Every year during China's Dragon Boat Festival, local governments distribute sticky rice dumplings (zongzi 粽子) to members of their communities. This cheery tradition, long under government management, is not free of challenges. Although zongzi are not expensive, there are often public complaints about uneven or unfair distribution. As a result, some local governments have outsourced the task of distribution to China's growing local CSOs (shehui zuzhi社会组织). Perhaps unsurprisingly, complaints in these localities have decreased. In an interview, one local official explained the logic behind his decision to outsource:
It is a good idea to outsource responsibilities for low-risk but potentially controversial tasks to community organizations. If there are any problems, they will take the blame; if there are no problems at all, they will take care of the cost and we might get the credit. Why wouldn't we be glad to have more of them?Footnote 2
Thinking about zongzi – besides making us hungry – should raise important questions about evolving state–society relations in China. Are different CSOs treated differently? If so, why? How can a collaborative relationship form between the authoritarian state and CSOs? If differentiation of CSOs exists, then what factors or motivations explain it?
The well-known ‘reform and opening’ policy of the Chinese Communist Party (CCP), and the parallel abandonment of Maoist totalitarianism, opened the way for nascent civil society to emerge in the 1980s.Footnote 3 Chinese CSOs have been developing rapidly over the past three decades. Following the 2008 Wenchuan earthquake and the subsequent Lushan and Ludian earthquakes, CSOs and related associational activities have exploded.Footnote 4 Optimistic scholars see such developments as tremendously promising.Footnote 5 According to a report by the Ministry of Civil Affairs, as of 2016, there were a total of 702,000 social organizations – a more than threefold increase since the data became available in 2001.Footnote 6 Among these organizations, 336,000 are social groups, 361,000 are people-run non-enterprise (minban feiqiye 民办非企业)Footnote 7 and 5,559 are foundations.Footnote 8 The government has loosened its control over the CSO sector and formally recognized the important role CSOs and other societal forces play in its five-year plans.Footnote 9 None of this detracts from the reality of state repression that bans some CSOs and controls others through laws, mandatory affiliations, personnel and funding.Footnote 10 Lacking formal institutions and mechanisms that can protect the rights of CSOs, China is still far from having a functioning civil society.Footnote 11
The common theme in the analyses of Chinese civil society's recent development is captured by what this paper calls “deliberate differentiation,” in which the local governments have the tendencies to intentionally treat different types of CSOs, sensitive versus not sensitive, with different policy responses,Footnote 12 although providing very different reasons for why deliberate differentiation exists.
Proponents of diversification of civil societyFootnote 13 and consultative authoritarianismFootnote 14 regard differentiation as the means by which the state obtains information, advice and support from key sectors of the population through a wide variety of consultative and deliberative forms of controls. Meanwhile, the graduated controls approach sees the CCP dividing CSOs into those that might turn antagonistic and problematic and those that can safely be allowed in order to enhance public goods provision,Footnote 15 with some arguing that the state's fear of a CSO's collective action potential may be the key determinant.Footnote 16 Despite this literature, few studies have explored whether such differentiation exists systematically and, if so, which factors or motivations actually drive the state's deliberate differentiation. Some factors may be more important than others, whereas some may be irrelevant. In short, I see an opportunity to test falsifiable claims about the sources of variation in China's authoritarian responsiveness to CSOs.
Drawing evidence from an online field experiment and 63 in-depth interviews with government officials and leaders of social organizations in Sichuan Province, this study demonstrates first, that policy differentiation towards CSOs exists and second, that at the local level differentiation aims primarily to increase productivity and outsource responsibility for public goods provision. While higher levels of government and more urban localities may be focused on collecting information from CSOs or disciplining potential regime opponents, local governments concentrate more on getting the most out of “good” CSOs than repressing “bad” ones. Under a decentralized system, grassroots county-level officials have the freedom and motivation to allow various social organizations to exist – particularly those that provide public goods and contribute to the stability of the regime. Local officials encourage and facilitate CSOs to share the workload and are happy to let the groups face public criticism if necessary. Furthermore, contrary to what Xueguang Zhou suggests,Footnote 17 self-organized grassroots CSOs are not seen as sources of spontaneous collective action, but as new sources of bureaucratic support. In sum, the variation in treatment of CSOs is real and determined in different ways than existing theories suggest.
