Four ways to avoid centripetal effects. How political actors escape institutional incentives in divided societies

ABSTRACT

For democracy in ethnically divided societies, political moderation is crucial. The centripetalist school recommends that countries should introduce institutions which offer parties and candidates incentives to rally for votes across ethnic lines. This article discusses the conditions of the centripetal effect of institutions, and actors’ strategies to dismantle this effect. Political elites might try to escape the pressure to seek inter-ethnic votes by building strategic electoral alliances that circumvent the centripetal effect, by engineering the composition of voters in electoral districts, engineering group identities, or mobilizing new voters. Empirically, this article analyses mayoral elections under the two-round majority vote, in a context where they display quasi-centripetal features: five South-Eastern European towns with ethnically heterogeneous populations, split evenly between two groups. Results show that political elites only exceptionally resort to centripetal strategies as expected by theory. Instead, alternative strategies, circumventing the centripetal effect, are predominant.

KEYWORDS:

• Ethnic politics
• institutional engineering
• local politics
• South-East Europe

1. Introduction

The question of whether political institutions can provide for political moderation and conciliation in divided societies remains salient. Multiparty elections can fuel mobilization along ethnic lines, leading to radicalization, political violence and ethnic conflict. Concerned with this spiral of radicalization, a lively debate has been sparked as to whether ethnic polarization can be avoided through institutions with a centripetal momentum.¹ Relying on a special version of spatial models of politics, tailored to multiethnic societies, centripetal institutions incentivize political elites to widen their basis of supporters across ethnic lines. The centripetal model is believed to bring to power candidates and parties with a broad multi-ethnic appeal, and with moderate platforms on ethnic issues. However, the analysis of the effect of centripetal institutions on strategic behaviour has so far remained limited to assessing the type of reaction along the lines of the main centripetal hypothesis and focusing on a single dimension of politics (the literature is reviewed in section 2.1).

This article investigates the political strategies that political parties and candidates opt for in elections under centripetal electoral systems. It identifies alternative reactions to centripetal institutions. Some of these lead to the desired centripetal effect and to the election of candidates with a moderate programme, but the article also identifies four alternative types of strategies, which allow candidates to be elected without a centripetal effect. As a result, the article intersects alternative models of ethnic politics. This includes, inter alia, secondary dimensions of politics, or ethnic engineering, with the model of centripetal institutions.

The theoretical argument is based on a formalized model of the centripetal effect, and its scope conditions: the article shows that some of the scope conditions can be subject to political strategies. Empirically, this article investigates mayoral elections in five ethnically divided towns of South-East Europe, which are all composed of two roughly equally sized ethnic groups. In similar contests – all countries have experienced ethnic tensions and violence in their recent history of democratization –, (two-round) majoritarian electoral rules for the mayoral elections offer incentives to political competitors to moderate, largely in line with the centripetal model.²

The cases presented in this article shed light on a surprisingly large variety of strategies played in the field. In some cases, cross-ethnic vote pooling for moderate candidates, i.e. the expected “centripetal effect” was identified. In most other municipalities, however, alternative strategies are used. Political elites build alternative strategic alliances in order to escape the need for moderation. They also attempt to change the composition of the voters in electoral districts, mobilize voters of their own group, or they try to re-shape the political saliency of ethnic markers, i.e. they engage in the re-construction of (salient) ethnic identities. Thus, this article offers new insights into the effect of institutions in divided societies,³ electoral coalitions,⁴ and on the endogeneity of the political alignment of cultural identities.⁵

The remainder of this article proceeds as follows: the next section discusses the literature on centripetal institutions. The third section discusses the underlying assumptions of the centripetal effect, and from this derives possible political strategies. Section four introduces the towns, and in section five the elections and electoral strategies used there are described.

2. Centripetalist effects in divided societies

2.1. Consociational and centripetal institutions

The effect of political institutions in heterogeneous societies has received major interest in the literature interested in political cooperation and democracy: Are ethnically heterogeneous societies governed best through all-party coalitions, where politics and political parties are segmented along identity lines, and negotiate over compromises, in systems addressed as “Consociational Democracies”?⁶ Critiques of the consociational model highlight its propensity to segment the electoral competition, the risk of radicalization,⁷ and political blockages.⁸ Instead, they advocate a model that would incentivize parties to seek votes across ethnic boundaries.⁹

Scholars use spatial models of voting with one or several dimensions of politics in order to understand election behaviour, or the formation of coalitions.¹⁰ Research interested in institutional engineering for divided societies relies on single-dimensional spatial models in order to understand party positioning and voter behaviour. They investigate centrifugal and centripetal dynamics on the ethnopolitical issue axis, the dominant political dimension in divided societies.¹¹

“Because ethnicity is a largely ascriptive affiliation”, the Downsian model of party politics not applicable: “The boundaries of party support stop at boundaries of ethnic groups. […] The near-impossibility of party competition for clientele across ethnic lines means an absence of countervailing electoral incentives encouraging party moderation on ethnic issues”.¹² In brief, elections by first-past-the-post will not lead to moderation, but to polarization and permanent exclusion.

In order to overcome this dilemma, the centripetal school advocates an alternative model that is aimed at reducing polarization in divided societies by making “politicians reciprocally dependent on the votes of groups other than their own”.¹³ They aim to facilitate cross-ethnic voting through ethnically heterogeneous districts and ballots that allows voters to express multiple candidate or party preferences, facilitating the expression of second choices across ethnic lines. Such a system, the Alternative Vote (AV) with mixed-ethnic districts, was practiced for short time periods in Fiji and Papua New Guinea.¹⁴ In light of the relatively rare applications of the Alternative Vote in multiethnic societies,¹⁵ primarily limited to the Asian-Pacific region,¹⁶ there is little empirical evidence about actors’ behaviour under the AV. Fraenkel and Grofman that the assumptions of the centripetal model might not always apply.¹⁷

The study of quasi-centripetal electoral systems, with similar or equal features, might provide new insights. Majoritarian elections with two-candidate runoffs are a frequent occurring sibling of the AV: both require an absolute majority of votes for a candidate to be elected in the first round of counting, and in both, voters have the option to express a second preference.¹⁸ The “mechanical” seat allocation principle is almost equivalent: candidates with few votes drop out from the race, and their votes are re-allocated to preferences further down the ballot (AV), or as voters in two-round elections shift their preferences (runoff).

