Political Selection Institutions and Policy Performance: Evidence from China

Abstract

Political selection institutions in non-democracies are usually conceptualized as mechanisms to co-opt competent agents for regime survival. Departing from this common emphasis, this article highlights their linkage function between informal politics and policy outcomes. Using multilevel modeling and error correction models, hypotheses on the determinants and implications of formal political selection rules are tested. Drawn from an original dataset of political selection rules in China, this analysis finds that coalitions with particular policy priorities strive to achieve desired policy outcomes through shaping formal political selection institutions. The geographic variation in specific features of the political selection rules is primarily driven by coalitional politics. In addition, the effect of policy performance on local leaders’ promotion prospects is not uniform but conditioned on the political selection rules. Under such incentive arrangements, local leaders are found to expand government spending in the policy area prioritized in formal political selection rules. These findings advance our understanding of the endogenous political nature of political selection rules and the relations between informal institutions and policy performance.

1 Introduction

If the Central Committee is deprived of the right to direct [rasporyazhat’sya] the allocation of personnel, it will be unable to direct [napravlyat’] policy.

--- Vladimir Lenin at the XI congress in 1922 (Rigby 1988).

The quality of politicians is essential in political life. V.O. Key argues that the nature of the workings of government depends ultimately on the men who rule it and more emphasis shall be placed on those we elect to office (Besley et al. 2005). Factors, ranging from the rents that politicians can earn while in office to the incumbency veto power, affect the supply of bad politicians (Caselli and Morelli 2003; Acemoglu et al. 2010; Gelbach et al. 2010). In particular, Egorov and Sonin (2011) point out an inherent “loyalty v.s. competence” trade-off in non-democracies. Because a more competent advisor is more prone to treason, the dictator, fearful of being betrayed, is unwilling to surround himself with competent associates, which causes the poor performance of dictatorship. However, such an argument is contradicted by recent studies on political selection in China suggesting that autocracies are more likely to promote technically competent officials than democracies (Bell 2015; Lee and Schuler 2020). Scholars also posit that the Chinese Communist Party (CCP) can effectively balance loyalty and meritocracy by devising distinct selection criteria for officials across different government levels.¹ This article contributes to the debate by further unpacking the intricate relationship between loyalty and competence in the Chinese context. Instead of suggesting personal connections and policy performance (i.e., meritocracy) are separate or at odds with each other, this research shows that the utility of competence and performance is endogenously determined.² Moreover, the type of competence that contributes most to promotion is contingent on the personal ties of rule-makers. In addition to replacing (Shih et al. 2012) or complementing (Jia et al. 2015) work performance in determining career advancement at the individual level, personal ties can enlarge or reduce (i.e., tune the magnitude of) the effects of particular types of performance on promotion prospects through shaping the formal performance evaluation rules.

Scholars are keen on political selection in China, yet remain divided about whether personal connections or policy performance is the key to career advancement for Chinese officials (Li and Zhou 2005; Shih et al. 2012). A large part of the confusion is attributed to our limited understanding on the intricate relationship between these two as well as on the decision-making process in personnel arrangement. A nuanced analysis on the formal political selection rules can help bridge the gap. Existing research on the target responsibility system (TRS), the written and official evaluation rules in selecting political leaders, places most attention on local officials’ strategic responses to fulfilling targets, such as data manipulation, selective policy implementation, and collusive behaviors (Gao 2009, 2015; Li 2015; Kostka 2016; O’Brien and Li 1999). Contributing to our understanding of the understudied target-setting process,³ this article starts from exploring the factors underlying the variation and shifts in performance targets. It then proceeds to empirically demonstrate the impacts of political selection rules on actual personnel arrangement and on local leaders’ spending decisions,⁴ revealing the real-world impacts of informal politics on policy performance and local leaders’ competence stocks.

