Divide-and-Conquer Strategy and Political Sociology of the Dictator’s Commitment Power in China

Abstract

This chapter extends the authoritarian institutions literature by showing that the dictator’s commitment ability is critical to whether he will be able to disrupt the elite’s collective actions to resist changes and reforms. Specifically, this chapter develops a conceptual framework for understanding the dictator’s commitment power in China’s political context, and also solves the puzzle about the timing of change by comparing Hu’s and Xi’s political tactics. Compared to Hu, the fact that Xi was able to come up with an unprecedented package for enforcing the anti-corruption campaign—ranging from the level of cadres who were purged to the scope of institutional changes that were introduced—enhanced dramatically his commitment power, which then allowed him to play the divide-and-conquer strategy and bring changes to China’s authoritarian institutions. Ultimately, the analysis completes the final piece of the jigsaw puzzle about how Xi was able to pull off the institutional engineering to tame the dictator’s growth curse.

Appendix: Statistical Properties

In fact, the analysis on one’s promotion expectation can not only reflect the strength of the dictator’s commitment ability, but can also be used to create a quantitative measure for further statistical tests. As a consequence, in this appendix, I elaborate on the statistical properties of such a measure (let’s call it the BEW index first).

1.1 Correlation and Weighted Composite Index

Notwithstanding that the argument for using jiguan has been provided above, there is a remaining issue to be dealt with. That is, if there is a high correlation between one’s actual birthplace and work experience, there is no reason to completely exclude the possibility that there is also a high correlation between one’s place of origin and work experience.

For example, presumably, Zhao’s appointment in Shaanxi must have something to do with the fact that his family was originally from there. As a result, just like chushengdi, one’s place origin is also correlated with one’s work experience. Ideally, indicators included in a composite index are supposed to be uncorrelated and independent of each other, so the index resulting from a certain aggregation process (addition, for instance) will capture the individual effects exercised by the included indicators. On the contrary, if some indicators are correlated with each other, the composite index thus constructed will then have a problem of double-counting the influence of those correlated ones.¹⁹

Generally speaking, giving indicators different weights itself is not something problematic in nature as long as we have good reasons to justify why certain indicators deserve more weights than others. For most questions, factional ties, one’s birth tie, is a much cruder measure and arguably contains less information than common work experience where both parties could develop trust based on shared responsibilities and risks associated with their positions in the same organization. Allowing for a non-unity weight on one’s work unit and therefore more influence on the likelihood for possible factional ties is both innocuous and helpful to make the information structure of the index more realistic. As a result, as far as the indicator of birthplaces is concerned, empirical researchers are advised to stick to the decision made above to choose people’s places of origin over their actual birthplaces.

1.2 Measurement Errors and Network Effect

In the following section, I show that the measurement errors created by the three traditional indicators—birth, education, and work experience (henceforth BEW index)—are by no means random. In fact, the measurement errors are systematic and we have to include additional indicators to fix the problem.

1.2.1 Random versus Systematic Measurement Errors

An easy objection that is often raised to quantitative measures of factional ties or promotion expectations such as the BEW index is that their crudity leads to erroneous pairings of faction leaders and followers. Admittedly, since factional affiliations are not directly observable, measurement errors do occur and, more significantly, as I show shortly, the type of measurement errors the BEW index introduces fundamentally calls into question the analytical leverage it promises to deliver. These two shortcomings notwithstanding, they are by no means a good reason to throw the baby out with the bath water.

Given the inevitability of measurement errors, the key issue here boils down to whether the errors are random or systematic and how it is addressed in a regression context in both scenarios. When there is no way to avoid errors in measurement, random measurement errors at least provide a possibility to make a consistent (causal) inference. If what we have is the opposite, then the entire rationale for using the BEW index has to be fundamentally challenged. As a result, the next key question to be answered is whether the measurement errors created by the BEW index are random or systematic. One possible answer comes from Shih (2008) where it is argued that the errors are random rather than systematic because they are due to “imperfect knowledge of the researcher rather than to some characteristics about China” (p. 67). In order to assess this statement, the following analysis examines the probabilistic properties of both types of errors. I first give a linear additive model of measurement errors:

$$ \mathbf{BE}{\mathbf{W}}_i={\mathbf{P}}_i+{\zeta}_i $$

(9.1)

