Journal Issue: Opportunity in America Volume 16 Number 2 Fall 2006
How Family Background Affects Educational Attainment
Theoretically if everyone, rich or poor, faces the same cost and reaps the same benefit from additional schooling, educational attainment should not differ by family background. In the real world, however, years of schooling completed, and educational achievement more generally, vary widely by family background. To illustrate we turn to data from the National Education Longitudinal Study (NELS) of 1988, which followed more than 20,000 eighth graders from 1988 through 1994 (for many, their sophomore year of college). This survey has rich information both about the educational experiences of the students and about their parents and schools. Figure 2 shows how students' educational achievements vary by family background. We have divided the students' families into four even groups (quartiles) based on an index of socioeconomic status. Those in the lowest quartile are the most disadvantaged, while those in the highest quartile are the most advantaged. The average family income in the lowest quartile is about $27,000 (in 2004 dollars), with an average family size of 4.6. In the second quartile the average family income is about $48,000 (average family size of 4.4); in the third quartile it is about $69,000 (average family size of 4.3); and in the fourth quartile it is nearly $110,000 (average family size of 4.4).
As the figure shows, children from families in the highest quartile have higher average test scores and are more likely never to have been held back a grade than children from families in the lowest quartile. Children from families in the top quartile are also more likely never to drop out of high school, and therefore much more likely to have a high school diploma six years after they entered the eighth grade.
Although these patterns are striking, it is not clear they reflect the causal effect of family background on a child's educational achievement. Inherited genetic ability confounds attempts to study the link between family background and educational achievement because to the extent that ability or intelligence is heritable, genetics helps determine whether children are successful in school. For example, evidence suggests that people with lower observed ability earn lower wages than those with higher ability.3 Thus, less able people will have lower socioeconomic status than more able people. Further, more able people likely find it less costly to get more schooling, in the sense that it is easier for them to master the knowledge required at each new step of school than it is for an individual of similar background but with lower ability. If it is also true that ability is genetically determined, then less able parents whose socioeconomic status is low will also have less able children who will get less schooling than the children of more able parents whose socioeconomic status is high. In this example, the heritability of ability combined with the link between ability and educational achievement means that low innate ability explains both the parents' low socioeconomic status and the children's lesser educational achievement.
To disentangle the effects of genetic makeup (which is not malleable) and family background (which is likely more malleable) on educational attainment, a researcher would ideally conduct an experiment. The experiment would begin with the random assignment of one group of children to disadvantaged families and another group to more advantaged families—without regard to the children's “innate” ability. Because assignment to families would be random, there would be no link between the genetic ability of the children and that of the parents. On average the only difference between the two groups of children would be their family background. Years later the researcher could compare the educational attainment of these children. With a large enough sample, differences between the two groups would provide a credible estimate of how much family background causally affects educational attainment.
In this experiment what the researcher wants to control is the wealth (or socioeconomic status more generally) of the family in which the child was raised. The researcher does not attempt to control which schools the children attended, whether the children had access to good medical care, their families' parenting practices, or other aspects of their lives that undoubtedly affect their educational attainment. Why not? Because the researcher is not interested in the effect of randomly assigning students to families of different backgrounds, assuming that the families do everything else the same.
Another way to see this is to consider possible policy implications. Suppose a new public policy aiming to increase the educational attainment of children were to give $10,000 to each family whose income fell below, say, the national poverty line. The policy's intent would not be for parents to put the money into the bank and not spend it on their children. Rather, the intent would be for them to use the money to buy nutritious food, enroll their children in better schools, purchase supplementary educational materials, get access to better medical care, or purchase other materials that would help their children's educational success. That is, the key policy question is not whether wealth or social advantage affects educational attainment per se, but whether the behaviors and resources made possible by that wealth and social advantage affect educational attainment.
