Conventional wisdom has it that organized criminal groups (OCGs)—drug cartels, gangs, and mafiasFootnote 1 —emerge to violently capture rents from illicit goods sold on informal markets.Footnote 2 Much of our knowledge about OCGs comes from illicit drug markets.Footnote 3 However, examples across time and space tell a different story: in the 1800s in Sicily, the mafia took over the market for lemons, and in the 1900s in New York, it sought to control the artichoke market. In the 1990s, street gangs in Cape Town encroached on the abalone market, and in Mexico in the 2010s drug cartels took control of the lime and avocado markets.Footnote 4 If trading in illicit goods is so profitable, why bother with artichokes, lemons, and shellfish?
Research finds that OCGs attempt to control profitable markets.Footnote 5 Yet diversification is costly: economies of scale are not always present, not all markets are complementary, and not all groups are equally risk tolerant.Footnote 6 Because research on the conditions for criminal diversification into licit markets is underdeveloped, it is unclear whether the drivers or the violence associated with diversifying into these markets are the same as those in illicit markets. To fill this gap, we develop a theory of how criminal groups diversify their economic portfolios into markets for licit commodities and when this diversification is likely to generate violence.
We argue that when the export value share of agricultural goods rapidly increases, OCGs seek to profit from these newly lucrative markets. Positive price shocks, specifically in agricultural goods exported in high quantities, present unique opportunities for profit, local territorial gain, and long-term control that merit the risk of diversification, or criminal market capture. To secure territorial and economic control and capitalize on these opportunities, organizations engage in violence against producers, the state, and rivals. In developing our argument we address two interrelated questions: When does criminal diversification into markets for licit commodities happen? And when does it produce violence?
Our work contributes to research on OCGs in four ways. First, our focus on markets for licit commodities beyond mere rent capture identifies another area of criminal involvement. This provides a more complete understanding of the motivations behind OCGs’ behavior and demonstrates that they are not limited to illicit markets.
Second, our theory of diversification and violence applies to markets for licit agricultural commodities that are territory-bound, contributing to new research on the drivers of criminal governance at the local level. To control production as part of the process of diversification, OCGs seek to control the territories where commodities are grown. To increase future profit, territorial control may require violence. Focusing on Mexico, we highlight the consequences of criminal attempts to control local life, furthering our understanding of criminal governance. Inasmuch as their behavior affects community behavior, it has important implications for our understanding of politics more generally.Footnote 7 Furthermore, OCGs today can produce violence comparable to that of civil wars,Footnote 8 so understanding when criminal violence escalates is central to the study of peace and security.
Third, we expand our case-study findings to generate a cross-national explanation of OCG behavior. Existing knowledge comes primarily from case studies, which help demonstrate causal mechanisms but are not well suited for developing a general account. We draw on this rich case literature to develop and test a cross-national account. This sheds light on international agricultural markets as an important trigger of criminal diversification.
Fourth, our study has significant policy implications. If local violence can be triggered by changes in international markets for licit goods, policies to combat OCGs should not be limited to illicit markets and local military or police responses.
Our empirical strategy combines analysis of the Mexican avocado market with statistical analysis of a global sample of countries between 1993 and 2018. To assess the validity of our argument we use data on agricultural production, criminal threats to the state, and homicides. We complement our statistical analyses with evidence from secondary sources to demonstrate the mechanics of criminal market capture of Mexican avocados.
Our strategy faces several challenges. First, we do not have a direct measure of our main theoretical mechanism—criminal market capture—at the global level. To mitigate this challenge, we leverage the case of Mexico. We zoom in on the avocado market to show how criminal market capture happens by (1) incorporating extensive field research from secondary sources as an illustration, and (2) leveraging location-specific data on criminal presence in Mexico. We address reverse causality with a plausibly exogenous global indicator of avocado export value and conduct a placebo test using corn, strawberry, and lime crops to test the importance of export value share.
Conversely, idiosyncratic features of the Mexican case may limit our ability to generalize, which we address with a cross-sectional country-level analysis. We devise a general theory and test its applicability to a large number of cases, ensuring that the implications of our argument are observable in a representative case. We consider our cross-national analysis as additional descriptive evidence, and encourage future testing of our proposed mechanism in this more generalized setting.
A second challenge arises from features of our dependent variable. Homicide statistics have limitations, including under-reporting and lack of granularity in terms of perpetrators and targets. These limitations may lead to downward bias in our results, and an inability to distinguish between violence involving OCGs and other societal violence. We accept this limitation and interpret our results with caution.
A third challenge stems from possible reverse causality. We posit that OCGs capture a market following increases in export price and export share. However, it could be that OCGs first manipulate production and supply to influence export value and profitability in international markets. To address this concern we introduce a plausibly exogenous measure of increases in the international demand for avocados: the popularity of the search term “avocado toast” on Google Trends.
The results from Mexico are in line with our expectations: increases in a municipality’s share of export value of avocados are associated with increases in homicides where OCGs are present. Our theory is also supported by the general statistical results: rapid increases in a state’s share of agricultural goods’ export value—an indicator of both immediate opportunities for profit and the plausibility of future market manipulation—are associated with more homicides, but only where OCGs threaten the state. Overall, we find that unexpected changes in international markets for agricultural commodities are an important driver of the domestic dynamics of criminal violence.
Cost-Effectiveness, Criminal Market Capture, and Violence
We build on current research suggesting that OCGs diversify into new markets when it is cost-effective. One way to reduce costs is by having a comparative advantage. Explanations emphasizing comparative advantages posit that OCGs will diversify into areas where their expertise allows them to reduce entry costs. For example, organizations that smuggle weapons over a border may use their knowledge and access to routes to also smuggle drugs or migrants.Footnote 9
But it is not clear that comparative advantages in illicit markets would apply to markets for licit goods in the same way. OCGs may encounter additional challenges in markets for licit goods: the regulated economy involves taxation and increased state controls, which drives up entry costs and reduces the benefits. It may also be that other costs, beyond standard economic inputs, operate differently, like the cost of controlling populations, or that of removing competitors. It is likely that such costs are higher in licit markets as well, given that they also increase the cost of using violence because it is more difficult to hide.
Our theory emphasizes changes in markets rather than the nature of groups as a means to generate cost-effectiveness. When a good becomes unexpectedly lucrative and offers opportunities for price manipulation in the future, organizations will prioritize that market because they can offset the costs of diversifying. When these goods are difficult to transfer to other territory, violent competition over growing areas becomes more likely.
Opportunities for expansion can generate violence as organizations compete with one another, the state, or other economic interests. Opportunities typically emerge from political shocks; for example, protection rackets—informal mechanisms for the selective application of the lawFootnote 10 —are destabilized when corrupt public officials are removed after elections, and disputes over the regulation of illicit markets trigger spirals of violence. Either OCGs dispute the control of illicit markets,Footnote 11 or they fight the state, depending on how it prosecutes them.Footnote 12 The focus on illicit markets does not provide specific predictions for violence in markets for licit goods. Nor does it consider how nonpolitical disruptions may shape patterns of violence (for an exception, see Dube).Footnote 13 We propose that changes in markets for licit goods can trigger criminal violence as groups diversify and compete for access to the production and sale of these goods.
