Surrounding the Cities from the Countryside: An Empirical Assessment of the Electoral Effects

Abstract

In this chapter, the actual electoral effect of the relationship vote-oriented grassroots strategy of Beijing-sponsored parties and the pandering of the pan-democrats to the protest vote are examined. Did the Beijing-sponsored parties’ grassroots strategy really achieve their intended effect? The chapter not only presents how their grassroots strategy led to their electoral successes in the District Councils but also how occupying the District Councils allowed them to marginalize the pan-democrats in the major battlefield, the Legislative Council. The empirical evidence challenges the conventional wisdom that the District Councils are politically insignificant.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

During its initial armed struggles against the Kuomintang (KMT), the Chinese Communist Party suffered from serious setbacks. Mao Zedong soon came to see the problem as a strategic mistake; the CCP had wrongly placed its focus on urban uprisings. As cities were the KMT’s stronghold, the CCP, while in its infancy, stood no chance of defeating the KMT. Mao later adopted an alternative strategy. He jettisoned the urban focus and moved his forces deep into the countryside, in the belief that before his party’s military capability grew stronger, it should avoid any head-on military confrontation with the KMT. The rural regions, he reasoned, provide a fertile ground to grow his revolutionary base and a shield against the KMT’s aggression. Once the CCP developed a bastion in the countryside, it could mount an offensive against the KMT in cities, the major battlefield.¹ Mao’s tactical maneuver is famously known as “surrounding the cities from the countryside,” which has been touted, at least by the official rhetoric, as the key to the CCP’s ultimate success in conquering the KMT.

In many ways, the Beijing-sponsored parties’ policy toward the District Council resembles Mao’s military strategy. Their major battleground is the Legislative Council. But these parties have difficulty making inroads into LegCo elections because political ideology plays a strong role in such elections, and their pan-democratic rivals have a solid issue ownership of political liberalization. For this reason, Beijing-sponsored parties shifted their focus to the “rural region,” a metaphor for the District Councils. As mentioned in the previous chapter, pan-democratic parties have no strategic advantage of occupying this elected tier. Peripheral as they seem, the District Councils are of paramount strategic importance, as they can act as a bastion against pan-democratic parties’ electoral expansion and, ultimately, undermine their support base from the ground up. In this chapter, I present empirical evidence to show how capturing District Council seats helps Beijing-sponsored parties “besiege” their pan-democratic rivals in LegCo elections.

6.1 The Causal Mechanisms

There is a popular belief in Hong Kong that prodemocracy voters’ behavior in LegCo elections is vastly different from that in District Council elections. In the latter elections, they may support pro-establishment candidates who offer quality constituency services. But when it comes to LegCo elections, these voters would always side with pan-democratic candidates, who are ideologically closer to them. As a result, pan-democratic parties have no need to worry even if they fare poorly in District Council elections. They may lose the battle, but they will win the war.

Not only ordinary citizens² but some pan-democratic politicians are also dogged believers in this optimistic view. I heard a party heavyweight of the Civic Party insist that her party was able to retain the same level of voter support, irrespective of its performance in District Council elections.³ Almost all of the District Councillors I have interviewed, however, believe otherwise.

There are four main reasons why District Council seats are strategically important to a party’s quest for higher elected offices:

• (1) They provide a stable source of income.

Running a LegCo election campaign is costly. Most pan-democratic parties are short of financial resources, partly because, as discussed in the previous chapter, the business community is reluctant to support these opposition parties for fear of antagonizing Beijing. Under such circumstances, the monthly salary contribution of elected party members, however meager, becomes a stable and important source of income for many pan-democratic parties. Losing District Council seats, thus, has a direct financial impact on a pan-democratic party’s ability to compete in LegCo elections.

• (2) They foster local support networks.

Behind the aforementioned popular belief about the disconnection between District Council and LegCo elections is an assumption that voters value ideological affinity dearly. To many District Councillors I spoke with, this is a fallacious assumption because personal rapport weighs more heavily than ideology in many voters’ minds. Sometimes people vote for a candidate simply because they had a chance to talk to her in person. In order for candidates to make an impression on local residents, they need to establish their contacts first. Collecting constituents’ contacts is, therefore, of paramount importance. Because the population size of the LegCo’s geographical constituencies is fairly large,⁴ it is not easy for LegCo candidates to obtain their constituents’ contacts, let alone their electoral support. In contrast, a typical District Council Constituency consists of only 17,000 residents. District Councillors, therefore, have a crucial role to play, as their local knowledge enables them to act as political brokers for LegCo candidates.

