Evaluating the effectiveness of Early Warning Based Anticipatory Action to mitigate adverse effects of droughts in Somalia

The intervention takes advantage of a randomly assigned variation in the payment structure of the cash transfers distributed prior to the harvest of an exceptionally dry agricultural season. Beneficiaries in the treatment group received the full amount of the transfer as a one-time lump-sum payment, whereas those in the control group received the same total amount distributed in three equal installments over three months. The treatment assignment was stratified at the district level and subsequently clustered at the village level to improve balance and, at the same time, prevent potential spillovers and social tensions between beneficiaries located in the same village.

Villages are randomized into treatment and control villages. All (eligible) HHs within the study villages will receive cash transfer for 220 USD. However, the control villages will receive it as three-instalments while the HHs in the treatment villages will receive it in the form of a one-time lump-sum payment. The amount remains the same in the 2 treatment arms.

2024-10-27

2025-01-10

Food Consumption Score (FCS) (discrete variable)
Probability of displacement (discrete variable)

Based on previous literature in the domain and planned interventions, food security will be the most important outcome of this intervention. Within food security, we are looking at Food Consumption Score (FCS) - a composite index based on households’ dietary diversity, food consumption frequency, and relative nutritional value of different food groups. The FCS is calculated by asking how often households consume food items from the 8 different food groups (plus condiments) during a 7-day reference period.  In addition to this, the FCS module collects data on sources of the consumed foods acquired by households.

Literature also shows that receiving anticipatory action before the disaster occurs can reduce the probability of displacement.

Livestock losses
Assets
Wealth Index
rCSI

Literature shows negative shocks can impact all material conditions of a household. We expect anticipatory action cash transfer to mitigate these adverse impact on assets. The impact of selling assets during shocks has two-fold impact, 1) loss of assets and 2) since everyone sells at the same time, the selling price is also compromised.

Wealth Index provides a composite index on all the assets available with the HH.

We are also looking at reduced coping strategy index (rCSI), an indicator used to compare the hardship faced by households due to a shortage of food. The index measures the frequency and severity of the food consumption behaviours the households had to engage in due to food shortage in the 7 days prior to the survey.

The study is conducted across three drought-prone districts in Somalia, employing a village-level randomization approach.

A cluster-stratified randomized controlled design was used to evaluate the effectiveness of different payment structures within the framework of Early Warning-Based Anticipatory Action (EWBAA). Beneficiaries located in villages targeted by the FAO Joint Drought Recovery and Resilience Project (JDRRP) program were cluster-assigned to either a lump-sum or a three-instalment payment structure. Specifically, within each village considered, only households that met the vulnerability criteria established by the FAO Country Office for JDRRP targeting were included in the beneficiary group. All eligible households in the treatment and control villages received an anticipatory action cash transfer, either in the form of a lump sum or in three instalments over approximately three months. At baseline, all the villages considered were classified as drought-prone, and the cash transfers were triggered based on vulnerability assessments conducted according to a predefined protocol implemented at the beginning of the agricultural season.

The study will evaluate the impact of different payment structures for cash transfers in an anticipatory action setting. The intervention consists of two different payment structures for anticipatory action cash transfers delivered to vulnerable households (HHs) in two drought-hit districts: Dhusamareeb, and Galdogob, both of which face IPC (Integrated Food Security Phase Classification) 3 or 4 food insecurity conditions. An approximation of the villages and observations is given below:

• Dhusamareeb: 6 villages – 387 HHs
• Galdogob: 3 villages – 396 HHs

Assignment to a specific payment structure was randomly conducted at the village level using a cluster-stratified random assignment, with the district serving as a stratum. Beneficiary households are vulnerable households targeted by other FAO programs that meet FAO Somalia (FAOSO) eligibility criteria. All beneficiaries received cash transfers either through a lump-sum transfer (treatment) or in three distinct monthly instalments (control).

Cash transfers were defined according to the minimum expenditure basket in the two districts. Beneficiaries in both Dhusamareeb and Galdogob received 210$. The total amount was either deployed as a single lump-sum transfer or distributed in three consecutive monthly transfers ($70 per month in the two districts). Simultaneously, the JDRRP program is expected to provide complementary inputs (seeds, fertilizers, and other critical farming supplies) and/or training (fodder production techniques, conflict mitigation strategies), as well as storage facilities and machinery, to HHs in both the treatment and control groups during the agricultural season.

The modality of the cash transfer has been assigned following a cluster-stratified randomization at village levels with districts serving as the strata. All eligible households within the villages received the cash transfer. Randomization was done through a statistical software.

Village

Yes

9 villages

783 HHs

5 control villages (approximately 317 HHs)
4 treatment villages (approximately 466 HHs)

FCS discrete (21): MDE = 0.138, Control mean: 0.931 (0.253) Probability of displacement: MDE = 0.068, Control mean: 0.971 (0.165)