Mauricio Rubio *
Paper presented at the PRIO Workshop on "techniques of violence in civil war"
Oslo – August 2004
The main purpose of this paper is to challenge the notion that kidnapping in Colombia “is just another business”. It is argued that, on the contrary, kidnapping has been inextricably linked to armed conflict. The paper is divided in four sections, beginning with a brief history of this activity in Colombia. In the second section the evolution of aggregate rates is analyzed, and two kidnapping booms are highlighted. In the third section three arguments are given to back the idea of a close relation between kidnapping and armed conflict. Emphasis is given to the precarious evidence about bands of common criminals as relevant perpetrators. Concluding remarks address the recent sharp drop in kidnapping rates, and the main issues that require further research.
1 - A brief history of kidnapping in Colombia 
Kidnapping in Colombia has two different roots. It was practiced in rural areas by some late “bandoleros” of the political violence period in the 50s. On the other hand, urban kidnapping of foreigners -diplomats and CEOs- was imported from groups such as Tupamaros (Uruguay) and Montoneros (Argentina). Perhaps the main contribution of Colombian armed groups to kidnapping technology was the capacity to fuse these two roots, catching bourgeois hostages in rural areas, and reproducing at the local level the scenario of alien victims. A highly unequal, almost class-segregated, society facilitated this innovation. Drug trafficking also contributed.
The relationship between drug activities and kidnapping has been quite complex. First, drugs made possible the replacement of a few wealthy, mostly foreign, hostages paying huge ransoms –the typical scenario of the 70s- by a higher number of victims paying less per incident. Drug lords buying haciendas had a by side effect of pushing up rural property prices. This speculative trend displaced poor peasants and attracted urban middle classes to the countryside. The week-end house owner became the typical victim during the eighties, when kidnapping rates first increased significantly. Rural guerrilla groups were able to get resources from kidnapping with a negligible effect on agricultural production and without deteriorating their relationships with peasants. The usual arrangement with rich farmers was extortion. Second, drug dealers gave a definitive stimulus to paramilitary groups as a private protection scheme against kidnapping. Third, some notorious drug lords began their criminal careers as kidnappers. Forth, drug traffickers have used kidnapping for purely political reasons and currently use it for collecting debts. Fifth, guerrilla groups easily substituted between kidnapping/extorsion and drugs to finance their military operations.
Indiscriminate massive kidnapping appeared when the group of middle class urban victims became depleted and city residents remained besieged within the city areas, by the late 90s.
Kidnapping has been mostly linked to politically motivated groups. Evidence not only from Colombia but also from Latin America helps corroborate this claim. A crucial issue in this activity is how to deal with agency problems. Several testimonies show that guerrilla leaders have been conscious of the need to indoctrinate their troops and to adopt highly centralized decision processes. Ideological training diffuses personal responsibility and reduces the individual tendency to appropriate big cash ransoms; it also reinforces the effect of class differences and helps assimilate the conduct to a tax-paying scheme by wealthy, alien victims. This may explain why non-political groups have not been able to kidnap in such a systematic manner as guerrilla groups.
2 - Evolution of kidnapping rates
Analysis of monthly time series –available at the national level from the early 60s- highlights two structural upward shocks in kidnapping rates. Both were followed by an also acute decline, defining two cycles that show little association with social or economic variables.
The first upsurge took place in the early months of 1987. This boom ended around 1991. Although many analysts have attributed the decrease to an anti-kidnap law that was passed in 1993, time series (intervention) analysis  shows that this was not the case: the decline began several months before that legal reform. Many possible factors were tested for their impact on this first high-kidnapping period. The only explaining variable that stands almost every statistical test was a change in criminal procedure at the beginning of 1987 –Ley 50- that was reversed in 1991.
The second kidnapping boom began by the end of 1997. Its length was also about three years. It was surprising to find that this second abrupt rise started months before the Pastrana administration conceded a demilitarized territory to guerrillas (FARC) to carry out peace talks. Its timing also precedes by a few months a significant change in kidnapping methods: the adoption by the two main guerrilla groups (FARC and ELN) of the “pesca milagrosa”, a term that was coined for random massive abductions on the roads. As the previous boom, this was a generalized phenomenon across the country.
Regional data -both at the departamento and municipio level- corroborate this assertion. The new rise cannot be adequately explained by exogenous social or economic aggregate variables. Nor any legal or procedural reform could be found to account for this second boom. The only reasonable explanation lies within the dynamics of the kidnapping industry itself. The significant reduction in the number of high-income victims eventually led perpetrators to look for a higher number of hostages. Simple statistical exercises based on a valuable database with a very brief description of almost every kidnapping reported from 1996 endorses this explanation . They also show that this endogenous strategic move backfired on guerrilla groups. Reactions, both private and public, against massive kidnappings help explain the last –and possibly definite- decline that began some three years ago. Although the recent and severe policy change for dealing with guerrilla groups –the so called “Política de Seguridad Democrática” of the Uribe administration- may have contributed to the decline, available data shows that it mainly reinforced a previous, autonomous, declining trend (see below).
