A Theory of Power Structure and Institutional

Compatibility: China vs. Europe Revisited∗

Ruixue Jia† Gérard Roland‡ Yang Xie§
September 16, 2021

Abstract
The literature on institutions and development contrasts inclusive societies with
extractive ones, while the scholarship comparing Imperial China and Premodern Eu-
rope defies this dichotomy. To reconcile these views, we model the institutional dif-
ferences between the two societies along two dimensions of their power structure: the
Ruler’s absolute power was weaker in Europe, whereas the Elite–People relationship
in terms of their power and rights was more balanced in China. Our model shows that
a more balanced Elite–People relationship can be compatible with a more absolutist
Ruler. The model also helps interpret differences in specific institutions and autocratic
stability.

Keywords: absolute power, Elite–People relationship, autocratic stability, inclu-

sive institution, rule of law, access to elite status

∗ We are grateful to Avner Greif and Joel Mokyr for their thorough and constructive feedback. We have
also benefited from thoughtful comments from Chris Bidner, Gary Cox, Wei Cui, Georgy Egorov, Patrick
Francois, Murat Iyigun, Mark Koyama, Ling Li, Weijia Li, Zhao Liu, Zhaotian Luo, Pietro Ortoleva, Al-
bert Park, Jean-Laurent Rosenthal, Jared Rubin, Larry Samuelson, Tuan-Hwee Sng, Michael Zheng Song,
Konstantin Sonin, David Stasavage, Guido Tabellini, Chenggang Xu, and Xueguang Zhou. We also thank
participants in online seminars hosted by BOFIT, Fudan, Harvard, Tsinghua, UBC, UCSD, and USC, the
economic history reading group at Monash, and the 2021 ASSA, CESI, CPSSS, and SIOE meetings for their
valuable comments. We thank Ming Zhang for his excellent research assistance. We appreciate the support
from the EDI grant “Endogenous Political Fragmentation: The Role of Property Rights in Historical Per-
spective” and the valuable feedback from EDI. An earlier version of the paper has been circulated as NBER
Working Paper 28403 and CEPR Discussion Paper 15700.
† UCSD and LSE, CEPR, and NBER; rxjia@ucsd.edu.
‡ UC Berkeley, CEPR, and NBER; groland@econ.berkeley.edu.
§ UC Riverside; yang.xie@ucr.edu.

1
1 Introduction
The very influential literature on institutions and development has often contrasted the
more inclusive, open-access, and equal societies with the more extractive, limited-access,
and unequal ones (e.g., North, 1989; North and Weingast, 1989; Acemoğlu, Johnson and
Robinson, 2001, 2005a,b; North, Wallis and Weingast, 2009; Acemoğlu and Robinson,
2012; Cox, North and Weingast, 2019). At the same time, the scholarship comparing Im-
perial China and Premodern Europe seems to contradict this dichotomic classification. It
has been emphasized that, although rule of law and protection of property rights against the
Ruler were weaker in China, some other institutional features of China were more inclusive
compared to Europe: for example, the access to elite status was primarily non-hereditary
and governed by the civil service exam, peasants enjoyed a greater degree of freedom,
and land ownership was less concentrated (e.g., Finer, 1997a,b; Fukuyama, 2011; Tackett,
2014; Zhang, 2017; Acemoğlu and Robinson, 2019; Stasavage, 2020; Greif, Mokyr and
Tabellini, Forthcoming; the survey by Qian and Sng, 2021). Are these views of institution-
al differences contradictory to each other? More generally, why is it possible for a society,
be it Imperial China, Premodern Europe, or some other society, to be quite inclusive in a
few institutional dimensions, but not so much in others?
In this paper, we reconcile these views and address these questions by providing a
framework that analyzes institutional differences along two, instead of one, dimensions of
the power structure of society, which is about how power and rights were allocated across
the Ruler, the Elites, which included primarily the lords in Europe and bureaucrats in China,
and the common People. First, in Europe, the power and rights of the Elites and People
were less conditional on the Ruler’s will compared to China, i.e., the absolute power of
the Ruler was weaker. This was reflected in, for example, the different strengths of rule of
law, property rights, and whether the king or emperor had ultimate ownership and control
over land and population. Second, in China, the Elite–People relationship in terms of their
power and rights was less asymmetric compared to Europe. This can be seen, for example,
by how the access to elite status was governed, how much freedom the peasantry enjoyed,
and how unequal land ownership was. The characterized power structure differences were
the most prominent during the 9–14th centuries, with persistence beyond, between the
society in the historical core of Imperial China and the western–central European society
where feudalism was once pervasive.
After examining rich comparative historical narratives based on our power structure

2
framework, we present a simple game-theoretical model of the relationship between the
two dimensions of the power structure. We start with a Ruler, who prefers to maintain a
particular status quo of autocratic rule, and a Challenger, who could try to alter it. Since
the Challenger can be either a foreign threat not under the Ruler’s rule, a conspiring elite,
or a rebellious population under the Ruler’s rule, since the Challenger’s goal does not nec-
essarily involve dethroning the Ruler, and since the challenge can be armed or nonviolent,
our model is sufficiently general to cover a wide range of threats that could destabilize an
autocratic rule. In the model, we assume that the success of a potential challenge in altering
the status quo depends on whether the Elites and People choose to side with the Ruler. In
the model, more symmetric power and rights between Elites and People is represented by
less unequal payoffs, if they have not defied the Ruler; we model a stronger absolute power
of the Ruler as a greater proportional reduction in the payoffs of the ruled, i.e., a heavier
punishment, after they unsuccessfully defied the Ruler.
Analysis of the model leads to a comparative institutional theory on the compatibili-
ty between the two dimensions of the power structure. In the analysis, we first take the
level of the Ruler’s absolute power as exogenous and analyze how it affects the stabili-
ty of autocratic rule and the Ruler’s perspective about the Elite–People relationship. In a
historical perspective, the Ruler’s absolute power is determined by a set of slow-moving
institutions that affect people’s expectations, values, and beliefs (e.g., Roland, 2004, 2008),
so it seems appropriate to start by taking this parameter as exogenous. To be precise, we
view the absolute power of the Ruler as the conditionality of power and rights of the ruled
on the Ruler’s will. Given any non-zero level of such conditionality, the more power and
rights the People enjoy when they have not defied the Ruler, the more they will lose if they
defy the Ruler, and, therefore, the more they will be willing to side with the Ruler during
a challenge. We call this the punishment effect of more power and rights of the People.
Knowing that a stronger alliance between the Ruler and People has worsened the prospect
of a challenge to the Ruler, the Elites will be more willing to side with the Ruler, too. We
call this the political alliance effect.1 The Challenger would then be deterred from chal-
lenging the status quo, stabilizing the autocratic rule and thus creating an incentive for the
Ruler to promote a more symmetric Elite–People relationship. Since a stronger absolute
power of the Ruler implies a greater aforementioned conditionality, it will make the initial
1 As remarked by Orwell (1947, p. 17), this idea of the Ruler and the People “being in a sort of alliance
against the upper classes” is “almost as old as history” in Europe; in China the same idea can be traced to not
later than Han Feizi (Watson, 1964, p. 87) from the 3rd century BC, which has been the most representative
text in the Chinese Legalist tradition since then.

3
punishment effect and, therefore, the total stabilizing effect stronger. The Ruler’s incentive
to promote a more symmetric Elite–People relationship will thus be greater when the Ruler
has a stronger absolute power. Having a more absolutist Ruler can thus be compatible with
a more symmetric Elite–People relationship, reconciling the seemingly contradictory views
in the literature on the institutional differences between Imperial China and Premodern Eu-
rope.
As we show in Online Appendix E, the insights and results from the theory are robust in
a Markov game in which the ruled covet the Ruler’s throne and all players take continuation
values into consideration. We also show in Online Appendix C that the compatibility result
also holds in the other direction, i.e., the stabilizing effect of a stronger absolute power of
the Ruler is increasing in the level of symmetry between the Elites and People.
A few additional implications arise about the power structure in the long run when we
extend the model by endogenizing the level of the Ruler’s absolute power. For example,
because a more absolutist Ruler can take away more of the power and rights of the People
once they unsuccessfully defy him, he is more willing to grant more power and rights
to them in the first place. Therefore, the People may prefer the Ruler to enjoy a stronger
absolute power, defying less often and enjoying their granted power and rights under a more
stable autocratic rule. This makes it possible for the power structure of a strong absolute
power and a relatively symmetric Elite–People relationship to be incentive-compatible for
both the Ruler and the People, and therefore to persist.
In another extension, we allow the current political stability, which has resulted from
the current power structure, to influence the future power structure, creating a dynamic
complementarity. We show that it is thus possible for two societies that differ slightly in
their power structure or autocratic stability to diverge into different steady states – one with
a stronger absolute power of the Ruler, a more symmetric Elite–People relationship, and a
higher stability of autocratic rule than the other.
Given these results, we further explore the historical relevance of our theory. We first
discuss how our theory can help understand specific institutions. For example, we can in-
terpret the use of the civil service exam to govern the access to the bureaucracy in China
and the rise of cities in Europe both as resulting from the Ruler’s efforts to reduce the Elite–
People asymmetry.2 Second, we examine the auxiliary predictions from our model about
the impact of the power structure on the stability of autocratic rule. We systematically com-
2 To be sure, the development of cities in Europe was to a large extent based on autonomous factors and

exogenous shocks, but as we show below, various kings acted to promote urban development.

4
pare Imperial China and Premodern Europe in the frequency of wars, the risk of deposition
for a Ruler in a given year, and the resilience of unified autocratic rule. Consistent with the
predictions of our model, the data show that autocratic rule was more stable in China than
in Europe over the 9–14th centuries, when the differences in the power structure were the
most prominent, with persistence in later centuries.
Our paper contributes to the political economy literature on institutions and develop-
ment by investigating the relationship between major components of inclusive institutions
(e.g., the aforementioned; Besley and Persson, 2011, 2014; Mokyr, 2016; Acemoğlu and
Robinson, 2019). The literature often analyzes society by categorizing it into two estates
(e.g., state vs. society, elites vs. mass, those with vs. those without access to political
and economic resources and decisions), and it has taught a general lesson that a strong
interdependence and synergy exists between institutional arrangements that are conducive
to sustainable political, economic, and social development, such as rule of law and prop-
erty rights on the one hand and a more open access to elite status on the other hand (e.g.,
North, Wallis and Weingast, 2009; Besley and Persson, 2011, 2014; Acemoğlu and Robin-
son, 2012). We extend the two-estate framework into a three-estate one. By that, we show
that the more repressive an institution is in the dimension of the Ruler’s absolute power,
the more inclusive it may be in the dimension of the power and rights equality between the
Elites and People, and this pattern may well persist. This seemingly paradoxical result is,
to our knowledge, new to the literature.
Our paper also contributes to the literature on the strategies that a ruling class can use to
fend off challenges to their rule. For example, democratization and enfranchisement as in
Acemoğlu and Robinson (2000, 2001) can serve as a credible commitment to redistribution
by shifting the decision power to the median voter, avoiding destructive revolution; Ace-
moğlu, Verdier and Robinson (2004) propose that an ad-hoc divide-and-rule policy given
weak institutional constraints can intensify the collective action problem among the ruled,
making the ruler safer; Padró i Miquel (2007) argues that the fear by the ruled of falling
under an even worse ruler external to the incumbent ruler can help discipline the ruled,
achieving stability for the incumbent ruler. Our analysis suggests that an absolutist Ruler
can co-opt the People and thus secure his autocratic rule by making the People’s power
and rights more comparable to the Elites’, for example by promoting meritocratization.
Compared with the literature, strategies of this type are unique in the sense that they do
not change where the absolute power lies, provide an ex-ante committed payoff schedule
through the institutional design of the power structure, and the Ruler’s incentive to engage

5
in such strategies depends on the absolute power of the Ruler in the same power structure.
Conceptually the closest to us in this thread of literature, Persico (2021) shows in a
general model for political regimes that as long as civil liberties are imperfectly protect-
ed, which is similar to our notion that the power and rights of the ruled are conditional
on the Ruler’s will, a politician will always have an incentive to promise equal treatment
across citizens, trying to win their coordinated support, which is similar to the more sym-
metric Elite–People relationship in our context with the political alliance effect involved.
Concurrently and independently developed, our paper and Persico (2021)’s paper comple-
ment each other: Persico (2021) focuses on policy treatment and provision of public goods,
whereas we focus on the compatibility within the power structure and its implications for
political stability.
It has been well documented in the literature that the unified autocratic rule of a domi-
nant state could hardly be maintained in Europe since the fall of the Roman Empire, while
in Imperial China it was relatively resilient (e.g., Finer, 1997a,b; Scheidel, 2019). A few
explanations of the political divergence have been proposed from the environmental, ge-
ographical, or geopolitical perspective (e.g., Wittfogel, 1957; Jones, 1981; Turchin, 2009;
Dincecco and Wang, 2018; Scheidel, 2019; the survey by Qian and Sng, 2021). Among
recent contributions, Ko, Koyama and Sng (2018) and Fernández-Villaverde, Koyama, Lin
and Sng (2020) show that the spatial distribution of external threats and the existence or
not of a high-productivity, traversable core geographical region could have played impor-
tant roles in facilitating the unification of China and fragmentation of Europe.
We contribute to the literature on the same topic but from the institutional front. First,
we introduce the power structure approach to the literature, and we show that the charac-
terized power structure differences between the two societies can explain the differences in
their autocratic stability. The same approach can also shed light on some important varia-
tions and changes within China and Europe and on other parts of the world, as we briefly
discuss in Sections 2.1 and 5.
Second, Acemoğlu and Robinson (2019) have emphasized the role of the state–society
struggle in understanding the political divergence. We complement their view by a richer
strategic dynamics: a political alliance could exist between the head of the state, i.e., the
Ruler, and the lower classes of society, i.e., the People, together against the state apparatus
members or higher classes of society, i.e., the Elites, especially when the Ruler’s absolute
power is strong. We can thus provide a more general understanding of a few specific institu-
tional arrangements, on which the literature has focused to explain the political divergence,

6
such as the wage of the tax-collecting agents, the tax on the masses, and the development
of fiscal capacity (Gennaioli and Voth, 2015; Ma and Rubin, 2019), the availability and
capacity of bureaucracy (Stasavage, 2020), and meritocracy as an informational solution to
the institution–power friction (Huang and Yang, 2021). That is to say, we can read the low
wage–low tax equilibrium and bureaucracy with meritocratic recruitment in Imperial China
as a more symmetric Elite–People relationship and thus a stronger Ruler–People alliance;
these were compatible with the stronger absolute power of the Ruler and higher autocratic
stability in China.
Finally, some studies have explored, along the technological, geographical, and eco-
nomic lines, the more exogenous factors behind the initial differences in the power structure
(e.g., McNeill, 1982; Roland, 2020; Stasavage, 2020). Although not focusing on the ori-
gin of these initial differences, we show that such differences can be incentive-compatible
and exhibit self-reinforcing dynamics over time, together with the persistent difference in
the autocratic stability. This dual divergence of the power structure and autocratic stability
complements the divergence of culture and its co-evolution with political institutions (Greif
and Tabellini, 2010, 2017; Greif, Mokyr and Tabellini, Forthcoming).
The paper is organized as follows. Section 2 briefly presents historical narratives in
the power structure framework on the institutional differences between Imperial China and
Premodern Europe. Section 3 presents the settings, analysis, and extensions of the model.
Section 4 explores the historical relevance of the theory with further discussion and stylized
facts. Section 5 concludes the paper.

2 Power Structure in Historical Narratives

2.1 Scope and Focus of Historical Narratives

In this section, we discuss historical narratives on the institutional differences between Im-
perial China and Premodern Europe along the two dimensions of the power structure of
society. In the narratives, by “China,” we consider the society in “the historical core of Im-
perial China,” i.e., “the traditionally agrarian part of China south of the Great Wall and east
of the Tibetan Plateau” (Fernández-Villaverde, Koyama, Lin and Sng, 2020, p. 8, 12). By
“Europe,” unless clarified otherwise, we follow the focus of Bloch (1962a,b), Finer (1997b,
Book III, Part III), and Blaydes and Chaney (2013), i.e., the Romano–Germanic influenced
or assimilated society in western and central Europe where feudalism was once pervasive.

7
This society was “[h]emmed in by these three blocs, Mohammedan, Byzantine, and Slav”
(Bloch, 1962a, p. xxvi) and “comprised principally the British Isles, the Scandinavian
countries, France, Germany, Italy, and northern Spain” (Finer, 1997b, p. 855).3
The most relevant period of the characterized power structure differences was the 9–
14th centuries, with persistence beyond. This period covered the rise and decline of feudal-
ism in Europe (e.g., Bloch, 1962a,b; Ganshof, 1952), with the Black Death taking place in
the middle of the 14th century; in Imperial China, it was since the Tang dynasty (618–907)
that political institutions had largely been stable, after the swings during the 800 preceding
years (e.g., Yan, 2009).
Admittedly, important variations and changes in the power structure existed across poli-
ties and over time within China and within Europe. At the same time, “over and above”
these variations and changes, historians have emphasized “the predominant quality of a
common civilization” in Europe and the “evolving axis” or “theme” of the institutional and
cultural characteristics of Chinese society during our focused period (e.g., Bloch, 1962a,
p. xxvi; Yan, 2009, p. 11–12). We follow this insight in our narratives: we try to identi-
fy the “ideal type” of the differences in the power structures between Imperial China and
Premodern Europe, sometimes discussing specific polities or periods as examples; we al-
so adopt the longue durée approach by focusing on significant, persistent features of the
power structures.
By no means denying the aforementioned important variations and changes within Chi-
na and Europe, our narratives can provide a benchmark to help interpret them in the power
structure framework. For example, within Europe, consistent with Proposition 3 below,
Bloch (1962a, p. 180) observes the co-existence of a stronger absolute power of the king
over his vassals and less significant class distinctions between the lords and the peasantry
in Germany from the 10th century to the end of the Middle Ages, compared with the power
structure in France at that time; within China, we discuss in Online Appendix G how our
framework can help interpret the dynastic cycles in Chinese history, each of which was
marked by a co-decline of the absolute power of the Ruler, the Elite–People symmetry, and
the stability and effectiveness of autocratic rule (e.g., Skinner, 1985; Usher, 1989; Dillon,
1998).
We summarize the historical narratives in Table 1, and we elaborate on them below.
3 Following Blaydes and Chaney (2013), “Europe” by this definition covered all countries under the section

“The Barbarian West” and the subsections “The British Isles,” “France,” “The Low Countries,” “Italy,” “The
Iberian Peninsula,” “The German-speaking States,” “Scandinavia,” and “Crusader States” under the section
“Europe” in Morby (1989).