How does a District Councillor acquire the local knowledge? Casework is an integral part of a District Councillor’s job. Based on my interviews, diligent District Councillors handle more than a thousand cases related to their constituents each year. In other words, they should be able to accumulate thousands of local residents’ contacts over a single term of office. Moreover, they may hold various cultural and recreational activities, through which they can reach out to even more constituents. Suffice it to say, even if District Councillors cannot make themselves a household name in their constituency, no one has more acquaintances in the neighborhood than they do.

District Councillors’ local contacts can assist a LegCo candidate’s election campaign in several ways. The most straightforward use is canvassing; for example, the District Councillors call their constituents on election day to promote the LegCo candidate. More importantly, their local knowledge helps the LegCo candidate allocate resources more efficiently. Who are the swing voters? How many are they? These are important electioneering questions that a LegCo candidate cannot answer without District Councillors’ inputs.

Another important function of District Councillors’ local support networks is to supply volunteers. A good rapport with local residents earns District Councillors not only more votes but also more volunteers to assist in their work. In one interview, a DAB District Councillor contends that “recruiting volunteers is the most vital part of the job [of District Councillors].”⁵ This is because constituency services are labor intensive and wages in Hong Kong are high. Getting loyal supporters’ voluntary help can significantly bolster a District Councillor’s chance of getting reelected.⁶ In the case of LegCo elections, running an election campaign is no less labor intensive. The number of volunteers LegCo candidates can mobilize is typically very limited, given their relatively weak linkages with the constituency. They, therefore, need to solicit help from District Councillors, who can bring together a large pool of volunteers for their deployment.

“My job is analogous to multi-level marketing,” a DAB District Councillor describes his role in the party’s LegCo election campaigns.⁷

District Councillors have their own incentive to help LegCo candidates from their own parties. Because LegCo elections take place about a year after District Council elections, the result of a LegCo election is indicative of a District Councillor’s ongoing performance. One DAB District Councillor points out, “If I can get 80 % of the residents I know to support my endorsed LegCo candidate, I would then know that I stand a good chance in the coming District Council election.”⁸ District Councillors are probably not the only ones interested in knowing how many votes they can contribute to a LegCo election. Party leaders also want to have that figure in order to gauge junior party members’ performance. It is plausible that peer pressure exists among District Councillors in this respect.

• (3) They facilitate vote-splitting .

A more technical reason why District Councils are able to influence LegCo elections is that District Councillors facilitate vote-splitting in LegCo elections. The electoral formula of LegCo elections is proportional representation (PR) with the largest remainder method. Because the PR system encourages small parties to participate, parties from both political camps field candidates to compete in each LegCo election. Parties from the same political camp court support from similar voters, so intra-camp competition is very much alive. At times, such competitions lead to suboptimal outcomes for the 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 can get 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 support parties, minimizing the aggregate vote loss due to intra-camp competition. The success of this vote coordination in part depends on how accurately a political camp knows the amount of votes it obtains in a district. With accurate information, the camp can calculate the number of lists to field in a legislative district in order to avoid excessive intra-camp competition. It can also make use of the information to decide how to split votes among the lists. One way to do this is to assign lists to District Council Constituencies (Ma and Choy 2003, pp. 175–179). For instance, if a legislative district consists of 20 District Council Constituencies, 10 of these DCCs are mobilized to support one list, while the remaining 10 to another. Each District Councillor in these 20 DCCs is responsible for mobilizing supporters to vote for their assigned legislative list.

• (4) They groom newcomers.

The long-term survival of a political party depends on its ability to continue to produce new candidates to occupy elected offices. Whether a party’s new blood is electorally competitive in turn hinges upon experience and tutelage. The District Councils provide a training ground for parties to groom junior members. How to gain local residents’ trust? Which government office to contact when a certain problem arises? How to run an effective election campaign? What would attract media attention? District Councillors are confronted with such questions all the time. By solving these questions, they can develop political savvy and practical knowledge, both of which are requisite skills for a potent LegCo member who needs to face a much larger constituency with more complicated issues.

