Why Rebels Stop Fighting: Organizational Decline and Desertion in Colombia's Insurgency

Abstract Desertion, or the unauthorized exit from an armed group, has major implications for counterinsurgency, war termination, and recruitment dynamics. While existing research stresses the importance of individual motivations for desertion, organizational decline, in the form of military and financial adversity, can also condition desertion. Organizational decline undermines a group's instruments to channel individual preferences into collective action. These instruments include selective incentives, ideological appeal, and coercion. When the binding power of these instruments diminishes, individual desires start to dominate behavior, making desertion more likely. The Revolutionary Armed Forces of Colombia (FARC) insurgency is used to examine this argument with a multimethod approach. First, a quantitative analysis employs unique data on more than 19,000 reported FARC deserters from 2002 to 2017, provided by the Colombian Ministry of Defense. Guarding against threats to causal inference, statistical analysis indicates that organizational decline drives desertion. Second, a qualitative analysis uses a large body of detailed reports on interviews with deserters conducted by Colombian military personnel. The reports demonstrate that organizational decline weakens selective incentives, group ideology, and a credible coercive regime, and fosters desertion through these mechanisms. These findings provide key insights for policymakers, given that desertion can both contribute to ending conflict and accelerate the recruitment of new combatants.

who deserted, 4 the desertion of locally recruited Islamic State ªghters accelerated the group's downfall. 5 Moreover, desertion can determine the nature of postconºict institutions. Governments involved in counterinsurgency in Nigeria and the Philippines have set up reintegration programs to encourage desertion among rebels. 6 Such programs can pave the way to military victory, but can later contribute to biased institutions favoring the conºict winners, undermining the prospects for long-term peace. 7 Much of the existing literature on desertion focuses on individuals' motivations-for example, a desire to see one's family, difªculty adjusting to a military lifestyle, personal insecurity, disillusionment with the group's ideology, or feelings of being an outsider. 8 Other research looks at aspects of group life-for instance, unit cohesion, loyalty among comrades, and mutual expectations among combatants. 9 A few scholars have examined the inºuence of structural factors; armed forces in democracies and countries with integrated ethnic groups, for example, have lower desertion rates than autocracies and countries that engage in ethnic discrimination. 10 centralized and hierarchical structures. Although both kinds of organizations may have different means to deal with deserters, they use similar instruments to bind their members to the collective, 15 and thus are similarly vulnerable to military and ªnancial decline. Our argument is less applicable to network-type organizations, such as transnational terrorist groups, criminal syndicates, and insurgencies with networked structures. 16 Members of such groups typically operate in small units or cells. At this level, collective action may be less problematic, as individual members can, in fact, make a difference, with each requiring fewer organizational incentives to do so. 17 Also, the collective action problem does not exist in the case of purely criminal organizations, as their primary goal is to produce private, not public, goods.
To test our theory, we examine a highly centralized and hierarchical communist rebel army, the Revolutionary Armed Forces of Colombia (FARC), 18 which existed from the 1960s until its collective demobilization in 2017, after the conclusion of a peace agreement with the Colombian government. We focus our analysis on the period 2002 to 2017, during which more than 19,000 members left the FARC. Our multimethod analysis is based on records of the government's Program of Humanitarian Attention for Demobilized Individuals (known by the Spanish acronym PAHD). The quantitative analysis uses systematic information on all FARC deserters provided by the PAHD. Employing statistical analysis, we ªnd that in the case of the FARC, organizational decline indicators-including leadership decapitation and income ºuctuations linked to the coca economy-drove the decision to desert. To probe the plausibility of the mechanisms operating between decline and desertion, we base our qualitative analysis on a large number of detailed reports on individual deserters created by the PAHD. We ªnd that changes in selective incentives, ideological appeal, and coercive behavior shifted the combatants' attention away from the collective and toward the individual, fueling the desire to desert.
The article makes three main contributions to academic research. First, it provides a broader framework for understanding the desertion phenomenon. Earlier theories on desertion focused on individual motivations, group cohesion, and, to a lesser extent, structural conditions. Organization-level dynamics-our main theoretical focus-help embed individual motivations in a meaningful context, where decline works as a conditioning factor.
Second, the exceptionally ªne-grained data from the Colombian conºict, including a treasure trove of qualitative material, allow for novel empirical insights. Given the near-absence of systematic evidence, desertion from rebel groups is difªcult to study. Although the dataset we use may contain some "fake" deserters who sneaked into the program, and may exclude some real deserters who did not want to hand themselves over to government authorities, it is to our knowledge the most complete dataset on desertion available for any internal conºict.
Third, the article's ªndings open new avenues for related research. For example, earlier research identiªed a "rebel resource curse," 19 whereby "rich" rebels attract opportunists who give up easily if the organization does not work to their beneªt. 20 Examining ªne-grained data on desertion provides additional insights into this supposed curse. Moreover, the article has relevance for the study of how civil wars end. 21 Organizational decline and desertion can help explain why one side emerges victorious and why leaders of armed groups change their minds about participating in negotiations. Although many post-Cold War conºicts ended only with a military solution or after the signing of a peace agreement, some simply dissolved as the parties gradually stopped ªghting. 22 A research agenda based on the study of organizational decline can help explain these instances of "organic" conºict termination.
The remainder of the article proceeds as follows. First, we discuss our theory in greater detail, focusing on the link between organizational decline and the decision of some armed group ªghters to desert. Second, we introduce the case of the FARC insurgency and discuss its suitability for our argument. Third, we present our empirical analysis in two steps: after quantitatively examining the relationship between decline and desertion, we qualitatively analyze PAHD's desertion reports to probe the mechanisms linking organizational decline to desertion. We conclude with a discussion of the implications of our ªndings for research and policy.

A Theory of Desertion
We argue that organizational decline increases the likelihood of desertion by members of armed groups, because the instruments that allow such groups to overcome collective action problems-selective incentives, ideological appeal, and coercion-are less effective during periods of decline (ªgure 1). 23 Armed groups use selective incentives to compensate members for their participation in the group. These incentives are important not only for attracting new members, but also for keeping existing ones, who depend on the organization for their subsistence and protection. 24 Salaries and opportunities for looting are the most common forms of selective incentives in armed groups.
