Cultivation and competition in Colombia: Disentangling the effects of coca price changes on violence

Stoelinga, Nicole Article — Published Version Cultivation and competition in Colombia: Disentangling the effects of coca price changes on violence Journal of International Development Provided in Cooperation with: John Wiley & Sons Suggested Citation: Stoelinga, Nicole (2023) : Cultivation and competition in Colombia: Disentangling the effects of coca price changes on violence, Journal of International Development, ISSN 1099-1328, Wiley, Hoboken, NJ, Vol. 36, Iss. 2, pp. 1007-1042, https://doi.org/10.1002/jid.3850 This Version is available at: https://hdl.handle.net/10419/293948 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc/4.0/ RESEARCH ARTICLE Cultivation and competition in Colombia: Disentangling the effects of coca price changes on violence Nicole Stoelinga Max Planck Institute for Research on Collective Goods, Bonn, Germany Correspondence Nicole Stoelinga, Max Planck Institute for Research on Collective Goods, Bonn, Germany. Email: [email protected] Abstract Though evidence indicates the presence of coca increases violence, changes to the price of coca products can have both negative and positive effects on conflict. This study addresses this matter. Using novel data on local prices of coca products, production and supply chains, this study disentangles the returns to employment in the agricultural sector (cultivators) and employment in the criminal sector (guerillas and paramilitaries). I identify each agents' respective exposure to price changes and estimate the effect of such changes on violence. The results show the presence of the opportunity cost effect: an increase in income from coca results in a reduction in violence. This reduction comes with an increase in school attendance for rural households. An increase in the objective prize (OP) leads to more violence. Moreover, guerillas flock to the area that witnesses such an increase, resulting in higher levels of competition which coincides with the timing of the increase in violence. Additionally, increasing expected returns to joining an armed group can lead to higher school dropout rates among children. KEYWORDS coca, competition, conflict, income, objective prize I am grateful for the comments of the editor and anonymous referees and insights and discussions with David Levine, Thomas Crossley and participants of the EUI working group. Received: 22 September 2022 Revised: 4 August 2023 Accepted: 20 September 2023 DOI: 10.1002/jid.3850 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2023 The Author. Journal of International Development published by John Wiley & Sons Ltd. J. Int. Dev. 2024;36:1007–1042. wileyonlinelibrary.com/journal/jid 1007 1|INTRODUCTION As coca cultivation in Colombia tripled between 2013 and 2017, there are signs of escalating violence (International Crisis Group, 2021). Though there is evidence that the presence of coca increases violence (Angrist & Kugler, 2008), changes to the value (price) of coca products can have negative and positive effects on conflict: increasing wages for farmers draw labour out of the criminal sector into the agricultural sector, thus reducing violence. However, violence can also increase as higher prices make coca a more valuable good, increasing competition. This paper disentangles these contradicting forces. It develops separate measures for the returns to being active in the agricultural sector and criminal sector, and estimates the effect of changes to these returns on violence. This is done by focusing on the 2007 –2018 time period, which witnessed fundamental changes in the Colombian coca sector: the breakdown of the large cartels and the demobilization of the United Self-Defence forces of Colombia (AUC). The resulting power vacuum changed the market drastically. The presence of fewer oligopolies in the market caused the bargaining power of international buyers to increase, while Colombian agents saw their position weaken (UNODC & Europol, 2021). Simultaneously, the government eradicated record amounts of hectares of coca and dismantled coca laboratories, and the demand for cocaine in the United States and Europe grew (Mejia & Restrepo, 2016; UNODC, 2007; UNODC & Europol, 2021). The changes affected the ability of armed groups to finance themselves through coca and increased competition for control over the market. Moreover, the effects of the effects trickle down, affecting the livelihood, safety and security of those cultivating coca. The new dynamics call for re-examining the relationship between coca, violence and the roles played by various actors. This study relies on a conceptual frameworks of Arjona et al. (2015), Becker (1968), Grossman (1999), Hirshleifer (1989) and Olson (1993) to adequately address the intrinsicalities of the conflict. Agents are considered to choose between employment in the agricultural and criminal sector. For those employed in the agricultural sector (farmers, cultivators and agricultural households), coca presents an important source of income. A relative increase in income from coca would draw labour out from the criminal sector, reducing violence: the opportunity cost effect (Angrist & Kugler, 2008; Collier & Hoeffler, 2005; Dube & Vargas, 2013). Agents active in the criminal sector (armed groups) finance themselves through coca (O'Connor, 2009). The larger the share of the coca market that is under control of an armed group, the larger their source of income. For these agents, coca forms an objective prize (OP). A positive shock to the expected returns to the criminal sector entices those active in the agricultural sector to join an armed group and forms a strong incentive for existing groups to expand their territory in order to gain access to the resource (as ‘roving bandits’; Olson, 1993) leading to an increase in violence: the rapacity effect (Collier & Hoeffler, 2005; Dal B o & Dal B o, 2011; Hirshleifer, 1989). Moreover, the increase in value encourages rebels to strengthen their hold over the areas they have under control (Arjona et al., 2015; Estancona, 2021; Olson, 1993). The latter allows armed groups to engage in rebel governance and create mutually beneficial partnerships with local agents, which are affected by the economic needs of the rebels (Arjona et al., 2015). Applying the framework to the Colombian conflict, I consider the effect of an increase in the income from coca (for farmers) and in the OP (for guerillas and paramilitaries) on the intensity of violence in Colombian municipalities. 1 The econometric analysis relies on rich and detailed data on the local prices of specific coca products (coca base, paste, leaf and hydrochloride [HCI]), coca production and supply chain differences, agricultural wages, coca cultivation, eradication efforts and the dismantling of laboratories, and measures of violence. The first step of the analysis consists of creating proxies for returns from the agricultural sector (income from coca) and returns from the criminal sector (OP). Second, the effect of a 1% increase in farmers' income from coca and of the OP for armed groups on local violent intensity is estimated. Violent intensity is measured as the total number of violent events, fatalities, guerilla and paramilitary attacks per 100 000 inhabitants. The results are explained through two channels: the opportunity cost and rapacity effect. 1008 STOELINGA I examine the presence of the two effects further. Since coca is frequently cited as the reason that parents can afford to send their children to school or university (Gutiérrez Sanín, 2019; International Crisis Group, 2021), I consider the impact of an increase in income on educational outcomes. Moreover, an increase to the returns to employment in the agricultural sector can lead to higher levels of income, self-employment or hours worked. Evidence of increasing labour market and/or educational outcomes would support the assumption that the opportunity cost effect drives the decrease in violence. Moreover, the data used in this study allow me to examine the impact of changes to the OP on local competition among armed groups. Doing so provides further insight into the role of competition as a determinant of violence. Concerns with respect to endogeneity are mitigated in various ways. First, the coca prices (originally observed on municipality-month level) used to construct the proxies are averaged to region-year level. As regions contain between 20 and a couple of hundred municipalities, local dynamics affecting these average prices is unlikely. Second, only lagged values of the proxies are used in the analysis. Furthermore, the proxies are constructed in a similar fashion to a Bartik instrument, fixed effects are included and the control variables are carefully selected to capture other determinants of violence, prices, supply and demand. I conduct extensive robustness tests to verify the results. The results show a clear presence of the opportunity cost effect: An increase in income of coca products for farmers (an increase in the returns to employment in the agricultural sector relative to the returns from employment in the criminal sector) is associated with a reduction in the intensity of violence on municipality level. The results indicate that for the average coca-producing municipality, the total increase in income from coca between 2007 and 2018 resulted in 16.9% fewer total violent events, 16% less fatalities and 11.6% fewer guerilla attacks per 100 000 inhabitants. Moreover, this higher income is associated with higher school attendance rates and less competition among armed groups. Similarly, higher relative returns to employment in the criminal sector can lead to an intensification of violence: The total increase in the OP for the time period considered is associated with 25.6% more violent events, 23.18% more fatalities and 27.8% more guerilla attacks. Further analysis of these effects indicates that armed groups, especially guerillas, flocked to areas that witnessed an increase in the value the OP, and competition among armed groups increases with approximately 10%. The timing of this increase in competition coincides with the rising intensity of violence in the municipality. Additionally, there is evidence that suggests that the higher relative returns to joining an armed group might lead to lower levels of school attendance. This paper is structured as follows. Section 3discusses the coca sector in Colombia and recent changes that affected the market. A conceptual framework is presented in Section 4, followed by a discussion of the data in Section 5. The methodology and empirical strategy are laid out in section 6. Section 7presents the results, interpreting these as elasticities and net effects. The section also includes robustness tests and relates the findings of this study to the literature. Section 8concludes the study. 