2.1 ONE-PARTY REGIME AND THE REGIME PARTY

One-party autocracies are in general more stable than other types of authoritarian regimes (Geddes Reference Geddes1999; Magaloni and Kricheli Reference Magaloni and Kricheli2010). The key mechanism, as Magaloni (Reference Magaloni2008) argues, is that authoritarian parties facilitate credible power sharing among the ruling elites. Authoritarian rulers, enjoying unchecked power, are unable to credibly commit to not abuse other members of the ruling coalition. Fearing the possibility of being exploited by the incumbent dictator, members of the ruling coalition tend to invest their loyalty in the subversion, not the survival, of the dictator. This commitment problem is mitigated when the incumbent dictator delegates his power of selecting ruling elite members to a parallel political organization, namely a political party, so that the ruling elites can also enjoy a share of power and spoils of office and without worrying too much about the changing whims of the ruler. Moreover, multi-party election also provides a potential exit option for the aggrieved ruling coalition members. Although most of these elections are restrictive and manipulated, defection of the ruling elites from the party would undermine the party's “image of invincibility (i.e., winning supermajority in the election to signal to the masses that the party is the “only game in town”),” and therefore constitutes a credible threat to the incumbent ruler who intends to exploit other elites. Cuauhtémoc Cárdenas’ defection from Mexico's long-time ruling Institutional Revolutionary Party (PRI) in 1987—just one year before the presidential election—was a vivid real-life demonstration of such a mechanism.

The theory of credible power sharing through an authoritarian party assumes the following conditions. The authoritarian party operates in a rather independent manner from the incumbent executive. It exists as a political organization in which career advancement within its hierarchies is sufficient to ensure the long-term loyalty of the rank-and-file. It also assumes that powerful elites have the option to defect from the party, and once they have left the party they can still maintain their political influence and impose a credible threat to the incumbent ruler (e.g., through multi-party elections). While these assumptions are reasonable and can explain a wide range of variance across the world of one-party regimes, there is a small, yet important group of one-party regimes which are not consistent with these assumptions.

In these regimes, the boundary between the incumbent executive and the political party is often blurred or even does not exist. Government positions are made into a part of the party hierarchy, and the ranks within the party determine access to government resources. The incumbent rulers concurrently hold the position of party leaders, and control the distribution of political offices both within the government and the party. It is unlikely for an elite to escape from the influence of the ruler by situating himself only within the party. More crucially, no mass elections—even nominal ones—are held in these regimes, and only the ruling party can legitimately appoint political offices. Under such circumstances, the ruling party is truly “the only game in town” and there are no exit options for elites who are facing the risk of being exploited by the incumbent. Not only are there no popular elections in which rival elites could potentially form (or join) an opposition party and challenge the incumbent rulers, publically defecting from the ruling party is a serious crime in these regimes. The ruling party itself is equivalent to the regime, and there is no independent political sphere (either in government or in opposition parties) beyond the organization of the ruling party.

The Chinese Communist Party (CCP) is a typical example of such “regime party.”Footnote ¹ The formation of CCP far predated that of the People's Republic of China. The party's leader is also the paramount leader of the regime. The Party rules the country with a nomenklatura system through which the Party not only makes government posts a part of the party hierachy, but also extends its reach into various public and private organizations (Burns Reference Burns1994; McGregor 2009). No competive, popular election exits in China, and the non-direct, carefully orchestrated elections of the people's congress delegates do nothing more than approving the candidates handpicked by the party committees at different levels (Manion Reference Manion2014; Truex Reference Truex2014). There are a handful, state-endorsed “democratic parties” under the close tutulage of CCP, serving as window dressings for the “CCP led multi-party collaboration” stipulated in the Chinese constitution. Forming independent political parties is outlawed in China, however, and unlike many other one-party states, the CCP even does not allow publically observable factions within the party, through which the elites could have organized collective actions against the ruler.Footnote ² Despite all these inconsistencies with the conditions under which power sharing is made possible in one-party regimes, CCP has steadily ruled China in the past several decades, and has achieved an impressive record of economic growth (Nathan Reference Nathan2003). Why and how is credible power sharing possible within the CCP?

