Social Mobility in Late Imperial China: Reconsidering the “Ladder of Success” Hypothesis

Abstract

Using unique data, we test whether the civil examination system in late imperial China (1796–1905) constituted a route to social mobility. We find that measures of both ability and “family type”—specifically the upper gentry family type—consistently predict success in the highest level exam, the jinshi, while direct proxies for wealth do not. Specifically, the higher the level of the father’s education the greater the odds of passing the first stage of the jinshi exam—the metropolitan exam. But in the final stage of the jinshi exam (the palace exam), where candidates were examined only on their knowledge of statecraft, it is the official rank of one’s father and ancestors, which we use to proxy for family-specific tacit knowledge (or “cultural capital”), that significantly predicts the final ranking and class of honors—both of which crucially determined entry into the Hanlin Academy and subsequent career trajectories. Thus, while it unwittingly facilitated mobility, the civil examination system transmitted hidden advantages among those endowed with more than just human capital.

Keywords

social mobility civil exam ability wealth tacit knowledge (cultural capital)

As an elite social class, the gentry stood head and shoulders above the rest of Chinese society and thus commanded the respect of the common people. The gentry were actually officials, who, for well over a thousand years and under the lasting influence of Confucianism, constituted a class of professionals at the top of the pecking order in China.¹ But they were more than just officials, for this highly regarded social status was attained only after one passed the grueling civil exam, which by and large was steeped in the tradition of Confucian classics and ethos (Elman, 2000; Miyazaki, 1976). Thus, China’s gentry were in fact learned or scholar-officials. And, while they may be regarded as having the same social status as the aristocrats in Europe, unlike the latter they could not bestow their distinctly privileged social status on their heirs. In order to “retain” this coveted social status, descendants of the Chinese gentry had to sit for and pass the civil exam themselves. It was this particular feature of the civil exam system that rendered social mobility in imperial China feasible, as the exam was open to all (males).

That was precisely what the late Ho Ping-ti (1967)—an eminent historian of China—contended. Ho found that more than half of those who successfully passed the juren 舉人 exam—the second highest degree in China’s exam pyramid—actually came from families where none of the ancestors had achieved the same level of exam success as they had or held the same official titles as they did. Ho’s finding is thus strong evidence that Chinese society in late imperial times was indeed highly socially mobile.²

Of course, not everyone agrees with this optimistic assessment. Given that passing all three levels of the civil exam—the shengyuan, the juren, and the jinshi (discussed in the following section)—required tremendous financial and other types of support (e.g., many had to sit for the exam repeatedly before they passed it), it was essentially an institution that favored resource-rich families, clans, and lineages. As such, the exam did no more than contribute to a “social reproduction of the status quo” in the words of Robert Hymes (1986: 42; see also Elman, 2013; Fairbank, 1983).

While the hypotheses implicit in these arguments are straightforward, the debate has not gone very far, as neither camp has constructed specific measures to test their respective claims. To rigorously test these competing claims, we draw upon a number of valuable historical sources from which to construct a unique data set and specifically a number of variables to proxy for ability on the one hand and family background on the other. Based on these measures, we test the determinants of success in the jinshi 進士 exam—the highest achievement under the civil exam system and one whereby entry into officialdom was guaranteed. Using one’s age on passing the provincial or juren exam and one’s ranking in it as proxies of “initial” ability, we find that they are consistently significant throughout the estimations, regardless of whether our dependent variable is simply measuring pass/fail in the jinshi exam or how well one actually performed in terms of ranking and class of honors. This latter measure is important because it affected the odds of entering the Hanlin Academy 翰林院, which admitted only the best jinshi for preparation for a future career in the government. In any case, these results lend strong support to Ho’s (1967) mobility claim.

However, family background, which we measure by exam and career achievements of one’s ancestors including one’s father, also matters. Specifically, “upper gentry” families enjoyed a significant advantage over not just “commoner” families, but also the other two intermediate social classes, namely, the “scholar-commoner” and “lower to middle gentry” families. This result thus powerfully suggests that family background also matters.

What then were the mechanisms through which family background affected civil exam success? To look into this, we classify family background into two types: the attainment of one’s father, grandfather, and great-grandfather in civil exams and official careers. In addition, owing to the lack of more direct measures of family wealth such as landholding, we use as proxies the average number of wives and concubines that one’s father and grandfather had, and whether a family had purchased a lower-level degree to give their son an earlier start. Perhaps they are noisy measures of wealth, or perhaps juren is already too high a level for wealth to matter for success in the metropolitan exam; in any case, we find that wealth as measured by the above metrics had no significant effect on exam success. What bore significantly on jinshi exam success was the father’s education, whereas the official ranks of the father, the grandfather, and great-grandfather significantly predict ranking and class of honor at the final stage of the jinshi exam (the palace exam).

While official rank may well be correlated with both education and wealth, it differed from the other two in that it affected one’s exam performance through the channel of tacit knowledge of statecraft transmitted across generations within the family.³ Unlike the metropolitan exam, which still emphasized the Confucian classics, the palace exam tested only one’s knowledge of statecraft. And unlike the Confucian classics, which could be studied on a systematic basis (and from a relatively tender age), knowledge pertaining to statecraft, particularly military affairs as the late Qing found itself increasingly caught in military conflicts both with the West and domestically, could only be acquired tacitly within the family. Specifically, families having members (father, grandfather, and so forth) who worked in the government were more likely to be able to share such knowledge with the younger generations. Hence, while wealth has no direct effect on exam success, as a primary source of statecraft knowledge family background, or specifically “cultural capital” (Bourdieu and Passeron, 1990), exerted a significant effect on just how well one did in the palace exam—a stage that importantly determined the entry point of one’s official career.

Our overriding goal is to clarify an issue of epic proportions in China’s social history via a systematic analysis of data assembled for this particular purpose. In doing so, we also hope to speak to an emerging literature that delves into the subtle but lingering importance of ancestry in shaping economic and social outcomes (Clark, 2014; Song and Campbell, 2017). For instance, we confirm and offer solid evidence for the importance of ancestry in transmitting certain competitive advantages across generations, in our case the transmission of tacit knowledge that in the great majority of instances could only be acquired from within the context of the family.⁴ Additionally, our study also bears on the importance of multigenerational analysis such as that propounded by Robert Mare (2011) and richly analyzed in Carol Shiue’s (2017) study of social mobility during the Ming-Qing period based on data constructed from Chinese genealogies, which represent an important source for studies on social mobility (see also Campbell and Lee, 2003, 2008).

The remainder of this article proceeds as follows. We begin by briefly reviewing the relevant literature, outlining the respective arguments and evidence presented by the proponents of the mobility thesis and their critics. We then introduce the data sources, explain how the variables are constructed, and account for the rationale behind the choice of empirical models. Before discussing the results of our empirical analysis, our next topic, we first present a mobility table based on our data to illustrate both the importance of family background for exam mobility and the possibility of downward mobility. The last section offers a conclusion.

China’s Imperial Exam and Social Mobility

The Civil Examination System: Origins and Goals

China’s civil examination system, or keju 科舉, was designed to recruit learned talent into the government, in a society where government service was considered the most honorable and worthwhile occupation of all. The civil exam system to which we refer throughout the article is the one in place during the Ming (1368–1644) and Qing (1644–1911) dynasties, as it was during this era that the system was consolidated and had become stabilized. The county examination 縣試 was the starting point. Success in this exam led to the qualification of a licentiate, or shengyuan 生員, followed by entry into a county or prefectural school as a “secondary student” 附生. In the Chinese exam system, as with all prestigious exam systems where the demand for qualifications outstrips the supply of rewards, a quota was imposed. Thus shengyuan faced competition for admission to county (and prefectural) schools. But it was at the provincial exam 鄉試 that competition flared up. As each province was assigned a quota for the juren 舉人, a shengyuan had to perform well in the qualifying exam held at the county school before he was allowed to take the provincial exam. Those who passed the exam became juren. But competition was fierce: for any given exam, only one out of every fifty to eighty candidates on average managed to pass. In short, as many as 98 percent of the candidates were destined to fail (Shang, 2004; Elman, 2000; Huang, 2016).