8
 Table 1: Power Structure in Imperial China and Premodern Europe

China Europe Examples of references

Absolute power of the Ruler

Bloch (1962b), Unger (1977), Mann (1986)
Ruler less constrained Ruler constrained Finer (1997a,b), Tamanaha (2004)
Strength of rule of law
by law by Church and law Fukuyama (2011), Acemoğlu and Robinson (2019)
Greif, Mokyr and Tabellini (Forthcoming)

Reserved for Ruler; Confiscation highly

confiscation legitimate constrained; Ruler Chao and Chen (1982), Levi (1988)
Ultimate ownership of land
when Ruler deemed expected to “live of Finer (1997b), Wang (2000)
it necessary his own”

Ruled considered Ruler’s Limited control; much

Bloch (1962a), Lander (1961), Levenson (1965)
Ruler’s control over population subjects; harsh penalty less harsh punishment
Mann (1986), Finer (1997a,b)
9

for disloyalty for disloyalty

Asymmetry in power and rights

between Elites and People
Elites a supreme class;
General comparison Much less unbalanced Bloch (1962b), Lü (1944), Finer (1997b)
oppressive to the poor

Non-hereditary, elite
Hereditary vs. non-hereditary Kemp (1970), Finer (1997b), Wickham (2009)
status governed through Hereditary nobility
access to elite status Yan (2009), Parish (2010), Tackett (2014)
the civil service exam

Serfdom common in
Mostly free and landowning Esherick (1981), Chao and Chen (1982)
Middle Ages; land
Inequality in land ownership peasantry; land ownership Beckett (1984), Finer (1997a), Wickham (2009)
ownership much more
less concentrated Tackett (2014), von Glahn (2016), Zhang (2017)
concentrated

Goody, Thirsk and Thompson (1976)

Primogeniture increasingly
Inheritance rule Partible inheritance Goldstone (1991), Bertocchi (2006)
more common
von Glahn (2016)
2.2 Absolute Power of the Ruler
The first difference we emphasize is that Chinese Rulers enjoyed a stronger absolute power
than their European counterparts, by which we mean that the power and rights of the ruled
were more dependent on the Ruler’s will in China compared to Europe. This difference
is first reflected in the strength of rule of law, and then in the ultimate ownership and
control over the most important assets in historical societies: land and population. Next,
we summarize the narratives and explain how we model the degree of absolute power based
on these narratives.

Strength of rule of law. As noted by many scholars, Chinese emperors were less con-
strained by rule of law (Finer, 1997a,b; Stasavage, 2016; Acemoğlu and Robinson, 2019;
Ma and Rubin, 2019, p. 227; Greif, Mokyr and Tabellini, Forthcoming).4 As put by Fin-
er (1997a,b, p. 455, 836), all the ruled, including the top bureaucrats, were “subjects not
citizens” and had only “duties not rights”; as observed by Fukuyama (2011, p. 290) and
Unger (1977, p. 104), “law was only the positive law that [the emperor] himself made” and
it “could be as general or as particular as the policy objectives of the rulers might require.”5
In contrast, European Rulers faced strong constraints from the Christian church (Mann,
1986; Fukuyama, 2011; Johnson and Koyama, 2019; Scheidel, 2019; Greif, Mokyr and
Tabellini, Forthcoming). Given the Pope’s threat to delegitimize and excommunicate them,
“[k]ings . . . could not defy the Pope for very long,” as shown in many examples (Southern,
1970, p. 130).6 The king also faced much tighter legal constraints. In the famous words
of Bracton (1968, vol. 2, p. 33), “[t]he king must . . . be under the law, because law makes
the king.” Having emerged from the 9th-century customary law, a man’s right to judge and
4 The Chinese Ruler had the obligation to act benevolently towards the ruled and to follow the “Mandate of

Heaven” (e.g., Zhao, 2009), but as noted by Stasavage (2016, p. 148), “the concept of a Mandate of Heaven
never extended to obtaining consent, nor did it involve assembling representatives to achieve this goal.” Finer
(1997a, p. 462) also notes: “[i]deally, government must be of the people, for the people: but, emphatically,
Mencius never for a moment hints that it can ever be by the people. Very much the reverse. . . . Nor did a
dissatisfied populace have the right to rebel.” Perry (2008) further contrasts the right to rebel in the Anglo-
American tradition, which is against tyranny and for liberty, i.e., about the Ruler’s absolute power and γ in
our model, with the People’s rights in the Chinese tradition, which is for livelihood and against poverty and
socioeconomic injustice, i.e., mainly about the Elite–People balance and β in our model.
5 For example, the founding emperor of the Ming dynasty created “law beyond the law” when he was

frustrated by the Great Ming code of his own, while insisting that only he could use the newly created law
(Brook, 2010, p. 87). Unger (1977, ch. 2) discusses the characteristics of law in Imperial China in detail.
6 Famous examples include the dramatic scenes of Henry IV of Germany at Canossa, Henry II of England

at Canterbury, and King John of England at Dover, and the destruction of Holy Roman Emperor Frederick
II’s family.

10
resist when his king had acted unlawfully had been repeatedly recognized by significant
legal documents through the Middle Ages (Bloch, 1962b, p. 172–173).7 Importantly,
this right was “not subject to the king’s justice” and “not upon the desires of the king”
(Tamanaha, 2004, p. 26).8

Ultimate ownership of land. While land could be owned by individuals in normal times
in China, the ultimate legitimacy of land ownership was always reserved for the Ruler,
so the emperor could re-centralize the ownership when he deemed it necessary (Chao and
Chen, 1982; Wang, 2000). Since even before the Qin dynasty unified China in 221 BC,
land confiscation from the noble families and landed gentry had been a common practice
of the Chinese Ruler to raise revenue for military projects (Ebrey and Walthall, 2013).9 De-
pending on the emperor’s will, systematic persecutions against Buddhism, Manichaeism,
and other religions also repeatedly happened, regularly entailing large-scale confiscation of
temple properties (de Groot, 1903, ch. 2).
In contrast, when European Rulers needed revenues, they could usually not confiscate
land from the Elites or the Church, at least during the 9–14th centuries.10 Instead, they had
to exchange rights or resources with revenues. Levi (1988, p. 99) states it clearly: “[d]uring
the medieval period, a monarch was expected to ‘live of his own’ (vivre du sien). That is,
funds for the monarch were to come from royal lands and customary dues. . . . Should
monarchs need more, even if it was to fund a campaign on behalf of the country as a whole,
they had to obtain assent to some form of ‘extraordinary’ taxation. They could neither
expropriate property at will nor rely on a regular levy.”11
7 Bloch (1962b, p. 173) raises examples of “the English Great Charter of 1215; the Hungarian ‘Golden

Bull’ of 1222; the Assizes of Jerusalem; the Privilege of the Brandenburg nobles; the Aragonese Act of
Union of 1287; the Brabantine charter of Cortenberg; the statute of Dauphiné of 1341; the declaration of the
communes of Languedoc (1356).”
8 For more extensive discussion on the rule of law, see Finer (1997b), Tamanaha (2004), Fukuyama (2011),

Vincent (2012), Fernández-Villaverde (2016), Acemoğlu and Robinson (2019), and Greif, Mokyr and Tabelli-
ni (Forthcoming).
9 Among famous early examples, Duke Xiao of the Qin state confiscated land from the feudal nobles in

the 340s BC, sharing it among the peasants; in 114 BC, Emperor Wu of Han confiscated land from nobles
and merchants to raise additional revenue to fund the Han–Xiongnu War.
10 European Rulers could become more capable of expropriating the Church as their absolute power grew,

but mainly in a later period, i.e., the 16–18th centuries, and especially during the Reformation. One may
also note that the famous dissolution of the English monasteries under Henry VIII (1536–1541) was ac-
companied by the rise of the gentry, who were commoners but their power and rights grew relative to the
peerage (Heldring, Robinson and Vollmer, 2021). This was consistent with Proposition 3 below about the
compatibility between a greater absolute power of the Ruler and a more balanced Elite–People relationship.
11 See also Finer (1997b, p. 887) for a similar observation. Besides, when Louis XIV managed to tax the

11
Ruler’s control over population. As the population were subjects of the Ruler in China,
the Ruler could reward or punish anyone arbitrarily, which precisely reflected his absolute
power (Levenson, 1965, p. 39; Finer, 1997a, p. 455). Consistent with the emphasis of Con-
fucianism on the loyalty of the ruled to the Ruler (Greif, Mokyr and Tabellini, Forthcom-
ing), one person’s rebellion, treason, or even slight disobedience, regardless of her social
status, would be punished extremely harshly, usually leading to eradication of the whole
family line (Finer, 1997b, p. 778).12 Sometimes mere suspicion from the Ruler could guar-
antee the calamity, as shown in the fall of Princess Taiping in 713.13 Following the Legalist
tradition in Chinese political philosophy, the absolute right to impose harsh punishment
on the ruled could effectively help the Ruler have strong control of both the Elites and the
People, despite sometimes significant administrative constraints (e.g., Watson, 1964; Sng,
2014).
In contrast, in feudal Europe, the king, in practice, did not have direct control over
peasants because the latter were controlled by their overlords; the peasants could, as a rule,
be punished by local courts controlled by their overlords, and the king did not have control
over these local courts (Bloch, 1962a). Although loyalty was also emphasized in Europe
and enforced through mechanisms like oaths, treason was punished much less harshly than
in China. First, although execution of the traitor and attainder could apply, killing the
family seldom applied, and the attainder would often later be reversed (Lander, 1961).14
Second, it was common in the feudal system for a vassal to have two or more overlords
(Bloch, 1962a) and when in conflict, he could simply choose which to follow (e.g., Cantor,
1964, p. 202; Tuchman, 1978; Mann, 1986). Eventually, as Finer (1997b, p. 881) observes,
the Ruler’s control over the population was “abysmal” and he “could not always count on
the fidelity of the vassal,” precisely because his lack of ability to punish them: “after all,
[they were] in possession of his lands and what could he do if defeated?”
nobility for the first time, the taxes happened only at the end of his reign and were insignificant in size and
subject to numerous exemptions (McCollim, 2012). Expropriations did happen but mostly under Eminent
Domain (Reynolds, 2010); in case of serious crimes like treason, the nature of the crime had to be determined
by law, not merely the Ruler’s will (Lander, 1961).
12 In a famous case, when Fang Xiaoru, a prominent minister, refused to write an inaugural address for

Emperor Yongle of Ming, the emperor sentenced 873 people to death, including Fang’s family, kinfolk,
friends, and students, before having Fang himself executed.
13 In 713, Emperor Xuan of Tang, merely suspecting that his aunt Princess Taiping had been planning a

coup, forced her to commit suicide and executed several dozens of her extended family and allies. Literary
inquisitions for merely potentially subversive attitudes to the Ruler were also conducted at a frequency and
scale much more significant than in Europe (e.g., Xue, 2021).
14 For example, during the reigns from Henry VI to Henry VII of England, 64% of the attainders were

eventually reversed (Lander, 1961, p. 149).

12
Formalization in our model. Motivated by these narratives, we assume that the Ruler,
Elites, and People are sharing a surplus of size π ; when the Ruler had survived a challenge
to the status quo of his rule in which the ruled did not side with him, he could punish
the defiers by having them enjoy only γ of their share of the surplus. Given the initial
distribution of the surplus, a lower γ indicates that more of the power and rights of the
ruled are conditional on the Ruler’s will, i.e., the Ruler has a stronger absolute power and
is more capable of exerting punishment on the ruled for defiance.

2.3 Asymmetry in Power and Rights between Elites and People

The power structures of Imperial China and Premodern Europe were also different in the
relationship between the Elites and the People. In Bloch’s words, the disparity between
“[a] subject peasantry” and “the supremacy of a class of specialized warriors” was one of
“the fundamental features of European feudalism” (Bloch, 1962b, p. 167), and his final
verdict on the system concerns only its constraints on the Rulers and its oppressiveness
towards the poor (Bloch, 1962b, p. 173). In contrast, prominent Chinese historian Lü
Simian summarizes the scenario in Imperial China elegantly: “once the father or elder
brother takes the throne, the sons and younger brothers,” who are princelings themselves,
“will become mere commoners” in terms of their power and rights (Lü, 1944, p. 347). Finer
(1997b, p. 836) also comments that “[i]n practice, China was a two-class society” where
the ruled, including “even the higher mandarins” and commoners, were having relatively
equivalent power and rights in front of the Ruler.
The difference in the Elites–People relationship was reflected by differences in, for
example, the dominance of the hereditary versus non-hereditary access to elite status, in-
equality in land ownership, and the inheritance rule. As above, we summarize the narratives
and then explain how we formalize them in our model.

Hereditary vs. non-hereditary access to elite status. In Medieval Europe, elite status
was governed primarily by hereditary nobility. As Finer (1997b, p. 879–880) explains,
“lineage [was] much more important than initiation,” while “the very right to be a vassal
(i.e. to hold a fief) [was] confined to those already noble!” Government positions, espe-
cially in courts and the army, were reserved for aristocrats. Although ordinary peasants
routinely performed military service as a privilege in the early Middle Ages, this stopped
to be the case later and was reserved for knights and higher titled nobles (for more dis-

13
cussion, see, e.g., Wickham, 2009). Access to priesthood and religious orders was not
forbidden to commoners, but even after the Gregorian reform in the 11th century, “the
abolition of . . . the hereditary ecclesiastical benefice” had remained a “formidable task” in
western Christendom until as late as the 13th century (Kemp, 1970, p. 1; Parish, 2010, p.
88–92).
In contrast, as early as during the 5–4th century BC, accompanied by reforms that
strengthened the absolute power of the Ruler, the Warring States in China had started to
abolish hereditary titles and make elite status open to the common People and dependent
solely on military merit (Yan, 2009, p. 23–24). To facilitate the fluid exchange between
the Elites and the People, the Sui dynasty (581–619) established the civil service exam to
regulate elite status, and the exam system was greatly developed during the Tang dynasty
(618–907). Notably, the exam was in principle open to almost all adult males, and elite
status gained via success in the exam could not be inherited. Following the destruction
of the aristocratic clans during the fall of the Tang dynasty, elite status in China had been
governed mainly by the exam system. Sustained by “a culture of merit,” the resulting
Chinese Elites were “more diffuse [and] justified . . . on the basis of talent and education”
instead of hereditary titles, which “would constitute one of the most striking distinctions
between Chinese and Western societies over the course of the subsequent millennium”
(Tackett, 2014, p. 3–5).
It may be worthwhile to comment here on the difference between the de jure and de
facto access to elite status. First, given the lack of comparative historical evidence on the
de facto difference in the access to elite status between Imperial China and Premodern
Europe for the 9–14th centuries, when our characterization of the power structure was the
most relevant, we do not take a strong stand on this subject.15
15 In Medieval Europe, with the system of hereditary aristocracy, the most visible way of social preferment
for commoners was the Church, which could have been comparable to the civil service exam in China.
According to Herlihy (1973) and Barrow (2015), however, it was still mainly the few landowners, patricians,
or clerics themselves, if allowed, who sent their children to the clergy, since only they needed to cut down
the numbers of heirs and could afford losing precious family labor. In practice the relevance of ecclesiastical
careers to commoners was thus limited. Herlihy (1973) also identifies three main patterns of social mobility in
Medieval Europe, and ecclesiastical careers were not among them. For later periods only scattered evidence
for China and England is available. Ho (1959) documents that during 1752–1938, 78%–88% of Cambridge
students came from elite families, whereas during the 13–19th centuries, only 50%–65% of the highest degree
holders (Jinshi) in the Chinese civil service exam system came from elite families; Clark (2014, p. 86) shows
that the surname-approach estimate of the intergenerational correlation of elite status for England during
1380–1858 is about 0.81–0.85, whereas Hao and Clark (2012) show that the estimate from the same approach
for Zhejiang and Jiangsu in China during 1645–1810 is about 0.81–0.89. These results suggest that during the
studied periods, the de facto social mobility in China was comparable to that in England, if not significantly

14
 Second, we emphasize in our framework the de jure difference in the access to elite
status. It is important to note that the de jure access to elite status alone can shape the belief
in society about the de facto access, affecting the stability of the autocratic rule. For exam-
ple, Bai and Jia (2016) show empirically that China’s abolition of the civil service exam in
1905 caused an increase in revolutionary activities against the Qing court, contributing to
the end in 1912 of not only the Qing dynasty but also the imperial era. One interpretation
for such evidence is that the People’s belief in the alliance with the Ruler was temporarily
broken when the abolition of the civil service exam shut down the main de jure access of
the commoners to elite status.

Inequality in land ownership. Circumstances on land ownership inequality are also sug-
gestive. In Imperial China, peasants “were mostly free” (Finer, 1997a, p. 205), “land-
owning peasantry had been the main agent and form of agricultural production,” and they
“had mostly enjoyed the freedom of choice” (Chao and Chen, 1982, p. 192–193).16 In
contrast, in early-Medieval Europe, mostly between the 8th and 10th centuries, small peas-
ants became gradually expropriated by rich aristocrats as well as by the Church, making
peasants gradually fall entirely under the control of landlords. This happened in many
ways, as documented by Wickham (2009): First, in the aftermath of the Viking incursions,
some landlords became richer and acquired more land, usually from poor peasants, either
through payment or expropriation. Tenant peasants faced higher rents and greater control
over their labor. They became gradually submitted to the judicial control of landlords and
completely lost their freedoms to become feudal serfs. The only escape route for encaged
peasants was to flee to the cities, a process that accelerated with the Black Death, but
those living in the countryside remained heavily under the control of landlords until much
later on.17 In the 17th century in England, around 70% of the land was still owned by
landlords and gentry (Beckett, 1984). Almost all scholars on China would agree that the
corresponding number remained below 45% from the 6th century to modern China (e.g.,
Esherick, 1981; Chao and Chen, 1982).18 Even during the Tang dynasty when the aristo-
cratic families still had considerable political influence, they “did not maintain large landed
estates over multiple generations” (Tackett, 2014, p. 12).
higher.
16 See von Glahn (2016, ch. 6, 8) for a similar observation from the mid-late Tang dynasty on.
17 It is important to note that the stronger property rights of land in Europe documented by historians in

reality concern mainly whether the rights of landlords were independent of the arbitrary will of the Ruler, not
whether small peasants enjoyed certain rights in their normal, everyday life.
18 For extensive discussion on the many works on England and China, see Zhang (2017).

15
Inheritance rule. The differences in land ownership concentration are partly related to
differences in inheritance rules. China gradually switched from primogeniture to partible
inheritance in the Qin and Han dynasties (221 BC–220), while primogeniture became more
common in Europe during the Middle Ages (Goody, Thirsk and Thompson, 1976; Bertoc-
chi, 2006; von Glahn, 2016, ch. 2, 8). The consequence of these rules on elite privilege is
intuitive: partible inheritance makes it more difficult for elite families to accumulate assets
over generations. As Goldstone (1991, p. 380) observed, in China, “land was generally
divided among heirs, and over a few generations such division could easily diminish the
land holdings of gentry families. At the same time, peasants, who could purchase clear and
full title to their lands, might expand their holdings through good luck or hard work. Thus
the difference between the gentry and the peasantry was not landholding per se, but rather
the cultivation, prestige, and influence that came from success in the imperial exams.”

Formalization in our model. Motivated by these narratives, we capture the relative pow-
er of the Elites and the People by a simple parameter β . With the surplus of size π men-
tioned above, the Elites will get a and the People will get β a, if they have been loyal to
the Ruler, where 0 ≤ β ≤ 1 and a higher β indicates a more symmetric Elites–People rela-
tionship. This approach allows us to avoid modeling the exact mechanism of each specific
institution, for example, the hereditary versus non-hereditary access to elite status, land
ownership, freedom of peasants, and the inheritance rule; we can instead focus on their
general implications on the Elite–People relationship in the power structure.