The term “outsourcing responsibility” in this paper has two dimensions: shouldering the burden of workload and being accountable in case of potential retribution. It is clear that local officials not only intend to solicit help from the “good” CSOs, but also to ensure the organizations take the blame when things go wrong. Therefore, what is outsourced to CSOs is not only the responsibility to assist local governance, but also the responsibility to protect the reputation of the local state by shielding it from blame.
Thus, the deliberate differentiation approach by the local Chinese governments is not merely tactics to collect information, to manage the society with a variety of tools (consultative authoritarianism) and to control and deter threats (graduated controls), but is intended to have “good” CSOs to be much more integrated in its local governance. The logic is similar to why township governments would need to set up village governments, even though the official five-level bureaucracy (Central, provincial, municipality, county and township) does not include the village level. “Good” CSOs are intended, just like village governments, to provide services, shoulder responsibilities, buffer criticism and channel controls.
At the level of theory, this research contributes to the literature on authoritarian responsiveness and authoritarian resilience. Responsiveness is a key marker of the local state's treatment of a CSO, and regime resilience may be an outcome of how shrewdly it manages civil society. There is a growing body of literature that aims to explain authoritarian responsiveness, whether in terms of fearing collective action,Footnote 18 gaining support from loyal insiders,Footnote 19 pleasing higher-ups for better job prospects,Footnote 20 collecting information,Footnote 21 or setting up mere window-dressing channels to increase public satisfaction.Footnote 22 However, the cases selected by existing studies tend not be systematic. When analyses are more systematic, including experiments, they are typically at the national level and assume that policy implementation is quite centralized. Most field experiments on related topics, though pre-tested, also do not originate treatment conditions from what actual individuals in such scenarios would do (that is, the sender or the receiver of the information).
The experiment in this research, thus, tests differentiation factors through government response to registration efforts by new CSOs. The language from pre-experiment interviews with real CSOs trying to get registered and newly registered CSOs was used directly to construct the treatment condition. Follow-up interviews were also conducted to investigate the underlying logic of the experimental results.
This research finds that the government is less responsive to politically sensitive CSOs, with slower and lower-quality responses to their requests for information, and is more responsive to those regime-supporting CSOs that could increase the bureaucracy's productivity and shoulder the burden and responsibility of the local state. Getting the help from regime-supporting CSOs would strengthen the state. The government's past interactions with leaders of such organizations also matter in this relationship. Sometimes, after interacting with a questionable CSO, government officials decide they would rather keep it close (perhaps by providing it office space at incubation centres) rather than risk setting it loose without supervision. Also, this research confirms that such relations are not static, as both the government and civil society organizations interact with each other and adjust their policies and behaviours accordingly.Footnote 23 These learning experiences create a dynamic, interactive process for both the state and the society.
With respect to authoritarian resilience, the contribution here is to suggest that an authoritarian regime might deliberately adopt some democratic institutions not because it is forced to by a contentious society or international actors, nor as baby steps towards more thoroughgoing political reform, but rather as tools to improve governance in the service of perpetuating control by the ruling elites. This article also makes a quantitative contribution to the vigorous debate among China scholars about the nature of the rights consciousness of Chinese citizensFootnote 24 by focusing specifically on the unacceptability of rights claims. The distinction demonstrated here is that, while organizing is acceptable, demanding rights is not – although how rights-claiming is framed still matters.Footnote 25
Research Design
To confirm whether such policy differentiation exists and to explore the motivations for it, I conducted in-depth interviews and a field experiment. The experiment assessed, via email enquiries, local governments’ relationships with CSOs.Footnote 26 Rival hypotheses were tested to demonstrate whether the state's main motivations were collecting information and preventing collective action or increasing productivity and outsourcing responsibility.
Interviews
Interviews are a key data source for this study, and 17 government officials and 46 CSO leaders were interviewed from 2012 to 2018. The officials interviewed were in charge of civil affairs or CSOs at various levels, including village, township, county, municipal and provincial positions. The interviews discussed the role of CSOs, state responses to CSOs, the past and future of the CSO sector, and policy towards new grassroots CSOs. CSO leaders were interviewed through referral by local scholars and academic institutions. Several of the leaders were interviewed multiple times. The overwhelming majority of the CSOs were not officially registered during the preliminary interviews and either had sought or were seeking registration.
