A Quantitative History of Regicide in China

Abstract

Regicide plagued imperial China throughout its history from 1046 BCE to 1911 CE. Of the 1,948 rulers of 448 states or polities that covered the area of what we now know as China, 35.7% died unnaturally, implying a regicide rate of 2,899.8 deaths per 100,000 ruler-years. With so many rulers either killed or forced to commit suicide, only 1.4% of the regicides were the result of peasant uprisings, with the rest committed by the ruler’s relatives and trusted ministers or invading armies. Aristotle once hypothesized that in an autocracy, both the inner court and the masses are motivated to kill its ruler, but the literature has not shown which of the two poses a bigger threat to the ruler’s life. We demonstrate that, at least in Chinese history, members of the court were far more likely to be perpetrators of regicide. Still, not all dynasties suffered regicide to the same extent because institutional reforms in the Tang and later dynasties altered the political calculus. The probability of regicide was lower for rulers who appointed a successor but higher for those who relied on a regency of ministers at the time of ascension or whose predecessor died unnaturally. Succession uncertainty was a clear threat to the ruler’s life. War, drought, flood, and other natural disasters—as well as the number of co-existing competing states—also increased the probability of a ruler’s unnatural death. The regicide rate exhibits significant variations over time and between different subsamples. Chinese historical regicide rate is generally higher and more volatile than those of European kings but lower than those of Islamic sultans.

4.1 Introduction

Imperial China is known for its autocratic governments. But was it truly static and stagnant over several millennia, with no noticeable change until modern times, as claimed by Hegel (1956) and Elvin (1972), among others? If there were changes, how effective were they? For at least the past two centuries, these questions have captivated scholars studying Chinese history, especially its institutional development history. In imperial China, the ruler monopolized almost all the power and played a dominating role in its governance (Wei1989; Xuand Zhu1996; Zhou 1999), although governance reforms were introduced from dynasty to dynasty. To study the effectiveness of China’s institutional development and address the above questions, this paper focuses on the history of regicide, as any positive progress in governance innovations should result in a sustained decline in the regicide rate. In the existing literature, studies of historical rulers are often based on the experience of a single emperor or emperors of a single dynasty, with only a few exceptions that cover the entire imperial period (Gong1992; Chen2007). In contrast, our analysis includes all the 1,948 rulers, both kings and emperors, from the entire period of 1046 BCE − 1911 CE.

In the existing Chinese historiography, especially economic-history work on the effect of state capacity, the quality of a dynasty’s institutions and the value of institutional reforms are typically evaluated using narrow but quantifiable outcomes such as economic growth, population increase, and productivity gains (e.g., Broadberry et al. 2018; Qianand Sng2022). Thus, since growth in economy and population was higher during the eighteenth century than in the nineteenth century, institutional quality and governance during the reign of Emperor Kangxi (1661–1722 CE) and Qianlong (1735–1796 CE) are considered far better, creating the era of Kangxi-Qianlong Prosperity, whereas nineteenth-century China was an empire in decline. Nevertheless, according to Chen et al. (2017), from 1660 to 1820 CE, China’s homicide rate for ordinary people steadily increased and personal security deteriorated. Based on this and other similar findings (Chen 2022), economic growth and productivity cannot be used alone to conclude the quality of governance institutions and the effectiveness of a dynasty’s system. We must also consider the level of personal security and other adverse outcomes caused by a dynasty’s institutions (see Chen 2022 for related discussions).

The homicide data used in Chen et al. (2017) are, however, limited to 1660 to 1885 CE, based on the Xingke Tiben (Criminal Case Memorials). Earlier historical records on violent crimes are rare and unsystematic, if available at all. According to Eisner (2011) and Baten (2021), in the absence of reliable violence data for long historical periods, one can instead focus on regicide data to investigate the history of human violence as the full sample of past rulers is relatively easier to obtain (as opposed to that of commoners or elites) and regicide represents an extreme form of violence that is related to the level of general violence in a given society. Therefore, constructing China’s quantitative history of regicide is not only of interest for its own sake but also for studying its history of institutional development.

A ruler may die naturally (due to illness or aging) or unnaturally (i.e., regicide). The ruler’s life risk is a function of personal, institutional, social, and environmental factors. A regicide can bring dramatic economic, social, and political turmoil with far-reaching consequences for the country, upending existing institutional arrangements and perhaps even leading to dynastic collapse (which can be good or bad). Still, well-functioning governance institutions should yield minimal regicide risk as such unexpected shocks can upset the larger society through interruptions to the public goods provided by the government and threaten the population’s livelihood. For example, using a sample of post-Second World War leaders, Jonesand Olken (2005) find that a sudden leader’s death affected economic growth negatively in autocratic countries but not in democracies. Given China’s long tradition of autocratic rule, a reduction in regicide through institutional reforms should mean higher welfare for the larger society, as that would bring more stability and security. Thus, the regicide rate serves as a key indicator of institutional quality.

A few studies have investigated Chinese regicides in history.¹For example, Zhu (2015) examines 261 usurpation and regicide cases from the Qin (221–206 BCE) to the Qing dynasty (1644–1911 CE), with detailed statistical analysis. His study is, however, limited to emperors who were usurped or murdered, excluding both those who were not and the kings who ruled peripheral states outside the mainstay state or empire.² Thus, his sample is relatively small and does not present the whole picture of Chinese regicide history and institutional development. Wang (2022) also conducts a quantitative analysis of Chinese regicide cases among 282 mainstay emperors but for the historical period beginning in the Han dynasty (206 BCE-220 CE).

Our ruler sample is much broader in scope and longer in the historical period covered, encompassing not only the emperors of the mainstream dynasties featuredin the Ershiwu Shi (Twenty-Five Histories), but also the kings of peripheral states that were folded into a Chinese empire at some point from 1046 BCE to 1911 CE, including nomadic polities (e.g., Xiongnu, ~ 215 BCE- ~ 220 CE), rebel regimes (e.g., Huangjing, 184–192 CE), exile regimes (e.g., Beiyuan, 1368–1389 CE), and polities established by peasant uprisings (e.g., Taiping Tianguo, 1851–1864 CE). Specifically, our dataset covers 1,948 rulers in 448 states or polities that existed on today’s territory of China, from King Wu of Zhou (the Western Zhou, 1046–771 BCE) to the last emperor of the Qing dynasty (1644–1911 CE), Xuantong.

Following Eisner (2011), we employ two indicators to measure the rulers’ violent death rate: regicideratio (i.e., percentage of rulers who died unnaturally) and regiciderate (probability of a ruler being killed per year, expressed as the number of unnatural deaths per 100,000 ruler-years). The latter is commonly used in criminology and allows us to establish a comparable index across time and space.

Based on the full sample, we find the regicide ratio to be 35.7%, or 695 rulers to have died unnaturally, which translates into a regicide rate of 2,899.8 deaths per 100,000 ruler-years, placing Chinese rulers as one of the most dangerous occupations in human history³ and in sharp contrast with the commoners’ homicide rate in eighteenth- and nineteenth-century China of slightly less than 3 per 100,000 person-years (Chen et al. 2017). This may explain why Chinese rulers had to be much more ruthless than ordinary people, or else they would be killed first. It is also why the reduction in the ruler’s life risk should be a major yardstick for evaluating the effectiveness of government institutional reforms (Chen 2022). In a society where even rulers are concerned about their safety, it is unlikely that ordinary people will be safe from violence.

