Getty Images

Open a newspaper in Colombia, and you’ll be sure to find headlines touting the success of the military and the public’s support of its counterinsurgency operations.

“The dominant view in Colombia is that the military is pretty widely supported,” said Aila M. Matanock, an assistant professor of Political Science at UC Berkeley who focuses on civil conflict and international intervention.


But a survey, conducted by Matanock and Miguel García Sánchez of Colombia’s Universidad de los Andes, suggests that this basic assumption might not be consistent with the reality on the ground, especially for those living in guerilla-controlled regions.

The survey drew inspiration from Nicaragua’s 1990 presidential elections, in which results deviated drastically from the pre-election polling. Lead-up surveys strongly suggested that Sandinista candidate Daniel Ortega would topple opposition candidate Violeta Chamorro by 16 points. Instead, Chamorro won by 14.

In the United States, analysts coined the term “the Bradley effect” to describe a similar phenomena: exit polls showed Tom Bradley, an African American politician who ran for California governor in the 1982, leading by a wide margin. But much to the pollsters’ surprise, Bradley lost; many theorized that voters falsely claimed they voted for Bradley because they did not want to appear bigoted.

In Nicaragua, too, “all the way up until the elections there was this very different measure of who was being widely supported,” Matanock said in a presentation on October 28 at the Center for Latin American Studies at UC Berkeley. “Almost all of these polls got it wrong, and got it pretty substantially wrong.”


The election polling may have been misleading because people were afraid to report their true preferences, Matanock concluded. In places like Colombia, where people feel pressure to show support for the military fear retribution if they don’t, creative polling methods are in order.

To measure support for the Colombian military, Matanock and García Sánchez conducted 1,900 face-to-face interviews with randomly selected subjects in May of 2010. “We employed direct and experimental measures of support for the military,” they wrote in the paper, Controlling Civilians? Examining Support for the Military in Colombia. “Individuals were randomized to receive a direct question about support for the military or an experimental question that employs a list experiment (also known as an item count technique)”


The direct question asked: “Some people believe that the Colombian military forces should have more freedom to defend the nation in the way they see fit. Do you think that the Colombian military forces should have more freedom to defend the nation in the way they see fit?”

While the question's framing could lead to lower responses of overall support (some people, for example, may support the military overall but do not think it needs more freedom to defend the nation), Matanock and Garcia Sanchez held that the framework represented a more concrete question for a country with an internal conflict. In these situations, they explained, militaries often have public positions on how to conduct counterinsurgency operations, which are contested and debated by civil society.


As Matanock and Sanchez explained, in Colombia, “there has been a long debate regarding the level of autonomy for the military in the counterinsurgency. Those close to the military have criticized mechanisms designed to limit its role countering the insurgency, which is contrary to organizations, often outside of the government, that demand more civilian design and oversight.” By using the experimental list approach, the duo hoped to measure, “active support for the military, as it is expressed in an existing debate regarding the institution."

The list experiment are designed to make subjects feel more secure about reporting their true preferences. Instead of simply answering “yes” or “no” to the question of military support, respondents would be given a list of questions with the military query included.


Matanock offered the following example to explain the concept: Give a person a list of three statements, and ask them how many of those they support; stress that they shouldn’t tell you which things they support, just the number. That list would be given to one half of the sample. The other half would get a list of four things, which include the item being measured — in this case, support for the military. They asked respondents: "I am going to present to you a list of four things (three, for the control group) that some people support and others do not. Please listen to these things and tell me HOW MANY you support. Do not tell me WHICH of these things you support, only how many of them you support.”

The list included these four items:

1) The South American nations creating a central bank.

2) The assessment of a special tax to finance the expansion of the parks and green spaces in your neighborhood.


3) The conservative ideology gaining more influence in the Colombian society.

4) The military forces having more freedom to defend the nation in the way they see fit (This statement was excluded for the control group.)