where $\mathbf{BE}{\mathbf{W}}_i$ is the BEW index for CC member $i$, which is supposed to be a proxy for ${\mathbf{P}}_i,$ promotion expectations (or factional ties in Shih (2008) since they are basically the same before we extend the BEW index), and ${\zeta}_i$ denotes errors in measuring ${\mathbf{P}}_i.$ When measurement errors are systematic, the bias can either over- or under-estimate the coefficient of interest. In other words, there is no way to justifiably exclude the possibility that the statistical significance of the variable of interest is due to the upward bias created by the measurement errors. By contrast, random measurement errors artificially make it more difficult to reject the null hypothesis. That is, the presence of random measurement errors is equivalent to raising the level of significance. If the coefficient of interest turns out to be statistically significant at the 95% level, the actual level of confidence for the variable of interest is in fact higher than 95%.

When the measurement errors are random, ${\zeta}_i$ will have full support—that is, $P\left[{\zeta}_i=z\right]>0,$ $\forall z,$ where $z$ denotes the set of all possible values ${\zeta}_i$ can take. Graphically, this is what the dotted curve (ignore the Hui-Zhang probability space for the moment) in Fig. 9.1 tries to visualize. Intuitively, when ${\zeta}_i$ has full support, all errors are likely to happen with some probabilities, and we don’t know which ones are going to or not going to happen for sure. In other words, a measure with random errors does not systematically exclude cases falling in a certain range of values that ${\zeta}_i$ might take.

Fig. 9.1

Probability density function of the BEW measurement errors: random versus systematic

By contrast, systematic errors do not have full support, and some of the values ${\zeta}_i$ might take have zero probability ($P\left[{\zeta}_i=k\right]=0,$ where $k\in z\Big).$ The extreme case is when the errors are a constant ($P\left[{\zeta}_i=\overline{z}\right]=1,$ where $\overline{z}$ is a constant.) Graphically, this is what the vertical line in Fig. 9.1 represents. As far as the BEW measurement errors are concerned, we surely won’t have this type of systematic errors for our lack of comprehensive knowledge about all CC members’ factional ties and owing to the non-uniform nature of the errors. Nonetheless, this does not mean we can then rest assured that the measurement errors the BEW index causes are random. In fact, as the case study on the network effect has shown above, it does systematically exclude a certain type of case.

1.2.2 Construction of Indirect Linkage Indicator

To fully appreciate network effect in probabilistic terms, formally, let’s first define

$$ {\mathbf{P}}_i=\Big\{{\displaystyle \begin{array}{ll}{\overline{\mathbf{P}}}_i& \sup \left\{{\mathbf{P}}_i\right\}\\ {}{\underset{\_}{\mathbf{P}}}_i& \operatorname{inf}\left\{{\mathbf{P}}_i\right\}\end{array}} $$

(9.2)

Given the fact that Hui and Zhang share the same political mentor and a similar career path, in order to simplify the analysis without losing generality, we assume that Hui’s and Zhang’s real factional tie (/promotion expectation) parameters are identical. In addition, because of their political eminence, there is every reason to believe their ${\mathbf{P}}_i$ to fall within the region above the upper bound.²⁰ As a consequence, the supremum of ${\mathbf{P}}_i,$ ${\overline{\mathbf{P}}}_i,$ is assigned to be their true parameter of political connections.

As far as their BEW scores are concerned, since Hui and Zhang do not share anything in common with any of the standing Politburo members, their $\mathbf{BE}{\mathbf{W}}_i^{Hui- Zhang}=0$ and their measurement errors can thus be given by

$$ {\zeta}_i^{Hui- Zhang}=-{\overline{\mathbf{P}}}_i $$

(9.3)

which is sure to fall in the region of extreme values on the left-tail of both curves in Fig. 9.1. However, based on our brief survey of Hui’s and Zhang’s personal political histories, we know for certain that their ${\mathbf{P}}_i$ is unlikely to be $0$ and $\mathbf{BE}{\mathbf{W}}_i^{Hui- Zhang}=0$ is wrong. Hence, it must be the case that $P\left[{\zeta}_i^{Hui- Zhang}=-{\overline{\mathbf{P}}}_i\right]=0$.

