Vote Buying as Rent Seeking: Land Sales in China’s Village Elections

What explains why vote buying occurs in some elections but not others? The phenomenon of vote buying is under-studied in authoritarian, single-party-dominant regimes, especially in non-partisan elections in which competition is candidate-centered rather than party centered. Village elections in China provide a valuable window on the dynamics of vote buying in these conditions. Employing both an in-depth case study and an original, panel survey to provide new, systematic measures of rents and vote buying, we develop and test the following hypothesis: the availability of non-competitive rents accessible by winning candidates explains the variation in the incidence of vote buying in local elections. Our causal identification strategy exploits the timing of land takings and the exogenous nature of formal land takings authorized in state land-use plans at higher administrative levels to test the vote-buying-as-rent-seeking hypothesis. We find that the lure of rents, mainly from government takings of village land, is a key driver of vote buying by non-partisan candidates for the office of village leader. The evidence suggests that vote buying provides information to the authoritarian state about which local elites it should recruit into the rent-sharing coalition.

1 3 Keefer and Khemani 2005;Muñoz 2014) and in which elections are implemented by authoritarian regimes (e.g., Blaydes 2011;Frye et al. 2019). Village elections in China provide a valuable window on the dynamics of vote buying in these contexts. 1 Employing both an in-depth case study and an original, panel survey to provide new, systematic measures of vote buying and rents, we develop and test the following hypothesis: the availability of non-competitive rents accessible by elected officials explains the variation in the incidence of vote buying in local elections. We find that the lure of rents, mainly from government takings of village land, is a key driver of vote buying by non-partisan candidates for the office of village leader. Our causal identification strategy exploits the timing of land takings and the exogenous nature of formal land takings authorized in state land-use plans at higher administrative levels to test the vote-buying-as-rent-seeking hypothesis.
Our findings build on and extend the existing literature on comparative vote buying. Scholars note that more personalized competition is less likely to be programmatic and more likely to rely on the exchange of individual benefits (Kitschelt and Wilkinson 2007). In non-partisan competition, individual candidates' promises to voters about future benefits lack the institutional commitment mechanisms that political parties provide, making pre-election vote buying an important electoral strategy (Hanusch and Keefer 2013). We examine rent-generating state interventions at the local level and make a theoretical contribution by showing that candidates invest personal resources to buy votes as an individual investment strategy repaid by the capture of rents once in office. Our study also makes an empirical contribution by providing novel and systematic evidence on how land takings-a common phenomenon in the developing world-generate rents and drive vote buying in grassroots elections, where parties do not structure or finance competition and where government intervention in the local economy varies across locales. In testing hypotheses involving rent seeking and vote buying, one challenge is measurement (Lehoucq 2003;Kramon 2016). Existing work examining the link between rent seeking and vote buying relies on anecdotal accounts or formal models. 2 Combining an in-depth case study with an original multi-wave survey that covers 1,240 household in 62 villages allows us to identify a scalable measure of rents and develop a unique measure of vote buying based on household reports.
While most studies of vote buying focus on multi-party elections, the prominence of vote buying in non-party elections has important implications for our understanding of authoritarianism. Existing studies demonstrate the informational value of vote buying to voters in democratic contexts (Kramon 2016;Muñoz 2014). This study highlights how successful vote-buying candidates reveal information to the state about their social and financial capacity to command support from the local community. Authoritarian regimes can use this information about local notables to decide whom to allow into the rent-sharing coalition and how to coopt these local elites for regime purposes (Blaydes 2011). Fieldwork evidence shows that the Chinese Communist Party (CCP) has tapped successful, non-partisan vote-buying candidates after their election to take on party roles within the village.
The paper proceeds as follows. The next section grounds the analysis in the comparative literature on vote buying. The following sections detail our argument that vote buying is best explained as a means of capturing rents and provide background on the institutional context in which rents are generated. We then present an in-depth case study, which allows us to identify a scalable measure of rent-seeking opportunities and to generate specific, testable hypotheses about vote buying. The remaining sections present the data and findings, discuss implications of the study, and conclude with an agenda for future research.

