Introduction

Convincing Colombians to support a peace agreement between the centre-right government of Juan Manuel Santos and the leftist Fuerzas Armadas Revolucionarias de Colombia (FARC) has proved to be a challenging task.² Having suffered the longest conflict in the western hemisphere with a tally of over 220,000 deaths and 6,000,000 victims (Comisión Nacional de Memoria Histórica, hereafter CNMH 2013), the population seems uneasy about the idea of granting prosecutorial and political concessions to the FARC guerrilla in exchange for its demobilization. Having achieved a final agreement with this guerrilla, on October 2, 2016, Santos submitted it for popular approval through a plebiscite, but failed to attain its endorsement by a razor-thin margin -50,21% of the population voted against it, while 49,78% voted to support it. Though surprising, these results were not entirely unpredictable: the FARC is regarded as one of the main culprits of the suffering of civilians and is highly discredited due to its history of kidnapping and drug trafficking (García, Rodríguez and Seligson 2014). Yet, the government remains adamant in stating that the long-term success of any peace agreement must hinge on the population's overall acceptance of its content. The details of the agreement thus will need to cater to the wants of the Colombian population lest it lack the political support to be implemented (Elespectador.com 2013).

But, what are these wants and how are they construed? Most surveys indicate that Colombians support striking a peace agreement but strongly oppose making any concessions to the rebels. That is, they want to end the conflict but are not willing to concede what is necessary to woo the rebels into an agreement (García, Rodríguez and Seligson 2014). How can these seemingly contradictory attitudes towards peacebuilding be explained? ³ Is it that Colombians do not know how to balance their interest in securing peace and safety with their righteous ideas about punishing those who have done them wrong? Further, are these preferences homogenous throughout the country? Though over the past decades conflict-related violence has been felt throughout the country, in the last ten years it has become increasingly peripheral and locally circumscribed (CNMH 2013). Hence, do people living in the still more heavily conflict-afflicted areas think differently from the average citizen? And what lessons, if any, can be learned from the Colombian case to advance peacebuilding in other countries stricken by armed conflict?

The article contributes to answering these questions based on the notion that political attitudes are the result of material interests (i.e. self-interest) and ideas -the former defined as the motivation to accumulate wealth, power, status and/or to secure the necessities of life, while the latter understood as the beliefs held by individuals (Goldstein and Keohane 1993). Building upon this, it proposes a theoretical framework wherein attitudes are shaped through an interaction amongst three components or filters that jointly allow an individual to give meaning to reality and to visualize the desirability of alternative courses of action. These are: (1) material interests, (2) ideas related to cause-effect relationships (causal beliefs), and (3) ideas related to principles or normative views (principled beliefs). The framework further suggests that the more directly linked self-interests are to a specific course of action, the weaker the relationship between ideas and attitudes will be, and vice versa.

Following this train of thought, the article posits two research questions and corresponding hypotheses at theoretical and practical levels. To prove them it uses literature related to political preferences, along with quantitative data on Colombians' political attitudes from the 2013 LAPOP survey.⁴ Further qualitative data was retrieved through interviews of experts on the Colombian conflict and public opinion to provide a stronger basis for the findings.⁵

At a theoretical level, the article asks: (RQ.1) Does the relationship between ideas and attitudes towards a given issue vary for individuals whose material interests are less directly affected by it? If so, how, and what factors may explain this variation? To which it answers that (H.1) the association does vary: though an association between ideas and attitudes is to be expected in complex policy issues, ceteris paribus, there is a stronger impact of ideas on attitudes among individuals whose material interests are less directly affected by a given policy issue.

At a practical level, it inquires: (RQ.2) What are the differences between individuals living close to conflict and the average citizen in terms of the ideational determinants of their attitudes towards peacebuilding issues? Focusing the main analysis on two ideational variables, namely ideology and the potential benefits that an individual believes a peace agreement would bring to her municipality, the article posits that (H.2) though generally speaking political attitudes towards these issues are guided by the supposed benefits of an eventual peace agreement, the more distant an individual is from the conflict, the more likely it is that her attitudes will be determined by her ideology.

The rest of the article is structured in the following way. The next section gives an overview of the features of the Colombian conflict and the political polarization that the peace process has generated. This is followed by a theoretical framework that explains attitudes as the result of the interplay between ideas and self-interest. The next two sections set out the empirical strategy and its results and interpretation. Finally, the article concludes with a discussion of the implications of these findings.

1. Context: The Colombian Armed Conflict and Political Attitudes towards Peacebuilding

Having lasted for more than fifty years and having involved at one point more than nine different armed forces including guerrilla groups, right-wing paramilitary groups, and drug cartels (Rettberg 2012), the complexity of the Colombian conflict by far surpasses the scope of this article. Acknowledging these limitations and preferring to focus more on recent events, this section simply summarizes the key features of the armed conflict, the FARC's role in it, and the linkage with political attitudes towards peacebuilding and the peace process that began in 2012.

