1. Introduction
Corruption undermines development in many ways. Yet, corrupt politicians have garnered high public approval and electoral support around the world (De Vries and Solaz, Reference De Vries and Solaz2017). One prominent explanation is that citizens lack credible information on corrupt officials' misdeeds (Ferraz and Finan, Reference Ferraz and Finan2008). However, even well-informed citizens have been found to support corrupt politicians in Africa (Chang and Kerr, Reference Chang and Kerr2017), Asia (Choi and Woo, Reference Choi and Woo2012), Europe (Konstantinidis and Xezonakis, Reference Konstantinidis and Xezonakis2013; Muñoz et al., Reference Muñoz, Anduiza and Gallego2016; Breitenstein, Reference Breitenstein2019), and the Americas (Vera, Reference Vera2020).
A similar observation has emerged regarding China's recent anti-corruption campaign. Since late 2012, the ruling party has cracked down on a number of officials at all levels of government. While Chinese citizens cheer at the fall of many corrupt officials, they pity and have “naïve sympathy” for others.Footnote 1 Given that people generally dislike corruption, why do they tolerate some corrupt politicians?
The corruption–competence tradeoff hypothesis argues that citizens knowingly forgive corrupt politicians in exchange for their competence (Winters and Weitz-Shapiro, Reference Winters and Weitz-Shapiro2013). Here competence is loosely defined as performing well in delivering public benefits, such as promoting economic development.Footnote 2 Citizens evaluate politicians based on multiple dimensions (Rundquist et al., Reference Rundquist, Strom and Peters1977), including honesty and the ability to provide public benefits. They assess politicians' performance or attributes in each dimension and sum up the weighted score across all dimensions, which further maps onto their approval and vote choices. Based on the multi-dimensional calculation, citizens may trade politicians' integrity for their competence without explicit negotiations.
To see how the implicit exchange works, consider for simplicity that competence and corruptness are both binary attributes. Thus, there exist four types of politicians with varying degrees of competence and/or corruption. While citizens presumably score politicians who are honest and competent the highest, this ideal type is often in short supply, especially in developing countries. More often than not, politicians are either corrupt or incompetent, or both. When citizens value politicians' competence more than their integrity, they will score the corrupt yet competent type higher than the honest but incompetent one; both will beat the corrupt and incompetent type. Thus, rational citizens (who prefer politicians with the highest score) may favor a corrupt over an honest politician because the former turns out to be the most competent among the realistic options—what we call the corruption–competence tradeoff. While another influential argument maintains that people tolerate corrupt politicians because it is hard to find clean ones (Muñoz et al., Reference Muñoz, Anduiza and Gallego2016), according to the tradeoff thesis, people still prefer a corrupt (yet competent) politician even when honest (but incompetent) officials are abundant.
Central to the corruption–competence tradeoff hypothesis are (1) an often implicit premise that the ideal type (i.e., honest and competent) of politicians is not always available and (2) an ordering of citizens' preference in which corrupt-yet-competent politicians outperform other sub-ideal types. As the former premise arguably holds in many contexts, several studies have examined the ordering of preferences by checking whether citizens prefer corrupt-yet-competent politicians over other sub-ideal types, notably those who are honest yet incompetent. For instance, Esaiasson and Muñoz (Reference Esaiasson and Muñoz2014) find that citizens prefer a corrupt-yet-competent mayoral candidate over an honest but incompetent one in Spain and Sweden. By contrast, Winters and Weitz-Shapiro (Reference Winters and Weitz-Shapiro2013) demonstrate that Brazilians prefer honest-yet-incompetent municipal mayors over corrupt-yet-competent ones, which refutes the corruption–competence tradeoff.
Several other studies have examined a softer version of the corruption–competence tradeoff—that is, whether people care less (or not at all) about a politician's corruption as long as they are competent. If so, it implies that corrupt-yet-competent politicians are almost equally favored as their honest and competent counterparts, and thus have a good chance of beating the other two sub-ideal types. In this direction, researchers have found that competent politicians receive no (or less) electoral punishment than their incompetent peers for engaging in corruption. This finding appears robust in several individual countries, such as Brazil (Pereira and Melo, Reference Pereira and Melo2015), Greece (Konstantinidis and Xezonakis, Reference Konstantinidis and Xezonakis2013), Moldova (Klašnja and Tucker, Reference Klašnja and Tucker2013), Peru (Vera, Reference Vera2020), South Korea (Choi and Woo, Reference Choi and Woo2012), and Spain (Muñoz et al., Reference Muñoz, Anduiza and Gallego2016), and cross-nationally (Zechmeister and Zizumbo-Colunga, Reference Zechmeister and Zizumbo-Colunga2013; Carlin et al., Reference Carlin, Love and Martínez-Gallardo2015; Rosas and Manzetti, Reference Rosas and Manzetti2015; Klašnja et al., Reference Klašnja, Lupu and Tucker2021). However, counterevidence also exists. For instance, Klašnja and Tucker (Reference Klašnja and Tucker2013) find that Swedish citizens impose similarly harsh electoral punishment on corrupt politicians, regardless of the economy. Thus, both the hard and soft tests have received mixed support.
This study contributes to the ongoing debate by examining whether (and why) the corruption–competence tradeoff exists in China. We conducted a vignette experiment involving 5527 ordinary Chinese citizens through a telephone survey from late 2018 to early 2019. While previous studies often arbitrarily select one level of government to analyze, we varied government levels in our design to examine whether the tradeoff varies when citizens evaluate officials at different levels. We generated eight fictitious profiles of officials at varying government levels and degrees of honesty and competence, and asked each respondent to evaluate one of them selected at random. Citizens evaluate these officials in three respects: demotion, promotion to their own jurisdictions, and corruption reporting.
