Studies in Comparative International Development

, Volume 45, Issue 1, pp 104–124

The Legacy of Violence on Post-Civil War Elections: The Case of El Salvador


    • Department of Political ScienceUniversity of Scranton

DOI: 10.1007/s12116-009-9056-x

Cite this article as:
Allison, M.E. St Comp Int Dev (2010) 45: 104. doi:10.1007/s12116-009-9056-x


Over the last several decades, numerous civil wars have ended as a consequence of negotiated settlements. Following many of these settlements, rebel groups have made the transition to political party and competed in democratic elections. In this paper, I assess the legacy of civil war on the performance of rebel groups as political parties. I argue that the ability of rebels to capture and control territory and their use of violence against the civilian population are two key factors explaining the performance of rebels as political parties. I test these hypotheses against the case of the Farabundo Martí National Liberation Front (FMLN) in El Salvador using one-way ANOVA and multivariate regression analyses. In analyzing the FMLN’s performance in the 1994 “elections of the century,” I find that, as a political party, the FMLN benefited both from the state’s violently disproportionate response and its ability to hold territory during the war.


El SalvadorGuerrillasCivil warElections

According to Fearon and Laitin (2003), there were 127 civil wars from 1945 to 1999. While the majority of all civil wars end in military victory by government or rebel forces, roughly 25% have ended as a result of negotiated settlements (Licklider 1995). These negotiated settlements typically involve not only a cessation of armed conflict and the disarmament and demobilization of former combatants, but social, economic, political, and military reforms. The settlements are designed to facilitate the transformation of a system based upon violent political conflict between armed groups to one based upon nonviolent political conflict between competing political parties. Unlike those civil wars that end through military victory, these negotiated settlements do not allow either the government or the rebel group to claim outright victory and impose its will upon the defeated. Elections are frequently used to determine who will rule in these post-civil war societies because they are supposed to transform a violent conflict into a nonviolent one: ballots take the place of bullets. They are expected to enable the former warring parties to pursue their conflicting ideologies and programs in a peaceful fashion. Elections give all factions an opportunity to present their agendas to the citizens, debate with their opponents, and mobilize public opinion to capture political power. Like other elements of a democratic system, elections contribute to the institutionalization of a conflict resolution mechanism in the body politic (Kumar 1998: 7).

In many ways, these rebel groups confront obstacles similar to those confronting any new political party. For example, they need to devise a political platform, raise money in order to cover the costs of full-time staff, advertising, and campaign materials, and select leaders and candidates. New political parties also have to overcome legal obstacles including the collection of signatures or the setting up of party offices in a certain number of electoral districts. But political parties with roots as rebel groups are not “new” organizations in the sense that they are starting from scratch. Some have long histories dating back several decades to political parties, unions, student organizations, and peasant groups in pre-civil war days. Each rebel organization is designed with a leadership and organizational structure conducive to conducting an irregular military campaign requiring secrecy, clandestine operations, and vertical command structures. Historically, these groups resorted to violence, relying upon extra-legal channels to raise money (selling drugs or weapons, “war taxes,” kidnapping, foreign donors). These are unlikely to be appropriate avenues of fundraising following the resolution of the war. And while each of the rebel groups negotiated a political settlement to its respective civil wars, all rebels are not equal. Some were much stronger militarily and counted on a significant number of armed combatants while others may not have been as militarily proficient, instead relying more heavily upon an extensive network of non-combatants. Consequently, the rebel histories of these groups and their performances during the conflict are likely to be at least as important to explaining their performance as political parties as those issues we typically associate with new political parties.

To date, however, there has been little systematic research on how former rebel groups have fared as political parties in postwar elections. Matthew Shugart (1992) argues that changes in electoral rules and institutional design are key concerns to guerrilla groups as they decide whether to accept a political settlement and, subsequently, whether they are likely to become relevant parties in the postwar political system. In a study of the transition of former rebels into political parties in Bosnia, Kosovo, and Mozambique, Manning (2004) argues that the extent to which former armed groups commit to the political settlement in the postwar period depends upon the challenges these new parties confront from intra-elite divisions and elite-base relations. In a study of the transition of former rebels groups into political parties in Central America, Allison (2006) finds that group level factors, such as the number of combatants and popular support for the rebels during a civil war, are more important to explaining the performance of new political parties than institutional factors, such as electoral rules and thresholds. Recently, two edited volumes have also taken a comparative approach to investigating rebel transitions to political parties (Deonandan, Close and Prevost 2007; de Zeeuw 2007).

None of this research, however, explains how successful rebels have been as political parties at the sub-state level. In this article, I assess the legacy of civil war on the performance of former rebel groups as political parties within the different regions of the area affected by the civil war. I argue that two key factors related to the legacy of the war explain this distribution of rebel electoral support within a country. First, a rebel group should perform better in those areas of the country in which it maintained a strong presence during the war than in those areas where it did not. During civil wars, rebel groups sometimes lay the groundwork for a national electoral campaign through political organizing in support of their goals in those areas of the country that they controlled. Second, rebel electoral support should be directly related to the use of violence perpetrated upon the civilian population by the rebels and by the government’s security forces. In those areas of the country where the rebels commit widespread human rights atrocities, we should see suppressed electoral support for the rebels, whereas in those areas where the government was more heavily responsible for abuses, we should expect higher levels of support for the new rebel political party.

In the next section, I develop the theoretical argument that connects the legacy of civil war to the performance of the former rebel group as a political party. I then discuss research design issues including a brief history of the Farabundo Martí National Liberation Front (FMLN) of El Salvador and how I operationalize and measure the variables used to test the hypotheses. Finally, I discuss the implications of this research and conclude with a number of suggestions for future research on the performance of former rebel groups as political parties.