The Colombian conflict has been a prolonged, degraded war that has operated with a logic of high-frequency and low-intensity violence. This means that, despite the manifold big massacres and outrageous terrorist attacks, the main damage to the civilian population has been done slowly but persistently through methods such as small massacres,⁶ kidnappings, and selective assassinations (CNMH 2013). Military actors have favoured this mode of violence because it allows them to obscure their crimes from public opinion at the national and international level, thus turning their deeds into catastrophic yet locally circumscribed affairs. Though the conflict-related violence has touched most of the municipalities in the country, its intensity has been particularly severe in some of them: out of 23,161 selective homicides registered between 1980 and 2012, 85% happened in just 30% of the nation's territory (CNMH 2013, 47). The municipalities with the highest toll have tended to be poor areas located in peripheral rural regions with minimal presence of the Colombian state (Yaffe 2011).

Founded more than 50 years ago, the FARC guerrilla has been by far one of the largest, most powerful and wealthiest illegal military actors (Aguilera 2013). With a clear hierarchical structure, a Marxist-Leninist ideology, and a small but stable rural base, this group has consistently aspired to seize power by force from what they regard as a ruthless oligarchy. Their "noble" intentions, however, have been tarnished by the inexcusable methods they have used against civilians and other military actors -terrorist attacks, forced recruitment, selective assassinations, and kidnappings have been their main modus operandi (CNMH 2013). In addition to these crimes, factors such as their involvement in drug trafficking, three unsuccessful previous peace negotiations in the 1980s and 90s, and domestic politics have meant that their historical favourability at the national level has tended to be low, never surpassing the one-digit level (García, Rodríguez and Seligson 2014).

Having reached the apex of their power in the late 1990s and early 2000s in terms of human and financial resources, the tide suddenly turned against this guerrilla when Álvaro Uribe, a popular far-right politician became president and imposed an assertive military strategy supported by US aid packages that allowed the government to regain the upper hand (Nussio, Rettberg and Ugarriza 2015). During his tenure (2002-2010), Uribe established security as the paramount public good, denied the existence of an armed conflict, and consolidated the image of the FARC as a terrorist, soon-to-be militarily defeated group with no political agenda (Aguilera 2013). Moreover, he managed to push the FARC rebels back into their mostly rural historical areas of influence and reduced their numbers by more than half, from 20,000 combatants in 2002 to 8,000 in 2010 (Efe 2010). Whereas 54% of the population thought the national army was losing the war against the FARC in 1995, by 2012, 57% of the population thought the FARC could be defeated militarily (Ipsos 2012).

Having finished his mandate with a 75% support rate, Uribe anointed Juan Manuel Santos as his successor. The latter was elected in 2010 and rapidly sought to distance his moderate-right government from that of his predecessor (Londoño 2010). Although the military pressure against the FARC persisted, a new law, which explicitly recognized the existence of an armed conflict, was passed and peace talks with the FARC rebels were unveiled in late 2012. Uribe then became the most vocal opponent of the Santos government and consistently denounced the new president for negotiating with terrorists (Colprensa 2013). Furthermore, taking advantage of his own popularity and most Colombians' mistrust of the FARC rebels, Uribe founded the Centro Democrático Alternativo party (hereafter CD), which became the second biggest political force in the Senate in 2014. CD presidential candidate Óscar Zuluaga -Uribe did not run since the constitution does not permit a third presidency- obtained 45% of the votes against 51% for Santos, even though the incumbent had the support of most of the country's elite and the media (Vélez 2014).

The discussion over the peace talks and concessions to be granted in a peace agreement have fragmented the country politically. Supporters of the negotiations claim that the concessions are a price worth paying since the entire nation would greatly benefit from a peace agreement (Lewin 2013). In turn, Uribe's party and mainly far-right voices posit that allowing impunity on such a wide scale would set a terrible precedent, forfeit victims' rights, and allow the FARC rebels to launder the wealth they have accumulated through criminal activities (Semana 2012a). As the 2016 plebiscite's worryingly polarized results showed, these different approaches to peacebuilding constitute to date one of the deepest political cleavages in Colombia (Semana 2012b).

Hence, in pursuit of an accurate explanation of the differences among Colombian attitudes towards these issues, the next section sets up a theoretical framework to serve as a basis for the discussion.

2. Theoretical Framework: Attitudes as Determined by Ideas and Self-Interest

Explaining political attitudes and behaviour has been a longstanding, highly debated issue in the social sciences. Though answers have been given from multiple vantage points, they can arguably be categorized according to whether their opinions are the result of self-interest -itself a consequence of individual socioeconomic factors- and/or of variables related to individual beliefs such as social norms, values, identities, symbols, etc. (Blyth 2002; Hainmueller and Hopkins 2014; Hay 2010; Schonhardt-Bailey 2006). Although most arguments can be categorized within these two broad categories, to date there is no consensus on what exactly the role of each classification is in shaping attitudes, how they interact with one another, or if their role varies conditional on the type of individual or policy realm under analysis (Hay 2002 and 2010).

