Considering concessions: A survey experiment on the Colombian peace process

Abstract

Designing peace agreements that can be signed and sustained can be difficult in civil conflict. Many recent cases of successful settlements include electoral provisions, often for rebel groups to participate as political parties. Engaging the electoral process, however, can also open the peace process to the population at large, potentially derailing a settlement or some of its provisions, perhaps especially those related to politics. In this paper, we examine popular support for peace processes, specific electoral provisions, and potential concessions that provide former rebels with protections, legitimacy, and power. Using a survey experiment in Colombia, we find that the peace process overall is more popular than its electoral provisions, and that rebel endorsement of the provisions further diminishes support. These results contribute to an explanation of why the 2016 Colombian plebiscite on the peace agreement failed and to an understanding of how design matters to agreement effectiveness.

Keywords

Colombia FARC peace agreement peace process post-conflict elections public support survey experiment

Introduction

Designing peace agreements that can be signed and sustained is difficult in civil wars (e.g. Collier, 2003). Many peace processes use elections to engage combatants as well as international actors over time to secure a stable settlement (e.g. Matanock, 2012, 2017a, b, 2018). While peace agreements are often considered elite deals, the population is also inherently involved.

We know little about popular support for peace agreements, however. Are plebiscites or other types of direct participation helpful in securing peace agreements? In this paper, we analyze attitudes toward peace agreements and their particular provisions, and, experimentally, we assess whether components viewed as concessions to rebels, and negative attitudes towards rebels based on their criminality, reduce support among average citizens.

We posit that, while a peace process may be popular, citizens object to providing for rebel participation in elections and other political representation, especially if these provisions are framed as concessions. Indeed, these components often offer rebels protection, legitimacy, and power—which rebels require in many cases to agree to peace. Given that settlements are often complex, and these contexts tend to have high polarization and low information, citizens may especially rely on elite cues to assess the settlement, including what components are concessions and how costly they are. Reliance on these cognitive shortcuts, in combination with highly unpopular rebel groups, potentially creates opportunities for opposition leaders to promote public resistance. Overall, even though political provisions are needed to secure settlements, we posit that they may not be popular, especially when framed as concessions to unpopular rebels.

We examine these attitudes in the context of Colombia’s peace process, using 2015 survey data, including an endorsement experiment and a prime experiment that we designed. Colombia offers an ideal setting to study these questions. The peace process, under negotiation when the study was fielded, was set to be approved by popular participation. It also included provisions for the unpopular FARC to participate as a political movement and increased political representation for areas most affected by the conflict, allowing us to test our argument.

The results suggest several important lessons for Colombia and potentially other peace processes, particularly those with electoral provisions and rebel criminality. First, support for the peace process overall is higher than support for its components, including electoral provisions.

Second, an endorsement by the FARC diminishes support for provisions, even in communities that would benefit from these provisions, for example through additional political representation. Specifically, when the FARC endorses a proposal to allocate special seats in Congress to conflicted areas—a proposal on which both sides’ negotiators agreed before the survey was fielded—support decreases from 44 to 31%. We interpret the endorsement of the proposal by the FARC to convey information about a complex proposal.

Third, many Colombians, especially those victimized by the conflict, link the FARC to criminality and have formed negative attitudes toward the rebels. Reminding respondents of FARC involvement in drug-trafficking therefore has no effect on victims; it does, however, have a negative and statistically significant effect among non-victims, the population we expect to be susceptible to such informational cues. Finally, these tests also indicate that certain subsets of the population respond differently to messaging about the peace process.

This paper has implications for how to build peace both in Colombia and in other countries, as well as how to design settlements in civil conflict and beyond. We discuss these implications in the conclusion.

How do elections fit into effective peace processes?

Civil conflicts now often occur in countries holding elections. Democratic countries, such as India, Kenya, and Turkey, are battling insurgencies (according to the latest available democracy data; see Marshall et al., 2002). Many of these conflicted countries with democratic leanings continue to contest power through elections.

Elections and peace processes have become intertwined. Peace agreements increasingly include provisions for combatant parties to participate in elections, for example, and these provisions are shown to help stabilize peace among elites (e.g. Matanock, 2012, 2017a, b, 2018).

More democratic states may consider including voters in peace processes that have typically included only elites (e.g. Barnes, 2002). Referenda were only used after 1996 and remain rare.¹ However, the Colombian case served as trial for a broader voter involvement (e.g. Monitor’s Editorial Board, 2016). Demand for such a mechanism often exists—in the Colombian case, for example, about 66% supported a referendum to approve the peace process when it was proposed (García-Sánchez et al., 2014)—and using such a mechanism therefore can improve the legitimacy of a settlement (Nilsson, 2012). A government may also attempt to use a referendum to take a more centrist line in negotiations, anticipating that the popular vote will overcome the objections of rebels and other radicals (Matanock and García-Sánchez, 2017). The population may have its own preferences about such a peace agreement, however, and fail to serve in roles such as being a counterweight against hardliners.