3 - Kidnapping and armed conflict in Colombia
The kidnapping cycles, and above all the upward shocks in 1987 and 1997 were not isolated phenomena in a few leading regions. On the contrary, they were quite generalized across the country. Kidnapping rose in every departamento during the first boom and in all but one departamento during the second. Furthermore, all across the nine regions with the highest rates, where two out of three kidnappings took place since 1981, the evolution of kidnapping looks quite similar. In particular, the two sudden rises took place in all these departamentos.
Whatever the ultimate reason for the drastic upward and downward changes in aggregate rates, they strongly suggest that kidnapping in Colombia has been a highly centralized and coordinated activity, very much correlated with the main actors of the armed conflict. Three analyses corroborate this idea. First, the quite precarious evidence about kidnappings perpetrated by so called “common criminals” as opposed to guerrilla or paramilitary perpetrators. Second, cross-section statistical exercises show that the regional pattern of kidnapping activities has been significantly associated with the geographical distribution of armed groups.
Third, the peculiar dynamics of the last decline in kidnapping rates that followed the decision of guerrilla groups to adopt massive kidnappings on the roads as their main technique for taking hostages.
3.1 – The myth of “common kidnappers” 
Cross-section data by municipalities, available from 1992, shows that kidnapping occurs mostly in small towns and villages. This concentration in rural areas is different to the one observed for homicide or other crimes. In the last decade, this pattern reinforced, and the three main cities lose participation in the activity, while the smallest municipalities, with 25% of the population, accounted for almost 45% of the kidnappings. Per capita rates are four times as high in the later. Official figures about kidnaps allegedly committed by “common criminals” show the same pattern.
Although no information about the place of residence of the victims is available, some indirect indicators suggest that kidnapping, a rural activity, has affected mainly urban victims. For example, the percentage of victims working in agriculture or farming is low and has decreased to 5%.
The sudden decline in kidnappings in the last few years has been generalized, although anti-kidnapping efforts, led by the military, have been mostly against guerrilla groups in rural areas. These figures are more consistent with the scenario of guerrilla rather than “common criminal” perpetrators.
Available data suggests that authorities have been using “common kidnapping” as a default category where incidents with unknown perpetrators are summed up.
A solid piece of evidence against the scenario of “common kidnapers” is the strong geographical association between massive kidnappings on the roads –incidents explicitly committed by guerrilla groups- and selective abductions in municipalities, allegedly perpetrated by common criminals.
The average number of selective kidnappings in towns where “pescas milagrosas” took place is nine times higher than the one observed in municipalities where no massive kidnapping by guerrillas occurred. And this high ratio is observed regardless of the alleged perpetrator.
Furthermore, the number of massive kidnappings on the roads looks like an accurate predictor not only of selective kidnappings by guerrilla groups but also of those supposedly committed by common criminals in the surrounding areas.
Another reason to be skeptical about non-guerrilla kidnappers is that the attribution of responsibility to the incidents depends crucially on the status of the victim. Common perpetrators are more frequently associated with incidents in which the victims are still captive or never reported their case. On the contrary, among hostages that gave more information to the police –i.e. what was the ransom paid- the proportion of kidnappings by guerrilla raises significantly.
Exceptions to this trend –the more information from the victim the less proportion of common criminals- are the cases in which the hostage was liberated in a raid by the army or the police, or when the hostage run away. In the former situation, the proportion of guerrilla perpetrators is as low as the one observed in cases where the victims are still captive. In fact, a significant number of kidnappings attributed to common criminals is related to cases that ended up with a rescue of the hostage.
Various complementary rationales can be given for this relationship. The first one is the fact that rescuing a hostage is a far more difficult task in rural areas. The second reason is that the probability of releasing a hostage by a raid dramatically decreases with time, so aborted kidnappings are mostly incidents in which there was not enough time to go through the different stages of captivity. Available information from the database of incidents corroborates this hypothesis. Successful raids were usually made during the same day of the kidnapping, or one or two days afterwards. After a critical period of three days the proportion of liberated victims is significantly reduced.
On the other hand, the number of days of captivity is, on average, much longer for complete kidnappings –when the ransom is paid- or when the alleged perpetrators are guerrillas.
These observations give support to the widespread idea that the main role of common criminals in the activity is some kind of outsourcing: catching the victim in the urban areas. The “selling” of the hostage is done afterwards to the guerrilla so the victim can be kept captive in a rural area, where rescue operations are much difficult.
