenber~ and Crawford (1980) suggest that those who remain at home are more likely than those who leave to obtain more schooling. Achievement The CPP and HES measured school achievement using the Wide Range Achievement Test, which includes subtests on spelling, reading and arithmetic. The purpose is to measure skills, not intelligence. A second measure of achievement is whether or not the child had repeated a grade of school by the~time of the 7 year exam, and the mother's evaluation of the child's success in school, as reported at the 7th birthday. ~ Using data from a white middle and working class Northeastern community, Kinard and Reinherz (1984b) looked at children prior to entry into school, at the end of kindergarten, at grade 3, and again at grade 4. The test used at time one was the Preschool Screening System. In grade 4 the Short Form Test of Academic Aptitude and the California Achievement Test 70 were used. In addition, ratings of school performance by parents and teachers were obtained at the end of grade 3. Parents rated school achievement on a five point scale and completed the Child and Adolescent Adjustment Profile (CAAP) of Ells- worth. Teachers assessed child's reading, arithmetic and overall academic achievement on a five point scale and also completed the Ellsworth scale.
189 Kinard and Reinherz found that, controlling for sex and maternal education, maternal age had a main effect on only one measure: infor- mation processing skills at preschool. Children of late adolescent (18-19) mothers had lower scores than children of either early adoles- cent or older mothers. However, this effect did not persist over time: no difference was found at grade three or grade four. There was no effect of maternal age on achievement and aptitude test scores or on teacher and parent ratings of performance. Maternal education was the major factor affecting cognitive and achievement scores, with substan- tial and consistent differences on almost every measure favoring chil- dren of better educated mothers. Vincenzi and Brewer (1982) used two samples of children, one in grade 4 and one in grade 6, from a primarily black low income area, to look at the school achievement of children of teen mothers. They found an impact of having a teen mother only for children with no preschool or kindergarten experience. Controlling for SES, AFDC receipt, and initial achievement level, children of teen mothers with no preschool or kindergarten experience tended to have lower reading scorers, more absences and were more likely to be retained in grade than children of non-teen mothers with no preschool experience. In addition, the achievement scores of children of teen mothers were helped more by preschool and kindergarten than the scores of children of non-teen mothers. ~ The results reported by Marecek based on the CPP show among blacks a linear relationship of age of mother with scores on the three sub- tests of the WRAT. Children of adolescent childbearers scored lowest. On the reading subtest age of mother at first birth was a stronger pre- dictor of scores for boys than girls. For whites, the age at first - birth of the mother affected scores only on the arithmetic test, and there was no sex difference. ~ Broman (1981) found that children of young~mothers (12-15) and 16-17 were more likely to have below average scores~on the arithmetic, reading and spelling subtest of the WRAT. For example, as the mother's age increased from 12-15 to 20-29 the percent of white children scoring low decreased from 8 to 4 percent. Among blacks the percent of low scorers decreased from 19-5 to 15-5 percent. These results controlled for SES differences between younger and older childbearers. Levin (1983) used the HES survey. Results are similar to those found with the DISC: significant effects of mother's age at first birth on the WRAT arithmetic and reading scores, as well as the full test, after controlling for sex and age of child. These relationships do not disappear when controls are introduced for race, birth order, income, education, household structure, household size and ecological factors. However, as the author points out, the sizes of the effects are small; mother's age at birth explains less than half of one percent of the variance in cognitive variables controlling for other factors. Total variance explained ranged from .18 to .33. Mother's age is also ,,,9
190 associated with ratings of child's exceptional performance, academic difficulty, and precociousness. However, when other variables are controlled, only the relationship of mother's age with exceptional performance is still significant. Again, the percent of variance explained is very small. Marecek found that, among whites, mother's age at first birth is unrelated to grade repetition. Among blacks she found a higher repe- tition rate among children of childbearers under 20 at first birth. Controlling for race, Davis and Grossbard-Schechtman found that having a mother who is young at birth of the study child does increase grade retention (indirectly through IQ). In addition, the impact of mother's schooling was found to be more significant for adolescent than for older mothers in keeping the child at grade level. That is, having an adolescent mother with one additional year of schooling decreased grade retention by almost 50 percent. In contrast, having an older mother with one additional year of schooling reduced grade retention by only about 10 percent. Moore et al. (1985) found a strong relationship among blacks and whites between age at first birth of the mother and being behind grade for age among 7 to 11 year olds: children of younger mothers were more likely to be behind grade. The difference was substantially weakened, although it did not disappear, when controls for other variables-- mother's education, family configuration, sex of child, family income, number of siblings, and number of recent moves--were added to the analysis. Among whites, Marecek found no relationship between age at first birth of mother and mother's reports of learning disturbances in their offspring. In the black sample there is a significant relationship with age at first birth among males. Nearly 12 percent of boys born to childbearers under 20 were rated as having a learning disturbance compared with 4 percent of the sons of older mothers. Thus there appears to be an effect of mother's age at first birth on school achievement and grade repetition. This effect is stronger for blacks than for whites and for boys than for girls. However, the effect is very small and is not found in every study. Adolescence Several studies have looked at adolescents: Levin (1983), Card (1978~; Belmont et al. (19801. The effects of maternal age do not appear to weaken as the children grow older. Levin (1983) finds simi- lar relationships between age of mother at birth of child and ~Q and achievement scores among 12 to 17 year olds as among 6 to 11 year olds. Net of sex and age of child, mother's age at birth is significantly associated with scores on the WISC and WRAT tests and subtests, with exceptional performance, and with ratings of academic difficulty.