Despite the pacifying effects for elites of electoral participation provisions,² including voters in these processes comes with the substantial potential cost that the population can reject or subvert a peace process. With direct involvement in the approval process, potentially through a referendum, voters can reject a settlement. In Colombia, for instance, voters did just that in the 2016 plebiscite. In Guatemala, also, after the 1996 peace process, voters rejected core components that would have changed the societal status quo when they rejected constitutional reforms in a national referendum (Seligson, 2005). Even without an opportunity to vote directly on the peace agreement, or aspects of it, the broader population can subvert provisions later by voting against politicians who supported them, for example, or by even endorsing violations of the peace agreement.

Understanding attitudes toward peace processes, their specific components and, especially, components viewed as concessions, is therefore important. Overall, we posit that, especially following a long civil war such as the Colombian conflict that has lasted for more than five decades, voters are likely to be supportive of ending the fighting. These contexts may be “ripe” for a settlement in particular (Zartman, 1985). While polarization may affect support for a particular settlement, polling the population on peace should produce the highest rates of support.

We posit that voters are less likely to support the specific components of a settlement. In many cases, voters will not want provisions that offer protection, legitimacy, or power to rebel groups, even though peace agreements inherently require some compromise. Combatants negotiating peace agreements expect terms that offer the same benefits as they could secure from fighting (Fearon, 1995; Reiter, 2003; Powell, 2002). Changing de jure power to match de facto power will inevitably involve some provisions that benefit the rebels. Voters’ attitudes toward approval may be influenced by how they feel about particular provisions: even if a peace process has popular support—as a promise to end civil war—these necessary provisions may not. Provisions that allow former rebels to participate in elections or even gain more representation in the areas of their influence are perhaps most associated with benefits for the rebels.³ Hence, these provisions should be less popular than a process for peace overall, even if the provisions are necessary to produce peace.

Finally, we posit that voters are even less likely to support provisions that are framed as concessions to the rebels—and especially if the rebels are involved in criminality or are otherwise especially unpopular. We posit that many voters are then likely to view provisions that rebel groups endorse as concessions and therefore costly to those who have sided with the state during the conflict. Rebel groups may also be especially unpopular. The blame attributed to the rebel group depends to some extent on how much the population blames them for the fighting and credits them for taking steps toward peace. Rebel groups, however, may also use tactics that make them more or less appealing. While kidnapping and targeting civilians indiscriminately will almost certainly lessen their support, other behaviors may or may not have similar effects. One such behavior is involvement in criminal activity for funding, including illegal mining in cases such as Angola and Sierra Leone and drug trafficking in Afghanistan and Colombia. Citizens in the communities close to rebel camps may benefit from goods and services provided with funds from these activities, and some may also rely on criminal activity for their own livelihood, potentially making them more sympathetic (Matanock and García-Sánchez, 2018). Others, however, may be unsympathetic to perceived profit-seeking.

Based on these intuitions, our theoretical expectations are that respondents should support particular provisions, especially electoral provisions, less than an overall peace process (tested observationally); support should be even lower if a rebel group endorses these provisions (tested experimentally); and, support should be lowest for the group, and for the provisions when endorsed, if they are reminded that the rebel group relied on criminal activity for funding (also tested experimentally).

Assessing voter attitudes toward aspects of the peace process, and the messaging around it, also fits into a larger theory about direct popular involvement. If provisions are unpopular, especially when seen as concessions (Assouline and Trager, 2017), the referendum process may produce special opportunities for opponents to use those sentiments to reject or subvert it. Previous research has shown that under some conditions—especially in contexts of low information and high polarization, which are often characteristic of countries with civil conflicts—direct democracy can actually inhibit deliberation (e.g. LeDuc, 2015). In these contexts, citizens do not have incentives to study complex settlements, and instead form their preferences based on endorsements or other types of cues available in the political environment (e.g. Lupia and McCubbins, 1998). Opponents then can capitalize on voters’ limited knowledge and polarization on the issue. Citizens with low levels of political awareness, in particular, may be susceptible to messaging such as endorsements by various actors.⁴

Prior work has shown some evidence of the elite influence on voter attitudes over time during the recent peace process in Colombia, where opponents, in part, were able to frame the provisions as concessions to rebels at the cost of a fair settlement (Matanock and García-Sánchez, 2017). In Guatemala’s referendum, too, observers attributed the failure to approve constitutional changes in part to opponents of the peace process who ran a considerable “No” campaign that won significant support and voter ignorance of the complexities of the issues (e.g. Stanley and Holiday, 2002: 21). The elite messaging in these campaigns appears to appeal to concessions and other cuing tactics that may be effective on voters in these contexts—voter attitudes on these topics, with similar framing, is therefore the topic of this paper.

The Colombian context

The Colombian case, prior to the plebiscite when our study was conducted, represented a unique opportunity to test attitudes about the peace process.

The FARC and drug-trafficking

The conflict, described in more depth in the introduction to this special issue, featured a very unpopular rebel group. Drug trafficking became central in the enduring conflict, and the FARC has benefited from it (Sánchez and Palau, 2006: 7–8). While the cocaine trade was generally seen as profit-seeking, and thus not suitable for an ideological group, it did provide some resources to communities, and many individuals in those communities benefited from the trade, so it was not as completely condemned as kidnapping, for example (Enrique Flórez, personal communication with one of the authors, 2010; Tickner et al., 2011).

