This project will predict political risks for more than 190 countries on a monthly basis for the duration of two years. We will build on my academic work with Christopher Rauh to provide these predictions. We have updated the methodology and are now exploiting supervised machine learning methods for our forecast. In particular, we are using random forest classifiers to produce predictions.
We have already downloaded around 4 million newspaper articles which we will update on a monthly basis to produce conflict predictions a year ahead and a quarter ahead. As we argue in our most recent working paper this allows us to predict conflict for countries which would otherwise be "off the radar". We call this the hard problem of conflict prediction. This means that in this project we will predict very subtle, low-level risks for a set of countries which includes OECD countries. The movement in this risk can be regarded as general political risk as it includes a multitude of factors in economics, society and politics - seen through the prism of conflict prediction.
This project will also contribute to the debates around fragility which has produced a lot of interesting and useful initiatives which try to measure it. Our approach is to take the concept of fragility very seriously, i.e. treat it as something that is linked to future, not yet realized, failure. This sets us apart from most existing approaches which do not explicitly rely on a forecast but are often coding existing conditions. Most fragility indexes are publicized with a delay, our index will be published monthly and will therefore be able to update on recent events rapidly.
We believe that linking fragility to past data publications is like the evaluation of the fragility of a vase by throwing it on the floor: if it breaks it was fragile. If we want to prevent failures we need to adapt our measures of fragility to the task.
Results will be published shortly. For declarations of interest in our forecast please use the contact form.