191 None of these relationships disappear when additional variables (race, birth order, income, education, household structure, household size and ecological factors) are controlled. As before, however, the pro- portion of variance explained by mother's age is very small--under 1 percent. Total variance explained ranges from 20 to 35 percent. Card (1978) looked at teens at age 15 and 17. An adolescent mother was defined as a mother less than 20 (for those who were age 15 in 1960) or a mother less than 18 (for those age 17 in 19601.3 In both cohorts children of adolescent parents had lower scores than their classmates on cognitive tests; they also had lower educational expecta- tions. The cognitive differences were about .4 standard deviation in magnitude. However, when other factors were controlled, these differ- ences declined. A change of one standard deviation in the proportion of the sample who were adolescent parents was associated with only about a one point change in IQ score. Card (1978) also developed a path model. Besides the small direct effect of having an adolescent parent on academic aptitude, there was a substantial indirect effect through family structure. Children of adolescent parents were much more likely to be living with only one parent than children of older parents, and children in one-parent families had significantly lower aptitude scores, grades and aspira- tions. Card concluded that the cognitive consequences of adolescent paren- tage were more severe for male than female children since she found that although in the comparison group males had higher mean academic aptitude scores than females, among children of adolescent parents, females had higher scores. Mednick and Baker used data from the Danish Longitudinal Study to examine youth age 17 to 19 in 1979. They, unfortunately, did not have test scores on youth. However, they obtained from teachers ratings on reading proficiency, math proficiency, reasoning ability and work or- ganization. From parents they obtained ratings of general problems in school and academic performance in general. In addition, they had substantial information on the health of the child at birth and in early childhood. Net of SES they found that for males the older the age of the mother at the birth of her first child the greater the reading pro- ficiency, math proficiency, reasoning ability, and the fewer the school problems. For females, the older the age of mother at first birth the greater reading proficiency, reasoning ability, work or- ganization, the fewer the school problems and greater academic per- formance in general. Mother's age at birth of study child was also related to measures of ability and achievement, but less strongly than age at first birth. Children whose mother was not a teen when born were better at math (males) and academic performance in general (girls) than those whose mother was a teen when they were born.
192 In a later analysis Mednick and Baker developed a path model to test causal relationships among mother's age and other variables in the model with child outcomes. Here they used only mother's age at birth of the study child. The models were developed separately for males and females. In these models, which controlled for a number of background factors (education, family size, birthweight, and SES) and a number of intervening factors (maternal characteristics, orderliness and contentment, family stability, crowding in the home, and father's criminality), effects of maternal age at birth were weakened. For males there was no direct or indirect impact of mother's age on either mother's or teacher's judgments of child's academic performance. Among females, mother's age had no direct effect but did have one indirect effect through mother's contentment. That is, older mothers were more content, and content mothers rated their children's school performance higher than discontented mothers. SOC I OE MOT I ONAL DEVELOPMENT The CPP data base and the home interview study conducted at the same time obtained ratings on a variety of behavioral dimensions. These ratings in the CPP were based on the examiner's direct obser- vations of the child during the 4 and 7 year psychological testing sessions. The home interviews obtained reports of the child's be- havior from the child's caretaker. The examiners' ratings were made on a five point scale, in which a "3" reflects appropriate behavior. The behaviors reported by caretakers were coded for the presence or absence of the behavior--for example, habitually bites nails. The variables available at seven years were similar to those available at the four year exam. The Marecek study used the following from the four year psychologi- cal testing sessions: emotional reactivity; irritability; degree of cooperation with examiner; degree of dependency on the examiner; dura- tion of attention span; goal orientation; response to directions; activity level; and indices of deviant stereotyped behavior. All 14 home interview variables were used: abnormality of behavior control; bedwetting; disruption in conduct; delay in the development of self- care; nail-biting; ingestion of non-food substances; phobic responses; sleep disturbance; thumb-sucking; social maladjustment; mechanical speech defects; defect in speech construction; stuttering; speech dis- turbance summary measure. (For a complete description of all measures obtained in the CPP, consult Marecek, 1979~. Variables from the 7-year examination used in the Marecek study were the following: separation from mother; fearfulness; rapport with examiner; self-conf idence; emotional reactivity; degree of coopera- tion; frustration tolerance; degree of dependency; assertiveness; hos- tility; duration of attention span; goal orientation; level of activ- ity; nature of activity; nature of communication; and indices of de- viant stereotyped behavior. In addition, all 14 home interview variables were used.