In a study that comes close to the ideal experiment just described, Bruce Sacerdote examines the educational attainment of children adopted from South Korea who were randomly assigned to U.S. adoptive families.4 Because the children are adopted and randomly assigned to their families, there should be no relationship between the mother's innate ability and the child's innate ability; thus any relationship between the mother's educational attainment and that of the children is causal. Because many of these families also have biological children, Sacerdote compares the link between a mother's schooling and a child's schooling for adopted and biological children and estimates how much the mothers' educational attainment determines that of the biological children. He calculates that only 23 percent of schooling transmitted from mother to child is the direct effect of the mother's education, suggesting a very large role for genetics. In contrast, he finds that nurture plays a much larger role than nature in transmitting health habits such as drinking and smoking: these habits pass along to biological and adopted children at equal rates. Sacerdote, an economist, notes that under very strong assumptions his finding means that 23 percent of educational attainment is determined by environment, implying that up to 77 percent is determined by nature. Most psychologists who examine how genetics affects academic achievement in young children find smaller estimates, in the range of 30 to 40 percent.5 Some also argue that adoption studies overstate the importance of genetics because adoptive families are not representative of families in the general population.6
Researchers have used other strategies to estimate the extent to which family income determines children's educational achievement. Again, because they cannot assume that family income is unrelated to other factors (such as inherited ability) that determine both children's socioeconomic status and their educational attainment, they must look for changes in family income that are unrelated to family characteristics such as whether the parents are highly educated or have high genetic “ability.” Pamela Morris, Greg Duncan, and Christopher Rodrigues take advantage of variations in family income caused by experimental welfare programs in the United States and Canada during the 1990s to examine how income affects children's achievement.7 The welfare programs were all designed to increase work, and several were also designed to increase income, either through wage supplements or by allowing participants to keep more of their welfare payments when they went to work. Because no direct family or child services (such as parenting classes or child care subsidies) were provided, any changes in children's achievement must be attributable to changes in their parents' em ployment, income, and welfare receipt generated by random assignment to the different programs.
Morris, Duncan, and Rodrigues look at how these differences in income (all generated by random assignment) affect children's achievement. They find that a $1,000 increase in annual income (over three to five years) increases achievement by 6 percent of a standard deviation for children who are two to five years old. However, it has no effect on achievement for older children (six to nine years old and ten to fifteen years old). The cost and benefit of the increased income for preschool-aged children compare favorably to the cost and benefit of direct educational interventions such as reducing class size. (In one experiment, Alan Krueger and Diane Whitmore Schanzenbach find that class-size reductions costing $9,200 per pupil for grades K–3 increased children's achievement by 13 percent of a standard deviation.)8
Addressing the question of how changes in family income affect children's academic attainment in yet another way, Gordon Dahl and Lance Lochner use the fact that increases over the past twenty years in the earned income tax credit for working families have caused increases in family income to examine how child achievement is affected.9 Families with two children with earned income of, say, $10,000 in 1993 would have been eligible for a tax credit of $1,511. That same family would have been eligible for a credit of $2,528 in 1994 and $3,110 in 1995. Thus with no change in nominal earned income, total family income would have increased each year. Did the added money improve student test scores?
Dahl and Lochner find that it did. A $1,000 increase in income raised math and reading scores by 2 to 4 percent of a standard deviation-- an improvement large enough to close roughly 3 to 5 percent of the achievement gap between children in the bottom income quartile (average family income of $14,214 in 2000 dollars) and those in the top quartile (average income $81,137).10 Furthermore, when Dahl and Lochner estimate how income affected test scores for various subgroups, they find even larger effects for children from disadvantaged families, who are more likely to receive the maximum increase in income.
Overall, the evidence suggests that parental socioeconomic status has a causal effect on children's educational outcomes. But the studies noted cannot identify precisely how increases in parental education or income improve children's educational outcomes. Economic theory suggests that people stay in school until the costs of doing so (direct costs as well as forgone earnings and the psychological costs of being in school) outweigh the benefits. Thus, if the children of advantaged families stay in school longer, it must be because they receive greater benefits or face lower costs than do less advantaged children (for example, forgone earnings are less important to a wealthy family than to a poor family). In the next sections, we investigate why the relationship between family background and educational attainment may be so strong.