The micro-foundations of elite politics: conversation networks and elite conflict during China’s reform era

Abstract

In this article, we explore the micro-foundations of elite politics by focusing on changes in network structures that emerge from informal conversations. Empirically, we offer a novel “situational conflict” explanation to account for the puzzle of why reformist leaders were periodically ousted during China’s reform era (1977–1992), emphasizing the unexpected power collision that catalyzed the violent crackdown on the Tiananmen movement in 1989. To do so, we employ network analysis and narrative to utilize an original dataset of elite conversations and primary sources that have only recently been made available. We find that ideological cleavage and manipulative brokerage produced each conflict to varying degrees but were contingent on the relational structure arising from elite conversational interactions. Furthermore, the actual unfolding of those conflicts often resulted from key actors’ discrepant understandings of the changing relationships via ongoing interactions at vital moments, such as during the Tiananmen movement. Integrating micro-sociological theories and network analysis, our work has methodological and theoretical implications for unpacking the black box of elite politics and its role in macro-historical change.

Historical events are often marked by memorable moments. The taking of the Bastille was an iconic incident during the French Revolution: it “invented” the modern revolution (Sewell, 1996). Two hundred years later, the defining moment for the 1989 Tiananmen movement—one of the largest social movements since WWII—was, unfortunately, the bloody massacre of demonstrators by the armed forces on June 4^th. It not only remains a traumatic memory for many Chinese but also left behind an insurmountable obstacle on China’s stumbling pathway towards democracy. The massacre, notably, was catalyzed by a decision of martial law on May 17^th, when popular protests on Tiananmen Square and elite conflicts nearby in Zhongnanhai—the central headquarters of the Chinese Communist Party—coevolved into upsurge.

Such coevolution is not uncommon in historical junctures, from the French Revolution to the Eastern European revolutions (Lachmann, 1997; Markoff, 1997). What makes the Chinese case extraordinary is that the massacre took place under the watch of the entire world, while communist leaders in Eastern Europe were simultaneously responding to protesters more modestly. Erich Honecker, for instance, had prepared for full-scale repression in East Germany similar to the “Chinese solution,” but he eventually conceded to protesters—who later tore down the Berlin Wall (Pfaff, 2006:115–122, 165–167). It is similarly puzzling why Deng Xiaoping, China’s paramount leader, did not take alternative options to pacify the student movement, since the Party General Secretary Zhao Ziyang—Deng’s long-time ally and designated successor—was proposing a concessive approach.¹ Yet instead of taking up Zhao’s suggestion, Deng broke his alliance with Zhao and brutally suppressed the movement. Explaining Deng’s deadly decision is thus tied up with this change in their relationship.

Elite politics, overall, is the politics of kingship and kinship: it takes place in both ceremonial courts and local communities, generating eventful decisions via everyday interactions. To understand how such politics work, we need to uncover not only the relational substrata undergirding formal institutions and ideologies but also how elite interactions continually redefine their relationships (Elias, 1983; Lachmann, 2000; Padgett & Ansell, 1993).

In this article, we argue that the changing network structures present elites with varying positions and dispositions in power conflicts and affect the actual unfolding of these conflicts via their conversational interactions. This explanation is different from two extant accounts of elite conflicts in China’s reform era, including the power collision during the Tiananmen movement. One account regards Chinese politics during this era as a series of ideology-driven conflicts between two rival groups—reformers and conservatives—which culminated in the decisive Tiananmen clash (Dittmer & Wu 1995; Nathan & Tsai 1995; Tsou, 1995; Yang, 2004). The other account instead considers most conflicts caused by a power manipulator in multiplex, crosscutting elite networks: Deng Xiaoping, who utilized his brokerage role to protect his status against his power contenders, including Zhao Ziyang in 1989 (Fewsmith, 1994; Huang, 2000; Padgett, 2012; Su, 2023). We find that ideological cleavage and manipulative brokerage both mattered to varying degrees, contingent upon the changing relationship structure arising from conversational interactions. Specifically, key actors’ discrepant conceptions of the network structure that was being altered by short-term conversations led to their overaction and overreaction, producing the otherwise unexpected collision between Deng and Zhao during the 1989 Tiananmen movement.

Our argument is built upon longitudinal network analysis of an original dataset comprised of the conversation records of top Chinese leaders over sixteen years (1977–1992) as well as on qualitative analysis of this dataset and other primary sources—including diaries, memoirs, autobiographies, and unpublished archives about Chinese leaders. To examine our dataset, we combine network analysis techniques and historical narrative to provide meaningful causal stories. Utilizing recently released primary sources, our work offers new insights to a new wave of historical research on China’s 1980s (Brown, 2021; Weber, 2021; Torigian, 2022; Su, 2023).

Situated at the intersection of micro-sociological theories, network analysis, and historical sociology, our research has significant implications for understanding how elite politics affects historical change (Collins, 1981; Emirbayer & Goodwin, 1994; Erikson & Occhiuto, 2017; Goffman, 1969, 1983; Tilly, 1995). Such micro–macro link is vital, insofar as “the role played by these particular individuals exemplifies the notion of ‘small’ causes yielding ‘big’ effects” (Ermakoff, 2015:66). By now, few empirical works have yet taken up the call to import micro-interactionism into historical sociology; even fewer have combined interactionism with network analysis to study politics (but see: McLean, 2007; Gibson, 2012; Schoots et al., 2020). By analyzing the elite conversation network, our article not only specifies the roles of different conversational brokerages in elite relationship but also fleshes out the micro-foundations—i.e., the relational and interactional foundations—of elite politics in authoritarian contexts.

Our work also makes methodological contributions by extending longitudinal network analysis of an original elite dataset to historical political sociology. Prior studies have used network analysis to examine elite politics in late medieval Italy (Padgett & Ansell, 1993), the English Civil War (Bearman, 1993; Hillmann, 2008), early modern Poland (McLean, 2004, 2011), and the 19^th-century Chilean Congress (Bro, 2023). Our work contributes to this literature by elaborating the effects of changing network structure on elite conflicts over time. We also offer an original approach to political sociology by employing longitudinal network analysis on a behavioral (i.e., conversational ties) rather than a background (i.e., kinship ties) dataset and offering a dynamic measurement of the network structure (also see: Padgett & McLean, 2006).

The paper proceeds as follows: the next section presents the empirical puzzles about elite conflicts in China’s reform era. We then outline three competing models—ideological cleavage, manipulative brokerage, and situational conflict—for explaining these conflicts. After that, we propose conversation network analysis as an analytic approach to substantiate the situational conflict model and specify our data and methods. The three subsequent sections present our empirical analyses: we first employ network analysis to reconstruct the changing relationship structure during this period. We then combine these network characteristics with narrative to explain the four key episodes of elite conflict under study. We further use the 1989 power collision as a critical test case to evaluate the three models. The conclusion summarizes our empirical, methodological, and theoretical contributions.

The puzzle of elite conflicts during China’s reform era

Power conflicts during China’s early reform era (1977–1992) were less violent but no less intense than during the Maoist period. The reform years not only saw factional struggles between reformers and conservatives but also intergenerational conflicts among elders who survived the Cultural Revolution (1966–1976) and younger leaders who rose to prominence during the reform era. Within one decade, three nominal top leaders (Hua Guofeng, Hu Yaobang, and Zhao Ziyang) were removed and a fourth (Jiang Zemin) was at risk. During each of these conflicts, not only was the top leadership reshuffled but many of their associates were also affected. It may not be surprising that Hua Guofeng—Mao’s designated successor—was forced to resign in 1980, but recent studies have found him to be less of a counter-reform conservative than had been previously believed (Torigian, 2022). Hu Yaobang and Zhao Ziyang were both Deng Xiaoping’s chosen successors and reformist allies, but it was Deng who ousted them in 1987 and 1989. Deng’s third chosen successor, Jiang Zemin, was attacked by Deng and almost lost his position in 1992—this time for not being reformist enough. It was only after Jiang’s consolidation of power in 1992 that the open political clashes characterizing the 1980s Chinese political scene seem to have diminished.

Against the backdrop of China’s successful reform, these intense elite struggles are puzzling in two regards. On one hand, scholars wonder why China’s economic reform advanced despite these periodic power struggles (Shirk, 1993; Padgett, 2012; Xiao, 2019; Weber, 2021). On the other hand, it is similarly puzzling that Deng Xiaoping—China’s “chief architect of reform and opening up”—not only played a vital role in those conflicts but also removed his reformist successors and allies from the top leadership. In this article, we focus on the second aspect: why did political conflicts occur (often among reformers) and how did these conflicts unfold? Notably, although Deng was always “victorious” in these conflicts, each of them proceeded in distinct ways: the gradual dethroning of Hua, the easy defeat of Hu, the abrupt removal of Zhao, and Deng’s compromise with Jiang. We hence aim to account for the varying courses of these conflicts in addition to their commonalities.

Of all these conflicts, the abrupt collision between Deng and Zhao during the 1989 Tiananmen movement is of most interest. As mentioned above, the long-time allies split at the May 17^th meeting in which it was decided to instate martial law. Deng rejected Zhao’s conciliatory approach toward the protestors and instead proposed a hardline approach, despite Zhao’s opposition. This relational breakdown was significant: martial law was declared, and Zhao was soon removed and replaced. Yet it was also surprising: before that point Deng had shown Zhao firm support several times. Why then did the long-term allies suddenly turn into opponents, making bloody suppression the only viable option?

Our empirical inquiry has broader implications for understanding the complicated process of elite conflicts. To clarify, we study the elite power struggles surrounding key leadership positions rather than policy disputes, insofar as politics is defined as “striving to share power or striving to influence the distribution of power, either among states or among groups within a state” (Weber, 1958:78). In democracies, such conflicts may be mitigated by electoral mechanisms and organized into competitive parties representing distinct ideologies. In non-democracies, the lack of electoral mechanisms means competition over succession and other vital positions is a constant driver of power struggles (Brownlee, 2007). Unlike policy disputes, such power struggles are about life and death and become the primary issue of elite politics.