Remarks. To be sure, both China and Europe experienced changes and challenges of
the power structure over the centuries. It should not be surprising that multiple Rulers in
Europe, especially during the early modern period, attempted to make the Elite–People
relationship more balanced. Nevertheless, the weaker de facto power of the Ruler and the
multiple checks on executive power by the Elites in Europe generally made it less possible
for the Ruler to consistently succeed in these kinds of endeavors.19 In Online Appendix D,
we show that our main model can be extended to accommodate this interpretation, where
we allow the current political stability to affect the future power structure. In Section 4,
helped by our theory, we discuss further the rise of cities in Medieval Europe, another
19 For example, Louis XIV insisted on depriving the nobility of actual power after the rebellions of the
Fronde, attempted to choose ministers and officials on merit, and used commoners to replace aristocrats. Even
though he succeeded temporarily, access to nobility through a judiciary and administrative office became
practically barred in 18th-century France.

16
phenomenon related to the Ruler’s hope to enlist the People as allies against the Elites by
granting more power and rights to urban commoners.

3 Comparative Institutional Analysis

Now we introduce the setting of our model. We assume that there is a Ruler (R), who
prefers a certain status quo of autocratic rule. The nature of the status quo is open to
interpretation: for example, it can be a peaceful, unified autocratic rule across the territory.
There is also a Challenger (C), who is unhappy about the status quo and can challenge
it. She could be one or a group of nobles, lords, or bureaucrats, or some common people
who are under R’s rule, or a foreign threat who is not under R’s rule; her challenge may or
may not seek to dethrone R or be violent. With such flexibility in interpretation, the model
is sufficiently general to accommodate different types of threats to autocratic rule, such
as external conflicts, elite revolts, coups, or secessions, popular uprisings, independence
wars, and other apparently non-violent attempts to alter the status quo, with or without a
competing claim over the ruling position.
Besides R and C, we assume that there are also the Elites (E), which represents the
nobles, lords, and bureaucrats, and the People (P), which includes peasants and urban com-
moners in the model. When interpreting E and P, depending on the identity of C, we
exclude the initial challenger from E and P. For example, if C were a group of elites, then
E would be the other elites; if C were a group of members of the people, then P would be
the other members of the people. We interpret E and P’s actions as whether all significant
members of each estate actively side with and fully support R to preserve the status quo
or not, focusing on the alliance across R, C, E, and P. Naturally, unanimous actions were
rare in reality both within E and within P; the model can be easily extended to analyze the
collective action problem within each estate.
As we have included both E and P in the model, there can be little doubt that, given
their significant political, economic, and military resources, the Elites’ position had a great
impact on autocratic stability in history, whereas concerns may arise about whether the
common People were relevant, especially in Medieval Europe. In response to the concerns,
we first provide in Online Appendix F historical examples where the People’s position was
critical in determining the outcome of a conflict, an important type of threat to the stability
of the Ruler’s rule in both Europe and China. Second, our model will be able to explain as
an equilibrium outcome the fact that in Europe the autocratic stability looked largely reliant

17
on the Elites but not the People: later analysis in Section 3.1.2 suggests that if the Elite–
People relationship is extremely asymmetric, as in Medieval Europe, then the People will
almost never actively support the Ruler when called upon, making their action seemingly
irrelevant and the Elites’ position apparently decisive to the status quo. Finally, one may
also note that even if we did not observe any significant move of the People in reality, it does
not suggest that the People were irrelevant; on the contrary, they may have been influential
on the off-equilibrium path, which we could not observe but may have been instrumental
in supporting the observed outcome as an equilibrium.
The four players in our model play a game of two stages. Stage 2 is about the stability
of the status quo of autocratic rule, where C, E, and P play a subgame while taking as
given the power structure. Stage 1 is about R’s design of the power structure. For reasons
discussed in Section 1, we assume that at this stage R chooses the degree of asymmetry
between E and P in terms of their power and rights, while foreseeing Stage 2 and taking
as given the level of his absolute power. Across the two stages, we assume that all players
maximize their own expected payoff. Given the two-stage structure, we now introduce in
detail and analyze Stage 2, and then move back to Stage 1.

3.1 Stage 2: Stability of Autocratic Rule

3.1.1 Setting

Figure 1 presents the setting of Stage 2. Nature (N) first randomly draws a state of the
world x ≥ 0, following the exogenous cumulative distribution function F(x). The state of
the world x will appear later in the game as the cost born by P if she sides with R.
Given x, C will decide whether to challenge the status quo, which is maintained by the
rule of R. If C does not challenge, then C will get her default payoff 0; E will get her status
quo payoff a > 0, which is exogenous; P will get β a, where β ∈ [0, 1] measures the power
symmetry between E and P in the status quo and is exogenous at this stage; R will get the
exogenous total surplus π net of the sum of E and P’s status quo payoffs (1 + β )a, which
is π − (1 + β )a in total. Stage 2 then ends there.
If C instead does challenge, then E will decide whether to side with R. If E sides with
R, then the status quo will survive. Stage 2 will end there with R, E, and P all getting their
status quo payoffs, respectively, while the failed challenge will incur an exogenous loss
y > 0 to C, leaving her the payoff −y.
If E instead does not side with R, then it will be P’s turn to decide whether to side with

18
 Nature (N) .
Draws state of the world
x per c.d.f. F(x)
Challenger (C)

Does not challenge Challenges the status quo, which

the status quo is maintained by Ruler (R)

Elites (E)
R gets π − (1 + β )a
E gets a Sides Does not side
P gets β a with R with R
C gets 0
People (P)
R gets π − (1 + β )a
E gets a Sides Does not side
P gets β a with R with R
C gets −y

N
R gets π − (1 + β )a
E gets γ a W.p. 1 − p, W.p. p,
P gets β a − x status quo status quo
C gets −y ends survives

R gets r R gets π − (1 + β )a
E gets a + w E gets γ a
P gets β a P gets γβ a
C gets z C gets −y

x ≥ 0, a > 0, π − 2a > r, 0 ≤ β ≤ 1, 0 ≤ γ ≤ 1, 0 < p < 1, w > 0, y > 0, z > 0

Figure 1: Stage 2: Stability of Autocratic Rule

R. If P decides to side with R, then the state of the world x comes in as the cost incurring to
P for the choice, while the status quo will survive. In this scenario, C will still get −y for
the failed challenge; R will still get his status quo payoff π − (1 + β )a; P will get her status
quo payoff β a but net of the cost x, which is β a − x in total; E will now suffer a punishment
because she has not sided with R, getting only γ a instead of her status quo payoff a, where
γ ∈ [0, 1] is exogenous. A lower γ measures a stronger absolute power of R to punish its
subjects who have defied him. For simplicity, we assume that the destroyed part of E’s

19
status quo payoff, (1 − γ )a, evaporates and is not going to R; assuming otherwise would
complicate Stage 1 with few additional insights. Stage 2 then ends there.
If P does not side with R either, then R will be left on his own. N will then determine
randomly whether the status quo will survive. With exogenous probability p ∈ (0, 1), the
status quo will survive, so C will still get −y for the failed challenge; R will still get his
status quo payoff π − (1 + β )a; E will be punished, getting γ a; P will be punished, too,
getting γβ a; as above, we still assume that the destroyed parts (1 − γ )a and (1 − γ )β a
evaporate and are not going to R. Stage 2 then ends there.
With probability 1 − p, the status quo will end, leaving C with an exogenous prize z > 0
and R an exogenous reservation payoff r, where we assume, intuitively, π − 2a > r so that,
given any β ∈ [0, 1], R would prefer the status quo to survive. P will still get her status quo
payoff β a, while E will now get an exogenous incentive w > 0 for having not sided with R,
in addition to her status quo payoff a, so her total payoff will be a + w. Stage 2 then ends
there.
About the random elements, we assume that N’s draws of x and whether the status quo
will survive on R’s own are mutually independent. About the informational environment,
we assume that in Stage 2 there is complete and perfect information. We will thus use
backward induction to solve for subgame perfect equilibria.
For simplicity, we assume that E and P will side with R if indifferent, respectively, and
C will not challenge if indifferent, ruling out mixed strategies. Online Appendix A shows
that the insights from our results would remain robust if mixed strategies were allowed.
Before analyzing Stage 2, we make a few remarks on the conceptual and technical
issues around the current setting.

Remarks. First, as discussed in Section 2, we interpret Imperial China as having a high β

and a low γ , while Premodern Europe had a low β and a high γ . The β –γ characterization
of the power structure captures the idea that power and rights are specific to estates and
scenarios, as β measures the E–P asymmetry and γ measures how much the power and
rights of the ruled depend on whether they have defied R.
Second, in the current setting, we have assumed that C, E, and P move sequentially.
As we will show, this has the advantage of simplicity when we highlight the political al-
liance channel through which the power structure affect E and C’s equilibrium strategies
by affecting P’s equilibrium strategy. The political alliance channel always exists, unless P
has moved strictly earlier than both C and E, which is unrealistic because, naturally, C the

20
Challenger must be among the first to move; any other sequence of moves, for example,
C, E, and P moving simultaneously or E and P moving simultaneously after C, would not
affect the insights of our analysis.
Third, about the specification of the payoffs, an alternative approach to model the E–P
( )
relationship is to assume that E and P’s status quo payoffs are 1 − β ′ a′ and β ′ a′ , re-
spectively, where β ′ ∈ [0, 1/2] measures the E–P symmetry and a′ > 0 measures the sum
of their status quo payoffs, instead of a and β a, respectively, as in our current approach.
Comparing the two approaches, first, as shown in Proposition 1 below, C and E will follow
P’s strategy in equilibrium in Stage 2, and all further results depend only on how γ and β
or β ′ would affect P’s best strategy in the equilibrium. Since P’s status quo payoffs have
the same form in the two approaches, i.e., either β a or β ′ a′ , the two approaches will thus
derive the same theoretical results. That said, as shown in Section 3.2 below, the current
approach will create a political–economic trade-off for R in Stage 1, making R’s problem
non-trivial. This is achieved without the help of any additional modeling device that would
be necessary if the alternative approach were adopted. In light of these considerations, we
opt for our current approach.20
Fourth, as mentioned, C can be an outsider or an elite member or part of the people;
the incentive for E not to side with R also depends on the specific context.21 Thus, for
generality and simplicity, we model any incentives of C and E that are additional to the
power structure via the exogenous variables w, y, and z that are added to C and E’s payoffs.
On the robustness of our analysis to this approach, first, modeling these incentives as mul-
tiplicative terms would not affect our results, since Proposition 1 below will suggest that in
the focal equilibrium, these additional incentives are irrelevant at the margin. Second, one
may suggest that these additional incentives can still be endogenous to the power structure,
and the potential endogeneity may depend on whether R will be replaced after a successful
challenge, and also on C and E’s identities.22 In light of this, in Online Appendix E, we
endogenize these additional incentives by collapsing C and E into a single player E under
20 On the empirical side, there is little historical evidence comparing the Elites’ power and rights between
Imperial China and Premodern Europe. This has already made it difficult for us to generate empirical impli-
cations related to a or a′ . The consequence of the different empirical implications from the two approaches
is thus limited, too.
21 For example, E could hope to replace R in the challenge, or simply to get more power, rights, or other

economic interests, or even to secede from the Ruler, without necessarily taking the ruling position; similarly,
C could hope to replace R, or to secede from R, or simply to loot a great fortune in the challenge.
22 For example, if C or E is to replace R after a successful challenge, then w or z will be endogenous to

the power structure; if C is a lord or provincial governor under R’s rule, then y will depend on the power
structure.

21
the autocratic rule, making her look forward infinitely in a Markov game, and allowing her
to replace R if her challenge succeeds. We show parallel results in Online Appendix E to
all results in the main text.
Fifth, P’s incentive not to side with R depends also on the specific context, for example,
P’s level and prospect of income, R’s level of legitimacy, whether and how severely R is
in a crisis, and whether P has an opportunity to revolt, all of which can be affected in
turn by many random factors. We thus model this random component of her incentive
that is additional to the power structure as a single, exogenously drawn, state-of-the-world
variable, i.e., the random cost x added to P’s payoff when he sides with R. Modeling it
alternatively as a reward for not siding with R would not affect our analysis.
Finally, one may propose two different types of commitment problems to be involved
within this stage. The first type concerns the credibility of the payoffs specified at the five
ending nodes. On this issue, we consider the power structure as a social contract that is,
once settled at Stage 1, difficult to break at Stage 2. As the specified payoffs are based
on the settle power structure, we assume away commitment problems about these payoffs
from this stage. That said, we discuss in Section 3.3 the implications if the power structure
can be changed between two repeatedly played Stages 2.
The other type concerns the credibility of any contract that R, C, E, and P could write
among themselves at Stage 2, taking the power structure as given. We understand that this
type of commitment problems can be severe: any threat R or C can exert upon E and P
depends on the status quo’s own survival or the success of C’s challenge, respectively, and
any reward R or C can promise to E and P is not too credible, since the need for cooperation
will disappear once the status quo survives or C’s challenge succeeds, respectively (e.g.,
Egorov and Sonin, 2011). Given this understanding, we have chosen not to focus on the
possibility of contracting among R, C, E, and P at Stage 2. That said, by Proposition 2
below, one can interpret R choosing a higher β at Stage 1 as an implicit contract between R
and P where R grants more everyday power and rights to P in exchange for support; when
players are bargaining over other potential contracts, the power structure can also serve as
the basis of their bargaining power. Meanwhile, the severity of this type of commitment
problems may be endogenous to the power structure. A more explicit exploration on the
contracting across R, C, E, and P could be interesting for future research.

22
3.1.2 Equilibrium Characterization

We start the backward induction from P’s strategy. In any subgame perfect equilibrium, P
will side with R if and only if

β a − x ≥ (1 − p) · β a + p · γβ a, (1)

i.e., the cost of siding with R is not greater than the probability-adjusted punishment for
not siding with R in case that C’s challenge fails:

x ≤ p · (1 − γ )β a ≡ x̂. (2)

As mentioned when introducing the players of the model, one may note here that if
the power structure has an extremely asymmetric relationship between E and P, i.e. if β is
close to zero, then the critical threshold x̂ will be extremely low, i.e., in equilibrium P will
almost never actively help R out when called upon, making R largely reliant on E. P may
thus look irrelevant to the fate of the status quo, but E must still consider P’s strategy when
solving for his own best strategy.
Now consider E’s best strategy while expecting P’s strategy in equilibrium, i.e., to side
with R if and only if x ≤ x̂. When x ≤ x̂, P would side with R, so E will side with R; when
x > x̂, P would not side with R, so E will not side with R if and only if

a < (1 − p) · (a + w) + p · γ a, (3)

i.e., the incentive for not siding with R is greater than the probability-adjusted punishment
in case C’s challenge fails:
p
w> · (1 − γ )a. (4)
1− p
This analysis implies that if this condition does not hold, then in any subgame perfect
equilibrium, E will always side with R so that it will be impossible for the status quo to end.
Such equilibria are empirically irrelevant, as in reality the chance for the status quo to end
was always strictly positive; such equilibria are also theoretically trivial, in the sense that
E and P will always side with R regardless of the state of the world. Therefore, to narrow
our focus onto empirically more relevant and theoretically less trivial scenarios, we assume
w > a · p/(1 − p) so that for any γ ∈ [0, 1], in any subgame perfect equilibrium, E will not
side with R if and only if x > x̂.

23
 Under this assumption, consider now C’s strategy while expecting these strategies of E
and P in equilibrium. When x ≤ x̂, E would side with R, so C will not challenge the status
quo; when x > x̂, E and P would not side with R, so C will challenge the status quo if and
only if
0 < (1 − p)z − py, (5)

i.e., the prize from a successful challenge is greater than the probability-adjusted loss from
a failed challenge:
p
z> · y. (6)
1− p
This analysis implies that if this condition does not hold, then in any subgame perfect
equilibrium, C will never challenge the status quo. Similar to the discussion above, such
equilibria are empirically irrelevant and theoretically trivial. Therefore, to further narrow
our focus onto empirically more relevant and theoretically less trivial scenarios, we further
assume z > y · p/(1 − p) so that in any subgame perfect equilibrium, C will challenge the
status quo if and only if x > x̂.
Note that under the two assumptions we have introduced, we have found the unique
strategy of each player in any subgame perfect equilibrium, so these strategies constitute a
unique subgame perfect equilibrium. To summarize:

Proposition 1. If w > a · p/(1 − p) and z > y · p/(1 − p), then for any β ∈ [0, 1] and
γ ∈ [0, 1], there exists a unique subgame perfect equilibrium at Stage 2, in which C will
challenge the status quo if and only if x > x̂, E will not side with R if and only if x > x̂, and
P will not side with R if and only if x > x̂, where x̂ ≡ p · (1 − γ )β a.

This equilibrium is indeed theoretically non-trivial, since in the equilibrium, whether C

will challenge the status quo and start a challenge and whether E and P will side with R all
depend on the state of the world; this equilibrium is also empirically relevant, since in the
equilibrium, a challenge of the status quo can happen and E and P may not side with R, i.e.,
the probability of challenge 1 − F(x̂) can be strictly positive and the survival probability of
the status quo
( )
S = 1 − 1 − F(x̂) · (1 − p) (7)

can be strictly lower than one. Therefore, to focus on this equilibrium, from now on we
assume that the condition in Proposition 1 holds, i.e., w > a· p/(1− p) and z > y· p/(1− p).

24
3.1.3 Impact of Power Structure on the Stability of Autocratic Rule

How does the β –γ power structure shape the probability of challenge and the survival
probability of the status quo in equilibrium?

Proposition 2. At Stage 2, a higher β and a lower γ decrease the probability of challenge

and increase the survival probability of the status quo of autocratic rule in equilibrium.

Proof. By Proposition 1, the probability of challenge is 1 − F(x̂) and the survival proba-
( )
bility of the status quo is S = 1 − 1 − F(x̂) · (1 − p), so a higher x̂ lowers 1 − F(x̂) and
raises S. Since a higher β and a lower γ increase x̂ ≡ p · (1 − γ )β a, the proposition then
follows.

The intuition of Proposition 2 deserves more discussion. In the model, β and γ influence
the stability of the status quo in equilibrium by their impacts on P, E, and C’s equilibrium
strategies. We discuss each of these impacts. First, the impacts of β and γ on P’s strategy
in equilibrium are straightforward: by Equation (2), P’s strategy hinges on the comparison
between her cost x for siding with R and the probability-adjusted punishment x̂ ≡ p(1 −
γ )β a for not siding with R in case C’s challenge fails; both a higher β and a lower γ impose
a heavier punishment (1 − γ )β a, making P more willing to side with R in equilibrium. We
can say that these impacts work through a generic, punishment channel.
Second, the impact of γ on E’s strategy in equilibrium generally has two channels. The
first is again the punishment channel: a lower γ imposes a heavier punishment (1 − γ )a
on E in case C’s challenge fails, making E more willing to side with R given any strategy
of P, including the one in equilibrium. The second, which is new, is a strategic, political
alliance channel: a lower γ makes P more willing to side with R in equilibrium, lowering
the chance for C’s challenge to succeed and, therefore, making E more willing to side with
R in the first place.23 Therefore, through both channels, a lower γ makes E more willing to
side with R in equilibrium.
In the specific case of Proposition 2, under the condition w > a · p/(1 − p), E always
prefers “both herself and P not siding with R” to “herself siding with R”, and further to
“herself not siding with R while P siding with R.” Meanwhile, P will always either side
23 To see the point, observe that when deciding whether to side with R, E compares the payoff of doing
( )
so,
( i.e., a, versus the payoff
) of not doing so, i.e., P[P sides with R|x, γ ] · γ a + 1 − P[P sides with R|x, γ ] ·
(1 − p) · (a + w) + p · γ a , where P’s strategy is represented by P[P sides with R|x, γ ]. There are two chan-
nels via which γ can influence this comparison: first, γ can affect γ a in the payoff of siding with R, which is
the punishment channel; second, γ can affect P[P sides with R|x, γ ], which is the political alliance channel.

