Introduction

Suppose that an opportunity arises for Israelis and Palestinians, the two Koreas or Colombians to reach a settlement in their decades‐old conflicts. Would holding a referendum be useful in the ratification process? Should opposing constituencies in divided societies be asked to ratify a peace agreement? The growing debate on referendums around the world has prompted work across area studies (Butler & Ranney Reference Butler and Ranney; LeDuc Reference LeDuc), yet few scholars have focused on referendums in peace processes or post‐conflict transitions (Johansson Reference Johansson; Kersting Reference Kersting, Reference Kersting; Lee & MacGinty Reference Lee and MacGinty). Likewise, most comprehensive studies of comparative peace processes, ethnic outbidding, post‐conflict democratisation, and federal or consociational transitions have only touched on peace referendums. Worse yet, despite increasing global interest in direct democracy, international organisations and governments often design peace referendums without considering comparative cases or relevant international expertise.

As governments and the international community are currently debating future referendums in conflict‐ridden societies, analysing the value of potential ratification processes is important for the stability and legitimacy of peace‐building operations. In a best‐case scenario, referendums could facilitate political transformation and become the litmus test of a society's capacity to overcome deep divisions and bring forward new political dynamics. Successful referendums in South Africa in 1992 and Northern Ireland in 1998 have been viewed as facilitating the respective peace settlements by engaging broader segments of the society in the peace process and limiting the role of violent opposition groups (Strauss Reference Strauss; Guelke Reference Guelke; McGarry & O'Leary Reference McGarry, O'Leary and Taylor). But referendums might also have unintended side effects and inflame already unstable ethnic relations as suggested in East Timor and Kyrgyzstan. In the case of the 1999 referendum in East Timor, Paris (Reference Paris: 219) criticises the United Nations administration for organising a premature referendum on independence before providing security on the ground. Thousands of East Timorese were killed and about 400,000 displaced (Paris Reference Paris: 219; Schulze Reference Schulze: 78), while in Kyrgyzstan the constitutional referendum of 2010 led to an estimated 400–500 dead and about 100,000 Uzbek refugees (Huskey & Hill Reference Huskey and Hill). Thus, premature and ‘wrongly’ designed referendums could inadvertently turn into conflict‐inducing, rather than conflict‐mitigating, opportunities.

Admittedly, referendums are not the only way of engaging the public in peace‐building or the sole ratification method. Putnam (Reference Putnam: 436) argues in his seminal contribution on diplomacy and domestic politics that ratification processes may include voting in parliaments or (in)formal negotiations with bureaucratic agencies and interest groups. His work provides a set of important theoretical tools with which to dissect international agreements, but his insights have yet to be adapted to the study of referendums. Likewise, Tsebelis (Reference Tsebelis) recognises alternative ratification processes, but argues that referendums add one more ‘veto player’ (the target population) in decision making and therefore make the status quo more difficult to change. More importantly, Lijphart (Reference Lijphart) suggests that referendums might serve as instruments of oppression against minorities. However, few studies have critically examined this claim. Nor have scholars or practitioners made a serious attempt to identify the linkages between alternative ratification processes and mediation outcomes in peace processes.Footnote

Research design

In a bid to fill the gap in the literature, the article focuses on referendums and the role of institutions in peace mediations, emphasising two contrasting episodes of referendums in South Africa and Cyprus. The article utilises the ‘least likely’ case study research design applied to cases of successful or failed conflict transformation. A case study is described as crucial when it is ‘least likely’ to fulfil a theoretical prediction yet does so, thereby going against current theories (Eckstein Reference Eckstein, Greenstein and Polsby; Gerring Reference Gerring). For instance, in the study of referendums in peace processes, a crucial or deviant case study is one that seems unlikely to follow a peaceful transformation, yet an enabling referendum design has provided the catalyst to nudge it to the right direction. Deviant cases, according to Lijphart (Reference Lijphart: 2), have considerable theoretical significance because of the light they can throw on the social conditions of stable and effective democracy. Thus, crucial cases of successful referendums are those which evidence high levels of ethnic polarisation prior to the referendum – as, for instance, pre‐apartheid South Africa – yet peacemakers have positively reversed their nation's fortunes, shaping a peaceful transition to a post‐conflict environment (Strauss Reference Strauss; Guelke Reference Guelke; Ross Reference Ross).