Two major findings—which are consistent across the multiple outcome variables—emerge from the analysis. First, citizens prefer corrupt-yet-competent over honest-yet-incompetent township leaders. This ordering of preferences indicates that Chinese people knowingly forgive grassroots-level corrupt officials as long as they deliver public benefits. Second, when evaluating municipal mayors, citizens' relative preference for corrupt-yet-competent officials over honest-yet-incompetent ones is attenuated to indifference. The different findings for township- and municipality-level officials suggest that citizens uphold a hierarchical corruption–competence tradeoff.
Our in-person interviews unpack the mechanisms driving these findings in practice. For instance, citizens are often more worried about the incompetence of grassroots bureaucrats than that of high-level officials, due to the former's proximity to their daily life. They are also more concerned with the adverse effect of high-level officials engaged in corruption, but less so for grassroots-level officials, because the former are more powerful in the government hierarchy.
Our study advances the literature on the corruption–competence tradeoff in two main ways. On the one hand, to the best of our knowledge, we are the first to document a hierarchical corruption–competence tradeoff across government levels. To complement Klašnja and Tucker's (Reference Klašnja and Tucker2013) finding that this tradeoff has varying salience across countries in “good” and “bad” corruption equilibria, we demonstrate that the tradeoff may also exhibit different degrees of salience across different government levels within the same country. The finding thus calls for a more sophisticated understanding of citizens' preferences. The hierarchical tradeoff also informs future empirical studies that the choice of government level studied will likely affect the conclusion.
On the other hand, by studying China, we extend the literature—which is primarily democracy based—to authoritarian regimes, which is both reasonable and important. This extension makes sense because the underlying mechanisms also apply to authoritarian regimes. Citizens similarly evaluate officials in multiple dimensions in these countries. Thus, their desire for honest officials could be overpowered by other, more important, considerations. As Nathan (Reference Nathan2020: 162) has noted, “authoritarian regimes, like democracies, are being watched and evaluated by their citizens, who still expect…that the governments in their countries will act effectively and fairly.”
Officials' competence is likely just such an alternative in autocracies because many authoritarian governments actively highlight their role in delivering good performance (e.g., fast economic growth) to legitimize their rule (von Soest and Grauvogel, Reference von Soest and Grauvogel2017). Accordingly, citizens view sound performance as a key indicator of good government, if not the most important one. Moreover, many alternatives for which people are willing to exchange politicians' integrity in democracies, such as partisanship, are less salient or meaningless in authoritarian regimes (Anduiza et al., Reference Anduiza, Gallego and Muñoz2013). Because authoritarian institutions often preclude these other alternatives, citizens are most likely to trade officials' integrity for competence if anything.
China is a suitable case to study because governance performance is a crucial source of regime legitimacy (Zhao, Reference Zhao2009). Moreover, its remarkable economic growth accompanied by widespread corruption and the increased bureaucratic slack amid anti-corruption crackdowns may have convinced the public that corruption and good performance often go hand in hand, and that honest officials sometimes deliver little (Ang, Reference Ang2020; Wang and Yan, Reference Wang and Yan2020). Thus, the comparison between corrupt-yet-competent and honest-yet-incompetent local officials—a crucial empirical test of the corruption–competence tradeoff—is familiar to Chinese citizens. Lastly, the strikingly similar organizational structure and governance duties of the multiple levels of Chinese local government help us explore the varying corruption–competence tradeoffs across levels of government.
Additionally, the corruption–competence tradeoff may undermine government responsiveness in authoritarian regimes in a similar way as it does in democracies. By allowing corrupt politicians to take office and enrich themselves, the corruption–competence tradeoff undermines electoral accountability and government responsiveness in democracies (Winters and Weitz-Shapiro, Reference Winters and Weitz-Shapiro2013: 418). This adverse effect is similarly applicable to authoritarian regimes, even though citizens often do not directly decide who takes office. By allowing local officials' good short-term performance to shield social grievances related to their corruption, the corruption–competence tradeoff undermines the fire alarm system that authoritarian rulers use to identify malfeasant agents (Huang et al., Reference Huang, Boranbay-Akan and Huang2019). It also reduces pressure on the rulers to dismiss them. If corrupt agents are not identified or punished at a low level, they may survive and thrive in the government, and ultimately enrich themselves at public expense.Footnote 3
Finally, the corruption–competence tradeoff has important implications for understanding the bases of political support in authoritarian regimes. Citizens' approval or disapproval of officials reflects their level of support for the regime, to whom the officials are agents (Li, Reference Li2016). A lack of regime support not only worsens the problem of authoritarian control by undermining the civil order; it also weakens elite coherence and insiders' loyalty, which in turn threatens regime stability (Svolik, Reference Svolik2012; Nathan, Reference Nathan2020). The corruption–competence tradeoff thus helps explain why citizens support the regime, despite widespread corruption, in an era of rapid growth. Yet, it also implies that anti-corruption measures may not so effectively boost regime support if they undermine economic performance.
The following section reviews the existing literature and discusses citizens' preferences for various bureaucratic types at different government levels in China. We introduce the research design in the third section, and present the statistical results and interview evidence in the fourth section. The final section discusses the implications of the findings and concludes.