Armed groups use ideology to appeal to their members. Ideological appeal is distinct from a selective incentive that attracts rational agents, as it establishes a normative basis for action that goes beyond self-interest. 25 Although not very prominent in original collective action theory, which was largely apo-  litical, later work stresses the importance of ideological appeal in fostering collective action in political conºict. 26 We understand ideology as a set of "ideas that identify a constituency, the challenges the group confronts, the objectives to pursue on behalf of that group, and a (perhaps vague) program of action." 27 Group members internalize ideas of what an organization stands for, and ªghts for, through training and daily practice. 28 Ideology also can create a set of common expectations that underpins a shared identity and solidarity among group members. 29 Hence, an armed group's capacity to overcome collective action problems depends on its ideological appeal.
Armed groups use coercion to make their members obey, often through the imposition of regulations and punishment of unwanted behavior. Members seeking to leave an armed organization without permission pose a particularly common challenge for both state and nonstate organizations; often, these individuals are sentenced to death. 30 Thus, coercion acts as a negative motivator, unlike positive motivators such as selective incentives and ideological appeal.
Together, selective incentives, ideological appeal, and coercion allow an organization to channel individual preferences into collective action. 31  instruments become less effective, however, once an organization goes into decline. Classic literature on organizational decline deªnes it as the "deterioration of the resource base and performance of an organization," caused by either environmental shocks or internal and managerial factors. 32 The related seminal argument by Albert Hirschman connecting organizational decline to exit applies, in principle, to any type of organization. 33 Companies may lose clients if the quality of their products deteriorates; states may experience increased emigration if their economies begin to decline; and armed organizations may lose members in periods of adversity. Despite the general applicability of Hirschman's theory, we need to specify the relevant forms of organizational decline in armed organizations and how these forms of decline affect the groups' instruments to promote collective action. Here, we focus on deterioration in armed groups' military performance and ªnancial resources. 34 Although military and ªnancial decline are mutually reinforcing processes, we separate them for analytical purposes.
First, military performance is the key indicator of success for any armed group. Armed organizations are created to achieve objectives such as secession from the state, the overthrow of an opponent's government, or the control of territory. Military confrontation is the method of choice to obtain these objectives, which depends crucially on relative military performance vis-à-vis enemies. Weakening military performance thus implies decline.
In theory, decline in military performance can affect all three instruments designed to promote collective action. It undermines the group's capacity to protect its members, an important selective incentive in conºict regions, 35 as it exposes members to a higher personal risk of being injured and killed. 36 Military decline makes the achievement of a group's common goals less realistic, which reduces its ideological appeal, because, as one observer notes, "even the most die-hard supporters can be discouraged by the failure of the revolution to materialize." 37 Also, military decline may limit an organization's control and oversight capacity, reducing its ability to sanction unwanted behavior. 38 When the capacity to control members declines, members' perceived certainty of punishment for wrongdoing may decrease and undercut the deterrence effect of punishment. In response to decline, armed groups may increase the severity of punishment, much like governments do when faced with a crime wave. Extensive research from criminology, however, suggests that the certainty of punishment is a more effective deterrent than is its severity. 39 Second, ªnancial resources are necessary for the operation of any organization, including armed groups. 40 Much of the economic literature on the causes of war has focused on the availability of lootable resources. 41 Our argument is about the sustainability of armed groups. Quite evidently, members need to be fed and equipped; many receive salaries, requiring the group to have sufªcient funds. Reduced ªnancial resources thus imply decline.
In theory, decline in an armed group's ªnancial resources has important consequences, especially for selective incentives, as it makes camp life harder. Decreasing ªnancial resources may cause dissatisfaction among members who depend on the organization for their subsistence, 42 or who were promised that the group would cover their living expenses. 43 In addition, ªnancial decline has downstream consequences on military performance and overall organizational capacity, potentially affecting the outlook of an organization to reach its common goals, and thus limiting its ideological appeal. Finally, decline may hinder an organization's ability to coerce its members.
Once these three instruments lose their effectiveness, group members may start to refocus on personal goals and desires. Although some members may miss their families or may want to have a family, as long as life in the collective is satisfactory, such feelings may remain latent. 44 For others, a desire to change one's course in life and pursue outside economic opportunities may become more pressing. 45 Accordingly, we agree with earlier research that individuals must be motivated to leave their group. We argue, however, that organizational performance conditions individual behavior and that desertion is more common during times of decline, when the instruments for promoting collective action are weakened (see ªgure 1). At this point, it is important to clarify the boundaries of our argument. First, our argument captures the difference between decline and "non-decline"; it is silent about the effects of other types of performance on the decision to desert, such as organizational rise or stability. 46 To be clear, desertion should thus be higher during periods of decline than during any other time.
Second, the behavioral complement of desertion (or leaving) is retention (or staying). Thus, our theory also should apply to retention, but it does not make predictions about the related phenomenon of recruitment. The relationship of recruitment to organizational decline is less clear. On the one hand, armed groups may, for strategic reasons, choose to increase recruitment to compensate for the effects of desertion. On the other hand, the decisionmaking of potential recruits depends on a range of factors that are often unrelated to the organization they are about to join. 47 Third, although we argue that organizational decline drives desertion, desertion also can drive decline, creating an endogenous relationship. For example, the departure of important leaders may motivate some members to desert, resulting in a self-reinforcing process. 48 Distinguishing between desertion and decline thus poses a challenge for empirical analysis. Focusing on exogenous indicators of decline and using a multimethod approach help circumvent this challenge. Theoretically, our argument is limited to the impact of decline on desertion.  50 We should thus be able to capture sufªcient variation in terms of organizational performance.
The FARC emerged from earlier communist guerrilla groups and had been involved in a low-intensity conºict with government forces since at least 1964. 51 In the second half of the 1990s, it had accumulated so much power that it could hold scores of regular soldiers in captivity and dominate large areas of Colombia. Its stated goal always was to take government power, and this was the period when it was closest to achieving it. After this period of growth in 2002, the FARC had roughly 21,000 armed guerrilla members and many thousands of urban militia members.
When efforts by the FARC and the government from 1998 to 2002 failed to reach a peace agreement, the stage was set for a new period of confrontation. The government received large amounts of military support from the United States, as part of "Plan Colombia," and started to push the FARC into increasingly remote areas of the country. It is in this context that the government strengthened its policy of providing members of nonstate armed groups, including the FARC, with incentives to desert. Rebels willing to do so were offered amnesty and an opportunity to join a reintegration program. They also received ªnancial compensation in exchange for providing authorities with strategic information. 52 The Program of Humanitarian Attention for Demobilized Individuals, located in the ministry of defense, was in charge of implementing the government's desertion policy, which included a sophisticated public relations campaign, television advertisements, and radio messaging that could reach rebels even in remote areas. 53  members of all nonstate armed groups, not just the FARC, were invited to hand themselves over to any civilian or military authority in the country and register as desmovilizado (demobilized). Among the more than 30,000 individuals who deserted during this period were more than 19,000 members of the FARC.