Preliminary interviews revealed that organizations working with vulnerable populations (such as mothers who lost their only child) and with groups claiming rights (such as workers owed significant backpay) were politically sensitive.Footnote 27 The government dislikes such organizations due to their potential to organize collective action, challenge the state's legitimacy, and disturb the stability of the regime.Footnote 28 Such organizations are perceived as anti-government and tend to receive no support (if not hostile treatment) during registration within their jurisdictions. This was a consistent finding across all levels of government, though officials may frame the differentiation in slightly different ways, calling one kind of CSO the “rights claim and advocacy group” and the other the “service provision only group.”Footnote 29
Less sensitive, yet still eliciting caution, are religious organizations and international organizations. According to several officials, “Religious organizations are ideology driven and provide an alternative to our socialist ideology.”Footnote 30 Even when dealing with religious organizations, however, officials suggested that local and more peaceful religions (here mainly referring to Buddhism and Daoism) should be promoted in place of foreign religions with anti-state potential (such as Christianity and Islam).Footnote 31 Christian and Islamic groups are seen to have more current foreign connections and therefore more subversive inclinations both on the ideological and security fronts.Footnote 32 Such differentiation based on political sensitivity informed the experimental design.
The experiment
Methodologically, this research resembles audit studies, which also measure discrimination directly with experimental fieldwork.Footnote 33 To assess the political logic of how different CSOs are treated, I sent emails to county governors to enquire about new CSO registration. This strategy was chosen because it is realistic and convenient. In 2007, the State Council announced the “Open Government Information Ordinance” (OGI), which required governments at the county level and higher to increase transparency, leading many local governments to establish official websites.Footnote 34 Most counties have a “county governor's inbox” so that people can voice their complaints. By emailing county governors, I could conduct a randomized experiment aimed at revealing the underlying logic of the policy behaviours of the Chinese authoritarian state.Footnote 35 In addition, enquiring about starting a new CSO is something done commonly at the local level, according to the CSO interviews. At the county level, administrative heads are still directly involved with CSOs. By looking at the creation of CSOs, this research also offers a unique perspective on how a nascent civil society is perceived and dealt with in China.
The internet in China is an important platform for public debate, problem articulation, and new kinds of protests.Footnote 36 The government is not promoting the internet to build participatory democracy, rather it sees the web as a useful tool for promoting development, setting policy agendas, supervising its bureaucracy, and increasing public legitimacy.Footnote 37
The CCP supports the internet as a legitimate yet controlled channel for communication between China's government and its people. There are many reasons for this, including the rising cost of stability maintenanceFootnote 38 and the need to acquire reliable information about the populace without face-to-face confrontation.Footnote 39 The cadre responsibility system motivates local officials to collect accurate information in the form of public feedback to improve their job prospects.Footnote 40 Therefore, the county governor's inbox takes not just complaints but also policy suggestions. A recent study found that officials are similarly receptive to citizens’ suggestions submitted through traditional channels and the internet, provided that there is no perception of hostile intent.Footnote 41
During the randomization process, each of the 114 county governors' inboxes had a 50 per cent chance of being assigned to the treatment group. In order to include a robust and less ambiguous treatment in the experiment, I included in the treatment email the phrase from the interviews, “organization … protect vulnerable groups’ (ruoshi qunti 弱势群体) individual interest and citizen rights (gongmin quanyi 公民权益).” The control group received an email enquiring about the process of potentially starting a social organization within the county's jurisdiction and asked the county governor to point the sender to the right resources. All of these words were meant to be politically sensitive and to indicate a misalignment of the government's and the CSO's interests and goals (see Appendix 1).
CSO leaders have reported that officials tend to be more responsive and supportive when the leaders are speaking the same “language.”Footnote 42 Preliminary interviews suggested that most CSOs (whether officially registered or not) have some connections with local governments, and such connections can be personal (shared experiences or strong ties) or institutional. CSOs actively learn about government documents and “key phrases” when interacting with the government. Therefore, the first few sentences of the emails reflected sender awareness of ongoing political developments and familiarity with newly issued government documents. This tactic made officials more likely to take the email seriously and gave a better approximation of what a CSO leader looking to register an organization would say. By conveying political sensitivity in a context-appropriate way, the treatment avoided being provocative in a way that might lead officials to doubt that the requests came from their localities.Footnote 43
The assignment of treatment was done before the emails were sent, so it is assumed that the variation in the response time, rate and content can be attributed directly to the treatment rather than other potential confounders related to the experiment. Since in China officials of the same bureaucratic rank generally avoid speaking to one another for fear of being accused of conspiringFootnote 44 and no major official meetings were ongoing during the period of the study, one can also safely assume that there was minimal interference between the control and the treatment groups. The task was designed to be minimally burdensome for the officials (or their office) as it increased their workload very little. Such a minor request also ensures that future researchers’ ability to utilize similar approaches will not be infringed.Footnote 45
Doing the experiment at the county level allowed me to see policy variation more clearly and have sufficient observations. Preliminary interviews with local officials revealed that policies related to civil affairs are mainly decided at or above the county level. Below the county level, officials are primarily involved in implementing policies. Policies sometimes vary at the township level, but the differences are mostly in implementation.