Here, the political selection institution and the nomenklatura system are used interchangeably. This analysis first makes clear the endogeneity of political selection institutions, and then demonstrates their policy implications through answering the following questions. First, how do we account for the variation in formal political selection rules and whether informal politics exerts any impact on the development of these rules? Second, what effects do formal political selection institutions exert on local policy? Informal institutions are defined as the unwritten, unofficial rules and procedures that constrain and enable actors’ behavior. In the realm of policy, coalitional politics remains the most relevant member of informal politics, and it is influential on varieties of political and economic outcomes in China.⁵ Drawn from an original dataset of political selection rules, this analysis finds that the variation in the content of provincial political selection rules is associated with provincial leaders’ coalitional affiliations. Additionally, these formal selection rules are effective to align subordinate city leaders’ policy priorities and spending decisions with those of provincial governments. Formal selection rules can achieve such an effect because local leaders are promotion-seeking and these formal rules are indeed followed in actual personal arrangements: those who pursue the policy priority reflected in formal selection rules are rewarded with better career prospects.

This paper contributes to the literature on political selection institutions in comparative politics as well as on cadre management in Chinese studies. Linking authoritarian institutions to outcomes beyond regime survival as well as sophisticated studies on politics within authoritarian institutions are still rare and in urgent need of expansion (Pepinsky 2013). In this article, I take up this suggestion by linking institutions to social spending and examining the impact of coalitional politics on the creation of formal institutions in contemporary China. Although the empirical analysis is restricted to China, the results have broader implications for understanding the impact of political selection institutions on policy and ultimately on communist party rule. As the party links policy performance to promotion, the “sunk investment” on the part of party members pressures them to generate policy outcomes desired by the top (Svolik 2012). Thereby, political selection rules can have broad policy implications beyond the mere control over the ruling coalition. In addition, by examining the determinants and implications of political selection rules, this research provides two novel findings regarding the communist rule: first, factionalism is not necessarily at odds with meritocratic recruitment, and the finding is consistent with recent studies stressing the positive role of factions in politics (Boucek 2009; Dewan and Squintani 2016); second, apart from relying on the patronage network as shown in Soviet studies (Rigby 1970; Willerton 1987), this analysis indicates that communist elites can also promote policy initiatives on the ground through adapting the nomenklatura system and strengthening performance-based promotion.

Mixed methods, including qualitative interviews, reading of Party documents, and statistical analysis, are used to establish the arguments. This article proceeds as follows. The next section presents the theory and hypotheses. Section III first describes formal political selection rules, and then investigates the political factors that drive the variation in the specific institutional features. Section IV examines the effectiveness of these rules within personnel arrangements and the impact of political selection rules on local spending policies. The last section discusses the results and concludes with implications.

2 Authoritarian Institutions and Policy

Only recently have scholars started to examine which and how autocratic institutions mediate policy processes and outcomes. Gandhi (2008) argues that autocratic legislatures facilitate policy compromises by providing a forum for bargaining and information-sharing, and shows that non-democracies with legislatures have greater civil liberties and lower levels of military spending. Miller (2015a) finds that multiparty autocratic elections produce better human development outcomes than non-electoral autocracy, and ruling parties use the information revealed by the electoral outcomes to calibrate policy concessions that prevents social revolts (Miller 2015b). Scholars have also highlighted the impact of informal institutions on policies. For example, Mccubbins and Thies (1997) posit that different factional combinations within Japan’s Liberal Democratic Party ruling coalition have divergent policy priorities and pursue distinct spending policies. In addition, scholars search for mechanisms by which informal institutions affect the creation of formal rules that shapes policy outcome. For example, Tsai (2016) and Grzymala-Busse (2010) examine how interactions between various state and non-state actors or elite competition generated by informal rules shape the development of formal institutions in communist and post-communist regimes. Therefore, I hypothesize that informal politics affects the design of formal political selection rules, and provincial elites’ coalitional tie shapes the policy priority reflected in political selection rules.

In particular, the within-party populist coalition that consists of former Chinese Communist Youth League officials (CYLs) is regarded as Hu Jintao’s power base and holds contrasting policy priorities from other coalitions (Li 2012). Different from Jiang Zemin’s followers and other officials who move up their career ladder during the economic reform era, CYLs is argued to be more welfare-oriented. Thus, the first hypothesis to be tested is that provinces with more CYLs in their party standing committees are more likely to assign more points to social welfare targets than to economic ones in TRS (H1).