Explaining Rebel Group Electoral Performance

Control of Territory

In general, one would expect that rebel groups that capture and control territory during a civil war should be more likely to succeed as political parties than rebel groups that fail to do so. A rebel group that captures and controls territory, develops a certain level and type of order within that territory, and provides benefits to those living within the territory is more likely to be able to mobilize the population within this area in order to support it as a political party than a rebel group that controls only unpopulated or remote areas of the country. As Kalyvas (1999: 259) discusses, “an insurgent organization which controls a given area (a ‘liberated area’) operates as a counter-sovereign authority, a ‘counter-state’. It provides protection, administers justice, collects taxes, and applies its social program.” In the controlled territory, the rebel group can acquire valuable experience by completing public works projects and working with local organizations to address the needs of the population living within its sphere of control, all critical to its future success as a political party. These organizational experiences should prove advantageous in making the transition to and succeeding as a political party.

On the other hand, if the rebel group is unable to capture and hold territory throughout the country, it will be at a disadvantage when preparing to compete as a national political party. While even a small rebel group might be able to inflict damage upon government forces or elude destruction, it will be more difficult for such a group to succeed as a political party relative to a group that has a presence throughout a larger area of the country. Obviously, success does not come simply from holding territory. It is certainly possible that the population living under rebel protection will not look upon occupying forces with appreciation. Rebel groups can be very violent towards civilians within liberated zones, particularly towards those suspected to be sympathetic to or actively aiding the government. While the rebel group can utilize its experience of having successfully governed controlled zones, or liberated areas, “it also enjoys a local monopoly of violence which it uses to punish its enemies and sanction uncooperative behavior, such as the refusal to supply food or pay the ‘revolutionary tax’” (Kalyvas 1999: 25). A rebel group that relies almost exclusively upon coercion to control territory is unlikely to be successful in the postwar electoral period. So, beyond the territory controlled by the rebels, we should expect the relationship between the rebel group and the population living within its zone of control to be a strong component explaining a rebel group’s performance as a political party.

Therefore, we have two competing expectations. First, popular support for the guerrillas should be greater in controlled zones relative to other areas because of the voluntary support provided by the citizens in these areas and the political work of the guerrillas in establishing legitimate authority in these zones. The second argument is that guerrilla support in controlled zones was coerced and attained through terror and fear. Therefore, there should be a negative relationship between guerrilla controlled zones and its electoral support.
  • Hypothesis 1 (territory): new political party support should be greater in areas of the country that it controlled during the war than in areas that it did not control.

Political Violence

During a civil war, one of the strategic goals, perhaps even the primary goal, of rebels and government, is to “win” the loyalty and support of the “people.” Unfortunately, not all loyalty or support is won through such “benevolent” means. Instead, support is accomplished through coercion. As recent history has made quite clear, neither governments nor rebels can claim a monopoly on the use of force against civilians during war. Governments and rebels from every region of the world and historical period have been known to use violence against civilian populations in order to deter potential and/or punish actual defectors. Violence perpetrated against civilian populations, while perhaps a more emotional and highly publicized issue today, is nothing new (Kalyvas 2001). In some wars such as the recent cases in Latin American history, government forces have committed the majority of civilian massacres (Wickham-Crowley 1990, 1992). Civilians have suffered at the hands of government forces disproportionately relative to the hands of rebels in Guatemala, El Salvador, Honduras, Chile, Argentina, Brazil, and Nicaragua during the Somoza dynasty. On the other hand, rebels have also targeted civilians equally to the government, or perhaps more so, in Peru (Shining Path), Nicaragua (the Contras), and Colombia (Revolutionary Armed Forces of Colombia and the United Self-Defense Forces of Colombia).

Unfortunately, human rights violations, and perhaps the massacre of civilian populations, occur in all wars. While the act of massacring civilians might serve some strategic purpose in the midst of civil war (Kalyvas 1999), it is unlikely to help a rebel group in the postwar period when, as a new political party, it must attract electoral support. Rebels and governments can vie for the people’s support by providing their supporters with benefits at the same time that they deny these goods to those that seek to remain neutral or actively support the opposing group. A rebel group (or government for that matter) that relied overwhelmingly upon coercion in its pursuit of victory is unlikely to fare as well as a rebel group that was able to capture the loyalty and support of the civilian population through less violent means. While rebel groups often attempt to avoid the intentional targeting of civilians because they rely upon the civilian population for food, recruits, and intelligence, this does not mean that rebel violence does not negatively impact the civilian population indirectly. Rebels often seek to undermine the government’s popular support through disrupting the country’s political and economic infrastructure. Rebels frequently target the government’s political (ministry offices, party/political headquarters, the electoral process) and economic (roads, bridges, water and utility plants) infrastructure. These tactics cause great hardship for the civilian population that depends on these services for their survival.

What is the expected relationship, then, between civil war violence and rebel support as a political party? As a new political party, support for the rebels should be strong in those communities where the rebels infrequently used violence against the civilian population in absolute terms as well as relative to government forces.
  • Hypothesis 2 (violence): new political party support should be greater in areas of the country that were more severely affected by the government directed violence than in areas that were less affected.

Political and Economic Factors

While territorial control and civil war violence are expected to be the key variables in explaining the spatial distribution of the new political party’s vote, there are also a number of political and economic factors which can help explain new political party performance.

New, or potential, political parties also need to be concerned with the presence of other opposition political parties. It is possible that while one group takes up the cause of armed conflict, another opposition group might establish itself as a legitimate opposition political party within the political system. These “legitimate” opposition parties might undermine support for the rebel group as a political party. The existence of a political party with similar ideological underpinnings and electoral support might make it more difficult for a new political party, such as one that has its roots as a former rebel group, to be successful. For example, the existence of a left or center-left legal opposition political party will make it more difficult for a Marxist-Leninist rebel group to become successful. Likewise, an ethnic-based rebel group might have a more difficult transition to political party if a similarly ethnic-based party already exists.
  • Hypothesis 3 (opposition parties): new political party support should be greater in areas where political parties with similar political platforms are weak.