Vote Buying in the Comparative Literature
Vote buying is a particular form of clientelism, in which voters receive a non-programmatic benefit-like cash or gifts-in exchange for a future vote for a particular candidate (Hicken 2011,292). Vote buying is prevalent in elections across globe. Roughly half of respondents in the sixth wave (2010)(2011)(2012)(2013)(2014) of the World Values Survey report that "voters were bribed very or fairly often in their country's elections"; in light of its prevalence, the question of "why politicians undertake vote buying in the first place" has begun to receive more scholarly attention (Keefer and Vlaicu 2017,775). While vote buying is typically analyzed as a political strategy to mobilize voters in multi-party competition (e.g., Diaz-Cayeros et al. 2016;Greene 2010;Lust-Okar 2006;Magaloni 2008;Stokes et al. 2013), it is also a common phenomenon in contexts where electoral competitions do not develop along partisan lines, such as local elections in one-party regimes (e.g., Zhao 2018). Theoretical analyses suggest that vote buying is more likely to take place when competition is candidatecentered rather than party-centered (Hicken 2007) and when the role of parties is weak or absent (Keefer and Vlaicu 2017;Muñoz 2014). Voters are less likely to trust candidates' policy promises that yield benefits only after the election. When party organizations that could discipline candidates are not involved in the elections, individual candidates' policy promises are less credible. Candidates instead choose to deliver pre-election benefits (i.e., vote buying) to mitigate the commitment problem (Hanusch and Keefer 2013). In an empirical analysis, Heath and Tillin (2018) find that when governance institutions are weak and inefficient, voters prefer benefits from direct, clientelistic exchanges (e.g., vote buying) over promises of uncertain programmatic rewards.
Elections and vote buying, where it occurs, can distort citizens' wishes but provide useful information to autocrats. Focusing on China, Martinez-Bravo et al. (forthcoming) show that local, non-partisan elections in autocratic regimes delegate selection and monitoring of village officials to citizens. They argue that rulers introduce elections when they lack information about local officials. However, they do not address the existence of vote buying or how it distorts the information that 1 3 regimes receive from elections. In describing parliamentary elections in Mubarak's Egypt, Blaydes instead suggests that vote buying in elections "works as a kind of market mechanism for the selection of individuals who will be allowed to extract state rents (2011,49)." Local elites who enter and win in the electoral marketplace reveal information to the regime about whom to coopt. Thus, vote buying serves as a mechanism to select local elites to be drawn into the rent-sharing coalition. As important and prevalent as vote buying is, the empirical evidence on the relationship between rent seeking and vote buying in non-partisan, authoritarian elections remains scant.

Vote Buying as Rent Seeking
We hypothesize that the incidence of vote buying in non-partisan elections is patterned on the availability of rents. We build our argument on the comparative literature that highlights the prominence of rents in shaping vote buying (Diaz-Cayeros et al. 2016;Keefer and Vlaicu 2017;Kitschelt 2000). Chandra (2007), in her conceptual overview of vote buying, begins by establishing the existence of rents, which stem from the state's "monopolistic control" over valued resources. Officials can sell resources for artificially high prices, generating non-competitive rents (Bates 1987;Buchanan et al. 1980). Thus, where the state controls significant resources, political control of the state is a valuable prize.
Identifying rents presents both a theoretical and empirical challenge. In his study of clientelism, Kitschelt (2000,862) is critical of studies that conceptualize rentseeking opportunities in terms of macro-level trade policies or the overall size of the public sector, since they fail to specify micro foundations or allow for sub-national variation. Hellman associates the existence of rents and rent seeking with partial reform in post-Communist transitions. "Partial reforms were expected to generate rent-seeking opportunities arising from price differentials between the liberalized sectors of the economy and those still coordinated by nonmarket mechanisms … these arbitrage opportunities have generated rents to those in a position to take advantage of these market distortions (Hellman 1998,219)." In the next section on China's institutional context, we establish the existence of non-competitive rents associated with village leadership positions in China.
Having established the existence of rents, scholars analyze how political actors seek to access them. 3 While bribery of regulators is a paradigmatic example of rent seeking (Aidt 2016), this paper highlights buying votes as an important pathway to controlling lucrative rents. Posner describes how shifting monopoly profit to officials "draws real resources into the activity of becoming the official (1975,812)." Politicians' access to rents rewards election winners with personal benefits. According to Aidt, "Controlling these rents is a contestable prize (2016,150)," and individuals will seek positions in government in order to gain access to them. Elections in authoritarian regimes are widely understood as a mechanism for sharing rents or spoils with elites who have the potential to threaten regime survival (Gandhi 2008).
Recent literature suggestively links vote buying and rent seeking. Blaydes argues that Egypt's non-partisan "rent-seeking elite" buy votes with cash or goods in a quest to win seats in parliament in order to access rents-not policy influence or party representation (2011,8). As she notes, "parliamentary hopefuls spend millions to reap billions (10)." High spending on vote buying is a rational choice, since elections in Egypt are a "sorting mechanism" for "access to state largesse," and "those who spend the most on campaigning are most likely to prevail (54)." While this account reflects comments from Egyptian informants, the study offers no test of the intuition. Zhao (2018), in a case study of village elections in China, similarly relates the availability of rents to vote buying on the part of candidates for local office. Keefer and Vlaicu (2017) develop a formal voting model in which parties use various strategies to win votes. They suggest that the larger the available rents, the greater the incentive to buy votes in order to control those rents. These studies derive their conclusions from anecdotes (Blaydes 2011;Ahram 2012), case studies (Zhao 2018), and theoretical models (Keefer and Vlaicu 2017).
A direct challenge in systematically testing the vote-buying-as-rent-seeking hypothesis is the measurement of rent-seeking opportunities and vote-buying activities; both remain largely hidden in most electoral contexts. We use an in-depth case study to reveal the nature of rent seeking and vote buying. Then, employing original survey data, we provide a quantitative test of the relationship between rent seeking and vote buying in China's village elections. Our measures of rents address Kitschelt's (2000) call for specifying micro foundations and accounting for subnational variations in rents in the context of elections.