25
with or not side with R, and her decision solely depends on x, so E does not face strategic
uncertainty about P. Therefore, a heavier punishment upon E brought by a lower γ would
not change the fact that E’s best response to P’s strategy in equilibrium is to “follow” P’s
strategy, i.e., to switch between to side or not to side with R at x = x̂. Therefore, the
punishment channel is muted and we observe only the political alliance channel.24
Finally, the impact of β on E’s strategy and the impacts of β and γ on C’s strategy in
equilibrium have only the political alliance channel: β does not affect E’s payoffs at any
of the five ending nodes of the game, and β and γ do not affect C’s payoffs at these nodes,
either, but a higher β makes E more willing to side with R by making P more willing to
side with R in equilibrium, whereas a higher β and a lower γ make C more reluctant to
challenge by making P and E more willing to side with R in equilibrium.
To summarize, Proposition 2 reveals that both a higher β and a lower γ will make P
more willing to side with R, thus E more willing to side with R, and, therefore, C more
reluctant to challenge the status quo in the first place. The probability of challenge is
then lowered and the status quo becomes more stable. In our specific setting, a generic
punishment channel appears in β and γ ’s impacts on P’s strategy; it exists in γ ’s impact
on E’s strategy but is muted, with only a strategic political alliance channel visible; in β ’s
impact on E’s strategy and β and γ ’s impacts on C’s strategy, only the political alliance
channel exists. All these make the impacts of β and γ on political stability come from only
their impacts on P’s switching threshold x̂, providing much simplicity for the result.
Proposition 2 thus highlights that how well R can form an alliance with P is critical in
shaping the stability of autocratic rule.25 This proves crucial in R’s design of the power
structure at Stage 1, which comes below. Also, by Proposition 2, compared with Europe,
both a higher β and a lower γ make an autocratic rule more stable in China. We will come
back to this implication in Section 4.
24 If E faced strategic uncertainty about P, the punishment channel would not be muted. For example, ∫
suppose E did not observe x when deciding
∫∞( ) whether to side with R. She would then compare a versus 0x̂ γ a ·
dF(x) + x̂ (1 − p) · (a + w) + p · γ a · dF(x). As a lower γ will strictly lower the latter sum by lowering γ a,
its impact on E’s decision via the punishment channel would be visible.
25 Chapter 17 in Han Feizi argues that “too much compulsory labor service” upon the People (low β ) would

make it easy for the Elites to shelter the People in exchange for their financial and political support against
the Ruler (low x̂), damaging the Ruler’s “long lasting benefit” (low S, Watson, 1964, p. 87). This argument
follows exactly the modeled impact of β on the stability of autocratic rule via the political alliance channel
in this analysis and Online Appendix E.

26
3.2 Stage 1: Design of Power Structure
3.2.1 Setting

This stage characterizes how R’s incentive to promote the symmetry between E and P can
depend on the level of his absolute power. We assume that R at this stage simply chooses
β , while foreseeing the equilibrium at Stage 2 and taking γ as given. R’s program is thus

max V R ≡ (π − (1 + β )a) · S + r · (1 − S), subject to (8)

β

( )
0 ≤ β ≤ 1, S = 1 − 1 − F(x̂) · (1 − p), x̂ = p · (1 − γ )β a, (9)

where V R is R’s expected payoff from Stage 2. Without losing generality, we also assume
that the state of the world x’s probability density function is always strictly positive while
finite in the relevant range, i.e., satisfies f (x) ∈ [ f , f¯] ⊂ (0, ∞) over x ∈ [0, pa].

3.2.2 Institutional Compatibility

How will R choose β by the program? There is a political–economic trade-off: by Proposi-

tion 2, a higher β will increase the survival probability S of the status quo at Stage 2, which
is political; at the same time, it will decrease the status quo payoff π − (1 + β )a at Stage 2,
which is economic.
The economic side of the trade-off is straightforward: a higher β will decrease the
status quo payoff at a marginal rate of a. The political side is less so, as it depends on
the impact of β on the survival probability, i.e., dS/d β . Intuitively, this impact is largely
governed by γ : a higher γ suggests that P will not lose much of her status quo payoff after
she has not sided with R and C’s challenge has failed, so any additional status quo payoff
would not make her much more loyal to R and, therefore, it will not make E much more
loyal toward R, and neither would C be much more reluctant to challenge.
The key assumption that leads to this intuition is that the punishment upon P, i.e., (1 −
γ )β a, is multiplicative between 1 − γ and β . We find this assumption uncontroversial, since
in reality, given the punishing institution against defying behaviors, the ones who own more
would often be more concerned about losing it.
We can formalize this intuition by showing that the impact of β on the survival prob-
ability of the status quo can be approximated by two positive and increasing functions of
1 − γ:

27
Lemma 1 (Impact of β on stability governed by γ ). There exist c ≡ (1 − p)p f > 0 and
c̄ ≡ (1 − p)p f¯ > c such that

dS
ca · (1 − γ ) ≤ ≤ c̄a · (1 − γ ). (10)
dβ

Proof. By Proposition 1, the marginal impact of β on S is

dS dF(x̂)
= (1 − p) · = (1 − p)p f (x̂) · a · (1 − γ ), (11)
dβ dβ

where x̂ ≡ (1 − γ )β p · a ∈ [0, pa]. By f (x) ∈ [ f , f¯] over x ∈ [0, pa], the lemma follows.

The proof of Lemma 1 also suggests that the approximation would be exact if and only
if the state of the world x followed a uniform distribution, i.e., f (x̂) is a constant. Such
an assumption could be arbitrary. Therefore, our approximating result captures the most
robust part of the intuition.
Lemma 1 suggests that R’s trade-off around β is largely governed by γ , too:
/
Proposition 3 (Institutional compatibility). At Stage 1, if γ < γ ≡ 1 − 1 (π − 2a − r)c,
/
then R will prefer β to be as high as possible, i.e., β ∗ = 1; if γ > γ̄ ≡ 1 − p (π − a − r)c̄,
then R will prefer β to be as low as possible, i.e., β ∗ = 0, where γ < γ̄ < 1. Further, if
π > 2a + r + 1/c, then γ > 0.

Proof. The marginal impact of β on R’s expected payoff in equilibrium at Stage 2 is

dV R ( ) dS
= π − (1 + β )a − r · − aS. (12)
dβ dβ

By Lemma 1, β ∈ [0, 1], and S ∈ [p, 1], we have

dV R ( )
≥ π − (1 + β )a − r · ca · (1 − γ ) − aS
dβ
( )
≥ (π − 2a − r) · c · (1 − γ ) − 1 · a, (13)

so if
(π − 2a − r) · c · (1 − γ ) − 1 > 0, (14)

i.e.,
1
γ < 1− ≡ γ, (15)
(π − 2a − r) · c

28
then dV R /d β > 0. At the same time, we have

dV R ( )
≤ π − (1 + β )a − r · c̄a · (1 − γ ) − aS
dβ
( )
≤ (π − a − r) · c̄ · (1 − γ ) − p · a, (16)

so if
(π − a − r) · c̄ · (1 − γ ) − p < 0, (17)

i.e.,
p
γ > 1− ≡ γ̄ , (18)
(π − a − r) · c̄
then dV R /d β < 0. Finally, note γ < γ̄ < 1, and γ > 0 is equivalent to π > 2a + r + 1/c.
The proposition is then proven.

The intuition of Proposition 3 is as follows. When γ is sufficiently low, a higher β

will increase the punishment P will face in case C’s challenge fails, so the increase in the
stability at Stage 2 will be significant; therefore, the political side of R’s trade-off at Stage
1 will always be dominant; R then prefers the highest possible β . If γ is sufficiently high,
the opposite will happen, and R will prefer the lowest possible β .
Proposition 3 explains the institutional compatibility within the power structure across
Imperial China and Premodern Europe: as in European history, a high γ , which represents
a weak absolute power of the Ruler, and a low β , which represents a highly unbalanced
relationship between the Elites and People’s rights and power, are compatible, while as in
Chinese history, a low γ and a high β are compatible.
One may wonder why we did not show a result for γ ∈ [γ , γ̄ ]. It is not straightforward
to derive such a result without further restrictions on the distribution of x. To see this point,
observe that

dV R ( ) dS dS
= π − (1 + β )a − r · − aS and = (1 − p)p f (x̂) · a · (1 − γ ). (19)
dβ dβ dβ

A lower γ increases S, 1 − γ , and x̂, but its impact on f (x̂) depends on properties of f (·).
Therefore, any unambiguous result about the impact of γ ∈ [γ , γ̄ ] on R’s preference over
β would rely on further restrictions on the distribution of x, which would have to be more
or less arbitrary. As an example, Online Appendix B derives a result that R will generally
prefer a higher β given a lower γ with an additional restriction on the distribution of x. For

29
theoretical robustness, Proposition 3 only touches upon the extreme cases and, therefore,
the first-order implications of γ . In light of all these, we provide a numerical example in
Figure 2, where we plot R’s choice β ∗ against γ : consistent with Proposition 3, β ∗ = 1 if
γ < γ , while β ∗ = 0 if γ > γ̄ ; silent in Proposition 3, given the specification of the example,
β ∗ weakly decreases with γ over γ ∈ [γ , γ̄ ].

Specification: F(x) = 1 − e−x , p = 0.8, π = 20, a = 0.6, r = 5. Under this specification, π − 2a > r.
The Ruler’s expected payoff in equilibrium at Stage 2 is denoted as V R . The blue, solid line plots
β ∗ when γ ∈ [0, γ ) ∪ (γ̄ , 1], which is consistent with Proposition 3. The red, dashed line plots β ∗
when γ ∈ [γ , γ̄ ], about which Proposition 3 is silent.

Figure 2: Ruler’s choice β ∗ a function of γ ∈ [0, 1]

3.3 Extensions: Endogenizing the Absolute Power of the Ruler

In the analysis above we have taken the level of the absolute power of the Ruler γ as
exogenous. Here we introduce two examples of extensions in which we endogenize γ and
derive additional implications.

People’s perspective on the Ruler’s absolute power. One may argue that γ would even-
tually depend on the legitimacy that P has granted to R in the first place. Along this argu-
ment, if before Stage 1 P has an opportunity to choose γ , how would her preference of γ
look like?

30
Corollary 1. If P could choose γ before Stage 1, then P would prefer any γ < γ over any
γ > γ̄ .

Proof. Given the β –γ power structure, P’s expected payoff at Stage 2 is

( ) ( ( ) )
V = γβ a · 1 − F(x̂) · p + β a · 1 − 1 − F(x̂) · p
P
( ( ) )
= 1 − 1 − F(x̂) · p · (1 − γ ) · β a. (20)

By Proposition 3, if γ > γ̄ , R will choose β = 0; if γ < γ , R will choose β = 1. Therefore,

 
V P γ <γ ,β =1 > 0 = V P γ >γ̄ ,β =0 . (21)

The corollary is then proven.

The intuition is as follows. On the equilibrium path at Stage 2, P will never side with R
when called upon. Therefore, she will receive either her status quo payoff β a or her post-
punishment payoff γβ a. Given a sufficiently high γ > γ̄ , R will prefer the lowest possible
β = 0 at Stage 1, so P will receive exactly a zero payoff eventually; any sufficiently low
γ < γ will induce R to choose β = 1, granting P a strictly positive payoff eventually. P will
then prefer any sufficiently low γ < γ over the sufficiently high γ > γ̄ before Stage 1.
To clarify, we focus on the extreme case to highlight that it is not always the case that P
will prefer a high to a low γ ; instead, P may tolerate a quite absolutist R. We will come back
to this insight in Section 4 when discussing the bureaucracy with the civil service exam in
China.

Allowing current stability to shape future power structure. One may also argue that
the European Rulers might have wanted to raise β but were not able to do so. Online
Appendix D provides a response to this argument in several steps. First, it is easy to see
that before Stage 2, R will prefer γ to be as low as possible, since a lower γ stabilizes the
autocratic rule without sacrificing the status quo payoff, as seen in Equation (8).
Second, Proposition 2 implies that, if the total surplus π is sufficiently big, then the po-
litical side of R’s trade-off with respect to β will be dominant, as long as the conditionality
of the power and rights of the ruled exists, i.e., γ < 1. In that case, any R would prefer β to
be as high as possible, as in Corollary D.1 in Online Appendix D.
Third, the last two results suggest that when the total surplus is sufficiently big, any R
would like to invest in a lower γ and a higher β at the same time. Given this preference, we

31
can consider an alternative setting in which Stage 2 gets played repeatedly over different
periods and, instead of letting R choose β only once, we can justify a mechanical link from
the current stability of autocratic rule in equilibrium to the future power structure, thereby
endogenizing the absolute power of the Ruler in the future: the more stable R’s autocratic
rule is today, the more successful he would be in investing in the power structure toward
the direction that he would favor, so the higher the degree of R’s absolute power and the
more symmetric the Elite–People relationship tomorrow. This effect on the future power
structure, by Proposition 2, would eventually lead to a higher future stability, creating a
dynamic complementarity.
Finally, given this dynamic complementarity, multiple stable steady states of (β , γ , S),
i.e., the power structure and stability of autocratic rule, may exist, and, among these steady
states, the stronger the absolute power of R and the more symmetric the Elite–People rela-
tionship, the higher the stability of autocratic rule, as derived in Proposition D.2 in Online
Appendix D. A dual divergence of the power structure and stability of autocratic rule from
slightly different initial conditions can thus appear, as shown in Proposition D.3 in Online
Appendix D. Here we summarize the implication as follows:

Corollary 2. Compared with Premodern Europe, Imperial China could have been given a
slightly lower γ , a slightly higher β , or a slightly higher S at very early times. This slight
difference could have led the two societies to diverge into different stable steady states,
where compared to Europe, China had a lower γ , a higher β , and a higher S.

4 Further Discussion and Stylized Facts

4.1 Understanding Specific Institutions

Bureaucracy and civil service exam in China. Our model can help us understand spe-
cific institutions without explicitly modeling them in detail. One such example is the Chi-
nese bureaucracy with the civil service exam, the hallmark of the Chinese imperial insti-
tutions (e.g., Finer, 1997a,b). Following our model, we can read it primarily as the Ruler
raising β by generalizing the access to elite status between the Elites and People. By Propo-
sition 3, Chinese Rulers had a great incentive to do so because they enjoyed a low γ , i.e.,
a strong absolute power. This is consistent with the fact that the civil service exam was
first introduced during the Sui dynasty (581–619) and greatly developed during the Tang
dynasty (618–907), when the absolute power of the Ruler had recovered from the low level

32
during the Six Dynasties period (220–589) (Yan, 2009). Given the bureaucratic system, the
Elites became mainly bureaucrats who were appointed by the Ruler, so they became further
reliant on the Ruler for legitimacy, making their power and rights more conditional on the
Ruler’s will, i.e., further lowering γ . Not only did the Ruler favor the stability of autocratic
rule, i.e., a high S, as the result of the combination of a consolidated generalized access to
elite status and a strong absolute power, i.e., a higher β and a low γ , but also by Corollary
1, the People might have been satisfied with the power structure and the resulting stability,
without too much appetite for stronger rule of law or property rights.

Cities in Medieval Europe. In a similar vein, we can also read part of the rise of cities in
Medieval Europe in relation to the Ruler’s effort to raise β , by issuing charters that granted
certain rights to the People in cities against other local Elites. This effort could eventually
help stabilize the Ruler’s autocratic rule. For example, “Philip [II of France] knew that
in recognizing a commune, he was binding the citizens of that town to him. At critical
moments in the reign the communes . . . proved staunch military supporters. . . . From the
point of view of the communes . . . the king was their natural ally, a counter to the main
opponents of their independence, the Church or the magnates” (Bradbury, 1998, p. 236).26
By Lemma 1 and Proposition 3, however, this stabilizing effect was not guaranteed
when the Ruler’s absolute power was as weak as it was in Medieval Europe. As a European
Ruler was generally constrained by his own charters, he would find it difficult to punish the
cities by retracting the granted rights. Because of this, granting more power and rights to
cities might not help the Ruler much in creating a political alliance with urban commoners
and securing his position. In this sense, when a Ruler in Europe freed a city from its feudal
lords, he ran the risk of having freed it also from himself. Notable examples of this risk
can be found during the rise of cities and boroughs in England and the free imperial cities
in the Holy Roman Empire.27
26 Philip II’s practice followed Louis VII, who “gave encouragement to the commune movement and re-

ceived reciprocal support from the communities, at the expense of local lords” (Bradbury, 1998, p. 32).
Relatedly, on the economic consequences of cities freeing peasants from local lords, see Cox and Figueroa
(Forthcoming).
27 In England, in May 1215, facing rebelling barons, John of England chartered the right of Londoners to

elect their own mayor, together with other rights, “[i]n a last attempt to win the city” (Williams, 1963, p. 6).
This proved futile: in June, still, “discontent citizens joined the barons in enforcing the signing of Magna
Carta; the Mayor [of London] was the only commoner whose name appeared among the signatories” (Porter,
1994, p. 25–26). Magna Carta eventually extended the city rights by confirming in Article 13 that “the city
of London [and] all other cities, boroughs, towns, and ports shall have all their liberties and free customs”
(McKechnie, 1914, p. 241). Similarly, in the Holy Roman Empire, “the free towns had been winning valuable

33
 The greater danger in the stabilizing effect under a less absolutist Ruler implies that this
kind of Ruler could be more reluctant to promote the development of cities. This is con-
sistent with the observation that within Premodern Europe, the urbanization rate increased
on average in the states that were governed with a weak or no assembly, which indicat-
ed a stronger absolute power of the Ruler, whereas decreased in the long run in the states
governed with a strong assembly, which indicated a weaker absolute power of the Ruler
(Stasavage, 2020, p. 191–192). Given the uncertainty of the stabilizing effect under the
generally weak absolute power of European Rulers, together with the dual divergence of
the power structure and stability as in Corollary 2 and Online Appendix D, the European
population that enjoyed cities’ privileges was eventually relatively small at the eve of the
modern times (Cantor, 1964; de Vries, 1984, p. 76). This reflected the limited success of
the Ruler’s effort to raise β in Premodern Europe.

4.2 Comparing Stability of Autocratic Rule

Proposition 2 states that a stronger absolute power of the Ruler and a more symmetric
relationship between the Elites and People, as in Imperial China compared to Premodern
Europe, imply a higher stability of autocratic rule. As the nature of the challenge and the
status quo of autocratic rule in our model are open to flexible interpretation, Proposition 2
can generate several auxiliary predictions that we could bring to data from historical China
and Europe.

Number of wars. First, if we interpret the challenge in our model as an armed conflict,
Proposition 2 then predicts that anyone in Europe who preferred an alternative to the status
quo would be more willing to start a war than her counterpart in China. Note that this
prediction does not depend on the challenger’s identity and her status in the respective
status quo: she could be either a foreign power, a rebellious local lord or regional governor,
or a group of commoners. We also find it difficult to argue for a systematic difference in
the number of all these possibly relevant entities between China and Europe in either way.
Therefore, we should compare the total number of wars that challenged a status quo in
the Chinese society with the number for the European society, regardless of the identity
of the challengers. As defined in Section 2.1, here the “Chinese society” is the society
privileges in addition to those which they already possessed, and the wealthier among them, like Lübeck and
Augsburg, were practically imperia in imperio, waging war and making peace, and ruling their people without
any outside interference” from the Emperor (Holland, 1911, p. 342).