The peace process and its electoral provisions

When the FARC and the Colombian government initiated the latest round of peace talks in 2012, they agreed to a “road-map” that established six points to be negotiated, including one on political participation. In 2013, the negotiating team provided for FARC electoral participation as a legal political party.⁵ They also agreed on the creation of “Special Transitory Peace Districts” to reserve special seats in Congress for areas most affected by conflict.⁶ These special districts would have potentially given the FARC more representation in Congress because, despite potential polarization in these areas, these included FARC strongholds and presumably its social bases (e.g. Tickner et al., 2011).

In January 2013, as negotiations progressed with the FARC, Santos proposed a plebiscite to approve a prospective settlement. This proposal differed from constituent assemblies used in the past in Colombia to approve deals. During the negotiations, factions had emerged among the elites on the government side: the Santos administration led, and strongly backed, the peace negotiations, while the popular former president, Álvaro Uribe, opposed it (e.g. Matanock and García-Sánchez, 2017). The plebiscite may have provided Santos with some leverage against the FARC because it made clear that any settlement would have to satisfy voters; it may have also been an attempt by Santos to overcome Uribe-affiliated hardliners (e.g. Matanock and García-Sánchez, 2017). By March and April 2015, when our data was collected, however, a plebiscite was not yet finalized.

After our survey, the ratification of the peace agreement through a plebiscite was approved. Once the combatants signed the 2016 September final agreement, the October vote asked about approval of the peace agreement as a whole. The plebiscite was rejected by a narrow margin of 0.43 percentage points.

These results surprised many, including the government, but also polling firms and some studying the process, who had expected a large margin of approval (e.g. Blu Radio, 2016). A revised settlement was ultimately approved by Congress in November 2016. However, while a resounding acceptance might have provided needed legitimacy to the peace process and overridden the opposition, a rejection added uncertainty (which may have also been the case with any close vote; Matanock and García-Sánchez 2017). We therefore examine survey evidence in this case, prior to the plebiscite, to test our theory about voter attitudes. The results also speak to whether holding referenda more broadly is risky, potentially allowing opponents an opportunity to undermine a settlement.

Design of the survey experiment

The sample

To study attitudes toward the peace process, we conducted two experiments embedded in the Colombia 2015 Latin American Public Opinion Project (LAPOP) survey. The survey sampled respondents from 59 municipalities prioritized by USAID,⁷ which are part of the government’s “National Consolidation and Reconstruction Policy” program,⁸ producing results representative of these municipalities.⁹ This sample suited our project because the oversampled areas most affected by the conflict also stood to benefit most from the settlement provisions we studied. We thus have a hard test of whether endorsement by the FARC can override support for potentially beneficial policies emerging from settlement provisions.

The sample consisted of 1390 individuals. Fieldwork took place in March and April 2015. All surveys were conducted face-to-face using tablets in most cases (unless the tablets posed security concerns). All questionnaires were administered in Spanish.¹⁰ Vanderbilt University and the Observatorio de la Democracia—a research center at the University of Los Andes—coordinated the survey, which was executed by the Centro Nacional de Consultoría, a well-known local polling firm.

The effect of attitudes about the FARC on the settlement

We first examine observational data on attitudes toward a settlement, specific provisions, and the FARC. We then use an endorsement experiment to assess the effect of attitudes toward the FARC on the peace agreement.

The peace process proposed the creation of special seats reserved for electoral districts in areas most affected by the civil conflict, as noted, based on the agreement of both negotiating sides when our survey was conducted.

Individuals in the control group were asked:

Some people have proposed to reserve seats in Congress for some of the regions that have been more affected by the armed conflict, with the purpose of making these regions have more representation in the Congress.

To what extent do you agree or disagree?

The treatment group received the same but beginning:

The FARC have proposed …

The only difference is the FARC endorsement for these special electoral districts.¹¹ So, with this design, we can measure attitudes toward the FARC by simply comparing the average support across experimental groups.

Endorsement experiments can be plausibly interpreted in two ways. First, differences across group averages may reflect social desirability bias stemming from individuals’ unwillingness to reveal their attitudes toward the FARC. This interpretation is the one that has recently been used in the conflict literature (e.g. Bullock et al., 2011;Lyall et al., 2013; Matanock and García-Sánchez, 2018). For these purposes, the policy proposal should be well understood by the population and not likely to have partisan effects. The experiment would then allow a fearful population to reveal its preferences.

We propose an alternative interpretation, however, when the policy does not meet these specifications: if citizens are uncertain about the policy, and they suspect that it will benefit one side, endorsements by controversial actors may convey information. Given the limited information most citizens had about the contents of the peace agreement in Colombia—especially in the periphery of the country (i.e. our subject pool)—we theorize that the FARC’s endorsement would provide voters with information about the policy, communicating potential beneficiaries and outcomes (Kuklinski and Quirk, 2000).¹²

The effect of emphasizing the link between the FARC and illegal activities

As discussed in previous sections, we also want to test whether a possible source of negative attitudes about the FARC stems from stronger associations with illegal activities. While multiple factors could shape individual attitudes (e.g. the group’s tactics or ideological platform), we focus on whether the links between the FARC and criminality permeate and harm public support. Little theory exists on citizen attitudes toward the criminal activities of rebel groups. While other behaviors may be clearly negative, this behavior may provide public goods to communities close to rebel camps, for example, and even employment for individuals in those communities (e.g. Tickner et al., 2011). Illegal activity is also a common source of funding for other rebel groups. We suspect that it may have a delegitimizing effect but want to test this. Thus, in our second experiment, prior to asking questions about support for electoral provisions for the FARC, a randomized treatment group was primed about the FARC’s ties to drug-trafficking.