2.2 TERM LIMITS AND POWER SHARING IN A REGIME PARTY

Term limits that effectively govern the leadership transitions serve as an important alternative institution that facilitates power sharing in regime parties like CCP (Svolik Reference Svolik2012, 2013). Studies of power sharing within one-party regime has recognized the role of institutionalized leadership turnovers (e.g., Magaloni Reference Magaloni2008), but have had less to say about how and under what conditions term limits matter. I define term limits explicitly in two dimensions, based on the features of CCP's leadership turnovers in the past decades. First, a well-established term limit system is one in which the leadership is subject to peaceful turnover following a fixed, periodic schedule. Second, the successor to the incumbent leader should be announced well before the turnover takes place.Footnote ³ Leadership turnovers lacking either of the two features are not sufficient to facilitate credible power sharing.

The term limits institution facilitates power sharing by mitigating the commitment problem between the ruler and the elites. Dictators worry about the security of their life and assets once they leave office (e.g., Debs Reference Debs2010; Escribà-Folch Reference Escribà-Folch2013; Albertus and Menaldo Reference Albertus and Menaldo2012), and such concern becomes more salient if they face a fixed schedule for stepping down. The term limit places the incumbent ruler who intends to exploit other elites under the shadow of future retaliation once his term ends. Elites who face the risk of being expropriated, on the other hand, can invest their loyalty and resources preemptively in the heir apparent who is capable of delivering such post-tenure punishment. These elites, meanwhile, will also enjoy a share of the power as a reward to their loyalty when the new leader assumes his office, and the political power these elites gain under the new leader could also impose a threat to the outgoing leader. A pre-appointed successor, combining with the fact that the ruler will have to step down in a future point, provides the elites an “exit option” from the incumbent ruler alternative to defection from the party. The threat of post-tenure punishment makes the ruler's commitment to share power with other elites more credible.

Not a single CCP paramount leader in the post-Mao era has been subject to post-tenure punishment so far. The absence of such punishment, however, does not mean that the institution is not working. In many cases, we see few consequences associated with deviations from the institutions, precisely because the relevant actors are behaving in accordance with the incentive structure induced by the institutions. Had the rulers (say Jiang Zemin or Hu Jintao) not been subject to term limits, they might have acted very differently from how they actually did. The recent fall of Zhou Yongkang also suggests that post-tenure punishment of top leaders is not an empty threat. Zhou is accused of “undermining party's solidarity and engaging in political activities not approved by the party” during his tenure as China's domestic security chief.Footnote ⁴ His continued support of Bo Xilai—his political ally—even when the party decided to take Bo down, caused the perhaps the most severe rift among the party's leadership since the Tiananmen Square incident. Although Zhou's activities raised concerns among other members of the party leadership and led to countermeasures even when he was still in power,Footnote ⁵ he was not formally arrested until two years after his retirement from the Politburo Standing Committee.Footnote ⁶

The term limits institution also facilitates power sharing by alleviating the monitoring problem faced by the elites. The secretive nature of authoritarian politics gives the ruler certain advantages in exploiting the elites: the elites often do not know for certain whether the ruler is transgressing the power-sharing agreement, and the ambiguity caused by such information scarcity incapacitates the elites from organizing successful collective actions against the ruler (Svolik Reference Svolik2012; Boix and Svolik Reference Boix and Svolik2013). A fixed term and a pre-appointed successor place the ruler under a situation in which whether he complies with the institution is easily observable. A dictator's attempt to overstay in office would immediately send a clear signal to the elites that is not committed to power sharing. The heir apparent, who has the single largest share of interest in the succession, also provides a focal point around which the elites could coordinate collective oppositions against the ruler. This becomes particularly crucial if there is no other alternative mechanism, such as a public and organized faction, by which the elites could have devised collective actions. Imagine during the CCP's 16th Party Congress, if Jiang Zemin was not subject to a two-term limits as the head of the state and if the party did not already have a well-prepared successor (Hu Jintao), it would have been much more difficult for Jiang's opponent to organize collective effort to demand that he step down (Miller Reference Miller, Lin and Shirk2002; Shirk Reference Shirk, Lin and Shirk2002). Jiang's decision to stay as the chairman of the Central Military Commission in the same congress also sent a clear signal of his intention to overstay in power, making himself a target of criticisms both within and outside the party, which eventually forced him to resign completely two years later (Mulvenon Reference Mulvenon2005).