Institutional Context
This section introduces the relevant institutions in which rents are generated and elections are contested. In contemporary rural China, rent seeking occurs overwhelmingly in the context of formal land takings, in which the state exercises a statutory monopoly over the sale of rural land into the urban market, making available non-competitive rents to state authorities, including village leaders (Zhao 2018). 4 The county land management bureau requisitions land from village collectives at below-market value, calculated in terms of the agricultural use-value of the land, and sells it to commercial developers at higher, market prices, making the rents quite sizeable (Chen and Kung 2016). Indeed, land takings have become a significant source of revenue for county governments (Su and Tao 2017). The difference between the price received from developers and the compensation paid to farmers, minus any public investment in basic infrastructure, represents the rent the state captures. Village leaders, who manage compensation funds for the village, share in these land rents by under-compensating rural households and by controlling compensation paid for land collectively owned and operated by the village.
Land takings are part of a top-down planning process. Since the 1990s, the Ministry of Land and Resources has assigned "annual land requisition quotas to each locality in the country (Su et al. 2019,761)." As concerns about loss of arable land have mounted, the ministry has established even tighter guidelines on land takings (Brandt et al. 2017(Brandt et al. ,1041. Regulations protecting arable land place strict limits on the area of farmland that may be converted into construction land in any given year (World Bank and Development Research Center 2014,113). Land-use plans received by the county are highly detailed, and "alteration [is] very rare (Su et al. 2019,762)." Ma and Mu (2020, 4) and Su et al. (2019,762) conclude that village leaders have no influence on formal state decisions to requisition land. Only after a decision is made, officials from the county, township, and village hold joint meetings to announce the decision and to present information to villagers regarding the purpose and size of the taking as well as the basis on which households are to be compensated (Brandt et al. 2017). Compensation funds, to be paid by the developer (the state in the case of takings for rail and highway projects, e.g.), reach the village via the local government late in the course of the project, since funds are generated only after land is taken, cleared, and sold through auction or negotiation by the state to the developer.
A related source of rents for village officials comes from illegal, informal sales of farmland by the village. Informal sales occur outside the purview of local government agencies that regulate the taking of village land for lucrative non-agricultural uses. Such informal sales are one response to state limits on rural-to-urban land conversion intended to protect arable land. In a survey designed to enumerate all changes to rural land holdings in a two-province sample between 1996 and 2011, Brandt et al. (2017Brandt et al. ( ,1036 find that only 12 percent of land takings recorded at the village level were initiated informally, i.e., by the township or village. Village-run enterprises have historically been a source of rents, stemming from privileged access to capital from local state-run banks and credit cooperatives and soft tax obligations (Hu 2005;Oi and Rozelle 2000;Takeuchi 2013). 5 However, Whiting (2001,289-295) shows that, following the banking and fiscal reforms of the 1990s, township-and village-run enterprises (TVEs) struggled to compete and survive. "TVEs, instead of being a tremendous asset, turned into a liability (Su and Tao 2017,240)." Leaders tasked with governing villages in China are selected through popular election for three-year terms. Village elections take place under the auspices of the 1998 Organic Law of Villagers' Committees and select the village head and members of the village leadership committee. 6 The entire village constitutes the electoral district, and any adult whose household registration is in the village may stand for election and vote. Elections are directly managed by election steering committees, which are themselves selected by popular bodies, including village assemblies, village groups, or village representative assemblies. According to O'Brien and Han, election steering committees are quite "independent [from the Chinese Communist Party] and responsive [to villagers], even though committees in most locations continue to be presided over by village party secretaries (2009,." Moreover, under rules in force in most provinces of China, "[CCP] control over the nominating process has been gradually loosened," and every voter is entitled to nominate candidates (364). Electoral rules mandate that there must be at least two candidates for village leader. Elections are candidate-centered and self-funded; party affiliation does not appear on the ballot, and the party does not finance campaigns of candidates who are party members. Indeed, CCP members may compete against each other. Most candidates campaign by making personal appeals in speeches before village assemblies and during visits to voters' homes.
In an intensive study of one county in Yunnan, Landry et al. (2010) show that the ruling party does not structure competition in village elections. They "confidently conclude that in practice both party and nonparty members were allowed [to] run in the …village elections and that nonparty candidates emerged from strata of villagers that were unlikely to be incorporated in the formal, non-electoral political institutions that dominated village life before the introduction of elections (2010,(13)(14)." Kennedy further argues that vote buying is unlikely to occur in elections where competition is absent or manipulated by the local party-state. "[Vote buying] is a problem that exists only because of reduced township interference and a fairer election process (2009,394)," i.e., absence of direct interference by the CCP in the selection of candidates and electoral choices of voters.
Scholars highlight the ambiguity of the 1998 Organic Law with respect to the legality of vote buying (Kennedy,(620)(621). The law does not mention vote buying (贿选); it states that the occurrence of bribery (贿赂) may invalidate an election, without defining what bribery entails. A 2009 central party-state notice acknowledges this problem and calls for "further clarification of the bounds of vote buying ( 进一步明确贿选的界限)" without final resolution. 7

Case-Study Evidence on Village Elections
Working within this institutional context, we combine an in-depth case study with an original survey in order to identify scalable measures of key concepts-rents and vote buying-and to generate specific, testable hypotheses relating rent seeking and vote buying. 8 The experience of villages in one satellite county outside a metropolis in an eastern coastal province of China is illustrative. In 2013, one of the authors conducted more than seventy in-depth interviews, including elected village leaders, unsuccessful candidates, villagers (voters), and township officials as well as informal discussions with county officials, focusing on village governance in more than a dozen villages in the county. 9

