ABSTRACT 
 
 
 
 
Title of Dissertation: THE POLITICS OF APPOINTING INSIDERS 
AND OUTSIDERS: APPOINTING 
PROVINCIAL GOVERNMENT AGENCY 
HEADS IN CHINA 
  
 Xiaonan Wang 
Doctor of Philosophy, 2022 
  
Dissertation directed by: Professor Margaret M. Pearson 
Department of Government and Politics 
 
 
 
 
My dissertation studies the politics of appointing insiders and outsiders in China?s provincial 
government agencies. This study is motivated by the following puzzles: why are most agency 
heads appointed from outside of the agency instead of from the inside, and why has the trend of 
appointing outsiders increased over time? This question is relevant beyond China, has important 
implications for the quality of government, and is related to the broad problem of bureaucratic 
authority and political control. I find that three existing theories of bureaucratic appointments ? 
the incentive-assignment conflict (the Peter Principle), patron-client relationships, and human 
capital or political experience ? cannot fully explain the patterns of appointing provincial 
government agency heads in China. I develop and empirically test a theoretical framework that 
draws insights from the literature on delegation and bureaucratic embeddedness. Provincial leaders 
 
 
face an information-collusion trade-off in appointing agency heads: while agency heads with prior 
experience inside the agency help reduce performance uncertainty, they can also use their 
information advantages to engage in corruption, especially the type of corruption that involves 
collusion with colleagues and business clients. Whether to appoint insiders or outsiders is a 
function of balancing the risk of performance uncertainty and collusive corruption. I argue that 
how provincial leaders balance the trade-off depends on two factors. First, provincial leaders? 
information on and connections with the candidates of appointees help mitigate the problems of 
adverse selection and moral hazard. When provincial leaders lack information and connections, 
they prioritize reducing corruption risks over performance uncertainty by appointing more agency 
heads without inside-agency experience. Second, according to the delegation model, increased 
monitoring capacity should alleviate provincial leaders? concerns about corruption risks at the 
appointment stage. However, authoritarian monitoring has distinctive features, including top-down 
control, uncertainty, and intra-party propaganda, that could lead provincial leaders to take more 
preventive measures by appointing more agency heads without inside-agency experience. Finally, 
I emphasize that the theoretical framework is more applicable to appointments in important 
agencies with high corruption risks.  
I built an original dataset of China?s provincial government agency heads appointed from 
1978 to 2020. To my best knowledge, this is the first systematic dataset on China?s sub-national 
government agency heads. I use provincial party secretaries? in-province time before making 
appointment decisions as a proxy for their information on and connections with sub-provincial 
bureaucrats. I leverage Xi Jinping?s anti-corruption campaign starting from late 2012 as an 
example that successfully strengthened the party?s monitoring capacity with the authoritarian 
approach. I categorize important agencies with high corruption risks based on both existing studies 
 
 
and a self-constructed index. I use the agencies that are not high-profile high-risk as the ?control? 
group. The empirical results provide supportive evidence to the theory. First, I find that when 
provincial party secretaries? in-province time is short, they are more likely to appoint agency heads 
without prior experience inside the agency in the high-profile high-risk agencies. Second, I find 
that after Xi Jinping?s anti-corruption campaign, provincial leaders appointed more agency heads 
without prior inside-agency experience in the high-profile high-risk agencies.  
This study makes four contributions. First, it contributes to the literature on bureaucratic 
embeddedness by providing and empirically testing a theoretical framework that explains how 
political leaders deal with the dilemma of bureaucratic embeddedness in appointments. Second, it 
contributes to the literature on delegation by unpacking how the mechanisms work differently in 
an authoritarian regime. Third, it contributes to the literature on state capacity by showing that 
authoritarian regimes face inherent contradictions in building state capacity. Finally, this study 
contributes to the literature on bureaucratic appointments in China by providing a strategic 
explanation for what hinders the Chinese bureaucracy toward a Weberian direction.     
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
THE POLITICS OF APPOINTING INSIDERS AND OUTSIDERS: APPOINTING 
PROVINCIAL GOVERNMENT AGENCY HEADS IN CHINA 
 
 
 
by 
 
 
Xiaonan Wang 
 
 
 
 
 
Dissertation submitted to the Faculty of the Graduate School of the  
University of Maryland, College Park, in partial fulfillment 
of the requirements for the degree of 
Doctor of Philosophy 
2022 
 
 
 
 
 
 
 
 
 
 
Advisory Committee: 
Professor Margaret M. Pearson, Chair 
Professor Ernesto Calvo 
Associate Professor Ethan Kaplan 
Associate Professor Ciqi Mei 
Associate Professor John F. McCauley 
Associate Professor Calvert W. Jones 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
? Copyright by 
Xiaonan Wang 
2022 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Acknowledgments 
 
Throughout my journey of pursuing a Ph.D., I am so lucky to be surrounded by people who are 
willing to guide me, help me, and inspire me in a way I can never repay. I want to thank the Chair 
of my dissertation committee and my mentor, Prof. Margaret Pearson. You gave me the freedom 
to explore but guided me in the right direction in each critical step. You not only provided 
suggestions of your own but also reached out for outside advice and opportunities on my behalf. 
You are honest and straightforward but also modest and encouraging. You helped me finish my 
dissertation. You also helped me start my academic career. I know that you have mentored many 
other students and junior scholars in the same way. I am grateful and proud that I can be one of 
them. I want to thank my advisor who inspired me to choose the academic road, Prof. Ciqi Mei. I 
came across two papers about China?s central-local relationship during my undergraduate. While 
all the authors are economists, I developed a vague interest in political science. It is you, your 
research, and your passion for your job that turned my vague interest into a committed choice. You 
showed me that political science and political scientists are pretty cool. I want to thank Prof. 
Ernesto Calvo. You always have creative ideas and approaches in doing research. I still remember 
your suggestion to get ?large, strange, informative, and underexplored datasets? in my first 
committee meeting with you and Margaret. You generously spent your time and provided 
resources to help me and other students get out of our comfort zone to learn new things. I want to 
thank ?the? economist on my committee, Prof. Ethan Kaplan. I only started to understand many 
fundamental issues in econometrics until taking your course. Your slides have been an invaluable 
resource to me ever since. You and Margaret brought together a group of Ph.D. students interested 
ii 
 
 
 
in Chinese political economy in the Department of Economics and the Department of Government 
and Politics to present ongoing projects each semester. I presented two chapters of my dissertation 
and got amazing feedback from you and the other participants. I want to thank Prof. John 
McCauley. Your comments on my early draft led me to expand the analysis in what later became 
my job market paper. Besides my dissertation, it was such a great experience to learn from you 
and work with you on the project of Chinese investment in Africa. I am excited that we will 
continue working on the project in the future. I want to thank Prof. Calvert Jones. My dissertation 
benefited enormously from your bright ideas and feasible suggestions. You even dissected the 
reviewers? comments for me. You led me to the comparative perspective and literature I previously 
neglected. It is my great pleasure to have you all as my committee members. I know I am eternally 
indebted to your wisdom and kindness. 
I am also indebted to many other professors and Ph.D. colleagues in the Department of 
Government and Politics. I will miss the time we had seminars and workshops together. I will miss 
the time we formed study groups to learn methods. I will miss the time we came to the office to 
study, chat, and help each other. Amid these everyday interactions, I grew up bit by bit: passing 
the comprehensive exams, passing the prospectus, and finally finishing and defending my 
dissertation.  
I want to thank my partner Yuruo. You are the first reviewer of my research. As a Ph.D. in 
epidemiology, your comments often make me wonder about the value of a Ph.D. in political 
science. But your intellectual contribution to my research, although significant, is not why I want 
to be with you. Life with you is simple but full of joy. 
Finally, I want to thank my parents. I cannot devote myself to pursuing a Ph.D. in the United 
States without your understanding and unconditional support. You support me because this is my 
iii 
 
 
 
choice, despite it may not be the choice you prefer. You support my choice because eventually, 
you want me to be happy. I do not take it for granted. For your understanding and unconditional 
support, I will be eternally grateful.      
iv 
 
 
 
Table of Contents 
 
Acknowledgments ii 
List of Tables viii 
List of Figures x 
Chapter 1: Introduction 1 
1.1 ?Whatever the Party Asks Me to Do? 1 
1.2 Provincial Government Agency Heads in China: 1978-2020 6 
1.3 Argument in Brief 13 
1.4 Contributions 22 
Chapter 2: Theory 28 
2.1 Why Appoint Agency Heads from the Outside: Existing Explanations 28 
2.2 The Delegation Problem 42 
2.3 The Appointment Dilemma 47 
2.4 Information, Authoritarian Monitoring, and the Appointment Dilemma 54 
2.5 Scope of the Theory: Not All Agencies are Created Equal 62 
Chapter 3: In-Province Time and the Appointment Dilemma 69 
3.1 Main Results 69 
3.2 Robustness Checks 81 
3.3 Alternative Explanations 84 
3.4 Summary 91 
Chapter 4: Authoritarian Monitoring and the Appointment Dilemma 94 
4.1 Main Results 94 
v 
 
 
 
4.2 Robustness Checks 101 
4.3 Alternative Explanations 106 
4.4 Summary 111 
Chapter 5: Outsiders in the Low-Profile Low-Risk Agencies 113 
Chapter 6: Conclusion 125 
Appendix A: Appendix for Chapter 1 134 
A.1 Agency in the Dataset 134 
A.2 Relevant Experience outside of the Agency 135 
A.3 Data Collection Process 136 
A.4 Descriptive Statistics 137 
Appendix B: Appendix for Chapter 2 138 
B.1 Coding Whether Agency Heads are Promoted into Their Positions 138 
B.2 Coding Patron-Client Relationships 139 
B.3 Categorizing Agencies 139 
B.4 Agency Insiders and Corruption Risks 142 
Appendix C: Appendix for Chapter 3 143 
C.1 Robustness Checks: Difference-in-differences Effects by Agency 143 
C.2 Robustness Checks: Appointments to Replace Retired Agency Heads 144 
C.3 Robustness Checks: Alternative Time Periods 146 
C.4 Robustness Checks: Tenure in Office and Appointments of Agency Insiders 150 
Appendix D: Appendix for Chapter 4 152 
D.1 Robustness Checks: Before-After Effects by Agency 152 
D.2 Robustness Checks: Before-After Effects by Province 153 
vi 
 
 
 
D.3 Robustness Checks: The Supply Shock 155 
D.4 Robustness Checks: Staggered Adoptions Using Central Inspections 157 
D.5 Robustness Checks: Replace the Dependent Variable as Relevant Experience 160 
Bibliography 164 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
vii 
 
 
 
List of Tables 
 
Table 2. 1: Summary of the Theoretical Framework 61 
 
Table 3. 1: In-Province Time and Appointments of Agency Insiders (1984-2012) 72 
Table 3. 2: Cross-Sectional Effects of In-Province Time on Appointing Agency Insiders 75 
Table 3. 3: Marginal Effects of In-Province Time on Specialists Outside of the Agency 91 
 
Table 4. 1: Before-After Changes 96 
Table 4. 2: Average DID Effects 100 
 
Table 5. 1: Before-After Effects (1998 Administrative Reform) 122 
 
Table A. 1: Provincial Government Agency in the Dataset 134 
Table A. 2: Relevant Work Experience and Education Backgrounds 135 
Table A. 3: Descriptive Statistics 137 
 
Table B. 1: Criteria of Promotions from Other Positions to Agency Heads 138 
Table B. 2: Promotion Links between Provincial Party Secretaries and Agency Heads 139 
 
Table C. 1: Appointments to Replace Retired Agency Heads (1984-2012) 144 
Table C. 2: Cross-Sectional Effects for Appointments to Replace Retired Agency Heads 145 
Table C. 3: Agency Heads Appointed Between 1978 and 1983 146 
Table C. 4: Cross-Sectional Effects for Appointments Made Between 1978 and 1983 147 
Table C. 5: Agency Heads Appointed Between 2013 and 2020 148 
Table C. 6: Cross-Sectional Effects for Appointments Made Between 2013 and 2020 149 
Table C. 7: Tenure in Office and Appointments of Agency Insiders 150 
Table C. 8: Cross-Sectional Effects of Tenure in Office 151 
 
viii 
 
 
 
Table D. 1: Before-After Effects Excluding the Cases with Deputy Heads Removed 155 
Table D. 2: Average DID Effects Excluding the Cases with Deputy Heads Removed 156 
Table D. 3: Average DID Effects Using Central Inspections 158 
Table D. 4: Before-After Effects Using Relevant Experience as the Dependent Variable 160 
Table D. 5: Average DID Effects Using Relevant Experience as the Dependent Variable 161 
Table D. 6: Before-After Effects of Agency Outsiders with Relevant Work Experience 162 
Table D. 7: Average DID Effects of Agency Outsiders with Relevant Work Experience 163 
 
 
ix 
 
 
 
List of Figures 
 
Figure 1. 1: Illustration of Appointing Provincial Government Agency Heads 9 
Figure 1. 2: Prior Work Experience by Agency (1978-2020) 10 
Figure 1. 3: Prior Work Experience by Time (1978-2020) 12 
Figure 1. 4: Provincial Party Secretaries? In-Province Time (1984-2020) 19 
 
Figure 2. 1: Proportion of Agency Outsiders Promoted into the Positions (1978-2020) 31 
Figure 2. 2: In-Province Time between Agency Insiders and Outsiders (1984-2020) 34 
Figure 2. 3: Patron-Client Relationships (2005-2020) 35 
Figure 2. 4: Agency Heads? Ages (2005-2020) 36 
Figure 2. 5: Agency Heads? Highest Education Levels (2005-2020) 39 
Figure 2. 6: Proportion of Agency Heads with Political Experience (1984-2020) 40 
Figure 2. 7: Indexes of Government Agencies 66 
Figure 2. 8: Agency Heads Investigated for Corruption (2005-2020) 68 
 
Figure 3. 1: Short In-Province Time and Appointments (Dynamic DID Effects) 79 
Figure 3. 2: Average DID Effects and Randomization Inference 80 
Figure 3. 3: Patron-Client Relationships and Appointments (2005-2012) 87 
Figure 3. 4: In-Province Time and Age of the Agency Heads 88 
Figure 3. 5: In-Province Time and Agency Heads? Highest Education Levels 89 
Figure 3. 6: In-Province Time and Agency Heads? Political Experience 90 
 
Figure 4. 1: Proportion of Having Inside-Agency Experience (2005-2020) 98 
Figure 4. 2: Nonparametric Dynamic DID Effects (2005-2020) 99 
x 
 
 
 
Figure 4. 3: Broadly Defined Patron-Client Relationships Before and After the Campaign 107 
Figure 4. 4: Strictly Defined Patron-Client Relationships Before and After the Campaign 108 
Figure 4. 5: Higher Human Capital? (Age) 109 
Figure 4. 6: Higher Human Capital? (Master?s degree and above) 110 
Figure 4. 7: Higher Human Capital? (Doctorate degree) 111 
Figure 4. 8: More Political Experience? 112 
 
Figure 5. 1: Proportion of Having Inside-Agency Experience (1990-2005) 119 
Figure 5. 2: Before-After Effects by Agency (1998 Administrative Reform) 120 
 
Figure B. 1: Officially Set Personnel Quota in Each Agency 140 
Figure B. 2: Alternative Measurements of the High-Risk Index 141 
 
Figure C. 1: Average DID Effects by Agency 143 
 
Figure D. 1: Before-After Effects by Agency 152 
Figure D. 2: Before-After Effects by Province 153 
Figure D. 3: The Relationship between ?Cleanness? and Provincial-Specific Changes 154 
Figure D. 4: Nonparametric Dynamic DID Effects Using Central Inspections 157 
Figure D. 5: Average DID Effects Using Central Inspections and Randomization Inference 159 
 
xi 
 
 
 
Chapter 1: Introduction 
 
1.1 ?Whatever the Party Asks Me to Do? 
Tang Zhihong was the former head of the Huanggang City Health Commission in Hubei Province. 
Her name probably would not be known to the public if not for the Covid-19 pandemic and a live-
streamed broadcast. When the virus hit Hubei in January 2020, a central inspection team was sent 
to Huanggang. When questioned by the inspection team, she was caught ignorant about the number 
of patients and the capacity of local hospitals. Tang was dismissed the next day. The scene was 
broadcasted by China Central Television, the most important state-owned broadcaster controlled 
by the Chinese Communist Party (CCP). It soon became a widely discussed topic among the public. 
Her previous experience caught particular attention ? she had not worked inside the agency or had 
any relevant experience before. She had a bachelor?s degree in law. After graduation, she worked 
in the Bureau of Justice, the Legal Office, and the Bureau of Archives. She was appointed as a 
deputy county mayor before being promoted to the head of the Huanggang City Health 
Commission.  
The case of Tang Zhihong is not unique. Liu Yingzi, the head of the Hubei Province Health 
Commission, had no agency-specific or relevant experience either. She was also dismissed during 
the pandemic. Similar cases are also not unique in Hubei Province or the Health Commission, an 
agency commonly viewed as relatively low-profile. Ouyang Bin was appointed as the head of the 
Hunan Province Department of Transportation in 2006. His educational background was in 
Chinese literature and history. He worked mostly as a generalist, including in the Hunan Party 
1 
 
 
 
General Office and as prefecture-level city party secretaries or mayors, with three years in the 
Hunan Province Department of Agriculture. After the collapse of an unfinished bridge which 
caused 64 deaths,1 he was punished and soon transferred to the Provincial People?s Consultative 
Conference, a position with no real authority. While whether their lack of relevant experience 
should be blamed for the mistakes is debatable, the phenomenon of ?non-professionals leading 
professionals? is often called into question when things go wrong. 
However, it is also easy to find where ?professionals lead professionals? in many other cases. 
You Qingzhong served as the head of the Jiangsu Province Department of Transportation from 
2008 to 2017. He graduated with a bachelor's degree in transportation engineering. After 
graduation, he worked in a state-owned corporation in charge of building transportation 
infrastructure for 20 years. From 2002 to 2008, he was the deputy head of the Jiangsu Department 
of Transportation. In some other cases, while the agency heads are appointed from outside of the 
agency, they have relevant experience. Duan Yufei was appointed as the head of the Guangdong 
Province Health Commission in 2016. Previously, he was the director of the Guangdong Food and 
Drug Administration. His primary work experience was split among a vocational medical school, 
the Health Commission of a city, and generalists? positions. 
The appointments of Tang Zhihong, Liu Yingzi, Ouyang Bin, You Qingzhong, and Duan 
Yufei represent divergent yet common career paths of typical bureaucrats in the party-state. The 
party controls appointments in the party branch and government agencies, state-owned 
corporations, public universities and hospitals, and other public organizations. It is common in 
Leninist systems for bureaucrats to have interweaving career paths in multiple organizations. As 
 
1 A bridge over the Tuojiang River in Hunan province collapsed on August 27, 2007. The collapse was identified as a 
man-made mistake and 32 officials were punished. See ?32 officials disciplined over bridge collapse?, China Daily, 
December 26, 2007. 
2 
 
 
 
Kornai (1992, p.39) describes, ?a factory manager may be promoted to city party secretary and 
then return to the state administration as a deputy minister. Conversely, he or she may begin as a 
party secretary, continue as a chief executive of a large company or a chief of police, and later fill 
a high office in the party?. Party members are expected to do whatever the party asks them to do 
? when assigned to unfamiliar positions, they are required to obey the order and quickly adapt to 
the new positions.2 The question is why the party assigns certain people to certain positions. The 
cases above suggest that the party?s appointment strategy is variegated. Why does the party 
sometimes assign a job to one without any relevant experience, sometimes to one with only 
relevant experience outside of the organization, and sometimes to one with prior experience inside 
the organization? 
This study attempts to explain the reasons behind the appointments of China?s provincial 
government agency heads in the reform era. Specifically, I aim to develop and empirically test a 
theoretical framework that helps understand the variations of appointment decisions.3 For example, 
why were Ouyang Bin, a generalist who had never worked in the related sectors, and You 
Qingzhong, a transportation specialist who had worked inside the agency for six years, both 
appointed as the head of the Department of Transportation at around the same time? Why did the 
successor of You Qingzhong, who was appointed in September 2017, spend most of his career in 
the Department of Water Resources without prior experience inside the agency?   
How bureaucrats are appointed is an important question to scholars of bureaucracy and the 
quality of government. First, the question of political appointments is relevant beyond China. Even 
in countries with independent civil service institutions, politicians can appoint many bureaucrats 
 
2 A popular party slogan is that ?I am a brick ready to be moved to wherever is needed?. 
3 My argument is informed by both related literature and a small number of elite interviews I conducted in 2020 and 
2021. Due to the travel restrictions during the Covid-19 pandemic, the interviews were conducted virtually. The 
interviews are approved by the University of Maryland College Park IRB (reference number: 1679526-1).  
3 
 
 
 
and they do not always follow Weberian standards (Grindle 2012). In post-communist countries, 
bureaucratic appointments still resemble more to their communist past (Goetz and Wollmann 2001, 
Scherpereel 2004, Meyer-Sahling 2008). In the United States, new presidents can fill close to 4000 
positions and many are political appointees appointed from outside of the agency (see a review 
article by Lewis 2011). When Hurricane Katrina hit the United States in 2005, the Federal 
Emergency Management Agency (FEMA) was filled with political appointees with very little 
experience in emergency management (Lewis 2008, p.160). Similar cases are so common that the 
federal government of the United States is called the ?government of strangers? (Heclo 1977) or 
?amateur government? (Cohen 1998). As political leaders share some similar problems when 
appointing bureaucrats, studying how provincial agency heads are appointed in China could shed 
light on the theory of political appointments in general.  
Second, agency heads? prior experience shapes their skills and preferences (Krcmaric, et al. 
2020) and leaders matter a lot for the quality of governance when institutions are weak (Jones and 
Olken 2005, Carreri 2021). While the CCP has tried to improve the institutionalization of the 
bureaucracy from time to time, there is an enduring force within the party to de-institutionalize 
and emphasize the role of flexibility, revolutionary spirit, and personal morality instead (Heilmann 
and Perry 2011, Zhi and Pearson 2017, Ding and Thompson-Brusstar 2021). In such contexts, as 
shown by the cases of Tang Zhihong, Liu Yingzi, and Ouyang Bin, which type of leader is selected 
has a great deal of influence on how government operates. The CCP cares about governance 
because its legitimacy hinges on performance (Zhao 2009, Zhu 2011, Yang and Zhao 2015). 
Selecting good leaders is always a priority of the party. The party has a sophisticated and rigorous 
procedure for evaluating and selecting leaders (Manion 1985, Burns 1994), and ?the selection of 
4 
 
 
 
bureau leaders is rarely random or haphazard? (Kostka 2013, p.56). However, the reasoning behind 
job assignments remains a black box. It is thus important to unpack the logic.        
Finally, whether agency heads have inside-agency or relevant experience relates to the 
general problem of bureaucratic authority and political control. According to Weber (1978), what 
differentiates modern from pre-modern states is the establishment of a merit-based and specialized 
bureaucratic organization independent from political interference. The growing complexity of 
governance in modern society makes such a professional organization indispensable. However, 
Weber (1978) also warns that advanced knowledge in specialized areas is a source of independent 
political power. If non-elected bureaucrats drift away from elected politicians or even dominate 
them, it conflicts with the principle of democracy. The tension exists similarly in non-democratic 
regimes. Lenin (2015[1918], p.157) describes bureaucrats as ?privileged persons divorced from 
the people and standing above the people.? Meanwhile, however, the revolutionary party needs a 
professional bureaucracy to govern. How political leaders balance the goal of using bureaucrats? 
knowledge and maintaining political control has important implications for state capacity and 
control by political leaders. 
The CCP?s cadre policy in the Mao era was a continuous struggle between the ?reds?, who 
were revolutionaries with little specialized knowledge, and ?experts?, who were intellectuals 
whose loyalty to the party?s ideology is suspicious (Harding 1981, Lee 1991). During the Cultural 
Revolution, both the ?reds? and ?experts? were severely attacked by the mass as the bureaucratic 
institutions they represented were considered to have betrayed the interests of the people (Lee 
1978). When Deng Xiaoping took power and started the reform in the 1980s, he tried to rebuild 
the bureaucracy and pushed for specialization. On multiple occasions, he urged party leaders to 
respect and identify those with specialized skills and assign them to positions where they can apply 
5 
 
 
 
their talents. For example, he said that ?the four modernizations cannot be achieved merely by 
keeping to the socialist road; we must also master professional knowledge and skills. No matter 
what job a person has, he must acquire the specialized knowledge it entails and become 
professionally competent; those who fall short of this standard must study.? (Deng 1980). It seems 
that after the Cultural Revolution destroyed China?s state capacity, the party attempted to rebuild 
the bureaucracy in a Weberian direction. However, the cases of Tang Zhihong, Liu Yingzi, and 
Ouyang Bin suggest that placing appointees in positions that match their skills is certainly not the 
only consideration when the party makes appointment decisions. 
Before presenting the summary of my argument, Section 1.2 introduces the data, the 
institutional background, and the pattern of provincial government agency heads appointed in the 
reform era.  
   
1.2 Provincial Government Agency Heads in China: 1978-2020 
I built an original dataset of China?s provincial government agency heads appointed between 1978 
and 2020. To what is my best knowledge, this is the first systematic dataset on China?s sub-national 
government agency heads. The dataset includes the agencies that are part of the provincial 
government organs, which is like a cabinet in the democratic systems. There have been nine 
administrative reforms since 1978, with old agencies terminated and new agencies created.4 In 
particular, the reform in 1998 significantly downsized the government and the number of cabinet 
agencies dropped from 40 to 29. Since 2018, the number of cabinet agencies is 26.5 The dataset 
 
4 See Lan (2000) for a brief history of China?s administrative reforms. 
5 See the website of the State Council of the People?s Republic of China, http://english.www.gov.cn/. 
6 
 
 
 
includes 21 agencies that are relatively stable in the cabinet departments.6 Table A.1 in Appendix 
A lists the names of the agencies and their policy jurisdictions and core functions. 
For these 21 agencies, the dataset contains all the agency heads (N=3689) appointed between 
1978 and 2020 in 22 provinces.7 I collected detailed information on their career paths, education, 
and other demographic variables. The main variables of interest are whether agency heads have 
prior work experience inside the agency and relevant work experience from the outside. I was able 
to find these two pieces of information through various sources for most agency heads. Due to data 
availability, the dataset only has systematic information on the age and the educational 
backgrounds of those appointed between 2005 and 2020. Table A.2 in Appendix A lays out what 
counts as relevant work and education experience for each agency. Section A.3 explains how I 
constructed the dataset in detail. Table A.3 reports the descriptive statistics. 
A theory of appointments requires a clear specification of the actors in charge of the 
appointments, the pool of candidates, and how vacancies are created. The authority of appointing 
local bureaucrats has been delegated to the sub-national government since 1984. Provincial 
government agency heads are appointed by the Provincial Party Standing Committee, the highest 
leadership group of the province. The standing committee normally has 10 to 15 members,8 with 
 
6 Agencies that are canceled in the early period or created recently are not included in the dataset. Some have changed 
names (often associated with some changes in functions) and some related agencies have been merged. In such cases, 
I treat them as a single agency before and after the changes. The Department of Public Security is not included because 
heads of Public Security often hold concurrent appointments with higher administrative ranks. See Section A.1 in 
Appendix A for details. 
7 Agency leaders in the four municipalities (Beijing, Shanghai, Tianjin, and Chongqing) and five autonomous regions 
of ethnic minorities (Tibet, Xinjiang, Inner Mongolia, Ningxia, and Guangxi) are not included in the dataset. The 22 
included provinces are more comparable in terms of the political status of the provincial leaders (e.g., provincial 
leaders of the four municipalities have higher political ranks than their peers in the 22 provinces) and the social 
conditions. Concurrent appointments (e.g., a deputy governor concurrently appointed as the head of the Finance 
Department) are not counted in the 3689 agency leaders. 
8 The Provincial Party Standing Committee often includes the Provincial Party Secretary, Governor, Deputy Party 
Secretary, Secretary of the Disciplinary Inspection Commission, Head of the Organization Department, Secretary of 
the Political and Legal Affairs Commission, Head of the Propaganda Department, Head of the United Front Work 
Department, and Secretary-General of the Party General Office.     
7 
 
 
 
Provincial Party Secretary being the ?number one? leader. While appointment decisions are 
collectively approved by members of the standing committee, Provincial Party Secretary is the 
most influential decision-maker. The head of the Organization Department also plays an important 
role as the department is directly in charge of personnel management. The Organization 
Department can recommend a list of candidates to the Provincial Party Secretary, but ultimately, 
the Provincial Party Secretary can make the final decision. Members of the Provincial Party 
Standing Committee are appointed and held accountable by the party center ? the Politburo 
Standing Committee of the CCP. The Provincial Party Secretary takes primary responsibility when 
things go wrong in appointments. Therefore, I also collected a dataset of Provincial Party Secretary 
and matched each agency head to the one that oversaw the appointment.  
Each position in the party-state is associated with an administrative rank. The position of 
provincial agency head is associated with the Bureau Rank. Anyone with the Bureau Rank or with 
the Deputy Bureau Rank for over two years is eligible to be the candidate for the provincial agency 
head. For example, deputy agency heads, prefecture-level city (deputy) mayors and (deputy) party 
secretaries, deputy heads of party departments, deputy general-secretaries of the provincial 
General Office, managers of state-owned firms, presidents of public universities, and deans of 
hospitals could all be considered as the candidates. Most provincial agency heads are selected from 
bureaucrats in the same province. Only in occasional cases are provincial agency heads ?sent down? 
from the corresponding central ministries.  
Vacancies emerge when the incumbent agency heads retire or when they are promoted, 
transferred, or dismissed. The standard term of office is five years and agency heads are not 
allowed to serve more than two terms. The actual tenure of many agency heads is shorter than five 
years. The party sets a strict retirement age for each administrative rank. It is mandatory for 
8 
 
 
 
provincial agency heads to retire from leadership positions at the age of 60 ? they are required to 
?take a back seat? and many are placed on the committees in the Provincial People?s Congress or 
the Provincial People?s Consultative Conference. It is also common for agency heads to be 
transferred to lateral positions or promoted in the middle of the term. Occasionally, vacancies 
emerge because the incumbent agency heads are investigated for corruption. According to my 
dataset, vacancies are normally filled quickly. Figure 1.1 illustrates how provincial government 
agency heads are appointed.  
 
 
Figure 1. 1: Illustration of Appointing Provincial Government Agency Heads  
9 
 
 
 
What is the pattern of provincial agency heads appointed between 1978 and 2020? Figure 
1.2 and Figure 1.3 show the proportions of two mutually exclusive types of agency heads ? those 
who have prior work experience inside the agency and those appointed from outside of the agency 
but with relevant work experience. The two types combined are the proportion of agency heads 
with any prior relevant experience. Figure 1.2 reveals two interesting patterns.  
 
 
Figure 1. 2: Prior Work Experience by Agency (1978-2020) 
First, most agency heads (66 percent) have relevant work experience inside or outside the 
agency. Contrary to what the examples of Tang Zhihong and Liu Yingzi suggest, only 15 percent 
of the agency heads in the Health Commission do not have prior work experience in health-related 
areas. Many of those who are appointed from outside of the Health Commission have experience 
in managing hospitals or medical schools. Although in some agencies, such as the Family Planning 
10 
 
 
 
Commission (canceled in 2013), the Ethnic Affairs Commission, and the Department of Civil 
Affairs, many agency heads have no relevant work experience at all, these are the agencies that do 
not require highly specialized skills. It is worth noting that in the Department of Land and 
Resources (Department of Natural Resources after 2018) and Environmental Protection ? two 
agencies that require certain specialized knowledge ? the proportion of agency heads with relevant 
work experience is also low. I will discuss these two cases in Section 2.1.9  But overall, the 
phenomenon of ?non-professionals leading professionals? exists but does not dominate.  
Second, far fewer agency heads (38 percent) have prior work experience inside the agency. 
Except for the Department of Finance, the Development and Reform Commission, and the 
Department of Water Resources, more than half of the agency heads in the rest of the agencies are 
appointed from the outside. For example, while 83 percent of the agency heads in the Economic 
Affairs Commission (Economic and Information Technology Commission after 2009) have 
relevant work experience, only 33 percent have prior work experience inside the agency. It is 
common to appoint agency leaders from related positions in the government, party, or other public 
institutions outside of the agency. In the Department of Education, many agency heads are selected 
from university presidents. In the Department of Justice, many are appointed from the government 
sectors of public security and courts or the party department of the Political and Legal Affairs 
Commission. In the Department of Human Resources and Social Security and the Department of 
Culture, many are appointed from the related party departments, i.e., the Department of 
Organization and the Department of Propaganda. 
 
9 See Kostka (2013) for an analysis of the career backgrounds of provincial agency leaders of the Department of 
Environmental Protection. The pattern in Figure 1.1 is consistent with her data that shows approximately three-
quarters of the agency leaders are appointed from outside of the environmental protection sector. 
11 
 
 
 
Figure 1.3 shows how the patterns change over time. First, the bias against those with inside-
agency experience is quite consistent over time.  There is not a single period from 1978 to 2020 
where the proportion of newly appointed agency heads with inside-agency experience exceeds 50 
percent. Second, the overall level of specialization ? measured by the proportion of agency leaders 
with relevant work experience either inside or outside of the agency ? does not increase over time. 
Instead, there is a slight trend of decrease. Among the agency heads appointed between 1978 and 
1988, approximately 77 percent have relevant work experience. For those appointed between 2013 
and 2020, the proportion drops to 60 percent.  
     