There is still the question of which markets to diversify into: some features of goods, particularly their lootability, can also increase cost-effectiveness and their attractiveness to nonstate actors. When natural resources are relatively isolated, they are easier to control.Footnote 14 But a market’s ease of capture is not enough. To be desirable, it must add significant value—including profit and opportunities such as territorial, political, or social control—to an OCG’s portfolio. Furthermore, not all natural resources are equally lootable. For example, the exploitation of raw materials such as oil or certain minerals requires higher capital investment than most agricultural commodities.Footnote 15 And within agricultural goods, some have a longer growing period and cannot be easily moved. For example, avocados—the main commodity discussed in this paper—must usually mature for three to five years before they consistently yield fruit.
Long-standing research indicates a strong association between commodities and conflict.Footnote 16 When commodities are highly profitable, they trigger competition over their control, which increases violence.Footnote 17 Other scholarly research casts doubt on high value as a trigger and points to price shocks as a catalyst for violence. For example, negative price shocks in the international market for maize drove Mexican laborers into cultivation of illicit drugs, which in turn generated incentives for OCGs to fight for control over the illicit crops.Footnote 18 Furthermore, in their meta-analysis, Blair, Christensen, and Rudkin find no evidence of a link between higher commodity prices and violence, but do find evidence that when the price of labor intensive commodities increases, violence is reduced, except when commodities are simultaneously labor intensive and lootable.Footnote 19 Still other research suggests that how changes in illicit commodity prices affect violence depends on whether such violence is produced primarily by the state or nonstate groups.Footnote 20
Since most of this research is specific to a case or a (usually high-value) commodity, we lack a general understanding of the conditions for diversification and violence across commodities and countries. We also know little about criminal actors’ motivations for expansion into markets for licit goods. Much of the research on resource lootability and territorial control or competition focuses on civil war violence, which differs from OCG violence in its strategic purpose and goals. Rebel organizations seek territorial control for political and military advancement and the ability to engage in state-like governance.Footnote 21 Regarding resource lootability, rebels may prefer stationary banditry but continue to loot if investing in resource production is too costly or detracts from their long-term political goals.
For OCGs, however, the reason to seize territory is to expand their economic empire by reducing the production and export costs of key goods and deterring state or competitor presence. Territorial control remains costly, so OCGs are unlikely to pursue it without significant opportunity for profit (whereas the rebels featured in previous theories obtain other political and reputational benefits—even other economic benefits, such as taxation).Footnote 22 Because territorial control itself has fewer benefits for OCGs than for rebels, we argue that OCGs pursue key growing territories only when they provide opportunities for long-term economic expansion. Building on earlier findings, we propose a theoretical refinement by considering the size of the export market. We argue that where a state or territory’s export value share is sufficiently large, international price increases incentivize OCGs to attempt market capture of these commodities. When such goods’ value rapidly increases, this is an opening for profit and eventual market capture.
To summarize, we do not observe criminal diversification into all profitable markets, nor does diversification entail the same levels of violence wherever it occurs. According to our argument, rapid increases in price in markets with opportunities for future control provide windows of extreme profitability. We demonstrate the importance of both short- and long-term incentives for diversification. We argue further that because these commodities are territory-bound, controlling production allows groups to engage in coercive governance, with the ultimate goal of maximizing profit.Footnote 23 Violence, in turn, is the result of how much resistance they face in the process within this territory. We develop this argument next.
A Theory of Criminal Diversification
In this section we develop a theory of criminal diversification. We explain the conditions under which OCGs decide to attempt diversification into licit markets for agricultural products and why this process generates violence. With this objective in mind, we develop a theory about the cost–benefit calculations OCGs face when deciding to engage in the process we call criminal market capture. We conceptualize the decision to diversify as resulting from considering the benefit from the potential long-term gains from market capture and the costs imposed by the resistance groups will face, which in turn shapes how much violence to expect in the territory where these agricultural products are grown. In doing so, we highlight that opportunities for long-term control amplify the benefits of territorial expansion.
OCGs trade in illicit commodities because of the immense profits derived from their control. To increase their profit, they may seek to diversify by capturing additional markets for goods or services. Diversification may provide direct profit in the short term, opportunities for increasing profit in the long term, or socioeconomic control that reduces costs in the long term. The section proceeds as follows. We first define relevant concepts, and discuss the scope of application of our theory. We then discuss our proposed causal mechanism and outcome. Finally we consider observable implications. The subsequent sections describe our identification strategy and analysis.
Concepts
Market capture is the process of establishing control over portions of the market for a commodity. Control is decision-making power over the activities involved in a market: determining prices, establishing volumes to be sold or amounts to be harvested, or selecting wholesalers. Criminal market capture means forcefully and illegally controlling portions of the market for a commodity, eventually manipulating production and prices.
Criminal markets emerge when brokers gain control, partial or complete, of the chain of production and distribution of a given product by evading regulations.Footnote 24 For example, in the 1920s the Morello-Terranova mafiosi in New York leveraged their control over the distribution of artichokes coming from California because they controlled the rail line and imposed a “tax” on artichokes arriving from California by train.Footnote 25 The Sicilian mafia sought to control the production of lemons and manipulated prices by controlling the groves,Footnote 26 and South African gangs leveraged their storage capacity to manipulate local prices in the abalone market.Footnote 27
Scope Conditions
Our theory is bound by a set of scope conditions, which include the type of goods, the kind of price shifts, and the size of the export market share. We focus on the production of territory-bound agricultural commodities in emerging markets, where producers use semi-skilled labor, and where there is a high degree of asset specificity in agricultural production (that is, fixed investments in groves, with relatively immobile labor used to harvest the products). Commodities in these specific markets require less investment and specialized knowledge than in contexts where agricultural commodities are produced but require high investments due to high technification and the need for skilled labor.Footnote 28
We would not expect OCGs to start exploiting a market in places with significant costs of entry. Because agricultural commodities are territory-bound and just a few producers control them, capturing and controlling production becomes easier, reducing costs of entry.Footnote 29 Further, in emerging markets, agricultural commodities typically require less advanced technology, labor, and tailored production knowledge. Existing shipping routes, infrastructure, labor, and local markets also reduce up-front costs.Footnote 30 These features reduce the cost of diversification and increase profitability, making them ripe for criminal investment. Because controlling the production of territory-bound commodities ultimately involves controlling producers, minimizing state involvement, and expelling other criminal competitors, we expect more violence at this point in the chain of production and distribution than at other points.Footnote 31
We argue that OCGs pursue diversification when there are unusually large increases in the export value of licit goods that can be grown locally. These disruptions bring attention to the new high value of a commodity and create an opportunity to increase short- and long-term profits. Our argument zeroes in on countries’ global market share of agricultural commodities as an overlooked factor in incentives for market capture. International price increases for agricultural commodities create a significant opportunity for immediate profit. But when OCGs operate in states that hold a large portion of the international market share of these goods, rapid increases in their price provide not just immediate profit but also an opportunity for eventual control and manipulation.