Individuals in the treatment group were shown the following vignette:

The FARC have been present in rural areas in the country for more than 50 years, and they have been involved in growing, producing and commercializing illicit drugs in Colombia.

The control group only saw the non-italicized first part. The only source of variation between treatment and control is the additional information. We can therefore assess the effect of priming respondents on an association between drug-trafficking and the FARC by comparing the average support across experimental groups.

Outcome variables

We are interested in measuring outcomes related to provisions included in the current peace process. Thus, after our prime experiment, we conduct the endorsement experiment explained above, but we also ask all respondents about electoral provisions. Specifically, we ask respondents whether they would agree with the following peace agreement provisions:¹³

creating new special seats in the Congress designated for political parties created by demobilized members of the FARC;

once demobilized, allowing the FARC to form a political party to run for seats in Congress, through elections.

Randomization

Randomization was done at the individual level, blocking on gender, age group, and municipality size to obtain balance on these variables. We divided our sample into four “experimental groups” resulting from the combinations of the two treatments just described (corresponding to the endorsement and prime experiments). Table 1 summarizes the size of each one of these experimental groups according to the initial randomization.

Table 1.

Matrix of treatments and size of experimental groups

	Endorsement	No endorsement
Prime	Group 1, N = 367	Group 2, N = 366
No prime	Group 3, N = 385	Group 4, N = 382

Empirical strategy

To calculate attitudes toward the peace process, its particular provisions, and what influence the FARC has on these attitudes, especially when respondents consider the FARC’s ties to drug-trafficking, we first provide summary statistics on basic support and trust questions. Next, we analyze each experiment separately.

For the endorsement experiment, we test whether support is affected by the endorser of the proposal, using the following equation:

\hat{τ} = {\bar{Y}}_{e = 1} - {\bar{Y}}_{e = 0}

(1)

where ${\bar{Y}}_{e = 1}$ is the average response of the treatment group for the endorsement question, and ${\bar{Y}}_{e = 0}$ is the average response in the control group.

For the prime experiment, we compare the means of the treatment and control groups for our main outcomes: support for participation and allocated seats provisions. Thus, the effect will be estimated by the following equation:

\hat{τ}_{m} = {\bar{Y}}_{m, p = 1} - {\bar{Y}}_{m, p = 0}

(2)

where ${\bar{Y}}_{m, p = 1}$ is the average response in the treatment group for outcome variable m. Likewise, ${\bar{Y}}_{m, p = 0}$ is the average response in the control group for this same variable. Our main outcome variable, m, corresponds to the questions presented in the previous section.

We also estimate whether the change induced by FARC endorsement of a provision is bigger when individuals are primed on the FARC’s involvement in drug trafficking. For this purpose, we use a difference-in-differences estimator according to the following equation:

\hat{τ}_{DiD} = ({\bar{Y}}_{e = 1, p = 1} - {\bar{Y}}_{e = 0, p = 1}) - ({\bar{Y}}_{e = 1, p = 0} - {\bar{Y}}_{e = 0, p = 0})

(3)

where e = 1 and p = 1 are treatment group indicators for the endorsement and prime experiments. $\bar{Y}$ is the average response to our outcome of interest, i.e. support for seats in Congress for municipalities most affected by the civil conflict.

All of our models are estimated using ordinary least squares (OLS) with robust standard errors to allow for different variances across experimental groups.¹⁴ Thus, the coefficient accompanying the treatment indicator (or their interaction) corresponds to the difference in means estimator in equation (2) (difference-in-differences estimator in equation 3; see Gerber and Green, 2012).

Attitudes about the peace process, particular provisions and the FARC

Initial evidence suggests that the peace process enjoys substantial support, but electoral provisions that potentially offer the FARC protection, legitimacy, and power do not. For instance, when asked about the best method of conflict termination, most supported a settlement over any other option (77%); in contrast, few supported military force (13.5%) (García-Sánchez et al., 2015). As expected, the promise of putting an end to more than 50 years of fighting that have left 6 million victims should result in high levels of public support for the overall peace agreement.

When asked about settlement provisions, Colombians were more tolerant of those that do not involve political representation or participation by the FARC. Public support for concessions from the government to the FARC, in particular, varies widely by type: a reduction in the sentences to rank and file FARC members received 3.82 on a 1–7 scale, where 7 is “a lot” of support. In contrast, allowing demobilized members of the FARC to participate in politics received 3.07 and providing seats in Congress for FARC electoral participation received just 2.92. In general, citizens seem more willing to provide economic benefits and even special forms of justice to the FARC, rather than to ease their entrance into the political arena.¹⁵

Low levels of support for the FARC may help to explain attitudes about these particular concessions that confer political benefits. As Figure 1 shows, trust for the FARC itself is close to “none at all”¹⁶ (contrasted in this graph with overall support for the settlement).¹⁷

Figure 1.

Average support for the peace process and the FARC.

This descriptive evidence confirms our expectations from the previous section. That is, according to the 2015 LAPOP data, Colombians seemed to broadly support the peace process, but not specific political provisions, perhaps particularly as concessions to the unpopular FARC.