Conversely, another set of crucial cases are those where domestic conditions and international engagement have been highly promising, yet the mismanagement of the referendum timing and design has led to a renewed stalemate. This article focuses on Cyprus as a paradigmatic case. The European Union has emerged since the 1990s as the new and promising actor in the island that could have used its power to ‘catalyse’ a peace settlement (Richmond Reference Richmond; Rumelili Reference Rumelili; and Tocci Reference Tocci). For one thing, the EU could have provided the ideal forum to change public attitudes in the conflict, by creating a ‘win‐win’ framework, which could subsequently set the stage for a settlement (Yesilada & Sozen Reference Yesilada and Sozen; Hadjipavlou Reference Hadjipavlou; Anastasiou Reference Anastasiou; Kaymak Reference Kaymak). For another, the Cyprus conflict was relatively peaceful preceding EU involvement, with almost zero casualties since 1974 (Doyle & Sambanis Reference Doyle and Sambanis: 257–281; Heraclides Reference Heraclides). This fact would have been arguably conducive for a peace settlement since, unlike pre‐apartheid South Africa, a generation of Cypriots has grown up with almost no direct exposure to violence. Still, despite these promising circumstances, Cyprus remains divided following the failure of the Annan Plan referendum in 2004.

Of course, there are crucial differences in the background conditions of Cyprus and South Africa. What is most important in this comparative design is that the ‘most critical differences’ in the level of external incentives, ethnic/racial polarisation and violence, could have predicted a ‘yes’ vote in Cyprus and yet another pro‐apartheid ‘no’ vote in South Africa. Admittedly, there was a widespread perception in South Africa that a ‘no’ vote would have had negative repercussions, leading to new waves of violence and international sanctions. But for the most part, violence could have turned South Africans away from the peace process as suggested, for instance, in Israel/Palestine (Kydd & Walter Reference Kydd and Walter). Lake and Rothchild (Reference Lake and Rothchild) have argued that violence frequently initiates a vicious cycle of ethnic fear and more violence while Galtung (Reference Galtung) has questioned the ‘naivety’ of sanctions. Likewise, in South Africa, both De Klerk (Reference De Klerk) and his advisors in interviews with the author argued that the fear of sanctions had very little effect on their decisions.Footnote

An alternative explanation points to the ANC's (African National Congress) moderate policies (including carefully crafted statements at the time of the referendum), which played a positive role in securing the ‘yes’ vote among South African whites (Butler & Ranney Reference Butler and Ranney: 9; Strauss Reference Strauss). But as argued in this article, the ANC's quiescence was partly due to the design of the referendum, which allowed it enough manoeuvring space to both reject in principle the white‐only initiative and encourage voters to support De Klerk. However, as the Cypriot experience suggests, it is questionable whether positive actions across the ethnic divide alone could be a sufficient condition for a positive referendum outcome. In the months preceding the 2004 referendum, the Turkish Cypriots community engaged in impressive peace rallies. Such mobilisations could have arguably influenced the ‘yes’ vote, especially among moderate AKEL voters. The nominally communist AKEL had historically built a strong identity around the prospect of reunification emphasising ‘common legacies’ with the Turkish Cypriots, particularly on the left. Its identity orientation would have predicted a ‘yes’ decision, particularly if the referendums were properly timed. Yet AKEL reversed its position and voted against the Annan Plan in a shocking last‐minute decision (Trimikliniotis Reference Trimikliniotis; Pericleous Reference Pericleous; Kaymak Reference Kaymak).

To what extent does the timing and design of the referendums explain the differences in the South African and Cypriot peace processes? To answer this question, the article turns to secondary and available primary sources, including legal documents, public opinion surveys, parliamentary debates and interviews with key advisors of leaders in the peace process, as well as individuals with privileged access to information. It also utilises comparative cases of referendums in other peace processes.Footnote Finally, based on the lessons that can be learned from the two case studies, it offers recommendations for the design, monitoring and implementation of peace referendums in general.