Term limits that effectively bind top leaders are rarely seen among dictatorships, however. The successful institutionalization of leadership turnover in the CCP should be attributed to several historical and political factors. First, the resignation of revolutionary veterans in the early 1980s set up a norm of retirement among the CCP top leaders (Manion Reference Manion1993). This norm was later reinforced through the introduction of the age-based mandatory retirement (Nathan Reference Nathan2003; Huang Reference Huang and Li2008; Shirk Reference Shirk2012). Second, the 1982 constitution of the CCP regime—the one currently in effect—stipulates a two-term limit for major political positions of the state, including the president, the prime minister, and the chairman of the National People's Congress Standing Committee. Since 1992, the paramount leaders of the CCP regime have been concurrently holding the position as the President of the state (Huang Reference Huang and Li2008), and are forced to relinquish other powerful positions (most notably, the General Party Secretary of the Central Committee) when his second term as the President ends (Svolik Reference Svolik2012, Chapter 4).Footnote ⁷

More importantly, the distribution of power was not highly skewed towards the incumbent leader even before the term limits were fully adopted. Deng, despite being a powerful leader, used his personal charisma to push forward a system that prevents over-concentration of party power in individuals. He reinforced the norm of retirement by voluntarily resigning from the chairmanship of the Central Military Commission. He also emphasized collective decision-making within party's top leadership, particularly in the Politburo Standing Committee, in order to weaken the role of individual leaders (Deng Reference Deng1993). He handpicked Hu as Jiang's successor and placed a number of his protégés (e.g., Zhu Rongji, Qiao Shi) into the Politburo Standing Committee to further keep Jiang's power in check. As a consequence, Jiang—Deng's chosen successor, did not enjoy the same level of personal authority as Mao and Deng, and was only “first among equals” (Svolik Reference Svolik2012, 8) during his tenure as the party's paramount leader. Such a condition is important, as the first leader whose tenure was governed by the term limits—Jiang in the case of CCP—would only comply with the institution when his personal power was sufficiently confined.

Once the term limit is in place (i.e., when a leader is expected to step down in accordance with term limits), it generates expectations about the leadership turnover, and such expectations lead to an evolving bargain over allocation of power among multiple generations of leaders who in turn keep the ruler's power in check.Footnote ⁸ A leader who intends to overstay in office will face opposition not only from the heir apparent, but also from those aspiring elites who have preemptively invested their loyalty in the incoming leader (Svolik Reference Svolik2013).Footnote ⁹ An incumbent leader who is about to leave office also has the incentive to appoint close associates to powerful positions to ensure the hands of the succeeding ruler will be sufficiently tied. Jiang Zemin, for example, publicly announced the promotion of Jia Qinglin and Huang Jü several weeks before he handed over the CCP general secretaryship to Hu Jintao in November 2002. Jia and Huang, along with Jiang's other allies, significantly confined the influence of Hu Jintao within the Politburo Standing Committee and allowed Jiang, who already retired, to continue to exert influence on important decisions (Fewsmith Reference Fewsmith2003; Miller Reference Miller2004, Reference Miller2011). Similar to what Jiang did, immediately before his retirement in November 2012, Hu Jintao promoted two generals into the Central Military Commission as the Vice Chairmen, a move considered to limit the influence of Xi Jinping—Hu's successor—within the military (Mulvenon Reference Mulvenon2013). The incoming leader, knowing that his influence would be limited by his predecessor, will also try to secure important positions for his followers beforehand, so that he has a legion of elites that he can rely on to carry out his policies when he assumes the office.Footnote ¹⁰

As noted in the introduction, the theory allows us to derive two testable hypotheses regarding elite interactions under the constraints of term limits. The first hypothesis predicts that there is a rather equal distribution of power among the ruling elites when the dictator's tenure is governed by term limits. The second hypothesis predicts that the effective adoption of term limits would initiate a series of actions by top-level elites in anticipation of future leadership turnover—such as active promotion of factional influence among multiple generations of leaders—that in turn keep the power of the incumbent ruler in check.