2. Integrity or competence? Citizens' preferences for bureaucratic types in China
2.1 Conditional punishment and relative weight
People evaluate politicians on multiple dimensions (Rundquist et al., Reference Rundquist, Strom and Peters1977; Fernández-Vázquez et al., Reference Fernández-Vázquez, Barberá and Rivero2016). For instance, citizens may value politicians who deliver public benefits even more than those who provide personal benefits such as jobs (Konstantinidis and Xezonakis, Reference Konstantinidis and Xezonakis2013; Zechmeister and Zizumbo-Colunga, Reference Zechmeister and Zizumbo-Colunga2013) because public benefits signal that the politician is a desirable type—an excellent public manager (Ferrer, Reference Ferrer2020: 25; Vera, Reference Vera2020: 657).
While all of this research builds on the role of competence in winning public forgiveness for corrupt politicians, prior studies of the corruption–competence tradeoff have proposed two distinct channels—which Breitenstein (Reference Breitenstein2019) succinctly summarizes as conditional punishment and relative weight. Conditional punishment highlights that competence reduces the relevance of integrity in people's evaluations of politicians, which often involves unconscious bias as an explanation (Breitenstein, Reference Breitenstein2019: 3). For instance, the good performance of corrupt politicians makes people less likely to believe they are the corrupt type. Relative weight emphasizes that citizens deliberatively place more weight on politicians' competence than their integrity, which closely follows the initial theoretical formalization of the implicit exchange of rational voters (Rundquist et al., Reference Rundquist, Strom and Peters1977).
While the two channels are sometimes compatible, our study more closely builds on the relative weight perspective for three reasons. First, we are not only interested in why citizens fail to severely punish corrupt politicians, but also why they sometimes prefer corrupt over honest officials. As discussed above, this preference can be easily explained by the public's implicit trading of integrity for competence due to different weights, but it is less obvious in the conditional punishment framework. Second, while insightful, the unconscious bias underpinning conditional punishment does not necessarily predict that citizens will forgive corrupt-yet-competent politicians. It is unclear why the bias could not operate in reverse: if people are less likely to believe politicians are competent when they are corrupt, the latter effect could offset the former.
Finally, the conditional punishment perspective has informed empirical tests of whether politicians' corruption and competence interactively affect their public approval. While these tests shed light on the public preferences for different types of politicians,Footnote 4 the direction and magnitude of the interaction coefficient cannot directly translate into the order of public preference—our main interest.
2.2 Types of officials in China
We classify officials into four types based on whether they are competent and/or corrupt (Table 1). The public probably likes corrupt and incompetent (CI) officials the least and honest and competent (HC) officials the most. Between them are two sub-ideal types: corrupt but competent (CC) and honest but incompetent (HI). If citizens place more weight on performance than integrity, they will like the former more than the latter (a preference ordering of HC > CC > HI > CI). This comparison thus informs whether the corruption–competence tradeoff exists.
Types of officials
The preferred HC ideal type may be rare, especially in developing countries where officials' corruption and their performance are positively correlated rather than independently distributed. For instance, prosperous local economies generate more rents for officials in China, and corrupt officials have used cronyism to “get things done” and promote development during the reform era (Ang, Reference Ang2020). In such a development trajectory, corruption and competence reinforce each other. Thus, many officials are paradoxically corrupt yet competent (i.e., the CC type)—what Ang (Reference Ang2019) describes as a “corrupt meritocracy.”
Wan Qingliang, the former Party secretary of the capital city of Guangdong, is an example of the CC type. He increased the annual GDP growth rate in one of his previous jurisdictions from 7.3 to 22.1 percent in just 5 years. However, he also brazenly engaged in corruption, taking more than 110 million RMB (15.68 million USD) in bribes.Footnote 5 More generally, between 2012 and 2017, 54 of the 331 officials who were municipal Party secretaries in 2011 were disciplined for corruption in China; 40 percent of those punished had recently been promoted—an indication of competence—before their fall (Ang, Reference Ang2020, chapter 6). Similarly, our data on exposed corruption cases of province-supervised officials show that 44 percent of them are promoted faster than average early in their careers (Appendix 3.1). Thus, CC officials are quite numerous in the bureaucracy.
There may also be many HI officials; they have attracted more attention from both the public and the ruling party in China's recent anti-corruption campaign. The anti-corruption campaign has gained unprecedented strength since the 18th Party Congress in 2012: 1.54 million officials were disciplined for corruption from 2013 to 2017.Footnote 6 While the campaign has deterred local officials from engaging in corruption, bureaucratic inaction has increased in some places, leading to administrative ineffectiveness and poor performance (Wang and Yan, Reference Wang and Yan2020; Wang, Reference Wang2022). The Party openly criticizes textbook examples to alert its local agents and calls for a fight against inaction.Footnote 7
Chinese citizens often encounter local officials who are either corrupt or incompetent in their daily lives. If they prefer CC over HI types, this implies they are willing to trade bureaucratic integrity for competence. Anecdotal evidence suggests that some refer to the CC as “good” corrupt officials and the HI as “bad” honest officials.Footnote 8 These labels reveal a preference for the former, which is consistent with the corruption–competence tradeoff theory.
2.3 Why preferring CC over HI types?
Three factors may explain citizens' preference for CC over HI officials in China. First, people greatly value competent local officials; citizens may appreciate the role they have played in the country's economic boom over the past four decades. Accordingly, citizens evaluate a local government favorably if it promotes economic growth and enhances social welfare (Chen, Reference Chen2017). Our interviews reveal that people often view competence as a defining feature of a “good” official. While they appreciate the rapidly growing economy, some believe that “everyone wants to get even better [economy].” To further enhance economic development, they demand competent officials.