193 Early Childhood At age 4, Marecek found no difference among white children in re- ported behavioral disorders by mother's age at first birth. Among blacks, children of adolescent childbearers showed excessive confor- mity, insufficient ability to communicate, abnormal control behavior (girls, not boys), and excessive nail-biting. Furstenberg found no difference between children of adolescent mothers and older mothers on four indices of interpersonal develop- ment: ability to defer gratification, efficacy, trust and self- esteem. The children were between 3 and 6 years old at the time. These were assessed by interviewers from responses to structured questions using a doll play game (see Furstenberg, 1976~. Middle Childhood At age 7, Marecek found no differences by age of mother at first birth in socio-emotional behavior of white children. Among blacks, in contrast, boys especially showed problems in social behavior and in expression of affect. Both girls and boys of black adolescent mothers tended to show more problems controlling their behavior than children of black older mothers. Among 7 year olds, girls of adolescent mothers exhibited more bedwetting and phobias, while boys exhibited more thumbsucking. Finally, sons of adolescent mothers exhibited more speech deficiencies. In summary, on average the effects of mother's age on socio- emotional development were more marked for boys than girls and for 7 year olds than for four year olds. The domain of strongest effect was that of social behavior. The second domain of effect was that of self- control. Children of black adolescents were at greater risk. The type of adjustment problems differed by sex, with maladjustment for boys ex- pressed as rebelliousness, aggression or undercontrol of anger, while maladjustment for girls expressed itself as fearfulness and other "neurotic" behaviors. The effects seemed to increase rather than de- crease over time. Male children of black adolescent childbearers were judged to be more openly hostile, aggressive and willful than male children of older childbearers and their mothers reported that the children had difficulty in relating to peers. These behavior ten- dencies may interfere with school learning and thus prevent the child from fulfilling his/her intellectual potential. In the Health Examination Survey (HES), a number of types of in- formation on socio-emotional development were obtained. A behavioral/ attitudinal history of the child was obtained in an interview with the parent or guardian. The child/youth's psychological status was asses- sed by direct examination by a staff psychologist. School achievement and adjustment were obtained in a self-administered questionnaire from the child's teacher. For youth 12-17, in addition, the parent's des
194 cription of the youth's health, behavior and attitudes were obtained from the parent or guardian in a self-administered questionnaire. The youth's own health behavior and attitudes and health habits and history were obtained from the youth through a self-administered questionnaire. For information on specific items see Levin (1983~. Using the HES, cycle II, Levin (1983) found mother's age at birth of the study child to be related to a few emotional and psychological adjustment variables: problems with going to bed, problems with speech articulation, and problems with individual and social activities. These were significant even after controlling for other factors among children 6 to 11 in the HES. Kinard and Reinherz (1984a) studied the effect of maternal age on the socio emotional development of a sample of predominantly white lower middle and working class children. Parental ratings on the Simmons Behavior Checklist were obtained at preschool screening, at the end of kindergarten and again at the end of the third grade. The Child and Adolescent Adjustment Profile (CAAP) was also used at the end of grade 3. Teacher ratings were obtained using the Preschool Behavior Questionnaire and the CAAP. Child self-concept was measured using the Piers-Harris Children's Self-Concept Scale. Finally, infor- mation on the number of years during elementary school the child used two types of school services--guidance (psychological) and special needs (academic)--was obtained. There was no difference on prenatal or neonatal conditions by mother's age at first birth (15-17, 18-19, or 20-24), nor was there any difference on childhood health and development by maternal age at first birth. Controlling for family structure, maternal age had a direct impact only on one variable the number of years special needs (academic) services were received. This relationship was opposite from the expected--children of older mothers received special services for more years than children of younger mothers. There was one inter- action: children of adolescent mothers in one-parent families had the highest mean score for attention problems while those of early adoles- cent mothers in two-parent families had the lowest. Controlling for maternal education, there was a direct effect of maternal age on third grade teachers' ratings of withdrawn behavior. Children of early adolescent mothers tended to have the fewest problems while children of late adolescent mothers tended to have the most. Kinard and Reinherz concluded that children of adolescent mothers were generally no different from children of mothers in their early twenties with respect to behavior and emotional functioning. Maternal education had the greatest impact on behavioral and emotional func- tioning. The extent to which children of adolescent mothers are at risk for behavioral and emotional maladjustment seems to be a function of the association between adolescent childbearing and low educational attainment.