Since now we know Hui’s and Zhang’s indices given by the BEW are wrong for sure, those cases where CC members also have a zero BEW score (which is the lowest possible value it can take) but even more robust factional ties and higher promotion expectations (namely, higher ${\mathbf{P}}_i$) are equally unlikely to happen. As a result, the entire probability space to the left of $-{\overline{\mathbf{P}}}_i,$ which we call the Hu i-Zhang probability space and is assumed to exist when the measurement errors are random, will collapse to a segment on the X-axis.

With this change, it is clear by now that the ${\zeta}_i$ represented by the solid curve in Fig. 9.1 no longer has full support (i.e., some of the possible values ${\zeta}_i$ can take have zero probabilities). In other words, using the BEW index as a measure of factional ties (or promotion expectations) is destined to create systematic measurement errors. Moreover, in a regression context, when the BEW index is used as an independent variable and its coefficient is statistically significant, we won’t be able to tell if the statistical significance should be attributed to the effect of the BEW index itself, or to the bias created by the systematic errors. As a consequence, we either have to give up using the BEW index, or must find a way to get rid of the systematic bias arising from the exclusion of the factional ties established through non-spatial channels.

From a pure probability/statistical point of view, we can restore the Hui-Zhang probability space by increasing the dimensionalities of the original index, or the data generating process. That is, if we can collect data for another indicator to capture the effect exerted by non-spatial networks, the measurement errors thus incurred will then be random/stochastic for the entire range of values the new BEW-plus index can take. More concretely, first of all, while I am not sure if all those cases falling in the Hui-Zhang probability space have positive scores according to the new BEW-plus index (denoted by $\mathbf{BE}{\mathbf{W}}_i^{\ast}\ge 0$), for those that do, their measurement errors can be given by

$$ {\zeta}_i^{\ast }=\mathbf{BE}{\mathbf{W}}_i^{\ast }-{\overline{\mathbf{P}}}_i>{\zeta}_i^{Hui- Zhang} $$

(9.4)

and, graphically, in Fig. 9.1, these points will fall in the region to the right of $-{\overline{\mathbf{P}}}_i$ instead.²¹ As for those points with ${\zeta}_i^{\ast }<-{\overline{\mathbf{P}}}_i$,²² their uncertainties are also restored. The reason is that, with this newly added indicator along with others I am going to introduce shortly, the index has now captured (with some stochastic errors) all systematic sources of influence on factional ties/promotion expectations, and therefore its residuals, ${\zeta}_i^{\ast },$ will have full support, which is a critical condition for the measurement errors to be random.

Now since we have explained the systematic bias created by the BEW index and how the bias can be corrected by introducing a new indicator of non-spatial networks, the current section details how exactly this indicator is constructed.

Figure 9.2 presents all major components of the BEW-plus index. The three indicators of the BEW index mainly capture the direct links between CC member $i$ and the Politburo’s standing member $j$ and their data are also readily available from China’s official publications. By contrast, non-spatial networks have no natural definitions according to which we can determine how close it has to be for CC member $i$ to qualify as a member of the faction led by Politburo member $j$.

Fig. 9.2

Construction of the composite index

This issue is also visualized in Fig. 9.2. The number of linkages $\left({\sum}_{i=1}^n{k}_i\right)$ represents the size of a particular faction, and, presumably, the distance between faction leader $j$ and other faction followers is not uniformly distributed. As a consequence, some ties between the leader and followers within the faction might not be directly built but through the introduction of a member who has more direct access to the leader (i.e., linkage ${k}_1$). The example of Hui Liangyu and Zhang Dejiang perfectly epitomizes this logic. Neither of them enjoyed direct ties with Jiang Zemin, so they gained their factional membership via an important linkage, Gao Di, who was their political mentor back in Jilin. If we define direct interactions through the BEW channels as the first-order relationship, the tie built via linkage ${k}_1$ in Fig. 9.2 can be called a second-order relationship since its coming into being is predicated on the former.

Issues arise when it comes to the question of whether we should also include $n$th-order relationship for $n\ge 3$ in our new BEW-plus index. Although including third-order relationships and beyond might enable the new index to capture a fuller picture of China’s political landscape at the elite level, we decide to limit our scope of research to second-order relationships for the following reasons. In the first place, compared to second-order relationships, the determination of third-order ones needs too much subjective interpretation, and the amount of information is not equally available across all CC members, either. The indicator thus constructed can be very misleading. Second, even for those cases we can be more certain about, the tenuousness in relationships above second order also makes their effects on promotion or policy outcomes highly questionable.