Ideally, one would prefer to control for additional social and economic factors. One would expect any rebel group, especially a Marxist–Leninist influenced group, to perform better in areas of the country where economic conditions are poor. Poor economic conditions provide a pool of potential recruits for rebel groups and most likely will provide a pool of potential voters. They also make it more likely that people will punish the incumbent and vote for another political party (Lewis-Beck and Stegmaeir 2000; Tucker 2002). This should not only help new political parties, but all other opposition parties not associated with the incumbent party (Lewis 2000). Countries emerging from civil war are often in poor economic shape due to the loss of life, damage to infrastructure, and the over concentration of a state’s resources committed to internal security. Normally, poor economic conditions would lead to the incumbent party being punished by the voters. However, the economic conditions are likely to vary throughout the country with some areas worse off than others. This makes it more likely that rebel parties will perform better in those areas suffering from greater levels of economic misery.
  • Hypothesis 4 (economic conditions): new political party support should be greater in areas where economic conditions are poor.

Finally, one would expect that when the existing political parties are doing a commendable job, voters will tend not to support new political alternatives. But, on the other hand, if voters do not believe that the system’s current parties are capable of resolving important issues, new political parties are bound to emerge (Hauss and Rayside 1978: 38). One way in which we can get a handle on whether citizens are satisfied with the existing political parties and whether voters might support a new political party is by looking at levels of voter abstention. High voter abstention rates mean that there might be a significant portion of the population that is not completely satisfied with the status quo and would be more open to supporting a new political party.

If a high level of voter abstention at the national level indicates that there is significant support available for a new political party, we should also expect this relationship to exist at lower levels. Therefore, I expect the rebels to perform better in those areas of the country with traditionally high levels of voter abstention.
  • Hypothesis 5 (abstention): new political party support should be greater in areas with high levels of voter abstention.

Why is a new political party that traces its roots back to its formation as a rebel group more successful in some areas of a given country than others? New political party support should be greater in areas of the country that rebels controlled during the war, that government forces used overwhelming violence against the civilian population, that do not have strong political parties with similar political platforms, that are characterized by poor economic conditions, and that suffer from high levels of voter abstention. In the following section, I discuss how I test these hypotheses.

Research Design

In the following analyses, I set out to explain how rebel control and management of territory and its use of violence impacts its performance as a political party in postwar elections through an analysis of the performance of the Frente Farabundo Martí para la Liberación Nacional (Farabundo Martí National Liberation Front, or FMLN) in El Salvador. First, I provide background on the FMLN in El Salvador. Second, I discuss the measurement of the variables used to test the hypotheses developed in the previous section. Third, I discuss the results of the statistical analyses.

Case Selection

On January 16, 1992, the FMLN and Salvadoran President Alfredo Cristiani concluded 2 years of United Nations-mediated talks with the Peace Accords signing at Chapultepec Castle, Mexico City. This “negotiated revolution” ended a decade of civil war that had cost the lives of over 75,000, the internal displacement of 750,000, and the international migration of an additional one million Salvadorans. Two years later, the FMLN and Nationalist Republican Alliance (ARENA) competed in the 1994 “elections of the century.”1 These elections marked a symbolic end to the war as well as the beginning of the FMLN’s life as a political party.2 It is also the election where the most immediate impact of the war on the FMLN’s performance should exist and the only election with which this analysis is concerned. During the intervening period between the signing of the accords and the holding of elections, great uncertainty surrounded the likely electoral performance of the FMLN. How would a former rebel group perform when forced to compete at the ballot box?

On the one hand, prospects for the FMLN looked favorable. The FMLN was probably the largest and militarily most formidable rebel group in the Western Hemisphere. It had successfully fought a US-trained, funded, and equipped Salvadoran military to a stalemate under geographic conditions highly unfavorable to a successful protracted rebel conflict. The FMLN also maintained extensive relations with unions, students’ and teachers’ groups, and peasants’ associations, which could provide party militants and voters. On the other hand, there were reasons to suspect that the FMLN would struggle as a political party. The FMLN was a rebel army organized to fight a military, not an electoral, campaign. In addition, while elections had returned to El Salvador in the midst of the civil war as part of the US-sponsored counterinsurgency program, the FMLN had rejected these “demonstration elections” (Herman and Brodhead 1984) as illegitimate tools of yanquí imperialism. Perhaps more detrimental to the political future of the FMLN, as the war progressed, the FMLN increasingly

attacked the elections in areas of the country not under their military control. They took away ID cards from voters, an implicit threat. They refused to declare a cease fire on election day, and some military actions were close enough to polling places so as to give the impression of an attack on the elections, though no voters were killed (Spence and Vickers 1994: 6).

Finally, the FMLN assassinated opposition mayors, forcibly recruited civilians to fight on its behalf, kidnapped wealthy businessmen for ransom, and engaged in widespread acts of economic terrorism. While the Salvadoran government and military were overwhelmingly responsible for the brutality committed against the civilian population, it remained unclear how an organization with such a history could be successful at the ballot box.

Variable Operationalization

The dependent variable used to measure electoral performance is the percentage of the vote captured by the FMLN in elections for the Legislative Assembly in the March 1994 elections of the century. I analyze electoral data from the Supreme Electoral Tribunal (Salvadoran electoral commission) at both the department level and the municipal level. I use the FMLN vote in legislative elections because the FMLN supported a coalition candidate in elections for the presidency and a number of municipal offices.

