Who’s to Blame? Post-conflict Violence, Political Messaging, and Attitudes Towards Peace Agreements

Abstract

How do episodes of post-conflict violence affect public support for peace? I argue that political messaging about who or what is to blame can influence how violence affects attitudes towards peace agreements. I test this argument in Colombia, a country which has experienced violence after a 2016 peace agreement, and where rival political camps debate whether government failures or noncompliance by rebels is to blame. In an experiment with 1466 respondents in conflict and non-conflict zones, I paired news about post-conflict violence with information supporting these competing messages. I find that emphasizing rebel culpability reduced support for peace agreements, but emphasizing poor government implementation did not have a strong countervailing effect. A probe of the mechanisms suggests that while emphasizing rebel culpability increased perceptions that rebels alone were to blame, emphasizing government implementation failures led respondents to conclude that both parties were to blame, limiting the effectiveness of this message.

How do episodes of post-conflict violence affect public support for peace agreements? Past work suggests that violence during and after peace negotiations undermines civilians’ confidence that rebels can be trusted to abide by a peace deal (Kydd andWalter 2002), and thereby increases support for political parties that reject peace. Yet an extensive literature on public opinion formation suggests that the effects of material shocks are rarely uniform across the electorate – rather, voters rely on cues or messaging from political elites to interpret and respond to such events (Zaller 1992; Lupia, McCubbins Arthur et al. 1998). In this paper, I posit that during episodes of post-conflict violence, political messaging plays an important role in influencing public perceptions of blame.

I argue that how citizens react to post-conflict violence depends on their beliefs about who or what is to blame; beliefs that are based on information that is often ambiguous or politically charged. If post-conflict violence is perceived as a symptom of rebel noncompliance, such violence may undermine confidence in peace processes. By contrast, if post-conflict violence is perceived as a symptom of poor government implementation, then it may have a less deleterious effect – galvanising citizens to demand better implementation from their government. In this context, political messaging about blame can be highly effective in influencing attitudes towards a peace agreement.

I tested this argument using a survey experiment conducted in Colombia in 2022, a context in which post-conflict violence was escalating and blame for the violence was contested and politicized. I developed interventions that paired recent news about post-conflict violence with competing political messages and evaluated their effects on support for peace agreements among a sample of 1,466 respondents from a mix of conflict and non-conflict zones.

My results indicate that emphasizing rebel culpability for post-conflict violence decreased the public’s appetite for future peace deals. Pairing news about recent violence with information implicating rebels reduced support for future peace agreements, an effect that was strongest for respondents with the most at stake, such as residents of former conflict zones. Contrary to my expectations however, the treatment intended to divert blame to the government by emphasizing poor implementation failed to have a countervailing effect, a result which suggests that a key messaging strategy of supporters of peace may be ineffective. Respondents appear to have inferred from such information that neither the government nor the rebels could be trusted to comply with and implement the peace agreement.

These results build on a growing body of research emphasizing the pivotal role political elites and messaging can play in either sustaining or dismantling public support for peace deals (Matanock, García-Sánchez and Garbiras-Díaz 2020; Haas and Khadka 2020). I contribute to this literature by demonstrating how political messaging weaponizes and interacts with developments on the ground in post-conflict settings, and, in particular, episodes of post-conflict violence. My results suggest that political messaging can exacerbate the negative effects of violence on public support for peace agreements, but that counteracting these effects through messaging may be more difficult.

Literature

Prior research on the topic of ‘spoilers’ argues that violence can be an effective strategy for groups opposed to peace (Stedman 1997). Attacks perpetrated during or after an agreement’s negotiation signal to the public that rebels cannot be trusted to comply with the agreement and refrain from violence (Kydd and Walter 2002; Braithwaite, Foster, and Sobek 2010; Findley and Young 2015). As a consequence, public support for negotiated solutions declines.

This literature relies implicitly on a model of public opinion in which members of the public observe changes in their material circumstances and update their views of the responsible parties or policies accordingly. Such a model is consistent with the finding that elected officials often lose support if they preside over economic downturns (Lewis-Beck and Stegmaier 2000), wartime casualties (Karol and Miguel 2007), or terrorist attacks (Bali 2007).

An alternative model suggests that members of the public rely on cues from trusted co-partisan political elites when forming opinions on policy issues (Zaller 1992; Lupia, McCubbins, Arthur, et al. 1998). Proponents of this viewpoint to evidence that even on seemingly objective material conditions such as the state of the economy, public opinion is often sharply divided along partisan lines (Bartels 2002). Similarly, studies that apply this model to the study of conflict have demonstrated that partisanship and political messaging influence public attitudes towards peace processes (Berrebi and Klor 2008; Matanock and Garcia-Sanchez 2017; Matanock, García-Sánchez and Garbiras-Díaz 2020; Haas and Khadka 2020). In particular, an experiment from Matanock, García-Sánchez and Garbiras-Díaz (2020) showing that elite cues about the contents of a peace agreement appear to outweigh the factual information provided by enumerators gives credence to the claim that the post-conflict setting is one where elite messaging can play a powerful role. Moreover, results from Berrebi and Klor (2008) indicating that rocket attacks had divergent effects in right and left-leaning constituencies in Israel are consistent with a model in which political dynamics influence public reactions to violence.