 
Figure 1. 3: Prior Work Experience by Time (1978-2020) 
A closer look at the data shows that the decrease is mainly driven by the proportion of agency 
heads with inside-agency experience. Before 1998, 47 percent of agency heads have inside-agency 
12 
 
 
 
experience. After 1998, only 31 percent have previously worked inside the agency. A large drop 
occurred sometime between 1998 and 2002 and the proportion remains low afterward. However, 
the proportion of those with only relevant experience outside of the agency does not decrease. 
Before 1998, 27 percent of those appointed from outside of the agency have relevant work 
experience. After 1998, the proportion slightly increases to 30 percent.    
The two figures suggest that provincial leaders seem to bias against those with inside-agency 
but not relevant experience when making appointment decisions. Despite Deng Xiaoping?s push 
for specialization in the 1980s, the appointments of provincial agency heads did not move in that 
direction. Why do provincial leaders prefer to appoint those without prior inside-agency 
experience as agency heads? Against this general pattern, however, provincial leaders do 
sometimes appoint agency heads with inside-agency experience. What are the factors that can 
explain when provincial leaders appoint agency heads with inside-agency experience and when 
they appoint those without? Section 1.3 provides a sketch of my argument.  
 
1.3 Argument in Brief 
Existing theories of bureaucratic appointments provide three possible explanations for why 
provincial leaders sometimes appoint agency heads from outside of the agency. First, all 
bureaucratic organizations face an incentive-assignment conflict that could result in bureaucrats 
being assigned to positions outside of their original organizations (Fairburn and Malcomson 2001). 
Promotion is used as an incentive to make bureaucracy work. But promotion also serves the role 
of matching people with certain skills to corresponding positions. The two roles of promotion 
conflict with each other. If there are no vacancies inside an organization for those who ?need? to 
13 
 
 
 
be promoted, they might be promoted to vacancies in another organization. Second, patron-client 
relationships play an important role in appointments and provincial leaders may prefer to appoint 
agency heads who are connected to them. If those who have patron-client relationships with 
provincial leaders are from outside of the agency, provincial leaders might appoint agency heads 
from the outside. Third, high human capital and political experience could be crucial for leadership 
success. If those from outside of the agency have higher human capital or more political experience, 
provincial leaders may appoint agency heads from the outside. As I will elaborate and empirically 
demonstrate in Chapter 2, the above three explanations cannot fully explain why provincial leaders 
sometimes appoint agency heads from outside of the agency. More importantly, they do not take 
provincial leaders? incentives, constraints, and strategies into account. Therefore, they fail to 
provide a unified theoretical framework that explains when provincial leaders appoint insiders and 
when they appoint outsiders, and why.    
My argument is that provincial leaders face an information-collusion trade-off in appointing 
provincial government agency heads. Provincial leaders? appointment decisions are a function of 
balancing the goals of using agency heads? expertise to reduce performance uncertainty and 
limiting their appointees? information advantages to reduce corruption risks. How provincial 
leaders balance the trade-off depends on two factors ? their information on the candidates of 
appointees and the monitoring capacity on the appointed agency heads. First, when provincial 
leaders hold more information on and connections with sub-provincial bureaucrats, this knowledge 
can help mitigate the principal-agent problems of adverse selection and moral hazard. This can 
thus encourage provincial leaders to appoint agency heads with prior experience inside the agency. 
But when provincial leaders lack information on and connections with sub-provincial bureaucrats 
? a result of the fact that they themselves are frequently rotated between provinces ? they prioritize 
14 
 
 
 
appointing agency heads from outside of the agency to reduce the information asymmetry that 
might facilitate corruption. Second, increased monitoring capacity is expected to alleviate 
provincial leaders? concerns about corruption risks at the appointment stage. If true, provincial 
leaders are expected to appoint more agency heads with inside-agency experience when the 
monitoring capacity increases. However, I argue that authoritarian monitoring has distinctive 
features, including top-down control, uncertainty, and intra-party propaganda, that might lead 
provincial leaders to take more preventive measures by appointing more agency heads from 
outside of the agency. 
A useful theoretical framework for understanding appointment strategies is the delegation 
model (e.g., Epstein and O'halloran 1994, Huber and Shipan 2002). The key insight of the model 
is that bureaucratic expertise reduces the uncertainty of policy outcomes.10 Therefore, politicians 
should be better off if they delegate to bureaucrats. However, the problem of delegation is the 
information asymmetry between principals and agents: bureaucrats? true intentions (adverse 
selection) and actions (moral hazard) are unobservable to politicians, and they can take advantage 
of the information asymmetry to act against politicians? interests. Appointment strategies can be 
used as an ex-ante preventive measure to mitigate the problem (Str?m 2000, Lewis 2008, Samuels 
and Shugart 2010, Dewan and Hortala-Vallve 2011, Mart?nez-Gallardo and Schleiter 2015, Hassan 
2020).  
When appointing agency heads, provincial leaders in China face the following dilemma: 
while agency heads with inside-agency experience have both subject-area and agency-specific 
expertise which help reduce the uncertainty of performance, their information advantages also 
 
10 As will be elaborated in Chapter 2, the key to delegation is not to make policy outcomes stochastically better but to 
make outcomes less risky (Bendor and Meirowitz 2004). 
15 
 
 
 
create greater challenges for political control. A particular problem of political control is 
corruption (Rose-Ackerman 1978), which is also a prevalent problem in China?s bureaucracy since 
the economic reform (Manion 2004, Wedeman 2012, Pei 2016, Ang 2020). Corruption in China 
often takes the form of collusion among colleagues and with businesses. Agency heads with prior 
inside-agency experience have greater opportunities and capacities to engage in this type of 
collusive corruption, which might implicate provincial leaders who oversaw the appointments in 
the first place. Provincial leaders? appointment decisions thus depend on how they balance 
reducing performance uncertainty and corruption risks. When provincial leaders appoint an agency 
insider, they delegate the work to the one with agency-specific expertise at the expense of higher 
corruption risks. When they appoint an agency outsider, they take a preventive measure to reduce 
corruption risks at the expense of higher performance uncertainty. 
This information-collusion trade-off is one form of the embeddedness dilemma that troubles 
bureaucratic appointments in developing countries where monitoring is weak (see two review 
articles by Pepinsky, et al. 2017, Grossman and Slough 2022). When bureaucrats are assigned to 
their familiar positions, they could use their local knowledge to improve efficiency. However, 
embedded bureaucrats could also use their information advantages for private gains ? their 
entrenched networks and local knowledge make them more susceptible to capture and corruption.  
Bureaucratic embeddedness often refers to the relationship between bureaucrats and local 
communities, such as home places (e.g., Bhavnani and Lee 2018, Xu, et al. 2021) and ethnic 
connections (Hassan 2020). It is also sometimes defined as internal social proximity among 
bureaucrats (Bozcaga 2020) or bureaucrats? internal cohesiveness and external ties with industry 
groups (Johnson 1982, Evans 1995). In this study, I define bureaucratic embeddedness as the 
relationship between bureaucrats and their work environment. Agency heads with prior inside-
16 
 
 
 
agency experience present the dilemma of embeddedness because of their agency-specific 
knowledge, their connections with other bureaucrats inside the agency, and their ties with the 
administrative clients. The embeddedness is not a result of indigenous connections, such as home 
places, ethnicity, alumni, or geographic proximity, it comes from previous work experience in the 
organization. In the context of appointing provincial government agency heads in China, how do 
provincial leaders balance the risk of performance uncertainty and the risk of collusive corruption?  
The delegation model suggests that two factors ? the divergence of policy preferences and 
the oversight capacity ?  are important in determining whether politicians delegate to bureaucrats 
or micromanage themselves (see a review article by Huber and Shipan 2011). First, if the policy 
preferences between bureaucrats and politicians are close (the ?ally principle?), politicians tend to 
delegate (e.g., Epstein and O'halloran 1994, Epstein and O'halloran 1999, Huber and Shipan 2002). 
But if politicians can only guess bureaucrats? intentions without knowing them precisely, they are 
less likely to delegate (Bawn 1995, Bendor, et al. 2001, Bendor and Meirowitz 2004). Second, if 
politicians can more effectively control bureaucrats ex-post, they are expected to grant more ex-
ante discretion (e.g., Epstein and O'halloran 1994, Bawn 1997, Huber and Shipan 2002).  
I adapt the theoretical framework to explain when provincial leaders in China appoint agency 
heads with inside-agency experience and when they appoint those without. First, provincial leaders 
have different levels of information on sub-provincial bureaucrats when making appointment 
decisions. When appointing provincial leaders, the party center faces a similar problem of 
balancing local embeddedness and political control (Bulman and Jaros 2020). On the one hand, 
some provincial leaders are frequently rotated and even parachuted into the province without prior 
in-province experience. When provincial leaders are rotated to new provinces, they are only 
allowed to select agency heads from the local candidates. This is a deliberate arrangement to 
17 
 
 
 
prevent them from developing local power bases that could endanger the authority of the center. 
On the other hand, some provincial leaders are selected from the localists who have worked in the 
province for some time. In these cases, the party relies on their local knowledge to improve 
governance.  
The different ways of reshuffling provincial leaders have a secondary consequence on how 
provincial leaders appoint agency heads. For example, Ouyang Bin, the former head of the Hunan 
Province Department of Transportation, was appointed by Zhang Chunxian, the then Provincial 
Party Secretary of Hunan, in March 2006. Zhang Chunxian was appointed as the Provincial Party 
Secretary of Hunan in December 2005. He had never worked in Hunan before: he spent most of 
his career in the central government and was appointed to Yunnan Province for two years. Zhang 
stayed in the province for three months before making the appointment decision. In the case of 
You Qingzhong, he was appointed as the head of the Jiangsu Province Department of 
Transportation by Liang Baohua in March 2008. Liang was appointed as the Provincial Party 
Secretary of Jiangsu in October 2007, but he had been in the Jiangsu Provincial Party Standing 
Committee since December 1994.  
Figure 1.4 plots provincial party secretaries? in-province time when making each 
appointment decision between 1984 and 2020.11 I calculate their in-province time as the years 
between becoming the Provincial Party Standing Committee members and appointing agency 
heads. For example, when Zhang Chunxian appointed Ouyang Bin, the in-province time is three 
months. When Liang Baohua appointed You Qingzhong, the in-province time is 13 years and three 
months. Each dot in Figure 1.4 represents the appointment of an agency head. The data show a 
 
11 Agency heads appointed between 1978 and 1983 are excluded because the authority of appointing agency heads 
was delegated to the provincial level in 1984. 
18 
 
 
 
great deal of variation in terms of the in-province time. Approximately one-quarter of the 
appointments were made when provincial leaders? in-province time was shorter than two years 
and another one-quarter of appointments were made when the in-province time was longer than 
seven years. For agency heads appointed between 1984 to 2020, the average in-province time did 
not increase.  
 
 
Figure 1. 4: Provincial Party Secretaries? In-Province Time (1984-2020) 
When provincial leaders? in-province time is short, they have little information on the 
candidates. In addition, they do not have time to build patron-client relationships to align their 
interests with their appointees. In this scenario, provincial leaders might be reluctant to appoint 
those with inside-agency experience because they present greater challenges for ex-post control. 
Instead, provincial leaders might take a preventive measure by appointing agency heads without 
19 
 
 
 
inside-agency experience to reduce corruption risks ex-ante. For example, with a short in-province 
time of three months, Zhang Chunxian appointed Ouyang Bin, an agency outsider.  
On the contrary, when provincial leaders have worked in the province for some time, they 
are more likely to have repeated interactions with the candidates. This helps them know their 
appointees? character better. They also have more time to form ties with some of their appointees. 
In this scenario, provincial leaders might be able to find candidates with inside-agency experience 
whom they can trust, which allows them to rely on agency insiders? local knowledge without 
worrying about corruption risks. If true, they are expected to appoint more agency heads with 
inside-agency experience to improve government efficiency. For example, with a long in-province 
time of more than 13 years, Liang Baohua appointed You Qingzhong, an agency insider. Chapter 
3 tests the argument by studying how in-province time affects provincial leaders? decisions of 
appointing agency insiders and outsiders. 
Second, increased monitoring capacity is supposed to alleviate provincial leaders? 
appointment dilemma. If their appointees? actions can be more effectively monitored, provincial 
leaders are expected to take fewer preventive measures at the appointment stage because they can 
rely on ex-post monitoring to reduce corruption risks. The CCP is good at employing authoritarian 
tools ? such as anti-corruption campaigns ? to increase its monitoring capacity over local 
bureaucrats. On the one hand, according to the substitution effect between ex-post and ex-ante 
control, provincial leaders are expected to appoint fewer agency outsiders after anti-corruption 
campaigns. On the other hand, I argue that authoritarian monitoring relies on distinctive 
approaches such as top-down control, uncertainty, and intra-party propaganda. In such contexts, 
provincial leaders? primary objective is to preserve themselves from becoming victims of the 
campaign in a dangerous and uncertain period. This might lead them to appoint more agency 
20 
 
 
 
outsiders as a preventive measure to reduce corruption risks, so they can avoid being implicated 
by their appointees? misbehavior. Chapter 4 uses Xi Jinping?s anti-corruption campaign as an event 
that increased the CCP?s monitoring capacity. I test how the anti-corruption campaign changes 
provincial leaders? appointment strategy. As an illustrative example, the successor of You 
Qingzhong, appointed in September 2017, is an agency outsider. The two successors of Ouyang 
Bin appointed in March 2008 and October 2010 are both agency insiders, probably due to the then 
provincial party secretaries? longer in-province time. However, the three agency heads appointed 
after the anti-corruption campaign began in 2013 are all agency outsiders. 
Third, as Figure 1.2 shows, many agency heads without inside-agency experience have 
relevant experience outside of the agency instead. I find that even when pressures to reduce agency 
heads? information advantages are high ? such as when provincial leaders face short in-province 
time or authoritarian monitoring ? there is a workaround to maintain bureaucratic expertise and 
reduce corruption risks at the same time. Specifically, provincial leaders can appoint agency heads 
from outside of the agency with relevant skills.    
Finally, the trade-off of balancing performance uncertainty and corruption risks is more 
relevant for positions where both policy outcomes are important and corruption risks are high, 
such as the Department of Transportation. Therefore, the delegation framework is more applicable 
in explaining provincial leaders? appointment strategies in these high-profile high-risk agencies. 
What explains the appointments of agency insiders and outsiders in low-profile agencies with low 
corruption risks, such as the Department of Culture? As in any hierarchical organization, there is 
an incentive-assignment conflict: the need to reward someone with a promotion may end up 
placing him or her in a different organization. Low-profile agencies with low corruption risks are 
used to place those who need promotions. Chapter 5 studies how the 1998 administrative reform, 
21 
 
 
 
which reduced the number of agencies in the government cabinet from 40 to 29 (reduced supply 
of positions), affects the appointments of agency insiders and outsiders in different types of 
agencies. 
 
1.4 Contributions 
This study makes contributions to the literature on bureaucratic embeddedness, delegation, state 
capacity, and political selection in China. First, it contributes to the literature on bureaucratic 
embeddedness by studying how political leaders deal with the dilemma of bureaucratic 
embeddedness ? defined as the relationship between agency heads and their work environment in 
this study ? in appointments. A competent and uncorrupt bureaucracy should be both embedded 
and not captured (Evans 1995), but how politicians balance the trade-off in appointments remains 
understudied (Grossman and Slough 2022). Abundant evidence shows that politicians are aware 
and try to avoid the risks of embeddedness. There is a strict rule of ?avoidance? that forbids 
bureaucrats to be assigned to their home places in ancient Europe (Beik 1985), Imperial China 
(Miyazaki 1976), and contemporary India (Bhavnani and Lee 2018, Xu, et al. 2021). Although the 
CCP does not implement the strict rule of home places avoidance (Chung 2011), they have 
intentionally assigned local leaders and central ministers to unfamiliar places without their former 
colleagues (Huang 2002, Bulman and Jaros 2020, Chan and Fan 2021). Provincial leaders could 
use the strategy of avoidance by appointing agency heads without inside-agency experience. These 
preventive measures, however, might come at the cost of government efficiency.  
Existing theories of appointments developed in democratic countries focus on how political 
leaders balance bureaucratic expertise, ideological divergence, and political patronage (Geddes 
22 
 
 
 
1994, Lewis 2008). Theories of appointments developed in authoritarian regimes focus on how 
autocrats balance the trade-off of competence and loyalty to stay in power (Egorov and Sonin 2011, 
Zakharov 2016). The few studies on the dilemma of bureaucratic embeddedness in appointments 
focus on how top leaders in either democratic countries or authoritarian regimes appoint 
subnational leaders in a way to remain in power (Bulman and Jaros 2020, Hassan 2020). However, 
provincial leaders in China face different incentives and constraints. As subnational leaders in an 
authoritarian regime, they face neither electoral pressures nor coup threats. Instead, they are under 
top-down control. In the chain of top-down accountability, they could be held accountable from 
above for their appointment decisions. In addition, provincial leaders themselves are being 
strategically assigned and thus have different levels of information on and connections with sub-
provincial bureaucrats. This study contributes to the literature by providing a theoretical 
framework and empirical findings on how subnational leaders in an authoritarian regime balance 
the trade-off of bureaucratic embeddedness in appointments.  
Second, this study contributes to the literature on delegation by unpacking how the 
mechanisms work differently in an authoritarian regime. The literature suggests an ?ally principle? 
in delegation and the substitution between ex-post and ex-ante control (see a review article by 
Huber and Shipan 2011). Although the ideological conflicts between provincial leaders and agency 
heads in China are limited, the logic of the ?ally principle? still applies. If provincial leaders can 
find trusted agency insiders, it may alleviate their concerns about corruption risks and encourage 
them to appoint agency insiders to reduce performance uncertainty. Specifically, this study argues 
that repeated interactions and connections with subordinates have a positive effect on mitigating 
the problems of adverse selection and moral hazard. On the contrary, frequent rotations ? a 
common political tool used in many authoritarian regimes to control elites (Migdal 1988, Hill and 
23 
 
 
 
L?wenhardt 1991, Snyder 1992, Siegel 2018) ? exacerbate subnational leaders? appointment 
dilemma. This study echoes the recent findings that, contrary to the conventional wisdom (e.g., 
Pollock 1937, Xu 2018, Colonnelli, et al. 2020), close relationships between politicians and 
bureaucrats could help solve principal-agent problems and result in positive effects on economic 
development (Jiang 2018) and public service provision (Toral 2021a, b, Akhtari, et al. 2022).  
The delegation model also suggests that increased monitoring is supposed to be a solution 
to the appointment dilemma. One of the robust findings from the delegation literature is that 
politicians take fewer preventive measures ex-ante if they have more effectively ex-post control 
on bureaucrats. Similarly, the maleficent face of local embeddedness is less worrisome when the 
level of accountability is high (Bhavnani and Lee 2018). However, by studying how Xi Jinping?s 
anti-corruption campaign affects provincial leaders? appointment strategies, this study shows that 
the substitution between ex-post monitoring and ex-ante precaution might not apply to 
authoritarian monitoring. Instead, the distinctive features of authoritarian monitoring might lead 
provincial leaders to take more preventive measures. 
Third, this study contributes to the literature on state capacity by showing that authoritarian 
regimes face inherent contradictions in building state capacity. The CCP?s firm control of 
appointments and the ability to rotate party cadres to different positions are considered as one of 
the institutional foundations of China?s state capacity (Edin 2003, Xu 2011, Svolik 2012). The 
literature suggests that top-down control of personnel appointments ensures that party cadres at 
various organizations and administrative levels have the incentive to faithfully carry out the party?s 
policy priorities. However, I find that effective control of provincial leaders, such as through 
frequent rotations, creates governance challenges for provincial leaders. Provincial leaders who 
are parachuted into the province lack information on and connections with sub-provincial 
24 
 
 
 
bureaucrats, which leads them to prioritize reducing the risk of corruption over the risk of 
performance uncertainty when making appointment decisions. For provincial leaders in this 
scenario, filling vacancies is not an active tool to facilitate state capacity but a passive task that 
must be finished. 
 Another tool the CCP uses to maintain state capacity is campaign-style enforcement. Most 
prominently, the CCP has regularly resorted to anti-corruption campaigns to increase its 
monitoring capacity (Manion 2004, Wedeman 2005). Xi Jinping?s anti-corruption campaign has 
been considered as an example that successfully increased the party?s monitoring capacity using 
the authoritarian approach (Carothers 2022). However, this study highlights that the success of the 
anti-corruption campaign relies on authoritarian features that might result in secondary 
consequences. Specifically, under top-down control, uncertainty, and intra-party propaganda, the 
primary goal of provincial leaders is to preserve themselves from being victims of the campaign. 
They thus appoint more agency heads from outside of the agency to reduce corruption risks. While 
provincial leaders? cautiousness may be effective in preventing corruption, it may not help improve 
government efficiency. Furthermore, from the perspective of building state capacity, this strategy 
might be overly cautious and unnecessary because agency heads can be more effectively monitored 
than before. Meanwhile, I find that provincial leaders appoint outsiders with relevant skills as a 
workaround to maintain bureaucratic expertise. The strategy of appointing outsiders with relevant 
skills has broad implications for building state capacity in countries with both a high level of 
corruption and a low level of bureaucratic expertise.         
Finally, this study contributes to the literature on political selection in China by studying the 
appointments of provincial agency heads. Existing literature has mainly focused on the 
appointments of regional leaders (e.g., Li and Zhou 2005, Shih, et al. 2012, Jia, et al. 2015, Landry, 
25 
 
 
 
et al. 2018). Provincial government agency heads are equally important positions. For example, 
the Development and Reform Commission makes economic policies and approves major 
investment projects. The Health Commission manages hospitals and leads the response to public 
health crises. However, how government agency heads are appointed is understudied. When the 
question is studied, the scope is limited to the central ministers (Huang 2002, Chan and Fan 2021) 
or a specific agency at the provincial level (Kostka 2013). This study fills the gap.  
The new dataset provides the first-ever overall picture of the career backgrounds of 
provincial government heads appointed in the reform era. On the one hand, it clarifies the myth of 
?non-professionals leading professionals?. On the other hand, it also shows how China?s provincial 
government agencies are different from and have not moved toward a Weberian direction. As I 
will elaborate in Chapter 2, the conventional wisdom that focuses on structural or mechanical 
factors cannot fully explain the pattern of appointments. My theoretical framework provides a 
strategic explanation of what hinders China?s provincial government agencies toward a Weberian 
bureaucracy.  
The rest part organizes as follows. Chapter 2 first examines whether existing theories can 
fully explain why provincial leaders sometimes appoint agency heads from outside of the agency. 
I then elaborate on the theoretical framework that explains when provincial leaders appoint agency 
insiders and when they appoint agency outsiders, and why. I also discuss the scope of my theory. 
Chapter 3 and Chapter 4 provide empirical evidence that supports the theory. Chapter 3 uses 
provincial party secretaries? in-province time to test how information on and potential connections 
with sub-provincial bureaucrats affect provincial leaders? appointment strategy. Chapter 4 uses Xi 
Jinping?s anti-corruption campaign to test whether authoritarian monitoring leads provincial 
leaders to take fewer or more preventive measures in appointments. Chapter 5 discusses appointing 
26 
 
 
 
outsiders in the low-profile low-risk agencies. Chapter 6 concludes with how the theoretical 
framework and empirical findings help explain the overall patterns of appointments in Figure 1.2 
and Figure 1.3, and how they relate to the comparative literature beyond China. Chapter 6 ends 
with implications for building state capacity in authoritarian regimes. 
 
  
27 
 
 
 
Chapter 2: Theory 
 
As Chapter 1 shows, most provincial government agency heads in China have relevant experience 
but far fewer have inside-agency experience (Figure 1.2). The proportion of agency heads with 
relevant experience did not increase from 1978 to 2020 and the proportion of those with inside-
agency experience has decreased over time (Figure 1.3). Why do provincial leaders sometimes 
appoint agency heads from outside of the agency? Why does the trend of appointing agency 
outsiders increase over time?12 The phenomenon of appointing outsiders is not unique to China 
and who serves the leadership positions has significant impacts on the quality of governance. More 
broadly, the question is related to the problem of bureaucratic authority and political control. This 
chapter provides a theoretical framework that explains when and why provincial leaders appoint 
agency heads with and without inside-agency experience. Section 2.1 first examines whether three 
prominent theories on bureaucratic appointments can explain the phenomenon. From Section 2.2 
to Section 2.4, I develop the theoretical framework. Section 2.5 discusses the scope of the theory. 
 
2.1 Why Appoint Agency Heads from the Outside: Existing Explanations 
Existing theories provide three possible explanations for why provincial leaders appoint agency 
heads from the outside: an incentive-assignment conflict, patron-client relationships, and human 
capital or political experience. Yet, as I will show, none of them seem to fully explain the pattern 
of appointments. First, matching people with certain skills to corresponding positions are 
 
12 There are no formal rules that directed this change. 
28 
 
 
 
constrained by the incentive mechanism and the hierarchical structure of bureaucratic 
organizations. Promotion rather than monetary compensation is the prevalent form of incentive in 
hierarchical organizations (Baker, et al. 1988) because performance outcomes are often difficult 
to verify (Prendergast 1993) and actors at different layers of the organizations have misaligned 
interests (Fairburn and Malcomson 2001). In this context, promotion can be an effective incentive 
by binding future career prospects with the success of organizations. However, the incentive role 
of promotion conflicts with its role of job assignments because promotion is supposed to assign 
people to positions where they can best apply their skills. The conflict is called the Peter Principle: 
?in a hierarchy, every employee tends to rise to his level of incompetence.? (Peter and Hull 1969, 
p.25). The dual role of promotion and its incompatibility are well recognized in labor economics 
(Baker, et al. 1988, Milgrom and Roberts 1992, Fairburn and Malcomson 2001). Specifically, if 
one needs to be promoted as a reward for good performance, where he or she will be placed 
depends on where the vacancy is. The pyramid structure of hierarchical organizations means that 
the number of positions decreases as the rank of positions increases. It is thus difficult to promote 
someone to a higher rank position in the same organization.   
Provincial leaders in China do face such a challenge when appointing agency heads. The 
party-state is an overarching hierarchical organization that manages government agencies, party 
departments, regional leaders, state-owned firms, universities, hospitals, and other public 
institutions. The CCP has relied heavily on promotion to make the bureaucracy work. For example, 
the promotion incentive is an important driver for rapid economic growth in China?s reform era 
(see a review article by Xu 2011). In addition, controlling the job assignments in important 
occupations and maintaining the prospect of promotion is essential for regime stability (Svolik 
2012). The problem, however, is that the supply of higher rank positions is small relative to the 
29 
 
 
 
number of people waiting to be promoted (Kou and Tsai 2014). Because of this, the CCP has 
created new positions and administrative titles and has been reluctant to downsize the government 
(Burns 2003, Chan and Gao 2018). 
In theory, the conflicting role of promotion is a possible explanation for why some agency 
heads are appointed from outside of the agency. When someone needs to be rewarded with a 
promotion but there is no vacancy in that organization, he or she may be promoted to a higher rank 
position in another organization. If this is the case, most of the agency heads appointed from 
outside of the agency should be promoted into their positions. Based on their previous positions, I 
coded whether each agency head was promoted from a lower rank or less important positions of 
the same administrative rank.13 As a reward for those who ?need? to be promoted, promoting them 
to a higher administrative rank is more effective than promoting them to a more important position 
of the same administrative rank. On the one hand, material benefits and privileges are associated 
with administrative ranks. On the other hand, the CCP has set a rule of age limit for promotion 
since the 1980s. For example, party cadres are not eligible to be promoted to the Bureau Rank over 
the age of 55 (Kou and Tsai 2014). Although the rule is not strictly enforced, it creates pressure on 
career advancement. From the perspective of providing effective incentives, promotions need to 
be timely.  
However, among the agency heads without inside-agency experience, only 40 percent were 
promoted from a lower administrative rank (the Deputy Bureau Rank). If those promoted from 
less important positions of the same administrative rank (the Bureau Rank) are included, the 
proportion increases to approximately 66 percent. The latter proportion is higher because except 
for prefecture-level city mayors and party secretaries, most of the other positions associated with 
 
13 See Table B.1 in Appendix B for details on how promotions are coded. 
30 
 
 
 
the Bureau Rank are deemed as less important than government cabinet agencies. The data 
therefore do not support the claim that agency heads appointed from outside of the agency are the 
ones who ?need? to be promoted. 
 
 
Figure 2. 1: Proportion of Agency Outsiders Promoted into the Positions (1978-2020) 
Figure 2.1 further breaks down the proportion in each agency. There is a great deal of 
variation across agencies. In the Family Planning Commission, the Department of Audit, the 
Foreign Affairs Office, the Ethnic Affairs Commission, and the Department of Culture, 60 percent 
of the agency heads without inside-agency experience are promoted from a lower rank on average. 
The proportion including those promoted from less important positions is also high (77 percent on 
average). The explanation of incentive-assignment conflicts seems more applicable to these 
agencies (see Chapter 5). In all the other agencies, however, the proportion of the agency outsiders 
promoted from a lower rank is lower than 50 percent. Except for the Department of Education, the 
31 
 
 
 
Department of Science and Technology, and the Development and Reform Commission,14 the 
proportion including those promoted from less important positions is also lower. What else can 
explain why sometimes those without inside-agency experience are appointed as heads of these 
agencies? 
The second possible explanation is that agency heads without inside-agency experience have 
political connections with provincial leaders. On the one hand, politicians prefer those who are 
loyal to them personally. Autocratic leaders appoint incompetent loyalists because competent 
subordinates are threats to their power (Egorov and Sonin 2011, Zakharov 2016). A more positive 
view is that the existence of patron-client relationships between politicians and their appointees 
could help solve the principal-agent problems when formal institutions are weak (Jiang 2018, Toral 
2021a). As the pool of loyalists is often limited, if provincial leaders cannot find candidates who 
are loyal to them among those with inside-agency experience, they may select agency heads from 
those without.   
On the other hand, public offices could be used as quid pro quos and these exchanges are 
often based on existing political connections. One form of the quid pro quos is the patronage 
system: politicians reward party loyalists and campaign contributors with public offices in 
exchange for their political support in electoral times (e.g., Grindle 2012, Hollibaugh Jr, et al. 2014, 
Oliveros 2016, Colonnelli, et al. 2020, Brierley 2021, Oliveros 2021). This type of exchange does 
not apply to the Chinese context because the careers of provincial leaders in China do not depend 
on elections. Alternatively, public offices can be used in exchange for money. Office selling is 
prevalent in the pre-modern states (Swart 1949) and developing countries (Kristiansen and Ramli 
 
14 The Development and Reform Commission is an exception due to its significance relative to the other government 
cabinet agencies. Therefore, different from the other agencies, those appointed from prefecture-level city mayors and 
leaders of the other government cabinet agencies are coded as promotions from less important positions. The higher 
proportion is a result of the coding criterion. 
32 
 
 
 
2006, Weaver 2021). Buying and selling of offices are also found to be a common type of 
corruption in China?s bureaucracy (Zhu 2008, Pei 2016) and military (Wang 2016). Therefore, a 
potential explanation for why those without inside-agency experience are appointed is that they 
bribe provincial leaders to do that. Although the actual transactions are hard to observe, the 
literature shows that bribery and office selling in China often occur between two parties that 
already have some connections (Luo 2008, Li 2011, Zhan 2012, Wang 2016). 
Are agency heads without inside-agency experience more likely to have connections with 
provincial leaders ? either for loyalty or quid pro quos? Empirical evidence in China suggests that 
political connections do matter for political selection, but the evidence pertains mostly to the 
appointments of provincial leaders (Shih, et al. 2012, Jia, et al. 2015, Landry, et al. 2018), not their 
subordinates (Landry, et al. 2018). It is unclear whether and how political connections matter in 
the appointments of provincial government agency heads. As has been shown in Chapter 1, a 
crucial difference in bureaucratic appointments at the sub-provincial level is that provincial leaders 
? those who make appointment decisions ? are frequently rotated and some are parachuted into 
the province. Therefore, provincial leaders have different levels of information on sub-provincial 
bureaucrats.  
If agency outsiders are more likely to have political connections with provincial leaders, 
provincial leaders? in-province time before appointing agency outsiders can be expected to be 
longer than their in-province time before appointing agency insiders, because longer in-province 
time provides more opportunities to build connections. However, Figure 2.2 shows that provincial 
party secretaries do not have longer in-province time when they appoint agency heads without 
inside-agency experience.  
 