In Politics, Aristotle remarked:

“[D]emocracy appears to be safer and less liable to revolution than oligarchy. For in oligarchies, there is the double danger of the oligarchs falling out among themselves and also with the people; but in democracies, there is only the danger of a quarrel with the oligarchs. … A government that is composed of the middle class more nearly approximates democracy than oligarchy and is the safest of the imperfect forms of government.”

In a democracy, it is the electorate that decides who becomes the ruler, so that even if someone were to assassinate the sitting ruler, he or she would not gain legitimate power to rule automatically. Thus, regicide is of little value, sharply reducing the incentives for such action. However, in a centralized regime, such incentives are high as the return from being the emperor is perceived high: not only are aristocrats motivated to eliminate and replace the emperor, but also the public can be enticed to start a revolution to remove him or her. In other words, both the inner court and the broader society pose threats to an autocratic ruler.

Given Aristotle’s hypothesis, the existing literature has not yet established which of the two, the inner court or the masses, poses a bigger threat to an autocratic ruler’s life. Historians have predominantly focused on the role of peasant uprisings and social turmoil in ruler assassination and dynastic collapse, rather than quantitatively assessing the extent of court members killing the ruler or forcing him to commit suicide. This emphasis may give contemporary readers the impression that it was the masses who posed the main threat to the ruler’s and his regime’s life. But our analysis reveals that of the regicides in Chinese history, 28.4% were ministers, 20% were the ruler’s relatives, 48.7% were the rulers or generals of invading peripheral polities, and only 1.4% (nine individuals) were the leaders of civilian or peasant uprisings. That is, other than external threats, almost all the internal threats to the ruler’s life came from the inner court (the ruler’s ministers and relatives, contributing 48.4%), with the mass uprisings and rebellions playing a negligible role. This indicates that Chinese rulers were largely successful in mitigating life risks from mass revolts, but not in coping with threats from their inner circles and foreign enemies. When restricting our sample to only mainstream emperors coveredin the Ershiwu Shi (Twenty-Five Histories) for 221 BCE-1911 CE, we find that of all the emperor killers, 73.9% were their ministers and relatives, 24.2% the rulers or generals of invading states, and only 1.3% were mass revolt organizers. Thus, court members were even a bigger threat in the mainstream empires than the peripheral states, perhaps due to most of the offshoot polities not lasting long enough for the rulers’ inner circle members to develop successful plots against the rulers.

Our findings of historical regicide patterns in China generally corroborate the experience in historical Europe and modern totalitarian nations. Among medieval European regicides, 16.1% were committed by their ministers, 17.4% by relatives, and 18.8% by external forces. The proportion of cases perpetrated by organizers of civilian uprisings was 4.9%. Similarly, Svolik (2012) analyzed a sample of 316 totalitarian leaders from 1946 to 2008 and found that 205 of them, or 64.9%, were overthrown or executed by insiders within their ruling circles.

Our regression exercise indicates that appointing a successor in advance reduced succession uncertainty, which lowered the likelihood of regicide, a finding that aligns with contemporary practice in corporate succession management (Zhangand Rajagopalan2010). Throughout Chinese history, rulers have debated the merits of appointing a successor. It remains unclear whether Emperor Taizu of the Song dynasty (960–1279 CE), who died unexpectedly, designated his younger brother as heir or if his brother ascended to the throne through regicide. But, upon assuming power, his brother, the future Emperor Taizong, posited that the disadvantage of appointing a successor in advance lies in the disruption of power dynamics within the court, as the successor might be motivated to assassinate the emperor to seize immediate control, forcing the emperor to take precautions against such a threat.⁴Later, Emperor Zhenzong of the Song expressed similar concerns. However, their doubts and rationalization do not align with our results, because succession planning not only mitigates uncertainty but also deters potential rivals from harboring ambitions for the throne, thereby preventing violent conflicts.

We also find that the probability of regicide would be higher if the ruler had a ministerial regency upon ascending the throne at a young age. Regent ministers, typically held administrative and military power, may harbor motives to undermine and eventually kill the rulers. It is also true that shorter-duration dynasties corresponded with higher rates of regicide. With 48.4% of ruler deaths attributed to court insiders, a higher rate of regicide within a dynasty signifies lower internal governance quality, culminating in shorter dynastic lifespans. Additionally, there is a significantly positive autocorrelation in the mode of death between consecutive rulers; that is, when one ruler died unnaturally, the likelihood that the next ruler would have the same fate would be significantly higher. Our data indicate that 41.6% of the regicide cases involved the killings of two or more consecutive rulers, consistent with Eisner’s (2011) observations on European kings. Moreover, the more competing polities co-existing, and the more wars, droughts, and floods, within each 50 years, the higher the regicide rate during the same period.

Using data on the deaths of 1,513 kings from 45 European kingdoms between 600 and 1800 CE, Eisner (2011) found a regicide ratio of 22%, corresponding with an average annual violent death rate of 1,547 per 100,000, both figures much lower than in historical China. This finding offers another perspective on the origins of the economic Great Divergence between the East and the West since the late eighteenth century. A closer examination illustrates that both the Chinese and European regicide rates declined from the seventh century onward, attributed to institutional innovations in both regions. However, the decline of Chinese emperors’ regicide rate was more relatively gradual but marked by greater fluctuations. Globally speaking, from the eleventh century to the end of the eighteenth century, the average duration of European rulers extended due to improved governance reforms, juxtaposed with a decline in the average tenure of Islamic rulers (Blaydesand Chaney2013). But China followed a different pattern: its ruler duration swung up and down wildly, implying that it never figured out how to institutionally apply checks and balances to ensure the peaceful transfer of power at the top.

Our study makes four contributions to the existing literature. First, we highlight the extremely high life risk faced by autocratic rulers, which worked to shape ruler behavior. This is in contrast with ruler behavior models in which the ruler’s objective is to maximize his long-term fiscal income, with rebellion from the commoners acting as a constraint on his actions, an assumption that is only valid when the ruler has a relatively stable reign and encounters a minimal risk of regicide (Sng2014; Maand Rubin2019). However, our finding indicates that autocratic rulers confronted a significantly higher life risk than assumed in existing studies. The primary life threat for such a ruler originated from members in his inner circles, rather than from the commoners.

Second, our research contributes to the state capacity literature by emphasizing the importance of ruler security and peaceful transfer of power. As defined by Bates (2020), a state comprises “a political hierarchy consisting of tax collectors to collect income, a police force, and a military to assist with security, and a judiciary to administer legal justice. This political body is led by a chief executive (sometimes referred to as “the monarch”) in conjunction with a noble or parliamentary council.” Existing work on state capacity largely focuses on fiscal extraction ability and the rule of law, while overlooking governance institutions at the top (Ma2013). It can be argued that a regime with high risk to the ruler’s life could hardly possess high state capacity (Wang 2022), a point especially important for studying imperial China.

Third, we offer another perspective on the Great Divergence in terms of institutions ensuring political stability or the lack thereof (Blaydesand Chaney2013). Our analysis introduces quantitative evidence highlighting the institutional differences among Eastern, Western, and Islamic polities and their institutional performance, as viewed through the lens of regicide rate and rule duration. Lastly, we construct a quantitative history of regicide for historical China since 1046 BCE, with the most comprehensive coverage of all polities that existed in today’s Chinese territory, adding to the Chinese history literature.

