Resource disparity and multi-level elections in competitive authoritarian regimes: Regression discontinuity evidence from Hong Kong

Electoral Studies 33 (2014) 200–219

Contents lists available at ScienceDirect

Electoral Studies journal homepage: www.elsevier.com/locate/electstud

Resource disparity and multi-level elections in competitive authoritarian regimes: Regression discontinuity evidence from Hong Kongq Stan Hok-Wui Wong* Department of Government and Public Administration, The Chinese University of Hong Kong, 3/F, T. C. Cheng Bldg., United College, Shatin, New Territories, Hong Kong

a r t i c l e i n f o

a b s t r a c t

Article history: Received 16 January 2013 Received in revised form 30 April 2013 Accepted 1 August 2013

Much research has been done to study how competitive elections affect autocracies and their opposition. Electoral institutions, however, may have different social and political effects. In this paper, I examine the effect of an understudied electoral institution: lowerlevel elections. I argue that elections at grassroots levels tend to favor the ruling party by allowing it to more fully utilize its resource advantage to buy political support, which would in turn undermine the opposition’s ability to develop a local support network that is important to its struggle for democratization as well as for elected offices. Evaluating the effect of lower-level elections is empirically challenging because the effect is likely to be confounded with voter preference. I tackle this identification problem by taking advantage of a quasi-experiment afforded by the electoral formula of Hong Kong, which allows me to use a regression discontinuity design to test my causal argument. I find strong statistical evidence supporting my argument; the ruling elite’s aggressive expansion in the District Councils, the lowest elected tier, aims to drive out the opposition elites, who, by occupying a District Council seat, are able to increase their vote share of that constituency by 4–5 percentage points in a subsequent legislative election. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Competitive authoritarianism Multi-level elections Authoritarian elections Political machine Regression discontinuity

1. Introduction Recent studies of authoritarian regimes find that holding regular elections helps the incumbent achieve various political objectives. Geddes (2005) argues that authoritarian elections provide a peaceful and relatively low-cost q I would like to thank two anonymous referees, Michael DeGolyer, Barbara Geddes, Siu-Hing Lo, Ellen Lust-Okar, Ngok Ma, Dixon Sing, Alvin So, Daniel Treisman, and the participants at the Conference on Trustbuilding and Governance in Hong Kong and Macao for their helpful comments on earlier versions of this article, and Tim Tang, Amina Yam, Yalin Hung, and Kiu-Yan Wong for research assistance. Special thanks go to Ngok Ma, who shared part of the election data. I am responsible for the remaining errors. Funding of this research was partly provided by the Academic IT Steering Committee and the Faculty of Social Science of the Chinese University of Hong Kong. * Tel.: þ852 3943 7485. E-mail address: [email protected]. 0261-3794/$ – see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.electstud.2013.08.003

means to deter potential challengers. Drawing on the experience of the Arab world, Lust-Okar (2005) points out that by excluding some opposition leaders from participating in elections, authoritarian leaders effectively undermine the unity of the opposition elite. In her study of the hegemonic party in Mexico, Magaloni (2006) finds that elections provide an opportunity for the ruling party to showcase its invincibility through orchestrating expensive election campaigns, which in turn discourage defection of the ruling elite. Gandhi and Przeworski (2007) suggest that dictators can co-opt the opposition elite by allowing them to contest and win some seats in the dictator-controlled legislature. Blaydes (2010) points out that elections in Mubarak’s Egypt ease the distributional conflicts among the ruling elite. These studies examine the generic effects of authoritarian elections. But as scholars of electoral studies have

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

long observed, different electoral arrangements and political institutions produce different economic, social, and political outcomes (for example, Persson and Tabellini (2005); Lijphart (2012)). Electoral institutions are not uniform across autocracies (Geddes, 1999; Hadenius and Teorell, 2006; Cheibub et al., 2010). Therefore, we have reason to suspect that their political effects also vary. This paper studies the effects of one underappreciated, albeit important, electoral institution in authoritarian regimes: lower-level elections. There are dictatorships such as China where competitive elections exist only in villages, whereas the authoritarian state of Malaysia runs elections at both national and local levels. Despite this variation, nearly all autocracies that hold regular elections at the national level run some form of local elections. Why would autocracies have incentive to hold local elections? This paper aims to advance a simple argument: competitive elections at a lower level of government favor autocratic regimes for three reasons. First, ideology plays a relatively insignificant role in shaping the electoral contests at the grassroots level, which implies that the opposition that advocates programmatic changes such as political liberalization would have difficulties using its cause to mobilize popular support. When programmatic policies fail to attract voters, electoral success often depends on the provision of constituency services and the distribution of spoils. Given its resource advantage, the ruling elite can easily outcompete the opposition in this respect. Second, low-level elections help solve a commitment problem of machine politics. In her study of political machines (or clientelist parties), Stokes (2005) points out that the functioning of an effective political machine requires voters not to renege on the implicit deal where the machine offers services and the recipient votes for the machine. In authoritarian regimes, low-level elections require the ruling party to gather voter preferences sometimes down to the neighborhood level, thereby improving the party’s ability to monitor voters in higher-level elections. Finally, capturing seats in low-tier governments is also conducive to warding off political challenges from the opposition because it can prevent opposition parties or their coalition from developing a local network from which they draw political and financial supports.1 To illustrate my argument, I provide a case analysis of Hong Kong. Since the city’s sovereignty transfer in 1997, pro-Beijing parties have aggressively expanded their political presence in the District Councils, the lowest elected tier in Hong Kong’s political structure. Thanks to their superior resource advantage over the pro-democracy opposition parties, they have achieved great electoral successes in the past decade. Capturing the seats in the District Councils is not an end in itself, however. More importantly, it serves as a stepping stone for the ruling elite to undermine, from the ground up, the electoral support for the pro-

1 This is not to say that the challenge coming from the opposition is the only source of threat against the political survival of authoritarian regimes. Various authors have pointed out that intra-elite power struggles are another common cause leading to authoritarian breakdown (Zolberg, 1966; Geddes, 1999; Svolik, 2012). In the current paper, I will not deal with the struggles within the ruling elite.

201

democracy opposition in legislative elections, which are the major battlefield. I conduct statistical analyses based on a regression discontinuity design to examine the importance of this bottom-up strategy. The empirical results indicate that by capturing a seat in a District Council constituency, the pro-democracy opposition elite is able to increase its vote share of that constituency by 4–5 percentage points in a subsequent legislative election. For this reason, driving out the opposition from the District Councils is a sensible move by pro-Beijing parties to curtail its rival’s political influences. A methodological advantage of using Hong Kong as the case is that it has a unique political structure that allows for a rigorous test of the electoral effect of capturing a District Council seat on a higher-level election. Simply showing that the vote share received by pro-democracy parties in District Council elections is positively correlated with the vote share they receive in legislative elections is insufficient to identify the effect of interest. The positive correlation may merely reflect idiosyncratic district characteristics. For example, pro-democracy districts are likely to record a low vote share for the incumbent party in District Council elections and a high vote share for the opposition party in legislative elections. In other words, voter preference is confounded with the effect of interest. I take advantage of the electoral formula of the District Councils, the plurality rule, which involves a distinct cutoff point to decide winners and losers, to apply a regression discontinuity design to isolate the District Council effect from other confounding factors such as voter preference. Hong Kong differs from a typical authoritarian regime in a number of important respects. First, it is not a sovereign state, but a city of the People’s Republic of China (PRC). Historically, no substantial decolonization movement ever took place in Hong Kong. Nor did the city undergo any violent pro-democracy movement before and after the sovereignty transfer. Politically, while the rest of China is under the tight control of the Chinese Communist Party (CCP), Hong Kong retains a high degree of autonomy and civil liberties under the ”one country, two systems” principle. Although these unique features limit the external validity of theoretical insights drawn from the experiences of this city to other competitive authoritarian regimes, it is problematic to say that Hong Kong as a polity is so unique and unprecedented that observations drawn from its political institutions can never yield any comparative value in the studies of politics. In fact, upon closer examination, Hong Kong in many ways is qualified as a competitive authoritarian regime. First, despite being a city of China, Hong Kong has a political system distinct from that of the mainland. The ”one country, two systems” principle has effectively set Hong Kong’s political system apart from that of the PRC. That principle is an institutional arrangement designed by Beijing back in the 1980s to reassure Hong Kong people, who had been gripped by the fear of pending Communist rule, that Beijing would tie its own hands after the reunification, such that neither the Chinese state nor the CCP would meddle with the city’s rule of law and small open economy. The CCP has indeed refrained from playing a visible role in Hong Kong’s politics (it does not even have a

202

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

Table 1 Election at different levels in Hong Kong.