Maintaining social order is an imperative target and the occurrence of massive protests annuls achievements in meeting other targets. Popular pressure becomes increasingly influential in the policy agenda setting (Wang 2008). When there is a greater intensity of social conflict, local leaders are willing to incorporate public demands into policy-making (Distelhorst and Hou 2017; Meng et al. 2017). Thus, a likely alternative hypothesis is that a greater intensity of social unrest is correlated with more points assigned to social welfare targets in TRS (H1a).

The political selection institution is mostly characterized as a co-optation instrument (Svolik 2012). However, it also holds deep policy relevance. As shown by Kung and Chen (2011), excessive grain procurement during China’s Great Leap Famine is driven by local leaders’ career incentives structured by political selection rules. Political selection rules are expected to be more closely tied to policy implementation at the lower administrative levels (Landry et al. 2018). The ability of political selection rules to align lower governments’ policy priorities with those of upper level governments is widely assumed without a systematic examination. Government spending is commonly used to capture policy priority (Gandhi 2008; Miller 2015b; Mccubbins and Thies 1997). I argue that well-enforced formal political selection rules, through shaping local leaders’ perceptions regarding whether and what policy performance matters most for promotion and thereby their policy priorities, affect local government spending. This generates two related hypotheses:

Hypothesis 2.1 (H2.1): Performance-based selection rules are well enforced: all else equal, the higher priority given to social welfare in provincial political selection rules, the greater the positive effect of performance in social areas on subordinate city leaders’ promotion prospects.

Hypothesis 2.2 (H2.2): Well-enforced political selection rules affect the policy priority of local leaders: all else equal, in subordinate prefectures where the provincial political selection rules attach a higher priority to social welfare, social spending is more likely to expand and to expand faster.

3 Adaptation and Underlying Forces of the TRS

The political selection institution is the key instrument for maintaining political control in this economically decentralized system (Manion 1985; Edin 2003; Whiting 2006; Landry 2008; Birney 2014). A regular evaluation system for local party and government leaders was instituted in 1979 with the aim to select competent officials for economic recovery after the Cultural Revolution. Since early 1980s, local leaders have been evaluated according to TRS, which provided quantifiable performance targets (see Online Appendices Table 1). The rules also specified the points assigned to each target in the scoring metrics. Autonomy is granted to provincial governments in setting the content, value and point of targets to fit local conditions. Local party committees mostly designate performance targets into several categories that can be regrouped into three blocks: economic, social, and political targets (see Online Appendices Table 1). In determining which targets are “social”, “economic” or “political”, I rely on official categories.⁶ The benefit of using these official categories vis-à-vis a manual coding is that they explicitly tell which targets are perceived by local governments as “economic” or “social”. The designation of specific targets into official categories is also consistent across localities, which makes the target categories comparable. Out of a total of 100 points, “political” targets, mostly about party building (dangjian), generally do not exceed 20 points.

As an official from the Central Organization Department described: “TRS evaluation rules are very important because they play a role of signaling and orienting. They convey credible information about what higher level authorities value” (Interview BJ10311201). However, scant attention has been devoted to the factors underlying the allocation of target points that signal policy priority. Different from target values that permit negotiations and inputs from subordinate units, TRS points are allocated in a top-down fashion and are impossible to negotiate over.⁷ According to interviews, TRS points are first drafted by the Provincial Office of Targets and Performance (POTP) which is established under the party organization department or the general office of the provincial government and is composed of some party standing committee members and heads of key provincial government departments. The draft is then submitted to the provincial Party standing committee for finalization (Interview JX19111502, Interview HN17121701). Based on the interview (Interview JX19111505) and the sample of evaluation regulations mentioned below, the target points are adapted and small changes are made every year.