33 
 
 
 
 
Figure 2. 2: In-Province Time between Agency Insiders and Outsiders (1984-2020) 
To examine the pattern of patron-client relationships between agency heads and provincial 
leaders more directly, Figure 2.3 shows the proportion of agency heads who have prior patron-
client relationships with the provincial party secretaries in charge of their appointments. I follow 
existing studies that identify patron-client relationships through promotion links (Keller 2016, 
Jiang 2018). Specifically, if the agency head had been promoted by the same provincial party 
secretary before, then the two are coded as having a patron-client relationship. For example, Yu 
Xiangdong was appointed as the head of the Shandong Province Department of Land and 
Resources in August 2020. The Provincial Party Secretary in charge of this appointment is Liu 
Jiayi, who assumed the position in April 2017. Before this appointment, Yu Xiangdong had been 
promoted from the deputy party secretary (the Deputy Bureau Rank) to the mayor of a prefecture-
level city (the Bureau Rank) in July 2018, with Liu Jiayi in charge of the promotion as well. Yu 
34 
 
 
 
Xiangdong is thus coded as having a patron-client relationship with the then Provincial Party 
Secretary. I coded both promotion links from the Deputy Bureau Rank to the Bureau Rank (strictly 
defined promotion links) and more important positions within the Deputy Bureau Rank (broadly 
defined promotion links).15 Due to data availability, I only include the agency leaders appointed 
between 2005 and 2020.  
 
 
Figure 2. 3: Patron-Client Relationships (2005-2020) 
Figure 2.3 shows that among the agency heads appointed between 2005 and 2020, fewer 
than 14 percent have prior promotion links (strictly defined) with the then provincial party 
secretaries. The number increases to 21 percent if promotions to more important positions within 
 
15 See Table B.2 in Appendix B for details on how promotion links are coded. 
35 
 
 
 
the Deputy Bureau Rank are included. The data suggest that patron-client relationships are quite 
uncommon, which is not surprising considering provincial party secretaries? in-province time is 
not long on average. Agency heads without inside-agency experience are indeed slightly more 
likely to have patron-client relationships, but most agency outsiders are not politically connected 
with the then provincial party secretaries. In other words, many agency outsiders are appointed for 
reasons other than political connections. What else can explain these cases?  
 
 
Figure 2. 4: Agency Heads? Ages (2005-2020) 
The third possible explanation is that agency heads appointed from outside of the agency 
have higher human capital and more political experience. The qualifications required for good 
leadership can be debated. On the one hand, subject-area and agency-specific experience are 
important for agency management. On the other hand, outsiders may have characteristics that can 
36 
 
 
 
contribute to agency management in different ways. For example, compared with professional 
bureaucrats, political appointees in the United States have less subject-area expertise and public 
management skills but higher levels of education, political experience, or private sector experience 
(Lewis 2008). Political appointees are also viewed as more risk-taking, better connected to 
interested groups and constituencies, and having broader views (Moe 1985, Maranto 2001, Bok 
2003). In addition, prior career experience in certain organizations, such as political parties, is 
found to be positively associated with leadership performance (Goetz 1997, Grotz, et al. 2021).  
Since the 1980s, the CCP has shifted its cadre policy to prioritize promoting younger and 
better-educated over the revolutionary zealots (Zhou 2001). Education levels are an increasingly 
important criterion for promotion for prefecture-level city mayors and party secretaries (Yao, et al. 
2020). When appointing agency heads, if the candidates from outside of the agency are younger 
and better educated than those with inside-agency experience, provincial leaders might appoint the 
outsiders as agency heads for their higher human capital. However, this is not what the data suggest. 
I compare the age and the highest education level16 of agency heads appointed between 2005 and 
202017 with and without inside-agency experience. Figure 2.4 and Figure 2.5 demonstrate that at 
the time of appointments, the human capital of these two types of agency heads is barely different.  
The CCP has always attached great importance to political experience. Mao Zedong once 
said that ?non-professionals leading professionals is a general rule? (Mao 1958, p.107). In 1976, 
the ?People?s Daily? published an article entitled ?the party has to lead everything? to criticize 
Deng Xiaoping?s ?absurd? claim that those who do not have related expertise cannot lead the 
health work.18 Even in the reform era, the preference for generalists has not faded. In 2003, when 
 
16 The highest education level includes degrees obtained on the job. 
17 Systematic demographic information is missing for quite some agency leaders appointed in earlier periods. 
18 ?The party has to lead everything ? criticizing Deng Xiaoping?s reactionary speech that opposing the party to lead 
the health work?, People?s Daily, July 4, 1976, Page 3. 
37 
 
 
 
Xi Jinping was the Party Secretary of Zhejiang Province, he said that ?a good politician cannot 
just know one thing but must be versatile?, and ?although they are not experts in anything, they 
stand on a higher place, see farther ahead, and can grasp the general trend ? that?s how they can 
lead many experts?19. Therefore, appointing agency heads from outside of the agency might reflect 
the preference for political experience. It might be the CCP?s intention to build a bureaucratic 
organization different from the Weberian type. In such an organization, specialized knowledge is 
less important than loyalty and sensitivity to the party?s changing policy doctrine (Rothstein 2015).  
This logic may explain why few agency heads in the Department of Environmental 
Protection and the Department of Land and Resources have inside-agency experience (see Figure 
1.2). To solve environmental problems, the CCP does not resort to technical solutions but relies 
heavily on political commands and ?blunt force? to meet ?hard? targets (Kostka and Zhang 2018, 
Van der Kamp 2021). Under the campaign-style enforcement, specialized knowledge is less 
important than a ?strong? leadership style that can help meet the ?hard? targets (Kostka 2013). For 
example, a good agency head for the Department of Environmental Protection is the one who can 
make determined decisions to shut down factories. Similarly, one of the crucial tasks of the 
Department of Land and Resources is to implement the strictest policy to stop the decline of arable 
land. In 2006, the central government set a ?red line? that the total arable land must be no less than 
120 million hectares. When selecting agency heads for the Department of Land and Resources, the 
important consideration is ?whether the candidates have the political will to hold the ?red line? 
rather than whether they have relevant skills?20. 
 
 
19 Xi Jinping, February 12, 2003, in Zhejiang, see http://politics.people.com.cn/n/2014/1128/c1001-26110565-2.html 
20 Interview on December 24, 2020. 
38 
 
 
 
 
Figure 2. 5: Agency Heads? Highest Education Levels (2005-2020) 
I coded whether the agency heads had previously worked as prefecture-level city party 
secretaries (or deputy party secretaries), mayors (or deputy mayors), and whether they had worked 
in the General Office of the provincial government and party. Regional leaders ? either as party 
secretaries or mayors ? are important generalists? positions with a wide range of responsibilities. 
The deputy party secretaries or mayors also manage multiple issue areas. The General Office is a 
vital department in both the government and the party. Those working in the General Office serve 
as ?organizational secretaries? (to distinguish from ?personal secretaries?) of the leaders. They 
coordinate and control the work of different bureaus, gather information, organize meetings, and 
complete other tasks instructed by the leaders (Tsai and Liao 2018). Therefore, those who have 
worked as regional leaders or have worked in the General Office are expected to have broader 
perspectives and better political acumen, which might help them ?grasp the general trend?. They 
39 
 
 
 
are also more likely to have already been tested in difficult tasks. If an agency head had worked in 
one of the above positions, I coded him or her as ?having more political experience?.  
 
 
Figure 2. 6: Proportion of Agency Heads with Political Experience (1984-2020) 
Figure 2.6 shows that agency heads appointed from outside of the agency are indeed more 
likely to have more political experience. The proportion of agency heads with political experience 
is 57 percent among those without inside-agency experience and 25 percent for those with inside 
experience. The proportion is even higher in the Department of Environmental Protection and the 
Department of Land and Resources: 74 percent of agency outsiders have political experience in 
these two agencies. The pattern is quite consistent for agency heads appointed in different periods 
between 1984 and 2020, although there is a slightly increasing trend for agency insiders to also 
have political experience. The data thus provide some support for the political experience 
explanation. 
40 
 
 
 
However, the data also raise further questions. The CCP?s cadre policy is multifaceted (Ding 
and Thompson-Brusstar 2021). For example, Xi Jinping said in 2014 that ?there are two types of 
cadres that should be valued?: one is the ?permanent? local type, and the other is the ?flying pigeon? 
type21. He said that local cadres ?love their homeland and are enthusiastic, we should help improve 
their quality and broaden their perspective to make them better serve their homeland?. In the 
context of bureaucratic embeddedness defined in this study, agency heads with inside-agency 
experience are the ?permanent? local type, and agency heads appointed from outside of the agency 
are ?flying pigeons?. The preference for political experience thus reflects only one side of the 
party?s guideline in cadre selections. Moreover, political experience and agency-specific 
experience are not supposed to be mutually exclusive. Figure 2.6 shows that one-quarter of the 
agency heads have both inside-agency and political experience. An agency head can start his or 
her career in the agency and then work in generalist positions for several years before being 
appointed back to the same agency. Or an agency head can start his or her career in generalist 
positions and then work in an agency for a few years. These agency heads seem to be the ideal 
type as they have both political and agency-specific experience. The question thus remains: why 
do provincial leaders sometimes (62 percent of the time) appoint those without any inside-agency 
experience as agency heads? 
Each of the above explanations offers some answers but also leaves important questions 
unaddressed. The incentive-assignment conflict suggests that the need to reward those with 
promotions could end up assigning them to different organizations where vacancies exist. While 
the theory may explain the appointments in some agencies (Chapter 5), most agency outsiders are 
 
21 Xi Jinping cited the names of the two most popular bicycle brands in China, the ?Permanent? bicycle and the ?Flying 
Pigeon? bicycle. See ?The ?Permanent? and the ?Flying Pigeon??, People?s Daily, June 16, 2014, Page 4. 
41 
 
 
 
not promoted into the agency. The patron-client relationships suggest that provincial leaders 
appoint agency outsiders because they are politically connected. I demonstrate that patron-client 
relationships between provincial party secretaries and provincial government agency leaders are 
quite uncommon. Even when connections matter, it is unclear whether they lead provincial leaders 
to appoint insiders or outsiders. Finally, while agency heads without inside-agency experience are 
not younger or better educated, they do have more political experience. The preference for political 
experience over specialized skills may explain the appointments in some agencies that are required 
to meet ?hard? targets. However, it does not explain why agency heads cannot always have both 
political and agency-specific experience.  
More importantly, provincial government agency heads are appointed by provincial leaders, 
but the above explanations do not take provincial leaders? incentives and strategies into account. 
As a result, these explanations fail to account for when provincial leaders appoint insiders and 
when they appoint outsiders, and why. My argument studies provincial leaders? appointment 
decisions as the strategies to achieve their objectives under certain constraints. Specifically, I 
provide a theoretical framework to understand how they make appointment decisions in different 
political conditions. A useful theoretical framework to understand appointments is the delegation 
model. 
 
2.2 The Delegation Problem 
In the delegation model (Epstein and O'halloran 1994, Huber and Shipan 2002), politicians decide 
whether to delegate a policy issue to bureaucrats or micromanage themselves. If politicians 
delegate, bureaucrats choose a policy outcome and politicians accept it. If politicians micromanage, 
42 
 
 
 
they choose the outcome they prefer. Two problems complicate politicians? decisions. First, the 
policy outcome is uncertain. In a unidimensional policy space, this means that for a chosen ideal 
policy point in a spectrum of ?1 to 1, the outcome locates within a range of this point.22 
Bureaucrats know the nature of the uncertainty better than politicians (the range of the potential 
outcomes is narrower) because they have expertise on the issue. Second, bureaucrats and 
politicians might have different interests, such as divergent policy preferences. If provincial 
leaders micromanage, they can choose their preferred policy, but they must bear the risk that the 
outcome is far away from their intentions. If provincial leaders delegate, they can expect a less 
uncertain outcome, but they must bear the risk that bureaucrats choose a policy that is different 
from their preferred one. The delegation problem is thus how to best use bureaucrats? expertise to 
reduce policy uncertainty and at the same time make sure bureaucrats act in politicians? interests.  
Although the divergence of interests between bureaucrats and politicians is often modeled 
as the distance of their ideal policy points (e.g., Epstein and O'halloran 1994, Huber and Shipan 
2002), the substance of the conflict could be more broadly defined. The conflict might occur 
whenever bureaucrats pursue their interests instead of politicians?. A common example is the 
problem of corruption ? ?while superiors would like agents to always fulfill the superior?s 
objectives, monitoring is costly, and agents will generally have some freedom to put their own 
interests ahead of their principals?. Here is where money enters. Some third person, who can 
benefit by the agent?s action, seeks to influence the agent?s decision by offering him a monetary 
payment which is not passed on to the principal? (Rose-Ackerman 1978, p.6).  
Politicians make decisions in two dimensions ? choose how much discretion to grant and 
whom to appoint (see two review articles by Bendor, et al. 2001, Gailmard and Patty 2012). Which 
 
22 For example, if the chosen policy is ?0.5, the outcome could be anywhere between ?0.8 and ?0.2. 
43 
 
 
 
decisions they make depend on which tools are at their disposal. The decision of whether to 
delegate or whom to appoint is further complicated by two types of information asymmetry ? 
adverse selection and moral hazard.23 If bureaucrats? preferences and intentions are observable and 
they commit to their promises after delegation, politicians? optimal strategy is to delegate the work 
to or appoint bureaucrats who share the same preference with them. Compared with 
micromanaging themselves, politicians can benefit from bureaucrats? expertise by getting a less 
uncertain outcome. This is called the ?ally principle? ? all else equal, when the preferences of 
bureaucrats and politicians become closer, politicians are more likely to delegate (Epstein and 
O'halloran 1994).24 For example, in the context of appointments, political leaders only assign 
localists who are embedded in the local communities to places that do not threaten their rule 
(Hassan 2020). In the United States, presidents are more likely to appoint loyal political appointees 
rather than career bureaucrats in federal agencies that do not share their policy agendas (Lewis 
2008). In these cases, partisanship and ethnicity are information shortcuts that help reveal 
bureaucrats? intentions and commitments. 
However, bureaucrats? true preferences and intentions are not always observable ex-ante. 
Moreover, bureaucrats have incentives to hide the information. For example, those who want to 
exploit the positions for private gains will pretend to be incorruptible. This leads to the problem of 
adverse selection where politicians might delegate the work to or appoint the wrong type of agents. 
The adverse selection problem is more severe when uncertainty about bureaucrats? character 
 
23 Besley (2006) reviews the adverse selection and moral hazard problem between elected politicians and voters. The 
adverse selection and moral hazard problems in appointments are discussed in Str?m (2000), Huber and Martinez-
Gallardo (2008), Indridason and Kam (2008), and Samuels and Shugart (2010). 
24 Evidence that supports the ?ally principle? is found in legislation (Epstein and O'halloran 1999, Huber and Shipan 
2002), budget allocation (Dahlstr?m and Holmgren 2021), and political appointments (Lewis 2008, Hassan 2020). 
Situations that might lead to violation of the ?ally principle? are discussed by Gailmard (2002), Bendor and Meirowitz 
(2004), and Huber and McCarty (2004).  
44 
 
 
 
increases. Even if bureaucrats? preferences and intentions are observable ex-ante, there is no 
guarantee that bureaucrats will deliver their promises ex-post and they cannot always make their 
commitments credible to politicians. After delegation, bureaucrats? actions are unobservable, and 
they might shirk or act against politicians? interests. This is the problem of moral hazard.  
Politicians can resort to ex-ante and ex-post tools to mitigate the problems of information 
asymmetry. Politicians make the first move when they choose how much discretion to grant or 
whom to appoint. They can affect bureaucratic autonomy ex-ante through laws and administrative 
procedures (McCubbins, et al. 1987a, Bawn 1995, Epstein and O'halloran 1999, Huber and Shipan 
2002), jurisdiction and budget allocation (Ting 2002, Volden 2002, Wood and Bohte 2004, 
Dahlstr?m and Holmgren 2021), and strategic appointments (Str?m 2000, Lewis 2008, Dewan and 
Hortala-Vallve 2011, Mart?nez-Gallardo and Schleiter 2015). When the uncertainty of bureaucrats? 
intentions and actions is high, politicians are expected to grant less discretion or take more 
preventive measures ex-ante (Bawn 1995, Bendor, et al. 2001, Bendor and Meirowitz 2004). For 
example, when partisanship and ethnicity cannot serve as information shortcuts in appointments, 
politicians often appoint some types of ?outsiders?. When the presidents and their party do not 
always share the same interests and thus the co-partisan ministers? loyalty is uncertain, presidents 
appoint non-partisan ministers from outside of the party (Mart?nez-Gallardo and Schleiter 2015). 
To reduce the risks of corruption and collusion, a common strategy is not to allow bureaucrats to 
be assigned to their home places (Miyazaki 1976, Beik 1985, Bhavnani and Lee 2018, Xu, et al. 
2021). For similar reasons, bureaucrats are sometimes intentionally assigned to unfamiliar places 
or organizations (Bulman and Jaros 2020, Chan and Fan 2021). 
Once the work has been delegated or after the appointment decisions have been made, 
political leaders can resort to ex-post controls. The tools of ex-post control include rewards and 
45 
 
 
 
punishments, dismissals and rotations, monitoring, and retaining the veto power to bureaucratic 
actions (e.g., McCubbins and Schwartz 1984, McCubbins, et al. 1987b, Bawn 1997, Huber and 
Shipan 2002, Bendor and Meirowitz 2004, Huber and Martinez-Gallardo 2008, Indridason and 
Kam 2008). The delegation model suggests that ex-post and ex-ante controls are supposed to be 
substitutes (Epstein and O'halloran 1994, Bawn 1997, Huber and Shipan 2002). Political control 
is costly: ex-ante preventive measures come at the cost of lost bureaucratic expertise, and ex-post 
oversight takes the resources that could have been used in other places. Therefore, if politicians 
can more effectively control bureaucrats? actions ex-post, they are expected to take fewer 
preventive measures ex-ante, and vice versa. For example, when legislators are members of the 
committee with jurisdiction over the agency, they are less likely to limit the agency?s discretion 
(Bawn 1997). Similar results are found in the presence of legislative veto over the agency?s actions 
(Huber and Shipan 2002). In the context of appointments, the CCP is found to appoint more 
outsiders to multi-task than single-task positions because the cost of monitoring the former type 
of positions is higher (Huang 2002). 
The delegation model provides a useful theoretical framework to understand how provincial 
leaders in China appoint agency heads. Agency heads, especially those who have worked in the 
agency before, have information advantages over provincial leaders. These agency heads? 
information advantages are a double-edged sword. Provincial leaders? objective is to best use their 
expertise while minimizing the risks of information asymmetry. How provincial leaders balance 
the trade-off depends on the uncertainty of agency heads? intentions and actions and the ex-post 
monitoring capacity. I elaborate on provincial leaders? appointment dilemma in Section 2.3 and 
explain how in-province time and authoritarian monitoring affect the appointments of agency 
46 
 
 
 
insiders and outsiders in Section 2.4. The appointment dilemma is expected to be more relevant in 
some agencies than others. I discuss the scope of the argument in Section 2.5. 
 
2.3 The Appointment Dilemma 
Should provincial leaders appoint agency heads from inside or outside of the agency? Provincial 
leaders face a dilemma: appointing agency heads with prior experience inside the agency reduces 
performance uncertainty at the cost of higher corruption risks. The dilemma resembles the 
information-collusion trade-off identified in theories of local embeddedness. In that context, local 
embeddedness means that bureaucrats share the same race and ethnicity with the local population 
or are born in the same community (see two review articles by Pepinsky, et al. 2017, Grossman 
and Slough 2022). Partly due to ?lower information-gathering costs, superior local knowledge, or 
access to better technologies (including cultural cues) with which to serve in-group members? 
(Grossman and Slough 2022, p.14), embedded bureaucrats produce better policy outcomes. 
Meanwhile, embeddedness also makes bureaucrats more susceptible to corruption and elite capture 
(Xu, et al. 2021).  
In this study, I define embeddedness as the relationship between bureaucrats and their work 
environment. The embeddedness of agency heads contains three elements: agency-specific 
knowledge, connections with colleagues inside the agency, and connections with administrative 
clients outside of the agency. First, agency insiders have subject-area and agency-specific expertise. 
While subject-area expertise can be gained through relevant experience outside of the agency, 
agency-specific knowledge ? including the decision-making process and agency culture ? can only 
be accumulated through working in the agency. Second, agency insiders have time to interact and 
47 
 
 
 
build trust with their colleagues before assuming leadership positions. Third, agency heads with 
prior experience in the agency are more likely to have interactions with their outside administrative 
clients. For example, the main administrative clients of the Development and Reform Commission 
are industry groups. Agency insiders are more likely to have connections with industries under the 
agency?s jurisdiction.  
On the one hand, appointing agency heads with prior experience inside the agency helps 
reduce the uncertainty of performance in two ways. First, their familiarity with the agency reduces 
the transaction costs of agency heads? turnovers. Agency heads who have worked in the agency 
before are ?more familiar with the agency?s policy decisions, personnel, process, and culture; this 
makes monitoring, planning, and leading significantly easier? (Lewis 2008, p.142). Prior 
experience in the agency also helps them develop informal ties with other bureaucrats inside the 
agency. Informal ties among bureaucrats provide channels for information exchange and 
cooperation, which improves government efficiency (Bozcaga 2020). On the contrary, when local 
leaders are appointed to new places, they are found to spend more than half of their time knowing 
the ground and building trust with their subordinates (Eaton and Kostka 2014). They also spend 
enormous time and money wining and dining because they need these social events to build close 
relationships with their colleagues (Wang and Yan 2020). As many agency heads do not have a 
long tenure in the agency, the transactions costs could significantly affect the performance25.  
Second, as evidenced by the East Asian developmental states, agency heads? connections 
with colleagues inside the agency and their ties with administrative clients outside of the agency 
could result in a ?reinforced Weberian? bureaucracy (Johnson 1982, Evans 1995). Close 
 
25 For agency heads appointed between 1984 and 2013 (some of the agency heads appointed after 2013 are still 
incumbent), the median length of their tenure is five years. One-quarter of the agency heads only worked for three 
years. 
48 
 
 
 
connections with the agency create a sense of corporate coherence, which increases bureaucratic 
autonomy. External ties with society provide channels for information gathering and policy 
negotiations. The combination of the two elements results in the ?embedded autonomy? where 
bureaucrats are both embedded with their administrative clients and not captured (Evans 1995). 
On the contrary, when agency heads are appointed from the outside, they do not have a sense of 
belonging, which is an important source of bureaucratic autonomy. They might also hurt agency 
morale and discourage subordinates to develop agency-specific expertise and upgrade their skills 
because their career prospects become unclear (Lewis 2008). For example, bureaucrats in the 
agency may not welcome agency heads from the outside. According to one interviewee, ?from my 
perspective, I want the agency head to be promoted internally because that opens positions to be 
filled inside the agency and if the agency head can be further promoted, I can benefit as well (we 
are familiar after all); but the higher-ups must have their considerations when appointing agency 
heads.? 26  Appointing agency heads from the outside also disrupts connections with the 
administrative clients. 
It is not hard to find examples where agency outsiders perform poorly. Ouyang Bin, the head 
of Hunan Province Department of Transportation who had never worked in the agency before, was 
punished not for corruption but for ?inadequate supervision? after the collapse of the Tuojiang 
bridge. The bridge collapse, as pointed out by many experts, was caused by technical ignorance27 
? the bridge was made with rocks and concrete rather than stronger steel to make it look ?in 
harmony? with the natural environment. To rush the speed, workers were ordered to take the 
scaffolding down before the bridge was finished. The failure of agency outsiders is especially 
 
26 Interview on December 31, 2020. The interviewee is a Division Rank bureaucrat in a government affiliated agency 
in a southwest province. 
27 ?Collapsed China bridge made of rocks, concrete?, Reuters, August 15, 2007.  
49 
 
 
 
noticeable during crises, such as the Covid-19 pandemic. Tang Zhihong, the head of Huanggang 
City Health Commission, showed her probable miscommunication with the hospitals. The 
manager of the Nanjing airport who was investigated after the mismanagement that let the cases 
spread nationwide from the airport28 had no relevant experience in transportation. The head of the 
Xi?an Big Data Management Bureau who was suspended from duty due to multiple breakdowns 
of the health code system was appointed from outside of the agency as well.    
It is worth noting that agency heads without inside-agency experience could also be good 
leaders and by no means are those with prior experience in the agency predestined to perform 
better. What matters is that agency-specific knowledge, internal connections with colleagues, and 
external connections with administrative clients reduce the uncertainty of performance. The 
traditional thought of delegation is that politicians who are ignorant of the issue should delegate to 
experts or specialists because they produce superior outcomes. But the key insight of the modern 
delegation model is that bureaucratic expertise helps narrow the range of policy outcomes. In other 
words, the focus of delegation is not to make outcomes stochastically better but to make them less 
risky (Bendor and Meirowitz 2004). While agency outsiders could be highly qualified and may 
eventually be good leaders, whether their skills are transferable and whether they can get along 
with their subordinates and administrative clients are uncertain ex-ante. The focus on performance 
uncertainty also helps distinguish the value of inside-agency experience and relevant experience 
outside of the agency. Compared with agency heads with no relevant experience at all, those with 
relevant experience outside of the agency have subject-area knowledge. But transferring someone 
from one organization to another increases the uncertainty of whether he or she can work 
 
28 ?Flight from Russia source of latest COVID-19 outbreak in east China city?, Xinhua, July 30, 2021 
50 
 
 
 
effectively in the new organization. Following this logic, appointing agency heads who have 
already proven to be able to work in the agency is a safer choice for provincial leaders.  
On the other hand, agency heads with prior experience inside the agency have greater 
opportunities and capacities to engage in corruption. The problems of political control could come 
in a variety of forms. The common problem in democratic countries is ideological divergence ? 
bureaucrats with different policy preferences present greater challenges for political control. For 
autocrats, the problem is to prevent subordinates from undermining their power. The ideological 
differences between provincial leaders and provincial agency heads in China are likely to be 
limited, at least publicly. The top-down control mechanisms ? including rewards, punishments, 
and rotations ? are found to successfully motivate local bureaucrats to faithfully implement the 
policies once the higher-ups set the priority clear (Edin 2003). The loyalty-competence trade-off 
is more relevant in the appointments of provincial leaders and less so for appointments at the sub-
provincial level (Shih, et al. 2012, Landry, et al. 2018). In the context of appointing provincial 
agency leaders, a more salient problem of political control for provincial leaders is corruption risks 
? ?one of the best indicators of the magnitude and the severity of control problems besetting 
Chinese system is corruption? (Huang 2002, p.75). Ever since the economic reform, corruption 
has become a prevalent problem for the CCP (Manion 2004, Wedeman 2012, Pei 2016, Ang 2020). 
Corruption in China?s bureaucracy has two features. First, among four types of corruption ? 
?petty theft?, ?grand theft?, ?speed money?, and ?access money?, the dominant type of corruption 
in China is ?access money?, defined as ?high-stakes rewards extended by business actors to 
powerful officials, not just for speed, but to access exclusive, valuable privileges? (Ang 2020, 
p.10). The root of corruption is the economic reform that creates market incentives while 
decentralizing the control of valuable resources to local government (Pei 2016). For example, in 
51 
 
 
 
the Department of Transportation, the major type of corruption risk is briberies from construction 
firms to agency heads for contracts. In the Department of Land and Resources, the main corruption 
risk is bribing for land and mining rights. Second, corruption in China often takes the form of 
collusion among elites, including collusion between bureaucrats and businesses and collusion 
between superiors and subordinates within the bureaucracy (Gong 2002, Pei 2016). As the 
dominant type of corruption is ?access money?, it is unsurprising that many corruption cases 
involve bureaucrats-businesses collusion. Corrupt officials do not act alone because collective 
corruption helps reduce the risk of being caught ? it divides the risk among multiple participates 
and facilitates cover-ups (Gong 2002). It is common to find multiple officials in the same agency 
caught corrupt in one case. In 2014, a high-profile case of collective corruption was uncovered in 
the Shanxi Province Department of Transportation with seven officials in the agency investigated, 
including the head and one deputy head. 29  The corruption was mainly related to high-way 
constructions in the province. 
For provincial leaders, the main challenge of political control is thus to prevent collusion. 
While agency heads with prior experience inside the agency are not inherently more corrupt, their 
entrenched networks inside and outside of the agency provide the soil of collusion. When that 
happens, their agency-specific knowledge helps them exploit the loopholes for private gains. The 
case of Li Jiangong illustrates how inside-agency experience assists in collusive corruption.30 Li 
Jiangong was the former head of the Shanxi Province Department of Land and Resources. He was 
sentenced to 15 years for corruption. He worked inside the agency for 12 years before being 
appointed as the agency head. Mining is the major industry in Shanxi Province and The 
 
29 See the report from China Youth Daily http://zqb.cyol.com/html/2014-05/09/nw.D110000zgqnb_20140509_4-
07.htm. 
30  The Shanxi Province Department of Natural Resources uses the case of Li Jiangong as an anti-corruption 
educational material. See https://zrzyt.shanxi.gov.cn/ztzx/chnl1021/chnl584/.   
52 
 
 
 
Department of Land and Resources controls the distribution of mining rights. Li Jiangong was a 
competent and clean official before becoming the agency head. Because of his prior experience in 
the agency, he was friends with many businesspeople in the mining industry. When he was just 
appointed as the agency head in 2009, he was approached by a mine manager, who was also his 
old friend, to get the mining license approved. The mine manager offered Li 30 million RMB to 
get it done. It became a routine for him to take money from the mining industry afterward. Li 
Jiangong also intervened in the appointments inside the agency. He appointed those who bribed 
him as his subordinates. After Li fell, multiple officials inside the agency appointed by him were 
investigated for corruption as well. In addition, Li Jiangong was found to be good at tailoring the 
regulations to benefit his friends in the industry without violating the formal procedure. 
To sum up, provincial leaders face an appointment dilemma of reducing two different risks. 
Appointing agency heads with inside-agency experience reduces the risk of performance 
uncertainty but increases the risk of collusive corruption. Provincial leaders? objective is to avoid 
both performance failure and corruption scandals. Provincial leaders care about both types of risks 
because they are held accountable by the center. They are often required to take ?leadership 
responsibility? in the presence of major performance failure or corruption scandals. For example, 
Jiang Chaoliang, the former Provincial Party Secretary of Hubei, was removed from the position 
after the Covid-19 outbreak in Hubei. After ?disappearing? for one year and a half, he was 
appointed as a committee member in the National People?s Congress ? an end of his political 
career. In another example, after mounting corruption scandals, Yuan Chunqing, the then 
Provincial Party Secretary of Shanxi, was dismissed. Multiple officials promoted under his watch 
were investigated for corruption. He made self-criticism in front of Liu Yunshan, the Politburo 
Standing Committee Member who came to Shanxi to announce his dismissal, that ?Comrade Liu 
53 
 
 
 
Yunshan reprimanded the serious problems in Shanxi?s anti-corruption struggles; as the Provincial 
Party Secretary, I take leadership responsibility?.31  Ironically, Yuan Chunqing launched two 
rectification campaigns that punished more than 1000 officials in his tenure.  
Although the empirical evidence of how performance affects provincial leaders? career 
perspectives is inconclusive (Li and Zhou 2005, Shih, et al. 2012, Jia, et al. 2015, Landry, et al. 
2018) and the party is often reluctant to punish senior officials (Cai and Zhu 2013, Mei and Pearson 
2014, 2017), the bottom line is that both performance failure and corruption scandals create extra 
trouble for provincial leaders. As they control the tool of appointments, they can use it to choose 
the ?right? type of agency heads to avoid potential trouble under their jurisdiction. No matter what 
motivations they have, ?the primary goal for politicians in China is to be safe?.32  Although 
dismissals and rotations are useful tools to correct the ?wrong? appointments (Huber and Martinez-
Gallardo 2008, Indridason and Kam 2008), the damage has already been done at that point. For 
provincial leaders who want to preserve themselves from taking ?leadership responsibility?, ex-
ante appointments are more effective than ex-post corrections. As reducing performance 
uncertainty and reducing corruption risks conflict with each other, how do they balance the trade-
off in appointments? 
 