Constituency Seat number Playing field Electoral formula Terms of office

Chief executive

Legislative council (LegCo)

Election Committee with 1200 membersa 1 Highly uneven Plurality 5 years

Functional Constituencies 35 Highly unevenc Mixed 4 years

District councils Geographical Constituencies 35 Relatively even PR 4 years

District Council Constituencies About 400b Relatively even Plurality 4 years

a Political institutions have undergone gradual and slight changes since 1997. Information presented in this table is based on Hong Kong’s political structure as of 2013. b In addition to some 400 directly elected seats, the government itself also appoints dozens of District Councillors. c One example of the uneven playing field is unequal representation (Ma and Choy, 2003). More than 90 percent of the eligible voters are represented by only 5 seats in the functional constituencies.

visible presence in the city). In addition, Hong Kong people have been allowed to choose some of their own deputies in the Legislative Council and the District Councils by relatively free and fair elections (For a brief overview of Hong Kong’s political structure, see Table 1).2 Reciprocally, being an elected official in Hong Kong does not automatically earn one any political office in the PRC. Under such circumstances, the politicians and the people of Hong Kong have over time become accustomed to this political distance with the PRC. For instance, when they fight for democratization, they are fighting for the democratization of their city, not that of the PRC. Second, opposition parties in Hong Kong are allowed to compete regularly in the city’s legislative elections that are generally considered free and uncorrupted.3 However, electoral competition is not on a level playing field as in other competitive authoritarian regimes (Levitsky and Way, 2010). The election of a significant number of legislative seats is skewed heavily in favor of the pro-Beijing elite (the functional constituencies), an arrangement effectively preventing opposition parties from controlling the legislature. In addition, although the regime in Beijing is not ruling Hong Kong directly, it holds the ultimate veto power over the implementation of universal suffrage in the election of the Chief Executive and the abolishment of the functional constituencies. Consistent with the standard behavioral assumption about autocrats’ preference for staying in power, Beijing thus far shows neither eagerness for nor commitment to Hong Kong’s democratization. Having spent years bargaining with Beijing, the opposition force managed to obtain only small political concessions.4 Puny as they are, these concessions give many people hopes that further political liberalization is possible. At the same time, pro-Beijing parties in Hong Kong have aggressively used their superior resource advantage to undermine the opposition’s popular support. Their move helps improve Beijing’s bargaining position vis-à-vis the pro-

2 For a more detailed analysis of Hong Kong’s politics after 1997, see Lam et al. (2007). 3 The reason why Beijing chose to tolerate multiparty elections rather than simply removing elected bodies to make Hong Kong conform to the rest of China's political system is that heavy-handed political controls would scare off entrepreneurial talent or investment (Wong, 2012). 4 For instance, in the 2010 political reform negotiation, Beijing agreed to increase the number of electors of the Chief Executive from 800 to 1200.

democracy opposition; by weakening the opposition’s popularity, pro-Beijing parties can diminish their rival’s presence in Hong Kong’s legislature, which can then reduce the political pressure for further political liberalization. Because the conditions examined here – the challenge from the opposition, a ruling elite with a resource advantage, and competitive elections – are not something so unusual that cannot be found in autocracies elsewhere, how the pro-Beijing camp in Hong Kong takes advantage of lower-level elections to achieve its political objective should offer some comparative insights to understand similar dynamics occurring in other authoritarian regimes. The rest of the paper is organized as follows. In the next section, I present my simple argument. In Section 3, I examine the political background of the legislative and District Council elections in Hong Kong in the posthandover period. Section 4 discusses the methodology used to estimate the electoral effect of capturing a District Council office. Section 5 presents the empirical results. The final section concludes. 2. Multi-level elections in authoritarian regimes Schedler (2002) argues that a pro-democracy opposition party participating in an authoritarian election is attempting to achieve two political objectives. It endeavors to capture elected offices inside of the autocratic system in order to gain political power, while struggling for an ultimate removal of that very autocratic system. Pursuing these objectives may involve different optimal strategies, but these strategies are likely to offer an effective complementarity with each other. On the one hand, by occupying a seat in the legislature, the opposition party can make their political cause more widely known because its official position allows it to reach out to more voters. The formal political office also entitles the opposition to a stable and often high salary, which is conducive to party development and political advocacy. On the other hand, the opposition’s call for an end of the political monopoly held by the incumbent may be conducive to winning seats because the cause of democratization should appeal to voters who detest the exclusionary nature of the political order that often produces corruption, unfair distribution of economic benefits, and unresponsive officials. Despite the electoral appeal of its political cause, the opposition party faces resource constraints that would significantly limit its electoral strength. A common

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

characteristic of autocracies is that the autocrat monopolizes not only the political power, but also the control of state resources. This gives the ruling elite a superior comparative advantage over the opposition in the distribution of spoils (Lust-Okar, 2009). While some autocrats use spoils to buy political support in districts that are not the ruling party’s stronghold (Molinar and Weldon, 1994), others take spoils away from disobedient constituencies as a punishment (Diaz-Cayeros et al., 2003). Suffice it to say, this resource advantage offers the incumbent an effective tool to counter the challenge of the opposition. From voters’ perspective, whether to vote for the opposition or the incumbent depends on the expected utilities followed from these choices. For most voters, although they are not a member of the ruling coalition, they, by voting for the incumbent party, can nevertheless gain access to some private benefits such as a bag of rice, a full meal or simply free flu shots. Inconsequential as these benefits may appear, they are tangible and probable. In contrast, if they vote for the opposition party, given the party’s resource constraints, voters are unlikely to receive the same level of spoils from elected opposition politicians. For pro-democracy voters, they can certainly hope for the most ideal scenario, in which the opposition camp defeats the ruling party in the voting booth. Such a ”democratization by elections” (Lindberg, 2009) would lead to a wholesale change of the political order, resulting in political enfranchisement of many who are excluded from sharing the political power and economic benefits under the current dictatorship. However substantial these benefits are, the chance that the opposition can oust the incumbent in the voting booth is by no means great in most competitive authoritarian regimes most of the times. In expectation, therefore, voting for the opposition is not necessarily an attractive option to voters. Although the incumbent can do little to alter voters’ payoff derived from democratization, it can find ways to suppress the likelihood of democratization (or the perceived likelihood), to reduce the voters’ expected utility of voting for the opposition. A prime example of this, as pointed out by Magaloni (2006), is that many authoritarian regimes would continue to stage elections despite the fact that they would always win a supermajority. The supermajority victory is intended to signal the incumbent’s invincibility, which implies that the likelihood that the opposition can defeat the incumbent is vanishingly small. The same logic can be applied to explain why an election of low-tier offices would favor the ruling elite rather than the opposition force. When the opposition candidates at the national level fail to pose a serious challenge against the ruling party, those at the local level, with their limited jurisdiction and constituency, would have even greater difficulties convincing voters that electing them would make democratization more likely. At the same time, it is unlikely that the opposition candidates running for local offices would face fewer resource constraints than those who seek national offices, which implies that they are unable to offer voters more material benefits to compensate for their relatively insignificant influences in the pursuit of regime change.

203

While low-tier elections may dampen support for the opposition party, they may give voters additional incentives to support the incumbent who would otherwise not have been able to receive in national elections. Given the small constituency size, local politicians maintain a closer relationship with local residents than national politicians. Suppose local constituencies are heterogeneous such that they have different demographic compositions and physical environments. Because local politicians know local conditions better than national leaders, they can tailor the spoils delivered according to local needs (Oates, 1999). As a result, spoils can be used more efficiently to generate electoral support. In addition, from a voter’s perspective, the chance that she can be a pivotal voter is greater in lowtier elections than in national elections. For this reason, she is probably able to demand more benefits from a local candidate than from a national candidate. Perhaps more importantly, from the ruling party’s perspective, lower-level elections help tackle a commitment problem associated with spoil distribution. Despite the ruling party’s ability to dole out private benefits, such benefits yield intended electoral effects only when the recipients really vote for the party. Most authoritarian regimes which hold regular elections adopt secret ballots (Geddes, 2005). Recipients of the ruling party’s largess thus incur a relatively low cost if they decide to renege on the implicit deal with the ruling party. Such a commitment problem can be more effectively dealt with if lower-tier elections are in place. The reason is that winning lowerlevel elections requires a party to have an active and close interaction with local residents. This is possible most probably because there exists a political machine that develops grassroots political networks.5 Such a political machine is able to collect important voter information, such as political preference, often down to the neighborhood level. Sometimes the party may even send grassroots agents who are familiar with local residents to polling stations to observe the progress of voting and check who shows up to vote. Such a monitoring method definitely cannot yield perfect compliance from the voters, but it gives the party a reliable estimate of its electoral chances.6 The voter information as well as the close monitoring by grassroots agents are useful not only to lower-level elections, but also to higher-level ones. This is why most authoritarian regimes that stage national level elections also have lower-level elections.7 The foregoing analysis suggests that the ruling party of an autocracy enjoys a superior advantage over opposition parties in capturing local elected offices. This has important implications not only on the opposition’s short-term objective of occupying elected offices, but also on its long-term objective of fighting for democratization. By driving out opposition parties from the grassroots level, the

5 Taylor (1996) explains how such grassroots political networks function in the form of caudillo politics in Honduras. 6 For instance, Stokes (2005) provides interesting anecdotal accounts of machine politics in Argentina; a grassroots agent can identify defectors by examining whom is unwilling to look him in the eye on election day. 7 I thank an anonymous reviewer for highlighting this point.

204

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

ruling party can effectively prevent opposition parties from developing an extensive local support network, which would in turn constrain the opposition’s ability to advocate its political cause, attract followers, and drum up donations. These constraints would leave substantial negative impacts on the opposition’s struggle for elected offices as well as for democratization.8 3. Multi-level electoral politics in Hong Kong On July 1, 1997, the British handed over Hong Kong to the People’s Republic of China. The new Hong Kong government was confronted with a challenging political situation; it had inherited from the colonial government a strong opposition party, the Democratic Party (DP), which enjoyed an enormous appeal in the city because of its prodemocratization stance. Two years before the handover, the DP, thanks to the winner-take-all nature of the plurality rule, an electoral formula adopted by the colonial government in the last legislative election, captured 12 of the 20 directly elected seats. Together with additional 7 indirectly elected seats of the functional constituencies, it emerged as the largest political party in the colonial legislature. This was not a desirable outcome from the Beijing government’s point of view for a good reason: the DP pressed Beijing to democratize Hong Kong, which, from Beijing’s perspective, was tantamount to relinquishing control of the city.9 To curtail the political influences of the DP, Beijing unilaterally declared, immediately after the handover, that the term of the DP-dominated legislature was over, and replaced it with a provisional legislature, the members of which were supported by Beijing. Although the provisional legislature finished its term within a year, followed by an election of a new Legislative Council (LegCo), Beijing exerted its influences by setting the rules of the game. In particular, it has continued to allow for the existence of the functional constituencies. Because the playing field of the functional constituencies has been skewed heavily toward the business elite and other pro-Beijing elements, Beijing has been able to wield more influences on the election outcomes. Beijing also replaced the plurality rule with proportional representation. Thanks to this rule change, the seat share of pro-Beijing elements significantly increased. A well-known political effect of proportional representation is that it tends to produce a multi-party system. Indeed, the post-

8 Despite these adverse effects, the opposition’s electoral loss is not automatically translated into the incumbent’s electoral gain. Even if the opposition does not appeal to voters, this does not imply that voters have to support the incumbent, as they can always abstain or in some cases, vote for candidates associated neither with the opposition nor with the incumbent. From the incumbent’s perspective, such outcomes are not necessarily undesirable, because driving down the electoral support for the opposition is sufficient to weaken, rather than strengthen, the democratization movement as a whole. 9 Another possible reason that Beijing was hostile to the DP is that the leading members of the DP such as Szeto Wah were simultaneously controlling another political group, the Alliance in Support of Patriotic Democratic Movement of China (ASPDMC), whose central tenets included ”ending one-party dictatorship in China” and ”building a democratic China.”