Mandate referendums

In a mandate referendum, a leader presents a general idea or a framework for a peace arrangement to the public in an attempt to secure early approval for the nascent agreement, thereby expediting a conditional or unconditional directive to finalise it. Such referendums can help initiate a peace settlement by offering advice, direction, and more importantly, a mandate to negotiating leaders. Although the literature in conflict resolution has long emphasised timing (Zartman Reference Zartman, Reference Zartman), there has been little application of this concept to the design of peace referendums. Through mandate referendums, leaders can better prepare the public for a peace settlement, incorporate early feedback into the negotiations, strengthen their credibility across ethnic antagonists and safeguard the peace process from subsequent and often unavoidable reversals in public opinion.

Mandate referendums should not be seen as a panacea, particularly if leaders prioritise personal or partisan gains over the peace process or if leaders are unwilling to take even ‘calculated risks’ for peace. Yet mandate referendums could provide visionary leaders more flexibility in choosing the ‘ripe moment’ to move the peace process forward. More importantly, such referendums could be more effective if the leadership proposing them maintains its overall popularity (as in the case of De Klerk's government) or if international conditions are favourable (e.g., the period before accession to the EU for the two Cypriot communities). If leaders wait until the completion of the negotiations, it might be too late for their own popularity to be effective or for external incentives to work, as suggested by the failed Cypriot referendum.

The South African experience

The South African white‐only referendum of March 1992 offers one of the best examples of a pre‐agreement mandate referendum. Given its apartheid origins, it has received little attention from political scientists (for exceptions, see Strauss Reference Strauss; Kersting Reference Kersting; Sussman Reference Sussman) while its transferability to other settings has not yet been widely debated, even though it legitimised and facilitated the end of one of the most repressive regimes in the twentieth century. Interestingly, South Africans reached a peace settlement in their first major attempt, unlike other deeply divided societies such as Cyprus where negotiations have failed for decades despite systematic UN mediation attempts. The peaceful transition from apartheid in South African has been described as a ‘miracle’ by experts in the field (Guelke Reference Guelke; Waldmeir Reference Waldmeir; Sandal Reference Sandal), particularly given the fate of other white ‘settler’ societies in sub‐Saharan Africa, major economic disparities, as well as mutual fears and grievances. In other words, South Africa, especially the timely political transformation of its white community, fits Eckstein's (Reference Eckstein, Greenstein and Polsby) criteria of a ‘least likely’ case study of a positive conflict outcome where institutional features, including mandate referendums, played a decisive role.

The South African referendum was triggered by a by‐election in Potchefstroom, which led to the defeat of De Klerk's National Party and challenged his authority as chief negotiator for the white community. As in other divided societies facing peace negotiations, opponents made every effort to de‐legitimise his policies, stating that the ruling National Party did not represent the nation and even branded de Klerk a traitor (Sussman Reference Sussman). In response, Klerk considered general elections, but South Africa's electoral system would have privileged rural areas where the opposition had an advantage among white farmers. However, his opponents' rhetoric emphasising the need for the nation to be properly represented in the negotiations, along with a statement from the deputy leader of the opposition Conservative Party, Dr Ferdi Hartzenberg, that his party could win a referendum, allowed De Klerk some space for electoral manoeuvring.Footnote In the end, De Klerk responded by initiating a snap referendum, arguing that a mandate was necessary and that the use of a referendum process would cause the least disruption in the ongoing negotiations (De Klerk Reference De Klerk). The mandate character of the referendum also allowed the National Party more flexibility in re‐wording the referendum question (Kersting Reference Kersting), thereby taking advantage of recent public opinion polls.Footnote De Klerk asked voters to reject or endorse his reform policies to negotiate an end to white minority rule through talks with the black majority.

The referendum asked: ‘Do you support continuation of the reform process which the State President began on February 2, 1990 and which is aimed at a new constitution through negotiation?’ (Kersting Reference Kersting; De Klerk Reference De Klerk: 232). Moreover, the referendum vote took place in the context of the apartheid regime: blacks were not allowed to participate in the ballot, nor were Indians or persons of mixed‐race. De Klerk argued that the referendum needed to be restricted to the white community which was the only one divided on the necessity of reforms and said he would regard majority support as granting him a mandate to sign ‘binding agreements’ in constitutional talks (Kersting Reference Kersting; De Klerk Reference De Klerk: 232). What makes the South African case particularly interesting is the ambiguity in having a second referendum at the end of the negotiations. On this issue, De Klerk said in advance that it would be meaningless to ask white voters to reauthorise such agreements unless the final settlement was substantially different from that currently envisaged by his party (Kersting Reference Kersting).

