FIGURE 15.1. Intelligence correlations with age for adoptees.
Studies looking at the outcomes in the adult lives of adopted children are rarer. They support the idea that biology dominates nurture but suggest that nurture has some effect. One elegant study is that of Bruce Sacerdote, who looked at various outcomes for Korean adoptees in the United States.7 These children were randomly assigned to approved families with varying degrees of education and economic resources. The adoptive families did not include those at the lower end of the income distribution: U.S. law required the adoptive families to have incomes of at least 125 percent of the poverty level. But in other respects the adoptive families spanned a wide range of income and education.
Table 15.1 shows the share of variation in each outcome for the Korean adoptees and their nonadopted siblings that is explained by nurture as opposed to nature. The proportion of outcomes explained by nurture is just the correlation between these outcomes for nonrelated adopted siblings, assuming that the family assignments were indeed random. The proportion explained by nature is derived from how much closer the correlations of biological siblings were for these same families.8 We can see that height is indeed largely biologically determined, whereas alcohol consumption is almost entirely socially determined.
TABLE 15.1. Proportion of outcomes explained by nature and nurture for Korean adoptees
Outcome
Proportion explained by nature
Proportion explained by nurture
Height
0.01
0.86
Family income
0.11
0.33
Four years of college
0.14
0.41
Smokes
0.15
0.27
Drinks alcohol
0.34
0.06
Selectivity of college
0.34
0.24
For characteristics other than height, adopted siblings were always significantly correlated, despite their absence of genetic connection. Their shared family environment had an influence. But family environment appears to have only a very modest influence on the later earnings of children. Genetic inheritance explains three times as much of children’s income variation as does family environment. Figure 15.2 shows an adopted child’s adult income relative to the adoptive parents’ income at the time of adoption. There is no connection. So the correlation of incomes for adopted siblings is due to aspects of their shared family environment other than parental resources. There is no sign here that giving extra income to families would result in higher incomes for the next generation.
For educational attainment, the correlation between children and their adoptive parents was higher than for income. But again, nature explains three times as much of the variation as does nurture. As figure 15.3 shows, the mother’s educational attainment has little relationship to the outcomes for adopted children. This confirms findings from a study in Norway on the effect of introducing compulsory schooling: additional years of education for parents do not in themselves predict more years of schooling for their children.9 Similarly, a study in Sweden has shown that for adopted children, years of education can be predicted from the educational attainment of both the biological parents and the adoptive parents. The effects of the adoptive parents are stronger in the Swedish case than in the case of the Korean adoptees. But the variation in biological parents’ education still explains twice as much variation in children’s educational attainment as that of the adoptive parents.10
FIGURE 15.2. Income for Korean adoptees versus parents’ income.
FIGURE 15.3. Education level for Korean adoptees and their mothers.
In the Korean adoptions study, families did play a substantial role in educational achievement in one respect: getting adopted children into more selective colleges (table 15.1). Here the correlations between unrelated sibling adoptees were substantial, and genetics played a subsidiary role. But the earnings outcomes suggest that getting into selective colleges had little effect on the future incomes of adoptees, for which the correlations were much lower.
These adoption studies suggest that even if we could make the familial environment of every child in the United States identical, we would reduce the intergenerational correlation of social outcomes by only modest amounts, perhaps one-quarter of existing values. Moreover, it is not clear that public policies can do much to change family environments in the ways that matter to the social outcomes for children. Public policy can change the amounts of income available to families and even the amounts of education parents receive. But the causal role that family income and parental education play in child outcomes is itself highly uncertain. Other elements of parental behavior that cannot be compensated for by public policy may be the crucial ones. And these other parental behaviors may well be associated with the genetics of parents. It may be impossible to reduce the influence of inheritance in determining social outcomes through government actions.
Confirming the idea that genetics potentially plays a substantial role in outcomes are the correlations in earnings reported in an interesting recent study of siblings of different types in Sweden. Assume that the correlation in earnings between siblings has separable contributions from shared environments and shared genetics. Assume also that the environment contribution to correlation for all siblings raised together is s, and for siblings raised apart it is zero. The genetic connection produces a correlation of g for those sharing all genes. Assume, finally, that mating is not assortative for the genes that matter to earnings.
This model has two strong simplifying assumptions. The assumption of zero environment correlation for siblings raised apart is unrealistic, but it turns out that this is not what mainly determines the fit of the model. The assumption of no assortative mating is more important. In this case it implies the correlation of adopted siblings should be s, half siblings reared together s + 0.25g, half siblings living apart 0.25g, full siblings and fraternal twins reared together s + 0.5g, full siblings reared apart 0.5g, and identical twins reared together s + g. Table 15.2 shows the predicted pattern of correlations under these assumptions, as well as the observed pattern. The authors report that the best fit under this model would be a family-environment contribution to correlation of only 0.02 and a genetic contribution of 0.26. Table 15.2 shows why these values come close to explaining the patterns observed. This would make genetics more than ten times as important as environment in explaining earnings outcomes.
