Ending civil conflict is difficult, particularly through political settlements. Conflicts now often occur in states with elections, and voters have sometimes been directly involved in the process, potentially in efforts to overcome elite divisions. Yet, according to evidence from the 2016 popular plebiscite in Colombia, referendums and other tools of direct approval by voters seem to amplify elite divisions and therefore are not a useful mechanism to strengthen peace processes in this way. Focusing instead on traditional elite-led negotiations that seek to satisfy each faction may have a better chance of producing signed settlements, although the Colombian case also suggests some alternative forms of inclusivity that may help increase the overall legitimacy of the process and improve the odds of implementation.

Ending civil conflict is difficult. While settlements negotiated between combatants have become the most common form of termination since the end of the Cold War–more common than victories by either side–they are especially hard to secure and stabilize.1 What will yield peace? Conflicts now often occur in states with elections, meaning that various actors may be involved in peace processes that seek settlements. Settlements can be approved by empowered elites alone, by institutional mechanisms like congressional votes, or by direct voter involvement, perhaps as part of an effort to overcome elite divisions or increase legitimacy. Direct voter involvement in the approval process may also be a component of a trend toward greater inclusivity around all aspects of settlements.2

In Colombia, direct voter involvement through a 2016 plebiscite was employed, in part, in an effort to offset an elite challenge and add legitimacy to a settlement. Our analysis of this case, however, suggests that a referendum may paradoxically provide an important platform for elites seeking to upend the peace process, and that it may be especially easy to mobilize voters against a settlement when components can be framed as offering concessions to rebels. Using case evidence, including survey data from 2004 to 2016, we show that elite opposition to the peace process, based on division among elites, could be part of the explanation of the plebiscite's rejection in Colombia.

We posit that referendums and other tools of direct voter approval can amplify elite divisions and, therefore, should not be employed to overcome elite opposition in order to strengthen peace processes. Focusing on traditional elite-led negotiations–seeking to satisfy the necessary factions and using the simplest approval processes available to provide for the required constitutional changes–may have a higher chance of producing successful settlements. Such negotiations remain the central component of most peace processes, and our results suggest maintaining that exclusive structure.3 The Colombian case, however, also suggests that other forms of inclusivity can help increase legitimacy for the process, potentially improving the odds of implementation, which merits further study.

In our examination of the 2016 Colombian popular plebiscite, which sought direct voter approval of a peace process, we first overview the Colombian conflict and how it compares with other civil conflicts. Next, we describe the elite division. We then present survey data on public opinion toward a settlement prior to the plebiscite and results from the plebiscite, demonstrating that support decreases with the elite division and suggesting that running such a referendum may paradoxically provide a platform for elites seeking to upend the peace process. We then show evidence from a survey experiment that indicates that components of peace agreements that are framed as concessions for rebels are especially unpopular, making referendums or other direct voter involvement a risky strategy. Finally, we address the implications of these arguments for other states seeking an end to civil conflict through a settlement.

In many ways, the Colombian case looks like other civil conflicts, but it also presents a unique opportunity to account for voter attitudes in the peace process. For more than fifty years, Colombia has experienced a bloody armed conflict between the government, left-wing guerrilla groups, and right-wing paramilitary bands. On the left, the Revolutionary Armed Forces of Colombia (farc) emerged in 1964, followed over the next two decades by other leftist guerrilla groups, including the National Liberation Army (eln), the Popular Liberation Army (epl), and the 19th of April Movement (m-19).4 The farc, however, secured a position of strength due to its expansion strategy, as well as its eventual involvement in drug trafficking.5 On the right, organized paramilitaries emerged in the 1980s, clashing with the leftist guerilla groups and, at times, the government.6 This internal confrontation resulted in thousands of deaths, millions of displaced citizens, and tremendous economic and environmental destruction.

While a complex and important case in its own right, Colombia is also very similar to other civil conflicts, despite having one of the longest-running insurgencies in the world. Colombia is a clear case of asymmetric conflict–the most common civil war type–and it has featured varying levels of conflict, including many strong combatant groups in the beginning, but fewer weaker groups more recently, reflecting the composition of most other wars in the current era.7 By the late 1990s, the United States and Colombia teamed up to fight insurgency, initially through broader regional programs and then through the targeted Plan Colombia. Between the 1990s and 2000s, most left-wing guerrilla groups signed negotiated settlements with the state, and most right-wing paramilitary bands demobilized, but the farc persisted.

Colombia is a most likely case for the use of direct voter involvement in the approval of a peace process, and it thereby serves as a potential example for other similar cases. A long-standing electoral democracy, Colombia's regime dates back to 1957, but it was further opened in 1991, when a new constitution reorganized state structures and promoted a more pluralist and competitive political system.8 Most armed actors developed a relationship with political parties and electoral politics. Even throughout the conflict, Colombia remained one of the most stable Latin American democracies.9 And as we will discuss later, civil conflict often occurs in states with elections, making Colombia an early but not unique case.

Prior attempts to establish a settlement between the farc and the government have failed. In the mid-1980s, the government negotiated with many of the leftist groups.10 The Belisario Betancur administration and the farc signed a 1982 agreement to transform the guerrilla group into a political party and to make the political system more competitive.11 As a result of this process, the farc formed the Unión Patriótica (up) party, and the government implemented reforms such as the popular election of mayors. Nonetheless, over just a few years, thousands of up members were assassinated, primarily by right-wing paramilitaries but with plausible government complicity, and the farc split from the party and continued fighting.12 However, other left-wing guerrilla groups signed settlements and became political parties in the democracy reshaped by the constituent assembly that changed the constitution in 1991.13 These concessions were tailored to these rebels who, in return, agreed to demobilize, disarm, and renounce violence. The agreements, however, did not include the farc.