The FARC reached its lowest point after a series of military defeats in 2008. Emboldened Colombian generals even talked about the "end of the end." 54 Starting in 2008, however, the FARC reinvented itself with the adoption of the Plan Renacer (Rebirth Plan) and later the Plan de la Segunda Independencia (Second Independence Plan). Both plans focused on a return to classic guerrilla tactics, less direct military confrontation, and the recruitment of new combatants from among the FARC's clandestine urban networks and youth in its zones of inºuence. At the same time, the government's military strategy was losing effectiveness, prompting the new government of Juan Manuel Santos to start negotiations that would ultimately lead to a peace agreement and the collective demobilization of the FARC in 2017.

Analyzing the Relationship between Decline and Desertion
To examine the relationship between organizational decline and desertion, we quantitatively analyze a dataset that includes all registered FARC deserters. If our theory is correct, we should observe elevated numbers of desertions during times of organizational decline.

data on desertion
The desertion data we used are based on interviews conducted by the PAHD with individuals who deserted from Colombian armed groups between August 2002 and August 2017. Among other things, the data contain information about demographics, the deserters' group, their duration of membership, and whether they were certiªed as desmovilizado and deserving of government beneªts. The interviews were conducted to verify whether a person was a member of an armed group and to obtain information relevant to the government's future military operations. The resulting dataset, which we received as an Excel spreadsheet, contains 33,592 individuals, of which 3,816 were not certiªed and 1,042 were arrested. 55 Of those who were certiªed but not ar-International Security 45:4 178 54. John R. Thomson and Dorotea Laserna, "El Fin del Fin," National Review blog, July 11, 2008, https://www.nationalreview.com/2008/07/el-ªn-del-ªn-john-r-thomson-dorotea-laserna/. 55. Combatants who had been arrested could join the desertion program if they were not responsible for a crime other than participation in the group. We exclude them from our analysis. rested, 19,504 belonged to the FARC; 56 of those, 4,236 were female, 57 and 2,801 deserted before they were eighteen years old.
The PAHD dataset also contains information about members' reasons for leaving. 58 Our theory is, in principle, compatible with any individual motivation, as we argue that organizational decline brings personal desires and preferences to the forefront and hampers collective action, but we do not specify which personal preferences become more prevalent in times of decline. For a fuller understanding, we list individual motivations for desertion, comparing rank-and-ªle members to members with some level of command authority (table 1). 59 A desire to change one's life and escape abuse within the group are the most commonly reported motivations. Armed pressure and missing one's family also are relatively common. Together, these four categories account for roughly 80 percent of members' responses. Members with some level of command authority mention ideological disagreement more often than do rank-and-ªle members, who more frequently mention a desire to change their life circumstances or escape abuse in the group; otherwise, the two groups are similar.
In our quantitative analysis, we focus on members' date of desertion and area of activity while in the group. The date of registration included in the dataset indicates when an individual was ªrst introduced into PAHD's information system, usually a few days to a couple of weeks after having de-   After 2008, the FARC regained momentum, though it never reached its previous levels of military performance. 61 From 2008 to 2012, the number of desertions decreased, with some exceptions, including a peak in 2010 after a second successful hostage rescue operation (Camaleón) and the killing of a key FARC leader known as Mono Jojoy. Membership numbers remained relatively stable during this period, as the FARC began to quickly replace deserters and those who had been arrested and/or killed with clandestine urban network members and new recruits.
With negotiations under way, FARC desertions began to stabilize starting in 2012, as most members now had the prospect of obtaining a potentially better deal through collective demobilization, without the need to break ties with the organization. Still, the number of desertions remained sizable, which suggests a base level of desertions largely independent of organizational dynamics.
In our quantitative analysis, we use the location of the deserter's activity, rather than the location of desertion. The location of desertion is not always related to members' activity in the group, because rebels often travel to a safe area before deserting. For example, 2,750 members of the FARC handed themselves over in Bogotá, even though the FARC had almost no active armed presence in the capital. The location of activity is relevant for our analysis because we want to examine the conditions of this geographic area. Although rebels constantly moved around, almost all FARC deserters (97.1 percent) indicate a department of activity in their interviews (Colombia has thirty-three departments); far fewer (59.6 percent) indicate a sole municipality of activity, which explains our focus on departments in the statistical analysis.
The FARC had an estimated 21,000 armed guerrilla members in 2002, making the total number of FARC deserters (more than 19,000) seem extremely large. Ideally, we would be able to compare desertion numbers provided by the PAHD with information collected by the FARC itself. The group did not keep track of desertions, however, especially after the central command ordered subordinates to destroy internal archives following the recovery of FARC computers by the Colombian armed forces during operations in 2008. 62 Still, we can critically examine the extent to which the PAHD data accurately represent FARC desertion numbers.
First, the data may contain false positives: individuals who acted as rebels, but were not part of the organization. The PAHD certiªed not only armed guerrilla members, but also individuals who belonged to supporting mili- tias (6,009 FARC deserters stated that they were so-called milicianos). 63 Although there are cases of desertions of individuals at the margins of the organization, members of the clandestine urban network, for example, do not constitute false positives, but instead demonstrate the different roles in an armed organization. 64 The problem of false positives is mitigated by the PAHD's certiªcation process, which involved several government institutions. Nevertheless, there may still be an unknown number of individuals who lied their way into the desertion program to receive reintegration beneªts. 65 Those who wanted to act credibly as FARC members, however, needed to have intimate local knowledge. Therefore, they had to come from areas where the FARC had a presence, and they had to have had some exchanges with the group. This kind of false positive may affect the total number of deserters, but less so the distribution of deserters across the country.
The Colombian armed forces may have further had an interest to show success by inºating the FARC's desertion numbers. During the 2000s, several military units engaged in extrajudicial executions of civilians to increase the body count of rebels, known as the infamous falsos positivos scandal. 66 To date, however, there is no clear evidence that the armed forces similarly inºated the FARC's desertion numbers. Also, the FARC did not use the desertion program to encourage the departure of injured and older members; 67 in fact, it saw the program as a counterinsurgency tool. 68 Second, there may be false negatives: rebels who deserted but were not registered as such, including individuals who deserted without registering with the government's program or who switched groups, for example, from the FARC to the National Liberation Army, another insurgent group in Colombia. 69 Although there were such instances, we have no information on the magnitude of this phenomenon. The absence of this information may pose a problem for individual-level analysis, as deserters who did not register may systematically differ from other deserters. Our analysis operates at the aggregate level, however, and we are unaware of systematic clustering of anonymous deserters in a given time period or area. Overall, underestimating the number of FARC desertions would make our estimates smaller and thus more conservative.