3. Actors’ strategies under centripetal institutions

In societies where exceptionally stable ties between voters and parties are formed along ethnic lines¹⁹, inter-ethnic voting is hard to achieve. In the centripetal model, inter-ethnic vote pooling for moderate candidates depends on a number of conditions. This section develops a formal model, explaining the centripetal effect, closely following the centripetal model of electoral engineering.²⁰ Apart from some more common assumptions about rational and informed voters with exogenous preferences, there are three particular scope conditions of the centripetal model. First, electoral districts are heterogeneous with regards to group identities. Second, it is easier to gain these votes across ethnic lines than winning additional co-ethnic votes. And third, a moderate position on ethnic issues allows candidates to pool votes across ethnic lines. However, all three conditions are subject to strategic actors’ behaviour, and hence not purely structural conditions.

3.1. Formal discussion and conditions

The centripetal effect materializes under a number of conditions, related to the identity distribution of the electoral districts, electoral thresholds, equal suffrage across ethnic groups, and the structure of voter preferences.

In an electoral district with two ethnic groups, the model distinguishes between four groups of voters and candidates, Radicals R and Moderates M of groups A and B, ordered along the ethnopolitical axis R_A, M_A, M_B, R_B, and with single-peaked preferences along this axis.²¹

Under the Alternative Vote, assuming the previously defined preference orders, moderate candidates inevitably win the election

1. if none of the two radical blocs holds an absolute majority of the vote, i.e. the overall share of voters of type R_A < 0.5 and R_B < 0.5 (under any different voting system, with the effective threshold t: R_A < t and R_B < t);

2. if at least one of the moderate candidates will enter the last round of counting, which is the case if M_A + M_B > R_A or M_A + M_B > R_B;

3. moderate voters have what Fraenkel and Grofman describe as “mild” preferences, i.e. they prefer moderate candidates rather than co-ethnics.²²

The three conditions may be fulfilled in electoral districts, where none of the two groups (A and B) dominates, and where the degree of radicalization (r_A, r_B), as defined by the three conditions (a) –(c), is not excessive.²³

3.2. Strategies derived

In this institutional context, candidates and parties can chose between different strategies. A centrist position on the ethnopolitical conflict is the most straightforward, but not the sole strategy that might lead to an election.²⁴ I elaborate on further options.

3.2.1. Political alliances and positioning – vote-pooling beyond one’s own constituency

A first set of strategies considers that voter preferences may be endogenous,²⁵ and electoral campaigns are aimed at altering the voters’ choice. Research has shown that the space for this is more limited in ethnically divided societies. Ethnic cleavages are much more stable than other political divides (but see “Ethnic engineering”). As a result, voter preferences within an ethnic segment represent a realistic (and preferred) target for campaign strategies. Strategies aiming for cross-ethnic voting instead target voters’ secondary preferences.²⁶

Radicalization: this first strategy aims at altering voter preferences on the ethnopolitical dimension of electoral competition, which are considered to be stable in the centripetal model. Polarizing campaigns on ethno-nationalist issues might raise the salience of this dimension and radicalize the voter first preferences.

Further alternative strategies build on alliance strategies between parties, or campaign elements that are oriented at gaining the secondary votes. Each of these strategies might be facilitated if such party alliances are in place at the national level of politics.

Ethnic blocs: Ethnic parties and candidates of one of the groups may agree of all parties of one ethnic group to form an alliance, and to call to their voters to support only the candidates of their own group.²⁷ The formation of ethnic blocs undermines the third assumption of the centripetal model, according to which moderate voters prime moderation over co-ethnicity. If group A forms an ethnic bloc, the chances of M_B to get elected become much lower, to the benefit of candidate R_A.²⁸ Moderates cannot profit from the formation of ethnic blocs.

Second dimension: Instead of positioning on ethnic issues, candidates or parties can emphasize other dimensions, such as economic (left-right) conflict.²⁹ As Fraenkel and Grofman demonstrate, strong secondary issues may interfere with voter ranking of political parties or candidates on the ethnonationalist issue dimension. As a result, the election outcome may deviate from the expectations of the centripetal model.³⁰ A secondary issue may motivate some to vote for a moderate candidate across ethnic lines. This would reinforce or coincide with the centripetal effect. To distinguish the effect of such a secondary dimension from the centripetal effect, we would need to have direct or indirect information about individual voters’ motives. Other voters, for whom secondary issues trump the ethnic dimension, may even decide to vote cross-ethnically for a radical nationalist. However, the latter scenario is probably less widespread, as it requires voters to vote for the least preferred option on the ethnic dimension.

Alliance of Radicals: In order to deter the centripetal effect, which threatens to erase radicals from political representation, radicals of all groups can forge electoral alliances. Usually, ethnic radicals of all sides will share conservative values and identity politics, and they can point to liberal and civic values as their common enemy. If successful, this alters the single-peaked preference order along the ethnopolitical axis.³¹ This will lead to the election of a radical in all situations where A·r_A + B·r_B > t.³²

3.2.2. Ethnic engineering

The centripetal model is based on the assumption that ethnic identities and ethnopolitical cleavages are fixed, and voters cast their first-preference vote for a co-ethnic. However, the assumption of the existence of exogenous identities and voting preferences, which define ethnic cleavages, is contested. Political competitors may have an active role in shaping conflicts and political alliances among ethnic groups.³³ Some of these strategies may have consequences for the centripetal model.

Changing the electorate: New citizens can be gained through migration, de jure or de facto enfranchisement, the latter addressing measures that facilitate the access of groups of voters to the polls who previously had not been able to exercise their voting rights. In the context of countries with out-migration, an expansion of the citizenry happens through the mobilization of voters from the diaspora.³⁴ Parts of the opposite group can be de-facto disenfranchized, i.e. decreasing B.

Altering the political salience of cultural identities: In a constructivist model of ethnic identities, elite politics can affect which cultural identities become politically salient, and thereby shift the balance of power.³⁵ Thus, political elites can aim to incorporate citizens who previously identified with the minority into the ruling group, so that it becomes numerically dominant. Thus, they substitute group A with a newly delimited group Ã, so that Ã > A, while $\tilde{B}=B-(\tilde{A}-A)$, i.e. $\tilde{B}<B$.

