Predicting Local Violence in Liberia Working Paper

It is interesting to note that we barely use forecasting models for conflict prevention programming at community level. Most of the time we either use a set of basic assumptions or chasing past conflicts as predictors of future ones (that, to be fair, tends to be a reliable one). An encouraging step is the new working paper from Blair, Blattman and Hartman: Predicting Local Violence. Below is the abstract:


"We use forecasting models and new data from 242 Liberian communities to show that it is to possible to predict outbreaks of local violence with high sensitivity and moderate accuracy, even with limited data. We train our models to predict communal and criminal violence in 2010 using risk factors measured in 2008. We compare predictions to actual violence in 2012 and find that up to 88% of all violence is correctly predicted. True positives come at the cost of many false positives, giving overall accuracy between 33% and 50%. From a policy perspective, states, international organizations, and peacekeepers could use such predictions to better prevent and respond to violence. The models also generate new stylized facts for theory to explain. In this instance, the strongest predictors of more violence are social (mainly ethnic) cleavages, and minority group power-sharing"

Main takeaways, from my point of view, are:
- They have identified a limited number of key indicators (so much for "complex root causes") as relevant. These may mean that even relatively small programmes/organizations could gather the data. The results may be Liberia specific, or even limited to the areas of the study (the usual How transferable this thing is? question), but the methodology seems promising
- What is it with power-sharing?!?! we may need to look further into the "quality" of the sharing and how that affects power dynamics. Does it raise expectations of minority groups and lead to more conflict? or majority groups do not accept the influence of minority groups? Maybe we should throw in some game theory!
- How much more accurate are our 'gut-feeling' predictive powers when selecting locations at the design and implementation stages? Forecasting tools may be specially useful when new staff with limited conceptual or contextual knowledge joins (it will help with the steep learning curve of the first months) in the design process
- The idea of machine learning or generation by the models of new facts. It may also help in identifying new trends over time. In a way this is connected to the fact that data-mining seems to have a bad image, yet it manages from time to time to surprise us and lead us to new questions.
- And of course, the question that is in the practitioner in me: how can I use forecasting as a baseline for impact measurement of conflict prevention? Aside from the attribution issue, will donors be on-board for non-existent conflicts (prevented?) that where previously forecast? Will they really accept forecasting as a planning tool? As we have seen from Somalia on the case of drought and food security, attentions sharpen at the early stages of the crises not when it was forecast or even the onset.


HT to the always recommendable Chris Blattman (One of the authors)