Theory

I build on and contribute to this literature by arguing that political messaging about blame is particularly powerful in shaping responses to post-conflict violence. One of at least two actors could be to blame for post-conflict violence: rebel leadership or government officials. Rebel leaders are responsible for complying with disarmament and demobilization, while government officials must implement agreed-upon concessions. Yet whether post-conflict violence is a result of rebel commanders reneging on disarmament, or of government officials reneging on implementation may be unclear. Indeed, one rationale for deploying international observers to post-conflict settings is that governments and rebels frequently accuse each other of violating agreements, necessitating monitoring by a neutral third party (Fortna 2004).

I contend that who citizens perceive as responsible for violence influences to a significant degree how violence affects their attitudes towards peace agreements. If a democratic government reneges on implementation, citizens have agency – they can demand that their elected officials implement the agreement and hold them accountable at the ballot box. However, if rebels renege on demobilization, citizens have less agency. Unlike government officials in democracies, rebel commanders are unelected and therefore more insulated from citizen pressure. I argue that because citizens lack agency over rebel commanders, post-conflict violence may undermine their confidence in the effectiveness of peace agreements if they perceive rebels as to blame.

But how do members of the public learn about who or what is to blame? In this uncertain context, political messaging about blame for post-conflict violence is potentially quite powerful. Proponents may frame violence as evidence that a peace agreement is being poorly implemented and requires more resources to succeed. By contrast, opponents may frame violence as evidence that rebels are untrustworthy and the conflict cannot be resolved through peaceful means.

Study Setting

In 2016, Colombia’s government ratified a peace agreement with the FARC, the country’s largest rebel group. This agreement has seen both major successes and significant setbacks since its ratification. Early on, it successfully disarmed and demobilized roughly 95 per cent of FARC members and reduced the level of violence in many conflict zones (Telesur 2017; Mora 2016). Within a few years however, factions of the FARC led by commanders who rejected the agreement called ‘FARC Dissidents’ reemerged in much of the FARC’s territory (Posso et al. 2021). In other former conflict zones, rival groups like the National Liberation Army (ELN) seized control. The result has been a resurgence of violent clashes, displacement, and political killings (El Espectador 2022b).

The question of who or what is to blame for post-conflict violence in Colombia has been a highly contentious one. This study took place during the presidential tenure of Iván Duque, a prominent opponent of the peace agreement. ¹ Duque, along with allies like former president Álvaro Uribe, sought to blame violence on the peace agreement itself, which they argued was too lenient towards the FARC. They noted that the agreement offered amnesty to FARC commanders, some of whom formed the factions that are now perpetrating violence. In the 2018 presidential campaign, Duque made his criticism of the 2016 agreement a key campaign issue (Alvarado 2018) and followed through while in office by allocating just a fraction of the funding required for implementation (Soto 2020).

Prominent supporters of the peace agreement like Gustavo Petro – who succeeded Duque as president – blame poor implementation by the government for the persistence of violence. They contend that the outgoing Duque administration neglected or mismanaged many of the agreement’s key provisions (El Espectador 2022a) and that the peace agreement could succeed if it were implemented fully. After taking office, Petro introduced a policy agenda that prioritized the implementation of the 2016 peace agreement, as well as negotiations with the ELN and other armed groups (Sánchez 2023).

Colombia’s experience may hold lessons that pertain to a large number of contemporary armed conflicts. Like Colombia, many democracies seek negotiated exits from internal armed conflicts. Of the 327 full or partial peace agreements in the UCDP database negotiated between 1975 and 2018 between governments and rebel groups, at least 30 per cent were negotiated by democracies. And, many of these countries face either multiple distinct armed conflicts, conflicts involving multiple armed factions, or conflicts with an armed group at risk of fragmenting into splinter groups (Duursma and Fliervoet 2020).

Among these cases are several examples that highlight the importance of political messaging about violence during peace processes. In Israel, politicians sceptical of a renewed peace process have cited the violence after the 1993 Oslo Accords as a justification for rejecting negotiations (Bachner 2023). By contrast, Northern Ireland’s peace process offers an example in which politicians responded to violence with a pro-peace narrative – when an IRA splinter group killed dozens of civilians in the notorious Omagh bombings, leaders from multiple sides argued that the violence demonstrated the need for implementation of the Good Friday Agreement (Darby 2006). And turning to the present day, elected officials in democracies like India and the Philippines are actively brokering peace deals with an array of rebel movements in contexts that feature sporadic outbreaks of violence as well as opposition from rival political camps (Flores 2023; Ramachandran 2024).

