Democracy in Divided Societies: Electoral Engineering to Reduce Conflict

Australian political scientist Benjamin Reilly reviews the evidence for using Ranked Choice Voting as a means to decrease electoral polarization

[Editor’s note:  Democracy is inherently difficult in societies with deep cleavages. Elections in such societies can encourage extremist appeals, zero-sum political behavior and conflict, and consequently often lead to the breakdown of democracy. In this article based on his important book Democracy in Divided Societies: Electoral Engineering for Conflict Management (Cambridge University Press), Australian political scientist Benjamin Reilly sets out the potential of “electoral engineering” as a mechanism of conflict management in divided societies. He focuses on the experience of divided societies which have used preferential, ranked ballot electoral systems – what are known in the United States as “ranked choice” systems. He shows that such systems can encourage bargaining between rival political actors, and present politicians with incentives to cast a broad net to attract votes, thus aiding the development of democracy even in polarized or divided societies.]

The question of whether, and how, democracy can survive in divided societies has long been a source of controversy in political science. Some of the greatest political thinkers have argued that stable democracy is possible only in relatively homogenous societies. John Stuart Mill, for example, believed that democracy was incompatible with the structure of a multi-ethnic society, as “free institutions are next to impossible in a country made up of different nationalities.”

Because ethnic identities tend to be invested with a great deal of symbolic and emotional meaning, aspiring politicians hungry for electoral success have strong incentives to harness these identities as a political force. Rather than converging on the mythical ‘median voter’ (Downs 1956), divided societies tend to exhibit ‘polarized pluralism’ (Sartori 1976), with competition for votes taking place at the extremes rather than at the centre. The logic of elections changes from one of convergence on policy positions to one of extreme divergence. Politics becomes a centrifugal game. With no median voters, politics can quickly come to be characterised by centrifugal forces, in which the moderate political centre is overwhelmed by extremist forces.

While this model was developed with ethnically-divided societies in mind, it also bears a strong resemblance to contemporary cases of political polarization, not least in the United States, where partisan allegiances have taken on an identity-like status for many Americans (Mason 2018). For instance, opinions polls routinely find Democratic and Republican voters hold sharply negative views of the other side. Questions that have long been used by researchers of race and ethnicity (such as “would you be comfortable with your son or daughter being in a mixed relationship?”), are now registering similar levels of social anxiety for party ID to those once reserved for the idea of interracial marriage (McCoy and Press 2022).

In an age of hard-line partisanship, with fewer “floating voters” between the two parties than in the past, many U.S. elections are less a battle of ideas and more a statement of identity -- like those in divided societies. Electoral campaigns promoting us-versus-them divisions have become an effective way to mobilize voters, and easier to instigate than those based on compromise, deliberation and restraint (Bartels 2020). Instead of broad-based representation, the result is increased partisan rancour, ideological rigidity and legislative polarization (Lublin and Reilly 2023).

Is it possible to design political systems which instead of rewarding such divisions promote accommodation, moderation and centripetal, centre-based politics? One promising approach is to strategically design institutions such as the electoral system. Divisive, zero-sum outcomes are not an inevitable characteristic of politics in divided societies, but often a reaction to the ‘rules of the game’ under which democratic competition takes place. Democracy in Divided Societies showed how changes to these rules – for example, the introduction of electoral systems which facilitate cross-partisan communication, bargaining and inter-dependence between rival politicians and the groups they represent – can promote more responsible politics, and enhance democracy, even in divided societies.

Democracy, Ethnicity and Conflict

All societies are inherently conflictual to some degree. Democracy itself operates as a system for managing and processing conflict, rather than resolving it (Przeworski 1991). Within certain circumscribed boundaries, conflict is considered legitimate, is expected to occur and is handled through established institutional means. Representative institutions allow conflicts to formulate, find expression and be managed in a sustainable way, via institutional outlets such as political parties and legislatures, rather than being suppressed or ignored. Changing formal political institutions can result in changes to political behaviour and political practice, and the design of political institutions is thus of paramount importance to the management of conflict in any democracy.

This insight raises the prospect of engineering political rules to improve the operation of political processes and institutions. For the political engineer, institutions change outcomes, and changing formal political institutions can result in changes in political behaviour and political practice. Most scholarly advocates of political engineering agree that the electoral system is a key mechanism in shaping the wider political arena. Different electoral systems can encourage politicians and candidates to pursue starkly different avenues to electoral success: civil or hostile campaigns, broad or narrow policy agendas, cooperative or divided legislatures. This explains the scholarly consensus that “if one wants to change the nature of a particular democracy, the electoral system is likely to be the most suitable and effective instrument for doing so” (Lijphart 1995, 412).