Competing explanations

Ideological cleavage

The first model considers power collision the cumulative result of ideological cleavage among elites. There is a longstanding belief that during the early reform era Chinese leaders divided into two rival groups—reformers and conservatives—with nearly fixed ideological orientations and memberships. Deng Xiaoping and Chen Yun, the two most prestigious leaders, each led a faction, and most power struggles occurred between the two rival blocs. Deng’s reformist and young allies, such as Hu Yaobang and Zhao Ziyang, were sacrificed in the factional struggles when Deng sought compromise with his conservative rivals. The student movement only accelerated the division between the two sides, incurring conservative backlash against Hu in 1987 and Zhao in 1989. In sum, this model finds factionalism resulted in periodic political clashes, including the power collision during the 1989 Tiananmen movement (Dittmer & Wu, 1995; Nathan & Tsai, 1995; Yang, 2004).

This account echoes the classical cleavage model in political sociology: politicians are organized into different ideological blocs, factions, coalitions, or parties, and elite conflicts often occur between rival sides when the cleavage intensifies. Furthermore, the elite cleavage reflects larger, societal cleavages, such as between the bourgeoisie and the proletariat or between conservative and liberal forces (Downs, 1957; Lipset & Rokkan, 1967; Marx, 1978). Insurrectionary mobilization and elite conflicts are therefore understood to mutually shape each other in many circumstances (Markoff, 1997:1113; Lachmann, 1997).

While ideological factionalism was a salient phenomenon during the early reform era, this model has problems unpacking the elite conflicts. Above all, the most intense political clashes did not occur between reformers and conservatives but among reformers. Thus, ideological affinity did not prevent antagonism: it was Deng, not conservative leaders, who made the decision to remove Hu and Zhao. Notably, this model cannot explain the 1989 collision: although ideological cleavage had indeed been intensified since the price reform in 1988, during the Tiananmen movement it was Deng who decided to abandon Zhao—even before he was pressured by conservative elders. Even if factional division mattered in this case, there is a gap between such cleavage and the sudden schism between two reformers.

There are also issues in using “ideological cleavage” as the explanatory factor: Above all, rather than two clear-cut ideological factions, leaders were reformist in some fields but conservative in others. Moreover, few politicians can be consistently assigned to either the reformist or conservative faction, and they rarely formed alliances based on ideological orientation alone. For example, when acting as Premier in 1982–83, Zhao Ziyang shared economic views with the purportedly “conservative” elder Chen Yun but was unwilling to discuss economic policies with Hu Yaobang, though both Hu and Zhao were regarded as reformers. Indeed, “elite politics was complex, but it was not polarized” for most of this era (Teiwes & Sun, 2007: xv). Finally, the ideological leanings of key figures are often (mis-)interpreted according to how their careers ended, meaning hindsight biases the unfolding of their actual orientations. We need to find a proper way to measure ideological factionalism to know when and how it mattered in these elite conflicts.

Manipulative brokerage

Another model instead considers elite conflicts are often caused by a manipulator operating in a multiplex power network where ideological orientation is confounded by crosscutting ties. Political clashes do not necessarily result from clearcut ideological cleavage but may be provoked by manipulative brokerage that turns crosscutting cleavages into conflicts. Such a broker is known as a tertius gaudens, who benefits from ongoing conflict among other parties and may even use the Divide et Impera (divide and conquer) strategy to “intentionally produce the conflict in order to gain a dominating position” (Simmel, 1950:162). In studies of early modern European politics this is referred to as the “royal mechanism,” which appeared when strong central authorities arose to balance antagonistic political groups (Elias, 1982: 320, 327–328). Most notably, kings often used their siblings and ministers to mutually control each other in their tripartite division of power (Bourdieu, 2004:22).

This idea was further developed in Padgett and Ansell’s (1993) now classical study of the “robust actor,” such as Cosimo de’ Medici, who controlled and balanced the multiplex network of Renaissance Florence. During China’s reform era, Deng Xiaoping is described as one such power broker (Huang, 2000; Torigian, 2022:3; Su, 2023:19). According to this account, Deng Xiaoping’s strength came from his structural position in the power network and his will and skill in exercising his brokering agency. In Padgett’s vivid account, “Deng was a sphinx, like Cosimo de’ Medici, ambiguous in his policies, in his offices and in his interests, sitting behind (and above) the scenes adjudicating disputes like a judge” (Padgett, 2012:311).² Deng was hence able to advance reform policies while making himself the ultimate authority across factions. When power contenders challenged his authority, Deng used his robust-actor role to pit them against their rivals, sacrificing one or even both sides while keeping himself unscathed (Fewsmith, 1994:245). This explains why Deng’s reformist successors were removed and conservative elders were marginalized, while Deng consolidated power. When this logic is applied to explain the Tiananmen collision and crackdown, Deng is alleged with plotting to oust Zhao Ziyang long before their final break. Given Deng’s military leadership, many observers “assume that a Machiavellian master plan was unfolding with cool precision at every stage of the assault… [and] see sinister plots beneath the surface of every Army action” (Brook, 1998:12).

In taking Deng’s centrality for granted, this model may explain why Deng was always the “winner” of his various struggles, but it cannot account for why it took him varying efforts to secure victory. The problem is rooted precisely in the model’s characterization of Deng’s brokerage. First, when did Deng become a prevailing robust actor and how did he remain so? After all, not every ruler always enjoys predominant advantage by sitting at the center of a “perfect” power network. Deng might have been a robust actor when his authority straddled multiple lines in 1981–89, but this role had not appeared until 1981 and was in decline after he semi-retired in 1987 and fully retired in 1989. As found elsewhere, brokerage may decay when brokers cease to nurture their existing bridges (Burt, 2002). When the robust actor’s brokerage role declines in the relationship structure, they may become locked in by rivals and allies and have diminishing flexibility in handling crises.³ Second, this model downplays the agency of other leaders, but there often exist different kinds of brokerage in any power network (Gould & Fernandez, 1989). In effect, non-central actors “can gain power that exceeds that of the most central actor” (Mizruchi & Potts, 1998:384). Although Deng was one of the most senior leaders, he was neither the oldest PSC member (Marshal Ye) nor the earliest, still living PSC member (Chen Yun). Deng was also never the nominal top leader. Thus, Deng’s authority was at times contested by other elders and young leaders (Guo, 2019:298–299, 302). In short, elite brokerage structures not only take on multiple forms but also change over time. As a result, while some conflicts during the reform era were provoked by a manipulative robust actor, others were induced by his weakening brokerage. A better explanation needs to specify the nature and effects of the changing brokerage.

Situational conflict

We propose a third model that considers social conflicts the results of situational interactions within changing relationship structure.⁴ Our model is built upon previous studies of elite politics and contentious politics regarding relational and processual mechanisms of group division and conflict escalation (McAdam et al., 2001; Slez & Martin, 2007; Tilly, 1995; Zhang, 2021).⁵Compared to the two models above, this model contends that elite relationship structure and their everyday interactions mutually constitute each other and such dynamic interplay is key to explaining elite conflicts during China’s reform era. Our explanation contains two components.

First, the variations in the relationship structure—such as cleavage and brokerage—shaped how each elite conflict unfolded in the 1980s. In politics, ideological cleavage represents only one of many frameworks for political cleavage and is itself subject to endogenous change in elite interactions. Even if some elite conflicts are produced by intensified ideological factionalism, previous alliances may become divided by new issues and drift into new sequences of cleavages and conflicts after common enemies are eliminated. Rather than considering ideological cleavage as given during China’s reform era, we should therefore examine when such cleavage became salient in elite conflicts. Likewise, both brokers and their brokerage roles are continually remade in unfolding interactions, rather than fixed by their structural positions (Bothner et al., 2010; Spiro et al., 2013; Obstfeld et al., 2014; Quintane & Carnabuci, 2016). Thus, even the brokerage of paramount leaders like Deng Xiaoping was fluctuating through interactions with other leaders, presenting Deng varying degrees of advantages in his battles with his three heirs. In addition to the “omnipotent” robust actor, other Chinese leaders exercised a variety of brokerage roles—ranging from coordinator, itinerant, to gatekeeper/representative (Gould, 1989; Gould & Fernandez, 1989:91–94)⁶—which offered them different advantages in the power struggles in 1980s.

Second, while the relationship structure offers elites different positions and dispositions in power conflicts, but how such conflicts unfold is contingent upon their interactions in a shorter time period. As found in micro-sociology, the nature of interpersonal conflict (including its avoidance) is often related to key actors’ discrepant understandings of a changing and ambiguous relationship structure (Fine & Kleinman, 1983:102; Gould, 2003:17). In elite politics, such discrepant conceptions may cause a series of (over)action and (over)reaction and even produce unexpected but decisive collisions. In particular, during the Tiananmen movement the schism between Deng and Zhao resulted from Deng’s reluctant reaction vis-à-vis a sequence of unexpected interactions that gradually altered the relationship structure in April and May 1989.

Table 1 summarizes how the three models explain elite conflicts during the reform era and the 1989 Tiananmen power collision. The “ideological cleavage” model considers periodic political conflicts the results of intensified ideological cleavage between reformers and conservatives. The “manipulative brokerage” model argues that Deng Xiaoping purposively produced clashes to dominate power contenders, including his reformist successors. The “situational conflict” model regards elite conflicts as the result of situational interactions among elites in a changing and differently perceived relationship structure. Rather than refuting the two other models, it specifies under what conditions factionalism or brokerage mattered.

Table 1 The conceptual frameworks of the three models in explaining elite conflicts

The analytic approach: conversation networks

There are two analytic challenges to substantiating the “situational conflict” model: how to measure relationship structure in dynamic terms and how to demonstrate that relational characteristics turn into actual conflicts. In response to these challenges, we employ conversation networks as a way of both quantitatively measuring the changing relationship structure and qualitatively representing actual elite interactions. Our analytic approach is built upon an enduring effort to bridge network analysis and interactionism, despite some ontological and methodological gaps.⁷ This strain of research focuses on “the dynamic construction and deconstruction of network relations through temporally unfolding processes of talk and interaction” (Mische, 2011). Notable examples include the correspondence network that shaped the patron-client relationships of Italian Renaissance elites (McLean, 2007).