Sichuan Province was chosen as the site for the experiment and the interviews because the massive 2008 earthquake triggered CSO development in the region, potentially increasing incentives for government responsiveness to CSOs. Since “not replying” in this experiment is treated as “one type of response” rather than missing data, the assumption is more valid in a region where the local government is generally expected to respond to emails of this kind. Therefore, a “non-reply” is more a deviation from the expectation. Previous studies of authoritarian responsiveness in China suggest the national response rate is around 30 to 40 per cent with the baseline group response rate being around 30 per cent.Footnote 46 Various treatments, such as the threat of collective action or reporting to superiors can increase the response rate. The response rate in the control group from Sichuan Province in this study was 60 per cent, a figure much higher than that of existing studies. This is probably because we developed the treatment condition directly from interviews with leaders of non-governmental organizations (NGOs). Government responses may be in the form of policy outcome, direct action, or information provided.Footnote 47 As a standard practice in this literature for ethical and practical reasons, this study only requested information for the benefit of the sender.
There are 165 county-level administrative regions in Sichuan Province. Seventeen of their website inboxes required a personal identification number to send an email (which the government can trace to the sender). Another 26 had pages where the email function did not work properly, or the link did not exist. There were eight websites that selectively posted responses from past emails, but the emails had not been updated within the prior six months. In this study, for obvious reasons, I chose to leave out those websites that required a personal identification number.Footnote 48 I also discarded the non-updated websites, as well as those without functional email inboxes. With the remaining 114 counties, it is reasonable to assume that, ceteris paribus, we could expect similar potential responses from them. The above exclusion process may affect the study's external validity since those offices that require an identification to send an email or do not have a properly functioning email mechanism may be different in their bureaucratic capacity or local conditions.
Since the sample size was not extremely large, it was worth checking the assignment outcome to see whether the procedure, ex-post, produced treatment groups correlated with county characteristics. Table 1 reports the results of a logistic regression of each county being randomized into the treatment group on eight different county characteristics, including area, number of families, number of townships, number of communities, population in units of 10,000, work population, agricultural population, and number of firms.Footnote 49 These geographic, demographic and socioeconomic covariates were also included in various models during analysis when assessing the duration of response and quality of response since such indicators may reflect governance capacity, request frequency and other confounders that should be controlled.
Logistic Regression of the Treatment Assignment
Notes:
t statistics in parentheses.
* p < 0.10, ** p < 0.05, *** p < 0.01
As expected, the results indicate that the county characteristics were not predictors of the treatments assigned, individually and jointly. The standardized differences for stratified comparisons would show the similarly balanced assignment of treatment having fewer assumptions.Footnote 50 For the 114 counties, 55 were assigned to the control group, and 59 were assigned to the treatment group. The emails were sent during the weekend of 17 August 2014.
Results and Analysis
As Table 2 shows, among the 55 counties in the control group, 33 responded and 22 did not. Among the 59 counties in the treatment group, 22 responded and 37 did not. The total response rate by officials was not very different from the results of similar designs utilized by other scholars.Footnote 51 Of the 55 total responses received, the majority were received during the first ten days. As we can see from Figure 1 below (response by the number of days), the number of responses also diminished as time went on.