2|LITERATURE This paper contributes to the extensive literature on the coca sector in Colombia. However, whereas most research on this topic studies the link between the presence of coca and conflict, I focus on the economic incentives for agents involved in the coca sector and on how these affect local intensity of violence. I consider the relationship between coca prices and violence, focusing on a time period marked by significant changes with respect to agents' bargaining power, control over the supply chain and competition. This time period has—to my knowledge—not been examined. By identifying the economic incentives and disentangling the separate effects, the results of this study present a more detailed and nuanced picture than is common in the literature, which frequently does not take STOELINGA 1009 regional differences in coca production technologies and prices into account or (solely) relies on the presence or intensity of coca cultivation. 2 In doing so, this study specifically extends the work of Angrist and Kugler (2008) who exploit an upsurge in coca cultivation and prices, and show this led to small but positive economic gains for selfemployed people living in rural areas as well as to higher levels of violence. The data used in this paper allow for studying these dynamics in much more detail. In addition to labour market outcomes, I consider educational effects of changes to income from coca. Moreover, by distinguishing income from coca cultivation and the OP, I am able to show the strength and opposing signs of the effects. I show that increase in the OP leads to heightened competition and that especially guerilla groups are likely to attempt to gain control over territory. These results illustrate the working of the rapacity channel and the extent to which competition contributes to violence, as well as the behaviour of different armed groups in the Colombian conflict (Estancona, 2021). Moreover, by estimating such effects, this paper is related to the crime-rebellion nexus narrative (Cornell, 2007), rebel governance (Arjona et al., 2015) and work of Anderson and Worsnop (2019) showing that the potential financial reward of engaging in drug production and trafficking can form an incentive to use violence to maintain control over cultivation and (production) processes. Importantly, this study contributes to the literature on the more general effects of coca as a source of income for agricultural families. The positive effect of coca on educational outcomes (Gutiérrez Sanín, 2019) as well as its negative impact (Bjørkhaug, 2010; Sviatschi, 2018) have been documented. This study disentangles the effects of a change in income and the OP, providing evidence of the simultaneous presence the positive and negative impact of coca on education. This study builds on work that focuses natural resources and conflict (Collier & Hoeffler, 2005; Ross, 2003) and drug-fueled conflict (Cornell, 2005,2007). The relationship between changes in the prices of agricultural goods and violence in Colombia has been documented by Dube and Vargas (2013) who consider the effects of changes in the coffee and oil price, and Rettberg (2010) who studied the breakdown of the International Coffee Agreement and the violence in Colombian coffee regions that followed. Though coca is an agricultural good as studied in those papers, it is an illicit good which cultivation presents a different dynamic and challenge and comes with severe risks for cultivators (UNODC, 2017). As this study is closely related to the mentioned works, the results of this study are compared with those of Dube and Vargas (2013) in Section 7. Finally, by using new data and focusing on a different time period, I contribute to and build upon the large literature on the coca-fueled conflict in Colombia. Other work in this field focuses on violence and the armed groups involved in coca trafficking, such as on coca eradication efforts (Fisher & Meitus, 2017; Moreno-Sanchez et al., 2003), modelling the war on drugs as a resource problem (Grossman & Mejía, 2008; Mejia & Restrepo, 2016), explaining different levels and types of violence (Holmes et al., 2006), armed groups involved in the coca sector (Bottía Noguera, 2003; Estancona, 2021; O'Connor, 2009; Suarez, 2000; Thoumi, 2002) and expansion of coca cultivation (Mejia & Restrepo, 2013). 3|INSTITUTIONAL BACKGROUND To understand the coca market in Colombia, it is important to understand the basic steps in the production process of cocaine. 3 The part of process that takes place in Colombia can be roughly divided into four stages (UNODC, 2010). In the initial three stages, coca leaf is cultivated, processed into coca base and eventually turned into coca paste. These stages are carried out by agricultural households, farmers or cultivators. In the fourth stage, coca paste is processed into coca HCI. This stage is carried out by armed groups that possess the capital to obtain the scarce ingredients (such as acid and acetone) and have the technical expertise required for production (O'Connor, 2009; UNODC, 2018,2021). Coca HCI is the product that is trafficked to its final destination by international wholesale buyers, where it is eventually processed into cocaine. 1010 STOELINGA 3.1 |Armed groups and cultivators This study focuses on the dynamics within the Colombian coca market and considers cultivators and armed groups as agents in this market. There are various armed groups involved in the Colombian coca sector, but these are generally divided into two categories: left-wing guerillas and right-wing paramilitary groups (Estancona, 2021; Gutiérrez Sanín, 2008). The groups have different origins and objectives and differ with respect to their involvement in the coca sector, relationships with cultivators, responses to changes to the value of coca and use of violence. The largest guerilla group active in Colombia to date was the FARC (Fuerzas Armadas Revolucionarias de Colombia). 4 FARC funded itself through kidnapping, extortion, illegal mining practices and involvement in the coca sector, with the majority of its drug-related income coming from a ‘protection tax’(Bruce-Jones & Smith, 2021). 5 The group took on the role of regulators of the coca paste market, arranging and controlling transactions between cultivators and cartels (Vargas, 2011). During the peace talks between the Colombian government and FARC that started in 2012, many dissident factions were formed (Flores & Vargas, 2018). These factions, together with other guerilla groups—such as the ELN and EPL—are the left-wing organisations currently active in Colombia. The Colombian government opposed guerilla groups through military means and also encouraged the formation and involvement of paramilitaries (Estancona, 2021; Gutiérrez Sanín, 2008). The paramilitaries were formed by landowners to protect their property from appropriation by guerillas and worked with cartels to protect coca trafficking routes and thereby prohibit left-wing groups from acquiring additional income (Estancona, 2021). AUC emerged as one of the largest paramilitary organisations in Colombia and relied on coca as a way of financing itself. 6 AUC disbanded in 2006, after which former members and other paramilitaries formed new, smaller organisations. These groups are often referred to as Bandas Criminales Emergentes or BACRIM. The BACRIM are generally considered to be continuations of paramilitary groups (Maher & Thomson, 2018; Saab & Taylor, 2009). The main difference between guerillas and paramilitary organisations is that guerillas are an oppositional group, whereas paramilitaries are pro-governmental organisations with connections to the state (Gutiérrez Sanín, 2008). 7 This structural difference has a significant impact on each groups' use of violence and the relationship it has with those who cultivate coca. One of the paramilitaries' main objectives is to protect cartels' interests and prevent cultivators from producing coca for left-wing organisations (Estancona, 2021). This is reflected in their use of violence: Paramilitaries employ selective violence aimed at controlling civilians and discouraging them from selling coca to guerillas. On the other hand, guerilla groups are often engaged in battles with the government and have significantly higher levels of violence. Guerillas are known to target civilians to intimidate and influence the population in order to maintain access to the coca grown in areas (Estancona, 2021; Gutiérrez Sanín, 2008). This emphasises the risk and uncertainty associated with cultivating coca: municipalities where coca is grown experience significantly more violence (Angrist & Kugler, 2008), cultivating coca brings farmers in direct contact with violent armed groups that often force cultivators to grow coca instead of other crops (Estancona, 2021; UNODC, 2021) and governments' eradication efforts can result in instant loss of the entire harvest. 8 However, coca forms a reliable and highly profitable source of income. It is an easy, and often, the most lucrative crop to grow since the market for coca products is almost guaranteed (Ibáñez, 2010; Mansfield, 1999; Thoumi, 2002). 9 As a consequence, about 125.000 to 170.000 households are involved in the production of coca leaves, base and paste (UNODC, 2019). 3.2 |The changing coca market The coca market in Colombia has changed significantly over the past years. Two main trends can be distinguished. First, the large cartels and armed groups are being replaced by smaller organisations (the aforementioned BACRIM) STOELINGA 1011 vying for control over the coca market (UNODC & Europol, 2021). The second trend is that the main destination markets for Colombian cocaine—the United States and Europe—are growing, increasing the potential profits to be made from producing, trafficking and selling coca for some of the agents involved (UNODC & Europol, 2021). These two shifts have significantly affected both armed groups and cultivators. The first change was set on by the gradual breakdown of the large cartels, after which the paramilitaries and guerillas took over drug trafficking (Flores & Vargas, 2018). 