Competition among junior members allows a party to identify promising candidates for LegCo elections. For young District Councillors, an effective way to stand out from their peers is to attract media attention, which is a challenge to most District Councillors, because their job is notoriously mundane and insignificant beyond their DCCs. If a District Councillor is able to bring a media spotlight to her work, she is demonstrating her ingenuity, if not also the potential for a more important office.

The internal competition in Beijing-sponsored parties is more intense because of their ability to recruit new members. Take the DAB as an example. As of 2013, the party has 1,400 members under the age of 35 (Wen Wei Po 2012e). Although this number is small compared with the party’s 23,000 members, it surpasses the total membership of the Democratic Party . Another example that illustrates the keen competition among the DAB juniors is that each year the party would organize an in-house training program for young members including the current District Councillors. Admission to the program is fairly selective; the party recruits only one-third of the applicants after multiple rounds of interviews (Wen Wei Po 2013). In addition, it is reported that since 2013, the DAB has introduced a “point system” to continuously appraise its young members who are interested in LegCo seats (Ming Pao Daily News 2013c). This system is analogous to a primary election except that it spans several years. Presumably, a party needs a primary only when it has more candidates than seats available.

Identifying promising candidates is only the first step. The next step for a party is how to promote these rising stars to compete for LegCo seats. The problem of newcomers is a lack of a citywide recognition. The District Councils again have an important role to play.⁹ Under the PR system, candidates can gain a LegCo seat as long as they can obtain a small percentage of votes in a LegCo district. This can be achieved with the help of political brokers; as discussed, each District Councillor is able to mobilize at least a fraction of the residents to vote for a LegCo list by dint of their own effort. For this reason, as long as a party has a sufficient number of District Council seats, it does not really need to count on the mass media to promote its own newcomers.

6.2 The Electoral Effects

6.2.1 A First Glance

In the previous section, I show various causal mechanisms that explain why controlling District Council offices can benefit a party in its bid for LegCo elections. Some mechanisms are concerned with a party’s long-term survival (e.g., grooming newcomers), while others deal with short-run electoral effect (e.g., facilitating vote-splitting ). In this section, I focus primarily on measuring the short-run effect of the District Councils for two reasons. First, the long-term effect is still unfolding; we lack data to evaluate its full impact. Second, the short-run effect is politically important. If a party cannot survive in the short run, it may not have a chance to improve its long-term well-being. Although Hong Kong’s political system is not fully democratic, popular elections are nevertheless held on a regular basis, which creates constant reelection pressure for political parties. The reelection pressure shortens the time horizon that a political party faces. A politician may not be impressed to learn that capturing a District Council seat can improve her chance of winning a LegCo seat in the distant future. But she would be concerned to hear that losing her current District Council seat would reduce her vote share in the following LegCo election by 5 % points.

To measure the short-run effects on LegCo elections through occupying District Council offices, I compare the LegCo vote shares of major parties in District Council Constituencies under their control with those not under their control. If the District Council factor does not matter, we would expect to see no significant difference in vote share between these two types of District Council Constituencies.

As may be seen from Table 6.1, the District Council factor actually has an enormous effect. Except for People Power , all parties received a significantly higher vote share in their controlled District Council Constituencies than in the uncontrolled ones. The effect size (uncontrolled DCCs – controlled DCCs) ranges from − 6. 74 to − 29. 35 %. More concretely, consider the Democratic Party . Its LegCo vote share in a DCC would increase by 10 % points if the DCCs were under its control. The difference is unlikely due to chance because the difference of means tests are all statistically significant.

Table 6.1 Average vote share received in Legislative Council elections by political party: controlled and uncontrolled District Council Constituencies (DCCs)

The data in Table 6.1 seriously challenges the popular belief that losing District Council seats has no effect on LegCo elections. It also explains why Beijing-sponsored parties are anxious to drive the pan-democrats out of the District Councils; many pan-democratic parties actually demonstrate a stronger ability to convert their District Council seats into an electoral advantage in LegCo elections than their pro-establishment counterparts (most notably, the ADPL and NWSC ). To diminish the pan-democrats’ presence in the LegCo, uprooting their District Council seats is the first and foremost step that Beijing-sponsored parties have to take.

