The Son Also Rises Read online

  Many of the other surnames in this group sound classically English and presumably were adopted in the slavery era from masters whose own families died out or left few descendants. One such surname is Doyley, which is recorded in the Domesday Book of 1088 as that of a substantial Norman landlord in England. Another is Rockingham, a high-status locative surname from medieval England. Others are more whimsical, such as Idlebird.

  On average, 91 percent of people with these surnames declared themselves black in the 2000 census, and 4 percent declared themselves white, with the majority of the rest identifying as mixed race. These surnames appear among physicians at one-third of their frequency in the population in 2000.

  NATIVE AMERICANS

  These are surnames of which 90 percent or more of the holders in 2000 identified themselves as Native American. Two names, Begay(e) and Yazzie, account for about two-fifths of this population. Many of the less common Native American surnames are quite distinctive: Manygoats, Roanhorse, Goldtooth, Fasthorse, Yellowman, Twobulls, Bitsilly, and Smallcanyon all had more than two hundred holders in 2000. These surnames are concentrated in the southwest of the United States. The sample includes nearly eighty thousand such surnames. They occur at an extremely low rate among physicians, about 6 percent of the expected rate.

  The remarkable differences in the representation of these groups of surnames among physicians imply low social mobility rates in the United States. The rich 1923–24 taxpayers, for example, are at least two generations removed from the majority of the physicians recorded in the current AMA directory, and those who attended Ivy League universities in or before 1850 are four generations removed; yet their names remain more common than expected among physicians.

  As figure 3.3 shows, other surnames associated with national origins also show significant variations in relative representation among physicians. Names of Japanese origin, as noted above, are heavily overrepresented. Next in frequency come German, Scottish, Irish, Italian, Scandinavian, and Dutch surnames (in all cases, no more than 5 percent of the holders of these names in the sample are black). French names are markedly underrepresented, creating another puzzle whose consideration is deferred for the moment.

  FIGURE 3.3. Differences in ethnic surname representation among U.S.-trained physicians.

  Social Mobility, 1920–2012

  The rate of change of over- or underrepresentation of the surname groups identified above among physicians and attorneys across generations of thirty years can be used to estimate the underlying persistence rate of social status, as explained in appendix 2.

  The AMA directory reveals how many physicians of each surname group graduated in each of three thirty-year generations. To estimate the relative representation of the surnames among physicians requires just dividing the share of each surname group among physicians with its population share among those age 25 in the same generation.8

  The relative representation of each surname group in three generations completing medical school is shown in table 3.1 and graphed in figure 3.4. All five surname groups exhibit a general convergence toward a relative representation of one in the later two generations observed. But, as the graph shows and as the estimates of the underlying persistence rate for each group confirm, this is a slow process that, for a number of these groups, will not be complete for many generations.

  TABLE 3.1. Relative representation by surname groups among doctors, by generation

  FIGURE 3.4. Relative representation of surname types among physicians, by generation.

  In the earlier generations, both the Jewish and black surname groups diverge from the mean in their representation.9 For the Jewish surnames, the likely cause was the policy of many medical schools between 1918 and the 1950s to limit admissions of Jewish students. The tightening of these quotas in the 1930s led to a decline in the number of Jewish students in AMA-approved medical schools: there were 794 in the class of 1937 but only 477 in the class of 1940.10

  The lifting of anti-Jewish quotas in the 1950s is reflected in the directory data. Data for the 1930s and the 1940s show a substantial decline in Jewish surname overrepresentation for physicians completing medical school. In figure 3.8 below, which shows relative representation by decade, there is a rise in Jewish relative representation among medical school graduates from the 1950s to the 1970s.

  For the black surnames, there was a decline in relative representation in the 1940s and 1950s, though the numbers of black medical graduates in these decades is so small that this may just be a random fluctuation. The AMA in these years recognized only two medical schools catering primarily to black students, Howard University’s College of Medicine and Meharry Medical College. Its reluctance to accredit more such schools in an age when many other institutions discriminated against blacks could explain why there is no increase of black surnames in the AMA directory for the 1950s and 1960s.

  The maintenance even now of a relative representation among physicians well above one for the Jewish surnames, the 1923–24 rich surnames, and the pre-1850 Ivy League surnames implies slow long-run social mobility. Those completing medical school in 1980 and later would be typically three generations removed from the rich taxpayers of 1923–24 (their great-grandchildren). Similarly, those completing medical school in 1980 and later are at least four or five generations removed from the early Ivy League graduates. At the intergenerational correlation of 0.3–0.5 conventionally believed to apply to educational and occupational mobility in the United States, these more recent graduates should show little trace of any advantage enjoyed by their forebears.

  The estimate of the underlying rate of persistence in status using these data makes two assumptions, analogous to those made for Sweden. The first is that physicians represent the upper 0.5 percent of the occupational status distribution.11 The second is that occupational status within each surname group has a normal distribution, with the same dispersion for all groups. The difference for elite surnames and lower-class surnames is just that the distribution is shifted upward or downward.