Results of the endorsement experiment

The endorsement experiment shows the effect of attitudes toward the FARC on the proposal to provide special seats to areas most affected by the conflict. When the control group assesses the proposal (see Figure 2), its support is almost a point lower than support for the peace process as a whole. It is, however, still close to the level of support for the most trusted institution in Colombia, the Catholic Church (see the Online Appendix). When the proposal is endorsed by the FARC, support is much lower still: the treatment group reports 0.74 points less support on a seven-point scale, bringing it under the midpoint (again, see Figure 2; also see the first column of Table 2).

Figure 2.

Support for the creation of special seats for areas most affected by conflict.

Table 2.

Heterogeneous treatment effects of the endorsement experiment

Variables	(1)Support	(2)Area	(3)Sex	(4)Victimization
Endorsement experiment	−0.740***	−0.754***	−0.511***	−0.644***
	(0.109)	(0.139)	(0.153)	(0.167)
Rural		−0.300*
		(0.161)
Rural × Endorsement		0.0370
		(0.222)
Woman			−0.242
			(0.155)
Woman × Endorsement			−0.457**
			(0.216)
Victim of conflict				0.335** (0.158)
Victim × Endorsement				−0.158 (0.220)
Constant	4.138***	4.255***	4.255***	3.936***
	(0.078)	(0.097)	(0.107)	(0.121)
Observations	1292	1292	1292	1290
R ²	0.035	0.039	0.052	0.039

Robust standard errors in parentheses.

***

p < 0.01; **p < 0.05; *p < 0.1.

Note: The dependent variable is coded on the same 1–7 scale.

Considering how to interpret these results, voters appear to be learning about the policy from the FARC endorsing it (e.g. that the provisions likely provide the rebels with benefits that can be construed as concessions) and then incorporating their attitudes toward the group into their assessment of this policy. Consistent with this interpretation, we find that the endorsement experiment is “radicalizing” responses: individuals in the control condition are more indifferent with respect to this policy, while individuals in the treatment show significantly more extreme responses (see Figure 3). Moreover, our findings are not consistent with the presence of social desirability bias for two reasons. First, if respondents were afraid of revealing their true attitudes toward the FARC, we should observe higher levels of trust in the direct question; instead we observe low levels (as depicted in Figure 1). Second, if such low levels reflect citizens’ unwillingness to report support, then we should expect higher averages in the treatment group; in contrast, we find a negative and significant effect of the endorsement experiment.

Figure 3.

Frequency of responses by treatment condition.

Overall, these findings are consistent with the endorsement conveying information to voters, which, in turn, is being used to assess their levels of support to the proposal. The results suggest that negative attitudes toward the FARC spill over into support for particular provisions, even among groups that would directly benefit from the policy.¹⁸

Heterogeneous treatment effects for the endorsement experiment

We also investigate whether the endorsement treatment has heterogeneous effects across socio-demographic characteristics that include gender, age, ideology,¹⁹ and urban vs rural areas of residency, as well as whether the respondent has been a victim of an attack by the FARC. These are cleavages along which support for peace processes often changes. First, existing work on conflict has shown that gender (female), age (older), and ideology (left) are generally different in terms of attitudes toward conflict (less supportive; e.g. Fair and Shepherd, 2006). Second, urban and rural regions in Colombia have been differently affected by the conflict, with most of the insurgency and cocaine trade having long been largely rural (Arias et al., 2014). Finally, past victimization should change respondents’ attitudes toward a peace process, although whether it makes them more or less conciliatory is debated (e.g. Hazlett, 2013).

We test for the presence of this heterogeneity by estimating the following model using OLS:²⁰

Y_{i} = β_{0} + β_{1} T_{i} + β_{2} D_{i} + β_{3} T_{i} D_{i} + ε_{i}

(4)

where $Y_{i}$ corresponds to individual i’s response to the endorsement question, $T_{i}$ is a treatment indicator for the endorsement experiment and, finally, $D_{i}$ is the subgroup variable under analysis that can be any one of these indicators. According to this specification, our coefficients of interest are $β_{1}$ and $β_{3}$ .

The second to fourth columns of the Table 2 present the results for the estimation of heterogeneous treatment effects. First, while the treatment effect shows no statistically significant variation by area of residency, its sign indicates that the negative effect of the FARC endorsement is attenuated in rural areas. Intuitively, this result is consistent because rural areas have been the most affected by the conflict and would benefit most from this provision.

Next, the negative effect of our treatment is larger for women than for men: among the latter, being treated reduces support for the proposal by 0.511 points,²¹ while among the former, it reduces it by almost one complete scale-point.

Finally, considering victimization, on one hand, the coefficient for having been a victim is positive and also statistically significant. This indicates that victims of conflict are more likely to support this provision. On the other hand, for both victims and non-victims, the treatment effect has a negative and statistically significant effect on public support for the proposal. From a statistical point of view, those victimized do not respond any differently to the FARC endorsement than individuals who were not victimized, but the sign of the interaction term is consistent with the idea that victims may have a larger bias against the FARC.²² These results should be taken cautiously given that we cannot directly manipulate these attributes (i.e. being victimized).

Results of the prime experiment

The prime experiment shows the direction of the overall effect that we expected—the reminder of the associations between drug-trafficking and the FARC, on average, reduces support for these provisions (see Table 3, showing the results from the difference-in-means estimators)—but it is not statistically significant.²³

Table 3.

