1 For the most recent comprehensive review on VCEs in rural China, see O'Brien, K.J. and Han, Rongbin, “Path to democracy? Assessing village elections in China,” Journal of Contemporary China, Vol. 18, No. 60 (2009), pp. 359–78CrossRefGoogle Scholar.

2 Manion, M., “How to assess village elections in China,” Journal of Contemporary China, Vol. 18, No. 60 (2009), pp. 379–83CrossRefGoogle Scholar.

3 O'Brien and Han, “Path to democracy?”

4 Besides Manion's piece, see Tan, Qingshan, “Building democratic infrastructure: village electoral institutions,” Journal of Contemporary China, Vol. 18, No. 60 (2009), pp. 411–20CrossRefGoogle Scholar; Alpermann, Bjorn, “Institutionalizing village governance in China,” Journal of Contemporary China, Vol. 18, No. 60 (2009), pp. 397–409CrossRefGoogle Scholar.

5 On how transparent VCEs have improved political accountability and political trust, see Manion, M., “Democracy, community, trust: the impact of elections in rural China,” Comparative Political Studies, Vol. 39, No. 3 (2006), pp. 301–24CrossRefGoogle Scholar. On how villagers have responded to electoral institutions in their participation, see Shi, T., “Voting and non-voting in China,” Journal of Politics, Vol. 61, No. 4 (1999), pp. 1115–39CrossRefGoogle Scholar. On how VCEs have improved local governance, see Luo, Renfu, Zhang, Linxiu, Huang, Jikun and Rozelle, Scott, “Elections, fiscal reform and public goods provision in rural China,” Journal of Comparative Economics, Vol. 35, No. 3 (2007), pp. 583–611CrossRefGoogle Scholar.

6 On the bottom-up momentum of VCEs, see O'Brien, K.J. and Li, Lianjiang, “Accommodating ‘democracy' in a one-party state: introducing village elections in China,” The China Quarterly, No. 162 (2000), pp. 465–89CrossRefGoogle Scholar. On different political actors with conflicting interests in VCEs, see Kennedy, John James, “The face of ‘grassroots democracy' in rural China: real versus cosmetic elections,” Asian Survey, Vol. 42, No. 3 (2002), pp. 456–82CrossRefGoogle Scholar.

7 On the role of clan and lineage organizations in VCEs, see Xiao, Tangbiao, Zongzu zhengzhi: cunzhi quanli wangluo de fenxi (Politics of Clans and Lineages: An Analysis of the Nexus of Power in Village Governance) (Beijing: Shangwu yinshu guan Press, 2010)Google Scholar. On the role of “evil forces” in VECs, see He, Xuefeng, Xiangcun zhili de shehui jichu (Social Foundation of Rural Governance) (Beijing: China Social Sciences Press, 2003)Google Scholar.

8 Manion, “How to assess village elections in China.”

9 The 2002 survey was part of the Asian Barometer Survey I project, examining political culture, attitudes and behaviour in Asia. The 2005 survey was sponsored by the Ministry of Civil Affairs and the Carter Centre, examining grassroots democracy and governance in rural China. For more information on the projects, see www.asianbarometersurvey.org and www.chinaelections.org respectively.

10 With a few exceptions, most quantitative studies on VCEs examine a large number of villages within only a few selected provinces. There are 325 villages from 25 provinces in Shi, T., “Economic development and election in rural China,” Journal of Contemporary China, Vol. 8, No. 22 (1999), pp. 425–42CrossRefGoogle Scholar, and 961 villages from 22 provinces in Hiroshi, Sato, “Public goods provision and rural governance in China,” China: An International Journal, Vol. 6, No. 2 (2008), pp. 281–98Google Scholar. Nevertheless, Shi only evaluates VCEs based on whether multiple candidates are used, and Hiroshi's data have little information on the quality of VCEs.

11 A stratified multistage area sampling procedure with probabilities proportional to size measure was used to select the sample in 2002. This is a national representative sample covering both urban and rural areas. Since VCEs are the focus, only the rural subsample is used in this article. In rural areas, the primary sampling units (PSUs) were counties, the second sampling units (SSUs) were townships and the tertiary sampling units (TSUs) were administrative villages. Altogether, 253 villages were selected from 22 provinces. A village questionnaire was used to collect information on various aspects of the selected villages, including the electoral procedures of the most recent VCEs. A total of 241 village questionnaires were completed and returned. The same stratified multistage area sampling procedure was used in 2005. The 2005 sample was a national representative sample only for rural areas. The PSUs were counties, the SSUs were townships and the TSUs were administrative villages. Altogether, 379 villages were selected from 27 provinces and a village questionnaire similar to that in 2002 was used. All 379 questionnaires were returned in 2005. Sampling weights have been adjusted accordingly for analysis, following best practice in analysing complex survey data.