2.4 Information, Authoritarian Monitoring, and the Appointment Dilemma 
The delegation model suggests that provincial leaders? prior information on the agency heads and 
the monitoring capacity ex-post are two important factors that could affect the appointments of 
insiders and outsiders. The ?ally principle? suggests that politicians are more likely to delegate to 
 
31 See the article in China News Weekly https://www.chinanews.com.cn/gn/2014/09-12/6586053.shtml. 
32 Interview on January 7, 2021. 
54 
 
 
 
bureaucrats when their policy preferences converge. But if politicians can only guess at 
bureaucrats? intentions and are not sure about what bureaucrats will do once given authority, the 
value of delegating decreases (Bawn 1995, Bendor, et al. 2001, Bendor and Meirowitz 2004). 
Therefore, we can expect provincial leaders to appoint agency heads from outside of the agency 
when they are unsure about the candidates? intentions and actions. Their bias toward agency 
insiders should attenuate when they have improved information on the candidates.  
Yet because bureaucrats? true intentions are never observable ex-ante, this results in the 
adverse selection problem. And because bureaucrats? actions cannot be perfectly monitored ex-
post, this results in the moral hazard problem. Nevertheless, some information shortcuts could help 
mitigate the problems. For example, partisanship and ethnicity often serve as information shortcuts 
that help reveal bureaucrats? preferences ex-ante (Lewis 2008, Hassan 2020). Bureaucrats who 
share the same partisanship or ethnicity with politicians also have aligned interests which help 
bind their actions afterward (Toral 2021a). However, as a single-party regime with a majority of 
Han ethnicity, partisanship and ethnicity are less informative in China. In such a context, shared 
work experience is more important in mitigating the problems of adverse selection and moral 
hazard (Keller 2016, Jiang 2018). 
In the Chinese context, shared work experience creates opportunities for repeated 
interactions, which helps provincial leaders to know the true character of sub-provincial 
bureaucrats better. Without shared work experience, an important source of information for 
provincial leaders is personal dossiers. Although the dossiers document detailed information of 
one?s lifetime experience, even with family backgrounds and ?black materials? supplied by 
55 
 
 
 
internal intelligence offices,33 it is not a substitute for knowledge gained through repeated personal 
interactions. In addition, the quality of the information in the personal dossiers is not always 
reliable.34 Faked dossiers were found in 15 out of the 20 provinces inspected by the center in 2014. 
The Organization Department of the CCP reported that 420 provincial-managed cadres were 
punished for faking their resumes and another 186 are suspended from getting promoted because 
their dossiers are ?problematic?. Another source of information is the Provincial Department of 
Organization, which has direct interactions with sub-provincial bureaucrats. But it is not a 
substitute for first-hand knowledge collected by provincial leaders themselves because they are 
unlikely to fully trust recommendations from the Provincial Department of Organization. 
Shared work experience also creates opportunities for provincial leaders to build patron-
client relationships with sub-provincial bureaucrats. A patron-client relationship is formed when a 
provincial leader promotes a sub-provincial bureaucrat (Keller 2016, Jiang 2018). Promotion links 
are more likely to exist when provincial leaders have shared work experience with sub-provincial 
bureaucrats. When formal institutions are weak, patron-client relationships help solve the moral 
hazard problem by aligning the interests between provincial leaders and their appointees (Jiang 
2018, Toral 2021a). On the one hand, it becomes the appointees? self-interest not to be 
troublemakers for provincial leaders. On the other hand, the shared political networks reduce 
information asymmetry and facilitate monitoring. It also binds provincial leaders to reward their 
appointees for good performance. The enhanced monitoring and credible commitments of rewards 
regulate the appointees? actions in turn. Overall, ?by fostering mutual trust and raising the value 
 
33 Interview on January 7, 2021. The interviewee is a former deputy Bureau Rank bureaucrat in a government agency 
in a southwest province. 
34 See the report from Xinhua News Agency http://www.xinhuanet.com//politics/2016-01/28/c_1117923755.htm. 
56 
 
 
 
of long-term cooperation, these relations help align the interests between government agents and 
their principals, and discourage short-term, opportunistic behaviors? (Jiang 2018, p.983). 
In sum, shared work experience between provincial leaders and sub-provincial bureaucrats 
mitigates the problem of adverse selection and moral hazard. When provincial leaders have prior 
shared work experience with the agency heads, they should be less concerned that the agency heads 
take advantage of the information asymmetry to engage in corruption. Therefore, they are expected 
to appoint agency heads with inside-agency experience to reduce performance uncertainty. On the 
contrary, when provincial leaders do not have prior shared work experience with the agency heads, 
they are more likely to appoint agency heads from outside of the agency to reduce information 
asymmetry. 
Another factor is the capacity to monitor appointees after appointments. Bureaucratic 
embeddedness increases corruption risks when monitoring capacity is low. If agency heads? 
actions can be more effectively monitored, provincial leaders are expected to worry less about 
corruption risks at the appointment stage. Instead, they should be encouraged to appoint agency 
heads with inside-agency experience to reduce performance uncertainty. The substitution effect of 
ex-post control and ex-ante precaution suggests that increased monitoring over provincial agency 
heads will solve provincial leaders? appointment dilemma. The CCP has launched at least seven 
waves of anti-corruption campaigns in the reform era to strengthen its monitoring capacity over 
local bureaucrats.35 Although some anti-corruption efforts are politically driven (Zhu and Zhang 
2017) and the effects of the previous campaigns are mixed (Manion 2004, Wedeman 2005), Xi 
Jinping?s anti-corruption campaign is considered a successful example that increased the CCP?s 
 
35 Manion (2004) studies the anti-corruption campaigns in 1982, 1986, 1989, 1993, and 1995. In 2006, the CCP 
launched an anti-corruption investigation against Chen Liangyu, the then Shanghai Party Secretary. In December 2012, 
Xi Jinping started the most recent anti-corruption campaign.  
57 
 
 
 
monitoring capacity using the authoritarian approach (Carothers 2022). The campaign aims to 
create an environment where ?no one dares to, is able to, and has no desire to corrupt?36. The 
campaign has been ?so? effective that bureaucratic slack became a prevalent problem as no one 
dares to do anything that contains the risk of making mistakes (Wang and Yan 2020, Wang 2021). 
To counter the fear, the CCP promoted a system to ?tolerant mistakes? in 2016. An effective anti-
corruption campaign is thus expected to alleviate provincial leaders? concerns about corruption 
risks and encourage them to appoint more agency heads with prior experience inside the agency. 
However, the authoritarian features that make the anti-corruption campaign effective also 
create secondary effects that might lead provincial leaders to take more preventive measures in 
appointments. Anti-corruption campaigns rely on top-down control. Top-down control means that 
the party center bypasses provincial leaders to directly monitor sub-provincial bureaucrats, 
including provincial agency heads. One of the most important approaches used in Xi Jinping?s 
campaign is the central inspections (Yeo 2016). From May 2013 to July 2014, the CCP sent four 
waves of teams to each province to conduct a three-month inspection. The members of each 
inspection team were intentionally selected from those with no connections to the targeted 
province. Under increased top-down control, corruption committed by sub-provincial bureaucrats 
is more likely to be detected. Once provincial leaders? appointees are caught corrupt, increased 
top-down control also means that the threats to hold provincial leaders accountable become more 
credible. After Xi Jinping?s campaign, the CCP stressed in the revised ?Regulations on the Work 
of Selecting and Appointing Leading Party and Government Cadres? that those in charge of the 
appointments will be held accountable if their wrong appointments cause serious consequences.37 
 
36 Speech by Xi Jinping in the fifth plenary session of the 18th Central Commission of Disciplinary Inspection. 
37 In 2014, the CCP revised the regulation for the first time since 2002. 
58 
 
 
 
In Guangzhou City, the occurrence of large-scale corruption is listed as one of the serious 
consequences where responsibility will be traced back to those who made the appointment 
decisions in the first place.38 From September to December 2014, 10 officials in the city?s 13 
chosen ?experiment points? were held accountable because of their previous appointment 
decisions.39 As corruption is more likely to be detected and the risk of being implicated by their 
appointees? misbehavior becomes real, increased top-down control might lead provincial leaders 
to prioritize reducing corruption risks by appointing agency heads without inside-agency 
experience. 
Anti-corruption campaigns, as any campaign-style enforcement that relies on extra-legal and 
ad-hoc instead of rationalized and routine institutions, increase political uncertainty. The reason 
to resort to campaign-style enforcement is that the routine force is weak. Anti-corruption 
campaigns can mobilize extra resources in a short time and send the signal that this time is different 
(Manion 2004). But also because campaigns break away from the routine and are not bound by 
the rule of law, it creates the uncertainty of who the targets are, when the next wave of crackdowns 
comes, and what constitutes mistakes (Montagnes and Wolton 2019, Li and Manion 2021). On the 
one hand, as the top leader is the only one who controls the wind of the campaign, it is hard to 
predict where and how far it will go. For example, although Xi Jinping?s anti-corruption campaign 
later proved to be more than just power struggles, the perceptions of the campaign?s motivation 
are widely divided at the beginning (see a review article by Kautz 2020). On the other hand, the 
increased uncertainty might be intentionally used as a tool to make deterrence effective (Lui 1986, 
Manion 2004, Wedeman 2005). When routine enforcement is weak, it cannot deter corrupt 
 
38 See https://www.chinanews.com.cn/sh/2015/09-21/7536256.shtml.  
39 See http://fanfu.people.com.cn/n/2015/0421/c64371-26878165.html. 
59 
 
 
 
officials because the probability of being caught is low. However, if corrupt officials need to take 
an unknown risk of getting caught in future crackdowns and an unknown frequency of crackdowns 
into consideration, ?fear of greater risk at some undetermined point in the future may cause risk 
adverse officials to reject bribes? (Wedeman 2005, p.97). Furthermore, a set of objective criteria 
that can keep officials from being investigated does not exist. For example, after bureaucratic slack 
became a prevalent problem after the campaign, the CCP included bureaucratic slack as a form of 
corruption. In addition, as the campaign disciplines more officials into compliance, it also becomes 
increasingly difficult for the party to detect non-compliers, which makes good performance no 
longer a guarantee of safety (Montagnes and Wolton 2019). Under extreme political uncertainty 
after anti-corruption campaigns, provincial leaders are likely to prioritize reducing corruption-
related risks in appointments, which might lead them to appoint more agency heads from outside 
of the agency.  
Another effective element of anti-corruption campaigns is intra-party propaganda. Similar 
to how authoritarian propaganda signals regime strength to citizens (Huang 2015), intra-party 
propaganda demonstrates the dominance of the party over individual cadres and thus induces 
submissiveness (Wedeman 2005, Mertha 2017). Examples of intra-party propaganda include 
criticism and self-criticism, political study sessions, and documentary series that put corrupt 
officials on TV to remorse their guilts with tears. These uncomfortable anti-corruption messages 
sow dread and fear that make no one feel safe. Flooded by the propaganda materials, scared 
provincial leaders might become single-minded in reducing corruption risks and appoint more 
agency heads without inside-agency experience. 
In sum, while the delegation model suggests that increased monitoring capacity should lead 
provincial leaders to take fewer preventive measures in appointments, authoritarian monitoring 
60 
 
 
 
has distinctive features that might lead them to take more preventive measures. Top-down control, 
uncertainty, and intra-party propaganda are important elements that make authoritarian anti-
corruption campaigns effective. These same elements, however, create secondary consequences 
on how provincial leaders make appointment decisions. 
Table 2. 1: Summary of the Theoretical Framework 
 Ex-ante   Ex-post 
Agency heads? character   Agency heads? actions 
Information asymmetry 
(Corrupt or clean)   (Collusion with colleagues and businesses) 
 Mitigate     
     
IV: Shared work experience (In-province time) 
     
Reduce incentives 
     
     
  Subst itu te or intensify?  
Preventive appointments   Authoritarian Monitoring 
Control strategy 
(DV: Appoint agency outsiders)   (IV: Xi Jinping?s anti-corruption campaign) 
 
Table 2.1 summarizes the theoretical framework. The dependent variable (DV) of this study 
is provincial leaders? strategy of appointing agency insiders and outsiders. I start with the two types 
of information asymmetry. Appointees? character and actions are unobservable to provincial 
leaders. In the context of appointing provincial government agency heads, the main risk of 
information asymmetry is that agency heads can take advantage of their information advantages 
to engage in collusive corruption. Provincial leaders can take preventive measures at the 
appointment stage to mitigate corruption risks by appointing agency heads from outside of the 
agency. Two independent variables (IV) explain whether provincial leaders appoint agency 
insiders or outsiders. First, provincial leaders? shared work experience with sub-provincial 
bureaucrats helps mitigate the problems of information asymmetry and thus should lead provincial 
61 
 
 
 
leaders to take fewer preventive measures in appointments. I use provincial party secretaries? in-
province time before making appointment decisions as a proxy for shared work experience. Second, 
provincial leaders? appointment strategies are affected by the ex-post monitoring capacity. On the 
one hand, if appointees? actions can be more effectively monitored, provincial leaders can take 
fewer preventive measures ex-ante. On the other hand, authoritarian monitoring can lead 
provincial leaders to take more preventive measures. I test the relationship using the case of Xi 
Jinping?s anti-corruption campaign. 
  
2.5 Scope of the Theory: Not All Agencies are Created Equal 
My argument builds on two assumptions. First, agency heads? prior work experience serves as an 
information shortcut for provincial leaders. Agency heads with prior experience inside the agency 
are considered to have information advantages that could be used to both reduce performance 
uncertainty and engage in collusive corruption. Second, provincial leaders care about the outcomes 
of the agency ? either policy outcomes or corruption scandals. These two assumptions do not 
equally apply to each agency. In some agencies, such as the Department of Transportation, the 
nature of the agency?s subject area determines that the corruption risk ? especially the risk of 
collusive corruption ? is high. In addition, both policy failure (such as the bridge collapse) and 
corruption scandals (such as collective corruption) of the agency could bring nationwide attention 
which provincial leaders prefer to avoid. In some other agencies, such as the Department of Culture, 
there is no trade-off between appointing agency heads with and without inside-agency experience 
because the risk of collusive corruption is low, and the policy outcomes of the agency are unlikely 
to be provincial leaders? priorities. In other words, two dimensions of agency characteristics are 
62 
 
 
 
relevant to my theoretical assumptions ? the risk of collusive corruption and the importance of the 
agency. The argument is more applicable to important agencies with high risks of collusive 
corruption.  
If the two dimensions are independent, the government agencies could be categorized into 
four types. However, the risk of collusive corruption and the importance of the agency are highly 
correlated in China?s reform era. On the one hand, since the 1980s, the top priority of the CCP has 
been economic development. Government agencies with policy jurisdiction related to economic 
development are thus more likely to be important agencies. For example, the Development and 
Reform Commission ? the agency that approves and manages major developmental projects ? is 
called the ?Mini State Council?. On the other hand, economic development breeds collusive 
corruption, especially in areas where local governments monopolize valuable resources and the 
market competes for access (Wedeman 2012, Pei 2016, Ang 2020). The defining nature of 
corruption in China is the marriage between power and money, ?this union manifests itself in the 
collusion between government officials who control the allocation and disposal of valuable state-
owned assets and economic resources and private businessmen trying to seize these assets? (Pei 
2016, p.118). For example, the Development and Reform Commission is also the place where 
corruption thrives, and the money involved is often huge. From 2013 to 2015, over 20 officials in 
the National Development and Reform Commission were investigated. In the case of Wei 
Pengyuan, the deputy director of the coal department, the money involved is the largest amount 
since the founding of the People?s Republic of China in 1949.40 More than 200 million yuan in 
cash was found at his home, which burned four cash-counting machines. The money was mostly 
briberies from firms in the coal industry. 
 
40 ?Former energy official found guilty of graft?, China Daily, October 17, 2016. 
63 
 
 
 
In other words, the agencies that are both important and have high risks of collusive 
corruption ? the type of agencies where the theoretical framework applies ? are the agencies that 
control valuable resources related to economic development. Although it is difficult to obtain an 
objective and systematic measurement, both quantitative and qualitative evidence have 
consistently identified which government sectors in China fit the category. Pei (2016) argues that 
government sectors related to land, mining, and state-owned firms are the soil of collusive 
corruption because these are the places that have decentralized the ?control over state-owned assets 
without complete clarification of their ownership rights? (Pei 2016, p.49). According to a study 
that analyzes 2802 corruption cases reported in ?Procuratorate Daily? from 2000 to 2009 (Gong 
and Wu 2012), the top five corrupt areas (except for personnel management which is controlled 
by the party instead of the government agencies) are government procurement and construction 
contracts (731 cases); financial management (315 cases); land, real estate, and urban planning (307 
cases); finance, investment, loan, and allocation of funds (298 cases); industry and enterprise 
management and reconstruction (247 cases). These five areas contain 67 percent of the cases.  The 
analysis of 142 city leaders and 54 province-minister leaders investigated for corruption since the 
reform era (Qiao 2013, 2014) find that one of the key features of these cases is collusion with 
businesspeople outside and colleagues inside. Among the 142 corrupt city leaders, 115 cases are 
related to land transfers, real estate development, corporate loans allocations, tax reductions, stock 
market listings, project management, corporate entry, ownership reconstruction, construction 
projects, and mining rights. Among the 54 corrupt province-minister leaders, the top two corrupt 
areas are real estate development and enterprise management. 
The government agencies that have policy jurisdictions over the above sectors fall into three 
categories: the agencies that control land and mining rights, the agencies related to construction 
64 
 
 
 
and infrastructural development, and the agencies that manage the enterprises. Based on Table A.1 
in Appendix A with policy jurisdictions and core functions of each agency, the corresponding 
agencies are the Department of Land and Resources (land and mining rights); the Department of 
Construction, and the Department of Transportation (construction and infrastructural 
development); the Development and Reform Commission, the Economic Affairs Commission, and 
the Department of Finance (enterprise management). These agencies control valuable resources 
related to economic development and are thus important agencies with high risks of collusive 
corruption. The theoretical framework should apply to these six agencies but not the others. Of all 
the agencies related to economic development, the Department of Commerce, the agency that 
manages trade and foreign investment, is the only one that is not included. This is because 
marketization in this area is more thorough ? ?there are few cases of collusion between officials 
and businessmen in the more marketized and highly competitive sectors such as retail, light 
manufacturing, and export-oriented industries? (Pei 2016). The data from 2802 corruption cases 
also support it: only 50 cases are related to trade and attracting businesses. 
Although the criterion of the categorization ? controlling land and mining rights, related to 
construction and infrastructural development, and managing enterprises ? is based on consistent 
findings from existing studies, it is not based on an objective measurement for each agency. As a 
validation process, I create an index that measures both the agency?s importance and its risks of 
corruption. I use the proportion of agency heads promoted to the higher administrative rank (the 
Deputy Provincial/Minister Rank) and the more important position within the Bureau Rank 
(prefecture-level city party secretaries in this case) as a proxy to measure the agency?s importance 
(the high-profile index). I use the total number of corruption cases discovered in the related 
government sectors as a proxy to measure the agency?s corruption risks (the high-risk index). The 
65 
 
 
 
corruption investigation dataset is from (Wang and Dickson 2022), which includes all corruption 
cases investigated between 2011 to 2016. Section B.3 in Appendix B introduces the details of the 
indexes and discusses the caveats and alternative measures of corruption risks. As the purpose is 
to identify government agencies that are both important and have high corruption risks, I multiply 
the high-risk index with the high-profile index and rank the agencies based on the index of high-
risk ? high-profile. 
 
 
Figure 2. 7: Indexes of Government Agencies 
     
Figure 2.7 plots the high-risk index, high-profile index, and the multiplication of the two. 
First, consistent with the categorization based on the existing literature, the number of investigated 
cases is particularly high in the construction, transportation, and land resources sectors. The 
number of cases is also high in the sectors related to enterprise management. Interestingly and 
66 
 
 
 
consistent with the existing studies (Gong and Wu 2012, Pei 2016), the number of investigated 
cases related to trade and foreign investment is even lower than those related to health and civil 
affairs. Second, government agencies that manage economic development are in general important 
agencies as measured by agency heads? career prospects. As two exceptions, the career prospects 
of the agency heads in the Department of Construction (ranked 12) and the Department of Land 
and Resources (ranked 11) are not particularly promising. 
 After multiplying the high-risk with the high-profile index, the top six agencies are the same 
as the six agencies categorized by the criterion of controlling land and mining rights, related to 
construction and infrastructural development, and managing enterprises. The Development and 
Reform Commission and the Department of Finance rank higher than the other agencies by a large 
margin because of their much higher importance and high corruption risks. The high-risk ? high-
profile index is almost identical for the agencies ranked fourth to sixth and drops by 22 percent for 
the seventh agency, the Department of Agriculture.   
Both the categorizations based on existing studies and my self-created index of high-risk ? 
high-profile suggest that the theoretical framework should be more applicable to the following six 
agencies: the Development and Reform Commission, the Department of Finance, the Economic 
Affairs Commission, the Department of Land and Resources, the Department of Construction, and 
the Department of Transportation.  
I further test whether agency heads with prior experience inside the agency are indeed more 
likely to engage in corruption in these six agencies. Note that what matters is provincial leaders? 
ex-ante assessments of the potential benefits and risks of appointing agency heads with inside-
agency experience. The actual ex-post outcomes depend on other factors, such as the length of 
tenure and the level of monitoring. Nevertheless, Figure 2.8 shows supportive evidence for the 
67 
 
 
 
assumption. Among the corrupt agency heads, the proportion of those with prior experience inside 
the agency is approximately 15 percentage points higher than the proportion among uncorrupt 
agency heads, but only in important agencies with high corruption risks. Section B.4 in Appendix 
B describes the details of the comparison.   
 
 
Figure 2. 8: Agency Heads Investigated for Corruption (2005-2020) 
In Chapter 3 and Chapter 4, I test how information on sub-provincial bureaucrats and 
authoritarian monitoring affect provincial leaders? appointment decisions in the six agencies 
identified as important agencies with high risks of collusive corruption (high-profile high-risk 
agencies), using the other agencies as the ?control group?. In Chapter 5, I discuss provincial leaders? 
appointment decisions in the agencies that are low-profile in the policy agendas with low 
corruption risks. 
 
 
68 
 
 
 
Chapter 3: In-Province Time and the Appointment Dilemma 
 
This chapter tests how provincial leaders? information on sub-provincial bureaucrats affects their 
appointment decisions. Chapter 2 elaborates on how shared work experience helps mitigate the 
problem of adverse selection and moral hazard. When provincial leaders have shared work 
experience with sub-provincial bureaucrats, they are expected to take fewer preventive measures 
(appointing outsiders) in appointments and instead to prioritize reducing performance uncertainty 
by appointing more agency heads with prior experience inside the agency. On the contrary, when 
provincial leaders do not have shared work experience with sub-provincial bureaucrats, we can 
expect them to take more preventive measures in appointments. Section 3.1 tests the theoretical 
argument by using provincial party secretaries? in-province time as a proxy for their shared work 
experience with sub-provincial bureaucrats. Section 3.2 reports the robustness checks on the main 
findings. Section 3.3 discusses and empirically tests the alternative explanations. Section 3.4 
summarizes.  
 
3.1 Main Results 
Figure 1.4 in Chapter 1 shows provincial party secretaries? in-province time calculated as the years 
between becoming members of the Provincial Party Standing Committee and appointing agency 
heads. I use in-province time as a proxy for provincial leaders? shared work experience with sub-
provincial bureaucrats. When in-province time is short, provincial leaders do not have enough time 
to know the character of sub-provincial bureaucrats through repeated interactions and form patron-
69 
 
 
 
client relationships with some of them. For example, when provincial party secretaries? in-
province time is within one year, it means that they have been parachuted into the province from 
the center or other provinces and need to make appointment decisions without knowing the ground. 
Therefore, short in-province time is expected to lead provincial leaders to take more preventive 
measures in appointments. As an observable implication, they are expected to appoint fewer 
agency heads with prior experience inside the agency in important agencies with high corruption 
risks. 
Variation in in-province time could be arguably considered as ?as-if? random. The variation 
is jointly determined by three factors. First, whether the provincial party secretary has prior work 
experience in the province or is parachuted in. Second, how long the provincial party secretary has 
assumed the position before appointing the agency head. The appointment and reshuffling of 
provincial party secretaries ? both decided by the party center ? are not exogenous.41 The party 
center faces a similar dilemma of balancing political control and local expertise in managing its 
provincial leaders. For example, provincial leaders are more likely to be selected from the localists 
in places with complex governance challenges and parachuted from the outside after major crises 
(Bulman and Jaros 2020). But neither appointments nor reshuffling alone determine the in-
province time before appointing provincial agency heads. Those who have prior work experience 
in the province may have long in-province time in their first year of assuming the position of 
Provincial Party Secretary. Similarly, those who are parachuted into the province may have long 
in-province time in their late tenure. Third, while provincial leaders can take initiatives in rotating 
agency heads to achieve their agendas, some vacancies are created due to the strict age limit of 
 
41 Appointment means whether the party center selects provincial party secretaries from inside or outside of the 
province. Reshuffling means when the party center rotates incumbent provincial party secretaries to another provinces 
or positions.  
70 
 
 
 
retirement as provincial agency heads are required to retire or ?take a back seat? at the age of 60. 
In other words, provincial leaders cannot completely manipulate the timing of appointments, and 
sometimes they need to fill the vacancies ?passively?. In sum, the variation of in-province time in 
different provinces in the same year is determined by different historical patterns of appointments 
and reshuffling of both provincial party secretaries and provincial agency heads. 
Table 3.1 shows provincial party secretaries? ??? year in the province and the proportion of 
newly appointed agency heads from 1984 to 2012 with prior experience inside the agency. I only 
include agency heads appointed before 2013 because appointments afterward are affected by 
increased monitoring due to Xi Jinping?s anti-corruption campaign. The total number of agency 
heads appointed during this period is 2234. Column (1) shows the number of newly appointed 
agency heads in the six high-profile high-risk agencies at provincial party secretaries? ??? year in 
the province. If provincial leaders manipulate the timing of appointments by not making 
appointment decisions when lacking information on sub-provincial bureaucrats, the number of 
appointments made in their first year in the province should be particularly small. But this is not 
what column (1) shows ? provincial leaders made 71 appointments in the high-profile high-risk 
agencies in their first year in the province, which is not small compared with the number at their 
other ??? year in the province. The average number of appointments made during their second, 
third, fourth, and fifth years in the province is 73. Interestingly, provincial leaders made 90 
appointments ? the highest number for the high-profile high-risk agencies ? in their second year 
in the province. The pattern suggests that after spending one year to know the ground, provincial 
leaders start to actively rotate agency heads in these important agencies to achieve their agendas. 
Column (3) shows the number of newly appointed agency heads in the other agencies at provincial 
71 
 
 
 
party secretaries? ??? year in the province. The number of appointments made in their first year in 
the province is not particularly small as well, but the pattern of ?strategic? appointments is muted. 
Table 3. 1: In-Province Time and Appointments of Agency Insiders (1984-2012)  
 High-profile high-risk agencies  The other agencies 
The proportion of  The proportion of 
??? year in the province N N agency insiders (%) agency insiders (%) 
 (1) (2)  (3) (4) 
1st year 71 38.02  167 37.72 
2nd year 90 48.88  160 33.33 
3rd year 67 56.71  184 32.78 
4th year 59 45.76  152 35.09 
5th year 74 50.00  165 32.31 
6th year 54 59.25  129 32.03 
7th year 40 52.50  122 36.88 
8th year 40 52.50  90 40.44 
9th year 30 50.00  81 25.92 
10th year 31 48.38  75 37.33 
11th year 27 59.25  80 42.50 
12th year and above 73 45.20  173 36.41 
  
Column (2) of Table 3.1 shows the proportion of agency heads with inside-agency 
experience appointed at provincial party secretaries? ??? year in the province in the high-profile 
high-risk agencies. The results show an obvious bias against agency insiders in provincial party 
secretaries? first year in the province. Among the agency heads appointed when in-province time 
is within one year, the proportion of those with prior experience inside the agency is 38 percent. 
Among the agency heads appointed when in-province time is longer than one year, the proportion 
of those with prior inside-agency experience is 51 percent. But the bias against agency insiders 
soon disappears when provincial leaders have spent one year in the province. Among the agency 
heads appointed in provincial leaders? second year in the province, the proportion of those with 
72 
 
 
 
prior experience inside the agency increases to approximately 49 percent. Column (4) shows the 
proportion of agency heads with prior inside-agency experience appointed to the other agencies. 
The first-year bias against agency insiders does not exist among appointments in these agencies. 
These results suggest that when provincial leaders lack information on sub-provincial bureaucrats, 
they tend to appoint agency heads without inside-agency experience to mitigate their information 
advantages as a preventive measure to reduce corruption risks, but only in important agencies with 
high corruption risks.  
 To more rigorously test whether short in-province time leads provincial leaders to appoint 
agency outsiders, I use two empirical strategies. First, I adopt a cross-sectional comparison. I 
compare agency heads in the high-profile high-risk agencies appointed in the same year in different 
provinces. I leverage the variation of in-province time across provinces in a certain year. For 
example, Ouyang Bin was appointed as the head of the Hunan Province Department of 
Transportation in 2006. The in-province time before appointing him is three months. In 2006, Guo 
Jianbiao was appointed as the head of the Zhejiang Province Department of Transportation. The 
in-province time before appointing him is three years and 10 months. The crucial assumption is 
that the then provincial party secretary of Hunan could be considered to have ?happen? to 
(randomly) face a shorter in-province time before appointing his agency head than the then 
provincial party secretary of Zhejiang. As the two agency heads were appointed in the same year, 
there are no time-varying differences that affect provincial leaders? appointment decisions. 
Although there are other factors across different provinces that might affect the appointment of 
agency insiders and outsiders, the identification strategy is valid if these factors do not correlate 
with the length of provincial leaders? in-province time. To allow for a larger number of 
comparisons within the same year, I do not restrict the comparisons to be among the same agency 
73 
 
 
 
in different provinces. Instead, I compare appointments made in the same year among all the six 
high-profile high-risk agencies. The assumption is that for a given in-province time, the 
appointment strategy is the same among the high-profile high-risk agencies. To test whether short 
in-province time affects the appointment of agency insiders and outsiders in the other agencies, I 
also compare agency heads in the other agencies appointed in the same year in different provinces.  
 
????????? = ?1???? + ?2???? ? ????? + ?rt+ ????               (3.1) 
 
Model (3.1) estimates the effects of in-province time for appointments in the high-profile 
high-risk and the other agencies in one linear probability model. ?, ?, and ? represent the province, 
agency, and year of appointments, respectively. ????????? is a dummy variable coded as 1 if the 
newly appointed agency head in province ?, agency ?, and year ? has prior work experience inside 
the agency. ???? is a dummy variable coded as 1 if the in-province time before appointing the 
agency head in province ? and agency ? in year ? is smaller or equal to one year. ?????  is a dummy 
variable coded as 1 if agency ? is one of the six high-profile high-risk agencies. ??? is the separate 
year fixed effects for the high-profile high-risk agencies and the other agencies, where ? represents 
the two agency types. In model (3.1), ?1 + ?2 is effectively the result of comparing appointments 
made to the high-profile high-risk agencies in the same year when in-province time is less than or 
equal to one year and when in-province time is longer than one year. ?1 is the same result for the 
other agencies. I cluster the standard errors at the level of provincial party secretaries because the 
74 
 
 
 
appointment decisions made by the same provincial party secretary in one province are not 
independent.42 
Table 3. 2: Cross-Sectional Effects of In-Province Time on Appointing Agency Insiders 
 DV: inside-agency work experience 
 T ? 1 T ? 1 year T ? 1 year T ? 1 year T ? 2 
year and 3 months and 6 months and 9 months years 
 (1) (2) (3) (4) (5) 
High-profile high-risk agencies      
Marginal effects: ?1 + ?2 -0.119 -0.115 -0.105 -0.088 -0.046 
 (0.065) (0.054) (0.054) (0.050) (0.047) 
Mean of the dependent variable 0.497 0.497 0.497 0.497 0.497 
N of agency heads appointed 
71 100 117 131 161 
within T 
The other agencies      
Marginal effects: ?1 0.070 0.072 0.049 0.052 0.050 
 (0.041) (0.035) (0.034) (0.033) (0.032) 
Mean of the dependent variable 0.350 0.350 0.350 0.350 0.350 
N of agency heads appointed 
167 221 245 294 327 
within T 
Note: Standard errors clustered at the level of provincial party secretaries are reported in the parentheses. 
 
Table 3.2 reports the marginal effects of short in-province time on the appointments of 
agency insiders in the high-profile high-risk agencies and the other agencies. In five separate 
models, I vary the dummy variable ???? to be coded as 1 if the in-province time is within one year, 
one year and three months, one year and six months, one year and nine months, and two years. 
 
42 The correlations of appointments made by the same provincial party secretary in one province could come from two 
sources. First, the same provincial party secretary may have a consistent preference to appoint one type of agency 
head to certain positions. Second, for a fixed pool of candidates, when one is assigned to an agency, he or she cannot 
be assigned to another agency. There are 155 provincial party secretaries in 22 provinces who appointed 2234 agency 
heads. On average, one provincial party secretary appointed 14 agency heads. As the number of observations per 
cluster is not particularly large, I adjust the clustered standard errors when fixed effects models are estimated by the 
dummy variable approach (Cameron and Miller 2015, pp. 14-15).    
75 
 
 
 
Columns (1) to (5) report the results, respectively. A caveat is that because the reference groups of 
these five models are different, the marginal effects from columns (1) to (5) are not directly 
comparable. 
For high-profile high-risk agencies, column (1) shows that the proportion of newly appointed 
agency heads with prior inside-agency experience is 12 percentage points lower when provincial 
party secretaries? in-province time is within one year, compared with those appointed in the same 
year with the in-province time longer than one year. The size of the effect is close to what Table 
3.1 shows. As the average proportion of agency insiders in the high-profile high-risk agencies is 
50 percent, the size of the effect is equivalent to a 24 percent drop. Due to the small number of 
agency heads appointed when the in-province time is within one year (N=71), the effect is only 
statistically significant at the 10 percent level. Compared with the agency heads appointed when 
the in-province time is longer than one year and three months, column (2) shows that the size of 
the effect is similar. As the number of agency heads appointed within one year and three months 
increases (N=100), the effect is statistically significant at the 5 percent level. Column (5) shows 
that when comparing agency heads appointed with the in-province time being within and above 
two years, the difference drops to fewer than 5 percentage points. For the other agencies, the effect 
is reversed: provincial leaders are more likely to appoint agency heads with prior inside-agency 
experience when their in-province time is short. The effects in columns (1) and (2) are 
approximately 7 percentage points.  
Table 3.1 and Table 3.2 show a consistent pattern of bias against agency insiders. When 
provincial leaders? in-province time is extremely short, they are more likely to appoint agency 
heads without inside-agency experience, but only in the high-profile high-risk agencies. According 
76 
 
 
 
to Table 3.2, the bias against agency insiders lasts approximately 18 months and then attenuates 
as provincial leaders? in-province time increases. 
Second, I use a difference-in-differences estimation. For each high-profile high-risk agency 
in each province, I consider the appointments with in-province time within one year as being 
?treated?. From 1984 to 2012, 60 high-profile high-risk agencies in different provinces 
experienced at least one appointment with the in-province time being within one year43. For each 
of the ?treated? agencies, I match each appointment with those made in the same year and the same 
province but to the agencies that are not high-profile high-risk. For example, five agency heads 
were appointed to the Hunan Province Department of Transportation in 1988, 1993, 2006, 2008, 
and 2010. The then provincial party secretary?s in-province time before appointing the agency 
head in 2006 is within one year. Therefore, the agency head appointed to the Hunan Province 
Department of Transportation in 2006 (Ouyang Bin) is considered as ?treated? while the other four 
appointments are ?untreated?. For each of the five agency heads, I match them with the agency 
heads appointed in the same year in the agencies of the Hunan Province that are not high-profile 
high-risk. This ensures that the matched agency heads are all from the ?control? group and 
appointed by the same provincial party secretary. Ouyang Bin is matched with the agency head of 
the Hunan Province Department of Agriculture who was also appointed in 2006. If there is more 
than one agency head appointed to the other agencies in the same year, I take the average on 
whether they have inside-agency experience or not.  
The identification assumption is that the appointments in ?treated? and ?untreated? agencies 
in the same province follow the same secular trend. For example, the trend of appointing agency 
 
43 The total number of agency heads appointed to the high-profile high-risk agencies when the in-province time is 
smaller or equal to one year is 71. This is because some of the 60 agencies experience more than one appointment 
with the in-province time being within one year. 
77 
 
 
 
insiders or outsiders in 1988, 1993, 2006, 2008, and 2010 is the same for the Hunan Province 
Department of Transportation and its matched agencies. The years of appointments are centered 
around the year of being ?treated?. In this case, 1988, 1993, 2006, 2008, and 2010 are coded as ?2, 
?1, 0, 1, 2. 
  
