McGill Journal of Political Studies 2014 April, 2014 | Page 12

Manion, for instance, uses game theory to
model the strategic calculation involved in
committing bribery and argues that in order
to completely move away from corrupt
equilibria, significant institutional change is
necessary4.
Although some existing studies do note
that high-level corruption (i.e. corruption
committed by high-ranking officials)
and low-level corruption (i.e. corruption
committed by low-ranking officials) exist
simultaneously in China5, none have actually
established the distinction or explained how
and why they are different. As aggregate
data treats all corrupt officials and their
cases the same way, existing literature tend
to overlook heterogeneity and potentially
important nuances within the group of
corrupt officials. This paper will use original
data comprising both high-ranking and
low-ranking corrupt officials and examine
whether the CCP does treat the two groups
in a systematically different manner.
Methodology
Multiple linear regressions were run in
R using original data collected specifically
for this research project. The goal is to first
examine whether there is a correlation
between the interested explanatory variable,
rank, and the dependent variable, duration of
internal investigation, and to assess whether
such correlation still exists when controlling
for an increasingly larger number of other
variables. Furthermore, I also performed a
paired t-test in R to find out whether the two
groups, high-rankings and low-rankings,
are significantly different from each other
Power and Democracy, Cambridge University Press,
2005; Hon S. Chan & Jie Gao, “Old Wine in New
Bottle: a county-level case study of anti-corruption
reform in the People’s Republic of China: Crime, Law
and Social Change, 2008.
4
Melanie Manion, “Corruption by Design: Bribery
in Chinese Enterprise Licensing,” The Journal of Law,
Economics & Organizations, V12 N1, 1996.
5
For instance Chan & Gao (2008). However, the
study focus solely on low-level corruption as it
examines the efficacy of anticorruption reform at
county-level.
12 | McGill Journal of Political Studies 2014

in terms of how long they were processed
during the internal investigation. The rest
of this section will lay out explanations of
the data collection, variable and the logic
behind the data analysis.
Data Collection
The data used in this project is original,
constructed based on the principle of pairing
every high-ranking corrupt official convicted
between 2008 and 2013 with a low-ranking
counterpart. In order to further control
for heterogeneity within the low-ranking
group, I chose county-level as the standard
for the pairing6. The reason why countylevel corruption cases, instead of prefecturallevel ones, are chosen to represent low-level
corruption is that the differences between
provincial- and county-level bureaucrats are
considerably larger and more noticeable in
terms of distance to central power, size of
informal network involved, and amount of
vested interest, etc.
The
latest
Supreme
People’s
Procuratorate report states that there have
been 32 judicially handled (i.e. prosecution
of ) high-level corruption cases between
2008 and 20137. However, the names of the
32 corrupt high-ranking officials were not
revealed. Thus, the first task was to find out
who they are. Through an extensive news
search using the Chinese search engine,
Baidu, I uncovered 30 out of the 32 highranking officials. After noting down each of
their names and their last official position,
I proceeded to the pairing operation. The
method of pairing is twofold, depending
on whether the high-ranking official held
An exception was made for pairing case 18, Wang Yi,
former Vice president of China Development Bank.
Since no detailed information on corrupt officials from
county-level banks were available, I instead found
a low-ranking counterpart at the prefectural level
(Zhang Ye, former president of Changsha City Commercial Bank). This is perhaps because, first, the media
lacks vested interest in scrutinizing bank managers for
corruption; and second, perhaps the kind of economic
crime county bank managers tend to commit is something other than receiving bribes.
7
Xinhua News, October 22, 2013
6

a position that is region-specific (e.g. head
of a province) or central (e.g. minister of a
ministry).
For every high-ranking official who held
power in a specific province, I looked for a
low-ranking official from the same province
and from the same xitong8 who was also
convicted of corruption between 2008 and
2013. Standard search keyword is “province
+ county-level position + corruption.” I read
through the first 20 pages on Baidu to note
down all usable results (i.e. within the 20082013 timeframe and with information on
all variables). Among all the usable cases, I
randomly select one to be included as the
low-ranking case for the corresponding
high-ranking official.
A different approach is adopted to match
a high-ranking official who held a position
that is not region-specific, such as the CEO
of a State-Owned Enterprise (SOE) or the
head of a ministry in Beijing. I pair the non
region-specific official with a low-ranking
counterpart who held power at a bureau of
similar function at the county-level9.
The result is 30 pairs of corrupt officials,
composing a sample size of 6010. After I
came up with the names and positions of
these 30 high-ranking officials and the
30 corresponding low-ranking officials, I
recorded the information and codified the
values of the six variables listed below. It is
acknowledged that there is certain degree
of selection bias in the process of pairing,
since only those low-ranking officials with
information on all variables are included
in the data11. In other words, there may be
Xitong means system or branch of government. The
concept is used as control variable, which will be more
thoroughly discussed later in the paper.
9
Note that “bureau with similar function” in this case
is more general than xitong and thus may transcend
the boundary of xitong. The reason why I took the
liberty to interpret function of a specific bureau and
to pair these non-region specific officials with others
of similar function is that, information on low-level
non-region specific corruption cases has proven to be
extremely unavailable.
10
The data set is available upon request.
11
There are three low-ranking officials, case 39, 53
8

implicit reasons why certain low-ranking
officials received full news coverage while
others did not. However, certain controls
(especially of locality and xitong) and
academic rigor persist throughout the data
collection process. Furthermore, considering
this is the first attempt to disaggregate
national statistics and scrutinize who these
corrupt officials are, I am confident that
the imperfect data can nevertheless shed
light on the question of how these cases are
handled by the Party.
Dependent Variables: Duration of
Internal Investigation (t)
The duration, t, is used to measure how
the CCP handles suspected corrupt officials,
because it captures observable characteristics
of how the Party deals with high-level
corruption vis-à-vis low-level corruption. t
denotes the length of time that an official
is under the internal investigation by the
Disciplinary Inspection Commissions
(DICs), which better reflects the interest
of the Communist regime than the other
anticorruption institution, the Procuratorate,
because the DIC is strictly controlled by
the Party whereas the Procuratorate is
technically an independent judicial body.
The process of investigation is considered
to have begun when the central or regional
DIC announces the suspension of an official
for investigation purposes and is considered
completed when the DIC announces
how the official is to be administratively
handled (e.g. removed from the Party and
the former government position) and that
the official has been transferred to the
judicial institution for upcoming judicial
investigation. t is measured in months.

and 54, whose information on age is not available.
These cases were nevertheless included in the data, for
there is simply no other usable data. Considering age
as a variable is only incorporated in model 5 during
the multiple linear regression test, the omission of
information does not greatly impact on the validity of
the test results.

Anticorruption Practices by the Chinese Communist Party | Pan | 13