Plan Colombia's Onset: Effects on Homicides and Violent Deaths *
Inicio do Plano Colombia: efeitos em homicidios e mortes violentas
Historically, the relationship between security policies and crime has been of great interest for a number of disciplines. Becker (1968) on the one hand, and Ehrlich (1974) on the other both offer analyses where criminals make the decision to commit crimes based on the potential benefits and the probability of receiving punishment in terms of apprehensions and/or monetary compensations to society. It was only recently that the increased availability of data and the new methodologies have enabled us to unravel the elusive relationship between security policies--in terms of increased financial and human resources--and crime rates. (1)
Most of the literature has focused on either developed countries or on the impact of security measures adopted for fighting criminal activities in urban areas (see Di Tella and Schargrodski, 2004 for Argentina, Klick and Tabarrock, 2005 for Washington D.C., Machin and Marie, 2005 for England and Wales, and Draca et al., 2009 for London). (2) Not many studies deal with these issues in developing countries, which are more likely to face both a wider range of criminal actors and higher levels of crime as such. Colombia, in particular, has been of major interest because of the coexistence of the world's oldest guerrilla group, the drug producers and dealers, and paramilitary groups as well as urban gangs. Over the past three decades, these criminals and their actions have mutated to survive despite the actions of the authorities.
The effects of the Plan Colombia program on homicides and other types of violent deaths are analyzed in this paper. (3) A key feature is that the policy seems to have had disproportionate effects on the different coca producing departments. Under these conditions, various strategies were considered for looking at homicides and violent deaths in those departments with high-coca destruction relative to low-destruction and non-producer ones before and after the implementation of the program.
Among the most recent studies of the situation in Colombia, Angrist and Kugler (2008), Baron (2008), and Perez (2012a,b) are the most notable. All of them looked at quasi-experiments to circumvent the corresponding endogeneity problems. Angrist and Kugler address the causal relationship between the increase in coca cultivation in the early nineties and violence as well as the potential effects on the labor market in the rural areas. They used the 1,000 hectare threshold under cultivation to classify the treatment and control groups and found consistent evidence supporting the hypothesis that more coca production brings about more violent deaths since coca supports organized criminals in rural areas.
Baron (2008) makes use of the Plan Colombia program to address the causal relationship between this program and violence, homicides, and partner abuse. To classify the treatment and control groups of departments he uses the 1,000 hectares of coca crops destroyed under the argument that this is a critical point from which coca activities might affect criminality in Colombia. The results show significant reductions in homicides in those regions where the program seems to have had stronger presence, and no significant effects on partner abuse were found.
Perez (2012a) analyzes the Democratic Security Policy (DSP) implemented by the incoming government in 2002 and its causal effects on a wide range of crimes. There are differentiated results based on the types of crimes committed and the type of perpetrator. Strong and significant reductions were found for those crimes commonly committed by organized crime such as terrorism, kidnappings, auto-theft, and terrestrial piracy, but no effects from the policy on crimes usually committed by common criminals in urban areas such as burglaries and street robberies.
The present study contributes to the literature in several ways. Previous studies on this topic analyzed the effects of Plan Colombia on violence using data from 2000 to 2005. This strategy disregards the effect caused by an overlap with a national policy implemented during 2002/2003 (the Democratic Security Policy, DSP), a strategy for which the main purpose was to reduce criminality (homicides included). This might be causing serious upward biases on the estimates and clouding the potential real effects of the program on the outcome variables. In this paper, a post-policy period free of the overlapping effects with the DSP is used to clarify which of the effects were the real effects of the program on homicides. (4)
The second contribution has to do with the classification of the regions into treatment and control groups. A relative measure (coca crops destroyed as a percentage of the total coca cultivation) and its distribution among the departments were used to classify the treatment and control regions. This approach naturally categorizes the two groups by means of a relative, comparable measure. I believe this strategy is stronger and a way that is even more consistent to look at the impact of economic and police resources on homicides and violent deaths. (5) Additionally, the departments were classified into three groups, instead of two: high-destruction, low-destruction, and non-producers. This approach made it possible to carry out additional exercises in order to test the robustness and assumptions of the empirical approach. (6)
The estimates show that there is no evidence supporting the hypothesis that the implementation of Plan Colombia significantly reduced homicides although it seems to have had differentiated positive/negative effects across gender and location on violent deaths in high-coca-destruction regions relative to the low-destruction and non-producing regions. This was the conclusion I reached after using a highly-disaggregated database for deaths and after carrying out a series of exercises and robustness checks testing for different specifications of the baseline model and different alternatives for the comparison groups.
My argument in this paper is that, whereas it is undeniable that Plan Colombia's economic resources have played an important role in enhancing security and reducing the areas under coca cultivation, it is also true that reductions in homicides were not an effect of Plan Colombia but perhaps a result of the Democratic Security Policy. (7) The main purpose of this latter strategy, which was implemented by the incoming government in 2002, was to cut down on the high levels of criminality in the country. Also note that it was precisely between 1999 and 2002, four years after Plan Colombia's implementation, that crime indicators reached the highest recorded levels in the country. The rest of the paper is organized as follows. The Plan Colombia program and its scope is briefly described in section 2. The nature of the classification of regions is explained in section 3 and some descriptive statistics of the data are provided. The empirical strategy and model are described in section 4. In section 5, the estimates of the impact of PC on homicides and violent deaths are presented, together with some robustness exercises. Section 6 is the conclusion.
2 Plan Colombia Program
Between 1994 and 1999, Colombia faced a steep increase of more than 350% in coca bush cultivation as a result of the decrease in the activities of the other two major producers, Peru and Bolivia. The consequences, apart from the obvious increase in cocaine production and export, were the strengthening of the illegal armed groups and the increase in violence and crime. In 1999, just one year after the start of a new government, the Plan Colombia program was carried out for the first time.
At the beginning, the main goals of Plan Colombia were: fighting against drug production and trafficking as well as related organized crime activities, improving the economy and democratic institutions, and moving ahead on peace-seeking talks with the insurgent groups. The government specifically sought to reduce the number of hectares of coca bushes by half by 2005. At the time the program was put into effect, there were 160,100 hectares of coca cultivation. (8)
The strategy of the Colombian-American partnership was to focus on aerial spraying and manual coca bush destruction. Even though there were several goals in the original plan that were humanitarian and defensive in nature, the involvement and cooperation of the American government and its specific resolve to reduce the amount of cocaine entering its territory changed the main scope of the program. As a whole, the focus was changed to military matters and this has been largely criticized (Ramirez, 2005). (9) With respect to this, Mejia and Restrepo (2009) show through a game theory approach that given the unlike main goals of the two governments (the Colombian and the American), the best strategies, those that either maximized the benefits or minimized the costs of the war against drug trafficking, are considerably different. For the American government, for instance, it would be optimal to focus the resources on the trafficking side of the problem only, whereas the Colombian government would very much prefer to concentrate on maintaining territorial control, which would mean focusing on the production side of the problem.
Assessments of the first phase of the Plan Colombia strategy showed controversial and unexpected results. Even though the Colombian-American partnership for reducing cocaine production was able to achieve the objective of a 50% reduction in coca crops, something they did not count on was the production process' adaptability to exogenous changes. As mentioned by Mejia and Restrepo (2009) and Mejia and Posada (2008), cocaine producers were able to produce virtually the same amount of coca leaf, and the corresponding amount in cocaine production even with only...