34
in the historical core of Imperial China, whereas the “European society” is the Romano–
Germanic influenced or assimilated society in western and central Europe where feudalism
was once pervasive.
We are not aware of systematic evidence on this subject that covers the period of our
interest. That said, Brecke (1999) provides comprehensive information on wars in Europe
from 900 onwards and in China from only 1400 onwards. We complement the data with
information from the Chinese Military History (2003) project from 900.28 We further iden-
tify for each war whether it was fought to challenge a status quo in the Chinese or European
society, respectively.
Figure 3 reports the number of wars breaking out in each given year that challenged a
status quo in the Chinese or European society. For robustness, we plot the retrospective
100-year moving-averages; when doing so, for each retrospective 100-year window, we
calculate the Olympic average in the window, i.e., taking the average in the window after
removing one of the highest and one of the lowest values in the window. Besides reporting
the result for wars of all lengths in Figure 3a, we also restrict our attention to more sig-
nificant wars that lasted longer than one year, three years, and five years, respectively, in
Figures 3b–3d.
Across Figures 3a–3d, we see the same pattern. First, Brecke (1999)’s data and our
data give comparable numbers of wars that challenged a status quo in the Chinese society
during 1400–1700, strengthening our confidence about our data. Second, the figures show
that the number of wars for Europe was consistently higher than that for China from 900
to 1700. We thus conclude that during 900–1700, there were significantly more wars chal-
lenging a status quo in Europe than those challenging a status quo in China, consistent with
Proposition 2.

Risk of deposition. Second, if we interpret the challenge in our model as to remove the
Ruler from the ruling position, Proposition 2 then predicts that a Chinese Ruler should have
faced a lower risk of deposition in each given year than a European Ruler. On the data, the
historical information of all monarchies in the world has been compiled by Morby (1989)
and some of it has been used in a few studies (e.g., Blaydes and Chaney, 2013; Kokkonen
28 The original data in the Chinese Military History (2003) project are at the level of individual battles. We
first compare the battle-level data from the Chinese Military History (2003) project with the war-level data
from Brecke (1999) to understand Brecke (1999)’s criteria of categorization. Complementing the criteria
with information from Wu (2016) and Tian (2019), we finally manually categorize the battles recorded in the
Chinese Military History (2003) project into wars.

35
 (a) All wars (b) One-year or longer wars

(c) Three-year or longer wars (d) Five-year or longer wars

Olympic average within each retrospective 100-year window. “Brecke Data: Europe” include wars
in Brecke (1999) that challenged a status quo in the European society, i.e., the Romano–Germanic
influenced or assimilated society in western and central Europe where feudalism was once per-
vasive; “Brecke Data: China” include wars in Brecke (1999) that challenged a status quo in the
Chinese society, i.e., the society in the historical core of Imperial China; “Chinese Data: China”
include wars in our Chinese war data that challenged a status quo in the Chinese society. For more
details of our Chinese war data, see Footnote 28.
Figure 3: Number of wars starting in a given year that challenged a status quo in the
Chinese or European society

and Sundell, 2014). Using the same data, to compare the risk of deposition between China
and Europe, we first calculate for each given year a measure of the risk of deposition in
that year, i.e., the share of the Rulers who were deposed in that year among all the Rulers
who had been in power in that year; we then visualize in Figure 4 the comparison between
China and Europe by plotting the retrospective 100-year moving-averages of the measure.

36
For robustness, again, the Olympic average is used.

Olympic average within each retrospective 100-year window. Following Blaydes and Chaney
(2013), “European Rulers” include all the ones who assumed power before 1500 and are under
the section “The Barbarian West” or the subsections “The British Isles,” “France,” “The Low Coun-
tries,” “Italy,” “The Iberian Peninsula,” “The German-speaking States,” “Scandinavia,” and “Cru-
sader States” under the section “Europe” in Morby (1989). “Chinese Rulers” include all the ones
under the subsection “China” under the section “The Far East” in Morby (1989).

Figure 4: Risk of deposition for a Ruler in a given year, China vs. Europe

Figure 4 first shows that the risk of deposition for a Chinese Ruler in a given year
declined rapidly from the high level during the 6th century, i.e., the late Southern and
Northern Dynasties period, to a lower level during the 7–8th centuries, i.e., the Sui dynasty
and the early and mid-Tang dynasty. As discussed in Sections 2.3 and 4.1, this decline
happened at the same time when, first, the absolute power of the Ruler first recovered
from a historical low and then was further strengthened and, second, the civil service exam
was first introduced and then greatly developed (e.g., Yan, 2009).29 These correlations are
consistent with Propositions 2 and 3.
Figure 4 further shows that during the 9–14th centuries, i.e., when the two differences
we emphasize in the power structure between China and Europe were the most prominent,
29 Yan (2009, p. 240–245) discusses the cultural and institutional elements behind the Northern dynasties-
led changes in the Chinese power structure during the 6–8th centuries.

37
the risk of deposition for a Ruler was generally lower in China than in Europe. That said, a
short period around the 10th century did exist when the risk in China appeared to be higher,
when China entered the Five Dynasties and Ten Kingdoms period (907–979). In light of
this, we conduct a Kolmogorov–Smirnov test to check whether the differences in the risks
between China and Europe during the 9–14th centuries are systematic. The test reports
that at a significance level of 0.1%, we can accept the claim that the risk of deposition for
a Ruler in a given year was generally lower in China than in Europe during the period,
whereas the opposite claim must be rejected. These results are consistent with Proposition
2.30

Resilience of unified autocratic rule. Finally, if we interpret the status quo of autocratic
rule in our model as a unified one across the territory, Proposition 2 then predicts that
a unified autocratic rule should have been more resilient in China than in Europe. As
discussed, the literature has well documented that China had been more unified than Europe
in history. Among many other measures, here we present in Figure 5 only a replication
of the comparison by Scheidel (2019, fig. 1.11) as an example, plotting the share of the
population in the continent Europe that was controlled by the largest polity in the continent,
together with the same measure for East Asia, where China is located.
As shown in the figure, since 800, in East Asia, the population share in the largest poli-
ty, which was the dominant empire in China, had usually been above 75%, except for short
subperiods of turbulence. In contrast, the number for Europe had been below 20%, con-
sistent with a more fragmented pattern. This comparison is consistent with Proposition 2.
Therefore, our model provides a power-structure approach to the unification–fragmentation
cleavage between Imperial China and Premodern Europe.
In sum, consistent with our model, we find a persistent difference in the stability of
autocratic rule, whether measured by the number of wars, the Ruler’s risk of deposition, or
the resilience or vulnerability of unified autocratic rule.31
30 One may recall that Hoffman (2015) shows that during the 16–18th centuries, major European sovereigns

were seldom deposed after losing a war. On the 7–14th centuries, when our characterization of the power
structure was more relevant, however, Eisner (2011) shows that the risk of regicide, which would surely lead
to but was not the only way to deposition, had remained high in Europe. Eisner (2011) also shows that at that
time battle death was a major risk for European rulers and being murdered with an external power involved
was also not rare.
31 To clarify, the dual divergence of the power structure and autocratic stability depicted in Corollary 2 and

Online Appendix D could have already taken place around 800. If so, we would not predict an increase in
stability differences but only the persistence of such differences.

38
Replicated from Scheidel (2019, fig. 1.11). According to Scheidel (2019, fig. 1.1), “Europe”
is defined primarily by the common geographical term, i.e., the continent Europe, whereas “East
Asia” includes nowadays China, Mongolia, North Korea, South Korea, and Japan.

Figure 5: Percentage of population claimed by the largest polity, Europe vs. East Asia

5 Conclusion
In this paper, we provide a power structure framework to reconcile a series of views on
the institutional differences between Imperial China and Europe that are seemingly con-
tradictory in the light of the literature on institutions and development. In this framework,
we read the institutional differences along two dimensions of the power structure: Chinese
Rulers had a stronger absolute power, while the relationship between the Elites and People
in terms of their power and rights was more asymmetric in Europe.
By building a model and analyzing how the power structure can shape the stability
of an autocratic rule, we show that, once we recognize that the Ruler’s absolute power
is about the conditionality of the power and rights of the ruled on the Ruler’s will, a more
symmetric Elite–People relationship will strengthen the political alliance between the Ruler
and the People, thus creating more loyalty to the Ruler, deterring potential challenges, and
stabilizing the autocratic rule. Importantly, this effect and, therefore, the Ruler’s incentive
to promote a more symmetric Elite–People relationship depend on the Ruler’s absolute

39
power. This suggests that an absolutist Ruler can be compatible with a more symmetric
Elite–People relationship. A society can thus be repressive in one institutional dimension
but inclusive in another at the same time, a new result to the literature.
This comparative institutional theory explains the coexistence of the two power struc-
ture differences between Imperial China and Premodern Europe. Besides guiding us to
understand specific institutions, our model also suggests a higher stability of autocratic
rule in Imperial China. This implication is supported by stylized facts about the number of
wars, risk of deposition, and resilience of unified autocratic rule.
Admittedly, our theory is highly stylized as we capture the power structure with only
two parameters, and we only examine the stability of autocratic rule as the outcome of the
power structure. The benefit of doing so is that we can deliver our key insights in a simple
manner. That said, our framework can be applied to understand other political, economic,
and social outcomes.
For example, on the one hand, as a result of the power structure, the too stable autocrat-
ic rule and lack of spatial competition in Imperial China may have hindered economic and
scientific innovations from happening or being adopted (e.g., Rosenthal and Wong, 2011;
Mokyr, 2016; Desmet, Greif and Parente, 2020). On the other hand, given the power struc-
ture in Premodern Europe, the profit from innovations flowed primarily to the Elites, while
the lack of pro-People institutions could not maintain a sufficiently stable social order for
sustainable growth until the early modern days (e.g., Greif and Iyigun, 2013; Greif, Iyigun
and Sasson, 2013). It could be worthwhile if the interplays between the power structure,
endogenous growth, and political and social stability are modeled explicitly.
As another example, about culture, the Chinese Legalist tradition had emphasized the
absolute power of the Ruler, while the apparent dominance of Confucianism in China had
been reflected by the institution and content of the civil service exam, a key component
of the more balanced Elite–People relationship. Our Proposition 2 thus explains why the
Chinese Rulers had promoted the Confucianism–Legalism confluence as the dominant po-
litical culture (e.g., Qin, 1998; Yan, 2004; Zhao, 2015); Proposition 3 sees Legalism as
the more fundamental side within the confluence; Corollaries 1 and 2 explain why such a
culture may have been accepted by the People and persistent over time, respectively.
There can also be more insights to gain if one applies our power structure framework to
other parts of the world or more recent periods. For example, Blaydes and Chaney (2013)
show that Christian kings in Western Europe enjoyed a higher political stability than Mus-
lim sultans during the 9–15th centuries. This difference can be explained in our framework:

40
lords in feudal Europe owned land and military forces on a regular basis, suggesting a high
status quo payoff a to the Elites, while Mamlukism in the Muslim world was designed to re-
move elite Mamluks “from the luxuries of settled life” (Blaydes and Chaney, 2013, p. 23),
suggesting a low a; one can show in our model that a higher a would increase the stability
of autocratic rule. As another example, for contemporary China, Proposition 3 is consistent
with the observation that the more absolutist the communist paramount leader is, the more
pro–People and anti-Elite his rhetoric and political initiatives are, with Mao Zedong and Xi
Jinping versus Jiang Zemin as the diametrical prototypes (e.g., Dickson, 2003; Francois,
Trebbi and Xiao, 2016; Lu, 2017; Gao, 2018; Shirk, 2018; Kositz, 2019; Li, Roland and
Xie, 2019). We thus hope that our study opens new avenues for future research.

References
Acemoğlu, K. Daron, Simon Johnson, and James A. Robinson. 2001. The colonial origins
of comparative development: An empirical investigation. American Economic Review
91: 1369–1401.
Acemoğlu, K. Daron, Simon Johnson, and James A. Robinson. 2005a. Institutions as a
fundamental cause of long-run growth. Handbook of Economic Growth 1: 385–472.
Acemoğlu, K. Daron, Simon Johnson, and James A. Robinson. 2005b. The rise of Europe:
Atlantic trade, institutional change, and economic growth. American Economic Review
95: 546–579.
Acemoğlu, K. Daron, and James A. Robinson. 2000. Why did the west extend the fran-
chise? democracy, inequality, and growth in historical perspective. Quarterly Journal of
Economics 115: 1167–1199.
Acemoğlu, K. Daron, and James A. Robinson. 2001. A theory of political transitions.
American Economic Review 91: 938–963.
Acemoğlu, K. Daron, and James A. Robinson. 2012. Why Nations Fail: The Origins of
Power, Prosperity and Poverty. New York: Crown Publishing Group.
Acemoğlu, K. Daron, and James A. Robinson. 2019. The Narrow Corridor: States, Soci-
eties, and the Fate of Liberty. New York: Penguin Books.
Acemoğlu, K. Daron, Thierry Verdier, and James A. Robinson. 2004. Kleptocracy and
divide-and-rule: A model of personal rule. Journal of the European Economic Associa-
tion 2: 162–192.
Bai, Ying, and Ruixue Jia. 2016. Elite recruitment and political stability: The impact of the
abolition of China’s civil service exam. Econometrica 84: 677–733.
Barrow, Julia S. 2015. The Clergy in the Medieval World: Secular Clerics, Their Families
and Careers in North-Western Europe c.800–c.1200. Cambrdige: Cambridge University
Press.

41
Beckett, John V. 1984. The pattern of landownership in England and Wales, 1660–1880.
Economic History Review 37: 1–22.
Bertocchi, Graziella. 2006. The law of primogeniture and the transition from landed aris-
tocracy to industrial democracy. Journal of Economic Growth 11: 43–70.
Besley, Timothy J., and Torsten Persson. 2011. Pillars of Prosperity: The Political Eco-
nomics of Development Clusters. Princeton: Princeton University Press.
Besley, Timothy J., and Torsten Persson. 2014. The causes and consequences of devel-
opment clusters: State capacity, peace, and income. Annual Review of Economics 6:
927–949.
Blaydes, Lisa, and Eric Chaney. 2013. The feudal revolution and Europe’s rise: Political
divergence of the Christian West and the Muslim world before 1500 CE. American
Political Science Review 107: 16–34.
Bloch, Marc. 1962a. Feudal Society, Volume I: The Growth of Ties of Dependence. London:
Routledge, 2nd edition. Translated by L. A. Manyon from La Société Féodale, Volume
I: La Formation des Liens de Dépendance, 1939, Paris: Éditions Albin Michel. Reprint,
2004.
Bloch, Marc. 1962b. Feudal Society, Volume II: Social Classes and Political Organization.
London: Routledge, 2nd edition. Translated by L. A. Manyon from La Société Féodale,
Volume II: Les Classes et le Gouvernement des Hommes, 1940, Paris: Éditions Albin
Michel. Reprint, 2005.
Bracton, Henry de. 1968. Bracton on the Laws and Customs of England. Cambridge:
Belknap Press. Translated with annotations by Samuel E. Thorne from Bracton De
Legibus et Consuetudinibus Angliæ, George E. Woodbine (ed.), 1915, New Haven: Yale
University Press.
Bradbury, Jim. 1998. Philip Augustus: King of France 1180–1223. London: Routledge.
Reprint, 2013.
Brecke, Peter. 1999. Violent conflicts 1400 AD to the present in different regions of the
world. Paper prepared for the 33rd Peace Science Society (International) North Ameri-
can Meeting, Ann Arbor, October 8–10, 1999.
Brook, Timothy. 2010. The Troubled Empire: China in the Yuan and Ming Dynasties.
Cambridge: Belknap Press.
Cantor, Norman F. 1964. The Civilization of the Middle Ages. New York: Harper Perennial.
Reprint, 1994.
Chao, Kang, and Chung-yi Chen. 1982. A History of China’s Land Institutions. Taipei:
Linking Publishing. Reprint, 2006, Beijing: New Star Publisher.
Chinese Military History. 2003. Chronicle of Wars in Chinese History. Beijing: People’s
Liberation Army Publishing House.
Clark, Gregory. 2014. The Son Also Rises: Surnames and the History of Social Mobility.
Princeton: Princeton University Press.
Cox, Gary W., and Valentín Figueroa. Forthcoming. Political fragmentation, rural-to-urban
migration and urban growth patterns in western Eurasia, 800–1800. European Review of
Economic History .

42
Cox, Gary W., Douglass C. North, and Barry R. Weingast. 2019. The violence trap: A
political-economic approach to the problems of development. Journal of Public Finance
and Public Choice 34: 3–19.
de Groot, Jan J. M. 1903. Sectarianism and Religious Persecution in China: A Page in the
History of Religions, Volume I. Amsterdam: Johannes Müller.
de Vries, Jan. 1984. European Urbanization, 1500–1800. Cambridge: Harvard University
Press.
Desmet, Klaus, Avner Greif, and Stephen L. Parente. 2020. Spatial competition, innova-
tion and institutions: The Industrial Revolution and the Great Divergence. Journal of
Economic Growth 25: 1–35.
Dickson, Bruce J. 2003. Whom does the Party represent? From “Three Revolutionary
Classes” to “Three Represents”. American Asian Review 21: 1.
Dillon, Michael O. 1998. Dynastic cycle. In Dillon, Michael O. (ed.) China: A Historical
and Cultural Dictionary. London: Curzon Press, 87.
Dincecco, Mark, and Yuhua Wang. 2018. Violent conflict and political development over
the long run: China versus Europe. Annual Review of Political Science 21: 341–358.
Ebrey, Patricia B., and Anne Walthall. 2013. East Asia: A Cultural, Social, and Political
History. Boston: Cengage Learning, 3rd edition.
Egorov, Georgy, and Konstantin Sonin. 2011. Dictators and their viziers: Endogenizing
the loyalty–competence trade-off. Journal of the European Economic Association 9:
903–930.
Eisner, Manuel. 2011. Killing kings: Patterns of regicide in Europe, AD 600–1800. British
Journal of Criminology 51: 556–577.
Esherick, Joseph W. 1981. A note on land distribution in prerevolutionary China. Modern
China 7: 387–411.
Fernández-Villaverde, Jesús. 2016. Magna Carta, the rule of law, and the limits on govern-
ment. International Review of Law and Economics 47: 22–28.
Fernández-Villaverde, Jesús, Mark Koyama, Youhong Lin, and Tuan-Hwee Sng. 2020. The
fractured-land hypothesis. National Bureau of Economic Research Working Paper No.
27774.
Finer, Samuel E. 1997a. The History of Government from the Earliest Times, Volume I:
Ancient Monarchies and Empires. New York: Oxford University Press.
Finer, Samuel E. 1997b. The History of Government from the Earliest Times, Volume II:
The Intermediate Ages. New York: Oxford University Press.
Francois, Patrick, Francesco Trebbi, and Kairong Xiao. 2016. Factions in nondemocracies:
Theory and evidence from the Chinese Communist Party. National Bureau of Economic
Research Working Paper 22775.
Fukuyama, Francis. 2011. The Origins of Political Order: From Prehuman Times to the
French Revolution. New York: Farrar, Straus and Giroux.
Ganshof, François-Louis. 1952. Feudalism. London: Longmans, Green and Co. Trans-
lated by Philip Grierson from Qu’est-ce que la Féodalité? 1947, Bruxelles: Office de
Publicité, S.C., 2nd edition.