1997 Hong Kong has observed a proliferation of political parties not only within the pro-democracy camp, but also within the pro-incumbent camp. The emergence of multiple pro-Beijing parties may have somewhat weakened the unity of the pro-incumbent bloc, but this does not necessarily run counter to Beijing’s political interest. The Basic Law, the mini-constitution of Hong Kong, stipulates that the Chief Executive cannot be a member of any political party. In other words, Beijing has never relied on any proBeijing party to run the Hong Kong government in the first place. To Beijing, the more important function of proBeijing parties, as I will discuss, is to counter the political influences of the pro-democracy elites. By dictating the rules of the election, Beijing might reduce the opposition politicians’ presence in the legislature, but it achieved little against their overriding popularity in town. As mentioned, the pro-democracy opposition (also known as the pan-democrats) enjoyed a wide appeal in Hong Kong, especially after 1989, when the entire city witnessed the brutal crackdown of a peaceful pro-democracy student-led movement in Beijing. In addition, the colonial government in Hong Kong had opened up the political system for local participation since the 1980s (Cheung, 2011). Many Hong Kong people gradually came to believe that building a democratic institution was perhaps the only effective way to check the Leviathan state of the PRC. For this reason, voters generally supported the prodemocratization stance of the opposition elite. In order to reduce the political clout of the prodemocracy camp, Beijing needs to rely on its political proxies to undermine the pan-democrats’ social support base.10 This is what pro-Beijing parties (also known as the pro-establishment camp) have been doing since 1997. Quite surprisingly, the pro-establishment camp has achieved arguably a great success within a short period of time. As can be seen from Fig. 1, the gap between the proestablishment and pro-democracy camp with respect to either vote share or seat share is closing steadily over time. More importantly, the data displayed in Fig. 1 come from the directly-elected geographical constituencies, which implies that the electoral success of the pro-establishment camp is by no means confined to the Beijing-influenced functional constituencies. In sum, this camp does enjoy a rising support among voters. A critical albeit understudied factor that contributes to the rapid ascendancy of the pro-establishment camp is its

10 As mentioned, under Beijing’s ”one country, two systems” principle, Hong Kong should be governed by the Hong Kong people. Officials of the People’s Republic of China or members of the Chinese Communist Party are not supposed to intervene in Hong Kong’s politics. For this reason, there are a number of Beijing-sponsored parties in Hong Kong. Although these parties disagree with each other over certain socio-economic issues, they all serve Beijing’s interests on major political issues such as obstructing the implementation of universal suffrage in the election of the Chief Executive. At times, these parties even have the same members. For instance, several leaders of the Democratic Alliance for the Betterment and Progress of Hong Kong (DAB) simultaneously hold executive positions in the Hong Kong Federation of Trade Unions (FTU), which is another Beijing-sponsored political party. As a result, it is instructive to treat these parties as a political camp representing Beijing’s interests visà-vis the pan-democrats.

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

205

Fig. 1. Over time Change of Legislative Seat and Vote Shares by Political Camp.

District Council strategy. The District Councils are the lowest elected tier of Hong Kong’s political structure. As an elected official, a District Councillor has very limited political power. This is in part because the geographical constituency that each District Councillor serves is tiny. Hong Kong is divided into some 400 District Council Constituencies. It is not uncommon that a housing complex contains a few District Council Constituencies, each of which spans no more than a dozen of apartment buildings. The private housing estate Mei Foo Sun Chuen is a case in point. In 2011 District Council Election, this housing complex is divided into three District Council Constituencies: Mei Foo South (F12), Mei Foo Central (F13), and Mei Foo North (F14) (see Fig. 2). In such minuscule constituencies, District Councillors have a weak incentive to pursue programmatic benefits, which exhibit little economies of scale. In addition, the job nature of the District Councils constrains what a District Councillor can offer to her constituency. According to the District Councils Ordinance: The functions of a District Council are (a) to advise the Government (i) on matters affecting the well-being of the people in the District; and (ii) on the provision and use of public facilities and services within the District; and (iii) on the adequacy and priorities of Government programs for the District; and (iv) on the use of public funds allocated to the District for local public works and community activities; and

(b) where funds are made available for the purpose, to undertake (i) environmental improvements within the District; (ii) the promotion of recreational and cultural activities within the District; and (iii) community activities within the District. The Ordinance indicates that a District Councillor has no decision-making power over government policies – even policies related to her tiny constituency (Holliday and Hui, 2007). What she can offer is often mundane community services such as fixing clogged pipes, providing free flu shots, and organizing a day trip to the outskirts. An oftcited mocking example is that a pro-establishment District Councillor once used street banners to advertise a major achievement during his term of office: he successfully got the authorities to lengthen the green signal of a traffic light by 3 s. This example further suggests that the benefits that a District Councillor can bring to her constituency is highly particularistic, rather than programmatic. On the surface, the office of a District Council seems politically insignificant. Upon closer examination, as I argue below, capturing a District Council seat yields less visible, but substantial electoral advantages to a political camp through three channels. 3.1. Channel 1: offer financial and political resources For many pan-democratic politicians, their District Councillor salary, however meagre, is an important source

206

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

Fig. 2. Size of District Council Constituencies: Examples of Mei Foo Sun Chuen.

of income for maintaining their parties’ operation. For instance, as of 2011, the DP has 47 District Council members and 8 Legislative Councillors, each of whom ought to hand in about 10 percent of their official salary to the party central. Their collective monthly contribution constitutes 15 percent of the party’s income in that year (Democratic Party, 2011). As the party’s income would be used in the campaigns for legislative elections, every seat lost in the District Councils has direct adverse effects on DP’s electoral chance in the LegCo. From the pro-establishment camp’s perspective, therefore, driving out the pan-democrats from the District Councils can deplete the pan-democratic parties of an important financial resource. In addition to a stable source of financial income, a District Council seat also provides politicians a close contact with their constituents, which is a valuable political asset. Many politicians from the pan-democratic camp started their political career from the District Councils as early as in the 1980s. Through the provision of constituency services as District Councillors, they became known in the constituency. Their political credentials also helped them develop a local support network, which consists of not only people who would vote for them, but also voluntary workers who assist their electoral campaigns. As a result, by taking away seats previously owned by the pandemocrats, the pro-establishment camp can effectively uproot its rival’s local support base.

3.2. Channel 2: improve electoral coordination The electoral formula of LegCo elections is proportional representation (PR) with the largest remainder. Because the PR system encourages small parties to participate, parties from both political camps field candidates to compete in each LegCo election. Consequently, the electoral formula creates pressures for intra-camp competitions, which may lead to suboptimal outcomes for a camp as a whole. For instance, when there are three candidates representing three different parties from the pan-democratic camp, all of them may be able to get elected if the votes they received are even. But if a candidate receives more than two-thirds of the votes of the pan-democratic camp, then only one candidate from this camp gets elected, as the votes ”in excess” of the winning pan-democrat cannot be transferred to the other two candidates. For this reason, the current electoral formula benefits the political camp that is able to coordinate votes among its constituent parties, minimizing the aggregate vote loss due to within-camp competitions. The success of this vote coordination depends on a political camp’s ability to calculate the amount of votes it obtains in a district. With accurate information, the camp is able to calculate the optimal number of lists to field in a legislative district to avoid excessive intra-camp competition. It can also make use of the information to decide how to split votes among

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

the lists. One way to do this is to assign lists among the District Council Constituencies it controls. To illustrate the idea, suppose a political camp fields two lists in a LegCo geographical constituency, and in that LegCo geographical constituency the political camp controls twenty District Council Constituencies. Then it may assign List A to ten District Council Constituencies and List B to the remaining ones. District Councillors assigned with List A would only help List A candidates to solicit votes during the election.11 In other words, District Councillors play a critical role in vote coordination. Because they have been serving the community at the grassroots level, they are the one who has the best information about the local residents. This explains why the pro-establishment camp has a strong incentive to drive out the pan-democratic camp from the District Councils, because doing so can disrupt its rival’s vote coordination at the LegCo level, while improving its own vote-splitting strategies. Thanks to its aggressive expansion in the District Councils, the proestablishment camp has improved its vote coordination, which can be reflected in Fig. 1. As can be seen from the figure, the seat-share gap between the two camp is narrower than their vote-share gap, suggesting that the proestablishment camp can attain proportionally more seats than the pan-democrats given a vote percentage. 3.3. Channel 3: utilize pro-establishment parties’ comparative advantage While the first two electoral advantages associated with occupying a District Council seat are not confined to any political camp, the last electoral advantage is specific to the pro-establishment bloc. One of the most salient electoral cleavages in Hong Kong’s legislative elections is political liberalization (Oksanen, 2011). In times of elections, pandemocratic candidates frequently exploit this cleavage by urging the government to implement universal suffrage in the election of the Chief Executive and to abolish the functional constituencies, while criticizing the proestablishment camp as a hurdle to democratization. In this heated confrontation, pro-establishment parties have difficulties presenting a convincing counter-argument against the pan-democrats’ call for political liberalization (Sing, 2010). Many pro-establishment legislators themselves were the beneficiary of the status quo; they managed to enter the LegCo through the uneven playing field of the functional constituencies. Their opposition to universal suffrage is easily perceived by voters as a defense of their vested interest, rather than as a genuine concern for the well-being of Hong Kong society. As political liberalization has been a recurring theme in legislative elections, this has put the pro-establishment camp in a weak position. The election of District Councils is different. Precisely because this office seems so politically insignificant, ideological confrontation such as the concern over political

11 This vote-splitting strategy was first used by the Democratic Party, who has had the largest number of District Council seats within the pandemocratic camp.