Nevertheless, the extraordinarily low passing rate for the juren exam meant that there was always a bottleneck. To deal with this, a separate status of “tribute students” 貢生 was created. These students, after receiving training at the imperial college 國子監 and successfully surmounting some further barriers, became qualified for entry into officialdom and earned themselves the respectable status of “lower gentry” (Ho, 1967). But still it was possible for some—presumably from wealthy families—to take a short cut, for the system also allowed the purchase of an imperial college degree for approximately a hundred taels of silver (this was available for their children as young as ten). Those who took this short cut then became a “state student,” or jiansheng 監生—a status that qualified them to take the juren exam later on (Hsu, 1950). This way, jiansheng could avoid the tormenting experience of being a shengyuan at a county or prefectural school (Zhang, 2010).

Upon passing the provincial exam, one became qualified to prepare for the ultimate national exam leading to the highest degree, the jinshi. The jinshi exam consisted of two parts: first, the metropolitan exam 會試 and, upon passing it, the palace exam 殿試, which took place before the emperor. While no one failed in the palace exam, examinees were ranked according to their performance and classified into a distinct class of honors. Only the top three candidates of each palace exam were given first-class honors 一甲; the next eighty to a hundred received second-class honors 二甲, and the rest third-class honors 三甲. Figure 1 summarizes the various levels of the civil exam and corresponding degrees in Qing China.

Figure 1.

The civil exam structure in Qing China.

A question of overriding importance is how competitive the jinshi exam really was. Depending on the number of juren taking the jinshi exam at any given time, which fell between four thousand (Guo, 2006) and six thousand (Weng, 1989 [Qing]) depending on the year, the passing rate ranged from a low of 4.2 percent to a high of 6.3 percent, assuming that each national exam produced approximately an average of two hundred and fifty jinshi. Ichisada Miyazaki’s (1976) estimate that only one out of thirty candidates passed this exam is thus consistent with the lower bound of this estimate.

While passing the jinshi (metropolitan) exam was clearly something of first-order importance to the candidates, once they overcame this hurdle their concern turned immediately to their performance in the palace exam, as it was the ranking achieved in this exam and the class of honors conferred on them that determined admission to the Hanlin Academy—an institution for grooming high-level officials (Lui, 1981; Elman, 2000). Those who ranked high in the palace exam could look forward to a plum appointment—for instance, as an official in the capital of Beijing, with further opportunities to serve as prefect in a prefecture. Even the worst performing graduates of the Hanlin Academy were immediately dispatched to counties to serve as magistrates without having to wait for a vacancy. In other words, candidates who ranked high in the palace exam immediately received a real job, whereas third-class jinshi had to wait in line until a position actually became available (Shang, 2004: 164). It is this crucial difference that made ranking and class of honors in the palace exam so important

Beyond Confucianism: The Tacit Knowledge of Public Affairs in the Palace Exam

Given that China’s civil examination system was steeped in the Confucian classics based on the Four Books 四書 and the Five Classics 五經 and their attendant ethos and style of expression,⁵ at least two-thirds of the contents of both the provincial (juren) and metropolitan (jinshi) exams were “codified knowledge” and were thus readily transferrable through writing or verbalization. In this regard, family background conferred little additional advantage.

But things were different when it came to the palace exam: a point had been reached where administrative competence and political leadership were now being called for if one were to be given a senior appointment in the government. The disproportionate significance of cewen 策問, or policy questions, in determining a jinshi’s performance in the palace exam can be gleaned from the fact that it was now the only subject to be examined. The cewen dealt with thorny issues with which the emperors were confronted. Moreover, as the Qing emperors found themselves increasingly immersed in conflicts with both foreigners (especially the British) and rebels, cewen became extraordinarily important. In the fifty or so palace exams that took place within our period of analysis, for instance, approximately half of the questions concerned military strategy and tactics, more than double the proportion in the early Qing period (Wang, 2013). A prominent example is the Opium War, which troubled Emperor Daoguang 道光 (r. 1821–1850) more than any other issue.⁶ The same was the case with Emperor Xianfeng 咸豐 (r. 1850–1861), who, facing the threat of the Taiping rebels, was desperate to find out what the jinshi candidates saw was the most effective way of defeating them. The answer that the emperor found most persuasive was provided by Optimus 状元 Sun Rujin 孫如僅, whose knowledge of both military tactics and geography was so thorough and detailed that the emperor considered an instantaneous application feasible (Deng and Gong, 2006: 1899).

But how exactly did one become distinctly more knowledgeable than others in this realm, given that there were literally no institutions or market from which one could acquire such knowledge, and that such knowledge assumed almost no importance prior to the palace exam? In the historical context of China, the family was the primary mechanism through which such tacit knowledge could be transmitted. The reason is straightforward, because knowledge of this nature was essentially distilled from experiences deeply grounded in officialdom. As such knowledge is difficult to transfer via writing down explicit codes, it could only be transmitted—in the majority of circumstances anyway—within the context of the family from one generation to another.⁷ For example, Zhang Zhidong’s 張之洞 excellent performance in the palace exam (he came in third) must have been inspired by his father’s experience of defeating the Taiping rebels as a prefect in Guizhou (Wu, 2009: 17). Long Qirui 龍啟瑞, the Optimus of 1841, offers another example. The son of a juren who not only possessed a wealth of experience in suppressing the pirates on China’s southeastern coastal seaboard but also displayed unquestionable competence in local governance (Long, 2002 [Qing]; Deng and Gong, 2006), Long was unwittingly given the question in the palace exam of how to deal with a naval intrusion (Deng and Gong, 2006: 1844).

We are not suggesting that knowledge of statecraft could only be transmitted within the family. Not everyone was born with such blessings. But those who could not rely on their family for knowledge of public sector management could acquire such knowledge by, for instance, working as assistants to experienced officials. In Qing China, many middle- to high-level officials set up their own commanding offices, essentially think tanks 幕府, and recruited assistants to help with administration, which turned out to be an effective training ground for candidates who performed poorly initially in the provincial and metropolitan exams (Shang, 1999: 11).

Zhang Jian 張謇, the famous Optimus of the 1898 palace exam, is a good case in point. Coming from a commoner background, Zhang did not have the wherewithal to prepare for even the provincial exam and was thus forced to work for an income (Zhang, 1994 [Qing]). Luckily, he was employed by General Wu Changqing 吳長慶—one of the leaders of the Huai Army who later was promoted to be the general in charge of Zhejiang province in 1880—to be his assistant. General Wu had keenly supported Zhang throughout his exam career. Altogether it took Zhang fully eight years to pass the juren exam (from when he was a stipend student 廩生), not to mention another four failed attempts at the jinshi exam, he eventually passed it. The knowledge, in particular that pertaining to military tactics which Zhang gained while working for General Wu, must have enabled his excellent performance in the palace exam.

Governor-general Lin Zexu 林則徐, the nation’s hero famous for burning the British opium, offers another example of this kind. In Fujian province, where his father taught in a private academy, Lin received the rare chance to work as an assistant to the governor, Zhang Shicheng 張師誠, for four years (Lai, 1997: 60). The position provided ample opportunities for him to acquire knowledge of the military and law—knowledge that would have been near impossible to acquire for those without ancestors holding middle-level offices or above. Lin ended up obtaining his jinshi qualification with flying colors and entered the Hanlin Academy at the young age of twenty-seven (Lai, 1997: 66).

Examples such as the ones enumerated above, though impressive, were actually few and far between. This adds to the difficulty of collecting these cases in a systematic manner for empirical testing. For this reason we focus on the family as the institutional context for transmitting cultural capital.

The Making of the Upper Gentry

Data and Variables

The data we employ for analysis have been obtained from two separate historical data sources. The first is A Compilation of Civil Examination Vermilion-Ink Essays in the Qing Era 清代硃卷集成 (Exam Essays hereafter), edited by Gu Tinglong 顧廷龍 (1992). This source contains a list of 4,035 juren who passed the provincial exam during the period of our analysis, accounting for about 8 percent of the total juren population (of fifty thousand).⁸ While being strikingly similar to the source that Ho (1967) used in his seminal study of social mobility in historical China, namely, the Record of Graduates of the Provincial Civil Examinations 鄉試同年齒錄, Exam Essays has the additional advantage of including detailed information on the maternal relatives of the previous three generations. This additional information addresses the biggest concern of Ho’s critics—the influence from the mother’s side (e.g., Hymes, 1986). Moreover, Exam Essays also contains detailed information on the location of one’s residence, which provides an additional check of the data’s accuracy.