4. Case selection

This analysis covers mayoral elections in five towns of South-East Europe in the 2000s and 2010s (in Romania, Serbia, Macedonia). The regional focus allows me to focus on countries with a recent history of transitions towards democracy, ethnopolitical tensions and violence (see next paragraph). Each of the five towns is segmented along ethno-religious lines. I selected the five towns in the region that are almost evenly split between two ethnic groups along a salient ethnic divide (according to census data), so that no group is the demographic majority in any of the towns.³⁶ However, mayoral elections occur by majority rule with a runoff, so that candidates are elected if they gain a majority of votes. Seeking votes across ethnic boundaries seems an almost inevitable condition for election success. The Serbian towns have in the course of the analysed period moved to indirect elections in the assembly, with very similar incentive structures. Different from direct elections, parties might have better control over the elections in indirect elections in the assembly.

The empirical study relies on multiple sources of evidence, including documentary evidence (systematic analysis of articles in the press; election statistics, partly at the level of polling stations³⁷; policy reports and secondary literature³⁸), and on interviews, which I conducted for all five cases, with experts and local political elites (see appendix C).

I analyse the mayoral elections over a period of 10 years after 2000 (Macedonian towns: 2005–2013). I include local elections once the countries qualify as democratic, and once violent conflicts (which have affected the three countries to different degrees) have stopped. While a generally democratic environment at the country-level is a condition for the inclusion of cases, I do not exclude that in such a strongly contested context, some of the local actors might resort to non-democratic practices.

4.1. The five towns

The two Macedonian towns, Struga and Kičevo, are divided between Albanians and Macedonians. Kičevo counts small Turk and Roma minorities, whereas the ethno-political boundary lines in Struga are fluid and regularly redrawn, featuring a small rural community with a transient identity: they are sometimes perceived as Macedonian, sometimes as Albanian, and sometimes as an own group, the Torbeši. Both towns used to be ruled by Macedonians, but in the course of the Ohrid peace agreement, which concluded a short violent conflict, the municipal landscape was changed: The merger of the Macedonian-dominated towns with Albanian-dominated rural villages lead to a roughly equal population share of both groups, and was intended to increase the political leverage of the ethnic Albanians.

Bujanovac is one of the three Serbian municipalities with a considerable Albanian population and was the centre of a violent uprising in 2000–2001.³⁹ Prijepolje is a Serbian-Bosniak split town in Southern Serbia. The town was in the midst of the hotspots of the violent conflicts of the 1990s, located in the Muslim-dominated Sandžak region between Bosnia and Kosovo. However, in Prijepolje, tensions have remained low.⁴⁰

The Transylvanian town Târgu Mureş was the centre of ethnic riots between Romanians and ethnic Hungarians in 1990, but since then the situation has calmed. The town is evenly split between the two groups. The 1992 census counted 46.1% ethnic Romanians and 51.3% ethnic Hungarians with most remaining inhabitants declared as Roma.⁴¹ By 2000, the figures were inversed, and ethnic Romanians were a narrow majority of 50.3%.

4.2. The institutions

In all the five towns investigated here, mayor elections are conducted by the two-round majority vote with runoff rule. While this electoral system is slightly different from the Alternative Vote (AV), on which the centripetal model of electoral engineering in divided societies builds, it is very similar: 50% election threshold, two voter preferences, and two rounds of counting (AV: several preferences, several rounds).⁴² While the two-round majority vote has some properties, such as the fragmentation of the vote, which might potentially undermine the centripetal effect,⁴³ these critiques would equally apply to the AV. Appendix C scrutinizes the differences between the two electoral systems, theoretically and empirically. The analysis suggests that for all the elections under study, and the present vote distributions, there is no reason to expect differences in the effect of the AV and two-round majority voting systems.

In all analysed towns, elections to the local assemblies and national parliamentary elections are conducted under proportional rules, which on the one hand provide for similar multiparty systems in all cases, and on the other hand allow the comparison of the outcomes of the mayoral elections to the party landscape in PR elections, and possible radicalization effects (appendix B2).

5. Strategies and results

This section discusses the different strategies which the political elites chose in the mayoral elections under investigation.

First, I assess the political scope conditions for the centripetal effect, or the existence of a sufficiently large pool of voters with moderate ethnonational preferences in order to allow for a moderate candidate to enter the second round of the vote count. A successful radicalization strategy, i.e. an election campaign that polarizes the contest, or a secondary dimension which leads to cross-ethnic votes for an ethnic radical, could undermine this scope condition. A comprehensive investigation of campaign effects on the voter first preferences would be beyond the scope of this article. However, aggregate election results show whether such a strategy might have been successful enough to reduce the support for moderate parties or candidates to such a degree that the centripetal strategy is no longer viable.

Table 1 identifies, for all elections, the distribution of votes between moderates and radicals, relying on the results of the simultaneously held elections to the municipal council by proportional rules. For this and the subsequent analyses, parties and candidates are distinguished between ethnic moderates (M) and an ethnic radicals (R), relying on comparative data on the party’s ethnopolitical positioning,⁴⁴ but also on additional information about their campaign and positioning (mentioned in the analysis in this section), where available. None of the municipalities is polarized to the extent that the scope conditions of the model would be violated: in none of the elections analysed are radical nationalists strong enough to dominate the first round of the elections. Also, in all five municipalities, the moderates win (much) more than one third of the votes. This is also reflected by the fact that in every single contest for the mayor office, at the second round was fought between a moderate (M) and a radical (R), or two moderates.

Table 1. Max. vote shares of Radicals (for the group, where Radicals are stronger), and joint vote share of the Moderates of the two groups, in elections to the Local Council.

Municipality	Year	max. Radical vote %	Moderate vote % (both groups)
Struga	2005	45.2	43.7
(Macedonia)	2009	35.5	39.7
	2013	32.2	66.7
Kičevo	2005	41.3	55.5
(Macedonia)	2009	40.6	47.3
	2013	45.8	51.7
Bujanovac	2002	17.9	62.4
(Serbia)	2006^a	29.3	48.8
	2008	24.5	56.2
	2010	24.5	56.2
Prijepolje	2004^a	38.6	42.1
(Serbia)	2009	29.3	64.6
Târgu Mureș	2000	19.9	66.3
(Romania)	2004	19.4	65.7
	2008	1.5	82.9

^aLocal Council seat %.

Hence, while I cannot exclude that polarization strategies were part of the parties’ campaigns, in none of the elections were these parties consequential enough to prevent moderate candidates from advancing to the second round of voting. This first scope condition for the centripetal model, related to the distribution of first vote preferences, holds for all elections under scrutiny.