Third, the high number of FARC deserters can be explained by the group's capacity for rapid replacement. Among all FARC deserters, 5,255 (more than 25 percent) had been in the group for less than one year, demonstrating the group's ability to constantly recruit new members, including through the forced recruitment of minors. 70 This ability to rapidly replace members shows that recruitment is responsive to a process different from desertion, although the two may be intertwined. 71 In sum, the PAHD data may contain some reporting biases, but they remain a largely accurate representation of the FARC deserter population. 72 independent variables Theoretically, military and ªnancial decline should drive desertion numbers in centralized armed groups, but empirical analysis requires a case-speciªc operationalization of these concepts. Also, because desertion can feed back into decline, we need indicators that are both valid representations of FARC's organizational decline and plausibly exogenous to the behavior we want to explain.
In examining military performance, we focus on the independent variables of killings of FARC leaders and rescue operations of kidnapped hostages with national-level importance, as these were broadly and quickly noticed by most FARC combatants. First, we include all killings by the armed forces of  (November 4, 2011). This variable takes a value of 1 for the three months after the killings and 0 otherwise (in our online appendix 3, we use different time windows). Second, we use major rescue operations of kidnapped hostages held by the FARC: Operation Jaque (July 2, 2008) and Operation Camaleón (June 13/14, 2010). Again, we code the three months after these events as 1 and 0 otherwise (we provide different time windows in online appendix 3). In the analysis below, we show results for aggregated indicators of FARC leader killings and government rescue operations (table A8 in the online appendix includes results for single incidents).
Killings of leaders and rescue operations are valid and plausibly exogenous indicators of military decline. Both were repeatedly mentioned in the PAHD reports as blows to the morale of FARC troops. Operation Jaque was celebrated as a particularly major demoralizing blow. 73 From a combatant's perspective, and among the broader Colombian population, all these events capturing military decline were surprising and shocking, thus generating a plausibly exogenous source of variation. They were partially planned, however, with information provided by deserters, 74 and by other intelligence sources. Even though the planning of such operations may have taken months or years, one could therefore argue that desertion may not only be a result of these events; it may also be a cause. A statistical association may thus reºect reverse causality. In our analysis, we guard against this concern by focusing on the aftermath of the events and adjusting for previous levels of desertion. One might further argue that successful major operations may be the culmination of broader military campaigns, and thus a relationship with desertion may be the result of omitted variable bias. We adjust for regular military operations, including both operations taken at the initiative of government forces and any type of contact between Colombia's armed forces and the FARC, to mitigate this concern. Regarding the FARC's ªnancial resources, we focus on two independent variables related to the coca economy, the group's primary source of income. Because coca is essentially an export product, we use exchange rates between the Colombian peso and the U.S. dollar as our ªrst indicator. (The use of cocaine in Colombia has been low and stable over the past two decades, 75 and the FARC generally prohibited its use in areas where it had inºuence. 76 ) Similar to exporters of other goods, participants in the coca economy in Colombia can therefore beneªt from a weak peso. On the other hand, the same quantity of exported Colombian coca will generate less revenue if the peso becomes stronger against the dollar. 77 Hence, we use the amount of Colombian pesos (1,000 per U.S. dollar) to capture this source of FARC revenue. Second, we use yearly prices for one gram of cocaine in the United States, as reported by the United Nations Organization on Drugs and Crime for the period 2002 to 2016. Lower-end market prices may lead to lower revenues for participants in the coca economy, including the FARC.
Exchange rates and cocaine prices are valid indicators of ªnancial resources. Both tap into the FARC's key income source. To sustain its activities, the group initially relied mainly on local extortion and engaged in kidnapping for ransom (especially around the year 2000). It also levied taxes on illegal mining, including gold mining, 78 and, most importantly, drug production and trade. The Colombian military estimated in 2005 that, at its ªnancial peak, the FARC had an annual income equivalent to 1 percent to 2 percent of Colombia's gross domestic product (or about $1 billion to $2 billion), roughly half of which was derived from the coca economy. 79 As the PAHD reports show, the FARC professionalized its management of the coca economy over the years. Many members were involved in it, with some urging peasants to cultivate more or less depending on price levels; managing and protecting cultivation sites;

Why Rebels Stop Fighting 185
overseeing peasants' work; buying basic coca paste from peasants; selling to drug trafªckers; and levying taxes on participants. 80 Also, the dollar-to-peso exchange rate shows a development largely parallel to cultivated coca areas in Colombia, suggesting that the coca economy is responsive to exchange rates (see ªgure A1 in the online appendix).
Both indicators are plausibly exogenous to desertion in Colombia. Exchange rates depend mainly on domestic monetary and ªscal policies and on international macroeconomic dynamics. Cocaine prices may be partially driven by supply from producing countries, and supply may be related to conºict in Colombia. In the analysis, we therefore adjust for coca production in Colombia to control for supply to the cocaine market. Cocaine prices in the United States, however, are still plausibly exogenous to desertion in Colombia, as they are driven mainly by domestic market conditions. 81 Further, although higher prices reduce the demand for cocaine, price changes clearly outweigh changes in demand. 82

statistical procedures
We employ linear regression analysis, using the number of deserters in a given department and month as the dependent variable. To account for omitted variable bias stemming from time-invariant characteristics and common shocks, we use two-way ªxed effects. 83 Department ªxed effects adjust for characteristics such as conºict history, terrain ruggedness, and remoteness. Month ªxed effects adjust for common shocks to all departments, including elections, international dynamics, and changes in national security strategy. 84 The plausibly exogenous indicators for military performance and ªnancial resources described above have temporal, but no spatial, variation. These variables would thus be collinear with month ªxed effects. To solve this issue, we interact our independent variables with a spatially varying component. For this purpose, we create a dummy variable that captures the spatial distribution of FARC areas. By including this component in an interaction, we effectively compare relative desertion in terms of decline in areas with and without a traditional FARC presence. Doing so makes sense, because desertion from traditional FARC areas should increase more during times of decline than should desertion from non-FARC areas. The coding of this variable is based on whether the FARC and the government's armed forces had any armed contact before 2002, when our study period starts, using data from the National Historical Memory Center (known by the Spanish acronym CNMH). Departments with numbers of contacts below the median receive a value of 0, and departments with a number above the median receive a value of 1. The variables for military performance and ªnancial resources are interacted with this FARC area dummy variable. In further analysis, we also interact the ªnancial resources variables with an indicator for land suitability for coca cultivation, based on climate and soil conditions (see table A20 in the online appendix). 85 Additionally, we show the results of an analysis including only the main independent variables without an interaction term or month ªxed effects (see tables A9/A23 in the online appendix).