Our basic analytic units are elite conversations—the most common and important type of interactions in elite politics—since “the basic element of politics is, quite simply, talk” (Hall, 1972:52). Elite conversations are thus a conduit for information sharing and decision making and a vehicle for producing historical change (Collins, 1981:998). For powerful people, conversation is power: their discussions are local but have far-reaching influence. Though this observation is universal, personal conversations are especially vital in regimes where informal authority trumps formal institutions. In brief, elite conversations affect politics by (re)structuring relationships. First, private and small-group conversations reproduce and reinforce elite relationships, insofar as “‘talk’ is an interactional vessel for making social connections” (McLean, 2007:39). Allies or friends generally tend to have more informal conversations to exchange ideas, build consensus, and strengthen group identity (Bearman & Parigi, 2004). Second, since “roles are not ‘givens’ that constrain interaction, but are something that actors must acquire through interaction” (Leifer, 1988:865), the nature and strength of interpersonal relationships are often expressed by and interpreted in conversations rather than the other way around. Third, pivotal conversations may cause relational disruption or even relational breakdown (Tavory & Fine, 2020).

Elite conversation networks can thus be constructed as an effective means to measure their relationship structure over time. Above all, if static background ties (e.g., kinship, marriage, friendship, work overlap, etc.) are indicators of durable relational roles, dynamic behavioral ties such as conversational ties are better measures of shifting relationships (Erikson, 2013). More specifically, private conversational ties are good indicators of elite relationship strength, while the frequency of conversational topics demonstrates sphere of influence. In addition, the structure of conversational ties reflects elite relationship structures, such as community and cleavage/polarization. Communities are close and recurrent interaction groups, and polarization or cleavage appears when two conversational communities become almost disconnected.

The structure of conversational ties also represents the positioning of key individuals in the network. In a network measured by static ties, a typical broker connects two otherwise disparate communities. In a dynamic communication network where most actors are connected via conversations, the key brokerage constitutes a hub-like center with wider-reaching influence thanks to their conversations with different groups of people while other types of conversational brokerages (i.e., brokerage roles established through conversational activities) also exist within the network. Moreover, such brokerage is a structure/strategy in process in the conversation network, since it is contingent upon relative interaction intensity among major actors and the timing of these interactions.

Finally, longitudinal conversational networks can reveal how network ties are constituted and negotiated over time, how brokerage forms and decays, and when communal polarization is absent or present. Taken together, this approach allows us to materialize the idea that elite relationships are reproduced via conversational ties and that power networks continually evolve from these interactive processes (Fuhse, 2022:235).

In addition to these relational effects, there are also sequential and contingent effects of conversational interactions. Elite relationships may abruptly change during key conversations, insofar as “conversational encounters are permeable to network effects but not entirely so” (Gibson, 2005:1561). Such sequential/contingent effects may appear within conversations (e.g., turn taking) and across a sequence of conversations within a short time period. To show these sequential and contingent effects, we need to zoom out on the conversation network at certain critical junctures.

Data and methods

Data

We modelled our dataset as a network of conversational ties. Given the nature of the data, we employ longitudinal network analysis and narrative as complementary tools for historicizing elite interactions. Our original dataset is comprised of conversation records extracted from the chronicles (“Nianpu” in Chinese) of top leaders, which were compiled by party historians for deceased leaders based upon the daily activities that had been recorded by their secretaries.⁸

To construct our dataset, we coded the chronicles of six top leaders from the reform era⁹: Deng Xiaoping, Chen Yun, Li Xiannian, Hu Yaobang,¹⁰ Ye Jianying, and Peng Zhen.¹¹ These chronicles all follow a similar format: each conversation item records essential information, such as participant names, their political positions, dates, locations, conversation forms, subject matters, and contents.¹² Notably, we only coded conversation records from informal meetings that took place at residences, hotels, and hospitals rather than records from public meetings or ceremonies.¹³ These informal conversations often lasted several hours and involved two or more participants. For each conversation, we constructed one conversational tie between each pair of participants, following the method used in studies of multi-way interactions (Moody, 2004).

From 1,336 unique records of conversations in more than 5,000 pages of chronicles, we extracted 6,114 time-stamped conversational ties between 1977 and 1992 among the six top leaders and between them and almost 600 other political elites. We do not have the complete communication data for individuals not belonging to the group of six top leaders, but we do have complete information for conversations among the six leaders (egos) and their conversational ties with other elites (alters).

Our use of conversational ties to measure relationship strength and structure may raise some questions about validity. First, informal conversations may still occur between rivals. We therefore do not assume conversational ties alone indicate factional alignment. However, statistically speaking, allies have a significantly higher number of private, informal conversations than do rivals. This is especially true in the case of the elder leaders, who preferred speaking with those they trusted in their decreasing number of private conversations. Second, leaders may sometimes meet because they are responsible for addressing the same policy issue without having a particularly strong relationship. Nevertheless, this confounder is less applicable to the six top leaders, who were able to intervene in literally any policy issues. Third, quantitative measures may reveal the general patterns of elite interactions and influences, but they cannot induce straightforward causal imputations or capture abrupt relational change during eventful moments such as the Tiananmen incident. We offset this limitation by making use of other primary sources to conduct refined network analysis and provide in-depth narrative accounts.

Although the chronicles are considered high-quality sources for studying Chinese politics and have been broadly used in scholarly works (e.g., Brown, 2021; Su, 2023; Teiwes & Sun, 2016; Vogel, 2011), there are a few issues challenging the reliability of our dataset. Above all, it is possible that some conversations were omitted, censored or concealed for unknown purposes. While this concern is legitimate, the chronicles do contain information that researchers generally consider sensitive: for instance, the chronicles record that these top leaders held informal meetings to discuss how to suppress the Tiananmen movement because suppression was considered a wise and appropriate choice by party leaders. While the records of conversation contents were often abridged or even omitted, our analysis was affected little by these missing contents because we mainly quantify the conversational ties for network analysis rather than conduct content analysis or topic modeling. Lastly, we use other primary sources to complement observations derived from the conversation dataset.¹⁴

Methods

We used a mixed-methods approach to analyze this conversation dataset. Instead of standard quantitative analysis for hypothesis testing, we integrate network analysis and narrative to triangulate on the complex political processes (Padgett, 2011). Longitudinal network analysis can reveal general trends of elite interactions and relational patterns, while narrative provides contexts, explanations, and interpretations to complement the otherwise simplified network visualization and quantification and zoom in and out at critical moments.

We first employed longitudinal network analysis to examine the elite conversation network over time, revealing the relative strength of their communications, the community structure within the network, and the change of their network roles (Knoke, 1994). We aggregated the conversational ties using different time windows appropriate for our analyses. In a time-aggregated conversation network, a conversational tie exists between two actors if they were involved in any conversation during the period, and the tie is weighted by the number of conversations held.

To illustrate patterns of conversation ties, we visualize the conversational network across different time periods and the “leader-topic” two-mode network. Our network visualizations are produced by the software Gephi, and the layout of the networks are determined by tie strengths using the Fruchterman-Reingold algorithm—meaning that nodes with stronger ties (i.e., more conversations) are closer so that we can visually inspect the relationships among actors. We also calculate various summary statistics of the conversations to substantiate the visual patterns. Alongside descriptive analysis, we also made use of some analytic methods: e.g., centrality analysis to identify the central actors in the conversation network over time and community detection to quantify the level of polarization of the conversation network.¹⁵

We complement the conversation network analysis with narrative and case study based upon fine-grained qualitative evidence—mostly firsthand observations. We have drawn from a variety of primary sources—elite diaries, memoirs, autobiographies, and declassified files—many of which have been only recently available. We decide not to cite directly from the Tiananmen Papers (Zhang, 2001)—a well-known collection of unverifiable, anecdotal personal meeting minutes pertaining to these top leaders during the Tiananmen movement—but we do use it for triangulation.¹⁶ We also made use of unpublished sources—including the Beijing chronicle of Zhao Ziyang and the diary of Li Rui (an important CCP official and an insightful insider)—from the Yenching Library at Harvard University and the Hoover Institute at Stanford University and interviewed a few party historians who have access to internal materials and elite informants.

Conversation networks: positioning and structure

This section uses conversation network analysis to reconstruct the relationship structure during this era. It shows that while Deng was the central actor in the network, he was still constrained by the ever-changing elite relationships. He was thus neither the single dominator of the network nor a passive adapter to factional struggles. This was both because other leaders also held (different types of) brokerage roles and Deng’s brokering capacity was contingent upon the varying degree of relational polarization among these leaders.

We organize our analysis into four time periods corresponding to the reigning eras of the four nominal leaders during the reform era: Hua (1/1977/-11/1980), Hu (12/1980–12/1986), Zhao (1/1987–5/1989), and Jiang (6/1989–12/1992). This periodization is thus associated with our explananda, i.e., the four major power reshuffles: the downfall of Hua Guofeng in December 1980, the forced stepdown of Hu Yaobang in January 1987, the brutal ousting of Zhao Ziyang in May 1989, and Jiang’s reversal and subsequent power consolidation at the 14^th Party Congress in late 1992.¹⁷ This temporally organized analysis enables us to explain both uniformity—i.e., Deng triumphed in all cases—and variation across the four political conflicts.

In our network analysis and narrative, we focus on the conversations between the six egos and other top leaders. During the reform era, China’s formal power structure was comprised of three institutional lines (Fig. 1). The power center was the Politburo (20 plus members) and the Politburo Standing Committee (PSC), which had 5–7 sitting members. The PSC was led by a nominal party leader, called Chairman (1977–1982) or General Secretary (1982–1992). Although Politburo and PSC members were reelected during the Party Congress every five years—in 1977, 1982, 1987, and 1992—top leadership transition could occur at literally any time (Wu, 2015). There were two other, parallel institutions at the top in addition to the Politburo and PSC. First, the Central Advisory Commission (CAC)—which was chaired successively by Deng Xiaoping and Chen Yun—was created in 1982 to comfort retiring party elders and ease generational transition. Second, the Central Military Commission (CMC) of People’s Liberation Army (PLA) was a vital source of power, especially under the chairmanship of Deng Xiaoping in 1981–1989. Notably, the top leaders of the CAC and CMC and a few marshals enjoyed PSC political status even when they were not PSC members. Finally, since real authority was exercised informally more than institutionally during this era, some prestigious elders remained behind-the-scenes players even after their retirement.

