The Son Also Rises Read online

  The small size of this group compared to other groups examined above raises the possibility of significant sampling error. Taking these academies to represent the top 0.1 percent of Swedish society, the implied persistence parameter over these 273 years is 0.87. There is little sign of an increased rate of regression to the mean for the entrants to the academies for the period 1980–2012 compared to 1950–79. The estimated persistence for elite surnames is still 0.83 for this last generation.

  Figure 2.18 also shows the relative representation of patronyms among academy members. Such surnames are still strongly underrepresented, but they have shown a slow but steady convergence toward proportional representation. However, the implied persistence rate for this group is 0.87, close to that for the elite surnames. The caveats above on such estimates apply here also.

  FIGURE 2.18. Elite surnames in the Swedish royal academies, 1740–2012.

  Interpretation

  Despite conventional estimates, our analyses suggest that Sweden appears to be a society with low rates of intergenerational mobility for income, occupation, and education. Moreover, rates of social mobility seem no higher now, in the modern inclusive, social-democratic Swedish state, than in the preindustrial era. Why do the results presented here differ so much from those of conventional mobility studies?

  One possible explanation is that the surname evidence presented here relates to the top 0.1 percent to 8 percent of the status distribution, whereas conventional studies look at mobility across the entire population. Could there be high persistence of status at the upper extreme of the distribution, but greater social mobility for 99 percent or more of families in Sweden? Björklund, Roine, and Waldenström (2012), for example, find an expected overall income intergenerational correlation for Swedish men of only 0.26. But for the top 0.1 percent of the income distribution, their estimate is 0.9.

  Assuming such a large disparity in status persistence rates to exist, consider what would happen to families with the surnames of the eighteenth-century elite—the noble and latinized surnames. Once descendants of such families fell out of the top 1 percent, the rapid social mobility in the bottom 99 percent of the distribution would cause their status to quickly fall to the social mean. Distribution of elite surnames across measures of status such as income would no longer be normal and might even be bimodal, with a cluster at the top and then a near-normal distribution around the mean (as in figure 2.19). In particular, there would be no marked deficiency of originally elite surnames at the bottom of the distribution.

  But, as is evident from the tax data, noble and latinized names are as underrepresented at the bottom of the income and wealth distribution as they are overrepresented at the top. Even when they fall out of the top 1 percent in various measures of status, they are still experiencing markedly slower rates of downward mobility than would be expected. Noble and latinized surnames are conspicuously absent from the bottom of the distribution of income, for example.

  FIGURE 2.19. Hypothetical bimodal status distribution of elite surnames.

  FIGURE 2.20. Frequency of noble surnames and the name Andersson relative to expected frequency, by kommun average house price, 2011.

  Another illustration of the absence of elite surnames in the bottom parts of the income and wealth distribution comes from evidence from the published tax returns on where people are living. The horizontal axis of figure 2.20 shows average house prices for six kommuns in the Stockholm region in 2008. Prices in Nacka are double those of Haninge and Botkyrka. Also shown for each kommun is the frequency of noble surnames relative to their frequency in the population of Swedish ancestral origin, and the same for the name Andersson.19 Taxpayers with noble surnames are found at half the expected frequency in poorer kommuns such as Botkyrka and at twice the expected frequency in rich kommuns such as Täby. Thus the relative frequency of noble surnames is four times higher in the highest-income than in the lowest-income kommuns. Along with the occupational segregation of surname types in Sweden, there is a social segregation.

  Figure 2.20 also shows, however, that taxpayers named Andersson are not underrepresented only in rich areas such as Täby; they are also underrepresented in poorer kommuns. There are only about half as many Anderssons filing tax returns as would be expected from their population share, and half as many more of the noble surnames in returns than would be expected from the population share. The differences in reported taxable incomes understate average income differences between these surname groups.

  Conclusions

  Generalized and long-term social mobility rates in Sweden in recent years are much lower than the rates reported in standard two-generation studies of the intergenerational correlation of income or education. Indeed, rates of long-run social mobility are so low that the eighteenth-century elite in Sweden have persisted to the present day as a relatively privileged group. There is little evidence that intergenerational mobility rates have increased within the last two or three generations compared to rates in the preindustrial era. The persistence rate for underlying social status is as high as 0.70–0.80. The implied social mobility rates are as low as those of modern England or the United States (see chapters 3 and 5).

  Nearly one hundred years of Swedish social democracy has created a more economically equal society, but it has been unable to change the underlying rate of social mobility. The strong intergenerational persistence of status in a country after many years of generous public provision of opportunities and funding for education, at a level similar to that of other countries without such equalizing expenditures, suggests that the forces that determine intergenerational mobility must be fundamental to the formation and functioning of families. These may be forces that are impossible to alter.

  1 Lindahl et al. 2012, table 5.

  2 When intergenerational mobility is estimated using surnames, the length between generations has to be specified. In this book, for convenience, it is always taken as thirty years.