The second source is a compilation by Zhu Baojiong 朱保炯 and Xie Peilin 謝沛霖 (1980) titled An Index to the Ming-Qing List of Jinshi Degree Holders 明清進士題名碑錄索引 (Index for short), which provides a list of all the jinshi who eventually passed the metropolitan and palace exams. Merging the two lists provides us with the useful information of who among the juren had passed the jinshi exam and specifically also their performance. In matching the two lists of names we rely also on the additional information regarding the examinees’ home county 籍貫縣 for accuracy (which is available in both Exam Essays and Index).

Exam Essays contains a total of 4,035 essays by juren from all eighteen provinces of Qing China who passed the provincial exam during the 1796–1895 period.

In the Qing dynasty the probability that a juren would become a jinshi was about 18 percent.⁹ Figure 2 divides the temporal distribution of the number of juren in our sample into nine periods. Although the number was much smaller before than after 1850, the passing rates, hovering around 18 percent, were similar.

Figure 2.

Passing rate of jinshi in the metropolitan exams.

We choose the year 1796 as the starting point of our analysis for the simple reason that the sample before 1796 is grossly underrepresented in the Exam Essays. Only 67 juren were enumerated during this period, amounting to a mere 1.63 percent of the overall sample spanning nearly two centuries. We use the passing rate (the number of juren who passed the jinshi exam as a percentage of all juren who took the jinshi exam) as a guide to gauge sample representativeness. Before 1796 the passing rate was only 10 percent—well below the average passing rate. In contrast, at around 18 percent the passing rate after 1796 is nearly identical with the average, reassuring us of the credibility of the sample we employ (see Figure 2 for details).

We end our analysis in 1895 because the civil exam was abolished in 1905 and assume that it would take up to ten years if not longer for a juren to attempt the jinshi exam again. Ending our analysis before the abolition of the civil exam should allow the last batch of candidates a more or less equal chance of passing the jinshi exam.

Concerning the geographic distribution of the juren population, Figure 3 shows that it was disproportionately concentrated in the eastern coastal provinces of Jiangsu, Zhejiang, and Anhui. Regional variations in the jinshi passing rates are analyzed in detail in the Appendix.

Figure 3.

Regional distribution of juren awardees (1796–1895).

In addition to providing individuals’ personal information, Exam Essays also contains the detailed vitae of all the jurens’ fathers, grandfathers, and even great-grandfathers, including the level of their exam success (if any) and official rank (if any). This detailed information thus allows us to construct a list of independent variables pertaining to one’s ancestors’ exam achievements and official rank. These variables together make up the analytical construct of “family type or background”—a variable to which Ho (1967) and others have resorted to in the debate over the extent of social mobility in late imperial China. As mentioned earlier, our data allow us to construct this variable on the mother’s side as well. In addition, we can employ exam achievements to proxy for family education and official rank for family-specific tacit knowledge. Finally, as Exam Essays also enumerated examinees’ age when they passed the juren exam as well as the ranking thus attained, we are further able to construct two separate measures to proxy for ability.

Dependent Variables: Passing the Metropolitan Exam and Performance in the Palace Exam

We have three dependent variables altogether. The first is whether a juren passed the metropolitan exam—a binary variable for which we simply assign the value of 1 to those who passed the metropolitan exam, and 0 otherwise, whereas the second and third pertain to, respectively, the ranking and class of honors attained in the palace exam. In the palace exam, each jinshi candidate received a ranking and was sorted into a particular class of honors. For instance, in the 1822 palace exam a total of 222 candidates passed the exam. The top three were classified as having attained first-class honors, the next one hundred second-class honors, and the remaining 119 third-class honors. To enable the comparison of jinshi ranking across time and region we employ the following formula—Equation 1—to calculate a standardized score for each jinshi, where the score ranges from 0 to 100:

s t d . r a n k i n g = \frac{t o t a l n o . o f j i n s h i i n a g i v e n e x a m - a c t u a l r a n k i n g}{t o t a l n o . o f j i n s h i i n a g i v e n e x a m} \times 100

(1)

For someone who was tenth in the second-class category (thirteenth in actual ranking since the three first class jinshi also rank ahead of him), for example, the standardized score is 94.14 ([222 − 13] × 100/222). For the class of honor variable, we assign 1 to first- and second-class honors, and 0 to those with third-class honors.

Independent Variables

Family Background/Type

The late Ho Ping-ti (1967) classified families in late imperial China into four distinct categories based on the exam achievements of a candidate’s paternal ancestors for up to three generations. First, “commoner” families were those whose ancestors had never attained any—not even shengyuan—academic status, and, thus never held a position in the government. Next were “scholar-commoner” families, whose ancestors had attained the qualification of a shengyuan or equivalent only, but still held no office. One step above were “gentry” families, which include all those who had either produced one or more degree holders higher than shengyuan—typically the juren, or held offices or had official titles. Finally, Ho singled out families with at least one ancestor who had attained an office in the government at the provincial level or above as “upper gentry.” To further differentiate “upper gentry” families from lesser counterparts in our empirical analysis, we redefine the “upper gentry” as those with one or more family members who attained the jinshi degree or held an official position at the provincial level or above, and the “lower to middle gentry” as those who had a member who obtained a juren degree or held a government position below the provincial level.

Ability

Ho’s mobility thesis implies that ability rather than family background was the key to civil exam success. To test this hypothesis, we construct two variables to proxy for ability.

Following Robert Marsh (1961), we use a candidate’s age upon passing the provincial exam to proxy for ability, presumably because the younger an examinee was when he passed the juren exam the more competent he was. Indeed, our juren sample does reveal that the candidates who subsequently passed the national exam were nearly three years younger than those who failed to advance further (28.88 years versus 31.23 years) at the time of passing the provincial exam. The mean age upon obtaining the juren degree was 31.24 (Table 1)—strikingly close to what others have found (e.g., Chang, 1955; Elman, 2000).

Table 1.

Descriptive Statistics of Key Variables.

	Mean	SD	Min	Max
Achievements in national exams
Passing metropolitan exam (fail = 0)	0.18	—	0.0	1.0
Ranking in palace exam^a	60.39	28.84	0.0	99.7
1st and 2nd class jinshi in palace exam (3rd class = 0)^a	0.58	—	0.0	1.0
Family background
Paternal family types^b
Commoner	0.12	—	0.0	1.0
Scholar-commoner	0.47	—	0.0	1.0
Lower and middle gentry	0.32	—	0.0	1.0
Upper gentry	0.09	—	0.0	1.0
Maternal family types^b
Commoner	0.36	—	0.0	1.0
Scholar-commoner	0.37	—	0.0	1.0
Lower and middle gentry	0.21	—	0.0	1.0
Upper gentry	0.06	—	0.0	1.0
Father’s education	2.02	2.15	0.0	7.0
Father’s official rank	1.68	3.78	0.0	19.0
Average education of grandfather and great-grandfather	1.38	1.52	0.0	7.0
Average official rank of grandfather and great-grandfather	1.09	2.67	0.0	18.5
Average number of wives and concubines of father and grandfather	1.52	0.63	1.0	8.0
State student as pre-juren qualification	0.09	—	0.0	1.0
Achievements at provincial exam
Age upon obtaining juren degree	31.24	8.24	14.0	64.0
Ranking^c	55.96	27.62	0.0	99.7
Observations	4,035

Source. Gu, 1992.

The number of observations for which information is available for these two measures is 728.

Paternal family type is constructed based on the highest educational level and official rank of a jinshi’s father, grandfather, and great-grandfather (Ho, 1967: 107), whereas maternal family type is constructed based on the highest educational level and official rank of a jinshi’s grandfather and one other important maternal relative—typically the maternal great-grandfather or a maternal uncle.

A candidate’s standardized ranking in the provincial exam is calculated based on his original ranking in the provincial exam and quota of that year.

The biggest drawback of using age as a partial indicator of ability is that it may potentially capture the effect of wealth, since a wealthy family could simply purchase a studentship from the imperial college, a qualification that enabled one to sit for the juren exam much earlier than did the shengyuan. However, only 9 percent of the 4,035 juren in our sample belonged to this particular category; in addition, we control for this possible effect in our analysis (see the section on pre-juren qualifications below).