Fig. 1

Formal Power Structure in China’s Early Reform Era

The 24 most important leaders were PSC members and those enjoying PSC-level political status during this period (Table 3 in Appendix 1 lists their career information). Demographically, these individuals fell into two generational cohorts, with the average birth year being 1912 (SD = 11). The 12 party elders were all born before 1910, joined the party in the 1920s, first became CC members in 1945 or 1956 (except for Chen Yun in 1931), reached the top status in the party before the Cultural Revolution, and returned to the top in the late 1970s. Even after (semi-)retirement in 1980s, they maintained enormous influence via informal meetings with each other and with their younger protégés who held formal positions (Guo, 2019:300). In contrast, the 12 younger leaders were all born after 1915, joined the party no earlier than 1932, first became CC members after 1956 (except for Hu Yaobang), and were promoted to the Politburo mostly during the reform era (except for Wang Dongxing and Hua Guofeng). The interactions of these 24 people not only accounted for the majority of our conversation records but also were the most politically consequential.

Deng’s overall dominance in the conversation network

Deng’s advantage in reform-era political conflicts was related to his dominance in the power structure, as reflected by the conversation network. First, insofar as the importance of leaders can be measured by the frequency of their conversations with important figures, Deng had the highest number of conversations with other PSC-level leaders. Table 2 provides descriptive statistics regarding the number of conversations that took place between each focal leader (ego) and the other 23 leaders.¹⁸ With 255 conversations in total, Deng was consistently the most active across the four periods. By comparison, Chen (156 conversations in total) was active only beginning in period II, with a relatively inactive status in Period I due to his illness. Hu Yaobang (120 in total) was active until he was removed from the role of General Secretary at the end of Period II. Li Xiannian (115 in total) and Peng Zhen (95 in total) were active across all four periods but they were not dominant actors. The oldest PSC member, Ye Jianying (44 in total) was the least active and died in 1986.

Table 2 Conversation numbers of the six egos with PSC-level leaders

Figure 2 visualizes the conversational ties of all 24 PSC-level leaders during this era. Deng had frequent conversations with all other “key” leaders; by comparison, Chen Yun mostly conversed with Li Xiannian, who, interestingly, had many conversations with both Deng and Chen. Hu Yaobang had more conversations with Deng and Zhao than with Chen or Li. Taking the 16 years together, there is little doubt that Deng was situated at the power center. In addition, the figure shows no ideological polarization (though we argue that there is no fixed ideological cleavage, for clearer visualization we adopt a widely recognized classification of the elites into reformers, conservatives, or neutralists—those without identifiable ideological positions).

Fig. 2

Conversation Network of the 24 PSC-level Leaders, 1977–1992. Note: The size of a node is proportional to the number of its conversational ties. The width of each edge is proportional to the number of conversations on the tie. Circle nodes are the six egos for whom we have the chronicles and triangle nodes are the other PSC-level leaders. Nodes are color-coded by their conventional ideological labels, with blue being reformers, red being conservatives, and grey being neutralists

Second, important leaders exerted more influence on important conversational topics. We coded the content of the conversation records using the topic labels of economy, ideology, personnel, security, military, diplomacy, united front,¹⁹legislation, education/science/culture, and general.²⁰ Among these topics, economic, ideological, and personnel issues were the most discussed in conversations among top leaders—and therefore the most important (Fig. 3).²¹ Figure 4 presents a two-mode network of actors and topics showing how the six egos engaged the ten topics in their conversations. Deng and Chen had the most conversations about economic affairs (Deng: 32; Chen: 28); Deng, Chen, and Hu were the leaders who had the highest number of conversations about personnel affairs (Deng: 18; Chen: 7; Hu: 8); Deng and Hu had the highest number of conversations about ideological affairs (Deng: 15; Hu: 9). Overall, Deng played a vital role in all three issue areas.

Fig. 3

Cumulative Frequency of Conversation Topic Among Top Leaders, 1977–1992. Note: This figure only counts conversations that involved at least two of the six top leaders, in order to exclude “insignificant” conversations between leaders and non-leaders such as their secretaries

Fig. 4

Two-mode Network of Six Egos and Ten Conversation Topics, 1977–1992. Note: The width of each edge is proportional to the number of conversations

Third, Deng also enjoyed unique structural advantages in terms of spheres of influence. As Fig. 4 shows, one source of Deng’s advantage came from the fact that he was the only top leader who commanded both military and civilian power lines between 1981 and 1989 (Periods II and III). Deng not only oversaw every field of civilian affairs but also almost monopolized military-related conversations (70%). Marshal Ye joined military-related conversations, but he was minimally present in civilian affairs.²² Other leaders were involved in several fields of civilian affairs—most notably, Chen Yun exerted a consistent influence on economic decisions almost equal to Deng’s (Fig. 4)—but were almost absent from military conversations. Deng thus possessed a strategic advantage over other prestigious civilian leaders (especially Chen Yun) and over army leaders (especially Marshal Ye) by occupying a unique “structural hole” between these two otherwise disparate groups, especially in 1981–1989 (Burt, 2002). Although some of these findings have been mentioned in previous studies (e.g., Huang, 2000: 363–65), our analysis clearly demonstrates that Deng’s dominant network position, which granted him a unique advantage in power struggles.

The changing network structure: centrality, brokerage, and polarization

To explain why it took Deng various efforts to achieve “victory” in the four vital struggles, we now turn to the change in the conversation network structure over time by examining a few network characteristics: centrality, brokerage, and polarization. The “centrality” measure indicates the relative importance of each actor in the conversational network; “brokerage” tackles different types of brokers in this network; the “polarization” measure shows whether the conversational structure was polarized.

First, Deng’s conversational centrality within the smaller group of six top leaders varied over time, constraining his structural advantage and brokering capacity. We employ network centrality measures to assess the “importance” of each actor in terms of their network position. In this type of analysis, central actors tend to be well-connected in the network, acting like hubs. To assess each top leader’s significance in their conversation network, we compute degree centrality—the number of ties an actor has, normalized by their total number of ties for comparison across networks—on each aggregated conversation network between the six top leaders.²³ Figure 5 shows the change in degree centrality of the six top leaders across the four periods. Deng’s degree centrality was highest in periods I and II, while it fell in periods III and IV. Structurally speaking, Deng was still the robust actor in Period III given that he continued to command the army and lead the power center, but his brokering capacity was weakened by the fact that he no longer maintained close communications with conservative elders. This also means Deng had decreasing degrees of freedom in exercising his agency in conflicts during the later periods, as shown in the next section.²⁴

Fig. 5

Degree Centrality of the Six Egos Over Four Periods

Furthermore, other leaders played different kinds of brokerage roles and exercised their own agency. One type is “coordinator” within the power center: an otherwise non-dominant coordinator can serve as an intermediator between less connected actors. As the conversation intensity analysis shows (Fig. 6), while there were few contacts between Deng and Chen, Li Xiannian had more conversations with both Deng and Chen, especially in periods I and II.²⁵ As such, Li served as a coordinator between the two most prestigious leaders, playing a vital role in Deng and Chen’s joint battle against Hua Guofeng. Likewise, Jiang Zemin served as a coordinator between Deng and Chen in Period IV, given the two elders no longer met while Jiang keeping close contact with both. As shown below, this role protected Jiang from being dethroned in the intensified power struggle in 1992. In both cases, it is obvious that Li or Jiang was not as powerful as Deng or Chen, but their coordinating role presented them unique advantages in critical political conflicts.

Fig. 6

Conversation Intensity Among the Six Egos Across Four Periods. Note: We calculate the conversation intensity for each pair of actors in the aggregated network of the six egos for each period and visualize them in the colormap in this figure, where darker red denotes higher intensity

In addition, even less powerful leaders such as Bo Yibo and Yang Shangkun can serve other brokerage roles. Bo Yibo, the executive director of CAC (1982–1992), served not only as an institutional interlocutor between top leadership and CAC members, but also as a key itinerant (messenger) between Deng and Chen—two increasingly hostile leaders—regarding personnel arrangement before the Party Congress in 1987. Heading the Leadership Appointment Group, Bo Yibo met Deng for six times (Feb. 4, Feb. 6, March 10, April 28, July 7, July 25, Sept. 6) and met Chen for eight times (Feb. 7, March 5, March 10, April 9, July 3, Sept. 6, Sept. 29, Oct. 9), mostly for arranging the (semi-)retirement of elders and the new Politburo and its standing committee for the 13^th Party Congress (Zhao, 2009:208–213; Wu, 2013:280–281; Lu, 2019:1033–1044). Meanwhile, Deng and Chen only met twice in the PSC meetings during this same period.²⁶ In the same period, Yang Shangkun was like Deng’s gatekeeper/representative rather than messenger. While this role reduced his mediating function between Deng and conservative elders, it granted him enormous power in the 1989 Tiananmen crisis, as shown below.

The last characteristic of the conversation network is polarization or lack thereof. We use modularity analysis to quantify the extent to which the conversation network breaks into communities of actors (Newman, 2006). Given a partition of the network into communities, modularity is defined as the fraction of edges that fall within the given communities minus the expected fraction if the network is random.²⁷ Thus, a high modularity score suggests that the network has dense connections between nodes within the same communities and loose connections between communities. In other words, a network with high modularity often consists of densely internally connected clusters. Figure 7 shows the change in modularity of the conversation network across the four periods, which reveals the polarization level of the network. There were not two clear conversation blocs in periods I and II as the modularity is low; modularity analysis also finds more than two small communities, indicating ideological differences had not evolved into two clear-cut factions at that time. Polarization only appeared in period III and continued in period IV, reflected by the high modularity shown in Fig. 7; modularity analysis also identifies exactly two communities in those periods.²⁸ For example, in period III, Chen Yun and Li Xiannian led one community and Deng Xiaoping and Zhao Ziyang were in the other. Closer read of the conversation records also indicates factional polarization only appeared in 1986, at the end of Period II. As we show below, this polarized relationship structure limited Deng’s capacity to exercise his power in the two conflicts in 1989 and 1992.