207

liberalization plays a weaker role in District Council elections. In fact, many voters tend not to evaluate a candidate based on her stance on democratization because voters seldom expect a District Councillor, who has difficulties effecting substantial policy change to her tiny constituency, can contribute anything to the democratization movement. Instead, voters would like to judge candidates by their constituency services.12 It is this area where proestablishment parties can outcompete the pan-democrats. Mundane as they are, District Councils’ constituency services are financially and administratively costly. Resource-rich parties therefore have an advantage in the provision of such services. Take the leading proestablishment party, the DAB, as an example. The DAB is arguably the wealthiest political party in Hong Kong. It reportedly obtained US$6 million political donations in the fiscal year of 2009/2010, while the largest two prodemocracy parties combined received less than a quarter of that amount (Ming Pao Daily, 2011). The great discrepancy in part results from the business sector’s reluctance to offer political donation to the pro-democracy camp for fear of provoking Beijing. This severely constrains prodemocracy parties’ financial resources. Pro-establishment parties are rich not only in financial capital, but also in human resources. The DAB has over 20,000 members, while the DP has some 700 only. The disparity is not surprising. As a Beijing-sponsored party, the DAB can offer many tangible and intangible political or economic benefits to its members. In contrast, members of the opposition movement face censure and ostracism directly and indirectly by Beijing.13 The FTU offers another illustrative example. This labor union-cum-party has 300,000 members. It sets up a team of volunteers for community services in virtually every single district. Four of these volunteer teams contributed over 10,000 manhours to do community services in 2002 (Hong Kong Federation of Trade Unions, 2012). As a comparison, the DP cannot afford setting up a party branch in every district, and some of the local branches it has do not even have a regular office. When it comes to the provision of constituency services, most pan-democratic District Council members can only count on a government subvention; each District Councillor is entitled to a monthly subsidy worth about US$3300 that can be used to run her office and organize local activities. Meagre as it is, it provides many pan-democratic District Councillors a stable source of income to finance community services. The District Councils, therefore, provide an ideal setting for pro-establishment parties to expand their support base. With ample resources, they offer more, if not better, constituency services to local residents than the resource-poor pan-democrats. As such, they establish political credentials

12 Corruption, which is an important electoral issue in many competitive authoritarian regimes, plays a much weaker role in Hong Kong’s elections, in part because Hong Kong has continued to keep a clean and efficient bureaucracy, despite the lack of democracy. 13 For instance, pro-Beijing business groups would not place advertisements in pro-democracy news media for fear of displeasing Beijing. In addition, many members of the Democratic Party are denied entry to the mainland.

208

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

as a capable companion who can provide solutions to practical problems related to livelihood. These credentials have gradually become the pro-establishment parties’ issue ownership in both District-Council and LegCo elections, diverting voters’ attention from the long-standing confrontation over political liberalization. The foregoing analysis suggests that the electoral advantages brought by the District Council office are substantial, though not necessarily obvious to ordinary voters. For this reason, pro-establishment parties have aggressively expanded their turf in the District Councils in the past decade. They have achieved great successes. Fig. 3 displays the over time change of the geographical distribution of District Council seats by political camps. As can be seen from these maps, the pro-establishment camp, denoted by the color red, enjoys an increasingly dominant position in District Councils. While in 2003 the proestablishment camp occupied only 28 percent of the elected District Council seats, it controls more than 76 percent of the seats eight years later. The rapid expansion of the pro-establishment camp at the District Council level requires a substantial amount of resource inputs. A natural question to ask is ”are District Councils a worthwhile political investment?” To answer this question, we need to examine whether occupying a District Council seat can really help the pan-democratic camp in the subsequent LegCo elections. If the answer is yes, then driving them out from the District Councils serves Beijing’s as well as the pro-establishment parties’ political interests. I will estimate the District-Council effect in the next section.

2003 District Council Election

2007 District Council Election

2011 District Council Election

4. Methodology If one wants to measure the District-Council effect on the outcome of a LegCo election (the popularly elected geographical constituencies), one needs to take into account of a potential spurious relationship. In the studies of competitive authoritarianism, few would dispute that coercion alone is insufficient to explain why people vote for the ruling party.14 There are voters who genuinely support the regime. For example, Geddes and Zaller (1989) point out that the least and the most educated voters are less likely to support an authoritarian regime because for the former, the state propaganda fails to reach them, and for the latter, they are politically knowledgeable enough to distinguish between propaganda and reality. Shi (1997) finds a similar curvilinear pattern in China. Based on his extensive research in rural China, Li (2004) provides a vivid picture of the powerful effect of the state media’s propaganda: ”villagers hear many beautiful promises from the central leaders; it is no wonder that many of them come to believe in the center and its policies (p. 235).” The empirical evidence presented in Kern and Hainmueller (2009) even suggests that exposure to quality television entertainment is sufficient to raise people’s tolerance of an authoritarian regime, thereby increasing its public support.

14 Geddes (2005) points out that nearly all authoritarian regimes that hold regular elections employ the secret ballot. It is therefore not costly for individual citizens to vote against the ruling party.

Notes: Each District Council Constituency elects a District Council member by the plurality rule. There are about 400 District Council constituencies in each election.

Fig. 3. Vote Shares of Pan-Democrats and Pro-Establishment Camp in District Council Elections.

When estimating the effect of the District Councils, one should distinguish this genuine support from the voters, known as the effect of voter preference, from the instrumental effect of the District Council office, which is our variable of interest. A legislative candidate gets elected sometimes because voters genuinely support her ideology or party (the effect of voter preference). She may also get elected because, as I argue in the previous section, her party has successfully captured the District Councils so that her rival’s income source is disturbed and vote coordination strategy is disrupted (the instrumental effect of the District Council office). Note that the effect of voter preference and the instrumental effect are not mutually exclusive. In fact, the instrumental effect is arguably intended to influence voter preference in the long run. When estimating the effect of carrying a District Council seat on the subsequent

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

legislative election, however, one needs to be cautious, in order not to mistake the effect for the cause. For this reason, showing that pan-democratic parties receive more votes in districts where the District-Council seats are controlled by themselves is insufficient to support the claim that winning the District Council seats improves the pan-democratic parties’ electoral support. This is because we cannot separate the effect of voter preference from the instrumental effect of the District Council office; for example, in a District Council constituency where a pro-establishment candidate is elected, voters may genuinely favor the pro-establishment camp over the pan-democrats. As a result, when it comes to the LegCo election, the pro-establishment camp is likely to receive more votes. The correlation between a political camp’s control of a District Council Constituency and the camp’s LegCo vote share in that constituency reflects only the underlying preference of the voters, rather than the instrumental effect of winning a District Council office. Because voter preference at the District Council level is unobservable independent of voting outcomes, one is confronted with a potential spurious relationship between occupying a District Council seat and the outcomes of a subsequent legislative election. To estimate the instrumental effect of winning a District Council seat on the election outcome of a subsequent legislative election, I employ a regression discontinuity design (RDD) to tackle this identification problem. The regression discontinuity design, as will be discussed in this section, is a quasi-experimental research design that mimics random assignment of treatment and control groups in a randomized controlled experiment. In a laboratory experiment, a researcher randomly assigns subjects to the treatment and control groups. Because randomization tends to produce relatively balanced control and treatment groups, the researcher can significantly minimize the risk of omitted variable bias. As such, the identified effect is more likely due to the treatment effect rather than the effects of other confounding factors. In the current context, an ideal research design is to randomly assign District Council seats to political parties. For instance, some proestablishment parties would land on districts that are ideologically predisposed to the pro-establishment camp, and some would land on districts in favor of the pan-democrats. In other words, we can avoid the situation where only districts that are ideologically inclined to pro-establishment parties would self-select to be led by pro-establishment parties. We can then examine the effect of occupying a District Council seat on LegCo elections by comparing the vote shares obtained by the pro-establishment camp and those obtained by the pan-democrats. In reality, I cannot affect the data-generating process of the District Council elections, but I can apply regression discontinuity, a quasi-experimental research design, to find out the causal effect. The idea of regression discontinuity is simple.15

15 There is a growing number of applied election studies using the regression discontinuity design. Notable examples include Fujiwara (2011), Gerber et al. (2011), Eggers and Hainmueller (2010), and Hainmueller and Kern (2008).