Results of the prime experiment

Variables	(1)Seats for FARC	(2)FARC Party
Prime treatment	−0.092 (0.102)	−0.004 (0.104)
Constant	2.808*** (0.0711)	2.810*** (0.072)
Observations	1279	1304
R ²	0.001	0.000

Robust standard errors in parentheses.

***

p < 0.01; **p < 0.05; *p < 0.1.

Note: The dependent variables are coded on the same 1–7 scale.

How should we interpret the negative but not statistically significant effect of the prime on individuals’ support for these electoral provisions? The effect could suggest the presence of a floor effect, where the effect may be so small, and statistically null, because respondents already hold a very low opinion of the FARC, so it cannot drop much lower. Figure 1 shows evidence consistent with this interpretation.

Another possible interpretation is that the experiment simply did not have sufficient power to pick up a statistically significant effect. However, ex-post power calculations (see the Online Appendix) indicate that our current sample should have had sufficient power (80%) to detect even a very small effect of the prime experiment (∼0.3 on a 1–7 scale).²⁴ From a substantive point of view, then, any treatment effect that we could not detect would be close to zero.

Beyond the floor effect, however, another component of the interpretation of these results could be that some individuals react to the prime, whereas others do not, and perhaps these groups are systematically distinguishable. We explore this possibility by sub-setting our sample by prior victimization in the conflict.²⁵ Victims of conflict may be more sensitive and socialized to the ongoing conflict, or simply more knowledgeable about its actors. In either case, the observational implication is that prior victimization should make respondents less reactive to the prime. Table 4 summarizes the results for the estimation of effect on attitudes toward the creation of special seats for the FARC.

Table 4.

Results of the prime experiment by victimization status

Outcome	Victim	Meancontrol	Meantreatment	Difference	SEdifference	p-value,two-tailed	p-value,upper tail	p-value,lower tail
Specialseats	No	2.806	2.488	−0.318	0.156	0.042	0.021	0.979
Specialseats	Yes	2.811	2.875	0.064	0.135	0.638	0.681	0.319
FARC Party	No	2.760	2.547	−0.214	0.155	0.169	0.084	0.916
FARC Party	Yes	2.843	2.989	0.146	0.139	0.294	0.853	0.147

As expected, these results indicate that the effect of the prime experiment is driven by those not victimized, where the prime operates as expected. These findings suggest that victims of conflict may already link the FARC and drug-trafficking, so the prime does not shape their attitudes. Others, however, oppose the provisions even more stringently once we elicit linkages between the FARC and this criminal activity (and they are already less supportive of electoral participation provisions).

Results of the prime and endorsement experiments

Finally, we test whether the effect of reduced support based on FARC endorsement is stronger for those primed on links between drug-trafficking and the FARC. The results from the difference-in-differences estimator are presented in Figure 4 (also see the Online Appendix for the formal test for the difference in differences). Being primed about the FARC’s involvement in drug-trafficking and receiving the FARC endorsement reduces average support by 0.16 points for the provision regarding special seats in Congress for areas most affected by conflict; however, this difference is not statistically significant (standard error, SE ≈ 0.22). Notice that within the prime treatment group, respondents who received the endorsement treatment showed an average lower support for the proposal (3.36) than those in the control group (4.18), and the difference between these two figures is 0.84 (with SE ≈ 0.16). Within the control group for the prime experiment, this difference is 0.66 (SE ≈ 0.15). While the signs and magnitudes of the coefficients are consistent with our main hypotheses, only the effect of the endorsement experiment shows statistical significance.

Figure 4.

Support for the creation of special seats for areas most affected by conflict with the FARC endorsement and drug-trafficking prime.

Conclusions

This paper explores attitudes toward the peace process to better understand the Colombian case, including the 2016 rejection of the plebiscite, but also settlements across contexts. In Colombia, our findings suggest that citizens were enthusiastic about peace, but that they did not want to explicitly reward rebel groups with concessions, especially if they viewed rebel groups negatively. Moreover, our findings indicate that endorsements, and even primes under particular circumstances, provide information to average citizens about the settlement, which is a complex issue to evaluate. These effects, however, are attenuated for certain populations that are likely better informed on related political issues (e.g. victims).

Our research indicates that the cues, and claims about concessions to the rebels, may have helped the opposition effectively make a case against a peace process that was, in theory, popular. The concessions from the government to the FARC, including electoral participation provisions, were the result of four years of negotiation between the two sides—and they were likely necessary to reach a settlement—but many citizens remained uncertain about the content of these negotiations,²⁶ and even when the settlement was final, most did not fully assess the causes and consequences of each component of the complex policy.²⁷ Low levels of information, followed by direct voter inclusion in the approval process through the plebiscite, may have led citizens to lean on elites’ cues. The plebiscite became a battle of narratives between divided elites, in which attitudes shifted over time as the former president (Uribe) split from the current president (Santos; again, e.g. Matanock and García-Sánchez, 2017). These results suggest a possible explanation for at least part of the failure of the 2016 plebiscite: by choosing a plebiscite, likely in part as an attempt to overcome elite divisions, the government actually left substantial opportunities for opponents to undermine it.