The debate has involved contradictory views at both the theoretical and practical levels. For instance, advocating a pure self-interest approach, Shepsle has argued that material interests are logically prior to any beliefs held by the actors, and thus only the former motivate individuals (cited in Goldstein and Keohane 1993). At the other end of the spectrum, Blyth (2002 and 2003) has posited that actions are not a direct reflection of material interests, but rather a particular reflection of what agents perceive these material interests to be. Therefore, he has argued, interests should be seen as "social constructs that are open to redefinition through ideological contestation" (Blyth 2002, 271) and thus logically ulterior to ideas. Similarly, Hay (2010) has favoured dispensing with the concept of material self-interest altogether, claiming that interests do not exist but that subjective and inter-subjective constructions of interests do.

More persuasively perhaps, other authors have based their arguments upon both interests and ideas to explain attitudes. Schonhardt-Bailey has defended both elements in determining attitudes and outcomes: "if interests may be seen as the engine that drives preferences, ideas comprise both the intellect and emotion that inspire actors" (2006, 25). Jacobsen (1995) has posited that ideas fill the means-end gap that self-interested agents face under conditions of uncertainty, while Campbell has stated that dismissing the analytical distinction between ideas and material interests would be to "ignore important differences in types of human motivation and determinants of political and other social outcomes" (1998, 402). Lastly, Goldstein and Keohane have stated that "ideas as well as interests have causal weight in explaining human action" (1993, 4), and "policy outcomes can be explained only when interests and power are combined with a rich understanding of human beliefs" (1993, 9).

In translating these theoretical analyses into practice, studies seem to favour the notion that both self-interest and ideas need to be included when analysing attitudes -more so given that their relevance may vary conditional on the policy realm. For instance, while Bakker (2014) has shown that self-interest is related to economic attitudes, Sears et al. (1980) and Hainmueller and Hopkins (2014) have claimed instead that attitudes on controversial issues such as immigration are more closely associated with ideology, party identification, and racial prejudice. Lau and Heldman (2009) have found that context-related variables such as the information environment moderate the impact of self-interest on public opinion, and Boninger, Krosnick and Berent (1995) have shown that self-interest, social identification, and values were all associated with the importance given to specific attitudes.

In short, though the debate is far from concluded, these empirical findings, as well as the theoretical notion that opinions cannot be fully understood without one of these two elements, seem to suggest that the analysis of attitudes is greatly enriched when both self-interest and ideas are included.

Therefore, this article conceives of public attitudes as being constructed by these two major components: material interests (i.e. self-interest) and ideas. The former are defined as the motivation to accumulate wealth, power, and status, and/or to secure the necessities of life, while the latter are defined as beliefs held by individuals. Following Goldstein and Keohane (1993), ideas are further divided into two sub-components to better isolate their impact on attitudes: principled beliefs, or "normative ideas that specify criteria to distinguish right from wrong and just from unjust" (1993, 9); and causal beliefs, which provide guidelines for individuals on how to achieve their objectives and shed light on cause-effect relationships.⁷ "Causal ideas respond directly to uncertainty by reducing it, whereas principled ideas enable people to behave decisively despite causal uncertainty" (1993, 17).

The theoretical framework proposes that these three components -material interests, causal beliefs, and principled beliefs- shape preferences by acting as the cognitive and material filters through which agents give meaning to their reality and visualize the desirability of given policies or courses of action (Blyth 2002; Schmidt 2008). The relationship amongst these components is dialectical -that is, they are always in contrast with one another- and through their interaction preferences are constructed to provide agents with guidelines for behaviour (Hay 2010).

Thus, when a course of action is proposed, the new input activates the agent's filters so as to allow her to process: (1) whether this course benefits her material interests; (2) whether it complies with the principles that she upholds; and (3) whether, in case of contradiction between these principles and benefits, she should give priority to one or the other. Having formed a preference towards a certain course of action (supporting or opposing it strongly or mildly, being indifferent, etc.), the agent acts upon it, thus having a material impact on the same context that had conditioned her filters in the first place. Figure 1 illustrates these relationships with greater clarity.

Source: Elaborated by the author.

Figure 1 Ideational and material filters in determining individual support for a policy alternative 

Depending on how directly the agent can link this course of action to her different filters, the relative weight of each of them in determining preferences will vary. For instance, if action A is proposing to legalize abortion, a Catholic priest is likely to rely heavily on his principled beliefs to decide that he opposes it strongly -his faith as determined by the Catholic church sets a clear direction in this regard. On the contrary, a scared, pregnant teenager ready to interrupt her pregnancy, despite the risk involved in it being illegal, is likely to base her decision primarily on her material interests. Finally, an expert demographer concerned about overpopulation might support the policy, based on the belief that it will provide a partial solution to that problem. Of course, moral issues and the potential (believed) outcomes of the course of action will be considered by all these individuals, but the relative importance given to each factor in deciding their support for the course of action will vary for each one of them.