Because our goal is to explain the initial impetus for market capture, rather than continued control, we focus on expansion into new markets following positive shifts in price. Criminals could capture a market after a negative shock, work to restrict production, and raise prices, but this process requires time, and its outcome is uncertain. A rapid increase in price provides confidence that there will be dividends in the immediate future. Groups could also attempt diversification into markets for high-price licit goods such as gold or gems in the absence of a shock, but this might require divesting from the illicit business. For example, when valuable minerals can be concealed, as with gold, OCGs need to develop additional surveillance and tax structures, which raise, rather than lower, the costs of control in the long run.Footnote 32 Focusing on high-price licit goods would also require knowledge about which markets to attempt to control, along with significant up-front costs. High-price licit goods also provide fewer benefits to organizations in terms of expanding their political or social control because such goods require only minimal criminal presence to manage production when compared to agricultural goods.
Although agricultural goods are not inherently valuable like diamonds, gold, or oil, they require less up-front investment in infrastructure while providing an opportunity to advance organizations’ long-term economic goals. Because agricultural goods require territory for growth, controlling their production can bolster OCGs’ ability to build informal governance structures through territorial control—which may deter challengers in the future and aid in additional economic investment. The cost of moving or replanting some crops with long growing periods (such as avocados, limes, olives, or even asparagus) makes the territory in which they grow particularly valuable—and sparks competition and violence over this territory following changes in their markets. Our theory thus explains how changes in the value of seemingly mundane goods can trigger diversification through the perceived opportunity for profit, expansion, and control. Since OCGs’ main goal is to maximize profit from illicit markets, we expect diversification to happen when a distinct opportunity arises: rapid increases in price in markets that criminals can reasonably seek to capture and eventually manipulate.
OCGs seek to maximize both short-term payoff and long-term financial investment. Consequently, price shifts alone are insufficient: the available export share of the global market matters. If OCGs are focused on maximizing profit, capturing a market that represents only a small share of the international economy might temporarily allow them to increase their profit by reducing that of producers, but it would not allow them to manipulate prices as effectively.
Empirical examples provide descriptive support for our argument. South Africa is the third-largest producer of abalone in the world; Italy in the nineteenth century was a leading producer of lemons; Mexico is the world’s largest producer of avocados and second-largest producer of limes. OCGs have targeted all of these markets.Footnote 33 Export share, in combination with upward shifts in price, creates a lucrative window for immediate profit and incentives for market capture. OCGs are aware of the international profitability of these markets and changes in price because they act as brokers between sellers and international buyers and as contract enforcers.Footnote 34 In synthesis, OCGs are most likely to enter a licit agricultural market when the state where they are present accounts for a sufficiently large share of global agricultural production, when the good is territory-bound, and when the price of the good rises significantly. We tie these scope conditions to examples of crops in Mexico in Table 1.
Characteristics of key Mexican agricultural exports
Outcome
OCGs are likely to face tensions when seizing control of newly profitable markets. They have an incentive to minimize violence (to avoid attracting unwanted attention to their activities),Footnote 35 but violence ensues in the process of market capture as a result of three kinds of tension: disputes with other groups, resistance from producers, and resistance from the state. If a good is newly lucrative, it can spark competition between groups to seize profit quickly and ensure their access to the territory in which the good is grown. This is particularly likely when the good in question has a longer growing period or is costly to replant elsewhere, making the particular territory worth competing over to quickly banish other would-be OCG investors. This rush may make OCGs more violent toward one another and toward producers in an effort to coerce cooperation with their particular group.Footnote 36 OCGs may invade plantations and forcefully remove producers,Footnote 37 regulate picking season, or force producers to sell their land to them.Footnote 38 Sometimes, groups can build cooperative relationships with communities and manage to keep violence low,Footnote 39 but these arrangements take time to build, and they are often fragile, short-lived,Footnote 40 and contingent on groups’ territorial control.Footnote 41 Given these factors, we expect more violence in the initial stages of market capture.
Variation in violence also results from the type of relationship between groups and the state.Footnote 42 When the state confronts OCGs, either to reclaim spaces under criminal control or because the security apparatus is protecting one group versus another, violence escalates.Footnote 43 States also establish cooperative or nonconfrontational relationships with OCGs.Footnote 44 If the state intentionally avoids confronting an OCG and there are no other competitors, or if law enforcement is absent from a territory, violence will be minimized.Footnote 45 Alternatively, when international demand for an important taxable product rapidly increases, the state may use force to limit criminal involvement. Because agricultural production happens in rural communities, which tend to be more isolated, state awareness of the presence of OCGs in new markets, and subsequent intervention, may take some time. OCGs have incentives to organize takeovers quickly, before the state has an opportunity to become an additional source of confrontation, and before it needs to be bought out.
Mechanism
The mechanism connecting our observable outcome (violence) to our independent variable (positive price shifts for agricultural products representing a large portion of the export market) is the process of market capture. The opportunity created by a price increase triggers OCG actions to facilitate criminal market capture. We take the following actions as evidence of the onset of a capture process: controlling production by regulating the picking season; taking over packing plants and transportation; and taking over land. These actions afford longer-lasting benefits beyond immediate access to rents through, for example, the imposition of a “tax,” and differ from looting due to their territorial semi-permanence.
Controlling the markets for licit agricultural goods also affords OCGs opportunities to build criminal governance in the long run—using mechanisms of population control to maximize profit.Footnote 46 Through market capture, OCGs can access the local communities where these markets are embedded and gain control over local institutions such as municipal governments and producers’ associations. Eventually, they may manipulate the electoral process, ultimately ensuring control over policies that make illicit activity easier.Footnote 47 In the long run, this dynamic can become self-reinforcing.
Although we cannot provide a global, systematic measure of market capture, in our discussion of the Mexican case we provide qualitative evidence suggesting the presence of this mechanism. The nature of our global data is such that we cannot test for the presence of the mechanism of capture at the global level. We note this as a limitation of our work.
Observable Implications
We expect more violence as groups expand economically and territorially. Changes in the profitability of territory-bound agricultural goods’ and the expansion of their export markets can spark competition from other organizations, resistance from producers, or challenges from the state—all of which increase the use of violence by OCGs in these territories. Based on these expectations, we develop observable implications, which we test by leveraging our case study. To assess the extent to which our argument might apply more generally, we also test the connection between changes in commodities’ export value share and criminal violence at the general level. We present observable implications here and formulate them as hypotheses within each section (case study and global sample, respectively). Table 2 summarizes our proposed evidence for the observable implications of our theory at the case and general levels.
Observable implications: change in export value share and criminal market capture
To test implications at the case level, we leverage different sources of data. Qualitative data allow us to probe market capture as our mechanism of interest. Quantitative data, including a plausibly exogenous measure of global avocado price and exploration of placebo crops (corn, limes, and strawberries), allow statistical assessment of how price and export share increases jointly function as a catalyst for OCG violence. And at the general level, we leverage quantitative data to provide evidence of a relationship between increases in export value share and violence, but only where criminal organizations threaten the state.
