Timing of Humanitarian Assistance

When Humanitarian Assistance Fades Out: Experimental Evidence from a Nation-Wide Program in Lebanon

Ghassan Baliki

ISDC - International Security and Development Center

A large literature has explored the inception and implementation of assistance programs – often experimenting with alternative program designs and targeting strategies – but there is little evidence to inform their scaling down. We study the scaling down of a nationwide humanitarian food assistance program in Lebanon. In January 2024 the number of beneficiaries had to be reduced from about 60,000 to 25,000 households. In partnership with the World Food Programme, we design a population-wide experiment to compare four popular targeting approaches to prioritise assistance among beneficiaries. We assess the performance of the four methods by relying on population-wide administrative data and two rounds of survey data collected in June 2024 and February 2025 among a stratified random sample of 6,400 households. Building on the random allocation across targeting arms, we also provide direct evidence of the consequences of ending assistance for those households that are eligible under some methods but not others (61% of the population). Finally, we study whether and how households prepare for the end of assistance by extending the warning window (between the announcement and the actual termination of assistance) for a random subset of households. Our results show that: 1) despite leading to different selections, the four methods do not lead to discernible differences in overall wellbeing; 2) the end of assistance leads to negative socio-economic consequences both in the short and medium run; 3) in spite of this, households do not manage to take preventive actions to cushion the negative consequences. Overall our findings highlight the negative socio-economic consequences of scaling down humanitarian assistance and provide concrete indications on how to minimize disruption.