China: The Blood and the Paper

The Y chromosome passes from father to son almost unchanged, so it is a poor historian of everything except one question: which men left sons who left sons. In China its answer starts early. Between about 6,800 and 5,400 years ago three men’s lines spread until they covered more than forty percent of Han men alive today, sixteen, eleven and fourteen percent apiece (Yan et al., 2014). The authors tie the timing to the graves of the period, which show “a gradual transition from an egalitarian matrilineal society into a hierarchical patrilineal one.” Whether the three could be the legendary Emperors Yan and Huang, they wrote, “is to be solved with more prudence,” a quality so rare in ancestry research that it deserves its own haplogroup.

Zeng, Aw and Feldman (2018) supplied the mechanism. Once men lived in clans of brothers and cousins, they also died that way: “Casualties from intergroup competition then tend to cluster among related males, and group extinction is effectively the extinction of lineages.” A war between clans was a war between chromosomes, and the losing chromosome did not sue for peace. David Reich’s phrase for the pattern, written about the steppe and not about China, is “a genetic scar that speaks volumes about past societies.” It is a scar from power, not from study. It opened more than four thousand years before anyone sat an exam.

The exam added its own entries in the same hand. Sijie Hu went through six lineage genealogies, 36,456 men from 1350 to 1920, and found that “degree and office holders, or the literati, produced more than twice as many surviving sons as non-degree holders.” His channels are fewer dead children and better access to marriage. That is rice, medicine and a wife. It did not take a harem: in the Tongcheng clan books, polygynous unions ran under two percent of the cases, and most second marriages followed a wife’s death, about twelve percent of all marriages (Shiue, NBER w19661).

What the blood never recorded is the reading. A chromosome can tell you a man had sons. It cannot tell you whether he could read the Mencius he had been told to obey, and nothing on this page says it can. The genes show who bred, the wall, and the cost of the wall. The habit traveled on paper.

From the eleventh century on, every candidate memorized Mencius, and Mencius had settled the matter of heirs in one line. The law codes then wrote it into the household, article by article.

A published English paraphrase says the same and adds the obvious: “neither official ideology nor the laws were seriously followed” (Zhao, 1995, p. 61). The Ming jurists could not even agree on whom the rule bound. One commentary, the 集解, says officials, clerks and government students “are all judged as commoners.” The other, the 纂註, notes that the clause names only commoners, so officials fall outside it, and gives the reason for the whole rule in three characters: 重無後也, “it is the gravity of having no heir.” The imperial clan had a quota instead of an age: a 1585 substatute let a prince of the blood keep at most ten concubines, an heir or commandery prince four, and so on down to two (Liang, 2019).

The Qing code dropped the age clause and kept the purpose. “No son” heads its list of the seven grounds on which a man could put away his wife. Another substatute divides family land “equally by the number of sons, without asking whether born of wife, concubine or maid” (our gloss of 不問妻妾婢生止以子數均分). Equal partition is the opposite of primogeniture. The land shrank with every brother. The degree was the one asset that did not divide.

The ink took sides. The clan books that counted the sons were kept by the houses that had sons to count, and a man who left none ended on the page the way he ended in the blood. The bare branches kept no books of their own.

Ho Ping-ti traced 12,226 jinshi from 48 examination lists back three generations in the father’s line. His numbers are the strongest fact in the Chinese record against the idea that the advantage lived in the blood of degree-holding houses. If it had, these shares would sit near zero.

A 2017 recount by Hsu Hong, on more Ming lists than Ho had, again found commoner origins for more than half. So the door let in new families. Now the corrections, and every one of them cuts the same way.

Shen Dengmiao (2019) shows that Ho’s high early-Ming figure rests on very few lists and a one-off: 64 of the 120 men who passed in 1371 came from households registered as Confucian under the Yuan, scholar families whose degrees the conquest had interrupted. Benjamin Elman (2009, p. 405) puts the bottom of the ladder plainly: “peasants, traders, and artisans, who made up 90 percent of the population, were not a significant part of the 2 to 3 million candidates who usually took the local biennial licensing tests.” And the ladder grew no rungs: one civil licentiate per 300 people in 1600, one per 1,000 by 1850 (p. 408). Jiang and Kung (2021), with data from 1796 to 1905, find that ability and an upper gentry family predict success at the top exam while direct proxies for wealth do not, and that at the final palace examination it is the official rank of a candidate’s father and ancestors that predicts his ranking.

Both halves are true and the page will not choose one. The degree could not be inherited, and about half the Ming men who reached the top came from houses with nothing above the lowest degree. The door did not let in peasants. It opened to families that could already pay a tutor, and the habit the exam rewarded could be built by money and kin in a generation or two.

Somebody paid for the literati’s surviving sons. Female infanticide and polygyny together meant that “during the Ming and Qing dynasties and the Republican Era, up to twenty percent of males remained single” (Jiang and Sánchez-Barricarte, 2011). The Chinese name for them was bare branches, guanggun, also bare sticks. Huang and Goldin put the arithmetic in one line: for “the overwhelming majority of males, this number would have been zero or one, but even those with one wife might have considered themselves fortunate.”

