Grassroots democracy in rural China, that is to say, village committee elections (VCEs), has attracted enormous attention from scholars and observers of China politics in the past decades, with “a mountain of evidence” accumulated.Footnote ¹ Unfortunately, as some criticism points out, China scholars can basically find the evidence to support any argument.Footnote ² In addition to the lasting debate on how to interpret contradictory empirical evidence, China scholars are also actively deliberating over the pros and cons of different approaches in assessing VCEs.

O'Brien and Han recently promoted a shift in the focus of contemporary research from “access to power” to “exercise of power,” and a move beyond examining electoral institutions.Footnote ³ Others acknowledge the significance of “exercise of power” in Chinese villages, but still argue for the salience of the procedural dimension of grassroots democracy, that is, “access to power.”Footnote ⁴ Though fully sympathizing with O'Brien and Han's research agenda, this article, in alignment with other China scholars, argues that it is important to assess the empirical findings on electoral institutions before moving on to the “exercise of power” in rural China, for the following reasons.

First, promoting liberal democracy in an authoritarian regime relies highly on the implementation of transparent and competitive elections. In other words, democratic “exercise of power” cannot be isolated from “access to power” following democratic procedures. Such procedures, once strictly followed, should be able to improve political accountability, facilitate participation and ultimately enhance the quality of governance.Footnote ⁵

Second, despite its initial bottom-up momentum, the evolution of grassroots democracy in rural China has primarily followed a top-down process involving multiple actors with conflicting interests.Footnote ⁶ A transparent and competitive VCE with village cadres held accountable to fellow villagers, when compared to a rigged VCE with township officials dominating the nomination process with their favourite candidates, has significantly different implications for the “exercise of power” in villages. Moreover, once VCEs become the only officially sanctioned arena for political competition, various social forces could be channelled into this arena.Footnote ⁷ Thus the process of implementing VCEs becomes a focal point in Chinese village politics, offering a great opportunity to examine the state–society relationship in rural China and providing a salient process for understanding the dynamics among numerous political forces in Chinese villages.

In response to Manion's appeal for more systematic and comparative work on the “varieties of elections” in rural China,Footnote ⁸ this article examines the implementation of VCEs with the help of two unique sets of data collected in 2002 and 2005.Footnote ⁹ The first covers 241 administrative villages in 22 provinces and the second covers 379 administrative villages in 27 provinces. This article tries to provide systematic information on the institutional quality of VCEs, examine how the practice has been evolving and explore the plausibility of different arguments on the existence of such dramatic variance.

Because of some limits in the data, all conclusions drawn are tentative and should be further evaluated against other high-quality representative data. Nevertheless, given all logistic and financial constraints, it is rare to be able to assess the quality of VCE electoral institutions using two comparable nationwide probability samples. Thus, all information in this article should be of significant value in enriching our understanding of the political ecology in rural China in general and democratic politics in these communities in particular.

Varieties of Electoral Institutions

In addition to promulgating that administrative leaders of any village should be directly elected by its residents and such elections should be held every three years, the Organic Law of Village Committees (OLVC, drafted in 1987 and amended in 1998) also provides detailed stipulations on how to implement VCEs:

To what extent have VCEs been implemented following such stipulations? Previous case studies and local surveys have generally agreed upon the following: first, the implementation is uneven across regions; and second, over the years, the quality of electoral institutions has steadily improved. To evaluate the first statement systematically at the national level, more information is required than that provided by case studies and local surveys. To assess the second statement effectively, longitudinal evidence collected through comparable surveys is needed. Contemporary literature on grassroots democracy in rural China has limited pertinent information for either purpose.Footnote ¹⁰

Fortunately, two village surveys, embedded in two nationwide surveys implemented by the Institute of Sociology at the Chinese Academy of Social Sciences in 2002 and 2005, provide valuable information to evaluate both statements.Footnote ¹¹ With richer data at hand, it is appropriate to examine whether “procedural features mandated by the revised Organic Law on Village Committees are the norm in village elections after 2000” or not.Footnote ¹² And it is also possible to scrutinize the evolution of grassroots democracy in rural China, particularly in institutional quality.Footnote ¹³

For most VCEs, four stages are involved: organization of a village election committee; candidate nominations; certain procedures to identify finalists if there are too many nominated candidates; and ballot casting.Footnote ¹⁴ In 2002 and 2005, village clerks (cunwenshu 村文书) of sampled villages were interviewed about the various measures adopted at different stages of their most recent VCEs.Footnote ¹⁵ Pertinent information is summarized in Table 1.

Table 1:

Electoral Institutions of VCEs in 2002 and 2005

Notes: Raw frequencies in parentheses. Because of the adoption of more than one measure at some stages, the summary of percentages is not equal to 100.

Sources: 2002 village survey (N = 241) and 2005 village survey (N = 379) in mainland China.