Because we use two-way ªxed effects, we include only control variables that vary across time and space (for more details, see online appendix 2). We control for yearly population per department (National Statistics Department-DANE); total yearly gross domestic product per department (DANE); yearly coca cultivation area (Colombian Observatory of Drugs); and indicators accounting for battleªeld dynamics based on a CNMH event dataset, including department-month counts of any contact between government forces and the FARC, and of military operations against the FARC initiated by the Colombian armed forces. We also use this last variable, government-initiated operations against FARC, as a separate and broader indicator of worsening military performance of the FARC vis-à-vis government forces (see table A18 in the online appendix). 86 In some models, we further include a lagged dependent variable. 87 In doing so, we capture the contagious nature of desertion, as the desertion of some members may cause others to take the same path. 88 and ªnancial resources, on desertion. We use killings of FARC leaders and government rescue operations of kidnapped hostages as proxies for military decline. Table 2 displays the results of the relationship between military decline and desertion using linear regression analysis. Models 1-3 focus on killings of FARC leaders, and models 4-6 focus on rescue operations. Models 1 and 4 include two-way ªxed effects, but no control variables; models 2 and 5 additionally include control variables; and models 3 and 6 also include a lagged dependent variable (hence, the slightly reduced number of observations).
Both indicators for military performance interacted with FARC area have consistently positive coefªcients that are statistically different from 0, meaning that unexpected, successful military operations against the FARC increase the number of desertions. Substantively, the number of desertions increases by more than 2 per department-month for leader killings (model 2) and almost 4 per department-month for rescue operations (model 5). These are large coefªcients, considering that the mean level of FARC desertion is 3.1 per department-month (see table A3 in the online appendix). Desertions thus roughly double in the wake of killings and rescue operations in traditional FARC departments. From these results, we infer that decline in military performance leads to desertion, which is in line with our theorizing.
We operationalize the FARC's ªnancial resources using two indicators for income from the coca economy. Table 3 displays the results for the relationship between ªnancial resources and desertion. Models 1-3 focus on dollar-to-peso exchange rates, and models 4-6 focus on cocaine prices in the United States. Note that the models for cocaine prices are assessed at the department-year International Security 45:4 188 level, given that these prices are reported yearly (hence, the smaller number of observations and years, not month ªxed effects).
Both indicators for ªnancial resources interacted with FARC area show consistently negative coefªcients that are statistically different from 0, meaning that less income from the coca economy increases the number of desertions. Substantively, the number of desertions rises by more than 2 per departmentmonth for every 1,000 peso reduction in the value of the dollar (model 2). The dollar cost is between 1,733 and 3,357 pesos during the period of study. With regard to cocaine prices, the number of desertions increases by more than 1 per department-year for every dollar that a gram of cocaine loses in value in the United States (model 5); the mean level of desertion per department-year is 30.9, and cocaine prices ºuctuated between $76 and $98 per gram. From these results, we infer that declining ªnancial resources lead to an increase in the number of desertions, an inference that accords with our theorizing. 89

Examining the Three Mechanisms Linking Decline to Desertion
Next, we qualitatively analyze detailed PAHD desertion reports. If our theory is correct, we should ªnd instances in these reports where, ªrst, organizational

Why Rebels Stop Fighting 189
89. Extensive robustness tests using different speciªcations are consistent with these ªndings (see online appendix 3). We also conducted an analysis for subsets of deserters and ªnd that the impact of organizational decline is largely generalizable across the organization. Decline is positively related to both desertion of armed guerrilla members and clandestine militia members (see tables A10, A11, A25, A26); rank-and-ªles and members with some level of command (see tables A12, A13, A27, A28); and short-term and long-term members (see tables A14, A15, A29, A30). decline affects selective incentives, ideological appeal, and coercive behavior; and, second, changes in these three mechanisms lead to desertion.
interview evidence on desertion The Colombian military interviewed all FARC deserters shortly after they had left the group and produced detailed and classiªed reports (for ethical considerations, see online appendix 6). We accessed a large sample of these reports through two channels: a 2010-12 repository, where we conducted a full review of each report on about 800 deserters, 90 and a browsing terminal at the PAHD ofªces. Full-text searches using key terms related to our mechanisms yielded an additional 41 deserters, mainly from the period 2002 to 2010.
A typical report is twenty to forty pages long, with most of the text covering information relevant for the Colombian military's operations against the FARC. 91 The reports do not focus on organizational decline, collective action instruments, and desertion. Instead, they contain lengthy descriptions of deserters' time in the group and several questions directly relevant for our study, including: Why did the deserters join the FARC? What made them stay? Why did they desert? What did they ªnd most demoralizing? What were the strengths and weaknesses of the organization? Some reports also cover desertion promotion campaigns, troop morale and discipline, drug-trafªcking activities, kidnap victims, and executions within the group. 92 The context in which these reports were produced is worth noting. 93 The interviews were conducted by military personnel, resulting in a clear power asymmetry between the two sides (further discussed in online appendix 6). 94 Given that the sides were antagonists until that point, we assume only a minimal level of trust. Both sides may have an interest in concealing certain information about their conduct and stressing other information. 95 On the one hand, deserters may want to hide incriminating information, yet reveal enough information to prove that they were FARC members and be certiªed as desmovilizado, thus entitling them to reintegration beneªts. Blatantly lying about factual events, therefore, could be risky. On the other hand, military staff International Security 45:4 190 90. In this repository, we found 1,033 documents; however, up to 20 percent of these documents are duplicates with different names and formats. Hence, they correspond to roughly 800 deserters. 91. An attempt to standardize guidelines in 2006 did not translate into homogenous reports, as interrogators prioritized some aspects to ªt their short-term military interests. 92. We support our ªndings with direct quotations from these reports and references to longer text passages, blinding all identifying information (reported in online appendix 7). may have an interest in showing that the government program to convince FARC members to desert is working and have leeway in formulating their reports. They might use terms such as "terrorist group," "alias," or desmovilizado (instead of deserter), and employ the third person, potentially distorting the original narrative. Aware of these issues, the military requires reporters to attest that the content of the reports is more or less credible. Given this context, we focus mainly on narrated experiences and motivations that are likely quoted directly from the deserters' statements.