The further steps of the analysis rely on information on either informal or official alliances, cross-ethnic and centripetal voting, and on efforts to mobilize the diaspora. With no voter surveys available, I use indirect evidence on cross-ethnic and centripetal voting. It is either based on interviews, newspaper articles and aggregate election results (which give some basic information about possible scenarios), and – for the Macedonian towns – on aggregate results at the polling station level, which allow me to conduct ecological analyses (see appendix A). Where relevant, the qualitative discussion refers explicitly to statistical inferences and sources. Table 2 offers a summary overview of the cases and their main characteristics.

Table 2. Mayoral elections in five evenly split towns, overview of variables.

Municipality	Year	Candidates qualifying for runoff (Groups A/B; R = Radical; M = Moderate)	Election type	Voting patterns		Electoral alliances		Ethnic engineering		Elected
				centripetal effect	inter-ethnic vote	of Radicals	ethnic bloc	changing the electorate	pol. salience of cult. identities
Struga	2005	RA MA^a	direct	no	no	no	no	(yes)	no	RA
(Macedonia)	2009	RA M $\tilde{B}$	direct	yes	yes (MBC)	yes (secret deal)	no	no	yes	RA
	2013	MA RB	direct	yes	yes (Torbeši)	no	yes (both sides)	yes (MA)	no	MA
Kičevo	2005	RA MA	direct	no	yes	no	no	no	no	RA
(Macedonia)	2009	MA elected in 1st round	direct	no	yes (Radicals)	yes	no	no	no	RA
	2013	MA elected in 1st round	direct	no	no	no	yes (both sides)	yes (MA)	no	MA
Bujanovac	2002	MA MB	direct	N.A.	N.A.	no	yes (Albanians)	no	no	MA
(Serbia)	2006	MA RB	direct	no	no	no	yes (Albanians)	no	no	MA
	2008		assembly	no	no	no	yes (Albanians)	no	no	MA
	2010		assembly	yes	yes	no	yes (Albanians)	no	no	MA
Prijepolje	2004	MB RB	direct	yes	yes	no	no	no	no	MB
(Serbia)	2009		assembly	no	yes	no	yes (Serbs)	no	no	RA
Târgu Mureș	2000	MA RB	direct	no	no	no	yes (Hungarians)	yes (RB)	(yes^b)	MA
(Romania)	2004	MA elected in first round	direct	no	no	no	yes (Hungarians)	yes (RB)	no	MA
	2008	MA elected in first round	direct	no	no	no	yes (Hungarians)	no	no	MA

^aIn the Struga elections of 2005, the candidate MA withdrew from the runoff, due to fears of violence.

^bMA targeting Roma votes.

5.1. Alliances and positioning

In the further parts of the empirical analysis, I summarize cases based on the strategies that candidates employed. This subsection discusses strategies related to the positioning of candidates and alliances.

5.1.1. Centripetal strategy

The centripetal effect might come into play either in the first or in the second round of mayoral elections. There is sparse evidence for of centripetal voting in the first round. Instead, in most of the first-round elections, several candidates for each group compete against each other, with little reasons for cross-ethnic voting, and there are few indications for strategic effects (see appendix B). Important exceptions include cases where the elites strategically support a moderate candidate who might win inter-ethnic votes with a centripetal effect in the second round, occurred in the Macedonian town of Struga (in 2009 and 2013). In Kičevo, cross-ethnic votes were cast for a radical candidate. Further, the Romanian voters in Târgu Mureș voted in some elections rather cohesively in favour of a moderate, but while there is no evidence that this was the result of a centripetal voting strategy, the further analysis highlights alternative explanations.

In the second round, a centripetal effect might either occur in the case of a runoff of moderates versus radicals from different groups, or between moderates and radicals from the same group. Among the cases analysed, elections of the latter type more often lead to centripetal voting.

A pattern of moderates versus radicals from different groups occurs in four municipalities, namely in Târgu Mureș (2000), in Bujanovac (2002, 2006), Kičevo (2013) and in Struga (2009, 2013). With Serbs and Albanians deeply split in Bujanovac, and no attempts to gain votes cross-ethnically, we can rule out a centripetal effect. In the 2000 elections in Tărgu Mureș, there is no apparent evidence for a centripetal effect involving cross-ethnic voting (see section ethnic blocs).⁴⁵ Inter-ethnic vote pooling was involved in the 2009 and 2013 elections in Struga however, and it is documented by the ecological analysis of the polling station results (appendix A). For Kičevo (2013), the same analysis does not show any inter-ethnic voting patterns.

Hence, in five out of six cases, the more moderate candidate was elected. However, the elections in Struga remain the only case with evidence for inter-ethnic votes. In four other elections, it appears that voters were strictly aligned behind their ethnic candidates (see subsection on ethnic blocs), with no centripetal effect at work.

A second type of constellation is present when two candidates, a moderate and a radical, from the same group enter the runoff. This occurred three times; in Struga (2005) between two Albanian candidates, in Kičevo (2005) between two Macedonians, and in Prijepolje (2004) between two Bosniaks. However, a centripetal effect only occurred once, in Prijepolje in 2004. Despite separate ethnic parties, polarization is considered to be low in Prijepolje. In the first round of the 2004 mayoral elections, the Serbian vote was fractionalized among several candidates in the first round, so that unexpectedly, two Bosniak candidates entered the runoff, on behalf of the radical Bosniak SDA and the more moderate SDP, which has a record of cooperation with Serbian parties in national politics. None of the Serbian parties issued an official recommendation for either of the Bosniak candidates, and turnout dropped considerably from 47.9% to 31.6%, most likely due to de-mobilization of Serbian voters. Still, the moderate SDP candidate, Nenad Turković, who entered the second round as the second-ranked candidate, won with a massive lead over his more radical competitor. Apparently, he benefitted from votes across ethnic lines. This corresponds to the expectations about centripetal effects. Local interlocutors also point to his local popularity as a school director. After the elections, he formed a centripetal coalition in the local government, including the “Democratic” Serbian party bloc.

In Kičevo and in Struga (both 2005), the centripetal effect did not materialize. In Struga, the moderate Albanian party DPA withdrew its candidate in the second round, under the accusation that its opponent, the radical Albanian DUI party, was buying votes and perpetrating violence.⁴⁶ In Kičevo, an influential local economic leader, Vladimir Toleski, entered the runoff on behalf of one of the two Macedonian nationalist parties (VMRO-NP), and won narrowly (50.9%). He was supported by the other Macedonian nationalist party, VMRO-DPMNE in the runoff, against the Social Democratic candidate.⁴⁷

Thus, despite a number of occurrences of a configuration of candidates prone to a centripetal effect, only in two such cases – Prijepolje in 2004 and Struga in 2013 – a moderate candidate is elected with cross-ethnic votes.