Second, due to widespread corruption, citizens may have a low disutility of accepting corrupt officials. Along with the fast economic growth, corruption has intensified in China since the reform (Wedeman, Reference Wedeman2012). While the recent anti-corruption campaign has cracked down on many officials, it may not have significantly reduced the public's perception of corruption yet. For instance, Wang and Dickson (Reference Wang and Dickson2022) demonstrate that corruption investigations, which reveal endemic corruption to ordinary citizens, lead them to downgrade their beliefs in officials' integrity. Several interviewees also said that pervasive corruption might persist despite the recent intensive anti-corruption efforts.
If people perceive corruption to be widespread, they are encouraged to engage in it due to the interdependence of corruption decisions (Persson et al., Reference Persson, Rothstein and Teorell2013; Corbacho et al., Reference Corbacho, Gingerich, Oliveros and Ruiz-Vega2016). They will also be reluctant to fight against the practice due to “corruption fatigue” (Peiffer and Alvarez, Reference Peiffer and Alvarez2016). In support of this reasoning, several individuals interviewed for our study said it would be “fussy” for them to evaluate an official primarily based on their integrity. One said, “Don't be too picky! How many people can keep their hands clean in officialdom?”
Third, the many corrupt yet competent local officials may have created and reinforced the impression that corruption is a necessary precondition for good performance. This impression, in turn, justifies officials' wrongdoing. One interviewee vividly explained to us, “Society is just like a giant dyeing cylinder. If you are too clean, you will be ostracized. Gradually, you cannot accomplish anything.”
In sum, Chinese citizens greatly value competent officials, may experience a low disutility of accepting corrupt ones, and sometimes view corruption as a necessary precondition for development. We thus expect that they are willing to trade officials' integrity for competence when dealing with corrupt or incompetent officials: they prefer CC over HI officials.
2.4 Varying public preferences across government levels
How does the corruption–competence tradeoff vary, if at all, across government levels? While prior research has not directly addressed this question, indirect evidence yields mixed expectations. On the one hand, a corrupt higher-level official can extort more due to their greater power, which imposes a higher cost on society (Fang et al., Reference Fang, Gu and Zhou2019). Based on 3057 corruption cases collected from The Procuratorial Daily of China (1992–2010), we find that the mean bribe is 2.3 million RMB (326,000 USD) for a township-level official and 4.5 million RMB (635,000 USD) for a municipality-level official. The larger scale of corruption should make people less likely to tolerate a corrupt official at a higher level; accordingly, the corruption–competence tradeoff is unlikely when citizens evaluate high-level officials.
Additionally, corrupt high-level officials are more detrimental in a bureaucratic hierarchy because they have a tremendous influence on the career prospects of those at a lower level. These “unprincipled principals” can actively use oversight tools to incentivize lower-level officials to engage in corruption (Brierley, Reference Brierley2020). Relatedly, lower-level officials often compete to offer higher-level officials extravagant hospitality to secure their endorsement in China (Gong and Xiao, Reference Gong and Xiao2017). Such behavior will arguably be amplified if the high-level official is a corrupt type. Corrupt officials will also be reluctant to punish corrupt low-level officials. Recent studies show that people welcome authoritarian leaders who uphold retributive justice by punishing corrupt wrongdoers (Tsai et al., Reference Tsai, Trinh and Liu2022). While citizens may tolerate low-level CC officials, they are unwilling to see many low-level CI officials go unpunished. Citizens will thus be less likely to trade integrity for competence in evaluating high-level officials, due to fears that they will “corrupt” many lower-level officials and comprise retributive justice.
On the other hand, ordinary citizens more easily fall prey to corrupt grassroots-level officials, with whom they have frequent direct contact (Rose and Peiffer, Reference Rose and Peiffer2015). In China, grassroots officials may directly damage citizens' well-being by embezzling their agricultural funds, retaining land acquisition compensation, and misappropriating poverty alleviation funds. Thus, opposite to the mechanisms above, citizens should be more likely to tolerate a corrupt higher-level official since they are more distant from their daily life, and thus should be more willing to trade their integrity for competence. We will examine the corruption–competence tradeoff at both low- and high-level governments to explore the possibility that the tradeoff varies across government levels.
3. Experimental design and data collection
3.1 Vignette design
We designed a vignette experiment to examine citizens' relative preferences for various bureaucratic types in China. The experiment was included in a nationwide telephone survey—the National Integrity Evaluation Survey (NIES)—conducted in late 2018 and early 2019. Using random digit dialing, the NIES aims to reach telephone users from each municipality and gauge their opinions on corruption and anti-corruption policies in China. The team used a Computer Assisted Telephone Interviewing system to ensure data quality. A total of 5527 adults, randomly selected from a larger pool of NIES respondents in 27 provincial jurisdictions (except Beijing, Tianjin, Shanghai, and Chongqing), completed the experiment.