As Putnam (Reference Putnam: 447; see also Raiffa Reference Raiffa) argues, issue linkage is extremely important in a two‐level negotiation process, particularly in altering feasible negotiation outcomes. De Klerk linked the referendum to peace in general and to his own personal fate and political future more specifically. He promised to resign and hold elections if his proposed reforms failed to win a significant majority. His threat targeted, on the one hand, the South African moderates and business leaders who feared a return to international isolation and black revolt and, on the other hand, the ANC leadership who informally rallied behind his campaign (Butler & Ranney Reference Butler and Ranney: 9). The ANC issued a statement saying that the referendum benefited the negotiating process and emphasised that this should be the last occasion on which South Africa was subject to the indignity of a racial or ethnic referendum.Footnote As Mandela had no simultaneous referendum to run, he felt comfortable intervening in the process, correcting the opposition's misinformation campaign. For instance, to confront the Conservative Party's claim that a ‘yes’ vote was tantamount to ‘white suicide’, the ANC leader reassured white farmers that: ‘President Robert Mugabe's plans to nationalize farm land in Zimbabwe had no bearing on the situation in South Africa’ (Strauss Reference Strauss: 347).Footnote

De Klerk's landslide victory included both rural Afrikaner communities and urban English‐speakers, surprising even his own party officials who had predicted a 55 per cent ‘yes’ majority, at best, well below the 68.7 per cent he received (Waldmeir Reference Waldmeir: 47). Admittedly, in South Africa, the mandate referendum option was not unproblematic. De Klerk later became more demanding in the negotiations while the ANC saw no reason to help him to pacify a weakened right wing opposition (Waldmeir Reference Waldmeir: 184–185). Equally, at the end of the negotiations, De Klerk was criticised for violating his promises in the referendum, particularly with regards to white demand for power‐sharing (Giliomee Reference Giliomee). Nonetheless, in the bigger picture, the referendum allowed the white community to make its exit from apartheid with some grace (Giliomee Reference Giliomee: 634) and transformed the image of the ‘other’ in South Africa's racial relations, opening the door for mutual political accommodation (Ross Reference Ross). According to key advisors of De Klerk, the referendum boosted both the confidence and legitimacy of his policies and determined without any doubt the question of who represented the white community in South Africa.Footnote In a particularly colourful image, one advisor described the period preceding the referendum as a ‘stagnant pool full of crocodiles and malaria’, arguing that the plebiscite was a decisive move by De Klerk to move ‘back to the river stream’ (Kersting Reference Kersting: 219).

The 1992 ‘yes’ vote added durability and institutional legitimacy to the post‐apartheid transition as well. Through his use of the referendum process, De Klerk delegitimised his opponents and expanded his own domestic and international credibility. Despite his weak negotiating position (having to negotiate under the threat of international sanctions and domestic violence), De Klerk secured credible constitutional provisions for minority groups and respect for property rights – key issues for the white constituencies (Waldmeir Reference Waldmeir: 243; Strauss Reference Strauss: 346). More importantly, the 1992 referendums created new cross‐cutting linkages, including a shared political vision of the future among supporters of the ‘yes’ vote which transcended ethnic and communal boundaries. De Klerk managed to keep his party together from the day he removed the ban on the ANC until he agreed to hand over power, thus bringing his constituency from the backwater of ethnic politics into the modern world (Waldmeir Reference Waldmeir: 217). Even though demands for permanent power‐sharing were not included in the final settlement, the endorsement of De Klerk's policy added an element of responsibility among ANC elites; this later translated into creating a new common symbolic landscape of racial relationships in South Africa while maintaining and reinforcing the inclusion of whites within ANC and the governing structures of the country (Ross Reference Ross: 257).Footnote

The 2004 failed Cypriot referendum

A contrasting case to South Africa, Cyprus demonstrates the weaknesses of post‐agreement referendums. The island of Cyprus has remained de facto divided into Greek and Turkish Cypriot controlled sections since 1974. Under the proposed 2002–2004 UN plan for reunification (‘the Annan Plan’), Greek and Turkish Cypriots would have retained autonomy over most of their affairs under a decentralised federal system. Turkish Cypriots promised to return land to Greek Cypriot displaced persons occupied by the Turkish military since 1974 in exchange for power‐sharing, EU membership and federal status within a reunited Cyprus (Pericleous Reference Pericleous; Sözen & Özersay Reference Sözen and Özersay; Faustmann & Kaymak Reference Faustmann and Kaymak).