Land Rents
Land development in Baijia Village (a pseudonym) was spurred in 2005 by the county's decision to site the new campus of a highly regarded secondary school on land near the village. The county government's economic development plan, designed to attract outside investment to the county, included mixed residential and commercial development of roughly 500 mu 10 of village land in the vicinity of the new school, which would be the anchor for a new, planned community. In the case of Baijia Village, urban-planning regulations necessitated dividing the entire 500-mu project into multiple, smaller sub-developments to be included in consecutive annual land-use plans in order to comply with the letter-if not the spirit-of provincial regulations limiting the pace of urbanization.
Three levels of local authorities-the county, township, and village 11 -shared in the lucrative rents generated by the series of land takings implemented over the following years. The county government controlled the lion's share of rent, while the township and village received smaller shares. Although the amount the developer actually paid to the county for the land was undisclosed, compensation to be paid to Baijia villagers was publicly announced and set at 18,000 yuan per mu. This amount reflects the average value of agricultural output on the land over the preceding three years (1,000 yuan), calculated over the number of years remaining on farm households' 30-year land-use contracts with the village (18 years, in this case). Interviews reveal that the village leadership received at least 21,000 yuan per mu, representing a direct rent to the village leadership of 3,000 yuan per mu. Thus, the village leader would have access to approximately 1.5 million yuan in rents at the village level over the course of the 500-mu project as a result of the lower compensation to villagers.
A related source of rent came from compensation paid directly to the village for collective land under village management, including land previously used for village offices, plazas, and roads as well as forest land managed directly by the village. Compensation for this land, roughly 50 mu at 21,000 yuan per mu, totaled approximately one million yuan and was paid to the village and managed by the village leader but not distributed among village households.
The village leader's role in negotiations between the government and real estate developers, on the one hand, and members of the village collective, on the other, presented further opportunities for generating rents in the process of the land takings. Villagers who lost housing in addition to arable land were slated to receive Studies in Comparative International Development (2022) 57:337-360 new housing provided by the developer based on a per-meter exchange negotiated between the developer and the village leader, acting on behalf of village households. For example, an exchange rate of 1:1.5 meant that 100 square meters of villagers' old housing could be exchanged in the future for 150 square meters of new housing in the new residential community. However, in return for side payments from the developer, the village leader reportedly lowered the negotiated rate. In 2009, he agreed with the developer to an exchange rate of 1:1.25, while other affected villages had settled on a rate of at least 1:1.3. Villagers believed that the leader was willing to accept a lower exchange rate because he had received more than two million yuan in side payments from the developer.
In sum, interviews suggest that the leader of Baijia Village could control nearly five million yuan from the development project from these three sources alone. Land takings represent the greatest opportunity for a village official to access rents.

Vote Buying as Rent Seeking
Prior to the advent of land takings, elections in Baijia Village had been a sleepy affair. However, the 2008 election for village leader saw unprecedented competition, with six candidates contesting the election, along with the introduction of vote buying to the village. Two leading candidates emerged. One was the incumbent village leader, who reportedly offered targeted voters 600 yuan (about 86 US dollars) each for their votes. The leader's wife's successful retail business in the township provided funds to buy votes. His main challenger was the local owner of a construction company who had never before stood for election. The owner operated his construction company out of the county seat, but because his household registration remained in the village, he was eligible to compete for the position of village leader. Through paid vote brokers, well-connected and respected members of the village, he offered 800 yuan (about 116 US dollars) to targeted voters in order to counter the appeal of the incumbent. This sum is equivalent to approximately ten percent of an individual's average annual income (8,000 yuan) at that time. 12 The remaining candidates, including a middle school teacher who reportedly borrowed money to buy votes, were quickly outbid and abandoned any attempt at vote buying. As one villager commented, "Nobody in our village would spend money on running for office in the past… But since we've had land takings, people have become crazy spending money on the election. They know that no matter how much money they invest, they can easily recoup it from the land takings if they win the election." 13 Another villager made a similar assessment: "What could make an owner of a construction company give up his comfortable life in the county and move back to our poor village? It can only be because he expected to earn more money from his position." 14

Vote-Buying Process
Vote-buying candidates took a number of steps to try to ensure electoral success. They targeted swing voters, recognizing that relatives and allies of competing candidates were unlikely to be swayed by payments of cash from other candidates. Swing voters targeted in the vote buying schemes reportedly cast their votes for the highest bidder. 15 Vote-buying candidates and their brokers used a variety of mechanisms to monitor vote choice. For example, brokers instructed recipients of cash for votes to include their own names as write-in candidates for the less important position of village committee member. These voters' names would be read in the public vote count but would not be elected with a single vote. In other cases, brokers instructed recipients of cash for votes to take pictures of their completed ballots with their cellphones and to send them to the brokers. 16 The construction company owner took other steps to try to ensure electoral success. He offered bribes of gift cards and expensive liquor to township officials overseeing the election to look the other way as vote buying occurred and to send cooperative officials to monitor the election itself. Interviews suggest that these actions inhibited township officials from reporting vote buying to the county civil affairs bureau, the government agency responsible for overseeing village elections. Rumors repeated by villagers in interviews suggested that the construction company owner also sought to curry favor with the deputy county leader overseeing elections by providing the finish work on his house for free. At the same time, these reports also indicate that township and county officials were aware of vote buying in the village election.