195 Adolescence Among children 12 through 17, Levin found age of mother at birth of the child to be related to a greater incidence of mental problems, to delinquency, and to difficulties with social contacts even after controlling for other variables. Before controlling for these variables, children of adolescent mothers also showed more speech problems, problems of parental control and self-centered problems, and problems with n socioability.~4 Mednick and Baker looked at the association between mother's age at birth of first or index child and measures of socioemotional devel- opment. They found that among males, with control for SES, a child of a mother young at first birth was more likely to exhibit criminal be- havior at ages 17-19. Among females, those born to mothers who were adolescents at first birth were more likely to be aggressive, impul- sive, emotional and to have poor peer relations. The relationship with mother's age at birth of the index child are generally consistent: among males, those with young mothers are more emotional and get along less well with adults. Among females, those with young mothers get along less well with peers, are more aggressive, and more impulsive. In the path model, controlling for mother's education, family size, family socioeconomic status, and birthweight, Mednick and Baker found that mother's age no longer has direct effects; however it has some indirect effects on children's socioemotional development. In particular, a younger mother's age is associated with family instabil- ity among both males and females. Family instability is associated with withdrawn behavior (poor peer interaction, fearfulness and feel- ings of inferiority) among males and with acting out behaviors (aggres- siveness, impulsivity) among girls. A younger mother's age also af- fects crowding in the homes for both males and females, but only among females does crowding in the home affect behavior: it is associated with withdrawal. Card also found several differences between children of adolescent and non-adolescent parents. Children of adolescent parents were less sociable, less tidy, less "cultured"6 and less mature than their classmates. Children of adolescent parents had greater interests in outdoor recreation activities, mechanical and technical matters, skil- led trades, and labor than their classmates. These psychological differences were not large, however, being only about .2 standard deviation in magnitude. Children of adolescent parents also had lower educational expectations and aspirations. However, when sex, race, SES, birth order, and head of the household were controlled, the rela- tionship between having an adolescent parent and personality traits, and interests and aspirations disappeared, whereas differences in academic aptitude remained significant.
196 Adulthood Card (1978) also explored the consequences of mother's age at first birth for children 11 years after high school, approximately age 30. Net of sex, race, socioeconomic status, birth order, and head of household, children who had an adolescent parent completed less school- ing, married at a younger age, and married more times than those who didn't have an adolescent parent. When, in addition, a control for academic aptitude was introduced, differences in schooling and number of marriages disappeared. The difference in age at first marriage re- mained. In a path model, Card also found a slight tendency toward earlier childbearing among children of adolescent parents. There were a number of indirect effects of adolescent parentage on later child- bearing history, educational attainment, occupation and income. These operated through family structure, family SES in 1960 and academic aptitude. For example, having an mother who gave birth while an adolescent affected the child's academic aptitude, which affected the child's own childbearing history. Several other studies (Presser, 1976; Newcomer and Udry, 1984) have also found that daughters of early childbearers are likely to be early childbearers themselves. Newcomer and Udry (1984) were unable to ex- plain much of this relationship in terms of transmissible attitudes, communication patterns or behavioral control attempts. Thus they hypothesized a biological mechanism such as age at physical maturation. However, they could not rule out causes (such as socioeconomic back- ground) that may be common to both mother and daughter. Maternal modeling is also a reasonable hypothesis: that the daughter tends to do what the mother does, rather than what she says. However, in the case of early sexual activity and childbearing, the behavior is not directly modelable since it precedes the birth of the daughter. The daughter cannot model what the mother did while she was a teenager, only what she does now. INTERVENING FACTORS The results from all the studies show fairly clearly that having a young mother does tend to have negative effects on a number of outcomes for the child--in particular, measured intelligence, achievement, and some aspects of socioemotional development--and these results appear not to decrease over time. However, the results also indicate that the direct effects of having a young mother are very small. Rather, most of the effects are mediated by other variables. In this section the evidence on these intervening variables is summarized by reviewing the path models researchers have developed. The four path analyses of interest are by Mednick and Baker (1980), Card (1978), Cohen et al. (1980), and Marecek (1979~. In the Marecek and Cohen et al. analyses, the dependent variable was the WISC score from the Collaborative Perinatal Project (at age 7 for Marecek and at