2
??????????? = ?? + ? ???? + ???               (3.2) 
?=?2
 
I report the dynamic difference-in-differences effects using model (3.2). ? represents the 60 
treated agencies. ? represents the years of appointments. ? is the distance of the appointed agency 
head to the ?treated? one. When ? =  0, it means the appointed agency head is ?treated?: the in-
province time is within one year. When ? =  1, the agency head is ?untreated? and appointed as 
the successor of the ?treated?. When ? =  ?1, the agency head is ?untreated? and appointed as 
the predecessor of the ?treated?. ??????????? is the difference between ???????? and the average 
of its matched agency heads. ?? is the fixed effects of the treated agencies. ?? is a set of dummy 
variables indicating whether the appointment is ?treated?, before the ?treated?, or after the 
?treated?. I include a window of ?2 to 2. ?0 estimates the ?treatment? effect. ??2 and ??1 test 
whether a pre-trend exists. ?1 and ?2 test whether a post-trend exists. If neither a pre-trend nor a 
post-trend exists, it helps validate the parallel trend assumption. 
Figure 3.1 plots the dynamic difference-in-differences effects. In the top panel, I consider 
?treated? appointments when the in-province time is within one year. In the bottom panel, I 
consider ?treated? appointments when the in-province time is within one year and three months as 
an alternative way to identify ?treated? appointments. This categorization includes more 
78 
 
 
 
observations as ?treated? appointments. The results show a clear picture of the normal appointment 
pattern being interrupted by short in-province time. When the appointed agency heads are ?treated? 
? appointments made in the high-profile high-risk agencies when the in-province time is short 
(within one year or one year and three months) ? provincial leaders are less likely to appoint agency 
heads with prior experience inside the agency. The agency heads appointed before and after the 
?treated? ones are not affected, which provides supportive evidence to the parallel trend 
assumption. 
 
 
Figure 3. 1: Short In-Province Time and Appointments (Dynamic DID Effects) 
I calculate the average treatment effects by comparing ???????????  of the ?treated? 
appointments in each ?treated? agency with their immediate predecessor (? =  ?1) and successor 
(? =  1) and then aggregate the effects across each ?treated? agency. The average treatment effect 
for the agency heads appointed with the in-province time smaller or equal to one year is ?0.165. I 
79 
 
 
 
conduct a randomization inference by simultaneously randomizing the assignment of six high-
profile high-risk agencies and the assignment of in-province time. I generate 10000 new datasets. 
For each randomly generated dataset, I repeat the matching process above and calculate the 
average treatment effect by comparing the placebo ?treated? agency heads with their immediate 
predecessor and successor. Figure 3.2 plots the actual effect (?0.165) with the distribution of the 
10000 placebo effects. The distribution of the placebo effects is centered around 0. The actual 
effect is located at the 6th percentile of the lower end.  
 
 
Figure 3. 2: Average DID Effects and Randomization Inference 
Overall, the results support the argument that shared work experience affects provincial 
leaders? decisions of appointing agency insiders or outsiders. When provincial leaders lack shared 
work experience with sub-provincial bureaucrats ? indicated as particularly short in-province time 
? they are less likely to appoint agency heads with prior inside-agency experience, but only in the 
80 
 
 
 
high-profile high-risk agencies. As argued in Chapter 2, the strategy of appointing outsiders is to 
mitigate agency heads? information advantages that provide them with more opportunities and 
capacities to engage in corruption, especially collusive corruption that involves businesspeople 
from the outside and colleagues from the inside. The bias against agency insiders disappears when 
it is provincial leaders? second year in the province, suggesting that increased shared work 
experience helps alleviate provincial leaders? concern about agency heads? information advantages 
and the associated corruption risks. Due to the relatively small number of observations of the 
?treated? appointments, some of the results are only statistically significant at the 10 percent level. 
I conduct several robustness checks on these main findings in Section 3.2. I discuss alternative 
explanations in Section 3.3. 
 
3.2 Robustness Checks 
First, I report the agency-specific difference-in-differences effect. For each of the high-profile 
high-risk agencies, I repeat the above process of matching and comparing ???????????  of the 
?treated? agency heads (in-province time within one year) with their immediate predecessor and 
successor. I aggregate the average difference-in-differences effect for each agency. For each of the 
?control? agencies that are not high-profile high-risk, I estimate the same difference-in-differences 
effect by matching with the other ?control? agencies excluding itself. Figure C.1 in Appendix C 
plots the agency-specific effects and orders them by size from positive to negative. Except for the 
Department of Land and Resources, the effects of the other five high-profile high-risk agencies are 
all negative and located at the lower end of the distribution. The top three agencies with negative 
effects are the Development and Reform Commission, the Department of Construction, and the 
81 
 
 
 
Department of Transportation. The exception of the Land and Resources might be driven by the 
smaller sample size as the provincial department was established in 2000. In addition, as discussed 
in Section 2.1, appointments in the Department of Land and Resources may also be affected by 
another logic that is different from what the theoretical framework suggests ? provincial leaders 
prefer to appoint agency heads with more political experience to ensure they are tough enough to 
meet the ?hard? targets. Therefore, increased in-province time does not necessarily lead provincial 
leaders to appoint agency heads from inside the agency because provincial leaders may opt to 
appoint agency outsiders for policy considerations. 
Second, the key identification assumption of the cross-sectional comparison in model (3.1) 
is that the variation of in-province time in different provinces in the same year and within the same 
type of agencies could be considered as ?as-if? random. As argued above, the variation of in-
province time is jointly determined by the past pattern of appointments and reshuffles of provincial 
party secretaries and provincial agency heads. Particularly, approximately 60 percent of the 
vacancies are created because the incumbent agency heads reach the retirement age. The variation 
of in-province time among the appointments to replace the retired agency heads could be more 
convincingly argued as ?as-if? random. Table C.1 in Appendix C shows provincial party 
secretaries? ??? year in the province and the proportion of newly appointed agency heads from 
1984 to 2012 with prior experience inside the agency, only including the appointments of replacing 
retired agency heads. Table C.1 presents the same pattern as in Table 3.1: in the high-profile high-
risk agencies, fewer agency insiders are appointed in provincial party secretaries? first year in the 
province, but the bias disappears in their second year in the province. Specifically, among the 
agency heads appointed in the high-profile high-risk agencies when the in-province time is within 
one year, the proportion of having prior work experience inside the agency is approximately 45 
82 
 
 
 
percent. Among those appointed when the in-province time is above one year, the proportion is 56 
percent. Table C.2 reports the marginal effects of the cross-sectional comparison as in Table 3.2. 
Due to the smaller sample size the number of fixed effects, the marginal effects of in-province 
time on the appointments in the high-profile high-risk agencies are not statistically significant. But 
the direction and size of the effects support the pattern of bias against agency insiders. The same 
as in Table 3.2, the effects are reversed for appointments in the other agencies. The average 
difference-in-differences effect calculated by comparing ???????????  of the ?treated? agency 
heads (in-province time within one year) in the high-profile high-risk agencies with the immediate 
predecessor and successor is ?0. 202, the size of which is larger than the effect using all the 
appointments. 
Third, I test whether the same pattern of bias against agency insiders exists for agency heads 
appointed from 1978 to 1983 and 2013 to 2020. As the authority of appointing provincial agency 
heads was delegated to the provincial level after 1984, the appointment of provincial agency heads 
was not decided by provincial party secretaries between 1978 and 1983. Therefore, the variation 
of provincial party secretaries? in-province time should not affect the 563 agency heads appointed 
during this period. The 892 agency heads appointed between 2013 and 2020 are affected by Xi 
Jinping?s anti-corruption campaign. As discussed in Chapter 2 and empirically tested in Chapter 
4, provincial leaders are expected to appoint more agency heads without prior experience inside 
the agency after increased authoritarian monitoring to avoid being implicated by their appointees? 
misbehavior. Under such a context, the effect of in-province time might be attenuated or even 
reversed (see Chapter 4 for details). Therefore, I use the periods from 1978 to 1983 and 2013 to 
2020 as two ?placebo? periods where the pattern of bias against insiders should not exist. If the 
evidence shows the same, the effect of in-province time in the period between 1984 and 2012 is 
83 
 
 
 
more likely to occur not by chance. As Table C.3 to Table C.6 present, the pattern of bias against 
agency insiders in Table 3.2 does not exist in either the period of 1978-1983 or 2013-2020.  
Finally, I replace in-province time with provincial party secretaries? tenure in office. For 
provincial party secretaries who have been members of the Provincial Party Standing Committee 
before assuming office, the two variables are not identical ? even when it is their first year in office, 
they are not completely clueless about their subordinates in the province. Therefore, for the 
provincial leaders with prior experience in the Provincial Party Standing Committee, there should 
not be a bias against agency insiders in their first year in office. For this subgroup of provincial 
party secretaries, Table C.5 reports their ??? year in office and the proportion of newly appointed 
agency heads from 1984 to 2012 with prior experience inside the agency. In provincial party 
secretaries? first year in office, the proportion of appointed agency heads with inside-agency 
experience is 53 percent in the high-profile high-risk agencies, which is even 10 percentage points 
higher than those appointed in their second year in office. For the same subgroup of cases, Table 
C.6 reports the marginal effects from the cross-sectional comparison using provincial leaders? 
tenure in the office instead of their in-province time. The marginal effects for both the high-profile 
high-risk agencies and the other agencies are small. The results suggest that what drives provincial 
leaders? bias against agency insiders is their in-province time rather than their tenure in office. 
 
3.3 Alternative Explanations 
The main results and robustness checks show that provincial leaders are more likely to appoint 
agency heads without prior experience inside the agency when their in-province time is extremely 
short, but only in the agencies that are high-profile high-risk. I interpret the results as provincial 
84 
 
 
 
leaders? strategy of reducing agency heads? information advantages and the associated corruption 
risks when they lack information on sub-provincial bureaucrats. The bias disappears soon when 
they have spent some time in the province because shared work experience helps mitigate the 
problem of adverse selection and moral hazard. Section 2.1 discusses three alternative explanations 
? incentive-assignment conflict, patron-client relationships, and human capital or political 
experience ? for why provincial leaders sometimes appoint agency heads from outside of the 
agency. To validate the causality between in-province time and the appointment of agency insiders 
and outsiders, what matters is whether other variables coincide with the length of provincial party 
secretaries? in-province time and thus confound the relationship. 
First, it is unclear how the incentive-assignment conflict correlates with the length of 
provincial party secretaries? in-province time. As discussed in Chapter 2, the conflict is the result 
of the dual role of promotion and the hierarchical structure of the party-state organization. It is 
more applicable to some agencies than others. As Figure 2.1 shows, the proportions of agency 
heads promoted from a lower administrative rank are smaller than 50 percent for all the six high-
profile high-risk agencies. The incentive-assignment conflict is thus not the main explanation for 
why provincial leaders appoint agency outsiders in the high-profile high-risk agencies. In addition, 
there is no theoretical reason to believe that the incentive-assignment conflict is more serious when 
provincial party secretaries? in-province is short.   
Second, do provincial leaders appoint agency outsiders because outsiders have political 
connections with them? My argument and findings suggest otherwise. On the one hand, it requires 
time to form patron-client relationships. When provincial leaders? in-province time is within one 
year, they are unlikely to have the time to build connections with some sub-provincial bureaucrats. 
One of the purposes of frequently rotating provincial leaders is to disrupt their local connections. 
85 
 
 
 
On the other hand, the theoretical framework in Chapter 2 suggests that patron-client relationships 
are expected to mitigate the problem of moral hazard and thus encourage provincial leaders to 
appoint more agency heads with prior inside-agency experience. The length of in-province time is 
a proxy for provincial leaders? shared work experience with sub-provincial bureaucrats. As many 
patron-client relationships build on shared work experience and more specifically promotion links 
(Keller 2016, Jiang 2018), in-province time also indirectly measures the probability of having 
political connections.  
Due to data availability, I only have direct measures for the patron-client relationships 
between the then provincial party secretaries and the agency heads for those appointed between 
2005 and 2020 (see Figure 2.3). Figure 3.3 shows the correlation between patron-client 
relationships and provincial leaders? appointment strategies for the agency heads appointed 
between 2005 and 2020. The panel above uses the broadly defined promotion links ? I code the 
existence of patron-client relationships if agency heads have experienced any sorts of promotions 
under the same provincial party secretaries. Among the 607 agency heads appointed during this 
period, 130 of them have patron-client relationships with the then provincial party secretaries. The 
results reveal different effects of having patron-client relationships for appointments in the two 
types of agencies. Only in the agencies that are not high-profile high-risk are outsiders more likely 
to be appointed when they have connections with the then provincial party secretaries. On the 
contrary, in the six high-profile high-risk agencies, the existence of patron-client relationships is 
associated with a higher proportion of agency insiders.  
 
86 
 
 
 
 
Figure 3. 3: Patron-Client Relationships and Appointments (2005-2012) 
The panel below shows the results using the strictly defined promotion links ? I code the 
existence of patron-client relationships only if agency heads have been previously promoted from 
the Deputy Bureau Rank to the Bureau Rank under the same provincial party secretaries. Only 83 
agency heads are considered politically connected using this standard. In the agencies that are not 
high-profile high-risk, there is a clear pattern of appointing outsiders who are politically connected. 
While in the high-profile high-risk agencies, patron-client relationships do not correlate with a 
higher proportion of agency insiders, the difference is small. Taken the results of Figure 3.3 and 
Figure 3.4 together, whether provincial leaders appoint agency outsiders for reasons of political 
connections depends on the type of agencies. There is strong supportive evidence among the 
87 
 
 
 
appointments in the agencies that are not high-profile high-risk. However, the relationship between 
political connections and the appointment of agency insiders and outsiders is ambiguous in the 
high-profile high-risk agencies, at least for the agency heads appointed between 2005 and 2012.  
 
 
Figure 3. 4: In-Province Time and Age of the Agency Heads 
Third, when provincial leaders are just parachuted into the province, do they seek agency 
heads from the outside with specific qualifications? For example, provincial leaders who are 
newcomers to the province may want to appoint agency heads with particularly high human capital 
in the important agencies with high corruption risks. If the candidates among agency insiders are 
old and not well educated, provincial leaders may seek agency heads from the outside who are 
younger and better educated. Section 2.1 compares various characteristics of agency heads with 
and without prior experience inside the agency. On average, agency outsiders are not different 
from insiders in terms of age and education levels. The question here is whether agency outsiders 
appointed when provincial leaders? in-province time is short (within one year) are younger and 
88 
 
 
 
better educated. Due to data availability, Figure 3.4 plots the age of the agency heads at the time 
of their appointments among those appointed between 2005 and 2012. There is no evidence that 
the agency outsiders appointed in the high-profile high-risk agencies in provincial leaders? first 
year in the province are younger ? the reverse is true. Figure 3.5 compares their highest education 
levels. Agency outsiders appointed in the high-profile high-risk agencies in provincial leaders? 
first year in the province are not better educated. Again, the reverse is true.  
 
 
Figure 3. 5: In-Province Time and Agency Heads? Highest Education Levels 
It is also plausible that provincial leaders who are newcomers to the province want to appoint 
agency heads with more political experience in the six high-profile high-risk agencies. As Figure 
2.6 in Section 2.1 shows, agency heads appointed from outside of the agency are indeed more 
89 
 
 
 
likely to have more political experience. The question here is whether the agency outsiders 
appointed in the high-profile high-risk agencies in provincial leaders? first year in the province are 
particularly more likely to have more political experience. Figure 3.6 shows that this is not the 
case ? the difference of political experience between agency insiders and outsiders does not covary 
with provincial leaders? in-province time. Therefore, the bias against agency insiders is unlikely 
driven by a particular preference for political experience when provincial leaders? in-province time 
is short. 
 
 
Figure 3. 6: In-Province Time and Agency Heads? Political Experience 
Relatedly, I test whether the bias against agency insiders is driven by a general preference 
against specialists when provincial leaders? in-province time is short. Among the agency heads 
without inside-agency experience, I repeat the cross-sectional comparison in model (3.1) using 
relevant experience as the dependent variable. I compare agency outsiders in the same type of 
agencies appointed in the same year in different provinces. If there is a general bias against 
90 
 
 
 
specialists when provincial leaders are just parachuted into the province, the proportion of agency 
outsiders with relevant experience should be lower when provincial leaders? in-province time is 
short. Table 3.3 shows that this is not the case for the high-profile high-risk agencies. Among the 
agency outsiders appointed to the high-profile high-risk agencies, those appointed when provincial 
leaders? in-province time is within one year are not less likely to have relevant work experience.  
Table 3. 3: Marginal Effects of In-Province Time on Specialists Outside of the Agency 
 DV: relevant work experience outside of the agency 
 T ? 1 T ? 1 year T ? 1 year T ? 1 year T ? 2 
year and 3 months and 6 months and 9 months years 
 (1) (2) (3) (4) (5) 
High-profile high-risk agencies      
Marginal effects: ?1 + ?2 0.014 -0.014 -0.043 -0.074 -0.086 
 0.079 0.081 0.071 0.066 0.063 
Mean of the dependent variable 0.492 0.492 0.492 0.492 0.492 
N of agency heads appointed 
44 62 72 78 90 
within T 
The other agencies      
Marginal effects: ?1 -0.094 -0.065 -0.045 -0.051 -0.054 
 0.063 0.056 0.052 0.043 0.042 
Mean of the dependent variable 0.430 0.430 0.430 0.430 0.430 
N of agency heads appointed 
104 138 158 187 210 
within T 
Note: The sample includes only agency heads without inside-agency experience. Standard errors 
clustered at the level of provincial party secretaries are reported in the parentheses. 
 
3.4 Summary 
This chapter empirically tests how provincial leaders? information on sub-provincial bureaucrats 
affect their decisions on appointing agency insiders and outsiders. I adopt two empirical strategies. 
First, I leverage the cross-sectional variation of in-province time in different provinces for agency 
91 
 
 
 
heads appointed in the same type of agencies in the same year. Second, I use a difference-in-
differences design by first matching the agency heads in each ?treated? agency (the high-profile 
high-risk agencies with at least one agency head appointed when the then provincial party 
secretary?s in-province time is within one year) with those from the ?control? agencies (agencies 
that are not high-profile high-risk) in the same province appointed in the same years. I subject the 
main findings to several robustness checks and rule out relevant alternative explanations. 
First, provincial leaders are more likely to appoint agency heads without inside-agency 
experience in the high-profile high-risk agencies when their in-province time is short (within one 
year or within one year and three months). The bias against agency insiders soon disappears in 
provincial leaders? second year in the province. Second, I discuss specific mechanisms that lead 
to the pattern of appointing insiders and outsiders. Based on the theoretical discussion in Chapter 
2, I interpret the strategy of appointing outsiders as a preventive measure to reducing agency heads? 
information advantages and the associated corruption risks when provincial leaders lack 
information on their subordinates. When provincial leaders have some shared work experience 
with their subordinates, they are encouraged to appoint agency insiders because shared work 
experience helps mitigate the problem of adverse selection (through repeated interactions) and 
moral hazard (through patron-client relationships). The results are unlikely driven by the incentive-
assignment conflict because this is a structural factor that does not covary with the length of 
provincial leaders? in-province time. The bias against agency insiders in the high-profile high-risk 
agencies in provincial leaders? first year in the province is contrary to the argument that outsiders 
are appointed because of their political connections. I also empirically demonstrate that the bias is 
not driven by provincial leaders? particular preferences on agency heads? specific qualifications ? 
such as higher human capital or more political experience ? when they are just parachuted into the 
92 
 
 
 
province. In addition, although provincial leaders appoint fewer agency heads with prior 
experience inside the agency when their in-province time is short, those from outside of the agency 
appointed when the in-province time is short are not less likely to have relevant work experience. 
I will discuss the implications of these findings in Chapter 6. 
   
  
   
 
 
 
 
 
 
  
 
 
 
93 
 
 
 
Chapter 4: Authoritarian Monitoring and the Appointment Dilemma 
 
This chapter tests how increased monitoring affects provincial leaders? decisions of appointing 
agency heads with and without prior inside experience. Specifically, I study how Xi Jinping?s anti-
corruption campaign starting from December 2012 affects the selection of provincial agency heads. 
The theoretical discussion in Chapter 2 presents two competing arguments. On the one hand, as 
Xi Jinping?s anti-corruption campaign increases the CCP?s ex-post monitoring capacity, it should 
ease provincial leaders? ex-ante concerns about preventing corruption and encourage them to 
prioritize improving government efficiency when making appointment decisions. If true, 
provincial leaders are expected to appoint more agency heads with prior experience inside the 
agency in important agencies with high corruption risks after the campaign. On the other hand, the 
authoritarian features of the campaign ? top-down control, political uncertainty, and intra-party 
propaganda ? could lead provincial leaders to take more preventive measures at the appointment 
stage to avoid being implicated by their appointees? misbehavior. If true, provincial leaders are 
expected to appoint fewer agency heads with inside-agency experience in the high-profile high-
risk agencies after the campaign. Section 4.1 tests the two competing arguments. Section 4.2 
reports the robustness checks on the main findings. Section 4.3 discusses and empirically tests the 
alternative explanations. Section 4.4 summarizes. 
 
4.1 Main Results 
I study the 1474 agency heads appointed between 2005 and 2020 ? eight years before and after Xi 
Jinping?s anti-corruption campaign. I adopt two identification strategies. First, I estimate a simple 
94 
 
 
 
before-after comparison. For agency heads of the same agency in the same province, I compare 
those appointed before and after the anti-corruption campaign. The identification assumption is 
that Xi Jinping?s anti-corruption campaign is the only factor that affects the appointments within 
the same agency of the same province before and after 2013. Specifically, I estimate the before-
after changes in the high-profile high-risk agencies and the other agencies using the following 
linear probability model, 
 
????????? = ?1???? ????????? + ?2???? ????????? ? ????? + ??? + ????               (4.1) 
 
?, ?, and ? are the indexes for province, agency, and year of appointment. ????????? is a dummy 
variable that takes the value of 1 if the newly appointed agency head in province ?, agency ?, and 
year ? has prior work experience inside the agency. ???? ????????? is a dummy variable that 
takes the value of 1 if appointments are made in or after 2013. ????? is a dummy variable that 
takes the value of 1 if agency ? is categorized as one of the six high-profile high-risk agencies. ??? 
is provincial-agency fixed effects. Model (4.1) effectively compares appointments made to the 
same agency of the same province before and after 2013. I estimate the model using five different 
bandwidths: 2009-2016, 2008-2017, 2007-2018, 2006-2019, and 2005-2020. A narrower 
bandwidth ensures that the comparison is made immediately before and after the campaign, which 
is more likely to satisfy the identification assumption that the campaign is the only factor that 
affects the appointments before and after 2013. But as new appointments do not occur every year, 
a wider bandwidth includes more observations. Different bandwidths strike a balance between the 
trade-offs. For each bandwidth, ?1 + ?2 is the average proportional change of newly appointed 
agency heads with inside-agency experience after the campaign in the high-profile high-risk 
95 
 
 
 
agencies. ?1 is the average proportional change of newly appointed agency heads with inside-
agency experience after the campaign in the other agencies. I cluster standard errors at the level of 
provincial party secretaries.44  
Table 4. 1: Before-After Changes 
 DV: inside-agency work experience 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign -0.057 0.013 0.014 -0.004 -0.004 
 (0.048) (0.031) (0.026) (0.025) (0.024) 
Post campaign ? High-
-0.083 -0.140 -0.130 -0.101 -0.102 
profile high-risk 
 (0.088) (0.060) (0.050) (0.048) (0.045) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Number of province-
405 434 440 440 440 
agency 
Number of observations 662 962 1213 1325 1474 
      
High-profile high-risk 
     
agencies: 
Mean (before) 0.434 0.445 0.447 0.427 0.425 
Before-after (?1 + ?2) -0.141 -0.127 -0.116 -0.106 -0.106 
 (0.070) (0.055) (0.046) (0.044) (0.040) 
The other agencies:      
Mean (before) 0.230 0.243 0.248 0.261 0.268 
Before-after (?1) -0.057 0.013 0.014 -0.004 -0.004 
 (0.048) (0.031) (0.026) (0.025) (0.024) 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
 
44 For the agency heads appointed between 2005 and 2020, there are 108 provincial party secretaries in 22 provinces. 
As the number of observations per cluster is small, I adjust the clustered standard errors when the fixed effects models 
are estimated dummy variable approach (Cameron and Miller 2015, pp. 14-15)  
96 
 
 
 
Table 4.1 reports the results from the before-after estimations. Columns (1) to (5) present 
the results from different bandwidths. For the high-profile high-risk agencies, the results 
consistently show that the proportion of newly appointed agency heads with prior inside-agency 
experience decreases after 2013. From the narrowest to the widest bandwidth, the before-after 
changes (?1 + ?2) are ?0.141, ?0.127, ?0.116, ?0.106, and ?0.106, respectively. The effects are 
all statistically significant at the 5 percent level. Substantively, the size of the decrease is large. 
Compared with the average proportion of agency heads with inside-agency experience before the 
campaign, it is equivalent to a 25 percent drop. The before-after changes in the other agencies (?1) 
are small. From the narrowest to the widest bandwidth, the effects are ?0.057, 0.013, 0.014, 
?0.004, and ?0.004. None of them are statistically significant. The before-after estimations suggest 
that provincial leaders tend to appoint fewer agency heads with prior experience inside the agency 
after Xi Jinping?s anti-corruption campaign, but only in the high-profile high-risk agencies. The 
results are contrary to the argument that increased monitoring could ease provincial leaders? 
concerns about corruption risks when making appointment decisions. Instead, the findings suggest 
that authoritarian monitoring leads provincial leaders to take more preventive measures. 
Second, I leverage variations from both the appointment time before and after Xi Jinping?s 
anti-corruption campaign and the different types of agencies. I use the appointments in the agencies 
that are not high-profile high-risk as the ?control? group. The estimation of interest is the relative 
differences between appointments made in the high-profile high-risk agencies and the other 
agencies before and after the campaign. The key identification assumption is that the appointment 
pattern in the high-profile high-risk agencies (?treated? agencies) follows the same trend of the 
pattern in the other agencies (?control? agencies). Figure 4.1 plots the proportion of newly 
appointed agency heads with inside-agency experience in the two types of agencies each year. The 
97 
 
 
 
proportion of having inside-agency experience is consistently higher in the high-profile high-risk 
agencies before 2013 except in 2006. The visual evidence also shows that the proportion of newly 
appointed agency heads with inside-agency experience follows a similar trend in the two types of 
agencies most of the time before 2013. The proportion of having inside-agency experience 
decreases after Xi Jinping?s anti-corruption campaign, but only in the high-profile high-risk 
agencies. A closer look shows that the drop did not occur immediately in 2013 but after 2014. I 
will discuss the phenomenon later.  
 
 
Figure 4. 1: Proportion of Having Inside-Agency Experience (2005-2020) 
Based on Figure 4.1, I calculate the dynamic difference-in-differences effects using the 
nonparametric approach proposed by Callaway and Sant?Anna (2021). Specifically, for agency 
heads appointed each year since 2013, I partition the data to include agency heads appointed in 
that year and those appointed in 2012 ? one year before the campaign. For each partition of the 
data, I calculate the two-period difference-in-difference effects. For example, to calculate the 
98 
 
 
 
difference-in-differences effect in 2014, I include the appointments made in 2014 and 2012. 
Agency heads appointed in the high-profile high-risk agencies constitute the ?treated? group and 
those appointed in the other agencies constitute the ?control? group. The year 2014 is considered 
as the ?post-treatment? period. The year 2012 is considered as the ?pre-treatment? period. For 
agency heads appointed each year before 2013, I partition the data to include those appointed in 
that year and one year before. The effects before 2013 estimate whether a pre-trend exists, which 
tests the parallel trend assumption more formally. The advantage of this approach is that it 
generates interpretable causal parameters that avoid issues of interpreting two-way fixed effects in 
multiple time periods in the presence of heterogenous treatment effects (Callaway and Sant?Anna 
2021, p. 201). A caveat is that new agency heads are not appointed every year. Therefore, the 
dynamic effect of each year is estimated from appointments in different agencies and provinces. 
 
 
Figure 4. 2: Nonparametric Dynamic DID Effects (2005-2020) 
 Figure 4.2 plots the nonparametric dynamic difference-in-differences effects. The 
parentheses report the total number of appointments in that year. There is not an obvious pre-trend 
99 
 
 
 
for the agency heads appointed before 2013. The effects from 2013 are consistently negative 
except in 2014. Although there seems to be a spike in 2014, the total number of agency heads 
appointed in 2014 is only 35 with 12 in the high-profile high-risk agencies. The pattern of 
appointing outsiders is more salient since 2015, although none of the single-year effects are 
statistically significant at the 5 percent level. The average effect for the agency heads appointed 
between 2005 and 2020 (a weighted average of all single-year effects with weights proportional 
to group size) is ?0.117. Relative to the ?control? group (agencies that are not high-profile high-
risk), the proportion of newly appointed agency heads with inside-agency experience in the 
?treated? group (agencies that are high-profile high-risk) decreases after Xi Jinping?s anti-
corruption campaign. 
Table 4. 2: Average DID Effects 
 DV: inside-agency work experience 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign ? High-
-0.106 -0.154 -0.135 -0.108 -0.110 
profile high-risk 
 (0.085) (0.059) (0.049) (0.047) (0.044) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Year fixed effects Yes Yes Yes Yes Yes 
Number of observations 662 962 1213 1325 1474 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
I also estimate the average difference-in-differences effects under different bandwidths using 
the two-way fixed effects model. In model (4.2), I estimate the interaction term 
???? ????????? ? ?????  with provincial-agency fixed effects ???  and year fixed effects ?? . 
Under parallel trend and linearity assumptions, ? could be interpreted as the average treatment 
100 
 
 
 
effect. Table 4.2 reports the results from model (4.2) using different bandwidths. The coefficients 
of ? range from ?0.106 to ?0.154. The ? from the bandwidth of 2005-2020 is ?0.110, the size of 
which is comparable to the effect calculated using the nonparametric approach. The effects are all 
statistically significant at the 5 percent level except for the model using the narrowest bandwidth. 
 
????????? = ??? + ?? + ????? ????????? ? Riskj + ????               (4.2) 
 
Both the before-after and difference-in-differences estimations show that provincial leaders 
tend to appoint fewer agency heads with prior inside-agency experience in the high-profile high-
risk agencies after Xi Jinping?s anti-corruption campaign. According to the theoretical discussion 
in Chapter 2, the strategy of appointing outsiders is a preventive measure for provincial leaders to 
avoid being implicated by their appointees? misbehavior. The authoritarian features of Xi Jinping?s 
anti-corruption campaign make provincial leaders take more preventive measures to avoid being 
victims of the campaign themselves. I conduct several robustness checks in Section 4.2 and discuss 
alternative explanations in Section 4.3. 
  
