The Son Also Rises Read online

  Second, as before, samurai surnames, even controlling for frequency, are now more overrepresented among elites than kazoku surnames. If samurai descendants never intermarried with descendants of commoners, then, assuming the same fertility, their descendants would now constitute 5 percent of the population. But figure 10.3 suggests that samurai descendants may be as much as ten times overrepresented among modern Japanese elites. That rate implies that half the modern elites are descended from the samurai. Intermarriage would greatly expand the share of the modern population of samurai descent. But if the samurai are really ten times overrepresented in modern elites, intermarriage must have been limited, so that their descendants constitute no more than 10 percent of the modern population.

  FIGURE 10.4. From samurai to salarymen.

  Tracking the most common surnames provides confirmation that the samurai and kazoku descendants are still heavily overrepresented among modern Japanese elites (figure 10.4). Most commoners acquired surnames only after 1868, when the government required all families to adopt surnames to aid military conscription, taxation, and postal delivery. Many previously high-status surnames were adopted by commoners. Thus the name Fujiwara, originally the name of a powerful and distinguished medieval family, is now held by more than three hundred thousand Japanese. It is clear that any very common surname in Japan must have been held disproportionately by commoners in the Meiji era. Thus, if we look at the status of the most common surnames, we can determine the relative status of the descendants of the commoners of 1868 in modern Japan.

  FIGURE 10.5. Relative representation of common surnames among elites, Japan, 1980–2012.

  Taking the fifty most common surnames in Japan, which together account for a quarter of the population, we find that their relative representation among elites in Japan is always less than one. Typically these surnames occur at only 85 percent of the rate expected from their population shares (see figure 10.5). Interestingly, they are equally underrepresented among physicians in Japan and among physicians trained in Japan who are now practicing in the United States.

  SOCIAL MOBILITY RATES, 1900–2012

  To calculate the relative representation of kazoku and samurai surnames among elites as far back as 1900 requires knowing the share of these surnames in the general population in earlier generations. There is some evidence of lower fertility among elites, which would imply smaller elite-surname population shares earlier. In the years 1940–67, for example, marriages involving highly educated husbands or wives produced one-fifth fewer births than those of couples with lower educational attainment.17 But the prevalence of adult male adoption among the elite would create a countervailing tendency for elite-surname population shares to increase, if some of the adoptees originally had average-status surnames. Absent better data, it is assumed that the population share of these surnames was the same in 1900 as in 2007.

  FIGURE 10.6. Relative representation of kazoku and samurai surnames among authors of scholarly publications, 1900–2012.

  Google Scholar gives measures of the number of publications associated with samurai, kazoku, and common surnames for the period 1900–2012.18 To calculate the relative representation of these surnames among publications by Japanese authors requires, however, that the total stock of such publications be estimated for every decade. This is done by measuring publications by authors holding the ten most common Japanese names in each decade.19

  Figure 10.6 plots the relative representation of samurai and kazoku rare surnames among authors of scholarly publications by decade from 1900. Both the samurai and kazoku surnames are heavily overrepresented in the initial decades, with publication rates of eleven and twenty-one times those of the average person in Japan, respectively. That relative representation declines over time, but is still more than four for both groups from 1990 to 2012. Indeed, for both groups, there is an increase in relative representation in the years 2000 and later.

  The implied intergenerational correlation of publication rates, assuming a standard thirty-year generation, is 0.71 for the kazoku and 0.88 for the samurai. There is no sign of any increase in mobility rates after the reforms of 1947. Again, adult male adoption may bias these estimates up by as much as .06. But even allowing for this, social mobility rates are clearly just as low in Japan as in the other countries studied, and they are just as low after 1900 as they were in the early Meiji period.

  The magnitude and pattern of these results for authors is confirmed by the representation of these surnames among medical researchers. Listings are available of medical researchers in Japan for 1965–66 and 1989–90, a twenty-five-year interval. Table 10.6 shows the relative representation of the kazoku and samurai rare surnames among medical researchers in each year. The names are again distinctly overrepresented. In the bottom row of the table are the implied persistence rates, adjusted to a thirty-year generation. The values are slightly lower than for the publications, at 0.63 for kazoku and 0.82 for samurai surnames. But this difference could be due to chance. Because of the small populations bearing these rare surnames (41,000 and 75,000 respectively circa 1990), the number of observed medical researchers in each period is small.

  Another estimate of social mobility in Japan comes from the 2012 American Medical Association directory. This lists more than a thousand Japanese-trained physicians practicing in the United States, a fifth of whom hold one of the fifty most common Japanese surnames. The AMA data also give the date of graduation of each registrant. The Japanese physicians can thus be divided into two generations, those graduating between 1950 and 1979 and those graduating between 1980 and 2012. The relative representation of physicians with these common surnames among this group is 0.77 for the period 1950–79 and 0.85 for 1980–2012. Thus, again, we see signs of regression toward the mean. But, assuming that doctors represent the top 0.5 percent of the society, the persistence rate implied by this rate of convergence is 0.65. This estimate has a very high potential measurement error, but it is another sign that the rate of social mobility in Japan is much lower in the modern era than conventionally estimated.