The Annan Plan required multiple concessions from both sides. Greek Cypriots were particularly concerned about the role of Turkey as a guarantor power and provisions for the massive naturalisation of post‐1974 Turkish settlers threatening the demographic composition of the island. But at the same time, Greek Cypriot moderates emphasised the importance of major territorial concessions after the settlement; this included the cities of Famagusta (Varosha) and Morphou,plus a large number of villages across the buffer zone.

Voting rights in the referendums were determined by the communities themselves: Greek Cypriots voted in the south while Turkish Cypriots voted simultaneously in the north. Although the Annan Plan had the support of the two main Greek Cypriot political parties (DISY and initially AKEL) representing two‐thirds of the electorate, it was eventually rejected by a landslide 76 per cent of Greek Cypriots while 65 per cent of Turkish Cypriots approved it during the twin April 2004 referendums.

In contrast to other deeply divided societies, both Cypriot communities were generally seen as pro‐settlement, particularly in light of EU accession, to the extent that UN diplomats initiated the idea of a referendum that would not require the formal endorsement of chief negotiators from each community.Footnote In 2004, UN mediators created a specific referendum formula to bypass the intransigent former leader of the Turkish Cypriot community, Rauf Denktaş, and acting under the assumption that Greek Cypriots were unlikely to reject the peace deal since AKEL and DISY had already signalled their tentative support for the Annan Plan. In general, moderate Greek Cypriot leaders had argued in favour of a compromise, pointing out that each day without a settlement intensified the waves of colonisation (from Turkey) and brought the occupied territories of Cyprus closer to ‘irreversible partition’. They warned that the continuation of the status quo would allow Turkey and the Turkish settlers to dominate the north while Turkish Cypriots would return to the South to reclaim their properties and constitutional rights (Pericleous Reference Pericleous).Footnote

As Cyprus fits the criteria of a crucial case study, what explains the failure of the ‘likely to succeed’ Cypriot referendum? Conventional wisdom assumes that referendums will fail if the content of the peace agreement does not satisfy the most fundamental needs and fears of the public. Yet peace agreements unavoidably contain undesirable provisions subject to nationalist framing, and these are often intensified in contested referendums (Hancock Reference Hancock2011b; Ross Reference Ross; Kaufman Reference Kaufman). So, too, ratification attempts might fail because of the nature and timing of the process (Putnam Reference Putnam; see also Kersting Reference Kersting; Kaymak Reference Kaymak). For instance, simultaneous post‐agreement referendums among antagonistic communities could be problematic as sides aim exclusively at their own constituencies, often successfully framing their own gains from the negotiations while weakening the position of moderates across the ethnic divide. Similar dilemmas apply to international mediators having to justify positions that could harm one of the two communities during decisive moments in their referendum campaigns.

To this Putnam adds the negative and positive aspects of preference restructuring and reverberation during international negotiations. Any bargaining situation, he says, ‘involves attempts by the players to restructure the game and to alter one another's perceptions of the costs of no‐agreement and the benefits of proposed agreements’ (Putnam Reference Putnam: 454). Mandela's assurances on property issues during the 1992 South African referendum are an example of positive reverberation across communal lines. Conversely, nationalist opposition to the Annan Plan in Turkey and the Turkish Cypriot community caused Greek Cypriots to fear that even the ‘few positive aspects of the Plan will not be implemented’.Footnote To this, the chairman of the ‘yes’ campaign in Northern Ireland, Quintin Oliver, adds that sides in the conflict present diametrically opposed visions of what the agreement means to their own constituencies, feeding each other's negative propaganda.Footnote Simply stated, mandate referendums could overcome this problem of negative reverberation if each side could identify its ripe moment to win a mandate in its own community, thus minimising negative input across the communal divide.

Cyprus also demonstrates that failed referendums can have long‐term path‐dependent effects, preventing moderates from utilising constructive ideas previously rejected in a popular referendum. Admittedly, mandate referendums are not immune to manipulation and failure; any negotiating strategy could backfire if leaders nurture unrealistic expectations about forthcoming international negotiations. Or spoilers might seek a mandate to prolong a conflict and deny concessions to out‐groups, as Bosnian Serb leaders did in 1994; yet such nationalist mandate referendums could count very little in international politics (particularly if condemned early) as suggested by the subsequent Dayton peace process in Bosnia.

