Executive Summary

Can quantitative methods of futures studies really be used to help policy makers decide when troops need to be activated? Can they actually provide us with warning signals about civil war, violent unrest, or epidemics before they occur? Or do the tools produced through such methods belong in a fairy tale, like a magician’s crystal ball? During the two years of this project, we heard these questions asked frequently by policy makers, members of the Bundeswehr, and in the beginning, even by this research team.

What we discovered, however, is that these methods can indeed be used to provide foresight about security risks. The tools developed as part of this project do not provide the time of day that a terrorist attack will occur, or identify the person who will commit it. What they do provide is a plethora of early warning tools for policy makers to choose from which emit signals for short, middle and long-term risks relating to state failure, adverse regime change, and civil war.

Second, we verified that quantitative data is not a replacement for qualitative expertise and decision making, and that the combined use of different tools developed in this project can provide a more detailed mural of potential risks which could occur in the future. The tools do not dictate a solution to a security problem. Rather, they provide decision makers with the opportunity to see where in the world, or where, on a sub-state level, a new conflict may break out, whether expected, or unexpected.

This project broke new ground in more ways than we could have imagined. While other organizations, such as Washington’s Fund for Peace mentioned in the travel report at the end of this paper, have created indicator-based early warning systems, the system created for this project is more flexible and offers a greater number of functions than any that we observed or reviewed. The prototype described in this chapter has the potential to be used by any government ministry for 20 different topics, from environment to health, offers different extrapolation methods, and includes over 6,000 indicators.

For purposes of the Bundeswehr and the Defense Ministry, the integration of political models allows the user to see world-wide where state failure, adverse regime change, or civil war could occur in years to come. A country comparison graph and a country index ranking those at greatest risk to least risk allow decision makers to see at a glance the hotspots they may need to prepare for in the future.

The Topic Monitoring tool described in chapter two adds an up-to-the-minute early warning system to compliment the long-term risk Indicator-based Early Warning System. By showing where a sudden increase in activity related to a certain keyword has occurred across the world, the policy maker can be aware of potential risks as they are reported in 63 different languages in real time.

The quantitative risk analysis forecasting models described in chapter five analyze risk from a sub-state and regional level. Here, fairly specific six to twelve month forecasts of conflict outbreaks, including number of people likely to be killed (greater than 10 or greater than 100) and potential regions to be affected, allow the policy maker to see what factors are leading to conflict, and when and where to expect violence in the near to mid-term.

The Risk Management Process Prototype described in chapter six brings all of the tools together in a three-step process. After noting the short, mid and long term risks produced by the tools, the policy makers have the opportunity to prioritize those, which through a combination of quantitative and qualitative analysis, are deemed most likely to occur, or those which are to take top priority. They then have the opportunity to run a simulation which can help correct any deficiencies in foresight analysis produced to that point.

This can be done through running a simulation using one of the two tools developed for the fourth work package and described in chapters three (Agent Based Model) and four (System Dynamics). These models can help identify what other factors not included to that point may play a role in sparking a conflict, and what individuals, corporations, groups or networks may be involved. Based on such specific analyses, the decision makers can reevaluate their priorities or how they would like to respond to the potential future crisis. They may also decide to rerun the desired extrapolation on the Early Warning System or to rerun the search on the Topic Monitoring tool to allow for new factors, topics, countries, or regions they may not have included the first time.

In this way, decision makers are empowered to have the maximum amount of data on risks which could occur on both a macro and micro level in its foresight analysis. Through the interim qualitative screening step using the Risk Management Process prototype, the analysis does not occur in an automotive void, but has better chances of mirroring reality. Through the final simulation step, a more detailed analysis can be developed to allow for a maximum number of influencing factors on the conflict to be included.

1.1      Work Package Two: Indicator-Based Early Warning System

The indicator-based early warning system contains over 6,000 indicators and 20 topics for decision makers to choose from.  Through the integration of various political models into the tool, it provides an early warning signal (red for the highest value, and blue for the lowest) when a country is at greater risk for civil war, state failure, or adverse regime change.

Figure 1 Adverse Regime Change 2011: Mali and Ukraine are in the danger zone


Different extrapolation methods are included, to allow for extrapolation into the mid or long-term, up until the year 2030. Policy makers have the opportunity to use different extrapolation methods. This has the advantage that when a policy maker wants to allow for a sudden, unexpected change to be included into the equation, such as a terror attack, he can use a different extrapolation method than when he is calculating national defense budgets, which only change once yearly. The tool has had fairly accurate forecasting results, including forecasting the Ukraine crisis three years in advance, and an increase in terrorism and violence in the countries of the Arab Spring in 2011. Further projection examples, such as what countries may be at risk for civil war in 2018, and other projections, can be seen in the user manual at the end of this report.