To explore how the corruption–competence tradeoff varies across government levels, we consider three sets of attributes—corruption versus honesty, competence versus incompetence, and township-/street-level director versus municipal mayor, which generates eight (2 × 2 × 2 = 8) vignette bureaucrats with different configurations of attributes. Each respondent heard descriptions of one fictitious bureaucrat randomly selected from the eight, and evaluated the bureaucrat in three aspects (i.e., demotion, promotion, and corruption reporting). Thus, respondents were divided into eight groups, each evaluating one official. We then compared the group average evaluations of bureaucrats to understand the public's relative preferences for various bureaucratic types. Thus, our design closely resembles a rating-based conjoint, in which respondents rate each profile, rather than a choice-based conjoint, in which respondents choose one from the paired profiles. The eight vignettes read as follows (the underlined phrases were substituted with those in parentheses to generate the eight fictitious profiles of officials):
Imagine a township/street director (municipal mayor), whose surname is Zhang, working in a municipality near yours. During Zhang's tenure in office, local economic growth is slow, environmental sanitation is bad, and public transport is inconvenient (local economic growth is fast, environmental sanitation is good, and public transport improves a lot). Zhang was disciplined by the Party for accepting gifts from an estate developer in his early years (has a reputation for honesty in his work and has no record of being disciplined).
Our vignette design is based on three considerations. First, we seek to maximize the gap between the two government levels to detect meaningful differences between them. Below the central government, local governments have four levels: provincial, municipal, county, and township levels. We use the leaders of the lowest-level government (i.e., township-level directors) to indicate the low-level officials.Footnote 9 We use municipal governments, rather than provincial or central governments, as the higher level because, given the large number of municipalities (around 300) and the significant variations among them, respondents are unlikely to question the plausibility of our vignette descriptions.
Second, we use economic development, environmental sanitation, and public transport to indicate excellent or poor performance. Third, we use an official's record of being disciplined by the Party for accepting gifts from estate developers to indicate corruption. While the vignette describes one minor corrupt exchange, it does not assert that the official has never engaged in other corrupt acts. We seek to convey to the study participants that a vignette official is a corrupt type—that is, a person with a high propensity to engage in corruption. In China, the Party disciplines malfeasant officials; those with severe violations are transferred to the judicial system for prosecution. Thus, our description of minor corruption clearly indicates malfeasance, but the punishment of Party disciplining (e.g., warning) does not necessarily imply the end of an official's political career. Officials still have a chance of promotion after disciplinary actions.Footnote 10
Table 2 shows the eight vignette officials. We are mainly interested in comparing CC and HI government leaders because the corruption–competence tradeoff predicts that people like the former more than the latter. We also compare these types with other types to check, for instance, whether respondents prefer honest over corrupt officials with the same performance record.
Eight vignette officials
Notes: This table displays the eight fictitious officials. The first two letters (CI, CC, HI, and HC) stand for corrupt-and-incompetent, corrupt-yet-competent, honest-yet-incompetent, and honest-and-competent types, respectively. The third letter indicates township leaders (T) or municipal mayors (M). The four highlighted vignette officials are honest-yet-incompetent township leaders (HIT), corrupt-yet-competent township leaders (CCT), honest-yet-incompetent mayors (HIM), and corrupt-yet-competent municipal mayors (CCM).
3.2 Outcome variables
We design three outcome questions to measure respondents' (dis)approval of a fictional bureaucrat. If these outcome variables yield similar patterns, this will increase our confidence in the findings. First, we asked respondents about their opinions of the career progress of the hypothetical leader, which we used to construct the first outcome variable—demotion—as a measure of disapproval. The binary variable takes a value of 1 if the respondent supports demoting the assigned leader. Second, if a participant chose not to demote the vignette bureaucrat, we further asked them whether they supported promoting Zhang to an important position in the respondent's own county-/district- or provincial-level government if there was a job opportunity. We personalized the question in this way to induce more careful answers. While Chinese citizens do not directly decide local officials' turnover, they understand that these questions ask about their attitudes, not whether they could literally choose officials (Tsai et al., Reference Tsai, Trinh and Liu2022), which our interviews also confirm.
The third outcome variable is designed to capture respondents' willingness to report the bureaucrat's corruption. To reduce the sensitivity of the question, we provided another vignette in which Zhang's subordinate comes across a clue that Zhang may have accepted a bribe of 20,000 RMB (2800 USD), and then asked respondents whether the subordinate should report Zhang's corruption to the anti-corruption agencies. We specified a 20,000-yuan bribe so that the misdeed would not constitute a crime or be too severe.Footnote 11 Yet, the recent anti-corruption campaign has encouraged corruption reporting in China, which may increase the overall corruption reporting rate and limit its variation. Thus, we consider this variable supplementary to the other two. While demotion and corruption reporting aim to measure disapproval, promotion indicates approval. The three outcome variables are all binary.
3.3 Methodological discussion
Before proceeding to the results, three aspects of our design and analysis merit discussion. First, despite the potential response biases, such as social desirability bias, in measuring the absolute (dis-)approval of each vignette bureaucrat, comparisons across the bureaucratic types are less likely to be prone to these biases; they can accurately reveal respondents' relative preferences among these types (Hainmueller et al., Reference Hainmueller, Hangartner and Yamamoto2015: 2399). Since our respondents evaluated one randomly selected vignette official rather than all eight, they were not fully aware of our experimental manipulation of different attributes of fictitious bureaucrats. Consequently, they are less likely to consciously bias their report of their relative preference to gain the interviewers' social approval (Wallander, Reference Wallander2009: 505–506). Moreover, the randomization procedure makes respondents similar in expectation in all conditions across groups. When comparing the average approval or disapproval of each official, the response bias incurred by respondents' individual characteristics is likely canceled out.