The remainder of the chapter is organized as follows: Sect. 4.2 offers the historical background and regicide examples. Sect. 4.3 presents descriptive statistics for data on the dynasties and rulers. Section 4.4 examines factors affecting rulers’ life risk, while Sect. 4.5 investigates the temporal variations in regicide, ruler duration and perpetrator type. Section 4.6 concludes the paper.

4.2 Historical Background and Regicide Examples

Constructing our ruler sample for 1046 BCE-1911 CE presents us with two immediate challenges. First, what states or polities should be included in what is called “China”? For the long period under study, China was always a “work in progress,” with the borders of the mainstream state or empire frequently redrawn, which means there were always competing polities that coexisted outside those territories. To address this, we use China’s map of 1820 CE as a reference point and include in our sample any state that existed within that 1820 territory over any subperiod of 1046 BCE-1911 CE, which produces 448 states or polities (including the empires),⁵ a significant increase compared to the 25 mainstream states documented in the official historical records, Ershiwu Shi(Twenty-Five Histories). For this reason, we will, for robustness checks, consider both samples: the full sample and the 25 mainstream dynasty sample.

Next, who can be treated as a “ruler”? During the formative centuries of historical China, or the Zhou dynasty (1046–221 BCE), there were no emperors but lots of kings, as there were many separate polities, each with a king, especially during the Spring and Autumn (770–476 BCE) and the Warring States period (476–221 BCE). While later eras witnessed the emergence of supposedly unified empires such as the Han (206 BCE-220 CE), the Tang (618–906 CE), the Song (960–1279 CE), and the Ming dynasty (1368–1644), all ruled by emperors, their territories were mostly smaller than that of the Qing dynasty in 1820, so they each had lingering peripheral states constantly knocking at their borders and threatening the mainstream emperor (Fairbankand Reischauer1989). Our full sample includes all these kings and emperors of the 448 states or polities based on Baten (2021), wherein a ruler is defined as an individual who “possessed a certain degree of autonomy in high-level decision-making”⁶ which gives us 1,948 rulers in total.⁷

Note that our primary source of information on the dynasties, polities, and rulers (including how they died) is the Zhongguo Lishi Jinian Zongbiao (Comprehensive Chronological Tables of Chinese History), on which Wilkinson (2013) commented that it is “on a huge scale and indispensable” and “puts all previous chronological tables out of business” (p. 277, p. 521). Additionally, we incorporate information from the Ershiwu Shi(Twenty-Five Histories) and the Zhongguo Lishi Dituji (Atlas of Chinese History). There are five rulers in our collection who ascended to the throne on two distinct occasions, in which case each ruler is treated as two separate occurrences in our sample. As in Eisner (2011), where possible, we cross-check our ruler data with other scholarly publications and in cases of discrepancies, we defer to what is in published scholarly work.

To facilitate comparison with existing work and for robustness checks, we divide the full sample into three subsamples. Subsample A has all the rulers (kings) from the pre-Qin period (1046–221 BCE), while subsample B comprises all the emperors of the mainstream empires and states from 221 BCE-1911 CE, as documented in Twenty-Five Histories, including the Qin, the Han, the Three Kingdoms (220–280 CE), the Jin (266–420 CE), the Sixteen Kingdoms (304–439 CE), the Northern and Southern Dynasties (420–589 CE), the Sui (589–608 CE), the Tang (618–906 CE), the Five Dynasties and Ten Kingdoms (907–960 CE), the Song (960–1279 CE), the Yuan (1279–1368 CE), the Liao (916–1125 CE), the Jin (1115–1234 CE), the Western Xia (1038–1227 CE), the Ming (1368–1644 CE), and the Qing (1644–1911 CE). This emperor sample has been widely examined in existing studies (e.g., Wang 2022). Subsample C collects all the rulers who were not emperors of any mainstream empire from 221 BCE to 1911 CE, comprising the rulers of peripheral states (e.g., Xiongnu and Dali), exile regimes (e.g., Beiyuan and Xiliao), rebel regimes (e.g., Huangjing), and polities established by civilian uprisings (e.g., the Da Shun regime established by Li Zicheng). These rulers are seldom included in existing regicide studies, but they coexisted with and affected those in subsample B.

For example, if we only consider the emperors, the Qing Dynasty has 12 of them in subsample B. However, for the same period, subsample C has 89 rulers from 62 other regimes that coexisted, including exile regimes (e.g., Southern Ming), frontier regimes (e.g., Yarkand Khanate), rebel regimes (e.g., Wu Zhou, founded by Wu Sangui), and regimes established by uprising organizers (e.g., the Taiping Tianguo).

Note that in Chinese intellectual discourse, “emperor,” or huangdi, means “the son of heaven”: “He is Heaven’s son and rules all under Heaven according to His will” (Zhou1999, 2). Historians generally agree that Qin Shi Huang, founder of the first Chinese empire—the Qin—was China’s first emperor. He was born in 259 BCE and died in 210 BCE. After unifying China at the age of 39 (221 BCE), he gave himself the title of emperor to convey his status as the empire’s sole monarch and to distinguish himself from former monarchs titled “king” (wang) (Loeweand Twitchett1986). That’s why 221 BC is of special importance in Chinese history and the dividing line between our subsample A and subsamples B and C.

The first emperor killed was also from the Qin dynasty, Qin Er Shi, whose name was Huhai. His successor was killed as well.⁸ After Qin Shi Huang, historical scholarship lacks consensus on which rulers can be classified as emperors. We follow Ershiwu Shi(Twenty-Five Histories) for our purpose. To provide more background on our regicide data, we present four cases, each representing a different type.⁹

Case1: Huhai of the Qin Dynasty

Qin Shi Huang, founding emperor of the Qin dynasty (221–206 BCE), declared that he would hand over the throne to his eldest son, Fusu, before his death. However, in collaboration with Chancellor Li Si, eunuch Zhao Gao falsified the imperial edict, forced Fusu to commit suicide and crowned Huhai, Qin Shi Huang’s younger son.

To consolidate the usurped throne, Huhai assassinated his brothers, Meng Tian and Meng Yi (who commanded the army), beheaded 12 of his siblings, and dismembered six brothers and 10 sisters.

Since Huhai blamed Zhao Gao for the fact that the uprising army led by Chen Sheng and Wu Guang was approaching the capital, Zhao Gao began plotting a coup. Yan Le, Zhao Gao’s son-in-law, led an army into the palace and forced Huhai to commit suicide in 207 BCE; Huhai died at the age of 24 and reigned for only three years.

