Measuring and testing vulnerability to food insecurity for prediction and targeting
利用新冠疫情的外生冲击，检验了基于理论的脆弱性衡量方法在预测粮食不安全家庭方面的表现，发现其不如简单的数据驱动方法，并指出数据稀缺环境下预测模型的局限性。
Taking advantage of the exogenous nature of the COVID-19 shock, we present an empirical test to assess the validity of theory-based methods for targeting household vulnerability to food insecurity. Specifically, we test the performance of an applied measure of vulnerability to food insecurity using the World Bank's multi-topic longitudinal survey data for Nigeria collected between 2010 and 2020, covering the period before and after the pandemic. The results show that this vulnerability measure severely underperforms in predicting food-insecure households out of sample relative to a simple, data-driven routine. Sensitivity tests using only the pre-pandemic data reveal that the poor forecasting performance is not simply due to the discontinuity in the data-generating process brought about by mobility restrictions. This evidence carries two important implications: i) from the methodological point of view, there is a need to enhance the effectiveness of targeting approaches employed by policymakers to identify vulnerability hotspots; ii) from a data-oriented perspective, this work underscores that predictive vulnerability models, regardless of their theoretical soundness or computational power, are constrained by data availability in data-scarce environments. Overall, we argue that greater methodological effort is required to address the limitations of current approaches in anticipating vulnerability to shocks.