Second, our goal is to compare public approval for the profiles of officials to establish the order of public preference but not estimate the effect of each attribute. Our design and analysis closely follow those of Winters and Weitz-Shapiro (Reference Winters and Weitz-Shapiro2013) and Esaiasson and Muñoz (Reference Esaiasson and Muñoz2014) to compare official profiles. We pay special attention to the comparison between the two sub-ideal profiles (i.e., corrupt-yet-competent and honest-yet-incompetent officials) because this comparison is more important for supporting or rejecting the corruption–competence tradeoff than comparing other pairs, such as honest-and-competent versus corrupt-and-incompetent officials.
Thus, our interest departs from the Average Marginal Component Effects (AMCE). In conjoint experiments, AMCE estimates reflect how much the probability of approving an official would change, on average, if one attribute (e.g., corruption or competence) changed from one level to another. We compare official profiles rather than estimate the effect of each attribute. Our focus on comparing the two sub-ideal profiles also differs from the AMCE estimates (reported in Appendix Table E), which average across all profiles.
Third, recent methodological advancements reveal that the AMCE estimates may conflate preference strength with direction in forced-choice conjoint experiments. Thus, one should be cautious about interpreting the sign of the AMCE of specific attributes as reflecting majority preference (Abramson et al., Reference Abramson, Koçak and Magazinnik2022). Our design is less prone to the misinterpretation problem because it is a rating-based rather than forced-choice conjoint (e.g., Costa, Reference Costa2021). Moreover, our primary interest is not the AMCE of each attribute, but rather the ranking of the public approval ratings of bureaucratic profiles. Since all three dependent variables are binary, we can estimate the proportion of the population that supports each profile by aggregating the responses and comparing them to establish a ranking of bureaucratic types.
3.4 Randomization and manipulation check
Each of the eight groups has around 690 respondents. Respondents are similar in observable variables across groups, indicating successful randomization (Appendix Table A).
Additionally, as a check of our experimental manipulation, we asked respondents about the vignette leader's type in two dimensions: competence and integrity. We find that respondents who heard about the fictitious leader's good performance and those presented with the party disciplining information were significantly more likely to regard the leader as a competent type and a corrupt type, respectively, than those who received the opposite information.
Over half (56 percent) of respondents who heard the party disciplining information regarded the hypothetical bureaucrat as a corrupt type, compared to only 8 percent of those presented with information about no party disciplining (a 48-percentage-point difference, p < 0.001).Footnote 12 Similarly, the outstanding performance information enhances the share of respondents who viewed the described officials as competent by 56 percentage points (p < 0.001) from 13 to 69 percent.Footnote 13 Thus, respondents paid attention to the vignette descriptions and interpreted the information provided as we intended.
3.5 Group interviews
To further probe the tradeoff, we organized and moderated six group discussions with 32 individuals in South China in 2020. The interview evidence was not intended to be conclusive but to understand the experimental findings' possible mechanisms. Thus, we set quotas on the interviewees' backgrounds—such as age, gender, education, occupation, and income—to increase diversity in the explanations (Appendix Table B).
4. Results
Table 3 summarizes the proportions of people who supported demoting each vignette official. In general, respondents value the integrity of leaders at both the township and municipality levels, given the level of performance. This is reflected by the significant row differences in each column in Table 3. Similarly, respondents also value the competence of leaders at both levels, given the level of integrity. They are much more likely to disapprove of a corrupt and incompetent official (i.e., the worst type) than an honest and competent one (i.e., the ideal type). While unsurprising, these findings lend credence to our vignette design. Although the absolute proportions reported in each cell may suffer from response bias or measurement errors,Footnote 14 the comparisons among bureaucratic types accurately reflect their relative preferences.
Proportions of respondents supporting demoting each vignette leader
Note: For each vignette official, we report the proportion of respondents that supported demoting the official. Robust standard errors are in parentheses. ***p < 0.01.
We repeat the analysis for the two additional outcome variables—promotion and corruption reporting (Appendix Tables C and D). The result of promotion closely mirrors Table 3. The results of corruption reporting are qualitatively similar: most of the differences are of the expected sign,Footnote 15 though some become statistically insignificant. The overall high rate of corruption reporting (on average, 92 percent) and the limited variation may have restricted this measure from fully revealing public preference. The AMCE estimates show similar patterns (Appendix Table E models 1 to 3). Overall, these findings are consistent with those reported in Table 3.
Next, we discuss the key findings on the corruption–competence tradeoff at the township and municipal levels—that is, CCT-HIT and CCM-HIM—in turn.
4.1 Main findings
We first evaluate the corruption–competence tradeoff at the township level, as presented in the left half of Table 3. The tradeoff thesis expects citizens to be less likely to disapprove of a corrupt-yet-competent township leader (CCT) than an honest-yet-incompetent one (HIT). Referring to the demotion proportion, it predicts CCT<HIT, that people are less willing to demote a CCT than an HIT. Indeed, CCT-HIT = 0.655–0.73 = −0.075. The decrease is substantial (0.075/0.73 = 10.3 percent) and statistically significant (p = 0.002).
Figure 1 more straightforwardly plots citizens' preferences for corrupt-yet-competent over honest-yet-incompetent officials as reflected by the three outcome measures. The first row presents their preference for CC-type over HI-type township leaders (i.e., CCT–HIT); panel (a) replicates the finding of demotion discussed above. Panels (b) and (c) consistently show that citizens prefer CCT over HIT: they are more likely to support promoting the former and less likely to support reporting their misbehavior. These findings support the tradeoff hypothesis.