Other Explanations
We propose that violent criminal expansion occurs following increases in export value share through the mechanism of market capture, but here we consider three other logics: loss of revenue, money laundering, and market disruption. First, if drug markets become less lucrative, perhaps OCGs will diversify into other markets to make up for lost revenue. But the United Nations Office on Drugs and Crime (UNODC) finds that drug prices have remained stable, and even increased, across a variety of substances, including cocaine and synthetic drugs.Footnote 48 And even if diversification stemmed from changes in drug markets, it would not invalidate our argument. We expect groups to diversify into markets that allow them to maximize profits, and thus we offer an explanation for which markets they choose. Further, diversification into new markets does not imply substitution. Changes in the value and export share of agricultural goods create an opportunity for market capture as a complement to, not a substitute for, drug markets. In fact, diversification may involve parallel processes of development in licit and illicit markets.Footnote 49 OCGs may respond to opportunities resulting from shifts in markets for licit commodities in addition to the profit that can be made in drug markets.
The second explanation is tied to surplus cash from illicit market activity that needs to be laundered. In this case, whether the market has value in itself may be less consequential, and diversification should happen without violence. OCGs may seek to launder their proceeds through legal businesses, and diversification can be a mechanism to facilitate this. One example is the acquisition of gas stations by drug cartels in Mexico.Footnote 50
But although the need to launder money may be a reason to diversify, there are strong incentives to engage in licit economies beyond smoke screens. Patterns of investment for money laundering are unlikely to be the same as investment for profit from the sale of goods. A recent report from the Financial Action Taskforce, an intergovernmental body working to prevent money laundering, suggests that there are more efficient ways to launder large sums of cash than capturing territory-based commodities, which may meet resistance from competitors, producers, or the state. These tools include smuggling cash, using money brokers, or using the banking system.Footnote 51 Money laundering depends on secrecy, minimizing surveillance and traceability; using newly profitable and popular goods would present additional, unwanted challenges.
Disruptions in the illicit economy are a third possible pathway. Governments frequently target illicit markets, and perhaps expansion into licit ones would minimize detection and loss of profit. The risk of disruption in criminal markets is high: cocaine seizures have increased worldwide over the last twenty years—but so has production.Footnote 52 Stricter law enforcement has not impacted these groups’ ability to continue to adapt and profit elsewhere. Other research has demonstrated that domestic drug prices temporarily increase as a result of disruptions in the illicit economy, further increasing OCGs’ profit.Footnote 53 Furthermore, states have strong added incentives to limit criminal involvement in licit markets because of loss of profit, which suggests we should expect greater disruption in these markets.
In the next section we present the case of Mexico as part of our theory-building exercise. We provide qualitative evidence of our mechanism of interest, complemented by a quantitative analysis of the avocado market. The subsequent section presents a test of our argument beyond the Mexican case, to probe its generalizability.
The Case of Mexico
To illustrate our mechanism of interest, we conduct subnational analysis in Mexico combining qualitative evidence from secondary sources with data on OCGs’ territorial presence and on avocado exports. The Mexican case offers important advantages: there is wide variation in agricultural production, as well as in the presence of OCGs and levels of violence, across subnational units. In our design, Mexico can be characterized as a typical case. Typical cases are representative of a population defined based on the scope of a theoretical argument.Footnote 54 Our argument is scoped to apply to states that are exporters of agricultural products and have active OCGs.
As a way to illustrate where Mexico lies in relation to other countries within the scope of our argument, Figure 1 plots average agricultural export value against average homicides per year for various countries. The countries that are labeled are above the threshold of ten homicides per 100,000 population per year—an epidemic level of violence, according to the World Health Organization’s definition, suggesting OCG presence.Footnote 55 Clearly Mexico is not the only possible case. However, it is one where data are available for all our central analytical indicators: agricultural production, violence, and OCG presence at the local level.Footnote 56 In addition, research on the dynamics of organized crime in Mexico is extensive, which allows us to triangulate our information and analyses with multiple other sources.
Case selection
Mexico is one of the world’s main producers of several commodities; OCGs are widespread there, and homicide levels are high. If our theory is correct, booms in market share of certain agricultural products should lead to criminal attempts to capture these lucrative markets. In turn, if we observe these dynamics in Mexico, we should expect to see them in comparable cases. In the next section, we zoom in on the avocado market. We provide qualitative evidence from earlier research in Michoacán, the state where most of the country’s avocado production is concentrated, and quantitative evidence of the association between criminal diversification and violence in the market for avocados, which extends beyond Michoacán. We also provide evidence for our proposed mechanism of market capture—OCGs responding to price increases in export markets that are substantial enough to allow eventual manipulation—by comparing avocados to other Mexican crops.
Avocado Production and Cartel Violence in Mexico
Scholars have posited multiple mechanisms to explain the growth and violent expansion of OCGs in Mexico: changes in the international drug markets; features of the democratic transition; changes in networks of protection from the authoritarian period; increased inter-cartel competition; violent confrontations with the state.Footnote 57
As profitable as illicit markets are, Mexican OCGs are involved in a variety of activities. Early accounts of connections between licit and illicit markets date back to the 1940s and the parallel development of the opium and tomato industries in the state of Sinaloa.Footnote 58 Recently, in states like Michoacán, OCG violence has been linked to the avocado, berry, and citrus industries. Because Michoacán produces the majority of Mexican avocados, we should observe the mechanics of our theory at work there.
OCGs have been present in Michoacán since the 1950s, due to its suitability for drug cultivation and its privileged position as a transportation route.Footnote 59 But despite this long-standing OCG presence, violence associated with the drug trade dates back only to the mid-1990s, and it is driven by group competition, as well as state and communal responses triggered by changes in the international drug markets. LeCour and Frissard, Herrera, and Moncada all document changes in criminal presence in Michoacán, resulting in spirals of violence as a result of groups moving to capture new markets.Footnote 60 In the late 1980s and into the 1990s, the Valencia family controlled the marijuana and poppy trade through a drug-trafficking group called Milenio. In the early 2000s, they were displaced by the Zetas, who took control over the lucrative market for illicit drugs, including cocaine and synthetic drugs. Years later, the Zetas were displaced by the Familia Michoacana. Avocado production, present in Michoacán since the 1950s, began to rapidly increase in the mid-2000s in response to changes in international demand, which boosted export revenue. Although OCGs were already present in the state, they had previously focused on controlling the drug markets.
