Factions in Nondemocracies: Theory and Evidence from the Chinese Communist Party

Transcription

1 Factions in Nondemocracies: Theory and Evidence from the Chinese Communist Party Patrick Francois 1 Francesco Trebbi 2 Kairong Xiao 3 1 University of British Columbia, CIFAR 2 University of British Columbia, CIFAR, NBER 3 Columbia University Political Economy I, Winter 2018

2 2 Motivation Political economy of development requires understanding nondemocratic regimes. Many of these regimes are autocracies which play important economic & geopolitical roles internationally. Single-party regimes are historically common (e.g. Falange in Spain; Communist Party of the Soviet Union in USSR; FRELIMO in Mozambique; KANU in Kenya; Ba'ath Party in Syria; Ba'ath Party in Iraq; National Liberation Front in Algeria; National Fascist Party in Italy) But data is sparse; internal dynamics are opaque Example: Chinese Communist Party

3 3 The Chinese Communist Party (CCP) The founding & ruling political party of the People's Republic of China Leninist constitutional architecture: Party = State Political linchpin of the second largest economy & the most populous country in the world today One of the oldest (founded in 1921) and largest by membership (89 million members in 2016) political party in the world

4 4 The Chinese Communist Party (CCP) A secretive, selective organization of about 65 million members who have positions of influence in all sectors of Chinese society Nathan and Gilley (2003, p.7)

5 5 Questions Is there anything systematic we can say about this type of regimes? What is the role of factions within the party? What are the incentives of individual politicians within a faction? Can economic theory guide us in producing reasonable predictions about current political equilibria in Chinese elite politics?

6 6 What we do Stylized facts on internal organization of the CCP. An organizational economics model of a single-party regime with competing factions & career motivated politicians designed to match empirical moments. Quantitative exercise: 1. Structural estimation of the model; 2. In-sample & out-of-sample performance; 3. Model selection tests; 4. Counterfactual analysis; 5. Critical juncture for China: Forecasts for 19 th Party Congress October 18-25th Is Xi Jingping a step towards personalistic autocracy?

7 7 Literature Review 1. Politico-economic literature on Chinese elite politics Nathan (1973); Pye (1981); Dittmer and Wu (1995); Dittmer (1995); Nathan and Gilley (2003); Shih (2004); Shih, Shan, and Liu (2010); Li (2012); Li (2013); Shih (2016); Meyer, Shih, and Lee (2016); Shih and Lee (2017) - On provincial promotions specifically: Li and Zhou (2005); Shih, Adolph, and Liu (2012); Jia et al. (2015); Fisman et al. (2017) 2. Factions in parties Persico, Rodriguez-Pueblita, and Silverman (2011); Dewan and Squintani (2015) 3. Internal organization of autocratic regimes Tullock (1987); Bueno de Mesquita et al. (2003); Besley and Kudamatsu (2008); Francois, Rainer, and Trebbi (2015; 2016); Acemoglu, Egorov, Sonin (2008)

8 8 Institutional Background: The CCP Hierarchy Top 2 Politburo Standing Committee (7) Politburo (25) Central Committee (205) Alternate Central Committee (171) Provincial/Local Party Organizations (88.8 millions)

9 9 Institutional Background: Diarchic Structure Position Number 1 Number 2 Politburo Standing Committee Central Military Commission Supreme People s Court Provincial Government General Secretary/ President Chairman President Provincial Party Secretary Premier Executive Vice Chairman Executive Vice President Governor

10 10 Institutional Background: Diarchic Structure National Provincial Provincial City-level City-level City-level City-level

11 Data Two biographical databases of Chinese politicians China Vitae: 4,494 Chinese elites since 1992 Shih, Shan, and Liu (2008): All CC and AC members since 1921 Four affiliation indicators Communist Youth League of China (CYLC or tuanpai) Shanghai Bang (Gang) Princelings Military 11

12 Institutional Background: Factions in current PB Table: Factions of 18 th Politburo Members (out of 25 members) Shanghai Gang CYLC Xi Jingping (General secretary/president of PRC) Yu Zhengsheng (Chairman of CPPCC) Wang Huning (Director of CCP Policy Research Center) Meng Jianzhu (Secretary of the Commission of Political Science and Law) Han Zheng (Shanghai party secretary) Li Keqiang (Premier) Li Yuanchao (Vice president of PRC) Liu Qibao (Head of CCP Propaganda Department) Liu Yandong (Vice Premier, State Council) Hu Chunhua (Guangdong party secretary) 12