Case 2: Emperor Ping of theWestern Han(206 BCE-9 CE)

Having learned the lessons about the consequences of power monopolization during the reigns of preceding emperors, Liu Ao (51–7 BCE), known as the Emperor Cheng of the Han dynasty, made his mother Wang Zhengjun Empress Dowager. He also appointed his uncle Wang Feng as marshal of the state and general in chief and gave him authority over the secretariat as well as overall responsibility for state affairs. And Empress Dowager Wang’s five younger half-brothers were all given the title of marquis. Both Emperor Cheng and his successor Emperor Ai (27–1 BCE) had no sons, so their nephews were made heirs. When Emperor Ai’s nephew, Liu Kan, later Emperor Ping (9 BCE-6 CE), took the throne at the age of eight, Empress Dowager Wang was still alive and appointed her nephew, Wang Mang (45 BCE-23 CE), marshal of the state. With Emperor Ping as his puppet, Wang Mang seized the highest power and forbade Emperor Ping’s consort kin from serving as officials in the court but kept them with some titles in the Zhongshan state, Emperor Ping’s former fiefdom when he was still a prince. As Emperor Ping grew older, he became increasingly dissatisfied with Wang Mang. So, feeling insecure, Wang Mang murdered Emperor Ping and later usurped the throne, establishing the Xin dynasty (9–25 CE).

Case 3: Emperor Taizu ofLater LiangDuring the Five Dynasties

Zhu Wen (852–912 CE) was a general in the Huang Chao Rebellion (874–884 CE), who later surrendered to the army of the Tang dynasty (618–906 CE), and was given the name Quanzhong, literally “full of loyalty,” by Tang Emperor Xizong (862–888 CE). Subsequently, he increased his forces and prepared to usurp the throne. In 904 CE, he assassinated Tang Emperor Zhaozong (867–904 CE) and declared the 13-year-old Li Zhu (892–908 CE) as crown prince and later Emperor Ai, who was only a puppet emperor under Zhu Wen’s control. Zhu Wen began to murder members of the Tang imperial clan, including Empress Dowager He, Emperor Ai’s mother. Zhu forced Emperor Ai to abdicate the throne in 907 CE and murdered him in 908 CE, putting the Tang dynasty to an end.

After the usurpation, Zhu founded the Later Liang state (907–923 CE), abolished the Bureau of Military Affairs, which had been controlled by eunuchs in the Tang, and established the Office for the Veneration of Governance, filled with literati rather than eunuchs. He then made his second son, Zhu Youwen, the crown prince. After hearing the news, his third son, Zhu Yougui (886–913 CE), led his army to kill Zhu Wen. In 912 CE, Zhu Yougui declared himself Emperor and named Zhu Wen’s fourth son, Zhu Youzhen (888–923 CE), governor of Kaifeng and regent of the Eastern Capital. In the following year, with the help of military commissioner Yang Shihou, Zhu Youzhen assassinated Zhu Yougui and ascended to the throne as Emperor Mo of the Later Liang.

Case 4: Emperors Huizong and Qinzong ofthe Song

Emperor Huizong (1081–1135 CE) of the Song dynasty was Emperor Shenzong’s eleventh son. Because his elder brother, Emperor Zhezong (1076–1100 CE), died at the age of 24 with no living male issue, Huizong became the eighth emperor of the Song dynasty. Despite his artistic achievements, he was deemed foolish by the ruling circles. The Song was invaded by the Jin State (or, the Jurchen) in 1125 CE. Out of fear, Emperor Huizong abdicated the throne in favor of his son, Zhao Heng (1100–1161 CE), who later became Emperor Qinzong. After the Jin army retreated in 1126 CE, Huizhong fled to Zhenjiang and was taken back to Kaifeng, the capital, by Qinzong. The Jin army staged a comeback and captured Kaifeng at the end of the year. Both Huizong and Qinzong were captured and escorted to the enemy Jin State in an episode known as the Jingkang Incident.

The Jin State’s ruler dubbed Emperor Huizong the “Duke of Faint Virtue” and Emperor Qinzong the “Marquis of Heavy Faint,” mocking both the father and the son for their repeated fainting. Tortured by the Jin, Huizong died at Wuguocheng in the Jin State (in today’s Heilongjiang) in 1135 CE, becoming the Song dynasty’s first emperor to die in a foreign country. Emperor Qinzong was tortured to death in 1156 CE.

4.3 Data Description

In addition to the ruler data above, we collect data on wars from Zhongguo Lidai Zhanzheng Nianbiao (A Chronology of Wars in Dynastic China) and on droughts and floods, for which data is available only for 115 BCE - 1911 CE from Zheng et al. (2006). More details will be provided when used in analysis.

Table 4.1 provides summary statistics of the dynasties and rulers for the full sample.The duration of rule for these dynasties ranges from 0.1 years to 957 years, with an average of 62.1 years. The longest-lasting regimes were mainly from the pre-Qin period and ethnic-minority regimes in the frontier areas. For example, the Eastern Zhoudynasty lasted for 515 years, and the kingdom of Khotan lasted for nearly 1000 years. The fact that these regimes are not among the much-celebrated unified empire polities is quite telling as it signals that the unified empires must have been much more challenging to rule peacefully (unless the rulers were able to adopt well-functioning governance systems). The number of rulers in each dynasty ranges from 1 to 40, with an average of 4.3. Each dynasty’s number of emperors killed ranges from 0 to 12, averaging about 1.6, implying an average regicide ratio of 56.8% by dynasty. This ratio is quite high because about 247 regimes have a single ruler, of which 70% are killed. In these regimes, the regicide ratio is 100%.

Table 4.1 Descriptive statistics by dynasty

Table 4.2 presents summary statistics on the 1,948 rulers. Ruler lifetime exhibits significant variation, with the youngest life ending at two years and the oldest at 103. On average, rulers lived for 42.2 years. The average age when assuming the throne was 27.7, with the youngest being one year old and the oldest 73 years. The average reign lasted for eight years, with the shortest at less than one month and the longest at 73 years. Among the 984 rulers with available data, the average number of sons was 3.4, with some having no sons and some over 40. Of the 1,243 rulers with clear records regarding whether he or she appointed a successor while alive, 50% did so. Detailed information on the initial enthronement process is available for 1,425 rulers, of whom 19% needed a minister’s regency (Guming Dachen). Out of the 1,948 rulers, 35.7% of them died unnaturally.

Table 4.2 Descriptive statistics of rulers

4.3.1 Regicide Ratio and Regicide Rate

In the literature, two indicators are used: regicide ratio (defined as the proportion of rulers who died unnaturally) and regicide rate (number of unnatural deaths per 100,000 ruler-years) (Eisner2011). Among the 1,948 rulers in our full sample, 695 were victims of regicide, accounting for 35.7%, which translates into a regicide rate of 2,899.8 per 100,000 ruler-years, calculated as the number of unnatural deaths in the sample divided by the sum of each ruler’s years of reign among all the rulers, multiplied by 100,000. Thus, the regicide rate so computed captures how many rulers would be killed for every 100,000 rulers during each reign year. For comparison, Chen et al. (2017) find that the homicide ratefor ordinary people was about 3 per 100,000 person-years between 1660 and 1885 in Qing China. Thus, the death risk for Chinese rulers is nearly 1,000 times as high as for ordinary people.

In Europe, homicide rates for the– common population were typically from 20 to 50 per 100,000 person-years during the High Middle Ages and declined to the 1–10 range in the eighteenth century (Eisner2011). Contemporary homicide rates are from 0.6 to 1.5 in Western Europe (Aebi et al. 2006), about 120 for young black men in the early 1990s United States(Blumstein 2000), and about 130 in Ciudad Juárez, Mexico, one of the most murderous cities in the world today.