  The highest-status group among the surname samples, those of Jewish origin, certainly exhibits a distribution of educational attainment that appears higher than the average for the United States (see figure 3.5). There are plenty of Jews with modest educational attainments: they just constitute a smaller share of the Jewish population than those with limited education do of the general population. The converse holds for the lower-status group, blacks (see figure 3.6). Black educational attainment in figure 3.6 looks as though it is shifted downward compared to the average.

  Given the assumptions that physicians represent the top 0.5 percent of the status distribution and that every group has a normal distribution of status, the numbers in table 3.1 allow us to fix the mean for social status of each group at any time. Figure 3.7, for example, shows the implied mean occupational status for Jews and blacks in the United States in 1980 and later. The heavy overrepresentation of Jews, and heavy underrepresentation of blacks, at the top of the status distribution does not require that means for these groups be far from the social mean. There is plenty of overlap in these distributions, but at the bottom or the top of the status distribution, one or the other group will heavily predominate.

  FIGURE 3.5. Educational attainment, Jewish versus general U.S. population, 2007.

  FIGURE 3.6. Educational attainment, black versus general U.S. population, 2007.

  FIGURE 3.7. Implied status distributions, Jewish and black names, 1980–2011.

  Table 3.2 shows the calculated persistence rates by generation. These rates for occupational status are remarkably high compared to conventional estimates. In the most recent generation (column 2), the persistence rate for the five groups averaged 0.74, ranging from 0.65 for the New French and Ivy League groups to 0.88 for Ashkenazi Jews. For the earlier generation in the three cases less affected by racial quotas for medical schools, the average rate of persistence is even higher, at 0.80. In the table persistence rates are not calculated for the groups affe
cted by quotas in earlier years.

  Table 3.2 also shows calculations of the average persistence rate for a generation of thirty years calculated for the 1970s and later. These calculations are included because estimated social mobility rates for Jews were clearly still being influenced by medical school quotas as late as the 1960s.

  Figure 3.8 thus shows the relative representation of each of the five surname groups for each decade from the 1940s onward. The peak frequency of Jewish surnames among physicians qualifying from domestic medical schools, at 7.6 times the expected rate, occurs in the 1970s. In the 1970s, blacks graduated from medical schools at a rate nearly three times higher than in earlier decades, in part as a result of affirmative action policies that have continued to this day.

  Figure 3.8 immediately suggests that these relatively high black mobility rates are likely partly a result of the dramatic institutional changes arising from the civil rights movement of the 1960s and have not been sustained. Similarly the regression to the mean of the Jewish population is underestimated by these generational estimates because the number of Jewish physicians was still being limited by racial quotas even in the 1950s.

  TABLE 3.2. Calculated intergenerational persistence for surname groups among doctors

  FIGURE 3.8. Relative representation of surname types among physicians, by decade.

  Table 3.2 also shows the estimated persistence rates for 1970 and later. The estimated social mobility of the Ashkenazi Jewish group increases, as expected, to a rate of 0.75 per generation. But this still implies remarkably slow mobility compared to conventional measures. For example, at this rate of mobility it will be three hundred years before the Ashkenazi Jewish population of the United States ceases to be overrepresented among physicians.12

  For the black population, the estimated recent rate of convergence toward the mean is even slower. The persistence rate per generation is 0.96. This implies that even in 2240, the black population will be represented among physicians at only half the rate of the general population. However, since the 1970s, rates of relative representation of blacks among physicians have likely been significantly influenced by affirmative-action policies at U.S. medical schools. The measured black persistence rate in this interval may thus also reflect a decline in the effects of such policies over time.

  Among descendants of the New France settlers, representation among physicians is also slowly approaching the mean for the general population. The persistence rate for this group is 0.78, again implying several generations before full convergence.

  The two elite white groups, the descendants of the 1923–24 rich and the descendants of pre-1950 Ivy League graduates, show very different rates of social mobility. The descendants of the rich show very high persistence rates, with no convergence on the average predicted even by 2316. But the Ivy League descendants exhibit rapid social mobility, with a persistence rate of only 0.23. However, under this approach, random error has a big effect on persistence rates for groups that deviate only by small amounts from the social average, as in the case of the Ivy League descendants.

  But as table 3.2 shows, even taking into account sampling errors, the overall rate of social mobility implied by surname persistence among physicians is very low. The persistence parameter averages 0.73 across these five groups for the last four decades.

  Attorneys

  We observe significant status differences and slow intergenerational social mobility using physicians as an indicator of social status. The pattern observed is indicative of a general one found across all high- and low-status occupations. Its broader applicability can be demonstrated in part by carrying out a similar analysis among attorneys.