Preference of corrupt-yet-competent and honest-yet-incompetent officials.
Note: For each column, the first and second rows display estimated differences between corrupt-yet-competent and honest-yet-incompetent officials (i.e., CC–HI) at the township and municipality levels for the indicated outcome variable. The bottom row further compares the two estimated differences across government levels. Horizontal lines denote 95 percent confidence intervals based on robust standard errors.
In Figure 1, the second row applies the same procedure to examine citizens' corruption–competence tradeoff in evaluating municipal mayors. In contrast to (a), panel (d) shows that respondents no longer disapprove of a corrupt-yet-competent mayor to a lesser extent than they disapprove of an honest-yet-incompetent one. The difference in demotion rates (0.008) is small and statistically insignificant. The insignificant difference is repeated on the other two outcome variables in (e) and (f). Thus, citizens' preference for CC-type over HI-type officials vanishes at the municipality level.
The third row compares the estimates from the first and second rows to rigorously test whether the relative preference for CC-type over HI-type officials is hierarchical. The positive coefficient (0.083, p = 0.019) in panel (g) suggests that, compared to citizens' preference for CC-type over HI-type officials at the township level, this preference significantly attenuates at the municipality level. We find a similar result when using promotion as the outcome in panel (h). We note that the estimate (0.036, p = 0.088) is not statistically significant at the conventional level for reporting in (i). Nevertheless, the coefficient has a positive sign that is consistent with the hierarchical tradeoff. Collectively, the evidence suggests that people uphold a hierarchical corruption–competence tradeoff.
We also collected a rich set of individual covariates, including age, gender, education, ethnicity, party membership, public sector employment, income, urban residence, hukou status, hometown origin, perceived corruption prevalence in their township- and municipal-level governments, and how frequently they interacted with the government. Controlling for these covariates yields the same findings (Appendix Figure A).
4.2 Exploring the mechanisms
We now explore why citizens prefer corrupt-yet-competent (CCT) over honest-yet-incompetent township leaders (HIT), but the relative preference attenuates in the comparison of corrupt-yet-competent (CCM) over honest-yet-incompetent municipal mayors (HIM) with the interview evidence.Footnote 16 Our interviewees frequently mentioned different proximities to ordinary citizens and positions in the power hierarchy in accounting for the hierarchical corruption–competence tradeoff, though the two are not necessarily the only mechanisms. (Appendix Table F summarizes their typical reasonings.)
On the one hand, citizens perceive competence as more consequential at the township level: competence has a larger AMCE at the township level than at the municipality level (Appendix Table E models 4 to 6). Table 3 likewise shows that the column differences (i.e., the effect of competence conditional on the corruption level) are larger at the township level. Our interviews reveal two reasons why this may be the case. First, lower-level officials' competence is perceived to be more important due to their proximity to ordinary citizens. As one woman explained, “We have more contacts with street-level officials. Their ability directly affects our daily interactions with the government. On the contrary, provincial and municipal governments are distant from us. Even if we file a petition, we have to visit the government from the grassroots.”
Second, incompetent township-level officials are more unpopular because citizens are concerned that the bureaucracy does not punish “lazy” officials as strictly as it punishes corrupt ones at the grassroots level. While citizens can report corruption, they feel less able to flag an inefficient or incompetent official. This problem partly reflects the difficulty of incentivizing grassroots-level officials in China. As the ruling party cares about performance legitimacy, officials who perform well are rewarded with promotions. Yet, due to their low position in the power hierarchy and large number, township-level officials face fierce competition; only 1 percent are promoted to the next rank each year (Zhou, Reference Zhou2017: 182). Given this slim likelihood of career progression, township-level officials may choose to minimize costly efforts to improve their performance. Also, compared to corruption, it is more difficult to precisely attribute unsatisfactory governance to a particular official and punish him or her accordingly.
In connecting these reasons, a rural woman who participated in the group discussions recalled that a grassroots-level leader had embezzled her family's share of the collective funds. Later, the leader got nervous after the anti-corruption campaign was launched, and sent someone to monitor them to prevent them from reporting the case to anti-corruption agencies. She explained, “You can report corruption if you collect some evidence. Although it is not necessarily useful, they [the corrupt officials] get scared in such an anti-corruption climate. However, if the official is incompetent and slack, it makes no sense to file a complaint.” They view citizens as less influential in punishing incompetent officials than corrupt ones at the grassroots level, so they hope grassroots-level officials can be intrinsically competent.
On the other hand, corruption is more consequential at higher levels of government. This is reflected by the larger row differences at the municipality level than at the township level in Table 3. Accordingly, corruption has a larger AMCE at the municipality level (Appendix Table E models 4 to 6). The interviews highlight two reasons. First, people view honest high-level leaders as more valued and corrupt ones as more detrimental. In recognition of these officials' high position in the power hierarchy, citizens believe that high-level officials, if corrupt, are more damaging due to the hierarchical structure of the bureaucracy. A respondent from the private sector said, “There is an old Chinese saying. If the upper beam is not straight, the lower ones will go aslant (shangliang bu zheng, xialiang wai). So, if the mayor is corrupt, he sets a bad example followed by subordinates. Then all levels of government within the municipal jurisdiction [including the municipality-level, the county-level, and the township-level governments] will be decadent.” Several other discussion participants agreed with this sentiment. This reasoning squares well with the idea that corrupt high-level officials will be “unprincipled principals” (Gong and Xiao, Reference Gong and Xiao2017; Brierley, Reference Brierley2020). Also, interviewees mentioned that high-level corrupt officials—due to their greater political power—can extort more than corrupt low-level officials.