Hypothesis 1 suggested that a rebel group would perform better in those areas of the country that it controlled during the war relative to areas it did not. Obviously, this is a difficult hypothesis to test in El Salvador and in any country emerging from civil war. The Salvadoran government contended that any serious offensive on its part would dislocate the FMLN from areas in which it had concentrated and, therefore, these could not be considered FMLN-controlled areas. To the contrary, others maintain that the FMLN controlled “the zones north of the Salvadoran departments of San Salvador, Chalatenango, Cabañas, and Morazán, and in the central zone of San Vicente—Volcán Chinchontepec” (Alvarez 1988: 85).3 During the war, FMLN presence also extended into the departments of San Miguel, La Unión, and Usulután. Given that the FMLN had a relatively strong presence in these eight departments, we might expect that its electoral support in the 1994 elections should have been higher there than in the other six departments, which had little history of FMLN activity. It is also possible that because the FMLN did not control any department in its entirety and most of the area that it did control was not heavily populated, the impact of FMLN presence might not be apparent in voter tallies at the department level. Therefore, I also explore FMLN control at a more localized level.

In order to systematically assess the impact of rebel controlled zones on the popular vote of the FMLN, I code FMLN-controlled zones as those municipalities where the FMLN prevented elections from being conducted in the first round of the 1984 presidential elections. All other zones are coded zero. In 1984, the war prevented 59 of the country’s 262 municipalities from holding elections. Employing municipalities where votes were not cast in 1984 provides a useful, albeit rough, indicator of FMLN controlled territory. This approach is relatively consistent with McClintock (1998)4 and with reports that the FMLN controlled 15–25% of the country in the early 1980s. Given that several authors have argued that the FMLN provided various services to the Salvadoran population and that FMLN support was widespread and voluntary, one expects FMLN-controlled zones to have had a positive impact on the FMLN vote in the 1994 elections.

However, there are a number of concerns with operationalizing FMLN-controlled zones in this manner. First, typically, municipalities did not conduct elections for two distinct reasons. One reason is that the FMLN-controlled certain zones and would not allow elections to be conducted. These are exactly the municipalities I am interested in. Second, the FMLN military strength was significant enough to thwart the electoral process even though the zone was not directly under FMLN control. Therefore, I am collapsing these two categories (FMLN control and FMLN zones of operation) into a single category. While these are important concerns, they should not preclude systematic analysis.

A second problem is that the FMLN frequently controlled municipalities for weeks, months, or years. The FMLN did not always permanently control these areas. Simply occupying the municipality for a number of days close to the date of the election might have been sufficient to prevent them from being held and, hence, having the municipality qualify as a FMLN-controlled zone. The ability of the FMLN to control territory and establish some sort of alternative governing structure capable of influencing its vote in the 1994 election is much more likely to have been a process of months and years, not days or weeks. On the other hand, including some municipalities in the analyses that were only influenced over a short period of time should prove an even harder test of the hypothesis, as we are including a number of municipalities with limited long-term FMLN presence.5

Hypothesis 2 addresses the relationship between the rebel group and the population. I hypothesized that rebel groups is more likely to succeed in those areas of the country that are more severely affected by government inflicted violence than those that are not. One way in which we can probe the impact of the legacy of the civil war on the electoral fortunes of the new political party is by comparing its electoral results against the geographic distribution of the violence. In El Salvador, Seligson and McElhinny (1996) conducted a survey whereby respondents were asked whether a family member had died as a result of the war. At the high end, over 70% of the respondents in Cuscatlán said that at least one family member died as a result of the war while at the low end less than 20% in Ahuachapán claimed to have lost a family member as a result of the war. At one level, this allows a basic assessment of FMLN performance relative to those departments where civil war victims originated.

A second way in which we can probe the violence hypothesis is by assessing whether war-torn areas of the country were more likely to support the new party than areas that were spared the bulk of the fighting. The United States Agency for International Development sponsored a program during the peace negotiations, the National Reconstruction Program (NRP), that set out to identify those areas of the country that had been most severely affected by the civil war. The goal was to identify the municipalities that were most in need of reconstruction funds and development projects and, therefore, places where the government should spend more heavily before the 1994 election. Out of the country’s 262 municipalities, the NRP identified 115 “ex-conflict” zones and 137 “non-conflict” zones.6 Comparing “ex-conflict” and “non-conflict” zones can provide insight into the electoral support provided to the FMLN by those municipalities most seriously affected by the war (“ex-conflict zones”) from those least affected (“non-conflict zones”). Given that the Salvadoran government most likely directed a disproportionate amount of reconstruction funds to these municipalities in order to better position itself for the 1994 elections, it is an even harder test of the hypothesis.

The data on voter abstention and strength of other left parties come from the Supreme Electoral Tribunal (TSE) in El Salvador while the data on economic conditions comes from the country’s 1998 census. I measure abstention and the strength of opposition parties at the municipal level. High voter abstention in the 1991 elections should be positively related to the FMLN vote in 1994. Abstention is measured by dividing the number of total votes from the 1991 legislative elections at the municipal level and dividing that number by the total voting age population for that municipality as reported in the 1992 census. I then subtract this number from one hundred to come up with the level of abstention for the municipality. I also expect that the high levels of support for the Democratic Convergence (CD) should be negatively related to the FMLN vote in 1994. I measure the strength of left parties using the CD’s vote percentage in the 1991 elections. Finally, I also test for the effects of economic conditions on the support for the FMLN. Unfortunately, poverty rates are only available at the department level beginning with a yearly survey of homes conducted by the Ministry of the Economy in El Salvador in 1998.7 These rates come from 4 years after the election. Therefore, I am less confident in their ability to explain FMLN performance. However, it is unlikely that these rates changed dramatically in the four years and at such an uneven rate across departments that the OLS results would be biased. I code total poverty rates as the combined percentage of people in each department living in relative and extreme poverty.

Empirical Analyses

In this section, I perform tests of the hypotheses at both the department and municipal levels for the 1994 elections in El Salvador. Politically, El Salvador is divided into 14 departments and further subdivided into 262 municipalities. First, I discuss the evidence for or against the two main hypotheses at the department level. Second, I test for the independent effects of the two main variables when controlling for other factors using multivariate regression using political violence, electoral, and economic data at the municipal level. Given the small number of departments, statistical analyses using analysis of variance (ANOVA) and multivariate regression are confined to the municipal level.