  Surnames are more difficult to track among attorneys. Like physicians, attorneys are licensed at the state level. But unlike the AMA, the American Bar Association, the main national association for attorneys, maintains no national directory of attorneys. The records of surname distributions among attorneys are contained in fifty different sources. To make the task feasible, the surname frequencies were checked in only the larger states. Also a smaller set of surnames for each group was used, and the surnames Olson and Olsen were taken as representative of the average frequency of names in the domestic population among attorneys by decade.13

  Bar associations and court systems in different states employ different practices with respect to recording inactive attorneys. Some, such as Illinois, record even attorneys first licensed in the nineteenth century. Others, such as Michigan, maintain records only of currently active attorneys. Because surname types are distributed differently across states, these different practices introduce potential error into the process. Attorneys can be licensed in multiple states, and there was no attempt to eliminate multiple listings.

  Using the records of just half the states, it is possible, based on the distribution of physicians with these surnames, to observe 88 percent of the expected attorney stock of a major Jewish surname, Katz; 86 percent of the most common surnames of the 1923–24 rich; 71 percent of Olson/Olsen; 82 percent of the most common New France surnames; and 72 percent of the common black surname Washington. The lower representation of Olson/Olsen comes from the fact that it is more evenly distributed across states than many of the other surnames examined, such as Katz, which is heavily concentrated in a few states.

  With the limitations noted, the same patterns found among physicians are seen for attorneys. Attorneys were assigned to generations by their first licensing date in each state. Usable attorney data actually goes back further than that for physicians, with reasonable numbers of observations even in the 1920s. As figure 3.9 shows, surnames are over- and underrepresented among attorneys in close proportion to their over- or underrepresentation among physicians for the most recent generation.

  There is perhaps a slight tendency among the descendants of the 1923–24 rich to prefer law to medicine, but otherwise the pattern is very similar. This finding suggests that there is nothing special about the occupations of physician and attorney as measures of status. High-status groups are equally disproportionately overrepresented in all elite occupations of equivalent social status. Low-status groups are equally underrepresented.

  FIGURE 3.9. Relative representation by surname types among attorneys and physicians, 2012.

  FIGURE 3.10. Relative representation of surname type among attorneys, by generation.

  To measure social mobility rates among attorneys, relative representations for surname types were calculated across three generations, as for physicians. The results are shown in figure 3.10. There is again a pattern of persistent but very slow regression to the mean for all groups.

  Table 3.3 shows the persistence rate implied for each surname type and period in figure 3.10.14 For the most recent generations of attorneys, the average implied intergenerational correlation is greater than for physicians, averaging 0.84. For the two earlier generations, the average implied correlation is even higher, at 0.94. The earlier estimates, however, are subject to substantial margins of error because of the small numbers of observations.

  Moving to the most recent measurement, which compares the 1990–2012 cohort to that of 1970–89, there is little sign of any improvement in mobility rates. The average persistence rate in this period is still 0.83.

  TABLE 3.3. Calculated intergenerational persistence for surname groups among attorneys

  Although the sampling for attorneys contains more possibilities for error, the attorney evidence is largely consistent with that from physicians and suggests even lower rates of social mobility. It confirms that the social mobility rates found for physicians indicate a generally slow rate of social mobility and are not just an artifact of the physician population.

  Some surname groups are significantly over- or underrepresented among both physicians and attorneys. Although that representation is gradually converging toward the average for all these groups, the rate of convergence is surprisingly low, given conventional mobility estimates. Social mobility is no higher for high
ly visible minorities, such as the Jewish and black population, than it is for less visible minorities: the descendants of the French settlers of Acadia and Quebec, the descendants of the rich of 1923–24, and the descendants of Ivy League graduates of 1850.

  New France Surnames

  The low representation of the surnames of New France settlers among physicians and attorneys is a surprise, as this group has not typically been identified as an underprivileged minority in the United States.

  By design, the surnames selected in this group were those for which less than 5 percent of holders in the census declared themselves black. Thus they largely exclude the common surnames of the Cajun population of Louisiana, such as Landry, for which 12 percent of holders were black. New France surnames instead tend to be concentrated in New England, as a result either of the takeover of parts of Acadia in the eighteenth century by the American colonies or of immigration between 1865 and 1920 of French Canadians from Quebec and New Brunswick. So low representation of these names in the physician and attorney elites cannot be attributed to their being geographically concentrated in poor areas of the United States. Moreover, because this group is not a highly visible minority, its low representation among the current medical and legal elites is unlikely to stem from acts of discrimination. No one bears a grudge against the Gagnons or holds prejudicial views of their abilities.

  FIGURE 3.11. Physicians per thousand surname holders, most common Irish and New France surnames.

  What, then, explains the low social status associated with these surnames? One possible explanation that George Borjas has emphasized in his work is the “cultural capital” of those of New French descent.15 Could this community have inherited a cultural legacy that impedes upward social mobility? There are claims that Franco-Americans were more committed to maintaining their language and religious practices than the assimilationist Irish and Italians. Certainly in 1970 a surprising number of Franco-Americans with parents born in the United States still retained French as their mother tongue.16