5.1.2. Second dimension

Secondary dimensions matter for the present analysis if they explain patterns of cross-ethnic voting.⁴⁸ While there is no systematic data on local party positions on secondary dimensions, appendix D displays two dimensions for all countries under study based on expert data (at the national level).⁴⁹ I complement this with information obtained from fieldwork for those elections where cross-ethnic voting mattered. In two cases where some centripetal moments occurred in Serbia (Prijepolje 2004, Bujanovac 2010), there were some programmatic affinities between the elected candidate and his (cross-ethnic) voters along the democracy-authoritarian dimension. However, in the Serbian structure of cleavages, a democratic orientation also implies ethnopolitically moderate positions, so that both explanations coincide.

Cross-ethnic voting patterns were also observed in other cases. In Prijepolje (2009) and in Kicevo (2009), cross-ethnic votes were cast for ethnopolitical Radicals. In the case of Prijepolje, secondary dimensions may, at best offer a very weak explanation for the unusual coalition. In Kičevo, the successful Macedonian nationalist candidate, Toleski, campaigned on good governance and economic issues, and had TV campaign adverts in Albanian, but local interlocutors primarily point to an informal alliance of the nationalist elites in order to explain cross-ethnic voting (see next subsection).

Also, the discussion of the respective cases highlights the importance of inter-ethnic national governing coalitions facilitating the creation of local coalitions (Kičevo, Struga, Prijepolje).

5.1.3. Alliance of Radicals

A special case of a secondary dimension are ties between radical parties across ethnic lines. Radicals might also agree in such alliances, oriented at exchanging second-preference votes, strategically, under threat of losing political ground under centripetal institutions

This scenario applies to the 2009 elections in Kičevo and Prijepolje, and I found evidence, which might point to such an agreement in 2009 in Struga. In the 2009 elections in Kičevo, there were three electorally significant parties, on the Macedonian side the moderate Social Democrats, and the nationalist VMRO-DPMNE, and on the Albanian side the radical DUI. The Albanian DUI did not present an own candidate, probably based on an election agreement between radical Albanians and Macedonians, which was not made public.⁵⁰ The analysis of polling station results (appendix A) shows that many DUI voters abstained in the mayoral elections, and a minority voted for the Macedonian radical candidate (see appendix A), securing his election in the first round. Local interlocutors and foreign experts described elections in Kičevo as primarily clientelist.

In indirect elections, alliances of Radicals were more frequent. Party elites might find it easier to secure voter discipline for a pragmatic deal than in direct elections. After the 2009 local elections in Prijepolje, the first elections with an indirect mayor election, a coalition was formed between the largest party, the moderate Bosniak SDP, with two Serbian parties/alliances: the Serbian nationalist SNS and a moderate Serbian alliance around DS. This allowed the election of a Serbian radical candidate, instead of the moderate Bosniak incumbent. This alliance was brokered by the national party offices in Belgrade, and some of the leaders of the involved parties reported to me that they would have preferred a different alliance.⁵¹

The alliances follow the consociational model, where in the election period, the parties address their own segment, and in the post-election period, they engage in elite accommodation, and split the offices among themselves.⁵² This is best expressed by the mayor of Prijepolje, Dragoljub Zindović, who governs in a coalition of radicals:

“Those parties which profit from most voter support shall form the government.”⁵³

5.1.4. Ethnic blocs

Closing ranks behind a single candidate can allow an ethnic group to secure the election of this candidate. Such alliances, also addressed as “seat-sharing arrangements”, can be agreed for the first round of elections, with a joint candidate for the parties of one group (Bujanovac, all years, Târgu Mureș 2004, Struga 2013 [Macedonians]), or for the second round (Târgu Mureș 2000, Struga 2013 [Albanians]). Ethnic blocs confined to a single group will lead to a victory of this group; if the other group follows suit (e.g. Struga 2013), ethnic mobilization will be decisive for securing an election victory.

In the Târgu Mureș, the ethnic Hungarians had been united behind the Hungarian Democratic Alliance UDMR as the only Hungarian organization throughout the 1990s, and consistently won the mayoral office. In 2000, the Romanian parties closed their ranks in the second round of the mayoral elections, after intense quarrels about the candidatures. The re-nomination of the incumbent Hungarian mayor, Imre Fodor (UDMR) was challenged both internally and from the ethnic Romanian side. His internal rival was the ethnic Hungarian MP László Borbély, perceived as the more vital and conciliatory candidate of the two, who would potential reach out to Romanians.⁵⁴ Eventually, the incumbent Fodor ran again, on behalf of the Hungarian bloc. He was perceived as an ethnic representative, with not very good conduct of Romanian language, and affiliations to the more radical wing of the party, and therefore unpopular with ethnic Romanians.⁵⁵ Seconding this within-Hungarian conflict, the popular incumbent district prefect, the Romanian Dorin Florea, announced that he would run as a mayoral candidate. He avoided positioning himself on ethnopolitical issues. This has led to a constellation of a conciliatory Romanian versus a polarizing Hungarian candidate. Despite this constellation that seems lending to a centripetal effect, I could not find any evidence for inter-ethnic voting from press reports, expert interviews or election results (polling station election results are not available). Instead, the pattern speaks of high mobilization efforts of both sides. While the Hungarian electorate was united behind Fodor in both rounds of the election, the moderate Romanian Florea was able to increase his number of votes in the runoff considerably, with turnout rising from 57% to 67%.

Four years later, in the run-op of the 2004 elections, the ethnic blocs and mobilization were reflected in the following call on the front page of one of the Romanian-language local newspapers:

Now you know who you vote for. In Târgu Mureș there are 110.000 citizen with voting rights. There are 55.000 Romanians and 50.000 Hungarians. Usually 38.500 Hungarians vote and the number of Romanians does not exceed 33.000. Vote on 6 June 2004.⁵⁶

The analysis of ethnic blocs shows that these often altered as a consequence of earlier electoral failures. For instance, in the 2009 elections in Struga, the Macedonian parties wanted alternation of office at all cost, and therefore cooperated, plus agreed on a centripetal strategy. After renewed failure, they formed an ethnic bloc in the 2013 elections, with joint candidacies (with a VMRO-DPMNE mayor candidate, and a joint list under the SDSM label for the council elections); a similar electoral pact was agreed on in Kičevo.⁵⁷ The ecological analysis of the election results (appendix A) suggests that voters adhered to this bloc in a disciplined matter.