Aftermath of Vote Buying
In the aftermath of vote buying, competition in the 2008 village election dwindled to two effective candidates, and victory went to the highest bidder-the construction company owner. After his victory in the village-wide election, the new village head was recruited by the local party apparatus to serve concurrently as village party secretary. 17 Finally, once land was formally requisitioned for the mixed residential and commercial development, the new village leader controlled the rents from the taking of land. As chairman of the village committee, his signature was required to implement the land taking and to access compensation funds deposited in the village account. As detailed above, the village leader controlled up to five million yuan (three quarters of a million US dollars) in rents through the land taking. The new village head's actions were not without controversy. Rumors circulated among villagers about under-compensation for their lost land. A small group of villagers-initially 5-6 people, ultimately growing to 15-mobilized to protest the low compensation by petitioning township and county authorities. They sought both higher compensation for land lost to urban development and dismissal of the elected village leader. However, neither township nor county officials took any action against the village leader. In the conclusion, we return to the issue of accountability in the context of vote buying in the agenda for future research.

Hypothesis Generation
The case of Baijia Village provides the basis for the following testable hypothesis. Specifically, we hypothesize that availability of rents is associated with the occurrence of vote buying in non-partisan, village elections. Candidates buy votes in order to gain access to rents controlled by the state. The case study also highlights the political context: in the absence of parties to structure and finance political competition, vote buying by candidates is a calculated, individual investment to capture rents.

Survey
We test this hypothesis with an original dataset. The dataset includes two waves of a multi-wave panel survey; the relevant waves were conducted in 2008 and 2014, in which two of the authors directly participated. The 2008 wave is based on a sample of 2,000 households in 100 villages selected from five provinces: Jilin, Hebei, Shaanxi, Sichuan, and Jiangsu. 18 Within each of the five provinces is a stratified random sample; five sample counties were randomly selected-one from each income quintile. Within each of the 25 counties, two townships were selected-one each from the top and bottom halves of the income distribution. Within each of the 50 townships, two villages were randomly selected for a total of 100 villages. In each village, 20 randomly selected households were interviewed, along with village and small group leaders. 19 The 2014 wave of the panel survey is based on a 60-village sub-sample of the original panel. To select the sub-sample, the counties from the high-and low-income quintiles were retained, and a third county was randomly selected from among the three remaining sample counties in each province. Two additional sample villages were included, bringing the total number of villages to 62. 20 The present study employs the 62-village sub-sample, with data from 1,240 households. Village leaders and household representatives were surveyed in faceto-face interviews. The 2008 wave included detailed batteries on village elections and other aspects of the household and village experience. The 2014 wave included extensive batteries on changes in household and village assets, including land. The unit of analysis for this study is the village election. Appendix I includes individuallevel analysis as a robustness check (Table A4); results are consistent with those reported, below. 21

Dependent Variable: Vote Buying
The dependent variable is operationalized through questions asked at the household level about the overall situation in the village's most recent election, specifically: Did anyone approach you [literally, go to your house] to buy your vote? (有人到 你家拉选票吗?) In the village-level analysis, the dependent variable is the share of households in each village responding affirmatively to the question about vote buying. The use of a common euphemism for vote buying (literally, "pulling votes" (O'Brien and Han 2009,365)) is intended to reduce its sensitivity for respondents. While official statements often use the more formal term "huixuan (贿选)" to refer to vote buying, "lapiao (拉票)" is the more colloquial term used by villagers in faceto-face interviews to refer to the payment of cash to influence vote choice. Lu (2011) reports that lapiao is a common usage by villagers in referring to material exchange for votes. A special feature on elections in the official People's Daily newspaper uses the combined term "lapiao huixuan" in analytical context but the more informal "lapiao" in providing specific accounts of vote buying in village elections (Yang and Li 2016). To check for sensitivity, we include an indirect measure of vote buying ("Did you hear about vote buying in the village?) (Appendix I, Tables A2, A3).

Independent Variables: Availability of Rents
The main independent variable of interest is availability of rents, measured in terms of whether there was a land taking within one term of a village election. 22 We use village responses to the following question: Since the second round of land contracting [in the mid-1990s], have there been any land takings in this village? Enumerators recorded all instances of land takings.
The identification strategy is two-fold. We are able to leverage multiple sources of variation, including formal and informal land takings and those preceding and following elections. Formal land takings involve decisions made at higher levels of government and are exogenous to the village. We contrast formal takings with all takings, including informal, illegal ones initiated by village leaders. We also exploit the timing of the land taking process, comparing takings initiated before and 21 Note that land takings and elections vary systematically at the village level. 22 Both village and household representatives were surveyed, and the data on land takings are mutually consistent. See Appendix I, Table A1. after the election. In order to influence the election, vote-buying candidates, who, we hypothesize, are motivated to control land rents that pass through the hands of village leadership, must know about land takings before the election. Moreover, delivery of compensation funds to the village comes late in the process. Thus, we test the relationship between land takings (available rents) and vote buying in three time periods: when land takings occur within one term before or after the election, three years before, and three years after the election. We also include other possible sources of rent, namely the number of enterprises in the village.