The authors report that in statistical terms, this simple model fails to explain the observed correlations: the correct model of the correlations has to be different. But although it fails, what is interesting is how close it comes to succeeding. The effects of family environment are more important than this simple model implies, but any explanation of these various sibling correlations requires a much larger genetic component than has typically been assumed.
If social status is largely transmitted through inherited genes or familial cultures, then shocks to wealth should have a much smaller effect on social status over generations than wealth that is gained through some inborn higher level of social competence. It is difficult, however, to find instances of random shocks to wealth that are uncorrelated with the characteristics of recipients for which we can observe the effects on the next generation. In an interesting and ingenious study, Hoyt Bleakley of the University of Chicago and Joe Ferrie of Northwestern University document one such random shock to wealth and its generational consequences. The removal of the Cherokee from the eastern part of the United States, following the passage of the Indian Removal Act of 1830, opened up for distribution large parcels of land in northwest Georgia. The state of Georgia organized a lottery to distribute eighteen thousand 160-acre parcels of land in Cherokee County in 1832.11
TABLE 15.2. Earnings correlations between siblings of different types, Sweden, 1987–93
Adult males resident for at least three years in Georgia were eligibl
e to one draw in this lottery, and almost all eligible men entered. The winners constituted just under one-fifth of the adult male population. The parcels of land had an average value equal to the median wealth in Georgia by 1850. Further, the land could be immediately sold: the winners did not need to take possession themselves or to homestead their property. So the lottery prize was equivalent to a large cash transfer (equivalent to nearly $150,000 today) to a random selection of adult males in Georgia.12
Tracking winners and their sons through the United States censuses of 1850, 1870, and 1880, Bleakley and Ferrie show, first, that by 1850, winners were indeed richer on average than losers. The value of the allotted land by then averaged $900, and the average wealth of winners was $700 higher than that of losers. So the winners were able to retain much of the benefit of winning for at least some years.
However, when we look at the children of the winners in 1870 and 1880, we see little sign that the good fortune of their fathers significantly changed their life chances. They were no more literate than the children of losers. Their occupational status was no higher. Their own children in 1880 (the grandchildren of the 1832 winners) were again no more literate. Worse, they were significantly less likely to be enrolled in school than the grandchildren of the losers.13
The comparative wealth of the children of lottery winners and losers is harder to estimate precisely; data on children’s wealth are available only for 1870. Wealth is not statistically significantly higher for lottery winners’ children, but the variance is so great that we cannot rule out the possibility that the wealth gains from the lottery were indeed transferred to the children. What we do observe is that a substantial shock to wealth alone did little to change the social status of families in nineteenth-century Georgia.
In an ironic reversal of the Georgia lottery, another set of windfalls has been created recently by Indian gaming profits. A recent study of child mental health in rural North Carolina, by design, oversampled children from the Eastern Band of Cherokee Indians. A casino opened on the Eastern Cherokee reservation in 1997, midway through the study. From 1998 on, parents of Cherokee children in the study received annual lump-sum disbursements of casino profits. Relative to the average incomes of these families, typically less than $30,000 per year, these annual payments of $4,000–$8,000 were large and were expected to be ongoing. Since households did not reduce labor-force participation in response to the payments, these families became substantially better off. Indeed, their anticipated lifetime gains were of a similar relative magnitude to those of Georgia land lottery winners.
The youngest children in the study experienced these family-income gains from age 14 onward, and there was information on their outcomes up to age 21. At age 18, conditional on their graduation from high school, they became eligible for their own annual payment of $4,000 from the casino proceeds.14 What effect did these payments have on the lives of these Cherokee children?
The study concludes that for those not living in poverty, the effects were limited. There was no measurable change in any educational outcomes, including high-school graduation rates, by age 21—despite the immediate cash gains that the children got for three years just for completing high school. Beneficiaries were less likely to commit minor crimes (but not major crimes) or sell drugs. Among those who were living in poverty before the income supplement, children in the youngest cohort were much more likely to graduate from high school, and they completed one to two more years of education by age 21. As with children from the more prosperous families, there were fewer minor crimes and less drug selling.15
In one respect these results confirm the “large effects” of exogenous shocks to income, as the authors describe them. But in another respect, they show that the influence of wealth on outcomes is limited. Children of families above the poverty line—78 percent of U.S. children—seem to gain little from a significant exogenous boost in family income or a cash incentive to complete high school. And while for the poorer children the combination of gains in family income and the extra cash inducement to complete high school did lead to more education, it is not certain that this will yield gains in living conditions for these children later in life. At age 21, we are still observing outcomes early in the life cycle.