In 2002, when it came to organizing village election committees, 42.7 per cent of sampled villages followed recommendations from their villager assemblies and 14.1 per cent took recommendations from villager small groups, both of which are compatible with the revised OLVC. In 2005, the figures dropped a little: 39.6 per cent counted on villager assemblies and 8.7 per cent followed recommendations from villager small groups. At the same time, other measures not endorsed by the OLVC were also used. In 2002, more than 37.8 per cent used villager representative assemblies, 17.8 per cent followed the instructions of Party branches, and a small number of villages even took orders from township governments (3.3 per cent) or outgoing village committees (3.3 per cent). In 2005, more villages (42.7 per cent) counted on villager representative assemblies while far fewer (3.96 per cent) took orders from Party branches. However, there were still some villages that followed instructions from township governments (2.37 per cent) or outgoing village committees (1.32 per cent).

Similar variance emerges when the measures used for nominating candidates are examined. In 2002, primary election (haixuan 海选) (36.5 per cent), joint nomination by eligible villagers (33.6 per cent) and self-nomination (17.8 per cent) were adopted in a fair number of villages. Nevertheless, Party branches (15.8 per cent), higher-level officials (4.6 per cent), villager small groups/villager representative assemblies (36.1 per cent) and village election committees (8.2 per cent) also played significant roles, activities that are expressly barred in the revised OLVC. In 2005, a much larger percentage of sampled villages (55.7 per cent) adopted primary elections, with significant drops in the adoption of the other two legitimate measures, with joint nomination at 16.9 per cent and self-nomination at 3.17 per cent. The influence of other local organizations and groups was also dramatically reduced: Party branches (3.9 per cent), higher-level officials (0.5 per cent), villager small groups/villager representative assemblies (16.6 per cent) and village election committees (2.9 per cent).

When there are too many candidates, the candidate pool has to be further narrowed. In 2002, 35.7 per cent of sampled villages used pre-election (yuxuan 预选). Other villages chose different strategies: villager representative assemblies (49 per cent), internal negotiations (xieshang yunniang 协商酝酿) (9.5 per cent), village election committees (9.1 per cent), Party branches (5 per cent) or higher-level officials (5 per cent) influenced the selection of finalists. The situation did not change much in 2005: 36.9 per cent used pre-election and 46.4 per cent authorized village representative assemblies to make decisions. The rest finalized their candidates through internal negotiations (5.0 per cent), decisions from village election committees (5.0 per cent), Party branches (0.5 per cent), higher-level officials (1.9 per cent) or other unspecified means (3.9 per cent).

With regard to adopting multiple-candidate elections (cha'e xuanju 差额选举), Table 1 shows much less variance. In 2002, 85.1 per cent used multiple-candidate elections to select the chairmen of their village committees and 86.3 per cent adopted the same procedure for selecting other members. In 2005, these two figures jumped to 91.6 per cent and 95.3 per cent respectively.

Since more than one measure could be used at some stages of VCEs, in order to evaluate the quality of electoral institutions rigorously I have strictly assessed each procedure against the OLVC. Any measure not officially sanctioned by the revised OLVC was coded as a violation. Table 2 shows the percentage of sampled villages that strictly followed the revised OLVC in implementing their most recent VCEs in 2002 and 2005 respectively.

Table 2:

VCEs Following the 1998 OLVC in 2002 and 2005

Note: Raw frequencies in parentheses.

Sources: 2002 village survey (N = 241) and 2005 village survey (N = 379) in mainland China.

Table 2 shows that in 2002, 42.3 per cent of sampled villages organized village election committees strictly following the revised OLVC. Slightly more than 51 per cent nominated their candidates through various legitimate means. With regard to finalist identification, 29.5 per cent did not violate the revised OLVC. Moreover, 79.7 per cent had multiple candidates for both the chairman and member elections. In 2005, varying progress was observed: 48.5 per cent strictly followed the revised OLVC in organizing election committees; 75.7 per cent had their candidates nominated through legitimate means; 36.9 per cent did not have any violations in identifying finalists; and 91 per cent adopted multiple-candidate elections. This comparison between 2002 and 2005 provides some new and stimulating information for evaluating the two general agreements in contemporary literature: the uneven implementation of VCEs and the improvement in the electoral quality of VCEs over the years.

The implementation of VCEs is obviously uneven across Chinese villages. However, the empirical evidence given above also suggests that the variation is much more than the regional variance suggested in contemporary literature; there is also significant difference among procedures in the same election. Comparatively speaking, adopting multiple-candidate elections has become something close to the norm in VCEs. However, other procedures mandated by the OLVC are not yet commonly used. More specifically, candidate nomination is very likely to be manipulated,Footnote ¹⁶ and the most serious manipulations in both 2002 and 2005 show up in finalist identification.