These points highlight the broader advantages of De Klerk‐style mandate referendums for other divided societies. To begin, mandate referendums are less risky or damaging for the peace process. If politicians lose their mandate, they could opt to step down without sacrificing the essence of the peace process. Additionally, popular directives allow communities to prepare themselves for peace negotiations; they can provide leaders with feedback and indicate the priorities to be addressed. Moreover, assigning a mandate to an elected leader balances democratic accountability with pragmatism in negotiations: thus, as implied by Putnam's analysis, De Klerk was able to negotiate a credible end to apartheid without fearing a risky referendum at the end of the process. Meanwhile, Greek and Turkish Cypriot negotiators had to factor into their cost/benefit calculations the possibility that one of the sides would renege from the agreement at some point in the future. More importantly, by restricting any mandate for later renegotiation, the Annan Plan allowed propaganda to dominate on issues that could have very easily been clarified in any subsequent version – for instance, on the status and territorial waters of the British bases. Finally, mandate referendums improve leaders' negotiating strategies and boost their odds of reaching a settlement. By providing certainty to the other side, more concessions can be demanded at the negotiating table.

EU engagement and counterfactuals in Cyprus

Could the EU have incentivised Greek Cypriots to unilaterally endorse the Annan framework in 2002–2003? A frequent critique of the Cyprus referendum is that EU accession was already secured by the time of the referendum, and therefore Greek Cypriots had no immediate incentive to vote ‘yes’. This line of thought implies that a segment of the Greek Cypriot electorate particularly in the left rejected the plan, hoping for an improved settlement.Footnote The timing of the referendum after accession was effectively secured and the lack of a firm endorsement of the Greek Cypriot leadership were arguably the worst decisions made by the UN during the Cyprus mediations of 2002–2004.

Admittedly, it is hard to know what would have happened, if Turkish Cypriots had agreed on a settlement in the decade preceding accession. Nonetheless, had a mandate referendum been considered earlier, the Cyprus Problem would have had a reasonable chance of ratification with a majority vote in the Greek Cypriot community. Admittedly, polls in the 2002–2003 period indicate widespread scepticism and large fluctuations in public opinion, but at this time, Greek and Turkish Cypriots were not significantly different in their attitudes towards the Annan Plan.Footnote According to a poll in February 2003, 50 per cent of both sides opposed the Plan,Footnote while a year earlier, 54 per cent of Greek Cypriots supported itFootnote (incidentally, a percentage close to early predictions for the 1992 South African referendum).

Following the December 2012 Copenhagen Summit and informal accession of Cyprus to the EU, both Cypriot President Glafkos Clerides and Greek Prime Minister Costas Simitis enjoyed unprecedented support in the Cypriot polls (82.6 and 85.6 per cent, respectively).Footnote Another poll right before the February 2003 elections indicated that Greek Cypriots saw Clerides as ‘the most competent to handle the Cyprus problem’.Footnote The polls imply that had he considered a referendum earlier, he could have possibly prevented the AKEL‐Papadopoulos alliance which led to the election of Papadopoulos as President in 2003. While Papadopoulos used his position to demonise the Annan Plan,Footnote Clerides was willing and capable to steer public opinion in favour of the settlement with the strong backing of the Greek government, particularly if additional concessions were made to Greek Cypriots in return for their ‘unilateral endorsement’ of the Plan. Such concessions might have reduced Turkish Cypriots support below the 65 per cent level it enjoyed in 2004, but a majority vote for the Plan in both communities would have been conceivable. At the time, Turkish Cypriots and Turkey were refusing to support the Plan, but a compromise might have been possible by 2004, when Turkish Cypriots and Turkey changed their policies. Since the latter still rejected the plan in 2002–2003, an alternative option would have been to ‘lock’ the Greek Cypriot side earlier into endorsing an ‘improved Annan Plan’ through a mandate referendum.

