Multi-Purpose Cash Assistance: Impact Evaluation on the Well-Being of Syrian Refugees - 2020

A total of 11,457 households were visited and used in this analysis, which constitutes one of the largest samples among impact evaluations conducted in Lebanon to date. The impact evaluation used a quasi-experimental fuzzy regression discontinuity design, which enables estimation of the causal impact of MPC after different enrolment periods in the programme (rather than producing simple correlations). For this purpose, a multi-sectoral household survey was collected across three regions in Lebanon (Bekaa, North, and Mount Lebanon), where 85 per cent of Syrian refugees and 94 per cent of MPC beneficiaries live, over three waves of data collection held in July/ August 2018, February/March 2019, and July/August 2019. The multiple rounds of data collection allowed for the measurement of effects that materialise at different points in time and enabled the validation of the detected impact. Furthermore, collecting data in the summer and winter made it possible to account for seasonal variations, and accessing administrative data from UNHCR on other assistance programmes targeting the population of interest made it possible to append this information and distinguish between the effect of MPC and other programmes. WFP and UNHCR cash assistance beneficiaries are selected based on a proxy means testing formula that constructs a vulnerability score for each household based on a set of socio-demographic characteristics from the UNHCR database. This score predicts the per capita expenditure level of households and is used to rank them from most to least vulnerable. All households with a score below the SMEB are eligible for MPC. However, due to budgetary constraints, only about 29 per cent of eligible households currently receive the MPC package. WFP follows a bottom-up approach by including in the programme households starting from the lowest scores and moving up the scores until the allocated funding is fully disbursed. UNHCR employs a geographical bottomup targeting approach by including the most vulnerable households in each geographic region until it reaches the regions allocated proportion given its budget constraints. Accordingly, the point at which the funding is fully disbursed creates an artificial cut-off line. This creates a quasi-natural experiment where households on either side of the cut-off are plausibly similar along observable and unobservable characteristics. As a result, the argument is that the only difference is the receipt of MPC. Thus, any differences in outcomes between households are attributable to the amount of cash assistance received, which allows for measuring the causal impact of the MPC. Notably, because the annual recalibration of the targeting formula that determines a households eligibility for MPC occurred prior to the collection of wave 2 data, this uniquely positioned the study to measure the impact of discontinuation on affected household.