Existing research finds that more points are assigned to economic targets and thus a bias toward economic development within the TRS (Edin 2003; Whiting 2006). The shift of ruling priorities in Beijing, from all-out economic growth in the Jiang era (1989–2002) to a more pro-people ideology under the Hu-Wen leadership (2003–2012), provides a unique opportunity to observe changes in the local TRS point distribution, if any. In particular, the 2006 and 2009 national guidelines for evaluating local leaders reduce the emphasis on economic targets (see Online Appendices Table 2). Post-2006 local personnel rules accordingly become more welfare-oriented in three ways: (1) some welfare indicators have become imperative targets; (2) the points of welfare targets in performance evaluation are rising and in some places even exceed those of economic targets (see Fig. 1); and (3) local governments start to reward local leaders who make achievements in social policy areas. Some local leaders question economic performance as the sole contributor to promotion.

Fig. 1

Source: Author’s dataset

Point Differences between Social Targets and Economic Targets in TRS for Prefecture-level leaders (2008–2012). This figure presents data in the following province-years: Guangdong (2008), Guangxi (2008), Yunnan (2008), Shandong (2008), Liaoning (2008), Shanxi (2008), Hubei (2011), Anhui (2011), Hunan (2012), Jiangxi (2012), Shaanxi (2012), and Gansu (2012).

“Performance in welfare enhancements affects the legitimacy of CCP rule. I am very confident that higher level authorities appreciate (xinshang) local leaders who focus on welfare improvement” (Interview YN12031201).

“Political promotion is not solely based on economic performance. In our locality, many local leaders were promoted because of welfare reforms” (Interview SX09051201).

Interestingly, provincial leaders respond differently to the priority shift: out of a total of 100 points, Guangdong and Shandong provincial party committees assigned 30 points more to social targets than to economic targets in their provincial TRS, whereas economic targets still received 10 points more than that of social targets in Jiangxi and Anhui provinces (see Fig. 1).⁸ Such variation in local TRS scheme is puzzling.

3.1 Data and Methods

Systematic studies on the nomenklatura system are rare due to political sensitivity and difficulties of gaining access to documents (Gao 2009). This empirical analysis relies on an original dataset of political selection rules (2003–2013) collected from field sites and government websites. Evaluation of local leaders mostly follows the one-level down pattern: party departments evaluate their immediate subordinates. In this analysis, I rely on TRS information from 60 provincial regulations in 16 provinces, covering city leaders in two thirds of all prefecture-level cities (see Online Appendices Fig. 1). The nature of data for all subsequent analyses is pooled cross-sectional. As most statistical analyses of Chinese politics do, ethnic autonomous prefectures are excluded from analyses due to a lack of consistent data.

Because the sample and non-sample groups are not independent of each other, I use bootstrap t-test to compare whether there are any statistically significant differences between these two groups (see Online Appendices Table 3.1). No systematic difference is found between prefectures in the sample of provincial regulations and non-sample ones. Yet, the 16 sample provinces are systematically different from the 15 non-sample provinces: they are economically worse-off, having a larger population size with slower population growth, allocating a greater proportion of government expenditure to social welfare, and having poorer welfare outcomes. Whether findings in this analysis can be generalized to economically better off provinces thus requires future investigation. These 16 provinces by no means constitute a representative sample, yet they exhibit considerable variation in socioeconomic conditions (see Table3.2 in Online Appendix) that are potentially correlated with evaluation target making, leaders’ promotion prospects and spending priorities. Bootstrap t-tests are also conducted within each province to test whether the sampled prefectures are systematically different from their non-sampled counterparts. 5 provinces out of the sampled 16 provinces have all subordinate prefectures covered in the sample. For the remaining 11 provinces, there are systematic differences in a few indicators within 6 provinces (see details in Table 3.3 in Online Appendix). Yet, no evidence indicates prefectures with certain characteristics are systematically oversampled.

