Journal Issue: School Readiness: Closing Racial and Ethnic Gaps Volume 15 Number 1 Spring 2005
Does Parenting Matter?
Despite all the studies reporting links between parenting and child well-being, we still need to question whether parenting matters.22 Our premise is as follows. Even though the literature is voluminous, it also has its limits, all of which comes down to the same problem: we do not know, in most cases, whether the so-called effects of parenting are caused by parental behavior or by something else that may complicate the causal link. We consider four different factors: family social, educational, and economic conditions; genetic similarities between parent and child; child characteristics; and other unmeasured characteristics (which we believe might be operating but have not measured, or do not know how to measure well). Although all four factors influence links between parenting and child well-being, they do not account completely for these links. (Another line of evidence supporting the premise that parents matter, reviewed later in this article, has to do with the potential of intervention programs to alter parenting.)
Parenting and Correlated Family Conditions
First, we know that parents differ in their social, economic, and educational backgrounds. And we know that variations in parenting are associated with such characteristics. The link between parental talking and child vocabulary is one example.23 Parents who talk a lot to their children, ask questions, use many different words, and discuss events are also more likely to be highly educated, to have high incomes, and to have few children, as well as to have children with large vocabularies. And these latter characteristics are themselves associated with child vocabulary. Thus in reality parental education might account for the link between parental talking and child language. If parents who talk a lot are more likely to be highly educated, we need to adjust for parental education to be sure the link between parental language use and child vocabulary is not inflated. It is relatively easy to measure parental education and make statistical adjustments to see if the link between parental talking and child vocabulary still exists, just as it is for other characteristics like family structure, income, parity (number of children), age, and the like. Studies that make such adjustments find that the link exists independent of parental education.
At the same time, the purpose of such studies is often to show how parental education, for example, influences children's language. Clearly, in that case, the parental education effect would have to translate into a specific parenting behavior, such as talking to the child. So we often consider parenting (in this case parental talking) to be a pathway through which parental education influences child language. That suggests two types of intervention strategies. One is indirect: to try to increase maternal education in the hope that more schooling would cause a mother to talk more to her child. The other is direct: to try to increase her talking with the child. The latter would target behavior directed toward the child (talking), rather than a more general characteristic of the adult (more education). The assumption is that it is possible to pinpoint the specific parenting behaviors that contribute to a specific child outcome. High levels of parental warmth, in the absence of much parental talking, for example, would not be expected to increase child vocabulary. Neither would parental monitoring, unless it involved lots of talking.
Parenting and Correlated Genetic Characteristics
Second, perhaps the most widely heralded causal issue is that parents and children are genetically related, which can, in part, account for links between parenting and child well-being. To continue with our earlier example, parents who talk a lot and have a large vocabulary are likely to have children who are predisposed toward language. That is, language facility is partly heritable.24 Even in the absence of parenting behavior, parent language test scores would be linked with child language scores.25 How can we tell to what extent the link is due to environment (here, language expressed to the child) and to what extent to genetics (here, the biological relationship between parent and child)? Studies informed by behavioral genetics are useful here.26 Two examples, one from studies of adopted children and the other from work with identical twins (monozygotic twins, whose genetic material is identical, so that any differences between them must be environmental), demonstrate that parenting influences child well-being, over and above genetic relatedness of parents and children.
Studies of adopted children show striking increases in cognitive abilities when the children leave institutional care to be placed with adoptive parents.27 Children in such studies, however, move from extremely deprived environments without consistent caregivers (orphanages) to stable, two-parent, largely middle- class homes. The studies speak to the powerful effect of having parents versus not having parents, but say little about the effects of varying levels of parenting behavior.
One study does address normal variation in parenting. Michel Duyme and colleagues identified a small sample of adopted children (fewer than seventy) from a review of more than 5,000 adoption cases in France.28 They selected all children between the ages of four and six who had been placed in prescreened adoptive homes, removed from their birth parents because of abuse or neglect, and put in foster care before their adoption. The children were given cognitive tests before their adoption and again between the ages of eleven and eighteen. Overall, the children showed striking gains in IQ test scores from early childhood to adolescence, from a mean score of 77 to 91 (14 points or almost one standard deviation on a test with a mean of 100 and a standard deviation of 15). The authors classified the adoptive households as low, middle, or high socioeconomic status (SES), based on paternal occupation. The gains were largest for those placed in high SES families (19 points) and smallest for those in the low SES families (8 points).29 The assumption is that the high SES families were providing more language, more teaching, and more materials, all of which facilitated the children's cognitive growth.