The Theory and Practice of Centripetalism

Centripetalism envisages democracy as a continual process of conflict management, a recurring process of dispute resolution in which contentious issues must ultimately be solved via negotiation and reciprocal cooperation, rather than simple majority rule. We know from the seminal work of Duncan Black (1958) that many of the issues which confront democracies are not resolvable by majority decision, but rather ‘cycle’ through an endless series of unstable temporary majorities. In such situations, differences need to come to be seen not as irreconcilable sources of conflict, but as part of a broader collective action problem, a problem which can potentially be overcome by bargaining and reciprocal trade-offs.

Under this scenario the role of democratic institutions, as the mediating agents which can process divergent interests and preferences into centripetal outcomes, becomes paramount. As argued by emeritus law professor Donald Horowitz, one of the most feasible paths to such inter-group accommodation is to present political parties and candidates with incentives to cooperate across ethnic lines. Most of Democracy in Divided Societies was devoted to empirical analysis of the institutional foundations of such arrangements across a variety of societal contexts, ranging from elections in the traditional societies of Papua New Guinea and Fiji to modern industrialized states like Northern Ireland and Australia.

A common theme across all cases was that presenting office-seeking politicians and their supporters with sufficiently strong institutional incentives towards cross-partisan behaviour can change not only their approach to campaigning but also affect larger changes on the nature of political competition as a whole. For example, legislatures are likely to be more functional when many politicians owe their victory not just to their own diehard supporters but also those from other parties, who they may need for their re-election as well. Politicians who have had to bargain with their counterparts for mutual support at elections are likely to be amenable to doing the same once in government.

By strengthening the ‘moderate middle’ at the expense of the extremes, such centripetal reforms can also help address the widespread problems of polarization, a major issue facing the United States in recent years. Many political scientists see the standard U.S. electoral arrangements—closed party primaries followed by plurality general elections—as exacerbating political polarization and extremism, by enabling motivated partisan ideologues (who usually comprise only a small share of the overall electorate) to choose a candidate. As most Congressional and State legislative races are relatively uncompetitive, victory in many cases is a foregone conclusion, giving candidates limited incentive to pitch appeals beyond their core supporters (Gehl and Porter 2020).

Plurality voting is also a classic zero-sum game: more votes for my opponent means fewer voters for me. Ranked choice voting, by contrast, offers the potential for a positive-sum game: a candidate can benefit from ballots cast initially for someone else, if those votes return to her in the form of second or later rankings. Over time, this can encourage the formation of pre-electoral coalitions, resulting in potentially enduring “coalitions of commitment” in government (Horowitz 1985, 365-95). There is evidence of both practices occurring under the century-long use of RCV in Australia, via both formal and informal pre-election coalitions underpinned by ranking exchanges (Sharman, Sayers and Miragliotta 2002).

Such mutual reciprocity should also, in theory, promote more moderate political outcomes (Horowitz 1991; Reilly 2001; Mann and Ornstein 2012; Diamond 2015; Drutman 2020; Lublin and Reilly 2023), as in most cases, the way to attract wider support is to adopt more centrist or “catch all” policy positions which appeal to the median voter. The exception is where more votes are lost by such moderation than are gained, which is always a possibility in very safe districts or in places with entrenched ethnic polarization. But in most cases, politicians seeking to gain additional votes from non-core supporters should have an incentive to moderate their political rhetoric and broaden their policy positions to pick up additional voter support. (Reilly 2001, 2018).

There are, however, preconditions for encouraging this search for greater support. First, elections need to be genuinely competitive. If elections are uncontested or outcomes otherwise pre-determined, there is little reason to expect change. Similarly, fairly-drawn, not gerrymandered, districts are needed – to avoid the same issue of pre-determined outcomes. And there needs to be a base level of competition with at least three candidates, at a minimum – any less and elections will be won on a first-past-the-post basis. If these conditions are met, the kind of cross-party “vote pooling” -- the exchange of preferences between supporters of different candidates or parties -- which underpins moderation can occur under a range of electoral systems. But they are most likely to occur if some kind of ranked-choice ballot is present.

