Policy diffusion in incomplete decentralized system the case of diffusions of housing provident fund (HPF) policy in urban china

Policy Diffusion in Incomplete Decentralized System:
The Case of Diffusions of Housing
Provident Fund (HPF) Policy in Urban China

CHEN TINGJIA
(B.Law, Nankai University)

A THESIS SUBMITTED
FOR THE DEGREE OF MASTER OF SOCIAL SCIENCES

DEPARTMENT OF POLITICAL SCIENCE
NATIONAL UNIVERSITY OF SINGAPORE
2013


DECLARATION

I hereby declare that this thesis is my original work and it has been written by
me in its entirety.
I have duly acknowledged all the sources of information which have been used
in the thesis.
This thesis has also not been submitted for any degree in any university
previously.

__

___
Chen Tingjia
12 October 2013

i

To my parents,
who gave me the best of their love during the past 24 years

ii


ACKNOWLEDGEMENTS

The topic of policy diffusion has interested me since my undergraduate. It is excited that
this research gives me an opportunity to make some contributions on the existing
knowledge of this topic. During my candidature, especially the period of thesis writing, I
have received numerous help from others.
First and foremost, I would like to express my deepest gratitude to my supervisor,
Professor Wang Cheng-Lung, without whom I could not even finish this paper. His
tremendous support, guidance, and encouragement stimulate me to concentrate on the
study and keep improving myself. The original idea of this thesis was conceived when I
did independent study on public policy with him. Also because of his guidance and
inspiration, I improved my argument more reasonable and much stronger. His
encouragement and intelligence have greatly prompted me to go further in academic
study. Luckily enough, I was just admitted by the University of Arizona to pursue my
Ph.D. in Public Policy. Besides the academic study, Professor Wang also gave me great
help on my life. As an international student without scholarship, I would have had a
much tougher time if he had not offered me the opportunity to be his research assistant.
But now I am so annoyed that my words are too thin compared with his vast support.
I also owe a debt of gratitude to Professor Naomi Aoki. Her course, decentralization
and local governance, has inspired me a lot in this thesis. Moreover, her comments on the
early draft of this thesis are greatly appreciated. I would also like to thank Professor Soo
Yoen Kim. Her training on research design allows this research to follow the scientific

method rigorously.
As the early version of this thesis has been presented in Mid-West Political Science
Association 2013 Conference, I also wish to thank Professor Jale Tosun, the discussant of
my paper. Her sharp questions have pushed me to have a further improvement on this
thesis.

iii


I thank my fellow graduate students at National University of Singapore, who have
enriched my life. The discussions with them always spark the ideas for my current and
future research. Among these individuals, I would like to mention two persons specially.
Li Xiang has introduced me several brilliant works on methodology which have
contributed a lot to this study. And the discussions with Zha Wen on the latest changes in
the judiciary system in China have stimulated both our interests. Thus, we would
collaborate on relative research in the near future.
Last but not the least, I am also grateful to my parents who give me the best of their
love all the time. I also want to thank my fiancé, Xu Chen, who is my dearest friend and
faithful companion. Whenever I encountered any difficulties, their warmest hands and
strongest support are always beside me. Nothing could compete with their support and
love.

iv


TABLES OF CONTENTS

SUMMURY ...................................................................................................................... vii
LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES ........................................................................................................... ix

Chapter 1

Introduction ..................................................................................................... 1

Chapter 2

Literature Review ............................................................................................ 3

2.1 General Discussion on Policy Diffusion ................................................................... 3
2.2 The mechanisms of policy diffusion and its conditions ............................................ 7
2.3 Policy innovation and diffusion in China.................................................................. 9
Chapter 3

Argument and Hypotheses ............................................................................ 13

3.1 Incomplete decentralization and its accountability locus ........................................ 13
3.2 Principal-agent theory and policy diffusion in incomplete decentralization........... 15
3.3 Central-local relations in China............................................................................... 16
3.4 Policy Diffusion of Housing Provident Fund in China ........................................... 19
3.4.1 Urban housing reform in China ........................................................................ 19
3.4.2 An explanation on the diffusion of the housing provident fund policy ............ 22
Chapter 4

Empirical Analysis ........................................................................................ 27