In line with our theory, studies of criminal involvement in the avocado market situate it around the late 2000s, after the boom in exports.Footnote 61 This temporal dynamic is central to our argument; it suggests that the mere presence of easily accessible commodities is not enough to motivate capture. By the 2010s, the Caballeros Templarios, who displaced the Familia Michoacana, were targeting avocado production directly. Because avocado producers kept detailed information on productivity, OCGs knew how much profit they could make. Case-study research shows that the Templarios had information on agricultural registers and cadastres in Michoacán.Footnote 62 Although the mechanism of criminal market capture is not directly observable, using a process-tracing approach,Footnote 63 we can provide causal observations using qualitative evidence of the actions OCGs took to take over and manipulate the production of avocados. If OCGs were only interested in the territory, or in extracting rents, they would not have sought direct control of production and packaging. As documented through extensive interview evidence from secondary sources, OCGs implemented a series of strategies to capture the avocado market: they took over packaging branches, intimidated farmers into giving up their lands, and threatened pickers into delaying or halting the harvest.Footnote 64 Those who resisted were killed. By 2013, the Templarios had taken over much of the avocado production chain in Michoacán.Footnote 65
At the time of the criminal market capture in Michoacán, prospects for continued control provided an important incentive. OCGs’ efforts to control commodities had the dual purpose of generating revenue and establishing criminal governance, ultimately affecting OCG’s plans to continue to profit from illicit markets.Footnote 66 As a result of the increased criminal market capture, violence rose dramatically.Footnote 67 Self-defense organizations sprung up in areas pressured by criminal activity in an attempt to protect farmers and communities. These areas experienced more violence as a result of clashes between such self-defense organizations and OCGs, as well as between OCGs competing for prime avocado-growing territory.Footnote 68
Beyond qualitative evidence at the micro level, there has been little systematic exploration of the relationship between changes in agricultural markets and criminal control in Mexico.Footnote 69 We test our argument by focusing on the avocado market beyond Michoacán, combining homicide data at the municipality level with data on exports of avocados from Mexican municipalities and on criminal presence. Figure 2 shows the spatial distribution of avocado cultivation along with criminal presence. We can see that Mexican avocados are also produced in states other than Michoacán, and that OCGs are present in many areas where avocados are grown for export.
Avocado production (tonnes) and criminal presence
According to our argument, Mexican cartels should be attracted to agricultural markets when there are large, rapid changes in the export value of goods for which Mexico represents an important portion of the global market. Mexico provides approximately 40 percent of the world’s supply of avocados, which it exports primarily to the United States.Footnote 70 Violence results from challenges from competitors, state security forces, or avocado producers, as OCGs expand to new markets and specific growing territory. In a later section, we conduct placebo tests replacing avocados with corn, limes, and strawberries to assess the importance of the size of the export market, changes in price, and mobility/territorial boundedness of the crop. Table 1 highlights the differences between these crops and avocados, clarifying why we expect more homicides in connection with avocados but not with these other goods. There is no statistical relationship between the production of these other crops and homicides.Footnote 71
Our Mexico-specific hypothesis is:
H1local: Increases in a municipality’s share of avocado export value are associated with an increase in OCG violence in that municipality.
Subnational Analysis: Change in Avocado Export Value Share and Homicides
We assess the impact on criminal violence of changes in the export market for avocados, an increasingly important export for Mexico. Our argument is that OCGs observe local changes in opportunity for economic capture and territorial control resulting from changes in a country’s position in the global market for exports. Our subnational test addresses this proposed mechanism by testing how increases in a commodity’s export value affect OCGs’ use of violence to seize economically strategic territory and capture the opportunity for market expansion. We then use a cross-national approach to look at the relationship between increases in countries’ value share of agricultural exports and criminal violence and assess the plausibility of our argument beyond Mexico.
Our data range from 2005 and 2010, with 11,440 municipality-year observations in the full model. Because our dependent variable is the number of homicides, we use OLS regression with standard errors clustered by municipality.Footnote 72 The dependent variable is led one year relative to all independent variables, with the exception of a dummy indicator for municipal elections in the current year.
Our independent variables of interest are the change in a municipality’s share of avocados’ export value, taken from the Servicio de Información Agroalimentaria y Pesquera, and the presence of OCGs, taken from Coscia and Rios.Footnote 73 We use increases in municipality-level export share as a local indicator of territorial and economic attractiveness to OCGs seeking short-term profit and long-term manipulation of an expanding portion of avocado exports. As we saw earlier in the case of Michoacán, OCGs violently compete to expand into up-and-coming growing areas. For each municipality-year, we calculate the change in that municipality’s share of the export value of avocados from Mexico:
$${\Delta _{{\rm{evsA}}}} = {{{\rm{municipality}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_t}} \over {{\rm{total}}\;{\rm{Mexican}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_t}}}\;\; - {{{\rm{municipality}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_{t - 1}}} \over {{\rm{total}}\;{\rm{Mexican}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_{t - 1}}}}$$
Larger, positive values of
${{\rm{\Delta }}_{{\rm{evsA}}}}$
indicate that a municipality’s importance as an exporter of avocados has increased substantially in the past year. Because avocados are grown in many Mexican municipalities and OCGs compete over control of these areas (see Figure 2), we can expect organizations to respond to new opportunities to seize profit and territory. When an avocado-producing municipality accounts for an increasing portion of avocado exports, it should be a lucrative prize that helps an OCG edge out competitors in the future. OCGs’ use of violence to capture avocado-growing territory and export profits depends on the existing competition and other municipal characteristics.
We capture the degree to which each municipality is threatened by OCGs with data from Coscia and Rios, who use web content to identify areas of operation of Mexican drug-trafficking organizations between 1990 and 2010.Footnote 74 However, our additional independent variables—including avocado export information—are recorded only from 2004 on, which limits our time frame to 2005–2010. To assess the extent of criminal threat to a municipality, we use the number of OCGs present in each year. We expect more violence in municipalities where OCGs are present, as they compete with the state, other producers, and one another for market access and territorial control; we expect no such increase in municipalities without OCGs.Footnote 75
For our dependent variable, we use homicides at the municipal level from the National Institute of Statistics and Geography (INEGI).Footnote 76 Here and in our cross-national case, we recognize that unattributed homicides is an imperfect measure of criminal violence. However, if increases in homicides following changes in the avocado market are driven by factors other than criminal violence, we should not observe a difference in criminally threatened municipalities versus municipalities without OCGs. The raw count of homicides likely undercounts strategic criminal violence, making it a difficult test of our argument. We also control for population size, which can impact the baseline level of violence.Footnote 77 In the appendix we also include models using homicide rate: homicide count divided by municipal population. We expect local state capacity to influence competition from the state or other producers. To account for this, we control for income and the number of prosecutors at the municipal level, following Trejo and Ley.Footnote 78 Because research has shown that OCG violence is linked to elections at the local level,Footnote 79 we include dummy variables for local election years.Footnote 80
Table 3 presents the results of our subnational analysis. It supports our Mexico-specific hypothesis. Increases in a municipality’s share of the export value of avocados are associated with increases in homicides, but only in areas where OCGs are present. Elsewhere, these increases in value have a dampening effect on homicides. Figure 3 shows the marginal effect of increases in avocado export share over the range of the criminal-threat variable. This effect is positive where OCGs are present.Footnote 81 Figure 4, produced using model 2 in Table 3, illustrates predicted homicides as the share of avocado export value increases in municipalities with no criminal threat versus municipalities with many OCGs.Footnote 82 Where OCGs are present—and likely to observe opportunities to boost their profit and influence—increases in the municipality’s share of avocado export value are associated with increases in the number of homicides. This effect, however, is not observed in municipalities with no criminal threat. Municipalities that are expanding their avocado exports and experience the maximum threat from OCGs are predicted to experience over fifty homicides per year—a sobering but substantively significant effect, especially considering that the median number of homicides per municipality-year is 0 and the mean is 9.Footnote 83
Change in avocado export value share, criminal threat, and homicides
Note: +
$p \lt .10$
; *
$p \lt .05$
; **
$p \lt .01$
; ***
$p \lt .001$
.