The second pertains to a candidate’s ranking in the provincial exam. But since the juren exam took place in different provinces, with multiple candidates across these provinces attaining essentially the same rank (in their own province), to allow for comparisons both across provinces and over time we need to standardize their ranking using a formula similar to Equation 1. According to this formula, a juren ranked first in a province with a quota of 120 would obtain a standardized score of 99.16 ([120 − 1] × 100/120), whereas the one who came last obtained a score of zero. The average standardized score based on our sample is 56.

Decomposing Family Background

Useful as it is, family background is likely a noisy measure because it lumps together the various advantages conferred by different types of resources across generations. For instance, while parental education may have a direct effect on the son’s ability, it may also have an indirect effect on his educational success by, for instance, increasing the time devoted to nurturing him (e.g., Becker, 1962). Moreover, while parental educational achievements are clearly related to official rank, the two may affect exam performance of the succeeding generation(s) via different mechanisms. Families with a history of distinguished careers in state bureaucracy may confer certain unique advantages such as knowledge of statecraft on their descendants. Owing to these reasons, we construct as many as four separate measures to proxy for family background; they include: the exam achievements and official rank of one’s ancestors, the average number of wives and concubines that one’s father and grandfather had, and whether one held a jiansheng degree (see below for details).

Family human capital (father’s/grandfather’s and great-grandfather’s examination achievements)

To proxy for a family’s human capital we code the father’s educational attainment as an ordinal variable ranging from no degree at all (0), a jiansheng (purchased) degree (1), a shengyuan degree of various ranks attained through the series of examinations in the county school (2 to secondary students 附生, 3 to extra students 增生, 4 to stipend students 廩生, and 5 to tribute students 貢生), all the way up to a juren degree (6) and a jinshi degree (7). In the case of the grandfather and great-grandfather, we compute the mean of their educational attainment instead.

The mean of the father’s education level is 2.02 (Table 1), equivalent to a low-level shengyuan. At 1.38, the average education level of the grandfather and great-grandfather is noticeably lower than that of the father, suggestive of mobility.

Tacit knowledge (father’s/grandfather’s and great-grandfather’s official rank)

To proxy for the possible effect of tacit knowledge (under the format of policy questions) we construct a variable based on the official rank of a candidate’s father and the average official rank of the grandfather and great-grandfather. Intuitively, the higher the official rank the more valuable such knowledge and experiences tended to be.

There were altogether nine official grades. But since corresponding to each of these grades there was a deputy assisting the chief, we code official rank on a 20-point scale from 0 to 19, with 0 assigned to those whose father held no official rank at all, and 1 to those whose rank was too low for classification 未入流. Next in the pecking order was the lower ninth rank (9b) corresponding to a score of 2 in our coding scheme, all the way up to the highest grade of upper first rank, having a score of 19 (1a). Based on this coding scheme the mean father’s official rank is 1.68, which is close to the lower ninth rank (9a)—very low in status indeed. At 1.09 the average official rank of the grandfather and great-grandfather was lower than that of the father, albeit not by much.

Wives and concubines of the father and grandfather

We also control for any indirect effect that family wealth may have on exam success. As data on landholdings are not available, we employ the number of wives and concubines that a juren’s father and grandfather had before he passed the juren exam as a crude proxy, on the grounds that in a premodern society wealth is usually highly correlated with the number of wives and children (e.g., Myers, 1970; Liu, 1992).¹⁰ The mean of this variable is 1.52 (Table 1).

Pre-juren qualifications

To mitigate the effect of wealth through the purchase of a jiansheng degree, we control for the pre-juren degree type. The usual binary (0, 1) classification applies.

Control Variables

In addition, we control for the variations in the number of male siblings, birth order, type of residence (urban versus rural), and regions and periods.

Siblings and birth order

To deal with resource dilution, we control for the number of brothers a jinshi candidate had and his own birth order in the family in the analysis. The mean number of male siblings in our sample is 2.16, with close to half of them (0.46) being the first-born.

Urban versus rural residence

We also control for this dimension just in case an urbanite enjoyed advantages over his rural counterparts in the civil exams.

Region fixed effects

To rule out the possibility that variations in our dependent variables may be due to the varying spatial impact of regional characteristics, particularly differences in the jinshi quota, we include regional fixed effects in the regression analysis. These broader regions include north China, south China, western China, central China, and the two special regions of Jiangsu/Anhui provinces, and Zhejiang province.

Period fixed effects

To control for the possible effect of the different time periods in which a candidate took the jinshi exam on exam performance, we separate the candidates according to the decade in which they took the exam. The first decade was from 1796 to 1805 and the last was from 1886 to 1895, with eight decades in between. As too few jinshi passed the exam during the two decades of 1796–1815, we combine them into one period (1796–1815), resulting in a total of just nine periods.

Estimation Strategy

In our analysis, where the dependent variable is whether one passed the jinshi exam, we employ a logit regression model, with specifications designed to test the competing hypotheses regarding the roles of ability and family background in determining success in China’s civil exams:

I n (\frac{p_{i}}{1 - p_{i}}) = X_{f a m i l y} β_{1} + X_{a b i l i t y} β_{2} + X_{w e a l t h} β_{3} + Z λ

(2)

where p is the probability of passing the metropolitan exam. In Equation 2, our baseline model, the key independent variable is family resources, which initially is measured by family type but is later measured by family education, official rank, and wealth in Equation 3. Ability measures enter into the analysis in both settings, and Z represents a number of control variables.

In estimations in which we employ a jinshi’s ranking attained in the palace exam as the second dependent variable, we employ the ordinary least squares method (OLS, Equation 3) in estimating ranking on a smaller sample that is conditional upon only those who passed the metropolitan exam.

r a n k i n g = X_{f a m - e d u} β_{1} + X_{f a m - o f f} β_{2} + X_{a b i l i t y} β_{3} + X_{w e a l t h} β_{4} + Z λ

(3)

Finally, to check the robustness of our results based on palace exam ranking we apply the sequential logit model to the determinants of passing the metropolitan exam first, and conditional upon passing it, to the determinants of attaining a certain class of honors in the palace exam (Mare, 1980; Buis, 2007). The sequential logit model can be written with notations similar to the ones in Equation 3, as follows:

I n (\frac{p_{i j}}{1 - p_{i j}}) = X_{f a m - e d u} β_{1 j} + X_{f a m - o f f} β_{2 j} + X_{a b i l i t y} β_{3 j} + X_{w e a l t h} β_{4 j} + Z λ_{j}

(4)

Analytical Results

Did the Keju Really Provide Social Mobility? Descriptive Evidence

Ho’s argument that Chinese society was highly mobile is premised on his observation that in the nineteenth century as many as 45.1 percent of juren and tribute students came from commoner and scholar-commoner families. While the figure for jinshi was slightly lower, it still accounted for a substantial 37.6 percent. This corroborating evidence led Ho to conclude that “during the Ming-Ch’ing period there were various institutionalized and noninstitutionalized channels which promoted the upward mobility of the humble and obscure” (Ho, 1967: 257).

Not everyone accepts Ho’s optimistic assessment of the civil examination system in facilitating social mobility in Qing China, however, because the civil exam was such a long-drawn process, taking an average of twenty years to complete (Chang, 1955). During this lengthy process, continuous financial support from the family and/or lineage was absolutely crucial. Moreover, the fact that no limits were imposed on age or number of attempts in taking the exam only prolonged the duration of financial support required, thereby putting the wealthy in an advantageous position (Elman, 2013; Fairbank, 1983). Thus, all things considered, the civil exam system is still considered by some as merely facilitating the transmission of social status between generations of the privileged.

To address this question we first construct a mobility table in the context of four generations using the Exam Essays data set. Multigenerational analysis is especially pertinent in our context because it allows us to examine whether the alleged advantages of family background could extend beyond one generation. For example, using genealogical records of the Ming and Qing dynasties (circa 1300–1900), Shiue (2017) finds that the transmission effect was weak for the grandfather but strong for the father and the uncle. However, using the household registration data of 1789–1909 in Liaoning province as their sample, Campbell and Lee (2003, 2008) find weak evidence of intergenerational mobility; both the grandfather and the uncle had only a limited influence on one’s chance of obtaining official titles. But a major limitation of these analyses is that they are confined to a particular region of China. We attempt to overcome this limitation by using a national sample.