The girls are counted, and the count is disputed, so here are both. James Lee and Cameron Campbell estimated that “from a fifth to a quarter of girls were victims of infanticide” in a banner village of Liaoning, Daoyi, 1792 to 1867, and that in the Qing nobility “about 10 percent of female infants were killed, but in the 1780s, the rate was as high as 20 percent” (as summarized by Lavely and Wong, 1998, p. 17). Cao Shuji and Chen Yixin (2002) answer that Lee and his colleagues 有所高估, “somewhat overestimate,” infanticide before the twentieth century, inferring it from the surface data of the Daoyi registers, and cite Lai Huimin’s finding that smallpox, not parents, was the great killer of the village’s children.

The women who were raised had a price, and it moved with grain. From 73,216 Qing criminal memorials on marriage and adultery, 1736 to 1896, Chen Zhiwu and his coauthors pulled 3,119 cases with prices. In drought years a ten percent rise in the price of grain lowered the price of wives and concubines by thirty-three percent. The authors’ term for her is 避险资产, a hedge asset. Matthew Sommer concludes that among the Qing poor “no clear distinction can be drawn between marriage and the traffic in women.”

The throne noticed the bare sticks only as a security problem. In 1705 the Kangxi emperor “excluded bare sticks from imperial pardons,” and in 1729 the Yongzheng emperor “commended officials who cracked down on bare sticks” (Buoye, 2014). North of the Huai River the problem raised an army. Famine had taught Huaibei families to kill their daughters, and the nineteenth-century count ran 129 men to every 100 women. In 1815 a magistrate sorted the local trouble into three categories, “bare sticks, smugglers, and bandits,” and found “the overlap between the three groups was great.” As the Nian they rose in the 1850s, and in 1862 the governor of Anhui reported “at least two thousand Nien forts in Huai-pei, each with one to three thousand inhabitants” (Perry, as quoted in Hudson and den Boer, 2002, pp. 16 to 19). Hudson and den Boer’s gloss on the young men says the rest: “Their parents had killed the girls who should have grown up to be their wives.”

The exam families wrote the strategy down, and they wrote it as advice to their sons.

Yan wrote that as a refugee who had watched war take everything else. He also recorded what a second wife did to a house in the north, where a concubine’s son was not counted as a person: after the father died, “a son calls his own mother a concubine, a younger brother degrades his elder to a hired hand” (our gloss of 子誣母為妾，弟黜兄為傭).

Fan Zhongyan made it policy. In 1050 he bought more than ten qing of land, over a thousand mu, in two counties of Suzhou and fed his clan from it: one sheng of rice a day for every man and woman five and older, a bolt of winter cloth a head, thirty strings to marry off a daughter, twenty to take a wife, and nothing for a second wife (再娶不支). In 1073 his sons added ten strings for any young man who sat the great examination, “paid only if he actually goes,” and reclaimed if he took it and did not sit. Two kinsmen who had passed the prefectural exam taught the clan’s boys for five shi of rice a month, cut to three if they had fewer than six pupils. A later rule raised the grant “so that the sons of every branch may know the beauty of study.” Another refused rations to any son adopted from another surname. The estate fed the paper line, not the chromosome.

Yuan Cai, in 1178, was blunter. For an official’s sons with no inherited salary and no land, 莫如為儒, nothing beats being a scholar: the gifted win a degree and wealth, the rest open a school and take fees. He thought Fan had it backwards: 置義莊不若置義學, better a charitable school than a charitable estate, because estate shares shrink as the lineage multiplies.

Zhang Ying of Tongcheng, the county whose genealogies Shiue and Hu counted, left his sons four rules and put reading first: 讀書者不賤, “he who reads is never base.” He had watched the great official houses: where one went desolate after the old men died, 其後無讀書之人也, “nobody after them reads.”

Zeng Guofan, the general who put down the Taiping, wrote to his son Jihong that most men want their sons to be high officials and 余不願為大官，但願為讀書明理之君子, “I do not; I want them to be gentlemen who read and understand.” In 1862 he arranged 400 taels through the Eastern Expedition Bureau, the office that raised money for his armies, for the boy’s next step up the exam ladder, and budgeted a hundred strings each for the clerks’ fees. The words say reading. The ledger says the exam.