4.3.2 Types of Murderers

Given the high regicide rate among rulers, it is of great interest to understand the sources of threat and identify perpetrators by type. In Table 4.3, we classify the perpetrator’s relationship with the ruler into two categories: internal ruling court members and external members. Internal members are further divided into relatives, ministers, and assistants (e.g., eunuchs and concubines), while external members consist of leaders of civilian uprisings and the rulers and generals of competing co-existing regimes.

Table 4.3 Relationship between murderers and rulers

For the full sample in Table 4.3, 48.7% of the regicides were from external co-existing regimes, with 49.8% of them coming from within the polity. Specifically, of the perpetrators, 28.4% were internal ministers, 20% relatives, and 1.4% leaders and organizers of civilian or peasant uprisings. Thus, internal court members (the ruler’s ministers and relatives) made up 48.4%,—a figure much higher than that from the civilian uprising organizers (1.4%)—and contrary to the impression generally conveyed to students by history books in Chinese classrooms that emphasize the role, and portray the importance, of the peasant class in toppling historical regimes. In total, there were only nine instances where civilian or peasant rebellions led to the ruler’s death, including the killing of King Ziying (?-206 BCE) of the Qin by Xiang Yu’s rebels during the Qin, of Wang Mang (Emperor of the Xin, 45 BCE-23 CE) by merchant Du Yu, and the suicide of Emperor Chongzhen (1611–1644 CE) of the Ming when Li Zicheng entered the capital city—Beijing—at the end of the Ming dynasty. This evidence also shows that between the two types of threat to an autocratic ruler’s life as postulated by Aristotle, members in his inner circle within the court were much more dangerous than the public, as organizers of commoner uprisings and rebellions hardly posed real threats to the ruler’s life.

The subsamples reveal variations in the patterns. For subsample A, which comprises the pre-Qin kings, 79.5% of perpetrators were from within the kingdom, with ministers accounting for 33.7%, relatives for 42.2%, the two groups which made up 75.9% of the total, and those out of civilian uprisings accounting for only 3.6%. Direct external threats to the ruler’s life were lower, contributing only 16.9% of the cases, than for the full sample. This is perhaps the case due to the smaller states in pre-Qin times having shorter borders to defend than the bigger unified empires starting from the Qin. It might also be caused by the underrepresentation of actual events in pre-Qin historical records.

For subsample B, including all the mainstream emperors from the Qin to the Qing, 75.2% of murderers were from within the empire, with ministers accounting for 40.1% of the total, relatives for 33.8%, and civilian uprising organizers for only 1.3%. The emperor’s inner circle was still the biggest threat, signaling the lack of proper power sharing, succession planning, and governance institutions, which resulted in the emperor having too much power that could not be peacefully transferred and was up for grabs upon, or even prior to, his death.

For subsample C, containing the non-emperor rulers for peripheral and rebel states from 221 BCE to 1911 CE, external threats from co-existing regimes (including the mainstream Chinese empire) were the key with them being responsible for perpetrating 65.6% of the regicides. Most of these ruler killings were committed by the troops and generals of the mainstream empires in subsample B. In these instances, the ruler’s ministers accounted for 22.4% of the regicides, relatives for 9.6%, and civilian uprising organizers for only 1%.

To contrast China’s historical experience with that of Europe, we use the European regicide data for 600–1800 CE from Eisner (2011) and decompose the perpetrators into similar categories. The results displayed in the last two columns of Table 4.3. demonstrate a similarity: the European ruler’s ministers, family members, and external enemies from other states constituted the three most significant risks to the ruler’s life. Of the 224 regicide cases, 18.8% were committed by external perpetrators, 20.1% by ministers, and 23.2% by relatives.

4.4 Factors Affecting Ruler Life Risk

Our analysis so far shows that the biggest group of regicide perpetrators consisted of people closest to the rulers, as well as foreign enemies. Next, we explore factors that influence the risk to a ruler’s life. We start with Table 4.4 where, of the full sample, the unnatural-death ruler group is compared with the natural-death ruler group. Note that rulers who died naturally lived longer with an average age of 46.3 years, as opposed to 34 years for those who were victims of regicide. The natural-death rulers had an average enthronement age of 28.3 years, versus 26.6 years for the unnatural-death rulers, ascended at an average age of 26.6 years; ruled on average for 16.5 years, versus 6.6 years for the other group; had an average of 3.8 sons, versus 2.4 sons for the other group; and 4.6 brothers, versus 4.9 brothers for the unnatural-death group, suggesting slightly more competition for the latter group as they had more brothers. The data also show that 62% of natural-death rulers appointed a successor before death, compared to only 24% of those who died violently, resulting in a 38-percentage point gap, and indicating a significant impact of advance succession planning. Moreover, 16% of rulers in the former group relied on a minister’s regency at the time of their ascension, whereas this figure was 25% for the latter group, suggesting that this may have been a determinant of the likelihood of regicide. Our T-test exercise confirms that except for the number of brothers, the mean difference between the two groups is statistically significant at the 1% level for all the other variables examined in Table 4.4.

Table 4.4 Comparison between natural- and unnatural-death rulers

War and natural disasters may have also played a role in regicide outcomes, either directly or indirectly, as such events may have generated enough distress in the larger society, providing an opportunity for a rival to act against the ruler. Due to the limitations of historical data, it is difficult for us to precisely match wars and disasters, especially those in early Chinese history, by year and with each regime. Therefore, we use 50-year intervals to examine the impact of wars and disasters within each interval on the regicide rate of that period.

Next, we examine the influence of the following factors: (1) whether the ruler appointed a successor in advance, (2) whether the ruler ascended to the throne with a minister’s regency, (3) whether his predecessor died from regicide, (4) the dynasty’s duration, (5) the number of external co-existing regimes, (6) the number of wars within the contemporary 50 years, and (7) the number of flood and drought events within the contemporary 50 years.

4.4.1 Effect of Successor Appointment

When a ruler appoints a successor, there are two effects: the chosen crown prince is eager to succeed, threatening the sitting ruler’s life, and there is also the barrier effect that the presence of a crown prince removes succession uncertainty, reducing the ruler’s life risk (Konradand Mui2017).

To illustrate the first effect, we can look at the example of Emperor Kangxi (1654–1722 CE) of the Qing, who appointed his one-year-old first son, Yunreng, as the crown prince when the emperor himself was only 23 years old. As Yunreng grew older, he and his father had a falling out, which resulted in the son being deposed twice and eventually imprisoned for life. That experience made Emperor Kangxi end the practice of appointing a successor for the rest of the Qing. Another example is Yang Guang, Emperor Yang of the Sui dynasty (581–618 CE), who murdered his father before being enthroned.

As for the barrier effect, when a successor is known, it shifts the life risk from the incumbent ruler to the crown prince as potential competitors will now focus on attacking the successor. For instance, Li Shimin (598–649 CE), who later became Emperor Taizong of the Tang, assassinated the crown prince and two brothers, Li Jiancheng and Li Yuanji, to gain control of the throne. In the example above, during the Qing, Kangxi’s other sons who aspired to assume the throne also attacked the crown prince.

To evaluate the net impact of these two effects, we conduct various analyses in Table 4.5. Of the 1,243 rulers with clear succession information, 619 appointed a successor, with 15.5% of the incumbent rulers suffering regicide. In contrast, among the 624 rulers who did not appoint a successor, 49.7% died unnaturally. The net impact of successor appointment is evident.