I test these hypotheses by exploring an important aspect of elite politics within CCP—how the opportunities to appoint close followers into higher ranks are distributed among the top party leaders. Specifically, I look at the selection of CCP's Alternate Central Committee members into the Central Committee between 1982 and 2012. The appointment of top-level cadres associated with the party leaders—especially those at the central committee level and above—is one of the very few institutionalized channels through which we can observe the allocation of power among top party leaders (e.g., Shih, Shan, and Liu Reference Shih, Shan and Liu2010). The promotions of elites into the Central Committee, which take place concurrently with leadership turnover during the National Party Congress, also provide a rare opportunity to examine whether the actions of the party leaders are consistent with the theorized manners. For this particular empirical exercise, we should expect a relatively equal distribution of the chances of promoting close associates between the incumbent paramount leader and his rivals during the entire period of observation, even before the term limits was fully effective (16th party congress). We also expect top leaders to exert more effort in promoting their followers as a result of power bargaining among multiple generations of leaders (i.e., “past, present, and future general secretaries (Shirk Reference Shirk2012)”) after the term limits was fully adopted. In the next section, I provide background information on the subject of the empirical analysis—CCP's Central Committee and its Alternate Committee.

4.1 DATA AND METHOD

I use an original dataset of CCP central committee members (Lu and Ma Reference Lu and Ma2015), which contains detailed biographical and career information of 1973 ACC and CC members from the 1st party congress (1921) to the 18th (2012).Footnote ¹¹ For this project, I use those observations that entered ACC in any of the seven party congresses during China's reform era (1982–2012), among which I delete a handful that were already ACC members before this period but entered the sample as a result of reelection. I also do not include those who were newly elected ACC members in the 18th party congress, since they have not been exposed to a period of risk for promotion.

The resulting sample contains 631 ACC members. Among these members, sixty two are female (9.8%), seventy seven are ethnic minorities (12.2%), 370 have college degrees (58.6%), 200 have graduate degrees (31.7%), and sixteen are so called princelings (i.e., whose parents are cadres at or above the ministerial level). The mean age of first-time ACC members is 52.8, and the average party experience (i.e., time since joining the Communist Party) for the first-time ACC members is 23.3 years.

The mean length that these ACC members have stayed as observations is 1.38 party congresses (the congress is held every five year). 265 of them (42%) eventually experienced the event of interest (i.e., promotion to CC membership), and the remaining 58% are right censored. For those who were eventually promoted, the mean time to promotion is 1.24 congresses, and for those right-censored observations, the mean time they have stayed as ACC members is 1.48 congresses. Figure 1 provides a visual display of the data. The left bands of each column represent the cumulative percentages of ACC members being promoted into the CC in each of the congress since they were first elected, and the bands at the right represents the cumulative percentages of those who were not re-elected as ACC members. The “residuals” are the ones who are re-elected as ACC members. The two lines should eventually add up to 100%. We see that for ACC elected in each congress, no more than 50% of them eventually got into CC, and such pattern remains relatively stable during the entire observational period.

FIGURE 1 Visualizing the Dependent Variable (promotion and dropout in each congress, in %)

I organize the data into the person-congress format, which allows me to accommodate time-varying covariates. The unit of analysis is ACC member-congress in which he experienced the risk of promotion. For example, if person A is elected as an ACC member in the 12th congress, reelected as the ACC member in the 13th and 14th, and promoted into the CC in the 15th, the observations would be A-13, A-14, and A-15. I code the observation as 1 if the ACC was promoted into the CC, and 0 otherwise. Because I am interested in the effect of covariates, I apply piece-wise logistic survival model to analyze the data. The point estimates can be interpreted as the marginal effect of the covariates on the odds of the outcome—promotion into CC in each congress.

4.2 COVARIATES

My key covariates measure official's connections with the top leaders. For the theoretical purposes discussed earlier, I divided connections into two categories: connection with the incumbent paramount leader, and connection with the main political rivals of the incumbent—the past and incoming general party secretaries (for example, Jiang Zemin and Xi Jinping are rivals to Hu Jintao at the 18th Party Congress).Footnote ¹² It is worth noting that we shall not take the word “rivals” in its literal sense, that the rival leaders necessarily have competing political objectives with the incumbent. Instead, as stated in my argument, the notion of “rivalry” tries to capture the political tension created in anticipation of leadership turnovers under the term limits. The” rivals” (the retired and incoming leaders) are the ones who have the genuine interest in and are also most capable of subjecting the incumbent to effective checks and balances.Footnote ¹³ Because of such categorization, the status of an ACC member's connection might change over time. For example, an official's connection to Jiang Zemin qualifies “connection to the incumbent” at 16th Party Congress (2002), but since Jiang stepped down after the 16th Party Congress, this connection changes to “connections to rival leaders” in the 17th and 18th Party Congresses. Later when I limit my analysis to observations in the 16th and 18th party congress—when the two peaceful leadership transitions took place--I further divide the rival leader category into past and incoming leaders.

