The Son Also Rises Read online

  Nina Benner, a reporter for Swedish Radio, tells a nicely illustrative story from her own family of how such surname changes took place. Her grandfather and his four brothers changed their surname from Andersson to Benner in 1916, when her grandfather was sixteen. His eldest brother was studying to become a physician, and his professor made it clear that Andersson wasn’t a suitable name in that profession. The name Benner stems from the small village of Bennebo, where her great-grandfather grew up.

  FIGURE 2.6. Percentage of men born in 1950–51 with surnames ending in -berg and -son, by age at death.

  The incidence of other Swedish names, however, remains constant among men in different age cohorts. As figure 2.6 shows, of men born in 1950–51, the percentage holding surnames ending in -berg (mountain) was the same among those dying before age 10 as among those dying after age 50. These topographical surnames can be used as a standard for measuring social mobility rates.

  Surnames and Current Earnings and Wealth

  Since the two sets of elite surnames, noble and latinized, were established before 1800, we would expect, given the rapid rates of social mobility reported for Sweden in the current and previous generations, that these surnames would have regressed completely to mean social status. They would not differ in any way from the average surname in Sweden. Their connotation of exalted status would have been totally lost.

  One way we can measure the status of different surnames, and also the distribution of status, in modern Sweden is from tax records. This information is publicly available and is even sold commercially under the slogan “Know what your neighbors earn.”

  Figure 2.7 shows for Stockholm kommun (municipality) some of the information for the noble surname Leijonhufvud. The publications conveniently give detailed addresses for each taxpayer. The first column of numbers shows earned income in Swedish kroner. The second shows capital income. The tax returns reveal clearly that the expected social entropy in Sweden has not occurred. Individuals with noble and latinized surnames have higher taxable incomes, both earned and capital incomes, than those with the common surname Andersson. The differences are not huge, but they are quite clear. Thus for six kommuns in the Stockholm region, the average taxable income of people with noble surnames in 2008 was 44 percent greater than for those named Andersson, and 27 percent greater for those with latinized surnames than for Anderssons.11

  Analysis of the tax data shows that those with noble and latinized surnames have higher incomes not only because there are more of them at the upper end of the income distribution but also because there are fewer of them at the bottom of the distribution. As figure 2.8 shows, the range of income among individuals with these names is just as great as among Anderssons, but the mean of that range is higher in each case.

  The income level for the top 1 percent of taxable incomes in these six kommuns was two million Swedish kroner or more. Among those with noble surnames, 2.6 percent have incomes in this top 1 percent. Thus the relative representation of noble surnames among the income elite is 2.6: these names are 2.6 times as likely as the average surname to be in the top 1 percent of incomes. Such estimates of relative representation among elite groups for Sweden and throughout the book are used as a convenient measure of the social status of surnames.

  As long as the intergenerational correlation of income is less than one, the mean income of those with noble surnames must be approaching the overall mean income. As this happens, the relative representation of noble surnames at the top of the income distribution will decline, and the relative representation of such surnames at the bottom will increase. Appendix 2 details how the speed with which the relative representation of surnames moves toward one at the upper tail of the status distribution gives us a measure of the intergenerational correlation of status. Below are shown these calculations for four elite groups in Sweden: attorneys, physicians, university students, and members of the Royal Academies. These estimates present a unified picture of very slow social mobility in Sweden, now and earlier.

  FIGURE 2.7. Sample of published tax returns for Stockholm, 2008.

  FIGURE 2.8. Distribution of taxable income within surname groups, 2008.

  ATTORNEYS

  The Swedish Bar Association maintains a register of seven thousand member attorneys that records each member’s date of birth. A comparison of the frequency of surname types in this register with the frequency of surname types in the general population reveals significant mismatches. As figure 2.9 shows, the surnames held by titled nobles—counts and barons—appear in the register at nearly six times the rate they occur in the general population.12 Other overrepresented surnames include those associated with untitled nobles and latinized surnames, both appearing at about three times the expected rate (their share in the population). Surnames beginning with Lund- appear at the expected rate. In contrast, surnames ending in -son appear at half the expected rate.

  These results again imply that the distant past has a surprising effect on the present even in Sweden. Surnames that were differentiated socially in the eighteenth century remain so even ten generations later. Noble surnames have retained their ranking in the social hierarchy: the surnames of counts and barons carry still higher status than those of the untitled nobles.

  Members of the bar can be divided into two generations, those born between 1930 and 1959 and those born between 1960 and later. Figure 2.10 shows the relative representation of each surname type in these two cohorts. It reveals, first, that each of the surname types has been regressing toward the expected mean representation of one. Second, however, it shows that the rate of regression to the mean is slow. Even among attorneys born in 1960 and later, those qualifying within the past thirty years, there are substantial differences in the relative representation of surnames.

