Dynamics

One methodological novelty of the ViEWS project is our software for dynamic simulation, called Dynasim. It allows us to model future conflict in a more dynamic way than classical time series forecasting techniques.

The procedure starts by estimating regression models of conflict and other variables of interest. It then simulates many instances of these models by sampling from the distribution of parameter vectors of the models. Each sampling of model parameters sets the rules for how a simulated state of the world will develop. Each of these simulated states of the world are then forecasted, one month at a time, with the software updating the state of the world at each step before predicting the next.

At each month Dynasim updates a large set of transformations of variables such as geographic neighbourhood effects and time series. These transformations allow the user to specify potential interdependencies between simulated variables (such as economic development, democracy and conflict) and by doing so capturing phenomena such as conflict traps and diffusion of conflict in a geographic neighbourhood.

By running many simulations Dynasim captures our uncertainty about the "true" model of conflict by recording the variation between the simulated outcomes. A larger observed difference between simulated outcomes represents not only uncertainty about the model parameters but also about the dynamic effects between the models, which are very difficult to analytically solve.

Dynasim, like the rest of the software powering ViEWS, is open source and in continuous development.