4.1 Data and measurement: the dependent variables .................................................... 27
4.2 Data and measurement: the independent variables ................................................. 28
4.3 Data and measurement: the control variables ......................................................... 29
4.4 Methodology ........................................................................................................... 33
4.4.1 Methodology for Binary Time-series Cross-Sectional Analysis ...................... 33

4.4.2 Methodology for Survival Analysis ................................................................. 34
4.5 Results ..................................................................................................................... 35
v


4.5.1 Results for Binary Time-series Cross-sectional Analysis ................................ 35
4.5.2 Results for Survival Analysis ........................................................................... 42
4.6 Discussion on the Statistical Results ....................................................................... 46
Chapter 5

Conclusion..................................................................................................... 49

Bibliography ..................................................................................................................... 52
Appendix ........................................................................................................................... 60

vi


SUMMURY

Most existing studies of policy innovations focus on the cases in which policy-making
process is decentralized. However, in many developing countries, such as China, the
extent of decentralization is limited. Thus, the determinants and mechanisms of
diffusions of policy innovations in such a less decentralized system could be different
from what has been established in the literature.
Since 1980s, China has launched a full aspect economic reform to transit its national
economy from a centrally planned system to a more market-based economy. Housing
reform, which was deigned to transform the welfare-oriented housing policy to a marketoriented one, was an important part of this policy transformation. This article specifically
analyzes how the housing provident fund (HPF) policy has been diffused in urban China
during 1990s.

This study intends to answer the following questions: (1) what are the key factors that
determine the adoption of policy innovation concerning HPF among Chinese local
governments (cities)? (2) How do central and local governments influence diffusion of
policy innovations in this process? (3) Why are some local governments more likely to be
positively affected by the central government to adopt the policy innovations? Using a
revised theoretical framework which integrates both local political economy factors and
national policy priorities of the central government, this study contributes to explain more
systematically what factors determine the adoptions of new policies in the less
decentralized system of China. In addition, this study applies binary time-series crosssectional analysis and survival analysis to study the large data set concerning policy
innovations and diffusions in China. The findings suggest that local revenue and
enterprises’ economic performance which represent the local level capacity have a
significant positive influence on the adoption of HPF policy, and that the effects of the
national promotion on HPF policy innovation is significantly affected by the condition of
local revenue. However, contradict to the expectation, the average wages of workers and
staff that may reflect the interests of local residents have a negative impact on the
diffusion of HPF policy among Chinese local governments.
vii


LIST OF TABLES

Table 1 Summary Statistics for the Binary Time-series Cross-sectional Analysis .......... 32
Table 2 Summary Statistics for the Survival Analysis ..................................................... 33
Table 3 Maximum-Likelihood Estimation for HPF Policy Innovation ............................ 36
Table 4 Changes in the Predicted Probabilities of HPF Policy Innovation ...................... 37
Table 5 Estimated Hazard Rate of HPF Policy Innovation .............................................. 43

viii

LIST OF FIGURES

Figure 1 China's experimentation-based policy cycle ...................................................... 10
Figure 2 Local Revenue Increases HPF Policy Innovation .............................................. 38
Figure 3 Economic Performance Increases HPF Policy Innovation ................................. 39
Figure 4 Average Wage Decreases HPF Policy Innovation ............................................. 40
Figure 5 National Promotion Decreases HPF Policy Innovation ..................................... 41
Figure 6 Effect of Interaction on Probability of HPF Policy Innovation .......................... 42
Figure 7 Effects of Local Revenue on the Hazard rate of Innovation .............................. 44
Figure 8 Effects of Economic Performance on the Hazard rate of Innovation ................. 44
Figure 9 Effects of Average Wage on the Hazard Rate of Innovation ............................. 45
Figure 10 Effects of National Promotion on the Hazard Rate of Innovation ................... 46
Figure 11 Effects of Interaction Term on the Hazard Rate of Innovation ........................ 46

ix


Chapter 1

Introduction

Since 1980s, China has launched a full aspect economic reform to transit its national
economy from a centrally planned system to a more market-based economy, among
which housing reform was regarded as one of the three main reforms. However, the
Chinese government did experienced a difficult time to pull out the housing system from
the old institutions. Fortunately, the policy innovation of Housing Provident Fund (HPF),
which was borrowed from Singapore, initiated by Shanghai municipal government, and
quickly diffused after the first adoption, broke the impasse and promoted the housing
reform in China. It is a form of long time compulsory housing saving under the
employees’ personal account, through which both employees and employers should