Control Variables
Scholars of clientelism identify a range of other factors, including income level, population size, and electoral rules that are conducive to vote buying. Comparative studies find that vote buying is more common among poorer communities (Stokes et al. 2013;Brusco et al. 2004), as these voters "discount the future, rely on short causal chains, and prize instant advantages (Kitschelt 2000,857)." By contrast, Kennedy notes with respect to the Chinese case "that vote buying is most prevalent in wealthier villages close to urban centers, such as large metropolitan areas (Beijing, Shanghai, or Shenzhen), provincial capitals, district cities, and even smaller county seats (Kennedy 2010,625)." However, the apparent association between wealthier, peri-urban communities and vote buying is also likely to be a reflection of the locus of land takings rather than the level of income. We address this debate by including an income measure in tests of our hypothesis.
Consistent with variants of clientelist theory that emphasize face-to-face interactions, smaller communities are also more likely to experience vote buying (Cox 1987). Candidates incur lower costs in smaller villages, because the number of votes to be bought is smaller in these villages (Takeuchi 2013). In what follows, we control for village population.
There is disagreement in the literature regarding the impact of the secret ballot on vote buying. Since the informational requirements of clientelist vote buying are high, secret ballot reforms are frequently associated with declines in vote buying (Schaffer 2007;Kam 2017). However, "many voters share the belief that their electoral choices can't be kept secret," even though they do not necessarily know how their votes are monitored (Stokes et al. 2013,101). Blaydes (2011,106) describes the ways that brokers in Egyptian parliamentary elections observe voters' choices on ostensibly secret ballots. To address this issue, we control for the use of secret ballots in the analysis. Other procedural factors arise in the case of village elections in China; some villages permit proxy voting, in which another person-often a husband or wife-may cast a ballot on behalf of another voter. Brandt and Turner indicate that "a village where looser proxy voting is permitted can engage in considerably more rent-seeking behavior than in a village where proxy voting is more restricted (Brandt and Turner 2007, 477)." Therefore, we control for proxy voting as well as the use of a secret ballot.
The intensity of competition is also commonly associated with vote buying (Golden and Chang 2001;Cox and Thies 1998). If candidates can easily win 1 3 elections or are not challenged by other candidates, they have less incentive to pursue vote buying. In studies of village elections in China, scholars use the number of candidates as an indicator to show that, in general, village elections have become more competitive over time (He 2007;O'Brien and Han 2009). We control for competitiveness using this measure.
Scholars have also noted the impact of higher-level interference on village elections (e.g., Fang and Hong 2020). To account for this possibility, we control for whether candidates are approved by local government officials and the numbers of official monitors on the day of election.
We also examine the effect on reports of vote buying when the winning candidate for village head is the CCP village party secretary. To the extent that the party influences electoral competition through vote buying, we would expect the election of the party secretary to be positively associated with vote buying.

Descriptive Statistics
The descriptive data (Table 1) allow us to paint a picture of both available rents and vote buying in the context of village elections. In 69 percent of the villages, one or more villagers reported vote buying. In the average village, 11 percent of respondents experienced vote buying. The highest percentage of respondents in any single village experiencing vote buying was 50 percent. 23 When asked less directly, 20 percent of villagers reported having heard of vote buying taking place. According to the case-study research, vote buying is limited by the fact that candidates believe their relatives and friends will vote for them without vote buying, while the relatives and friends of opposing candidates will behave the same way. 24 Over the period 1996-2013, 85 percent of the sample villages have reported at least one instance of land taking. These land takings affect more than 22 percent of all households in our sample. The average village experienced two takings during that period, and the average area of land lost was 133 mu. With compensation for arable land ranging in the tens of thousands of yuan per mu in sample villages, these takings presented clear opportunities to access rents.
The descriptive data also show that, in the period covered by the study, elections are meaningful contests for the selection of village leader. 25 For example, villages held between one and four rounds of nominations for candidates, with eight candidates on average coming from open nominations and at least 23 These percentages are not extreme by international norms. Keefer and Vlaicu (2017) note that, "Approximately 52 percent of respondents to Wave Six of the World Values Survey (2010)(2011)(2012)(2013)(2014) reported that voters were bribed very or fairly often in their country's elections." They also note significant cross-national variation: "The fraction of respondents who said that vote buying occurred often in their country ranged from 12 percent in the OECD to 56 percent in South Asia; and from 4.3 percent of Dutch respondents to 75.8 percent of Brazilian (773).". 24 Author's interviews. See also Nichter (2008) and Stokes (2005). Mattingly (2016) argues that lineage leaders, particularly those who become village leaders, exercise effective influence over their kinship group. 25 See full descriptive data in Appendix I Table A1. two candidates contesting the final election. Typically, not all candidates were members of the Chinese Communist Party. In 58 percent of sampled villages, candidates made campaign speeches. All elections were supervised by at least one representative of the township government. Seventy-five percent of villages employed secret ballots. Eight-five percent of villages allowed proxy voting; for voters using proxies, their proxy was a relative in the vast majority (85 percent) of cases. On average, turnout was over 80 percent of eligible voters. In all sample villages, the winner was decided by simple majority. The elected village head was concurrently party secretary in less than a third of villages.