209

If occupying a District Council Constituency has some effect on the outcome of the legislative elections, the relationship between the pan-democrats’ LegCo vote shares and their District Council vote shares, for example, should be best characterized by a function discontinuous at a certain threshold (such as 50 percent in a two-party vote of the District Council elections). The discontinuous jump is the causal effect of capturing a District Council seat on the outcome of the LegCo elections. The reason is that no matter how close a pan-democratic party’s District Council vote share gets to the threshold, the party would not get elected, and hence cannot occupy the District Council office, until its vote share just surpasses the threshold. The validity of regression discontinuity hinges upon the assumption that districts are very similar to each other in the neighborhood of the discontinuity. The only difference that sets them apart is whether they happen to receive the ”treatment” by chance; that is, whether the pan-democratic party obtains, due to random uncontrollable factors, barely sufficient votes to carry the District Council Constituencies. In other words, we have balanced treatment and control groups in the neighborhood of the discontinuity as if in a randomized experiment, so that the causal effect identified are more likely due to the effect of the treatment rather than the effect of other factors such as the district’s ideological predisposition. This argument is formalized in Lee (2008). To estimate the electoral effect of the District Council office, I examine the relationship between the pandemocratic camp’s LegCo vote share in each District Council Constituency (the dependent variable) and this camp’s margin of victory in the same District Council Constituency (the independent variable or the forcing variable). More formally, we can express the regression discontinuity design in the following way:

 Di ¼

1 if xi > 0 0 if xi  0

where Di is the treatment status of District Council Constituency i, with the value “1” denoting the constituency controlled by the pan-democratic camp and “0” otherwise, and xi is the pan-democratic camp’s margin of victory in the District Council Constituency i. This leads to the main regression specification:

yi ¼ f ðxi Þ þ dDi þ mi

(1)

where yi is the pan-democratic camp’s LegCo vote share in constituency i, f(xi) is a polynomial function, d is the causal effect of interest, and mi is an error term assumed to be independent and identically distributed. The polynomial function is intended to provide a flexible functional form to model the relationship between yi and xi, which is not necessarily linear, in order to avoid mistaking non-linearity for discontinuity.16

16 Angrist and Pischke (2009) provide a detailed discussion on this point.

210

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

Using a political camp’s vote share (as opposed to a party’s) as the unit of observation requires additional data work. In each District Council election, there are a few districts in which more than one pro-establishment (pandemocratic) parties compete against a pan-democratic (pro-establishment) candidate, which results from proestablishment (pan-democratic) parties’ occasional failure of co-ordinating among themselves. This gives its archrivals an opportunity to capture districts that are supposed to be its stronghold. In other words, we observe the pro-establishment camp as a whole beats the pandemocrats in vote share (xi > 0), but loses the seats (Di ¼ 0). Such cases, which account for less than 2 percent of all cases, violate the assumption of deterministic treatment assignment. I therefore drop them from the sample.17

5. Empirics 5.1. Data issues The Hong Kong government’s Electoral Affairs Commission publishes data on District Council and the Legislative Council elections on its Web site. I use the outcomes of the 2003, 2007, and 2011 District Council elections to predict the outcomes of the 2004, 2008, and 2012 LegCo elections. The arrangement of District Council elections differs from that of LegCo elections in a number of respects. Take the 2007 District Council election as an example. There are 405 District Councillors elected by the plurality rule from 405 districts. As for the 2008 LegCo election, there are only five LegCo geographical constituencies electing 30 LegCo members by the method of proportional representation.18 The five LegCo districts are supersets of the 405 District Council Constituencies, and there is no District Council Constituency that cuts across the boundary of LegCo districts. Because the Electoral Affairs Commission of the Hong Kong government publishes electoral results of the LegCo at the District Council Constituency level,19 this allows me to measure the effect of a party capturing a District Council seat on that party’s vote share in the same constituency in a subsequent LegCo election. Note also that District Council elections are held in November, while LegCo elections take place in the September of the following year. A party that succeeds in capturing a seat in a District Council should have at least several months to bring private benefits to its constituency (or a longer time if the party is already an incumbent),

17 For the majority of these cases, the pro-establishment camp’s margins of victory lie far away from the neighborhood of the threshold. Hence, excluding them has little impact on the regression discontinuity analysis. 18 Although incumbent District Council members are allowed to compete for legislative seats, the difference in the seat numbers between these two levels suggests that only a few District Council members can simultaneously hold a seat in the LegCo. 19 The electoral results are available from the Commission’s Web site: http://www.eac.gov.hk/.

which may affect its chances of success in the ensuing legislative election.20 In Hong Kong, the law stipulates that no political party is allowed to control the government. Despite this unique political arrangement, voters have little difficulty distinguishing between the pro-incumbent parties and the pan-democratic parties.21 One simple litmus test is whether a party supports the ”June 4 Incident” legislative motion that asks for a vindication of the 1989 prodemocracy student movement in Beijing, of which the Chinese government firmly disapproves and for which Hong Kong society at large has an enduring sympathy. This motion is proposed every year in the LegCo on the eve of the anniversary of the June 4 Incident, but has never been passed since Hong Kong returned to the PRC. If a party supports the ”June 4 Incident” legislative motion, it is clearly in opposition to Beijing. Based on the LegCo’s roll call data, I classify such parties as the pan-democrats and those who do not support this motion as proestablishment. The list of the pro-establishment parties and the pan-democratic parties is presented in Table 2. Unambiguous as it is, this party-based classification may downplay the actual presence of pro-Beijing elements in the District Councils. As discussed, voters in District Council elections pay far less attention to the issue of political liberalization. Candidates running for these elections can easily get away with not revealing their stance on this issue. In fact, many pro-Beijing candidates deliberately run under an independent label to avoid arousing the animosity of prodemocracy voters.22 For this reason, I provide an alternative classification of the pro-establishment camp, which includes both parties and individual politicians. To classify pro-Beijing District Council candidates, I check whether the candidate is a member of, or has received an open endorsement from, pro-Beijing social groups. There exist a number of pro-Beijing mass associations (see Appendix A) led by key social or political figures who hold important positions in the formal institutions of the PRC such as the National People’s Congress. Each of these mass associations consists of more than a hundred umbrella organizations, which help facilitate the penetration of pro-Beijing elements into Hong Kong society (Lo et al., 2002). These mass associations would place advertisements in major proBeijing news outlets (such as Wenhui Pao) prior to elections, District Councils’ included, to announce which candidates they endorse. By examining this list of endorsement and checking if a candidate belongs to any of these mass associations, I can uncover a more subtle political affiliation of about half of the so-called independent candidates. Readers, however, should use this alternative classification with caution, as it may still fail to uncover more covert political connections or may overestimate the electoral significance of the identified political connections.

20 With only a few months, the party cannot undertake large public projects, which may limit what it can offer to its constituency. Nevertheless, District Council members have no formal decision power to carry out such projects. 21 Voters can find candidates’ pictures and party logos on ballots. 22 About 40 percent of candidates are self-identified independent in each District Council election.

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219 Table 2 Major political parties in Hong Kong. Party

Pro-establishment Camp Civil Force (CF) Democratic Alliance for the Betterment and Progress of Hong Kong (DAB) Hong Kong Federation of Trade Unions (FTU) Hong Kong Progressive Alliance (HKPA) Kowloon West New Dynamic (KWND) Liberal Party (LP) New People’s Party (NPP) Pan-democratic Camp Association for Democracy and People’s Livelihood (ADPL) Civic Party (CP) Democratic Party (DP) Hong Kong Confederation of Trade Unions (CTU) Labor Party (LAP) League of Social Democrats (LSD) Neighborhood and Workers Service Centre (NWSC) Neo Democrats (ND) People Power (PP)

Number of Number of legislative district council seats council seats 15 134

0 13

13

6

. 22 9 4

. 1 5 2

15

1

7 47 .

6 6 .

0 0 5

4 1 1

8 1

1 3

Notes: Data on seat numbers come from the 2011 District Council Election and the 2012 Legislative Council Election. Only parties that have participated in the popularly elected geographical constituencies and District Council elections are included. The HKPA merged into the DAB in 2005. The CTU is the predecessor of the Labor Party.

5.2. Results 5.2.1. Estimation by polynomial regression Table 3 displays the regression results based on Equation (1). The independent variable MARGIN is the pandemocratic camp’s margin of victory in a District Council election held at time t. The subscript “þ” (“–”) denotes margin above (below) the threshold. The dependent variable is vote shares received by the pan-democratic camp in a Legislative Council election held ten months later. The causal variable of interest is D, which is a dichotomous variable with the value “1” indicating the pan-democratic camp controls the District Council Constituency and “0” otherwise. I include different polynomial fits to model the data. First, consider the polynomial regressions using the party-based classification of the pro-establishment camp (the first four specifications of Table 3). The causal variable of interest is statistically significant across specifications. Consistent with my expectation, the sign is positive, indicating that the pan-democratic camp can improve a District Council Constituency’s support for itself in LegCo elections if it can occupy that District Council seat. Because the jump measured by D occurs at the 0 percent mark, it is less likely due to other factors such as a constituency’s prevailing ideology. The effect is also of great substantive significance. According to the cubic model, which seems to be the best fit because it gives the lowest Akaike information criterion (AIC) of model selection, the pan-democratic camp can increase its votes in LegCo elections by about 5 percentage

211

points by capturing a District Council seat. As a comparison, the total loss of vote support for the pan-democratic camp between 1998 and 2012 is about 8 percentage points. It is also noteworthy that the coefficient on D captures the short-term effect of winning a District Council seat, namely, the effect exerting on a legislative election that occurs ten months after a District Council election. One may interpret the effect as large because the pan-democratic camp spends only a few months in the District Council office, which suffices to boost the pan-democrats’ votes by 5 percentage points. The last four columns of Table 3 show the results of polynomial regressions using an alternative definition of the pro-establishment camp, which includes both parties and individuals having connections with Beijingsponsored social groups. The cubic model is again the best model by AIC. The coefficient on the variable of interest, D, is 4.193, which is both statistically and substantively significant. 5.2.2. Estimation by local linear regression An alternative way to estimate the effect under the regression discontinuity design is local linear regression, which is amount to running standard linear regression within the neighborhood on both sides of the cutoff point. Lee and Lemieux (2010) suggest that this nonparametric estimation strategy serves as a complement to parametric estimation such as polynomial regression (p. 284). A key issue with this estimation strategy is how to select the appropriate bandwidth to define the neighborhood. If the bandwidth is too large, the estimate may become biased because the observations within the bins are no longer ”as good as randomized.” If the bandwidth is too small, the resulting estimate may become imprecise due to the lack of observations. Imbens and Kalyanaraman (2012) propose one data-driven method to find appropriate bandwidths. Applying their method to select bandwidths, I run local linear regressions based on Gaussian kernel and rectangular kernel. As can be seen from Table 4, the estimates of local linear regressions show larger negative effects than the estimates of polynomial regressions. They are also statistically significant at 10 percent or less. The Imbens– Kalyanaraman bandwidths are 4.318 and 9.269 based on Gaussian and rectangular kernels, respectively. These bandwidths are not small, compared to the ones used in other electoral studies. To check sensitivity, I re-run the local linear regressions using a bin size half that bandwidth. The resulting estimates are more significant, both substantively and statistically. The results presented in Tables 3 and 4 indicate that the pan-democratic camp’s occupation of District Council seats is conducive to its electoral performance at the LegCo level. They also explain the pro-establishment camp’s motivation to drive out the pan-democrats from the District Councils: to undermine the pan-democrats’ presence in the LegCo by uprooting their local support network. In fact, had the pro-establishment camp remained complacent about its rival’s presence in the District Councils, it would have incurred a high political cost in LegCo elections. In the first four columns of Table 5, I regress the pro-establishment camp’s LegCo vote shares on the

212

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

Table 3 Effect of controlling a district council by pan-democratic camp on its vote share of legislative council election: polynomial regressions. Dep. Var.