For the control group, which was asked the first three questions, the response would be between zero and three, while the for treatment group, which was also asked the fourth question about the military, the response would be between zero and four. To analyze the results of all the responses, Matanock and Sanchez compared the statistical means of the control group with the the means of the treatment group, with the difference in the between the two reported as support for the additional item — in this case, support for the military. Finally, they contrasted the list experiment’s results with that of the experiment with direct questions.

The list experiment model was meant to lower respondents’ fear of expressing their true opinion. By not outwardly saying whether they support the military having more autonomy, but rather, expressing it indirectly through the list question, people may be more likely to accurately report how they feel.


Matanock pointed to the success of list experiments in the United States: more people report racism, drug use, and other socially condemned behaviors through list experiments. “People don’t feel the pressure to report the dominant view in these surveys,” she said.

Overall, they found that lower levels of support were reported for the military when measured experimentally as opposed to directly. They also discovered that the differences were most pronounced in municipalities with guerrilla control and coca production.


The conflict in Colombia dates back to 1964, as the country’s military faced off with emergent leftist groups. Later, right-wing paramilitary organizations were formed, which battled the leftist guerrilla groups, such as the FARC, with both groups becoming heavily involved in drug trafficking. Throughout the conflict, the military has grown increasingly powerful and organized in its counterinsurgency operations. In 2002, newly elected President Álvaro Uribe secured significant aid funds from the United States to quash the paramilitaries and guerrillas. Shortly after, Plan Colombia, which gives the military substantial autonomy, was born. Most of the paramilitaries demobilized in 2005 and 2006, but have been replaced by smaller groups that the Colombian government refers to as “criminal bands”.


The military, as well as the FARC, paramilitary groups, and other armed groups, have been accused of committing widespread human rights violations. One scandal, often referred to as “falsos positivos,” grabbed Colombian headlines in February of 2010. Reportedly, military officials lured impoverished Colombians into rural areas with the promise of work, but instead shot them and labeled the victims as guerrilla fighters who had been successfully killed in counterinsurgency operations. One might expect that such publicity would lead to declining military support, Matanock said.

But in reality, they found that reported confidence in the military vacillated only slightly, and the numbers remained surprisingly stable. “Only the U.S. and Canada have more trust in the military than Colombia.” From 2004 to 2012, civilian trust in the Colombian military hovered around 65 percent. The military is consistently rated the third most-trusted organization in Colombia, trailing only the Catholic Church and the president.


The continued high level of support for the military suggests two options, Matanock said. The first is that the reports are true, and support is high. The country’s security situation is improving, so it’s possible that people are actually very supportive of the military. The second possibility is what she called “preference falsification” — people know that support for the military is the dominant view in Colombia, and so they report that they also support the military.

Ultimately, Matanock and García Sánchez found that reported rates of support for the military did differ when measured with direct questions versus with the list approach, and, rates of support varied depending on if a person lived in paramilitary or guerilla controlled regions.


Across the board, support for the military was lower when measured with list experiments, Matanock noted. In paramilitary-controlled zones, the rates of support for the military when measured with list questions ve-rsus with direct questions were not as profound as when measured in guerrilla-controlled areas – where the results were astounding.

When measured with the experimental list questions, only 6.3 percent of people in guerilla-controlled regions reported support for the military. But, when measured directly, an average of 58 percent reported support for the military. “You have an over 40 percent difference in the reported rates of support measuring directly versus measuring experimentally in the guerrilla-controlled regions,” Matanock said. “This is what our theory predicts based on preference falsification.”


In municipalities where cocoa is cultivated, results were similar. “What I hope we’ve shown you is that there is a statistically significant difference across experimental and direct measures,” Matanock said. “The experimental measures tend to be much lower than the direct measures. We see this especially in regions with guerrilla control and cocoa cultivation.”

Erica Hellerstein is a Master’s student at UC Berkeley’s Graduate School of Journalism. This article is part of an ongoing series curated through the Univision News bureau at UC Berkeley’s Center for Latin American Studies.

Share This Story

Get our newsletter