Instead, the government launched a major offensive against the farc in 1992.14 In 1999, President Andrés Pastrana initiated a new cycle of peace talks with the farc. During this period, the organization created a new political wing and even held territorial control, but the talks failed to produce a settlement as each side accused the other of focusing instead on strengthening itself on the battlefield. In 2002, President Álvaro Uribe recognized the political status of right-wing paramilitary bands and initiated peace talks with these groups, disassembling most of these organizations.15 But, with regard to the farc, the administration established an aggressive counterinsurgency strategy that debilitated, but did not defeat, the remaining guerrillas.16

A new peace process began in 2012, but elite divisions threated to upend it, despite its reliance on a popular plebiscite for approval. After decades of failed negotiations, the farc and the Colombian government returned to peace talks in 2012. The government announced a “road-map” (Acuerdo General para la terminación del conflicto y la construcción de una paz estable y duradera) that established six points of negotiation: rural development policy; political participation; end of the conflict; solutions to the problem of illicit drugs; victims; and the implementation, verification, and referendum to put the deal in place. A negotiating team representing each side, facilitated by multilateral mediation, met in Oslo and then Havana.17 In May 2013, a joint communique from the team showed agreement on the first point, and, by November, reports stated that political participation had been negotiated: the farc was to be designated as a legal political movement, a provision that has facilitated peace in other contexts,18 and political representation in territories most affected by the conflict was to be expanded, potentially reducing grievances but also representing farc constituencies.19 Over the next two years, the negotiating team worked through the remaining points, despite pauses, and the government and the farc showed their commitment to the process by declaring ceasefires. Other actors, including delegations of victims, were also consulted during the process.20

Even prior to the negotiations, however, the elites on the government side fractured, led by President Juan Manuel Santos against his predecessor President Uribe, the former ultimately supporting the settlement and the latter opposing it. Before the Santos administration, the popular, and populist, Uribe administration held office; Santos had been the defense minster during the Uribe administration, and he had implemented the hardline security policies that were part of Plan Colombia. President Uribe, who was denied a constitutional amendment that would have allowed him to seek a third term, initially backed Santos. Santos won with 69 percent of the vote in the 2010 presidential elections.21

However, relations soured between Santos and Uribe by early 2011. A rift first appeared in 2010, only a few weeks after his inauguration, when Santos reestablished diplomatic ties with Venezuela, a decision that Uribe criticized. As Santos took a more conciliatory approach, including moving toward peace negotiations with the farc and loosening laws used to prosecute members of the group, relations between the two politicians deteriorated.22 Santos's 2012 announcement of negotiations with the farc, however, triggered a formal rupture with Uribe, who created an organization (Colombians against Terrorism) and later a party (Centro Democrático) to oppose Santos.23 Uribe called the government insufficiently patriotic, claimed the settlement gave too many concessions to the farc, and, ultimately, accused Santos of treason against his legacy.24

Peace talks with the farc progressed, however, and, in January 2013, Santos had proposed a referendum to approve a prospective settlement.25 This proposal stood in contrast to a constituent assembly that had been used to make the 1991 changes to the Constitution, which the farc preferred.26 Indeed, when Santos sent legislation on the referendum to Congress in August of that year, the farc called for a pause in negotiations to examine it.27 Although the process was meant to be inclusive, especially once the comprehensive settlement was negotiated, this mechanism for approving that final deal was unexpected. Uribe also came out against this proposal, suggesting that Santos was using a referendum on peace as an electoral ploy (and it was initially set to coincide with the next elections).28

By the 2014 election, and without a comprehensive settlement yet negotiated, Santos finished behind Uribe's new choice, Óscar Iván Zuluaga, in the first round of voting; in the runoff, however, he clinched a reelection with 50.25 percent of the vote.29 This election merely marked what had become a clear division between a camp united behind Uribe's hardline agenda against insurgency, and a pro-peace coalition that included various parties led by President Santos.30

Despite farc opposition, and Uribe's skepticism, Santos succeeded in establishing a plebiscite, which was approved by Congress in 2015 and by the Constitutional Court in 2016. During the process, he referred to Uribe and his supporters as “enemies of peace,” saying that those opposed to the settlement were “trying to demonise the process and create fear in the country,” but that voters would have their say, suggesting that voter approval of the peace process might overcome these elite divisions.31 A popular plebiscite that succeeded may indeed have overridden the Uribe opposition and provided the needed legitimacy to the peace process.

Attitudes toward the peace process shifted as the elites split. Although the 2012–2016 negotiations between the Santos administration and the farc produced an agreed-on settlement, the plebiscite failed, reflecting opposition from the Uribe camp. But to what extent did this division among elites shape voter attitudes so that a narrow majority rejected the plebiscite, a mechanism paradoxically designed in part to overcome these very divisions? And what about the plebiscite made it so easy for elites to lead an effective opposition campaign?