Pan-democratic camp’s vote share in LegCo election

Definition of pro-establishment

Parties only

D Margin Marginþ

Parties and individuals

Linear

Quadratic

Cubic

Quartic

Linear

Quadratic

Cubic

Quartic

5.516*** (0.753) 0.026** (0.011) 0.019** (0.009)

3.581*** (0.948) 0.103** (0.051) 0.148*** (0.054) 0.001* (0.000) 0.001*** (0.001)

5.031*** (0.994) 0.149 (0.143) 0.187 (0.137) 0.008** (0.004) 0.010** (0.004) 0.000** (0.000) 0.000*** (0.000)

4.286*** (0.870) 0.015* (0.009) 0.067*** (0.015)

0.929 (1.269) 0.069 (0.048) 0.261*** (0.056) 0.000 (0.000) 0.002*** (0.001)

4.193*** (1.454) 0.144 (0.125) 0.167 (0.134) 0.006* (0.003) 0.011*** (0.004) 0.000** (0.000) 0.000*** (0.000)

31.018*** (1.157) 756 0.55 5308.4

31.171*** (1.264) 756 0.55 5301.2

30.687*** (1.501) 756 0.57 5284.9

4.943*** (1.026) 0.318 (0.280) 0.069 (0.268) 0.016 (0.014) 0.004 (0.013) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 29.801*** (1.750) 756 0.57 5287.7

31.688*** (1.087) 856 0.52 6043.8

32.239*** (1.224) 856 0.53 6032.7

31.368*** (1.476) 856 0.54 6015.1

3.260** (1.592) 0.191 (0.247) 0.231 (0.255) 0.007 (0.012) 0.011 (0.012) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000* (0.000) 30.844*** (1.694) 856 0.54 6015.5

Margin2 Margin2þ 3

Margin Margin3þ Margin4 Margin4þ Constant N R2 AIC

Notes: D is the treatment status, with the value “1” denoting the District Council controlled by the pan-democratic parties and “0” otherwise. MARGIN is the pan-democratic camp’s margin of victory in a District Council election held at time t. The subscript “þ” (“–”) denotes margin above (below) the threshold. All regressions control for year and LegCo constituency fixed effects. Standard errors are in parentheses. *<0.10, **<0.05, ***<0.01.

share on the pro-establishment camp’s margin of victory in District Council elections. In the middle four columns of the table, the pro-establishment camp is defined as Beijingsponsored political parties only, whereas in the last four columns, it includes both parties and pro-Beijing individuals. Interestingly, the variable of interest, Dpro-establishment, yields little statistically significant effect with the more restrictive definition. But if we include also the ”undercover” pro-Beijing elements in the definition of the proestablishment camp, the impact appears. Based on the cubic model, the one with the lowest AIC, the pro-establishment camp is able to reduce its rival’s LegCo vote share by about 4 percentage points if it could capture a District Council seat. Such an effect is statistically significant at 1 percent. Fig. 4 presents a visualization of the effect of controlling a District Council seat, based on the cubic polynomial regression of Table 3. The most striking feature of the figures

pan-democrats’ margin of victory in the District Council elections. The coefficient on the variable of interest, Dpandemocrat, is negative and statistically significant across specifications. The cubic model shows that by capturing a District Council Constituency, the pan-democratic camp is able to reduce the pro-establishment’s LegCo vote share by 3.4 percentage points. Juxtaposing this number with the pan-democratic camp’s vote gained according to Column 3 of Table 3, one can see that a significant share of the electoral gain of the pan-democratic camp due to the occupation of a District Council seat could have gone to the proestablishment camp had the pan-democrats failed to capture that seat. A closely related question is ”has the pro-establishment camp’s aggressive expansion at the District-Council level been able to bring down the LegCo vote share of the its rival?” In Table 5, I regress the pan-democrats’ LegCo vote

Table 4 Effect of controlling a district council by pan-democratic camp on its vote share of legislative council election: local linear regressions. Dep. var.

Pan-democratic camp’s vote share in LegCo election

Definition of pro-establishment

Parties only

D Bandwidth

Parties and individuals

Gaussian Kernel

Rectangular Kernel

Gaussian Kernel

Rectangular Kernel

11.96* (6.60) 4.32

12.88* (7.46) 9.27

14.38** (6.36 3.79

9.05 (7.91) 8.13

19.74*** (6.66) 2.16

20.61** (8.78) 4.63

24.58*** (6.00) 1.89

25.39*** (7.35) 4.07

Notes: D is the treatment status, with the value “1” denoting the District Council controlled by the pan-democratic parties and “0” otherwise. Bandwidths are calculated using the Imbens–Kalyanaraman method with specified kernel. The left column under a specific kernel uses the I–K bandwidth, while the right column half that bandwidth. All regressions control for year and LegCo constituency fixed effects. Standard errors are in parentheses. *<0.10, **<0.05, ***<0.01.

Table 5 Effect of controlling a district council on the rival camp’s vote share of legislative council election: polynomial regressions. Dep. var.

Pro-establishment camp’s vote share in LegCo election

Pan-democratic camp’s vote share in LegCo election

Definition of pro-establishment

Parties only

Parties only

Linear

Quadratic

Cubic

Quartic

Dpan-democrat

4.630*** (0.581)

2.572*** (0.725)

3.403*** (0.766)

3.062*** (0.789)

Margin Marginþ

0.030*** (0.008) 0.032*** (0.007)

0.154*** (0.039) 0.133*** (0.041) 0.001*** (0.000) 0.001*** (0.000)

0.082 (0.110) 0.048 (0.106) 0.006** (0.003) 0.002 (0.003) 0.000** (0.000) 0.000 (0.000)

42.480*** (0.892) 756 0.37 4915.3

41.817*** (0.967) 756 0.38 4896.1

42.787*** (1.155) 756 0.39 4889.5

Margin2 Margin2þ 3

Margin Margin3þ Margin4 Margin4þ Constant N R2 AIC

0.034 (0.215) 0.240 (0.206) 0.002 (0.011) 0.010 (0.010) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 42.855*** (1.345) 756 0.40 4890.1

Linear

Quadratic

Cubic

Quartic

Linear

Quadratic

Cubic

Quartic

4.541*** (1.121) 0.008 (0.011) 0.037*** (0.012)

1.904 (1.570) 0.277*** (0.059) 0.211*** (0.059) 0.003*** (0.001) 0.002*** (0.001)

1.627 (2.032) 0.068 (0.159) 0.008 (0.167) 0.007* (0.004) 0.004 (0.004) 0.000** (0.000) 0.000 (0.000)

3.758*** (0.874) 0.059*** (0.016) 0.023*** (0.009)

0.678 (1.169) 0.270*** (0.053) 0.072 (0.049) 0.002*** (0.001) 0.001 (0.000)

3.981*** (1.360) 0.180 (0.136) 0.130 (0.123) 0.011*** (0.004) 0.006* (0.003) 0.000*** (0.000) 0.000** (0.000)

36.887*** (1.095) 756 0.52 5344.4

32.056*** (1.412) 756 0.55 5314.1

34.298*** (1.720) 756 0.55 5309.9

0.062 (2.370) 0.356 (0.308) 0.054 (0.324) 0.015 (0.014) 0.001 (0.016) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 32.557*** (2.039) 756 0.55 5311

35.825*** (1.017) 856 0.52 6049.5

33.001*** (1.201) 856 0.53 6032.9

35.424*** (1.450) 856 0.54 6014.9

3.057** (1.493) 0.210 (0.260) 0.168 (0.242) 0.010 (0.012) 0.006 (0.011) 0.000 (0.000) 0.000 (0.000) 0.000* (0.000) 0.000 (0.000) 34.031*** (1.685) 856 0.54 6015.5

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

Dpro-establishment

Parties and individuals

Notes: Dpan-democrat (Dpro-establishment) is the treatment status, with the value “1” denoting the District Council controlled by the pan-democratic (pro-establishment) parties and “0” otherwise. For the first columns, MARGIN is the pan-democratic camp’s margin of victory in a District Council election held at time t. For the last eight columns, MARGIN is the pro-establishment camp’s margin of victory in a District Council election held at time t. The subscript ”þ” (”-”) denotes margin above (below) the threshold. All regressions control for year and LegCo constituency fixed effects. Standard errors are in parentheses. *<0.10, **<0.05, ***<0.01.