1.3 Additional Factors

The inclusion of non-spatial networks above is mainly justified on the grounds that neglecting them will lead to a systematic bias. The following factors, by contrast, are primarily incorporated because of more substantial issues related to our question of interest, promotion expectations.

1.3.1 Age

While the law of the jungle is in general the only rule in town in China’s elite politics, there are two impersonal rules regarding the retirement age of both Politburo standing members and the government officials at full ministerial and provincial levels (zhengshengbuji). Regarding the former, since the 16th National Congress of the CPC in 2002 when Li Ruihuan, one of the nine standing Politburo members back then, retired at 68, retiring at this age has become a convention for Politburo members.²³ As for the latter, since 1982 (Kou et al. 2006), they are required to retire at 65 if they are not promoted to the leader positions at the national level.²⁴

However, while this age rule is generally the case for most of China’s political elites, the rule for non-standing Politburo members is slightly different. As one Chinese political pun, “Seven, Go up; Eight, Come Down” (Qishang Baxia),²⁵ has it, those non-standing Politburo members who are aged 67 or below can still be considered for the membership of the Politburo’s standing committee, but those who are 68 or above will have to retire from their current positions once they finish their tenure.²⁶ In other words, those political elites who have joined the “succession echelon” enjoy two extra years and don’t have to be forced to retire even when they reach the 65-limit at the end of their term of office. As a result, we adopt a different coding rule for non-standing Politburo members.

1.3.2 Minorities

The inclusion of this indicator in our BEW-plus index is motivated by the particular form of power sharing in China, which can be a factor orthogonal to the faction-formation the original BEW index tries to model. While China is not in any sense close to what Lijphart (1999) calls a consensual democracy, it nonetheless also designs a certain way to share power with its minorities in both ethnic and gender terms without undermining the communist monopoly over political power²⁷ (Wang 2005a; Zhang 2008). The first implication of this affirmative action is that the promotion trajectories for minority CC members will be qualitatively different from those for male and Han-Chinese ones.

As to how different they are, two patterns can be readily identified. There are symbolic positions at different levels of government reserved for minority political elites (Zhao 2004). For instance, the leader positions in All China Women’s Federation and Ethnic Affairs Committee of National People’s Congress are all held by women and representatives of ethnic minorities. In addition, in China’s ethnic autonomous regions, ethnic minorities are overrepresented in the subnational governments. However, what needs to be borne in mind is that key Party positions are still held by Han-Chinese and the vast majority of ethnic minority government officials are functionaries (Sautman 1999). These two types of positions add up to a minority-only career track that is not open to Han-Chinese.

Nonetheless, while most minority political elites are constrained to positions belonging to the two aforementioned categories, this does not mean the door to political prominence at the national level is completely shut for them. In terms of ethnicity, prominent examples of minority government officials who broke the glass ceiling include Hui Liangyu of the Hui ethnic group, current Vice Premier as well as a non-standing Politburo member, and Dai Bingguo of the Tuijia ethnic group, the current State Councilor.²⁸ As for gender, Wu Yi, former Vice Premier and a non-standing Politburo member who retired in 2007, makes a good example of what a female government official can achieve on the non-minority career track.

While it takes a statistical test to show if there is really a “minority factor”²⁹ at play in the promotion of non-minority-only positions, as far as the construction of our promotion expectation index is concerned, there is a good theoretical reason for why the index should incorporate this information. That is, for those minority political elites who make it to the pool of candidates for non-minority-only positions,³⁰ when one of them competes for the position of this sort with a Han candidate, the decision-maker will be more likely to choose the former since this helps forestall potential ethnic tension and strengthen the Chinese identity of minority groups by showing that they can also be “real” national leaders (instead of those decorative ones).