The Observatorio de la Democracia of the Universidad de los Andes and the Latin American Public Opinion Project (lapop) of Vanderbilt University collected public opinion data from twelve national representative surveys between 2004 and 2016.32 To assess public attitudes toward a peaceful solution to the conflict, and thereby probe the plausibility of elite divisions in producing changes in public opinion, we examined the evolution of two questions that have been regularly included in the Americas Barometer survey, before and after the elite division: the first captures the percentage of Colombians who support a negotiated solution to the conflict with guerrillas, compared with a military solution or a combination of both strategies; the second measures the percentage of individuals who think forgiveness and reconciliation with farc members is possible. Complementing these data is the actual vote in the 2016 plebiscite.

We expected to see a downward trend in these attitudes, primarily after the division between Santos and Uribe, but before any components of the settlement were negotiated and made public by political camps (Uribe's camp versus other political camps). Our expectations build on previous research showing that the public is responsive to elites' opinions and their cues to voters.33 Referendums and other mechanisms for direct voter involvement may be especially afflicted by elite framing, as we will discuss further.34

The majority of Colombians have supported a peaceful solution since data collection began in 2004. Such support was well above 60 percent before the 2011 elite division, but then dropped to 55–58 percent, reaching its low in 2011, before finally rising again in 2016, after the settlement was actually signed. Similarly, attitudes toward forgiveness and reconciliation with the farc were initially high, ranging from 58 to 64 percent between 2006 and 2008, before decreasing to their lowest at 40 percent in 2014 (the surveys in intervening years did not ask this question, unfortunately), and then increasing slightly in 2016.35

These national averages have shown the expected downward trends, reaching their lowest points after the Santos-Uribe division (2011 onward). The decreases are apparent before particular components of the settlement were negotiated and announced (the first point made public in 2013), suggesting that the elite division rather than the revelation of the settlement's specific policies may account for the changes. These trends, of course, cannot prove that Uribe's opposition was the cause–other factors such as the visibility of farc members and their crimes after the start of the peace process may have played a role–but the evidence is suggestive of the public responding to the cues of a divided elite.

To further probe the plausibility of this argument, we map our variables by political camp in order to see if there are differences in opinions between Uribe supporters and other respondents. Using vote choice reports for the previous presidential election, we created a variable for the political camps of respondents, a dichotomous indicator that takes the value of one for Uribe supporters and zero otherwise.36

The comparison by political camp demonstrates the expected relationship with respect to support for a political solution to the conflict (Figure 1). The percentage of those in the Uribe camp with favorable attitudes toward a settlement has been greater than 50 percent for most of the period, though it was lower than the percentage of non-Uribistas supporting this option. In 2011, the two lines converged, perhaps because of mixed signals from the elites: the distance between the politicians' views on negotiations was not as evident until the next year, when talks began. Thereafter, Uribistas' support drops off, reaching its lowest level in 2014.

Figure 1

Percentage Supporting a Negotiated Peace with Insurgents by Political Camp, 2006–2016

Figure 1

Percentage Supporting a Negotiated Peace with Insurgents by Political Camp, 2006–2016

Close modal

Attitudes toward forgiveness and reconciliation with the farc show a similar pattern (Figure 2). Between 2006 and 2008, these attitudes were not significantly different between political camps. Uribe promoted a peace process with the paramilitaries during that period, so part of the convergence may be explained by a contamination effect across armed actors. By 2014, when peace talks with the farc were in motion, the camps had substantially diverged and, by 2016, when the settlement was signed, only 44 percent of those in the Uribe camp believed forgiveness and reconciliation with the farc was possible.37

Figure 2

Percentage with Positive Attitudes Toward Forgiveness and Reconciliation with the farc by Political Camp, 2006–2008 & 2014–2016

Figure 2

Percentage with Positive Attitudes Toward Forgiveness and Reconciliation with the farc by Political Camp, 2006–2008 & 2014–2016

Close modal

Finally, we examined the extent to which votes in the recent plebiscite also reflected elite divisions. At the municipal level, we ran a simple correlation between the 2016 plebiscite results and the outcomes for the 2014 presidential election.38 The vote share for the Uribista presidential candidate, Zuluaga, positively correlates with the percentage of “no” votes, and Santos's vote share with the percentage of the “yes” (both are statistically significant).39

These attitudes and votes in the popular plebiscite show evidence of the possible impact of the Santos-Uribe division on voters, even though it was meant to overcome elite divisions.

Other factors contributed to opposition to the popular plebiscite, but they do not seem to explain the shifts in camps that coincide with the division between elites; rather, if anything, they further reinforce the risk of directly involving voters in the approval process.

While attitudes in the Uribe camp began dropping, turning against a settlement, even before specific components were announced, the support rates dropped to the point of producing a statistically significant difference between the camps only after some of those specific components were made public (for example, Figure 1 shows less than 50 percent support in the Uribe camp in 2014, which was after the announcement of the first provisions). All settlements include concessions to the rebels, wherein de jure power is brought more in line with de facto power, making these settlements easy for elites or other actors to oppose. Even Colombians who were generally supportive of a settlement prior to the plebiscite were less supportive of particular provisions that may have been construed as concessions. For instance, while a majority in both camps typically supported a settlement as the solution to civil conflict (Figure 1), that support did not extend to creating the conditions to allow the farc to participate politically (just 13 percent of Uribistas and 35 percent of other camps supported this provision, according to the 2016 ab-lapop survey) or forming a political party (11 and 23 percent support, respectively).