contribute an equal percentage of the employees’ salary to the funds by month. More
importantly, the character of local initiation rather than central pilot selection of this
policy, presents the uniqueness of this policy.
Actually, not only in the case of HPF policy in this study, the diffusion of innovation
has been widely and deeply influencing the policy making process of the governments.
The political arrangement, social development, and policy practice in one region could
easily be observed, understood, and learned by another place. For decades, the practice of
the policy innovation and diffusion largely facilitated the relative researches in academia.
Most of these studies on policy diffusion among local governments are embedded in the
context of the U.S. (Baybeck, Berry, & Siegel, 2011; Berry & Berry, 1990, 1992;
Boehmke & Witmer, 2004; Cohen-Vogel & Ingle, 2007; Gilardi, 2010; Gray, 1973;
Grossback, Nicholson-Crotty, & Peterson, 2004; McLendon, Heller, & Young, 2005;
Shipan & Volden, 2008; Walker, 1969), where the decision making powers are usually
fully decentralized and suits the Tiebout model well. However, as Bardhan (2002) argued,
the assumptions for Tiebout model are too stringent for most developing countries.
According to this, the existing theories on policy innovation and diffusion should be
further revised and improved.
Thus, China, as the largest developing country in the world, should be a prefect
country for the study of policy diffusion in an “incomplete decentralization” system. Also,
1


since the reforms in China, especially the political reform, will still carry on, and local
policy innovation could be a breakthrough of further changing, it is important to deeply
understand how the reform carried forward in the past. Given the uniqueness and
importance of the HPF policy which have been briefly described at the beginning of this
chapter, it would be worth to explore the diffusion process, which could further inspire
other reform of China in the future. Therefore, this research applies binary time-series
cross-sectional analysis and survival analysis to study the large data of HPF diffusion in
China by answering the following questions: (1) what are the key factors that determine

the adoption of policy innovation concerning HPF among Chinese local governments
(cities)? (2) How do central and local governments influence the diffusion of policy
innovations in this process? (3) Why are some local governments more likely to be
positively affected by the central government to adopt the policy innovations? The
findings suggest that the local revenue and enterprises’ economic performance that
represent the local level capacity has a significant positive influence on the HPF policy
innovation, and that the effects of national promotion on HPF policy innovation is
significantly affected by local revenue. However, contradict to the expectation, the
average wage of local workers and staff that reflects the local interests has a significantly
negative impact on the HPF policy innovation. Detailed analysis and discussion will be
provided in the following chapters.
The organization of this paper is as follows. Chapter two reviews the literatures on
policy diffusion and positions the current research studies in the existing literatures. In
Chapter three, several hypotheses and related causal mechanisms are constructed by
analyzing the case of HPF policy innovation. Chapter four empirically tests the
hypotheses that have been raised. Chapter five concludes.

2


Chapter 2

Literature Review

This chapter reviews the main body of literature related to the current research. I first
focus on the general discussions on the topic of policy diffusion, especially the horizontal
and vertical diffusion. Then, I turn my attention to the mechanisms and conditions of the
diffusion of innovations. With a careful scrutiny, this research shows the necessity to
further explore the determinants and process of policy diffusion in an incomplete
decentralized system. Finally, as China, the largest developing country in the world, is

the main focus of this research, I present the existing studies which analyze the policy
innovation and diffusion in China. The previous literature has drawn a general picture on
the process of policy diffusion in China and emphasized the importance of the central
government; however, they ignore the active role of the local governments and the
interaction between the central and local governments during the diffusion process. This
research will contribute to the existing literature by studying the policy diffusion process
in a developing country and by examining the interaction between central and local
governments in the diffusion of innovations.