10 This was followed by the disbandment of the AUC in 2006 and the peace negotiations and eventual demobilisation of FARC in 2017. Especially the disbandment of AUC had a large impact on the Colombian coca market. Prior to its disbandment, AUC was involved in every part of the production process within Colombia and trafficked cocaine to international markets. According to some estimates, the AUC controlled about 40% of the cocaine exports from Colombia (Saab & Taylor, 2009). Along with the gradual disappearance of the cartels and the AUC in Colombia, the monopoly positions of these international trafficking organisations (such as the Italian 'Ndrangheta) in their respective markets deteriorated (UNODC & Europol, 2021). This made it possible for other (new) international organisations to enter the market. Brokers in Latin America started playing an increasingly important role in supplying these international traffickers with bulk quantities of HCI, which they obtained from Colombian groups. The increasingly fragmented market and presence of brokers caused the bargaining power of guerillas and paramilitaries involved in the coca sector to deteriorate, turning the latter into price takers (Millán-Quijano, 2020; UNODC & Europol, 2021). Especially the BACRIM faced a competitive environment in which they had less bargaining power: In many areas where the AUC was initially present, “two or more emerging criminal bands with no obvious relationship and often exhibiting an environment of mutual tensions and retaliations would occupy the vacated space”(Saab & Taylor, 2009, p. 463). On average, guerilla groups were less involved in all the aspects of coca production. Their focus was primarily on the taxation of coca products and regulation of the sale of coca products to cartels and criminal organisations in their territory (Vargas, 2011) and therefore already had a working relationship with various criminal organisations and brokers. This implies that guerilla groups and their activities were somewhat less affected by the changing landscape then the paramilitaries. The loss of bargaining power lowered profits for these groups, causing them to compete over areas where coca is produced to secure access to the resource. Aside from access to coca, territorial control is of fundamental importance as it also guarantees trafficking routes, a social base—in terms of a local labour force and farmers to buy products from—and income from taxation (Díaz & Sánchez, 2004). The recent changes in the coca market exacerbated the uncertainty faced by those cultivating coca. One of the ways the recent trends have affected cultivators is through the prices they receive for their coca products. Coca prices are affected by other factors than (the usual) supply and demand dynamics, and are often set by buyers in an area (International Crisis Group, 2021; UNODC, 2017). 11 As the weaker party, cultivators have little to no bargaining power and have to accept a price offer—Trying to sell coca products to another group is dangerous and can lead to retribution. As armed groups find it increasingly difficult to finance themselves through coca due to the loss of bargaining power, they raise taxes or extract more resources from farmers and cultivators, for example, by decreasing the prices paid for products. Additionally, groups might use violent means to pressure farmers to increase coca production or prohibit them from cultivating other crops (Estancona, 2021; International Crisis Group, 2021; UNODC, 2018). Moreover, the larger number of armed groups vying for territorial control exposes cultivators to changing buyers and different prices for coca products, causing uncertainty with respect to the income from coca for cultivators. The uncertainty and risk are illustrated by a situation recorded by the UNODC (2017): A drop in market activity due to a sudden absence of buyers in a local market lead to an oversupply of leaf, base and paste and a negative income shock for cultivators. The markets started working again only when new buyers appeared. 1012 STOELINGA 4|CONCEPTUAL FRAMEWORK Common frameworks in the conflict literature consider price shifts to affect violence through the opportunity cost and rapacity channel. This is based on the notion of civilians joining the rebels when the opportunity arises and rebels as ‘roving bandits’. On the other hand, rebels can act as long-term, stationary agents (or ‘stationary bandits’) who make choices to maintain access to resources present in the territory under their control in order to survive as a group (Estancona, 2021; Olson, 1993; Sánchez De La Sierra, 2020). Combining the above, this section describes a conceptual framework that addresses the intrinsicalities of the coca market, economic incentives of agents, and the connection to violence. 4.1 |Roving and stationary rebels, guerillas and paramilitaries: The opportunity cost and rapacity effect Following the work of Becker (1968) as described in Dube and Vargas (2013) and Wright (2016), agents are considered to choose employment in the agricultural or criminal sector, depending on where the expected returns are the highest. Employment in the agricultural sector focuses on the production of a good or resource and the criminal sector on extracting value through (illegal) taxation, appropriation, rent-seeking and theft (Grossman, 1999). In this setting, goods and resources can lead to violence and conflict through two channels: the opportunity cost or rapacity channel. First, a relative increase in expected returns to employment in the agricultural sector would draw labour from the criminal sector into the agricultural sector, reducing violence. This is the opportunity cost effect (Angrist & Kugler, 2008; Collier & Hoeffler, 2005; Dube & Vargas, 2013). Vice versa, a positive shock to the expected returns to the criminal sector would draw labour from the agricultural sector into the criminal sector, leading to an increase in violence. This is the rapacity effect (Collier & Hoeffler, 2005; Hirshleifer, 1989). This framework is applied to the coca market in Colombia, as to understand the relationship between returns to cultivating coca and producing leaves, base and paste (the agricultural sector), the returns to being a member of a guerilla or paramilitary group that finances itself through trafficking and producing coca HCI (the criminal sector), and violence. 12 The expected returns to being employed in the agricultural sector are presented by the income from cultivating coca and selling coca products (leaf, base and paste). An increase in the expected returns to cultivating coca has the potential to draw labour out of the criminal sector, that is, members of armed groups might forego criminal activities to work in the agricultural sector, reducing violence. The presence of this opportunity cost effect will be estimated empirically by using a proxy for returns to being active in the agricultural sector. Moreover, as there is evidence that positive income shocks can affect the number of hours worked and employment rates (Angrist & Kugler, 2008) or educational outcomes such as education level and school attendance (Gutiérrez Sanín, 2019; International Crisis Group, 2021; Sviatschi, 2018), the relationship between shifts in coca income and these outcomes will be investigated. The returns to employment in the criminal sector are captured by the value that can be obtained through appropriation and rent-seeking and will be referred to as the OP. An increase in the relative value of the OP with respect to the income from coca cultivation can cause the rapacity effect. One way in which this can occur is through a negative income shock for those employed in the agricultural sector. Facing such a shock, these agents can choose to increase coca production, cultivate other crops or—especially day-labourers, being more flexible and not tied down by farmownership—potentially find employment in the criminal sector. 13,14,15 The flow of labour from the agricultural sector into the criminal sector could lead to an increase in violence. Additionally, an increase in the OP can also motivate those already employed in the criminal sector to act as roving bandits (Olson, 1993). These rebels may attempt to expand their territory in order to gain access to the OP. Heightened competition might increase violence: Violence is linked to armed groups vying for territory and access to resources (UNODC & Europol, 2021) as well as instilling fear in the local population after having taken over control (Díaz & Sánchez, 2004; International Crisis Group, 2021). STOELINGA 1013 Stationary rebels secure long-term access to coca by establishing a permanent presence in areas and creating partnerships with local coca cultivators (Estancona, 2021). They engage in rebel governance (Arjona et al., 2015; Kalyvas, 2006; Mampilly, 2012), providing security and stability to cultivators, who supply the rebels with coca in return. As stationary rebels, guerillas and paramilitaries are concerned with maintaining access to the resources under their control in order to extract rents and aim to prevent the opposing group from obtaining access to that resource. When the value of coca decreases, a group might no longer be able to pay predetermined prices to the cultivators in its area, forcing them to resort to violence: ‘rebels may employ selective violence—lethal or nonlethal—against suppliers during temporal windows in which more costly tactics to prompt continued cooperation become too expensive’(Estancona, 2021, p. 1113). 16 An increase in the OP—aside from making it more attractive to act as a roving bandit—can cause groups to use violence to discourage cultivators from selling coca to a competing group that might promise higher prices. According to Estancona (2021), guerilla groups are most likely to use violence to ensure continued production of the resource in response to a price decrease; paramilitaries use violence to discourage cultivators from selling coca to guerillas, when the latter try to encroach on an area under paramilitary control in response to a price increase. 17 Empirically, the rapacity effect will be examined by estimating the effect of a change in a proxy that captures the OP on various measures of violence. Due to the differences between guerillas and paramilitaries, the measures include the attacks carried out by either group separately. Moreover, the presence of paramilitaries or guerillas will be accounted for in the analysis as it can affect the use of violence by either group. Finally, aside from examining competition as the number of different groups active in an area, I will consider the likelihood of a guerilla or paramilitary group entering an area due to an increase in the OP. 5|DATA AND SAMPLE Data on coca cultivation and eradication are obtained from the United Nations Office on Drugs and Crime (UNODC) and SIMCI. 