The perception that the government was making concessions seemed to have shaped voter attitudes: in a survey experiment run in areas most affected by the conflict, Aila M. Matanock and Natalia Garbiras-Díaz show that support for a proposal to provide more political representation to those areas is much lower when it is reported that the farc had endorsed the proposal (than when the proposal had simply been made).40 Endorsement by the farc produced a drop in the percentage of respondents supporting the proposal from 44.4 percent to 31 percent.41 These results are even more surprising given that all respondents would have directly benefited from increased political representation (because this sample covers regions set to receive more seats). Overall, the revelation of these components may have helped solidify voters' preferences against the settlement, and they were framed as concessions by Uribe during the opposition campaign (framing the transitional justice as not sufficient, for example: “the lack of justice doesn't produce a feeling of reconciliation”).42 But the timing of the downturn in attitudes, beginning prior to the announcement of the components, tentatively suggests that the elite division played a central role.

Another possible explanation for the split is that those in Uribe's camp turned against the settlement because they predicted that land reform provisions would be a component of it, rather than cue off Uribe's attacks on aspects of the peace process that he labeled “concessions.” However, while Uribe and some of his political coalition are against land reform, he did not often attack this component of the settlement–and with good reason, as land reform is very popular among Colombians, receiving approximately 80 percent support in the 2016 ab-lapop survey. Socio-economic status and preferences toward redistribution (something land reform would accomplish) also do not correlate with support for the settlement.43 Land reform was always likely to be a component of a settlement with the farc, due to its popularity and the farc's leftist platform. Attitudes toward it do not seem to be an omitted variable in our analysis. Likely knowing these preferences among the population, Uribe's attacks focused mainly on the transitional justice and farc political participation provisions.

Despite the fact that the failed plebiscite was seemingly established in part to overcome elite divisions, this evidence suggests that it amplified those divisions instead.

But the Colombian case also provides important implications for other peace processes. Modern civil conflicts often occur in countries with elections, even in democratic countries, so other states may be tempted to follow Colombia's lead in using referendums and other tools of direct approval by voters. Among ongoing civil conflicts that reach a twenty-five battle-death threshold,44 the mean level of democracy rose six points on a nineteen-point scale from 1974 (the beginning of the third wave of democratization) to 2010, and a majority of countries experiencing such conflict in 2010 were more democratic than authoritarian.45 The Arab Spring may have decreased the relative share of civil conflicts in democracies somewhat, but many fully democratic countries are still fighting their counterinsurgencies, including India, Kenya, and Turkey (as of the latest democracy data in 2013).46

So far, the use of referendums to approve peace processes has been relatively rare: fewer than 20 percent of the settlements in the ucdp Peace Agreement Dataset over the past four decades.47 Most of these cases are in territorial conflicts wherein voters later weigh in on succession, such as in South Sudan, rather than an approval mechanism for the settlement overall.

Other states with elections, however, may be tempted to use referendums and other forms of direct voter participation in the approval of a peace process, perhaps especially when elites are divided and the government is less than popular.

Just as lessons may be drawn from successful dimensions of settlements, unsuccessful dimensions also hold important implications for settlement design in other cases.48 Specifically, we posit that this failed popular plebiscite suggests that, if elite divisions exist, these mechanisms for direct voter approval may amplify splits, rather than provide additional legitimacy to and strengthening of the peace process. While mass action is crucial in many stages of conflict and postconflict contexts–for example, during wartime, when civilians can provide essential information and resources to combatants–this type of inclusivity at the approval stage of a settlement may not be one of them.49

Focusing instead on traditional, elite-led negotiations that seek to satisfy necessary factions may be more likely to yield a signed peace agreement. In fact, to secure a settlement, leaders on each side of a conflict must perceive the share of power they will receive through a settlement as comparable to what they would receive from continued fighting.50 Similar to any negotiated regime transition, elite pacts will create new state structures, producing changes that are acceptable to elites even if they slow the speed of change.51 In contrast to recent recommendations on inclusivity during peace processes, this case suggests that focusing on meeting the expectations of sufficient elite factions–so either all factions that may wield a veto or a sufficient number of factions to override any vetoes–may be the best option to obtain a signed settlement in many cases.

Ultimately, the Colombia case sought to follow a similar strategy, although the failed plebiscite made it more difficult. After the vote, the government called meetings with the opposition to discuss their objections to the agreement. Santos and Uribe finally met face-to-face to talk about peace. Later, the two negotiating teams met again in Havana to renegotiate the agreement. After a few weeks, the farc and the Colombian government announced a new deal that included modifications reflecting some points highlighted by Uribe and the opposition. Finally, the new agreement was approved in Congress at the end of 2016. The opposition, however, remained unsatisfied with the settlement and now accuses the government of betraying the people's will and democratic principles. There are, of course, cases in which it will be difficult to get necessary elite factions on board, as it was in Colombia, and having a failed plebiscite certainly does not help. But there remain some inclusivity strategies to deal with minor elite factions that are still opposed.