2.1 General Discussion on Policy Diffusion
The study of policy diffusion has emerged for decades long. Rogers (1962) gave a
comprehensive and systematic analysis on the general diffusion of innovations. In his
book, he presented the definition, elements, and process of diffusion. “Diffusion is the
process in which an innovation is communicated through certain channels over time
among the members of a social system.” (Rogers, 1962, p. 5) In terms of the diffusion in
the field of the political science and public policy, Walker defined the policy innovation
as “a program or policy which is new to the states adopting it” (Walker, 1969, p. 881),
meaning that it is still an innovation for the states who do not have such a policy before
no matter whether or not some other states have already adopted it. So it is different from
“policy invention”, which means the formation of original policy ideas (Berry & Berry,
2007). The spread of the innovative policy from the earlier adopters to the later ones is
called policy diffusion.
3


There are mainly two types of policy diffusion. One is horizontal; while the other is
vertical. The horizontal policy diffusion usually means the diffusion across nations or
among the local governments in a country. For this group of study, the existing literature
focuses on either the internal or external factors that lead to the policy diffusion.
Traditionally, the domestic conditions could strongly dominate the policy adoption within

its jurisdiction, so when people analyze the adoption of a policy innovation in a
jurisdiction, the internal factors are usually considered. In numerous empirical studies,
these internal factors, such as ideological position and prior beliefs (Gilardi, 2010;
Grossback et al., 2004), socioeconomic conditions (Walker, 1969) , election period
(Berry & Berry, 1990, 1992), the character or capacity of the leaders (Teodoro, 2009),
have been demonstrated their significant roles in the adoption of policy innovation.
However, the domestic determinants alone could not explain the story of policy diffusion
as a whole.
Walker opened a window to explore the diffusion of innovations from the external
factors, namely, the “regional reference groups”. In his pioneering study, Walker found
that in each of the regional clusters, the states usually followed the comparable leading
state to adopt a new program. Followed this study, Gray (1973) analyzed the adoption of
a new policy from the interaction between the adopters and the potential adopters. By
analyzing the cumulative frequencies of states adopting the law in the areas of civil rights,
welfare, and education, she pointed out that various issues could have different diffusion
patterns which had not revealed significant variations in Walker’s study, and a state may
adopt policies in various arenas with distinctive speeds. Even though Walker (1973)
criticized this article as its “completely intermixed population” did not take the regional
barriers (namely, whether there is an effective communication channel between two
states from different broader regions) into consideration and may not provide stable
feature of states’ innovative capacity, these two studies did pioneer the policy innovation
analysis from the perspective of the regional factor, namely, how the policy adoption of a
state is influenced by the regional leader or previous adopters. Influenced by their
valuable explorations, many following studies have emphasized the importance of the
neighbor’s policy adoption, either the neighboring state or local governments within a
country or the neighboring country (Baybeck et al., 2011; Berry & Berry, 1990, 1992;
4


Boehmke & Witmer, 2004; Chien & Ho, 2011; Clark, 2009; Cohen-Vogel & Ingle, 2007;

Linos, 2011; McLendon et al., 2005).
It is worth mentioning that Berry and Berry’s study (1990, 1992) on the diffusion of
state lottery and tax collection unified the internal determinants model and regional
diffusion model and combined the internal characters (i.e. the political, economic, and
social factors) and the regional influence together, by using the event history analysis
(EHA). Based on Mohr’s analysis of organizational innovation, they proposed that an
organization’s motivation to adopt an innovation is affected by its internal characters as
well as the information of other organizations, since the obstacle, namely, uncertainty,
that it may face can be overcome by the resources such as the information of relative
innovation practice by other organizations.
However, what should be noticed is that the influence of the neighbors may not always
be autonomous, the catalysis by the third party as a facilitator or a communication
channel is thus necessary for the diffusion. In terms of the cross national policy diffusion,
overwhelming studies have highlighted the function of international organizations in
policy diffusion because the international organizations convey valuable information
(Cao, 2009, 2012), and also the less-developed countries are more likely to rely on the
financial aid from the international organizations, such as the World Health Organization
(Clark, 2009). In additional to the international organizations, some transnational
corporations can also promote such diffusion by coercing the host country to adopt a
certain policy (Dolowitz & Marsh, 1996). These outside promoters have largely
accelerated the spread of innovative policies.
Surely, such outside promotions by the third parties not only exist in the cross-national
policy diffusion, but also exist in the domestic policy diffusion, which is the focal point
of this study, as well. Policy entrepreneurs play a significant role in the process of the
diffusion of innovation, since they are the people “who are willing to invest their
resources in pushing their pet proposals, are responsible for prompting important people
to pay attention and coupling solutions to problems” (Kingdon, 1984, p. 21). For instance,
Mintrom (1997) stated that the presence and action of policy entrepreneurs can
significantly increase the probability of the state legislative approval of school choice,
5