43
Gao, Hua. 2018. How the Red Sun Rose: The Origin and Development of the Yan’an
Rectification Movement, 1930–1945. Hong Kong: Chinese University of Hong Kong
Press. Translated by Stacy Mosher and Jian Guo.
Gennaioli, Nicola, and Hans-Joachim Voth. 2015. State capacity and military conflict. The
Review of Economic Studies 82: 1409–1448.
Goldstone, Jack A. 1991. Revolution and Rebellion in the Early Modern World: Popula-
tion Change and State Breakdown in England, France, Turkey, and China, 1600–1850.
Berkeley: University of California Press.
Goody, Jack, Joan Thirsk, and Edward P. Thompson (eds.) . 1976. Family and Inheritance:
Rural Society in Western Europe, 1200–1800. Cambridge: Cambridge University Press.
Greif, Avner, and Murat Iyigun. 2013. What did the Old Poor Law really accomplish? A
redux. Working paper, Stanford University.
Greif, Avner, Murat Iyigun, and Diego L. Sasson. 2013. Social institutions and economic
growth: Why England rather than China became the first modern economy. Working
paper, Stanford University.
Greif, Avner, Joel Mokyr, and Guido Tabellini. Forthcoming. Two Paths to Prosperity:
Culture and Institutions in Europe and China, 1200–2000.
Greif, Avner, and Guido Tabellini. 2010. Cultural and institutional bifurcation: China and
Europe compared. American Economic Review 100: 135–140.
Greif, Avner, and Guido Tabellini. 2017. The clan and the corporation: Sustaining cooper-
ation in China and Europe. Journal of Comparative Economics 45: 1–35.
Hao, Yu, and Gregory Clark. 2012. Social mobility in China, 1645–2012: A surname study.
Working paper, University of California, Davis.
Heldring, Leander, James A. Robinson, and Sebastian Vollmer. 2021. The long-run impact
of the Dissolution of the English monasteries. Quarterly Journal of Economics .
Herlihy, David. 1973. Three patterns of social mobility in medieval history. Journal of
Interdisciplinary History 3: 623–647.
Ho, Ping-ti. 1959. Aspects of social mobility in China, 1368–1911. Comparative Studies
in Society and History 1: 330–359.
Hoffman, Philip T. 2015. Why Did Europe Conquer the World? Princeton: Princeton
University Press.
Holland, Arthur William. 1911. Imperial cities or towns. In Chisholm, Hugh (ed.) Ency-
clopædia Britannica, volume 14. Cambridge: Cambridge University Press, 11th edition,
342.
Huang, Yasheng, and Clair Z. Yang. 2021. The great political divergence. Working paper.
Johnson, Noel D., and Mark Koyama. 2019. Persecution and Toleration: The Long Road
to Religious Freedom. New York: Cambridge University Press.
Jones, Eric. 1981. The European Miracle: Environments, Economies and Geopolitics in
the History of Europe and Asia. Cambridge: Cambridge University Press.
Kemp, Brian R. 1970. Hereditary benefices in the medieval English Church: A Hereford-
shire example. Historical Research 43: 1–15.
Ko, Chiu Yu, Mark Koyama, and Tuan-Hwee Sng. 2018. Unified China and divided Eu-

44
 rope. International Economic Review 59: 285–327.
Kokkonen, Andrej, and Anders Sundell. 2014. Delivering stability – Primogeniture and au-
tocratic survival in European monarchies 1000–1800. American Political Science Review
108: 438–453.
Kositz, Bryce. 2019. Mao and Xi: Story of the man, story of the people. In Golley, Jane,
Linda Jaivin, Paul J. Farrelly, and Sharon Strange (eds.) Power: China Story Yearbook.
Acton: ANU Press, 9–12.
Lander, Jack Robert. 1961. Attainder and forfeiture, 1453 to 1509. Historical Journal 4:
119–151.
Levenson, Joseph R. 1965. Confucian China and Its Modern Fate, Volume 2: The Problem
of Monarchial Decay. Berkeley: University of California Press.
Levi, Margaret. 1988. Of Rule and Revenue. Los Angeles: University of California Press.
Li, Weijia, Gérard Roland, and Yang Xie. 2019. Crony capitalism, the party-state, and
political boundaries of corruption. Working paper, University of California, Riverside.
Lü, Simian. 1944. A General History of China, Volume 2. Shanghai: Kai Ming Bookstore.
Reprint, 1992, Shanghai: East China Normal University Press.
Lu, Xing. 2017. The Rhetoric of Mao Zedong: Transforming China and Its People.
Columbia: University of South Carolina Press.
Ma, Debin, and Jared Rubin. 2019. The paradox of power: Principal-agent problems and
administrative capacity in Imperial China (and other absolutist regimes). Journal of
Comparative Economics 47: 277–294.
Mann, Michael. 1986. The Social Sources of Power, Volume 1: A History of Power from
the Beginning to AD 1760. Cambridge: Cambridge University Press.
McCollim, Gary B. 2012. Louis XIV’s Assault on Privilege: Nicolas Desmaretz and the
Tax on Wealth. Rochester: University Rochester Press.
McKechnie, William S. 1914. Magna Carta: A Commentary on the Great Charter of King
John. Glasgow: James MacLehose and Sons, 2nd edition.
McNeill, William H. 1982. The Pursuit of Power: Technology, Armed Force, and Society
Since AD 1000. University of Chicago Press.
Mokyr, Joel. 2016. A Culture of Growth: The Origins of the Modern Economy. Princeton:
Princeton University Press.
Morby, John E. 1989. Dynasties of the World: A Chronological and Genealogical Hand-
book. New York: Oxford University Press.
North, Douglass C. 1989. Institutions and economic growth: An historical introduction.
World Development 17: 1319–1332.
North, Douglass C., John J. Wallis, and Barry R. Weingast. 2009. Violence and Social Or-
ders: A Conceptual Framework for Interpreting Recorded Human History. Cambridge:
Cambridge University Press.
North, Douglass C., and Barry R. Weingast. 1989. Constitutions and commitment: The
evolution of institutions governing public choice in seventeenth-century England. Jour-
nal of Economic History 49: 803–832.
Orwell, George. 1947. The English People. London: Collins.

45
Padró i Miquel, Gerard. 2007. The control of politicians in divided societies: The politics
of fear. Review of Economic Studies 74: 1259–1274.
Parish, Helen. 2010. Clerical Celibacy in the West: C.1100–1700. Farnham: Ashgate
Publishing Limited.
Perry, Elizabeth J. 2008. Chinese conceptions of “right”: From Mencius to Mao – and now.
Perspectives on Politics 6: 37–50.
Persico, Nicola. 2021. A theory of non-democratic redistribution and public good provi-
sion. Working paper.
Porter, Roy S. 1994. London: A Social History. London: Hamish Hamilton.
Qian, Jiwei, and Tuan-Hwee Sng. 2021. The state in Chinese economic history. Working
paper, National University of Singapore.
Qin, Hui. 1998. The “large-community standard” and traditional Chinese society: Part I.
Sociological Studies 13: 12–21.
Reynolds, Susan. 2010. Before Eminent Domain: Toward a History of Expropriation of
Land for the Common Good. Chapel Hill: University of North Carolina Press.
Roland, Gérard. 2004. Understanding institutional change: Fast-moving and slow-moving
institutions. Studies in Comparative International Development 38: 109–131.
Roland, Gérard. 2008. Fast-moving and slow-moving institutions. In Kornai, János, Lás-
zló Mátyás, and Gérard Roland (eds.) Institutional Change and Economic Behaviour.
Basingstoke: Palgrave Macmillan, 134–159.
Roland, Gérard. 2020. The deep historical roots of modern culture: A comparative per-
spective. Journal of Comparative Economics 48: 483–508.
Rosenthal, Jean-Laurent, and Roy B. Wong. 2011. Before and Beyond Divergence: The
Politics of Economic Change in China and Europe. Cambridge: Harvard University
Press.
Scheidel, Walter. 2019. Escape from Rome: The Failure of Empire and the Road to Pros-
perity. Princeton: Princeton University Press.
Shirk, Susan L. 2018. China in Xi’s “New Era”: The return to personalistic rule. Journal
of Democracy 29: 22–36.
Skinner, George W. 1985. Presidential address: The structure of Chinese history. Journal
of Asian Studies 44: 271–292.
Sng, Tuan-Hwee. 2014. Size and dynastic decline: The principal-agent problem in late
imperial China, 1700–1850. Explorations in Economic History 54: 107–127.
Southern, Richard W. 1970. Western Society and the Church in the Middle Ages. Baltimore:
Penguin Books.
Stasavage, David. 2016. Representation and consent: Why they arose in Europe and not
elsewhere. Annual Review of Political Science 19: 145–162.
Stasavage, David. 2020. The Decline and Rise of Democracy: A Global History from
Antiquity to Today. Princeton: Princeton University Press.
Tackett, Nicolas. 2014. The Destruction of the Medieval Chinese Aristocracy. Cambridge:
Harvard University Asia Center.
Tamanaha, Brian Z. 2004. On the Rule of Law: History, Politics, Theory. Cambridge:

46
 Cambridge University Press.
Tian, Zhaolin. 2019. History of Warfare in China. Nanjing: Jiangsu People’s Publishing.
Tuchman, Barbara W. 1978. A Distant Mirror: The Calamitous 14th Century. New York:
Alfred A. Knopf.
Turchin, Peter. 2009. A theory for formation of large empires. Journal of Global History
4: 191–217.
Unger, Roberto M. 1977. Law in Modern Society: Toward a Criticism of Social Theory.
New York: The Free Press.
Usher, Dan. 1989. The dynastic cycle and the stationary state. American Economic Review
79: 1031–1044.
Vincent, Nicholas. 2012. Magna Carta: A Very Short Introduction. Oxford: Oxford Uni-
versity Press.
von Glahn, Richard. 2016. An Economic History of China: From Antiquity to the Nine-
teenth Century. Cambridge: Cambridge University Press.
Wang, Jiafan. 2000. A General History of China. Shanghai: East China Normal University
Press.
Watson, Burton D. 1964. Han Fei Tzu: Basic Writings. New York: Columbia University
Press.
Wickham, Chris. 2009. The Inheritance of Rome: A History of Europe from 400 to 1000.
London: Penguin.
Williams, Gwyn A. 1963. Medieval London: From Commune to Capital. London: Athlone
Press.
Wittfogel, Karl A. 1957. Oriental Despotism: A Comparative Study of Total Power. New
Haven: Yale University Press.
Wu, Guoqing. 2016. History of Warfare in China. Beijing: People’s Press.
Xue, Melanie M. 2021. Autocratic rule and social capital: Evidence from imperial China.
Working paper.
Yan, Buke. 2004. Bureaucratic political institutions during the beginning of the imperial
era – Qin and Han dynasties. In Wu, Zongguo (ed.) Studies on the Bureaucratic Political
Institutions in Ancient China. Beijing: Peking University Press, 19–85.
Yan, Buke. 2009. Peaks and Troughs: Political Civilization from the Qin Dynasty to the
Southern and Northern Dynasties. Beijing: Peking University Press.
Zhang, Taisu. 2017. The Laws and Economics of Confucianism: Kinship and Property in
Preindustrial China and England. Cambridge: Cambridge University Press.
Zhao, Dingxin. 2009. The Mandate of Heaven and performance legitimation in historical
and contemporary China. American Behavioral Scientist 53: 416–433.
Zhao, Dingxin. 2015. The Confucian-Legalist State: A New Theory of Chinese History.
New York: Oxford University Press.

47
For Online Publication: Appendix

A Allowing for Mixed Strategies

In this section we allow for mixed strategies at Stage 2 by dropping the earlier assumption
that E and P will side with R when indifferent and C will not challenge when indifferent. We
then characterize all the subgame perfect equilibria that are empirically relevant (possible
for the status quo to end) and can involve mixed strategies at a strictly positive share of the
states of the world. We then examine whether the main insights from the main text would
maintain.
When doing so, we adopt a few additional assumptions without losing much generality.
We consider only the empirically relevant, nontrivial case γ < 1. We also assume that x is
a continuous random variable so that its distribution does not have any mass point, and that
F(ap) < 1 so that 1 − F(x̂) > 0 always holds.
By backward induction, in any subgame perfect equilibrium at Stage 2, P will side with
R when x < x̂ and not side with R when x > x̂.
Taking this into consideration, in any subgame perfect equilibrium, E will side with R
when x < x̂; when x > x̂, E will side with R if

p
w< · (1 − γ )a; (1)
1− p

E will not side with R if

p
w> · (1 − γ )a; (2)
1− p
E will side with R with probability qE (x) if

p
w= · (1 − γ )a, (3)
1− p

where qE (x) is a function and satisfies qE (x) ∈ [0, 1] for any x > x̂.
Taking this into consideration, in any subgame perfect equilibrium, C will not challenge
when x < x̂. When x > x̂, C will not challenge if

p
w< · (1 − γ )a; (4)
1− p

1
C will also not challenge if

p p
w> · (1 − γ )a and z < · y; (5)
1− p 1− p

C will challenge if
p p
w> · (1 − γ )a and z > · y; (6)
1− p 1− p
C will challenge with probability qC (x) if

p p
w> · (1 − γ )a and z = · y, (7)
1− p 1− p

where qC (x) is a function and satisfies qC (x) ∈ [0, 1] for any x > x̂; if

p
w= · (1 − γ )a, (8)
1− p

however, C will compare

( )
0 vs. qE (x) · (−y) + (1 − qE (x)) · (1 − p) · z − p · y , (9)

i.e.,
( )
0 vs. (1 − qE (x))(1 − p) · z − 1 − (1 − qE (x))(1 − p) · y, (10)

so C will challenge with probability qC (x), where, for any x > x̂, qC (x) = 1 if

1 − (1 − qE (x))(1 − p)
z> · y, (11)
(1 − qE (x))(1 − p)

qC (x) = 0 if
1 − (1 − qE (x))(1 − p)
z< · y, (12)
(1 − qE (x))(1 − p)
and qC (x) ∈ [0, 1] if
1 − (1 − qE (x))(1 − p)
z= · y. (13)
(1 − qE (x))(1 − p)
We have then specified all equilibrium strategies at any x ̸= x̂. Therefore, at Stage 2, the
only families of subgame perfect equilibria that are empirically relevant and can involve
mixed strategies at a strictly positive share of the states of the world are:
p p
• When w > 1−p ·(1− γ )a and z = 1−p ·y, in any subgame perfect equilibrium, if x < x̂,

2
 then C will not challenge, E would side with R, and P would side with R; if x > x̂,
then C will challenge with probability qC (x) ∈ [0, 1], E will not side with R, and P
will not side with R.
∫∞
In any equilibrium of this family, the probability of challenge is x̂ qC (x)dF(x),
while the survival probability of the status quo is
∫ ∞
S = 1− qC (x)dF(x) · (1 − p). (14)
x̂

All impacts of γ and β on political stability still come from their impacts on x̂. All
main insights from the main text would then remain.
p
• When w = 1−p · (1 − γ )a, in any subgame perfect equilibrium, if x < x̂, then C will
not challenge, E would side with R, and P would side with R; if x > x̂, then C will
challenge with probability qC (x), where qC (x) depends on

1 − (1 − qE (x))(1 − p)
z vs. · y, (15)
(1 − qE (x))(1 − p)

E will side with R with probability qE (x), and P will not side with R.
∫∞
In any equilibrium of this family, the probability of challenge is x̂ qC (x)dF(x),
while the survival probability of the status quo is
∫ ∞
S = 1− qC (x)(1 − qE (x))(1 − p)dF(x). (16)
x̂

Still, all impacts of γ and β on political stability come from their impacts on x̂. All
main insights from the main text would then remain.

B Institutional Compatibility under Additional Restric-

tion on the Distribution of the State of the World
Proposition B.1. If the distribution of x satisfies

x · f ′ (x) a
ε ≡− ≤ ε̄ ≡ 1 − (17)
f (x) π − 2a − r

over x ∈ [0, pa], then a lower γ ∈ [0, 1] would make R prefer a higher β ∈ [0, 1].

3
Proof. Observe that

dV R ( ) dS dS
= π − (1 + β )a − r · − aS, = (1 − p)p f (x̂) · a · (1 − γ ), (18)
dβ dβ dβ

and
( )
S = 1 − 1 − F(x̂) · (1 − p). (19)

Therefore,

∂ 2V R ( ) ∂S dS
= π − (1 + β )a − r · −a·
∂ γ∂ β ∂ γ∂ β dγ
( ) ( ) dS
= − π − (1 + β )a − r · (1 − p)pa · (1 − γ ) f ′ (x̂) · paβ + f (x̂) − a ·
dγ
( ) ( ′ )
= − π − (1 + β )a − r · (1 − p)pa · f (x̂) · x̂ + f (x̂) + a · (1 − p) f (x̂)pβ a
(( ) ( ) )
= −(1 − p)pa · π − (1 + β )a − r · f ′ (x̂) · x̂ + f (x̂) − f (x̂)β a
(( ) ′ ( ) )
= −(1 − p)pa · π − (1 + β )a − r · f (x̂) · x̂ + π − (1 + 2β )a − r · f (x̂) .
(20)

Therefore, ∂ 2V R /∂ γ∂ β ≤ 0 if and only if

( ) ( )
π − (1 + β )a − r · f ′ (x̂) · x̂ + π − (1 + 2β )a − r · f (x̂) ≥ 0, (21)

i.e.,
f ′ (x̂) · x̂ π − (1 + 2β )a − r βa
ε ≡− ≤ = 1− . (22)
f (x̂) π − (1 + β )a − r π − (1 + β )a − r
Since [ ]
βa a
∈ 0, , (23)
π − (1 + β )a − r π − 2a − r
we have [ ]
βa a
1− ∈ 1− ,1 . (24)
π − (1 + β )a − r π − 2a − r
Therefore, ∂ 2V R /∂ γ∂ β ≤ 0 can be guaranteed by

a
ε ≤ 1− ≡ ε̄ , where ε̄ < 1. (25)
π − 2a − r

The proposition then follows.

4
C Dependence of the Stabilizing Effect of the Ruler’s Ab-
solute Power on the Elite–People Relationship
Lemma 1 shows that the impact of the Elite–People symmetry on the stability of the status
quo is generally increasing in the absolute power of the Ruler. We can show a parallel result
when analyzing the impact of the Ruler’s absolute power on the stability of the status quo:

Lemma C.1. The stabilizing effect of the Ruler’s absolute power is increasing in the level
of the Elite–People symmetry, i.e., dS/d(1 − γ ) > 0 is increasing in β .
( )
Proof. By x̂ = p(1 − γ )β a, S = 1 − 1 − F(x̂) · (1 − p), and f (x) ∈ [ f , f¯] ⊂ (0, ∞) over
x ∈ [0, pa], we have

dS d x̂
= f (x̂) · (1 − p) · = f (x̂) · (1 − p) · pβ a > 0, (26)
d(1 − γ ) d(1 − γ )

which is increasing in β .

D Allowing Current Stability to Shape Future Power Struc-

ture
Based on the equilibrium at Stage 2, R’s preference over γ is straightforward: a lower γ
stabilizes the status quo (higher S) without any impact on R’s status quo payoff; therefore,
R will prefer the lowest possible γ .
Proposition 3 also implies:

Corollary D.1 (Higher β almost always preferred by R). As π − r → ∞, γ → 1− .