While many studies of spoiling in peace processes (that is, upending a bargain that the major factions would otherwise agree on to end the civil conflict) focus on the rebel side, the Colombian case makes it clear that factions on the government side can also spoil a settlement.52 Again, this suggests that incorporating the elites of as many major factions as possible before isolating minor ones may be the best path forward. Other studies have suggested a similar strategy, arguing that ensuring the leaders of the major government and rebel factions can find an option they prefer to conflict may require exclusivity, in order to limit the number of actors who have to agree and, thereby, to provide those crucial elites with the most possible options to terminate conflict.53

Many cases, including El Salvador and South Africa, for instance, match this template: both feature a coalition of elites who accepted negotiations and, ultimately, a settlement (and those elites who were recalcitrant were neutralized through a wide pro-peace coalition that included middle-class segments).54

Beyond the main implication that a focus on elite factions may be useful in securing a signed settlement, we draw two important lessons from the Colombian case about seeking inclusivity in this step of the process to help overcome minor elite factions that remain in opposition to the agreement. First, we suggest not using a referendum or other direct vote on approval of the peace process. These mechanisms generally may not overcome elite divisions, perhaps in part because peace processes are complex issues, so voters look for elite cues. Given the uncertainly in these processes, elites opposed to the settlement may have the easiest time framing terms as concessions and the status quo as the safest option (factors like elite popularity seem to play an important role in these contexts, rather than the issue itself).55 Some have noted that referendums and the like are “risky” strategies.56 When components of settlement can be framed as concessions, which are unpopular, as the Colombian case makes clear, the strategy may be even riskier.

Second, the Colombian case also suggests that some inclusivity may be possible, even at this stage of the conflict. Other work has suggested that inclusivity, although not yet common at most stages of ending a conflict, is useful for increasing legitimacy and even improving the odds of implementing (if not securing) a settlement.57 In terms of process, Colombia suggests that including representatives of the voters, either through a constituent assembly as in the 1990s or directly through Congress as after the failed plebiscite, may be a way to achieve some degree of inclusivity without the same risk of amplifying elite divisions. This proposition, however, would need to be further tested.

In terms of audience, the Colombia case also suggests that if a referendum is held, it could be restricted to certain areas–specifically those areas most affected by the conflict–to achieve direct voter involvement with less risk of elite cues driving the outcome. Colombians directly affected by armed conflict, particularly at the hands of the farc, measured through displacement and attacks in particular areas, have been among the most supportive of the peace process.58 Both victims and non-victims in these areas tend to have more positive opinions about peace and reconciliation than do those in areas less affected by political violence. Rural regions that have most recently been the areas most affected by violence also show strong support for the peace process.59 More important, these regions may be least affected by elite framing because they live the conflict and thus are more likely to seek a deeper understanding of a settlement that will affect them on a day-to-day basis. This would fit with theory on elite framing in other contexts, which suggests that topics that voters tend to know less about, such as foreign policy for U.S. voters, is more susceptible to this type of influence. Colombians living in Bogotá would fit this model, since they currently experience very little of the conflict's violence and have weaker incentives to pay close attention to its potential solutions. This proposition, too, should be further tested.

Aside from these potential lessons from the Colombian case, mediators may be able to find other ways to increase inclusivity, perhaps at other stages in the process, even while focusing on elite factions at the stage of settlement approval.60

The deference to solving elite divisions in many settlement processes may indeed be why such settlements are often successful. The Colombian case suggests that a popular plebiscite or similar mechanism may not solve elite divisions but may actually amplify them. While this essay counters policies recommending inclusivity at every stage of a peace process, and instead recommends focusing on satisfying necessary elite factions when seeking to approve a settlement, it nonetheless identifies other mechanisms by which voters, especially in conflict areas, can still be included, potentially increasing the legitimacy and even the chances of success of a peace process.61

1

Joakim Kreutz, “How and When Armed Conflicts End: Introducing the ucdp Conflict Termination Dataset,” Journal of Peace Research 47 (2) (2010): 243–250; James D. Fearon, “Civil War & the Current International System,” Dædalus 146 (4) (Fall 2017); Paul Collier, V. L. Elliott, Håvard Hegre, Anke Hoeffler, Marta Reynal-Querol, and Nicholas Sambanis, Breaking the Conflict Trap: Civil War and Development Policy, World Bank Policy Research Report (Washington, D.C.: World Bank and Oxford University Press, 2003); Barbara F. Walter, “Does Conflict Beget Conflict? Explaining Recurring Civil War,” Journal of Peace Research 41 (3) (2004):371–388; and Monica Duffy Toft, Securing the Peace: The Durable Settlement of Civil Wars (Princeton, N.J.: Princeton University Press, 2009).

2

See Desirée Nilsson, “Anchoring the Peace: Civil Society Actors in Peace Accords and Durable Peace,” International Interactions 38 (2) (2012): 243–266.

3

Catherine Barnes, Owning the Process: Mechanisms for Political Participation of the Public in Peacemaking (London: Conciliation Resources, 2002); David E. Cunningham, “Responding to Multi-Party Civil Wars: Designing Peace Processes That Make Peace More Likely,” paper presented at the American Political Science Association's Annual Meeting, Chicago, Illinois, August 30–September 2, 2007; and Anthony Wanis-St. John and Darren Kew, “Civil Society and Peace Negotiations: Confronting Exclusion,” International Negotiation 13 (1) (2008): 11–36.

4

Lawrence Boudon, “Colombia's M-19 Democratic Alliance: A Case Study in New-Party Self-Destruction,” Latin American Perspectives 28 (2001): 73–92.

5

Juan G. Ferro and Graciela U. Ramón, El orden de la guerra: las FARC-EP, entre la organización y la política (Bogotá: ceja, Centro Editorial Javeriano, 2002); and Vanda Felab-Brown, “Organized Crime, Illicit Economies, Civil Violence & International Order: More Complex Than You Think,” Dædalus 146 (4) (Fall 2017).

6

Mauricio Romero, Paramilitaries y autodefensas, 1982–2003 (Bogotá: Editorial Planeta Colombiana, 2002).

7

David E. Cunningham, “Veto Players and Civil War Duration,” American Journal of Political Science 50 (4) (2006): 875–892.