4.2 Robustness Checks 
First, I report the before-after changes in each agency. I calculate the simple average differences 
between the agency heads appointed before and after 2013 for each agency. I rank the agencies by 
the direction and size of the changes. Figure D.1 in Appendix D plots the before-after changes for 
each agency and rank the agencies by the results from 2006 to 2019. For all the six high-profile 
high-risk agencies, the proportions of newly appointed agency heads with inside-agency 
101 
 
 
 
experience all decrease after 2013. Compared with the other agencies, they are all located at the 
lower end of the distribution. The proportion of having inside-agency experience decreases 
especially salient in the Department of Housing and Urban-Rural Development (Department of 
Construction), the Department of Land and Resources, the Economic and Information Technology 
Commission (Economic Affairs Commission), and the Department of Transportation. Among the 
14 agencies that are not high-profile high-risk,45 there are only obvious decreases in the Health 
Commission, the Department of Science and Technology, and the Ethnic Affairs Commission. The 
changes in the other agencies are either small or slightly positive. Interestingly, the proportion of 
newly appointed agency heads with inside-agency experience increases significantly in the 
Department of Audit. On the one hand, according to Figure 2.7, the Department of Audit is among 
the low-profile low-risk agencies. On the other hand, there may be an increased demand for the 
expertise of auditing after the anti-corruption campaign. 
Second, among the high-profile high-risk agencies, I calculate the before-after changes in 
each province. I rank the provinces by the direction and size of the changes. Figure D.2 in 
Appendix D plots the before-after changes among the high-profile high-risk agencies in each 
province and rank the provinces by the results from 2006 to 2019. The before-after changes in 16 
out of 22 provinces are consistently negative under different bandwidths, which suggests that the 
decrease of newly appointed agency heads with inside-agency experience is not driven by a small 
number of provinces. Furthermore, Figure D.3 plots the binary relationship between the ranking 
of a provincial ?cleanness? index and the ranking of before-after changes. I use the ?cleanness? 
index constructed by (Nie, et al. 2018). For each province, the index is calculated as the weighted 
 
45 The Family Planning Commission is not included because the agency is merged into the Health Commission after 
2013.  
102 
 
 
 
average from the number of officials investigated (12.5 percent), the number of news reports on 
corruption (12.5 percent), the level of administrative information disclosure (37.5 percent), and 
the level of financial transparency (37.5 percent). The result does not show any strong correlation 
between the two. For several provinces that are ranked as ?clean? ? such as Zhejiang, Hainan, 
Hebei, and Jiangsu ? the proportion of newly appointed agency heads with inside-agency 
experience does not change much or slightly increase. For several provinces that are ranked as 
?corrupt? ? such as Jilin, Hunan, and Qinghai ? the proportion of having inside-agency experience 
decreases by a large size. Henan and Hubei are two exceptions. But even for the provinces with 
relatively low or medium levels of corruption ? such as Shandong, Shaanxi, Fujian, and Guizhou 
? the proportion of newly appointed agency heads with inside-agency experience also decreases. 
These results suggest that the strategy of appointing outsiders is driven by a commonly shared 
incentive of reducing corruption risks among most provincial leaders. 
Third, the anti-corruption campaign may concurrently affect the supply of agency insiders. 
If one or more deputy agency heads in an agency are investigated, the candidate pool of agency 
insiders is reduced in the first place. As bureaucrats in the high-profile high-risk agencies are more 
likely to be investigated, the decrease of having inside-agency experience in these agencies could 
be a result of the supply shock. For each appointment made in and after 2013, I code whether there 
are any deputy agency heads removed from the positions within the last three years. There are 25 
cases in the high-profile high-risk agencies and 30 cases in the other agencies where appointments 
were made when there are deputy agency heads removed from the positions because of anti-
corruption investigation. As expected, the proportion of agency heads with inside-agency 
experience is low in these cases. In the high-profile high-risk agencies, 5 out of the 25 
appointments are those with prior experience inside the agency. In the other agencies, the 
103 
 
 
 
proportion is 7 out of 30. To show that the results are not only driven by the supply shock, Table 
D.1 and Table D.2 report the before-after and the average difference-in-differences effects after 
excluding these cases. Compared with the results in Table 4.1 and Table 4.2, the size of the 
coefficients only drops by about 1 percentage point. The results are still statistically significant at 
or around the 5 percent level. Even after dropping the cases where appointments are concurrently 
affected by the reduced supply of agency insiders, the pattern of appointing outsiders in the high-
profile high-risk agencies holds. 
The anti-corruption campaign might reduce the supply of agency insiders through another 
mechanism ? increased monitoring could push some bureaucrats out to private sectors. If those 
with inside-agency experience are more likely to leave, the candidate pool of agency insiders will 
also be smaller after the campaign. However, this is unlikely to be the case because resignations 
are rare for bureaucrats at or above the Deputy Bureau Rank. Although systematic data is 
unavailable, the media sources suggest that the number of resignations does not reach the scale 
that could change the supply. According to ?Shanghai Observer?, a total of 10 officials from the 
central and local government left for private sectors from 2014 to 2015.46 According to ?Jiemian?, 
a total of 7 officials in the Shandong Province resigned and joined private sectors from 2015 to 
2019.47 
Fourth, the identification strategies in Table 4.1 and Table 4.2 assume that the anti-
corruption signals spread equally to each province since 2013. While it is a reasonable assumption, 
one can alternatively assume that different provinces receive the signals at different times. Among 
the 22 provinces, the central inspection teams visited 9 provinces in 2013 and visited the other 13 
 
46 https://www.shobserver.com/news/detail?id=6562 
47 https://m.jiemian.com/article/3251181.html 
104 
 
 
 
provinces in 2014. If the effects of central inspections do not spillover across provinces, the 
?treatment? period should start differently for the two cohorts of provinces. As a robustness check, 
I use one year after central inspections as the start of ?treatment? for the high-profile high-risk 
agencies in each province. Figure D.4 reports the dynamic difference-in-differences effects using 
the nonparametric approach (Callaway and Sant?Anna 2021). Starting from one year after the 
central inspection team visited the province, the effects are consistently negative. Comparing 
Figure D.4 with Figure 4.2, the reason why the proportion of having inside-agency experience only 
starts to drop after 2014 may be the result of the timing of central inspections. Table D.3 reports 
the average treatment effects using the two-way fixed effects models under different bandwidths. 
The effects range from ?0.104 to ?0.166, which is close the effects in Table 4.2 where the post-
treatment period is the same for all provinces. In addition, I use a randomization inference 
by randomly assigning 9 provinces to be inspected in 2013 and the other 13 provinces to be 
inspected in 2014. I also randomly assign 6 agencies to be in the high-profile high-risk type. 
I repeat the randomization 10000 times.48 Figure D.5 plots the actual effects together with the 
10000 placebo effects. The placebo effects center around 0 and the actual effect is located at the 1 
percentile of the distribution. 
Finally, I replace the outcome variable from having inside-agency experience to relevant 
experience ? either inside or outside of the agency. As discussed in Chapter 2, higher corruption 
risks associated with agency heads who have prior experience inside the agency come from the 
information advantages of knowing the colleagues, protocols, and business clients of the agency 
well. Agency heads with relevant experience outside of the agency do not share the advantages. 
Therefore, the before-after changes and the difference-in-differences effects should be largely 
 
48 I report the results using the agency heads appointed between 2006 and 2019. 
105 
 
 
 
attenuated if the outcome variable is replaced by having relevant experience. Table D.4 and Table 
D.5 show that this is indeed the case. Instead, among the agency heads appointed from outside of 
the agency, the proportion of having relevant experience increases after the campaign in the high-
profile high-risk agencies (see Table D.6 and Table D.7). Similar to the findings on in-province 
time, these results suggest that the bias against agency insiders after Xi Jinping?s anti-corruption 
campaign is not driven by a change of preference against specialists in general. It seems that 
provincial leaders attempt to reduce corruption risks but compensate for the losses of agency-
specific expertise by appointing outsiders with relevant experience. 
 
4.3 Alternative Explanations 
The main results and robustness checks show that provincial leaders are more likely to appoint 
agency heads without prior experience inside the agency after Xi Jinping?s anti-corruption 
campaign. I interpret the results as provincial leaders? cautiousness to reduce information 
advantages and thus higher corruption risks associated with agency insiders. However, if some 
factors covary after the anti-corruption campaign, the strategy of appointing agency outsiders 
could reflect the change of these factors instead of the cautiousness to reduce corruption risks. This 
section examines whether the incentive-assignment conflict, patron-client relationships, and 
human capital or political experience confound the causal mechanism. 
First, if the severity of the incentive-assignment conflict changes after Xi Jinping?s anti-
corruption campaign, the increased proportion of agency heads from outside of the agency may be 
a result of the incentive-assignment conflict. For example, if the number of government agencies 
decreases after the campaign, there will be a reduced supply of positions (see Chapter 5). If the 
106 
 
 
 
number of candidates increases after the campaign, there will be an increased demand of positions. 
Both changes make the incentive-assignment conflict severer and may result in more agency heads 
appointed from outside of the agency. However, there is no evidence that either the supply or the 
demand of positions changes after Xi Jinping?s anti-corruption campaign. 
 
 
Figure 4. 3: Broadly Defined Patron-Client Relationships Before and After the Campaign 
Second, if provincial leaders prefer to appoint agency heads who have patron-client 
relationships with them after the campaign and if agency outsiders are more likely to have political 
connections, the strategy of appointing outsiders may be driven by the factor of patron-client 
relationships. However, as Chapter 2 demonstrates, prior patron-client relationships between 
provincial party secretaries and provincial agency heads are quite uncommon and are thus unlikely 
to be the dominant explanation. In addition, whether provincial leaders prefer to appoint those 
connected with them after the campaign is unclear. In the appointments of prefecture-level city 
leaders,  Li and Manion (2021) find that provincial leaders intentionally avoid promoting those 
who are connected with them to signal to the party center that they are not building their own 
107 
 
 
 
factions. Figure 4.3 reports the proportion of having patron-client relationships (broadly defined) 
among agency heads with and without inside-agency experience. For agency heads appointed after 
the campaign (2013-2020) in the high-profile high-risk agencies, those without prior work 
experience inside the agency are not more likely to have patron-client relationships. Agency 
outsiders in the high-profile high-risk agencies appointed after the campaign are also not more 
likely to have patron-client relationships compared with the outsiders appointed before the 
campaign. Therefore, the increased proportion of agency heads without inside-agency experience 
is unlikely to be driven by the factor of patron-client relationships. The results are similar when 
patron-client relationships are strictly defined (Figure 4.4).  
 
 
Figure 4. 4: Strictly Defined Patron-Client Relationships Before and After the Campaign 
In Figure 4.3, comparing the agency insiders appointed before and after 2013 in the high-
profile high-risk agencies, the proportion of having patron-client relationships decreases by 
approximately 10 percentage points. This is consistent with what Li and Manion (2021) find as the 
?decline of factions?. The pattern is similar but less salient in Figure 4.4. The finding suggests that 
108 
 
 
 
even when provincial leaders appointed agency heads with inside-agency experience after the anti-
corruption campaign, they wanted to distance themselves from corruption risks by intentionally 
appointing those not connected with them. However, the ?decline of factions? does not occur in 
the other agencies or among the agency outsiders appointed in the high-profile high-risk agencies. 
Such a ?selective? bias against factions suggests a different mechanism of how provincial leaders 
preserve themselves in Xi Jinping?s anti-corruption campaign: instead of signaling to the party 
center, provincial leaders are sincerely worried about corruption risks and the consequence of 
being implicated.   
 
 
Figure 4. 5: Higher Human Capital? (Age) 
Third, if provincial leaders want to appoint agency heads with higher human capital in the 
high-profile high-risk agencies after the campaign and if those from outside of the agency are 
younger or better educated, the increased proportion of agency outsiders may not be driven by the 
cautiousness to reduce corruption risks. However, Figure 2.4 and Figure 2.5 already show that on 
average this is not the case. Agency heads appointed from the outside are not younger or better 
109 
 
 
 
educated than those with inside-agency experience. Figure 4.5 plots the age of agency heads at the 
time of appointments for the two types of agencies before and after the campaign. In both the high-
profile high-risk and the other agencies, the ages of agency heads at the time of appointments are 
similar. Figure 4.6 and Figure 4.7 plot the highest education level for the two types of agencies 
before and after the campaign. 
 
 
Figure 4. 6: Higher Human Capital? (Master?s degree and above) 
If provincial leaders want to appoint agency heads with more political skills after the 
campaign, they may want to seek agency heads from outside of the agency. As Figure 2.6 shows, 
agency heads appointed from the outside indeed have more political experience. Although this 
alternative explanation is more difficult to rule out, Figure 2.6 shows that among the agency heads 
appointed from the outside, the proportion of those with more political experience decreases after 
the anti-corruption campaign. Figure 4.8 shows the proportion of having more political experience 
110 
 
 
 
in the two types of agencies before and after the campaign. The results show that the decreased 
proportion of having political experience among agency outsiders after the campaign is driven by 
  
 
Figure 4. 7: Higher Human Capital? (Doctorate degree) 
high-profile high-risk agencies ? the proportion decreases by more than 10 percentage points. 
Therefore, the strategy of appointing outsiders cannot be explained by an increased preference for 
political skills. Together with Table D.6 and Table D.7, the results suggest that provincial leaders 
appoint outsiders with relevant experience rather than political experience to substitute for insiders. 
 
4.4 Summary 
This chapter empirically tests how authoritarian monitoring affects provincial leaders? 
appointment strategies. Specifically, I leverage Xi Jinping?s anti-corruption campaign as a shock 
that strengthened the CCP?s monitoring capacity and compare provincial agency heads appointed 
111 
 
 
 
before and after the campaign. Both the before-after and difference-in-differences estimations 
using agencies that are not high-profile high-risk as the ?control? agencies show that provincial 
leaders appointed fewer agency heads with inside-agency experience in the high-profile high-risk 
agencies after the anti-corruption campaign. The results are contrary to the argument that increased 
monitoring capacity should ease provincial leaders? concerns about corruption risks at the 
appointment stage. Instead, the results suggest that authoritarian monitoring scares provincial 
leaders to prioritize reducing corruption risks by playing it safe and appointing outsiders. I also 
demonstrate that the increase of agency outsiders after the campaign cannot be explained by the 
incentive-assignment conflict, patron-client relationships, or seeking candidates with higher 
human capital and political skills. Interestingly, provincial leaders appointed outsiders with 
relevant experience to compensate for the losses of agency-specific expertise. I will discuss the 
implications of these findings in Chapter 6. 
 
 
Figure 4. 8: More Political Experience? 
112 
 
 
 
Chapter 5: Outsiders in the Low-Profile Low-Risk Agencies 
 
The delegation framework assumes that provincial leaders face an information-collusion dilemma 
when appointing agency heads. Agency heads with prior work experience inside the agency have 
information advantages ? they have agency-specific knowledge, connections with colleagues 
inside the agency, and connections with administrative clients outside of the agency. On the one 
hand, appointing agency heads with prior experience inside the agency could help reduce 
performance uncertainty. On the other hand, agency heads can also use their information 
advantages to engage in corruption. Provincial leaders? appointment strategy is to balance the risk 
of performance uncertainty and the risk of collusive corruption. As Section 2.5 discusses, this 
framework is more applicable to agencies that are important and susceptible to corruption because 
provincial leaders only face the information-collusion dilemma in these agencies. Chapter 3 and 
Chapter 4 empirically test the theory on the six high-profile high-risk agencies using the other 
agencies as the ?control? group. 
This chapter discusses why agency heads without prior inside-agency experience are 
appointed in the agencies that are not high-profile high-risk. The literature on bureaucratic 
appointments suggests that government positions can be used for multiple objectives and these 
objectives sometimes conflict with each other. Politicians need to balance government 
performance, power consolidation, and patronage when making appointment decisions (Geddes 
1994). Politicians often use a compartmentalization strategy to balance these objectives by 
applying different appointment standards in different types of positions. Empirical evidence from 
the United States, Latin America, and Africa shows that politicians arrange patronage 
113 
 
 
 
appointments in less important or low-skill agencies while maintaining meritocracy in important 
agencies (Geddes 1994, Hollibaugh Jr, et al. 2014, Brierley 2021).  
Provincial government agencies in China also serve more than one objective. In addition to 
their policy functions, provincial government agencies in China help maintain a certain rate of 
promotion and rotation of the large number of cadres. This non-policy function is pivotal for 
regime stability and the operation of the party-state. The party-state is a giant bureaucratic 
organization with many state employees and the CCP controls the appointments in important 
positions. The control of the vast majority positions in society helps limit state employees? exit 
options. The use of promotion as an incentive helps align state employees? interests with the party. 
Both are important in maintaining regime stability (Svolik 2012). However, the pyramid structure 
of bureaucratic organizations means that the limited supply of positions in higher ranks cannot 
meet the demand. Section 2.1 has discussed the incentive-assignment conflict or the Peter Principle: 
when someone needs to be rewarded with a promotion but there is no vacancy in that organization, 
he or she may be promoted to a higher rank position in another organization. This may explain 
why sometimes agency heads are appointed from outside of the agency. In this scenario, provincial 
government agencies are used to place those from outside of the agency who ?need? to be 
promoted. 
My interviews also reveal another mechanism that makes the positions of provincial 
government agency heads in particularly high demand.49 Not every cadre in the party-state will 
end up being promoted, but every cadre needs to be rotated. For example, there could be a 
prefecture-level mayor who is not competent enough to be promoted but who has made no 
 
49 Interview on December 24, 2020. The interviewee is a Division Rank bureaucrat in a government agency in a 
southwest province. He used to be a member of a county party standing committee. 
 
114 
 
 
 
mistakes and has not reached the age of retirement. Meanwhile, the mayor has been in the position 
for quite some time and thus needs to be rotated for the purpose of political control.  The position 
may also need to be vacated to place someone else. In this case, the most suitable arrangement for 
the mayor is a position in the provincial government agency. Provincial government agency has 
the same administrative rank as the prefecture-level mayor ? the party cannot place the mayor in 
a lower rank and do not want to promote him or her to a higher rank. As the mayor has proven to 
be not competent enough as a local leader, it is not suitable to rotate him or her to be the mayor of 
another prefecture-level city. Finally, provincial government agencies locate at the capital city, 
which has better medical and education resources. Although the party cannot promote the mayor 
to a higher administrative rank, the party can reward loyal cadres by taking care of them and their 
family before they retire. In this scenario, provincial government agencies are used to place those 
from outside of the agency who ?need? to be rotated. 
From a comparative perspective, government positions are used for non-policy functions in 
both the case of placing those who ?need? to be promoted or rotated and the case of patronage 
appointments in democracies. However, there are distinct mechanisms behind these two types of 
non-policy appointments. In democracies, elected politicians use government positions to reward 
campaign workers and campaign workers in turn use government positions to help politicians win 
reelection (e.g., Grindle 2012, Hollibaugh Jr, et al. 2014, Oliveros 2016, Colonnelli, et al. 2020, 
Brierley 2021, Oliveros 2021). As the career prospects of these patronage appointments depend 
on whether their connected politicians or political parties can stay in power, the patronage 
appointments are an incentive-compatible arrangement. The mechanism of patronage 
appointments does not work in China because politicians at all administrative levels do not face 
electoral pressures. Instead, the mechanism of placing those who ?need? to be promoted or rotated 
115 
 
 
 
is a politics of ?numbers?. As has been discussed above, public positions are used as a resource to 
control and coopt the society, which plays a crucial role of regime stability (Svolik 2012). On the 
one hand, to achieve the purpose of control and cooptation, the regime must employ enough people 
and maintain a certain rate of promotion and rotation. On the other hand, as in any hierarchical 
organizations, the demand for positions outstrips the supply. Therefore, to sustain the system, the 
regime must create new positions or use existing positions to arrange appointments for this purpose.    
Provincial leaders in China are expected to use a compartmentalization strategy to place 
those who ?need? to be promoted or rotated. Theoretically, provincial leaders have the incentive 
to place them in unimportant agencies because the consequences of bad performance are less likely 
to be significant in these agencies. This compartmentalization strategy is similar to placing 
patronage appointments in less important agencies in Latin America (Geddes 1994), Africa 
(Brierley 2021), and the United States (Hollibaugh Jr, et al. 2014). Empirically, as Figure 2.1 
shows, not every agency is used to place those who need promotions. In the six high-profile high-
risk agencies, the proportion of agency outsiders promoted to the positions is low, suggesting that 
these agencies are not used for the purpose of arranging placements. Except for the Family 
Planning Commission, the Department of Audit, the Foreign Affairs Office, the Ethnic Affairs 
Commission, and the Department of Culture, the proportion of agency outsiders promoted to the 
positions in each of the other agencies is below 50 percent. Interestingly, these five agencies also 
rank among the lowest in the high-profile high-risk index (see Figure 2.7). As these agencies have 
low risks of corruption, the strategy of appointing outsiders cannot be interpreted as reducing 
corruption risks. As they are among the least important agencies, it is also hard to interpret 
appointing outsiders as seeking candidates with higher human capital or more political experience. 
116 
 
 
 
Instead, agency heads appointed from the outside are more likely to be those who need promotions 
or rotations. 
To further test whether provincial leaders use a compartmentalization strategy to place those 
who ?need? to be promoted or rotated, I leverage the 1998 administrative reform as a shock that 
significantly reduced the supply of positions. The National People?s Congress approved the plan 
of administrative reform in March 1998. One of the main themes of the administrative reform is 
to transform the functions of the government to adapt to the marketization reform. All the 
government agencies with the function of industrial management and planning are cancelled. The 
total number of government cabinet agencies reduces from 40 to 29. This is the only administrative 
reform since the 1980s that significantly changed the number of government agencies. The 
decreased number of agencies reduces the supply of positions. Although the 1998 administrative 
reform also aimed to downsize the public employment, the outcome is less successful because the 
efficiency concern which requires the party to streamline conflicts with the stability concern which 
requires the party to hire as many public employees as possible (Burns 2003). As a result, the 
number of party cadres who ?need? to be promoted or rotated did not decrease in the same rate as 
the number of positions. This means that the government agencies that are not cancelled face 
higher pressures to place those who ?need? to be promoted or rotated. If provincial leaders use a 
compartmentalization strategy to arrange such appointments, more agency outsiders are expected 
to be appointed to the low-profile low-risk agencies but not the other agencies after the 1998 
administrative reform.  
I categorize the Family Planning Commission, the Department of Audit, the Foreign Affairs 
Office, the Ethnic Affairs Commission, and the Department of Culture as the low-profile low-risk 
agencies. As mentioned above, these five agencies rank the lowest in the high-profile high-risk 
117 
 
 
 
index. The selection of the lowest five agencies is not arbitrary. The high-risk ? high-profile 
indexes of these five agencies are 0.467, 3.196, 3.405, 3.474, and 6.533. The agency with the next 
lowest high-profile high-risk index is the Department of Science and Technology and its index is 
20.667. Specifically, according to the high-profile index, the Department of Science and 
Technology cannot be categorized as unimportant agencies. The indexes of the next six agencies 
following the Department of Science and Technology range from 23 to 31. The size of each 
agency?s index suggests that the selected five agencies indeed represent a unique group that is 
different from the other agencies. In the context of the 1998 administrative reform, they are the 
?treated? agencies. I include the agency heads appointed between 1990 and 2005 ? eight years 
before and after 1998.  
Figure 5.1 plots the proportion of newly appointed agency heads between 1990 and 2005 
with inside-agency experience in the low-profile low-risk agencies, high-profile high-risk agencies, 
and the other agencies. In general, more agency heads were appointed from outside of the agency 
after the 1998 administrative reform. But the relative changes differ in the three types of agencies. 
The proportion of having inside-agency experience in the low-profile low-risk agencies is not the 
lowest among the three agency types until 1997. Interestingly, starting from one year before the 
1998 administrative reform, the proportion decreased sharply in the low-profile low-risk agencies 
Different from Xi Jinping?s anti-corruption campaign, the administrative reform ? including which 
agencies were going to be cancelled ? was not unexpected before 199850. The formal approval by 
the National People?s Congress in March 1998 means that discussions and decisions had been 
made before. If it had already been known that some agencies were going to be cancelled in 1998, 
 
50 According to an interview by ?Caixin? with Li Tie, who worked in the National Economic Reform Commission in 
1998, he mentioned that they had already known about the administrative reform and their agency was going to be 
canceled before 1998. He also mentioned that the State Council froze the establishment of new government units in 
1997 as a preparation for the 1998 administrative reform. 
118 
 
 
 
positions in these agencies would become ?useless? in 1997. This may explain why the change 
started one year before 1998. The results indicate that positions in these agencies are increasingly 
used to place outsiders who ?need? to be promoted or rotated starting from 1997.  
 
 
Figure 5. 1: Proportion of Having Inside-Agency Experience (1990-2005) 
On the contrary, the proportion of newly appointed agency heads with inside experience 
does not decrease much in the high-profile high-risk agencies. The results indicate that provincial 
leaders did not use these agencies to place those who ?need? to be promoted or rotated, which is 
consistent with the compartmentalization strategy. The other agencies sit in the middle. There is a 
decreasing trend of appointing agency insiders in these agencies after 1998, but the decrease is less 
salient compared with the low-profile low-risk agencies. 
 
119 
 
 
 
 
Figure 5. 2: Before-After Effects by Agency (1998 Administrative Reform) 
Figure 5.2 plots the before-after change in each agency using different bandwidths.51 Agency 
heads appointed in and after 1997 are coded as ???? ??????. The agencies are ranked by the size 
and direction of the changes based on the agency heads appointed between 1991 and 2002. The 
results are consistent with the pattern in Figure 5.1. There are more agency outsiders in all the 
agencies after the 1998 administrative reform, but the changes are more salient in the low-profile 
low-risk agencies and the other agencies that are not high-profile high-risk. Except for the Foreign 
Affairs Office, the other four low-profile low-risk agencies all locate in the lower end of the 
distribution. The average change (weighted by number of appointments) in the low-profile low-
risk agencies is ?0.302, which means that the proportion of newly appointed agency heads with 
 
51 The Department of Environmental Protection and the Department of Land and Resources are not included because 
both agencies are established in the 2000s. 
120 
 
 
 
inside-agency experience decreases by more than 30 percentage points in these agencies. On the 
contrary, none of the high-profile high-risk agencies locate in the lower end. The top ten agencies 
with the largest decrease of agency insiders are all agencies that are low-profile low-risk and the 
agencies that are not high-profile high-risk. The average change in the agencies that are neither 
low-profile low-risk nor high-profile high-risk is ?0.164, the size of which is quite large but 
smaller than the change in the low-profile low-risk agencies. The average change in the high-
profile high-risk agencies is ?0.097. Although the proportion of agency insiders still decreases, the 
size is the smallest among the three types of agencies. 
Table 5.1 reports the before-after changes in the three types of agencies using model (5.1). 
 
????????? = ?1???? ??????? + ?2???? ??????? ? ???? + ?3???? ??????? ? ????? + ???
+ ????                (5.1) 
 
???? ??????? is a dummy variable coded as 1 if the agency heads are appointed in or after 1997. 
????  is a dummy variable coded as 1 if the agency is the Family Planning Commission, the 
Department of Audit, the Foreign Affairs Office, the Ethnic Affairs Commission, or the 
Department of Culture. ????? is a dummy variable coded as 1 if the agency is the Development 
and Reform Commission, the Economic Affairs Commission, the Department of Finance, the 
Department of Transportation, the Department of Construction, or the Department of Land and 
Resources (no observations before 2000). ??? is provincial-agency fixed effects. 
121 
 
 
 
Table 5. 1: Before-After Effects (1998 Administrative Reform) 
 DV: inside-agency work experience 
 1993-2000 1992-2001 1991-2002 1990-2003 
 (1) (2) (3) (4) 
Post reform -0.166 -0.168 -0.185 -0.101 
 (0.064) (0.061) (0.055) (0.053) 
Post reform ? Low-profile low-risk -0.092 -0.070 -0.065 -0.154 
 (0.125) (0.106) (0.097) (0.072) 
Post reform ? High-profile high-risk -0.025 -0.025 0.017 -0.050 
 (0.094) (0.102) (0.081) (0.073) 
Province-agency fixed effects Yes Yes Yes Yes 
Number of province-agency 421 432 437 438 
Number of observations 698 791 933 1194 
     
Low-profile low-risk agencies:     
Mean (prior to the campaign) 0.444 0.489 0.495 0.500 
Before-after changes (?1 + ?2) -0.257 -0.238 -0.250 -0.255 
 (0.100) (0.097) (0.084) (0.060) 
High-profile high-risk agencies:     
Mean (prior to the campaign) 0.560 0.560 0.570 0.575 
Before-after changes (?1 + ?3) -0.190 -0.193 -0.168 -0.151 
 (0.080) (0.064) (0.049) (0.039) 
Other agencies:     
Mean (prior to the campaign) 0.456 0.430 0.430 0.420 
Before-after changes (?1) -0.166 -0.168 -0.185 -0.101 
 (0.064) (0.061) (0.055) (0.053) 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level of provincial 
party secretaries. 
 
  
Consistent with the previous findings, the results show that the proportion of newly 
appointed agency heads with inside-agency experience decreases after 1997 in all three types of 
the agencies, but the size of the change is largest in the low-profile low-risk agencies. Different 
122 
 
 
 
from the previous findings, the size of the change in the high-profile high-risk agencies is not 
smaller than the other agencies. On the one hand, the inclusion of provincial-agency fixed effects 
restricts the before-after comparisons within the same agency in the same province, which drops 
the cases where no new agency heads were appointed before or after the reform. Therefore, the 
differences could be a result of the different sample compositions. On the other hand, the before-
after changes should be interpreted with caution because the reduced supply of positions is not the 
only factor that changed after 1997. As Figure 1.4 shows, provincial party secretaries? in-province 
time also drops sharply after 1997. For the agency heads appointed between 1990 and 1997, the 
median in-province time is more than seven years. For the agency heads appointed between 1998 
and 2005, the median in-province time is a little more than four and half years. According to my 
theoretical framework, the shorter in-province time should affect the appointments of agency heads 
in the high-profile high-risk agencies. Chapter 3 finds that provincial leaders are less likely to 
appoint agency heads with inside-agency experience in the high-profile high-risk agencies when 
their in-province time is short. Therefore, the decreased proportion of having inside-agency 
experience in these agencies may be a result of the concurrent change of the in-province time rather 
than the administrative reform. 
Despite the caveats, the overall empirical results support the argument that provincial leaders 
in China use a compartmentalization strategy when appointing government agency heads. 
Specifically, in addition to their policy functions, positions of provincial government agencies are 
used to place party cadres who ?need? to be promoted or rotated. This function is important to 
maintain regime stability. However, provincial leaders are more likely to use low-profile low-risk 
agencies for this purpose because the consequences of bad performance in these agencies are less 
likely to be significant. The 1998 administrative reform reduces the number of government 
123 
 
 
 
agencies from 40 to 29. I find that agency heads are more likely to be appointed from the outside 
in all the agencies, but the change is mostly salient in the low-profile low-risk agencies. Due to the 
nature of the low-profile low-risk agencies, the strategy of appointing outsiders is unlikely for the 
purpose of reducing corruption risks or seeking higher human capital and political experience. The 
results suggest that provincial leaders use positions in the low-profile low-risk agencies to place 
those who ?need? to be promoted or rotated and increasingly so after the number of agencies is 
reduced.       
 