Fig. 7

Polarization of the Conversation Network over Time. Note: Modularity works in a relative sense: we can compare modularity scores of a network at different times, but there is no general rule of thumb for a good modularity score. To supplement the cold numbers, we also mark the nodes in Fig. 8 by their conventional reformer or conservative labels, so as to show the level of factional division over time

Conversation networks and elite conflict in China, 1977–1992

This section examines how the network characteristics affected each conflict and provide a narrative account of the four main power struggles by drawing upon the conversational network dataset and other sources. In brief, the varying pathways of the four elite collisions were largely explained by changing network centrality, brokerage and polarization. To better illustrate the explanation, Fig. 8a, b, c, d includes the aforementioned 24 PSC-level leaders and other sitting Politburo members in each of the four periods, with their conventional ideological labels. By referring back to this figure and previous network analysis, the subsequent narrative will flesh out this explanation vis-à-vis the two extant explanations.

Fig. 8

The Conversation Network of PSC and Politburo Members across Four Periods, 1977–1992. a Period I, 1/1977–11/1980. b Period II, 12/1980–12/1986. c Period III, 1/1987–5/1989. d Period IV, 6/1989–12/1992. Note: The legend of Fig. 8 is the same with Fig. 2

Dethroning Hua

Period I began with Hua Guofeng’s power consolidation following Mao’s death and ended with his downfall in December 1980. Meanwhile, senior survivors of the Cultural Revolution, especially Deng Xiaoping and Chen Yun, returned and ascended to top positions, while younger leaders (e.g., Hu Yaobang, Zhao Ziyang, and Yao Yilin) rose to prominence. During this period, the factional division of reformers and conservatives had not yet appeared, and Deng only gained centrality gradually in the network. Thus, the power transition from Hua to Deng unfolded less as a victory of reformers over conservatives (or radicals) than a diverse coalition led together by Deng and Chen, who were coordinated by Li Xiannian (Li-Ogawa, 2022; Teiwes & Sun, 2016).

Policy wise, there was not such difference between Deng and Hua, who was the earliest designer of China’s reform and open-up policy. Figures 7 and 8a show that there were not two polarized conversational blocs during this period. However, dethroning Hua—Mao’s chosen successor—soon became the common goal of many elders who had suffered during the Cultural Revolution. Though Hua maintained close interactions with most other PSC members until his downfall, he had fewer conversations with Chen Yun and rising young leaders such as Hu Yaobang and Zhao Ziyang (see Fig. 8a).²⁹ Entering the PSC in mid-1977, Deng kept close conversational ties with both elders (e.g., 18 conversations with Li Xiannian and 11 with Wang Zhen) and young leaders—including future reformers (e.g., 32 conversations with Hu Yaobang and 8 with Zhao Ziyang) and future conservatives (e.g., 34 conversations with Hu Qiaomu and 22 with Deng Liqun). As Fig. 5 shows, Deng’s normalized degree centrality ($\approx 0.31$) among the six egos was also highest during this period. Chen Yun had only 2 conversations with Hua and almost no private conversations with Wang Dongxing, Mao’s bodyguard and Hua’s major PSC ally.³⁰ Li Xiannian maintained frequent conversations with both Chen (7 conversations) and Deng (18 conversations), serving as a coordinator between the two (also see Fig. 6 and discussions above). The Deng-Chen-Li triangular coalition played a vital role in dethroning Hua via such mode of interactions.

Once Deng, Chen, and Li forged a consensus about leadership turnover sometime in 1979, Hua’s days were numbered. 1980 offered nothing but loss for Hua, who had few personal conversations with elders. Precisely because of this relationship structure, the power transition was “smooth”: in February, Hua’s four associates in the Politburo resigned; Hua resigned from Premiership in August and stepped down from Party and CMC Chairmanship in December 1980. Deng succeeded Hua in chairing the CMC, thus becoming formally positioned as a robust actor straddling the civilian and military power lines.

Ousting Hu

Period II began with a power structure shared among six PSC members—four senior (Deng, Chen, Li, and Ye) and two younger leaders (Hu as General Secretary and Zhao as Premier)—and ended with the forced resignation of Hu, Deng’s first chosen successor. The ideological cleavage account, which depicts Hu as a sacrificial figure in intensified conflicts between reformers and conservatives, misses the cleavages crosscutting economic policy, ideological orientation, generational differences, and peer competition. Sitting above this power structure, Deng was the robust actor across frontline leaders, party elders, and army leaders, as well as the arbitrator between the two young reformers (Hu and Zhao) and between reformers and conservatives. Holding such a robust-actor role in a crosscutting power network allowed Deng to mobilize diverse groups of leaders to remove Hu without triggering much elite or popular resistance.

Conversational patterns show that Deng’s robust-actor position traversed the crosscutting elite relationships and all vital issue topics during this period. Figure 5 shows Deng’s degree centrality as highest among elders. As shown in Fig. 8b, Deng maintained close conversational relationships with both economic reformers Zhao Ziyang (24 conversations) and Wan Li (14 conversations) and conservatives Wang Zhen (3 conversations) and Bo Yibo (7 conversations). Deng also had frequent conversations with two leftist ideologues Hu Qiaomu (17 conversations) and Deng Liqun (14 conversations). His coordinating role was only reinforced by resolving the split between the two reformers—Hu and Zhao—over economic policy and leadership.³¹

Conversational blocs were not polarized until the end of this period (Figs. 7 and 8b). Like Deng Xiaoping, Chen Yun also had conversations with both conservative leaders such as Yao Yilin (14 conversations), Song Ping (4 conversations), and Li Peng (2 conversations) and young reformers Zhao Ziyang (9 conversations) and Hu Yaobang (10 conversations). In particular, bearing close connections with frontline economic leaders, Chen’s interventions in economic decision-making rivaled Deng’s. In the Hu-Zhao dispute, Chen Yun also helped Zhao take over economic leadership from Hu (CYNP, 2012: vol.2, p. 371–72). Other primary sources also show factional split over economic and ideological policies only began to appear at the end of this period (e.g., Deng, 2006). According to Zhao Ziyang’s memoir (2009: 93–94): “as for Comrade Chen Yun, I had enormous respect for him in the years when I first started working in the center. … In the initial few years, the two of us had a good relationship. I was even able to mediate and ease communication between Deng and Chen… Problems developed as reform deepened. …The distance between us grew greater.”

The actual collision between Hu and Deng was thus not a polarized conflict over economic or ideological policy but over the power arrangement and did not take place between reformers and conservatives but between two reformers, Deng and Hu. Eleven years younger than Deng, Hu might have considered the 1987 party leadership turnover as a succession, as Deng had occasionally “promised”. Yet when Hu implied Deng should step down in an interview in 1985, he not only put their dyadic relationship in jeopardy but also threatened the status quo of all the elders. After this interview was exposed, Deng made his disapproval widely known in informal conversations with party elders, whose gossip furthered Deng’s distrust of Hu’s intention (Zhao, 2009:169). After the death of Marshal Ye—Hu’s last backer— Deng, Chen, and Li decided during an informal meeting in October 1986 to remove Hu at the upcoming Fall 1987 Party Congress (Chen, 2016, vol.1:144–46). The student protests in December 1986 only accelerated Hu’s downfall.³² Following an order from Deng, Hu was attacked by nearly all senior leaders and sacked in disgrace during the party meetings in January 1987. Deng’s robust-role position at this time secured an easy victory (Chung, 2019).

The Deng/Zhao falling out

The third period witnessed the sudden falling out between Deng and Zhao. We will analyze elite interactions during the Tiananmen movement in the next section; here we only outline the change in the relationship structure which set the groundwork for the schism. This period saw a polarized conflict between “reformers” and “conservatives,” but the final collision occurred between two reformist allies. Commanding the army, Deng remained the robust actor, but his brokering capacity declined given his decaying relationship with conservative elders (see Figs. 4 and 5). As a result, Deng had to make a hard choice in factional conflicts, especially after he was pushed by Zhao, his ally and heir.

Three features of the power structure in Period III were reflected by the conversational network. First, Deng led a new triangular power nucleus with Zhao Ziyang and Yang Shangkun. This period began with an inter-generational leadership transition in 1987: the PSC was replaced by a younger cohort led by Zhao, while Deng remained the CMC chair; most other prestigious seniors (semi-)retired in compensation for merely honorary titles. Excluding Chen Yun and Li Xiannian, the new ruling nucleus included Deng (who wielded the ultimate, coercive power), Zhao (General Secretary, i.e., the nominal leader of the Party), and Yang Shangkun (Deng’s protégé and gatekeeper/representative, who supervised the army on Deng’s behalf). As Fig. 8c shows, frequent conversations occurred within the Deng-Zhao-Yang axis (13 conversations between Deng and Zhao, 16 between Deng and Yang, and 12 between Zhao and Yang). Li Peng, despite being Premier, was not included, as illustrated by his far fewer conversations with this core trio (8 with Deng, 2 with Zhao, and 3 with Yang).

Second, ideological/economic positioning was polarized during this period (Figs. 7 and 8c). Zhao no longer consulted other elders: he had far fewer conversations with Chen (6 conversations, but none in 1989) or Li Xiannian (1 conversation) than with Deng (13 conversations). According to Zhao, during his years as General Secretary: “Li Xiannian claimed that I only listened to what Deng Xiaoping said, while ignoring him” (Zhao, 2009:244). Deng’s relationship with the elders also worsened, as reflected by their significantly reduced conversations (e.g., 2 with Chen Yun and 3 with Li Xiannian). In contrast, Chen and Li Xiannian continued to have frequent conversations (12 conversations). A parallel polarization existed among the younger leaders: a reformist bloc with Zhao, Hu Qili, and Tian Jiyun and a conservative bloc entailing Li Peng, Yao Yilin, and Song Ping. Though Deng was still the supreme leader, Li, Yao, and Song had more conversations with conservative elders.