The absence of such effects at most levels of income is confirmed by a study of the effects of the Norwegian oil boom on educational outcomes for children. Incomes for all families in some regions of Norway increased during the 1970s as the result of increased demand for labor, driven by the exploitation of North Sea oil. The study compared children born between 1967 and 1969 in Rogaland, a county on the southern coast of Norway with extensive connection to oil extraction, with those born in counties unaffected by the boom. The income gains in Rogaland had no effect on the years of education achieved by children there.16
The author of the study, Katrine Løken, wonders if this lack of difference is a result of Nordic social-welfare programs: “Norway has very high public investment in children. All students in higher education are eligible for grants and subsidies from the government to finance their education. … It is possible that family income would have an impact on children’s educational attainment if all of these government interventions were removed.”17 But if this were the case, then social mobility rates in countries with similarly extensive educational and social-welfare programs, such as modern Sweden, should be much higher than elsewhere. We have seen, however, that underlying social mobility rates in Sweden are just as low as in more laissez-faire economies.
At least one study, however, found much stronger effects of income shocks on children’s outcomes. Phil Oreopoulos and my colleagues Marianne Page and Ann Stevens looked at the effects on children’s income after fathers lost their jobs when a firm closed in Canada. Such job losses have permanent effects on workers’ future earnings, and can be regarded as random negative shocks to income that affect individual workers. Each of the male workers chosen for the study had a son between twelve and fourteen years old at the time of the closing.
The study found that six years after the firm closed, the family income was still an average of 9 percent lower than before the closure. Thus the sons experienced a period of lower income in their youth, relative to a control group. By age 28, the sons affected by the closure had incomes 8 percent lower than those of sons in the control families. The income shock propagated across generations with an intergenerational correlation close to one.
This is a very puzzling result. Income differences associated with differences in parents’ education, personality, drive, and capabilities were, as would be expected, weakly inherited by these sons. A doubling of a father’s income from these sources would be associated with a rise of less than half in a child’s earnings. But income changes from the random shock of a firm closing, which are transmitted to children only through limited pathways, such as reduced financing for education, are almost fully inherited.18 This is not a demonstration of the independent effects of income changes on children’s prospects: there must be some mechanism other than income at work here to cause such significant effects.
On balance, for the bulk of families in the middle of the status distribution, feasible social interventions, such as income transfers or boosts to education, appear unlikely to significantly change child outcomes. James Heckman and others show evidence that among the most disadvantaged families, early brain development can be substantially influenced by childhood environment.19 Supporters of social interventions, such as Heckman, point to two well-known randomized trials of the effects of preschool programs: the Perry Pre-School Program and the Abecedarian Project. Both demonstrated statistically and quantitatively significant effects on the subsequent adult lives of the participants. The economic gains to the participants and society as a whole per dollar spent were substantial.20
But no matter how efficacious these programs were, there is no strong evidence that the widespread adoption of early interventions such as these would substantially improve outcomes a
t the bottom end of the status distribution. A recent evaluation of the large U.S. Head Start Program, which incorporates aspects of these two preschool programs, found that at the end of the third grade, randomly chosen Head Start participants showed no better cognitive or noncognitive performance than the randomly chosen nonparticipants.21 The $10 billion or so annually spent trying to improve outcomes for one million poorer children in the United States appears to have no measurable lasting benefits. The programs may still have effects on adult outcomes, but in the Perry and Abecedarian programs that did have successful adult outcomes, program effects were always evident at younger ages. So although some interventions may be shown to be beneficial, the ones actually in place in the United States are of dubious value.
Suppose that in the “good society,” a society that looked more like Sweden than the United States, we equalized the social environment for all children. This would produce a period of increased social mobility and a general gain in social and economic outcomes for children at the lower end of the status distribution. Inequalities in education, income, wealth, and health would all narrow.
However, in the new equilibrium, after this transition, what would happen to rates of social mobility? The upper and lower classes would now be sorted purely based on their genetic heritage. Would social mobility rates be higher or lower in this good society than in our current, imperfect one? It is impossible to say. It would all depend on how strongly the genes that determine social success are inherited compared to family ethos and behaviors. But we cannot predict that in the good society, inheritance of status would be any weaker than it is now. Thus in the good society, it is quite possible that social and economic outcomes would be just as predictable as they are in the imperfect and unjust societies that we observe. Low rates of social mobility are not in themselves indicators of social failure or misallocation of potential talent.
The Son Also Rises Page 27