To test H1, I use both a simple province-year random-effect model and the multilevel modeling because prefecture-level leaders within the same province can receive different TRS point schemes.⁹ For the second multilevel modeling, the dependent variable (Y) is constructed as the ratio of social target points over economic ones in formal TRS metrics for prefecture-level leaders, which is called TRS point ratio hereafter. The sample consists of TRS metrics for 205 prefectures at level 1, nested within 16 provinces at level 2 and 11 years at level 3. The likelihood test indicates that the three-level model is a better fit than the two-level model (with only prefectures nested within provinces), and the variance at the provincial-level accounts for 80 percent of the total variance in the outcome variable. Likelihood ratio tests between the model with and without the random slope for each variable indicate a lack of any cross-level interaction effects. Thus, the general form of the three-level modeling is a random intercept model, as follows:

$$ Y_{ijt} = \alpha_{jt} + \beta X_{ijt} + \varepsilon_{ijt} \;({\text{Level - 1}}\;{\text{Model}}) $$

$$ \alpha_{jt} = \gamma_{t} + \delta W_{jt} + u_{jt} \;({\text{Level - 2}}\;{\text{Model}}) $$

$$ \gamma_{t} = \mu + \partial M_{t} + e_{t} \;({\text{Level - 3}}\;{\text{Model}}) $$

where subscript i denotes prefectures, j denotes provinces, and t denotes years. The explanatory variable of interest is the level-2 or provincial-level variable that captures coalitional politics within the provincial rule-making body. CYLs’ strength is gaged by the portion of leaders within the provincial party standing committee who held CYL leadership positions at the provincial level or above between 1982 and 1985, or between 1992 and 2002, when Hu Jintao was in charge of the CYL and its related xitong.¹⁰ Alternative measures of coalitional ties, that is provincial leaders’ ties with top leaders (including Jiang Zemin and Hu Jintao), are measured using Shih’s method (Shih 2008a): if the provincial party secretary or governor shares the same birthplace, college, or work unit with the top leader, then the binary ties with the top leader variable is coded as “1”.

For the simple provincial-level model, the hausman test indicates that random effects (RE) model is preferred over fixed effects model. The simple province-year random-effects models (see models 1–3 in Table 1) only include provincial-level variables. Provinces with differentiated TRS point distribution for subordinate cities are thus dropped. Substantive results are similar to those of three-level regressions.

Table 1 Factors correlated with TRS point distribution (2003–2013).

To test H1a, I use a dataset from news reports on local protests gathered by the Chinese Academy of Social Sciences (CASS).¹¹ Using media reports to estimate social instability will likely introduce measurement errors, since the media reports are conditioned to local media environments. However, media reports’ data reflect the expressed grievances heard by a wide audience, and thus provides a valid proxy for societal and political pressure imposed on local leaders. Control variables include stationary city-level socioeconomic variables, including GDP per capita and population size.¹² Provincial-level control variables contain economic structure, the gross dependency ratio, and the existing social welfare ranking that are likely related with priority to social welfare. All explanatory variables are lagged to reduce the endogeneity problem. T-tests are also conducted to test whether officials who are a high-flyer for various reasons, including having strong personal networks, are appointed to leading positions of particular localities that can more easily produce better work performance in the first place. No systematic differences are found in the size and the growth of GDP per capita, or the size of revenue in the year prior to their tenure between the promoted prefectural leaders and non-promoted ones (see Table 3.4 in Online Appendix).

3.2 Results

Results are presented in Table 1. Provinces with more CYL leaders in their party standing committees assign more points to social targets relative to economic ones in TRS rules. Variation in CYL strength accounts for 40 percent of variation in the TRS point ratio. Provinces with more

CYLs in their leadership follow Hu’s pro-people ideology more closely, reflected in the distribution of TRS points. Every ten-percentage-point increase in CYLs’ strength is associated with around ten-percentage-point increase in the ratio of social target points over economic ones. Pearson’s correlation test indicates that the CYL strength is not correlated with the prior level of wealth, nor with that of social welfare, which rules out the possibility that CYL officials are appointed to less developed places in the first place to promote the pro-people ideology.

Faction members’ behavior is conceptualized as being motivated by individuals’ pursuit of tangible and immediate incentives (Nathan and Tsai 1995). The regression results nonetheless provide some evidence of faction members’ behavioral “stickiness” (see Online Appendices Table 4). Jiang’s allies keep favoring economic growth over welfare development under Hu-Wen administration despite a clear priority shift in Beijing. Two possible explanations lie behind such a behavioral “stickiness”: first, long-term loyalty to the patron remains important and can generate foreseeable gains because of the none zero-sum nature of leadership transition and a decent respect to the prior leader in political life; second, different coalitions can develop distinct ideological preferences either due to the self-selection in the early formative phase of coalition or/and an internalization of the patron’s ideology at a later stage.