213

214

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

is the discontinuity of the cubic polynomial fits at the 0 percent mark. This discontinuity is evidence of the effect of gaining a District Council seat (top graph) and that of losing the seat to the pro-establishment camp (bottom graph). 5.3. Robustness checks The central assumption of regression discontinuity is that observations in the neighborhood of the threshold are so similar that those immediately below the threshold (the bare losers) provide a valid counterfactual to how those immediately above the threshold (the bare winners) would have behaved had they not passed the threshold. One way to test this assumption, as recommended by Imbens and Lemieux (2008), is to make a density plot of the forcing variable. If one observes a discontinuity in the distribution of the forcing variable at the threshold, one can suspect that cases immediately below and above the threshold may have intrinsic differences such that we cannot assume the two groups are a valid counterfactual to each other. I employ the density test proposed by McCrary (2008). The forcing variables used in the foregoing analyses pass

the density test. For a detailed discussion of the test, see Appendix B. Another way to check whether the treatment assignment is random in the limit is to compare the average values of other related factors in the neighborhood of the threshold. If we observe a significant difference between the treated and control districts over a certain covariate, we may suspect whether there is a certain factor affecting the outcome other than treatment assignment. I conduct means comparison tests over numerous potential confounding factors. The comparison is based on a window size of 1 percent. Treated cases refer to District Council constituencies controlled by the pan-democratic camp. I further run polynomial regressions using these confounding factors as the dependent variable. Appendix C presents the regression and means comparison test results. Almost none of these major covariates observes a significant p value in the means comparison test. The only exception is the male population. The population share of male residents in treated cases is on average lower than that in control cases by 3.4 percentage points. The regression discontinuity design is known as a quasi-experimental design because cases in the neighborhood of the discontinuity are very similar to each other, and the only systematic difference between them is that some happen to receive the ”treatment” by chance, while others do not. In other words, the treatment assignment is ”as good as randomized.” In randomized experiments where the treatment and control groups are balanced there is no need to control for other effects. In the current context, although Appendix C shows that constituencies around the neighborhood of the discontinuity are indeed very similar, it also indicates that District Council Constituencies controlled by the pan-democratic camp tend to have a slightly lower population share of male residents. To ensure that the empirical results are not biased due to the omission of this variable and others, I re-run the regression specifications in Table 3 by adding relevant control variables. Table 6 shows that the effects of the variables of interest remain statistically significant across all specifications, despite the inclusion of these controls. Finally, a dozen of District Council members serve simultaneously in the LegCo. These ”dual representatives” are better known, and should be more competent than average District Council members. They may also carry some kind of incumbency advantage as a lawmaker in the LegCo that leaves a positive impact on their likelihood of capturing the District Council office. To tackle the potential endogeneity influences, I re-run the above analysis, excluding these individuals from the sample. The results, which are available from the author, are consistent with those identified in the previous tables. 6. Discussion and conclusion

Fig. 4. Effect of Controlling a District Council on Pan-democrats’ Vote Share of Legislative Council Election fits.

Despite their political power, dictatorial rulers face resource constraints. But few would dispute that opposition elites in an authoritarian regime face a greater

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

215

Table 6 Effect of controlling a district council by pan-democratic camp on vote shares of legislative council election: adding controls. Dep. var.

D Pan-Democrat Vote Sharet1 DC Turnout DC Total Voters Pan-Democrat Margint1 Post-secondary Income < 10 K Income 20 K to 40 K Population 15 < Age < 24 Age > 65 Male Born in Hong Kong Employees Elementary Workers Professionals Constant N R2 AIC

Pan-democratic camp’s vote share in LegCo election

Pro-establishment camp’s vote share in LegCo election

Linear

Quadratic

Cubic

Quartic

Linear

Quadratic

Cubic

Quartic

4.265*** (0.725) 0.255***

3.022*** (0.956) 0.249***

3.980*** (1.023) 0.239***

3.970*** (1.063) 0.240***

3.945*** (0.592) 0.380***

2.779*** (0.779) 0.374***

3.763*** (0.831) 0.363***

3.440*** (0.861) 0.365***

(0.037) 0.035 (0.051) 0.000 (0.000) 0.011 (0.011) 0.324*** (0.109) 0.006 (0.081) 0.111 (0.082) 0.000 (0.000) 0.169* (0.102) 0.329*** (0.095) 0.836*** (0.238) 0.157*** (0.056) 0.226* (0.118) 0.099 (0.101) 0.037 (0.223) 52.003*** (16.556) 445 0.68 2866.6

(0.037) 0.039 (0.051) 0.000 (0.000) 0.013 (0.011) 0.317*** (0.109) 0.009 (0.081) 0.111 (0.082) 0.000 (0.000) 0.169* (0.102) 0.328*** (0.095) 0.814*** (0.238) 0.153*** (0.056) 0.238** (0.118) 0.098 (0.101) 0.047 (0.222) 50.179*** (16.529) 445 0.68 2865.8

(0.037) 0.048 (0.051) 0.000 (0.000) 0.012 (0.011) 0.325*** (0.108) 0.016 (0.081) 0.117 (0.082) 0.000 (0.000) 0.168* (0.101) 0.352*** (0.095) 0.826*** (0.237) 0.161*** (0.055) 0.241** (0.117) 0.090 (0.100) 0.049 (0.221) 50.601*** (16.431) 445 0.69 2861.6

(0.038) 0.047 (0.051) 0.000 (0.000) 0.012 (0.011) 0.321*** (0.109) 0.017 (0.081) 0.117 (0.082) 0.000 (0.000) 0.166 (0.102) 0.352*** (0.095) 0.827*** (0.237) 0.160*** (0.056) 0.241** (0.117) 0.088 (0.101) 0.056 (0.223) 42.918*** (16.367) 445 0.689 2865.5

(0.030) 0.133*** (0.042) 0.000 (0.000) 0.018** (0.009) 0.208** (0.089) 0.054 (0.066) 0.049 (0.067) 0.000 (0.000) 0.159* (0.083) 0.134* (0.077) 0.070 (0.194) 0.063 (0.045) 0.246** (0.096) 0.221*** (0.082) 0.018 (0.182) 86.953*** (13.511) 445 0.61 2685.7

(0.030) 0.137*** (0.042) 0.000 (0.000) 0.021** (0.009) 0.202** (0.089) 0.051 (0.066) 0.048 (0.067) 0.000 (0.000) 0.158* (0.083) 0.132* (0.077) 0.092 (0.193) 0.059 (0.045) 0.260*** (0.096) 0.220*** (0.082) 0.027 (0.181) 85.215*** (13.457) 445 0.62 2682.8

(0.030) 0.144*** (0.042) 0.000 (0.000) 0.021** (0.009) 0.207** (0.088) 0.050 (0.066) 0.046 (0.066) 0.000 (0.000) 0.155* (0.082) 0.150* (0.077) 0.093 (0.192) 0.067 (0.045) 0.257*** (0.095) 0.210** (0.081) 0.035 (0.179) 85.117*** (13.336) 445 0.63 2675.9

(0.030) 0.145*** (0.042) 0.000 (0.000) 0.020** (0.009) 0.201** (0.088) 0.053 (0.066) 0.048 (0.066) 0.000 (0.000) 0.147* (0.083) 0.146* (0.077) 0.080 (0.192) 0.070 (0.045) 0.259*** (0.095) 0.203** (0.082) 0.043 (0.180) 81.805*** (13.253) 445 0.630 2677.7

Notes: D is the treatment status, with the value “1” denoting the District Council controlled by the pan-democratic camp and “0” otherwise. Unless specified otherwise, the units of all socio-economic variables are a share of the population. “DC Turnout” is the voter turnout of District Council elections. All regressions control for year and LegCo constituency fixed effects. Standard errors are in parentheses. *<0.10, **<0.05, ***<0.01.

resource constraint vis-à-vis the ruling elite. This severely limits the opposition’s ability to challenge the ruling elite. Given the political significance of resources, how the ruling elite translates its resource advantage into a political strength deserves scholarly attention. Scholars of authoritarian politics find that holding regular elections is one way for the ruling elite to distribute government spoils to buy political support. Electoral institutions, however, are not uniform across autocracies, which implies that the effectiveness of distributing resources may vary from regime to regime. This paper examines one underappreciated institutional feature relevant to resource distribution in authoritarian regimes: lower-level elections. I argue that elections at grassroots levels favor the ruling party by allowing it to more fully utilize its resource advantage to buy political support, which would in turn undermine the opposition’s ability to develop a local support network that is

important to its struggle for democratization as well as for elected offices. Empirical investigation of the causal effect of authoritarian elections is challenging because of the existence of other confounding factors. For instance, as Gandhi and LustOkar (2009) cogently point out, there exist two possibilities why opposition parties managed to oust the incumbents through elections in some authoritarian regimes: ”It may be that electoral coalitions among opposition parties lead to their victory and control over the chief executive office, but it is equally plausible that alreadyweakened incumbents both allow opposition coalitions and desist from using fraud and manipulation as part of a predetermined ‘step out’ of power. Elections, in that case, have little causal force.” In the context of this study, a positive correlation between occupying an elected office at a lower

Guangdong Communities in Hong Kong Community Groups in Hong Kong Island Sources: The Web sites of the mass associations.

Community Groups in New Territories Working Class in Hong Kong

Community Groups in Kowloon

250 147 307 246

148

1996 100,000 179 1999 25,000 121 1997 30,000 43 1985 70,000 128 1947 310,000 173

Founding Year Number of Members in 2004 Number of Affiliated Organizations in 2004 Number of Affiliated Organizations in 2012 Major Cooptation Targets

Hong Kong Island Federation of Associations (HKIF) Kowloon Federation of Associations (KFA) New Territories Associations of Societies (NTAS) Hong Kong Federation of Trade Unions (FTU)

government level and a good electoral performance in an election of a higher-level does not necessarily imply that capturing a lower office is conducive to capturing a higher office. It may merely reflect voter preference, which is often unobservable independent of election outcomes. To tackle this potential spurious relationship, I take advantage of a quasi-experiment afforded by the electoral formula of Hong Kong’s District Council elections, the plurality rule, which allows me to apply a regression discontinuity design to evaluate the instrumental effect of occupying a lower-level government office on the outcomes of a higher-level election. The Hong Kong case illustrates how the ruling elite can use elections of a lower-tier government to undermine the political support of a pro-democracy opposition force. Despite inheriting a strong and popular prodemocracy opposition force, Hong Kong’s ruling elite, known as the pro-establishment camp, has significantly marginalized the political influence of the pandemocrats in the legislature since the sovereignty transfer. Remarkably, its success is not based on any heavy-handed suppression of the opposition’s political freedom such as denying the opposition’s right to stand for elections. Rather, pro-establishment parties actively participate in elections that are widely considered free and uncorrupted. They compete for voters’ support in the same way as the pan-democrats do. It is these direct and competitive elections through which the proestablishment camp gradually undermines the pandemocrats’ popular support. I argue that an important factor contributing to the electoral successes of the proestablishment camp is its aggressive expansion in the District Councils, the lowest elected tier in Hong Kong that has been used by the pro-democracy opposition elites to develop their political support base. I find statistical evidence based on the regression discontinuity design that by occupying a District Council seat, the prodemocracy opposition camp is able to increase its LegCo vote share in that constituency by 4–5 percentage points. The results explain the motive behind the proestablishment camp’s aggressive expansion at the District-Council level in the post-handover period; it has aimed to uproot the pan-democrats’ local support base by driving them out of the District Councils. Although elections of lower-tier governments favor the ruling party, this does not mean that holding such elections can guarantee regime survival. As I argue above, an important function of lower-tier elections is to enable the ruling party to more fully utilize its resource advantage over the opposition. When the regime’s resource advantage dissipates due to, for example, a precipitous decline in state resources, the ruling party may lose its ability to distribute spoils, and hence unable to exploit its advantage of participating in lowertier elections. In addition, such adverse structural conditions also favor, as Wahman (2011) argues, the formation of a cohesive coalition among opposition parties, which is considered an important factor contributing to democratization by election (Howard and Roessler, 2006).