13 13 Stylized Facts National Provincial Provincial City-level City-level City-level City-level

14 14 Stylized Fact 1: Cross-Factional Mix National Provincial Provincial City-level City-level City-level City-level

15 15 Stylized Fact 1: Not Factional Strongholds National Provincial Provincial City-level City-level City-level City-level

16 Stylized Fact 1: Cross-Factional Mix 16

17 Stylized Fact 2: Leadership Premia in Power Scores (Bo, 2010) 17

18 18 Other Stylized Facts 1. Shanghai Gang & CYLC are national political players (i.e. they are present in every province in substantial number) 2. Post-Deng period (1992-present) is a stable political environment. We can only focus on this Moreover: 3. Leadership premia present in promotion rates of factional members (consistently with power shares result) 4. Balance of punishment in anti-corruption/purge campaign

19 19 Model Our attempt is to produce a tractable model of cadre behavior: a) Microfounded & based on individual incentives; b) Matching all reduced-form factional patterns simultaneously & matching the promotion profiles observed within the CCP in the data; c) Falsifiable; d) Useful for forecasting.

20 20 Model: The Party Hierarchy L level hierarchy. Each level l has M(l)/2 leadership nodes Each leadership node has two positions, No.1 and No.2 Career concerns. Instantaneous utility at level l: u l > u(l + 1) Continuos time. Each individual politician dies with instantaneous probability, δ The opening at level l is filled by applicants from the level immediately below, l + 1

21 21 Model: State Space & Promotion Chain National Provincial Provincial City-level City-level City-level City-level

22 22 Model: State Space & Promotion Chain National Provincial Provincial City-level City-level City-level City-level

23 23 Model: State Space & Promotion Chain National Provincial Provincial City-level City-level City-level City-level

24 24 Model: State Space & Promotion Chain National Provincial Provincial City-level City-level City-level City-level

25 Model: Factions Two factions B (Blue) and R (Red), the remaining N (Neutral) Factions are a quid-pro-quo contract A faction member receives support in promotions If the paramount leader is from his faction, the member will receive additional support If he eventually becomes the paramount leader, this faction member will provide the same support to the juniors When there is an opening Contest for promotion: faction supports only one member A faction member not supported by his faction cannot win 25

26 Model: Promotion Contest Blue 1 Blue 2 Blue Candidate Red 1 Red 2 Red Candidate Open Position Neutral 1 Neutral 2 Neutral Candidate Neutral 3 26

27 Model: Promotion Contest Blue 1 Blue 2 Blue Candidate Red 1 Red 2 Red Candidate Open Position Neutral 1 Neutral 2 Neutral Candidate Neutral 3 Within-faction Selection Between-faction Contest 27

28 Model: Between Faction Contest Tullock (1980) contest function: W I = i β + ρ + η where i = β, if I = B; i = ρ, if I = R; i = η, if I = N. If the paramount leader is from faction B, then β l > β, & if from faction R, then ρ l > ρ. The leadership premium is therefore λ B β l /β or λ R ρ l /ρ. The total contest weight of neutral is η. 28

29 Model: Within Faction s Support in Promotion The collective interest of a faction To increase the probability of attaining the paramount leadership To have as many members as possible Individual incentives To increase own chances of future promotion The closest co-factional in the hierarchy will always support its subordinate co-factionals unless the subordinate competes for a position in the same level, in which case, the subordinate will be asked to wait. Implies natural pecking order within faction. 29

30 30 Model: Seniority in Faction and Localized Support National Provincial Provincial City-level City-level City-level City-level

31 31 Model: Seniority in Faction and Localized Support National Provincial Provincial City-level City-level City-level City-level

32 32 Model: Seniority in Faction and Localized Support National Provincial Provincial City-level City-level City-level City-level

33 33 Model: Seniority in Faction and Localized Support National Provincial Provincial City-level City-level City-level City-level

34 Additional Stylized Fact: Evidence of Slowdowns 28

35 35 Model: Individual Decision Individual politician chooses the support for his cofactionals in a promotion within his jurisdiction to maximize his discounted utility: V t = t e δs v s ds

36 36 Model: Equilibrium Strategies Proposition 1: i) A politician from faction J at level l will limit the support of a cofactional member ascending to his level from l + 1 at t if there are members of both I J and neutrals, N, at level l. ii) If there are no members of faction I at level l, a politician from faction J I at level l will support a member of his own faction from l+1 at t if the number of cofactional members at level l is such that J t l > j+η i where i = β, j = ρ if J t = R, and i = ρ, j = β if J t = B.