A study of children exposed to cocaine prenatally also illustrates the power of change in parenting.30 The study recruited more than 400 mothers following delivery. All the new mothers were considered at high risk for cocaine use; about half had biological indications of cocaine use when they and their infants were tested at delivery. When the children were four years old, researchers gave them an IQ test, observed them in their homes, and gave their caregivers a vocabulary test. In the group of children who had been exposed to cocaine before birth, only 55 percent were living with their biological mothers at the follow-up, as against 95 percent of those in the group that had not been exposed. The cocaine-exposed preschoolers living with their mothers or with a relative had significantly lower IQ scores than their counterparts who were living with an adoptive or foster care mother—even though (not surprisingly) the latter group had been exposed to more cocaine than those who were not removed from the mother. Furthermore, the IQ scores of the exposed children living with an adoptive or foster mother were comparable to those of the children who were not exposed to cocaine prenatally. For example, the share of the cocaine-exposed children living with their mothers who had IQ scores under 70 (the mild mental retardation range) was 25 percent, as against 10 percent for the exposed children who lived with nonrelatives and 16 percent for the nonexposed children. As might be expected, the homes of the groups differed; cocaine-exposed children living with adoptive or foster mothers had more stimulating environments, and their mothers had higher vocabulary scores, than the cocaine-exposed children living with their biological mothers or relatives.31
The second class of studies does not rely on change in parents (from orphanage to family or from biological to adoptive parent). Instead, it uses genetic similarity to delve into parental effects. In a sample of 500 five-year-old identical twins, mothers were asked to talk about each of their children. Mothers tended to describe one twin in more negative terms than the other. When the children were in elementary school, their teachers were asked to rate their behavior.32 Teachers reported that the twin for whom the mother had more negative feelings had more behavior problems than the other twin.33 Because the children had identical genetic endowments, it is highly likely that maternal behavior accounted for the differences in behavior problems between the twins.
Parenting and Correlated Child Characteristics
A third causal issue is that parenting behavior may be in part contingent on the behavior of the child. That is, not only does parenting affect child behavior, but also children can influence parents.34 We provide two examples, the first having to do with reading, the second with behavior problems.
Children of mothers who read to them frequently have large vocabularies, as countless studies have shown.35 In an evaluation of the Early Head Start Program, Helen Raikes and her colleagues have found the expected links between shared book reading and child vocabulary in more than 1,000 children seen at age fourteen months, twenty-four months, and thirty-six months, even after adjusting for differences in mothers' verbal abilities.36 (The adjustment is necessary because mothers with higher verbal abilities are likely to enjoy reading more than other mothers, which could influence their shared book reading with the child, and because language ability is partly heritable.) Of more interest is their exploration of the pathways through which language at age thirty-six months was influenced. More shared reading at fourteen months was linked with higher vocabulary scores at twenty-four months, which affected the amount of reading at twenty-four and thirty-six months. Thus, mothers whose children knew and used more words were reading more to these children as they developed, over and above their reading levels at fourteen months.
One of the best-known examples of child-to-parent effects is an intervention geared toward children with conduct disorders and their parents.37 Half the children participated in a family program, which was effective in that the children displayed less aggression after the intervention. But the positive impact on the children was primarily due to changes in parenting behavior. That is, the parents in the intervention group stopped reacting negatively to their children's aggressive behavior by learning other techniques for dealing with outbursts. In contrast, the parents in the control group did not alter their responses to their children's outbursts, and therefore the children's problem behavior showed no change.
The point here is that child characteristics can influence parenting. But the existence of differences among children themselves does not totally account for parenting effects on children
Parenting and Unmeasured Correlates
The final complicating causal issue involves possible correlates of parenting that have not been measured. Even studies that adjust for family conditions and child characteristics may fail to measure other sources of variation in parenting and children's school readiness, perhaps because of limits of cost or time or the lack of a reliable indicator.
One characteristic often associated with parenting and child outcomes is parents' mental health. Mothers who are diagnosed with clinical depression or as having high levels of depressive symptoms engage in less nurturance and more punitive discipline, as has been demonstrated countless times for preschoolers as well as older children.38 And these mothers' preschool children have more behavioral problems and (sometimes but not always) lower cognitive test scores.39 But even when analysts adjust for maternal depression, parenting still contributes to these indicators of child wellbeing. Indeed, maternal depression, as well as other measures of mental health (anxiety, irritability), is thought to act on children through its effect on parenting behavior.40 Vonnie McLoyd, Rand Conger, and their colleagues have proposed a family stress model that traces the pathways from low income, financial instability, and material stress through parental mental health to parenting to child outcomes.41 Jean Yeung and her colleagues have shown that this pathway is stronger for behavioral problems than for cognitive and language test scores in young children.42
Even if a study measures many potential correlates, it is impossible to be sure that it includes all that are relevant. So scholars use a variety of statistical techniques to minimize the likelihood that results are due to something besides parenting.43 But the most convincing evidence is gleaned from experiments where families enter a treatment or a control group through random assignment. We present evidence from experiments designed to test the efficacy of parenting programs later in the article.