even consider the potential determinants, such as state politics, the choice in neighbor
states. As a brunch of policy entrepreneur, the behaviors of the bureaucrats have drawn
some scholars’ attention. Some empirical research shows that the governments which
tend to hire the chiefs from other organizations and the diagonal promoted bureaucrats
are more likely to make the policy innovations (Teodoro, 2009).
Another important type of policy diffusion is vertical diffusion, which emphasize the
influence of the federal or superior governments on the state/local governments. Berry
and Berry (2007) maintained that national government plays the analogous role to the
“widely respected leader state”, while the former one has more instruments to intervene
the local policies. These instruments include administrative mandates and creating
financial incentives. Daley and Garand (2005) made a contribution on exploring the
determinants of variation in the strength of state hazardous waste programs. They
examined how the action of the federal government (i.e. financial support) as long as the
spillover effect between state governments and the internal determinants, such as problem
severity, the state political and economic conditions, and the interest group pressure, can
affect the policymaking in one state. Besides the financial incentive, Karch (2006)
examined how the national legislations affect the state-level enactment of policy
innovations under no coercive pressure or financial stimulation, which is called catalytic
control by him. He maintained that the influences of national intervention on the state
innovation are determined by whether the national intervention lowers, or raises, or has
no impact on the obstacle. And he continued to study whether the national activities, such
as the president’s national televised address and national wide debate, can significantly
increase the probability that states would introduce a relative agenda-setting (Karch,
2012). Interestingly, Shipan and Volden found that the release of the state law on
antismoking will decrease the probability that a city in that state to adopt a same law
(Shipan & Volden, 2008), which implies that centralization may become an obstacle for
the local policy innovation (Shipan & Volden, 2012).
Besides the research above, some other scholars pay attention to the bottom-up policy

diffusion (Clavier, 2010; Shipan & Volden, 2006). For example, Shipan and Volden
(2006) tested the city-to-state diffusion as well as the state-to-state and national-to-state
6


diffusion by examining the diffusion of antismoking laws passed between 1975 and 2000.
They maintained that the legislative professionalism of a state and policy advocates at the
state level would affect the probability of the statewide adoption of a local law. These
multilevel diffusion studies open up our minds in studying policy diffusion analysis.
However, when analyzing the influence of federal government, these limited studies
overlook the interaction between the national intervention and the internal factors of the
states government.

2.2 The mechanisms of policy diffusion and its conditions
The popular determinants discussed above provide a space to further detect the driving
force of policy diffusion. The scholars also endeavor to detect the mechanisms of policy
diffusion. They either use these mechanisms as independent variables to explain the
features of policy diffusion, such as the temporal or conditional diffusion (Shipan &
Volden, 2008), policy adoption and expansion (Boehmke & Witmer, 2004), or detect the
factors determining the diffusion or adoption of an innovation under a certain mechanism
(Baybeck et al., 2011; Berry & Berry, 1990; Gilardi, 2010). In general, as Shipan and
Volden (2008) presented, the mechanisms of policy diffusion include social learning,
competition, coercion, and imitation. Social learning always refers to that a policy in one
region or organization is adopted by another region or organization, since they meet
similar problems and difficulties, or the adopters admire to the success policy outcome of
others. Because of the limited time, intelligence, and information, just as Lindblom said,
the existing knowledge of the regarding issue is important for the decision makers
(Lindblom, 1959), so they tend to adopt the policy solving the similar issues as theirs. In
terms of the competition mechanism, it usually happens under the situation when two
regions or organizations compete for scarce resources. In this situation, the adoption of

policy innovation is tried to overcome the negative externalities (Shipan & Volden, 2008).
Boehmke and Witmer (2004) distinguished the mechanisms of social learning and
economic competition in the process of diffusion. They argued that the social learning
theory is usually suitable to those states which are searching for the solutions to the
problems that they encountered, while economic competition mechanism is always
7