This comparison over time for each stage also reveals the progress in VCE institutional quality over the years. Nevertheless, the situation is still worrying when the overall quality of VCEs is evaluated. In this case, to be an overall high-quality VCE, all stages of the election must have been implemented strictly following the revised OLVC. Applying this criterion, as shown in the bottom of Table 2, only 11.2 per cent of sampled villages had high-quality VCEs in 2002. Although this number climbed to 16.6 per cent in 2005, a large majority of Chinese villages still violated the revised OLVC at some stage or other. Basically, seven years after the revised OLVC, only a little more than 16 per cent of Chinese villages had high-quality elections, ensured by non-compromised electoral procedures. This discouraging reality prompts a natural question: why is there such variance in the electoral institutions of VCEs?Footnote ¹⁷

Different Arguments on the Variance of VCE Electoral Institutions

Contemporary literature on China's rural democracy suggests four primary explanations for the uneven implementation of VCEs: socioeconomic features,Footnote ¹⁸ intervention from higher-level governments (particularly township governments),Footnote ¹⁹ the influence of social groups (especially clan/lineage organizations) in rural communities,Footnote ²⁰ and political learning.Footnote ²¹ Fortunately, the 2002 survey includes some data on the features of sampled villages. This enables a preliminary exploration of the validity of different arguments in an integrated model.Footnote ²²

The average per capita income was used as a proxy for the level of economic development.Footnote ²³ Distance to the township seat and the percentage of villagers who were CCP members were used as a rough guide to the possible influence of township officials.Footnote ²⁴ Regarding the influence of clan/lineage organizations, the clan structure of each sampled village was divided into three: a community without clans, with a dominant clan or with competing clans.Footnote ²⁵ Finally, whether rounds of VCEs had been held before the most recent one, and whether they were held before the trial of the OLVC was announced, were used as two proxies for villages’ experience with VCEs.Footnote ²⁶

In addition to the aforementioned four primary explanations, China scholars have pointed out another two structural features that might affect the quality of VCEs: the average level of political sophisticationFootnote ²⁷ and outward migration.Footnote ²⁸ The former was captured by the percentage of villagers who had completed at least primary education,Footnote ²⁹ and the latter was measured by the percentage of villagers working as migrant workers in other cities.Footnote ³⁰

Before moving on to regression analysis, two methodological concerns need to be addressed: the operationalization of the dependent variable and missing values in the data. Two possible approaches for operationalizing the dependent variable are a summary indexFootnote ³¹ and a binary variable.Footnote ³² The first assigns equal weight to different procedures, and the second assigns “veto power” to each procedure in determining the democratic nature of VCEs. To minimize the possible influence of the operationalization and check the robustness of the explanatory power of different variables, I have adopted both approaches and specified parallel models for comparison. A logistic model was chosen for the dichotomous approach. For the summary index approach, both ordered logistic and OLS models were used.Footnote ³³ Moreover, to reduce the possible influence of missing values, all missing values have been filled in with predicted values based on multiple imputations.Footnote ³⁴ The average results over five completed data sets are summarized in Table 3.

Table 3:

Results of Logistic, Ordered Logistic and OLS Models

Notes: Entries are averaged regression coefficients based on five imputed data sets. Averaged robust standard errors, corrected for clustering, in parentheses. * p < 0.1 ** p < 0.05 *** p < 0.01.

Source: 2002 village survey in mainland China (N = 241).

Among all four primary explanations, only social groups have shown persistent and significant influence across different model specifications and operationalizations. Villages with competing clans are significantly more likely to have high-quality electoral institutions. Basically, VCEs have effectively attracted and channelled the efforts of social groups within rural communities. Moreover, the involvement of social groups in VCEs has pushed for more transparent and competitive elections. This should not be a surprise given the lack of officially endorsed platforms for organized campaigns in VCEs, as well as the power of clan and lineage organizations in mobilizing collective efforts, enforcing social sanctions and spreading information.

The socioeconomic explanation is not significantly supported by the regression analysis. Regardless of model specifications, neither per capita income nor its squared term shows any significant influence. Both proxies for the possible intervention from township officials have also failed to achieve even the lowest significance level in all model specifications.

The argument for political learning has received some non-robust conformation from the regression analysis. Villages where the latest elections were their very first VCEs were significantly more likely to have high-quality electoral institutions. It seems that “rookies” may be faster learners whereas veterans find it more difficult to make changes to their previous flawed procedures. Nevertheless, this pattern does not show up in either the ordered logistic model or the OLS model. In addition, a higher level of outward migration may hurt the quality of grassroots democracy in rural China. The larger the percentage of villagers working as migrant workers, the less likely the most recent VCE was implemented strictly following the revised OLVC. However, this variable is not significant in either the ordered logistic model or the OLS model.