Empirical Strategy

I designed a survey experiment with interventions that mirror the competing political narratives around post-conflict violence in Colombia. The Post-conflict Violence Treatment provides a typical excerpt of news about escalating violence to elicit respondents’ reactions to violence alone, ² the Government Culpability Treatment pairs the same excerpt with information about the government’s poor implementation of the peace agreement, and the Rebel Culpability Treatment pairs the excerpt about violence with information about rebel commanders who violated the agreement (Table 1).

Table 1. Treatment texts

The information provided in all three treatments is factual and does not cite partisan sources. However, the government culpability treatment supports the political narrative of the peace agreement’s supporters, which blames inadequate implementation for violence, while the rebel culpability treatment includes the type of information amplified by the peace agreement’s political opponents as part of their message blaming rebels for violence and arguing that the peace agreement was too lenient.

My main hypotheses concern the effect of political messaging: I hypothesize that the effect of the rebel culpability treatment on support for peace outcomes will be more negative relative to information about post-conflict violence alone, while the effect of the government culpability treatment on attitudes towards peace agreements will be less negative compared to the post-conflict violence treatment alone.

I tested the effects of these treatments on three main outcomes of interest, each of which is measured on a 4-point scale. The first outcome question seeks to elicit respondents’ confidence in the 2016 peace agreement with the FARC by asking respondents whether and to what extent they believed the agreement could succeed if it were implemented. The second outcome question asked respondents to rate support for negotiations with the ELN, Colombia’s other major guerrilla group, and the third outcome question asked respondents to rate their level of support for negotiations with the FARC Dissidents. ³ I also created an index variable that combines the three questions as a measure of overall attitudes towards peace agreements. ⁴

I preregistered four main subgroup analyses around ideology, political engagement, security, and conflict proximity. I also prespecified a subgroup analysis with respect to past conflict victimization – however, I classified this analysis as exploratory due to potential non-response.

The survey reached 1,466 respondents from twenty-eight Colombian municipalities selected using a stratified random sampling strategy that ensured representation of each of Colombia’s regions and drew half of the sample from former FARC territory. ⁵ The sample was 52.7 per cent female. The median respondent was between 36 and 45 years old, had a high-school level education, and was politically centre-left. 48.6 per cent of respondents had family members who had been displaced by the conflict. At baseline, 64 per cent of respondents expressed confidence in the 2016 peace agreement, 71.4 per cent were favourable towards a peace agreement with the ELN, and 68.5 per cent towards a peace agreement with the FARC Dissidents.

Treatment effects were estimated with the Lin 2013 estimator, which interacts treatment conditions with centered covariates. To increase the precision of estimates and adjust for any random imbalances, I controlled for a set of pre-registered covariates that included age, urban/rural, ideology, political engagement, local security trajectory, former FARC presence, and geographic region. ⁶

Results

Figure 1 plots the average treatment effects of each of the three treatments. For the rebel culpability treatment and government culpability treatment, I show both the difference between a pure control group and the difference between the group that received the post-conflict violence treatment.

Figure 1. Main treatment effects.

Notes: The three main outcomes are scaled relative to the control group. The peace attitudes index is a z-score index of the three main outcomes. Models are estimated using the controls described in the text, with HC2 standard errors. Bars represent 90 per cent and 95 per cent CIs.

Across most of the outcomes of interest, the effects of the post-conflict violence treatment are negative but statistically insignificant. As hypothesized, the rebel culpability treatment has a larger negative effect on confidence in the 2016 peace accord as well as support for current and future peace processes, with the largest effects observed on support for peace with the ELN. The estimated effect of the rebel culpability treatment on this outcome is -16.6 per cent of a standard deviation relative to control and -16.0 per cent of a standard deviation relative to the post-conflict violence treatment. However, I do not find that blaming government implementation mitigates the negative effects of information about post-conflict violence. For the three main outcomes, the effect of the government culpability treatment is weakly negative and indistinguishable from the post-conflict violence treatment.

These patterns are largely consistent across the preregistered subgroups. In control, right-leaning voters had more negative attitudes towards peace than left-leaning voters and politically engaged respondents had more favourable attitudes towards peace than the politically disengaged; however, there were no major differences in treatment effects across these groups (SI Figure A.8). In exploratory analyses, I also evaluate whether the treatments triggered negative polarization among respondents, but find no evidence for such an effect (SI Figure A.10).