18 Data on coca derivative prices and production processes come from the productivity studies from the UNODC. These productivity studies focus on three general areas: coca cultivation and general productivity trends, socio-economic characteristics of agents active in coca cultivation and the persistence of coca cultivation. The data are collected through interviews conducted with cultivators, and harvest trials are carried out to estimate the potential coca yield. The studies provide the UNODC with insight into local price fluctuations, long-term trends of production processes and technological developments in the coca sector. There are a few important things to note with respect to the prices of coca products. First, production studies and reports that focus on the prices of coca products in Colombia, such as the UNODC (2013), mention that these prices “retain no logical relation to variations in supply and demand”and that “buyers can pressure the production exchange, presenting a controlled price scenario […]. This situation forces producers to adopt the enforced condition”(p. 56). These observations support the fact that cultivators are price takers with limited to no bargaining power. Second, prices of the coca products considered in this study vary across municipalities. They depend on the ease of reaching markets, presence of buyers and the rules set by them, quality of the product and municipal or village-level supply and demand for coca products (UNODC, 2013). The UNODC collects municipality-month level price data and reports the average monthly prices for each region. Every region contains a varying number of municipalities. In the final sample, the smallest region contains 23 and the largest region a few hundred municipalities (the average number of municipalities per region is 159). This study uses annual averages of the regional prices. The risk of prices being endogenous to municipality-level dynamics is thereby mitigated (this is discussed further in section 6.2). All prices are adjusted for inflation. Figure 1shows the annual changes in the prices for the four different coca products considered in this study for the various Colombian regions. Data on violence come from the Global Terrorism Database (GTD) and the Centro Nacional de Memoria Hist'orica (CNMH). Both the GTD and CNMH are event-based data sets. The GTD (LaFree & Dugan, 2007) spans 1014 STOELINGA The potential presence of endogeneity in this model is addressed in various ways. First, lagged values of both proxies are used as controls: Income i,d,t 1 is the lagged value of the proxy for income from coca for cultivators and OP i,d,t 1 the lagged value of the OP for armed groups. Second, the original municipality-month level data on coca product prices and production technology underlying the proxies have been averaged to region-year level as discussed in Section 5. By using these averaged prices, it is unlikely that the local municipal circumstances affect prices or technology the data on the level used in this study. 31 Third, the correlation between the two proxies is accounted for in the analysis by treating both proxies as independent variables and simultaneously estimating their effects on violence. This allows for estimating the effect of one proxy on violence while controlling for the effect that the variance of the other proxy has on the outcome variable, that is, examining the effect of shifts in income for cultivators on violence given the effects of shifts in the OP on violence (and vice versa). This addresses, for example, scenarios in which lower values of the OP would lead to armed groups suppressing wages for cultivators. Combined, these measures—as well as additional measures discussed below—ensure that endogeneity concerns are mitigated. Many drivers of coca-related violence are also determinants of prices of coca products. Z i,d,t captures control variables on municipal and departmental level that are measures of these drivers. First, local government intervention in the coca sector, affecting both prices and violence, is addressed by including time-varying, municipality level controls for the number of hectares (in thousands) of coca fields that are eradicated (both manually and through aerial eradication), the number of coca processing laboratories that were dismantled or destroyed, as well as growth rates capturing changes in the levels of the kilos (in thousands) of coca leaf, base, paste and HCI that have been confiscated. Controlling for such types of government interference in the coca sector also captures whether government forces are active in a municipality, and potential violence or conflict stemming from this presence. Moreover, though the analysis will focus on the first lag of each proxy, additional lags of the proxies are included as controls. The final number of lags (four in total) included in the model is chosen based on the Bayesian Information Criterion (BIC), in order to select the model that presented the best fit. The leading principle was to include sufficient lags to address endogeneity and autocorrelation concerns while ensuring to not over-fit the model. In addition, changes to the presence of guerilla or paramilitary groups—especially when a guerilla group enters paramilitary territory or vice versa, as discussed in section 4—can lead to heightened levels of violence, lower prices for farmers' products or an increase in coca production through competition. Following Estancona (2021), this is controlled for by including time-varying variables that capture the presence of paramilitary or guerilla groups in the municipality at t1, when examining the effect of changes in income or the OP on paramilitary or guerilla attacks. 32 Furthermore, Z i,d,t includes linear time trends for municipalities that produce coca, gold and coffee. The latter two are included to account for outside options of coca cultivators to change to another sector, as well as potential alternative sources of financing for armed groups (both coffee production and [illicit alluvial] gold mining are associated with conflict (Dube & Vargas, 2013; International Crisis Group, 2021). Finally, rain and temperature are included as important measures of potential production levels as well as determinants of conflict. Year (λ t ) and municipality (ϕ i ) fixed effects are included to capture any additional (un)-observed time-invariant variables. ϵ i,d,t is the error term. The coefficients of interest are δand ψ.δcaptures the differential effect of the income from coca for farmers, cultivators and agricultural households on violence in municipalities that cultivate relatively more coca. Similarly, ψcaptures the differential effect of the OP for guerillas and paramilitaries violence in municipalities that cultivate relatively more coca. By specifying the baseline model as above, I allow for continuous variation in both independent variables of interest (Income i,d,t 1 and OP i,d,t 1 ). Hence, the identification strategy relies on variation between the cocaproducing municipalities stemming from the proxies. The inclusion of noncoca-producing municipalities in the sample contributes to the estimation of the year fixed effects and the control variables, capturing noncoca-related factors that potentially affect violence. However, the estimation can be conducted without the inclusion of these noncocaproducing municipalities. To address concerns that the inclusion of the noncoca-producing municipalities drives the STOELINGA 1021 results, these municipalities are excluded from the sample in a robustness analysis; see Section 7.3. Additional tests are carried out to verify the results. First, the lagged value of each dependent variable (violent events, fatalities, paramilitary or guerilla attacks) is added as a control, following an Arellano and Bond (1991) framework. Secondly, department-level linear trends, to capture potential time-varying social or political trends, are included. The latter two are also combined in one analysis. The results are robust to these specifications. 6.3 |Opportunity cost and rapacity channel If the results show that an increase in income from coca leads to lower levels of violence, the conceptual framework posits that this might be due to the opportunity cost effect. The opportunity cost effect suggests that an increase in potential returns to cultivating coca leads to an increase in employment in the agricultural sector. This can imply, first, an increase in the number of hours worked by those active in the sector in order to reap the higher returns (Angrist & Kugler, 2008). I examine this by regressing the income proxy on the number of hours worked per week. Second, the income from coca cultivation is frequently cited as the reason that parents can afford to send their children to school (International Crisis Group, 2021), with one national survey finding that more than half of the families involved in coca cultivation invested their profits mainly in education (Gutiérrez Sanín, 2019). However, Sviatschi (2018) shows that an increase in coca prices leads to higher levels of child labour in areas suitable for coca cultivation, as well as children (especially of secondary school age) dropping out of school to help their family. This implies that the effect of increased income from coca cultivation on education can go both ways. Moreover, there is anecdotal evidence that an increase in the potential returns to joining an armed group has been associated with children dropping out of school (Bjørkhaug, 2010). 33 This suggests that an increase in the OP does potentially not only draw labour out of the agricultural sector, but can also entice children to drop out of school. These potential effects are examined by regressing the income and OP proxies on school attendance rates and average educational attainment. Importantly, since the feasibility of joining a paramilitary and guerilla group differs starkly as discussed earlier, the analysis includes controls capturing the presence of either or both groups. Third, one of the main predictions of the conceptual framework is that an increase in the OP causes an increase in violence through higher levels of competition among armed groups. The baseline analysis accounts for the presence of guerilla or paramilitary group in the municipality as this by itself can strongly affect the levels of violence. However, additionally, the extent to which a shift in the OP affects changes in the number of individual groups active in the area is examined. 34 This measure is referred to as competition and differs from the measure for the presence of a guerilla and/or paramilitary group in the municipality that is included as a control in Z i,d,t, . I examine these outcomes and present a more detailed picture of the effects that changes in income and the OP have and through what channels the opportunity cost and rapacity effect work. Yi,d,t¼λtþϕiþβIncomei,d,t1þψOPi,d,t1þθZi,d,tþϵi,d,tð5Þ where Y i,d,t is the municipal average of the number of hours worked per week, school attendance rates or the highest education achieved for those living in rural areas and that indicate that they, or their household, work(s) in agriculture as a day labourer or is/are self-employed. When considering the rapacity channel, Y i,d,t captures the level of competition in the municipality and the presence (entry) of a paramilitary or guerilla group. All other variables are as before. However, in the case that labour market or educational outcomes are considered, gender and age are included as additional controls. 