Cutting the Head Off the Snake - An Analysis of Kingpin Strategy in Combating Drug Trafficking Organisations

By: Jurgis Augunas

The Kingpin Strategy has been a widely used approach to fighting organised crime, particularly in the Mexican Drug War. It is a strategy used to weaken and disrupt Drug Trafficking Organisations (DTOs) by targeting the highest-ranking figures in the chain of command. Anticipating disruptions in the organizational structure caused by the capture or assassination of the leaders of a DTO, the strategy is backed by the belief that the drug trafficking hierarchy cannot be easily replaced. Advocates of this theory proclaim that by provoking disorganisation in the structure, levels of criminal activity are reduced. Critics, however, argue, that such disorder may increase violence, as individuals and rival gangs compete to succeed the eliminated leader (Lindo & Padilla-Romo, 2018).

“Kingpin Approaches to Fighting Crime and Community Violence: Evidence from Mexico’s Drug War” by Lindo & Padilla-Romo (2018) aims to investigate the relationship of the kingpin strategy on community violence in the Mexican drug war. Most of the existing literature of drug-related interventions on drug abuse focuses on stages closer to final consumers (“downstream markets”). This paper, however, aims to analyse the causal effects of the interventions in areas where the associated costs are most relevant – communities where production and distribution take place (“upstream communities”).

Working closely with Dell (2015), who demonstrates that drug-trade crackdowns positively affect drug-trade-related homicides in both the communities where the crackdown took place and where drug trafficking is likely to be diverted, this paper contributes to the existing literature by focusing explicitly on the effects of the kingpin strategy on homicide rates. Successfully carrying out such an analysis could prove to be a helpful tool for governments and law-enforcement agencies considering future kingpin captures as a drug trafficking combatting policy. Thus, this essay critically assesses the methodology used by Lindo and Padilla-Romo (2018), while considering the results and limitations that arise when analysing the causal link between such variables.

Two main empirical challenges arise when attempting to analyse such a causal effect. Firstly, the captures of kingpins are rare. Although the war on drugs lasted for almost a decade, only five main DTOs and their respected kingpin captures are considered, thus the actual events are few in numbers. Secondly, policies implemented to address organised crime typically involve a variety of different elements - the use of multiple strategies, of varying degrees of intensity, across different periods in time. Considering both challenges and seeking to understand the effect of kingpin captures, each event (major capture) must be thoroughly analysed.

While data on drug-related homicides is available from December 2006 to October 2011, the authors of this paper choose not to rely on it out of concern for the endogeneity of homicides being classified as “drug-related” or “not drug related”. Such an approach is used to mitigate errors associated with the way violence might be classified, which can be influenced by different institutions and their incentives to change public perception regarding the success of the war on drugs. Hence, this paper uses data gathered from multiple sources spanning from January 2001 through December 2010.

For the outcome variable - homicide rate, data on the number of deaths at the municipal level is obtained from the National Institute of Statistics and Geography (INEGI). To calculate per capita rates, municipality population estimates are sourced from the National Council of Population (CONAPO) and El Colegio de Mexico (COLMEX). Additionally, information on DTO presence at the municipality level is collected using the MOGO framework, which identifies DTO areas of operation between 1990 and 2010 from reliable web sources. Lastly, data on kingpin captures are extracted from Navy (SEMAR), Army (SEDENA), and Office of the Attorney General (PGR) press releases, focusing on the timing and location of captures.

The authors choose to consider only captures of a leader or lieutenant of a DTO, where the former is positioned at the very highest level of the organization and the latter is next in rank. The decision not to consider lower-level kingpin captures is based on the belief that such captures might not be completely endogenous to those of higher rank, potentially violating the independence assumption. Some constraints on the data associated with DTOs must also be acknowledged. Firstly, it is highly unlikely that the presence of a DTO in a selected municipality is captured efficiently. Data on organised criminal activities is difficult to capture and cannot be precisely measured as it is often misleading and under-reported (Blattman et al., 2014). Secondly, the exact intensity of the DTOs' presence in a municipality is not captured. Such a drawback can result in measurement invariance and heterogeneity between municipality features and cause the variable of interest to be prone to measurement errors (Lindo & Padilla-Romo, 2018).

The approach that the authors employ for this analysis is the generalized difference-indifference (DiD) method. The DiD technique is not a perfect replacement for randomized experiments (RCT), but it often represents an alternative method to analysing casual relationships (Wing et al., 2018). It involves comparing two groups – the control and treatment across two time periods – before and after a policy or event takes place. While one of the groups (treatment group) is exposed to the treatment, the other group (control group) is not and acts as the counterfactual to the treatment group. The difference in outcomes between the two groups is considered as the causal effect of the event.

In this study, four municipalities act as the treatment group – the municipality where the capture occurred, a neighbouring municipality where captured kingpin DTO had presence in, a non-neighbouring municipality where captured kingpin DTO did not have presence in and a neighbouring municipality where captured kingpin DTO did not have presence in. The estimated effects (𝛿) of the kingpin capture are evaluated by comparing the differences in homicide rates over time across the “treated” municipalities with the changes observed over time in municipalities that are not linked to any DTO and municipalities which are linked to a DTO but are yet to experience a kingpin capture. These two groups act as the comparison groups and approximate the counterfactual. Such an approach employed by the authors successfully mitigates potential biases that might arise due to fixed differences between municipalities and by the effects brought upon by external policies or unexpected shocks which occur at the municipal level.