Uncertainty -due either to lack of information or the complexity of the course of action or the policy realm- plays a role in the linkage between a course of action and material interests. Specifically, the more uncertain the context, the less likely the individual is to relate a specific action to her own material standing, and the more likely she is to rely on ideational variables (Blyth 2002). Principled and causal beliefs will then act as cognitive buffers that cushion uncertainty: the former by enabling the agent to "foresee" the consequences of an action despite her lack of comprehensive information, and the latter by providing self-explanatory road maps to indicate the way when material cause-effect relationships are not available. Therefore, the relative weight of material interests in shaping individual preferences is likely to be reduced with uncertainty, while that of causal and principled beliefs is likely to increase.

Based on this analytical framework, determinants of attitudes should follow certain patterns conditional on context and issues. These patterns, here labelled propositions and corollaries, are the following:

Proposition 1: In complex environments, attitudes towards a course of action are likely to be associated with the perceived consequences (i.e. causal beliefs) that course of action will have in relation to the individual's material interests.

Proposition 2: The less material wellbeing is directly linked to a course of action, the stronger the association between causal/principled beliefs and attitudes towards that course of action will be.

Corollary 1, following Proposition 2: The relationship between ideational and material factors in shaping individual attitudes is an interactive one; their degree of influence in determining these attitudes is conditioned by the context and how material interests relate to it.

When applied to the Colombian context, these propositions predict patterns in political attitudes. Following Proposition 1 and taking into account the complexity of the Colombian conflict, people's causal beliefs ought to be positively associated with preferences: the more benefits an individual believes a specific peacebuilding approach will bring to her, the more support she should express for it. Following Proposition 2, when comparing individuals living close to conflict with the average citizen, there should be a stronger association between ideational variables (principled beliefs and causal beliefs) and preferences towards peacebuilding approaches for the latter. The next chapter presents an empirical strategy to test these predictions.

3. Empirical Strategy and Data

Previous studies have shed light on the material and ideational determinants of attitudes towards conflict and peacebuilding policies, actors, and actions. In Spain, Aguilar, Balcells and Cebolla-Boado (2011) found that ideology and family victimization were highly relevant in explaining individual attitudes towards transitional justice policies. For the same country, Balcells found that "victimization experiences led to the rejection of the perpetrators' identities along the political cleavage that was salient during the war (i.e., left-right)" as opposed to others that were less visible (e.g. centre-periphery) (2012, 311). In the case of the Israeli-Palestinian conflict, Reifen et al. found no ideological differences in accepting "collateral damage," but "those on the left and right endorsed polarized policy preferences only when their positions about the conflict were held with moral conviction" (2013, 117).

Interesting studies have also been published regarding the case of Colombia. Gutierrez and Wood (2014) and Oppenheim et al. (2015) have posited that ideology influences combatants' behaviour with regards to the likelihood of their defecting and the strategic choices they made in warfare. Garcia, Rodríguez and Seligson (2014) have argued that Colombian citizens think more about political rather than economic issues when positioning themselves on the ideological scale, and that their ideology determines not only their position regarding illegal left-wing actors, but also their party affiliation to a certain extent. These authors have also found important differences between people living in areas close to the conflict as compared to the average Colombian citizen in terms of their socioeconomic features and attitudes towards different policy issues. Lastly, Nussio, Rettberg and Ugarriza (2015) have divided the population into victims and non-victims and found almost no statistically significant differences in attitudes towards transitional justice mechanisms between them.

Building upon these previous findings, the empirical strategy of this article makes two important contributions to the literature on political attitudes towards peacebuilding. First, from a perspective of ideas and self-interest, it mainly focuses on studying the differences in attitudes of individuals living in the most heavily conflict-afflicted municipalities and the national population as a whole. This approach contrasts with that of previous studies where most of the analysis was based only on the divide between victims and non-victims.

Secondly, it aims to expand the way in which the relation between ideas and attitudes can be empirically tested. Most studies tend to approach ideational determinants of attitudes by using ideology as the main indicator, despite the fact that this variable on its own is unlikely to encompass the full impact of ideas. Admittedly, ideology is useful in capturing a coherent set of ideas including both principled and causal beliefs -and has been identified as one of the main variables fuelling the Colombian conflict (Aguilera 2013). But, as Schonhardt-Bailey puts it, ideas should not be reduced to ideologies: while ideas form "a critical component of the mix that comprises the preferences of actors, ideologies usually imply some fairly coherent collection of the former" (2006, 25). With this in mind, in addition to ideology, this article includes two other important variables as well.