In the Serbian town of Bujanovac, the Albanian parties have agreed repeatedly on alliances, and the principle of Albanian unity was upheld also throughout their period in municipal government.⁵⁸

5.2. Ethnic engineering

5.2.1. Altering the political salience of cultural identities: the mayoral elections in Struga, 2009

The mayoral elections in Struga of 2009 include multiple strategies analysed in this article, and they are the only instance of a strategy to alter the political alignment of cultural identities.

The municipality of Struga is the product of the merger of a Macedonian-majority town and Albanian-majority rural villages. In 2005, an ethnic Albanian, Ramiz Merko (Democratic Union for Integration, DUI), was elected to the mayoral office. Merko polarized on ethnic issues, so that the main target of both the relevant Macedonian parties in place, the moderate Social Democrats and the nationalist VMRO-DPMNE, and the moderate Albanian parties was to replace him. However, the ethnic Macedonian parties considered it unlikely that any strategy might allow them to win Albanian votes in substantial numbers.

Instead, the Macedonian parties targeted a small, politically and economically widely marginalized local rural community, the Torbeši. Due to their Muslim faith, they used to be politically mobilized by Albanian political parties. This alignment is reinforced by numerous inter-marriages between Torbeši and ethnic Albanians. The most influential member of the Torbeši community, the local entrepreneur Fijat Canoski, is even married to an Albanian. However, the Torbeši share their native language, Macedonian, with the Macedonians.⁵⁹ Based on either of the two markers of identity, religion and language, Macedonians and Albanians consider the Torbeši as co-ethnics.

To remove the Albanian hardliner Merko from the mayoral office, the Macedonian parties needed to count on votes from the Torbeši community, and ideally from moderate Albanians. To achieve the impossible – the political re-alignment of the Torbeši with the Macedonians – the Macedonian parties de facto moved their support to a candidate of the Torbeši community. According to this plan, Fijat Canoski, the local entrepreneur with a Torbeši identity, was to enter the second round of the vote against the Albanian Ramiz Merko. This would allow Macedonians and Torbeši to unite politically. Moreover, Canoski would also be a candidate with the potential to reach out to parts of the Albanian electorate who disapproved of the incumbent mayor (i.e. allow for an inter-ethnic coalition between Macedonians and moderate Albanians).

The strategy went beyond an ordinary political alliance – in particular it implied a re-engineering of ethnopolitical cleavages, causing social tensions within the Torbeši community, which had developed strong social and family ties to the Albanians.⁶⁰ The political strategies are also perceived as a struggle over the Torbeši identity.

In order to allow the Torbeši candidate Canoski to enter the runoff, the Macedonian Social Democrats did not present an own candidate for the mayoral office, whereas the more radical party (VMRO-DPMNE), ran with a weak candidate, who would not be able to gather a large number of votes. Data from the 75 electoral wards allows me to compare the mayoral elections to the local assembly elections on the same day. Canoski by far outnumbered the votes of his party (PEI) in the assembly elections, as he received most moderate Macedonian votes. In the second round, Canoski won most of the Macedonian votes, the Torbeši votes, and a part of the moderate Albanian party’s (DPA) vote (see appendix A). In exchange, Canoski offered support for the VMRO-DPMNE candidate in the presidential elections.⁶¹ Nevertheless, he narrowly failed to be elected. Several interlocutors refer to irregularities in the conduct of the elections, which allowed the incumbent Albanian to stay in power.⁶² The aggregated election data suggest that many VMRO-DPMNE voters voted for Canoski, while in a few strongholds of the same party, votes shifted to the Albanian hardliner candidate in the second round of the elections. These very local vote shifts could indeed be the fruit of a backroom concession of VMRO-DPMNE to DUI, its junior coalition partner in the national government, to allow DUI to keep the mayoral office in Struga.

Four years later, in 2013, the political leaders of the Torbeši community switched sides again and joined an alliance with the Albanian party DUI. However, the Torbeši no longer presented an own mayoral candidate.

5.2.2. Changing the electorate

When ethnic blocs are formed, the electoral competition becomes a game of numbers. Accordingly, jointly with the formation of ethnic blocs, attempts to increase the size of parties’ and candidates’ own ethnic constituency occurred in several cases (Târgu Mureș, 2000 and 2004; Struga 2013 and Kičevo 2013).

The most obvious cases regard the 2013 elections in the Macedonian towns of Struga and Kičevo. Macedonia has a large diaspora, with a disproportional share of ethnic Albanians, many of whom migrated to Switzerland. They can cast their votes only in person in the municipalities where they are registered. In Kičevo, ethnic Albanian organizations organized charter flights from Switzerland on election day, in order to allow the diaspora to vote.⁶³ The goal was expressed by the slogan “we are 55% Albanians, but only including our diaspora”.⁶⁴,⁶⁵ However, the issue is controversial, with numerous competing narratives which cannot be verified. These range from pressure on diaspora members to turn out on election day, up to an attempt by border police to prevent the diaspora members from crossing the border, which in turn allegedly was prevented by ethnic Albanian staff of the mixed-ethnic border guards. Similar accusations surround the 2009 elections, including accusations of the inclusion of fictitious names on voter lists in Kičevo, the omission of some ethnic Macedonians on the voting lists in Struga, and the obstruction of the organization of two charter flights from Switzerland for 350 Albanian voters.⁶⁶,⁶⁷

In the Romanian town of Târgu Mureș, in the elections in 2000 and 2004, the Hungarian community organized bus transport at subsidized rates, or even free of charge, for the local diaspora in Hungary to return home for the elections.⁶⁸ The Romanian parties, vice-versa, involved a Roma leader in the campaign, in order to target Roma voters.⁶⁹

6. Conclusion

This article investigates very similar mayoral elections in five towns in South-East Europe, which are split between two equally sized ethnic groups. In these very similar cases, where neither of the major groups has a clear majority of votes ex-ante, centripetal theory would predict that candidates with a moderate position on ethnic issues are at an advantage, as they can win votes across ethnic lines.⁷⁰ And indeed, I find examples of how this effect plays out, including an electoral manoeuvre in the Macedonian town of Struga to bring a candidate from a small, marginalized community with a mixed identity into the runoff, who can appeal to voters of both groups. Elsewhere, even in very clear-cut centripetal onsets, where an ethnic Moderate from one group enters the runoff against a Radical from the other group (e.g. in the Târgu Mureș election of 2000), there is often no evidence for cross-ethnic voting.