1022 STOELINGA 7|RESULTS 7.1 |Baseline model Table 1shows the results of the baseline model 4. The coefficients capture the effect on violence in response to a 1 percentage point increase in income and OP for the average coca-producing municipality with 396 ha of coca. The results show a negative, significant differential effect from an increase in the coca income for cultivators on violent intensity. This suggests the presence of the opportunity cost effect. Additionally, there is a positive, significant differential effect from an increase in the OP on violent intensity, which suggests that the rapacity effect is present. Interpreting the results as elasticities and the coefficients as deviations from the mean, the results imply that a 1% increase in income from cultivating coca results in 1.9% fewer total violent events, 1.8% fewer fatalities and 1.3% fewer guerilla attacks per 100 000 inhabitants for the average coca-producing municipality. Similarly, a 1% increase in the OP leads to 2.1% more events, 1.9% more fatalities and 1.2% more guerilla attacks (Table A3). Across the time period considered, the average coca-producing municipality witnessed an annual increase in the proxy for coca income of 2.4% and an average annual increase in the OP of 3.81%. This implies that the average increase in income from coca for cultivators—or returns to employment in the agricultural sector—is associated with 4.6% fewer violent events, 4.3% fewer fatalities and 3.1% fewer guerilla attacks. Similarly, the average annual OP increase is associated with 7.6% more violent events, 7.2% more fatalities and 4.6% more guerilla attacks. 35 The results for the different types of attacks are shown in Table A10. Interestingly, an increase in the income from coca seems to lead to an increase in the number of assassinations, while an increase in the OP is associated with a decrease in assassinations. Paramilitaries are known to carry out more assasinations than guerillas: This result might be related to the argument of Estancona (2021), who notes that increases in farmers' coca prices can result in paramilitaries using coercive (violent) means to prevent cultivators from growing coca for guerillas. The number of bombings per 100 000 inhabitants is positively affected by an increase in the OP and negatively by the increase in income. TABLE 1 Income, OP and violence. Total events Fatalities Guerilla attacks Paramilitary attacks Income t1 1.303*** 0.869** 1.722** 0.029 (0.368) (0.405) (0.681) (0.023) OP t1 1.345*** 0.915** 1.518** 0.023 (0.416) (0.356) (0.594) (0.022) Controls Yes Yes Yes Yes Year FE Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes N8857 8857 8857 8857 Cluster 32 32 32 32 Note: The table shows the results for estimation 4. Dependent variables capture violent intensity on municipal level, measured as number of events, fatalities, guerilla or paramilitary attacks per 100 000 inhabitants. The full results, including controls, are shown in Table A4. The results for the dependent variables measured in levels are presented in Table A11. Standard errors, clustered by department, in parentheses. Abbreviation: OP, objective prize. *p<0.1, **p<0.05, and ***p<0.01. STOELINGA 1023 7.2 |Opportunity cost and rapacity channel Having found evidence that supports the presence of the opportunity cost and rapacity effect, the effects of changes in income and the OP on additional outcomes are considered. Labour market and educational outcomes, as well as competition among armed groups on a local level, are examined following Equation (5). The results are presented in Table 2. Columns 1–3 show the results for labour market and educational outcomes. First, considering the effect of an increase in income from cultivating coca, there do not seem to be (significant) effects on either the number of hours worked or level of education achieved. However, an increase in income from coca cultivation does seem to be associated with higher municipal school attendance rates. Considering that the average coca-producing municipality witnessed an annual increase in the proxy for coca income of 2.4%, this corresponds to a 3.1% increase in school attendance the year following the increase. In response to an increase in the OP, school attendance decreases: The average annual 3.81% increase in the returns to the criminal sector leads to a 5.3% decrease in municipal school attendance the following year. These findings are in line with literature that cites that income from coca in rural areas is frequently invested in education of children. However, the results also support the evidence presented by Bjørkhaug (2010): When the potential returns to joining a militia are relatively high, it can be attractive—especially for young males—to drop out of school and join an armed group. By having separated the effects from an increase in income and an increase in the OP, the results present the magnitudes of each effect individually and confirm both the positive and negative impacts of coca on education. The conceptual model explained the relationship between competition and violence through the rapacity channel. Agents, active in the criminal sector, compete for control over the market as ‘roving bandits’(Olson, 1993). The role of competition with respect to a change in income or OP is examined by estimating Equation (5), and the result is presented in column 4 of Table 2. The level of competition in a municipality seems to decrease in response to an increase in the returns from employment in the agricultural sector: controlling for the presence of guerilla and paramilitary groups, a cocaproducing municipality witnesses a decrease in the number of armed groups active in an area. However, this effect is quite small when considered as a deviation from the mean—the 2.4% average annual increase in income is associated with a 1.8% decrease in local competition. The average OP increase of 3.81% is associated with an increase in local competition of approximately 3.3%. However, there is evidence that an increase in the OP leads to guerilla groups TABLE 2 Opportunity cost and rapacity effect. Hours worked Highest level of education Attending school Local competition Guerilla presence Paramilitary presence Income t  1 2.200 0.126 0.013*0.160** 0.159*** 0.014 (2.944) (0.185) (0.006) (0.073) (0.050) (0.012) OP t 1 0.463 0.107 0.014*0.183** 0.185*** 0.011 (2.835) (0.196) (0.007) (0.068) (0.040) (0.011) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes Yes Yes N3437 3437 3437 8857 8857 8857 Clusters 24 24 24 32 32 32 Note: The table contains the results for estimation of Equation (5). The full results, including controls, are shown in Table A12. Standard errors, clustered by department, in parentheses. Abbreviation: OP, objective prize. *p<0.1, **p<0.05, and ***p<0.01. 1024 STOELINGA expanding their territory—most likely driving the increase in competition. Interestingly, the coefficient for an increase in the income of coca for cultivators is negative. This can be due to reverse causality, with guerillas being less likely to enter or be present in an area where income from cultivation is higher on average. These higher prices might make it less feasible for guerillas to finance themselves through coca. 7.3 |Robustness and comparison of magnitudes The results are tested for robustness. First, all noncoca-producing municipalities are excluded from the sample. The results of this estimation are shown in Table A5. The results are very comparable with those of the baseline estimation. Second, values of each dependent variable are included, as an additional measure for addressing endogeneity in dynamic panel data (Arellano & Bond, 1991). The results, shown in Table A6, are robust to this specification. Third, though the baseline analysis includes time-variant linear trends for coca-producing municipalities, a time-variant department-level linear trend is included to control for potential departmental level trends (see Table A7). Finally, both the lagged values and department-level trends are included simultaneously (Table A8). I compare the results from this study to similar estimates and results in the literature. Most importantly, Dube and Vargas (2013) report similar effects on violence when considering a change in prices of a labour-intensive agricultural good (coffee) in Colombia. The sizes of the estimates of the two studies are similar (Table A9). For example, Dube and Vargas (2013) find that a 1% increase in coffee price—for the average coffee producing municipality— results in 0.611 fewer guerilla attacks, while I find that a 1% increase in income from cultivating coca results in 0.77 fewer attacks. 36 Moreover, the evidence shown in this study supports various findings with respect to the activity and behaviour of guerilla groups. For example, this study supports the results of Estancona (2021), who presents evidence of an increase in the number of guerilla attacks in response to a positive price shock of coca. Similarly, the results suggest that guerilla presence has a positive effect on violent events (shown in columns 1 and 3–6 of Table A10), which is in line with earlier findings of Angrist and Kugler (2008). However, whereas Estancona (2021) uses a variable capturing shocks in the street price of cocaine in the United States and Angrist and Kugler (2008) rely on the size of the municipality-level coca production, the measures used in this study are more fine-grained. I use local-level proxies that incorporate the local intensity of coca production to account for the prevalence of the crop and technology and—most importantly—distinguish between the prices that cultivators and armed groups are exposed to, thereby providing more nuanced results. Using these measures, I explicitly estimate the impact of an increase in the OP on the likelihood of the presence of a guerilla group. By doing so, I support the assumption of Estancona (2021) that guerilla groups expand their territory in response to a positive price shock. Finally, Angrist and Kugler (2008) note that an increase in the coca price can lead to increases in (agricultural) wages. This study presents quantitative evidence to support the link between the price of coca and agricultural wages using the novel price data which was not available for the time period studied by Angrist and Kugler (2008). 