B1358
Title: The short-term dynamics of conflict-driven displacement: Bayesian modeling of disaggregate Data from Somalia
Authors: Gregor Zens - IIASA (Austria) [presenting]
Lisa Thalheimer - United Nations University (Germany)
Abstract: Understanding the immediate effects of conflict on forced displacement is crucial for timely policy intervention, yet quantitative analyses in this realm are sparse. This is primarily due to the scarcity of high-frequency data on displacement and the methodological challenges that arise when analyzing imperfect data collected in conflict zones. Addressing this gap, this paper develops a Bayesian dynamic linear distributed lag model to quantitatively assess the short-term impact of conflict on displacement in Somalia, using weekly panel data capturing over 8 million displacements and 19,000 conflict events from 2017 to 2023. State space and factor modeling techniques are employed to handle the complex spatio-temporal dependencies and data gaps that characterize high-resolution datasets from conflict zones. The model integrates regularization priors to mitigate overfitting due to noise, inherent to survey data collected in volatile environments. Computational scalability in panels of large dimension is ensured by employing efficient simulation smoothers for posterior simulation. Findings reveal a rapid and non-linear displacement response post-conflict, with heterogeneity in effects dependent on the nature of conflict events. The utility of the model is further demonstrated through a forecasting exercise, where it outperforms standard benchmarks, underscoring its relevance for informed decision-making in crisis scenarios.