  
   
124 
 
 
 
Chapter 6: Conclusion 
 
This study is motivated by the divergent career paths of China?s provincial government agency 
heads. Tang Zhihong, Liu Yingzi, and Ouyang Bin were all appointed from outside of the agency. 
None of them had relevant work experience as well. You Qingzhong served as the deputy agency 
head for six years before being promoted to the agency head. His earlier work experience was also 
in related areas. Although Duan Yufei had no prior work experience inside the agency, he had 
relevant experience outside of the agency. Their divergent career paths represent different types of 
agency heads in China. Although it is common in Leninist systems for bureaucrats to have 
interweaving career paths in multiple organizations, why certain types of bureaucrats are assigned 
to certain positions is a question that is relevant beyond China, has important implications for the 
quality of government, and is related to the broad problem of bureaucratic authority and political 
control. From an originally collected dataset of China?s provincial government agency heads 
appointed between 1978 and 2020, I find the following patterns of appointments. Most agency 
heads are appointed from outside of the agency, but many outsiders have relevant work experience 
(Figure 1.2). In addition, the trend of appointing agency heads from the outside increases over time 
(Figure 1.3). What are the factors that explain when and why provincial leaders appoint agency 
heads with or without inside-agency experience? This chapter summarizes the theoretical 
framework and empirical findings, discusses how they help explain the overall patterns of 
appointing provincial government agency heads in China?s reform era, and how they relate to the 
comparative literature beyond China. This chapter ends with implications on building state 
capacity in authoritarian regimes. 
125 
 
 
 
To explain the puzzle, I first examine three existing theories in bureaucratic politics: the 
incentive-assignment conflict (the Peter Principle), patron-client relationships, and human capital 
and political experience. I find that they cannot fully explain the above patterns of appointments. 
I then develop and empirically test a theoretical framework that draws insights from the literature 
on delegation and bureaucratic embeddedness. Appointees with prior experience inside the agency 
are more embedded in the agency. The embeddedness comes from their familiarity with their work 
environment. On the one hand, appointing agency heads who are familiar with the agency, the 
personnel, and the clients helps reduce performance uncertainty. On the other hand, agency heads 
who are embedded in the agency can use their information advantages to engage in corruption, 
especially the type of corruption that needs collusion with their colleagues and business clients. 
Provincial leaders? decisions on whether to appoint agency insiders or outsiders are a function of 
balancing the risk of performance uncertainty and collusive corruption. I argue that two factors are 
important in determining how provincial leaders balance the trade-off. First, provincial leaders? 
information on and connections with the candidates of appointees help mitigate information 
asymmetry. When provincial leaders have little shared work experience with the candidates, they 
are more likely to prioritize reducing corruption risks over performance uncertainty. Second, while 
monitoring is supposed to reduce the need to take preventive measures at the appointment stage, 
authoritarian monitoring could lead provincial leaders to take more preventive measures due to its 
distinctive features of top-down control, uncertainty, and intra-party propaganda. I also emphasize 
that the theoretical framework is more applicable to important agencies with high corruption risks. 
I categorize high-profile high-risk agencies based on both existing studies and a self-constructed 
index. The Development and Reform Commission, the Department of Finance, the Economic 
126 
 
 
 
Affairs Commission, the Department of Land and Resources, the Department of Construction, and 
the Department of Transportation are categorized as high-profile high-risk agencies. 
In Chapter 3, I find that when provincial party secretaries? in-province time before 
appointing agency heads is short, they are more likely to appoint agency heads without inside-
agency experience in the high-profile high-risk agencies. Their bias against agency insiders soon 
disappears when they have longer in-province time. In Chapter 4, I find that after Xi Jinping?s anti-
corruption campaign ? an example that successfully strengthened the CCP?s monitoring capacity 
? provincial leaders appointed more agency heads without inside-agency experience in the high-
profile high-risk agencies. This finding is contrary to what the delegation model predicts. In both 
chapters, I rule out the alternative explanations of incentive-assignment conflict (the Peter 
Principle), patron-client relationships, and human capital or political experience. Interestingly, in 
situations where provincial leaders tend to appoint agency outsiders, they do not bias against 
specialists and in favor of generalists. Specifically, the proportion of agency heads with relevant 
work experience ? both inside and outside of the agency ? does not decrease. The agencies that 
are not high-profile high-risk are used as the ?control? group in Chapter 3 and Chapter 4. In 
Chapter 5, I explain why provincial leaders sometimes appoint agency heads from outside of the 
agency in the low-profile low-risk agencies. In addition to their policy functions, provincial 
government agencies could also be used to place party cadres who ?need? to be promoted or rotated. 
I argue that provincial leaders adopt a compartmentalization strategy by placing those who ?need? 
to be promoted or rotated in the low-profile low-risk agencies. I find that after the 1998 
administrative reform that reduced the supply of positions (the number of government agencies 
was reduced from 40 to 29), provincial leaders appointed more agency heads from outside of the 
agency, especially in the low-profile low-risk agencies. 
127 
 
 
 
How do my findings help explain the overall patterns of appointing provincial government 
agency heads in China?s reform era (Figure 1.2 and Figure 1.3)? First, while provincial leaders 
bias against agency insiders in the high-profile high-risk agencies when they lack information or 
when they face authoritarian monitoring, they do not bias against specialists in general. In fact, 
after Xi Jinping?s anti-corruption campaign, agency heads appointed from outside of the agency 
are more likely to have relevant work experience compared with the outsiders appointed before. 
These results help explain why most agency heads are appointed from outside of the agency while 
many have relevant work experience. Second, Xi Jinping?s anti-corruption campaign, the 1998 
administrative reform, and the decreased in-province time all lead provincial leaders to appoint 
more agency heads from outside of the agency, although in different types of agencies and for 
different reasons. Provincial leaders appointed more agency heads without inside-agency 
experience in the high-profile high-risk agencies after Xi Jinping?s anti-corruption campaign as a 
preventive measure to preserve themselves. Provincial leaders appointed more agency heads 
without inside-agency experience after the 1998 administrative reform in the low-profile low-risk 
agencies because the number of positions that can be used to place those who ?need? to be 
promoted and rotated reduced. Provincial party secretaries? average in-province time decreases 
after 1998 (Figure 1.4). This is consistent with the finding that the party center has intensified the 
control of provincial leaders over time (Bulman and Jaros 2020). The decreased in-province time 
on average should result in more agency outsiders appointed in the high-profile high-risk agencies. 
In other words, Xi Jinping?s anti-corruption campaign, the 1998 administrative reform, and the 
party center?s increased control of provincial party secretaries help explain why the trend of 
appointing agency outsiders increases over time. 
128 
 
 
 
How do the theoretical framework and empirical findings relate to the comparative literature 
beyond China? First, although this study is about bureaucratic appointments in an authoritarian 
regime, the level of information on appointees affects how democratic leaders make appointment 
decisions in similar ways. Newly elected political leaders could be unfamiliar with the bureaucrats 
in the executive branch. This is especially true in the presidential system where political parties 
are more likely to nominate presidential candidates from outside of the party, legislature, and 
cabinet (Samuels and Shugart 2010). While these candidates are good at winning popular votes, 
they lack the skills and information to run the government, especially during their initial tenure in 
office. This affects how elected political leaders appoint bureaucrats. For example, Krause and 
O'Connell (2016) find that newly elected presidents in the United States are more likely to appoint 
loyalists as agency leaders to counterbalance the information advantages of the career bureaucrats; 
but as presidents? tenure increases, they will rely less on this strategy. Compared with politicians 
in democracies, provincial leaders in China face different constraints. When they are reshuffled 
from one place to another, they cannot bring their old subordinates to the new place. Therefore, 
they need to rely more on appointees? past career experience as an information shortcut to mitigate 
information asymmetry. 
Second, this study reveals a secondary consequence of monitoring on the type of bureaucrats 
selected. The unintended and mixed results of monitoring on bureaucratic selections are not unique 
in China. While strengthened monitoring capacity is supposed to improve the quality of 
government, increased monitoring could change the type of people that self-select to enter 
government. Grossman and Hanlon (2014) find that in Uganda?s farmer associations, increased 
monitoring pushes higher-ability members to opt out of candidacy. Szakonyi (2021) finds that 
politicians are less likely to seek reelection after the requirement to disclose their financial 
129 
 
 
 
situations, especially for those with suspicious financial records. In contrast to the self-selection 
effects, this study focuses on how increased monitoring changes politicians? preference in selecting 
different types of candidates. Cavalcanti, et al. (2018) find that audits in Brazil do lead political 
parties to select better-educated candidates when high levels of corruption are revealed. The career 
prospects of provincial leaders in China do not depend on elections but on top-down appointments. 
Monitoring in China takes the form of campaign-style enforcement, which is different from the 
democratic approach. In such contexts, the effects of monitoring on bureaucratic appointments are 
more nuanced. 
In the end, what are the implications of this study? The Covid-19 pandemic has put many 
agency heads in China under the spotlight. When things go wrong, the public and the media often 
blame the agency heads who lack expertise in the subject-area. In fact, whether non-professionals 
can lead professionals (?reds? versus ?experts?) has been a consistent debate within the CCP in 
the Mao, Deng, and Xi eras. Although the question is debated even in the highly specialized area 
of controlling the pandemic (Li, et al. 2022, Liu and Ma 2022), we know that different types of 
leaders affect policy decisions and outcomes in important ways. The descriptive patterns of 
China?s provincial government agency heads appointed in the reform area shed light on this debate. 
This study shows that while the phenomenon of ?non-professionals leading professionals? exists, 
it is not as common as the high-profile cases suggest. However, most agency heads do not have 
prior experience inside the agency, and the trend of appointing agency heads from outside of the 
agency increases over time. Although many agency outsiders have relevant work experience, the 
Chinese bureaucracy has clearly not moved in a Weberian direction. On the one hand, the trend 
seems to defy Deng Xiaoping?s vision of how to rebuild the bureaucracy in the 1980s. On the other 
hand, contrary to the argument that the Chinese cadre organization is built on a different logic from 
130 
 
 
 
the Weberian organization (Rothstein 2015), this study suggests that moving away from a 
Weberian bureaucracy is not a deliberate choice of the CCP either. Instead, I find that it is a result 
of provincial leaders? strategic appointments under the constraints set by the party center. The 
trend toward a Weberian bureaucracy is unintentionally interrupted by the periodical reshuffling 
of provincial leaders that aims to prevent localism, the anti-corruption campaign that aims to 
strengthen monitoring, and the administrative reform that aims to streamline the government.  
This study not only reveals and explains the patterns of bureaucratic appointments in China 
but also has broad implications for state capacity, especially for authoritarian regimes with weak 
institutions. This study suggests that authoritarian regimes face inherent dilemmas in building state 
capacity. First, although newly elected politicians in democracies may also lack information when 
making appointment decisions in their initial tenure, the low information environment is a built-in 
situation in authoritarian regimes. Frequently rotating subnational elites is a crucial tool for 
autocratic leaders to maintain power (Migdal 1988, Hill and L?wenhardt 1991, Snyder 1992, 
Siegel 2018). It is also an important source of state capacity because rotations are effective in 
ensuring subnational elites? compliance (Edin 2003, Xu 2011, Svolik 2012). However, provincial 
leaders who lack information make appointment decisions that prioritize reducing corruption risks 
over performance uncertainty. This is not necessarily good for the state capacity building at the 
subnational level. In other words, the policy tool that strengthened the autocratic leader?s control 
and thus one dimension of the state capacity conflicts with the conditions that help build state 
capacity in other dimensions. Recent studies have discussed the dilemma the party center faces in 
appointing provincial leaders (Bulman and Jaros 2020). Appointing provincial leaders from the 
outside increases the central control at the expense of efficient governance. This study reveals a 
131 
 
 
 
specific mechanism ? provincial leaders from outside of the province also appoint agency heads 
from outside of the agency. 
Second, the creation of an efficient and uncorrupt bureaucracy requires effective monitoring. 
Authoritarian regimes are no exception in trying to strengthen their monitoring capacity. They 
sometimes resort to semi-democratic tools such as autocratic elections (Miller 2015), party 
organizations (Charron and Lapuente 2011), and partially free media (Egorov, et al. 2009, 
Lorentzen 2014). The common mechanism behind these approaches is to allow some bottom-up 
channels of accountability and information revealing. These approaches, however, are inherently 
risky for autocratic leaders (Brancati 2014) and thus are unlikely to last long. For example, 
investigative journalism and partially free media have only been tolerated by the CCP for some 
time until they touched the red line.52 Alternatively, authoritarian leaders can resort to what they 
do best ? the authoritarian approach. Authoritarian regimes have launched at least 25 anti-
corruption campaigns since 1950 and the successful ones all used the authoritarian rather than the 
democratic approach (Carothers 2022). The success of Xi Jinping?s anti-corruption campaign and 
similar campaigns in other authoritarian regimes seem to show that authoritarian monitoring could 
strengthen autocratic leaders? monitoring capacity without exposing them to unwanted risks. My 
findings, however, show that authoritarian monitoring is not without costs. Provincial leaders make 
appointment decisions in a way to preserve themselves, which may not be the best choice to 
improve the quality of government. 
This study also reveals a workaround that might both reduce corruption risks and maintain 
some bureaucratic expertise. The overall patterns of appointments show that while most agency 
 
52 For example, see Repnikova, Maria, and Kecheng Fang. ?Behind the fall of China?s greatest newspaper.? Foreign 
Policy. January 29, 2015  
132 
 
 
 
heads are appointed from outside of the agency, many of them have relevant experience. When 
provincial leaders appoint agency outsiders, they do not bias against the candidates with relevant 
experience outside of the agency. After Xi Jinping?s anti-corruption campaign, provincial leaders 
appointed more outsiders with relevant experience to take the place of agency insiders. For 
countries with both a high level of corruption and a low level of bureaucratic expertise, rotating 
bureaucrats across different but similar organizations might be an alternative way to build an 
efficient and uncorrupt bureaucracy. 
133 
 
 
 
Appendix A: Appendix for Chapter 1 
 
A.1 Agency in the Dataset 
 
Table A. 1: Provincial Government Agency in the Dataset 
Agency Policy jurisdictions and core functions 
Development and Reform Commission (Planning 
Economic and industrial policy; review and approval 
Commission and Development and Planning 
of investment projects 
Commission before 2003) 
Department of Education Education policy; universities and schools 
Department of Science and Technology Innovation and high-tech policy and projects 
Economic and Information Technology Commission Industry and row materials; enterprises and high-tech 
(Economic Affairs Commission before 2009)  sectors 
Ethnic Affairs Commission Ethnic and religious groups 
Department of Civil Affairs Civil society organizations; social work and welfare 
Department of Justice Prison management and legal education 
Department of Finance Budget and state-owned assets; procurement 
Department of Human Resources and Social Security 
(Department of Human Resources and Department of Jobs and employment; personnel management in public 
Labor and Social Security as two separate agencies institutions; labor and social security 
between 1988 and 2008) 
Department of Land and Resources (established in 
1998, restructured as Department of Natural Resources Land, mine, and marine resources 
in 2018) 
Department of Environmental Protection (incorporated 
into government cabinet in 2008, Department of Pollution control and monitoring 
Ecology and Environment since 2018) 
Department of Housing and Urban-Rural Development Real estate and housing sectors; municipal construction 
(Department of Construction before 2008) projects 
Transportation management and infrastructure 
Department of Transportation 
construction 
Water resources protection and irrigation infrastructure 
Department of Water Resources 
construction 
Department of Agriculture and Rural Affairs 
Agricultural and rural economic development 
(Department of Agriculture before 2018) 
134 
 
 
 
Department of Commerce (Department of Foreign 
Trade and foreign direct investment 
Trade and Economic Cooperation before 2003) 
Department of Culture (restructured as Department of 
Art production and cultural relics 
Culture and Tourism after 2018) 
Family Planning Commission (canceled and merged 
Population and birth control 
into Health Commission in 2013) 
Health Commission (Health and Family Planning 
Hospitals; public health; pharmaceutical policy 
Commission between 2013 and 2018) 
Auditing government institutions, state-owned firms, 
Department of Audit 
and government leaders 
Receiving foreign visitors; planning foreign exchanges; 
Foreign Affairs Office 
consular affairs 
Note: For the agencies that experienced changes in names (often associated with some changes of functions) and 
those that have been merged, I treat them as a single agency before and after the changes. Policy jurisdictions and 
core functions are based on the official documents from the websites of central ministers. Provincial government 
agencies often follow the central ministers to draft their core duties. The policy jurisdictions and core functions of 
the same agency are also similar across provinces. 
 
A.2 Relevant Experience outside of the Agency 
 
Table A. 2: Relevant Work Experience and Education Backgrounds 
Other relevant work experience 
Related Agencies Relevant education backgrounds 
outside of the agency 
Development and Reform 
Commission; Economic and 
State-owned firms; tax bureaus (for 
Information Technology Economics and finance 
Department of Finance only) 
Commission; Department of 
Finance; Department of Commerce   
Universities (both research and 
Department of Education administrative positions) and Education 
schools 
Department of Science and Research institutions and 
Engineering and natural science 
Technology universities (research positions) 
Department of the United Front 
Ethnic Affairs Commission Ethnology and religious studies 
Work 
Department of Civil Affairs Red Cross  
Department of Public Security; 
Department of Justice Political and Legal Affairs Law 
Commission; courts 
135 
 
 
 
Department of Human Resources 
Department of Organization Human resources management 
and Social Security  
Department of Land and Resources   Geology 
Department of Environmental Environmental science and 
 
Protection  engineering 
State-owned construction or 
Department of Housing and Urban- transportation corporations; urban Civil engineering; transportation 
Rural Development; Department of planning institutes (for Department engineering (for Department of 
Transportation  of Housing and Urban-Rural Transportation only) 
Development only) 
Hydraulic engineering (for 
Department of Water Resources; State-owned farms (for Department Department of Water Resources 
Department of Agriculture and of Agriculture and Rural Affairs only); Agriculture (for Department 
Rural Affairs only) of Agriculture and Rural Affairs 
only)  
Department of Propaganda; 
Department of Culture  Humanities and arts 
professional art institutes 
Hospitals and medical schools (both 
Health Commission; Family 
research and administrative Medical science 
Planning Commission  
positions) 
Department of Audit Department of Finance Finance; certificate of audit 
Foreign Affairs Office; Department Trade associations; embassy (for Foreign languages (for Foreign 
of Commerce Foreign Affairs Office only) Affairs Office only) 
Note: Work experience in related agencies is counted as relevant outside experience. In addition, working in the 
same agency at different administrative levels (county, city, or the central government) is another common relevant 
experience outside of the agency. 
 
A.3 Data Collection Process 
To build the dataset, I first obtained the names of the agency heads and their tenures from 
provincial yearbooks and provincial government agency chronicles. For the agency heads 
appointed in the 2000s, their career paths and demographic information are largely available in 
?Baidu Baike?. For the agency heads appointed earlier and those whose information cannot be 
found in ?Baidu Baike?, I mainly relied on two sources ? the digitalized local gazettes in the 
?Wanfang? database and an online database of revolutionary figures called ?Fenghuo Home?53. 
 
53 See http://www.wphoto.net/ for details. 
136 
 
 
 
Some provincial government agency chronicles also provide the names and tenures of deputy 
agency heads, which also helps me find whether an agency head had previously worked inside the 
agency. Finally, if the information is still missing (especially the career experience), I searched 
their names in the ?CNKI?, ?Wanfang?, and ?Chaoxing? databases. Their career experience may 
be mentioned in news articles and government documents. For 34 agency heads, I did not find 
whether they have any prior relevant work experience. For 14 of them, I did not find whether they 
have prior work experience inside the agency. 
 
A.4 Descriptive Statistics 
 
Table A. 3: Descriptive Statistics 
Variables Time frame N Mean St. Dev. Min Pctl (25) Pctl (75) Max 
Inside-agency 
1978-2020 3675 0.375 0.484 0 0 1 1 
experience 
Relevant work 
1978-2020 3655 0.665 0.472 0 0 1 1 
experience 
Age at the time of 
2005-2020 1470 52.331 3.742 38.608 50.201 54.933 58.625 
appointment 
Master?s degree and 
2005-2020 1454 0.742 0.438 0 0 1 1 
above 
Doctorate degree 2005-2020 1454 0.210 0.408 0 0 0 1 
 
  
137 
 
 
 
Appendix B: Appendix for Chapter 2 
  
B.1 Coding Whether Agency Heads are Promoted into Their Positions 
 
Table B. 1: Criteria of Promotions from Other Positions to Agency Heads 
Type of promotion Examples 
Promoted from the positions associated with the Deputy Bureau 
Rank: e.g., deputy prefecture-level mayors, deputy prefecture-level 
party secretaries, deputy heads of other government cabinet 
Promoted from a lower administrative rank 
agencies and government-affiliated institutions, deputy heads of 
party departments, deputy secretaries-general of the provincial 
government or party, etc. 
Except for the positions of prefecture-level mayors, prefecture-level 
city party secretaries, and the heads of government cabinet 
agencies, those appointed from other positions associated with the 
Promoted from less important positions of 
Bureau Rank are counted as promoted from less important 
the same administrative rank 
positions: e.g., directors of Industrial and Commerce Bureau 
(agencies not in the government cabinet), manages of state-owned 
firms, university presidents, hospital deans, etc. 
Note: (1) The Development and Reform Commission is considered the ?small state council?, many agency heads 
are rotated from prefecture-level city party secretaries. In this case, those rotated from prefecture-level mayors and 
leaders of other government cabinet agencies are coded as promoted from less important positions of the same 
administrative rank. Those appointed from prefecture-level city party secretaries are coded as lateral moves. (2) 
Sometimes those who work in the Deputy Bureau Rank positions have the administrative rank as the Bureau Rank. 
For example, some deputy secretaries-general of the provincial government or party hold concurrent appointments 
in positions associated with the Bureau Rank or are granted the Bureau Rank. In such cases, they are coded as 
promoted from less important positions of the same administrative rank.   
 
 
 
 
138 
 
 
 
B.2 Coding Patron-Client Relationships 
 
Table B. 2: Promotion Links between Provincial Party Secretaries and Agency Heads 
Type of promotion links Examples 
Promoted to the prefecture-level city mayors and 
party secretaries, government cabinet agency heads, 
and other positions associated with the Bureau Rank 
Promoted from the Deputy Bureau Rank to the Bureau 
(e.g., heads of government-affiliated institutions, 
Rank before being appointed as agency heads 
managers of state-owned firms, hospital deans, 
university presidents, some deputy secretaries-
general of the provincial government or party, etc.)    
Promoted from other Deputy Bureau Rank positions 
to the prefecture-level city Party Standing 
Promoted to a more important position within the Deputy 
Committee or Deputy Party Secretary, executive 
Bureau Rank before being appointed as agency heads 
deputy mayors, executive deputy heads of 
government or party departments, etc. 
 
 
B.3 Categorizing Agencies 
I rank the agencies by the self-created index of high-risk ? high-profile (standardized between 0 
and 100). I use the index as an auxiliary approach to validate the categorization criterion of 
controlling land and mining rights, related to construction and infrastructural development, and 
managing enterprises. The high-profile index is the proportion of provincial government agency 
heads appointed between 1984 and 2013 who were promoted out from the agency. Agency heads 
who were appointed after 2013 are not included because many are still incumbent. The high-risk 
index is the total number of investigated corruption cases in related government sectors in the 22 
provinces, which include government agencies and bureaus in the administrative units of province, 
city, county, and township.  
 
139 
 
 
 
 
Figure B. 1: Officially Set Personnel Quota in Each Agency 
I acknowledge that the index is not perfect. First, the high-risk index measures both 
corruption risks and enforcement intensity. As Figure 2.7 shows, the number of investigated cases 
is high in the government sector of environmental protection, which is probably due to the 
increased enforcement in the sector since 2013. But overall, the number of investigated cases 
should be positively correlated with corruption risks. Second, the total number of cases does not 
capture whether the corruption is collusive and whether its influence is large enough to cause 
trouble for provincial leaders. For example, provincial leaders are less likely to take the risk of 
petty corruption into account when making appointment decisions. An alternative measurement is 
the number of investigated officials at or above the Deputy Division Rank. Third, there might be 
a concern that the total number of cases is not normalized by the size of the government sector. 
But when provincial leaders assess the corruption risks ex-ante, what matters for them is not 
140 
 
 
 
corruption risks per capita ? small agencies, even with high corruption risks per capita, are unlikely 
to trouble them. Furthermore, the size of the government sector is not exogenous ? some important 
agencies also tend to be larger. As an alternative measurement, I also normalize the number. 
 
 
Figure B. 2: Alternative Measurements of the High-Risk Index 
Figure B.1 plots the officially set personnel quota for each provincial government agency as 
a proxy for the size of the government sector at all administrative levels. The data should be viewed 
as a crude proxy because the official quota is based on different years for different provinces. For 
each agency, I average the number across provinces. Figure B.1 shows that the Development and 
Reform Commission, the Department of Finance, and the Economic Affairs are among the 
agencies that have the largest officially approved size. According to Figure 2.7, they are also 
among the most important agencies.  
Figure B.2 ranks the agencies by three different measurements of corruption risks: total 
number of investigated officials (the same as the high-risk index in Figure 2.7), number of 
141 
 
 
 
investigated officials at or above the Deputy Division Rank, and the total number normalized by 
size.  Only the normalized number affects the ranking. But as the normalized number also 
?normalize? the importance of some agencies, it is not an ideal index for the purpose.  
 
B.4 Agency Insiders and Corruption Risks 
For provincial government agency heads appointed between 2005 and 2020, I found 62 of them 
investigated for corruption by 2020. The agency heads that were held accountable for other reasons 
are not included. For each corrupt agency head, I compare whether he or she has inside-agency 
experience with the experience of a synthetic ?agency head?, calculated as the average of the 
uncorrupt agency heads from the same agency (different provinces included) appointed in the same 
year, one year before, and one year after. For example, for Li Jiangong, who was appointed to the 
Shanxi Department of Land and Resources in 2009, I calculate the proportion of having inside-
agency experience among the other uncorrupt agency heads in the Department of Land and 
Resources appointed in 2008, 2009, and 2010, and match it to Li Jiangong. Finally, I compare the 
proportion of having inside-agency experience among the corrupt agency heads with the average 
proportion of the synthetic ?agency heads?.  
  
142 
 
 
 
Appendix C: Appendix for Chapter 3 
 
C.1 Robustness Checks: Difference-in-differences Effects by Agency 
 
 
Figure C. 1: Average DID Effects by Agency 
 
 
 
143 
 
 
 
C.2 Robustness Checks: Appointments to Replace Retired Agency Heads 
 
Table C. 1: Appointments to Replace Retired Agency Heads (1984-2012) 
 High-profile high-risk agencies  The other agencies 
The proportion of  The proportion of 
??? year in the province N N agency insiders (%) agency insiders (%) 
 (1) (2)  (3) (4) 
1st year 33 45.45   92 39.13 
2nd year 41 58.53   91 36.66 
3rd year 33 60.60  116 34.48 
4th year 27 59.25   86 34.88 
5th year 28 53.57  110 30.27 
6th year 23 60.86   78 28.57 
7th year 19 36.84   75 42.66 
8th year 16 56.25   55 49.09 
9th year 17 58.82   52 25.00 
10th year 17 47.05   47 36.17 
11th year 14 57.14   55 50.90 
12th year and above 32 59.37  109 35.77 
 
  
144 
 
 
 
Table C. 2: Cross-Sectional Effects for Appointments to Replace Retired Agency Heads 
 DV: inside-agency work experience 
 T ? 1 T ? 1 year T ? 1 year T ? 1 year T ? 2 
year and 3 months and 6 months and 9 months years 
 (1) (2) (3) (4) (5) 
High-profile high-risk agencies      
Marginal effects: ?1 + ?2 -0.071 -0.118 -0.109 -0.074 0.020 
 0.083 0.083 0.081 0.073 0.071 
Mean of the dependent variable 0.550 0.550 0.550 0.550 0.550 
N of agency heads appointed 
33 46 53 61 74 
within T 
The other agencies      
Marginal effects: ?1 0.100 0.121 0.100 0.092 0.083 
 0.042 0.041 0.040 0.038 0.036 
Mean of the dependent variable 0.363 0.363 0.363 0.363 0.363 
N of agency heads appointed 
92 119 131 160 183 
within T 
Note: Standard errors clustered at the level of provincial party secretaries are reported in the parentheses. 
 
 
  
145 
 
 
 
C.3 Robustness Checks: Alternative Time Periods 
 
Table C. 3: Agency Heads Appointed Between 1978 and 1983 
 High-profile high-risk agencies  The other agencies 
The proportion of  The proportion of 
??? year in the province N N agency insiders (%) agency insiders (%) 
 (1) (2)  (3) (4) 
1st year 17  82.3  48 29.16 
2nd year 16  80.0  60 40.67 
3rd year 19  55.5  64 38.70 
4th year  9  55.5  35 40.00 
5th year 18  50.0  51 42.00 
6th year 14  46.1  35 38.23 
7th year  7 100.0  30 40.00 
8th year  3  66.6  14 50.00 
9th year  1 100.0   7 71.42 
10th year  9  55.5  18 16.66 
11th year  3  66.6  15 26.66 
12th year and above 19  52.6  51 49.01 
 
  
146 
 
 
 
Table C. 4: Cross-Sectional Effects for Appointments Made Between 1978 and 1983 
 DV: inside-agency work experience 
 T ? 1 T ? 1 year T ? 1 year T ? 1 year T ? 2 
year and 3 months and 6 months and 9 months years 
 (1) (2) (3) (4) (5) 
High-profile high-risk agencies      
Marginal effects: ?1 + ?2 0.229 0.185 0.247 0.264 0.238 
 0.082 0.075 0.060 0.057 0.052 
Mean of the dependent variable 0.629 0.629 0.629 0.629 0.629 
N of agency heads appointed 
17 23 29 31 33 
within T 
The other agencies      
Marginal effects: ?1 -0.099 -0.102 -0.076 -0.059 -0.055 
 0.060 0.063 0.047 0.044 0.046 
Mean of the dependent variable 0.392 0.392 0.392 0.392 0.392 
N of agency heads appointed 
48 70 85 101 108 
within T 
Note: Standard errors clustered at the level of provincial party secretaries are reported in the parentheses. 
 
  
147 
 
 
 
Table C. 5: Agency Heads Appointed Between 2013 and 2020 
 High-profile high-risk agencies  The other agencies 
The proportion of  The proportion of 
??? year in the province N N agency insiders (%) agency insiders (%) 
 (1) (2)  (3) (4) 
1st year 39 43.58  71 21.12 
2nd year 37 27.02  71 25.35 
3rd year 34 29.41  90 24.44 
4th year 41 43.90  85 29.41 
5th year 34 26.47  70 27.14 
6th year 34 35.29  72 27.77 
7th year 16 25.00  29 31.03 
8th year  5 40.00  16 18.75 
9th year  9 11.11  12 25.00 
10th year  6  0.00  16 31.25 
11th year  2  0.00   3  0.00 
12th year and above 25 36.00  75 29.33 
 
  
148 
 
 
 
Table C. 6: Cross-Sectional Effects for Appointments Made Between 2013 and 2020 
 DV: inside-agency work experience 
 T ? 1 T ? 1 year T ? 1 year T ? 1 year T ? 2 
year and 3 months and 6 months and 9 months years 
 (1) (2) (3) (4) (5) 
High-profile high-risk agencies      
Marginal effects: ?1 + ?2 0.127 0.103 0.085 0.054 0.026 
 0.080 0.076 0.067 0.067 0.067 
Mean of the dependent variable 0.326 0.326 0.326 0.326 0.326 
N of agency heads appointed 
39 43 55 67 76 
within T 
The other agencies      
Marginal effects: ?1 -0.073 -0.019 -0.028 -0.053 -0.045 
 0.048 0.045 0.045 0.043 0.043 
Mean of the dependent variable 0.264 0.264 0.264 0.264 0.264 
N of agency heads appointed 
71 79 98 119 142 
within T 
Note: Standard errors clustered at the level of provincial party secretaries are reported in the parentheses. 
  
149 
 
 
 
C.4 Robustness Checks: Tenure in Office and Appointments of Agency Insiders 
 
Table C. 7: Tenure in Office and Appointments of Agency Insiders 
 High-profile high-risk agencies  The other agencies 
The proportion of  The proportion of 
??? year in office N N agency insiders (%) agency insiders (%) 
 (1) (2)  (3) (4) 
1st year 59 52.54  165 31.90 
2nd year 64 40.62  159 33.12 
3rd year 57 57.89  157 42.67 
4th year 38 63.15   92 30.43 
5th year 41 46.34   96 42.10 
6th year and above 86 55.81  216 40.46 
Note: Only the cases where provincial party secretaries? in-province time does not equal their tenure 
in office are included. 
 
 
 
  
150 
 
 
 
Table C. 8: Cross-Sectional Effects of Tenure in Office 
 DV: inside-agency work experience 
 T ? 1 T ? 1 year T ? 1 year T ? 1 year T ? 2 
year and 3 months and 6 months and 9 months years 
 (1) (2) (3) (4) (5) 
High-profile high-risk agencies      
Marginal effects: ?1 + ?2 0.031 0.010 0.017 -0.043 -0.030 
 0.071 0.067 0.063 0.061 0.054 
Mean of the dependent variable 0.525 0.525 0.525 0.525 0.525 
N of agency heads appointed 
59 77 95 104 123 
within T 
The other agencies      
Marginal effects: ?1 -0.022 0.013 0.003 -0.002 -0.031 
 0.041 0.040 0.037 0.035 0.033 
Mean of the dependent variable 0.371 0.371 0.371 0.371 0.371 
N of agency heads appointed 
165 203 246 288 324 
within T 
Note: Only the cases where provincial party secretaries? in-province time does not equal their tenure in 
office are included. Standard errors clustered at the level of provincial party secretaries are reported in 
the parentheses. 
 