8|CONCLUDING REMARKS Coca cultivation in Colombia tripled between 2013 and 2017, reaching an all-time high and violence escalated (International Crisis Group, 2021). It is a difficult matter: On the one hand, coca is a very lucrative crop and stable source of income in a country with a large rural agricultural sector, where approximately 36% of people live in poverty (Worldbank, 2018). Simultaneously, coca is an important source of financing and a good worth competing for among armed groups, who are involved in a conflict considered responsible for at least 260 000 deaths (Reuters, 2020). The complexity emphasises the need for a deep understanding of the economic incentives for agents involved in the sector and how these incentives affect their choices and behaviour. This is illustrated by the STOELINGA 1025 extent to which many policies aimed at decreasing cultivators' reliance on coca are based on economic incentives, such as subsidy or crop substitution programmes. This study contributes to the understanding of the conflict by identifying incentives, disentangling the separate effects, examining the impact on labour market and educational outcomes and (violent) competition among and behaviour of the various armed groups involved in the conflict. This paper focuses on 2007–2018, a time period during which the Colombian coca sector changed fundamentally. First, the large cartels were replaced by an increasing number of smaller criminal groups vying for control over the coca market. The second trend is that the main destination markets for Colombian cocaine—the United States and Europe—are growing, increasing the potential profits to be made from producing, trafficking and selling coca for some of the agents involved (UNODC & Europol, 2021). These two trends have significantly affected armed groups and cultivators. The first are confronted with less bargaining power and higher levels of competition; for the latter, these changes have significantly impacted their livelihood, safety and security. The changing market structure and shifting roles of agents in the Colombian coca market emphasise the need for understanding agents' behaviour and how they respond to (economic) incentives. This study aims to shed light on how these recent changes have impacted the link between coca and violence. It examines the effect of changes in the returns to being active in the agricultural sector and in the criminal sector on violent intensity in Colombian municipalities. This is done by developing a proxy for the income from coca for those employed in the agricultural sector and a proxy for the OP, presenting the returns to being employed in the criminal sector. By constructing separate measures, the paper identifies and disentangles the (contradicting) effects of changes to the value of coca on violent intensity. The results indicate that, with respect to cultivators and farmers active in the agricultural sector, increases in the income from coca result in lower levels of violence (the opportunity cost effect). Moreover, there is evidence that increasing income can lead to higher school attendance rates, supporting previous findings that a share of the proceeds of coca is invested in education. There is suggestive evidence that a relatively higher—compared with the OP—income from cultivation might entice members of armed groups to seek employment in the agricultural sector, as well as reduce the pressure on groups to compete for territory and financial gain. Furthermore, the results indicate the presence of the rapacity effect: An increase in the OP is associated with an increase in the intensity of violence. This increase in the potential return from joining an armed group might entice children or young adults to drop out of school, lowering school attendance rates. Finally, armed groups—especially guerillas—tend to gravitate towards areas where the expected returns are higher: The increasing value of the OP leads to higher levels of local competition. ACKNOWLEDGEMENTS Open Access funding enabled and organized by Projekt DEAL. DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from DANE and UNODC. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from https://www.dane. gov.co/ and https://www.unodc.org/, with the permission of DANE and UNODC. ENDNOTES 1 Administratively, Colombia is made up of 31 departments and 1 capital district. Each department consists of various municipalities: There are 1122 municipalities in total. The UNODC (2017) distinguishes seven different regions in the coca cultivation surveys. These are not official administrative units but are ‘natural’or geographical regions. These are Amazon'ıa, Central, Meta-Guaviare, Orinoqu'ıa, Pacifico, Putumayo-Caqueta´ and Sierra Nevada. 2 Though the latter has been shown to have an effect on violence by Angrist and Kugler (2008), the presence of coca is not necessarily a good indication of the local value of coca. As noted by the UNODC (2010), coca prices do not follow the ‘regular’rules of supply and demand as they are affected by many other factors. Moreover, there is a considerable amount of variety in the extent to which farmers process the coca leaves they harvest and thereby the value of the 1026 STOELINGA product and their income. Finally, it is common that farmers suffer a drop in income due to government interference (Thoumi, 2002) or are pressured by local militias to sell their products for a lower value (UNODC, 2018) while the price of coca chlorhydate (HCI)—armed groups' OP—remains constant. 3 For an excellent in-depth discussion on the history, violence related to and production of coca in Colombia, see Díaz and Sánchez (2004). 4 Other groups are the Ejército de Liberaci on Nacional (ELN), Ejército Popular de Liberaci on (EPL), and M-19. 5 Coca has been an important source of funding for decades. According to Vargas (1999), almost half of the income of guerilla groups in the period 1991–1996 came from drug trafficking. About one fourth came from extortion and robbery, one fifth from kidnapping and a small fraction originated from municipal funds. However, illicit alluvial gold mining has also proved to be a source of financing for these groups since 2007 (Rettberg & Ortiz-Riomalo, 2016) and is linked to coca production (UNODC, 2013). When the international gold price collapsed in 2015, more groups were pushed into the cocaine market to make up for lost profits, exacerbating the conflict further (International Crisis Group, 2021). 6 AUC obtained about 70% of its funding from drug activities and 30% from voluntary contributions from landowners (Saab & Taylor, 2009). 7 Even though the BACRIM receive significantly less support than the AUC, there is evidence that there is some cooperation between local political, judicial and military institutions (Hristov, 2009). 8 The chemicals used for eradication can affect coca as well as other crops and render land unusable for years. Additionally, the chemicals used present significant health risks (Veilette & Arvelo-Velez, 2003). 9 According to Ibáñez (2010) coca crops are three to five times more lucrative than other licit crops. Moreover, Ibáñez (2010) presents evidence that coca is often cultivated due to not being able to make a living from legal crops. 10 Such as the Medellin and Cali cartel in the 1990s and eventually the Norte del Valle cartel around 2008. 11 For example, FARC was known to set a fixed price for coca products that was guaranteed to cover the basic costs incurred for farmers (Cook, 2011). Paramilitaries, however, maximise profits, and wages in areas where paramilitaries are in control are lower than in areas controlled by guerillas (Felbab-Brown, 2005). 12 Cultivating coca is illegal and can be considered ‘criminal’activity. However, for simplicity and in line with the literature, this study refers to engagement in the cultivation of coca as being employed in the agricultural sector, and involvement with guerillas or paramilitaries as being involved in the criminal sector. 13 Employment in either sector is not mutually exclusive—it is possible that cultivators divide their efforts between the agricultural and criminal sector. 14 The link between coca production and gold mining is discussed in Rettberg and Ortiz-Riomalo (2016) As it is a feasible option in regions where gold is present, the empirical analysis will account for this; see Section 6. 15 Militias are known to pay wages to their members (Angrist & Kugler, 2008). This has also recently been reported by the International Crisis Group (2021). However, whereas there is very little to no opportunity for personal enrichment in the guerillas and the groups require life-long commitments from its members, the paramilitaries rely on contracts—and members can technically resign, though reality seems to be more complicated—and pay wages and allow for personal enrichment. This implies that both offer very different prospects for potential recruits (Gutiérrez Sanín, 2008). 16 FARC, the largest guerilla group, was known to pay fixed prices for coca products such as leaf, base and paste in the areas it controlled. 17 It is worth noting that the type of violence a group engages in can be affected by their capacity (income). Moreover, guerillas and paramilitaries are known to traditionally engage in different types of violence due to systemic differences between the two groups as noted earlier. This is also discussed in Estancona (2021) and Wright (2016), among others. Though not at the core of this study, part of the empirical strategy will examine the type of violence groups engage in, in response to increase in the OP. 18 These data are also commonly used in other studies focusing on coca production in Colombia and are the source of much of the coca-related data from the Centro de Estudios sobre el Desarrollo Econo'mico (CEDE). 19 In the time period considered, there was a low frequency of paramilitary attacks, but it is considered due to its importance given the relationship between violence and coca in Colombia. 20 Results focusing on the type of attacks (bombings, assassinations, kidnappings, attacks on infrastructure, etc.) are included in the Appendix. 21 The UNODC distinguishes seven regions: Amazonia, Central-Sur de Bolivar-Catatumb, Meta-Guaviare, Orinoquia, Pacifico, Putumayo-Caqueta and Sierra Nevada. STOELINGA 1027 22 There is evidence that the decreasing gold prices around 2015 pushed armed groups into the coca market (International Crisis Group, 2021). First, this does not seem to drive the levels of violence presented in the graph as these levels, across the board, are decreasing around 2015. However, to allow for the possibility that some violence is explained by this, the presence of gold in the municipality will be accounted for in the analysis. 23 In UNODC production, surveys farmers are asked whether they only produce and sell leaf or process leaf to base or paste. The extent to which farmers process leaf varies greatly within regions. For this reason, the prices are weighted by this ‘prevalence of production’rate to create a proxy for the average price that farmers receive for their products. 24 Illicit alluvial gold mining (Rettberg & Ortiz-Riomalo, 2016) is linked to coca production (UNODC, 2013) and can be a feasible alternative source of income for cultivators in certain areas. 25 As there are only 16 clusters in the subsample, the robustness of the results is tested by using the wild bootstrap method proposed by Cameron et al. (2008). The results presented are robust to this specification. 26 The value of HCI is also used by the UNODC (2018) as a variable that captures the value of coca for producers. 27 Note that the similarity of the trends in the figure is also largely due to averaging of the proxies across all municipalities. 28 Farmers can be pressured by local groups to sell their products for a lower value while the price of coca chlorhydate (HCI)—armed groups' OP—remains constant (UNODC, 2018). 28 The fact that income can change but the OP can stay constant can also partially be explained by the fact that bargaining power shifted to international resellers. Additionally, shocks that affect the OP but not necessarily the income for cultivators stem from activities of other guerilla or paramilitary groups—that, for example, expand their territory in order to achieve a stronger bargaining position—or changes in demand for cocaine in foreign markets. 29 The results are robust for specification of violence in levels; see Table A.13. 