Having known the classification of connections, a practical question then arises: how do we measure an official's connection with the top leaders? Like many other authoritarian regimes, elites politics in China is a black box, and personal relations between elite politicians are seldom reported publically. There are also no open and observable political factions within CCP like those in LDP or PRI that can provide a basis for coding elite connections. Arbitrary coding of connections based on already well-known patron-client relations (e.g., Cheng Liangyu, the former Party Secretary of Shanghai and an ACC member of the 15th Congress, is known to be a close associate of Jiang Zemin) would result in serious selection issues: we can only handpick those we know, but would miss a large number of potential connections we do not know. Estimation based on such selected sample would produce seemingly desirable yet largely biased results.

The probabilistic approach suggested by Shih, Shan, and Liu (Reference Shih, Shan and Liu2010) and Shih, Adolph, and Liu (Reference Shih, Adolph and Liu2012) provides a reasonable solution to this problem. It measures an official's connection with a top leader based on the following criteria: whether the two were born in the same province, graduated from the same institution, or worked in a same bureaucratic agency within two ranks at a given period of time. Although officials who have shared birthplace, education or workplace experience might not necessarily be connected, such probabilistic measure mitigates concerns for selection bias. I therefore follow this strategy and code ACC official's connection with the top leaders.

This approach is not without pitfalls, of course. Because China is a fairly large and populous country, being born in the same province or graduating from the same university might not be very informative criteria to determine connections. Following the suggestions in recent work that explore more accurate measures of factional ties (e.g., Keller Reference Keller2015 and article in this issue; Meyer, Shih, and Lee Reference Meyer, Shih and Lee2015), I pay additional attention to shared workplace experience between officials. I create two additional variables that are based on whether the official has a shared workplace experience with the top leaders (incumbent and rival), and whether the official shares 10% or more of his/her career with the top leaders. The rationale for the latter is that since top CCP cadres are career bureaucrats, they rarely stay in the same position or locale for an extended period of time,Footnote ¹⁴ and a 10% or more shared work experience (or more than three or four years) suggests that the two have worked together for an extended period of time, or might have worked as colleagues in multiple places—which is a strong indicator that the two are closely connected.Footnote ¹⁵ Results based on each of the three connection measures will be presented and compared in the estimation section.

The first control variable is the average provincial fiscal revenue relative growth during the four-year interim when the congress is not held as proxies for performance.Footnote ¹⁶ Fiscal revenue data has been considered a more credible indicator of local economic performance than the noisy GDP figures reported by local officials (Lü and Landry Reference Lü and Landry2014).Footnote ¹⁷ Following Shih, Adolph, and Liu (Reference Shih, Adolph and Liu2012), the relative growth is calculated by subtracting rest-of-China growth from each province's growth rate. Intuitively, this measure reflects how better (or worse) a province's revenue performance is relative to the average of all other provinces. It has several advantages. First, it allows the inclusion of ACC members who did not have provincial administrator experience, such as central government bureaucrats, military officials, university administrators, or star athletes into the analysis. Because provincial governor and party secretary account for less than 20% of all ACC members, using absolute growth data will result in list-wise deletion of more than 80% of the observations. These observations are assigned a score of zero under the relative growth calculation, assuming they are doing neither better nor worse than their peers in terms of generating revenue growth. Second, it makes comparison of growth rates over time feasible. The substantive meaning of absolute growth might change over time. For example, a 10% provincial revenue growth rate might have been only mediocre in the early 1990s, but it would be quite impressive in the late 2000s. The relative growth measure assumes the promotion is a zero-sum game (Shih, Adolph, and Liu Reference Shih, Adolph and Liu2012), and mitigates the inconsistency problem of absolute growth rate over time.