  TABLE 10.6. Relative representation of samurai and kazoku among medical researchers, 1965–89

  Year

  Samurai

  Kazoku

  Observed 1965–66

  30

  13

  Observed 1989–90

  70

  23

  1965–66

  5.99

  4.95

  1989–90

  4.69

  2.94

  Implied persistence rate

  0.84

  0.64

  SAMURAI VERSUS KAZOKU MOBILITY RATES

  The persistence rates estimated for the kazoku for 1900 and later are in the range 0.64–0.71, and for common surnames the rate is 0.65. This is similar to but perhaps a bit lower than in other countries. However, the persistence rates for the surnames associated with the samurai are 0.84–0.88, higher than elsewhere. There is also evidence that the samurai surnames are now more elite than the kazoku surnames, even though circa 1900 the kazoku were the most elite group in Japanese society. Why has regression to the mean across all these measures been faster for the kazoku than for the samurai descendants?

  One interesting difference between the two groups is that all the samurai surnames were identified as belonging to at least one samurai family in 1812, whereas the kazoku surnames were identified as belonging to families added to the kazoku between 1869 and 1946. Only 42 percent of the kazoku derived from the hereditary nobility; the other 58 percent were people awarded titles in recognition of distinction in the military, administration, commerce, and other professions between 1869 and 1946.

  According to the theory posited in chapter 6, observed status on any dimension is related only with some degree of error to an underlying social status, which regresses slowly to the mean. When we observe samurai publication rates for 1900 onward, using Google Scholar, to get a measure of observed status, the selection of surnames is based on
samurai names from 1812. Thus the expected value of the underlying status equals that of observed status for 1900 and later. Publication rates are an accurate indicator of samurai status overall. We thus observe in the publication rates the rate of regression of underlying status, even though publications are just a partial indicator.

  However, in the case of the kazoku, many of whom achieved prominence only after 1880 and constituted the initial cohort of this elite, those showing up as authors between 1900 and 1949, and even in the 1950s and 1960s, include many individuals newly appointed to the kazoku because of their distinction in scholarly or technical fields. Thus for this group, the expected error relating observed status (measured by number of publications) to underlying status is positive on average. The observed regression to the mean for observed status is thus greater than that for underlying status, because in the succeeding generations the average error linking publication rates to underlying status is once again zero.

  Despite this observed faster regression to the mean of the kazoku, then, there may well be no inherent difference in the social mobility rates of the kazoku and the samurai descendants after 1947. If that is the case, however, we would expect that by the 1970s, when most of the first-generation kazoku were dead, then the rate of regression to the mean of both groups should be the same. The expected error term relating their observed status on any measure to underlying status is now zero for both groups. However, even in this later period, the samurai descendants regress more slowly to the mean than the kazoku descendants. The reason for this difference remains a puzzle.

  Korea

  In South Korea, as in Japan, since 1948 there has been a profound remaking of the political and social order that has displaced previous elites from any privileged social position. The period of Japanese colonial rule, 1910–45, and the disruptions created by the regime in North Korea after 1945 could be expected to have led to a profound dislocation by 1948 of the traditional elites of the long-lasting Joseon dynasty. This was followed by a period of sustained and rapid industrialization, with substantial migrations of people to urban areas. Korea is also, like Japan, a society of great ethnic and cultural homogeneity, a fact that, again, would lead us to expect a great deal of social mobility in South Korea in the period 1948–2013. Korea also appears by conventional measures to be a society of high social mobility: a recent study of income mobility across generations reports a persistence rate of 0.35.20

  Korea does not initially seem like a promising society in which to study social mobility using surnames. The three most common surnames in Korea—Kim, Lee, and Park—are held by two-fifths of the population. However, as in China, the lack of variety in surnames led to the convention of identifying people both by surname and by family place of origin. Thus, for the surname Kim, which is held by a quarter of all Koreans, in 2000 the census in Korea recorded 348 different clans, or bon-guan. Membership in these clans is patrilineal. Clan membership was important in Korea for delineating possible marriage partners: traditionally, marriage had to be to someone outside the bon-guan. Indeed, until 1997, it was not legal for Koreans to marry within their bon-guan.

  In total, these surname–place of origin combinations provide 3,783 distinctive family names by 2000. There have been claims that although clan membership is supposed to descend strictly through the male line, in the nineteenth century many arrivistes from lower-status groups affiliated themselves fraudulently with clans of distinguished lineage. Even if that is correct, by 1898, under the Japanese Family Registration Law all family names became fixed in Korea, so the modern surname-bon-guan combinations indicate with high fidelity relationships to people born more than a hundred years ago.

  In a recent paper, Christopher Paik applies the methods of this book to estimate educational mobility rates in South Korea for the years 1955–2000, using these surname–place of origin designations.21 Korea, like China, had a national examination system in the Joseon imperial era (1368–1894). Extensive records exist of the names of successful candidates in these exams. From these records, Paik assigns to each surname and bon-guan combination (taken from the 1985 census) a weighted average exam-success rate, the weights being based on the prestige of the exam. These rates vary from 0 to as high as thirty-six per thousand for the Suh of Daegu, a clan that constituted 0.34 percent of the 1985 Korean population. The largest surname–place of origin group in Korea today is the oldest branch of the Kim of Gimahae, to which nearly 1 percent of Koreans belong. But their exam-success rate was only 0.05 per thousand.