Third, although still being the paramount leader, Deng lost brokering capacity among elders in their informal relationship structure. Figure 5 shows that Deng’s degree centrality in the conversation network dropped behind Chen’s, Li’s, and Peng’s; conversations records also show that Li Xiannian continued to converse only with Chen and almost ceased to interact with Deng. In other words, while the new ruling nucleus marginalized conservative elders, it also alienated Deng from informal conversations of these elders. That’s partially why Deng needed an itinerant (Bo Yibo) for communication with Chen Yun in the 1987 power transition, as mentioned above.

The key to making this ruling structure function was thus the solidarity between Deng and Zhao. The two leaders had sustained waves of attack from conservatives before the Tiananmen movement. In 1988–89, the abrupt but aborted price reform advanced by Deng and Zhao caused explosive inflation and social panic, thus drawing broad criticism from the conservative bloc (Fang, 2004, vol. 11; Weber 2021:247–258). Zhao was called to resign by the conservative bloc at the turn of 1989. At that time, Zhao received a firm endorsement by Deng, who commended the conservative Premier Li Peng and Executive Vice Premier Yao Yilin to respect Zhao’s leadership (Lu, 2019: 1156). However, this fragile relationship structure eventually led to power collision and a disastrous crisis response during the Tiananmen movement, as shown later.

The reversal of Deng/Jiang tension

Period IV began with a reactionary response to the Tiananmen movement, but this direction was eventually reversed after Deng’s legendary Southern Tour in early 1992. With declining brokerage in the power network after retirement, Deng was unable to build another coalition to “legally” remove General Secretary Jiang Zemin but only forced Jiang to reorient his policy by implying the use of coercive power. Jiang switched his position to avoid a clash with Deng and later consolidated power with Deng’s support. Neither ideological cleavage nor manipulative brokerage can explain this dramatic reversal of the Deng/Jiang relationship.

After the Tiananmen crackdown, the new power nucleus was composed of Deng (who resigned all formal positions in November 1989), Jiang (General Secretary and Chairman of CMC), Li Peng (Premier), and Yang (State President and Permanent Vice Chairman of CMC, and Deng’s representative in the army). As shown in Fig. 8d, Deng, Jiang, Li Peng, and Yang constituted a conversational community (12 conversations between Deng and Jiang, 10 between Deng and Li, 15 between Deng and Yang, 14 between Jiang and Li), while Chen and Li Xiannian continued to remain close (8 conversations). The rift between Deng and his conservative comrades returned after the Tiananmen crackdown (Figs. 7 and 8d): for example, public records show that Deng and Chen had only one private conversation during this period (Baum, 1994:319–20). Having witnessed the dismissal of two predecessors, however, Jiang understood the support of leading elders was vital to his survival (Kuhn, 2004:204). He thus maintained informal conversations with Chen (5 times) during these years (CYNP, 2012). Likewise, Li Xiannian wrote nine letters to Jiang between July 1989 and March 1992 to voice his opinions, though the two had only one private conversation. Unlike Zhao Ziyang, who had sided with Deng, Jiang became a de facto coordinator between Deng and conservative elders.

In other words, Deng was no longer the robust actor during this period. This relationship structure affected how Deng reversed Jiang’s policy orientation. In reaction to the Tiananmen movement, Jiang cautiously adopted a stance of ideological conservatism and economic retrenchment,³³ which had disastrous economic effects in 1989 and 1990 (Fewsmith, 1994:87–122; Naughton, 1995:274–88; Xiao, 2019:260). The influence of conservative elders on Jiang was considerable and consistent: just three days before Deng’s Southern Tour, Jiang and Chen had a conversation in Shanghai (CYNP, 2012, vol. 3:500); the next day, Li wrote to warn Jiang and other PSC members of “bourgeois liberalization” (LXNNP, 2011, vol.6:569).

Disappointed by his successor again, Deng made it clear that “whoever is against reform must be driven out of his office” during his Southern Tour in January 1992. Deng’s message was threatening, especially because Yang Shangkun—Deng’s representative—openly signaled to use the army to “protect and escort” the economic reform. Jiang was at a crossroads: defer to Deng or die politically. Sensing their perilous relationship, Jiang called Deng on the Chinese New Year (Kuhn, 2004:214)³⁴ and then made Deng’s speeches the focus of the PSC and Politburo meetings in early March. At the PSC meeting, Premier Li Peng also apologized for his overcautious economic policy (Li, 2007, vol.2:911). Through conversations on April 3^rd, June 18^th, and August 17^th, Jiang and Li reached a consensus regarding how to adjust their economic stance (Li, 2007, vol.2:917, 930–31, 943). On June 9^th, Jiang coined a bold reformist slogan, “Socialist Market Economy,” which was approved by Deng three days later in their first meeting after the Southern Tour. The slogan became the goal of China’s economic reform during the 14^th Party Congress in October 1992, when Jiang became the Party’s third-generation core (Tsou, 1995:100–102).³⁵

The way this reversal unfolded is causally related to the elite network structure. The new Deng/Jiang ruling nucleus did not favor Deng, because his relationship with conservative elders continued to deteriorate while Jiang maintained close interactions with both sides. Therefore, unlike in the ousting of Hu and Zhao, Deng was unable to make another (even temporary) alliance with party elders, who in fact sided with Jiang (Guo, 2019:302). Although Deng eventually converted Jiang, he did so by implying the use of coercive power (i.e., a coup, since Deng held no formal position at that time)—a signal of his declining brokering capacity. Once Jiang shifted his policy orientation in 1992, he was endorsed, perhaps reluctantly, by both sides.

In sum, elite relationship structure and ongoing interactions mutually constituted each other. This section focuses on how changes in network characteristics—centrality, brokerage, and polarization—affected Deng struggled with the four nominal leaders: the gradual dethroning of Hua in 1981, easy ousting of Hu in 1987, abrupt removal of Zhao in 1989, and forced reversal of Jiang’s policy orientation in 1992. In particular, we find that different kinds of brokerage emerged in this network (Gould & Fernandez, 1989; Spiro et al., 2013; Stovel & Shaw, 2012). While Deng was the robust actor in Periods II and III, other leaders such as Li Xiannian, Bo Yibo, Yang Shangkun, and Jiang Zemin also exercised brokerage—as coordinator, itinerant, or gatekeeper/representative—by controlling vital communication routes (Mizruchi & Potts, 1998:384). As both brokers and their brokerage roles changed over time, even the paramount leader was constrained by the shifting relationship structure.

The Tiananmen power collision: a critical case

This section uses the actual unfolding of power collision during the 1989 Tiananmen movement as a critical case to further test the three competing models: the Tiananmen power collision as the result of 1) ideological cleavage, 2) Deng’s robust action, or 3) situational conflict. The political climax of the reform era, the Tiananmen movement tragically began with the unexpected death of Deng’s ousted first successor Hu Yaobang and ended with the downfall of Deng’s second successor Zhao Ziyang. Unlike Hu’s deteriorating relations with Deng more than one year before his ousting, there had been little sign of tensions between Deng and Zhao until May 1989. Instead, as the previous section showed, their alliance had been strengthened within a polarized power structure. Why then did Deng abruptly abandon his long-time ally and reformist heir?

With new primary sources and high-quality historiographies (e.g., Zhao, 2009; Li, 2010; Chen, 2016; The Last Secret, 2019; Lu, 2019; Brown, 2021), we are able to assess the three models by reconstructing the interactive process that took place at the political center. The “ideological cleavage” model has well explained the intensified factional tensions between reformers and conservatives, but this was not the determining factor for the final split of the two reformist allies. The “manipulative brokerage” model correctly positions Deng in the analytic center, but little evidence suggests that Deng strategically plotted the ousting of Zhao or called for military suppression in advance. Our “situational conflict” model argues that the Deng/Zhao split resulted from their overactions and overreactions based upon their discrepant understandings of the changing relationship structure. In what follows, we amass ample evidence to demonstrate why our model offers a better explanation for the relational breakdown between Dang and Zhao and Deng’s decision to invoke martial law.

Before elaborating our argument, we briefly summarize the Tiananmen movement from a top-down perspective. On April 15^th, the unexpected death of Hu Yaobang—broadly considered an unfairly treated reformer—prompted students to demonstrate on Tiananmen Square. After Zhao Ziyang left for North Korea for a state visit on April 23^rd, Premier Li Peng immediately took a hardline stance and convinced Deng to define the protest as “turmoil.” Deng’s view was expressed in a front-page editorial, entitled “It is Necessary to Take a Clear-cut Stand against Turmoil,” in the official newspaper The People’s Daily on April 26^th, which antagonized student protesters. Though initially supporting this editorial from North Korea, Zhao changed his view after returning to Beijing on April 30^th and conveyed a reconciliatory stance in public speeches on May 3^rd and 4^th. After that, party leaders disputed how to handle the protest. On May 13^th, when Beijing students began launching a hunger strike, Zhao tried to persuade Deng to soften the hardline stance during their sole private meeting since late April, but he was unable to change Deng’s position. The student occupation of Tiananmen also disrupted Gorbachev’s visit to Beijing starting from May 15^th, the first summit between USSR and China in thirty years. In a meeting convened by Deng on May 17^th, Deng began attacking Zhao’s early May speeches as a “bad” turn and proposed martial law, despite Zhao’s opposition. On May 19^th, Li Peng declared martial law, which became effective on May 20^th. Meanwhile, party elders selected Jiang Zemin as Zhao’s replacement. The student movement was bloodily suppressed on June 4^th.

Relational and sequential effects of conversations

Like the “ideological cleavage” model, we acknowledge that factional division played an important role in the Tiananmen movement and find clear patterns of polarized interactions during this period. Factional conflict placed great pressure on the ruling group, as open clashes between Zhao Ziyang and Li Peng in May made political compromise difficult. However, closer examination reveals a more nuanced picture of conversational interactions during the movement.