The first variable refers to individual perceptions regarding the effect that a peace agreement would have on the respondent's material wellbeing. Contrary to ideology, these perceptions do not necessarily have a sense of coherency and are likely to hinge mostly upon causal beliefs rather than on a mix of principled and causal ones. The second one is an interaction term that builds upon previous work on victims/non-victims and ideology by linking these two concepts together -i.e., the effect of ideology conditional to the fact that the individual is a victim of the conflict. It was included to expand on whether ideational determinants varied amongst individuals with different experiences within the conflict, and to further explore the attitudes of victims given their political salience in the Colombian peace talks (CNMH 2013).⁸

The data used to implement this empirical strategy comes from the LAPOP survey, which has been measuring Colombian attitudes, perceptions, beliefs and experiences since 2004. For its 2013 version, the survey included an over-sample of 111 municipalities that have been severely affected by the armed conflict, which means that it contained two samples: a national one (NS) representative of all the adults living in Colombia (n=1511), and a special one (n=1505) representative of all the adults who live in these 111 municipalities (CAMS) that constitute nearly 10% of all the country's municipalities. As explained by García, Rodríguez and Seligson (2014), who provide a comprehensive description of the survey's methodology and variables, the latter were selected based on a national security strategy dubbed Espada de honor (Sword of Honor) and other studies on the incidence of the conflict at the municipal level. The same questions were asked for both samples so as to allow for comparisons between them.

Ideology was measured as an individual's self-declared political ideology on a left- to-right scale from 1 to 10. Benpeace, or the potential benefits that an individual believes the demobilization of the FARC would bring to her municipality, was obtained through the latent factor in responses to three different questions. These refer to whether the individual thought that demobilization of the FARC would (1) increase or decrease security, (2) whether or not it would improve the economic situation in the respondent's municipality or community, and (3) whether it would strengthen or weaken Colombian democracy. As these questions were categorized inversely in the survey, lower values indicate greater perceived benefits. Lastly, the interaction term was constructed by multiplying Ideology with a dummy variable taking the value of 1 when the respondent was classified as a victim of the conflict (labelled ideology#victim).⁹ Table 1 shows this information with greater clarity.

Seven variables were chosen as dependent variables to assess individual overall attitudes towards the current peace process and the policy arrangements that will stem from it. Six of these variables refer to questions asked in the survey, while the seventh one was constructed. The chosen topics range from support for the current peace process and concessions to the rebels to support for specific transitional justice measures such as truth-seeking. The constructed variable, the Transitional Justice (TJ) Index, focuses on the individual's support for transitional justice measures as a whole, involving matters of justice, memory, truth-seeking and reparation of victims. Table 2 explains these variables in greater detail.

Naturally, control variables were used in the model to isolate the impact of ideational variables. Following the model of Nussio, Rettberg and Ugarriza (2015), these included: (i) age, (ii) education, (iii) gender, (iv) Catholic religion or not, (v) trust in the FARC, (vi) trust in the government, and (vii) urban residence or not. Three further control variables were added based on conversations with the co-directors of the LAPOP survey and García, Rodríguez and Seligson (2014): (viii) household income, (ix) interest in politics, and (x) whether they had children or not. Lastly, as it appeared in the interaction term, (xi) the question of whether the individual was or was not a victim was also included. Details on how all of these variables were measured appear in the 2013 LAPOP questionnaire.

Due to the ordinal scale of the dependent variables, ordinal logistic regression was chosen as the regression analysis method (Aguilar, Balcells and Cebolla-Boado 2011). In the case of the TJ Index, however, ordinal least squares were used due to the fact that the variable is not categorical. Since the survey had two separate samples -as opposed to having one wherein oversampling of CAMS is accounted for with sampling weights- regression analysis was done separately for each sample. In other words, for each dependent variable, each model was run for the national sample and for the CAMS, as had been done previously by García, Rodríguez and Seligson (2014), and Rettberg (2012).

Two models were run for each dependent variable for each sample. The first one included only the Ideology and Benpeace and the control variables (Model 1), while the second one included the variables from Model 1, plus the interaction term Ideology#victim (Model 2). To ensure the accuracy of the findings, different combinations of control variables and other regression methods were also used, though these are not reported in the article.

Based on the Three-Filter framework, the following results are expected: (1) Since they both encompass causal beliefs, at least one of the two independent variables should prove to be regularly associated with individual preferences throughout the different dependent variables. Furthermore, (2) both independent variables will be more closely associated to policy issues and/or have stronger associations for the NS than for the CAMS. As the FARC's ideology is clearly at the left of the spectrum and political parties who tend to criticize the process are at the right, (3) there will be a negative association between preferences and ideology, and (4) a positive association between Benpeace and attitudes. Lastly, since victims are normally expected to have stronger principled beliefs about the conflict (CNMH 2013), (5) the interaction term of ideology conditional on being a victim is expected to be associated with such preferences, and (6) this association is more likely to be stronger in the NS than in the CAMS, since the same difference in the direct link between material interests and course of action applies.

4. Results and Interpretation

In general terms, people living in conflict-afflicted areas tend be rural, less educated, more likely to have children, and have lower incomes than the national population as a whole. They also categorize themselves more to the right of the ideological spectrum, are less interested in politics and, in terms of their relation to the armed conflict, display a 13- percentage-point higher rate of victimization and a 9-percentage-point higher rate of guerrilla-induced victimization than the rest of the population.