Table 4.5 Impact of successor appointment on regicide risk

Excluding rulers who died young and did not have the opportunity to have a son, we focus on the remaining sample of 794 rulers who had at least one son. In this sample, 592 rulers appointed a successor, and 202 did not. Among those who did appoint, 14.9% experienced regicide, while 35.1% of those who did not face the same fate.

Furthermore, we can also restrict our attention to regimes that had at least three separate rulers with at least one son. This led to a sample of 731 rulers, of which 558 rulers appointed a successor and 173 did not. The regicide ratio was 14.5% for those who appointed a successor and 33.5% for those who did not.

Finally, we remove the last ruler of each regime from the preceding sample as their reign was typically short, so they might not have had enough time to appoint a successor, which may have affected our analysis. After these removals, we are left with 666 rulers, of whom 517 appointed a successor (with a regicide ratio of 14.3%) and 149 did not (with a regicide ratio of 34.9%).

The above analysis thus confirms that the violent death risk for rulers who appointed a successor was less than half of that for those who did not. Succession planning would clearly improve the ruler’s safety.

4.4.2 Minister’s Regency

The presence of a minister’s regency was often associated with young rulers. These ministers frequently held military and political power, making it relatively easy for them to usurp the throne, as exemplified by Wang Mang (45 BCE-23 CE) in the late Western Han Dynasty. Table 4.6 shows that a ruler with a minister’s regency during their initial ascension faced a significantly higher probability of regicide. Of the 1,425 rulers with complete information on regency, 272 were assisted by ministers, and 44.9% of them died unnaturally. In contrast, 32.4% of the other 1,153 rulers suffered regicide.

Table 4.6 Impact of Minister’s regencyon regicide

The difference between the two groups becomes even more pronounced when only the emperors in subsample B are considered. Among the 341 emperors with regency information, 136 had a minister’s regency, of whom 52.9% died from regicide. In contrast, 29.8% of the other 205 emperors died unnaturally.

4.4.3 Regicide Continuity

We address the question of whether a ruler’s fate was affected by that of his predecessors by investigating the correlation in death mode between consecutive rulers in the same regime. Our analysis reveals a significant autocorrelation in death mode, that is, when a ruler dies from regicide, there is a significant chance that his successor would die in the same way, implying that the regicide rate reflects a regime’s quality of governance institutions that can be changed only through institutional reforms.

In our full sample, there are 136 instances (68 pairs) of two consecutive rulers being assassinated; for example, Qin Er Shi and King Ziying of the Qin dynasty (221–206 BCE) both died from regicide. We also find 69 instances (23 groups) of three consecutive rulers being murdered; for example, Emperor Wu (or Taizu), Prince Ying (Zhu Yougui), and Emperor Mo of the Later Liang state (907–923 CE). In total, 289 rulers were consecutively murdered, accounting for 41.6% of the 695 regicide cases.

These findings are consistent with Eisner’s (2011) based on European regicide history. Thus, within each dynasty, regicide occurrences tended to cluster, reflecting the wrongs of a regime’s underlying institutions.

4.4.4 Dynasty Duration and Regicide Ratio

The longevity of a regime often reflects its success in governance institutions. Consequently, a longer dynasty duration should be accompanied by a lower regicide rate within that dynasty. Figure 4.1 plots the relationship between the two, where the horizontal axis represents a regime’s duration and the vertical axis is its regicide ratio, showcasing a significant negative correlation, that is, longer-lasting regimes had significantly lower regicide rates.

Fig. 4.1

Correlation between dynasty duration and regicide ratio (full sample). Notes The regicide ratio is defined as the proportion of rulers who died unnaturally

4.4.5 Competing Regimes, War, and Regicide

Invading armies from competing polities accounted for a large portion of regicide cases. In Fig. 4.2, the horizontal axis stands for the number of competing regimes that existed in the 1820 CE land of Qing China and within each 50-year interval while the vertical axis is the regicide ratio in the given period. It is evident from the chart that more co-existing polities meant higher regicide rates.

Fig. 4.2

Relation between competing regimes and regicide ratio (full sample). Notes The regicide ratio for each 50-year interval is the percentage of the period’s rulers who died unnaturally

Similarly, Fig. 4.3 shows that more wars in a 50-year interval were associated with a higher regicide ratio.

Fig. 4.3

Relation between the number of wars and regicide ratio (full sample). Notes The regicide ratio for each 50-year interval is the percentage of the period’s rulers who died unnaturally

4.4.6 Natural Disasters and Regicide

Natural disasters play a significant role in triggering conflicts (Baiand Kung2011). It is expected that natural disasters can cause much distress, economic and otherwise, in the larger society, which can translate into severe political distress in the ruler’s court and lead to regicide. To examine the possible relationship between natural disasters and regicide, we select two types of disasters: drought and flood. Figures 4.4 and 4.5 respectively illustrate that over each given 50-year interval, higher numbers of drought and flood were associated with a higher regicide ratio. By combining both types of disaster into a summary indicator, Fig. 4.6 demonstrates a positive correlation between natural disasters and the regicide ratio in each period.

Fig. 4.4

Correlation between drought and regicide ratio (sample after 115 BCE). Notes The regicide ratio for each 50-year interval is the percentage of the period’s rulers who died unnaturally. Due to the availability of disaster data, the sample used here is from 115 BCE onwards

Fig. 4.5

Correlation between flood and regicide ratio (sample after 115 BCE). Notes The regicide ratio for each 50-year interval is the percentage of the period’s rulers who died unnaturally. Due to the availability of disaster data, the sample used here is from 115 BCE onwards

Fig. 4.6

Correlation between natural disasters and regicide ratio (sample after 115 BCE). Notes The regicide ratio for each 50-year interval is the percentage of the period’s rulers who died unnaturally. Natural disasters are drought and flood. Due to the availability of disaster data, the sample used here is from 115 BCE onwards

4.4.7 Regression Analysis

This subsection builds upon the univariate analysis of the preceding subsections by incorporating the different factors into a multivariate regression model. Table 4.7 presents the results of a probit model in which the dependent variable takes a value of 1 if a ruler died from regicide and 0 otherwise. The primary explanatory variables include: (1) whether the ruler ever appointed a successor, with a value of 1 indicating yes and 0 indicating no; (2) whether the ruler ascended to the throne with a minister’s regency, with a value of 1 representing yes and 0 otherwise; (3) the number of wars occurring in the year of the ruler’s death; (4) the number of co-existing competing regimes; (5) years of dynasty duration; (6) the number of floods and droughts in the ruler’ death year; (7) numbers of the ruler’s sons and brothers, respectively.

Table 4.7 Probit regression analysis of regicide

The main control variables are regional and century-fixed effects. We divide China into seven regions: Northeast, East, Central, North, South, Northwest, and Southwest. For a regime and for each of these regions, its regional dummy takes a value of 1 if its capital or main territory was inside the region, and 0 otherwise.

Due to varying degrees of missing data, we organize the regression analyses according to sample size, from the largest to the smallest. In Table 4.7, successor appointment in each regression has a statistically significant (at the 1% level) and negative effect on regicide, which means that ruler succession arrangement was almost a significant institution for ensuring smooth and peaceful power transition and that efforts to add power transfer uncertainty would increase the ruler’s life risk. While practical, regency arrangements for a young ruler would also put the ruler’s life in danger; so, would more wars. In addition, the greater the number of competing regimes, the higher the probability of regicide.¹⁰ The longer a regime lasted, the lower its regicide rate. Droughts and floods did not have a statistically significant impact on regicide, partly because these events were exogenous and did not reflect the quality of a regime’s institutions. The numbers of the ruler’s sons and brothers, respectively, did not have a statistically significant impact, as the two variables’ importance is probably captured by the ruler’s other variables.

