The Son Also Rises Page 14

  The evidence of similarly slow rates of social mobility in modern Sweden, preindustrial Sweden, medieval England, modern England, and the United States is thus at variance with the human-capital account of intergenerational mobility. Instead it seems that social status is transmitted within families independently of the resources available to parents. This raises the possibility that it is nature, much more than nurture, that propagates social status so persistently across the generations. Bryan Caplan is potentially correct when he concludes, “While healthy, smart, happy, successful, virtuous parents tend to have matching offspring, the reason is largely nature, not nurture.”8

  Demography and Social Mobility

  One clear prediction of the human-capital theory is that, other things being equal, the more children parents have, the poorer the children’s outcomes. The more children there are, the fewer family resources can be devoted to bolstering the human capital of each. Over time, however, there have been remarkable differences in the correlation of fertility and social status. Sometimes it has been strongly positive, at other times strongly negative.

  Currently, in high-income societies such as the United Kingdom, Sweden, and the United States, the correlation is relatively weak, with high-status parents having as many children as lower-status parents, or modestly fewer. But in other historical periods there was a strong negative association between social status and fertility, as in England for parents who married between 1890 and 1960. High-status families had much lower fertility than those of low status. In contrast, in the preindustrial world, fertility was typically strongly positively associated with status. In England before 1780, this effect was so strong that the wealthiest parents had twice as many children as the average family. Between these two periods, for parents who married between 1780 and 1880, there is no association between fertility and social status.9

  Figure 7.2 illustrates the number of surviving children for men who first married between 1500 and 1780 and those who married between 1780 and 1879, classified by their wealth at death. Marriages of high-status people before 1780 in England led to six births on average, with four children surviving to adulthood. At death, the richest men in the earlier period left more than four children; the poorest of the testators (in the middle of the overall status distribution since only richer men were probated) left only half as many. But for men marrying in the later period, there is no correlation of fertility with wealth.

  The lineage of Charles Darwin is a nice illustration of how large the families of the middle and upper classes could be in preindustrial England. He descended from a line of successful and prosperous forebears. His great-grandfather Robert Darwin (1682–1754) produced seven children, all of whom survived to adulthood. His grandfather Erasmus (1731–1802) produced fifteen children (born to two wives and two mistresses), twelve of whom survived to adulthood. His father, Robert Waring (1766–1848), produced six children, all of whom survived to adulthood.10

  FIGURE 7.2. Surviving children per father by wealth at death.

  In a social environment where all these children had to be privately educated, dowries needed to be provided for daughters, and estates were divided among children at death, human-capital theory would predict that the heedless fecundity of the English social elites of these years would lead to rapid downward social mobility. The lower classes of preindustrial society, producing only modestly more than two surviving children per family on average, would be able to concentrate resources on the care and education of their offspring and see them rise rapidly on the social ladder.

  In contrast, by 1880 in England, upper-class men seem to have produced far fewer children than those of the middle or lower classes. Indeed, from 1880 to 1940, the richest English families seem to have been dying out. Based on the rare-surname samples of chapter 5, the upper-class males produced, on average, fewer than two children who survived to adulthood. At the middle and bottom of society, however, men were producing an average of 2.5–3 children who survived to adulthood, in reversal of the pattern observed before 1780. Figure 7.3 shows, by twenty-year periods, the estimated total number of children surviving to adulthood per adult male for two wealth cohorts: initial rich and initial poor or average-wealth rare surnames. Fertility for the richer lineage is consistently less than that of the poorer in the years 1800–1959.

  FIGURE 7.3. Surviving children per adult male by twenty-year period, rich and poor.

  This major change in the relationship between fertility and status can be illustrated again by the Darwin family. Charles Darwin (figure 7.4), marrying in 1839, had ten children, though only seven survived childhood. These seven children produced only nine grandchildren, an average of only 1.3 per child. (This figure is unusually low for this era, but there was great randomness in individual fertility.) The nine grandchildren produced in turn only twenty great-grandchildren, 2.2 per grandchild. This figure was less than the population average for this period. The great-grandchildren, born on average in 1918, produced 28 great-great-grandchildren, 1.4 each.11 Thus by the time of the last generation, born around 1918, average family size for this still rather elite group had fallen to substantially less than replacement fertility. The Darwin lineage failed to maintain itself in genetic terms.

  FIGURE 7.4. Charles Darwin, 1881.

  Interestingly, with respect to social mobility rates, the twenty-seven adult great-great grandchildren of Charles Darwin, born on average nearly 150 years after Darwin, are still a surprisingly distinguished cohort. Eleven are notable enough to have Wikipedia pages, or the like, such as Times obituaries, devoted to them. They include six university professors, four authors, a painter, three medical doctors, a well-known conservationist, and a film director (now also an organic farmer).12

  But we see no signs that social mobility rates in England slowed as the upper-class groups produced fewer children. Instead, as chapter 5 shows, the intergenerational correlation of status remained constant for education and wealth. By implication, human-capital effects on social mobility must be modest. Status is strongly inherited within families mainly through genetic or cultural transmission, or both.