3.3 Robustness

Regressing variables on the raw value of social target points or the point difference between social targets and economic ones, the coefficients of CYL strength variable remain positive and statistically significant (see Online Appendices Table 4). A 10 percent increase in the proportion of CYL leaders within the provincial party standing committee enhances the TRS point difference by over 3 points out of a total of 100 points (see models 1–4 in Online Appendices Table 4). An alternative measurement of social unrest is constructed using data from the China Labor Bulletin (CLB) Project that has cataloged labor protests on the basis of news media reports since 2011.¹³ Labor unrest is a particular salient form of social instability due to the increasing number of migrant workers (Distelhorst and Hou 2017). Main findings remain unchanged using CLB data (see model 3 in Online Appendices Table 4). Adding other provincial-level economic variables, such as the growth rate of GDP per capita, population, and urbanization rate, the coefficient of CYL strength variable stays positive and statistically significant. Using available nightlight brightness data (2003–2013) generated from the DMSP-OLS satellite images that are independent of the Chinese statistical system to gage the levels of wealth, substantive results remain unchanged (see model 4 in Online Appendices Table 4).¹⁴

4 Enforcement of TRS

To what extent do formal TRS rules actually matter in personnel arrangements where personal connection and favoritism are influential? Due to the difficulty in data availability, few studies empirically examine the effectiveness of formal rules in China. The availability of TRS point information provides a unique opportunity to test the enforcement of personnel rules.

To test hypothesis 2.1, the following analysis uses the TRS point information and tests whether performance in social policy areas contributes more to career advancement in places that prioritizes social welfare in TRS but less so where local TRS prioritizes economic development. All prefectural party secretaries and mayors who have experienced career moves in sample province-years are considered.¹⁵ Interview evidence justifies using “leader-tenure” instead of the “leader-year” panel data in the study of promotion.¹⁶ Leaders’ biographical data are collected from the government online archive. To avoid transitory arrangements, the binary dependent variable uses the political outcome of one year after the completion of tenure, instead of the immediate one, with “1” indicating promotion, and “0” for all other position changes (see coding schemes in Appendices).¹⁷

The explanatory variable of key interest is a set of interaction terms between prefectural leaders’ policy performance (average welfare or economic growth during tenure) and priority reflected in the TRS point distributions (priority in TRS). An ideal measurement of policy performance is how well TRS target values are achieved. However, in the TRS dataset, only one province’s regulations reveal information on target values. Due to such a data limitation, I take the performance in education, public health, and environmental protection to gage social policy performance. The welfare growth during tenure variable is measured by the average of four growth variables, the growth rate of teacher–student ratio, licensed doctors per capita, clinical bed-population ratio, and the average growth rate of three environmental measures.¹⁸ These welfare measures encompass all available welfare indicators that are consistent and systematic in statistical yearbooks. Economic growth during tenure is measured by the average GDP growth rates during the leader’s tenure. Using growth rates of GDP per capita, nightlight brightness, and fiscal revenue generate similar substantive results. Priority in TRS is a dummy variable with “1” indicating more points are assigned to social targets relative to economic targets, and “0” for all other conditions. Based on existing research, a set of control variables is included: (1) leaders’ individual attributes, because younger, more educated, male Chinese who are earlier in their tenure have better odds of political advancement (Walder 2004; Landry 2008); and (2) one’s personal connection with superiors.¹⁹

The full model with interaction effects are presented in Table 5 in Online Appendix. Interaction effects between a dummy variable and a continuous variable in logistic models are best presented through graphs (Jaccard 2001). Figure 2 plots the marginal effects of work performance on the probability of promotion in places with different policy priorities. For mayors and prefectural party secretaries, good social performance, indicated by the large increase in social welfare indicators, in welfare-oriented places has a larger positive effect on promotion probability than in places where TRS prioritizes economic growth (see Fig. 2a-1, b-1); meanwhile, extraordinary economic performance, indicated by the big jump in economic indicators, in places where TRS prioritizes economic growth has a larger positive effect on promotion probability than in welfare-oriented places (see Fig. 2a-2, b-2). Other things equal, splendid performance in the policy area that receives priority in the TRS is indeed rewarded with better promotion prospects. Adding other variables, such as cities’ economic and political importance, generate similar findings.