Marginal effect of change in avocado export share (from model 2 in Table 3)
Predicted homicides per municipality by change in avocado export share (from model 2 in Table 3)
Avocado Toast and Endogeneity
For a causal interpretation of our results, endogeneity between OCG presence, production of avocados, and their international price is a concern. It may be that, rather than seizing a newly lucrative market, OCGs seize the market first, and then manipulate supply and price. In the case of Mexico, this would mean that OCGs first manipulated the production of avocados to influence their export value in global markets.
To address this concern, we sought an alternate measure of export value share (capturing the international demand for avocados) that is also exogenous to OCG manipulation. The second condition eliminates international avocado price, which varies with both demand for avocados (plausibly exogenous) and supply (manipulable by OCGs). Instead, we use the popularity of the web search term “avocado toast” according to Google Trends.
Consumer behavior drives the demand for avocados in international markets; the popularity of avocado toast illustrates this behavior. Avocado toast gradually became a staple at trendy cafes in the mid-to-late 2000s, taking hold in the United States—Mexico’s main avocado importer—in 2006 and 2007. Figure 5 demonstrates the divergence in homicides by municipality type in Mexico following this shift in demand.Footnote 84 Google Trends captures “interest in a particular topic from around the globe.”Footnote 85 Searches for “avocado toast” reflect the world’s obsession with a new way of eating avocados, which corresponds to overall trends in consumption.Footnote 86 In the United States, per capita consumption of avocados more than doubled from 2010 to 2020.Footnote 87 The Trends data scale the relative popularity of a search term over time from 0 to 100, providing information on global demand for avocados that is plausibly exogenous to other factors driving changes in criminal homicides. Although the overall popularity of avocados and avocado toast increases over this period, there are fluctuations in the popularity of the search.Footnote 88 To mimic our previous sections using changes in agricultural or avocado export value, we assess how increases in “avocado toast” search popularity affect criminal violence in municipalities that export avocados as compared to elsewhere.Footnote 89
Avocado production, criminal threat, and homicides over time
Because the popularity of this search term reflects the international demand for avocados, the measure is not municipality-specific. However, changes in international demand should have more of an effect on criminal violence in avocado-exporting municipalities than they do elsewhere. The number of OCGs, our other independent variable, varies by municipality but does little to ease concerns about endogeneity. The factors driving changes in avocado exports also provoke changes in criminal presence and competition. To avoid this complication, we use an alternate measure of criminal threat: the proportion of years prior to our observation when OCGs were present in the municipality. This modification allows us to avoid post-treatment bias from changes in the attractiveness of a municipality when assessing the impact of increases in international demand for avocados in municipalities that are prone to criminal activity.Footnote 90
Table 4 and Figure 6 provide support for our causal explanation: increases in global demand for avocados provided an opportunity for violent criminal capture of a newly lucrative business. An increase in the popularity of “avocado toast” corresponds to about seventy more homicides in avocado-exporting municipalities that face the highest level of criminal threat. Note that the median criminal threat is zero. A smaller increase in homicides is also predicted in municipalities that do not export avocados. However, in Figure 6(b), the marginal effect on homicides of an increase in search term popularity is statistically indistinct from zero across many of the values of criminal threat. Increases in international demand for avocados have a positive and substantively meaningful effect on homicides in avocado-exporting municipalities with consistent criminal presence, but may also affect criminal violence elsewhere as groups compete for territory and future profit.
Homicides, “avocado toast” search popularity, and criminal threat
Note: +
$p \lt .10$
; *
$p \lt .05$
; **
$p \lt .01$
; ***
$p \lt .001$
.
Marginal effect of increases in “avocado toast” search popularity
Change in Price Versus Change in Export Value Share: The Cases of Corn, Strawberries, and Limes
Global or local increases in the price of a commodity, all else equal, should make it more attractive for OCGs to invest in that particular commodity.Footnote 91 However, we argue that increases in the value of commodities that also provide an opportunity for OCGs to eventually control a substantial portion of the market uniquely provoke more competition and violence. To probe this difference, we use placebo tests,Footnote 92 replacing our treatment of changes in avocado export value share with changes in the value of other key agricultural commodities in Mexico.
The first, corn, had a significant increase in global price, but Mexican exports constitute a small portion of the global market. Between November 2005 and June 2008, the global price of corn tripled, from USD 95.98 to USD 287.13 per metric ton.Footnote 93 Brazil and the United States are the primary exporters of corn, capturing a substantial portion—roughly 50 percent—of the global market. Mexico is not a top exporter of corn, meaning that changes in global price do not present an opportunity for OCGs to become price makers rather than price takers. If our argument about the importance of export value share holds true, we should not observe any relationship between changes in corn production value and criminal violence. The second commodity, strawberries, is a significant export for MexicoFootnote 94 but did not experience a bump in global price during this period that would have made it more attractive for capture.Footnote 95
We also include limes to explore an additional characteristic of agricultural goods. Corn and strawberries grow quickly—their time from planting to profit is minimal, meaning that they could be quickly planted in new territory to capitalize on global market shifts. But limes, like avocados, require two to three years of growing time before their fruit can be harvested and sold, making the territory in which they grow particularly worthy of competition. Although Mexico is the main global exporter of limes,Footnote 96 their price did not change significantly during the time of our analysis.Footnote 97 Comparing these commodities to avocados allows us to vary key characteristics—international market share, significant global price increases, and competition over specific growing territory—to provide evidence for our particular mechanism of export market expansion.
As seen in Table 5, even during the years of significant increase in global price, there was no increase in violence where criminal groups were active and these alternative crops are grown.Footnote 98 Although imperfect, this test provides additional evidence that OCGs engage in territory-claiming violence for a specific purpose: eventual market capture and manipulation of newly lucrative exported agricultural commodities. Assessed separately, price increases (corn), a large export market share (strawberries or limes), and territory-boundedness (limes) have no effect on increases in OCG homicides in growing territories.
Corn, strawberry, and lime production, criminal threat, and homicides in non-avocado-growing municipalities
Note:+
$p \lt .10$
; *
$p \lt .05$
; **
$p \lt .01$
; ***
$p \lt .001$
.
External Validity: Changes in Export Value Share and Homicides at the Cross-National Level
In the previous section, we provided evidence suggestive of our theory’s plausibility. In this section, we extend our analysis beyond Mexico with the aim of corroborating our argument in a general setting. We hypothesize that large, unexpected increases in a state’s share of a commodity’s export value—how much profit OCGs could expect in the future from capturing a substantial portion of the good’s international market—are associated with increases in criminal violence. To test our hypothesis at the cross-national level, we compiled data on the value of commodities, OCGs’ threat to the state, and the number of homicides per year. OCGs seize opportunities for diversification into booming licit markets for agricultural goods. To test the validity and generalizability of our claims, we analyze the relationship between changes in a country’s share of a commodity’s export value and changes in homicides at the country-year level. We use a global sample of countries between 1993 and 2018.Footnote 99 We include this test as descriptive, cross-national evidence of the relationship between changes in agricultural export value share and criminal violence.
