Bayesian spatiotemporal modelling of political violence and conflict events using discrete-time Hawkes processes

Monitoring of conflict risk in the humanitarian sector is largely based on simple historic averages. To advance our understanding, we propose Hawkes processes, a self-exciting stochastic process used to describe phenomena whereby past events increase the probability of future events occurring. The overarching goal of this work is to assess the potential for using a more statistically rigorous approach to monitor the risk of political violence and conflict events in practice and characterise their temporal and spatial patterns. The region of South Asia was selected as an exemplar of how our model can be applied globally. We individually analyse the various types of conflict events for the countries in this region and compare the results. A Bayesian, spatiotemporal variant of the Hawkes process is fitted to data gathered by the Armed Conflict Location and Event Data (ACLED) project to obtain sub-national estimates of conflict risk over time and space. Our model can effectively estimate the risk level of these events within a statistically sound framework, with a more precise understanding of the uncertainty around these estimates than was previously possible. This work enables a better understanding of conflict events which can inform preventative measures. We demonstrate the advantages of the Bayesian framework by comparing our results to maximum likelihood estimation. While maximum likelihood gives reasonable point estimates, the Bayesian approach is preferred when possible. Practical examples are presented to demonstrate how the proposed model can be used to monitor conflict risk. Comparing to current practices that rely on historical averages, we also show that our model is more stable and robust to outliers. In this work we aim to support actors in the humanitarian sector in making data-informed decisions, such as the allocation of resources in conflict-prone regions.