related to the policies which are pushed by other states' pressure, such as lost business,
tax revenue. Moreover, the coercion mechanism states that a region or organization
adopts a policy coerced by another region or organization in exchange of acquiring
relative aid or avoiding the penalties. Finally, the pioneering members are likely to be
imitated and followed by other membership in a stable group. That is, the followers just
want to “copy the actions of another in order to look like other” (Shipan & Volden, 2008).
Most of these explanations on horizontal policy diffusion are under the conditions of
Tiebout model 1, while those vertical diffusions are analyzed under the federal system in
which there is a clear scope between the power of the federal and state governments. For
example, state governments learn new policy from other states to win votes from their
residents. Also, the competition is possible since the local residents could “vote by feet”
by freely moving from one state to another. This is a quite decentralized system which is
supposed that the governments at the lower tiers have better information on local
preference and conditions. Hence, the local governments could provide the public goods
reflecting the preferences of the local residents to achieve economic efficiency (Oates,
1972, pp. 11-12; Tiebout, 1956) and have a quick action on local affairs (Rondinelli,
Nellis, & Cheema, 1983) once they enjoy the decision-making and fiscal autonomy and
take the responsibility within the regions. This is based on the assumptions of high
political accountability and consistently aggregated preferences in a locality. However, as
Bardhan (2002) argued, the assumptions for Tiebout model, such as population mobility,
mechanisms of political accountability, similar administrative capacity across different
government tiers, are too stringent for the developing countries. Consequently, it is

necessary to

analyze

the diffusion

process

in

developing

countries

whose

decentralizations are usually incomplete. The practice in China, the largest developing
country in the world, may provide us a further understanding on the diffusion process in
an incomplete decentralized system.

1

Tiebout model is a model on local expenditure proposed by Tiebout in 1956. It argues that local
expenditure could be more adequately reflected the local preference than the national level based on the
following assumptions. They are: full mobility of the voters, full knowledge of the local issues of the voters,
large number of communities, no restrictions on employment opportunities, diseconomies between
communities, having and keeping optimum size of the communities. (Tiebout, 1956)

8

2.3 Policy innovation and diffusion in China
A sprawling literature on policy innovation in China has been published in Chinese
journals since 2000, which probably resulted from the initiation of the prize for the
innovation of the Chinese local governments. Compared with the empirical studies on
policy diffusion among the local governments in the U.S. and other different countries,
these researches lack the empirical analysis for the diffusion of a specific policy
innovation. However, they could provide some intuitionistic knowledge on the practice in
China. The motivations of the local governors, such as promotion (B. Wang, 2009;
Jiannan Wu, Ma, & Yang, 2007), the desire to innovate, and the pressure from the society
(G. Chen & Li, 2009; He, 2011; Jiannan Wu et al., 2007; X. Yang, 2008) are usually
mentioned. Also, the innovations could vary subject to geographical areas. Although the
proportion of innovations in one region is related to its level of economic development (B.
Wang, 2009), it shows that the eastern area is the most active region in policy innovation
followed by the western provinces, while the middle region is surprisingly the least active
(He, 2011; Jiannan Wu et al., 2007). Besides, some scholars find that the governments in
city and county levels are more likely to have innovative policies (He, 2011; Jiannan Wu
et al., 2007). Nevertheless, little attention was paid to the diffusion of these innovations.
Fortunately, several recent studies could help us to draw a big picture on the diffusion
process in China.
Policy experimentation is an important policy practice in China. It defined as a policy
process “in which experimenting units try out a variety of methods and processes to find
imaginative solutions to predefined tasks or to new challenges that emerge during
experimental activity” (Heilmann, 2008). It differs from the traditional policy diffusion or
spontaneous policy diffusion that the scholars usually studied; Rather, just as Heilmann
said, it is “a purposeful and coordinated activity geared to producing novel policy options”
in which the local experiences or innovation is the important resources for the national
policy formulation (Heilmann, 2008). Even though with these discrepancies, however,
the studies of policy experimentations do contain the same attributes to the diffusion of
policy innovation. In Rogers (1962)’s definition, there are four key elements for diffusion:

an innovation, communicated channels, time, and social system. Since there is a sequence
of time for different jurisdiction to adopt a specific innovative policy in the social system,
9


and the propaganda of the central government is a strong communication channel, policy
experimentation should be regarded as a special kind of policy diffusion.
In his research, Heilmann argued that “experimentation under hierarchy”, the pattern
of experimental governance in China, did and do stimulate the socioeconomic and even
some political changes in China (Heilmann, 2008). His works contribute to the
explanation of how experimentation itself works in China. Three types of China’s
experimentation are given by him, which are experimental regulation, experimental
points, and experimental zones. And China’s experimentation-based policy cycle is as
follows.
Figure 1 China's experimentation-based policy cycle