At the general level, we assess one specific hypothesis:
H1general Unexpected increases in a state’s share of an agricultural commodity’s export value are associated with an increase in OCG violence in that state.
To estimate whether meaningful changes in commodity markets are associated with increases in OCG violence we first address some empirical challenges. The areas where these commodities are present may be very different from those where they are not. If these differences are correlated with our dependent variable they can induce bias in our analyses. To ameliorate these endogeneity issues, following Herrera and Martínez we assume that global demand shifts provoking higher global price, which drive changes in exporting countries’ global share of markets, are exogenous to domestic violence levels.Footnote 100 We then analyze how unexpected changes in agricultural commodities’ global markets affect violence within and across countries. We lead our dependent variable by one year relative to all predictors, and cluster our standard errors at the country level to account for idiosyncratic variation across countries and over time.Footnote 101
Another limitation is that we lack information on OCGs’ location relative to each product within each state. We also do not have ex ante knowledge about which products will experience significant shifts in global demand, and coding each prominent commodity for each country lies outside of the scope of this project. Our cross-national results should be viewed as evidence of a positive relationship between changes in a country’s share of products’ export value and homicides in the following year, which should be present in only countries with an active criminal threat. Building on our extensive subnational tests in Mexico, this global test is a proof of concept that opportunities for expansion into export markets for licit goods can provoke increases in homicides where OCGs threaten the state.
Combined, our sources yield a data frame of 47,652 observations covering a global sample of countries from 1993 to 2018. Our dependent variable—the count of homicides—takes on a wide range of positive values, making standard OLS, with standard errors clustered by country, an appropriate modeling strategy. We discuss data sources, concept measurement, and confounders next.
Dependent Variable
We operationalize our dependent variable—OCG violence—as the count of homicides per year by country, and we measure it using UNODC’s yearly homicide data.Footnote 102 A notable limitation is that these data capture all homicides, rather than only those attributable to OCGs. While the UNODC also collects data on homicides attributed to OCGs, there are concerning levels of bias and extensive missingness. When data are available, they represent a small fraction of the broader homicide data, missing or undercounting observations from countries where criminal presence is rampant.
Still, using counts of homicides offers an important advantage: it captures the multiple manifestations of OCG violence. Homicides attributable to OCGs result from confrontations between groups, with the state, or producers, or from the death of bystanders. OCGs may operate without homicides, but when homicides occur on a large scale, they result from an alteration in the status quo. This is important for us, as we seek to capture the impact on domestic violence levels of unexpected changes in international markets. Previous work, such as that of Trejo, Albarracin, and Tiscornia, Yashar, and Yoo, has used general homicide data as an indicator of OCG activity.Footnote 103 We follow these scholars’ lead in our use of this measure.
To validate our choice we assess the correlation between robberies resulting in homicides and changes in export value share for a limited sample of countries with available data. Homicides resulting from robberies are different from OCG homicides and unlikely to follow the same patterns. If more “random” violence, such as robberies, is associated with changes in our independent variable, the effect we observe may be the result of alternative pathways. As expected, robbery-driven homicides and changes in export value share are not significantly correlated in our sample. Furthermore, in Colombia, where homicide rates are high and OCGs pose a significant threat to the state, the association between robbery-driven homicides and increases in export value share is negative, albeit not statistically significant. Although this is one case and one alternative driver of homicides, it provides more confidence in our assertion that higher homicide numbers following increases in the share of agricultural products’ export value are driven by OCGs rather than other sources of violence.
Our data may undercount homicides from OCGs due to difficulty in observing clandestine violence or the risk of retribution from reporting, but under-reporting of homicides should bias against our findings. Finally, homicide data offer an advantage for cross-national analyses over time, as they provide comparable information for many countries across many years.
Independent Variables
In our theory, our primary independent variable must reflect changes in OCGs’ perceived opportunity to profit from newly lucrative agricultural products. We make use of annual International Trade Data from the Atlas of Economic Complexity,Footnote
104
which tracks countries’ export value for individual goods, export diversity, and trade sophistication, based on the United Nations Statistical Division (COMTRADE) Standard International Trade Classification. We use the export value of a set of goods: nonprocessed food products, raw paper materials, and animal- or vegetable-based oils.Footnote
105
We make several adjustments to capture moments of lucrative economic opportunity that OCGs may seize. Our primary measure mimics the
${{\rm{\Delta }}_{{\rm{evsA}}}}$
used to capture changes in the municipal share of avocado export value. For each country-year-product, we calculate a measure of the change in the share of the export value of each good in those categories:
$${\Delta _{{\rm{evs}}}} = {{{\rm{country}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_t}} \over {{\rm{global}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_t}}}\;\; - {{{\rm{country}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_{t - 1}}} \over {{\rm{global}}\;{\rm{export}}\;{\rm{valu}}{{\rm{e}}_{t - 1}}}}$$
Larger, positive values of
${{\rm{\Delta }}_{{\rm{evs}}}}$
indicate that the country’s portion of the growth in export value of a particular agricultural good has increased substantially in the past year. They represent windows of opportunity for OCGs to enter these newly booming markets to gain immediate profit and to employ strategies such as those outlined in the Michoacán case to engage in long-term capture.Footnote
106
We expect OCGs operating in or near territory that produces these commodities to violently seize control of production and exports as a means of increasing their income and local clout.
Our proposed mechanism centers on the presence of OCGs in countries profiting from increased agricultural export wealth. Increases in global export share of agricultural goods in countries without criminal presence should not be associated with an uptick in homicides. To distinguish countries under criminal threat from those where criminal presence is not strong, we rely on data from the Varieties of Democracy (V-Dem) project.Footnote 107 We use the variable coding “anti-system” movements that are “heavily engaged in criminal activity, e.g., narcotics, bootlegging, illegal exploitation of natural resources, extortion, kidnapping.”Footnote 108 The variable is averaged over coders’ decisions of no criminal threat (0) versus the threat to the state is criminal in nature (1). Values close to 0 indicate that it is unlikely that OCGs are a significant threat to the state, while values closer to 1 indicate consensus that OCGs are present and active.
Because in most states OCGs do not represent a full “anti-system” threat to the central state, the mean value of this measure is close to zero (0.07). States where OCGs present a significant challenge to governance score higher: Mexico averages 0.38, Honduras 0.36, Colombia 0.72, and the Philippines 0.54. We take this index as a conservative measure of criminal presence and threat because countries where OCGs operate but do not significantly challenge the state are likely to be underscored (for example, South Africa averages only 0.04). However, underscoring criminal threats would bias against any statistically significant findings rather than amplify the relationship between criminal threat, changes in export markets for licit goods, and homicides.Footnote 109
Our argument is that OCGs engage in violence when expanding their territorial and economic presence in markets for licit goods. Thus, we expect increases in homicides in countries with a clear criminal threat following growth in the country’s share of a commodity’s export value. In contrast, countries without active OCGs should not experience a significant uptick in violence. Given this expectation, we interact the criminal presence variable with the change in products’ export value share.