The second set of control variables is a series of personal characteristics. They include ACC member's gender (1 if female), ethnicity (1 if ethnic minority), princeling status (1 if parents are cadres at the ministerial level or above), and education attainment (two dummy variables for college and graduate degree). I also include a dummy variable for whether the official is above the age of 60 at the time of congress, since the majority of the ACC members are at deputy ministerial/governor level or below, and the mandatory retirement age for officials at this level is 60.

The last covariate is the ACC voting rank. In each party congress, ACC members are ranked based on the number of votes they received during the election in National Party Congress. Extant studies suggest the voting rank being an indicator of ACC member's latent political strength (e.g., Shih, Adolph, and Liu Reference Shih, Adolph and Liu2012), but whether the voting rank would affect the prospect of an ACC member's promotion into CC remains unclear. I take a one-congress lag of the ACC rank, in the spirit of using the voting rank one receives in the last congress to predict the probability of promotion in the current congress. All ACC ranks are standardized into percentiles (a score ranging from 0 to 1, with 1 denoting rank at the last in each congress).

5.1 POOLED SAMPLE ANALYSIS

I first estimate the result using pooled dataset for the entire observational period. Table 1 (in Online Appendix) presents models with the measures for connections only. In model 1 I use the original probabilistic measure for connection in Shih, Adolph, and Liu (Reference Shih, Adolph and Liu2012) with three components (shared birthplace, education institution, and workplace experience). In model 2 I use a more conservative measure that looks only at job overlaps, and in model 3 I further constrain the criteria by looking at whether an official shares 10% or more of his career with the top leaders. In all of the three models, connections with the incumbent and rival leaders have positive and statistically significant coefficients. Because the coefficients from logistic model are notoriously hard to interpret, following King, Tomz and Wittenberg (Reference King, Tomz and Wittenberg2000), I present a visual display of the covariates effect in Figure 2. Connection with the incumbent leader increases one's chance of promotion by about 12% to 15%, whereas connection with the rival leaders increases the chance by 7% to 9%.

FIGURE 2 Comparison Between Three Coding Strategies of Connections (Without Other Covariates)

I now turn to examine if the associations observed above are sensitive to potential confounders. The comparison of the effects of three measures after including the complete set of control variables is presented in Figure 3. It is evident that the positive effects of incumbent and rival connections remain largely similar as to the ones presented in the baseline models.Footnote ¹⁸ Because we are using the pooled data, the results lend strong support to the first hypothesis, that the opportunities to promote followers are distributed rather equally between the incumbent and the rival under the term limits. Neither the incumbent nor the rivals enjoy a decisive advantage (or disadvantage) in promoting their close associates during the entire observational period.

FIGURE 3 Comparison Between Three Coding Strategies of Connections (With All Covariates)

5.2. SPLIT SAMPLE ANALYSIS

As noted in the theory section, the expectation of future leadership turnover under the established term limits institution will stir a series of elite actions—i.e., active promotion of factional influence—that in turn keep any one faction from dominating the others. This leads to the second hypothesis: that we should observe the party leaders exerting more effort to promote their followers following the establishment of the term limits. Although several important elements of the term limits institution already emerged in the 1980s and 1990s (e.g., retirement of revolutionary veterans in 1980s, transition of leadership from Deng to Jiang in early 1990s, selection of Hu Jintao as the heir-apparent in the late 1990s), the term limits were not fully adopted until 2002—when Jiang stepped down as the general party secretary in the 16th party congress. Our analyses so far used the pooled data during the entire post-Mao period, which might have lumped together the dynamics of promotion under various degrees of institutional constraint. To explore such dynamics, I create two sub-samples of the data. The first sample includes only those observations since the 16th party congress (2002). The second sample includes those observations until the 15th party congress (1997). I run the full model with all control variables on the two samples and present the estimated marginal effects of the key covariates in Figures 4 and 5 (See Tables 2 and 3 in Online Appendix for the complete regression results). We need to bear in mind that the number of observations in each sub-sample is only half of that of the full sample. We should expect the degree of uncertainty (i.e., standard errors) to increase as the sample size reduces.

FIGURE 4 Comparison Between Three Coding Strategies of Connections (With All Covariates, Since 2002)

FIGURE 5 Comparison Between Three Coding Strategies of Connections (With All Covariates, until 1997)

Despite of having a much smaller sample size, the effects of the two key covariates still remain positive and statistically significant across all three different measures for the period since 2002 (Figure 4). Moreover, the connection to the rival leader increases one's chance of promotion by about 10% to 16% during this period, which is larger than the effect we observe in the aggregated sample (7%–9%).