Analysis and Findings: Vote Buying as a Form of Rent Seeking
In this section, we report tests of the hypothesis that the presence of one or more land taking within one term of a village election will be associated with a higher incidence of vote buying.
We estimate the following specification using ordinary least square to test our hypothesis: where VB i is the proportion of voters reporting vote buying in a given village, land_taking i is the number of land takings within three years (one term) before and after the election in that village, X includes controls, and K denotes county dummies. Column 1 of Table 2 provides a baseline by examining the correlation between all land takings-formal and informal-within one term-both before and after an election-and vote buying. In the bivariate model, the number of rent-seeking opportunities in the three-year period both before and after the election is positively and significantly associated with the prevalence of vote buying. Each additional rent-seeking opportunity is associated with a 2.6 percent increase in the share of villagers reporting vote buying. 26 Columns 2-5 show that the number of firms in the village, hypothesized to be another possible source of rents to office holders, is negatively related to vote buying, substantively small, and not significant. This result is consistent with Su and Tao's (2017) assessment that village-run enterprises have become a liability. Moreover, in our dataset, the majority of firms registered in sample villages are privately owned and operated, meaning that village leaders cannot extract rents from the firms as easily. In case-study research, one of the authors finds that the presence of more private firms, especially ones that make significant contributions to the local tax base, is associated with closer monitoring of the village leader by the county and township governments. 27 This context makes it more difficult for village leaders to capture rents in regulating firms.
Column 2 also introduces controls for village characteristics. As hypothesized in both the comparative and China-specific literature, the larger the population of the electoral district, the less likely the occurrence of vote buying. However, this factor is not significant when county fixed effects are included, while land takings remain significant. The results also show no significant relationship between income levels in the community and vote buying. While the finding that poorer communities are more likely to experience vote buying is well established in the literature (Stokes et al. 2013), it does not find support here. One possible reason is that income measures in surveys in rural China may be unreliable due to social desirability bias. Kennedy's (2010) suggestion that vote buying is prevalent in wealthier communities is not supported by the analysis; rather, his hypothesis likely reflects the role of land takings, which are more likely in peri-urban areas. In this dataset, land takings are significantly negatively correlated with distance from the village to the county seat. 28 The analysis also controls for multiple aspects of electoral rules and design (columns 3-5). None of these variables is consistently significant, while land takings are positive and significant and robust to all specifications. Although measures of electoral competitiveness-number of candidates and number of rounds of nominations-are positively associated with vote buying as expected, they do not attain statistical significance. Similarly, winning candidate vote share, a measure of (non) competitiveness, is negatively and significantly related to vote buying, as expected, but loses significance after controlling for county fixed effects. Two other variables, higher-level government approval of candidates and the number of government officials monitoring village elections, reflect local government intervention or oversight of the election, but neither is significant. Based on one author's interviews with township officials, monitors arrive in villages only on polling day to supervise elections, while candidates or their brokers usually make their deals with villagers at earlier times. 29 As a result, it is difficult for officials to know about the existence of these deals. The officials also point out that, even when they do know about vote buying, they may still prefer to not report it so as to avoid unnecessary trouble. In some situations, if the government receives complaints from villagers about vote buying prior to the election, it might send a higher number of officials to monitor elections to assure that the elections will not be disrupted by conflicts between villagers and candidates. 30 Among electoral rules, the positive (but insignificant) relationship between secret ballots and vote buying is most striking. This finding stands in contrast to much of the existing literature, which documents the importance of secret balloting in undermining the ability of candidates and their brokers to monitor vote choice (Lehoucq 2007). However, qualitative evidence produced by one author's fieldwork reveals strategies that candidates use to monitor votes even in the context of secret balloting. Blaydes (2011) describes similar techniques in Egyptian elections. Thus, candidates, motivated by the desire to capture large rents while in office, buy votes even when monitoring is difficult and devise new strategies to monitor votes. As Kam concludes, the secret ballot is "a powerful institutional force in improving… elections but not a decisive one (2017,626)." In columns 3 and 4, proxy voting is negatively and significantly related to vote buying. It remains negative but loses significance when controls for county fixed effects are introduced in column 5. Proxies reduce the number of targets for vote buying, since relatives cast ballots on behalf of their family members.
Finally, column 4 introduces a variable to assess whether the election of the party secretary as village head is associated with vote buying. When the party secretary runs in the election for village head, he must compete with other candidates (including, potentially, other party members) and may do so by buying votes. The relationship is positive, although after controlling for county fixed effects, the coefficient is not significant. This result is consistent with vote buying being an individual rather than institutional practice. Using the less sensitive measure of vote buying (Appendix I, Tables A2, A3), this variable remains positive and regains significance. This suggests that voters may hesitate to report having directly experienced vote buying in the election for village head when the party secretary participates in vote buying but are more willing to report having heard about vote buying in this situation.