Since they represented the two rival sides, we use Zhao Ziyang’s and Li Peng’s meetings to exemplify the patterns of elite communications during this period.³⁶ Overall, their meetings reflected and reinforced the polarization of the larger power structure (Fig. 9). Among the five PSC members, Zhao and his reformist ally Hu Qili upheld the most conciliatory stances, while Li Peng and Yao Yilin were hardliners; Qiao Shi was located somewhere between the two sides. After returning to Beijing, Zhao met with several reformers to consolidate their support and ask them to communicate his stance to larger groups, while Li met conservatives to strengthen their alignment. Notably, except for his meeting with Deng on May 13^th, Zhao did not meet or call any (semi)-retired elders. In contrast, Li visited Wang Zhen and had phone conversations with the secretary of Chen Yun, Bo Yibo, and other elders. By middle May, the two rival stances had been solidified: Zhao and Li had heated debates during the PSC meetings. Zhao was unwilling to communicate with Li personally but asked his associates to forward his position to Li, who also refused to back down. There was thus little room for compromise in their visible rivalry.

Fig. 9

The Conversation Network of Zhao Ziyang and Li Peng, April 15 – May 17, 1989. Sources: ZZYZNHSNJS, 2005; Li, 2010; Chen, 2016; Wu, 2019:108–519; Lu, 2019:1184–1251

A closer look, however, reveals the sequential effects of the conversational interactions on political realignment to have extended beyond the factional division (in Appendix 2, Fig. 10a and b lists Zhao’s and Li’s meetings chronologically between April 15 and May 17). Above all, the two divergent positions first appeared during Zhao’s foreign visit between April 23^rd and April 30^th. Before Zhao’s visit, Zhao, Li, and other leaders expressed no significant difference in their response to the protests.Zhao proposed three principles regarding the student protests, which were verbally supported by Deng, Li, and other PSC members (Zhao, 2009:5–10). According to Li, Zhao also opposed the agenda of “liberal democracy” held by some protesters and the “illegitimate student organizations” (Li, 2010:73). According to Li Rui’s diary on May 6, 1989 (Li, 2019, vol. 21, p. 46), before the North Korea visit Zhao had called the student protests a “turmoil” (dongluan), the same term that was later used by Deng to define the movement. There was thus little disagreement at that time.

It was Li Peng’s conversational interactions immediately following Zhao’s visit that facilitate the formation of a hardline stance. On the same evening as Zhao’s departure on April 23, Li visited Yang Shangkun to schedule a personal meeting with Deng. On April 24^th, Li had several conversations with top leaders to forge consensus, delivered that consensus to Beijing’s municipal leaders, and convened a PSC meeting to legitimize this consensus without notifying Zhao’s chief of staff. On the morning of April 25^th, Li convinced Deng to take a hardline stance in their sole conversation.³⁷ This “consensus” led to the April 26^th editorial that infuriated the students and escalated the protests. After Zhao’s return on April 30^th, Zhao and Li quickly diverged into moderate and hardline positions and struggled over how to (re)define and react to the student movement.

Moreover, Li Peng also created a favorable alignment beyond his conservative faction during Zhao’s absence. In other words, some reformers aligned with Li’s stance, making Zhao’s subsequent reaction more difficult. Zhao later admitted: “When I had visited him on April 19^th, he [Deng] had agreed with my position. On the 25^th, after being briefed by Li Peng and Yang Shangkun, he had changed his mind to agree with their assessment” (Zhao, 2009:10). Wan Li—then the Chairman of National People’s Congress and Zhao’s reformist ally—supported Li during this short period: “Wan Li fell for their trick. (Wan Li and I had been in total agreement in our view of the student protests.)” (Zhao, 2009:9). Likewise, the Beijing Mayor Chen Xitong, who had not been close to Li Peng and instead shared Zhao’s reformist view, supported Li’s hardline stance (CXTQS, 2012:40–42, 75).

Lastly, the most determining factor of the final collision was Deng’s and Zhao’s discrepant conceptions of the changing relationship structure and their overactions and overreactions, to which we will now turn.

Deng: robust action or reluctant reaction?

Like the “manipulative brokerage” model, our model recognizes Deng’s critical role in shaping the outcome of the Tiananmen event, both as a social movement and a power struggle. Moreover, both models disagree with the “ideological cleavage” explanation, which argues that Deng was manipulated by conservatives into imposing martial law (Brown, 2021: xii; Su, 2023). Where the two models disagree is with respect to whether Deng’s break with Zhao constituted strategic robust action or situational reluctant reaction. The “manipulative brokerage” model depicts Deng as a Machiavellian dictator sitting above the two factions, strategically pitting one against the other, and plotting the ousting of Zhao before their open split in middle May. Some of Zhao’s supporters even believed that Deng had deliberatively capitalized on the student protests to dethrone Zhao from the very beginning of the movement (for a summary of such interpretation: Brown, 2021:92; Su, 2023:113).³⁸ Our “situational conflict” model instead argues that while Deng gradually shifted being Zhao’s ally to a more neutral position, he did not make a final verdict to remove Zhao until the middle of May. Meanwhile, it was Zhao’s overconfidence in Deng’s trust that caused Zhao to make a sequence of “mistaken” actions and overactions, in turn provoking Deng’s reluctant but deadly overreaction.

As mentioned above, the relationship between Deng and Zhao had only been reinforced in the 1988 price reform and Deng gave Zhao firm support when Zhao was attacked by conservatives earlier in 1989. As shown in Fig. 5, Deng’s brokering capacity declined among top elders, unwilling and perhaps unable to collaborate with conservative elders to ambush Zhao. Zhao admitted in his memoir that he did not think Deng had plotted against him in advance: “Before the June Fourth incident, I had always felt that, overall, Deng had treated me very well and shown a lot of trust in me” (Zhao, 2009:48). On April 19^th, for example, the 85-year-old Deng had proposed the transfer of his military leadership to Zhao in their conversation preceding Zhao’s visit to North Korea (Lu, 2019:1157). Little evidence supports the claim that Deng deceived Zhao. Even Zhao did not question Deng’s intentions.

It was precisely because of Zhao’s confidence in their relationship that Zhao insisted on making his North Korea visit at such a critical moment, despite being suggested by both reformers (e.g., Tian Jiyun, Du Runsheng, and Chen Xiaolu) and conservatives (e.g., Li Tieying and Li Ximing) to postpone his visit until after handling the crisis in Beijing (Li, 2007:61; Wu, 2013:447; Wu, 2019:71, 99). It is also possible that Zhao was inclined to leave Li Peng to (mis)handle the crisis during his absence (Lu, 2019: 1170–1172; Brown, 2021:91–93; Gewirtz, 2022:226). Thinking counterfactually, if Zhao had stayed in Beijing, a more moderate approach would likely have come to the fore.³⁹ In sum, Zhao’s decision to leave based upon his (mis)-judgment became the first turning point in the making of the final collision.

The split between Zhao and Li Peng over movement response appeared soon after Zhao’s return to Beijing on April 30 (Zhao, 2009:30–31; Lu, 2019:1169, 1183–1185). Zhao’s two public speeches on May 3^rd and 4^th further made this division publically visible, although it is unclear whether Zhao sought to use this opportunity to scapegoat Li (Brook, 1998:31). Clearly, Zhao understood the key to winning this battle was converting Deng. In his personal meetings with Yang Shangkun, Wan Li, Yan Mingfu, and Xu Jiatun on May 2^nd and 3^rd, Zhao requested them to lobby Deng for retracting the editorial (Xu, 1993:370–74; Zhao, 2009:74–76; Li, 2010:90–91).

Having noticed Zhao’s changing stance, Deng shifted to a neutral position hovering somewhere between Zhao and Li and did not meet with either—but he did not abandon Zhao. Zhao still had many opportunities to indirectly deliver his views to Deng. For example, Zhao met Yang Shangkun—Deng’s gatekeeper/representative—five times before their joint meeting with Deng on May 13^th and asked Yang to convert Deng each time. Meanwhile, Deng straddled Zhao’s and Li’s positions: on April 30^th and May 11^th, Deng’s secretary Wang Ruilin called Li on behalf of Deng. These conversations delivered a significant message to Li, emboldening him to confront Zhao in the PSC meetings. However, Zhao was unaware of Deng’s shift from backing him to a balancing the two rival sides.

After the April 26^th editorial, Deng did not meet with any leader until May 11, when he talked to Yang Shangkun, according to the unverified record in The Tiananmen Papers. Deng is said to have made his final verdict during his “disappearing” period and to have told Yang in their meeting.⁴⁰ However, the unverified meeting record at most suggests that Deng adhered to the hardline stance but gave no signal to remove Zhao or crackdown on the movement (Chen, 2016: v.3, 36). Neither Yang nor Zhao interpreted this meeting as the end of Zhao’s career.

In effect, Zhao still aimed to seize the opportunity presented by Gorbachev’s visit to persuade Deng on May 13^th—just as the radical students were capitalizing on the visit’s global audience by using the hunger strike to demand government concession (Zhao, 2001:162).⁴¹ Notably, in the evening of May 12^th, Zhao visited Yang to prepare for this vital conversation, signaling they still remained certain level of mutual trust (ZZYZNHSNJS, 2005, p. 987). Accompanied by Yang, Zhao met Deng at Deng’s residence at 10am and suggested Deng consider a conciliatory approach. Deng supported Zhao’s resolution of the crisis, but he did not withdraw his earlier stance (DXPNP, 2004:1275; Zhao, 2009:21; Li, 2010:150–151). After the meeting, Yang reported the information to Li Peng, another signal of Deng’s neutral position. Zhao and Li had different interpretations of this meeting: Zhao did not find it to be unusual and interpreted it positively to his chief adviser, but Li sensed Deng’s dissatisfaction with Zhao (Chen, 2016: vol. 2, 447; Lu, 2019:1220). When the meeting information spread in the evening, some of Zhao’s supporters considered it a positive signal, since Deng had given Zhao verbal support (Zhang, 2009:162). Still, no explicit evidence shows that Deng had already decided to remove Zhao.

