Elections are regularly held in countries facing ongoing civil wars, among them Iraq, India, Nigeria, Afghanistan, the Philippines and Colombia. Occasionally, electoral choices amidst conflict are stark: Candidates propose two different visions for their country’s future, and their campaigns are defined by positions on the civil war itself. Such was the case May 25 when Colombia voted to elect its next president.

 How did legacies of violence affect vote choice in the Colombian election? This post provides evidence that President Juan Manuel Santos — who opened negotiations 19 months ago with the Revolutionary Armed Forces of Colombia (FARC), the country’s oldest and strongest guerrilla group — performed poorly at polls in areas where armed groups have historically committed high levels of violence. The anti-negotiation candidate, Óscar Iván Zuluaga, in contrast, performed far better in communities that have traditionally been hard-hit by violence. Below I offer background on the electoral campaign, statistical evidence illustrating the relationship between vote shares and violence, and conclude with some thoughts on what might account for the connection.

 Contrasting perspectives on whether to negotiate with the FARC quickly became the most important issue of the presidential election campaign. Santos has bet his campaign on a “peace agenda”: slow but steady progress at the negotiating table (agreement has been reached on three of six agenda items), providing historic reparations to victims of the armed conflict, and initiating a land restitution process for those forcibly displaced from their homes. The latter addresses core grievances articulated by the FARC.

Óscar Iván Zuluaga has opposed these initiatives, stating that he would suspend negotiations with the FARC until the group met certain key demands, including a unilateral ceasefire (the candidate has recently altered his stance slightly). Zuluaga is supported by ex-president Álvaro Uribe, figurehead of Zuluaga’s party and the person who gave him the nod to contend for executive office. Uribe uses his Twitter account to personally insult Santos, to batter him from the right, and to spread rumors about how Santos will break up the armed forces and sell out the country to the FARC if he is reelected.

And so, with the conflict front and center in the electoral campaign, Colombians went to the ballot box May 25. Because no candidate received 50 percent of the vote — Santos received 25.69 percent and Zuluaga 29.25 percent — the two will compete in a run-off June 15. Candidates from smaller parties garnered anywhere from 8 percent to 16 percent of the vote.

A few days after the votes from the first round were counted, a respected Colombian journalism outfit published an article that suggested a positive correlation between electoral support for Santos and FARC areas of strength over the past 10 years. What happens when we test that relationship more rigorously?

I statistically examine how the average number of FARC and paramilitary attacks per year in each Colombian municipality between 1988 and 2010 affected the percent of votes cast for Santos and Zuluaga in those municipalities on election day in May. I use import.io to scrape first-round electoral returns at the municipality level, data from both governmental and non-governmental sources to measure levels of violence, and regression analysis to control for factors that might influence both armed group attacks and electoral support. Factors I control for include population size, area, elevation, distance from Bogotá, whether a municipality includes an international border, poverty levels, the percent of a municipality’s area under coca cultivation and the presence of sites for mining gemstones. The outcome we seek to explain is the percent of votes for a particular candidate, either Santos or Zuluaga, in a given municipality.

The findings (see figures below) are striking and counter those from the Colombian journalism outlet referenced above. Where the FARC committed high levels of attacks between 1988 and 2010, Santos performed poorly in the 2014 elections. For instance, in towns where the FARC committed an average of four attacks per year, Santos is estimated to have received 31 percent of the vote, yet where the FARC committed an average of 28 attacks per year, he is estimated to have received only 21 percent  of the vote. A similar picture emerges for paramilitary attacks: At higher levels of paramilitary attacks, Santos received lower levels of electoral support.

The results for Zuluaga — displayed below — demonstrate the opposite tendency. In municipalities where the FARC attacked most in the past, Zuluaga tended to perform better on election day. For example, where the FARC committed an average of four attacks per year, Zuluaga is estimated to have received nearly 37 percent of the vote, yet in municipalities where the FARC committed an average of 28 attacks per year he received nearly 46 percent of the vote. A similar pattern holds for paramilitary attacks: Zuluaga performed far better in areas where paramilitaries were particularly violent, when compared to those where they were not. (An interesting additional finding is that municipalities on an international border are far less likely to vote for Zuluaga, possibly fearing that his aggressive rhetoric toward Ecuador and Venezuela could prompt renewed hostilities.)

 Although I have controlled for some alternative explanations, these results cannot establish a causal relationship, nor can we pinpoint precisely why Santos seems to be punished for legacies of violence while Zuluaga seems to benefit. I venture a number of plausible explanations.

First, those most affected by violence may not want to grant to the FARC the terms they anticipate Santos will offer. That negotiations have been conducted in Cuba, shrouded in secrecy, has created a void of information surrounding the possible terms of the agreement. This has allowed the right to speculate loudly, for instance, that a peace deal would ensure impunity for atrocities, and that the government is granting far more concessions to the FARC than it should.

Second, although a 2008 survey of victims in Colombia, conducted by Angelika Rettberg of the Universidad de los Andes, suggests a preference for material compensation and truth-telling, victimized communities are also interested in punitive or retributive justice for harms done. If victimized communities believe prosecutions and punishment of perpetrators will be less likely under a Santos administration, this may be limiting his support in such areas. Third, despite 50 years of proof to the contrary, there still exists a belief that the FARC can be defeated militarily. Direct exposure to violence may sharpen those beliefs, pushing individuals and communities to favor a continued military path forward.

Finally, citizens may simply blame incumbents for grievances for which they are not responsible, whether in the recent or distant past. To some, Zuluaga represents a break from that past (despite the fact that he was finance minister under ex-President Uribe), and certainly from the current administration. The promise of change may resonate particularly with those who continue to live with the painful memory of violence.

In summary, historical legacies of violence appear to have significantly shaped vote choices in the Colombian election. This is likely pronounced in the Colombian context because of the salience of the conflict to this particular campaign. Further work needs to be done — in Colombia, following the run-off June 15, and in other countries — to establish the precise mechanisms that lead past victimization to drive citizens’ decisions many years later at the ballot box.

Michael Weintraub is a predoctoral fellow at Yale University’s Program on Order, Conflict  and Violence and, beginning in fall 2014, will be an assistant professor of political science at Binghamton University (SUNY).