197 ages 6 to 8 in the Cohen et al. analysis). The Cohen et al. analyses also included children 6 to 11 and 12 to 17 from the HES. The Card and the Mednick and Baker analyses referred to teens 15 to 17 and 17 to 19. In the Card analysis, the dependent variables were academic aptitude in high school and childbearing history, education, occu- pation and income at age 30. In the Mednick and Baker analyses the outcomes were mother's and teachers' judgements of child's academic performance at age 17-19, child's acting out behaviors, and withdrawn behaviors. Two of the studies (Card, 1978, and Cohen et al., 1980) found a small direct impact of age of the mother at the birth of the study child on academic achievement and IQ as a teenager. In the remaining studies the impact was only indirect. The studies differed consider- ably in whether education of mother and father were included as con- trol variables (Card; Marecek; Mednick and Baker) or whether they were included as potential intervening variables (Cohen et al., 1980~. In the Cohen et al. study, education was the most important variable intervening between teen maternity (versus later maternity) and child's IQ, with family structure contributing, but less important. In con- trast, when looking at the relationship between age of mother in years and child's later achievement and IQ, family size was the most impor- tant intervening variable. The other studies used an indicator of SES instead of maternal education as intervening variable. The Mednick and Baker study used crowding in the home, which was highly associated with SES. In both the Marecek and the Card studies, the most important intervening factor was household structure, which affected academic aptitude both directly and indirectly through family socioeconomic status. In the Mednick and Baker study, mother's age had no indirect effect on the academic performance of males. However, it did affect that of females. For females, however, the most important intervening factor was mother's contentments. Unfortunately, Mednick and Baker did not have actual test scores for their youth. As a result, the path analysis was not directly comparable to that of Card and of Cohen et al. Since only parent and teacher evaluations were used, it could be anticipated that response tendencies, which are affected by personal- ity and environmental influences, might have influenced the results. That is, perceptions or evaluations by teachers might have been con- taminated by their knowledge of the socioeconomic status and teen parenthood status of the mother and her family. Mednick and Baker were the only ones to also explore the inter- vening factors predicting socioemotional characteristics of the chil- dren of adolescent and nonadolescent childbearers. Controlling for education of the mother, socioeconomic status of the family, family size, and birth weight, they found, as for cognitive development, that mother's age did not directly affect the behavior of boys or girls; however, it did have indirect effects. For boys and girls the strongest effects operated through family stability. For girls, in addition, there was an indirect impact through crowding in the home. Thus, in the area of academic achievement it was the characteristics
198 of the mother, such as her contentments and orderliness8, which appeared to exert the most direct influence; in the area of socio- emotional functioning, it was the stability of the family situations that was the crucial determinant of child behavior. Family structure appears to be one of the most important factors intervening between parental age, background factors such as SES and race and the outcomes of the child's cognitive development, such his/ her later achievements. Besides the Card and Cohen et al. analyses, several other studies have examined selected parts of the model and have found family structure to be an important intervening factor. Both Menken and McCarthy (1979) and Kellam et al. (1982) found that children of mothers who were teenagers at first birth were themselves more likely to spend time with only one parent than children of older childbearers. Kellam et al. (1977), Furstenberg (1979) and Mednick and Baker (1980) show that the family structure of the child, in turn, has very important effects on the health, cognitive development, social adapta- tion and psychological well-being of the child. However, the relation- ship is not simple. Kellam et al. found that children in mother-alone families were at highest risk of maladaptation to school, with children in mother/father or mother/grandmother families at least risk, and children in mother/stepfather families similar to children in mother- alone families in risk. Furstenberg and Crawford (1980) found that among young mothers who remained unmarried, those who lived with their parents were much better off than those who left home. Young mothers who remained unmarried and stayed with their parents were more likely to return to school and to graduate from high school; a larger pro- portion were employed; and a smaller proportion were on welfare. Furstenberg (1979) found that although there was little difference by family structure, children of unmarried mothers who lived in a house- hold with kin {usually grandparents) tended to outperform those who lived with their mothers alone on one measure of cognitive skills, even though the latter children were more apt to have gone to school. Results from the Mednick and Baker analysis also supported the argu- ment in favor of family support. They suggested that the health of infants of teens who received help from other family members was better than that of infants of teens who did not have such assistance, and that this might have explained in part why it was hard to show a difference between the children of young teens and older mothers at one year of age: young teenagers were more likely to have family assistance. The impact of a child born to a teenager on other members of her family of origin may also be important. Furstenberg found no long-term consequences for the socioeconomic and marital and family careers of the members of the adolescent mother's family of orientation, including occupational mobility of the father, marital dissolution of the par- ents, and siblings' educational attainment, freedom from welfare de- pendency and marriage. Similarities among siblings' life courses were