8

Daniel L. Nielson and Matthew S. Shugart, “Constitutional Change in Colombia: Policy Adjustment through Institutional Reform,” Comparative Political Studies 32 (3) (1999): 313–341; and Erika Moreno, “Whither the Colombian Two-Party System? An Assessment of Political Reforms and Their Limits,” Electoral Studies 24 (3) (2005): 485–509.

9

Jonathan Hartlyn and Arturo Valenzuela, “La democracia en América Latina desde 1930,” in Historia de América Latina: política y sociedad desde 1930, vol. 12, ed. Leslie Bethell (Barcelona: Crítica, 1997).

10

Elizabeth J. Bilyeu, Guerrilla Groups in Colombia: Prospects for the Future (Leavenworth, Ky.: U.S. Army Command and General Staff College, 1995); and Boudon, “Colombia's m-19 Democratic Alliance.”

11

Steven S. Dudley, Walking Ghosts: Murder and Guerrilla Politics in Colombia (New York: Routledge, 2004); and author interview with Alberto Rojas Puyo, Bogotá, Colombia, December 3, 2009.

12

Dudley, Walking Ghosts.

13

Jaime Zuluaga Nieto, “De guerrillas a movimientos políticos (Análisis de la experiencia colombiana: el caso del m-19),” in De las armas a la política, ed. Ricardo Peñaranda and Javier Guerrero (Bogotá: tm Editores, 1999); author interview with Enrique Flórez, Bogotá, Colombia, March 16, 2010; and Boudon, “Colombia's m-19 Democratic Alliance.”

14

Lawrence Boudon, “Guerrillas and the State: The Role of the State in the Colombian Peace Process,” Journal of Latin American Studies 28 (1996): 279–297.

15

Enzo Nussio and Kimberly Howe, “When Protection Collapses: Post-Demobilization Trajectories of Violence,” Terrorism and Political Violence 28 (5) (2016): 848–867.

16

The government was attempting a defeat like the one in Sri Lanka. See Sumit Ganguly, “Ending the Sri Lankan Civil War,” Dædalus 147 (1) (Winter 2018).

17

Richard Gowan and Stephen John Stedman, “The International Regime for Treating Civil War, 1988–2017,” Dædalus 147 (1) (Winter 2018).

18

Aila M. Matanock, “Bullets for Ballots: Electoral Participation Provisions in Peace Agreements and Conflict Recurrence,” International Security 41 (4) (2017): 93–132; Aila M. Matanock, Electing Peace: From Civil Conflict to Political Participation (Cambridge: Cambridge University Press, 2017); and Aila M. Matanock, International Insurance: Why Militant Groups and Governments Compete with Ballots Instead of Bullets (Ph.D. Diss., Stanford University, 2012).

19

Juanita León, “El Acuerdo de participación política: lo acordado, lo aplazado y lo ganado,” La Silla Vacía, November 6, 2013, http://lasillavacia.com/historia/el-acuerdo-de-participacion-politica-lo-acordado-lo-aplazado-y-lo-ganado-46035.

20

See, for example, Luis Fernando Arias and José Antequera Guzmán, “Peace Proposals from Victims of Colombia's Armed Conflict,” presentation for the Colombia Peace Forum, United States Institute of Peace, Washington, D.C., July 29, 2014.

21

Mónica Pachón and Gary Hoskin, “Colombia 2010: análisis de las elecciones presidenciales y legislativas,” Colombia Internacional 74 (2011): 9–26.

22

“Santos v Uribe,” The Economist, April 7, 2012, http://www.economist.com/node/21552204.

23

Margarita Batlle and Gustavo Duncan, “Colombia: un panorama menos confuso,” Revista de Ciencia Política 33 (1) (2013): 101–116.

24

Stathis N. Kalyvas, “Jihadi Rebels in Civil War,” Dædalus 147 (1) (Winter 2018); and “Uribe acusa a Santos de traición y mentira,” El Espectador, November 18, 2014, http://www.elespectador.com/noticias/politica/uribe-acusa-santos-de-traicion-y-mentira-articulo-528107. The division was primarily between Juan Manuel Santos and Álvaro Uribe, but they were backed by various political, economic, and social leaders. The “yes” campaign in the plebiscite was ultimately supported by most political parties in Congress (the progovernment coalition), many local politicians, the business elite, the unions, indigenous organizations, and various news outlets. The “no” coalition included Uribe's party (Centro Democrático) and several civic groups linked to the right, including the Public Prosecutor, the Association of Cattlemen, various evangelical leaders, and one tv station (rcn). Civic groups arose on both sides to promote the campaigns. Victims' groups, veterans, the Catholic Church, and the Conservative Party were divided between the yes and no vote. See “Estos son los activos del No y del Sí en la recta final del plebiscito,” La Silla Vacía, September 28, 2016, http://lasillavacia.com/historia/estos-son-los-activos-del-no-y-del-si-en-la-recta-final-del-plebiscito-58128.

25

“Santos propone ‘refrendar popularmente’ acuerdos con las Farc,” El Tiempo, January 16, 2013, http://www.eltiempo.com/archivo/documento/cms-12520841.

26

See, for example, Moreno, “Whither the Colombian Two-Party System?”

27

Simon Romero, “Rebels Delay Peace Talks in Colombia,” The New York Times, August 23, 2013, http://www.nytimes.com/2013/08/24/world/americas/colombian-rebels-put-talks-with-government-on-pause.html.