Conversational contingency and the final falling out

What drove Deng to make his final decision to break with Zhao? We find it was neither driven by the pressure from conservatives nor Deng’s premeditated plot. Rather, despite their different stances, their split became inevitable only after Zhao’s meeting with Gorbachev on May 16^th. Zhao’s misjudgment of the changing relationship structure led to his further overactions, which in turn caused Deng’s (mis)understanding of his intention and deadly (over)reaction.

By mid-May, Zhao still believed he had Deng’s trust and continued pressuring him to change his stance, even publicly. In a televised meeting with Gorbachev on May 16^th, Zhao revealed the “secret” that Deng was still China’s ultimate behind-the-scenes decision maker—a secret which “was not published but I [Zhao] am informing you [Gorbachev] about it today” (History and Public Policy Program Digital Archive, 1995). This remark was immediately regarded as a deliberate betrayal of Deng by relevant actors, since it implied that Deng was responsible for the mishandling of movement. Zhao immediately clashed with hardliners at the PSC meeting on the evening of May 16^th (Zhao, 2009:26–27; Li, 2010:161–164). His public remark moreover incited accusations about Deng’s gerontocratic dictatorship by intellectuals and protesters on May 17^th (History and Public Policy Program Digital Archive, 1999; Lu, 2019:1233–1235). Zhao later insisted that he had mentioned Deng out of goodwill (Bao, 2012; Zhao, 2009). However, even Zhao’s sympathizers considered his comments to invite a showdown with Deng (Zhang, 2009: 190–191; Chen, 2013:596; Chen, 2016, vol. 2:433–442). So did Zhao’s rivals (Li, 2010:161) and Deng’s family (Zhao, 2009:48; Lu, 2019:1229).⁴²

Surprised by Deng’s frustration with the comment, Zhao and his advisers attempted to arrange last-minute meetings with Deng to remedy their relationship. In the morning of May 17^th, Bao Tong received an alarming message about Zhao’s remarks and immediately contacted Deng Rong, Deng Xiaoping’s daughter and de facto secretary, to apologize for Zhao. However, Deng Rong rejected Bao’s request and sent back a message: “there is no need for more conversation (between us)… My father is ready to prepare to be overthrown for the fourth time” (Wu, 2013:505; Lu, 2019:1229). Meanwhile, Zhao called Deng’s office “to express my views personally, in a face-to-face meeting,” but Deng instead decided to convene a PSC meeting that afternoon, upsetting Zhao, who “realized that things had already taken a bad turn” (Zhao, 2009:27). From an interactionist viewpoint, the difference between a one-on-one conversation and a group meeting could not have been greater, signaling that any sense of trust was gone. The confrontation on Tiananmen and in Zhongnanhai co-evolved into the peak.

In the May 17^th meeting convened by Deng, Zhao again proposed to “somewhat relax the judgment from this [April 26^th] editorial,” but only his ally Hu Qili supported the revision. Other PSC leaders—Li Peng, Yao Yilin, and even Qiao Shi—pushed back against Zhao and criticized his moderate turn. As mistrust eclipsed their common ideological stance, policy preferences, or long-term friendship, Deng harshly criticized Zhao’s concessive approach and proposed martial law—despite Zhao’s opposition. After the split with Deng, Zhao’s support structure quickly collapsed, and he chose to resign (Chen, 2016, vol.2:291). For example, Yang Shangkun—who just coordinated the publication of a concessive editorial in the People’s Daily a few hours ago—“changed his position” during the meeting and took distance from Zhao afterwards (Lu, 2019:1229; Zhao, 2009:27–29; Li, 2010:168–174). Unsurprisingly, conservatives zealously joined Deng to dethrone Zhao. Military repression soon became “consensus” among party leaders. As Li Peng acutely noted in his diary, the May 17^th meeting “changed China’s destiny” (Li, 2010:168).

In hindsight, if the April 25^th Deng-Li conversation initiated this sequence of relational realignments, Zhao’s revelation to Gorbachev on May 16^th led to Deng’s final falling out with him (Brown, 2021: xiv). Many years later, Zhao still “believed that a series of misunderstandings about his intentions had caused the rapid breakdown of Deng’s trust in him” (Gewirtz, 2022:252). Yet Deng would not have gained such (mis)understanding without Zhao’s misjudgment of the changing relationship structure. In a conversation with the new leadership after the Tiananmen massacre, Deng also offered a sentimental reflection on his changing relationship with his chosen successors: “Since the establishment of [the second-generation] leadership, I have been continuing considering the succession matter. Although the two successors both failed to hold on … they were the only choices that I could make, not to mention people are changing.” (Deng, 1993: 309–10) After all, Zhao’s overaction and Deng’s overreaction co-produced their unexpected showdown. Thinking counterfactually, if Deng had not broken with Zhao, would he still have used the army to massacre the protesters? We argue that the movement would have ended quite differently. It might still have been suppressed, but there would have been far fewer casualties. Zhao and his aides could have played a stronger role in mediating between the state and protesters, facilitating a compromise between the two sides. As Chen Xitong, mayor of Beijing at the time, later reflected: the June 4^th incident is “an avoidable, should-be-avoided but not being avoided tragedy” (CXTQS, 2012:35).

In sum, ideological cleavage was an important but not the decisive factor in the Tiananmen collision; Deng’s role was critical, but he did not manipulate the split in advance. Rather, the collision resulted from key actors’ discrepant conceptions of the changing relationship structure altered in a short-term sequence of conversational interactions in April–May 1989. This “situational conflict” logic, we argue, can also be applied to explain other episodes of power struggle: in an opposite way, Jiang Zemin’s acute judgment of the changing situation and proactive conversational interactions following Deng Xiaoping’s 1992 Southern Tour reversed their relationship and rescued Jiang’s career.

Conclusion

Elite politics occur not only within institutional structures but also in relational networks that are reproduced via everyday interactions. Empirically, our article finds the variations of elite conflicts during China’s early reform era can be explained by brokerage and polarization embodied by the conversational network, while the actual unfolding of these conflicts often resulted from key actors’ discrepant understandings of the changing relationship structure arising from conversational interactions. Taken together, it is the relational and interactional effects of the conversation network that jointly produced the political outcomes.

Methodologically, our article makes two contributions to the historical sociology of elite politics: we apply longitudinal network analysis to examine an original conversational dataset. Most extant research measures elite relationship by static ties such as kinship relations or work overlaps, which cannot predict changing political alignment and choices. In contrast, we offer a dynamic, behavioral approach to measure elite relationship by quantifying their communications over time. We further measure characteristics of relationship structure such as network centrality, brokerage and polarization as processual properties so as to explain sequential political outcomes. Our approach has broad implications for the examination of elite communication data—such as informal conversations, correspondence, conference minutes, and even tweets—through network analysis, narrative, and conversation analysis.

We moreover make theoretical contributions by integrating two otherwise loosely connected fields: micro-sociology and historical sociology. First, we have analytically differentiated brokerage roles (robust actor, coordinator, itinerant, and gatekeeper/representative) and applied them with both quantitative and qualitative measures in explaining elite conflicts (Gould & Fernandez, 1989; Stovel & Shaw, 2012; Spiro et al., 2013; Obstfeld et al., 2014). Different from Padgett and Ansell’s conceptualization of brokerage as strategic agency possessing favorable structural positions in the networks, we have emphasized the processual nature that Leifer (1983) initially attributed to “robust action” and materialized this idea in our analysis of a well-known modern robust actor.

Second, we continue the theoretical agenda of micro-sociology by synergizing social network analysis and interactionalism and substantializing the notion that social networks are “processed in communication” (Fuhse, 2022:15–16) and recursively structure future communications. Without examining ongoing interactions, previous structural network analysis tends to assume relationship patterns determine people’s choices and behaviors, leaving little space for contingency in vital moments (see critique: McLean, 2007:11–14). Our theorization of conversation networks has mixed the relational and interactive aspects together to offer adequate micro-foundations for analyzing both the structural and the contingent facets of elite politics.

Third, we demonstrate the underexplored value of micro-sociological insights for study of historical change. Historical sociology should not be limited to macro-level categories such as the state and regime; nor is micro-foundation only about rational choices or utilitarian strategies. In line with previous studies, we underscore analytical sensitivity to the micro-level interpersonal relations and interactions that underpin sociopolitical structures and macro-historical outcomes (Ermakoff, 2015; Gibson, 2012; McLean, 2011; Slez & Martin, 2007; Xu, 2013; Zhang, 2021).

It would not be surprising to apply our approach to other monarchical or authoritarian contexts, where informal interactions of elites often trump formal institutions or procedures. Do elite interactions such as informal conversations also matter in genuinely two-party or multi-party systems? In democracies, parties function as institutional containers, fixing politicians’ ideological positions, policy preferences, and legislative behavior. Yet behind those formal institutions and processes, elite politics unfold via informal relationships, extra-state institutions, and everyday interactions. In the nineteenth century, for example, the interactions of American congressmen in Washington DC boarding houses—an “extra-institutional” site—fostered informal relationships that affected their voting preferences (Parigi & Bergemann, 2016). Today, when institutional partisanship is eroded by strongmen and populism in democracy and becomes increasingly fluid, we believe our approach can yield valuable insights into the understanding of contemporary politics as much as historical events like the taking of the Bastille and the Tiananmen massacre.

Appendices

Appendix 1. Background information of the PSC-level leaders during the early reform era

Table 3

Table 3 All 24 PSC-level leaders in China in 1977–1992 (ranked by birth year)

Appendix 2. The timeline of meetings during the Tiananmen movement

Figure 10

Fig. 10

a Meetings of Zhao Ziyang and Li Peng during the Tiananmen movement, 4/15–4/30. b Meetings of Zhao Ziyang and Li Peng during the Tiananmen movement, 5/1–5/17. Sources: ZZYZNHSNJS, 2005; Li, 2010; Chen, 2016; Wu, 2019:108–519; Lu, 2019:1184–1251