6.2.2 A Closer Look

Can Beijing-sponsored parties really undermine the pan-democrats’ electoral support by capturing more District Council seats? This question is tricky to answer, despite the seemingly compelling evidence presented in Table 6.1. The reason is that the data in Table 6.1 only suggests what would have happened to a party’s LegCo vote share had it not been able to occupy a District Council Constituency. Again, take the DP as an example. If the DP loses a DCC, its vote share in the following LegCo election is expected to decrease by about 10 % points. But Table 6.1 does not show where the 10 % points would go. If 10 % of voters end up voting for the CP or other pan-democratic parties, the DP’s loss is not necessarily the pan-democratic camp’s loss.

The DP example suggests the importance of taking into account voters’ preferences when estimating the instrumental effect of occupying District Council offices. More concretely, legislative candidates sometimes get elected because voters genuinely support their ideology or party (the effect of voter preference). They may also get elected because, as I argue in the previous section, their party has successfully captured the District Councils so that their rival’s source of income 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. But because I am primarily interested in identifying the short-run instrumental effect of the District Councils, I need to distinguish voters’ genuine support from the District Councils’ instrumental effect, in order not to mistake the effect for the cause.

For this reason, showing that pan-democratic parties receive more votes in districts where they control the District Council seats is insufficient to support the claim that winning the District Council seats would improve their 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 candidate from a Beijing-sponsored party is elected, voters may genuinely favor Beijing-sponsored parties over the pan-democrats. As a result, when it comes to the LegCo election, the Beijing-sponsored party 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.

The unique institutional setting of Hong Kong allows us to tackle this estimation problem with a specific research design known as regression discontinuity . The regression discontinuity design (RDD) is a quasi-experimental research design that mimics random assignment of treatment and control groups in a randomized controlled experiment. The details of the application of the regression discontinuity design are discussed in the appendix to this chapter.

The Hong Kong government’s Electoral Affairs Commission publishes data on District Council and 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 were 405 District Council members elected by the plurality rule from 405 districts. As for the 2008 LegCo election, there were only five LegCo geographical constituencies electing 30 LegCo members by the method of proportional representation .¹⁰ The five LegCo districts (LCD) are supersets of the 405 DCCs, and there is no DCC which cuts across the boundary of LCDs. Because the Electoral Affairs Commission of the Hong Kong government publishes electoral results of the LegCo at the DCC level,¹¹ this allows me to measure the effect of a party capturing a District Council seat on that party’s vote share in a LegCo election.

Note as well that District Council elections are held in November, while LegCo elections take place in 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 already has an incumbent District Councillor), which may affect its chances of success in the ensuing legislative election.¹²

The statistical results based on the regression discontinuity design are presented in Table 6.2. The table consists of twelve different regression specifications, which vary by their functional forms and variables of interest. In the first four specifications, I regress the pan-democratic camp’s vote share in LegCo election on the treatment variable D _{Beijing-sponsored}, which is assigned the value of “1” if District Council Constituencies are controlled by Beijing-sponsored parties and “0” otherwise. Interestingly, the variable of interest is statistically insignificant in three of the four specifications, suggesting that Beijing-sponsored parties cannot reduce, at least in the short run, the electoral support of the pan-democrats by their occupation of District Council seats.

Table 6.2 Effect of controlling a District Council Constituency on a rival camp’s vote share in the Legislative Council election: a regression discontinuity design

However, if we look at the next four specifications, in which I regress the pan-democrats’ LegCo vote share on another treatment variable D _{pro-establishment}, we see a powerful effect of the District Councils emerge. Take the cubic polynomial specification as an example. By capturing a District Council Constituency, a pro-establishment District Councillor – who can be anyone from a Beijing-sponsored party, a non-Beijing-sponsored pro-establishment party, or just a pro-Beijing “independent” – is able to reduce the pan-democrats’ overall vote share by 5.378 % points. This effect is not only statistically significant but also substantively important. As mentioned in Chap. 4, the district magnitude of the LegCo geographical constituencies is fairly large. In some LegCo constituencies, a candidate can win a LegCo seat with as low as 6 % of the vote. The effect size of 5.378 % points is therefore too big to ignore by any party or political camp.