Electoral Performance

The FMLN performed relatively well in the 1994 elections. In elections for the presidency, the FMLN formed a coalition along with the CD and the National Revolutionary Movement (MNR) in support of Rubén Zamora. The coalition captured 25% of the national vote. This forced a second round of voting under a majority runoff formula, since no candidate secured an outright majority in the first round. In the second round, the coalition candidate captured 32% of the vote, but was defeated by the ARENA candidate, Armando Calderón Sol, who captured 68%. In municipal elections, the FMLN won 15 municipalities out of 262 nationwide (5.7%).8 Finally, in legislative elections, the FMLN became the country’s second largest party by winning 21 out of 84 seats (25%). Unlike the single member district (SMD) voting used to elect the municipality’s mayor and councils, membership in the Legislative Assembly is determined via proportional representation. While the FMLN did not capture the majority of the Salvadoran vote, it was obvious that it had significant popular support, more than the US and Salvadoran governments had recognized.

The first hypothesis states that the new political party should be more successful in rebel-controlled zones than in uncontrolled zones. In the case of the FMLN and its zones of operation, there exists evidence of both coercive and non-coercive relationships.9 On the one hand, there has been much written about FMLN attempts to transform the areas under their control to a more just society based upon cooperation and solidarity (Alvarez 1988; Department of Social Science 1987; Hammond 1998; Luciak 2001; McClintock 1998; Wood 2000, 2003). Following the failed 1981 offensive, one of the goals of the FMLN was to capture national territory in order to dismantle “the existing state’s infrastructure of local government offices, jails, telephone exchanges, police stations, military garrisons, and counterinsurgency military and civil defense facilities” (Alvarez 1988: 79). As the FMLN proceeded to dismantle these institutions, it would simultaneously establish “alternative governing authority” based upon “widespread” and “voluntary” campesino support (Wood 2003). Therefore, a new society would emerge in these FMLN zones based upon “cooperation” (McClintock 1998: 74) and “solidarity” (Alvarez 1988: 87) thus preparing them for the day that the FMLN would capture state power.

In FMLN-controlled zones, there existed “numerous collective initiatives, including collective agricultural production, and a strong sense of community among the FMLN militants” (McClintock 1998: 74). Collective agricultural production would serve the needs of the guerrilla army as well as the needs of those who could not adequately produce for themselves (Pearce 1984). Another collective initiative in the zones under FMLN control involved the construction of popular education programs. These programs “established close links between education and other practices and goals not normally considered part of it—community organization, political struggle, and social transformation” (Hammond 1998: 5). The political, social, and economic benefits of living in FMLN-controlled zones should have resulted in an increased level of electoral support relative to areas of the country where the FMLN had little control or organizational presence.

On the other hand, several sources have portrayed less than ideal conditions within FMLN zones. Perhaps a strong FMLN presence should not have had a positive impact on its postwar electoral support if it ruled these areas through fear and intimidation or if the population blamed the FMLN for the violence even if they were not directly responsible. Bracamonte and Spencer argue that while the FMLN initially relied upon ideological education in its recruitment efforts, the FMLN increasingly relied upon forced recruitment of those living in war zones as it continued to suffer heavy casualties (1995: 8). Bosch also argues that following the failed offensive of 1981, “terrorism in the countryside would replace combat in major cities” (1999: 109). Instead of ruling with the voluntary participation of the people, the FMLN controlled these zones through terrorizing the inhabitants. Likewise, the United Nations Truth Commission on El Salvador details a number of incidents of forced recruitment by the FMLN, particularly by the People’s Revolutionary Army (ERP). Given this evidence, one would expect lower levels of electoral support in FMLN controlled zones during the 1994 elections relative to the rest of the country, or simply low levels across both zones.

Do the analyses provide us with any insight into which set of arguments seems to have had the greatest impact on electoral support for the FMLN in the 1994 elections? As seen in Fig. 1, the FMLN captured 30% of the legislative vote in San Salvador, by far its best performance in any single department. In Chalatenango, the FMLN captured 19% of the vote, its fifth best performance. These results, in addition to the FMLN’s performance in San Vicente (23%) and La Libertad (22%) where it also maintained strong presences during the war, provide relative support for the hypothesis. On the other hand, Morazán was a FMLN stronghold, yet the party only captured 16% of the vote, its 11th best showing out of fourteen. In addition, Santa Ana, a department with little FMLN military activity or outright signs of support for the rebels during the civil war, supported the FMLN with 22%. This was good for the fourth greatest level of FMLN support. At the departmental level, then, it does not appear as if there is a very strong association between FMLN rebel activity and its electoral support in the 1994 elections.
Fig. 1

FMLN vote in 1994 legislative election (departments with FMLN presence are in capital letters)

The fact that one does not find a strong correlation between support for the FMLN and the departments where it had the strongest presence during the war is not terribly surprising.10 While Chalatenango and Morazán were two departments that showed the most outwardly visible signs of FMLN support during the civil war, the rebel presence was largely confined to the more northern areas of each department. These areas were both more rural and less populous than the remaining areas of the department that appeared to remain more sympathetic to the government. Therefore, as I mentioned previously, it is critical to analyze the legacy of the civil war on the FMLN’s electoral performance at the municipal level.

How does FMLN rebel activity relate to its electoral performance at the municipal level? Table 1 presents a comparison of FMLN support in the 1994 legislative elections between FMLN-controlled zones and all others, using one-way analysis of variance (ANOVA).11 The difference between FMLN support in these controlled zones from the rest of the country’s municipalities is statistically significant (F = 17.625, 1 df, p < .0001). FMLN support in formerly controlled zones averaged 23% while FMLN support in the country’s other municipalities averaged 15%. FMLN support is indeed much greater in controlled zones.
Table 1

FMLN-controlled zones and FMLN performance in the 1994 elections


FMLN performance

Number of zones

Mean (S.D.)

