Predicting Food Crises

Andree, Bo Pieter Johannes ; Chamorro, Andres ; Kraay, Aart ; Spencer, Phoebe ; Wang, Dieter

World Bank, Washington, DC

关键词: FAMINE;    FOOD SECURITY;    FOOD INSECURITY;    EXTREME EVENT;    COST-SENSITIVE LEARNING;   

DOI  :  10.1596/1813-9450-9412
RP-ID  :  WPS9412

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

Globally, more than 130 million peopleare estimated to be in food crisis. These humanitariandisasters are associated with severe impacts on livelihoodsthat can reverse years of development gains. The existingoutlooks of crisis-affected populations rely on expertassessment of evidence and are limited in their temporalfrequency and ability to look beyond several months. Thispaper presents a statistical foresting approach to predictthe outbreak of food crises with sufficient lead time forpreventive action. Different use cases are explored relatedto possible alternative targeting policies and the levels atwhich finance is typically unlocked. The results indicatethat, particularly at longer forecasting horizons, thestatistical predictions compare favorably to expert-basedoutlooks. The paper concludes that statistical modelsdemonstrate good ability to detect future outbreaks of foodcrises and that using statistical forecasting approaches mayhelp increase lead time for action.