The Importance of Ranking

By giving voters the opportunity to express their preferences not just for but also between parties and candidates, ranked ballots have attracted significant enthusiasm from political thinkers. The first proposal for a ranked ballot in Britain in the 19^th century, for instance, was hailed by John Stuart Mill as “a scheme of almost unparalleled merit” for democratic governance, which would encourage people to behave in politics as rational individuals, forming their own opinions and considering their own interests. Contemporary political theorists agree: McLean, for instance, argues that some facility for preference ordering is one of three basic requirements of a good voting system (1987, 154). Similarly, Blais and Massicotte write that “the more information the ballots reveal about voters’ preferences, the more accurate the representation of preferences is likely to be” (1997, 76).

For most voters who are not political scientists or philosophers, the major advantage of a ranked ballot is that they do not need to be strategic about expressing their true choices. In plurality elections, by contrast, voters for smaller parties (and sometimes larger ones too) are often faced with an acute dilemma: should they vote sincerely for their true choice, even if that party or candidate is unlikely to win? Or should they instead abandon their favourite and strategically vote for the least-worst option amongst those who have a chance? While a ranked ballot doesn’t completely eliminate the potential for strategic voting, in general, voters can simply express their preferences honestly in the knowledge that voting sincerely can never hurt their chosen candidate.

Of the ten major electoral systems used in the world today, three enable electors to rank-order candidates on the ballot in this way: the alternative vote (AV), the most widely used form of RCV in the U.S. and also used for over a century in Australia, and also in Ireland for president and (in slightly different form) Papua New Guinea; the supplementary vote (SV), previously used for mayoral elections in the United Kingdom and presidential elections in Sri Lanka; and the single transferable vote (STV), as used in Ireland, Malta and most upper houses in Australia, which is a system of proportional representation. Another version of ranked ballot, in which rankings are counted as fractional votes, is used in Nauru and for some seats in Slovenia. All share a common feature: they enable electors to not only choose their favored candidate but also indicate their preference between others. It is this particular feature that distinguishes preferential voting from other electoral system options.

My book evaluated the case for centripetalism and ranked choice electoral methods by examining the electoral history of all the divided societies which utilise such institutions -- Papua New Guinea, Northern Ireland, and Sri Lanka as well as ‘one-off’ or short-lived cases such as Estonia, Bosnia, and Fiji -- plus other examples of established democracies in Australia, Europe and North America. Examination of these different cases suggested that, while vote pooling does indeed encourage viable steps towards inter-ethnic cooperation and conflict management in some countries, not all preferential systems are equally effective at promoting accommodation in divided societies.  

Case study:  Papua New Guinea

One key test of centripetalism came from a somewhat obscure case, that of Papua New Guinea. An ethnically-fragmented state in the South Pacific, Papua New Guinea’s first three elections – in 1964, 1968 and 1972 – were conducted under similar AV rules to that used in Australia, its colonial administrator until independence in 1975. These elections using AV revealed a distinctive approach to campaigning in what is a traditional clan-based society. Candidates from smaller clans or those without a large ‘core’ vote often campaigned outside their home base area for other voters’ second preferences. In other cases, traditional tribal alliances enabled aligned candidates to cooperate and aggregate support without the vote being ‘split’ several ways among competing candidates.

These accommodative trends were positive factors in a young and barely-developed democracy. But these advantages declined sharply when AV was replaced by plurality voting in 1975, leading to increasingly unrepresentative victors and increasing campaign violence (Reilly 1997, 2002). With incentives for campaign cooperation removed, ethnic groups reverted to their traditional hostilities. The return to conflict was magnified by the way plurality elections work to reward vote-splitting in a fragmented society. Candidates who previously campaigned broadly and encouraged the swapping of rankings with supporters of other candidates instead focussed their energies on maximising their clan-based vote, and in many cases restricting the campaigning of opposition candidates to their own home areas. This led to to a sharp increase in electoral violence, increasing number of split votes among “spoiler” candidates, and politicians being elected on alarmingly small pluralities, sometimes as little as 6 percent of the vote.

In 2003, Papua New Guinea changed back to a ‘Limited Preferential Vote’, essentially AV with three mandatory rankings. Studies of the 2007 elections found a sharp decline in campaign violence, at least part of which can be attributed to the new electoral system. In what is still an underdeveloped, fragmented and volatile democracy, the return of ranked voting has been applied not just for parliamentary elections but also for presidential elections in the autonomous province of Bougainville, reducing tensions and increasing cooperation during election campaigning and electing more broadly-supported and representative leaders than was the case under plurality voting.