4.5 Temporal Variation in Regicide and Ruler Duration

The analysis so far has treated all rulers as if they came from the same cross-section, regardless of era, which clearly does not take into account the effect of institutional changes during the 3,000-year sample period. Thus, this section investigates the evolution of both the regicide rate and ruler duration over time. In the presence of quality governance institutions, the regicide rate should be low and ruler duration should be long, at least prior to the advent of democracy.

4.5.1 Long-Term Trends in Regicide

Figures 4.7 and 4.8 present the temporal evolution of the regicide rate based on the full sample and subsamples A and B combined, respectively. To focus on the long-term trends, we use a 300-year moving average.

Fig. 4.7

Evolution of regicide rate (full sample). Notes The regicide rate line is based on a 300-year moving average. The two vertical black lines respectively stand for the establishment of the Qin dynasty in 221 BCE and of the Tang dynasty in 618 CE

Fig. 4.8

Evolution of regicide rate (subsamples A and B). Notes The regicide rate line is based on a 300-year moving average. The two vertical black lines respectively stand for the establishment of the Qin dynasty in 221 BCE and of the Tang dynasty in 618 CE. Subsample A includes kings from 1046 to 221 BCE, while subsample B all the emperors from 221 BCE to 1911 CE

Figure 4.7 shows that for the full sample, the regicide rate exhibits noticeable fluctuations: It was low during the pre-Qin period, gradually increasing following the Qin and Han dynasties (221 BCE-220 CE), peaked during the Three Kingdoms (220–280 CE), the Eastern and Western Jin (266–420 CE), and the Northern and Southern Dynasties (420–589 CE), then gradually declined from the late Tang onward, and rose again during the Ming (1368–1644 CE) and the Qing (1644–1911 CE). Therefore, with all the kings and emperors of states included, big and small, long- and short-lived, institutional development did not converge to a stable condition.

However, one may argue that many of the polities in the full sample existed for a short life and their rebel and even some peripheral states never amounted to much, so it is not surprising that they did not develop a workable political order, let alone well-functioning governance institutions. For this reason, Fig. 4.8 uses only subsample A and subsample B, which contain all the emperors from the mainstream dynasties covered in the Ershiwu Shi(Twenty-Five Histories). In this case, the regicide rate started a long-term declining trend after the middle of the Tang dynasty, with sharp declines during the second half of the Tang, then the Song and the Ming dynasties.

These changes suggest much institutional improvement in China, rather than stagnation as stated by others (e.g., Hegel1956; Elvin1972), since the late eighth century. This is consistent with the general view among historians that statecraft innovations introduced during the Tang and Song transition created the imperial Chinese state as we know it (e.g., Fairbankand Reischauer1989; Wang 2022). “In more than two thousand years of feudal society, despite fluctuations in imperial power, the general trend has been the continuous strengthening of the emperor’s autocratic rule. Absolute monarchy, with unlimited imperial power, has dominated history since the Qin and Han dynasties.“ Still, succession arrangements, the civil service examination system (keju) for selecting senior officials being opened to grassroots candidates since the early Song dynasty (Wang 2022), and other reforms must have been responsible for the gradually improving institutional environment, making it increasingly safer for the rulers.

Note that Figs. 4.7 and 4.8 diverge after the Song dynasty, with the regicide rate for the full sample rising thereafter in Fig. 4.7, whereas the regicide rate for subsample B continues its downward trend during the Yuan (1279–1368 CE), Ming (1368–1644 CE), and Qing (1644–1911 CE) in Fig. 4.8. This divergence is caused by the rulers of the rebel and peripheral regimes included in subsample C; especially during the Ming and Qing periods, many of these regimes were crushed by the imperial troops of the Ming and Qing, typically with their rulers executed, which exacerbated the regicide rate of these peripheral rulers.

Eisner’s (2011) data cover 1,513 kings of 45 European kingdoms between 600 and 1800 CE, from which he finds a regicide ratio of 22% for European kings, implying a regicide rate of 1,547 unnatural deaths per 100,000 ruler-years. To compare historical China with Europe, Fig. 4.9 plots the respective regicide rates of European and Chinese rulers, calculated using 100-year intervals. Clearly, from 600 to 1800 CE, the European regicide rate experienced a continuous downward trend, whereas the Chinese regicide rate was consistently higher and fluctuated.

Fig. 4.9

Regicide rate by century—comparison between China (full sample) and Europe. Notes Regicide rate stands for the number of regicides per 100,000 ruler-years. Data for Europeare from Eisner (2011)

If we limit our attention only to the Chinese emperors covered in subsample B (together with subsample A), Fig. 4.10 tells us that between 600 and 1800 CE, the regicide rate for Chinese rulers was about the same as for European kings, both following a declining trend. However, the Chinese regicide rate was far more volatile than its European counterpart, implying more unstable or unsettled institutional development in China’s mainstream empires even in modern times. It begs the question: why didn’t China’s institutions evolve and improve, to learn from past lessons and improve ruler security? This topic is obviously beyond the scope of the present chapter.

Fig. 4.10

Regicide rate comparison between China (Subsample B) and Europe. Notes Regicide rate stands for the number of regicides per 100,000 ruler-years. For China, subsample B comprises all emperors from 221 BCE to 1911 CE. Data for Europeare from Eisner (2011)

The differences between Figs. 4.9 and 4.10 remind us that when comparing China with Europe, we need to be careful with sample selection and should apply the same standard to define rulers; otherwise, biases would occur. We provide more comparisons using different subsamples in Appendix B.

4.5.2 Changes in Ruler Duration Over Time

Ruler duration is another metric for measuring the quality of a regime’s governance institutions. Before elections and term limits were introduced in modern times, the higher the quality and stability of institutions, the longer the state’s average ruler duration. In Table 4.4, we saw that, for natural-death cases, Chinese rulers had an average reign duration of 14.4 years, while rulers who died from regicide had a reign duration of only 7.4 years. This indicates a significant negative correlation between regicide rate and ruler duration.

Figure 4.11 plots both the 100-year and 300-year moving averages of China’s ruler duration, based on the full sample. Comparing both Figs. 4.7 and 4.11, we see an inverse relation between regicide rate and ruler duration: high regicide rates were associated with low ruler reign durations, and vice versa. It is intriguing that though the Spring and Autumn (770–476 BCE) and then the Warring States period (476–221 BCE) were supposed to be the unruliest times in early Chinese history, the pre-Qin rulers experienced the lowest regicide rates and the highest ruler durations. But, as shown in Fig. 4.11, ruler duration started a sharp downward trend after the Qin unified China, which continued until the Tang dynasty’s statecraft innovations encouraged the ruler duration to move upward. This finding suggests that when China was divided into many kingdoms, the incentive to murder a king was not high, making the regicide rate lower and ruler duration longer; but once the divided regimes were unified into one large empire, the emperor controlled a large territory and much more resources, raising his power to a new level; as a result, the incentive to kill the emperor and take over his throne became tremendously higher unless institutional innovations were introduced to constrain such behavior; hence, regicide rates were higher and ruler durations were shorter during periods of unified empires and before statecraft development advanced. Prior to the Qin, the average ruler duration fluctuated between 17 and 25 years; however, from the Qin to the Qing Dynasty, the average ruler’s reign moved between 10 and 15 years.