This result suggests that when the surplus R would enjoy is sufficiently large, given any
γ < 1, R will prefer β to be as high as possible. This result and R’s preference over γ allow
us to consider the following setting:

• At t:

– The ruling position’s historical strength St−1 is given.

– γt = γ (St−1 ) and βt = β (St−1 ) are realized, where γ (S) and β (S) satisfy γS (S) <
0 and βS (S) > 0, respectively.

5
 – The modeled Stage 2 plays out St = 1 − (1 − F(x̂)) · (1 − p) ≡ S (βt , γt , θ ) as in
the unique subgame perfect equilibrium; θ include all factors that conditional
on St−1 , affect St but do so not through γt or βt .

• At t + 1: The same happens.

The dynamics then follows

βt = β (St−1 ), γt = γ (St−1 ), St = S (βt , γt , θ ) , (27)

or just
( )
St = S β (St−1 ), γ (St−1 ), θ . (28)

Steady states are then defined by

( ) ( ) ( )
S∗ = S β ∗ , γ ∗ , θ , β ∗ = β S∗ , γ ∗ = γ S∗ , (29)

or just ( ( ) ( ) )
S∗ = S β S∗ , γ S∗ , θ . (30)

Existence and stability of steady states. The defining equation of steady states can help
establish a few technical results. The first result is about the possible range of St in the
dynamics:

Lemma D.1. Any St in the dynamics must satisfy S ≤ St ≤ S̄, where S = p and S̄ = 1 − (1 −
p) · (1 − F(pa)) < 1.

Proof. Note that Sβ ≥ 0 and Sγ ≤ 0. Therefore, the minimum S is reached when βt = 0 and
γt = 1 and the maximum S̄ is reached when βt = 1 and γt = 0. The lemma then follows.

The first result helps establish the second result, which is about the existence of a steady
state given a reasonable assumption about β (·) and γ (·):

Lemma D.2. If β (S), γ (S), β (S̄), and γ (S̄) are all within the range (0, 1), then there exists
( )
at least one steady state S∗ , at which St = S β (St−1 ), γ (St−1 ), θ crosses St = St−1 from
St > St−1 to St < St−1 , and 0 ≤ Sβ · βS + Sγ · γS ≤ 1.

Proof. Note that Sβ > 0 and Sγ > 0 for any β > 0 and γ < 1. Therefore, by β (S) > 0 and
( )
γ (S) < 1, we have S β (S) , γ (S) , θ > S; by 0 < β (S̄) < 1 and 0 < γ (S̄) < 1, we have

6
 ( ( ) ( ) ) ( )
S β S̄ , γ S̄ , θ < S̄. Since S β (s), γ (s), θ is continuous in s, the defining equation
( ) [ ]
S∗ = S β (S∗ ) , γ (S∗ ) , θ must have a solution S∗ ∈ S, S̄ , i.e., a steady state exists, at
( )
which St = S β (St−1 ), γ (St−1 ), θ crosses St = St−1 from St > St−1 to St < St−1 . Moreover,
note that ( )
dS β (s) , γ (s) , θ
= Sβ · βS + Sγ · γS ≥ 0, (31)
ds
( )
so St = S β (St−1 ), γ (St−1 ), θ is increasing in St−1 . Therefore, at S∗ ,
( )
dS β (s) , γ (s) , θ
0≤ = Sβ · βS + Sγ · γS ≤ 1. (32)
ds

The third result is the condition for a steady state to be stable:

( )
Lemma D.3. A steady state S∗ is stable if and only if at S∗ , St = S β (St−1 ), γ (St−1 ), θ
crosses St = St−1 from St > St−1 to St < St−1 and 0 ≤ Sβ · βS + Sγ · γS ≤ 1.
( )
Proof. First, suppose a steady state S∗ is stable, then at S∗ , St = S β (St−1 ), γ (St−1 ), θ
crosses St = St−1 and
( )
dS β (S∗ ) , γ (S∗ ) , θ
−1 < = Sβ · βS + Sγ · γS ≤ 1. (33)
dS∗

Note that
Sβ · βS + Sγ · γS ≥ 0, (34)

so
0 ≤ Sβ · βS + Sγ · γS ≤ 1. (35)

Therefore, the crossing must be from St > St−1 to St < St−1 .

The other direction of the lemma is straightforward. The lemma is then proven.

The last two results establish the existence of stable steady states:

Proposition D.1. If β (S), γ (S), β (S̄), and γ (S̄) are all within the range (0, 1), then there ex-
( )
ists at least one stable steady state, and at all the stable steady states, St = S β (St−1 ), γ (St−1 ), θ
crosses St = St−1 from St > St−1 to St < St−1 and 0 ≤ Sβ · βS + Sγ · γS ≤ 1.

7
 ( )
Multiplicity of stable steady states. Multiplicity will appear if St = S β (St−1 ), γ (St−1 ), θ
crosses St = St−1 more than once. The conditions governing single- or multi-crossing con-
cern the second-order properties of β (·) and γ (·), which depend on their micro-foundation.
In light of this, we do not specify the conditions here; instead, we take the possibility of
multiplicity as given and explore the implications under this possibility:

Institutional compatibility under multiple steady states. Assuming β (S), γ (S), β (S̄),
and γ (S̄) are all within the range (0, 1), we can have the following result: if multiple steady
states exist given θ , then any two different steady states must be different in a certain way,
i.e., follows institutional compatibility:

Proposition D.2. Given θ , if there are two steady states {S∗ , β ∗ , γ ∗ } and {S∗ ′ , β ∗ ′ , γ ∗ ′ },
then any one among the following three statements will imply the other two: 1) S∗ ≥ S∗ ′ ;
2) β ∗ ≥ β ∗ ′ ; 3) γ ∗ ≤ γ ∗ ′ .

Proof. The result follows the three defining equations of steady states and their monotonic-
ity.

Given multiple steady states, the second result is about the divergence of compatible
institutions:

Proposition D.3. If there are N ≥ 2 different stable steady states S1∗ < · · · < SN∗ , then there
are N − 1 different unstable steady states S̃1 < · · · < S̃N−1 , they satisfy S < S1∗ < S̃1 < S2∗ <
∗
S̃2 < · · · < SN−1 < S̃N−1 < SN∗ < S̄, and the institutional dynamics is determined by the
initial strength of the ruling position S0 :
∗
• if S̃n < S0 < S̃n+1 , where n = 1, . . . , N − 1, then St → Sn+1 as t → ∞;

• if S ≤ S0 < S̃1 , then St → S1∗ as t → ∞;

• if S̃N−1 < SN∗ < S̄, then St → SN∗ as t → ∞.

( )
Proof. As eventually St = S β (St−1 ), γ (St−1 ), θ has to cross St = St−1 from St > St−1
to St < St−1 , we can rank the stable and unstable steady states as proposed. Neighboring
unstable steady states then divide the possible range of S into sub-ranges, starting from
each of which St will converge to the stable steady state in it.

8
 This result implies that the institutional difference between China and Europe can
be thought as different stable steady states given the same primitives but different initial
strengths S of the ruling position in history, which is compatible with different β and γ at
very early times.

E Endogenizing the Challenger and Elites’ Incentives in a

Markov Game
In this extension of Stage 2 we collapse C and E into a single player E, make her look
forward in a Markov game with an infinite number of discrete periods, and allow her to
replace R. Figure 1 shows each period of the Markov game.
Compared with Figure 1, Stage 2 will now continue after each period; the prize z for C
to challenge and the incentive w for E not to side with R are replaced by the aspiration of E
to replace R at the end of this period; the loss y for C if her challenges fails is replaced by
the punishment that would reduce E’s payoff from the status quo level a to γ a. About the
stochastic elements of the game, we assume that N’s draws of x and whether R will survive
the challenge on his own within each period and across periods are mutually independent.
About the dynamic elements of the game, we assume that all the players have an infinite
horizon with an exogenous intertemporal discount factor δ ∈ (0, 1). All other assumptions
in the main text remain here.
We will adopt the Markov perfect equilibrium as the solution concept in our analysis.
For simplicity, we still assume that E will not challenge and P will side with R if they are
indifferent in their decision, respectively, ruling out mixed strategies. Online Appendix E.3
shows that allowing for mixed strategies would accommodate a mixed-strategy equilibrium
when and only when pure-strategy equilibria do not exist, while the key insights would
remain robust.

E.1 Equilibrium Characterization

Now we analyze the extended Stage 2 by first characterizing all possible Markov perfect
equilibria and finding the conditions under which they exist. We denote the net present val-
ues that the players enjoy at the beginning of each period as V R , V E , and V P , respectively.
We have a first result to partially characterize all Markov perfect equilibria:

9
 .
Game continues

Nature (N)
Draws state of the world
x per c.d.f. F(x)
Elites (E)

Does not challenge Challenges the status quo, which

the status quo is maintained by Ruler (R)

R gets π − (1 + β )a
E gets a People (P)
P gets β a
Sides Does not side
Game continues with R with R
R gets π − (1 + β )a
E gets γ a N
P gets β a − x
W.p. 1 − p, W.p. p,
status quo status quo
Game continues ends survives
R gets r R gets π − (1 + β )a
E gets a E gets γ a
P gets β a P gets γβ a

R exits without future payoffs Game continues

E becomes R
New E enters
Game continues

x ≥ 0, a > 0, π − 2a > r, 0 ≤ β ≤ 1, 0 ≤ γ ≤ 1, 0 < p < 1

Figure 1: Extended Stage 2: Each period in the Markov game

Lemma E.1. In any Markov perfect equilibrium, P will side with R if and only if x ≤ x̂ ≡
(1 − γ )β p · a, where x̂ ∈ [0, pa]; when x ≤ x̂, E will not challenge the status quo, and when
x > x̂, E will challenge if and only if the aspiration to replace R in equilibrium dominates
the probability-adjusted punishment in case of a failed challenge:

p
V R −V E > · (1 − γ )a. (36)
(1 − p)δ

10
Proof. In any Markov perfect equilibrium, P will side with R if and only if

β a − x + δ V P ≥ (β a + δ V P ) · (1 − p) + (γβ a + δ V P ) · p, (37)

i.e.,
x ≤ (1 − γ )β p · a ≡ x̂. (38)

Given this strategy of P and the continuation strategy of E in the equilibrium, E will not
challenge if x ≤ x̂, since
a + δV E ≥ γa + δV E (39)

holds for any γ ∈ [0, 1] and V E ; when x > x̂, E will challenge if and only if

a + δ V E < (a + δ V R ) · (1 − p) + (γ a + δ V E ) · p, (40)

i.e.,
p
V R −V E > · (1 − γ )a. (41)
(1 − p)δ
The lemma is then proven.

Note that the analysis is parallel to Section 3.1.2, the definition of x̂ is the same as in
Section 3.1.2, and Condition (36) is parallel to Conditions (4) and (6).
By Lemma E.1, only two Markov perfect equilibria are possible. The first one is a
secured-R equilibrium:

Proposition E.1 (Secured-R equilibrium in the Markov game). If

π − (2 + β )a p
h(β , γ ) ≡ − · (1 − γ )a ≤ 0, (42)
1−δ (1 − p)δ

then “E never challenges the status quo; P would not side with R if and only if x > x̂” is a
Markov perfect equilibrium; in this equilibrium, the survival probability of the status quo
is S = 1.

Proof. For “E never challenges the status quo; P would not side with R if and only if x > x̂”
to be a Markov perfect equilibrium, the condition

p
V R −V E ≤ · (1 − γ )a (43)
(1 − p)δ

11
must hold, where, given E and P’s strategies in this equilibrium,

π − (1 + β )a a
VR = and V E = . (44)
1−δ 1−δ

The condition is then equivalent to

π − (1 + β )a a p
− ≤ · (1 − γ )a, (45)
1−δ 1−δ (1 − p)δ

i.e.,
π − (2 + β )a p
h(β , γ ) ≡ − · (1 − γ )a ≤ 0. (46)
1−δ (1 − p)δ
The proposition is then proven.

The intuition of the result / is as follows: the function h(β , γ ) measures E’s aspiration
V R − V E = (π − (2 + β )a) (1 − δ ) to replace R given the specified strategies, net of the
( )
probability-adjusted punishment p/(1 − p)δ · (1 − γ )a on E in case the challenge fails.
The condition h(β , γ ) ≤ 0 then suggests that the aspiration cannot dominate the punish-
ment. Lemma E.1 then implies that we have the secured-R equilibrium.
Note that this equilibrium is parallel to the scenario in Section 3.1.2 when Conditions
(4) and (6) do not hold. Following the same argument as in Section 3.1.2, this equilibrium
is empirically not much relevant, as in reality the chance for R to be ousted was always
strictly positive; it is also trivial, in the sense that no challenge will happen in equilibrium.
The second equilibrium is an unsecured-R equilibrium:

Proposition E.2 (Unsecured-R equilibrium in the Markov game). If

( )
π − (1 + β )a · S + r · (1 − S) a p
g(β , γ ) ≡ − − · (1 − γ )a > 0, (47)
1−δS 1 − δ (1 − p)δ

where
( )
S = 1 − 1 − F(x̂) · (1 − p) ∈ [p, 1] and x̂ ≡ (1 − γ )β p · a, (48)

then “E will challenge the status quo if and only if x > x̂; P would not side with R if and
only if x > x̂” is a Markov perfect equilibrium; in this equilibrium, R’s stability is S ≤ 1.

Proof. For “E will challenge the status quo if and only if x > x̂; P would not side with R if

12
and only if x > x̂” to be a Markov perfect equilibrium, the condition

p
V R −V E > · (1 − γ )a (49)
(1 − p)δ

must hold, where, given E and P’s strategies in this equilibrium,

( )
V R = π − (1 + β )a + δ V R · S + r · (1 − S)
( )
= π − (1 + β )a · S + r · (1 − S) + δ V R · S
( )
π − (1 + β )a · S + r · (1 − S)
= (50)
1−δS

and
( ( ) ) ( )
V E = a · 1 − 1 − F(x̂) · p + γ a · 1 − F(x̂) · p + δ V E · S + δ V R · (1 − S)
( ( ) )
= a · 1 − (1 − γ ) · 1 − F(x̂) · p + δ V E · S + δ V R · (1 − S)
( ( ) )
a · 1 − (1 − γ ) · 1 − F(x̂) · p + δ V R · (1 − S)
= , (51)
1−δS

with
( )
S = 1 − 1 − F(x̂) · (1 − p) ∈ [p, 1]. (52)

The condition is then equivalent to, with some algebra,

( )
π − (1 + β )a · S + r · (1 − S) a p
g(β , γ ) ≡ − − · (1 − γ )a > 0. (53)
1−δS 1 − δ (1 − p)δ

The proposition is then proven.

Again, the intuition of Proposition E.2 follows Lemma E.1: the function g(β , γ ) indi-
cates, given the specified strategies, how E’s aspiration V R −V E to replace R is compared
with the punishment in case the challenge fails. The condition g(β , γ ) > 0 then suggests
that the aspiration dominates the punishment. Lemma E.1 then implies that we have the
unsecured-R equilibrium.
Following the same argument as in Section 3.1.2, the unsecured-R equilibrium is em-
pirically relevant and nontrivial. We thus now explore the conditions under which it always
exists and is the unique equilibrium. The following result first shows that the secured-R
equilibrium and the unsecured-R equilibrium cannot exist simultaneously:

13
Corollary E.1. Given r ≤ π − 2a, if g(β , γ ) > 0, then h(β , γ ) > 0, i.e., if the unsecured-R
equilibrium exists, then the secured-R equilibrium does not exist.

Proof. Observe that, by r ≤ π − 2a, for any S ∈ [p, 1], g(β , γ ) ≤ g(β , γ )S=1 = h(β , γ ).
Therefore, if g(β , γ ) > 0, then h(β , γ ) > 0.

The intuition of Corollary E.1 is as follows. Since R is safer in the secured-R equi-
librium than in the unsecured-R equilibrium, E’s aspiration to replace R is stronger, too.
Therefore, if E’s aspiration is already so strong that the unsecured-R equilibrium is sup-
ported (g(β , γ ) > 0), then given the strategies specified in the secured-R equilibrium, E’s
aspiration must be too strong to support the secured-R equilibrium (h(β , γ ) > 0).
This corollary helps derive a set of conditions under which the unsecured-R equilibrium
will generally exist and be the unique equilibrium, parallel to Proposition 1:
( /
Proposition E.3 (Focus on unsecured-R equilibrium in the Markov game). If (1− δ p) (1−
)
δ )(1 − p)δ · a ≤ r ≤ π − 2a, then given any β ∈ [0, 1] and γ ∈ [0, 1], the unsecured-R e-
quilibrium exists and is the unique Markov perfect equilibrium.
( / )
Proof. For any β ∈ [0, 1] and γ ∈ [0, 1], by 0 < (1 − δ p) (1 − δ )(1 − p)δ · a ≤ r ≤
π − 2a and S ∈ [p, 1], we have

(π − 2a) · S + r · (1 − S) a p
g(β , γ ) ≥ − − ·a
1−δS 1 − δ (1 − p)δ
r (1 − p)δ + p(1 − δ ) r 1− p+ p
≥ − ·a > − ·a
1−δS (1 − δ )(1 − p)δ 1 − δ p (1 − δ )(1 − p)δ
r 1
≥ − · a ≥ 0. (54)
1 − δ p (1 − δ )(1 − p)δ

Therefore, g(β , γ ) > 0, i.e., the unsecured-R equilibrium exists, and by Corollary E.1, the
secured-R equilibrium does not exist. Therefore, the unsecured-R equilibrium is the unique
equilibrium.
( / )
In this result, (1 − δ p) (1 − δ )(1 − p)δ · a ≤ r is parallel to w > ap/(1 − p) and
z > yp/(1 − p) in Proposition 1, guaranteeing that E’s aspiration to replace R is sufficiently
strong so that E will challenge if P will not side with R.

14
E.2 Analysis of the Unsecured-R Equilibrium
To focus on the empirically relevant, nontrivial unsecured-R equilibrium(in our analysis, /
from now on we assume that the condition in Proposition E.3 holds, i.e., (1 − δ p) (1 −
)
δ )(1 − p)δ · a ≤ r ≤ π − 2a, so that the unsecured-R equilibrium exists and is the unique
Markov perfect equilibrium. Without losing generality, as in Section 3.2, we also assume
that the state of the world x’s probability density function satisfies f (x) ∈ [ f , f¯] ⊂ (0, ∞)
over x ∈ [0, pa].
Now we can derive parallel results to Sections 3.1.3 and 3.2. First note that as in
Section 3.1.3, in each period, the probability of challenge is still 1 − F(x̂) and the survival
probability of the status quo is still
( )
S = 1 − 1 − F(x̂) · (1 − p), (55)

where x̂ ≡ (1 − γ )β p · a, so Proposition 2 still holds in this Markov game.

Now examine R’s preference over γ and β . The net present value of R’s payoffs in
equilibrium is
( ) ( )
R π − (1 + β )a · S + r · (1 − S) π − (1 + β )a − r · S + r
V = = , (56)
1−δS 1−δS

which differs from Equation (8) only at that it includes the future payoffs. Therefore, R
will still prefer γ to be as low as possible.
On R’s preference over β , first, since Proposition 2 still holds in this Markov game, the
political–economic trade-off still appears and Lemma 1 still holds. We can then derive the
following result parallel to Proposition 3:

Proposition E.4 (Institutional compatibility in the Markov game). If γ < γ ≡ 1 − (1 −

/( )
δ ) (1 − δ (1 − p)) (π − 2a − r) + δ r c, then R will prefer β to be as high as possible; if
/( )
γ > γ̄¯ ≡ 1 − (1 − δ )p π − a − r(1 − δ ) c̄, then R will prefer β to be as low as possible,
( )/( )
1−δ
where γ < γ̄ < 1 and if π > 2a + c + (1 − δ (2 − p))r
¯ 1 − δ (1 − p) , then γ > 0.