The last four specifications reveal a yet more subtle relationship between capturing a District Council Constituency and competing for a LegCo seat. In this set of regressions, I swap the positions of the pan-democrats and Beijing-sponsored parties by regressing Beijing-sponsored parties’ LegCo vote share on D _pan-democrat, which takes the value of “1” if District Council Constituencies are controlled by the pan-democratic camp and “0” otherwise. The coefficients on this variable of interest are statistically significant at 1 % across the four specifications. The negative signs suggest that the powerful effect of the District Councils is not confined to the pro-establishment camp; the pan-democratic camp is also able to lower Beijing-sponsored parties’ support in LegCo elections by capturing District Council Constituencies. The effect size is somewhat smaller, however. For instance, as indicated by the cubic polynomial specification, Beijing-sponsored parties’ LegCo vote share would shrink by 3.657 % points in District Council Constituencies that the pan-democrats control, falling short of what the pro-establishment camp can do to the pan-democrats in reverse.

These twelve specifications in Table 6.2 together paint an interesting picture of the electoral dynamics between the two elected tiers. In particular, the Beijing-sponsored parties’ occupation of a District Council Constituency poses no direct threat to pan-democratic parties’ voter support in LegCo elections in that DCC. But the indirect effect is very much potent. For Beijing-sponsored parties, if they do not crowd out the pan-democrats in the District Councils, their pan-democratic rivals will then be able to make use of the District Council offices to undermine the LegCo support of Beijing-sponsored parties. Another important implication we read from Table 6.2 is that non-Beijing-sponsored, pro-establishment parties matter. From Beijing’s perspective, it should not rely solely on Beijing-sponsored parties to weaken the pan-democrats, because it is the pro-establishment camp as a whole, rather than Beijing-sponsored parties alone, which can deliver Beijing’s desired outcome. And why would the pro-establishment camp achieve a better short-term result than Beijing-sponsored parties alone? My conjecture is that the ideological position of Beijing-sponsored parties is too salient, which prevents their District Councillors from neutralizing moderate voters in LegCo elections. By contrast, non-Beijing-sponsored pro-establishment parties or those candidates who are pro-Beijing and “independent” are less ideologically discernible, so that the District Councillors with this background are able to disguise their political leaning and, thus, take away moderate votes from the pan-democrats.

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. But to ensure that the empirical results are not biased due to the omission of other variables, I rerun the first eight specifications in Table 6.2 by adding control variables pertaining to District Council Constituencies, including, but not limited to, population, voter turnout in the last District Council election, the pan-democrats’ vote share in the last LegCo election, and the population share of males, college graduates, and the elderly. Table 6.3 shows that the statistical results of Table 6.2 carry over to these regression specifications, despite the inclusion of a variety of control variables. In particular, the pro-establishment camp as a whole fares better than Beijing-sponsored parties alone in the use of District Council offices to undermine the pan-democratic camp.

Table 6.3 Effect of controlling a District Council on the rival camp’s vote shares of Legislative Council election: adding controls

To Beijing-sponsored parties, one important function of weaving a big network of District Councillors is to facilitate vote coordination in LegCo elections. Effective vote coordination would reduce the amount of “waste votes” generated under the proportional representation rule. With their dominant position in the District Councils, do Beijing-sponsored parties perform better than the pan-democrats in vote coordination? We can find out the answer by examining how efficiently each political camp translates its votes into seats. One way to measure this is to divide its seat share by its vote share in a given election. If the number is smaller than one, this implies that achieving a given level of seat share requires a higher level of vote share. In other words, some votes would be “wasted.”

Figure 6.1 shows the respective seat share to vote share ratio by political camp. The pan-democratic camp achieved its most efficient allocation in 1998. Since then, its seat-to-vote ratio has never surpassed 1.09. Its ratio even drops below 1 in 2004 and 2012. In contrast, the performance of the pro-establishment camp has improved over time. Its seat-to-vote ratio has never dropped below its 1998 level. As already discussed, within the pro-establishment camp, Beijing-sponsored parties are the most successful in terms of grassroots penetration. Not surprisingly, their efficiency in vote allocation is also the most impressive. As may be seen from Fig. 6.1, their seat-to-vote ratio has been consistently greater than 1 since 1998. Even in 1998, the figure was very close to 1. Remarkably, in 2012, its ratio reached 1.26. The result reflects its success in coordinating votes among the District Council Constituencies it controls.