Figure 2 Libya, Syria, Egypt and Tunisia show a strong, sudden decrease in the indicator value for “Absence of Political Violence and Terrorism” even before 2010

In the last year, additional data bases have been added, and the Internet-based Interface has been linked with the new Topic Monitoring Tool.

1.2      Work Package Three:  Topic Monitoring Tool

At the click of a mouse on the Indicator-based warning system interface, one can change from looking at long-term developments world-wide to sudden occurrences. While both display results on a map of the world with red representing the highest value, in the Topic Monitoring Tool, the highest value stands for recent sudden increase in texts containing the keyword used in a search. The tool provides short term early warning functions through its identification of sudden trend activity. At present, the prototype has the capability to search the Internet in 63 languages when one enters a term or combination of words. It can conduct simultaneous monitoring of several countries, and can perform the automated identification of activities, trends, and lower level keywords.

1.3      Work Package Four: Virtual African Societies Agent-Based Model and Bundeswehr Recruitment Model

Two models were created in work package four: a multi-agent system called Virtual African Societies and a System Dynamics model that simulates the impact of demographic change on the personnel recruitment of the Federal Armed Forces up to 2030 and beyond. The agent-based model was created to simulate when ethnic conflict breaks out. It focuses on Guinea and is parameterized with real-world GIS data. The model exhibits a high level of generalizability and could potentially be applied to other countries, in particular to countries in which ethnic group interrelations are likely to play a role in future conflict outbreaks.

The System Dynamics model shows that from the year 2040 on there might be a personnel shortage if the selection criteria stay the same. Increasing the percentage of women in the unit “sergeants and their teams”, from the current state of 12% to 20-33%, would change this scenario, according to the model. The research showed that internal factors, especially the attractiveness of the Federal Armed Forces and the recruitment profiles, must be adapted to the candidates’ detailed metrics. Improving areas such as job security, family friendliness and career opportunities within the Federal Armed Forces would help with the personnel shortage.

1.4      Work Package Five: Architecture Development Method for the Prototypes

Work package five documented the functional requirements for the web-based Interface created for this project. The analysis was based on the Indicator-based Early Warning System and the Topic Monitoring prototypes developed in work packages two and three. The documentation was carried out using the framework of the Architecture Development Method of the Bundeswehr creating the System View, Technical View and Operational View required.

1.5      Work Package Six: two Sub-State Risk Assessment Forecasting Models

Two risk assessment forecasting models for the regional and sub-state level were developed, with a focus on Africa and the Gulf of Guinea region. Indicators for a total of 768 regions were taken into account in the conceptual model. Future developments occurring six to twelve months in the future could be projected from data that was either structural – such as economic development, regime type, number of ethnic groups – or dynamic – such as conflict status, number of neighboring regions with armed conflict, or behavioral patterns of those agents involved in an armed conflict network.

Figure 3 Violence spill-over prediction for Mali

Two models were used: a logistic regression model and a Random Forest Classifier. The latter proved to have the stronger prediction power in this study, showing for example, that there would be an outbreak of conflict in the high risk region Kidal long before it actually occurred in January 2012. The model showed an increased risk for the neighboring Timbuktu and Gao regions through spill-over effects. Likewise, the model correctly predicted that Guinea’s Nzérékoré would be a high risk region for armed conflict before 70 people were killed there in July 2013.

1.6      Work Package Seven: Risk Management Process Modell

In the final work package, a risk management process model allows for the combined usage of all tools developed in this project so that a decision maker has a clearer picture of developments in the short, middle and long term. In this risk management process, the decision maker himself has the opportunity to rate the likelihood of a risk scenario, based not only on the results from the quantitative method tools used, such as the Early Warning System and the Topic Monitoring Tool, but also on qualitative data, such as country expertise gained from analysts. This information can then be used to launch a second quantitative layer: running an agent-based model simulation or a system dynamics analysis to see how specific risks, such as ethnic conflict, could play out on an agent, subnational, national, or regional level.

1.7      Conclusion

Through this combination of tools, the decision maker is empowered to not only have quantitative data and analysis of potential future risks on a micro and macro level, but is able to actively rate which may be most likely, and which should be prioritized for risk prevention.