Source: Heilmann, 2008
Heilmann’s finding in economic area has been verified by the research on policy
diffusion in other areas, such as public service and social welfare, in China. Foster (2004)
made a case study on the diffusion process of "Service Promise System" (fuwu chengnuo
zhi). Though it was initiated under the pressure of the masses, which is different from
Heilmann’s studies, it shows somewhat similar way of the spread of the policy. Namely,
local officials posit a new policy program through the way of policy learning; with a
10


good performance, such a policy program may further draw the attention of the provincial
leaders and even relative central officials. Hence, the higher levels of governments tended
to regard this policy innovation as a model, and promoted this policy to the governments
or agents in other areas or sectors, so that the policy diffusion could happen. Recently,

based on the interplay between central and local governments, Shi (2012) used three
types of diffusion in China, which are coercion, competition, and pre-emption diffusions,
to explain the local activism and national unification in the process of social policies
diffusion.
All the studies above emphasize the importance of local pioneering practice or
experiments and the general coordination and further promotion of the central
governments, yet the intervention of the central government was discussed in a general
way. In a newest research, it identifies two different mechanisms of this central-local
interaction, “mandatory policy diffusion” and “championship policy diffusion” (X. Zhu,
forthcoming). According to Zhu, these two mechanisms are determined by two
intervention methods of the central or superior government, which are called intervention
with administrative command, and the contest with performance evaluation. For the
former one, since under the unitary instruction of the central or superior government, all
the local governments should adopt the same innovative policy; while for the latter one,
the local governments tend to choose some different policy instrument with the same
policy goal, so that they would not lose the champion as followers.
These scholars’ researches provide us valuable knowledge and a way to study policy
diffusion in China. However, there are two major weaknesses for these studies. First, they
discussed the diffusion process in a general and more descriptive way that pointed out the
importance of local pioneering actions and national coordination. Nevertheless, the
general national factor cannot explain the variations of localities in adopting a new policy.
Therefore, with no analysis on the detailed conditions of the localities, we still cannot
understand why and how some regions adopt certain policy faster than the others. Second,
these studies overemphasize the central intervention in the diffusion process. The
interaction between the central and local governments is ignored. Despite a recent
research analyzing the driving force of the diffusion of police departments’
microblogging in China from the perspective of organization innovation (Ma, 2013), it
11

discusses the predominance of the upper government in general without paying attention
to the possible variations in the effect of the upper government influence depending on
the local conditions.
To fill the existing gaps, the research intends to answer the following questions. What
are the key factors that determine the adoption of policy innovation concerning HPF
among Chinese local governments (cities)? How do central and local governments
influence diffusion of policy innovations in this process? Why are some local
governments more likely to be positively affected by the central government to adopt the
policy innovations? By answering these questions, this study tries to explore the specific
factors that influence policy diffusion in China. Shipan and Volden (2012) has
hypothesized that decentralization may be more suitable for policy diffusion than
centralization; however, the meticulous conditions should be scrutinized.

12


Chapter 3

Argument and Hypotheses

In this chapter, I answer three questions First, what are the key factors that determine
the adoption of policy innovation concerning HPF among Chinese local governments
(cities)? Second, how do central and local governments influence diffusion of policy
innovations in this process? Third, why are some local governments more likely to be
positively affected by the central government to adopt the policy innovations? In the
nutshell, my answers to the above questions are as follows. Under the conditions of
incomplete decentralization, both national intervention and local level capacity and
interests determine the adoption of a policy innovation. In present-day China, the local
governments gain the autonomy to some extent that enable them to represent their
interests in local decision making process, while they still have strong accountability to

the central and upper level government which implies the national intervention is
important. As for the case of the Housing Provident Fund (HPF) policy, this chapter
hypothesized that the central government promotion and local level capacity and interests,
such as the local revenue, the performance of the enterprises, and the average income of
the employees, could together determine the diffusion of HPF policy. Furthermore, as the
localities have different interests, national intervention may have various influences on
these localities. For those localities which have stronger incentive to initiate the HPF
policy, they would be more likely to be affected by the promotion the central government.
Thus, national intervention could magnify the influence of the local level capacity and
interests.