The Colombian government limited the plebiscite to a single vote on an entire agreement, the most popular part, rather than each component. However, it could have also used targeted messaging campaigns that highlighted the benefits of the settlement, and even its specific components, for particular groups of Colombians. It might also have limited the vote to victims or those in areas most affected by the conflict. The government might also have studied attitudes toward particular concessions, and likely messaging around these, when negotiating the settlement as well.

Our results also indicate possible hurdles to implementation of the peace process. While a revised settlement that took into account the attitudes of more of the leaders of the factions in Colombia was ultimately approved by Congress in November 2016, citizen resistance to some of the concessions included in the settlement may still be present. Indeed, post-plebiscite Colombia is still characterized by elite polarization (Matanock and García-Sánchez, 2017), and so splits among the elites during this phase may produce endorsements and other types of cues that can still derail aspects of the deal. Thus, the lessons about concessions and messaging hold in the post-plebiscite period.

This paper has a series of implications for other conflicts. Lessons from the FARC in Colombia may apply to potential peace processes in other cases, perhaps especially those with criminalized rebel groups like the Taliban in Afghanistan. Our findings are important to understanding how to design settlements. Settling a civil conflict through a peace process that sticks is a difficult proposition. A transformation in the dominant approach to crafting peace agreements to terminate civil conflict has meant that many cases involve electoral provisions (Matanock, 2018). Participation provisions, for example, have been demonstrated to produce more enduring peace (Matanock, 2017a). Engaging the electoral process as part of the peace process, however, may also mean engaging average citizens in its approval, especially in more democratic contexts with regular elections.

All settlements require concessions to an opponent, as we discuss, but this study suggests that voters may also take the acceptance of the settlement by the opponents as a signal of how much was conceded. Such a dynamic may apply even outside of civil conflicts: for example, in the 2015 Joint Comprehensive Plan of Action signed with Iran, similar recriminations were experienced during the 2016 elections campaigns in the USA. This paper therefore also speaks to the broader difficulties of signing deals in democracies.

Supplemental Material

analysis_colombia_pp@7 – Supplemental material for Considering concessions: A survey experiment on the Colombian peace process

Supplemental material, analysis_colombia_pp@7 for Considering concessions: A survey experiment on the Colombian peace process by Aila M Matanock and Natalia Garbiras-Díaz in Conflict Management and Peace Science

Supplemental Material

Colombia_Sobremuestra_v5 – Supplemental material for Considering concessions: A survey experiment on the Colombian peace process

Supplemental material, Colombia_Sobremuestra_v5 for Considering concessions: A survey experiment on the Colombian peace process by Aila M Matanock and Natalia Garbiras-Díaz in Conflict Management and Peace Science

Supplemental Material

colombia-electoralprovisions-onlineappendix_05.26.2018 – Supplemental material for Considering concessions: A survey experiment on the Colombian peace process

Supplemental material, colombia-electoralprovisions-onlineappendix_05.26.2018 for Considering concessions: A survey experiment on the Colombian peace process by Aila M Matanock and Natalia Garbiras-Díaz in Conflict Management and Peace Science

Footnotes

Acknowledgements

We are very grateful to the Latin American Public Opinion Project (LAPOP) hosted by Vanderbilt University for allowing us include these experimental questions on their 2015 Colombia survey. We especially want to thank Miguel García-Sánchez and appreciate his valuable suggestions. We also received excellent comments from the reviewers and editors of this issue and journal, participants at the Electoral Violence Conference at the University of California, Berkeley, the Conflict Dynamics and Challenges to Building a Resilient Peace in Colombia Conference at the Universidad del Rosario, the Center on Democracy, Development, and the Rule of Law Research Seminar at Stanford University, the Seguridad, Cuidadanía, y Violencia en América Latina Seminar at ITAM, the University of California’s Conference on International Cooperation, and the Folke Bernadotte Academy’ Peacebuilding after Armed Conflict workshop. This project has been preregistered on EGAP (no. 20150518AA). We relied on an Inter-Institutional Agreement signed between Vanderbilt University and the University of California, Berkeley (Vanderbilt’s IRB Protocol no. 110627).

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

Álvarez Vanegas

Garzón

(2016) El país que develó la ventaja del NO. Report, Fundación Ideas para la paz, Bogotá, Colombia. Available at: http://www.ideaspaz.org/publications/posts/1411.

Arias

Camacho

Ibáñez

Mejía

Rodríguez

(2014) Costos económicos y sociales del conflicto en Colombia: ¿cómo construir un posconflicto sostenible?Ediciones Uniandes-Universidad de los Andes.

Arjona

(2016) War dynamics and the “no” vote in the Colombian referendum. In: Political Violence at Glance. Available at: https://politicalviolenceataglance.org/2016/10/20/war-dynamics-and-the-no-vote-in-the-colombian-referendum/

Arriola

Choi

Gichohi

(n.d) Political endorsements and cross-ethnic voting in Africa. Working Paper. Available at: https://www.dhdannychoi.com/uploads/2/6/1/3/26136995/acg_endorsement.pdf

Assouline

Trager

(2017) Concessions for concessions sake: Injustice, indignation and the construction of intractable conflict in Israel–Palestine. Available at: http://www.roberttrager.com/Research_files/BalancingTheScalesMPSA.pdf

Barnes

(2002) Owning the process: Mechanisms for political participation of the public in peacemaking. ACCORD Project: International Review of Peace Initiatives 13.