Regarding their attitudes, CAA individuals display greater support for: (1) ending the civil conflict with the guerrilla via a peace agreement; (2) the current peace process with the FARC; (3) allowing the rebels to participate in politics and (4) avoid prison if they confess their crimes; and (5) establishing ways in which the crimes committed by the FARC rebels can be publicly remembered (see Table 3). In other words, in four of these seven attitudes, CAA individuals are more willing on average either to grant concessions to the FARC rebels or to push for more transitional justice measures like truth-seeking. Although these differences are not large in magnitude (never more than 1 on a scale of 7), they do contradict the idea that populations more victimized by the rebels and ideologically inclined to the right would be less willing to support measures facilitating a peace agreement with the FARC.

Various patterns emerge in examining the determinants of attitudes and the role of ideational variables. Table 4 shows them in a summarized version of the results obtained after running a total of 28 regressions.

The first emerging pattern is that Benpeace is associated with individual preferences in six out of the seven dependent variables analysed for both samples. In all these cases but one (the politics question for CAMS), p-values indicate with 99% confidence that this association exists. Interestingly, the only question in which this variable is not statistically significant is the one referring to punishment. As this is arguably one of the most emotionally charged and morally sensitive issues in peace processes (Rettberg 2012), these results could indicate that principled beliefs unrelated to ideology, rather than ideas on the long-term material benefits of the policy, are the main determinants of support for allowing a certain degree of impunity.

Taking into account Benpeace's inverse categorization, the negative association between this variable and preferences is as expected: the stronger the belief that the FARC's demobilization will bring benefits, the stronger the support for policies to facilitate the peace agreement and transitional justice measures. The magnitude of the association is generally similar for both samples throughout the six alternatives where it exists -controlling for other variables, one additional unit in Benpeace is associated with a 20% to 30% decrease in odds of a response that indicates higher levels of agreement with the statement.

As no large differences were found in the magnitude of the associations amongst the samples -except for the politics question- there is no strong evidence indicating a stronger impact of ideas at the national level for this variable. There is, nonetheless, evidence to conclude that causal beliefs are a relevant variable in explaining support for conflict-related policies. In this sense, Proposition 1 is validated.

A second relevant pattern is that Ideology does indicate a clear-cut difference in determinants of preferences between the two samples. While the association between Ideology and the dependent variables is statistically significant in five out of the seven questions for the NS (peace process, politics, punishment, truth, and TJ Index), the same is true only for two questions in the CAMS (memory and TJ Index). Furthermore, while there is 95% and 99% confidence in this association in two of the questions (truth and politics) in the NS, it is never higher than 90% in the CAMS. In other words, whereas ideology is a significant predictor of individual support for most of the dependent variables in the NS, this is seldom the case in the CAMS. This clear distinction supports Proposition 2 of the analytical framework.

As expected, the nature of this association is negative: the more ideologically inclined to the right, the less likely a citizen is to show support for policies that facilitate a peace agreement with the FARC rebels. The magnitude is also fairly constant: one additional unit of Ideology is associated with a 6-10% decrease in odds of indicating a higher level of agreement, or with a decrease of 0,04, in the case of the TJ Index. Even if this impact is not as strong as that of Benpeace, it is still relevant. For instance, according to the models' predicted probabilities, an individual who claims to have an extreme right-wing ideological position (Ideology=10) is 10% more likely than a centrist (Ideology=5) to express strong disagreement with guaranteeing political participation for the rebels. When compared to an extreme left-wing individual, the difference in the predicted probability is as high as 20%. Figure 2 illustrates these differences.

Source: Elaborated by the author, based on LAPOP Colombia 2013.

Figure 2 Predicted probabilities of individuals in both samples answering "strongly disagree" (=1) to the politics question vis-à-vis Ideology and Benpeace 

The only dependent variable for which Ideology is statistically significant in both samples is the TJ Index -not a question in itself but rather a latent factor in responses to four other questions. This suggests that even if Ideology is not associated with all the policies underpinning a transitional justice system, it might still provide a (weak) guideline towards a citizen's overall support for the idea. It is also worth commenting on the fact that the reparation question does not display any statistically significant association with Ideology for either of the two samples. Two possible explanations might account for this: (i) since the military actors in the conflict espouse ideologies that span the entire ideological spectrum, even if political inclinations influence an individual's sympathy for a particular type of victims, they will probably not affect her support for the reparation of victims in general. Alternatively, it can be argued that (ii) since the status and relevance of victims has grown significantly in the past few years (CNMH 2013), they have managed to garner a degree of public support that is not conditioned by political inclinations. If this is correct, it would strengthen the role of victims as independent actors in the peace process.

At first glance, the condition of being a victim seems to be equally associated with attitudes in both samples. However, as the results with the interaction term seem to suggest, this might be misleading: conditional on ideology, the impact of being a victim is stronger and the difference in determinants between the two populations widens. Whereas in the NS, Ideology#victim is associated with individual preferences in three out of the seven questions (Punishment, Truth, and TJ Index), in the CAMS the term is only statistically significant for the reparation question. This reaffirms the notion that ideational predictors vary depending on the material context in which the agent is embedded. It also confirms Proposition 2: the role of ideas is weaker in the CAM, where material interests are more directly affected by the armed conflict.