37 37 Model Prediction: Arrival Rate of Promotions Proposition 2: The instantaneous arrival rate of promotions at each level of the hierarchy is as follows: Let 1 J t = 1, iff J t l > 0 and 1 J t = 0, otherwise for iff J = B,R,N. For an N member: R t l 1 δ + δ R p t l 1 1 B t β + η + δ N t l = η N t l N t l 1 δ + δ N p t l 1 1 B t β + 1 R t ρ + η + B t l 1 δ + δ B p t l 1 1 R t ρ + η δ I p t l 1 : the instantaneous arrival rate of a promotion for a politician sharing a node with an I politician at level l 1 I t l : the number of positions held by faction I at level l, I {R, N, B}, at time t. (Details)

38 38 Model Prediction: Arrival Rate of Promotions Proposition 2 (continued): For a B member: δ B t l = β B t l R t l 1 N t l 1 p δ + δ t R l 1 β + 1 t + N η p δ + δ t N l 1 β + 1 t R ρ + 1 t N η The instantaneous arrival rate of promotions for a R member is defined in a similar way. δ I p t l 1 : the instantaneous arrival rate of a promotion for a politician sharing a node with an I politician at level l 1 I t l : the number of positions held by faction I at level l, I {R, N, B}, at time t. (Details)

39 39 Model: Entry At time t at the entry level, an entering politician chooses the faction with the highest expected utility stream: max I {B,R,N} Et V t B L, E t V t R L, E t V t N L where the expected utility stream is determined by the continuous time Bellman equation δe t V I t l = u l + δ I t l E t V I t l 1 V I t l

40 40 Model Prediction: Equilibrium Entry Behavior Proposition 3: with M(l) large enough for all l, any equilibrium necessarily involves politicians in factions B, R, and N.

41 From Model to Data The model is in continuous time The data is observed at low frequency, i.e. five CCP National Congresses from 1992 to 2012 Promotion chains and their order are latent Simulation methods to link two subsequent state variables S T, S T+1 1. A group of individuals retired between two snapshots T, T Randomly select one retired individual, start a promotion chain 3. Update S T to S T at each move to reflect the promotion 4. Repeat step 2 and 3 for all retired individuals until S T =S T+1 5. Repeat step 4 for R K times with randomized order of retirement 41

42 Maximum Simulated Likelihood Estimation The unconditional likelihood of observing Y given X is: R K C L f Y X = δ R K r=1 J p l 1 I l (l) c=1 l=l 0 Where Y is the observed data on career outcomes; X is the observed data on the hierarchy plus a set of individual characteristics; C is number of chains part of a set of chains that can rationalize Y; R K is number of simulated sets of chains. This is the Maximum Simulated Likelihood (MSL) estimator. (Details) 42

43 MSLE Parameter Estimates 43

44 44 Alternative Specifications Faction is more important for promotions at higher CCP levels than that at lower levels Magnitude of leadership premia are similar across factions

45 45 Actual & Predicted Seat Shares Note: Mean model seat share prediction in 100 simulations.