30 The analysis focusing on alternative measures of violence capturing the type of attack—assassinations, assaults, mass homicides, kidnappings, bombings and attacks on infrastructure—can be found in Table A.12. 31 In a way this is similar to Dube and Vargas (2013), who use international prices for coffee and oil while, ensuring that Colombia's internal price does not affect these international prices. 32 Following Estancona (2021), the presence of a guerilla group is controlled for when considering paramilitary activity and vice versa. 33 In the same vein, Vargas and Restrepo-Jaramillo (2016) show that lack of access to education contributes to child recruitment. 34 Measured as the number of different perpetrators responsible for at least one violent event per year in a municipality. 35 The average coca-producing municipality witnessed, across the entire time period studied (2007–2018) witnessed, an increase of 8.9% in the income from cultivating coca and an increase of 12.2% of the OP. This increase in income is associated with 16.9% fewer violent events, 16.02% fewer fatalities and 11.57% less guerilla attacks. Similarly, the increase in the OP is associated with 25.62% more events, 23.18% more fatalities and 27.82% more guerilla attacks. The average increases in income and the OP across the entire time period, broken down by region, are shown in Table A.11. 36 Interestingly, the authors find an effect of the coffee price shock on paramilitary attacks whereas this study does not. Importantly, the time period considered by Dube and Vargas (2013) is 1988–2004. This time period was significantly more violent and differs fundamentally from the period considered in this study (with one of the main differences with respect to paramilitary activity is that AUC was not disbanded in the time frame studied by Dube and Vargas, 2013). Finally, the fact that violence seems higher in this study, focusing on coca, versus the work of Dube and Vargas (2013)is most likely due to the fact that coca has a more violent history than coffee. Moreover, the scenario with respect to the prices is quite different; whereas Dube and Vargas (2013) study one extremely large, sudden drop in prices, this paper focuses on a time period that is not characterised by such large and unexpected price shifts. REFERENCES Anderson, N., & Worsnop, A. (2019). Fatality thresholds, causal heterogeneity, and civil war research: Reconsidering the link between narcotics and conflict. Political Science Research and Methods,7(1), 85–105. https://doi.org/10.1017/psrm. 2016.22 Angrist, J. D., & Kugler, A. D. (2008). Rural windfall or a new resource curse? Coca, income, and civil conflict in Colombia. The Review of Economics and Statistics,90(2), 191–215. https://doi.org/10.1162/rest.90.2.191 Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies,58(2), 277–297. https://doi.org/10.2307/2297968 1028 STOELINGA Arjona, A., Kasfir, N., & Mampilly, Z. (2015). Rebel governance in civil war. Cambridge University Press. https://doi.org/10. 1017/CBO9781316182468 Becker, G. S. (1968). Crime and punishment: An economic approach. In The economic dimensions of crime (pp. 13–68). Springer. Bjørkhaug, I. (2010). Child soldiers in Colombia: The recruitment of children into non-state violent armed groups. MICROCON Research Working Paper. Bottía Noguera, M. C. (2003). La presencia y expansio'n municipal de las farc: Es avaricia y contagio, ma's que ausencia estatal. Bruce-Jones, T., & Smith, M. (2021). Coca, Clausewitz, and Colombia: The inadequacy of micro-level studies in explaining farc violence against civilians during the Colombian civil war. Studies in Conflict & Terrorism,44(12), 994–1021. https:// doi.org/10.1080/1057610X.2019.1616927 Cameron, A. C., Gelbach, J. B., & Miller, D. L. (2008). Bootstrap-based improvements for inference with clustered errors. The Review of Economics and Statistics,90(3), 414–427. https://doi.org/10.1162/rest.90.3.414 Collier, P., & Hoeffler, A. (2005). Resource rents, governance, and conflict. Journal of Conflict Resolution,49(4), 625–633. https://doi.org/10.1177/0022002705277551 Cook, T. R. (2011). The financial arm of the FARC: A threat finance perspective. Journal of Strategic Security,4(1), 19–36. https://doi.org/10.5038/1944-0472.4.1.2 Cornell, S. E. (2005). The interaction of narcotics and conflict. Journal of Peace Research,42(6), 751–760. https://doi.org/10. 1177/0022343305057895 Cornell, S. E. (2007). Narcotics and armed conflict: Interaction and implications. Studies in Conflict & Terrorism,30(3), 207– 227. https://doi.org/10.1080/10576100601148449 Dal B o, E., & Dal B o, P. (2011). Workers, warriors, and criminals: Social conflict in general equilibrium. Journal of the European Economic Association,9(4), 646–677. https://doi.org/10.1111/j.1542-4774.2011.01025.x DANE. (2011). Gran encuesta integrada de hogares 2013, 2012, 2011, 2010, 2009, censos. Díaz, A. M., & Sánchez, F. (2004). A geography of illicit crops (coca leaf) and armed conflict in Colombia. Dube, O., & Vargas, J. F. (2013). Commodity price shocks and civil conflict: Evidence from Colombia. The Review of Economic Studies,80(4), 1384–1421. https://doi.org/10.1093/restud/rdt009 Estancona, C. L. (2021). Rebel primary commodity markets, price shocks, and supplier victimization. International Studies Quarterly,65(4), 1111–1123. https://doi.org/10.1093/isq/sqab073 Felbab-Brown, V. (2005). The coca connection: Conflict and drugs in Colombia and Peru. Journal of Conflict Studies,25(2), 104–128. Fisher, D. G., & Meitus, A. A. (2017). Uprooting or sowing violence?: Coca eradication and guerrilla violence in Colombia. Studies in Conflict & Terrorism,40(9), 790–807. https://doi.org/10.1080/1057610X.2016.1239430 Flores, T. E., & Vargas, J. F. (2018). Colombia: Democracy, violence, and the peacebuilding challenge. Conflict Management and Peace Science,35(6), 581–586. https://doi.org/10.1177/0738894218787786 Grossman, H. I. (1999). Kleptocracy and revolutions. Oxford Economic Papers,51(2), 267–283. https://doi.org/10.1093/oep/ 51.2.267 Grossman, H. I., & Mejía, D. (2008). The war against drug producers. Economics of Governance,9(1), 5–23. https://doi.org/ 10.1007/s10101-007-0036-1 Gutiérrez Sanín, F. (2008). Telling the difference: Guerrillas and paramilitaries in the Colombian war. Politics and Society, 36(1), 3–34. https://doi.org/10.1177/0032329207312181 Gutiérrez Sanín, F. (2019). Tensiones y dilemas de la produccio'n cocalera. Ana'lisis Pol'ıtico,32(97), 71–90. Hirshleifer, J. (1989). Conflict and rent-seeking success functions: Ratio vs. difference models of relative success. Public Choice,63(2), 101–112. https://doi.org/10.1007/BF00153394 Holmes, J. S., De Pineres, S. A. G., & Curtin, K. M. (2006). Drugs, violence, and development in Colombia: A department-level analysis. Latin American Politics and Society,48(3), 157–184. https://doi.org/10.1111/j.1548-2456.2006.tb00359.x Hristov, J. (2009). Blood and capital: The paramilitarization of Colombia (Vol. 48). Ohio University Press. Ibáñez, M. (2010). Who crops coca and why? The case of Colombian farmers (tech. rep.). Discussion Papers. International Crisis Group. (2021). Deeply rooted: Coca eradication and violence in Colombia (tech. rep.). Kalyvas, S. N. (2006). The logic of violence in civil war. Cambridge University Press. LaFree, G., & Dugan, L. (2007). Introducing the global terrorism database. Terrorism and Political Violence,19(2), 181–204. https://doi.org/10.1080/09546550701246817 Maher, D., & Thomson, A. (2018). A precarious peace? The threat of paramilitary violence to the peace process in Colombia. Third World Quarterly,39(11), 2142–2172. https://doi.org/10.1080/01436597.2018.1508992 Mampilly, Z. C. (2012). Rebel rulers. Cornell University Press. Mansfield, D. (1999). Alternative development: The modern thrust of supply-side policy. United Nations International Drug Control Programme Vienna, 19. STOELINGA 1029 Mejia, D., & Restrepo, P. (2013). Bushes and bullets: Illegal cocaine markets and violence in Colombia. Documento CEDE, (2013-53). Mejia, D., & Restrepo, P. (2016). The economics of the war on illegal drug production and trafficking. Journal of Economic Behavior & Organization,126, 255–275. https://doi.org/10.1016/j.jebo.2015.11.003 Millán-Quijano, J. (2020). Internal cocaine trafficking and armed violence in Colombia. Economic Inquiry,58(2), 624–641. https://doi.org/10.1111/ecin.12771 Moreno-Sanchez, R., Kraybill, D. S., & Thompson, S. R. (2003). An econometric analysis of coca eradication policy in Colombia. World Development,31(2), 375–383. https://doi.org/10.1016/S0305-750X(02)00192-4 Murray, D., Hoell, A., Hoerling, M., Perlwitz, J., Quan, X.-W., Allured, D., Zhang, T., Eischeid, J., Smith, C. A., Barsugli, J., McWhirter, J., Kreutzer, C., & Webb, R. S. (2020). Facility for weather and climate assessments (facts): A community resource for assessing weather and climate variability. Bulletin of the American Meteorological Society,101(7), E1214– E1224. https://doi.org/10.1175/BAMS-D-19-0224.1 O'Connor, D. (2009). The political economy of Colombia's cocaine industry. Papel Pol'ıtico,14(1), 81–106. Olson, M. (1993). Dictatorship, democracy, and development. American Political Science Review,87(3), 567–576. https://doi. org/10.2307/2938736 Rettberg, A. (2010). Global markets, local conflict: Violence in the Colombian coffee region after the breakdown of the international coffee agreement. Latin American Perspectives,37(2), 111–132. https://doi.org/10.1177/0094582X09356961 Rettberg, A., & Ortiz-Riomalo, J. F. (2016). Golden opportunity, or a new twist on the resource–conflict relationship: Links between the drug trade and illegal gold mining in Colombia. World Development,84,82–96. Reuters. (2020). Violence across Colombia leaves 17 dead. Ross, M. L. (2003). Oil, drugs, and diamonds: The varying roles of natural resources in civil war. The political economy of armed conflict: Beyond greed and grievance, 47–70. Saab, B. Y., & Taylor, A. W. (2009). Criminality and armed groups: A comparative study of farc and paramilitary groups in Colombia. Studies in Conflict & Terrorism,32(6), 455–475. https://doi.org/10.1080/10576100902892570 Sánchez De La Sierra, R. (2020). On the origins of the state: Stationary bandits and taxation in eastern Congo. Journal of Political Economy,128(1), 32–74. https://doi.org/10.1086/703989 Suarez, A. R. (2000). Parasites and predators: Guerrillas and the insurrection economy of Colombia. Journal of International Affairs,53(2), 577–601. Sviatschi, M. M. (2018). Making a narco: Childhood exposure to illegal labor markets and criminal life paths. Manuscript, Department of Economics, Princeton University. https://rpds.princeton.edu/pdf/files/sviatschimakinganarco march2018.pdf Thoumi, F. E. (2002). Illegal drugs in Colombia: From illegal economic boom to social crisis. The Annals of the American Academy of Political and Social Science,582(1), 102–116. https://doi.org/10.1177/000271620258200108 UNODC. (2007). Colombia coca cultivation survey 2007 (tech. rep.). UNODC. (2010). World drug report 2010 (tech. rep.). UNODC. (2011). Colombia