In contrast, the marginal effects of connection with either the incumbent leader or the rivals, although both are positive, are not statistically significant for the period until the 15th party congress (Figure 5). There are several interpretations for this. First, as noted earlier, the increased uncertainty might just be a result of the reduced sample size. Second, as term limits were not fully adopted during this period, top-level leaders were not subject to the expectation of imminent leadership turnover, and thus were less engaged in promoting their followers to place other leaders under check and balance. Third, several members in the rival category in this period (Hu Yaobang in the 13th party congress, Zhao Ziyang in the 14th party congress) did not fit perfectly with the account of power sharing. Although both Hu and Zhao had held the position of general secretary, neither were paramount leaders. In addition, they did not leave office as a result of the binding term limits, but were instead forcefully removed from office due to their failure in the party's internal struggle, which suggests that their ability to promote factional members after they stepped down was very much limited. This also explains why the effect of connection with rivals is smaller than that of the incumbent during this period. However, the positive sign of the coefficients indicates that the rivals and their factions were not completely precluded from promoting their followers during this period, suggesting some informal elements of the term limits might already at work during this period.

Finally, I further restrict my analysis to observations in the 16th and 18th Party congress, when the two peaceful leadership turnovers (Jiang to Hu, and Hu to Xi, respectively) took place under the current term limits institution. I am able to code top leaders into three categories for these two congresses: the incumbent paramount leader who is about to step down (Jiang Zemin, Hu Jintao), the heir apparent who is about to succeed the office (Hu Jintao, Xi Jinping), and the past/retired paramount leader (Deng Xiaoping, Jiang Zemin). As discussed in the theory section, we should expect more intensive bargains over allocation of power among the present, past and future paramount leaders during the time of transition. I run the full model with all the control variables, and present the marginal effects of the three key covariates in Figure 6 (for full regression result, see Table 4 in Online Appendix). The effects of connections with the outgoing incumbent leader and incoming leader are both positive and statistically significant across all three measures. The substantive effects (around 25% to 35%) are much larger than the one we observe from the aggregated data. The effect of connection with retired leader is not statistically significant for the most broadly defined measure, but is significant for the two stricter measures. This uncertainty might be explained by a potential “cohort effect” (Meyer, Shih, and Lee Reference Meyer, Shih and Lee2015): as the influence of the retired (past) leader's factions declines over time, they tend to focus on promoting a handful of core factional members, instead of trying to send anyone that is vaguely connected into higher ranks.

FIGURE 6 The Two Transitioning Congresses (16th and 18th)

Results from these split sample analyses further shed lights on our understanding of the relationship between term limits and power sharing within the CCP. The evidence suggests that there was already a substantial degree of power sharing among the Party's top leadership before term limits were fully adopted. Neither the incumbent leader nor his rivals enjoyed a systematic edge over each other in promoting their close followers during this period. This is an important initial condition, as Svolik (Reference Svolik2013, 9) points out, that the effective adoption of the term limits relies on the even balance of power among the ruling elites. The first leader whose tenure is governed by term limits (Jiang Zemin in the case of CCP) would only comply with the institution when there is a relatively even balance of power among the elites. Once adopted, the term limits initiates expectation of leadership turnover among the elites that in turn enhances power sharing. Both the incumbent ruler and his rivals have clearly exerted more effort in promoting their close followers in the two transitioning congresses to ensure the hands of the future party leader would be tied. Patronage-based promotion, in this case, not only helps to advance leader's own factional interest, but also serves to place leaders under the checks by each other.

5.3. OTHER INDIVIDUAL ATTRIBUTES

Although the control variables are not the primary focus of this study, it is worth discussing their effects to better our understanding on the selection mechanism within the Central Committee. I run the model with full specification and present the simulated effects of all covariates in Figure 7 (for full regression result, see Table 5 in Online Appendix).

