When marriage is endogamous within an elite group, however, high status can be maintained forever. Witness the Brahmin class in Bengal, or the Copts of Egypt. Conversely, endogamous marriage can condemn low-status groups such as the Muslims of West Bengal to perpetual deprivation.
These findings imply that to maximize mobility in a society, we want assortment in marriage to be based just on observed current status. If religious or ethnic background, or skin color, is correlated with social status at the group level, and marriage rates are much higher within religious, ethnic, and skin-color groups, social mobility will be slowed.
Countering Charles Murray’s concerns about marriage becoming more assortative in modern America, endogamous marriage is clearly in decline. Thus the National Jewish Population Survey of 2001 found increasing rates of intermarriage for Jews in the United States. Before 1970 only 13 percent of Jews married non-Jews. By 1991–2001 the rate was 45 percent.10 Rates of exogamy for Jews in the United States have risen to very high levels. In a strange irony, the achievement in the United States, finally, of a society largely free of discrimination against the Jewish population will eventually end a near two-thousand-year tradition of unusual Jewish social and intellectual achievement through the mechanism of greater intermarriage between the Jewish and gentile communities. Across U.S. society as a whole between 1980 and 2008, using the U.S. Census Bureau definitions of race and ethnicity, rates of exogamous marriage rose from 7 percent to 15 percent.11
If the way to produce children of the highest possible social phenotype is to find a partner of the highest possible social genotype, the path is clear for those whose aim in life is to produce the highest-achieving progeny possible. To discover the likely underlying social genotype of your potential partner, you need to observe not just their characteristics but also the characteristics of all their relatives. What is the social phenotype of their siblings and their parents? And what is the observed status of their grandparents and cousins?
The point here is not that any of these relatives will contribute anything directly to the social and economic success of your child. As far as can be observed, they will not. But the social status of the relatives indicates the likely underlying social status of your potential mate. This social genotype, rather than the observed social phenotype, is what your children will inherit.
These observations of the status of relatives can be formed into an aggregate with predictable weightings. If social status turns out to be mainly genetically determined, as for heights, we can also determine the weights to attach to each relative for the best prediction. If mating is not assortative, for example, then, in line with the simple model of table 15.2, the potential partner gets a weight of one. Their siblings and parents thus get a weight of one-half.12 Grandparents and aunts and uncles get one-quarter. Great-grandparents and cousins get a weight of one-eighth. However, since mating is highly assortative, the shared genetics of siblings, parents, grandparents, aunts, uncles, cousins, and so on is correspondingly much greater than suggested here. The entire lineage becomes strongly predictive of the underlying status of a potential marriage partner.
In line with this reasoning, a recent study in Japan examined the effects of the educational attainment of grandparents, aunts, and uncles on both sides of a family on children’s probability of going to university. Controlling for the parents’ education, there was a positive correlation between the education level of all four sets of relatives and the child’s probability of attending university.13
As noted in chapter 6, for a group of more than four thousand people in England with rare surnames, we know their wealth at death as well as all their familial connections over more than four generations (1858–2012). These data allow for the estimation of the correlation of wealth at death not just for parent and child but also for great-grandparent and child, cousins, and even second cousins. What is remarkable about this wealth information is the persistence of wealth connections with increasingly distant relatives. The parent-child correlation in wealth averages 0.43 and the sibling correlation 0.56. But the correlation for cousins, who in genetic terms are only one-quarter as related by descent as siblings, is still 0.34. And second cousins, one-sixteenth as related by descent, still have a correlation of 0.22. If genetics underlies social status, then mating must be highly assortative, so that second cousins are much more closely connected than might be expected.14
This implies that even very distant relatives are surprisingly closely related in terms of social status. That information can be used to predict the likely outcomes for the children of anyone in this lineage. It is this fact that underlies cases such as that of the Darwin family, mentioned in chapter 7, in which the twenty-seven adult great-great-grandchildren of Charles Darwin still form a surprisingly distinguished cohort.
All this implies that if the weighted score for the relatives is as high as for your potential mate, who is of high status, his or her underlying social genotype is as high as the observed phenotype. For the purpose of producing high-quality children—and for this purpose alone—this potential partner is a bargain on the marriage market. If the weighted score of the relatives is even higher than that of the potential partner, then he or she is a marital fire sale. Conversely, if the relatives are, on average, of lower status, this marriage is unlikely to produce children with social potential as high as the partner’s, because the partner’s social phenotype is better than the genotype.