Confounders
Other country-level factors may influence changes in the number of yearly homicides in states. We account for these confounders by including GDP per capita to account for connections between country wealth and violence; population, given that more populous states experience more homicides; and proportion of the labor force employed in agriculture as an indicator of the importance of agricultural commodity production for a state’s welfare. These data are from the World Bank’s World Development Indicators.Footnote 110
We expect that the importance of a given commodity to a state’s economic well-being impacts whether OCGs are able to compete economically and territorially for its control. Export diversification promotes economic growth,Footnote 111 and diversified export portfolios lead to stronger states, which are better able to provide access to more employment and economic opportunities, disincentivizing crime. Stronger, more capable states might also be more effective at crime detection and deterrence. Conversely, when countries are dependent on a few commodities, states may guard market entry for key products more closely, making it more difficult for OCGs to take advantage of demand shocks for goods that are central to the economy. Therefore, we should expect stronger competition, and more violence, in countries with minimal export diversity. We include the Economic Complexity Index from the Atlas of Economic Complexity as an indicator of a country’s export diversity.Footnote 112
Whether a state is embroiled in violent conflict influences overall levels of violence, so we include a dummy variable taking a value of 1 if a country experiences civil war in a given year, and 0 otherwise.Footnote 113 Finally, a state’s ability and willingness to enforce the law influence both OCGs’ ambitions and the homicide rate. We include two measures from V-Dem to capture the extent of law enforcement and accountability of public officials: an index of the rule of law, and an indicator of how likely public officials are to engage in corruption.
Results
We find strong evidence for a positive relationship between changes in export value share and homicides in criminally threatened states. When OCGs threaten a state, increases in that state’s share of agricultural goods’ export values are associated with a significant increase in homicides (Table 6). Model 1 provides basic evidence of this relationship; model 2 includes the confounders discussed in the previous section. Larger increases in a country’s export value share of agricultural goods correspond to much higher counts of homicides in countries where OCGs are active in the following year.
Change in export value share, criminal threat, and homicides
Note: +
$p \lt .10$
; *
$p \lt .05$
; **
$p \lt .01$
; ***
$p \lt .001$
.
Figure 8 illustrates the predicted value of homicides over the range of
${{\rm{\Delta }}_{{\rm{evs}}}}$
produced using the complete model in Table 6.Footnote
114
Sharp increases in homicides result from large, positive changes in products’ export value share in countries where OCGs threaten the state. Two issues are relevant for interpretation. First, the median homicide count is just over 300, while the maximum is 65,000. In countries experiencing even a moderate criminal threat, a 15 percent change in a country’s export value share of key products is associated with a predicted 10,000 homicides in the following year (Figure 8). When there is no change to export value share, fewer than 3,000 homicides are predicted. Further, in countries where the threat from OCGs is high, the predicted increase in homicides after a 15 percent change in an agricultural product’s export value share in the previous year is stark: over 40,000 at the maximum level of threat. The states falling above the “mean threat” level are prime examples of OCG hubs—El Salvador, Honduras, Colombia (Figure 7). Thus, in environments where OCGs are a salient threat, a larger increase in these states’ export value share of agricultural products is associated with substantial increases in homicides.
“Avocado toast” searches and predicted homicide count
Predicted homicides, export value change, and criminal threat (from model 2 in Table 6)
These results support the plausibility of our theoretical argument in a cross-national setting.
Conclusion
Research has focused on understanding diverse features of OCGs and violence in connection with markets for illicit goods. We shed light on an understudied phenomenon: the conditions under which OCGs violently target markets for licit commodities. We develop and test a theory of criminal diversification into markets for licit goods, focusing on agricultural commodities. We show that when there are increases in the local share of a good’s export value, OCGs seek market control, resulting in more violence. This seems to be particularly true for territory-bound goods with a long growing horizon, such as avocados and limes. We provide multiple sources of evidence, including quantitative analysis of Mexican avocado exports, qualitative evidence of market capture of avocados, and a proof-of-concept descriptive test of the relationship between export value share and criminal homicides at the cross-national level. In all analyses, the evidence supports our argument. Cross-national, cross-temporal data provide generalizable evidence of our proposed theoretical relationship. The case of Mexico provides direct evidence of the presence of OCGs and their attempts to control markets beyond illicit ones. A placebo test with other commodities—corn, strawberries, and limes—in Mexican municipalities helps isolate and provide evidence of our mechanism. Finally, measuring international demand with the rate of web searches for “avocado toast” helps mitigate concerns about reverse causality.
We show that violence is not limited to the underworld of illicit goods. Scholars have demonstrated that criminal violence can be as deadly as civil war violence. However, these analyses focus on illicit markets. By providing evidence of the conditions for violent takeovers of licit markets we extend our current understanding of OCG behavior and its consequences. We propose a specific mechanism—increases in price where significant export share enables eventual market capture—and argue that it is an important measure of groups’ motivation to seize markets for immediate profit as well as eventual control and manipulation.
We generate a systematic explanation of OCG behavior in a cross-national setting, currently an underdeveloped area of research. Taking advantage of an area where there is access to cross-national, cross-temporal data to test our argument, we rely on knowledge from case studies as rich sources of theory development and use it to generalize beyond individual cases.
Our project sparks questions about the nature of diversification and highlights the need for systematic data collection. There is little information on OCGs’ revenue streams. Criminal markets are “global chains of local operations,”Footnote 115 and revenue flows at the top may not reflect the bottom of the chain. Besides, not all OCGs are powerful cartels; diversification can help OCGs better distribute cash among members at the local level. Because diversification is not necessarily about substitution but about long-term investment opportunities,Footnote 116 research using fine-grained data on group features is necessary to better understand what types of groups can diversify.
Finally, we contribute relevant insights for future research on criminal violence and governance. Dominant accounts describe organized crime as primarily urban and drug related. We show that this understanding might be biased. OCGs deal in a wide variety of products beyond drugs, and the capture of natural resources implies that these dynamics extend to rural areas. OCGs compete to regulate licit and illicit markets, but the nature and degree of violence depend on the market and the point in the chain of production and distribution. Controlling agricultural production may require a different level of territorial control and violence than certain illicit drugs. Variation in economic behavior has implications for criminal governance; whether the same mechanisms apply in rural and urban settings and across market segments is a question for future research.
Acknowledgements
We are grateful to the following readers and groups for their comments on drafts and presentations: Javier Osorio, Cecilia Farfán-Méndez, Sandra Ley, Jeff Harden, Michael Weintraub, Brown University Watson Institute for International and Public Affairs Security Studies Seminar, the Carolinas Conflict Consortium, the Vanderbilt Political Economy of Conflict Workshop, and UNU WIDER Development Conference (UNIANDES).
Data Availability Statement
Replication files for this article may be found at <https://doi.org/10.7910/DVN/HNA4FP>.
Supplementary Material
Supplementary material for this article is available at <https://doi.org/10.1017/S0020818325100763>.
Open access