  Biology versus Culture

  If nature dominates nurture in the transmission of status, to what extent is the transmission genetic as opposed to cultural? The evidence presented here cannot answer this question. The best we can do is suggest some tests that would rule out genetic transmission as a significant cause of the high intergenerational correlation of status. There is an extensive list of these tests, which are discussed here and in subsequent chapters.

  First, we can ask whether social mobility mimics processes that we know to be largely genetically driven, such as the intergenerational transmission of height in affluent societies. That is, does social status show a constant rate of regression to the mean regardless of a family’s position in the status distribution? Or is social mobility higher at one or the other end of the status distribution?

  To illustrate the nature of biological transmission of traits such as height, which are the product of many different genes acting in combination, we can consider the data from Francis Galton’s famous study of the connection between the heights of parents and children. Presented to the Royal Society in 1885, this study introduced the concept of regression to the mean. Figure 7.5 shows Galton’s 928 observations, grouped into seven clusters of average parent heights (at one-inch intervals for the five central groups).13 Also shown is the best linear fit of the 928 observations, where the implied persistence rate is 0.64. All across the parent height distribution, the observations lie close to the fitted straight line. One persistence rate predicts intergenerational height mobility for those close to the mean as well as for the extremes of short and tall.

  In comparison, consider the transmission of wealth across generations. Simon Boserup, Wojceich Kopczuk, and Claus Kreiner have assembled a wonderful data set from Danish tax records that allows them to compare the wealth of 1,155,564 children with that of their parents.14 The Danish state appar
ently keeps its Big Brother eye on the economic fortunes of its citizens. The only limitation of this data is that because the wealth of parents comes from the years 1997–99 and that of the children from 2009–11, the children’s wealth is observed much earlier in the life cycle than that of parents. But the authors control for the age profile of wealth in making the comparison.

  FIGURE 7.5. Francis Galton’s observations of biological inheritance of height.

  The huge size of the Danish wealth data set means that the authors can divide the parents into percentiles and look at the average wealth of children for one hundred sets of parents, measured again as a percentile of the child wealth distribution. Other than the top and bottom 3 or 4 percent of parental wealth, the picture has the same linear character as that for height inheritance. One persistence rate, 0.20, describes inheritance across the middle 90 percent of the distribution (figure 7.6).

  The greatest deviation appears in the bottom 4 percent of parental wealth, where the children are much richer than we would expect. But the parents at the bottom of the distribution in Denmark have negative wealth—that is, debt. This suggests not chronic, grinding poverty (no one, after all, lends much to the truly poor), but more likely indebtedness to finance a business venture or training. The fact that this is not truly the bottom of the wealth distribution explains the breakdown of the stable relationship.

  Children in the top 3 percent of the parental-wealth distribution also show slightly greater wealth inheritance. Although this effect is statistically significant, it represents only modest deviations from the single persistence rate in real terms: the persistence rate implied for the top percentile is 0.24, as opposed to 0.20 for the rest of the distribution. For the second percentile, it is 0.23, and for the third 0.22.

  FIGURE 7.6. Social inheritance of wealth, Denmark, 1997–2012.

  This result, showing the same intergenerational correlation all across the status distribution, is found also in England. For the period 1858–2012, we can measure social mobility by looking at the representation of surnames among Oxford and Cambridge students, which typically describes the upper 1 percent of the educational distribution. We can also measure social mobility by looking at probate rates: the probated elite was typically 15–45 percent of the population, much closer to median status. The estimated mobility rates are the same from these two sources.

  There is at least one exception to this rule of uniform persistence rates across the status distribution. A study by Anders Björklund and others found that in Sweden, although the overall correlation of income across one generation was 0.26, for the top 0.1 percent of the income distribution, it was much higher, at 0.9.15 However, their study found that this effect was caused largely by the exceptionally high levels of wealth among the sons of the very high-income fathers. There was no unusually strong transmission among the top 0.1 percent of educational attainment, high cognitive abilities, or high noncognitive abilities. Thus while there may be exceptions to the rule of a constant persistence rate across the status distribution, it may only be found for the very wealthy, and only for wealth transmission (not for other status indicators).