coca cultivation survey 2011 (tech. rep.). UNODC. (2013). Colombia coca cultivation survey 2013 (tech. rep.). UNODC. (2017). Colombia coca cultivation survey 2016 (tech. rep.). UNODC. (2018). Colombia coca cultivation survey 2018 (tech. rep.). UNODC. (2019). World drug report 2019 (tech. rep.). UNODC. (2021). Cocaine: A spectrum of products (tech. rep.). UNODC & Europol. (2021). The illicit trade of cocaine from latin america to Europe: From oligopolies to free-for-all? (Tech. rep.). Vargas, G. A. (2011). Drugs, hearts and minds: Irregular war and the coca economy in south Bolivar, Colombia (1996–2004). Civil Wars,13(01), 21–39. https://doi.org/10.1080/13698249.2011.555690 Vargas, G. A., & Restrepo-Jaramillo, N. (2016). Child soldiering in Colombia: Does poverty matter? Civil Wars,18(4), 467– 487. https://doi.org/10.1080/13698249.2017.1297051 Vargas, R. (1999). Drogas, m'ascaras y juegos: Narcotra'fico y conflicto armado en Colombia. TM Editores. Veilette, C., & Arvelo-Velez, J. E. (2003). Colombia and aerial eradication of drug crops: U.S. policy and issues (tech rep.). Worldbank. (2018). Poverty and shared prosperity in Colombia (tech. rep.). Wright, A. L. (2016). Economic shocks and rebel tactics. HiCN WP, 232, 1714. How to cite this article: Stoelinga, N. (2024). Cultivation and competition in Colombia: Disentangling the effects of coca price changes on violence. Journal of International Development,36(2), 1007–1042. https:// doi.org/10.1002/jid.3850 1030 STOELINGA TABLE A6 Robustness 2: Including lagged value of dependent variable. Total events Fatalities Guerilla attacks Paramilitary attacks Income t1 1.308*** 0.986** 1.568** 0.030 (0.362) (0.449) (0.605) (0.024) OP t1 1.340*** 1.060** 1.382** 0.023 (0.405) (0.395) (0.574) (0.022) Guerilla presence t1 0.009 (0.006) Paramilitary presence t1 6.773** (3.054) Total events t1 0.037 (0.031) Fatalities t1 0.095*** (0.029) Guerilla attacks t1 0.145*** (0.033) Paramilitary attacks t1 0.130*** (0.015) Controls Yes Yes Yes Yes Year FE Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes N8857 8857 8857 8857 Cluster 32 32 32 32 Note: The table shows the results for estimation 4, including the first lag of each dependent variable. Dependent variables capture violent intensity on municipal level, measured as number of events, fatalities, guerilla and paramilitary attacks per 100 000 inhabitants. Standard errors, clustered by department, in parentheses. *p<0.1, **p<0.05, and ***p<0.01. TABLE A7 Robustness 3: Including department-level trend. Total events Fatalities Guerilla attacks Paramilitary attacks Income t1 1.315*** 0.851** 1.767** 0.029 (0.370) (0.403) (0.687) (0.023) OP t1 1.364*** 0.883** 1.596** 0.023 (0.421) (0.340) (0.611) (0.022) Dept. trend Yes Yes Yes Yes Controls Yes Yes Yes Yes Year FE Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes N8857 8857 8857 8857 Cluster 32 32 32 32 Note: The table shows the results for estimation 4, including a time-varying linear trend for each department. Dependent variables capture violent intensity on municipal level, measured as number of events, fatalities, guerilla and paramilitary attacks per 100 000 inhabitants. Standard errors, clustered by department, in parentheses. *p<0.1, **p<0.05, and ***p<0.01. STOELINGA 1037 TABLE A8 Robustness 4: Including lagged value of dependent variable and department-level trend. Total events Fatalities Guerilla attacks Paramilitary attacks Income t1 1.318*** 0.965** 1.606** 0.029 (0.363) (0.446) (0.611) (0.024) OP t1 1.358*** 1.025** 1.447** 0.023 (0.409) (0.375) (0.587) (0.022) Total events t1 0.036 (0.031) Total events t1 0.036 (0.031) Fatalities t1 0.095*** (0.029) Fatalities t1 0.095*** (0.029) Guerilla attacks t1 0.144*** (0.033) Paramilitary attacks t1 0.130*** (0.015) Paramilitary presence t1 6.822** (3.059) Guerilla presence t1 0.009 (0.006) Dept. trend Yes Yes Yes Yes Controls Yes Yes Yes Yes Year FE Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes N8857 8857 8857 8857 Cluster 32 32 32 32 Note: The table shows the results for estimation 4, including a time-varying linear trend for each department and the first lag of the dependent variable. Dependent variables capture violent intensity on municipal level, measured as number of events, fatalities, guerilla and paramilitary attacks per 100 000 inhabitants. Standard errors, clustered by department, in parentheses. *p<0.1, **p<0.05, and ***p<0.01. TABLE A9 Average percentage point change in proxies. Region Income OP Amazon'ıa 9.230 29.44 Central-Sur de Bol'ıvar-Catatumb 6.825 9.058 Meta-Guaviare 8.112 7.834 Orinoqu'ıa 11.54 9.040 Pac'ıfico 11.65 9.484 Putumayo-Caquet'a 14.00 25.01 Sierra Nevada 0.136 8.304 Note: The table shows the average (across 2007–2018) percentage point change in the proxies for income from farming and the OP, per region as defined by the UNODC. Abbreviation: OP, objective prize. 1038 STOELINGA TABLE A10 Type of attack. Assassinations Mass homicides Bombings Kidnappings Assaults Infrastructure Income t1 0.096*** 0.121 0.743*0.043 0.124 0.137 (0.030) (0.333) (0.397) (0.099) (0.121) (0.153) OP t1 0.083*** 0.135 0.664** 0.007 0.105 0.241 (0.024) (0.307) (0.295) (0.088) (0.080) (0.173) Paramilitary presence t1 0.026 0.172 4.373** 1.027*1.485** 0.898** (0.092) (0.157) (1.671) (0.560) (0.683) (0.354) Guerilla presence t1 0.158*** 0.023 2.722*** 0.564*** 0.693*** 0.439** (0.055) (0.034) (0.347) (0.115) (0.190) (0.186) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes Yes Yes N8857 8857 8857 8857 8857 8857 Cluster 32 32 32 32 32 32 Note: The table shows the results for estimation 4, for different types of attacks. Dependent variables capture the type of violent event on municipal level, measured as number of events per 100 000 inhabitants. Standard errors, clustered by department, in parentheses. Abbreviation: OP, objective prize. *p<0.1, **p<0.05, and ***p<0.01. TABLE A11 Income, OP and violence—levels. Total events Fatalities Guerilla attacks Paramilitary attacks Income t1 0.548*** 0.383*** 0.772*** 0.015 (0.190) (0.126) (0.201) (0.012) OP t1 0.798*** 0.674*** 0.972*** 0.012 (0.249) (0.206) (0.207) (0.012) Competition, % growth 0.338*** 0.370*** 0.485*** 0.009** (0.044) (0.053) (0.020) (0.004) Laboratories 0.018*** 0.011 0.022*0.000 (0.005) (0.010) (0.012) (0.000) Gold producer 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Coffee producer 0.005 0.004 0.008 0.000 (0.004) (0.004) (0.007) (0.000) Coca producer 0.000 0.000 0.000 0.000* (0.000) (0.000) (0.000) (0.000) Man. eradication 0.084*** 0.166** 0.195*** 0.001 (0.023) (0.068) (0.053) (0.001) Eradication 0.078*** 0.075 0.062*0.002 (0.026) (0.102) (0.035) (0.002) (Continues) STOELINGA 1039 TABLE A11 (Continued) Total events Fatalities Guerilla attacks Paramilitary attacks Confisc. leaf 0.000*** 0.000*** 0.000*** 0.000 (0.000) (0.000) (0.000) (0.000) Confisc. base/paste 0.000 0.000*0.000 0.000 (0.000) (0.000) (0.000) (0.000) Confisc. HCI 0.000*** 0.000*** 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Rain 0.000 0.000 0.000 0.000* (0.000) (0.000) (0.000) (0.000) Temperature 0.008 0.017 0.011 0.003 (0.026) (0.037) (0.054) (0.002) Income t2 1.188*** 0.924*1.542*** 0.031 (0.415) (0.512) (0.298) (0.028) Income t3 0.060 0.262 0.156 0.010 (0.269) (0.450) (0.292) (0.009) Income t4 0.075 0.635 0.065 0.012 (0.220) (0.542) (0.222) (0.013) OP t2 1.189*** 0.488 1.571*** 0.028 (0.387) (0.346) (0.313) (0.025) OP t3 0.045 0.015 0.138 0.007 (0.264) (0.470) (0.290) (0.006) OP t4 0.127 0.312 0.167 0.006 (0.213) (0.415) (0.218) (0.009) Paramilitary presence t1 1.832 (1.186) Guerilla presence t1 0.009 (0.006) Controls Yes Yes Yes Yes Year FE Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes N8857 8857 8857 8857 Cluster 32 32 32 32 Note: The table shows the results for estimation 4. Dependent variables capture the level of violence on municipal level, measured as number of events, fatalities, guerilla and paramilitary attacks. Standard errors, clustered by department, in parentheses. Abbreviations: HCI, hydrochloride; OP, objective prize. *p<0.1, **p<0.05, and ***p<0.01. 1040 STOELINGA TABLE A12 Opportunity cost and rapacity effect. Hours worked Highest level of education Attending school Local competition Guerilla presence Paramilitary presence Income t1 2.200 0.126 0.013*0.160** 0.159*** 0.014 (2.944) (0.185) (0.006) (0.073) (0.050) (0.012) OP t1 0.463 0.107 0.014*0.183** 0.185*** 0.011 (2.835) (0.196) (0.007) (0.068) (0.040) (0.011) Age 0.112*** 0.025*** 0.002*** (0.028) (0.003) (0.000) Gender 14.041*** 0.513*** 0.024*** (2.771) (0.070) (0.007) Competition, % growth 0.326 0.000 0.002 0.820*** 0.545*** 0.009** (0.205) (0.010) (0.002) (0.023) (0.051) (0.004) Gold producer 0.000 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Coffee producer 0.138 0.002 0.001*0.000 0.000 0.000 (0.120) (0.006) (0.001) (0.001) (0.001) (0.000) Coca producer 0.000 0.000 0.000 0.000 0.000 0.000* (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Man. eradication 0.944** 0.045** 0.002*0.007 0.021 0.001 (0.340) (0.021) (0.001) (0.011) (0.019) (0.001) Eradication 0.784** 0.029*0.001 0.010*0.011 0.002 (0.295) (0.014) (0.001) (0.005) (0.007) (0.002) Rain 0.012*0.000 0.000 0.000 0.000 0.000* (0.006) (0.000) (0.000) (0.000) (0.000) (0.000) Temperature 0.435 0.012 0.001 0.018 0.002 0.003 (0.883) (0.032) (0.003) (0.018) (0.012) (0.002) Confisc. leaf 0.000 0.000 0.000*0.000*** 0.000*** 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Confisc. base/ paste 0.000 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Confisc HCI 0.000 0.000 0.000 0.000 0.000*** 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Income t–2 3.181 0.105 0.013 0.227*** 0.425*** 0.030 (2.153) (0.119) (0.016) (0.073) (0.095) (0.028) Income t3 3.204 0.067 0.004 0.041 0.104 0.010 (3.305) (0.185) (0.022) (0.061) (0.062) (0.009) Income t4 3.071 0.140** 0.006 0.066*** 0.065*** 0.011 (2.993) (0.056) (0.012) (0.021) (0.019) (0.013) OP t2 4.199*0.081 0.013 0.218** 0.463*** 0.028 (2.353) (0.136) (0.017) (0.087) (0.128) (0.025) OP t3 2.556 0.088 0.005 0.046 0.085 0.007 (3.467) (0.225) (0.022) (0.065) (0.062) (0.007) (Continues) STOELINGA 1041 TABLE A12 (Continued) Hours worked Highest level of education Attending school Local competition Guerilla presence Paramilitary presence OP t4 3.186 0.130*0.007 0.061** 0.089*** 0.006 (2.772) (0.065) (0.011) (0.025) (0.026) (0.009) Paramilitary presence t1 3.908*** 0.028 0.032 0.869*** 0.381*** 0.136*** (1.016) (0.044) (0.036) (0.069) (0.068) (0.005) Guerilla presence t1 0.469 0.023 0.002 0.839*** 0.563*** 0.009 (0.446) (0.020) (0.003) (0.031) (0.047) (0.005) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Municipality FE Yes Yes Yes Yes Yes Yes N3437 3437 3437 8857 8857 8857 Clusters 24 24 24 32 32 32 Note: results for estimation of Equation (5). Competition growth is the annual relative change in the number of different groups active in a municipality, the control variable included in each analysis. Standard errors, clustered by department, in parentheses. Abbreviations: HCI, hydrochloride; OP, objective prize. *p<0.1, **p<0.05, and ***p<0.01. 1042 STOELINGA