Blu Radio (2016) Colombianos votaron con emociones y es difícil medir eso: responden encuestadoras. Available at: https://www.bluradio.com/nacion/colombianos-votaron-con-emociones-y-es-dificil-medir-eso-responden-encuestadoras-118252

Bullock

Imai

Shapiro

(2011) Statistical analysis of endorsement experiments: Measuring support for militant groups in Pakistan. Political Analysis 19(4), 363–384.

Collier

(2003) Breaking the conflict trap: Civil war and development policy. Washington, DC: World Bank Publications.

10.

Fair

Shepherd

(2006) Who supports terrorism? Evidence from fourteen Muslim countries. Studies in Conflict and Terrorism, 29(1), 51–74.

11.

Fearon

(1995) Rationalist explanations for war. International Organization 49(3): 379–414.

12.

García-Sánchez

Rodríguez-Raga

Seligson

Zeichmeister

(2014) Cultura Política de la Democracia en Colombia, 2014. Dilemas de la democracia y desconfianza institucional en el marco del proceso de paz. Observatorio de la Democracia. Bogotá, Colombia. LAPOP Vanderbilt University.

13.

García-Sánchez

Montalvo

Seligson

(2015) Cultura política de la democracia en Colombia, 2015. Actitudes democráticas en zonas de consolidación territorial. Observatorio de la Democracia. Bogotá, Colombia. LAPOP Vanderbilt University.

14.

Gerber

Green

(2012) Field Experiments: Design, Analysis, and Interpretation. London: WW Norton.

15.

Harbom

Högbladh

Wallensteen

(2006) Armed conflict and peace agreements. Journal of Peace Research 43(5), 617–631.

16.

Hazlett

(2013) Angry or weary? The effect of physical violence on attitudes towards peace in Darfur. Working paper. University of California, Los Angeles, CA.

17.

Kuklinski

Quirk

(2000) Reconsidering the rational public: Cognition, heuristics, and mass opinion. In: Elements of Reason: Cognition, Choice, and the Bounds of Rationality. New York: Cambridge University Press, pp. 153–182.

18.

LeDuc

(2015) Referendums and deliberative democracy. Electoral Studies 38: 139–148.

19.

Lupia

McCubbins

(1998) The Democratic Dilemma: Can Citizens Learn what they Need to Know?Cambridge: Cambridge University Press.

20.

Lyall

Blair

Imai

(2013) Explaining support for combatants during wartime: A survey experiment in Afghanistan. American Political Science Review 107(4): 679–705.

21.

Marshall

Jaggers

Gurr

(2002) Polity IV Project: Political Regime Characteristics and Transitions, 1800–2009. College Park, MD: University of Maryland.

22.

Matanock

(2012) International insurance: Why militant groups and governments compete with ballots instead of bullets. PhD dissertation, Stanford University.

23.

Matanock

(2017a) Bullets for ballots: Electoral participation provisions in peace agreements and conflict recurrence. International Security 41(4): 93–132.

24.

Matanock

(2017b) Electing Peace: From Civil Conflict to Political Participation. Cambridge: Cambridge University Press.

25.

Matanock

(2018) External engagement: Explaining the spread of electoral participation provisions in civil conflict settlements. International Studies Quarterly 62(3): 656–670.

26.

Matanock

García-Sánchez

(2018) Does counterinsurgent success match social support? Evidence from a survey experiment in Colombia. The Journal of Politics 80(3): 800–814.

27.

Matanock

García-Sánchez

(2017). The Colombian paradox: Peace processes, elite divisions and popular plebiscites. Daedalus 146(4): 152–166.

28.

Monitor’s Editorial Board (2016) Colombia creates peace before ending a war. The Monitor’s View. Available from: http://www.csmonitor.com/Commentary/the-monitors-view/2016/0622/Colombia-creates-peace-before-ending-a-war

29.

Nilsson

(2012) Anchoring the peace: Civil society actors in peace accords and durable peace. International Interactions 38(2), 243–266.

30.

Powell

(2002) Bargaining theory and international conflict. Annual Review of Political Science 5(1), 1–30.

31.

Reiter

(2003) Exploring the bargaining model of war. Perspectives on Politics 1(1), 27–43.

32.

Sánchez

Palau

(2006) Conflict, decentralisation and local governance in Colombia, 1974–2004. Bogotá: CEDE.

33.

Seligson

(2005) Democracy on ice: The multiple challenges of Guatemala’s peace process. In: The Third Wave of Democratization in Latin America: Advances and Setbacks. New York: Cambridge University Press, pp. 202–231.

34.

Stanley

Holiday

(2002) Broad participation, diffuse responsibility: Peace implementation in Guatemala. In: Ending Civil Wars. The Implementation of Peace Agreements. Boulder, CO: Lynne Rienner, pp. 421–462.

35.

Tickner

García

Arreaza

(2011) Actores violentos no estatales y narcotráfico en Colombia. In: Política Antidrogas en Colombia. Bogotá: Ediciones Uniandes, pp. 413–445.

36.

Zartman

(1985) Ripe for Resolution: Conflict and Intervention in Africa. New York: Oxford University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

15.90 MB

0.02 MB

0.26 MB