Moving on, Ideology#victim in the NS exhibits a negative association with the punishment and truth questions, and the TJ Index. In other words, ceteris paribus, a victim who considers herself left-wing is more likely to support courses of action that facilitate a peace agreement than a non-victim with the same ideological position. Figure 3 shows how these magnitudes affect probabilities of support conditional on ideology for some of the questions. For instance, after controlling for all other variables, the predicted probability of an extreme leftist victim (Victim=1 and Ideology=1) answering "strongly disagree" on the punishment question is 20% lower than that of an extreme leftist who is not a victim doing so.

Source: Elaborated by the author, based on LAPOP Colombia 2013.

Figure 3 Predicted probabilities of individuals in both samples answering "strongly disagree" (=1) to the punishment and truth questions conditional on being a victim 

Similarly, Victim#ideology is statistically significant and positive for the CAMS on the reparation question (see also Figure 3). That is, victims positioned on the right side of the ideological spectrum are more likely to support this policy measure. As this positive association contradicts the patterns detected thus far between ideology and transitional justice measures, an alternative explanation -even if speculative- is needed. Arguably, it can be said that right-wing victims living in conflict-afflicted areas are faced with the dilemma of finding a way to balance their desire to hold the FARC rebels accountable for their crimes, while still being willing to make concessions to them so as to increase the likelihood of a peace agreement. As concessions regarding political participation and punishment are known to be the rebels' most cherished demands (Nasi 2003), it is likely that these agents will opt instead to channel their ideological preferences through issues such as reparation, which are relevant but not so critical as to impede a successful negotiation.

Overall, the analysis with this interaction term sheds light on the varying role that ideas play in different contexts. Firstly, in line with the framework's expectations, it confirms that ideology is often more determinant of attitudes at the national level than at the local, conflict-afflicted level. Secondly, it shows that despite having faced conflict-related hardships, the support of victims for certain policies varies depending on their ideology, all other things being equal. In this sense, having had their own material interests severely compromised in the past by conflict-related events can paradoxically enhance the degree of influence that ideas will have on them in the future.

Conclusions

This article has addressed two research questions. At a theoretical level, it has delved into the question of whether associations between ideas and attitudes vary amongst individuals who are more directly affected by a specific policy issue, and the factors that may explain this variation. At a practical level, it has sought to assess, from an ideas and interests perspective, whether there are any significant differences in the determinants of attitudes towards peacebuilding between Colombians living close to conflict and the rest of the national population.

At a theoretical level, the findings suggest that the Three-Filter framework, which conceives of individual attitudes as the outcome of interplay amongst material interests and causal and principled beliefs, can be used to explain political attitudes. Hinting at the complexity of the Colombian conflict, causal beliefs are positively associated with individual preferences in almost all the variables evaluated (Proposition 1) for both populations -though no significant variation in the magnitude of this effect between the two samples was found. When considering both causal and principled beliefs as encompassed in ideology, clear-cut differences were found in the determinants of attitudes between those living in the conflict-ridden areas and the national population at large. This suggests that the degree of influence of ideas on opinions varies depending on how directly linked material interests are to that specific policy realm. In particular, the less directly affected they are, the greater the relative influence of ideas in shaping attitudes (Proposition 2). As a consequence of both of these findings, it can be argued that ideational and material factors are interdependent and that their relative degrees of influence in moulding attitudes vary depending on the context and how the individual relates to it (Corollary 1). All this renders hypothesis one (H.1) valid as an answer to the first research question (RQ.1).

These findings constitute a contribution to the literature on political attitudes. Firstly, they highlight the analytical usefulness of coupling interests and ideas to explain individual preferences. Even if the former can be traced back in history as social constructions, they are still a relatively stable and pervasive motivation of human behaviour (Campbell 1998). Therefore, dispensing altogether of this concept (Hay 2010) would deprive scholars of an analytical tool that, when understood within its limitations, is likely to clarify rather than blur the understanding of political behaviour. Furthermore, ideational factors also have limitations that are hard to surmount without introducing material interests into the equation. Their inherent subjectivity can easily lead to vagueness and difficulty in measurement for analytical purposes, but by conceiving of both as independent and interdependent variables, it is possible to retain some of the predictive power that renders material interests so useful, while allowing for the much-needed flexibility that ideas can deliver.

Secondly, it suggests that further theoretical research should focus on explaining how these ideational and material elements condition one another. Due to the limited scope of this study, the article does not take into account external factors that condition the degree of influence of ideas and self-interest on human behaviour -information, complexity of the policy realm, and uncertainty are but a few of these. Future studies should research the ways in which attitudes are constructed as a result of the interactions of these multiple factors and identify the nature and magnitude of these relationships. Differentiating ideas by type could also provide more conceptual clarity in future ideational analyses (Schmidt 2008). These theoretical efforts should be supported through empirical analysis in order to prove relationships of causality. To facilitate all of this, greater use of longitudinal surveys to account for unobservable variables and explore the use of instrumental variables should be a starting point in this field.