In contexts where these institutionally engineered pressures are significant, we alternative ways of dealing with centripetal incentives are getting more important. This study offers insights into a multi-faceted menu of alternative electoral strategies. Some of these are not unknown in the election literature:⁷¹ political parties address voter preferences and second issue dimensions, engage in clientelist alliances, or radical parties join in alliances. In other cases, parties aim to re-construct the political salience of identities, i.e. ethnic engineering (e.g. Struga), or they form ethnic blocs, and engage in mobilization along ethnic lines. These strategies, if applied to the context of centripetal institutions, may allow political elites to undermine the supposed centripetal effect of electoral institutions.

If political parties close the ethnic ranks, in order to prevent their ethnic communities from voting across ethnic lines, elections become a horserace of ethnic mobilization. Not only party strategies, even election slogans rely on the fragile balance of demographic numbers, and call voters to turn out, in order to secure the office for their own group. Defection from the ethnic bloc is labelled as “betrayal”. Instead of promoting conciliation or running on a political platform, or even with a party label, candidates present themselves as representatives of their groups. Candidates and parties try to shift the ethnic balance in favour of their own ethnic group. This includes activities to bring the diaspora back on election day, either in special buses (Târgu Mureș), or even with charter flights (Kičevo, Struga).

When cooperating across ethnic lines, radical parties occasionally reached out to each other. Often, however, political alliances are clientelist or supported by coalitions in the national government (Kičevo, Prijepolje).

There is a remarkable menu of diverse actor strategies observed across elections and towns, though these occur in a very similar and particular context: all five cases are characterized, among other factors, by a high degree of polarization. Further research should seek to uncover whether the results can be generalized for a larger set of cases, including cases with a lower degree of ethnic polarization, and whether the insights from two-round majority ballots also apply for the Alternative Vote. Certainly, high salience of ethnicity and a heterogeneous population are the context centripetal electoral institutions are usually recommended for, and where they might be potentially be successful. The present shows that the centripetal incentives mostly fail to reach their goals. And it highlights that the centripetal model does not sufficiently consider alternative actor strategies.

Acknowledgments

I am very thankful to Olimpija Hristova, Nikola Savić, and Tibor Toro for their excellent support, especially for the press clippings. I wish to thank my interviewees for patiently providing me with detailed and thrillingly interesting background information and anecdotes. Many thanks to Arjan Schakel, Bernard Grofman, Matthijs Bogaards and conference participants for helpful comments. And I am thankful to the Political Science Institute at the University of Belgrade for inviting me for a research stay which enabled this study. Earlier versions of this paper were presented at the Congress of the Francophone Political Science Associations 2011, at Masaryk University, Brno, 2015, ASN European Conference, 2016, and the ECPR General Conference 2016.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author's work has generously been supported by the Swiss National Science Foundation (NCCR Democracy).

Notes on contributors

Daniel Bochsler

Daniel Bochsler is an Associate Professor at Central European University (CEU), and a Full Professor at the University of Belgrade. His research interests include political institutions in heterogeneous societies, democratization, and elections. He conducts comparative analyses and studies on South-East European countries.

Notes

1 Reilly, “Democracy in Divided Societies,” 7; Horowitz, “A Primer for Decision Makers”; Zuber and Szöcsik, “Ethnic Outbidding”.

2 Sellers, Between Nations.

3 Lijphart, Democracy in Plural Societies; Horowitz, Democracy in Divided Societies.

4 Golder, Pre-Electoral Coalition.

5 Posner, Political Salience; Sambanis and Shayo, Social Identification and Ethnic Conflict.

6 Lijphart, Democracies: Forms; Mitchell, O'Leary and Evans, Northern Irelands.

7 Reilly, Democracy in Divided Societies.

8 McCulloch and Vandeginste, Veto power.

9 Horowitz, “A Primer for Decision Makers,” 122–3; Reilly, “Democracy in Divided Societies”; Reilly, “Electoral Systems for Divided Societies,” 167.

10 E.g. Downs, An Economic Theory of Democracy; Adams et al., Unified Theory.

11 Horowitz, Ethnic Groups, 346.

12 Ibid.

13 Reilly, Democracy in Divided Societies, 11; Horowitz, A Primer for Decision Makers, 122–3.

14 Horowitz, A Primer for Decision Makers, 122–3. Similar: Fraenkel and Grofman, The Borda Count. Other centripetal electoral rules include the Single -Trans­ferable Vote, with a similar logic, but a lower electoral threshold and thus a weaker centripetal incentive, and rules requiring that the winning candidates’ votes are spread across (ethno-)territorial units. Bogaards, Electoral Choices; Bochsler, Non-discriminatory rules.

15 Australia has the longest tradition of the Alternative Vote, but without sufficiently heterogeneous districts. Similar, European elections by the AV were held in mono-ethnic minority districts in Slovenia or in presidential elections a rather homogeneous federal unit in Bosnia and Herzegovina. Bose, Bosnia After Dayton, 234–8; Fraenkel and Grofman, The Borda Count.

16 Reilly, Democracy and Diversity.

17 Fraenkel and Grofman, A neo-Downsian model; Fraenkel and Grofman, Does the Alternative Vote.

18 Sartori, Comparative Constitutional Engineering, 63–5; Reilly, Cross-Ethnic Voting.

19 Birnir, Ethnicity and Electoral Politics.

20 Horowitz, Ethnic Groups, 346; Reilly and Reynolds, Electoral Systems, 32–6.

21 Also: Fraenkel and Grofman, Does the Alternative Vote. The model applies equally for elections with one candidate for each of the four resulting political orientations, R_A, M_A, M_B, R_B, for elections where there are several candidates with the same political orientation (under the assumption that voters prefer candidates with a similar orientation over those with a different political orientation), or for elections, where one or several positions are not represented by any candidate.

22 E.g. pM_A ≻ pM_B ≻ pR_A ≻ pR_B. Fraenkel and Grofman, A neo-Downsian model.

23 Reilly and Reynolds, Electoral Systems, 32–6.

24 Horowitz, A Primer for Decision Makers, 123–4.

25 Steenbergen et al., Who’s Cueing Whom; Hellwig and Kweon, Taking cues.

26 Reilly, Democracy in Divided Societies, 10–1.

27 Such bloc voting has been discussed in the literature, partly also as a coalition strategy, and dependent on group sizes, though previous work does not identify it as a response to centripetal institutions. Barry, Political Accommodation, 399–400; Horowitz, Ethnic Groups, 323. Ishiyama, Explaining ethnic bloc voting in Africa, 765.