To conduct a multigenerational analysis, we first tabulate the exam outcomes involving four generations based on the juren sample. For example, Panel A of Table 2 makes use of the distribution of four generations of candidates across the four levels of exam achievements, including those who became a jinshi and those who failed to achieve even the lowest degree—the commoners.¹¹ But we go one step further. In Panels B through D we compute the probability of one’s exam success conditional upon one’s ancestors’ own success, for example the father’s achievement depending on the grandfather’s achievement.

Table 2.

Distribution and Mobility of Juren and Their Ancestors’ Exam Achievements.

	No degree	Below juren	Juren	Jinshi	Total
	No degree	Below juren	Juren	Jinshi	%	N
A. Distribution of exam achievements, by generation
Great-grandfather (GGF)	45.48	48.50	4.34	1.69	100	4,035
Grandfather (GF)	39.23	51.67	6.64	2.45	100	4,035
Father (F)	30.16	56.98	9.05	3.82	100	4,035
Son (S)^a			81.91	18.09	100	4,035
B. Distribution of the grandfather’s exam success conditional on the great grandfather’s education
GF\|GGF = No degree	58.91	35.86	3.54	1.69	100	1,835
GF\|GGF = Below-juren^b	23.15	67.19	7.41	2.25	100	1,957
GF\|GGF = Juren	22.29	44.57	25.71	7.43	100	175
GF\|GGF = Jinshi	14.71	50.00	19.12	16.18	100	68
	Pearson chi²(9) = 729.8184 Pr = 0.000
C. Distribution of the father’s exam success conditional on the grandfather’s education
F\|GF = No degree	46.56	45.17	5.50	2.78	100	1,583
F\|GF = Below-juren^b	20.05	67.24	9.74	2.97	100	2,085
F\|GF = Juren	16.42	51.87	21.64	10.07	100	268
F\|GF = Jinshi	18.18	43.43	17.17	21.21	100	99
	Pearson chi²(9) = 506.7355 Pr = 0.000
D. Distribution of the son’s (juren) exam success conditional on the father’s education
S\|F = No degree			82.91	17.09	100	1,217
S\|F = Below-juren^b			82.86	17.14	100	2,299
S\|F = Juren			76.16	23.84	100	365
S\|F = Jinshi			73.38	26.62	100	154
	Pearson chi²(3) = 17.9245 Pr = 0.000

Source. Gu, 1992.

Of the 4,035 juren, about 18 percent passed the metropolitan (jinshi) exam. In their “essays,” the exam achievements of their father, grandfather, and great-grandfather were reported.

“Below-juren” degree holders refers to licentiates, tribute students, and state students (see Figure 1 for details).

Before we examine Table 2 in detail, we use the specific example of Pan Zengshou 潘曾綬 of Suzhou prefecture in Jiangsu to illustrate how to interpret the results therein (Gu, 1992: 98.149–53). Pan’s great-grandfather had not attained any exam success whatsoever, but his grandfather became a tribute student, an attainment just below that of a juren. On the heels of his grandfather’s humble achievement, Pan’s father became a jinshi (actually he attained the Optimus, the highest honor), while Pan himself slipped but still managed to become a juren eventually. With four generations having attained different levels of exam success, Pan’s family demonstrates the possibility of both upward and downward mobility across generations.

Panels B through D then reveal the probability of exam success in the civil exam, conditional on the success (or failure) of one’s immediate ancestor. In Panel B, for example, the probability of Pan’s grandfather becoming a tribute student was 35.86 percent, conditional upon his great-grandfather being a commoner or non-degree holder. In Panel C, the probability of Pan’s father becoming a jinshi was a much smaller, 2.97 percent, conditional upon his grandfather being a mere tribute student. Finally, in Panel D, the probability of Pan himself slipping down the mobility ladder was a huge 73.38 percent, testifying to the enormous difficulty of staying at the top once the family had reached the highest rung on the ladder of success in the exam system. Pan’s example thus clearly illustrates how in late imperial China a family could conceivably experience both upward and downward mobility through the civil exams.

Going beyond the sheer probability of exam success, two results stand out from this analysis of intergenerational mobility. Foremost is that family background indeed mattered. Specifically, the chances of the next generation ending up as a juren or jinshi were far greater among those whose fathers were also a juren or jinshi (e.g., Panels B and C). For example, while 21 percent of the sons of jinshi managed to retain the jinshi status, only 2.78 percent of those with a commoner father became a jinshi themselves (Panel C). As for the juren in our sample, those with a father who was also a juren or jinshi respectively enjoyed a 6.75 percent (23.84% – 17.09%) or 9.53 percent (26.62% − 17.09%) higher passing rate in the jinshi exam than those whose father was a commoner (Panel D). There is thus an element of truth in the “status reproduction” thesis.¹²

However, Table 2 also reveals that there was substantial downward mobility in the process. Using the same two panels for the purposes of illustration, only slightly more than 30 percent of the juren families were able to retain the same status (or higher) in the next generation (e.g., 25.71% + 7.43% in Panel B, 21.64% + 10.07% in Panel C), suggesting that fathers who were a jinshi themselves had a harder time ensuring their children performed equally well in the exam (about 16% in Panel B and 21% in Panel C). On the whole, in our sample more than 60 percent of those with a juren or jinshi father failed to attain the same or higher status, i.e., they experienced downward mobility. These results therefore suggest that upward mobility—achieved presumably via one’s ability—was clearly also a feasible option. While our analysis may have overstated the extent of upward social mobility given that the juren were already winners in the civil exam, the fact that jinshi status had to be attained and could not be purchased and the evidence of many jinshi having slipped from the apex of the civil exam pyramid cast doubt on the claim that the entire Qing bureaucracy was monopolized by the several hundred famous lineages who invariably were winners of the civil exam tournament over the generations (Zhang, 2003).

Family Type versus Ability

To examine the importance of family background relative to ability in determining success in the metropolitan examination, we turn to regression analysis. Table 3 reports the results of a logit model that examines the probability of passing the jinshi (metropolitan) exam based on family background or type on the one hand, and on the candidate’s measurable ability on the other hand. Using the “commoner” family as the reference group for comparison, we construct the variable family type based on family members on both the paternal (three immediate ancestors) side and the maternal side (but only the grandfather and the uncle) with the highest academic-cum-official career achievements. Columns 1 and 2 report the baseline model results using only family type from one side as the explanatory variable, whereas column 3 combines both paternal and maternal sides. In column 4 we report the results of respectively the two ability measures and the wealth measures. Column 5 is the full model, which includes a number of control variables outlined in the previous section.

Table 3.

Social Class and Metropolitan Exam Performance (Logit).

DV = Whether one passed the metropolitan exam	(1)	(2)	(3)	(4)	(5)
Paternal family type^a
Scholar-commoner	−0.092		−0.100		−0.082
Scholar-commoner	(0.136)		(0.140)		(0.144)
Lower and middle gentry	0.093		0.088		0.009
Lower and middle gentry	(0.140)		(0.146)		(0.152)
Upper gentry	0.587***		0.586***		0.477**
Upper gentry	(0.168)		(0.176)		(0.189)
Maternal family type^a
Scholar-commoner		0.018	0.034		0.010
Scholar-commoner		(0.097)	(0.100)		(0.104)
Lower and middle gentry		0.068	−0.002		−0.065
Lower and middle gentry		(0.111)	(0.116)		(0.120)
Upper gentry		0.198	0.009		−0.026
Upper gentry		(0.170)	(0.178)		(0.183)
Age upon obtaining juren degree				−0.046***	−0.046***
Age upon obtaining juren degree				(0.006)	(0.006)
Ranking in provincial exam				0.008***	0.008***
Ranking in provincial exam				(0.002)	(0.002)
Average number of wives and concubines				−0.025	−0.056
Average number of wives and concubines				(0.067)	(0.072)
Pre-juren qualification (State students =1, else = 0)				−0.038	−0.154
Pre-juren qualification (State students =1, else = 0)				(0.142)	(0.149)
Controls	No	No	No	No	Yes
Constant	−1.563***	−1.545***	−1.570***	−0.543**	−0.597
Constant	(0.121)	(0.069)	(0.124)	(0.226)	(0.428)
Observations	4,035	4,035	4,035	4,035	4,035
Pseudo R²	0.007	0.000	0.007	0.028	0.039
Log likelihood	−1,895.08	−1,906.92	−1,895.00	−1,854.72	−1,832.32

Note: Standard errors are in parentheses. Control variables include number of brothers, birth order, residence type (urban vs. rural), period, and region dummies. DV = dependent variable.