At the practical level, the main results validate the second hypothesis (H.2) by indicating a consistent association between individual preferences and the believed benefits of a peace agreement for both samples, while showing that ideology tends to be mostly associated with preferences for the national sample, but not for those living close to conflict. These findings contribute to a better understanding of peacebuilding in countries that have been unevenly affected by (civil) war. In the specific case of Colombia, they can contribute to the current pursuit of a peace agreement and the chances of it being finally ratified and successfully implemented in the long run. In particular, they put forward three key notions that, though linked here to the Colombian peace process and peace agreement, are also likely to stand in other countries with strongly localized conflicts.

The first one is that part of the legitimacy of the peace process depends on convincing Colombians of the potential benefits of signing a peace agreement (causal beliefs). This finding holds true for both the population in the conflict-afflicted areas and for the nation as a whole, and for all policy issues except for the question of concessions regarding punishment. In simple words, Colombians need to believe that the price they are to pay for peace is reasonable, as measured against what they think will obtain in return for these concessions. Hence, in order to boost the legitimacy of the peace process and the peace agreement, political actors should focus on properly informing the citizenry and generating optimistic yet realistic expectations of what may follow the achievement of such an agreement.

The second notion is that, notwithstanding the seeming homogeneity of Colombian attitudes towards specific conflict-related issues such as transitional justice, people living in conflict-afflicted areas do think differently from those in the rest of the country -at least insofar as their attitudes are less likely to be determined by ideological factors. This difference may be the result of the fact that the security and economic interests of individuals in these municipalities are more directly affected by the conflict than those of people in other parts of the country. Though the limited scope of this article does not allow it to prove causality, said disparity does explain the outcomes of the 2016 plebiscite and the 2014 presidential elections: while the Centro Democratico's discourse against the peace process resonated primarily amongst right-wing individuals located mainly in the central, highly-urbanized provinces, poorer and more conflict-ridden municipalities voted to support President Santos' peace process (Azuero 2014; Fergusson and Molina 2016).

Lastly, a third notion is that in matters pertaining to punishment, truth-seeking, and transitional justice measures in general, ideology does have a different impact on a victim's attitudes than on those of a non-victim. This is a noteworthy difference when taken into account against previous findings (Nussio, Rettberg and Ugarriza 2015), which suggested that the condition of being a victim was not associated with attitudes towards transitional justice. Consequently, it should contribute to the literature on what victims want and why they want it, and should also allow for the design of policies that are more tailor-made to suit their needs.

All these results come, of course, with limitations. One of the most important is the fact that, since the Colombian conflict has not been a clear-cut war where only one portion of the municipalities has been affected, it is hard to have one clearly defined group of conflict-afflicted municipalities to compare with the rest. Furthermore, since the internal dynamics of the conflict have varied significantly through time, some municipalities have been affected to a greater or lesser degree at different times. Another noteworthy caveat is that, given the features of CAMS respondents (e.g. they have lower levels of education and are more likely to be violently targeted for their political ideas), it is impossible to fully ensure that their answers to some of the complicated survey questions cited in this article were always accurate and reliable. Therefore, some nuances are probably not fully accounted for in the results.

That being said, the difference found between the determinants of attitudes amongst the two populations is still worthy of attention as it has political and social consequences for the current peace talks. Politically, it suggests that people living in conflict-afflicted areas are less likely than the national population in general to decide their support for these policies based on their party-affiliation, political identities or politically-related principles. In this sense, winning their hearts and minds will require a different kind of effort on the part of politicians and policymakers.

Socially, it highlights the difficulty of constructing and implementing a peace agreement that caters to the wants of all Colombians. The majority of the population may decide to support the peacebuilding efforts based on factors that are different from those on which the minority bases its decision -despite the fact that the latter is the most directly affected by it. A potential solution to this could be that of promoting dialogue between the national and local, conflict-afflicted level, and providing a greater say to the representatives of conflict-afflicted areas in the peacebuilding discussions. Also, increasing awareness of these differences in attitudes within the safer and richer parts of the country could greatly facilitate a consensus.

This article has sought to transpose the discussion on ideas and interests from the limits of academia to the historical context of a peace agreement that is likely to determine the future of Colombia for decades to come. The results will provide valuable lessons for other conflict-ridden countries poised to undergo similar transitions in the future, and they are of course relevant and timely all the same for Colombia's own peacebuilding process. Having withstood a 50-year-long conflict that has repeatedly scarred their bodies and minds, Colombians are now bound to decide on how to end it, and to endure both the good and the bad consequences of this decision. In finding a way to accommodate their past in what seems to be a hopeful future, their only compass will be the ideas they hold to be true and their inherent motivation to secure a better life.