Federation of Guangdong Community Organizations (FHKGCO)

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

A Major Pro-Beijing Mass Associations in Hong Kong

216

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

217

B Density of the Forcing Variables

The top graph shows the distribution of the forcing variable, the pan-democrats’ margin of victory in District Council elections, while the middle graph the proestablishment camp’s margin of victory. The discontinuity estimates at the threshold are 0.22 and 0.27, respectively. Their respective standard errors are 0.38 and 0.40; we fail to reject the null hypothesis that the forcing variables are continuous at the cutoff point. Caughey and Sekhon (2011) advise that a density plot of the incumbent margin is more useful than a density

plot of a particular party’s margin because the latter plot tends to conceal the incumbent advantage that affects who can barely win or barely lose. The bottom graph shows the incumbent camp’s margin of victory. Similar to the top and middle graphs, there exists no abrupt change in the distribution at the threshold. The discontinuity estimate is 0.21 with a standard error 0.34. Again, the null hypothesis of continuity cannot be rejected.

218

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219

C Covariate Balance

References Angrist, J.D., Pischke, J.S., 2009. Mostly Harmless Econometrics: an Empiricist’s Companion. Princeton Univ. Pr. Blaydes, L., 2010. Elections and Distributive Politics in Mubarak’s Egypt. Cambridge Univ. Pr. Caughey, D., Sekhon, J.S., 2011. Elections and the regression discontinuity design: lessons from close us house races, 1942–2008. Polit. Anal. 19 (4), 385–408. Cheibub, J.A., Gandhi, J., Vreeland, J.R., 2010. Democracy and dictatorship revisited. Public Choice. ISSN: 0048-5829 143 (1), 67–101. Cheung, C.Y., 2011. How political accountability undermines public service ethics: the case of Hong Kong. J. Contemp. China 20 (70), 499– 515. Democratic Party. http://www.dphk.org (accessed 08.09.12.), 2011. Diaz-Cayeros, A., Magaloni, B., Weingast, B.R., 2003. Tragic Brilliance: Equilibrium Hegemony and Democratization in Mexico. Manuscrito. Disponível em http://papers.ssrn.com/sol3/papers.cfm. Eggers, A.C., Hainmueller, J., 2010. Mps for sale? Returns to office in postwar British politics. Am. Polit. Sci. Rev. 103 (4), 513. Fujiwara, T., 2011. A regression discontinuity test of strategic voting and duvergers law. Quart. J. Polit. Sci. 6 (3–4). Gandhi, J., Lust-Okar, E., 2009. Elections under authoritarianism. Ann. Rev. Polit. Sci. 12, 403–422. Gandhi, J., Przeworski, A., 2007. Authoritarian institutions and the survival of autocrats. Comp. Polit. Stud. 40 (11), 1279. Geddes, B., 1999. What do we know about democratization after twenty years? Ann. Rev. Polit. Sci. 2 (1), 115–144. Geddes, B., 2005. Why parties and elections in authoritarian regimes?. In: Annual Meeting of the American Political Science Association, pp. 456–471. Geddes, B., Zaller, J., 1989. Sources of popular support for authoritarian regimes. Am. J. Polit. Sci., 319–347. Gerber, A.S., Kessler, D.P., Meredith, M., 2011. The persuasive effects of direct mail: a regression discontinuity based approach. J. Polit. 73 (1), 140–155. Hadenius, A., Teorell, J., 2006. Authoritarian Regimes: Stability, Change, and Pathways to Democracy, 1972–2003. Helen Kellogg Institute for International Studies. Hainmueller, J., Kern, H.L., 2008. Incumbency as a source of spillover effects in mixed electoral systems: evidence from a regressiondiscontinuity design. Elect. Stud. 27 (2), 213–227. Holliday, I., Hui, G.K.H., 2007. Local, advisory, and statutory bodies. In: Contemporary Hong Kong Politics. Hong Kong University Press. Hong Kong Federation of Trade Unions, 2012. FTU Eagerness about International Year of Volunteers. http://www.ftu.org.hk/zh-hant/about? id¼17&nid¼97 (accessed 09.08.12.).

Howard, M.M., Roessler, P.G., 2006. Liberalizing electoral outcomes in competitive authoritarian regimes. Am. J. Polit. Sci. 50 (2), 365–381. Imbens, G., Kalyanaraman, K., 2012. Optimal bandwidth choice for the regression discontinuity estimator. Rev. Econ. Stud. 79 (3), 933–959. Imbens, G.W., Lemieux, T., 2008. Regression discontinuity designs: a guide to practice. J. Economet. 142 (2), 615–635. Kern, H.L., Hainmueller, J., 2009. Opium for the masses: how foreign media can stabilize authoritarian regimes. Polit. Anal. 17 (4), 377–399. Lam, W.M., Lui, P. L. Tim, Wong, W., Holliday, I., 2007. Contemporary Hong Kong Politics: Governance in the Post-1997 Era. Hong Kong University Press. Lee, D.S., 2008. Randomized experiments from non-random selection in us house elections. J. Economet. 142 (2), 675–697. Lee, D.S., Lemieux, T., 2010. Regression discontinuity designs in economics. J. Econ. Lit. 48, 281–355. Levitsky, S., Way, L.A., 2010. Competitive Authoritarianism: Hybrid Regimes after the Cold War. Cambridge University Press. Li, L., 2004. Political trust in rural china. Modern China 30 (2), 228–258. Lijphart, A., 2012. Patterns of Democracy: Government Forms and Performance in Thirty-six Countries. Yale University Press. Lindberg, S.I., 2009. Democratization by Election: a New Mode of Transition. Johns Hopkins University Press. Lo, S.-hing, Yu, W.-yat, Wan, K.-fai, 2002. The 1999 district councils elections. In: Crisis and Transformation in China’s Hong Kong. M. E. Sharpe. Lust-Okar, E., 2005. Structuring Conflict in the Arab World: Incumbents, Opponents, and Institutions. Cambridge University Press. Lust-Okar, E., 2009. Legislative elections in hegemonic authoritarian regimes: competitive clientelism and resistance to democratization. In: Democratization by Elections: a New Mode of Transition, vol. 2009, pp. 226–245. Magaloni, B., 2006. Voting for Autocracy: Hegemonic Party Survival and Its Demise in Mexico. Cambridge University Press. Ma, N., Choy, C.-K., 2003. Political Consequences of Electoral System: The Hong Kong Proportional Representation System. City University of Hong Kong Press. McCrary, J., 2008. Manipulation of the running variable in the regression discontinuity design: a density test. J. Economet. 142 (2), 698–714. Ming Pao Daily, 2011. Foxy Data Leaks Donators of the pan-democracy camp. October 18. Molinar, J., Weldon, J., 1994. Electoral determinants and consequences of national solidarity. In: Cornelius, Wayne, Craig, Ann, Fox, Jonathan (Eds.), Transforming State-society Relations in Mexico. The National Solidarity Strategy. La Jolla: Center for US-Mexican Studies, UCSD. Oates, W.E., 1999. An essay on fiscal federalism. J. Econ. Lit., 1120–1149. Oksanen, K., 2011. Framing the democracy debate in Hong Kong. J. Contemp. China 20 (70), 479–497.

S.H.-W. Wong / Electoral Studies 33 (2014) 200–219 Persson, T., Tabellini, G.E., 2005. The Economic Effects of Constitutions. The MIT Press. Schedler, A., 2002. The nested game of democratization by elections. Int. Polit. Sci. Rev. 23 (1), 103–122. Shi, T., 1997. Political Participation in Beijing. Cambridge Univ Press. Sing, M., 2010. Explaining mass support for democracy in Hong Kong. Democratization 17 (1), 175–205. Stokes, S.C., 2005. Perverse accountability: a formal model of machine politics with evidence from Argentina. Am. Polit. Sci. Rev. 99 (3), 315. Svolik, M.W., 2012. The Politics of Authoritarian Rule. Cambridge University Press.

219

Taylor, M.M., 1996. When electoral and party institutions interact to produce caudillo politics: the case of Honduras. Elect. Stud. 15 (3), 327–337. Wahman, M., 2011. Offices and policies–why do oppositional parties form pre-electoral coalitions in competitive authoritarian regimes? Elect. Stud. 30 (4), 642–657. Wong, S.H.-W., 2012. Authoritarian co-optation in the age of globalisation: evidence from Hong Kong. Journal of Contemporary Asia 42 (2), 182–209. Taylor & Francis. Zolberg, A.R., 1966. Creating Political Order: The Party-states of West Africa. Rand McNally & Company, Chicago.