FMLN-controlled zones

22.85 (18.81)


Rest of country

15.28 (9.39)



16.97 (12.49)



F = 17.625 (1 df, p < 0.0001)


The second main hypothesis postulates that new political party support should be greater in those areas of the country more severely affected by government orchestrated violence. How did the role of violence play out in El Salvador’s conflict? Following the end of violent activities in El Salvador, the United Nations Truth Commission Report determined that repeated “serious acts of violence” had been committed by the FMLN in only 5% of the cases, while the state and related security forces were identified as the perpetrators roughly 90% of the time. Many guerrilla recruits often cited government repression as the reason for supporting or joining the FMLN (McClintock 1998; Wood 2003). As a result, we would expect the FMLN to perform better in those areas where the conflict occurred relative to those areas less affected by the war given the disproportionate violence committed by the government against the civilian population.

On the other hand, it is likely that the crimes or violations committed by the FMLN were not accounted for in the Truth Commission Report. As a result, the report probably underestimated the manner in which guerrilla violence affected ordinary Salvadorans. While there is no credible evidence that the FMLN targeted civilians on a scale comparable to the Salvadoran military, the FMLN frequently targeted the country’s infrastructure—electricity grids, bridges, dams, water stations, highway chokepoints, etc.—and, less frequently, civilians aligned with the government. FMLN attacks against the country’s infrastructure were attempts to undermine support for the government by proving that the government could not provide for the general population and that the billions of dollars in US economic aid was wasted. During the first half of 1980 alone, Bosch estimates that the FMLN “committed 3,140 acts of violence, including arson, assault, assassination, and the destruction of bridges, electric power towers, and private business establishments” (1999: 60). Some argue that this violence was part of a grand strategy of the FMLN to increase recruitment levels and eventually bring down the government (Bracamonte and Spencer 1995: 27). These attacks upon the country’s infrastructure would not have shown up as examples of human rights violations.12

It is also possible that FMLN attacks upon the infrastructure had the opposite effect than what the FMLN had envisioned. Instead of encouraging civilians to support the rebels, the FMLN could have made life so unbearable, that their actions pushed many non-aligned civilians to support the government. Similarly, Bracamonte and Spencer argue that increasing terrorist tactics employed by the FMLN as the 1980s progressed, particularly in the urban areas, brought some additional support to the guerrillas, but not nearly enough to compensate for those that were pushed away from the more moderate Christian Democratic Party to the more hard line ARENA party, as seen in the strengthening of ARENA and the weakening of the PDC in the 1989 presidential elections (1995: 33). Even sympathetic leaders from the Frente Democrático Revolucionario (Democratic Revolutionary Front or FDR) distanced themselves from the rebels by criticizing the FMLN’s less discriminate use of violence against the civilian population in the latter half of the 1980s (Baloyra-Herp 1995; McClintock 1998).

In addition to the attacks upon the state’s infrastructure, one of the groups that comprised the FMLN (the ERP) resorted to forced recruitment of personnel and the assassination of public officials, resulting in further alienation from the population. McClintock (1998: 60) estimates that the FMLN killed about 40 civilians each year throughout the mid-1980s. Therefore, the forced recruitment of civilians, assassination of government officials, and attacks upon the country’s infrastructure, and the FMLN’s perceived responsibility for provoking government repression, could have had a negative effect upon guerrilla support in war-torn areas of the country.

Do the statistical analyses provide us with any insight into which set of arguments seems to have had the greatest impact on electoral support for the FMLN in the 1994 elections? Figure 2 plots the vote for the FMLN by department against the percentage of the department’s population responding that they had a relative killed as a result of the war. If one divides the seven most severely affected departments from the seven least affected, there does not appear to be much of a relationship.13
Fig. 2

Distribution of political violence and FMLN performance in the 1994 legislative elections

At the department level, two stand out—La Unión and Cuscatlán. La Unión ranked as the seventh most conflict prone department with 38% of its residents having had a family member killed as a result of the war. Given that six departments were more conflict-prone and seven were less, we would have expected FMLN support in La Unión close to the median. Instead, less than 7% of the department’s votes were cast for the FMLN making it the department that provided the least support for the FMLN. In Cuscatlán, on the other hand, over 70% of its residents reported having a family member killed as a result of the war. This was far and above, the most for any single department. Given the hypothesis, we would have expected FMLN support to be greater than the mid-range of all departments—yet Cuscatlán finished 6th most.

But perhaps the significance of the legacy of the war on FMLN performance is stronger at the municipal level than can be discerned at the department level. Table 2 presents a comparison of voting in non-conflict and ex-conflict zones for the FMLN in the 1994 legislative elections using ANOVA. The FMLN averaged 21% of the vote in ex-conflict zones and only 13% in non-conflict zones. On average, the FMLN received about 60% more votes in zones that were severely affected by the war than they received in zones that were not. The difference between zones is also statistically significant (F = 28.167, 1 df, p < .0001). These results support the hypothesis. Those areas of the country more affected by the war provided greater support for the FMLN than areas less affected by the conflict.
Table 2

FMLN performance in “Ex-Conflict” and “Non-Conflict Zones” in the 1994 elections


FMLN performance

Number of zones

Mean (S.D.)