The PAHD reports have several advantages. First, the number of reports and sheer wealth of information would be hard to reproduce with any other datacollection effort. Second, the interviews are conducted right after a FARC member has deserted, which limits the possibility of recency bias and distorted memories. 96 Deserters are asked about an experience that is still fresh in their mind and not inºuenced by later experiences. Third, interview conditions are similar for all deserters, which makes the reports largely comparable. Fourth, the reports allow us to go beyond narrow operationalizations and use a broader understanding of organizational decline to examine both desertion and retention.
At the same time, there are potential limitations. First, restricted access to the reports did not allow for systematic sampling. Although our selection provides sufªcient information to examine relevant mechanisms, it is not representative of the larger universe of deserters. We have no reason to believe, however, that PAHD ofªcials preselected reports that were most favorable to them or that biased our sample in some other way. Second, our sample does not include members who stayed until the FARC's collective demobilization in 2017. This exclusion becomes a limitation when assessing why some members stayed, rather than left, the group. Third, most of the reports are from 2010 to 2012. Parts of this period were clearly marked by organizational decline, particularly after the 2010 killing of FARC leader Mono Jojoy. Also, it was a time when the coca economy experienced relatively low proªtability. However, we have only a few reports from 2008 and preceding years that correspond most clearly with a phase of organizational decline. All reports contain numerous references to earlier developments starting as early as 2002 (but with a focus mainly on 2010 to 2012), thus partially mitigating this lack of information. Fourth, even if deserters are inºuenced by organizational decline, they may not be aware of it, or they may not explicitly mention it in the interviews. To partially address this concern, we include additional material drawn from sources offering ªrsthand accounts of deserters, including newspaper articles, scholarly work, and a survey with FARC former combatants. 97 analysis strategy We identify evidence in the PAHD reports that captures the three mechanisms that we argue link organizational decline to desertion. The evidence and the strength it provides to our theory depend on the structure of our argument. 98 We divide our argument into two parts of a causal chain (see ªgure 1). In the ªrst part, we argue that organizational decline undermines selective incentives, a group's ability to offer members an ideology, and its capacity to coerce members. This part of our argument amounts to a relationship of sufªciency, meaning that decline is invariably followed by weaker instruments to promote collective action. The implication is that whenever we ªnd reports about organizational decline, we should observe a weakening of some of the instruments used to promote collective action. Observing a strengthening of any of the collective action instruments as a result of decline would contradict our theory. The absence of contradictory evidence is thus particularly important for this part of our argument.
In the second part of the causal chain, we argue that when an armed group's instruments to promote collective action are weakened, desertion becomes more likely. In this probabilistic relationship, weakened collective action instruments raise the likelihood that group members will desert. The implication is that whenever we ªnd reports describing a weakened collective action instrument, we may observe cases of desertion. Cases of deserters who left for reasons unrelated to organizational decline do not, however, contradict our theory. In fact, deserters need to be individually motivated to take action. The presence of conªrmatory evidence thus increases the plausibility of the second part of our argument, but it does not provide a strong test.
We therefore interpret the evidence presented in the article as a plausibility probe of our three mechanisms. Our argument, which ultimately makes a probabilistic prediction about individual behavior, does not allow for a stronger test using qualitative interview reports. We believe, however, that a plausibility probe of the mechanisms, together with a quantitative analysis of the relationship between decline and desertion, provides a solid empirical basis for our argument.

qualitative ªndings
In one PAHD report, a FARC deserter explains that, before deciding to desert, he chose to remain in the group "because once he was in the structure, he was interested to get his food permanently, and because of the comments that the leaders made about the state. In addition, they [the leaders] constantly threatened them." 99 The deserter's comments illustrate how selective incentives, ideological appeal, and coercion provide the organizational backdrop to individual behavior. 100 Below, we present our qualitative ªndings involving these three mechanisms and ªnd support for each. selective incentives. As stated, armed groups often offer selective incentives to their members to address collective action problems. Many groups provide salaries or opportunities to loot. The FARC avoided use of such incentives. 101 Only members of the FARC's clandestine urban network sometimes received monetary beneªts; rank-and-ªle combatants were not allowed even to keep money on their persons. The FARC did, however, offer incentives designed to appeal to recruits from rural communities. Several deserters report having been engaged in the local coca economy before joining the FARC, either as day laborers picking coca leaves or as eradicators of coca plants. 102 In this environment of scarce economic opportunity, the promise of basic goods such as food and clothing, as well as protection from security threats, incentivized participation in the group.
The PAHD reports show that organizational decline undermined the FARC's capacity to provide selective incentives. "Feeling safe" and being protected from threats, rather than being a defenseless outsider, was a selective incentive to join and stay in the group. 103 This relative feeling of safety was the ªrst casualty of the government's military pressure on the FARC. One deserter stated that "having been present during two bombings, when alias . . . commander of the . . . Front was killed," 104 was ultimately his reason for leaving the group. The feeling that "one could die in any moment," 105 and the risk of "being taken down by the army," 106 were commonly mentioned reasons for desertion.
In addition to producing supply shortages, military pressure undermined the FARC's capacity to satisfy members' basic needs: 107 "It obliged them to search for new and old areas of inºuence unknown for the majority of new recruits," and where "support by the civilian population was lacking." 108 As a result, even if monetary resources were available, guerrillas often did not receive military and logistical assistance from locals. During these times, deserters from camps located in the jungle complained about hunger, having sometimes received "only one meal, on others two, or they didn't eat at all." 109 Moreover, the high-proªle killings registered in the PAHD reports appear to have had a debilitating effect on the FARC's capacity to provide selective incentives. The 2010 killing of the widely admired commander Mono Jojoy in an aerial bombing was perhaps the most devastating military strike against the FARC. From then on, "every structure had to ªnd its own supplies." 110 Deserters complained that "food and equipment did not arrive any more as before." 111 One deserter stated that after Mono Jojoy's death, the "mentality has changed a lot and people want to desert." 112 Personal desires and motivations moved to the forefront. Another deserter described the situation in his unit after the killing as "everybody pushing into their own direction." 113 Another noted how he and a comrade decided to escape this situation of disorientation and sadness. 114 The 2007 killing of Negro Acacio, leader of the 16th Front, had similar consequences: "With the death of comrade Acacio, everything fell apart, as one noticed how the supplies of clothes and food started to become less and less, and if one got sick, this was a problem, as one could die and nobody would care. In the year 2008, we saw how people started demobilizing, in the beginning it was few, but then it was a lot of people. By then, many in the company and the militia networks had the idea to escape as soon as they could." 115 Thirteen members of Acacio's inner circle deserted shortly after his killing, including two midlevel commanders. 116 The killing of the leader of the 37th Front, Martín Caballero, which Juan Manuel Santos called the most important blow to the FARC in 2007, produced a similar dynamic: "After Commander Martín Caballero's death, I thought staying in the organization had no sense any more. He was a good commander. When he died, everything changed for me. That was the most important reason for me to think of desertion, [as well as] bad living conditions, food, clothing, demotivation due to living in the mountains. . . . Now I wanted to claim back the time I lost in the mountains." 117 Thus, the killing of a FARC leader and weakened selective incentives produced a desire to return home and, ultimately, the motivation to desert.