FIGURE 7 Effects of All Covariates

Two variables that show significantly negative effects are female and ethnic minority. Being a female or ethnic minority ACC member reduces the chance of promotion to CC by almost 20%. This is consistent with the intuition that ACC plays a partial “window dressing” function for the Party.Footnote ¹⁹ Women and ethnic minorities are appointed as ACC members to showcase the Party's commitment in promoting equality along the gender and ethnicity lines. Yet once it comes to the competition into more powerful positions like CC, these members are hugely disadvantaged or are even not considered for promotion at all.

A higher relative fiscal revenue growth does not contribute to one's chance of advancement. Despite ample evidence suggesting that the CCP tends to reward officials who can bring in more rents, especially at the lower tiers of government (Whiting Reference Whiting, Naughton and Yang2004; Lü and Landry Reference Lü and Landry2014), such logic seems to not apply at a higher level. This is consistent with what Lü, Landry, and Duan (Reference Lü, Landry and Duan2014) find, that the ability of an official to claim the credit for better economic performance becomes weaker and weaker as one climbs up the administrative ladder. At a level as high as Alternate Central Committee member, loyalty becomes the dominant concern when determining an official's further advancement into the inner circle (Egorov and Sonin Reference Egorov and Sonin2011; Jia, Kudamatsu, and Seim Reference Jia, Kudamatsu and Seim2014).

One's chance of promotion is reduced significantly if he is above the age of 60 at the time of party congress. This age is also the mandatory retirement age for cadres at the deputy minster/governor level—the ones that constitute the majority of ACC members. The negative and significant effect of age suggests the end of lifelong tenure among the CCP elites during the reform era (Manion Reference Manion1992, 1993). The adoption of age-based retirement ensures important party positions routinely become available, which is critical to “create an enduring stake in the regime's survival (Svolik Reference Svolik2012, 193)” among those rank-and-file with strong career aspirations.

College and post-graduate degrees do not bring advantage to ACC member's promotion, as suggested in Model 6 of Table 2. This result seems counterintuitive, as many other studies have found a positive correlation between political ranks and educational attainment (e.g., Shih, Adolph, and Liu Reference Shih, Adolph and Liu2012). Yet in the case of ACC's promotion into CC, education qualification might work in opposite directions. Higher level in education might indeed accelerate the speed of those career bureaucrats’ promotion into the CC. But for those social elites with very high education qualifications (e.g., intellectuals, entrepreneurs), ACC is perhaps the highest rank they can make, or more precisely, the Party co-opted them into the ACC but with no intention to further promote them into more powerful positions. Zhang Yanzhong, who holds a Ph.D. from Oxford and is one of China's leading aeronautical scientists, was elected an ACC member in the 13th (1987) and 14th (1992) Party congress, but he never made it to the more powerful CC. Zhang Ruimin, CEO of Haier Group and a former master student at Chinese University of Science and Technology, has been elected ACC member for consecutively three congresses (16th, 17th, and 18th). Such examples again suggest the dual function of the ACC. On the one hand, ACC is a career step stone for those who arise within the regime and are potential candidates for the selectorate. On the other hand, ACC performs the role of co-opting influential non-regime elites into the high ranks of the Party, providing them with considerable political status but little substance in power sharing.

Princelings, who have long enjoyed unmatched political and economic privileges in China, do not seem to have a systematic edge in terms of promotion from an ACC member to a full CC member. While the sign of the coefficient suggests a positive effect (see Figure 3), such effect is not salient at the 90% level. One potential explanation for the large uncertainty is perhaps there are so few princelings in the sample (16 out of 631).

ACC voting rank is positively associated with the prospect of promotion at a marginally significant level. Although ACC election is largely orchestrated and follows the order from the top leaders, it does not mean the voting rank reflects accurately the potential strength of each individual ACC member. The central leader's order can safely manipulate who will be selected into the ACC, but is unlikely to control precisely how many votes each ACC member would receive from close to 3000 National Party Congress delegates (Lu and Ma Reference Lu and Ma2015). The relative large uncertainty could also be driven by that fact that several ACC members who were at the bottom of the rank have performed surprisingly well. Xi Jinping, who ranked the last in the 15th ACC, was promoted to the CC in the 16th Party Congress, entered the most powerful Politburo Standing Committee in the 17th Party Congress, and ascended to the position of General Party Secretary—China's paramount leader—in the 18th Party Congress. Jia Tingan, former general party secretary Jiang Zemin's personal associate, ranked the last in the 17th ACC. He was promoted to the CC in the 18th Party Congress.