28

Edulfo Peña and Daniel Valero, “¿Qué más hay detrás del referendo sobre paz impulsado por el Gobierno?” El Tiempo, August 22, 2013, http://www.eltiempo.com/archivo/documento/cms-13009187.

29

Felipe Botero, “The Legislative and Executive Elections in Colombia, 2014,” Electoral Studies 36 (2014): 225–229.

30

The coalition included the Liberal Party, Cambio Radical, the u Party, the Greens, segments of the Conservative Party and the left, and some social movements.

31

“Colombia Santos: farc ‘to Keep Weapons until Referendum,”’ bbc, September 7, 2013, http://www.bbc.com/news/world-latin-america-23998567.

32

Samples are representative of citizens of voting age. Information was gathered using stratified probability samples, with regions, municipality size, and urbanization as the main strata. Clusters were defined as house blocks (manzanas) for urban areas and hamlets (veredas) for rural areas. They were selected randomly in each municipality. In each cluster, six participants (three male, three female, divided into three age groups) were randomly selected by the enumerators. All samples consisted of approximately 1,500 face-to-face interviews.

33

For example, on foreign policy, which may be removed from voters in a similar way to the conflict for many in Colombia (since it has been largely confined to rural regions in recent years), see John Zaller, “Elite Leadership of Mass Opinion: New Evidence from the Gulf War,” in Taken by Storm: The Media, Public Opinion, and U.S. Foreign Policy in the Gulf War, ed. W. Lance Bennett and David L. Paletz (Chicago: The University of Chicago Press, 1994), 186–209; John Gray Geer, From Tea Leaves to Opinion Polls: A Theory of Democratic Leadership (New York: Columbia University Press, 1996); Adam J. Berinsky, “Assuming the Costs of War: Events, Elites, and American Public Support for Military Conflict,” The Journal of Politics 69 (4) (2007): 975–997; and Matthew A. Baum and Philip B. K. Potter, “The Relationships between Mass Media, Public Opinion, and Foreign Policy: Toward a Theoretical Synthesis,” Annual Review of Political Science 11 (2008): 39–65.

34

See, for example, Amanda Taub and Max Fisher, “Why Referendums Aren't as Democratic as They Seem,” The New York Times, October 4, 2016, https://www.nytimes.com/2016/10/05/world/americas/colombia-brexit-referendum-farc-cameron-santos.html.

35

See figures a1 and a2 in the authors' online appendix at https://dataverse.harvard.edu/dataverse/matanock.

36

The lagged measure of choice means that we reduce any risk of seeing differences between groups driven by sorting into different camps once the division becomes clear. Camps are held constant until the next election, making Uribe's potential camp more than his actual camp, which, if anything, should reduce the size of our effects. There is no consistent ideology question that is applicable to Colombia that we could use instead. This is, therefore, the best political camp measure. Individuals interviewed between 2006 and 2010 were coded as in his camp if they voted for Uribe in the previous presidential election (2002 and 2006). For the 2011 and 2013 surveys, the Uribe camp consists of those who voted for Santos in 2010. For the 2014 and 2016 studies, the Uribe camp consists of those who intended to vote for Zuluaga. The 2013 survey was excluded from the analysis because the question on previous vote choice was not included in that year.

37

Using support for a settlement as the dependent variable, we also more rigorously examined these patterns through logistic regression analysis (see the authors' online appendix for complete results and predicted probabilities). The interaction between two crucial independent variables, “political camp” (a binary indicator coded one for Uribe camp and zero otherwise) and “breakup 2011” (a binary indicator coded one for interviews conducted after the split in 2011 and zero otherwise), we found that after 2011 and among the Uribe camp, the predicted probability of supporting a peaceful solution with the farc reached its lowest point: about five out of every ten supporters of the former president agree with a negotiated peace. We control for being a resident of an urban area, years of education, age, and gender (male) in these models.

38

Electoral results, at the municipal level, for the 2014 presidential elections and the 2016 plebiscite vote were obtained from the website of the National Registry Office (Registraduría Nacional del Estado Civil).

39

Respectively, the correlations are 0.77 (p < 0.00) and 0.61 (p < 0.00).

40

Aila M. Matanock and Natalia Garbiras-Díaz, “Running Candidates after Using Violence? A Survey Experiment on the Colombian Peace Process,” working paper at the University of California, Berkeley, 2016.

41

See Figure a4 in the authors' online appendix.

42

“El plebiscito en las frases de Santos y Uribe,” La Nación, October 2, 2016, http://www.lanacion.com.co/index.php/actualidad-lanacion/item/277462-el-plebiscito-en-las-frases-de-santos-y-uribe.

43

Nicolás Galvis Ramírez, Omar David Baracaldo Orjuela, Miguel García Sánchez, and Catalina Barragán Lizarazo, Barómetro de las Américas Colombia: Paz, Posconflicto y Reconciliación, 2016 (Bogotá: Observatorio de la Democracia, Universidad de los Andes, Colombia, and usaid, 2017). However, there are certain regions where opposition to land reform was concentrated and that voted at high rates against the plebiscite; see Juan Tellez and Kyle Beardsley, “Santos' Gambit and the Shadow of Domestic Politics in Peace Processes,” Political Violence @ a Glance, November 9, 2016, https://politicalviolenceataglance.org/2016/11/09/santos-gambit-and-the-shadow-of-domestic-politics-in-peace-processes/.

44

Kreutz, “How and When Armed Conflicts End.”