Commoner is the reference group.

***

p < 0.01, ** p < 0.05, *p < 0.1 (two-tailed tests).

Our foremost finding is that, while neither the “scholar-commoner” nor “lower to middle gentry” families exhibited significant advantages over “commoner” families (the reference group) in passing the metropolitan exam, “upper gentry” families clearly did (column 1). Specifically, juren from “upper gentry” families enjoyed a 61.1 percent advantage (exp^0.477 − 1, model 5) over commoners in the net odds of passing the metropolitan exam—a finding that casts doubt on the mobility thesis.

But the same does not apply to the family type based on the maternal side (column 2). This may suggest that influence from the mother’s side is not significant—a finding that seemingly reflects the patriarchal nature of the Chinese society. However, the insignificance may have been due to the coarseness of the data on the maternal side. The Qing government required a candidate to report the educational level and official position of his three immediate paternal ancestors but not those of his maternal ancestors. Besides, as the mother (rather than her brothers and/or father) was most likely the provider of her son’s education, the academic and career credentials of a candidate’s maternal relatives would be a poor proxy for measuring influences on the maternal side. These findings do not change in the third model (column 3 of Table 3); families classified as “upper gentry” based on the father’s side remain just as significant (in fact with a slightly larger coefficient), whereas the same social class on the mother side continues to be insignificant.

But Ho’s thesis that Chinese society in late imperial times was upwardly mobile is supported by the finding that the two proxies for ability both bear significantly upon exam success (columns 4 and 5). In terms of magnitude, a 10 percent increase in ranking in the provincial exam has the effect of increasing one’s chance of passing the jinshi exam by 8.3 percentage points (exp^0.008*10 − 1, calculated based on the full model, column 5). Likewise, a one-year increase in age upon passing the juren exam has the effect of decreasing the odds of passing the jinshi exam by 4.5 percent (exp^−0.046 − 1, column 5).¹³

Did wealth improve one’s chances of success in the exams? Using the number of wives and concubines that one’s father and grandfather had and whether one had purchased his jiansheng degree as proxies, our finding is decidedly in the negative (columns 4 and 5, Table 3). One possible explanation is that at the highest level of competition material support from a family weighed much less than sheer intellect and determination. In other words, juren holders were a group of highly accomplished individuals who already met a certain threshold in terms of the wealth needed to advance to candidacy, so that additional wealth would not matter much. This view is highly consistent with a classic sociological finding that the influence of family background on progression to the next rung in the educational system is distinctly smaller for the higher rungs (Mare, 1980).¹⁴ It could also be that, once a candidate achieved the status of a juren chances were great that his relatives and friends would be eager to invest in his future exam and career. In addition, there was also the so-called binxing organization 賓興會, essentially a community chest to provide travel expenses for one to sit for the jinshi exam, leveling the playing field (Ho, 1967; Mao, 2007). Of course, we cannot rule out the possibility that these proxies are simply too noisy to capture the effect of family wealth. Regardless of which of these explanations best fits the Chinese situation, there is no denying that wealth still played a role in the earlier stages of the civil exam.

Indeed, in the nineteenth century, as the Qing state began to face a dire need of revenue to counter threats to its survival—from the Taiping rebels, for instance (Wakeman, 1975)—wealth arguably played an increasingly influential role in officialdom. Those with abundant wherewithal were now in a position to voluntarily exit the jinshi tournament by purchasing an office in the government instead (Zhang, 2010).¹⁵ However, evidence suggests that county magistrate posts, the position to which the juren holders were entitled but which was not always immediately available, were increasingly accounted for by jiansheng—state students who had purchased their degrees. Their numbers jumped from 16.9 percent in the Jiaqing 嘉慶 reign (c. 1796–1820) to 35.9 percent in the Guangxu reign (c. 1875–1908) (Lee, Chou, and Hsu, 1975). This finding suggests that office purchase by some juren was more likely a forced response to the heightened competition from lower-degree holders who were also entitled to the same level of appointment as the juren, rather than a conscious strategy conceived to exit the jinshi tournament. Of course, a more conclusive answer will have to await further research.

Finally, in the full model (column 5), which includes a number of important controls, the upper-gentry variable remains significant, although the level of significance is now reduced from 1 percent to 5 percent and with a smaller coefficient. The same is true of the two ability measures, which remain significant in the full model. To check whether “upper gentry” families enjoyed an advantage over “commoner” families we changed the reference group from “commoner” to “upper gentry” and found that the latter continued to enjoy a significant edge over all the other three social groups (and hence we do not separately report the results), reaffirming our previous result.

Summing up, regression analysis based on the analytical construct of family type and ability does lend partial support to Ho’s thesis that Chinese society in late imperial times was mobile. However, inequality in social origins did have an important bearing on jinshi exam success. Specifically, and consistent with the findings summarized in Table 2, jinshi from an upper gentry background did enjoy a distinct advantage over their counterparts of much humbler origins. We now turn to examine the possible mechanisms by which these “upper gentry” families enjoyed their advantages by decomposing family type into two distinct dimensions—human (educational) and cultural (official rank) capital.

Identifying the Advantages of “Upper Gentry” Families

Table 4 similarly shows the odds of passing the metropolitan exam with respect to the exam achievements and official rank of both a candidate’s father, his grandfather and great-grandfather, the various measures constructed to proxy for the effect of family wealth, and the same control variables.

Table 4.

Paternal Education and Office and Metropolitan Exam Performance (Logit).

DV = Whether one passed the metropolitan exam	(1)	(2)	(3)	(4)
Father’s education	0.062***		0.062***	0.054**
Father’s education	(0.021)		(0.021)	(0.022)
Father’s official rank	0.024**		0.018	0.015
Father’s official rank	(0.011)		(0.012)	(0.012)
Average education level of GF and GGF		0.019	−0.014	−0.028
Average education level of GF and GGF		(0.031)	(0.033)	(0.033)
Average official rank of GF and GGF		0.038**	0.033*	0.029
Average official rank of GF and GGF		(0.017)	(0.017)	(0.018)
Average number of wives and concubines				−0.060
Average number of wives and concubines				(0.072)
Pre-juren qualification (State students = 1, else = 0)				−0.149
Pre-juren qualification (State students = 1, else = 0)				−0.149)
Age upon obtaining juren degree				−0.045***
Age upon obtaining juren degree				(0.006)
Ranking in provincial exam				0.008***
Ranking in provincial exam				(0.002)
Controls	No	No	No	Yes
Constant	−1.687***	−1.583***	−1.696***	−0.583
Constant	(0.058)	(0.056)	(0.064)	(0.419)
Observations	4,035	4,035	4,035	4,035
Pseudo R²	0.006	0.003	0.007	0.040

Note. Standard errors are in parentheses. Control variables include number of brothers, birth order, residence type (urban vs. rural), period, and region dummies. DV = dependent variable; GF = grandfather; GGF = great-grandfather.

***

p < 0.01, **p < 0.05, *p < 0.1 (two-tailed tests).

Given that the father’s level of education is likely correlated with those of the grandfather and great-grandfather, we first enter them separately in columns 1 and 2, before putting them together in column 3. We do the same for official rank. The results reveal that, while the father’s education and official rank are both significant in the baseline model (column 1), only the former remains significant in the full model; moreover, the level of significance in the full model is reduced from 1 percent to 5 percent (column 4). In terms of magnitude, a one-level increase in the father’s education has the effect of raising the odds of passing the metropolitan exam by 5.5 percent (exp^0.054 − 1, column 4). While completely different in context, the finding that the father’s exam achievements have a significant impact on the son’s achievements nonetheless is consistent with conventional wisdom (e.g., Holmlund, Lindahl, and Plug, 2011).