12 Manion, “How to assess village elections in China,” p. 380.

13 Though these two surveys do not follow a panel design, at the national level they are legitimate repeated evaluations of VCEs in rural China across time. For methodological justification on this longitudinal evidence, see Chambliss, Daniel F. and Schutt, Russell K., Making Sense of the Social World (Washington, DC: Pine Forge Press, 2010), pp. 31–37Google Scholar

14 Since the survey was not designed for examining electoral institutions, some critical aspects, e.g. using roving ballot boxes, were not included. Nevertheless, if the quality of electoral institutions is evaluated all-inclusively on whether all procedures are transparent, this article provides a liberal estimate.

15 In the 2002 survey, multiple adopted measures were recorded for organizing village election committees, nominating candidates and identifying finalists. However, in 2005 only one measure was recorded for such procedures. Therefore the 2005 data could overestimate the percentage of villages strictly following the revised OLVC, due to the possible adoption of both legitimate and undemocratic measures. However, these are the best data now available.

16 O'Brien and Han suggest that primary elections have spread to 26 provinces. However, this conclusion is based on election procedures and legal documents made by different provinces, rather than practice in reality. O'Brien and Han, “Path to democracy?” p. 364.

17 Here I focus on the regional variance to facilitate the dialogue with contemporary literature.

18 For detailed arguments on this, see Shi, “Economic development and election in rural China”; Hu, Rong, “Economic development and the implementation of village elections in rural China,” Journal of Contemporary China, Vol. 14, No. 44 (2005), pp. 427–44CrossRefGoogle Scholar.

19 For detailed arguments on this, see Kennedy, “The face of ‘grassroots democracy' in rural China”; O'Brien, K.J., “Implementing political reform in China's villages,” Australian Journal of Chinese Affairs, No. 32 (1994), pp. 33–59CrossRefGoogle Scholar.

20 For detailed arguments on this, see Tangbiao Xiao, Politics of Clans and Lineages. The influence of clan and lineage organizations is not unique to south China. For similar cases in north China, see Hu, Zongze, “Power to the people? Villagers' self-rule in a north China village from the locals' point of view,” Journal of Contemporary China, Vol. 17, No. 57 (2008), pp. 611–31CrossRefGoogle Scholar.

21 For detailed arguments on this, see Lu, Jie and Shi, Tianjian, “Political experience: a missing variable in the study of political transformation,” Comparative Politics, Vol. 42, No. 1 (2009), pp. 103–20CrossRefGoogle Scholar.

22 For unknown reasons, the 2005 survey did not collect information on villages’ clan structure and migrant workers. Thus, a parallel model cannot be estimated with the 2005 data.

23 Villages were categorized into four groups according the sample distribution of per capita income. Villages falling between the lowest value and 25% of the sample distribution were categorized as low income; those between 25% and 50% were categorized as medium low income; those between 50% and 75% were categorized as medium high income; and those between 75% and the largest value were categorized as high income. The 25%, 50% and 75% percentiles of this variable in the 2002 sample were 800, 1,264, and 2,150 yuan respectively. To capture the possible curvilinear relationship, its quadratic item was also included.

24 These two indicators are not perfect but they are the best available information in the 2002 survey. This percentage of CCP members was transformed into an ordinal variable. Villages were categorized into four groups according to the sample distribution of this percentage. The 25%, 50% and 75% percentiles of this variable in the 2002 sample were 1.79%, 2.29% and 3.01% respectively. The number of CCP members has also been tried for analysis, and the results do not change much.

25 It was transformed in to two binaries for analysis. The base-category for comparison is villages without clans. This measure is not perfect, but offers critical advantages over other measures for cross-regional analysis. Surname index is a misleading indicator of the clan/lineage structure in a village, since people sharing a surname may come from different clans/lineages. The number of ancestral halls is a good indicator of clan/lineage structures in south China, but much less meaningful in north China. Other voluntary associations like the Senior Association may also play some role here. Unfortunately, the 2002 survey did not have related information.

26 More specifically, the existence of VCEs before 1988 was used to control for the bottom-up momentum in VCEs.

27 For detained arguments on this, see Shi, “Voting and non-voting in China.”

28 For detained arguments on this, see Xu, Yong and Xu, Zengyang, Liudong zhong de xiangcun zhili (Governance and Floating Population) (Beijing: Chinese Social Sciences Press, 2003)Google Scholar.

29 This variable was transformed into an ordinal variable according to its sample distribution. The 25%, 50% and 75% percentiles of this variable in the 2002 sample were 35.99%, 48.95% and 61.91% respectively.

30 This variable was transformed into an ordinal variable according to its sample distribution. The 25%, 50% and 75% percentiles of this variable in the 2002 sample were 4.21%, 10.37% and 17.85% respectively.

31 Since all four procedures examined in Table 2 have been evaluated against the 1998 OLVC and coded as dummies, the summary index ranges from 0 to 4.

32 In this dichotomous approach, elections with any procedural violations are categorized as rigged.

33 Given the nature of summary index, it should be more appropriately analysed as an ordinal variable. Nevertheless, similar summary indexes have also been analysed as continuous variables in contemporary literature.

34 On multiple imputation, see King, Gary, Honaker, James, Joseph, Anne and Scheve, Kenneth, “Analyzing incomplete political science data: an alternative algorithm for multiple imputation,” American Political Science Review, Vol. 95, No. 1 (2001), pp. 49–69Google Scholar.