3.1 Incomplete decentralization and its accountability locus
Last chapter has argued that the full aspects of decentralization, which is an important
condition for the existing policy diffusion analysis, may not be suitable for most
developing countries as the decentralization in the majority of them is incomplete.
Decentralization means “the transfer or delegation of legal and political authority to plan,
make decisions and manage public functions from the central government and its

13


agencies to field organizations of those agencies, subordinate units of government, semiautonomous public corporations, areawide or regional development authorities;
functional authorities, autonomous local governments, or nongovernmental organizations”
(Rondinelli, 1980, p. 137). Decentralization, which is a relative concept to centralization,
was raised to solve the problems in the later system (Schneider, 2003). It is a continuous
variable rather than a constant. That is, any points between the two ends (extreme
centralization and anarchy) of the continuum of decentralization could be regarded as
“decentralization”. On this continuum, there are mainly different degree of
decentralization, including deconcentration, delegation, and devolution (Rondinelli,
1980). Among these, deconcentration is the minimal degree of decentralization, while

devolution is the most substantial one. The World Bank (2004) makes a detailed
discrimination on these three degrees of decentralization. In a deconcentration system,
the officials at the local level are the employees of the central government rather than of
the local government itself (World Bank, 2004). Therefore, these officials should have
stronger upward accountability than the downward. As for the officials in the delegation
system, they could be the employees of central or local government; they have autonomy
to some extent, while they should also fully or partially be accountable to the center
(World Bank, 2004). Compared with the first two forms, devolution endows the localities
autonomy to the largest extent. The localities in this system could be regarded as an
independent or separate power with little or no direct control of the central government
(Rondinelli, 1980). This research calls the first two forms of decentralization as
“incomplete decentralization”.
As we can see, decentralization in the U.S. belongs to the forms of devolution. It has
the separation power between the federal and state government, high downward
accountability because of the political decentralization and local elections, and free
movement among the states. All these conditions are the foundation of the multitudinous
studies on policy diffusion. However, for most developing countries, local autonomy is
limited. Even with the process of decentralization, the local governments have to partly
or largely rely on their counterpart national government. For many of them, the local
officials are appointed by the central or upper level government, which may impair the
downward accountability in decentralization. Thus, for most of the developing countries,
14


the subnational governments “are not wholly controlled by the central government but are
ultimately accountable to it”, which could be characterized as principal-agent relationship
between the central and local government (Litvack, Ahmad, & Bird, 1998, p. 9). This
accountability locus, as an important premise, may significantly influence the diffusion
process in most of the developing countries.

3.2 Principal-agent theory and policy diffusion in incomplete decentralization
Principal-agent theory is an important branch of literature that focuses on organization
control. In his seminal article, Moe (1984) introduced the principal-agent theory in
political science studies. Since then, the principal-agent theory has been widely used to
discuss the organization control from various perspective, such as bureaucracy (e.g.
Gailmard, 2010; K. Yang, 2009), the relationship between politics and administrative
system (e.g. Worsham & Gatrell, 2005), and even the relationship between the states and
international organizations (e.g. Worsham & Gatrell, 2005). As what has just been talked
about, it also provides us an angle to analyze the central-local relations under the
conditions of “incomplete decentralization”.
According to this theory, principals and agents usually have distinct interests, since
they need to make the rational choice for themselves; Also, the agents have incentives to
hide the information from principals to maximize their own interests (Miller, 1992, p. 2;
Moe, 1984). Therefore, “information asymmetries” and “conflict of interest” are the two
main reasons for the “agency problem” (Moe, 1984). To reduce the “agency problem”,
the principals need to build monitoring devices and incentive structures to ensure the
agents work in a proper way.
In a decentralized system that locality has limited autonomy and high upward
responsibility, the central or upper level government and the local governments together
form a principal-agent relationship. On the one hand, because of the endowment of
autonomy through decentralization, the local governments would like to follow their own
interests when they make a decision. Also, the subordinates may not always behave as
what their superior expected. So whether adopting a new policy or not is highly likely to
15