Neither the exam achievements nor official rank of the grandfather and great-grandfather is significant in the full model (column 4), suggesting that the grandfather’s education had no additional effect on the grandson’s education. In column 4, in which all controls are included, the two ability measures still dominate over the two wealth measures, a result consistent with those shown in Table 3.

Success in the Palace Exam

We now examine the more specific outcome of how jinshi fared in the palace exam, using their exam ranking first and then class of honors to check robustness. Table 5 reports the OLS results on ranking. First, the father’s education is no longer significant in determining a jinshi’s ranking in the palace exam (columns 1 and 2), but the father’s official rank now has a distinctly significant effect (1%) on a jinshi’s ranking in the OLS estimates (columns 1 and 2). This new result suggests that the father’s official rank can predict how well a candidate performed at the final stage of the jinshi exam rather than simply predicting pass or fail. The result is consistently similar in the case of the grandfather and great-grandfather (column 2). Unlike human capital, which did not have a multigenerational effect, tacit knowledge or cultural capital had a cumulative effect across generations.

Table 5.

Determinants of Ranking and Class of Honors in the Palace Exam.

	OLS			Sequential Logit
	Ranking	Ranking	Ranking	Jinshi vs. Juren	1st and 2nd vs. 3rd class
	(1)	(2)	(3)	(4)	(5)
Father’s education	0.541	0.589	0.589	0.051**	0.057
Father’s education	(0.505)	(0.508)	(0.508)	(0.022)	(0.043)
Father’s official rank	0.877***	0.876***		0.016	0.042*
Father’s official rank	(0.276)	(0.278)		(0.012)	(0.025)
Average education level of GF and GGF	−0.391	−1.046	−1.055	−0.027	−0.087
Average education level of GF and GGF	(0.806)	(0.816)	(0.821)	(0.033)	(0.070)
Average official rank of GF and GGF	0.976**	1.272***	1.310***	0.030	0.112***
Average official rank of GF and GGF	(0.429)	(0.433)	(0.437)	(0.018)	(0.040)
Father’s office level (no office = 0)
Lower (>0, <7a)			6.752*
Lower (>0, <7a)			(3.797)
Middle (≥7a, <3b)			7.301**
Middle (≥7a, <3b)			(3.304)
Higher (≥3b)			15.634**
Higher (≥3b)			(6.054)
Average number of wives and concubines	0.718	0.375	0.327	−0.044	−0.053
Average number of wives and concubines	(1.769)	(1.796)	(1.814)	(0.071)	(0.153)
Pre-juren qualification (State students = 1)	1.479	1.056	0.854	−0.135	0.292
Pre-juren qualification (State students = 1)	(3.487)	(3.499)	(3.517)	(0.149)	(0.317)
Age upon obtaining juren degree	−1.241***	−1.284***	−1.277***	−0.045***	−0.093***
Age upon obtaining juren degree	(0.140)	(0.144)	(0.144)	(0.006)	(0.013)
Ranking in provincial exam	0.090**	0.077**	0.074**	0.008***	0.007**
Ranking in provincial exam	(0.037)	(0.037)	(0.037)	(0.002)	(0.003)
Controls	No	Yes	Yes	Yes
Constant	85.545***	75.770***	75.692***	−0.577	1.304
Constant	(5.542)	(9.916)	(9.938)	(0.419)	(0.852)
Observations	728	728	728	4,035
R²	0.163	0.203	0.204	–2,270.42

Notes. Standard errors are in parentheses. Control variables include number of brothers, birth order, residence type (urban vs. rural), period, and region dummies. OLS = ordinary least squares; GF = grandfather; GGF = great-grandfather.

***

p < 0.01, **p < 0.05, *p < 0.1 (two-tailed tests).

What then is driving these results? Does the answer turn on whether a jinshi had a father who held an official appointment, or on differences in the level of official appointment? To find out, we divide the nine official ranks into three categories: low (i.e., lower seventh rank and below [7b], equivalent to county school instructors at the most), middle (upper seventh to upper fourth rank [7a–4a], equivalent to county- and prefectural-level officials), and high (lower third rank [3b] and above, equivalent to officials at the provincial level and above). As reported in column 3 of Table 5, the results show that the differences in the palace exam ranking were driven not only by whether a jinshi’s father was an official (> 0), but also whether his father was at least a mid-level official (county and above, ≥7a). In particular, the coefficients in column 5 suggest that this advantage was more than doubled when a jinshi’s father was a high-level official (15.634 vs. 7.301 and 6.752). In any case, these results suggest that a family’s resources attained and accumulated beyond human capital (exam achievements) are a better predictor of how well a jinshi candidate performed in the palace exam than they are of whether or not he passed the metropolitan exam.

We now turn to the sequential logit model to check the robustness of our results (columns 4 and 5 of Table 5). We confirm that the father’s education fails to significantly predict a candidate’s odds of obtaining first- or second-class honors in the palace exam (columns 4 and 5); only official rank determined the class of honor (albeit only at the 10 percent level of significance). Also consistent with the earlier results, the grandfather’s and great-grandfather’s education has no effect beyond that of the father in predicting jinshi exam success but official rank significantly (1 percent) predicts class of honor (column 5). Together, these results hint at the importance of cultural capital in transmitting the uncodified knowledge crucial for one’s success in the palace exam.

Conclusion

We have set out in this article to shed light on an old, intriguing question regarding whether the civil exam system actually facilitated social mobility in late imperial times. By drawing upon various valuable historical sources and constructing a unique data set, we have confirmed that there was downward as well as upward mobility in late imperial China, testifying to the fluidity of social mobility in this period. Furthermore, we have found that a candidate’s measurable ability is always a significant predictor of whether he passed the jinshi exam and of how well he performed in the exam, supporting Ho’s (1967) thesis that imperial Chinese society was highly mobile by virtue of the civil exam. Moreover, his argument receives further support from our finding that a few crude proxies of wealth had no significant effect on exam success.

Ho’s assessment needs qualification, however, in light of the finding that family background also significantly influenced exam success. In particular, “upper gentry” families enjoyed a distinct advantage over not just their “commoner” counterparts but also families that belonged to the other two higher social classes. By the same token, we arrive at the same conclusion when we break down family type into the more refined aspects of human and cultural capital using, respectively, education and official career as proxies. While the finding concerning the significance of human capital may not be all that surprising, we would not have easily predicted the role played by the transmission of tacit knowledge within the family if not for the radical difference in the content of the palace exam exclusively emphasizing the importance of statecraft. This unexpected finding also sheds light on studies of social mobility from a multigenerational perspective, as it directly suggests that cultural capital has a far stronger, enduring effect than human capital.

We hope that our study has illuminated a major historical puzzle of Chinese society in late imperial times and will inspire similar studies in the context of Chinese history.

Footnotes

Appendix

Acknowledgements

We thank the two referees of the journal, Ying Bai, Cameron Campbell, Ting Chen, Ho-fung Hung, James Z. Lee, Yi-min Lin, Guanglin Liu, Chicheng Ma, Debin Ma, Yusheng Peng, Xi Song, Donald Treiman, Xiwei Wu, Xuande Wu, and participants at various seminars and conferences for helpful comments and suggestions on earlier drafts and presentations. We are responsible for any remaining errors.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by Sein and Isaac Souede and the Fundamental Research Funds for the Central Universities (16JXYB02).

Notes

Author Biographies

Qin Jiang is an associate professor in the Department of History at the Shanghai Jiao Tong University. He holds a PhD in social science from the Hong Kong University of Science and Technology. His research focuses on the social and economic history of Qing China.

James Kai-sing Kung is Sein and Isaac Souede Professor in Economic History at the University of Hong Kong. His research interests span the economic history and political economy of China. He serves on the editorial board of the Journal of Comparative Economics and Explorations in Economic History.

References

Becker

Gary S.

(1962) “Investment in human capital: a theoretical analysis.” J. of Political Economy 70, 5: 9–49.

Bourdieu

Pierre

Passeron

Jean Claude

(1990) Reproduction in Education, Society and Culture. London: Sage Publications.

Buis

Maarten

(2007) “SEQLOGIT: stata module to fit a sequential logit model.” http://ideas.repec.org/c/boc/bocode/s456843.html.