Response Table by Treatment Assignment
Response Frequency by Number of Days
The estimated treatment effect (difference in means) is -0.23 while the p-value for the t-test result is 0.0151. The reply rate from the control group and the treatment group are statistically different with political sensitivity leading to a lower response rate from the county governors.Footnote 52 The effect can also be estimated through permutations. With statistical packages developed by Bowers, Fredrickson and Hansen, and with 1,000,000 permutations, a confidence interval was constructed with the upper bound of −0.05 and lower bound of −0.40 for the effect.Footnote 53 This means that when the treatment was included, the probability of getting a response from government officials will be reduced by 5 to 40 percentage points. The duration of the reply time also varied among the responses (see Table 3). It took the treatment group longer to respond than the control group. Here the response duration variable is an ordinal variable based on reply time. “No response” technically means it took an infinite number of days for a response, but on a continuous scale, disproportionate influence will be created from these observations. Therefore, an ordinal variable was created to turn the number of days into seven ordered categories. Grouping by week also reflects the typical work cycle as the expectation of response during the weekend can be different from that during the weekdays. The variable is therefore coded from 0 to 6. Response within 1 day is coded as 6; 2–7 days is coded as 5; 8–14 days is coded as 4; 15–21 days is coded as 3; 22–28 days is coded as 2; 29 and above is coded as 1; no response is coded as 0. Since the latest response was received on the 35th day, each category covers one week. A smaller number indicates longer response time.
Experimental Results by Different Dependent Variables
Note:
* p < 0.10, ** p < 0.05, *** p < 0.01
The response speed models suggest that, when sensitive CSO requests were made, the negative treatment effect was present in all models (see Appendix 4). Therefore, politically sensitive CSOs had to wait longer for a reply. When a reply was received, there were several different kinds. The quality of the responses was incorporated into the study as some responses were less meaningful than others. It is important to note that non-response in this study has meaning and should not be treated as “data missing.” Simply responding by seeking more information from the sender has less quality (and different intention potentially) than detailed answers. Referring the question to another person also has less quality than directly answering the question. Therefore, the response quality variable is coded with the instructions shown in Table 4.
Instructions for Coding Response Quality
Such coding is ordinal because each higher number satisfies, and approves upon, the conditions of the previous number. The basic ordinal logistic regression indicates that the treatment email negatively and significantly affected the quality of the response. In other words, there were responses received from both the control and treatment groups, but the quality of response differed between the two groups. All but one detailed personalized response was from the control group. A research assistant who was not aware of the details of the research performed an inter-coder reliability check. The result was a Cohen's Kappa inter-coder agreement of 99.24 per cent.
Appendix 2 presents estimates of the treatment effects controlling for other covariates, such as county area, population size, worker population, agricultural population, and the number of townships, families, communities and firms. The treatment effect of lower response quality is significant across models. This result indicates that replies to requests from sensitive CSOs tend to be less meaningful. The population size of the county is positively correlated with responsiveness. This point may require further investigation, as the size of the population may be directly related to the size of the staff and, therefore, the capacity to govern.
A simple word count was also used as a more direct way of approximating quality of response. Since “no response” has meaning in this study, the word count was turned into an ordinal variable in which no response was coded as 0 with successive categories covering a range of 500 words (that is, 1–500 was coded as 1; 501–1000 was coded as 2, etc.). The quality of response was consistently lower when more sensitive CSOs made requests. As in the coded model, the word count model indicates the negative treatment effect (see Appendix 3).
The effects, if the experiment can reveal any, may be underestimated because the government has incentives to allow some politically sensitive organizations to operate. The option of being able to crush such organizations would not only set an example to others but also be seen by officials as a routine exercise of their authority. One government official, describing hostile actions against religious organizations (such as taking down crosses from the top of churches in Zhejiang Province), said that the government's actions are sometimes purely for the purpose of testing the effectiveness of its policy commands. Similarly, officials may test their authority and power with CSOs they do not like.Footnote 54 The literature also suggests that authoritarian governments tend to want to collect information about their bureaucracies, as well as about potential threats.Footnote 55 Therefore, even a small treatment effect is worth examining because it is likely to be underestimated, which supports the argument of this research.
These experimental results differ from, and are complementary to, a recent national study in China by Chen, Pan and Xu, albeit with different treatments.Footnote 56 The key difference is that Chen's study introduced an “immediate” threat, while the threat this research may pose to the government is in the long run (social organizations might form, and in the future, organize collective action). This contrast provides potential evidence, which may need further exploration in future research, for a temporal dimension in authoritarian response to political threats (whether it is collective action, bureaucratic competitiveness, insufficient loyal support or lack of information). The response rate will be lower when the potential threat is not imminent.