Nevertheless, the co-opting logic epitomized by the second pattern above has its caveat when applied to ethnic minorities. That is, under the officially defined umbrella of shaoshu minzu (ethnic minorities), there is a wide variation depending on how sinicized they are, which greatly affects the Chinese government’s strategy of co-optation. Even though we do see successful stories such as Hui Liangyu and Dai Bingguo, they nonetheless both belong to highly sinicized ethnic groups (Hu i and Tujia, respectively),³¹ and the similar co-optive logic in fact does not apply to the less assimilated Tibetans and Uighurs.³² There are two main reasons for this discrimination against them. First, the low degree of sinicization prevents them from establishing trust with the Hans at all levels of society. Without trust at the social level, it becomes extremely difficult for them to join any political networks that can help them switch to the normal career track. Second, their marginalization in China’s political system has an international aspect as well. The fact that there are outspoken foreign-funded leaders in exile for both groups, that is, Dalai Lama for the Tibetans and Rebiya Kadeer for the Uighurs, also makes China’s political leaders feel unable to trust them politically.³³

More importantly, the sidelining of both the Tibetans and the Uighurs means more than their bleak promotion prospects but also the reduced political agency that would otherwise be complete given their Central Committee membership. In other words, many political economy implications derived from ideas such as fragmented authoritarianism and reciprocal accountability also do not really apply to them. They simply don’t enjoy much political agency in China’s political system. I draw two conclusions here. In the first place, the political dynamics for minorities are qualitatively different from other non-minority elites. Second, to capture this difference within the officially defined shaoshu minzu, we exclude the less sinicized ethnic groups, the Uighurs and the Tibetans, from our dataset.

1.3.3 Native Province

The inclusion of this indicator is in fact a natural extension of the logic underpinning ${\mathbf{B}}_i$ as one of the political–sociological bases for a factional tie to develop. That is, in addition to people with the same birthplaces, senior CPC leaders also prefer those who are the provincial leaders of their native provinces in recruiting their new faction members (Shih 2008, p. 67). To take this factor into account, we add to the new index the following dummy variable

1.4 Aggregate Promotion Expectations

In the previous sections, I have explained at length various potential sources of influence on individual CC members’ promotion expectations. Since the analysis on the dictator’s growth curse in Part II is aggregate at the sectoral level, the following analysis discusses each institutional channel of sectoral interest in detail. Here is the general schema³⁴:

Figure 9.3 shows four possible channels through which a sector gains its political agency and how promotion expectations can affect the adopted payment schemes. These four channels can be further classified according to two dimensions, spatial and temporal. In terms of the spatial dimension, a CC member could be interested in and be a representative for a particular sector either when she holds a position in the central government directly supervising that sector, or when she is the leader of a province that heavily depends on it for revenue.

Fig. 9.3

Promotion expectations and payment schemes

As for the temporal dimension, the two ways in which a sector is “represented” extend across one’s present and past institutional affiliations. That is, while one’s current affiliation always overrides one’s past ones, for reasons to be specified shortly, the sectoral interest for a CC member whose present position has nothing to do with any sectors can still be derived from her historical institutional affiliations. For example, although, institutionally, the former Head of Organization Department of the CPC (Li Yuanchao) was supposed to have no sectoral preferences whatsoever, in fact, his previous work experiences in different departments and provinces (Jiangsu, for instance) are potential sources for gleaning information about his sectoral interests.

1.5 Current Institutional Affiliations

With the general schema explicated above, I continue to explain the concept and measurement issues of each channel individually. The significance of the discussion below comes from the very fact that we need to know whose promotion expectations matter to the policy of interest.

1.5.1 Bureaucratic

Drawing on the long tradition of bureaucratic bargaining in the China filed (Liebertha l 1992; Liebertha l and Oksenberg 1988; Shirk 1993) as well as, despite its exiguity, the literature on China’s trade policymaking (Chen and Feng 2000; He and Yang 1999; Sheng 2002, 2006), my theory posits that, given the fusion of business and government in China’s political economy, the interested bureaucrats of a particular sector include those who currently hold leader positions (at the ministerial level (zhengbuji)) in the sector-specific ministries or bureaus supervising it.

For example, the former minister of railroad, Liu Zhijun, was a clear case of having a sectoral interest in industries related to railroad. Moreover, Liu had also been a member on the Central Committee since 2002 and therefore enjoyed political agency according to the idea of reciprocal accountability. While Liu’s case is fairly straightforward, in actuality, the cases that allow such an easy interpretation are fairly few in number because most sectoral supervisors are not even CC members. Among the CC members elected by the 17th CPC National Congress, for instance, only 19 out of 204 had direct institutional affiliations with sector-specific departments. In other words, from the perspective of bureaucratic bargaining alone, sectoral interests are under-represented within the Central Committee.