Fig. 2

Interaction effects between policy performance and TRS priority (2003–2013)

Further supporting H2.1, good economic performance improves local leaders’ promotion prospects only in places that TRS prioritizes economic development (see Table 2). In cities that provincial TRS prioritizes social welfare, the effect of economic performance on promotion prospects is negative, whereas that of social policy performance is positive. The direction of policy performance’s effect on promotion prospects varies according to the TRS rules.

Table 2 Logit models on TRS Rules’ effectiveness in political advancement.

5 Implications of TRS

Extant research relies on case studies to examine the policy implication of formal TRS rules (Edin 2003; Heimer 2016; Whiting 2006). Quantitatively testing the policy implication of TRS rules (H2.2), I use the error correction model (ECM), a standard empirical approach for predicting government spending (Gandhi 2008; Miller 2015a; De Boef and Keele 2008). Econometric theory suggests that time-series analyses should always start with a general model without adhoc restrictions (Hendry 1995; De Boef and Keele 2008). ECM is one of the two classes of regression models appropriate for stationary data and weakly exogenous regressors.²⁰ ECM refers to any model that directly estimates the rate at which Y_t changes to return to equilibrium after a change in X_t. Error correction is known as the rate of return to equilibrium. The form of the ECM is the following:

$$ \Delta Y_{it} = \alpha Y_{it - 1} + \beta \Delta X_{it} + \gamma X_{it - 1} + \varepsilon_{it} $$

(1)

where subscript i denotes prefectures and t denotes year. ${Y}_{it}$ is a proxy of shifts in spending priority, measured by the change in the proportion of social spending in total prefectural government expenditures (Spending Ratio_it – Spending Ratio_it-1). Changes in the proportion of economic spending are also examined under the assumption that there may be a trade-off between economic growth and social welfare spending. Social spending includes spending on education, social security, and public health.²¹ Using the broad social spending category instead of any specific one, such as education and healthcare spending, is due to the spillover across social policy areas. For example, targets on social security may also affect medical and healthcare spending, and increase in job creation target point might also have implications for. education spending. ${X}_{it}$ is a set of independent variables, and ${\varepsilon }_{it}$ is an error term. The ECM model can distinguish direct, short-term effects ($\beta$) of changes in the independent variables from their longer term effects($\gamma$) on changes in the outcome variable. Social spending data are either unavailable or inconsistent before 2005, thus the analyzed period lies between 2006 and 2013.

The main explanatory variables of interest are the change and lagged level of the TRS point difference variable. Their coefficients are expected to be positive. Control variables include the lagged level of social spending ratio as required by the ECM framework, and the lagged levels and changes in stationary socioeconomic variables which have consistent and systematic city-level data, including GDP per capita, and are likely correlated with social spending expansion.²² Table 3 presents the results. The increase in TRS point differences is positively associated with the increase in the proportion of social spending in total government expenditures, which support H2.2. The hausman test indicates that fixed effects (FE) model is preferred over random effects model. Additionally, I employ estimations with panel-corrected standard errors (PCSEs) and first-order autocorrelation to correct within-group heteroskedasticity and cross-section as well as serial correlation of errors (Beck and Katz 1995). For all models, the substantive results are robust after controlling for the lagged levels and changes in revenue, the ratio of fiscal transfer to total revenue, population, urbanization, economic openness, and marketization that are likely correlated with the change in social spending.²³ Using the net change in social or economic target points, instead of the TRS point difference, generates the same substantive result.

Table 3 Policy implications of TRS point scheme.