Evidence from the PAHD reports on the link between ªnancial decline and desertion is less clear, perhaps because changes in income are not as memorable as military attacks. One example, however, involves the government's coca eradication efforts. Several deserters report how coca eradication strained their ªnancial situation, as even coca farmers, the basis for a proªtable coca economy, left areas of cultivation. 118 This process had important knock-on effects on selective incentives, contributing to supply shortages. To prevent eradication and protect an important income source, the FARC engaged in extreme measures. One deserter explained that the mission of his unit "was to neutralize this activity [eradication], placing landmines in the area where they were eradicating and throwing cylinder bombs day and night." 119 In sum, evidence from the PAHD reports shows that organizational decline undermined the FARC's capacity to offer members protection and provide for basic needs. We found no reports of a strengthening of FARC's capacity to offer selective incentives in the wake of such decline. Also, we found numerous cases of deserters who explained their decision to desert as a result of the FARC's weakening selective incentives.
ideological appeal. Ideological appeal is crucial for rebel groups that claim to ªght for common goods such as justice, equality, or independence. 120 Even though the Colombian government, and Colombians more generally, often described the FARC as narcoterroristas, 121 the group adhered to a relatively consistent set of beliefs and goals grounded in Marxist ideology. 122 Members regularly listened to political teachings, 123 and participated in ideology-based practices to instill a sense of solidarity. The FARC even created a musical archive of more than 500 songs that refer to the group's foundational myth and ideals. 124 According to several deserters, ideology was the main reason why

Why Rebels Stop Fighting 195
they joined and stayed in the group. 125 One noted that the main strength of the organization was "the permanent study of all the materials and statutes, which is what maintains the combatants' conviction of the armed ªght." 126 Ideology not only kept members in line, but made them willing to endure hardships. 127 The PAHD reports show that organizational decline undermined the FARC's ideological appeal. Its goals could be fully realized only in victory or, to a lesser extent, a negotiated settlement. Except perhaps for the late 1990s, it seemed impossible that the FARC would take government power. 128 Yet, many deserters believed in this possibility when they were members. 129 The expectation of victory conditioned the binding power of the FARC's ideology, but in times of decline, the possibility of victory became increasingly unrealistic. One deserter stated as early as 2003 that "the guerrilla can launch an offensive but it is not in the condition of taking power. During the time of the demilitarized zone [from 1999 to 2002], one noted the growth of the movement, but once this was over, one sees mass desertion on a daily basis." 130 He is thus connecting military performance, ideological appeal, and desertion.
The killings of FARC leaders dampened members' expectations of victory. For example, rank-and-ªle members perceived the 2011 killing of top leader Alfonso Cano as "a strong blow to the organization, leading to pessimism and low morale," 131 and the killing of Mono Jojoy affected the "commitment to the politics and plans of the group." 132 Without the possibility of victory, the future for FARC ªghters seemed bleak. Many deserters complained that as the organization had no future, they had "no future" within the organization. 133 "Fighting for nothing, only for risking one's life," 134 ultimately convinced one member who had been with the group for eighteen years to desert.
In the absence of an achievable group goal, many members began thinking more about their personal objectives. For example, one deserter said that "the FARC structures will never rise, as their leaders say. Rather on the contrary, the public forces hit them more severely every day. That is why he thought that it was a better option to hand himself over to the authorities, so that they could help him become a good person and, like this, dedicate himself to his family." 135 The FARC's weakening military performance also affected its ideological appeal in two indirect ways. First, when the FARC had to retreat to areas where the group had little local support, mistreatment of civilians, 136 including "torturing and assassinating peasants," 137 became more common. In the eyes of some deserters, this implied a contradiction with FARC ideology. 138 One deserter stated that he was "taking conscience of how badly they were treating society and in addition, that the ideology, for which they were ªghting, was basically dead." 139 During times of military success and better ªnances, the FARC's relationship with the civilian population was more harmonious, and ideological principles were easier to put into practice. For example, when income from the coca economy was sufªcient, the FARC did not need to engage in practices such as extortion of local business owners. 140 Second, in response to continuous military pressure and cases of mass desertion, the FARC boasted about the recruitment of new members, including children. 141 One deserter explained, however, that he did not "ªnd proper backing to continue a revolutionary ªght," given the "lack of ideological conviction of recently recruited guerrillas, who will not have the convincement and capacity to withstand war situations against state forces." 142 Limited ideological training, as a result of quick personnel turnover, thus diluted the ideological appeal of the organization for long-standing members.
Another cause of the FARC's weakening ideological appeal was a Colombiaspeciªc dynamic not contemplated in our theory. Interest in personal ªnancial gain derived from drug trafªcking led to ideologically inconsistent leadership. One deserter stated that "the only thing that matters today is the money from drug trade and it goes exclusively into the pocket of the principal leaders." 143 Misbehavior related to the drug trade was an enduring challenge for the FARC. Reportedly, leaders often did not live up to the ideals of equality and

Why Rebels Stop Fighting 197
justice; instead, they enjoyed "privileges," 144 and "spend the money for what they want." 145 Thus, organizational decline reduced the FARC's ideological appeal. We ªnd no evidence of decline being followed by strengthened ideological appeal. We do note, however, several cases of deserters who pointed to weakened ideological appeal as their reason to desert.
coercion. When selective incentives and ideology are insufªcient to keep members in line, group leaders may resort to coercive measures. In the case of the FARC, a war council (an ad hoc court of justice) could sentence members suspected of "desertion with money or weapons" to death. 146 Although it is unclear how often capital punishment was used in such instances, anecdotal evidence suggests that it was common. Indeed, many deserters reported having witnessed executions of other members who had attempted to desert, including one who noted thirteen executions in her unit from 2005 to 2012, at least seven of which were for attempted desertion. 147 Media reports describe a series of executions of deserters in 2005, 148 as well as more than 400 executions of FARC members from 2008 to 2010. 149 Hence, the likelihood of severe punishment was commonly known among potential deserters. 150 According to one eventual deserter, he remained in the group "due to the fear that if he escaped and was discovered, he would be brought to a war council and that they would do harm to his family." 151 Here, being "brought to a war council" is synonymous with being executed.