While the DiD design is one of the most used methods in empirical research, it does have limitations (Roth et al., 2023). According to Wing et al. (2018), the most challenging aspect is assessing the validity of the parallel trend assumption, which states that if no treatment was given to either of the groups, the difference between the outcome of the groups would be constant over time. As this cannot be statistically observed, the analysis might be prone to invalid causal inferences or biased estimates. However, this assumption can be partially validated with graphical evidence by comparing trends in the outcome variable prior to the treatment across the groups. (Wing et al., 2018)

To validate the parallel trend assumption, the authors analyse the timing of kingpin captures for various DTOs and consider the four municipalities of interest. In Figure 5 each graph shows the average difference in homicide rates between the municipality of interest and its comparison group – non-neighbouring municipalities in the same state with no DTO presence. It is evident that there is a constant difference in the homicide rates of the analysed municipalities and their comparison group leading up to the kingpin capture. This provides support that the parallel trend assumption holds and that the authors have identified a suitable control group that follows a similar trend to the treatment group prior to the capture taking place.

To confirm that the difference-in-difference method is the correct approach to analysing the kingpin capture effect on community violence, the authors also note that all the captures take place at least a year after the war on drugs started. This addresses the potential problem that any war-related activities and the outcomes they might have produced at a broader level would have influenced all municipalities of interest and their comparison groups.

Fixed effect control variables are included in the model to account for time-invariant characteristics that might bias the estimates. Municipality fixed effects (𝛼𝑚) ensure that the Word count: 2191 ECO3008 6 estimated effects are driven by within municipality variations, while month-by-year fixed effects (𝛾𝑡 ) control for time-specific factors that remain constant within each year and month but may vary across different periods. Additionally, the model considers different time intervals post-capture (0-5 months, 6-11 months, and 12+ months) to capture the short, medium, and long-term effects.

The authors reinforce their main findings with additional fixed effects and controls in Table 2. Column 2 includes state-by-year-by-month fixed effects for spatial heterogeneity, column 3 controls for war effects specific to municipalities with DTO presence, and columns 4-5 explore variation in homicide rates from expected levels. Statistically insignificant estimates further support the causal effect of the analysis. The paper finds that kingpin captures significantly increase community violence in municipalities where the capture occurred, estimating a 61% increase in homicide rates. Smaller but significant increases in community violence are also seen in non-neighbouring municipalities where the DTO has presence in. Interestingly, there is little-to-no effect in the short run on the variable of homicide rates. However, there is a significant increase of 12% in the homicide rate 12+ months following the capture, indicating that the kingpin strategy may have a destabilising effect throughout the organisation over a longer period.

The estimates also show decreasing rates of homicides in neighbouring municipalities in which the kingpin does not have presence in 0-5 months after the capture and municipalities where kingpin does have presence in 12+ months post-capture. In addition, the authors find that kingpin captures led to an additional 4880 homicides between 2007 and 2010, accounting for 7.1% of the murders throughout the period. Around 30% are due to spill-over effects to non-neighbouring municipalities where captured kingpin DTO had presence in.

The authors of the paper are particularly successful at providing a robustness check to support the causal interpretation of their main findings. This ensures that the conclusions are not driven by specific modelling choices.

The analysis of kingpin captures during the war on drugs faces a potential issue where simultaneous military operations could intertwine with the effects of the captures. However, this concern is effectively addressed through evidence presented in Figure 6. The figure demonstrates that military operations related to the war on drugs do not increase homicides in municipalities with a DTO presence compared to those without. In each of the eight states with a military operation, the municipality homicide rates, with and without DTO presence, are plotted. The graphical panels show no significant variation in municipality homicide rates postmilitary operation, supporting the conclusion that military operations do not conflate the effects of kingpin captures.

This paper extends the analysis to demographic characteristics linked with homicides from kingpin captures, as presented in Table 3. Results indicate that males aged 15-44 are most affected, aligning with findings that this demographic group is more likely to engage in drug trafficking (Fairlie, 2002; Vilalta and Martinez, 2011).

While overall birth rates show no statistically significant impact from kingpin captures, specific subgroups of women are affected, as indicated in Table 5. Estimates reveal a decline in the number of infants born to married mothers in both the municipality of the capture and non-neighbouring municipalities without DTO presence. The statistically significant effects on pregnancy-related outcomes emerge 0-8 months post-capture, suggesting that the decision to become pregnant was unlikely influenced by the capture. This may instead reflect increased migration among married women in these regions or a rise in the number of abortions.

Examining a broader health impact, particularly on mental health, the study analyses the effect of kingpin captures on suicide rates. However, the estimates are statistically insignificant, suggesting no influence on suicide rates. It's important to note the challenge in measuring effects on mental health, and while these findings don't necessarily imply no effect, they specifically indicate that such an extreme outcome as suicide remains unaffected.

In summary, this paper successfully establishes a causal effect of kingpin captures on homicide rates during the Mexican Drug War. Although limitations associated with the data and the DTO presence in municipalities cannot be dealt with, the authors address threats to validity by providing alternative methods of proving causal inference. Overall, estimates demonstrate an immediate 61% increase in homicide rates in the captured municipality and a longer-term 12% increase in non-neighbouring municipalities without DTO presence. Even though further analysis and more reliable data on DTO presence at the municipal level should be considered, the findings of this analysis indicate that kingpin captures may be ineffective in decreasing community violence and future policymakers combatting drugtrafficking should take this into consideration.