199 probably due to homogeneity of background and not early parenthood. Taking a new child into the home did have consequences for the dynamics of the family, some positive and some negative (see Furstenberg, 1979~. A second pregnancy did appear to propel the young mother from the home. These results suggest the importance of considering whether the teen childbearer remains with her parents or starts her own household in determining the risk to the child. What factors are associated with reliance on parents, particularly remaining in the parental home? Furstenberg (1979) found the factors associated with an adolescent mother remaining in the home to include: 1) younger age, 2) a strong affective bond, 3) desire to remain in school, 4) both parents in household, 5) higher level of income/SES, 6) more physical space and less crowding, 7) no subsequent childbearing, and 8) remaining single. One of the consistent and most important findings in the study of the effects of mother's age at first birth and child outcomes is that the education of the mother has a consistent positive impact on the intelligence and achievement of her child. The effect is consistently large, regardless of how it was included as part of the model: about 1 IQ point for each year of schooling of the mother in several of the studies (Cohen et al., 1980; Davis and Grossbard-Schechtman, 1980; Edwards and Grossman, 1979~. Previous research (see Chapter 6) has shown a strong relationship between an early first birth and educa- tional deficits among young women. Not only does lack of schooling hinder the prospects for a young woman's future employment, economic well-being, and life success, but it appears to have very detrimental effects on her children. Unfortunately, we still have very little information about exactly what education means. If we had some better understanding of what it is about education that improves children's cognitive and socioemotional performance, then we could better target programs to teen mothers. One possible link is through parenting be- haviors of such mothers. This is the topic of the following section. In conclusion, having a young parent, on average, is harmful to children; there is a small direct effect, but there is an even larger indirect effect which is due to-differential characteristics of the mother (such as orderliness), to her lesser schooling, to less stable family structure, to lower family socioeconomic status, and to larger family size. The size, types of effects and causal pathways of effects differ for girls and boys, and for blacks and whites. It is clear that future analyses should develop separate models by race and sex. PARENTING BEHAVIORS OF ADOLESCENT AND NON-ADOLESCENT PARENTS Recently there has been increased attention paid to explaining dif- ferences between children of adolescent and non-adolescent mothers in terms of differential parenting behaviors. The rationale is that dif- ferences between children of adolescent and non-adolescent mothers might be explainable by differences in their parents' childrearing be
200 havior and practices. Such behaviors (following Elster et al., 1983) result from differences between adolescent and older parents in: 1. Stress and coping, 2) social support, 3) cognitive development, 4) attitudes toward childrearing, 5) knowledge of child development, and 6) infant characteristics. This all seems very reasonable except that the differences between children of adolescent and non-adolescent parents were found to be very small. Most of the differences we observed were due to indirect ef- fects through other factors. Thus, for example, it may be more rele- vant to compare the childrearing practices of mothers living with a husband or another relative compared with living alone, or of mothers with low versus high levels of schooling and so on. There is a sub- stantial literature developing in this area (see Hetherington et al., 1981~. The small direct effect of having an adolescent mother implies that the chance of finding much difference in childrearing practices is probably very small. And this is, in fact, what the studies show. Neither Sandler et al. nor McAnarney found significant differences in mothering behaviors during the first several days after birth. Of 48 comparisons made by Sandler et al., three were significant. They showed that the older the mother, the more time she was likely to spend out of contact with the baby, the more the total amount of voca- lizing by the mother, and the less the amount of silence in the mother-infant interactions during the first days after birth. In the McAnarney study, no relationship was found between mother's age at first birth among adolescent mothers and any of eight major maternal behaviors or the counts of one major infant behavior category. Sandler also used the Cohler Scale of Maternal Attitudes toward their infants. No difference was found between adolescent and non-adolescent mothers on this scale. McAnarney (1984) reports that some differences in parenting practices begin to show up at one year, but such results are still tentative and based on a very small sample of teen mothers. Furstenberg {1976) found no difference in maternal interest, maternal performance or maternal success by age at first birth in a sample of black teen childbearers. There is a growing body of research looking at the infant parenting behaviors of teenage and older mothers, which finds small differences between the groups (Roosa and Vaughan, 1984; Osofsky and Osofsky, 1971; de Cubes and Field, 1984; Field et al., 1985; Elster et al., 1983~. One problem with the research is obtaining comparable samples of older mothers; the latter are more likely to be married, and of higher SES, for example. A second problem is sample attrition which has proven to be a problem in studying teen mothers and their infants over time (McAnarney, 1983~. A third problem is that of rater bias. Since it is fairly easy to distinguish older and younger mothers, the research designs to date can't eliminate the possibility that the age of the mothers affects observer ratings. This field appears to be growing; it is really too early to judge what the results will be. For a good review of the research to date, see Elster et al., 1983.