Formal Land Takings as an Exogenous Source of Variation
The hypothesis-testing strategy employed here relies on the grounded assumption that land takings are exogenous to village electoral politics. If village leaders themselves can initiate informal land takings outside of official land-use plans, then the causality may be reversed: successful vote-buying candidates may drive land takings.
To further explore the hypothesized causal relationship between land rents and vote buying, Table 3 limits the analysis to only formal land takings, as measured in the survey by the presence of a permit accompanying the land taking. These takings are unambiguously initiated outside the village; formal land takings are approved and included in land-use plans at the county level and above and are implemented with the oversight of the county land bureau. Columns 1 and 2 of Table 3 present both simple bivariate results and control variables for all elections that occur within one electoral term of a land taking. In the first bivariate model, the number of formal land takings within one electoral term is positively and significantly associated with the prevalence of vote buying. Each additional formal land taking is associated with a 2.9 percent increase in the share of villagers reporting vote buying. In the fully specified model, each additional formal land taking is associated with a 3.1 percent increase in the share of villagers reporting vote buying. This effect appears to be driven by formal land takings that were initiated outside the village and not other factors.

Land-Taking Information Prior to an Election
The information environment in which electoral candidates operate is important. Our hypothesis-testing strategy also relies upon the logic that candidates for village office have information about the land rents available to office holders before they engage in vote buying. Columns 3-6 of Table 3 disaggregate rent-seeking opportunities into the electoral term immediately preceding and following an election. Column 3 reports the positive bivariate relationship between rent-seeking opportunities during the term before the election and the proportion of villagers reporting vote buying. Column 4 includes variables controlling for other possible causes of vote buying identified above and including county fixed effects. In this specification, only available rents through land takings, among all other relevant factors, is significant. Each additional formal land taking is associated with a 2.9 percent increase in the share of villagers reporting vote buying. This result reinforces the initial findings that the presence of land rents motivates vote buying on the part of individual candidates. The number of formal land takings in the three years after the election is not significantly correlated with vote buying. 31 Reliable information about land takings to be initiated in the future is not likely to be known or available to election candidates and therefore is unlikely to influence their electoral ambitions. This result is consistent with the information environment and the process of formal land takings described in the case study and supports the hypothesis that the lure of land rents drives vote buying.

Conclusion
This study makes an empirical contribution to the vote buying literature by introducing a new measure of available rents that captures sub-national variation and reflects the micro-foundations of vote buying. In theoretical terms, rent-seeking opportunities created by state intervention in the local economy shape vote-buying behavior by candidates. We find that the availability of rents-as reflected in the occurrence of lucrative state land takings-explains variation in vote buying in village elections in China. Furthermore, the Chinese case reveals the dynamics of elections in which competition is non-partisan and candidate-centered, rather than structured and financed by any political party. The evidence suggests how hegemonic political parties use authoritarian elections. The party as an institution does not use vote buying to elect its preferred candidate. Rather, vote buying provides information to the authoritarian state about local elites' economic capacity to buy votes and their socio-political standing to win elections. These distorted electoral contests provide information to the authoritarian state about whom it should recruit into the rent-sharing coalition. As reflected in the qualitative data, local economic elites who successfully buy their way into the office of village head may be recruited to the post of village party secretary. Similarly, village party secretaries may position themselves as individuals to try to win elections for village head in order to control the rents flowing through the village.
Recent policy changes since 2018 calling for village party secretaries to serve concurrently as the-presumably elected-heads of village committees raise new questions about the future of village elections. 32 One possible scenario is suggested by the findings of the present study. The lure of lucrative land rents may continue to drive vote buying as an individual strategy among competing village notables inside and outside the party. Non-party elites may seek to buy victory in competitive elections for the position of village head in the hopes of being tapped to join the party, undertaking the concurrent role of village party secretary, and thereby entering the rent-sharing coalition. Alternatively, vote buying may become a prominent phenomenon in intra-party elections for village party secretary.
The emergence of vote buying also has important implications for governance; comparative studies highlight the effect of vote buying on political accountability. Leight et al. (2019) find that voters who receive payment for their votes are more likely to tolerate takings by politicians and to tolerate expropriation in larger amounts. Moreover, politicians who pay for votes are likely to expropriate more resources than they would in the absence of vote buying.While multiple studies of village elections in China highlight their positive impact on accountability of leaders and on the quality of governance (e.g., Birney 2007; Kennedy et al. 2004;Manion 2006), the ways in which vote buying erodes local accountability remain an unexplored topic.
Further research is needed to assess the accountability of elected leaders who engage in both vote buying and rent appropriation. Two types of data will be useful in this regard. First, in future work, data on subsequent electoral contests involving incumbent leaders who previously bought votes, presided over the distribution of land takings compensation, and captured sizeable rents will shed light on the willingness and ability of voters to hold leaders accountable. Second, data on the presence or absence of villager protests following land takings will allow us to 32 See note 1.

3
test the results of laboratory experiments suggesting that voters who are paid for their votes are more likely to tolerate rent appropriation by the leaders they elect (Leight et al. 2019). Given the scale of land grabbing and rent appropriation under electoral authoritarianism as well as in electoral democracies in Asia, Africa, and Latin America (White et al. 2012), it is important to explain not only the causes of vote buying but also the relationship among rents, vote buying, and political accountability.