Fig. 6.1

Seat-to-vote ratio by political camp and by LegCo election (Source: Author’s calculation based on election data from the HKSAR Electoral Affairs Commission)

6.3 Conclusion

In Chinese, the term for political brokers is zhuangjiao. Literally, it means piles. The term is a vivid characterization of the undertaking of voter mobilization; it is analogous to the construction of a house, which entails driving piles into the ground to support the vertical structure. Beijing-sponsored parties have attempted to erect a large political structure in postcolonial Hong Kong’s legislature, in hopes of marginalizing the popular opposition force. To lay the groundwork for their edifice, they need to plant a lot of “piles” into the ground. This is the origin of their District Council strategy; they aggressively expand into this lowest elected tier, with a view to have their District Councillors act as political brokers in order to strengthen the parties’ position in LegCo elections and, thus, undermine their pan-democratic rivals’ electoral support from the ground up.

The empirical results presented in this chapter show that the District Council strategy of Beijing-sponsored parties has paid off handsomely. By occupying more District Council seats, the pro-establishment camp can erode the pan-democrats’ grassroots support, which in turn undermines the latter’s performance in LegCo elections. Beijing-sponsored parties also benefit from their dominant position in the District Councils by preventing the pan-democrats, especially the moderate prodemocracy parties who have an elaborate grassroots support base, from playing out exactly the same bottom-up strategy. In addition, controlling the District Councils allows Beijing-sponsored parties to improve their vote-coordination strategies during LegCo elections. An efficient allocation of votes would benefit the entire political camp as a whole.

It is worth noting that the empirical analysis of this chapter covers only the short-term effect of controlling the District Councils. Its long-term effect is no less important. For one thing, failing to control at least some District Council seats is likely to hinder a party’s long-term development because the party will be deprived of an effective mechanism to groom and cherry-pick new candidates for higher elected offices. Given the Beijing-sponsored parties’ inexorable rise in the District Councils, the harmful long-term effect on opposition parties will gradually emerge in the coming elections.

An important policy implication for pan-democratic parties is that they should devise an appropriate District Council strategy in response to Beijing-sponsored parties’ encroachment. Failing to do so would hurt not only the opposition camp as a whole but also individual opposition parties. As clearly shown in Table 6.1, controlling a District Council Constituency can increase an opposition party’s LegCo vote share by as many as 30 % points. The cost of abandoning the District Councils is simply too high for any opposition party serious about getting a LegCo seat.

Appendix: Details on the Regression Discontinuity Design

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 would be to randomly assign District Council seats to political parties. For instance, some pro-establishment 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 Beijing-sponsored 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 Beijing-sponsored parties and those obtained by the pan-democrats.

In reality, I cannot affect the data-generating process of the District Council elections, but if I can apply regression discontinuity, a quasi-experimental research design, I would be able to find out the causal effect. The idea of regression discontinuity is simple.¹³ 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 the Beijing-sponsored parties’ District Council vote shares, for example, should be best characterized by a function discontinuous at a certain threshold (such as 50 % 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 pro-establishment’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 or not they happen to receive the “treatment” by chance; that is, whether the Beijing-sponsored party obtains, due to random uncontrollable factors, barely sufficient votes to carry the districts. 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 that is identified is 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 the study by Lee (2008).

To estimate the electoral effect of the District Council office, I examine the relationship between the pan-democratic camp’s LegCo vote share in each District Council Constituency (the dependent variable) and a Beijing-sponsored party’s (or a pro-establishment District Councillor’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:

$$\displaystyle{D_{i} = \left \{\begin{array}{l l} 1&\quad \mbox{ if $x_{i} > 0$} \\ 0&\quad \mbox{ if $x_{i} \leq 0$}\\ \end{array} \right.}$$

where D _i is the treatment status of District Council Constituency i, with the value “1” denoting the constituency controlled by a Beijing-sponsored party (or a pro-establishment District Councillor) and “0” otherwise, while x _i is the pan-democratic camp’s margin of victory in the District Council Constituency i.

This leads to the main regression specification:

$$\displaystyle{ y_{i} = f(x_{i}) +\delta D_{i} +\mu _{i} }$$

(6.1)

where y _i is the pan-democratic camp’s LegCo vote share in constituency i, f(x _i ) is a polynomial function, δ is the causal effect of interest, and μ _i 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 y _i and x _i , which is not necessarily linear, in order to avoid mistaking nonlinearity for discontinuity.¹⁴