Improved Local Implementation under Central Inspection? Evidence from Social Mandates in China

Abstract

While the principal-agent problem has been challenging to public administration in general, it is particularly acute in China, where implementation gaps abound. This study examines whether central inspection, one important form of central monitoring in China, helps reduce implementation output gaps at the provincial level. Based on an analysis of a unique dataset of provincial governments’ documents formulated to implement central mandates in social policy areas from 2003 to 2017, it finds mixed results. While central inspection does help speed up provincial implementation outputs, it does not improve the quality of provincial implementation outputs.

Keywords:

• Central inspection
• Chinese local government
• implementation gap
• implementation output
• principal-agent problem

Acknowledgments

I want to thank Dorte Sindbjerg Martinsen, Kasper Ingeman Beck, Anders Woller Nielsen, Wiebke Marie Junk, Jørgen Delman, Chunrong Liu and the three anonymous reviewers, for their useful comments. I would also like to thank the Fudan-European Centre for China Studies for its language support.

Notes

1 This article uses both the term “implementation” and “compliance.” However, there are key differences between them. Implementation is a process of putting a policy into practice. In contrast, compliance is a state of conformity between an implementor's behavior and a specified rule. Implementation is typically a critical step towards compliance, but compliance can occur without implementation.

2 In western democracies, media and other forms of fire-alarm monitoring help reduce asymmetric information, thus mitigating the agency problem. In comparison, those forms of fire-alarm monitoring are relatively weak in China, because they face several limitations in their activities (see Anderson et al., 2019 for a discussion).

3 Indeed, the previous solutions work largely via incentive schemes or monitoring. Replacing an agent could be considered as an extremely harsh form of sanction that the principal can use to shape the incentive structure. Co-optation helps reducing monitoring costs and achieving better alignment of the incentive structure of the agent (Sheng, 2009, pp. 75–76). Similarly, administrative procedures also help lowering the costs of monitoring and sharpening sanctions (McCubbins et al., 1987, p. 273).

4 Note however, the recent performance management reform in China aims not only to induce more compliance from the local government, but also to improve citizen satisfaction with government performance (see Zhang et al., 2021 for a discussion).

5 These implementation outputs are important for both citizens and lower-level governments. For citizens, they often have direct influence on their lives. A good example is the college matriculation policy for migrant children (yidi gaokao) reform. Each province had to formulate its own reform plan, as required by Notice of the General Office of the State Council on transferring the opinion of the Ministry of Education and other ministries on children of migrant workers to enter senior high schools and sit college entrance exams locally after receiving compulsory education. These provincial reform plans attracted great attention from ordinary citizens, because they have huge consequences for migrant children’s rights to education (Luo & Jin, 2013). For lower-level governments, these provincial-level implementation outputs provide the needed guidance for them. This is because the provincial government is the gatekeeper guarding and providing access to the local levels (Lieberthal & Oksenberg, 1988, p. 350). In other words, these implementation outputs make the provincial governments’ positions on a central policy known to the lower-level governments, so that they can take actions accordingly (see O’Brien & Li, 1999, p. 71 for a discussion).

6 According to the Regulation on Government Inspection Work, central inspection teams collect information mainly by themselves. However, this is a bit different in the anti-corruption campaign, as the inspection teams rely not only on their own efforts but also a certain degree of fire-alarm (that is, reports from the public). Still, this is very contingent on the initiatives from the inspection teams, so it is not like the independent fire-alarm proposed by McCubbins and Schwartz (1984).

7 The specification of central inspection clearly is not random, but the factors determining it are extremely hard to assess. Therefore, the most plausible way is to include relevant observable control variables in the data analysis, as will be explained later.

8 When a provincial government formulates a document to implement a central opinion, it normally refers to the central opinion in the introduction, using terms such as “to implement (a specific central opinion),” or “according to (a specific central opinion).” This is the criterion for finding the provincial document. In a very few cases where there exist more than one provincial document referring to the central opinion, then the most relevant one was chosen.

9 In total, there are around 22% of missing data. Those missing data can happen mainly due to: (1) the central opinion is not relevant for a province, as this province has already taken initiatives earlier; (2) the province is non-compliant and does not formulate a document; (3) the document exists but is unavailable online. While acknowledging that missing data can be a serious challenge, the author believes nevertheless that it is still the best result one can get in the case of China.

10 A deeper investigation of the missing data shows the number of missing data on the timing measurement is in total 772 (449 not under central inspection, 323 under central inspection). Similarly, the number of missing data on the quality measurement is in total 788 (454 not under central inspection, 334 under central inspection). It seems that the distribution of missing data is rather balanced. If all missing data were present in the group of observations under (or not under) central inspection, then this might create more serious biases in the results.

11 Bo’s data only cover the period before 2012. For the period during the Xi-Li administration, the data were coded manually based on Bo’s (2010, p. 99) system.

12 See Online Appendix 5 for results from a pooled and a non-pooled OLS model.

13 Despite higher political priority, the State Council’s opinions may also be more difficult to implement, because they deal with more challenging problems. This might be the reason why in comparison to the General Office’s opinions, the State Council’s opinions are implemented with a longer time.

14 This may be because during the Xi-Li administration, there was in total a much larger number of central opinions than the previous period. This might create delay in formulating needed policy documents at the provincial level.

Additional information

Notes on contributors

Yi Ma

Yi Ma, Ph.D., holds a postdoctoral position at the Department of Cross-Cultural and Regional Studies, University of Copenhagen. His research interests include policy design and implementation, central-local relations, and soft law governance in China.