Additional information about the likely outcomes for your offspring can be drawn from factors such as the potential mate’s ethnic or social group. The more the individual deviates positively from the average social phenotype of that group, the more likely their current status is to be the product of accident, higher than their underlying social genotype. The more they fall below the average for the group, the more likely it is that this status is the result of chance: the person’s underlying social genotype is likely to be better adapted for success.15
Suppose you are faced with a choice of two marriage partners, both of whom have a high-status phenotype. They are both graduates from elite colleges and have PhD’s in philosophy, for example, or both are board certified in rhinoplasty. But one partner is of Ashkenazi Jewish background and the other of New France descent. Then the predicted status of your children will be higher if you select the Jewish partner.
Since Coptic surnames are those that stand out in figure 13.10 as the highest-status group in the United States, all else being equal, if you want high-status offspring, find yourself a partner named Girgis, Boutros, or Shenouda. Chinese and black African surnames also stand out as particularly high status. So again, all else being equal, choose Chen over Churchill, Okafer over Olson.
In Selfish Reasons to Have More Kids, Caplan points out correctly that upper-class parents pointlessly invest too much time in the rearing of their children. In his view, genetics is what matters, so you might as well have more children, invest less in each, and enjoy being a parent more. That all seems sensible and humane.
Caplan does, however, address the stark corollary outlined above. If genetics determines child outcomes, then we can determine just from lineage which potential partners have (on average) the best genes. So the current competition to produce high-status offspring will be displaced by competition to mate with someone of the highest genetic potential. A better, more humane, less competitive, social world than ours may exist, but it is not obvious how we will attain it in a world where people have such strong aspirations for the social success of their own children.
1 Murdoch 1973, 98.
2 “She’s Warm, Easy to Talk To” 2011.
3 “Private Schools Are Expected to Drop a Dreaded Entrance Test” 2013.
4 In my second year as an assistant professor at Stanford University, I was assigned the task of mentoring six freshmen. Each appeared on paper to have an incredible range of interests for an eighteen-year-old: chess club, debate club, history club, running team, volunteering with homeless shelters. I
soon discovered that these supposed interests were just an artifact of the U.S. college admission process, adopted to flesh out the application forms and discarded as soon as they had worked their magic.
5 Caplan 2011.
6 Murray 2012.
7 The empirical evidence for increased sorting is actually weak. See, for example, Kremer 1997, 126, which reports a modest decline in the correlation of years of education of spouses between 1940 and 1990.
8 Love of course plays its part, but the wisdom of the ages is that reciprocal love flourishes best between socially matched partners.
9 This does not imply that the social genotype is actually derived from genetics, just that it behaves in a way that mimics genetic transmission of characteristics.
10 United Jewish Communities 2003, table 14.
11 Passel, Wang, and Taylor 2010.
12 Because of dominance effects, genetically transmitted traits are slightly more highly correlated between siblings than between parents and children, so the weighting should be somewhat greater.
13 Aramaki 2013.
14 Cummins and Clark 2013.
15 Given the findings above suggesting the importance of genetics in predicting outcomes for upper-class children, the same considerations would apply for those seeking donors or eggs or sperm for in-vitro fertilization. A recent study of the implicit market for human eggs found that despite guidelines from the American Society of Reproductive Medicine that recommend a fixed level of compensation for donors “to avoid putting a price on human gametes or selectively valuing particular human traits,” compensation for donation was strongly correlated with the average SAT scores for admission to the colleges that the potential donors attended (Levine 2010, 28–31).
APPENDIX 1: MEASURING SOCIAL MOBILITY
INTERGENERATIONAL SOCIAL MOBILITY is a staple of sociology and economics. The preferred tool of sociologists in the study of mobility, because social classes are not easily assigned a numerical status value, is the transition matrix. Parents and children are divided into ranked groups according to social class, occupation, income, or wealth. The standard occupational classification used in the United Kingdom until recently, for example, placed people into six categories:
A.
Higher managerial, administrative, or professional workers
B.
Intermediate managerial, administrative, or professional workers
C1.
Supervisory or clerical and junior managerial, administrative, or professional workers
C2.
Skilled manual workers
D.
Semiskilled and unskilled manual workers
E.
Casual or lowest-grade workers, pensioners, and others who depend on the state for their income
To measure father-son mobility, for example, each father and each son is assigned a status. The transition matrix shows the fractional distribution of outcomes for fathers of each status category, as in table A1.1 (where the numbers are hypothetical, chosen purely to illustrate the appearance of a typical transition matrix). Each row shows the probability of a son achieving a certain status given the father’s status. The numbers in each row add up to one. The table shows that in this example the chances of a son of a father of the lowest class ending up in the highest class, and vice versa, are low.
TABLE A1.1. Sample transition matrix
Table A1.2 shows the case of complete immobility, in which occupational status of all sons is the same status as the fathers.’ In contrast, table A1.3 shows complete mobility. The distribution of sons’ occupational status is the same for all ranks of fathers, and thus the fathers’ status provides no information about the sons.’
