The Dynamic Conflict Risk Model

The Dynamic Conflict Risk Model (DCRM) estimates the monthly risk of violent conflict in sub-national regions, with a forecasting window of six months. The model distinguishes between four conflict categories and currently covers Africa, the Middle East, Asia and Eastern Europe.

To view visualizations of the DCRM’s most recent results, click here. For a user manual of how to explore and navigate these visualizations, visit the Help section of the Science 4 Peace Portal.

Data and methodology

The DCRM calculates the likelihood of either the presence or absence of conflict in each sub-national administrative unit on a monthly basis. In line with the Global Conflict Risk Index (GCRI), the DCRM uses conflict definitions and data from the Uppsala Conflict Data Program (UCDP). It differentiates between the following conflict categories:

  • State-based conflict (SBC): Armed conflict between a state government and at least one organized group. Fighting either revolves around central government power or over territory. The use of force results in at least 25 battle-related deaths in a calendar year.
  • Non-state conflict (NSC): Armed conflict between two groups, neither of which is directly affiliated with a state government. This includes conflict between rebel groups and militias, fighting between supporters of different political parties as well as conflict between ethnic, religious or other social groups. Fighting results in at least 25 battle-related deaths in a calendar year.
  • One-sided violence (OSV): Direct and deliberate killing of civilians, carried out either by a state government or a non-state armed group, resulting in at least 25 deaths in a calendar year.
  • Any conflict (ANY): This category includes all instances of internal conflict that belong to the following categories: State-based conflict, Non-state conflict and One-sided violence.

For each type of conflict and world region, the DCRM uses a separate model with its specific input variables and parameters. While the Global Conflict Risk Index (GCRI) only covers conflicts with at least 25 battle-related deaths in a year, the DCRM’s monthly analysis can include conflicts that result in fewer fatalities. The model treats any incident with at least one fatality as an instance of violent conflict for the month and administrative unit where the incident took place.

In general, the DCRM uses two types of input variables to predict conflict:

  1. Event data, including past conflict events, protests, and related incidents in each administrative unit and neighboring units.
  2. Structural variables, which include local demographic and geographical data, as well as political and economic variables at the country level.

The Joint Research Centre is developing the DCRM in cooperation with the Conflict Prevention and Mediation Support Division within the EU External Action Service (EEAS) and with funding from the Service for Foreign Policy Instruments (FPI).

The DCRM is continuously evolving, with ongoing work to improve modelling techniques, refine the selection of input variables, and find ways to quantify and communicate uncertainty in the model's predictions. In the coming months, the JRC plans to extend the DCRM’s coverage to Latin America and the Caribbean and develop an additional model to predict monthly conflict fatalities.

DCRM documentation

There are no published reports on the Dynamic Conflict Risk Model yet, but upcoming publications will be made available here.