Inevitably, as internal referendums target only one of the many communities in a conflict, it is easier to initially commit only one community to holding a referendum during a particular moment in the negotiation process. Each community might have its own peculiarities, needs and preferred timing in addressing its domestic constituencies. In fact, only weeks before the April 2004 referendum, the undecided AKEL leadership finally pointed to the timing issue and asked the UN to postpone the referendum. Interviews in Cyprus suggest that AKEL actually considered abstaining from the 2004 vote to minimise the damage that would be caused by an overwhelming negative outcome, aiming to initiate another referendum vote once constituencies were less polarised.Footnote Indeed, Greek Cypriot public opinion appeared to have shifted in favour of a compromise by 2008 following the victory of AKEL's Demetris Christofias; then in 2013, the main advocate of the Annan Plan, Nikos Anastasiades, won the presidential elections with the largest popular mandate for a Cypriot president in decades. And speaking with the wisdom of hindsight, members of the UN team in Cyprus told the author that it was a mistake not to encourage Christofias to run a South African‐style mandate referendum shortly after making the first breakthroughs in bi‐communal negotiations with the Turkish Cypriot community in 2009.Footnote

Conclusions

Overall, the study of peace referendums offers an important addition to academic and public policy literature, particularly as the international community is currently debating even advocating the efficacy of referendums in conflict‐ridden societies around the globe (Goddard Reference Goddard). For instance in 2013, the Israeli government took a new law to the Knesset requiring that all territorial concessions be ratified by voters in a special referendum (Goddard Reference Goddard), while Colombia's congress approved a referendum to seal an eventual peace deal with the country's FARC rebels.Footnote Moreover, peace referendums have been debated in the past decades among other places in Sri Lanka (Edrisinha Reference Edrisinha), Western Sahara (Zunes & Mundy Reference Zunes and Mundy), Abyei (Johnson Reference Johnson), Nagorno‐Karabakh (Johansson Reference Johansson) and Kashmir (Bose Reference Bose) potentially affecting the future of some of the world's most volatile geopolitical regions. The findings of this article are even more relevant for those conflicts around the globe where referendum options have not entered the public debate yet and, as a result, leaders aiming for a negotiated settlement have more flexibility in deciding among the various alternatives suggested here.Footnote

Leadership is of critical importance in contested peace processes. South Africa and Cyprus suggest how leaders could either mobilise voters for peace or alternatively frame potentially promising peace settlements as catastrophic. Another example is President Charles De Gaulle in France, who held a pre‐referendum on the self‐determination of Algeria in 1961. In an intense political campaign, including three nationally broadcast speeches, he threatened to resign if his proposal to continue negotiations was defeated. Like De Klerk, he used his personal leverage to promote a settlement. De Gaulle warned the public that ‘a weak or negative response would cause the downfall of much he had achieved and trigger renewed subversion and political instability’ (Lustick Reference Lustick: 289). He received a massive 72 per cent ‘yes’ vote and proceeded to implement his policies to negotiate with Algerians, sidelining army radicals and pied‐noir resistance in the process.

As this article concludes, timing and early consultation with the public are also extremely important. Besides Cyprus and South Africa there are other examples. For instance, the 1998 referendum on the contested issue of granting citizenship to Russian‐speakers in Latvia took place years before the country's accession to the EU (Morris Reference Morris). Likewise, in the case of Northern Ireland's mostly post‐settlement referendum, constructive ambiguities in the agreement offered stakeholders (including London and Dublin) the mandate to implement the agreement and redesign provisions as necessary, thus safeguarding the agreement from any subsequent reversals in public preferences.

Of course, it is hard to predict what would have happened to Israeli Prime Minister Yitzhak Rabin had he initiated a referendum for the Oslo I Accord in 1993. Rabin ignored protests and relied exclusively on slim support from the Knesset (also dependent upon Arab voters) (Haklai Reference Haklai). Rabin could have likely defeated his opponents in a referendum and won the mandate to proceed with a comprehensive settlement as polls suggested that 53 per cent of the electorate supported the Oslo Agreement in late August 1993 (Waxman Reference Waxman, Elman, Haklai and Spruyt). However, even if he had won the referendum, religious extremists would have argued that no one has the right to concede territory as Israel is God‐given and beyond the decisions of people at a certain time and place (Haklai Reference Haklai; Waxman Reference Waxman, Elman, Haklai and Spruyt). Nonetheless, even in the case of Israel, an informal mandate referendum might have been a safer instrument for mediation than a post‐settlement one, even when it had failed. For example in the case of the failed 2004 Gaza referendum among Likud party members, Ariel Sharon opted to ignore his party and the (unofficial) negative outcome (Shamir Reference Shamir). Such responses might be more difficult in the future for Israeli leaders, given existing or newly proposed legally binding referendum clauses legislated in the Knesset.