Campbell

Cameron

Lee

James

(2003) “Social mobility from a kinship perspective: rural Liaoning, 1789–1909.” International Rev. of Social History 47, 1: 1–26.

Campbell

Cameron

Lee

James

(2008) “Kin networks, marriage, and social mobility in late imperial China.” Social Science History 32, 2: 175–214.

Chang

Chung-Li

(1955) The Chinese Gentry: Studies on Their Role in Nineteenth-Century Chinese Society. Introduction by Michael

Franz

. Seattle: Univ. of Washington Press.

Clark

Gregory

(2014) The Son Also Rises: Surnames and the History of Social Mobility. Princeton, NJ: Princeton Univ. Press.

Clark

Gregory

Cummins

Neil

(2015) “Intergenerational wealth mobility in England, 1858–2012: surnames and social mobility.” Economic J. 125, 582: 61–85.

Jiang

Chu

楚江 (2012) “清代舉人額數的統計” (Statistics on the number of juren in the Qing dynasty). MPhil thesis, Hunan Univ.

10.

Deng

Hongbo

鄧洪波 and Gong

Kangyun

龔抗雲 (2006) 中國狀元殿試卷大全 (A complete compilation of optimus essays in the palace examination in China). Shanghai: 上海教育出版社.

11.

Elman

Benjamin

(2000) A Cultural History of Civil Examinations in Late Imperial China. Berkeley: Univ. of California Press.

12.

Elman

Benjamin

(2013) Civil Examinations and Meritocracy in Late Imperial China. Cambridge, MA: Harvard Univ. Press.

13.

Fairbank

John K.

(1983) The United States and China. 4th ed. Cambridge, MA: Harvard Univ. Press.

14.

Tinglong

顧廷龍 [ed.] (1992) 清代硃卷集成 (Compilation of civil examination vermilion-ink essays in the Qing era). Taipei: 成文出版社.

15.

Guo

Peigui

郭培貴 (2006) “明代科舉各級考試的規模及其錄取率” (English title: “The Scale of Different Level Imperial Examinations in the Ming Dynasty and Its Enrollment Rate”). 史學月刊 9: 24–31.

16.

Ping-Ti

(1967) The Ladder of Success in Imperial China. New York: Columbia Univ. Press.

17.

Holmlund

Helena

Lindahl

Mikael

Plug

Erik

(2011) “The causal effect of parents’ schooling on children’s schooling: a comparison of estimation methods.” J. of Economic Literature 49, 3: 615–51.

18.

Hsu

Ta-Ling

許大齡 (1950) 清代捐納制度 (The system of purchasing offices via contributions during the Qing period). Beijing: Harvard Yenching Institute.

19.

Huang

Yifei

(2016) “Essays in economic history and applied microeconomics.” PhD diss., California Institute of Technology.

20.

Hymes

Robert

(1986) Statesmen and Gentlemen: The Elite of Fu-Chou, Chiang-Hsi, in Northern and Southern Sung. Cambridge: Cambridge Univ. Press.

21.

KRACKE

EDWARD

Jr. (1947) “Family vs. merit in Chinese civil service examinations under the empire.” Harvard J. of Asiatic Studies 10, 2: 103–23.

22.

Lai

Xinxia

來新夏 (1997) 林則徐年譜 (A chronicle of Lin Zexu). Tianjin: 南開大學出版社.

23.

Lee

Kuo-Chi

李國祁, Chou

Tien-Sheng

周天生, and Hsu

Hung-Yee

許弘義 (1975) 清代基層地方官人事嬗遞現象之量化分析 (A quantitative analysis of the careers of prefects and magistrates in the Qing dynasty). Taipei: 行政院國家科學委員會.

24.

Lefgren

Lars

Lindquist

Matthew J.

Sims

David

(2012) “Rich dad, smart dad: decomposing the intergenerational transmission of income.” J. of Political Economy 120, 2: 268–303.

25.

Hongzhang

李鴻章 and Huang

Pengnian

黃彭年 (2002 [1885]) 畿輔通志 (Provincial gazetteer of Zhili). Shanghai: 上海古籍出版社.

26.

Liu

Ts’Ui-Jung

劉翠溶 (1992) 明清時期家族人口與社會經濟變遷 (Lineage population and socioeconomic change in the Ming-Qing period). Taipei: 中央硏究院經濟硏究所.

27.

Long

Qirui

龍啟瑞 (2002 [Qing]) 經德堂文集 (Collected works of Jingdetang). In 續修四庫全書 1542. Shanghai: 上海古籍出版社.

28.

Lui

Adam Y.

(1981) The Hanlin Academy: Training Ground for the Ambitious (1644–1850). Hamden, CT: Shoe String Press.

29.

Mao

Xiaoyang

毛曉陽 (2007) “清代賓興禮考述” (On the binxing ceremony in the Qing dynasty). 清史研究 3: 45–54.

30.

Mare

Robert

(1980) “Social background and school continuation decisions.” J. of the American Statistical Association 75, 370: 295–305.

31.

Mare

Robert

(2011) “A multigenerational view of inequality.” Demography 48, 1: 1–23.

32.

Marsh

Robert

(1961) “Formal organization and promotion in a pre-industrial society.” American Sociological Rev. 26, 4: 547–56.

33.

Miao

Quansun

繆荃孫 (1991) 江蘇省通志稿 (Provincial gazetteer of Jiangsu). Nanjing: 江蘇古籍出版社.

34.

Miyazaki

Ichisada

(1976) China’s Examination Hell: The Civil Service Examinations of Imperial China. New Haven, CT: Yale Univ. Press.

35.

Mokyr

Joel

(2002) The Gifts of Athena: Historical Origins of the Knowledge Economy. Princeton, NJ: Princeton Univ. Press.

36.

Myers

Ramon

(1970) The Chinese Peasant Economy: Agricultural Development in Hopei and Shantung, 1890–1949. Cambridge, MA: Harvard Univ. Press.

37.

Polanyi

Michael

(1966) The Tacit Dimension. Chicago: Univ. of Chicago Press.

38.

Shang

Xiaoming

尚小明 (1999) 學人遊幕與清代學術 (Positions as aides and staff in Qing dynasty academia). Shanghai: 東方出版社.

39.

Shang

Yanliu

商衍鎏 (2004) 清代科舉考試述錄及有關著作 (Review of the imperial examination in the Qing dynasty). Tianjin: 白話文藝出版社.

40.

Shiue

Carol

(2017) “Social mobility in the long run: an analysis with five linked generations in China, 1300–1900.” Mimeo. https://spot.colorado.edu/~shiue/Social%20Mobility.pdf.

41.

Song

Campbell

Cameron

(2017) “Genealogical microdata and their significance for social science.” Annual Rev. of Sociology 43, 1: 75–99.

42.

WAKEMAN

FREDERIC

Jr. (1975) The Fall of Imperial China. New York: Free Press.

43.

Wang

Xiaoyong

王曉勇 (2013) “清代殿試軍事類策問研究” (A study of the cewen in the Qing dynasty final examination). 福建師範大學學報 (哲學社會科學版) 182, 5: 126–30.

44.

Weng

Tonghe

翁同龢 (1989 [Qing]) 翁同龢日記 (Diary of Weng Tonghe). Beijing: 中華書局.

45.

Jianjie

吳劍傑 (2009) 張之洞年譜長編 (Chronicle of Zhang Zhidong). Shanghai: 上海交通大學出版社.

46.

Shaosong

余紹宋 (2010 [1942]) 重修浙江通志稿 (Draft Zhejiang general gazetteer, revised edition). Beijing: 方志出版社.

47.

Zhang

Jian

張謇 (1994 [Qing]) 張謇全集 (Complete works of Zhang Jian). Nanjing: 江蘇古籍出版社.

48.

Zhang

Jie

張傑 (2003) 清代科舉家族 (The imperial-exam families in the Qing era). Beijing: 社會科學文獻出版社.

49.

Zhang

Lawrence

(2010) “Power for a price: office purchase, elite families, and status maintenance in Qing China.” PhD diss., Harvard Univ.

50.

Zhu

Baojiong

朱保炯 and Xie

Peilin

謝沛霖 (1980) 明清進士題名碑錄索引 (An index to the Ming-Qing list of jinshi degree holders). Shanghai: 上海古籍出版社.