In addition, it has been assumed in the existing literature that authoritarian responsiveness attenuates or eliminates potential threats. The treatment condition in this research is unique because responsiveness by the government might enlarge the threat by assisting the regime-challenging CSOs. The treatment can also tease out whether officials are more interested in avoiding political sensitivity and potential collective action or in collecting information. If they are more interested in collecting information, then the response rate should be higher in the treatment group; if they are more concerned about the potential for collective action, then the response rate should be lower in the treatment group. The results accord with King, Pan and Roberts’ finding that the Chinese government is more concerned about potential collective action even at the expense of losing information.Footnote 57
A Dynamic Process of Deliberate Differentiation
The preliminary interviews with government officials in Sichuan Province revealed a common theme that they are not worried about sensitive organizations because there were plenty of constraints to prevent their existence; what officials needed was help from “good” organizations:
We treat different social organizations differently. However, we are confident that the existing institutions are sophisticated enough to deter those “troublemakers,” who are quite rare in rural areas like where we are – you would find many of them in Beijing or Guangdong but not here. What we want is to motivate people who have time, money, and the will to initiate their organizations to provide public services complementing the government. The government is not good at everything, and we are tired of being the judge and the athlete at the same time. We would be happy to only play the role of the judge and let social organizations use their expertise to contribute to the society. The problem is that we do not have enough of those social entrepreneurs.Footnote 58
Government officials’ policy focus, therefore, was not eliminating regime-challenging groups, but cultivating potential partners among regime-supporting groups to extract productivity while outsourcing responsibility to those partners. In other words, the primary concern for many officials was not the fear of lack of information and the control of politically sensitive groups, but the lack of sufficient support from service-provision groups that could ease the burden of the local governance.Footnote 59 If a group is seen as politically non-sensitive and has the potential to provide the services needed by the locality, and to share the responsibility of the local government, that group is likely to be invited into a mass organizations centre (quntuan zhongxin 群团中心) or an incubation centre, where the actual productivity of the group will be monitored and assessed. Through this process, the local governments are able to first identify the regime-supporting groups through the differentiating strategy, and then cultivate the groups with higher productivity to form partnerships and outsource responsibilities.
Such policy differentiation was not always the case; before 2008, there were limited CSOs operating in China, more commonly referred to as NGOs by officials. The translation in Chinese – feizhengfu zuzhi 非政府组织 – often made government officials misunderstand them as “anti-government organizations.” Local governments frequently had an antagonistic attitude towards the CSO sector as a whole.Footnote 60 There was not much differentiation at the time, and all of the organizations faced similar control, manipulation and penetration from the government. To reduce the interference from local governments, CSOs would often register as private companies or keep a low profile by not registering at all.
Since the 2008 Wenchuan earthquake, local governments have had the opportunity to witness closely the work done by CSOs during quake relief efforts; some CSOs did an impressive job providing social stability and local governance. Many local governments have since realized how valuable such “good” social organizations can be, especially if CSOs do not take resources (money, time, staff, etc.) away from the government.Footnote 61
Why would local governments not profess concern about regime-challenging CSOs and the need to actively and heavily monitor them? Officials interviewed typically pointed to existing institutional constraints. Before the 2016 charity law, domestic organizations needed to have a “sponsor organization,” either a government department or government-organized non-governmental organization (GONGO), to be registered.Footnote 62 Such “sponsor organizations” could usually threaten to shut down the operations of the CSO by rejecting its annual budget should the CSO cross any lines.
Even after the 2016 charity law, CSOs still need to submit annual reports. A local government might keep an “abnormal list” to identify individual organizations with bad records. The “Draft Regulations for the Registration of Social Associations” accompanying the new charity law further shows that the Chinese government intends to provide the necessary leeway, but also retain constraints, within which “good” social associations need to operate.Footnote 63
Another powerful constraint on CSOs is financial support. The 2016 charity law did not grant public fundraising rights to social service organizations or CSOs in general. This means that CSOs mainly get money from foundations (usually government-affiliated) or through governmental purchase of services.Footnote 64 Because the new foreign NGO law has cut off foreign support for local CSOs, the government has even stronger control over CSO funding.
Besides legal status and funding, the government has also reshaped people's behaviours. Members of CSOs (whether regime-supporting or potentially regime-challenging) said in interviews that they intentionally put on a cooperative face when interacting with the government. Such behaviour change was evident from the 2008 Wenchuan earthquake to the 2013 Lushan earthquake. Many of the CSO leaders have since formed the habit of reporting to local governments and putting themselves on the record (bei'an 备案) before starting their operations.Footnote 65 For example, scholars have documented positive changes happening in environmental protests across China, capturing how the anti-PX (paraxylene) campaign might be changing the landscape of state–society relations in China.Footnote 66 However, in interviews, almost all environmental groups mentioned that they have made the conscious decision to pick their fights and no longer want to be involved in anti-PX campaigns since the government is sensitive to those specific protests.Footnote 67 Such adaptation is consistent with Hildebrandt's argument that CSO emergence in China may not weaken the state but could effectively strengthen it.Footnote 68 CSO leaders reported such practices to be beneficial because they made local governments less suspicious of CSO activity.
