Rebel Groups Protection Rackets: Simulating the Effects of Economic Support on Civil War Violence

Rebel groups engage in a series of economic transactions with local populations during a civil war. These interactions resemble those of a protection racket, in which aspiring governing groups extort the local economic actors to fund their fighting activities. Seeking security in this unstable political environment, these economic actors may decide to pay the rebels in order to ensure their own protection or to flee, impacting the outcomes of the civil war. We propose an agent-based simulation model that represents the decision-making of these economic actors during protection racket interactions as well as the inter-rebel groups’ competition to control territory. Our model reveals insights about the mechanisms that are helpful for understanding violence outcomes in civil wars, and the conditions that may lead rebel groups to prevail or achieve a stalemate.

Group Members
Frances Duffy, University of North Carolina at Chapel Hill, United States
Kamil Klosek, Charles University, Czech Republic
Luis Gustavo Nardin, Brandenburg University of Technology, Germany
Gerd Wagner, Brandenburg University of Technology, Germany

Kamil Klosek
Luis Gustavo Nardin

Thanks for the presentation. This seems like a very challenging topic for modeling purposes. Since you motivate your presentation with some empirical examples, can you say something on how you plan to connect your model with empirical data? How will you evaluate the performance of your model? I also like to poke some comses members who have made models on similar topics to have a look at your talk (@salwaismail @crooks @jwkim ).

Thanks for your question, it is very relevant.

First I would like to clarify that despite motivating the presentation with empirical examples, we are developing a general model to capture the main entities and dynamics involved in extortion and violent activities in the context of civil wars. Thus, the model is supposed to be configurable and able to represent different civil war scenarios.

Because we intend to use the model to analyze the Somalian civil war, we will need to connect the model with empirical data. We plan to use empirical data for two purposes: (1) to calibrate the input parameters and (2) to evaluate the performance by fitting it with aggregate statistics of the model.

To calibrate the model, we will use some demographic and economic indicator to define the size of the economy and number of economic actors (i.e. Enterprises). Moreover, these indicators will allow us to develop a distribution function that will generate the Enterprises’ income during the simulation. These indicators will probably be extracted from datasets available at the World Bank [].

The calibration of the rebel groups will be more challenging as there are little or no information about them. Several of their behaviors will be based on theoretical studies in the fields of conflict and international relations instead of empirical data.

However, we will use data from Uppsala Conflict Data Program (UCDP) [] to characterize the different existing rebel groups in Somalia. We also found some studies describing the financial aspects of these groups that may help us to calibrate the percentage extorted from Enterprises. One example is the report

Keatinge, T. (2014). The Role of Finance in Defeating Al-Shabaab. Whitehall Report 2-14. London: Royal United Services Institute for Defence and Security Studies.

that estimates such percentage, which is characterized as corporate taxes, around 2.5% of the Enterprises’ profit.

Once we calibrate the input parameters, we will assess the performance of the model by fitting some aggregate statistics with empirical data. Our initial thought is to assess and validate the model based on the patterns of Enterprises fleeing the region and patterns of fights among rebel groups.

The empirical data about fleeing patterns will be extract from UNHCR (United Nations High Commissioner for Refugees) []; while the empirical data about fights pattern will be extracted from the ACLED (Armed Conflict Location and Event Dataset) []. ACLED provides a dataset with all the violent events that ocurred in Somalia from 1997 to 2018 specifying the date, type of violent event, actors involved, and the region where it took place.

We are not advanced in the calibration and validation process yet and we know it is very challenging, e.g., the lack of data. There is still much uncertainty whether our plan on how to use the empirical data to calibrate and validate the model will work. Hence, I would appreciate any feedback about previous experiences trying to perform these tasks in similar models.

Thank you,