201 Another area of increasing interest to researchers is that explor- ing the knowledge of teen mothers about child development relative to that of older motherse Early research (De Lissovoy, 1973) found teen mothers to be ignorant of developmental norms for children's behavior and to hold unrealistic expectations. Recent research also suggests teen mothers to be less knowledgeable than older mothers about child development; however, the differences are relatively small. That is teen mothers do not rate much lower than older mothers (Roosa, 1983; Stevens, 1983~. Field and colleagues (1982; 1985) have developed a series of inter- ventions to increase parental knowledge of child development and imp prove parenting, which appears to have been successful among low in- come and teenage mothers. Again, it is difficult to sort out the effects of SES, education and age of mother, since teen mothers are disadvantaged on all factors. Lack of control for SES may explain the inconsistencies in results from study to study and the failure to iden- tify strong age effects. In addition, one study found that an effect of age disappeared as the mothers matured into their twenties (Stevens, 1983~. A third area in which research appears to be increasing is that of identifying the relationship of knowledge of child development to parenting practices (see, for example, Stevens, 1984; Johnson et al., 1982; Le Resche et al., 1983; Roosa, 1983~. Levin (1983) is the only one so far to find significant differences in parental childrearing practices between young and older mothers of elementary school age children. Net of sex and age of child, race, birthorder, income, education, household structure, household size and ecological factors, he found younger mothers of children 6-11 to be significantly less likely than older mothers of children 6-11 to moni- tor their children's behavior and to control their bedtimes. Parental monitoring refers to the last time a doctor or dentist was seen and the number of child's friends parents know well. Parental control of bed- time refers to reported problems getting child to bed and reported naps taken when child was little. The parental control variable is not statistically significant with controls for other variables among youth 12 to 17. There is very little agreement in the child development literature on the impact of maternal behavior, if any, on child development, since the interplay between various factors is quite complex, including the influence of the child on the parent. Strong conclusions from the re- search on parenting among teen mothers (and fathers) are not warranted at this time. Although a number of writers and authors have suggested a connec- tion between adolescent parenthood and the abuse and neglect of chil- dren, there is little evidence to substantiate this link. A recent review (Kinard and Klerman, 1980) of the published papers in this area
202 points out that the findings of many studies are conflicting. The authors suggest that the main reason for a link, if any, is the socio- economic status of the families, not the age of the mother per se. That is, both births to adolescents and reported cases of child abuse are more common among lower SES families. Poverty may contribute both to early pregnancy and to child abuse. More work is needed in this area. ME:THODOLOG ICAL I SSUES There are several inconsistencies across studies that make it some- what difficult to compare results. These are discussed in Kinard and Reinherz (1984) and will be only briefly summarized here. The first is that age of the mother is variously defined as age of mother at first birth or age of mother at the birth of the index child. The Marecek study avoids the problem by selecting only first born children. The Kinard and Reinherz study, in contrast, uses age of mother at first birth, but the study child is not necessarily the first. The majority of studies (see Table 1) use age of mother at birth of index child. This is more likely to be the first child for adolescent than for older mothers. Thus birth order and family size are important variables confounded with age of mother at first birth. A second issue is the categorization of age groups in comparing teenage and older mothers. The most common division appears to be under 18, 18-19 and 20 to 24. However, in some studies, the first two groups are collapsed; in some studies the third group consists of all those 20 and older. The former could be a problem if results differ between early and late teens. The latter is a problem because some of the outcomes are poorer for children of mothers 40 and older. The analyses using path models generally specify a linear age variable. However, the Cohen analysis uses both a linear and a dummy variable for age of mother. Some studies do not even specify the age groups that were used in the analysis. Finally, depending on how age of mother is defined, some mothers who are categorized as 20+ at this birth, could have been adolescent mothers at an earlier birth. A third issue is the source of data. Studies using standardized tests and other standard scales are the easiest to compare. Most problematic are those studies that rely heavily on parent and teacher reports of behavior, since these seem so easily contaminated by paren- tal attitudes, beliefs, and well-being. It would be very useful to have some methodological analyses that attempt to sort out the factors that contribute to response sets on such ratings and evaluations. Mul- tiple measures by a variety of raters and evaluators would be most use- ful. Such measures appear in the large data sets such as the CPP and HES. This review has relied heavily on the studies using these data sets for this reason.
203 The fourth issue is that of control and intervening variables, and the appropriate identification of each. Most studies did control for confounding background variables such as SES. Several, in addition, looked at the influence of intervening factors such as family structure and family size. One problem is whether to include parental education as a background or intervening factor, and it is included in different ways in different studies. But probably the most important methodo- logical conclusion is that the analyses really must be conducted separately by both race and sex (or interaction terms used to sort out the different effects). The only study to do this was Marecek. The Mednick-Baker study in effect does so since the sample is all white. Dividing the sample by sex is important in looking at cognitive out- comes, since males' and females' verbal and performance abilities are subject to different influences, and in looking at socioemotional out- comes as well, since behavior disturbances are manifested differently among boys and among 9 iris. A fifth and final issue is that of def ining what the intervening factors such as "prenatal care,n Education," and "family structure" really mean. That is, what is it about them that affect children's health, cognitive and socioemotional development? SUMMARY AND CONCLUSIONS Although a relationship between an early first birth and the child's health at birth has been found, this appears to be a result of less than adequate prenatal and perinatal care rather than biology, since it appears to disappear in special hospital populations that receive excellent health care. Unfortunately, here again, what pre- natal care contributes is not clearly defined. Children of older mothers are consistently less healthy at birth than children of average age mothers. This is likely to be a true biological effect. The few studies that have looked at the health of infants of adolescent and older mothers find few direct effects of age on infant health. One study, however, did find the death rate from accidents within the first year of life was much higher for infants of teenage than older mothers, even controlling for maternal education and family size. The age of the mother at birth of a child does appear, on average, to affect her child's intelligence scores on standard tests, achieve- ment scores on standard tests, retention in grade, and other parental and teacher evaluations of performance. This appears to hold for both blacks and whites, for children of all ages beyond the infant level, and for both boys and girls. The direct effects, however, are very small in all the studies. This probably explains why studies using large samples (such as the CPP and the HES) do obtain results that are statistically significant while other studies of smaller samples obtain only occasionally signif icant results (for example, Kinard and Rein- herz). The studies appear to be consistent in this regard. However, even in the very large data sets the differences between children of adolescent and older mothers are very small.