TABLE A1.2. Sample transition matrix showing no mobility
TABLE A1.3. Sample transition matrix showing complete mobility
Although such transition matrices offer the most complete description of social mobility in any society, they can be hard to interpret and compare. How much mobility does table A1.1 imply, for example? Is it closer to table A1.2, showing no mobility, or table A1.3, showing complete mobility? Another approach to measuring mobility, typically favored by economists and psychologists, is to rank aspects of social status—such as income, wealth, years of education, cognitive skills, and longevity—on a numerical scale. Even occupations can be represented in this way by assigning to each occupation a status score corresponding, for example, to the average earnings of each occupation or the average years of schooling required for each occupation.
If we are measuring mobility by comparing the earnings of fathers and their sons, we might observe the pattern pictured in figure A1.1. The line that best fits this pattern of data will be of the form
where y is the measure of status, vt is some random component, and t indexes the initial generation. Then b measures the persistence of status over a generation. For a given sample of fathers and sons, b could potentially have any value. A b of 0 indicates no persistence of status: no prediction can be made about sons’ status from the fathers’. The larger is b, the greater the predictive power of fathers’ status for sons’ status.
FIGURE A1.1. Earnings of fathers versus earnings of sons.
However, if the status measure has constant variance across generations, as is typically the case for societies as a whole, then b has special properties. In this case, b is also the intergenerational correlation of y and has a value between −1 and 1. In figure A1.1, which is drawn with constant variance, the slope of the line that best fits these observations, b, describes the intergenerational persistence rate of earnings. In this case it is 0.4. (1 − b) is thus the obverse, the rate of social mobility. As can be seen, with a b of 0.4, sons’ earnings can vary substantially from the fathers.’ For fathers with mean earnings of $45,000, sons’ earnings range from $24,000 to $63,000. Here b describes just the systematic components of inheritance. The lower is b, the more important are the random components.
When a measure of status has constant variance across generations, b2 measures the share of variance predictable at birth. The reason for this is that if σ2 measures the variance of the status measure y, and measures the variance of the random component in status, then, from equation A1.1,
Thus random components explain a share of the current variance of status of (1 − b2) and inheritance the other share of variance, b2. That is also why b has to lie between −1 and 1 in this case. Figure A1.2 shows what happens as b approaches 1. In this case, for a stable variance of status, it has to be the case that the random component in status becomes zero. Child status is perfectly predictable from parent status.
This intergenerational correlation is the simplified measure of mobility employed throughout this book. This simplification is not appropriate if social mobility rates vary at different points on the social scale, as some have argued. But one of the arguments of the book is that social mobility rates seem to be constant across the whole range of social status. This measure also requires that we assign a cardinal measure to social status at all social ranks.1
Because of data limitations, standard measures of social mobility often focus only on fathers and sons. But people, of course, have two parents. Women’s earnings, education, and wealth have become increasingly important to the social status of families in recent generations, and mothers’ status also contributes to their children’s outcomes, independently of fathers’ status on these measures. Does the traditional focus on fathers produce distorted estimates of intergenerational correlations?
FIGURE A1.2. Social mobility when b = 1.
Suppose mating were completely assortative with respect to social position. Then the intergenerational correlation of fathers and sons, or fathers and daughters, with respect to earnings, wealth, or education would be the same as if we took the average of fathers’ and mothers’ status as the measure for the earlier generation. So the conventional b measures would still summarize overall social mobility. Because mating is not completely assortative, these individua
l b measures tend to overestimate overall social mobility. However, even if mating were completely random, the correlation of children’s characteristics with the average of the parents’ characteristics would still just be 1.4 times the individual correlation.2 Assuming a correlation of 0.5 in the characteristics of the parents on any measure, the correlation between children’s characteristics and the average of parent characteristics is 1.15 times the single correlation. This is only a little higher than the single-parent correlations typically measured.
The simplified measure b used in this book, the intergenerational correlation of characteristics, makes it very easy to compare mobility rates across societies and across different measures of social status. It also has a simple natural interpretation. Conventional estimates of these intergenerational correlations suggest that modern societies exhibit high rates of social mobility for any particular measure of status. Thus the intergenerational correlation for such attributes, including features that we think of as largely biologically inherited in high-income societies (such as height), is typically in the range 0.13–0.54 for a single parent. Even when this figure is increased to account for inheritance from both parents, the typical correlation between parents and children in income, education, wealth, IQ, height, body mass index, and longevity is only 0.25–0.75.3 This implies that typically only 6–50 percent of all variation in these characteristics among children is predictable from the characteristics of parents. Parents of extreme characteristics typically see their children revert toward the mean by substantial amounts.
The Son Also Rises Page 29