In our theory section, we argue that organizational decline should loosen a group's coercive grip on its members, as the group experiences weakened control capacity and members' perceived certainty of punishment for committing an offense is reduced (although, in some cases, the result may be more severe punishment). The PAHD reports add nuance to this argument. During times of military decline, the FARC's coercive grip strengthened. "When the [military] was that "sometimes they kill innocent people or some milicianos based on rumors." 159 Criticizing the war council rulings was not an option. One deserter complained that combatants were executed for minor infractions and that the application of justice was inconsistent: "In this moment, [the commander] stood up and questioned the opinion of the interviewee, saying that he did not agree and that the decisions of the Secretariat could not be criticized." 160 This sense of defenselessness led several FARC members to desert, with some stating that they left because "they were very certainly going to be killed" by their comrades for prior wrongdoings. 161 In one case, two members deserted because one "received the order to kill" the other. 162 Thus, despite an increase in the severity of punishment, a generalized sense of "persecution" and "injustice" contributed to the decision of some FARC members to desert. 163 In sum, organizational decline inºuenced the FARC's willingness to use coercion against its members. Our qualitative ªndings are more nuanced, however, than our initial argument. Weakened coercive capacity does not appear to have reduced the certainty of punishment, but it does seem to have increased the severity of punishment. Increased punishment motivated members to desert, as it undermined the legitimacy of FARC's internal justice procedures.

Alternative Explanations for Desertion
In this section, we examine three alternative hypotheses that focus on outside opportunity rather than inside organizational dynamics to explain the decision of some members of armed groups to desert.
According to the ªrst alternative hypothesis, the possibility of living in safety after leaving an armed group, not the organizational forces inside the group, may drive a member's decision to desert. Desertion programs offer one such possibility, and several studies have found a relationship between campaigns that publicize these programs and the likelihood of desertion. 164 Since the 1990s, Colombia has offered members of nonstate armed groups the option of joining a desertion program, and it has invested resources in campaigns publicizing the program. 165 From 2002 to 2017, however, the program offer did not vary much, but publicity campaigns about this program may have led FARC members to desert.
The PAHD reports show that most deserters had heard about the program mainly through radio messaging. 166 They rarely mentioned it, however, as a factor in their decision to desert; some explicitly denied that radio messages were decisive. 167 In part, this may be because FARC leaders spread rumors about the program. As one deserter recounted, "They [the commanders] tell them that if they hand themselves over to the troops, they [the military] just extract the information and then kill them." 168 This was a common fear throughout the entire period of study. 169 Also, the distribution of leaºets and use of local radio stations to spread the word about the governent's desertion program often accompanied military operations. It is thus difªcult to isolate the effect of propaganda from military confrontation.
Our quantitative data allow us to examine the effect of major campaigns during the Christmas seasons of 2010, 2011, and 2012 to promote desertion. The size of these campaigns afforded them the greatest opportunity to inºuence combatants. Yet, the results show no link between them and desertion levels (see table A34 in the online appendix). 170 A second alternative hypothesis holds that contact with enemy forces may provide members with an opportunity to desert. Hence, the number of desertions should rise during military encounters. The PAHD reports show that several deserters did hand themselves over upon encountering government troops. However, quantitative analysis using indicators from the National Historical Memory Center for combat between FARC troops and government forces, and for any contact between government forces and the FARC, shows little support for this hypothesis (table A35 in the online appendix).
A third alternative hypothesis is that better economic opportunities outside the group may incentivize desertion. 171 Deserters interviewed by the PAHD often spoke of a desire to pursue outside economic activities, but they never Why Rebels Stop Fighting 201 mentioned a particular outside economic activity as a reason to leave. Also, they were likely unaware of speciªc economic opportunities at the time of leaving, as, feeling threatened, almost all deserters moved away from the regions where they had been active group members. 172 An earlier study on Colombia showed a link between coffee production and the opportunity cost of ªghting. 173 Our quantitative analysis using international coffee prices shows no support, however, for a link between this indicator and desertion (table A36 in the online appendix).

Conclusion
We have argued that organizational decline undermines the capacity of armed groups to promote collective action using selective incentives, ideological appeal, and coercion. As a consequence, members begin to think more about their personal desires and motivations and less about the collective, raising the likelihood of desertion. We ªnd broad support for this argument from quantitative and qualitative analysis of ªne-grained data on why members of Colombia's FARC insurgency chose to desert.
Our ªndings have important implications for policy and research. First, as the case of the FARC shows, an effective program to promote desertion may have dramatic and undesired consequences. Deserters were quickly replaced with a new generation of recruits. Also, preliminary analysis suggests that desertion is associated with the forced recruitment of minors (see online appendix 5). Colombia's successful policy to promote desertion among members of armed groups may thus have inadvertently contributed to increasing recruitment of child soldiers. Such programs should therefore be accompanied by effective policies to prevent new recruitment.
Second, this study reveals the entanglements between international dynamics and desertion. Foreign governments involved in fragile countries have been strategizing about the most effective policies to deactivate armed conºicts, through, for example, counterinsurgency; countering of violent extremism; disarmament, demobilization, and reintegration efforts; and reduction of community violence. 174 They may be less aware, however, of the direct connection between their economies and conºict. Research by economists has shown a relationship between price ºuctuations of internationally marketed commodities and conºict. 175 The present study further speciªes this linkage by showing that the decreasing proªtability of the international cocaine trade increases the number of deserters. Given this ªnding, governments in consumer countries should be aware of their impact in faraway conºicts, rather than viewing drug consumption merely as a domestic policy issue.
In Colombia, the peace agreement of 2016 ended the armed confrontation between the FARC and government forces. The FARC's collective demobilization has been accompanied by a new cycle of violence, however, with dissident groups, large criminal organizations, and the National Liberation Army guerrilla group dominating local communities. 176 These organizations have no shortage of recruits. Hence, Colombia's desertion promotion policy will likely continue for years to come.