But the Chinese government is more wary when it is less able to verify what CSOs are doing and is currently testing new regulations and constraints to uncover where CSO funding comes from, who is operating what projects, and what the actual objectives of a group might be. The new “Foreign NGO Management Law”Footnote 69 is one example. Within the legal domain, we see a differentiation between domestic and foreign CSOs. Control over domestic organizations seems to be loosening while control over foreign NGOs is tightening.Footnote 70 That being said, there are exceptions to these trends, and the situation is a dynamic one. There were 221 foreign NGOs officially registered in China during the first nine months of 2017, and 291 events operated by foreign NGOs have been officially documented.Footnote 71 We could potentially see an increase in the number of “good” CSOs helping out governance in China despite the concern of foreign influence.
Differentiating CSOs is a dynamic process, proceeding in three overlapping stages: an ambiguous first meeting, initial differentiation, and the decision whether or not to outsource responsibility. In stage one, the government is exposed to a new CSO. When officials are unsure of an organization, they like to keep it close by, perhaps by providing the group with office space at incubation centres.Footnote 72 Several cities in China have piloted social service organization incubation centres that nurture and assist newly organized grassroots organizations, providing resources that the groups desperately need but also making sure the government gets to know them. The government, the party, or other entities trusted by the government may organize the incubation centres.
Once they know more about a CSO, the officials start to differentiate based on the logic and actions described above. However, differentiation does not necessarily mean repressing all regime-challenging organizations. A leader of one rights-claiming group mentioned that he is very clear about the type of reaction he will face, based on the number of people he organizes to hold a meeting or protest. As long as the number of people is less than 200 and no major political event is ongoing, the government will not care too much. Of course, this measure would vary by locality, but it indicates that even with anti-government collective actions, participation needs to be large enough before the government takes action.Footnote 73
CSOs can also utilize stage one and stage two of the differentiation process to create a good relationship with local governments and adjust their operations so that they are seen to be regime-supporting rather than regime-challenging. Natural disasters, like Sichuan's major earthquakes, can be catalytic for increasing trust and building relationships between CSOs and local governments.
The gradual adaptation of CSOs to state preferences is also apparent when comparing interviews from 2011 and 2017. In 2008, CSOs were not as organized and would send people directly to the quake-stricken region to provide services, instead of first coordinating with other CSOs or with the government. After the government regained control of the localities in the immediate aftermath of the disasters, some CSOs faced severe scrutiny and interference while others started to build trust and a working relationship with the local government. By 2013, many CSOs had adjusted their working procedures and would voluntarily register with local governments to report about their intended activities before entering a jurisdiction. This comforted many government officials and built trust between the government and the organizations.Footnote 74 The CSOs also reported that by voluntarily becoming subordinate (or at least appearing to be subordinate), they enjoyed a much larger space to conduct their projects, and fewer suspicions and confrontations arose.
Voluntary subordination leads CSOs to adjust their goals, as well as the way they conduct their projects. In interviews, several religious organizations emphasized that during the quake relief period, they entered the quake-stricken region by abandoning or downplaying the religious aspects of their organizations.Footnote 75 For example, a Christian organization involved in relief focused on helping the local people, not spreading the gospel. Adjusting one's goals to align with those of the government and eliminating dis-aligned projects might help a CSO survive in an authoritarian regime like China's, but it, unfortunately, makes CSOs less diverse.