45

Monty G. Marshall, Keith Jaggers, and Ted R. Gurr, “Polity iv Project: Political Regime Characteristics and Transitions, 1800–2004” (College Park: Center for International Development and Conflict Management, University of Maryland, 2006).

46

Democracy, or audience costs more broadly, seems to have surprisingly little effect on the termination type of civil conflicts, although some have hypothesized that settlements may be less likely, including because citizens in democracies may see concessions as undesirable. See Joakim Kreutz, “Why Can't Democracies Settle Civil Wars?” Political Violence @ a Glance, May 20, 2014, https://politicalviolenceataglance.org/2014/05/20/why-cant-democracies-settle-civil-wars/. Initial evidence, however, counters this proposition: we examined whether democracies were less likely to have a settlement that provides for concessions (the closest measure to our question of interest in available data). See Jason Lyall, “Do Democracies Make Inferior Counterinsurgents? Reassessing Democracy's Impact on War Outcomes and Duration,” International Organization 64 (1) (2010): 167–192. We found no difference compared with other regime types (see Table 1 in the authors' online appendix).

47

Natalia Garbiras-Díaz and Aila M. Matanock, “Elite Cues and Endorsement Experiments in Conflict Contexts,” working paper at the University of California, Berkeley, 2017.

48

On Colombia, see, for example, Javier Ciurlizza, “Post-Conflict in Colombia: The International Potential of Peace,” Open Democracy, March 9, 2016, https://www.crisisgroup.org/latin-america-caribbean/andes/colombia/post-conflict-colombia-international-potential-peace; and “Colombia Creates Peace before Ending a War,” The Christian Science Monitor, June 22, 2016, http://www.csmonitor.com/Commentary/the-monitors-view/2016/0622/Colombia-creates-peace-before-ending-a-war.

49

Eli Berman and Aila M. Matanock, “The Empiricists' Insurgency,” Annual Review of Political Science 18 (2015): 443–464.

50

See Matthew Soberg Shugart, “Guerrillas and Elections: An Institutionalist Perspective on the Costs of Conflict and Competition,” International Studies Quarterly 36 (2) (1992): 121–151.

51

See, for example, Terry L. Karl, “Democracy by Design: The Christian Democratic Party in El Salvador,” in The Central American Impasse, ed. Giuseppe DiPalma and Laurence Whitehead (London: Croom Helm, 1986), 195–217; Guillermo O'Donnell and Philippe C. Schmitter, Transitions from Authoritarian Rule: Tentative Conclusions about Uncertain Democracies (Baltimore, Md.: John Hopkins University Press, 1986); and Barry R. Weingast, “The Political Foundations of Democracy and the Rule of Law,” The American Political Science Review 91 (2) (1997).

52

Stephen J. Stedman, “Spoiler Problems in Peace Processes,” International Security 22 (2) (1997): 5–53; and Andrew H. Kydd and Barbara F. Walter, “The Strategies of Terrorism,” International Security 31 (1) (2006): 49–80.

53

Wanis-St. John and Kew, “Civil Society and Peace Negotiations”; and Cunningham, “Responding to Multi-Party Civil Wars.”

54

Elizabeth J. Wood, Forging Democracy from Below: Insurgent Transitions in South Africa and El Salvador (Cambridge: Cambridge University Press, 2000).

55

Arthur Lupia and Mathew D. McCubbins, The Democratic Dilemma: Can Citizens Learn What They Need to Know? (Cambridge: Cambridge University Press, 1998); and Lawrence LeDuc, “Referendums and Deliberative Democracy,” Electoral Studies 38 (2015): 139–148.

56

See, for example, Taub and Fisher, “Why Referendums Aren't as Democratic as They Seem.”

57

See Nilsson, “Anchoring the Peace: Civil Society Actors in Peace Accords and Durable Peace”; and Barnes, Owning the Process.

58

Miguel García-Sánchez, Jorge Daniel Montalvo, and Mitchell A. Seligson, Cultura política de la democracia en Colombia, 2015: Actitudes democráticas en zonas de consolidación territorial (Bogotá: United States Agency for International Development, 2015); Ana Arjona, “War Dynamics and the ‘No’ Vote in the Colombian Referendum,” Political Violence @ a Glance, October 20, 2016, http://politicalviolenceataglance.org/2016/10/20/war-dynamics-and-the-no-vote-in-the-colombian-referendum/; and Michael Weintraub, “¿Qué Pasó?” Foco Económico, October 4, 2016, http://focoeconomico.org/2016/10/04/que-paso-colombia/.

59

Nicolas Liendo and Jessica Maves Braithwaite, “Un paso hacia la paz? Determinants of Colombian Attitudes toward the Peace Process,” working paper at the University of Arizona, 2016; and Weintraub, “¿Qué Pasó?” Liendo and Braithewaite also find that political preference overwhelmingly predicts support; beyond that dominant predictor, they find some support for education and religion.

60

See, for example, Thania Paffenholz, “Civil Society and Peace Negotiations: Beyond the Inclusion–Exclusion Dichotomy,” Negotiation Journal 30 (1) (2014): 69–91.

61

Indeed, since peace agreements under democratic regimes cannot escape from the popular vote, even if a plebiscite is not used, general elections may easily turn into a referendum on the settlement, and such inclusivity may help shape attitudes when they do arise.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license, which permits copying and redistributing the material in any medium or format for noncommercial purposes only. For a full description of the license, please visit https://creativecommons.org/licenses/by-nc/4.0/legalcode.