  One argument for a potentially important role for genetic transmission in the inheritance of underlying social status is the tendency of status indicators to regress to the mean. Genetic processes, unlike other inheritance processes, have built into them an inherent tendency for characteristics to so regress. Why would the transmission of cultural traits also show such a consistent tendency to regress toward the mean? Even if family culture were transmitted with some error between generations, as long as positive errors were as likely as negative errors, the result would be increased dispersion of outcomes for an elite group, not consistent regression to the social mean.16

  If there were an important genetic influence on intergenerational correlations, furthermore, groups that marry endogamously (within the group) would not regress to the mean in social status. Suppose, for example, we were to take everyone over six feet tall in the United States and decree that from then on they could only mate with other people in this group and their descendants. The first generation of descendants would regress toward mean height, because the six-foot-and-above club includes many people whose genotypic height is below their phenotypic height. But after this first generation, there would be no further regression of the height of the descendants toward the mean. For a genetic trait like this, endogamy would ensure persistence. If genetics is important in transmitting social status, the degree of endogamy is an important controller of the rate of social mobility.

  For even if marriage is perfectly assortative, those in an elite group who choose marriage partners from the general population experience more downward mobility than those who marry within the elite. This is because the partners from the general population, even if they have observed characteristics identical to those of alternative partners from within the elite, have a greater random component to their status on average, since they are drawn from a population with a lower mean underlying status.

  If, to the contrary, we think that the advantage of elite families is a cultural trait, then endogamous marriage would not lead to any more faithful a transmission of that trait than exogamous marriage to someone with the same observed characteristics as the elite group. We show below that endogamy is associated with a complete absence or a slowing of the process of regression to the mean for elite groups.

  In different social systems where marriage is exogamous, are rates of social mobility the same? Again, if genetic transmission of innate abilities is the key driver of status, then we would expect it to be constant.

  Is social mobility Markov in character?17 That is, is all the information useful for predicting the status of the next generation contained in the parent generation, or does the more extended lineage matter in determining outcomes for children? If genetics is the most important element, the process has to be Markov.

  Chapter 6 presents important evidence from the Oxford and Cambridge cohorts showing that the history of an elite group does not seem to affect its subsequent social mobility. It also shows that the extended lineage can appear to matter even within a framework where the true process is Markov. Such a process, as specified by the law of mobility proposed here, further implies that mobility rates for any group must be constant over time. If the longer history is not significant, every period will look the same as any other. Thus persistence rates must be constant across generations. All of these characteristics are observed in the mobility data over longer periods for Sweden and England.

  Finally, if genetics is an important vehicle of social and economic success, then elite and underclass groups should always be formed as a selection from the upper or lower tails of a larger population. No entire population will become elite or underclass through ideological or cultural transformation. If genes are important carriers of social competence, elite groups will be formed only as a selection from a larger population.

  Under these conditions, individual families will become elite as a result of a series of random accidents. Below we test whether that prediction is borne out by the trajectories of several generations of such families. But if the world is characterized by persistent regression to the mean, balanced only by random shocks to maintain status dispersion, how do large social groups end up systematically above or below the mean in the first place? If the differences between groups are genetic as opposed to cultural, one process that could produce such differences would be selective affiliation to a social group by people at the top or the bottom of the status distribution. Again we examine below whether there is any historical evidence of such processes at work.

  1 This approach was first formalized in Becker and Tomes 1979, 1986. Goldberger (1989) argued that the Becker-Tomes economic model did not imply anything distinct from the simple regression to the mean in all human characteristics posited by Galton (1889). Mulligan (1999) tried to find features of inheritance that would disprove the Galton hypothesis, but with
little success.

  2 Gary Solon recently formulated a simple model of the contribution of each of these paths of inheritance that concludes that the intergenerational correlation of earnings, assuming a constant variance of earnings, will be , where τ is the parent-child correlation of abilities independent of investments in children, and γ is the elasticity of earnings with respect to human capital investment. If such investments have no effect on income, then γ = 0, parents do not invest, and by = τ, from the biological and cultural correlation. But if γ > 0, by > τ so that investment increases child income (Solon 2013).

  3 Using the Solon 2013 model, for example, if correlation of incomes over one generation is 0.47, over two generations it will be 0.16, and over three 0.06. Table 6.3 shows that the correlation of wealth in England is in fact 0.43 across one generation and 0.26 over three generations.

  4 My parents, for example, paid nothing for my thirteen years of elementary and secondary education in state schools. Nor did they pay fees or maintenance for the four years I spent at Cambridge University, in those days a deluxe education that included a servant to make your bed and clean your room.

  5 Recent evidence suggests that attending private school improves the A-level results for a student of a given ability over those of an equivalent student at a state school. This conclusion is derived, however, from the fact that students from state schools perform better than those from private schools with the same A-level scores once admitted to university. Thus private students gain some advantage in admissions, but it is eroded in terms of degree results.

  6 In 2013–14 the most expensive school in the United States was the Lawrenceville School in New Jersey, with fees of $44,885 for day students (“The 50 Most Expensive Private High Schools” 2013).