Proof. The marginal impact of β on R’s net present value in equilibrium is

( ( ))
δ (π −(1+β )a−r)S+r
π − (1 + β )a − r + 1−δ S · ddSβ − aS
dV R
= . (57)
dβ 1−δS

15
 ( / )
By Lemma 1, β ∈ [0, 1], S ∈ [p, 1], and 0 < (1 − δ p) (1 − δ )(1 − p)δ · a ≤ r ≤ π − 2a,
we have
( )
(1−δ (S−p))(π −2a−r)+δ r
dV R 1−δ S · ca · (1 − γ ) − a
≥
dβ 1−δS
( )
a (1 − δ (1 − p)) (π − 2a − r) + δ r
≥ · · c · (1 − γ ) − 1 , (58)
1−δS 1−δ

so, if
(1 − δ (1 − p)) (π − 2a − r) + δ r
· c · (1 − γ ) − 1 > 0, (59)
1−δ
i.e.,
1−δ
γ < 1− ( ) ≡ γ, (60)
(1 − δ (1 − p)) (π − 2a − r) + δ r · c
then dV R /d β > 0. At the same time, we have
( )
dV R π − a − r + δ (1−
π −a)
δ · c̄a · (1 − γ ) − ap
≤
dβ 1−δS
(( ) )
a π −a
= · − r · c̄ · (1 − γ ) − p , (61)
1−δS 1−δ

so, if ( )
π −a
− r · c̄ · (1 − γ ) − p < 0, (62)
1−δ
i.e.,
(1 − δ )p
γ > 1− ( ) ≡ γ̄¯, (63)
π − a − r(1 − δ ) · c̄
(
1−δ
then dV R /d β < 0. Finally, note γ < γ̄¯ < 1, and γ > 0 is equivalent to π > 2a + +
)/ ( )
c

(1 − δ (2 − p))r 1 − δ (1 − p) . The proposition is then proven.

Proposition E.4 differs from Proposition 3 only at that γ and γ̄¯ are differently defined
from how γ and γ̄ are defined, respectively, due to the change in the expression of V R . This
is then followed by parallel results to Section 3.3. We have then shown that we can derive
all the parallel results to the main text from the Markov game.

16
E.3 Allowing for Mixed Strategies
Here we allow for mixed strategies by dropping the earlier assumption that E will not
challenge and P will side with R if they are indifferent between their options. We then re-
characterize all the Markov perfect equilibria of the game at Stage 2 and examine whether
the main insights would remain. As in Online Appendix A, we assume γ < 1; we also
assume that x is a continuous random variable so that its distribution does not have any
mass point, and that F(ap) < 1 so that 1 − F(x̂) > 0 always holds.
In any Markov perfect equilibrium, P’s strategy is then “not to side with R when x > x̂
and to side with R when x < x̂.” As x is a continuous random variable, we can leave P’s
strategy when x = x̂ unspecified without much real consequence.
By γ < 1, given P’s strategy and E’s continuation strategy in the equilibrium, E’s s-
trategy is then not to challenge when x < x̂; when x > x̂, E will challenge with a given
probability qE (x) ∈ [0, 1], which is a function of x > x̂, and we denote
∫∞
qE (x)dF(x)
q̄E ≡ x̂
∈ [0, 1]. (64)
1 − F(x̂)

In particular, if in equilibrium

p
V R −V E > · (1 − γ )a, (65)
(1 − p)δ

then qE (x) = 1 for any x ≥ x̂, with q̄E = 1; if in equilibrium

p
V R −V E < · (1 − γ )a, (66)
(1 − p)δ

then qE (x) = 0 for any x ≥ x̂, with q̄E = 0; if in equilibrium

p
V R −V E = · (1 − γ )a, (67)
(1 − p)δ

then qE (x) should make this condition hold. Again, as x is a continuous random variable,
we can leave E’s strategy when x = x̂ unspecified.
In the equilibrium with such strategies, we must have
( )
R π − (1 + β )a · S + r · (1 − S)
V = , (68)
1−δS

17
 (( ) ) ( )
V = a · 1 − 1 − F(x̂) · q̄E · p + γ a · 1 − F(x̂) · q̄E · p + δ V E · S + δ V R · (1 − S)
E
( ( ) )
a · 1 − (1 − γ ) · 1 − F(x̂) · q̄E · p + δ V R · (1 − S)
= , (69)
1−δS
and
( )
S = 1 − 1 − F(x̂) · q̄E · (1 − p). (70)

By some algebra, the function that governs the existence of the equilibrium turns out to be
p
V R −V E − · (1 − γ )a
(1 − p)δ
(( ) )
1−δ π − (1 + β )a · S + r · (1 − S) a p
= · − − · (1 − γ )a . (71)
1−δS 1−δS 1 − δ (1 − p)δ

Now define
( )
π − (1 + β )a · S + r · (1 − S) a p
k(β , γ , q̄E ) ≡ − − · (1 − γ )a, (72)
1−δS 1 − δ (1 − p)δ

where
( )
S = 1 − 1 − F(x̂) · q̄E · (1 − p) and x̂ = (1 − γ )β pa. (73)

Note that by F(pa) < 1 and π −2a > r, k(β , γ , q̄E ) is strictly decreasing over q̄E ∈ [0, 1].
We can then characterize the Markov perfect equilibria in three scenarios, except for E and
P’s strategies when x = x̂:
1. When k(β , γ , 0) < 0, the unique family of Markov perfect equilibria that can exist
must satisfy:

• P will side with R when x < x̂ and will not side with R when x > x̂;
• E will never challenge when x ̸= x̂.

2. When k(β , γ , 1) > 0, the unique family of Markov perfect equilibria that can exist
must satisfy:

• P will side with R when x < x̂ and will not side with R when x > x̂;
• E will not challenge when x < x̂ and will challenge when x > x̂.

3. When k(β , γ , 0) ≥ 0 and k(β , γ , 1) ≤ 0, there exists a unique q̄E ∈ [0, 1] such that

k(β , γ , q̄E ) = 0, (74)

18
 and the unique family of Markov perfect equilibria that can exist must satisfy:

• P will side with R when x < x̂ and will not side with R when x > x̂;
• E will challenge with a given probability qE (x) ∈ [0, 1], where the function
qE (x) satisfies ∫∞
x̂ qE (x)dF(x)
= q̄E , (75)
1 − F(x̂)
when x > x̂ and will not challenge when x < x̂.

Note that Scenario 1 corresponds to Proposition E.1, where h(β , γ ) ≡ k(β , γ , 0), and S-
cenario 2 corresponds to Proposition E.2, where g(β , γ ) ≡ k(β , γ , 1). Now examine whether
our main messages remain in Scenario 3.
In Scenario 3, in equilibrium, we always have
( )
π − (1 + β )a · S + r · (1 − S) a p
k(β , γ , q̄E ) ≡ − − · (1 − γ )a = 0, (76)
1−δS 1 − δ (1 − p)δ

i.e., ( )
( ) 1 p · (1 − γ )
π − (1 + β )a − r · S + r = + · a · (1 − δ S). (77)
1−δ (1 − p)δ
This implies
pa(1−δ S)
(1−p)δ · d(1 − γ ) + aS · d β
dS = ( ) . (78)
γ)
π − (1 + β )a − r + 1−1 δ + p·(1−
(1−p)δ δ a

By π − 2a > r, we see that a higher β and a lower γ will increase in equilibrium the survival
probability of the status quo S, corresponding to Proposition 2, which is for Scenario 2.
This result also suggests that in equilibrium

dS aS
= ( )
dβ π − (1 + β )a − r + 1−1 δ + p·(1− γ)
(1−p)δ δ a
(( ) )
p·(1−γ )
a· 1−δ + (1−p)δ · a − r
1

=( ( ) )2 . (79)
p·(1−γ )
π − (1 + β )a − r + 1−δ + (1−p)δ δ a
1

19
This implies
(( ) )
p·(1−γ )
a· 1
1−δ + (1−p)δ ·a−r
dS
≤ ( )2 ≡ b̄(γ ) (80)
dβ
π − 2a − r + 1
1−δ ·δa

and
(( ) )
p·(1−γ )
a· + 1
1−δ ·a−r(1−p)δ
dS
dβ
≥( ( ) )2 ≡ b(γ ), (81)
p
π − a − r + 1−δ + (1−p)δ δ a
1

where both b̄(γ ) and b(γ ) are decreasing in γ . The insight in Lemma 1 then maintains. A
result similar to Proposition E.4 would then follow.
To summarize, allowing for mixed strategies would allow the mixed-strategy, Scenario-
3 equilibria to exist, in which the main insights from Scenario 2 would maintain, but with
more technical complexity. In light of this, we can rule out mixed strategies from the
Markov game, gaining in simplicity without losing much intuition.

F Relevance of Elites and People in Conflicts

There existed a wide range of conflicts in both Chinese and European histories. Having
carefully examined significant examples, we argue that the positions taken by the Elites
and the People were critical in determining the outcome of the conflict. Below we discuss
some examples.1
History has shown that given the Elites’ political, economic, and military resources,
whether they sided with the Ruler when the Ruler was challenged was critical to the out-
1 An incomplete list of the examples we examine include, for China, the Qin–Han turnover, Rebellion of
the Seven Prince States, Western Han–Xin turnover, Xin–Eastern Han turnover, Eastern Han–Three King-
doms turnover, Western–Eastern Jin turnover, Eastern Jin–Southern Dynasties turnover, Sui–Tang turnover,
Tang–Zhou turnover, An Lushan Rebellion, Huang Chao Rebellion and Tang–Five Dynasties and Ten King-
doms turnover, Northern–Southern Song turnover, Yuan–Ming turnover, Ming–Qing turnover, and Revolt of
the Three Feudatories; for Europe, the Rebellion of Robert de Mowbray, Henry I’s invasion of Normandy,
1215 Magna Carta, Second Barons’ War, Hundred Years’ War, Jacquerie, Wat Tyler’s Rebellion, Richard
II–Henry IV of England turnover, Jack Cade’s Rebellion, Wars of the Roses, German Peasants’ War, Dutch
Revolt, and Thirty Years’ War. Some examples include more than one entries of examination. These cover
15 and 14 entries for China and Europe, respectively, and 29 in total.

20
come of the challenge. For example, the fate of the French throne during the Hundred
Years’ War closely followed whether the Duke of Burgundy, first John the Fearless and
later his son Philip the Good, allied with the English or veered back to the French ruler
(Seward, 1978). During the Wars of the Roses (1455–1485), “crucially, Thomas, Lord S-
tanley, refused to answer Richard [III of England]’s summons” in the Battle of Bosworth
in 1485, and his brother “Sir William Stanley committed his men, tipping the battle de-
cisively in Henry [Tudor, later Henry VII of England]’s favour,” delivering the demise of
Richard III and the coronation of Henry VII (Grummitt, 2014, p. 123). In China, during the
civil war at the end of the Sui dynasty (611–618), Emperor Yang was killed in a coup by
Yuwen Huaji, the commander of the royal guard and the son of Duke Yuwen Shu; during
the late Tang dynasty, after Qiu Fu, Wang Xianzhi, and Huang Chao led peasants to revolt
all over the country (859–884), it was the regional governors, such as Wang Chongrong
and Li Keyong, who fought hard to recover Chang’an, defeated the uprisings, and restored
the throne of Tang.
The People’s position was more than often crucial, too, as we can see in the history
of not only China but also Europe. In Chinese history, in the final years of the Qin, Xin,
Sui, Tang, Yuan, and Ming dynasties, following the initial rebellion within the country or
invasion from the outside, peasants revolted and contributed to the end of these dynasties.
In Europe, for example, Morton (1938, p. 46, 63) comments on the English history: “the
king was able to make use of the peasantry in a crisis when his position was threatened
by a baronial rising,” and “even the strongest combination of barons had failed to defeat
the crown when, as in 1095 [Robert de Mowbray’s rebellion] and in 1106 [the challenge
of Duke Robert Curthose of Normandy over the throne of Henry I], it had the support of
other classes and sections of the population.” Finer (1997, p. 901) also observes that the
English fyrd, largely mobilized from the freemen, “was retained, and even called out by
the Norman kings against their rebellious Norman barons.” In the Hundred Years’ War,
the turning point toward the eventual French triumph was the rise of Joan of Arc, as she
inspired the common people of France to join the war.2 In England, shortly before and
during the Wars of the Roses, popular support was generally important in determining how
firmly Richard II, Henry IV, Henry VI, Edward IV, and Richard III could hold the throne
(e.g., Morton, 1938; Bennett, 1999; Grummitt, 2014).3 In the German Peasants’ War, as
2 For more details on the French throne’s lack of popular support before Joan of Arc, the change after that,
and the implications of the change on the development of the war, see Morton (1938) and Seward (1978).
3 During the Richard II–Henry IV turnover, “Richard found himself without supporters,” as the common

“merchants [have been] alienated” (Morton, 1938, p. 115); Grummitt (2014, p. 5) comments that “Lancas-

21
the status quo was challenged by peasants across southwestern Germany, the uprisings
were eventually defeated by the Swabian League, given that the support from the common
people in cities were inconsistent.
These examples show that both the Elites and the People are highly relevant in conflicts.
This gives us confidence to link the power structure among the Ruler and both the Elites
and the People to the stability of autocratic rule.

G Dynastic Cycle in Chinese History

Our model also provides an interpretation of the dynastic cycle, a prominent observation
in Chinese history. In brief, students of Chinese history often observed that each dynasty
started with a relatively stable autocratic rule, but would see over generations declining
trian legitimacy was based on an appeal to popular support,” and Bennett (1999, p. 204) states that “it was
widely believed that Henry had been raised to the throne on the basis of a covenant with the people.” During
Henry VI’s reign, he relied heavily on the support of the people, i.e., “the willingness of the political nation
to act for the common good,” but later in his reign “[t]he Commons could have little confidence in the king”
(Grummitt, 2014, p. 21). Over 1449–1454, “the defeat in France led to a popular groundswell of opinion a-
gainst the Lancastrian regime; the appeal to the commons that had been one of the foundations of Lancastrian
rule would, in part, prove its undoing” (Grummitt, 2014, p. 13, 21, 23); in 1450, Jack Cade’s Rebellion broke
out, during which Londoners played a decisive role by first siding with the revolt but later deserting them
for looting, eventually “shutting off Cade and his men . . . from the City” (Morton, 1938, p. 123); Grummitt
(2014, p. 161) comments that “[i]n 1450 politics were driven by an agenda that was unmistakably set by the
commons,” which would continue “[t]hroughout the following two decades” when Richard, Duke of York,
would have, “to all intents and purposes, become an opportunistic ‘Cadist’, jumping on a popular bandwagon
to end his self-imposed political exile” (Grummitt, 2014, p. 161, also 24–28). Following the strategy, Edward
IV’s taking of the throne depended on the support of the common people: “[t]he most significant aspect of
the series of events that led to Edward’s accession was . . . the judgment made by the assembled crowds at
St John’s Fields . . . that Henry VI . . . should be deposed for his recent misdeeds. . . . The assembled crowd
then merely assented to Edward’s de jure right to be king of England. . . . The notion of popular support for
the new king emerged as an important principle of Yorkist propaganda,” and “the Yorkists owed their suc-
cess in no small part to the support they enjoyed from the commons. . . . This popularity was also crucially
transformed into tangible expressions of support in terms of men and, more importantly perhaps, money”
(Grummitt, 2014, p. 71, 76–77). At that time, the importance of the people made them also a force to be used
by other political players: for example, “[m]ore than any of his contemporaries perhaps, [Richard Neville]
was able to exploit the idea of the commonweal and mobilise the commons’ sophisticated understanding
of the constitution and the nature and duties of kingship in support of his own aims” (Grummitt, 2014, p.
102). During the Edward IV–V–Richard III turnover, “[a]ll the chronicles agree that [Richard III’s] usurpa-
tion evoked no popular enthusiasm” initially (Wood, 1975, p. 270), but what followed proved again what
the common People could do: Buckingham’s rebellion failed because it “failed to mobilise popular support,”
while Richard III’s lack of popular support when he was killed in the decisive Battle of Bosworth suggested
that “[t]he disengagement from politics [of the common People] was probably the most damning indictment
of Richard III’s short reign” (Grummitt, 2014, p. 117, 124). In general, the Wars of the Roses “had witnessed
the incorporation of the commons as a legitimate partner with the king, lords and parliamentary Commons in
the process of politics and government" (Grummitt, 2014, p. 163).

22
power of the emperor, increasing dominance of the elites over the emperor and common
people, increasing concentration of land ownership, and decreasing effectiveness of gov-
ernance, eventually slipping to chaos and leading to the end of the dynasty (e.g., Skinner,
1985; Usher, 1989; Dillon, 1998).
To interpret the observation, guided by our model, we start by noting that the founding
emperor of each dynasty often enjoyed a strong absolute power, i.e., a low γ , since he was
bestowed with a high level of legitimacy by receiving himself the Mandate of Heaven to be
the new Ruler (e.g., Zhao, 2009; Jiang, 2011). By Proposition 3 and Online Appendix D,
he would have been more willing or able to restrict the asymmetry between the Elites and
People, maintaining a relatively high β ; as a result of the high β –low γ power structure,
the stability of autocratic rule would have been relatively high. Over generations, however,
later emperors became more and more distant from the act of receiving the Mandate of
Heaven; an increasing number of precedents also placed further constraints on their behav-
iors; all these led to a decline of the Ruler’s absolute power, i.e., a higher γ . By Proposition
3 and Online Appendix D, again, the later emperors would be less willing or able to en-
force a more symmetric Elite–People relationship, leading to a lower β . The low β –high
γ power structure would then lead to a lower stability of autocratic rule and, eventually, a
downward spiral to its collapse.

References
Bennett, Michael J. 1999. Richard II and the Revolution of 1399. Stroud: Sutton.
Dillon, Michael O. 1998. Dynastic cycle. In Dillon, Michael O. (ed.) China: A Historical
and Cultural Dictionary. London: Curzon Press, 87.
Finer, Samuel E. 1997. The History of Government from the Earliest Times, Volume II: The
Intermediate Ages. New York: Oxford University Press.
Grummitt, David. 2014. A Short History of the Wars of the Roses. London: Bloomsbury
Publishing.
Jiang, Yonglin. 2011. The Mandate of Heaven and the Great Ming Code. Seattle: Univer-
sity of Washington Press.
Morton, Arthur L. 1938. A People’s History of England. London: Lawrence and Wishart.
Reprint, 1999.
Seward, Desmond. 1978. The Hundred Years War: The English in France 1337–1453.
London: Penguin. Reprint, 1999.
Skinner, George W. 1985. Presidential address: The structure of Chinese history. Journal
of Asian Studies 44: 271–292.

23
Usher, Dan. 1989. The dynastic cycle and the stationary state. American Economic Review
79: 1031–1044.
Wood, Charles T. 1975. The deposition of Edward V. Traditio 31: 247–286.
Zhao, Dingxin. 2009. The Mandate of Heaven and performance legitimation in historical
and contemporary China. American Behavioral Scientist 53: 416–433.

24