  3 Appendix 2 provides the technical details of this calculation.

  4 Riddarhuset 2012.

  5 Boyce 2005, 154.

  6 There was concern that disreputable people were adopting noble surnames.

  7 Because a new latinized name only recently adopted would not have time to grow to have forty holders by 2011, this criterion narrows names to those in existence much earlier.

  8 As Watson and Galton famously demonstrated, rare surnames over many generations tend to either die out or survive at a relatively higher frequency (Watson and Galton 1875).

  9 FamilySearch, n.d.

  10 This trend is due in part to a substantial increase in children born to immigrants in this period. Of males born in 2000 who died by 2009, one in ten had a Muslim first name, and another one in ten had a name suggesting an immigrant parent.

  11 2008 tax returns for the kommuns of Botkyrka, Huddinge, Haninge, Nacka, Stockholm, and Täby (Kalenderförlaget 2008a,b,c).

  12 For details of how these data were treated, see Clark 2013.

  13 “Every Other Doctor in Sweden from Abroad” 2009.

  14 See again Clark 2013 for details of these calculations.

  15 For details of this calculation, see Clark 2013.

  16 Clark 2013.

  17 In the first thirty-year period, 1700–1729, a larger fraction of students adopted latinized surnames, but this trend does not affect the calculated intergenerational correlation, which is affected only by the fraction of students who changed their surnames later.

  18 Some acquired latinized names from their mothers.

  19 Clark 2013 details how this calculation was performed.

  THREE

  The United States

  Land of Opportunity

  THIS CHAPTER EXAMINES SOCIAL MOBILITY RATES in the United States using the same methods as those applied to Sweden, again using surnames as the diagnostic. Measured by surname distributions, U.S. social mobility rates are also low. But, importantly in light of recent political debates, they are no lower than Sweden�
��s, and they show no sign of a decline in recent years.

  Using surnames, we identify a variety of elite and underclass groups whose mobility can be tracked across three generations. The elite groups are the descendants of Ashkenazi and Sephardic Jews, the descendants of the wealthy individuals as of 1923–24 who had rare surnames, the descendants of individuals with rare surnames who graduated from Ivy League universities in or before 1850, and people of Japanese descent. The underclass groups are Native Americans, black Africans whose ancestors came to the United States before the Civil War, and, surprisingly, the U.S. descendants of the French settlers who came to the French colonies of North America between 1604 and 1759.

  This chapter examines mobility rates across three generations—1920–1949, 1950–1979, and 1980–2012—using two sources. The first is the American Medical Association’s Directory of Physicians in the United States. This lists more than a million licensed physicians in the United States. About a quarter of them are of foreign origin. But because the AMA directory records the medical school attended by each physician, it can identify physicians who are likely of domestic birth. Here Caribbean and some Central American medical schools are counted as domestic.

  As a guard against the fraudulent impersonation of retired and deceased physicians, the current directory lists many physicians who completed medical school as early as the 1920s. The directory thus presents a view of the surname composition of the U.S. medical profession from the 1920s to the present, though with small numbers of observations in the earliest decades.

  The second source is lists of licensed attorneys, with year of licensure. Attorneys are licensed by state agencies, with no central register, so this information is contained in fifty state websites. Using a selection of twenty-five more populous states allows similar measures of social mobility rates for attorneys as for physicians. Attorneys are a less exclusive elite than physicians.

  As in Sweden, the measure of average social status used for each surname group, and for birth cohorts of each group, is their relative representation among physicians and attorneys. That is just the frequency of these names among physicians or attorneys in relation to their frequency in the general population. If this ratio is greater than one, this surname group constitutes an elite. If it is less than one, it forms an underclass.

  Elite and Underclass Surnames

  To measure social mobility using surnames requires estimates of the frequency of surnames in the United States by cohort. The main source for this information is a file produced by the U.S. Census Bureau giving the frequency of all surnames that appear at least one hundred times in the 2000 census. This source also records the fraction of holders of each surname declaring themselves members of the census categories white, black, Asian/Pacific Islander, Native American, and Hispanic.1

  To infer the frequency of surnames that appear less than a hundred times in the U.S. census in 2000, the Social Security Administration’s Death Index is employed.2 This lists those who have died in the United States in 1962 and later by name and year of birth. To estimate surname frequencies by birth cohorts, earlier surname frequencies are estimated using the Death Index or information on the size of ethnic groups over time from the U.S. censuses. The Death Index correction, however, is biased by the differential death rates of social groups at each age.

  The analysis below uses the following surname groups.

  FIGURE 3.1. Relative representation of surname types among physicians.