Ex-conflict zones

21.07 (15.59)


Non-conflict zones

13.25 (7.03)



16.97 (12.50)



F = 28.167 (1 df, p < 0.0001)

I also test the effects of these two primary independent variables while controlling for a number of social, economic, and political conditions using ordinary least squares (OLS) regression. These variables measure the strength of other left parties, economic conditions, and the level of abstention. Again, while including a number of socioeconomic variables at the municipal level would be ideal, they are not available for El Salvador at this level during the time period under consideration. I do report results using the 1998 departmental poverty rates in Model B. Table 3 presents the results of the multivariate regression models. Overall, the first model (A) is a good fit (R2 = 0.61) with all but one of the independent variables reaching conventionally accepted levels of statistical significance.
Table 3

Accounting for FMLN Performance in the 1994 Elections


Model A coefficient

Model B coefficient

FMLN-Controlled Zone

1.855 (1.312)

2.991 (1.359)*

Conflict Zone

2.767 (1.108)*

3.028 (1.098)**

Opposition Strength

0.924 (.055)***

0.900 (.055)***


−0.117 (.059)*

−0.116 (.058)*


−0.123 (0.055)***


17.59 (4.717)***

23.897 (5.188)***


N = 259

N = 259


R2 = .61

R2 = .62


F (4, 255) = 97.642

F (5, 254) = 81.66


Prob > F = 0.0000

Prob > F = 0.0000

*p < 0.05, **p < 0.01, ***p < 0.001

The first hypothesis expects FMLN control over an area to be positively related to its electoral performance. Though the coefficient for FMLN control is in the predicted direction, it is not statistically significant (p < 0.159). FMLN control during the war is not a systematic predictor of postwar electoral performance when controlling for the effects of the other variables in the model. The second hypothesis concerns the geographic distribution of civil war violence. I expect conflict zones to be positively related to FMLN electoral support. The coefficient is in the predicted direction and statistically significant (p < .05). This finding is consistent with those who argue that the violence perpetrated by the state resulted in many Salvadorans providing support to the FMLN. Their support appears to have been voluntary and to have continued into the postwar period.

The two other variables in Model A were also statistically significant. I had predicted that the existence of a strong left party would negatively impact the performance of a leftist rebel party, but this has proven not to be the case. The coefficient is statistically significant (p < 0.001); however, it is positive. In this case, I explored the relationship between the CD’s performance in the 1991 elections and FMLN electoral performance in 1994. There are several possible reasons for this positive relationship. One possible reason is that the CD had only participated in two previous elections. The limited electoral history of the CD did not allow for the party to develop a committed group of voters. Alternatively, this could indicate that the people in these areas were more inclined to support leftist ideals, so they were a better ideological fit for the FMLN than areas where the CD had weak support. A third reason is that the CD, in another form, had previously played the role of the FMLN’s political wing. For some voters, a vote for the CD in 1991 was a vote in favor of the FMLN. Lastly, the effects of abstention (p < 0.05) is statistically significant, but in the opposite direction from what had been predicted. Those municipalities that had experienced lower voter turnout levels in previous elections did not turn out in greater support of the FMLN.

Model B presents the results of multivariate regression with the addition of poverty rates at the department level. In controlling for the level of poverty at the department level, all of the variables reach conventionally accepted levels of statistical significance, including the variable for FMLN-controlled zones. The variable for poverty is statistically significant, although its direction is negative. Departments with greater poverty provided less support to the FMLN. This goes against our expectation that former revolutionary groups would perform better as a political party in poorer areas of the country. One possible explanation for this unexpected relationship is that the ruling ARENA party employed its extensive patronage network to heavily influence the vote in the poorer areas of the country. The variable for FMLN-controlled zones is significant, although marginally, and in the hypothesized direction. The FMLN was somewhat successful in transforming zones under their control into electoral support in the postwar period. Finally, the variables for conflict zones, opposition strength, and abstention remain statistically significant and in the same direction.

Conclusions and Implications

Following a decade of political violence and a negotiated settlement to the Salvadoran civil war, the FMLN competed for the first time as a political party in the 1994 “elections of the century” and secured roughly 20–25% of the vote. In this paper, I tested two main variables expected to explain the geographic support for the FMLN in these elections. At the departmental level, we see a weak relationship between FMLN controlled zones and the distribution of political violence and the FMLN support in the 1994 elections. The FMLN performed well in departments where it had a strong presence (San Salvador, Chalatenango, and San Vicente) and in those where its presence was relatively weak (Santa Ana). Likewise, FMLN support was strong in departments with high (Chalatenango, Cuscatlán, and San Vicente) and low (Santa Ana and La Libertad) levels of political violence.

In order to analyze the legacy of the Salvadoran civil war on the electoral performance of the FMLN, I also conducted a number of statistical analyses at the municipal level. I separately ran one-way analysis of variance (ANOVA) to determine whether there existed a statistical relationship between the support for the FMLN in controlled zones and non-controlled zones and between ex-conflict and non-conflict zones. In these analyses, FMLN support was indeed greater in controlled zones (relative to non-controlled zones) and ex-conflict zones (relative to non-conflict zones). The municipal level results strongly support the hypothesized relationships. When analyzing FMLN electoral support including both FMLN control and conflict zones while controlling for a number of political factors using multivariate regression analysis, both coefficients remain statistically significant.

What conclusions can we draw from the preceding analysis? On the one hand, the FMLN did perform better in zones it controlled during the war than in those it did not. This supports the finding that the ability of the FMLN to develop alternative authoritative structures during the civil war had positive implications for the FMLN in the postwar period. Many scholars cite the importance of civilian support in FMLN-controlled zones and these zones do tend to favor the FMLN more so than non-controlled zones. In terms of the impact of civil war violence, FMLN support is greater in ex-conflict zones versus non-conflict zones using ANOVA and OLS regression. Those areas that suffered disproportionately from the war’s violence also tended to support the FMLN in greater percentages than those areas that escaped relatively unscathed. These findings also emphasize that researchers need to focus on local political units to the greatest extent possible. While the FMLN maintained a strong presence in more than half of the country’s 14 departments, though, the effects of its presence appears only to be significant at the municipal level.