204 The studies are also consistent in suggesting that there may be important indirect effects: through family structure, maternal edu- cation and family size. An early birth is associated with a greater probability that the family will be headed by a single parent, that the mother will complete less schooling and that there will be a larger number of children. And these factors have also been shown to have effects on the cognitive development and achievement of the child. Schooling appears to be the most consistently important of these, with family structure a close second, although not all of these have appear- ed in the same way in all models. More work could be done comparing the relative contributions of these three factors. It is especially important to look more at the contribution of schooling, since it (and family size) is most subject to manipulation. Since there is very little understanding about exactly what schooling contributes to an individual's capabilities, more work is needed to define what it is about the amount of schooling the mother completes that improves the cognitive ability and performance of her children. Differential school completion may simply reflect differential motivation or capabilities, for example. An important issue is that of identifying the age of the mother at which effects on the child are most severe, for example, among younger teen or older teen mothers. Most studies show that age has effects that are continuous. That is the negative effects on children de- crease gradually as mother's age increases; there is no sharp line distinguishing the intelligence or achievement of a child of a 17 versus 18 year old mother, or a child of a 19 year old from that of a 20 year old. Thus it is not possible to draw sharp age of mother distinctions in child outcomes. In fact, one study (Moore et al., 1985) suggests that, if anything, outcomes for children of very youngest mothers may be slightly above those expected. This may be due to the likelihood that the girl's mother participates in the rearing of the child. One studies suggest that such participation improves child outcomes (Mednick and Baker, 1980; Field, 1984~. Effects of mothers' age at first birth on the socio-emotional development of their children have been found, but appear to be very weak. Several studies found that children of adolescent childbearers are at risk of social impairment and mild behavior disorders, par- ticularly undercontrol of behavior. The pattern differs between the sexes, however. One study showed boys more likely to show rebellious- ness, aggression or under control of anger while girls showed fearful- ness and other "neurotics behaviors (Marecek, 19879:204.5~. Another study (Mednick and Baker, 1980) found just the reverse, with daughters of early childbearers exhibiting greater aggressiveness and impulsivity while sons exhibited withdrawal, fearfulness and feelings of inferior- ity. As with cognitive outcomes, most effects are indirect, which, according to Mednick and Baker, operate through family structure. That is, children of young mothers experience unstable family situa- tions, which are associated with problem behaviors in their children. Again, what is needed here is an adequate explanation for the effects of family structure on child outcomes.
205 Only one study shows differences in mothering behaviors between adolescent and older mothers. Finally, there appears to be no consis- tent relationship between mother's age at first birth and child abuse and neglect net of differential socioeconomic status of the family. Notes 1 The definition of death at different stages of life as used in this chapter are the following: Fetal death: 20+ weeks of gestation Neonatal death: infant less than 28 days of age Infant death: infant 28 days to 1 year old Perinatal death: from 28 weeks of gestation through either the first 27 days of life or the first week of life. 2 The U.S. Collaborative Perinatal Project consisted of the complete population or random samples of all patients qualifying for prenatal care in the 12 participating centers during 6 years of intake, 1966- 1972. The study is not representative of all prenatal care patients as the particular hospitals selected, primarily teaching hospitals associated with medical schools, tended to be located in predominantly low income inner city areas and attracted low income clientele. As a result, black and low income prenatal care patients are overrepresented in the study. Such patients are at higher risk of poor pregnancy out- comes to begin with. These hospitals may also have attracted (or had referred) more of those clients with potential pregnancy problems. 3 Two groups with different definitions were used because the age of the respondent's parents was obtained in 5 year age categories, rather than by single year of age. This grouping of age restricted the abi1- ity of researchers to infer age at first birth. The method described was used to obtain the best approximation of teenage versus older childbearing. For more information see Card (1978~. 4 "Socioability. is defined by responses to three items: 1) re- action to school in the first year, 2) ease in making friends (at present), and 3) how much trouble the child was to bring up. 5 Crowding in the home is simply the number of people per room in the house, directly coded. 6 '~Cultured" is a subtest of a personality inventory on the Project Talent Data inventory. 7 Mother's contentment is a scale based on the following items: mother's attitude toward child, mother's isolation, family isolation, mother's overall contentment, and mother's acceptance of her situation.
206 8 Mother's orderliness is a scale based on the following items: 1) dress, 2) grooming, 3) home appearance, and 4) home furnishings. 9 Stability of family is representated by the total number of family constellations since birth of index child, recoded on a five point scale, with a score of 5 including all cases with 5 or more constel- lations.