  The implied intergenerational correlation of occupational status from the attorney data by surname group is as follows: titled nobles, 0.79; untitled nobles, 0.72; latinized surnames, 0.71; and patronyms, 0.69. (Note, however, that interpreting intergenerational correlations for patronyms requires caution because significant numbers of people switched from the patronyms of their birth to other types of surnames. If it was mainly the more socially successful who changed their names in this way, the estimated correlation could be significantly higher than the true correlation.) The average intergenerational correlation reported for attorneys in table 2.1 is 0.73 for the three elite surname groups. The estimated correlations do differ by surname group, but because the numbers of attorneys in each surname group are typically less than fifty, these variations in estimated correlations could easily stem from chance alone.

  FIGURE 2.9. Relative representation of surname types among attorneys, 2012.

  FIGURE 2.10. Relative representation of surname types among attorneys, by birthdate, 2012.

  PHYSICIANS

  A second source for measuring social mobility rates is the list of physicians in Sweden registering first between 1890 and 2011, which covers four generations. Starting with currently registered physicians, we see in figure 2.11 the same differences in relative representation of surnames that we see among attorneys. The surnames of the three elite groups of the eighteenth century are overrepresented relative to their share of the population. Patronyms are greatly underrepresented.

  Analyzing surname types for Swedish physicians is complicated by the fact that a substantial proportion of currently registered physicians in Sweden are of foreign origin. Physicians with a medical license from any other European Union country can register in Sweden without further required training. Thus in 2007, almost one in five of all physicians registered in Sweden were trained abroad, including Swedes who attended foreign medical schools. But of those registering first in 2007, excluding Swedes trained in foreign medical schools, two of every five were foreign.13 One consequence is that even surnames such as Lund-, which have an average representation among attorneys, are underrepresented among physicians.

  FIGURE 2.11. Relative representation of surname types
among registered Swedish physicians, 2011.

  FIGURE 2.12. Relative representation of surname types among Swedish physicians, 1890–2011.

  To correct for this complication and calculate the relative representation of Swedish surname types among Swedish-born physicians in Sweden, it is assumed that all foreign physicians were registered in 1980 or later, and that the relative representation of the surnames Lund- and Berg- averaged one between 1980 and 2011. These assumptions imply that in this cohort, only 70 percent of all physicians are Swedish born—a reasonable estimate. The overall domestic physician population for these years is calculated accordingly. For the years before 1980, it is assumed that all registered physicians in Sweden were Swedish born.

  Figure 2.12 shows relative representation of the four surname types—titled noble, untitled noble, latinized, and patronyms—among physicians in thirty-year cohorts, by registration date, beginning in 1890. To make clearer what is happening with the patronyms, a logarithmic scale is used in figure 2.12. All three groups regress toward the mean, but their rate of regression is again very slow among all cohorts. Figure 2.13 shows the best-fitting relative representation for all those in the three high-status groups across the four generations. The estimated persistence rate in this case is 0.74, and the fit, as can be seen, is good. The rate of regression to the mean was no faster in the past thirty years than in earlier years. To a first approximation, it was the same as in the years before 1980.

  The corresponding persistence rate for the patronyms is similarly high, at 0.74. Again, however, we must be cautious about the estimate for patronyms. Because of the abandonment of patronyms, which was likely more common among the upwardly mobile, the intergenerational correlation estimated here may overestimate the persistence of status among those with patronym surnames. However, the persistence rate estimated for this group is the same as for the three elite surname groups.

  Thus the representation of surnames among both attorneys and physicians in Sweden suggests a similar pattern: social mobility in Sweden is much slower than the conventional estimates suggest, even for very recent generations. A second surprising finding from the surname distribution of Swedish physicians is that not only are true social mobility rates slower than conventionally estimated, but they are no faster now than they were in the early twentieth century. The enlargement of the political franchise and the institutions of the extensive welfare state of modern Sweden, including free university education and maintenance subsidies to students, have done nothing to increase rates of social mobility.

  FIGURE 2.13. Estimated persistence rate for Swedish physicians with elite surnames.

  FIGURE 2.14. Surnames of Uppsala students submitting master’s theses, 2000–2012.

  Educational Mobility, 1948–2012

  The ineffectiveness of free university education in increasing social mobility is borne out by patterns of surname distribution among university graduates, even in recent decades. Figure 2.14, for example, shows the relative representation of the surname groups among those completing master’s theses at Uppsala University from 2000 through 2012. Taking surnames of the form Lund- or Berg- as having an average representation, the noble and latinized surnames, largely originating before 1800, are again overrepresented by 60 to 80 percent. The most common patronyms appear at half their expected representation.14

  The differences between the elite surnames and patronyms among university graduates are less pronounced than among attorneys and physicians. But master’s degree programs, even at elite universities such as Uppsala, are less exclusive than the professions of attorney and physician. Indeed, based on the numbers of master’s theses submitted annually at Sweden’s most selective universities—Gothenburg, Lund, Stockholm, and Uppsala—we can predict that 8 percent of Swedes born in 1990 will complete a master’s thesis at one of these universities.15

  If a surname type, such as a latinized name, is at a relative representation of two for the top 8 percent of the population, then its relative representation among the top 1 percent (approximating the selectivity of the legal and medical professions) would be 2.8.16 Thus the information for university students is consistent with the evidence for physicians and attorneys in the most recent generations and suggests again that there is currently very slow regression to the mean for elite and underrepresented surnames.