Fig. 4.11

Moving averages of Chinese ruler duration (full sample). Notes The two vertical black lines respectively stand for the establishment of the Qin dynasty in 221 BCE and of the Tang dynasty in 618 CE

Once we change our focus to the mainstream states, that is, subsamples A and B, Fig. 4.12 reveals a different story than the one based on the full sample. The general pattern for the average ruler duration remains similar between Figs. 4.11 and 4.12, for the 1046 BCE-618 CE period. But, as mentioned earlier, after the Tang dynasty was founded in 618 CE, it took statecraft development to a new level, followed by further institutional innovations during the Song (960–1279 CE). Consequently, the average ruler duration reversed its long-term declining trend after the Qin and started rising from the Tang dynasty onward and thus noticeable progress in governance institutions since the Tang.

Fig. 4.12

Moving averages of Chinese ruler duration (subsamples A and B). Notes The two vertical black lines respectively stand for the establishment of the Qin dynasty in 221 BCE and of the Tang dynasty in 618 CE. Subsample A includes all kings from 1046 to 221 BCE, while subsample B comprises all emperors from 221 BCE to 1911 CE

Finally, we conduct a comparative analysis of ruler durations among Chinese, European, and Islamic rulers. Again, the data are from Eisner (2011) for European kings and from Blaydesand Chaney (2013) for Islamic sultans. The overlapped period between their datasets is 600–1500 CE, so we employ our subsample B of Chinese emperors for this same period. Figure 4.13 plots the 100-year moving averages of ruler duration for the three sets of rulers. Based on Fig. 4.13, over the given period, the average ruler duration rose steadily in Europe, reflecting continuing improvement in governance institutions, whereas for Islamic societies, their average ruler duration was initially rising but, after the mid-ninth century, followed a clear downward trend partly because of the Sunni Revival. In contrast, China’s average ruler duration did not just go in one direction; instead, it swung up and down, implying that though there was cumulative progress in institutional innovations, China never found a solution to the power-governance challenge.

Fig. 4.13

Ruler duration comparison among Chinese (Subsample B), European, and Islamic rulers. Notes The ruler duration lines are based on 100-year moving averages. For China, subsample A includes all kings from 1046 to 221 BCE, while subsample B comprises all emperors from 221 BCE to 1911 CE. Data for Europe and Islam are from Blaydes and Eric Chaney (2013)

4.6 Conclusion

This paper presents a quantitative analysis of Chinese regicide history from 1046 BCE to 1911 CE. Our findings reveal that Chinese rulers faced a significantly higher risk of regicide than European kings but fared better than Islamic sultans. The life risk for an autocratic ruler in historical China was determined by a host of factors, such as how much power he monopolized, whether a successor was appointed, whether he assumed the throne at a young age and received assistance from a minister or ministers, how many competing polities were surrounding the regime, how many wars and natural disasters were happening, as well as dynastic characteristics. Still, statecraft innovations since the Tang dynasty did improve the quality of governance, leading to a long-lasting downward trend for the regicide rate, and an upward trend for ruler duration—at least for the mainstream dynasties. Thus, starting with the Tang-Song transition, China did not seem to stagnate in institutional development, contrary to the claims by Hegel (1956), Elvin (1972), and Jinand Liu (2011). However, China’s long-term decline in regicide occurred later and at a slower and much more volatile pace than Europe’s.

The existing literature on state capacity has largely focused on its impact on economic growth, rather than on such outcomes as regicide, homicide, and general violence (e.g., Broadberry et al. 2018; Qianand Sng2022). Broadberry et al. (2018) estimate that China’s real per capita gross domestic product peaked in the Northern Song (960–1126 CE) and declined thereafter, implying that whatever happened in state capacity development since then did not amount to any meaningful improvement in living standards. Figures 4.10 and 4.12 show that the regicide rate declined and ruler duration increased during the Song, Ming, and Qing dynasties (see also Figs. 4.14 and 4.15 in Appendix B). Therefore, while governance innovations in these periods may not have improved state capacity for economic development, they brought about progress in terms of institutional quality for political governance and social stability, especially for the peaceful transfer of power. Perhaps in this sense, the Ming and Qing emperors put too much emphasis on political stability and holding onto power at the expense of economic growth for the people.

Appendices

Appendix A. Summary Statistics for Subsamples

Tables 4.8–4.10 present the dynasty/regime-level statistics for the three subsamples, while Tables 4.11–4.13 display ruler-level information.

Table 4.8 Statistics by dynasty: subsample A (kings, 1046–221 BCE)

Table 4.9 Statistics by dynasty: subsample B (emperors, 221 BCE-1911 CE)

Table 4.10 Statistics by dynasty: subsample C (non-emperors, 221 BCE-1911 CE)

Table 4.11 Statistics by ruler: subsample A (kings 1046-221BCE)

Table 4.12 Statistics by ruler: subsample B (emperors, 221 BCE-1911 CE)

Table 4.13 Statistics by rulers: subsample C (non-emperors, 221 BCE-1911 CE)

Appendix B. Evolution of Regicide Rate Over Time

Based on the full sample, Fig. 4.14 presents the 100-year moving average of regicide rate per 100,000 ruler-years, exhibiting significant volatility; during the Three Kingdoms, the North and South Dynasties, and the Qing Dynasty, the regicide rate was notably high. Figure 4.15 portrays the regicide ratio by century, displaying a similar trend to Fig. 4.14.

Fig. 4.14

Regicide rate over time (Full Sample). Notes Regicide rate stands for the number of regicides per 100,000 ruler-years. The two vertical black lines respectively stand for the establishment of the Qin dynasty in 221 BCE and of the Tang dynasty in 618 CE

Fig. 4.15

Regicide ratio by century (full sample)

Figures 4.16 and 4.17 are constructed similarly to Figs. 4.14 and 4.15, except that they are based on subsamples A and subsamples B. The main difference between Figs. 4.14 and 4.16 is that for the mainstream emperors from the Tang dynasty onward, their life risk started a long-term decline because of the statecraft innovations that began in the Tang, with the regicide rate reaching its low in the Qing. Figure 4.17 shows a similar pattern as in Fig. 4.16 but based on the regicide ratio within each century, where the regicide ratio peaked during the Eastern and Western Jin dynasties as well as the Northern and Southern dynasties and reached its lowest level during the Ming and the Qing.

Fig. 4.16

Regicide rate over time (subsamples A and B). Notes Regicide rate stands for the number of regicides per 100,000 ruler-years. The two vertical black lines, respectively, stand for the establishment of the Qin dynasty in 221 BCE and of the Tang dynasty in 618 CE. Subsample A includes all kings from 1046 to 221 BCE while subsample B comprises all emperors from 221 BCE to 1911 CE

Fig. 4.17

Regicide ratio by century (subsample A&B). Notes Subsample A includes all kings from 1046 to 221 BCE while subsample B comprises all emperors from 221 BCE to 1911 CE