On the other hand, some might argue that even in formerly FMLN-controlled and ex-conflict zones, the FMLN managed less than one-quarter of the total vote. FMLN support was perhaps “widespread” and “voluntary,” but it was nowhere near universal. Most residents continued to support other political alternatives as evidenced by the fact that the FMLN did not capture a plurality of the vote in any single department. While this is accurate, the vote percentage also speaks highly of the FMLN’s support and might actually have underrepresented its actual support. Several developments both during the electoral campaign and on election day might have depressed FMLN support. First, there was the widespread belief that had the FMLN won, the peace process would collapse and war would resume (Wantchekon 1999). Second, there is evidence that voter registration irregularities and the allocation of state campaign funds tended to disadvantage the FMLN relative to other competitors (Stahler-Stolk 1995; Vickers and Spence 1994). Even with these electoral advantages for the governing party, one-quarter of the voters supported the FMLN.

In terms of the other variables, all appear to be helpful in explaining the performance of the FMLN as a new political party. Of particular interest, the FMLN did well in those municipalities where the left had performed well previously. For rebel groups contemplating the transition to political party, it would appear that they should seriously consider the performance of like-minded political parties prior to making their decision. New political parties may benefit more from cultivating voters in those areas where previous opposition groups have already shown promise than by venturing out into areas that have not previously shown much sympathy for similar ideologies. However, as mentioned earlier, the link between the CD and the FMLN in the public’s eyes was quite strong. This might not be the case for all rebel groups competing as political parties. Future research should distinguish between existing opposition parties that share a similar ideological tendency from those existing political parties that are seen by many as appendages of the rebel organization. Given that the organizations and individuals that comprised the CD had an extensive history of coordination with the rebels meant that its electoral support was more of a proxy for the FMLN than it was an independent political alternative.

Given the concerns surrounding the first postwar election, future research on the long-term effects of the Salvadoran civil war are strongly recommended. Following the 1994 “elections of the century,” the FMLN went on to compete in several additional presidential (1999, 2004, and 2009) and municipal and legislative (1997, 2000, 2003, 2006, and 2009) contests. While one might expect that the effects of the war on electoral results to weaken over time and to be replaced by voters’ evaluations of the FMLN’s platform and actual governance, it would still make a valuable contribution to the emerging literature on post-civil war analyses. While it is important to investigate the longer-term implications of the Salvadoran civil war, there is still more that needs to be understood about the 1994 elections. For example, while the FMLN performed well throughout much of the country, particularly in areas where it had a strong presence during the war, why did it only win in 15 municipalities, and why those municipalities? This research does not directly address these questions. However, they are important questions that should be explored. It is possible that the FMLN’s organizational efforts were much greater in these communities during the war and that state-directed violence in these areas was more intense than the country’s other 247 municipalities.

A second extension of this research would be to broaden the analyses to other rebel groups that have undertaken the transition to political party. Obviously, not all rebel groups are the same. Many possess more extensive histories of brutality towards civilian populations than that the FMLN. Others have controlled larger swaths of territory for both longer and shorter periods of time. Finally, others shunned political work almost entirely during the civil war and instead focused upon military objectives. These characteristics are likely to produce different outcomes for rebel groups as they compete as political parties following the end to hostilities.


These elections have been referred to as the “elections of the century” because, in addition to being the first election in the postwar period, it was the first time since the adoption of the 1983 constitution that the election for each level of government was held simultaneously. Elections for the presidency are held every 5 years while those for legislative and municipal office every three.


Since the signing of the Peace Accords in 1992, the FMLN has competed in four presidential elections (1994, 1999, 2004, and 2009) and six legislative and municipal elections (1994, 1997, 2000, 2003, 2006, and 2009).


Alvarez cites an undated document, El Poder Popular en El Salvador, by J. Ventura(e) as the source of FMLN activity.


In addition, McClintock supplements these municipalities with others where elections were held, but the mayor was forced to govern in exile because of the presence of the FMLN.


Future studies are encouraged to refine the operationalization of FMLN-controlled zones. First, more extensive research might systematically divide these two categories to further distinguish between FMLN controlled territory and “zones of FMLN operation.” The degree to which the FMLN controlled these 59 zones varied significantly as did the length of time over which the FMLN operated in these municipalities.


During the initial coding of the municipalities most severely affected by the civil war, ten municipalities were not coded because they were too dangerous. In the following analyses, I have coded these ten as conflict zones. I ran the analyses with and without these ten municipalities as ex-conflict zones. There were no differences in the results.


Ministerio de Economía, Dirección General de Estadística y Censos. Encuesta de Hogares de Propositos Multiples, 1998.


The FMLN won 13 municipal elections by itself and two in an alliance with the CD.


While the FMLN was guilty of assassinating civilian government officials and engaging in sporadic forced recruitment, there is no credible evidence that the FMLN participated in large-scale attacks against the civilian population. One might make the argument, that had the FMLN captured state power, it would have ruled with an iron fist and through the use of terrorism, but there is little evidence that during the war the FMLN controlled the population within its sphere of influence through violence.


I would like to thank an anonymous reviewer for further refining this expectation.


In the 1984 elections, the FMLN prevented elections from being held in 59 municipalities. San Luis de la Reina in San Miguel drops from all analyses, though, because there is no available electoral data from 1994. San Luis would have counted as an FMLN controlled zone and conflict zone.


Readers should be aware that Brian J. Bosch is a former U.S. military attaché whose analysis of the civil war in El Salvador is self-described as “principally from the San Salvador government’s perspective” (1999: xi), José Angel Moroni Bracamonte is the “pen name of a combatant in the war” on the Salvadoran military’s side, and David E. Spencer served in the U.S. Army and National Guard while working as a political consultant to the armed forces.


One possible explanation, of course, is that the FMLN was responsible for more of the deaths in these areas than the government. Unfortunately, at this point, I am unable to distinguish between the two.



The author would like to thank Paul Hensel, Christina Fattore, Joe Young, Stephen Shellman, Len Champney, William Parente, and the editors and anonymous reviewers of Studies in Comparative International Development for their constructive and helpful comments.

Copyright information

© Springer Science+Business Media, LLC 2009