  For both 1985 and 2000, the Korean census records surname-bon-guan frequencies for all of the country’s 192 districts. Paik can thus calculate the average exam status of surnames within each district. The censuses of 1955, 1970, 1985, and 2000 also give the average years of education attained per district for those age 25–39. Paik can thus also assign to each district in each of these years a standardized educational achievement score.

  If the past history of families before 1898 has no effect on current outcomes, then there should be no correlation between the average exam status score by district and current educational achievement. But for all four years examined, Paik finds a significant association.22 The higher the exam lineage status for the average surname in the district, the higher was the educational attainment of the 25–39 age group in each census year.

  But Paik also observes that the relationship between exam status and district education levels for the 25–39 cohort is strongest for 1955 and weakest for 2000. This weakening is due to social mobility. From the rate of decline in correlation, it is possible to infer the underlying persistence parameter for modern Korea. The implied intergenerational persistence rate is 0.86 from 1985 to 2000, 0.86 from 1970 to 1985, and 0.74 from 1955 to 1970. On average, the implied persistence rate for educational status in Korea from 1955 to 2000 is thus 0.82.

  Thus despite the homogeneity of Korean society and the remaking of social institutions on a large scale several times in the past century, social mobility in Korea is no faster than in other countries and potentially slower. Like Sweden and Japan, Korea illustrates how widely conventional estimates of social mobility can vary from underlying rates of social mobility.

  Conclusion

  The samurai and kazoku show surprising persistence as elites in modern Japanese society, despite the samurai having lost any legal privileges by 1871 and the kazoku having lost theirs by 1947. In particular, if the descendants of the samurai constitute 5 percent of the modern Japanese population, then they could still constitute anywhere between 20 and 50 percent of modern Japanese elites. The homogeneity of Japanese society does not lead to a higher rate of social mobility than we observe in more ethnically, racially, and religiously diverse societies, such as the United States. Reinforcing this finding, Korea, another highly homogeneous society, again shows rates of social mobility that are no faster than in socially and ethnically fractured societies such as the United States.

  1 Inglehart and Welzel 2010, 554.

  2 Jones, Kojima, and Marks 1994.

  3 Clark and Ishii 2013.

  4 Ueda 2009.

  5 Amano 1990, 192.

  6 Clark and Ishii 2013.

  7 Sonoda 1990, 103.

  8 Harootunian 1959.

  9 Takayanagi, Okayama, and Saiki 1964.

  10 Ando 1999, 259.

  11 Kitaoji 1971, 1046.

  12 Mehrotra et al., 2011, table 1. The number of adult adoptions in 1955 suggests that about 7 percent of each male cohort were adopted as adults.

  13 Lebra 1989, 106–32.

  14 Moore 1970.

  15 See Clark and Ishii 2013.

  16 No sources were available listing only regular physicians.

  17 Hashimoto 1974, S184.

  18 Clark and Ishii (2013) detail some complications that arise in performing this exercise.

  19 This stock of all publications is adjusted upward by 13 percent to allow for the expected underrepresentation of common surnames among elite groups such as authors
, as seen in figure 10.5.

  20 Ueda 2013.

  21 Paik 2013.

  22 For 1970 and 1955, exam status in each district is assumed to be the same as in 1985.

  ELEVEN

  Chile

  Mobility among the Oligarchs

  CHILE IS NOW ONE OF SOUTH AMERICA’S more prosperous economies. Its income per capita equals that of Argentina, although it is still only about one-third that of the United States. But like its neighbors, Chile is characterized by inequalities in income and wealth that are among the highest in the world. The Gini coefficient measuring income inequality is 0.55 for Chile, compared to 0.26 for Sweden (see figure 1.3). Despite Chile’s relatively high average income, poverty is highly visible, as illustrated by the shanty town in the port city of Valparaiso shown in figure 11.1.

  A trope of modern discussions of mobility has been that inequality breeds immobility. Empirical studies of the connection between inequality and mobility rates across countries appear to confirm this assertion. In these studies, South American countries register high degrees of inequality and low levels of social mobility.1

  However, the mechanisms that might explain this empirical association between inequality and immobility are unclear. One explanation is an analogy: the greater the distances between the rungs of the economic ladder, the harder it is for people at the bottom to climb up the ladder. But this analogy is entirely spurious as an argument for lower social mobility. In a society such as the United States or Chile, where the rungs in the income ladder are further apart, the same rate of occupational mobility as in Sweden produces much greater changes in income. So the ordinary processes of occupational mobility ensure the same proportional movements in income toward the mean as are seen in Sweden, where the rungs of the income ladder are closer together. If the rungs on the income ladder are farther apart in the United States or Chile, the upward steps that are possible through occupational mobility are correspondingly greater. So the distance between the analogical ladder rungs explain nothing about rates of social mobility.