As a consequence, if we follow this view strictly, one paradoxical result is that there isn’t really too much agency problem in China’s policymaking process since most sectoral representatives are simply excluded from the process and do not have any power to bargain at all. Nonetheless, this view has a blind spot that results in the under-estimation of agency problem in China’s policymaking process. While it is certainly true that the chief of the Bureau of Light Industry under the Ministry of Industry and Information Technology does not enjoy much political leverage against the policymakers in the Politburo, it does not mean the light industry has no other sources of political agency at all. As our following analysis is going to show, a sector can also gain its political agency from provincial leaders and CC members’ historical affiliations.

1.5.2 Provincial

In fact, in addition to the bureaucracy at the central level, provincial leaders who are mostly CC members also exert great influence on how national policies are made (Bo 2002; Cheung et al. 1998). To incorporate this feature, there are two major issues. First of all, how do we define “leaders” of a province? As far as one’s current³⁵ institutional affiliation is concerned, a leader of a province unequivocally refers to either its Governor or Party Secretary, since it’s nearly unlikely for provincial government officials with administrative ranks lower than ministerial (zhengbuji) or, equivalently, provincial (zhengshengji) level to be appointed as CC members.

Second, the more critical issue is how to define provincial leaders’ default sectoral interests. Regarding this, there are multiple economic indicators from which their sectoral interests could be derived such as number of employees, ownership structure, and tax payables. Among these indicators, this study chooses tax payables since this indicator reflects most directly how much a sector contributes to a provincial government’s revenues. In other words, the leaders of a particular province will side with the sector that makes the greatest contribution to the government revenue.

1.6 Historical Institutional Affiliations and the Formation of Sectoral Coalitions

1.6.1 Theoretical Argument

Besides central bureaucrats and provincial leaders, there is another subset within the group of CC members that can be a source for sectoral political agency. While their present institutional affiliations have no bearing on any sectors, they nonetheless had work experiences in either sector-specific departments or provinces in the past.

The real effect of one’s past institutional affiliations consists instead in the lasting power of sector-related networks even after one has left her former positions. That is, after one has been removed from those positions, the benefits attached to them will disappear, but the networks established previously remain and make one a potential supporter of the sector one used to be affiliated with when she is lobbied by those who have a stake in the same sector at present. In other words, we have to take into account the possibility that a lobbying coalition for a certain sector could be formed among CC members who have and used to have institutional affiliations with it.

Suppose now CC member $i$’s previous sectoral interest coincides with member $j$’s current interest owing to their past and present institutional affiliations respectively. The formation of such an ad hoc lobbying coalition between them is based on the two aforementioned effects. First of all, if both members $i$ and $j$ were colleagues in the past, then it is natural for $j$ to take advantage of her previous connection with $i$ to form an ad hoc coalition to lobby for sector-specific favors. This kind of relational effect is exactly what the work experience tries to capture. Second, if members $i$ and $j$ have no common work history in the past, a theoretical possibility arises from the mechanism of indirect linkage demonstrated in Fig. 9.3. That is, in spite of their lack of direct interactions in this case, the trust needed for such a lobbying coalition to take shape can still be built up by a third party whom both $i$ and $j$ are acquainted with.

The theoretical point made here can be substantiated by a recent controversy over local governments’ liaison offices in Beijing (zhujingban). In 2010, rumor had it that the State Council decided to close all local governments’ liaison offices in Beijing in order to curb various corrupt practices arising from local officials’ efforts to affect national policies through illegal means. As a popular pun in China, paobu qianjin (running into ministries to grab money), goes, to snatch more resources for their jurisdictions,³⁶ local officials exploit various common networks they share with the bureaucrats in charge of resource allocation in the central government. Likewise, many foreign business associations in China (e.g., the American Chamber of Commerce (AmCham) and the US Information Technology Office (USITO)) also hire people who used to work in the government to deal with the government relationship for them. For those CC members whose current institutional affiliations imply no sectoral biases, the networks developed from their institutional affiliations in the past give rise to alternative channels of rent extraction, which incentivize them to form lobbying coalitions with other interested CC members.