The fiction of the period knew all of this and said it with less arithmetic. In Jin Ping Mei, Ximen Qing keeps a wife and five concubines, and the only son is a concubine’s. A rival concubine handles the problem:

A polygynous house is a war over one heir, and the novel says so with a cat. In Dream of the Red Chamber the law’s equal share for the concubine’s son buys him nothing at table. Jia Huan, son of the concubine Aunt Zhao, complains that “you all look down upon me for not being the child of my lady,” and his sister Tanchun refuses her own mother’s brother as kin because her ritual uncle is the wife’s brother: “Who’s my maternal uncle? My maternal uncle was at the end of the year promoted to be High Commissioner of the Nine Provinces!” The hero Baoyu calls the men who study for the exams “the salaried worms” (Joly, chs. 20, 55 and 19). Near the end he passes the provincial exam anyway, and does not come home: 中了第七名舉人。王夫人道：「寶玉呢？」家人不言語。 In our working gloss, not a published translation: “Placed seventh.” “And Baoyu?” The servant said nothing (ch. 119).

Pu Songling, who failed the provincial exam for most of his life, gave the hall its physiology. A candidate is at dinner when the list comes out and “rushed out with his mouth full to ask if his name was there or not; and on hearing that it was not,” what he had swallowed “stuck fast like a lump in his chest and made him very ill” (“The Great Test”). A man who throws out his wife for a fox girl loses his place at the head of the list, and “at forty he had no wife, and was so poor that he was glad to pick up a meal where he could” (“The Husband Punished”; both Giles, 1880).

Shen Fu had no degree and no post, and in 1808 he wrote down what that meant in a household. His wife Yun spent years finding him a concubine, a singing girl named Hanyuan; a richer man took the girl with a thousand in gold and a promise to keep her mother, and Yun never recovered. His painting shop took in less in three days than the house spent in one. Then their son died. The Chinese is Shen Fu’s; the English is our working gloss, not a published translation:

The concubine Yun died seeking arrived as a gift after she was gone, sent because the son was dead.

A Chinese surname is a claim, not a chromosome, and the geneticists have now audited three famous ones.

Cao Cao, the third-century warlord, had a palace eunuch for a grandfather, Cao Teng, who adopted his heir. For eighteen hundred years people asked whether the adopted son was a Cao at all. Fudan’s team found that six of eight Cao lineages whose genealogies name Cao Cao share one Y type, against under five percent in other Cao families, and a tooth from the tomb of his great-uncle Cao Ding carries the same line (Wang et al., 2012; Wen et al., 2016). The adopted heir was most probably taken “from his grandfather’s own clan rather than from beggardom” (Wang and Li, 2013). The same study punctured the family’s grander claim: descent from the Han chancellor Cao Can “was not supported by genetic evidence.” The eunuch’s paperwork was honest about the adoption and flattering about the ancestors, which is about the usual ratio.

The Kong of Qufu claim Confucius, and their genealogy is the most famous in China. Of 1,118 Kong men tested, the main male lines split three ways, 46, 27 and 21 percent, and the authors could only say that either of the first two “may be” Confucius’s. Their own explanation: 外孔原為劉姓, “the outer Kong were originally surnamed Liu,” and changed their name (Hou et al., 2016). The best-kept family tree in the country, and the men who carry the name do not share one father.

The Aisin Gioro, the house that held the throne from 1644 to 1911, enjoyed the most extreme reproductive privilege in the empire for 267 years. Their Y chromosome is identifiable today because “the haplotype is rare” (Yan et al., 2015). Not the star cluster. Three centuries at the top of the most populous state on earth left no dominant male line. The great Chinese bottleneck is older than any dynasty.

About 4,000 surnames cover the Han population, and Wang, Li and Zhang alone are carried by “more than 270 million people (21 percent of the population)” (Clark et al., 2014). A surname carried by a hundred million men cannot be one patriline. In Spain, where it has been measured, shared ancestry within a surname falls as the surname gets commoner (Martinez-Cadenas et al., 2016).

Gregory Clark and his coauthors still found something in the names. Thirteen surnames heavily overrepresented among Qing jinshi stayed overrepresented among Republican officials, elite professors, company chairs and central officials, with an implied persistence “0.96 between the Imperial and Republican era, and still 0.92 between the Republican and Communist eras.” All thirteen cluster in the lower Yangzi, so Clark re-ran them against regionally favored names and got 0.89 for the Republican era and 0.82 for the Communist. The persistence is real and the book already tells it.

The cause is another matter. Clark has since argued that status passes like an additive genetic trait (Clark, 2023), and in 2025 that the English pattern is “consistent with direct genetic inheritance, but not with cultural transmission of status.” The critics are specific. Benning and colleagues (2025) reanalyzed the 2023 paper and found its conclusions “based on the conflation of genetic and non-genetic transmission (for example, of wealth) within families.” Torche and Corvalan (2018) show that surname averages measure persistence between groups, not the mobility of individuals. Lala and Feldman (2024), Feldman being the same geneticist who explained the Chinese Y bottleneck as culture, set out the gene-culture case against determinist readings. And Chen, Kung and Ma, whose 2010 census result the book relies on, measured no genes; they list “region-specific endowments (natural or genetic)” as a confound and design around it.

The book does not adopt a genetic reading of status, and neither does this page. Clark measured the persistence and then guessed at a cause his measurement cannot reach. The claw mark is in the surname data. The claw is not.