Time is not the only factor affecting the probability of rescue of a hostage. Some characteristics of the victim and of the incident itself also determine how the incident might come to an end. These variables also help discriminate kidnappings supposedly perpetrated by common criminals. It is found that the less ground there is for a political justification of an abduction, the smaller the probability of being attributed to guerrillas. Particularly reprehensible incidents, for example when the victims are kids, are generally ascribed to common criminals.
3.2 - Geographical distribution of kidnapping and conflict 
Cross-section analysis of kidnapping rates at “departamento” (state) level shows a rather heterogeneous geographical distribution. There is a slight correlation with economic activity and a stronger one with the presence of different armed groups. Rates are higher in those regions where multiple armed actors meet. Spatial-correlation exercises show spillover and contagion effects.
However, the regional concentration of kidnapping, as measured by Gini coefficients, has been continuously declining. A progressive spreading of the activity took place, from high-rate departments to their neighbors.
Data at the municipality level corroborates this scenario. The number of towns affected by this activity continuously increased until the end of the 90s, when rates began to fall. Furthermore, kidnapping rates became more equally distributed among municipalities. Cartographical analysis of municipal data corroborates the notion of a spreading of kidnapping all over the country. A similar trend has been observed for the armed conflict.
As mentioned above, kidnapping in Colombia has been mostly a rural activity with urban victims, which means that hostages are usually non-residents of the place where the incident occurs. This characteristic makes particularly difficult to posit and to interpret statistical correlations of kidnapping rates with local variables. Not to mention the basic difficulty of dealing with kidnapping rates in per capita terms. With this cautions in mind some cross-section econometric exercises were done. Statistical models show that the capacity of local factors to explain regional differences of kidnapping rates is not only small but has been decreasing. Also, that the regional patterns of kidnapping are mainly explained by the geographical expansion of armed groups.
4 – Massive kidnappings: a boomerang effect 
Kidnapping rates have been steadily falling for the last three years. Colombian authorities claim this is due to the “Política de Seguridad Democrática” (PSD) of the Uribe administration. It seems too early to make a systematic evaluation of Uribe’s PSD. But available information shows that the decline in kidnapping began earlier.
In 1998, the two biggest guerrilla groups (FARC and ELN) almost officially adopted a new technique, massive kidnappings on the roads, the so called “pescas milagrosas”. There is no available information about victims’ financial means, but indirect estimates based on their occupation suggest that this tactical change of guerrillas was the result of the progressive draining of wealthy hostages. Average estimated income of victims continuously decreased, and testimony about kidnappings of really poor people became less rare. Another symptom of the depletion of traditional victims was the increasing share of vulnerable groups such as infants or elderly people.
For a couple of years massive kidnappings provoked soaring rates. The share of foreigners, the most looked-after kind of hostage, also increased. Average ransom paid by foreigners has always been greater than local’s. The adoption of the new technique was far from marginal. By 2002, for example, half of ELN’s victims were captured in massive kidnappings.
Selective kidnappings also increased dramatically, following the same geographical pattern of massive ones. As was mentioned above, regional information about massive versus selective kidnappings challenges the conventional view of an activity where common criminals play a leading role.
Quite soon, however, the new kidnapping technique backfired. It was very costly for guerrillas in political terms. And it also facilitated the state’s military response against perpetrators. Successful rescue raids increased both for massive-kidnapping and for selective incidents in nearby locations. In spite of the fact that selective and massive kidnappings are highly associated, and that rescue operations are also positively correlated with the latter, some deterrence effect of rescues can be detected.
Two issues require further investigation. First, the background, rationale, lobbying actors and main consequences of Ley 50 of 1987, the criminal procedure reform that shows a statistically significant effect on kidnapping rates. Second, a close look at the recent response to massive kidnappings. In particular, it is worth knowing if prosecutors and judges are backing military raids and if victims are bringing more formal criminal charges.
* This is a summary of findings of an ongoing research, “Kidnapping in Colombia”, financed by the Guggenheim Foundation.
 For the complete set of references See Rubio, Mauricio (2003) “Del rapto a la pesca milagrosa. Breve historia del Secuestro en Colombia”. Documento CEDE 2003-36
 For details time series estimation procedures see Rubio, Mauricio and Daniel Vaughan (2004). “Evolución del Secuestro en Colombia. Un análisis de series de Tiempo”. Documento CEDE Forthcoming
 See section 4
 See Rubio, Mauricio (2004) “El mito colombiano de las bandas de secuestradores comunes”. Documento CEDE Forthcoming.
 See Rubio, Mauricio and Daniel Vaughan (2004). “Geografía del secuestro y del conflicto en Colombia”. Documento CEDE Forthcoming
 See Rubio, Mauricio (2004). “La caída del secuestro o el milagro de las pescas”. Documento CEDE Forthcoming