  
151 
 
 
 
Appendix D: Appendix for Chapter 4 
 
D.1 Robustness Checks: Before-After Effects by Agency 
 
 
Figure D. 1: Before-After Effects by Agency 
  
152 
 
 
 
D.2 Robustness Checks: Before-After Effects by Province 
 
 
Figure D. 2: Before-After Effects by Province 
  
153 
 
 
 
 
Figure D. 3: The Relationship between ?Cleanness? and Provincial-Specific Changes 
  
154 
 
 
 
D.3 Robustness Checks: The Supply Shock 
 
Table D. 1: Before-After Effects Excluding the Cases with Deputy Heads Removed 
 DV: inside-agency work experience 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign -0.063 0.019 0.015 -0.004 -0.003 
 (0.047) (0.032) (0.026) (0.025) (0.023) 
Post campaign ? High-
-0.060 -0.136 -0.123 -0.092 -0.090 
profile high-risk 
 (0.087) (0.064) (0.055) (0.052) (0.049) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Number of province-
394 428 436 439 439 
agency 
Number of observations 631 923 1163 1274 1419 
      
High-profile high-risk 
     
agencies: 
Mean (before) 0.434 0.445 0.447 0.427 0.425 
Before-after (?1 + ?2) -0.123 -0.117 -0.108 -0.096 -0.093 
 (0.072) (0.059) (0.050) (0.048) (0.043) 
The other agencies:      
Mean (before) 0.230 0.243 0.248 0.261 0.268 
Before-after (?1) -0.063 0.019 0.015 -0.004 -0.003 
 (0.047) (0.032) (0.026) (0.025) (0.023) 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
 
  
155 
 
 
 
Table D. 2: Average DID Effects Excluding the Cases with Deputy Heads Removed 
 DV: inside-agency work experience 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign ? High-
-0.075 -0.147 -0.128 -0.098 -0.096 
profile high-risk 
 (0.083) (0.063) (0.053) (0.050) (0.047) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Year fixed effects Yes Yes Yes Yes Yes 
Number of observations 631 923 1163 1274 1419 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
  
156 
 
 
 
D.4 Robustness Checks: Staggered Adoptions Using Central Inspections 
 
 
Figure D. 4: Nonparametric Dynamic DID Effects Using Central Inspections 
  
157 
 
 
 
Table D. 3: Average DID Effects Using Central Inspections 
 DV: inside-agency work experience 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post inspection ? High-
-0.129 -0.166 -0.121 -0.104 -0.105 
profile high-risk 
 (0.083) (0.056) (0.049) (0.046) (0.042) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Year fixed effects Yes Yes Yes Yes Yes 
Number of observations 662 962 1213 1325 1474 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
  
158 
 
 
 
 
Figure D. 5: Average DID Effects Using Central Inspections and Randomization Inference  
  
159 
 
 
 
D.5 Robustness Checks: Replace the Dependent Variable as Relevant Experience 
 
Table D. 4: Before-After Effects Using Relevant Experience as the Dependent Variable 
 DV: relevant work experience 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign -0.031 0.031 0.017 0.015 0.017 
 (0.047) (0.033) (0.027) (0.026) (0.024) 
Post campaign ? High-
-0.010 -0.055 -0.020 -0.010 -0.018 
profile high-risk 
 (0.068) (0.043) (0.037) (0.035) (0.031) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Number of province-
400 432 440 440 440 
agency 
Number of observations 646 941 1186 1289 1436 
      
High-profile high-risk 
     
agencies: 
Mean (before) 0.793 0.765 0.759 0.746 0.742 
Before-after (?1 + ?2) -0.041 -0.024 -0.003 0.005 -0.001 
 (0.066) (0.040) (0.037) (0.037) (0.034) 
The other agencies:      
Mean (before) 0.639 0.643 0.629 0.632 0.634 
Before-after (?1) -0.031 0.031 0.017 0.015 0.017 
 (0.047) (0.033) (0.027) (0.026) (0.024) 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
  
160 
 
 
 
Table D. 5: Average DID Effects Using Relevant Experience as the Dependent Variable 
 DV: relevant work experience 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign ? High-
-0.027 -0.055 -0.017 -0.010 -0.021 
profile high-risk 
 (0.069) (0.043) (0.039) (0.037) (0.033) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Year fixed effects Yes Yes Yes Yes Yes 
Number of observations 646 941 1186 1289 1436 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
  
161 
 
 
 
Table D. 6: Before-After Effects of Agency Outsiders with Relevant Work Experience 
 DV: relevant work experience outside of the agency 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign 0.000 0.031 -0.004 0.006 0.002 
 (0.061) (0.038) (0.033) (0.033) (0.030) 
Post campaign ? High-
0.056 0.059 0.090 0.075 0.058 
profile high-risk 
 (0.105) (0.067) (0.058) (0.055) (0.050) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Number of province-
316 369 391 400 406 
agency 
Number of observations 458 657 827 900 995 
      
High-profile high-risk 
     
agencies: 
Mean (before) 0.630 0.573 0.563 0.554 0.549 
Before-after (?1 + ?2) 0.056 0.090 0.086 0.081 0.060 
 (0.087) (0.057) (0.048) (0.047) (0.046) 
The other agencies:      
Mean (before) 0.525 0.524 0.502 0.496 0.494 
Before-after (?1) 0.000 0.031 -0.004 0.006 0.002 
 (0.061) (0.038) (0.033) (0.033) (0.030) 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
 
  
162 
 
 
 
Table D. 7: Average DID Effects of Agency Outsiders with Relevant Work Experience 
 DV: relevant work experience outside of the agency 
 2009-2016 2008-2017 2007-2018 2006-2019 2005-2020 
 (1) (2) (3) (4) (5) 
Post campaign ? High-
0.046 0.065 0.093 0.073 0.049 
profile high-risk 
 (0.104) (0.066) (0.059) (0.055) (0.050) 
Province-agency fixed 
Yes Yes Yes Yes Yes 
effects 
Year fixed effects Yes Yes Yes Yes Yes 
Number of observations 458 657 827 900 995 
Note: Clustered standard errors are reported in parentheses. Standard errors are clustered at the level 
of provincial party secretaries. 
  
163 
 
 
 
Bibliography 
 
Akhtari, Mitra, Diana Moreira, and Laura Trucco.  2022.  "Political Turnover, Bureaucratic 
Turnover, and the Quality of Public Services."  American Economic Review 112: 442-93. 
Ang, Yuen Yuen.  2020.  China's Gilded Age: The Paradox of Economic Boom and Vast 
Corruption.  New York, NY: Cambridge University Press. 
Baker, George P, Michael C Jensen, and Kevin J Murphy.  1988.  "Compensation and 
Incentives: Practice Vs. Theory."  The Journal of Finance 43: 593-616. 
Bawn, Kathleen.  1997.  "Choosing Strategies to Control the Bureaucracy: Statutory Constraints, 
Oversight, and the Committee System."  The Journal of Law, Economics, and 
Organization 13: 101-26. 
???.  1995.  "Political Control Versus Expertise: Congressional Choices About 
Administrative Procedures."  American Political Science Review 89: 62-73. 
Beik, William.  1985.  Absolutism and Society in Seventeenth-Century France: State Power and 
Provincial Aristocracy in Languedoc, Cambridge Studies in Early Modern History.  New 
York, NY: Cambridge University Press. 
Bendor, Jonathan, Amihai Glazer, and Thomas Hammond.  2001.  "Theories of Delegation."  
Annual review of political science 4: 235-69. 
Bendor, Jonathan, and Adam Meirowitz.  2004.  "Spatial Models of Delegation."  American 
Political Science Review 98: 293-310. 
Besley, Timothy.  2006.  Principled Agents?: The Political Economy of Good Government.  New 
York, NY: Oxford University Press. 
Bhavnani, Rikhil R, and Alexander Lee.  2018.  "Local Embeddedness and Bureaucratic 
Performance: Evidence from India."  The Journal of Politics 80: 71-87. 
Bok, Derek.  2003.  "Government Personnel Policy in Comparative Perspective."  In For the 
People: Can We Fix Public Service, eds.  John D. Donahue and Joseph S. Nye Jr.  
Washington DC: Brookings Institution.  255-72. 
Bozcaga, Tugba.  2020.  "The Social Bureaucrat: How Social Proximity among Bureaucrats 
Affects Local Governance."  Working Paper. 
Brancati, Dawn.  2014.  "Democratic Authoritarianism: Origins and Effects."  Annual Review of 
Political Science 17: 313-26. 
Brierley, Sarah.  2021.  "Combining Patronage and Merit in Public Sector Recruitment."  The 
Journal of Politics 83: 182-97. 
Bulman, David J, and Kyle A Jaros.  2020.  "Loyalists, Localists, and Legibility: The Calibrated 
Control of Provincial Leadership Teams in China."  Politics & Society 48: 199-234. 
Burns, John P.  2003.  "'Downsizing' the Chinese State: Government Retrenchment in the 
1990s."  The China Quarterly 175: 775-802. 
???.  1994.  "Strengthening Central CCP Control of Leadership Selection: The 1990 
Nomenklatura."  The China Quarterly 138: 458-91. 
Cai, Yongshun, and Lin Zhu.  2013.  "Disciplining Local Officials in China: The Case of 
Conflict Management."  The China Journal: 98-119. 
Callaway, Brantly, and Pedro HC Sant?Anna.  2021.  "Difference-in-Differences with Multiple 
Time Periods."  Journal of Econometrics 225: 200-30. 
Cameron, A Colin, and Douglas L Miller.  2015.  "A Practitioner?s Guide to Cluster-Robust 
Inference."  Journal of human resources 50: 317-72. 
164 
 
 
 
Carothers, Christopher.  2022.  "Taking Authoritarian Anti-Corruption Reform Seriously."  
Perspectives on Politics 20: 69-85. 
Carreri, Maria.  2021.  "Can Good Politicians Compensate for Bad Institutions? Evidence from 
an Original Survey of Italian Mayors."  The Journal of Politics 83: 1229-45. 
Cavalcanti, Francisco, Gianmarco Daniele, and Sergio Galletta.  2018.  "Popularity Shocks and 
Political Selection."  Journal of Public Economics 165: 201-16. 
Chan, Hon S, and Jie Gao.  2018.  "The Politics of Personnel Redundancy: The Non-Leading 
Cadre System in the Chinese Bureaucracy."  The China Quarterly 235: 622-43. 
Chan, Kwan Nok, and Shiwei Fan.  2021.  "Friction and Bureaucratic Control in Authoritarian 
Regimes."  Regulation & Governance 15: 1406-18. 
Charron, Nicholas, and Victor Lapuente.  2011.  "Which Dictators Produce Quality of 
Government?".  Studies in Comparative International Development 46: 397-423. 
Chung, Jae Ho.  2011.  "Central-Local Dynamics: Historical Continuities and Institutional 
Resilience."  In Mao's Invisible Hand: The Political Foundations of Adaptive 
Governance in China, eds.  Sebastian Heilmann and Elizabeth J Perry.  Cambridge, MA: 
Harvard University Press.  297-320. 
Cohen, David M.  1998.  "Amateur Government."  Journal of Public Administration Research 
and Theory 8: 450-97. 
Colonnelli, Emanuele, Mounu Prem, and Edoardo Teso.  2020.  "Patronage and Selection in 
Public Sector Organizations."  American Economic Review 110: 3071-99. 
Dahlstr?m, Carl, and Mikael Holmgren.  2021.  "Loyal Leaders, Affluent Agencies: The 
Budgetary Implications of Political Appointments in the Executive Branch."  The Journal 
of Politics DOI: 10.1086/717756. 
Deng, Xiaoping.  1980.  "The Present Situation and the Tasks before Us."  In Selected Works of 
Deng Xiaoping.  Beijing: Foreign Languages Press. 
Dewan, Torun, and Rafael Hortala-Vallve.  2011.  "The Three as of Government Formation: 
Appointment, Allocation, and Assignment."  American Journal of Political Science 55: 
610-27. 
Ding, Iza, and Michael Thompson-Brusstar.  2021.  "The Anti-Bureaucratic Ghost in China's 
Bureaucratic Machine."  The China Quarterly 248: 116-40. 
Eaton, Sarah, and Genia Kostka.  2014.  "Authoritarian Environmentalism Undermined? Local 
Leaders? Time Horizons and Environmental Policy Implementation in China."  The 
China Quarterly 218: 359-80. 
Edin, Maria.  2003.  "State Capacity and Local Agent Control in China: CCP Cadre Management 
from a Township Perspective."  The China Quarterly 173: 35-52. 
Egorov, Georgy, Sergei Guriev, and Konstantin Sonin.  2009.  "Why Resource-Poor Dictators 
Allow Freer Media: A Theory and Evidence from Panel Data."  American political 
science Review 103: 645-68. 
Egorov, Georgy, and Konstantin Sonin.  2011.  "Dictators and Their Viziers: Endogenizing the 
Loyalty?Competence Trade-Off."  Journal of the European Economic Association 9: 
903-30. 
Epstein, David, and Sharyn O'halloran.  1994.  "Administrative Procedures, Information, and 
Agency Discretion."  American Journal of Political Science: 697-722. 
???.  1999.  A Transaction Cost Politics Approach to Policy Making under Separate Powers: 
Cambridge: Cambridge university press. 
165 
 
 
 
Evans, Peter B.  1995.  Embedded Autonomy: States and Industrial Transformation, Princeton 
Paperbacks.  Princeton, N.J: Princeton University Press. 
Fairburn, James A, and James M Malcomson.  2001.  "Performance, Promotion, and the Peter 
Principle."  The Review of Economic Studies 68: 45-66. 
Gailmard, Sean.  2002.  "Expertise, Subversion, and Bureaucratic Discretion."  Journal of Law, 
Economics, and Organization 18: 536-55. 
Gailmard, Sean, and John W Patty.  2012.  "Formal Models of Bureaucracy."  Annual Review of 
Political Science 15: 353-77. 
Geddes, Barbara.  1994.  Politician's Dilemma: Building State Capacity in Latin America, 
California Series on Social Choice and Political Economy.  Berkeley, CA: University of 
California Press. 
Goetz, Klaus H.  1997.  "Acquiring Political Craft: Training Grounds for Top Officials in the 
German Core Executive."  Public Administration 75: 753-75. 
Goetz, Klaus H, and Hellmut Wollmann.  2001.  "Governmentalizing Central Executives in Post-
Communist Europe: A Four-Country Comparison."  Journal of European Public Policy 
8: 864-87. 
Gong, Ting.  2002.  "Dangerous Collusion: Corruption as a Collective Venture in Contemporary 
China."  Communist and Post-Communist Studies 35: 85-103. 
Gong, Ting, and Muluan Wu.  2012.  "Research Report on Cases of Corruption in China During 
2000?2009."  Sociological Studies [In Chinese] 215: 204-20. 
Grindle, Merilee Serrill.  2012.  Jobs for the Boys: Patronage and the State in Comparative 
Perspective.  Cambridge, MA: Harvard University Press. 
Grossman, Guy, and W Walker Hanlon.  2014.  "Do Better Monitoring Institutions Increase 
Leadership Quality in Community Organizations? Evidence from Uganda."  American 
Journal of Political Science 58: 669-86. 
Grossman, Guy, and Tara Slough.  2022.  "Government Responsiveness in Developing 
Countries."  Annual Review of Political Science 25. 
Grotz, Florian, Ferdinand M?ller-Rommel, Jan Berz, Corinna Kroeber, and Marko Kukec.  2021.  
"How Political Careers Affect Prime-Ministerial Performance: Evidence from Central 
and Eastern Europe."  Comparative Political Studies 54: 1907-38. 
Harding, Harry.  1981.  Organizing China: The Problem of Bureaucracy, 1949-1976.  Stanford, 
CA: Stanford University Press. 
Hassan, Mai.  2020.  Regime Threats and State Solutions: Bureaucratic Loyalty and 
Embeddedness in Kenya.  New York, NY: Cambridge University Press. 
Heclo, Hugh.  1977.  A Government of Strangers: Executive Politics in Washington.  
Washington, DC: Brookings Institution Press. 
Heilmann, Sebastian, and Elizabeth J Perry.  2011.  "Embracing Uncertainty: Guerrilla Policy 
Style and Adaptive Governance in China."  In Mao's Invisible Hand: The Political 
Foundations of Adaptive Governance in China, eds.  Sebastian Heilmann and Elizabeth J 
Perry.  Cambridge, MA: Harvard University Press.  1-29. 
Hill, Ronald J, and John L?wenhardt.  1991.  "Nomenklatura and Perestroika."  Government and 
Opposition 26: 229-43. 
Hollibaugh Jr, Gary E, Gabriel Horton, and David E Lewis.  2014.  "Presidents and Patronage."  
American Journal of Political Science 58: 1024-42. 
Huang, Haifeng.  2015.  "Propaganda as Signaling."  Comparative Politics 47: 419-44. 
166 
 
 
 
Huang, Yasheng.  2002.  "Managing Chinese Bureaucrats: An Institutional Economics 
Perspective."  Political Studies 50: 61-79. 
Huber, John D, and Cecilia Martinez-Gallardo.  2008.  "Replacing Cabinet Ministers: Patterns of 
Ministerial Stability in Parliamentary Democracies."  American Political Science Review 
102: 169-80. 
Huber, John D, and Nolan McCarty.  2004.  "Bureaucratic Capacity, Delegation, and Political 
Reform."  American Political Science Review 98: 481-94. 
Huber, John D, and Charles R Shipan.  2002.  Deliberate Discretion?: The Institutional 
Foundations of Bureaucratic Autonomy.  New York, NY: Cambridge University Press. 
???.  2011.  "Politics, Delegation, and Bureaucracy."  In The Oxford Handbook of Political 
Economy, ed.  Robert E. Goodin.  Vol. 7.  New York, NY: Oxford University Press. 
Indridason, Indridi H, and Christopher Kam.  2008.  "Cabinet Reshuffles and Ministerial Drift."  
British Journal of Political Science 38: 621-56. 
Jia, Ruixue, Masayuki Kudamatsu, and David Seim.  2015.  "Political Selection in China: The 
Complementary Roles of Connections and Performance."  Journal of the European 
Economic Association 13: 631-68. 
Jiang, Junyan.  2018.  "Making Bureaucracy Work: Patronage Networks, Performance 
Incentives, and Economic Development in China."  American Journal of Political 
Science 62: 982-99. 
Johnson, Chalmers.  1982.  Miti and the Japanese Miracle: The Growth of Industrial Policy, 
1925-1975.  Stanford, CA: Stanford University Press. 
Jones, Benjamin F, and Benjamin A Olken.  2005.  "Do Leaders Matter? National Leadership 
and Growth since World War II."  The Quarterly Journal of Economics 120: 835-64. 
Kautz, Carolin.  2020.  "Power Struggle or Strengthening the Party: Perspectives on Xi Jinping?s 
Anticorruption Campaign."  Journal of Chinese Political Science 25: 501-11. 
Keller, Franziska Barbara.  2016.  "Moving Beyond Factions: Using Social Network Analysis to 
Uncover Patronage Networks among Chinese Elites."  Journal of East Asian Studies 16: 
17-41. 
Kornai, Janos.  1992.  The Socialist System: The Political Economy of Communism.  Princeton, 
NJ: Princeton University Press. 
Kostka, Genia.  2013.  "Environmental Protection Bureau Leadership at the Provincial Level in 
China: Examining Diverging Career Backgrounds and Appointment Patterns."  Journal 
of Environmental Policy & Planning 15: 41-63. 
Kostka, Genia, and Chunman Zhang.  2018.  "Tightening the Grip: Environmental Governance 
under Xi Jinping."  Vol. 27: Taylor & Francis.  769-81. 
Kou, Chien-wen, and Wen-Hsuan Tsai.  2014.  "'Sprinting with Small Steps' Towards 
Promotion: Solutions for the Age Dilemma in the CCP Cadre Appointment System."  The 
China Journal: 153-71. 
Krause, George A, and Anne Joseph O'Connell.  2016.  "Experiential Learning and Presidential 
Management of the Us Federal Bureaucracy: Logic and Evidence from Agency 
Leadership Appointments."  American Journal of Political Science 60: 914-31. 
Krcmaric, Daniel, Stephen C Nelson, and Andrew Roberts.  2020.  "Studying Leaders and Elites: 
The Personal Biography Approach."  Annual Review of Political Science 23: 133-51. 
Kristiansen, Stein, and Muhid Ramli.  2006.  "Buying an Income: The Market for Civil Service 
Positions in Indonesia."  Contemporary Southeast Asia: 207-33. 
167 
 
 
 
Lan, Zhiyong.  2000.  "Understanding China's Administrative Reform."  Public Administration 
Quarterly: 437-68. 
Landry, Pierre F, Xiaobo L?, and Haiyan Duan.  2018.  "Does Performance Matter? Evaluating 
Political Selection Along the Chinese Administrative Ladder."  Comparative Political 
Studies 51: 1074-105. 
Lee, Hong Yung.  1991.  From Revolutionary Cadres to Party Technocrats in Socialist China.  
Berkeley, CA: University of California Press. 
???.  1978.  The Politics of the Chinese Cultural Revolution: A Case Study.  Berkeley, CA: 
University of California Press. 
Lenin, Vladimir Il?ich.  2015[1918].  State and Revolution.  Chicago, IL: Haymarket Books. 
Lewis, David E.  2011.  "Presidential Appointments and Personnel."  Annual Review of Political 
Science 14: 47-66. 
Lewis, David E.  2008.  The Politics of Presidential Appointments: Political Control and 
Bureaucratic Performance.  Princeton, NJ: Princeton University Press. 
Li, Hongbin, and Li-An Zhou.  2005.  "Political Turnover and Economic Performance: The 
Incentive Role of Personnel Control in China."  Journal of Public Economics 89: 1743-
62. 
Li, Ling.  2011.  "Performing Bribery in China: Guanxi-Practice, Corruption with a Human 
Face."  Journal of contemporary china 20: 1-20. 
Li, Xun, Weizheng Lai, Qianqian Wan, and Xi Chen.  2022.  "Role of Professionalism in 
Response to the Covid-19 Pandemic: Does a Public Health or Medical Background 
Help?".  China Economic Review 71: 101733. 
Li, Zeren, and Melanie Manion.  2021.  "The Decline of Factions: The Impact of Purge on 
Political Appointments in China."  Available at SSRN 3446354. 
Liu, Yang, and Liang Ma.  2022.  "Expert or Non-Expert Leadership: The Impact of Expert 
Leadership on Organizational Performance: An Empirical Study Based on Prevention and 
Control of the Covid-19 Epidemic in Chinese Cities."  China Public Administration 
Review [in Chinese] 4: 78-105. 
Lorentzen, Peter.  2014.  "China's Strategic Censorship."  American Journal of political science 
58: 402-14. 
Lui, Francis T.  1986.  "A Dynamic Model of Corruption Deterrence."  Journal of Public 
Economics 31: 215-36. 
Luo, Yadong.  2008.  "The Changing Chinese Culture and Business Behavior: The Perspective 
of Intertwinement between Guanxi and Corruption."  International Business Review 17: 
188-93. 
Manion, Melanie.  1985.  "The Cadre Management System, Post-Mao: The Appointment, 
Promotion, Transfer and Removal of Party and State Leaders."  The China Quarterly 
102: 203-33. 
???.  2004.  Corruption by Design: Building Clean Government in Mainland China and 
Hong Kong.  Cambridge, MA: Harvard University Press. 
Mao, Zedong.  1958.  "Speeches at the Second Session of the Eighth Party Congress (8-23 May 
1958)."  In Selected Works of Mao Tse-Tung.  Vol. 8.  Paris: Foreign Language Press. 
Maranto, Robert.  2001.  "Why the President Should Ignore Calls to Reduce the Number of 
Political Appointees."  The Heritage Foundation Backgrounder 1413. 
168 
 
 
 
Mart?nez-Gallardo, Cecilia, and Petra Schleiter.  2015.  "Choosing Whom to Trust: Agency 
Risks and Cabinet Partisanship in Presidential Democracies."  Comparative Political 
Studies 48: 231-64. 
McCubbins, Mathew D, Roger G Noll, and Barry R Weingast.  1987a.  "Administrative 
Procedures as Instruments of Political Control."  Journal of Law, Economics & 
Organization 3: 243. 
???.  1987b.  "Administrative Procedures as Instruments of Political Control."  Journal of 
Law, Economics, and Organization 3: 243-77. 
McCubbins, Mathew D, and Thomas Schwartz.  1984.  "Congressional Oversight Overlooked: 
Police Patrols Versus Fire Alarms."  American journal of political science: 165-79. 
Mei, Ciqi, and Margaret M Pearson.  2017.  "The Dilemma of ?Managing for Results? in China: 
Won't Let Go."  Public Administration and Development 37: 203-16. 
???.  2014.  "Killing a Chicken to Scare the Monkeys? Deterrence Failure and Local 
Defiance in China."  The China Journal: 75-97. 
Mertha, Andrew.  2017.  "'Stressing Out': Cadre Calibration and Affective Proximity to the CCP 
in Reform-Era China."  The China Quarterly 229: 64-85. 
Meyer-Sahling, Jan-Hinrik.  2008.  "The Changing Colours of the Post?Communist State: The 
Politicisation of the Senior Civil Service in Hungary."  European Journal of Political 
Research 47: 1-33. 
Migdal, Joel S.  1988.  Strong Societies and Weak States: State-Society Relations and State 
Capabilities in the Third World.  Princeton, N.J: Princeton University Press. 
Milgrom, Paul R., and John Roberts.  1992.  Economics, Organization, and Management.  
Englewood Cliffs, NJ: Prentice-Hall. 
Miller, Michael K.  2015.  "Electoral Authoritarianism and Human Development."  Comparative 
Political Studies 48: 1526-62. 
Miyazaki, Ichisada.  1976.  China's Examination Hell: The Civil Service Examinations of 
Imperial China.  New Haven, CT: Yale University Press. 
Moe, Terry M.  1985.  "The Politicized Presidency."  In The New Direction in American Politics, 
eds.  John E. Chubb and Paul E. Peterson.  Washington DC: Brooksings Institution.  235-
71. 
Montagnes, B Pablo, and Stephane Wolton.  2019.  "Mass Purges: Top-Down Accountability in 
Autocracy."  American Political Science Review 113: 1045-59. 
Nie, Huihua, Dongling Han, Liang Ma, and Diyang Zhangnan.  2018.  "State-Business Relations 
Rankings of Chinese Cities 2017."  National Academy of Development and Strategy, 
Renmin University. 
Oliveros, Virginia.  2016.  "Making It Personal: Clientelism, Favors, and the Personalization of 
Public Administration in Argentina."  Comparative Politics 48: 373-91. 
???.  2021.  "Working for the Machine: Patronage Jobs and Political Services in Argentina."  
Comparative Politics 53: 381-427. 
Pei, Minxin.  2016.  China?s Crony Capitalism: The Dynamics of Regime Decay.  Cambridge, 
MA: Harvard University Press. 
Pepinsky, Thomas B, Jan H Pierskalla, and Audrey Sacks.  2017.  "Bureaucracy and Service 
Delivery."  Annual Review of Political Science 20: 249-68. 
Peter, Laurence J., and Raymond Hull.  1969.  The Peter Principle: Why Things Always Go 
Wrong.  New York, NY: William Morrow and Company. 
169 
 
 
 
Pollock, James Kerr.  1937.  "The Cost of the Patronage System."  The Annals of the American 
Academy of Political and Social Science 189: 29-34. 
Prendergast, Canice.  1993.  "The Role of Promotion in Inducing Specific Human Capital 
Acquisition."  The Quarterly Journal of Economics 108: 523-34. 
Qiao, Defu.  2014.  "The Characteristic, Trend, and Innovation in Corruption Control - Report on 
54 Cases of Top Province-Minister Leaders since the Reform and Opening."  Theory and 
Reform [In Chinese] 3: 36-42. 
???.  2013.  "Research Report on 142 Cases of Corruption among Top Officials in Cities 
since the Reform and Opening."  Theory and Reform [In Chinese] 5: 65-71. 
Rose-Ackerman, Susan.  1978.  Corruption: A Study in Political Economy.  New York, NY: 
Academic Press. 
Rothstein, Bo.  2015.  "The Chinese Paradox of High Growth and Low Quality of Government: 
The Cadre Organization Meets Max Weber."  Governance 28: 533-48. 
Samuels, David J, and Matthew S Shugart.  2010.  Presidents, Parties, and Prime Ministers: 
How the Separation of Powers Affects Party Organization and Behavior.  New York, 
NY: Cambridge University Press. 
Scherpereel, John A.  2004.  "Renewing the Socialist Past or Moving toward the European 
Administrative Space? Inside Czech and Slovak Ministries."  Administration & Society 
36: 553-93. 
Shih, Victor, Christopher Adolph, and Mingxing Liu.  2012.  "Getting Ahead in the Communist 
Party: Explaining the Advancement of Central Committee Members in China."  
American political science review 106: 166-87. 
Siegel, David.  2018.  "The Political Logic of Cadre Rotation in Post-Soviet Central Asia."  
Problems of Post-Communism 65: 253-70. 
Snyder, Richard.  1992.  "Explaining Transitions from Neopatrimonial Dictatorships."  
Comparative Politics: 379-99. 
Str?m, Kaare.  2000.  "Delegation and Accountability in Parliamentary Democracies."  European 
journal of political research 37: 261-90. 
Svolik, Milan W.  2012.  The Politics of Authoritarian Rule, Cambridge Studies in Comparative 
Politics.  New York, NY: Cambridge University Press. 
Swart, K. W.  1949.  Sale of Offices in the Seventeenth Century.  The Hague: Martinus Nijhoff. 
Szakonyi, David.  2021.  "Indecent Disclosures: Anticorruption Reforms and Political 
Selection."  American Journal of Political Science. 
Ting, Michael M.  2002.  "A Theory of Jurisdictional Assignments in Bureaucracies."  American 
Journal of Political Science: 364-78. 
Toral, Guillermo.  2021a.  "The Benefits of Patronage: How Political Appointments Can 
Enhance Bureaucratic Accountability and Effectiveness." Paper presented at the Annual 
Meeting of the American Political Science Association, Boston. 
???.  2021b.  "Turnover: How Electoral Accountability Disrupts the Bureaucracy and 
Service Delivery."  Working Paper. 
Tsai, Wen-Hsuan, and Xingmiu Liao.  2018.  "The Authority, Functions, and Political Intrigues 
of the General Office of the Chinese Communist Party."  The China Journal 80: 46-67. 
Van der Kamp, Denise S.  2021.  "Blunt Force Regulation and Bureaucratic Control: 
Understanding China's War on Pollution."  Governance 34: 191-209. 
Volden, Craig.  2002.  "Delegating Power to Bureaucracies: Evidence from the States."  Journal 
of Law, Economics, and Organization 18: 187-220. 
170 
 
 
 
Wang, Erik H.  2021.  "Frightened Mandarins: The Adverse Effects of Fighting Corruption on 
Local Bureaucracy."  Comparative Political Studies. 
Wang, Peng.  2016.  "Military Corruption in China: The Role of Guanxi in the Buying and 
Selling of Military Positions."  The China Quarterly 228: 970-91. 
Wang, Peng, and Xia Yan.  2020.  "Bureaucratic Slack in China: The Anti-Corruption Campaign 
and the Decline of Patronage Networks in Developing Local Economies."  The China 
Quarterly 243: 611-34. 
Wang, Yuhua, and Bruce J Dickson.  2022.  "How Corruption Investigations Undermine Regime 
Support: Evidence from China."  Political Science Research and Methods 10: 33-48. 
Weaver, Jeffrey.  2021.  "Jobs for Sale: Corruption and Misallocation in Hiring."  American 
Economic Review 111: 3093-122. 
Weber, Max.  1978.  Economy and Society: An Outline of Interpretive Sociology.  2 vols.  
Berkeley, CA: University of California Press. 
Wedeman, Andrew Hall.  2005.  "Anticorruption Campaigns and the Intensification of 
Corruption in China."  Journal of Contemporary China 14: 93-116. 
???.  2012.  Double Paradox: Rapid Growth and Rising Corruption in China.  Ithaca, NY: 
Cornell University Press. 
Wood, B Dan, and John Bohte.  2004.  "Political Transaction Costs and the Politics of 
Administrative Design."  The Journal of Politics 66: 176-202. 
Xu, Chenggang.  2011.  "The Fundamental Institutions of China's Reforms and Development."  
Journal of economic literature 49: 1076-151. 
Xu, Guo.  2018.  "The Costs of Patronage: Evidence from the British Empire."  American 
Economic Review 108: 3170-98. 
Xu, Guo, Marianne Bertrand, and Robin Burgess.  2021.  "Organization of the State: Home 
Assignment and Bureaucrat Performance."  Journal of Law, Economics, and 
Organization. 
Yang, Hongxing, and Dingxin Zhao.  2015.  "Performance Legitimacy, State Autonomy, and 
China's Economic Miracle."  Journal of Contemporary China 24: 64-82. 
Yao, Yang, Tianyang Xi, Lixing Li, He Wang, Feng Wan, Qian Zhang, Songrui Liu, and 
Shundong Zhang.  2020.  "Selection, Training, and Incentivizing: Evidence from the 
CCER Officials Dataset."  China Economic Quarterly 19: 1017-40. 
Yeo, Yukyung.  2016.  "Complementing the Local Discipline Inspection Commissions of the 
CCP: Empowerment of the Central Inspection Groups."  Journal of Contemporary China 
25: 59-74. 
Zakharov, Alexei V.  2016.  "The Loyalty-Competence Trade-Off in Dictatorships and Outside 
Options for Subordinates."  The Journal of Politics 78: 457-66. 
Zhan, Jing Vivian.  2012.  "Filling the Gap of Formal Institutions: The Effects of Guanxi 
Network on Corruption in Reform-Era China."  Crime, law and social change 58: 93-
109. 
Zhao, Dingxin.  2009.  "The Mandate of Heaven and Performance Legitimation in Historical and 
Contemporary China."  American Behavioral Scientist 53: 416-33. 
Zhi, Qiang, and Margaret M Pearson.  2017.  "China's Hybrid Adaptive Bureaucracy: The Case 
of the 863 Program for Science and Technology."  Governance 30: 407-24. 
Zhou, Xueguang.  2001.  "Political Dynamics and Bureaucratic Career Patterns in the People's 
Republic of China, 1949-1994."  Comparative Political Studies 34: 1036-62. 
171 
 
 
 
Zhu, Jiangnan.  2008.  "Why Are Offices for Sale in China?? A Case Study of the Office-Selling 
Chain in Heilongjiang Province."  Asian Survey 48: 558-79. 
Zhu, Jiangnan, and Dong Zhang.  2017.  "Weapons of the Powerful: Authoritarian Elite 
Competition and Politicized Anticorruption in China."  Comparative political studies 50: 
1186-220. 
Zhu, Yuchao.  2011.  "'Performance Legitimacy' and China's Political Adaptation Strategy."  
Journal of Chinese Political Science 16: 123-40. 
 
172