28 In particular, this is the case if r_A > 0.5 and p_A>p_B and (1-r_A) p_A + (1-r_B) p_B > r_Bp_B.

29 e.g. Rovny, Communism, Federalism, and Ethnic Minorities.

30 Fraenkel and Grofman, Does the Alternative Vote.

31 pR_A ≻ pR_B ≻ pM_A ≻ pM_B, for voters of R_A.

32 In particular, this will lead to different outcomes in all cases, where p_B·r_B > t - p_A·r_A > 0, and where M_A + M_B > R_A or M_A + M_B > R_B. This is the case in ethnically heterogeneous municipalities, where Radicals jointly have a majority of the votes. These are the cases, which are typically addressed by the centripetal model.

33 Posner, Political Salience; Sambanis and Shayo, Social Identification and Ethnic Conflict.

34 Caramani and Grotz, Beyond citizenship.

35 Posner, Political Salience.

36 The municipalities have between 40,000 and 150,000 inhabitants; the largest population shares of one group are 57% Serbs in Prijepolje and 57% Albanians in Struga. Dozens of evenly split municipalities with populations between 850 and 17,000 inhabitants can be found in Croatia, Romania and Serbia (Čoka), but they were not included due to concerns about data availability. Three Bulgarian towns were not included because there have been no similar levels of ethnic violence since transition.

37 Obtained from the national electoral commissions, except for Bujanovac: OSCE Odihr Office for Democratic Institutions and Human Rights, Early Municipal Elections.; B92, 5 June 2006, “Mustafi Preševo, Draškoviću Medveđa”.

38 Brunnbauer, Partis albanais; Dikici, Torbeshes; International Crisis Group, Southern Serbia; International Crisis Group, Serbia's Sandžak; European Stability Initiative, Kicevo; Taleski, Minorities and Political Parties; Huszka, Presevo Valley.

39 International Crisis Group, Southern Serbia.

40 International Crisis Group, Serbia's Sandžak.

41 Árpad, Maros megye településeinek etnikai.

42 This can lead to differences in voting behaviour, in alliance strategies, and in the mechanical effect of the system.

43 Birch, Two-round Electoral Systems; Adams, Merrill and Grofman, Unified Theory.

44 Information is gained through field research and cross-validated with expert-coded party positions Szöcsik and Zuber, “A New Dataset on Ethnonationalism in Party Competition in 22 European Democracies.”. The positioning is always relative to other parties in the same country, and not comparable across cases.

45 Four and eight years later, the incumbent mayor of Târgu Mureș was elected in the first round. This could potentially also be related to the fact that he gained votes across ethnic lines due to his moderate position. However, in both cases, the Hungarian candidate won roughly the same number of votes in the first round as the Hungarian minority party, so that it is implausible that a significant number of ethnic Hungarians might have voted across ethnic lines.

46 Dnevnik, 24 March 2005, DPA i PDP bokotiraat’; Utrinski, 24 March 2005, “Koalicijata DPA-PDP se povlekuva od vtorog krug”.

47 European Stability Initiative, Kicevo.

48 Secondary dimensions, which affect within-group vote choice only, may affect the distribution of votes between radicals and moderates. However, they do not alter any assumption of the centripetal model, in particular regarding the single-peakedness of preferences along the ethnopolitical dimension, and hence do not affect the present analysis.

49 Ethnonationalism in Party Competition (EPAC), Szöcsik and Zuber, A New Dataset on Ethnonationalism in Party Competition in 22 European Democracies.

50 Interview with Dino Seferi, member of assembly DUI, Kičevo, 4 June 2015. Interviews with experts.

51 Interviews with officials of different parties, Prijepolje, 16–17 January 2011.

52 Typically, the junior partner nominates the assembly speaker or the vice-mayor. Horowitz, Ethnic Groups, 365–88; Horowitz, A Democratic South Africa? 167–77.

53 Interview with Dragoljub Zindović, mayor of Prijepolje, Prijepolje, 17 January 2011.

54 Cosmeanu, Şi noi, belgienii?

55 A few years later, when splits within the Hungarian minority became more manifest, he became the president of the more radical wing, the Szekler National Council.

56 Zi de zi, 15 May 2004.

57 http://www.novamakedonija.com.mk/NewsDetal.asp?vest=31513843563&id=9&setIzdanie=22830. See also Stewart, From Adversaries.

58 This even remained the case when the coalition was extended to include ethnic Serbs in 2010; the coalition was extended to an oversized coalition, including all Albanian parties. This coalition allowed the Albanians to still rely on an absolute majority of Albanian councillors, and not accord the Serbian coalition partners the pivotal vote. “Throwing out one of the Albanian parties from the coalition would be considered as a betrayal of Albanian national interests”. (Interview with Sherif.Abdili, Bujanovac, 23 December 2010. Similar wording by Jonuz Musliu, Bujanovac, 23 December 2010.) There was a previous mixed-ethnic coalition from 2002 to 2006, under mayor Nagip Arifi of the moderate Albanian party (Party for Democratic Action, PDD).

59 Dikici, Torbeshes.

60 Interview with Sefer Canoski, 5 June 2015.

61 A1, 1.4.2009, “Party bargaining before the finish”.

62 A1, 29.3.2009 “Threats for PEI’s leader, Canoski”; Vreme, 30.3.2009, “Merko is tormenting Canoski’s family”. Interviews with local leaders, 2 June 2015, Struga; expert interviews, 2011, Skopje.

63 Utrinski Vesnik, 24.3.2013, “Mass voting in Kichevo and Struga”.

64 Interview with Dino Seferi, member of assembly DUI, Kičevo, 4 June 2015.

65 Uskana.org, a website with extensive material to facilitate the vote registration from abroad, and displaying lists of voters (last accessed on 2 July 2015).

66 A1, 19.3.2009, “DPA: The government is obstructing Zurich flights”.

67 Utrinski Vesnik, 24.3.2013, “Mass voting in Kichevo and Struga”.

68 Népújság, “Jöjjenek haza!”, 1.6.2000.

69 24 de ore muresene, 26.2.2000, “În Tîrgu Mureș rromii vor vota un primar român”.

70 Sartori, Comparative Constitutional Engineering, 63–5; Reilly, Cross-Ethnic Voting.

71 E.g. Adams et al., Unified Theory; Birch, Two-round Electoral Systems.