46 Alternative Promotion Models Faction models: Baseline model Adding individual covariates (details) Alternative models: Random promotion Pure seniority-based promotion Meritocracy Factional balance across the hierarchy as driver of promotions [next] 46

47 47 In-Sample Fit: Faction Model vs. Alternative Models (Share of total promotions by faction)

48 48 In-Sample Fit: Faction Model vs. Alternative Models (Share of total promotions by faction)

49 Out-of-Sample Fit: Faction Model vs. Alternative Models (Share of total promotions by faction in 18 th Congress) Note: Model estimated on Party Congresses 14 th -17 th. 49

50 Specification Tests Against Alternative Models 50

51 Summary of Alternative Promotion Models Faction model outperforms alternative models in fitting promotion dynamics in Mean Squared Error: in sample and out of sample Model selection tests - LR tests for nested models; Vuong tests for non-nested models - can reject alternative models against faction model 51

52 52 Summary of Counterfactuals We produce several counterfactuals, e.g.: More winner-take-all leadership in Deng s design [Nathan (2016) suggests that Xi may be overturning Deng s system of collective leadership]. Double leadership premia λ. Counterfactual General Secretary. Current premier Li Keqiang instead of Xi Jinping.

53 Are Princelings a CCP Faction? Debate among scholars Bo (2008) argues that princelings are not a political faction Li (2013) defines Xi as a princeling associated with Jiang s camp Our model allows for a formal test. Modified contest function: W P = π β + π + ρ + η 53

54 Are Princelings a CCP Faction? 54

55 Is President Xi Jinping Affiliated to the Shanghai Gang? 55

56 56 Out-of-Sample Forecast: 19th Party Congress, October 2017 (Share of total promotions) Note: Factional model + individual covariates estimated on Party Congresses 14 th -18 th. Average of 100 simulations.

57 57 Politburo Forecast We are collecting the rest of the data, but results so far encouraging. E.g. Predicted Shares 24%, 66%, 9% for N, B, R, respectively, versus Actual Shares 25%, 65%, 10%

58 58 Conclusions Understanding inner workings of autocracy crucial to political economy of development We study elite organization in a large & complex nondemocracy: China Stylized facts: Leadership premia, Cross-factional mix support a factional perspective Propose a microfounded model of CCP internal organization and estimate it Promotion mechanisms induce specific slow-moving dynamics of elite composition Counterfactuals of institutional changes under Xi. Relevant as 2017 is crucial turning point in Chinese political equilibria (Shih and Lee, 2017). Evaluation after October 18 th -25 th Other applications to complex nondemocracies, e.g. Russia

59 59 Appendix: Instantaneous Arrival of Promotions The instantaneous arrival of promotion for a I {R, N, B} from level l is δ I l = R l 1 δ + δ R p l 1 p I R l +N l 1 δ + δ N p l 1 p I N l +B l 1 δ + δ B p l 1 p I B l δ I l : the instantaneous arrival rate of a promotion for an I politician at level l δ I p l 1 : the instantaneous arrival rate of a promotion for a politician sharing a node with an I politician at level l 1 p I J l : the probability that an I faction member at level l is promoted to an opening paired with a J faction member at level l 1 I l : the number of positions held by faction I at level l, I {R, N, B}

60 60 Appendix: Probability of Promotion Conditional on Opening p I J l : the probability that an I faction member at level l is promoted to an opening paired with a J faction member p I J l 1 I l i i + k + k if K l, K l > 0 i i + k if K l > 0, K l = 0 i i + k ifk l = 0, K l > 0 1 ifk l = 0, K l = 0 Where I, i, K, k, K, k are B, β, R, ρ, {N, η} with respectively. (Back)

61 61 Appendix: Maximum Simulated Likelihood Estimation Y: the observed data on career outcomes X: the observed data on the hierarchy plus a set of individual characteristics k: a set of promotion chains, k = k 1, k 2,, k C The unconditional likelihood of observing Y given X is: where f Y X = E k f(y X, k) = E k f Y k c X k c, k = δ C c=1 L l=l 0 f(y k c X k c, k) J p l 1 I l (l)

62 62 Appendix: Maximum Simulated Likelihood Estimation Many sets of promotion chains consistent with Y But we do not observe which set of chains actually occurred and the order of occurrence within the set We simulate R K sets of chains, and the Maximum Simulated Likelihood (MSL) is f Y X = 1 R K C f(y R k c X k c, k) K r=1 c=1 (Back)

63 63 Appendix: Adding Individual Covariates So far we treat within-faction selection as random Define q I,s (l) as politician s of faction I is selected as candidate q I,s (l) = 1 I l We can add individual characteristics, Z, to the within-faction selection exp(γz s ) q I,s (l) = s A I (l) exp(γz s ) A I (l) is the set of the members of faction I at level l. (Back)