Local governments also realize that maintaining a bureaucracy is very costly. Since the tax and fees reform, local governments’ sources of income have been reduced tremendously. For tasks such as quake relief, it may not be realistic to keep a government-run team. For other tasks, such as mental-health assistance, the government lacks expertise. Therefore, it is better to let the CSOs take care of such tasks, and the government can provide money to support the CSOs.Footnote 76
The differentiation of CSOs may not solely depend on the issues the CSO is promoting, because CSO leaders’ personal backgrounds and past government-CSO interactions can have a significant impact. For example, one CSO in the study is led by a married couple, and the husband is a retired military officer. In interviews with his counterparts in the government, officials mentioned that such organizations run by former government officials or retired military officers are more trustworthy and easier to work with than other organizations. The authoritarian government sees democratic activities facilitated by this CSO (such as community deliberation, freely elected autonomous governance groups, etc.) as promoting the stability of local communities and therefore condones the efforts, even if the activity is creating an alternative source of power and decision-making mechanism. The organization now trains government officials routinely on facilitation and deliberation skills. In this way, the CSO passes on expertise to government officials thereby gaining more government trust along the way.Footnote 77
It is worth pointing out that only the CSO leader's personal expertise and reputation, not the CSO's role in society, have been legitimized in the eyes of the local government. This phenomenon explains why, during the county governor email experiment, several responses sought to learn more about the individual behind the request and the details of the project before making any decisions regarding the organization.
Conclusion
In this study, I used an online field experiment to establish that the authoritarian Chinese state allows some CSOs to exist (under a zone of indifference) and varies its treatment of CSOs systematically depending on whether the organizations and the government share policy goals and the burdens of maintaining stability. This study adds to the graduated control and consultative authoritarian approaches by revealing that local governments differentiate among CSOs less to collect information and assert control over regime-challenging groups, and more to exploit productivity and outsource responsibility to regime-supporting groups. These findings clarify the relative importance of the various governmental motivations examined in the existing literature.
The qualitative data also capture the dynamic process of deliberate differentiation by demonstrating three overlapping stages of the adaptive relationship: ambiguous treatment, initial differentiation, and outsourcing responsibility to regime-supporting CSOs. When unsure, the government would rather keep organizations close by to observe and understand their real objectives, hoping to turn some groups into useful partners. For this reason, officials send groups to incubation centres and provide them with resources. With organizations whose goals are misaligned with those of the government, officials also start by leaving the group in the zone of indifference. If those CSOs reach a critical point or begin to cause trouble, the government will exercise its authority by cracking down. When the government has an excellent relationship with an organization, it seems to be due more to the CSO leader's individual history and ties to the government than any governmental desire to permanently support the organization itself. The CSOs that are eventually allowed to operate are those that share the burden of the local regime and (at least appear to) subordinate themselves to the government. There are, however, some compromise positions, such as the semi-institutionalization of certain useful organizations with the Communist Youth League as a liaison.
It is clear that whether a CSO is regime-challenging or regime-supporting is not the only determinant in the local government's decision to allow or repress it. The type of group did not motivate the ultimate outcome as much as the group's actions and the potential for outsourcing responsibilities. There are more variation and diversity of differentiation than previously described by the literature and the CSOs have much more agency in final outcomes than that imagined. Further research and observation are needed to see whether this trend is temporary or might lead to the eventual institutionalization of the CSO sector in China.
Biographical note
Taiyi SUN is an assistant professor in political science at Christopher Newport University, USA.
Appendix 1: The Email Sent in the Experiment in English Translation and Original Mandarin Chinese
Dear County Governor,
Greetings!
After studying the “Chinese Communist Party's Third Plenum decision about deepening reform” and reading “to ask social organizations to set up various enterprises in villages,” “motivate the vitality of social organizations,” and “let social organizations provide public service” and other quotes, I am greatly encouraged. I want to enquire about the procedure of how to set up a social organization in the county, whom I should contact, and what materials I should prepare? Are there any existing social organizations that you can refer me to? I particularly wish to enquire about how to start an organization to protect vulnerable groups’ individual interest and citizens’ rights. Footnote †
Thank you very much for answering my questions.
Please reply to my email: xxx@xxx.com
Thank you.
尊敬的领导:
您好!
学习了《党的三中全会关于深化改革的若干决定》里提到的 “社会组织在农村兴办各类事业”、“激发社会组织活力”、以及 “交由社会组织提供公共服务” 等词条我深受鼓舞。特别想请教一下咱们这组建社会组织的流程是怎么样的, 可以找谁来办理此事, 需要哪些材料？已经有哪几个组织可以学习、借鉴的？我特别希望了解如何办一个社会组织来维护弱势群体的个体利益, 保障公民的权益。
非常感谢您百忙之中为我答疑。
请回复我邮箱: xxx@xxx.com
谢谢!
Appendix 2: Ordinal Logistic Regression of Response Quality
Appendix 3: Response Quality by Word Count
Appendix 4: Response Speed by County Officials
