Conflict, food inflation, and climate change as time-varying determinants of food availability in Somalia: Evidence from machine-learning and VAR techniques

Le résumé fourni par la source

Food insecurity in Somalia has been attributed primarily to civil conflict, yet the relative contributions of conflict, food price inflation, and climate change to food availability—and how these contributions have shifted over time—remain insufficiently quantified. Somalia's predominantly rain-fed agropastoral economy, heavy dependence on cereal imports, and protracted history of insecurity make it an instructive case for understanding how multiple stressors interact and evolve within a structurally vulnerable food system. Using monthly data from January 2001 to December 2021, this study employs machine-learning techniques and a vector autoregression (VAR) framework with time-varying Granger causality to evaluate the evolving determinants of food availability in Somalia. Across all models, civil conflict is the dominant predictor, followed by food inflation (0.210), temperature (0.138), and precipitation (0.094). Temporal decomposition reveals a structural transition: conflict accounted for 70.74% of feature importance during 2001–2007, a period of intense conflict, but declined to 11.95% by 2015–2021, when recurrent droughts increasingly shaped food availability. Meanwhile, food inflation's share rose from 14.61% to 45.8%, and temperature's from 8.9% to 31.69%. The Seasonal-Trend decomposition using Loess (STL) confirms that trend effects outweigh seasonal and residual variation. Interestingly, time-varying Granger causality illustrates that inflation and temperature become increasingly influential over time.Food insecurity in Somalia has been attributed primarily to civil conflict, yet the relative contributions of conflict, food price inflation, and climate change to food availability—and how these contributions have shifted over time—remain insufficiently quantified. Somalia's predominantly rain-fed agropastoral economy, heavy dependence on cereal imports, and protracted history of insecurity make it an instructive case for understanding how multiple stressors interact and evolve within a structurally vulnerable food system. Using monthly data from January 2001 to December 2021, this study employs machine-learning techniques and a vector autoregression (VAR) framework with time-varying Granger causality to evaluate the evolving determinants of food availability in Somalia. Across all models, civil conflict is the dominant predictor, followed by food inflation (0.210), temperature (0.138), and precipitation (0.094). Temporal decomposition reveals a structural transition: conflict accounted for 70.74% of feature importance during 2001–2007, a period of intense conflict, but declined to 11.95% by 2015–2021, when recurrent droughts increasingly shaped food availability. Meanwhile, food inflation's share rose from 14.61% to 45.8%, and temperature's from 8.9% to 31.69%. The Seasonal-Trend decomposition using Loess (STL) confirms that trend effects outweigh seasonal and residual variation. Interestingly, time-varying Granger causality illustrates that inflation and temperature become increasingly influential over time.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.