In a bid to assist relevant future academic and policy research, I offer in this article a number of possible referendum designs, noting both their positive and problematic aspects. On the one hand, the design and timing of referendums could lead to successful outcomes in peace processes despite conditions that predict otherwise, as in the case of South Africa. On the other hand, poorly designed referendums and/or poor timing could inflame already unstable situations or lead to continuous stalemates as in Cyprus. Although adopting a set of fixed recommendations is not a good idea, mandate referendums have a great deal to offer. By securing early approval of a settlement in the form of conditional or unconditional mandates, leaders can incentivise a subsequent peace process. Mandate referendums are more likely to succeed in the first place and once successful, they could extend the durability of the peace process for the long‐term future. Overall, prior endorsement of a settlement by the public could sustain peace in difficult times, especially during unavoidable reversals in public opinion.

Mandate referendums are not risk‐free (particularly if polls are indecisive) yet such risks are unavoidable if referendums are already dictated by normative, constitutional or political necessities. Often taking a risk is necessary to move a political process forward. But as most experts and advisors to De Klerk have pointed out, the South African mandate referendum allowed the South African leader to take calculated risks rather than face a volatile and uncertain future.Footnote An alternative option, of course, is for policy makers to avoid referendums at all costs, hoping that the subsequent implementation of the settlement will increase public legitimacy through normal electoral process. But in cases where referendums are already promised or nearly decided due to legal/constitutional provisions, mandate referendums should be part of available options familiar to specialists.

The article extends Putnam's influential work on ratification processes to the study of peace referendums. Although it builds on his two‐level game theory model, it departs from his emphasis on domestic politics as a constraint in international negotiations. By way of contrast, it makes a contribution to the relevant literature by identifying a missing link in Putnam's theorising through which domestic politics could create positive opportunities for peace settlements. The South African case demonstrates how participation fosters social learning, leading to more creative solutions to seemingly intractable conflicts.Footnote Overall, by treating domestic constituencies as a resource rather than a constraint, leaders could maximise their available pathways for positive transformation.

The broader literature and debates on violent conflict in emerging democracies could gain significant insights from this analysis as well (Paris Reference Paris; Mansfield & Snyder Reference Mansfield and Snyder; Jarstad & Sisk Reference Jarstad and Sisk). For scholars and practitioners advocating unqualified support for democratisation, this study suggests that benign intentions should not be a substitute for the proper strategising and timing of democratic initiatives, including popular referendums. As some form of public ratification of peace processes might be unavoidable due to legal, constitutional or normative considerations, governments and international organisations should consider ‘least destructive’ (even potentially promising) institutional options such as mandate referendums. Moreover, for those interested in leadership studies, the analysis suggests pathways and linkages through which visionary and determined leaders could calculate their risks and put forward a successful peace agenda. As demonstrated in this article, leaders often hold the key in securing the passage of a peace referendum.

While cautioning against adopting a fixed set of recommendations, the article highlights the multiple advantages of securing advance mandates in peace negotiations. Questioning the absence of relevant academic and policy literature on the topic, it demonstrates how governments and international organisations frequently propose and implement referendums in conflict‐prone societies without systematically observing past failures and successes. Future research could identify relevant applications of these concepts not only in peace‐building operations, but also in international mediations on issues of major global concern, such as humanitarian intervention, global warming and the sovereign debt crisis (a highly timely issue for Cyprus and other countries in the eurozone crisis). Consistent failures in addressing potentially catastrophic challenges in world politics suggest the need for novel institutional designs in international mediations in general and peace processes more specifically.

Acknowledgements

I am grateful to Adrian Guelke, Allan Leonard, Antonis Ellinas, Brendan O'Leary, Cherrel Africa, Christoforos Fokaides, Claudio Radaelli, Ioannis Papadopoulos, Han Dorussen, Iosif Kovras, John McGarry, Keith Gottschalk, Lawrence Leduc, Luis Fernando Medina, Oded Haklai, Quintin Oliver, Stuart Kaufman, Richard Whitman, Tom Schumann (South Africa Parliament Library) and the anonymous reviewers of the EJPR for their comments and support on this project.