  ASHKENAZI JEWS

  This group consists of individuals with the surnames Cohen, Goldberg, Goldman, Goldstein, Katz, Lewin, Levin, Rabinowitz, and variants, who numbered nearly three hundred thousand in 2000. These surnames are common in New York City, the area of the greatest Jewish population share in the United States. However, in the 2000 census, nearly 4 percent of people bearing these surnames declared themselves black (5.5 percent for Cohen). This mostly stems not from intermarriage but from black Americans’ independently adopting these surnames because of their Biblical resonance. These names appear among physicians at a rate nearly six times higher than in the general population, the highest frequency of any domestic surname group, as shown in figure 3.1.3

  SEPHARDIC JEWS

  Certain surnames are associated with the Sephardic Jewish community: Abecassis, Baruch, Saltiel, Salomone, Sarfaty, Sasson, and variants. These are much less common than the Ashkenazi surnames, being held by only five thousand people in 2000—not enough to measure social mobility rates. But Sephardic Jewish names appear among physicians at a rate more than four times higher than among the general population, making them the second most elite group among long-established populations in the United States (see figure 3.1).

  1923–24 RICH

  These surnames were chosen from those appearing in the New York Times lists of federal taxpayers in New York in 1923 and 1924. Congress passed a provision for public inspection of income-tax returns in 1924. Before the effective repeal of that provision in 1926, major newspapers across the country printed thousands of names and tax payments for the tax years 1923 and 1924. The Times alone reported the tax payments of more than thirty thousand people.

  The sample was formed of rarer surnames held by at least one taxpayer reported per ten births recorded prior to 1900 in the Death Index. The modal such surname was held by less than one hundred people in 2000. The ten most common of these names were Vanderbilt (1,717), Roosevelt (961), Winthrop (727), Colgate (616), Guggenheim (512), Sonn (480), Bloomingdale (467), Plaut (455), Kempner (436), and Pruyn (421). This group thus includes descendants of the Puritan settlers of New England, the colonial Dutch, and the Jewish populations.

  The number of people with these surnames in 2000 is more than one hundred thousand. These surnames appear among physicians at a rate nearly three times higher than among the general population.

  JAPANESE AMERICANS

  As figure 3.1 shows, Japanese surnames also signal an elite group, appearing at a higher than expected frequency among domestically trained physicians. The overrepresentation seems to apply to all Japanese surnames, since the thirty most common Japanese surnames all have an above-average representation among physicians. The surnames used here are the most common Japanese surnames in the United States, representing 145,000 people in 2000. They appear at more than double the expected frequency. But because the mobility behavior of this group is unusual, the discussion of it is deferred until the end of the chapter.

  “IVY LEAGUE” GRADUATES, 1650–1850

  These surnames are the rare surnames of graduates of Brown University, Columbia University, Dartmouth College, Harvard University, Princeton University, Rutgers University, the University of Pennsylvania, the College of William and Mary, and Yale University from 1850 or earlier.4 The bulk of the sample graduated in the early nineteenth century, a period when university education expanded and the population grew. The surnames selected from this group were those held by fewer than three hundred people in the 2000 census and fewer than two hundred in the 1850 census.5 This set consists of a thousand surnames, with an average estimated frequency in 2000 of only 83.

  Of these surnames only a very few are familiar, for example, Rutgers and Rensselaer. The rest are obscure surnames of largely English, Dutch, German, and Irish origin. These surnames still appear among physicians at more than twice the expected rate.

  NEW FRANCE SETTLERS

  These are surnames derived mainly from the descendants of the colonists of New France. They arrived in the United States through the takeover of parts of Acadia by the English, the expulsion of Acadians to Louisiana, and the movement of French Canadians to New England in the years 1865–1920 to take up employment in factories. The surnames chosen were those more common in Canada than in France. A further restriction was that at least 90 percent of holders of these names in 2000 declared themselves as white and fewer than 5 percent declared themselves black.

  Examples include Gagnon, whose distribution across Canada and the United States is shown in
figure 3.2. The highest incidence of Gagnon is in New Brunswick, Canada (part of the old French colony of Acadia), and in New Hampshire and Maine in the United States, where it accounts for 0.2 percent of surnames. While its frequency in Canada overall is 633 per million, in France it is a rare surname, with a frequency of 15 per million. These surnames in the United States represent an underclass, occurring among physicians at less than three-fifths of the expected rate. There are nearly seven hundred thousand people in this sample. The most common of these names, each with between forty and sixteen thousand holders in 2000, are Hebert, Cote, Gagnon, Bergeron, Boucher, Delong, and Pelletier.

  FIGURE 3.2. Map of the distribution in North America of the surname Gagnon, 2012.

  BLACK AMERICANS

  This group is identified as surnames of English or German origin of which 87 percent of more of the holders identified as black in the 2000 census. The English-or-German criterion enabled us to exclude surnames belonging to more recent immigrant groups of black African origin who are actually social elites within the United States.6 Of the four hundred thousand people in this group, about two-fifths have one name, Washington, presumably because it was widely adopted by emancipated slaves lacking surnames after the Civil War.7 The other predominantly black surnames in this sample include, in order of frequency, Smalls, Cooks, Gadson, Merriweather, Broadnax, Boykins, and Pettaway.