Case study:  Northern Ireland

Northern Ireland too can be judged a successful case of vote-pooling, i.e. exchanges of ranked preferences among candidates and their supporters, under its proportional ranked choice system, the Single Transferable Vote. Unlike Papua New Guinea, the Northern Ireland case has been extensively researched, with several major studies published (McGarry and O'Leary 2004; O’Leary 2013, Mitchell 2014). Most find little vote-pooling prior to the 1998 elections which ushered in the power-sharing executive under the ‘Good Friday’ Agreement. Analysis of this first election suggests that preference-vote transfers served to give voice and representation to the ‘moderate middle’ sentiment for peace that existed within the community, and to translate this sentiment into an electoral majority for ‘pro-agreement’ parties.  After that election (1998–2007) transfers across the sectarian divide between the main moderate unionist party and the main moderate nationalist one increased sharply, particularly from unionists to nationalists. However, most transfers remain within ethnic blocs, particularly since the emergence of Sinn Féin as the largest party at the 2022 elections.

Other cases are less positive. Sri Lanka has used a preferential system to elect its president for over 40 years, throughout devastating civil wars and natural disasters, but the frontrunning candidate has always won an outright majority of first rankings, meaning that the system has yet to be really tested. Bosnia and Estonia both used ranked voting in one-off uses, whether at parliamentary elections in Estonia (1990) or sub-regional presidential polls in Bosnia (2000), before reverting to unranked systems. Fiji ran two elections under a hybrid AV system in 1999 and 2001 which gave parties, not voters, the main role in allocating rankings – a move which some blamed for the 2000 coup against the elected government. Following another coup in 2006, Fiji abandoned the system, and has recently return to elected government using list proportional representation. While varying widely, one lesson from all of these cases is that stability matters for electoral rules, with voters and candidates alike needing several iterations of using ranked systems to properly utilize its potential.

Critiques

Critics of centripetalism tend to focus on these short lived cases, sometimes arguing that vote-pooling systems are inherently unstable or unpredictable. But the century long Australian and Irish experiences, and indeed the growing body of evidence from the United States, suggests the contrary – that ranked voting can actually make politics more stable and predictable. Other critics of centripetalism argue that promoting stability and centrism is not what is needed in an age of multiple overlapping crises, and that more drastic changes are needed. It is true that ranked choice voting is more an evolutionary than a revolutionary reform – which is also why it has emerged as a viable reform option in long-established democracies such as the United States.

A more subtle critique holds that the idea of centrist politics relies on simplistic median voter models which elide the complexity and multi-dimensionality of contemporary politics in advanced democracies such as the U.S. (Santucci 2021). But even in a multi-dimensional policy space featuring voters and activists with competing preferences or ideologies, a vote-maximizing equilibrium position exists (Miller and Schofield 2008), and ranked systems are far better at identifying this than a straight plurality contest.

Centripetalism draws upon core political science ideas about the nature of social cleavages, particularly Seymour Martin Lipset’s classic arguments about the virtues of cross-cutting cleavages for promoting stable democracy. Normatively, the virtues of political aggregation and centrism are appealing for those schooled in the Anglo-American tradition of two-party politics, where centripetal reforms can be seen as being compatible with majoritarian political models. This may help explain its current popularity in the United States, where single-member RCV has been adopted in two states – Maine and Alaska – and dozens of cities, and studies have shown a very similar impact in terms of civility and countering polarization to that found in the literature on ethnically-divided societies (see for instance Donovan et al 2016; John et al 2018; Reilly 2021; Reilly, Lublin and Wright 2023).

Conclusion: The future of elections in divided, multi-ethnic cities and nations

Beyond these cases, the centripetal model of inter-ethnic accommodation is likely to attract attention in many global regions in the future due to the increasingly inter-mixed nature of ethnic group distribution around the world. According to the United Nations, over three-fifths of the world’s population will be urban by 2030. This worldwide trend of rural-urban migration towards multi-ethnic ‘world cities’ appears to be leading inexorably towards the development of massive, ethnically-heterogeneous urban metropolises as models of human settlement in the 21st century.

As ethnic groups increasingly find themselves in close physical proximity but separated by growing distinctions between rich and poor, and as both education levels and voter sophistication continue to rise, so the centripetal model of inter-ethnic accommodation based on the foundation of ranked ballot elections is likely to become an increasingly attractive option for constitutional engineers worldwide.

Benjamin Reilly, Adjunct Senior Fellow, East-West Center, Honolulu