An enhanced priority in social welfare, reflected in the TRS point distribution, is associated with a faster rate of social spending growth relative to other spending categories. A one point increase in the point differences between social and economic targets enhances the social spending ratio by around 5 percent. Variation in the target point differences variable explains nearly 7 percent of total variation in the outcome variable. Additionally, the statistical significance of the coefficient of the lagged level of TRS point differences is not robust,²⁴ indicating that the TRS point scheme has only short-term effects on social spending. The reaction of social spending to the change in TRS point distribution is ephemeral-influencing social spending this year, but not next year. This goes along with the fact that the TRS point scheme is updated on a regular yearly basis.

Table 4 TRS point scheme and agricultural spending.

Growth-related spending, such as infrastructure expenditure, is not publicly available after 2006. It is thus infeasible to test whether the enhanced points in social targets affect growth-related spending. The only economic spending data available at the city level is spending on agriculture, forestry and water conservancy. Preliminary evidence shown in Table 4 suggests that the increase in social target points is indeed negatively correlated with the expansion of agricultural spending. Yet, such an effect is not robust, as indicated in model 6 (see Table 4).

6 Conclusion

The nomenklatura system in China promotes performance-based selection, which ultimately helps deliver on ruling elites’ policy promises on the ground. Formal performance-based promotion rules are not always overshadowed by cronyism and thus useless. They can be effective in aligning local policy priorities with those of upper level governments. As shown in this analysis, the enhanced priority given to social welfare in local political selection rules translates down to a faster expansion of social spending by subordinate governments. The political selection institution is more than a co-optation mechanism. By elevating the importance of policy performance for promotion, well-enforced political selection rules can mold local leaders’ perceptions and policy priorities, which contribute to generating policy outcomes that are favorable to the images of top ruling elites.

Findings in this research cast doubts on the pervasive belief that factions only care about life-and-death power struggles. Rather, it lends supporting evidence to factions’ policy relevance in authoritarian rule. This analysis finds that the allocation of TRS points is immune from bottom-up pressure. Factions are political forces underlying local variation in political selection rules. Coalitional politics should not be narrowly conceptualized as relevant only in the uncommon time of power struggle. Joining studies on the interaction between formal and informal institutions (Helmke and Levitsky 2004), this research shows that factions also matter in normal politics, especially in the design of formal institutions. In the situation where the patron emphasizes performance legitimacy, factions do not merely rely on administrative directives or propaganda campaigns to generate desired policy outcomes. They also strive to affect policy implementation through shaping formal political selection institutions that propel performance-based promotion. In this way, factionalism becomes compatible with meritocratic selection. Although it is hard to pinpoint whether CYL officials advocate the pro-people policy initiatives out of calculated career considerations or the populist ideology, factional affiliation is found relevant and important beyond the search for power, and also matters in the development of formal rules.

By unveiling the association between coalitional ties and the allocation of TRS points that exert impacts on local spending policies and personnel arrangement, this analysis demonstrates the endogenous nature of formal political selection rules and the intricate relationship between competence and loyalty. The type of competence valued by higher level leaders is conditioned on their factional affiliation and political loyalty. Far from being separate, competence and loyalty are closely related. The type of competence developed and strengthened by the political incentive system is deeply rooted in coalitional politics.

Why does the nomenklatura system perform differently across communist regimes? In Soviet Union, nomenklatura rules were ignored and cadres’ departments were subjugated to economic ministries. As a matter of common practice, cadres were selected on the basis of personal acquaintance, position sale, and on grounds of common ethnicity (Lewin 2003). The greater emphasis on meritocracy in contemporary Chinese personnel system may be attributed to the historical legacy of imperial rule (Xi 2019), including the imperial examination system, and Deng Xiaoping’s pragmatism in transforming the cadre corps in the 1980s.

Performance-based promotion does not offer a panacea for all problems in communist rule. Nor do results suggest that the current Chinese nomenklatura system is problem-free. As accurately pointed out by Landry et al. (2018), the well-functioning of political meritocracy depends inordinately on the ruler’s ability to make the “right” policy decisions and on an objective performance evaluation system that in practice can be subject to strategic responses and data manipulation by opportunistic local politicians. Moreover, the influence of coalitional politics on political selection institutions, albeit brings in policy pluralism at the local level, might at the same time introduce unevenness or discontinuity in local developmental trajectories.