Second, a lack of proximity makes it more difficult for ordinary citizens to monitor high-level officials and report corrupt behavior than it is to detect corruption among grassroots-level officials. For instance, the woman mentioned above from a rural area said the village head was so afraid of the intense anti-corruption campaign because she had evidence of his wrongdoing: “It is easier for us to figure out and report the corruption of grassroots officials. On the contrary, it is hard to collect evidence of a high-level official's corruption. Even if we have [the evidence], we don't know where to file a report against such a high-ranking official.” Thus, citizens desire intrinsically honest leaders at high levels of government.
Therefore, contrary to the prediction that citizens dislike corrupt low-level officials more due to their proximity (Section 2.4), the interview evidence described above suggests that citizens are often more tolerant of corrupt low-level officials because the proximity allows citizens to monitor and report their corrupt conduct when necessary. However, it is more difficult, if not impossible, to monitor high-level corrupt officials. Accordingly, citizens are more willing to implicitly trade integrity for competence at the grassroots level than at the high level.
5. Conclusion
This paper investigates why citizens tolerate corrupt officials. Using experimental and interview evidence from China, we examine the prominent corruption–competence tradeoff explanation, which maintains that citizens knowingly tolerate corrupt officials in an implicit exchange for their excellent governance performance. It discusses two main findings that have several important implications.
First, we find that Chinese citizens like corrupt-yet-competent township leaders significantly more than honest-yet-incompetent ones, which is at odds with the intuition that people universally dislike corrupt officials, but strongly supports the corruption–competence tradeoff explanation.
Second, the public's preference for corrupt-yet-competent over honest-yet-incompetent officials disappears when evaluating municipal mayors. The different township- and municipality-level findings indicate that people uphold a hierarchical corruption–competence tradeoff. While we have not explicitly studied government levels beyond the municipality level, this finding suggests that citizens' relative preference for corrupt-yet-competent officials at higher levels (i.e., who are more distant from ordinary citizens and more powerful) is likely even weaker.
The hierarchical tradeoff sheds new light on prior studies that have reported conflicting evidence regarding the corruption–competence tradeoff. Our findings suggest that, net of the differences in contexts and empirical strategies, the salience of this tradeoff—examined in the same country and operationalized similarly across levels of government—varies when people evaluate officials at different levels.
The hierarchical corruption–competence tradeoff also has significant implications for understanding anti-corruption policies and regime support in China and beyond. The public's relative preference for corrupt-yet-competent over honest-yet-incompetent officials at lower levels of government suggests that the ruling party has reasons to avoid fighting corruption at the cost of hampering governance efficiency by, for instance, replacing corrupt-yet-competent officials with those who are honest but incompetent. In local governments where lazy governance becomes noticeable, administrative ineffectiveness will likely counteract the intended positive effect of anti-corruption reforms on enhancing regime support. Yet, citizens welcome punishing high-level corrupt “tigers” so that corruption will not spiral out of control.
Our paper is the first to examine the corruption–competence tradeoff in authoritarian China. We adapted the vignette description of corruption to real politics by choosing a malfeasance that indicates corruption but is not too severe to render our outcome measures (e.g., promotion) unrealistic to respondents. One possible response to our findings is that they are unique to minor corrupt acts that respondents consider trivial. If an official engages in more severe corruption that imposes a higher cost on citizens, they may be less willing to uphold the corruption–competence tradeoff.
While future research may examine this possibility, our findings are unlikely to be entirely driven by the description of corruption for three reasons. First, conceptually, we are interested in understanding citizens' willingness to forgive corrupt officials rather than particular corrupt acts. Our description successfully and equally increased people's belief that the vignette officials at both government levels are corrupt. A corrupt official may engage in other more severe corrupt acts. Second, empirically, we find that respondents are significantly more likely to disapprove of corrupt than honest vignette officials at both government levels. Thus, the corruption behavior we describe is not inconsequential. Finally, our key finding that the corruption–competence tradeoff varies across government levels is less affected by our particular corruption description because we used identical descriptions for township leaders and municipal mayors. If respondents prefer corrupt-yet-competent over honest-yet-incompetent township leaders primarily because the described corruption is too minor, we should expect them to exhibit the same preference when evaluating municipal mayors because people who perceive a corrupt act as trivial for low-level officials will probably regard it as similarly trivial, if not more so, for high-level officials. Yet, contrary to that expectation, we do not find the same preference for municipal mayors.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2023.7. To obtain replication material for this article, please visit https://doi.org/10.7910/DVN/GQFQJJ
Acknowledgement
Both authors contributed equally to this work and have decided to rotate the authorship order in our recent collaborations. We thank Yongshun Cai, Hao Dong, Barbara Geddes, Manoel Gehrke, Chad Hazlett, Fengming Lu, Xing Ni, Ankui Tan, James Tong, Yu Xie, Li-an Zhou, as well as participants in seminars and conferences at Peking, Princeton, and Tsinghua Universities. We also thank Editor Paul Kellstedt and two anonymous reviewers for their constructive comments and suggestions. We are grateful for the generous support of the Institute of State Governance and the Center for Integrity and Governance at Sun Yat-sen University. Zhu and Yang would like to acknowledge financial support from the National Natural Science Foundation of China (Grant No. 71704193) and Peking University (Grant No. 7100603771), respectively. All errors are ours.