I do find, however, that the negative effect of the rebel culpability treatment was more pronounced among groups more exposed to the conflict, such as residents of former FARC territory and conflict victims affected by displacement. Figure 2 disaggregates the effect of the rebel culpability treatment by former FARC territory.

Figure 2. Stronger effects of rebel culpability treatment in former FARC territory.

Notes: Models are estimated using the controls described in the text, with HC2 standard errors. Bars represent 90 per cent and 95 per cent CIs.

Among residents of former FARC territory, the rebel culpability treatment has a negative and statistically significant effect on all the major outcomes of interest. Focusing on support for peace with the ELN, this treatment has a negative and statistically significant effect of -23.2 per cent of a standard deviation for respondents in former FARC zones, compared to -9.4 per cent of a standard deviation for respondents in non-FARC zones.

In further analyses, I show that a similar pattern holds for respondents whose families were victims of conflict-related displacement (SI Figure A.3). And, in exploratory analyses, I test multiple alternative measures of local-level conflict intensity both during the conflict and in the years following the peace agreement (SI Figure A.9). I find that in general, respondents in areas with more conflict exposure were more supportive of peace agreements in the control group, but that respondents in such areas also had stronger negative reactions to all three treatments. In other words, while respondents in areas of high conflict intensity appear to have more favourable attitudes towards peace agreements at baseline – a finding consistent with some prior research on the topic (Hazlett 2019; Tellez 2019) – their support for peace agreements appears potentially more sensitive to cues about post-conflict violence than the general population.

To evaluate why the rebel culpability treatment had the anticipated effect while the government culpability treatment did not, I next turn to the mechanisms. For one question, respondents rated their confidence that the government was implementing the 2016 peace agreement, and for another question, respondents rated their confidence that the rebels were complying with the 2016 peace agreement. Consistent with my theoretical expectations, respondents in the control group who were more confident that the rebels were complying with the peace agreement also had more positive attitudes towards peace agreements (t-statistic = 3.7), while respondents who were more confident the government was implementing the peace agreement had more negative attitudes (t = −2.0).

Figure 3 plots the effects of each treatment condition on these mechanisms. While most estimates are imprecise, their size and direction offer clues as to why some treatments worked as intended and others did not. Consistent with my expectations, the rebel culpability treatment yields negative – though statistically insignificant – effects on beliefs that the FARC were complying with the agreement, but has no effects on beliefs about government implementation. However, the government culpability treatment did not have a countervailing effect. Rather, it reduced respondents’ confidence that the government was implementing the peace agreement, with a decrease of -12.6 per cent of a standard deviation relative to control, but it also had a strong and unanticipated negative effect on beliefs that the FARC was complying with the agreement (19 per cent of a standard deviation). This response may explain why the government culpability treatment did not have the intended effect; on the whole, the message failed to shift respondents’ perceptions of blame away from rebels and onto the government.

Figure 3. Effects of treatments on beliefs about compliance and implementation.

Notes: Models are estimated using the controls described in the text, with HC2 standard errors. Bars represent 90 per cent and 95 per cent CIs.

Conclusion

This paper investigated how post-conflict violence affects public support for peace agreements. I hypothesized that the effect of violence can be influenced by political messaging about who or what is to blame. My results suggest that political messaging about blame can indeed be effective, but it has important limitations. As anticipated, whereas news about violence alone had little noticeable impact on attitudes towards peace agreements, pairing this news with messaging that explicitly blamed rebel behaviour decreased support for future peace processes, particularly among residents of conflict zones who might be most vulnerable to renewed violence. Contrary to expectations, however, messaging that blamed post-conflict violence on poor government implementation failed to have a countervailing effect. Instead, a large number of respondents inferred that neither the government nor the rebels could be trusted to fulfil the peace agreement’s provisions.

Taken together, these results provide some support for the idea that political messaging can influence how post-conflict violence affects public attitudes. Yet they also highlight important limitations or asymmetries in the arguments available to defenders of peace agreements compared to their opponents. For opponents of peace agreements, episodes of post-conflict violence provide an opportunity to make a relatively simple argument that blames rebels for violence and casts doubt on the effectiveness of peace agreements – an argument the results in this paper suggest is persuasive. By contrast, post-conflict violence appears to create a more difficult messaging environment for supporters of peace. They may argue, as proponents of Colombia’s agreement did, that it is the government that is ultimately to blame for violence due to its failure to implement the peace agreement’s provisions. Yet whether this argument is too indirect or whether civilians are predisposed to blame rebels more than the government, this defence seems ineffective.

This result may be discouraging for proponents of peace agreements and suggests that more research is needed to identify a messaging strategy that can sustain public confidence in the face of violence. Further research on this topic could, for example, investigate whether including stronger monitoring and verification provisions in proposed peace agreements and emphasizing these provisions in messaging might be a more effective antidote to citizens’ real fears about the return of violence.