  These data imply that if 8 percent of all twenty-two-year-olds in Sweden now get a master’s degree from one of these four elite universities, the rate for those with elite surnames is 13–14 percent. The status differences signaled by Swedish surnames will not end soon.

  There are extensive records of those enrolled at the only two Swedish universities established before 1954: Uppsala (founded in 1477) and Lund (founded in 1666). These records include the surnames of more than two thousand members of three student “nations” (dining and residence associations) at Uppsala between 1942 and 1966. These records show the relative representation of different surname types at Uppsala circa 1948 and circa 2008 (two generations later). Figure 2.15 shows, again, a clear convergence of all four groups toward the mean across these two generations.

  To calculate the intergenerational correlation for education implied by the data in figure 2.15, we need to take into account that Uppsala and Lund were much more elite institutions in the 1940s than in 2000–2012. The fraction of Swedes attending Uppsala and Lund in the late 1940s can be roughly estimated as still only 1 percent of the population, compared to an estimate for master’s theses of 8 percent of the population today. The estimated persistence rate for each of the three elite groups, allowing for this shift in the upper proportion of the population being observed at universities, is 0.72 for the titled noble surnames, 0.75 for the untitled noble surnames, and 0.57 for the latinized surnames.

  FIGURE 2.15. Relative representation of surnames at Uppsala University, 1948–2012.

  However, because the sample size for these surnames at Uppsala in the years 1942–66 is small, there is significant sampling error in these estimates. Combining these groups into one elite implies an overall intergenerational correlation across these two generations of 0.66. Yet the two subsequent generations of students matriculated after major reforms in 1977 that greatly expanded access to universities. Tuition is now free, and grants and loans are available to students to cover living costs.

  For the patronym surname group, here estimated on the basis of the surnames Andersson, Johansson, Karlson, and Nilsson, the implied intergenerational correlation, 0.87, is even lower. The caveats detailed above for such estimates apply here also.

  Educational Mobility, 1700–1908

  There are good data available on the surnames of Lund attendees for the period 1666–1908: sources include a register of all students for 1732 through 1830 and detailed biographies from a number of the student nations that all students had to enroll in. For Uppsala there is complete registry data for the period 1477–1817, but data from only one student nation for the period 1817–1902.

  Figure 2.16 shows the relative representation of latinized surnames at Lund by thirty-year cohorts, starting in 1700. In the first generation observed, 14 percent of Lund students had latinized surnames, compared with an estimated 0.13 percent of the general population. Such names were thus 122 times more common among students than in the general population. The share of latinized surnames among students fell to 1.1 percent by 1880–1909. They were 5.3 times as frequent among Lund students as among the general population. The pace of this decline in representation implies a high persistence of this group, however. The persistence rate estimated for 1700–1909 is 0.78, assuming that university students represented the top 0.5 percent of the status distribution.

  One complication in calculating persistence is surname changing. If students born with the surname Andersson were changing this to Wigonius as they entered the university elite, then persistence would be exaggerated. The biographical sources for some of the student nations at Lund and Uppsala, which list the parents’ surnames for
most students, allow us to estimate the fraction of latinized surnames newly adopted in each generation. Figure 2.17 shows what percentage of students in each generation inherited rather than adopted a latinized surname.17 Between 1730 and 1819, 96 percent of students acquired latinized names by inheritance. However, in the period 1820–1909, that proportion fell to 88 percent (even though, by design, these are all surnames that existed before 1800).18 The tendency of new members of the university-educated elite in nineteenth century Sweden to switch to such latinized surnames means that the persistence rate estimates for these years represent an upper bound. The true persistence rate is likely lower. Thus there is no good evidence of any decline in the persistence rate for status between preindustrial and modern Sweden, despite the enormous institutional changes that have taken place.

  FIGURE 2.16. Relative representation of latinized surnames, Lund and Uppsala university students, 1700–2012.

  FIGURE 2.17. Percentage of latinized surnames inherited, 1730–1908.

  A more elite group of academics than Lund and Uppsala students is the members of the various Royal Academies of Sweden. There are nine such academies. Comprehensive membership lists are available for the Swedish Academy of Sciences, founded in 1739; the Swedish Academy of Music, founded in 1771; and the Royal Academy, founded in 1786. Together these three have had nearly three thousand domestic members.

  Figure 2.18 shows the relative representation of latinized and noble surnames among the members of these three academies by thirty-year cohorts, starting in 1740 and ending in 2012. In the earliest period, such surnames were held by half of the members of the academies. By the last generation, this figure had declined to 4 percent. But these surnames in 2011 were held by only 0.7 percent of the general Swedish population, so they were still strongly overrepresented among academy members.