Madagascar Human Development Cash Transfer Impact Evaluation

Eligible poor, rural Malagasy households will be given a cash transfer by the Government of Madagascar in a program called the Human Development Cash Transfer (HDCT). The HDCT will last for three years, with cash transfers of approximately 10-12 USD being issued to beneficiary households once every two months. Receipt of the cash transfer will be conditioned on 80% school attendance for households with children ages 6-10 and will be unconditional for households with children ages 0-5. A portion of the transfer beneficiaries will be organized into “Mother Leader” groups, or a group of approximately 10-15 women from the same community who are led by a community-elected female leader called a “Mother Leader” (ML). The MLs, who are beneficiaries themselves, will be trained to learn about and implement best practices in health, nutrition, and parenting. The MLs will guide their groups through modules on these various subjects as part of the cash transfer program. In addition, a portion of the Cash + ML recipients will also be exposed to behaviorally informed “nudges”, or organized activities that encourage beneficiaries to follow through on their intentions and actively work on their goals toward better household outcomes in health, productive activities, and well-being.

This trial will seek to evaluate the impact of the HDCT on household economic well-being, child health, and educational outcomes.

2016-09-01

2020-04-23

Key outcome variables to measure include the following:

+ Food-related household consumption and spending
+ Non-food household consumption and spending
+ Household agricultural production
+ Household income sources
+ The role of women / female empowerment
+ Parental interactions with children
+ Household food security
+ Savings behaviors
+ Child health
+ Child school attendance and educational outcomes
+ Cognitive development of children

Subjects
The subjects in this experiment were individuals in poor rural regions of Madagascar that have children between the ages of 0 and 10 years AND have been deemed as among the lowest income individuals in the community.

Study Design
The impact evaluation will randomly sample a balanced number of beneficiaries from across the control regions and the various program groups outlined below. A survey will be conducted on the random sample to capture outcome measures on the variables listed above.

HDCT Program Groups
Subjects were randomized into five groups, namely:

0. Control (No cash transfer)
1. Cash Transfer
2. Cash Transfer + Mother Leader
3. Cash Transfer + Mother Leader + Nudge 1
4. Cash Transfer + Mother Leader + Nudge 2

Cash Transfer
The monetary transfer is the basis for the HDCT program. A cash transfer is disbursed to program beneficiaries on a bi-monthly basis. A base amount is given to all beneficiary families, with additional amounts allocated for each of up to two children. For families with children who are eligible for primary school based on their age (6-10 years old), the transfer amount is conditional on the child having at least an 80% school attendance record. For families with children who have not yet reached primary school age, the transfer is unconditional.

Mother Leader Program
Borrowing from an expansive literature on the efficacy of community-based health promotion workers and building upon the recent experience of empowering beneficiary women in a cash transfer project in Colombia, the HDCT program includes a “Mother Leader” (ML) component. Beneficiary women are assigned to small groups within their fokontany and each group elected one of the group members to serve as a Mother Leader for a term of one year, a voluntary position. MLs are trained in child development activities, health promotion activities, and civics, lessons that they then impart in monthly group sessions. MLs also act as a liaison between the program administrators and the beneficiaries. The ML program also includes group-wide activities related to savings, community participation, or financial empowerment activities.

Behaviorally-Informed Nudges
Behaviorally-informed nudges will be implemented at the ML group level and will be designed to enhance the effects of the mother leader and cash transfer program treatment interventions.

This is a multi-level cluster randomization design. Receipt of the cash transfer will be randomly assigned at the commune level by a public lottery. Once the communes have been determined, we will use a computer to randomly assign the cash beneficiary fokontany to the treatment conditions.

This is a multi-level cluster randomization design. Receipt of the cash transfer will be randomly assigned at the commune level by a public lottery. Once the communes have been determined, we will use a computer to randomly assign the cash beneficiary fokontany to the treatment conditions.

Yes

318 fokotany (villages)

~8300 beneficiaries surveyed at endline

0. Control (No Cash Transfer): ~1600 beneficiaries surveyed (13 communes)
1. Cash Transfer: ~1650 beneficiaries surveyed (from 77 fokontany)
2. Cash Transfer + Mother Leader: ~1650 beneficiaries surveyed (from 77 fokontany)
3. Cash Transfer + Mother Leader + Nudge 1: ~1650 beneficiaries surveyed (from 77 fokontany)
4. Cash Transfer + Mother Leader + Nudge 2: ~1650 beneficiaries surveyed (from 77 fokontany)

The Cash Transfer effects will be determined by measuring differences between the 13 control communes and the 38 beneficiary communes. We will sample 13 communes with 5 village per commune and 16 households per village in the control population. For any of the treatment arms, we will sample 74 villages, with 5 Mother Leader groups per village and 4 households per Mother Leader group. The MDE we calculate for the cash effects is 0.3 standard deviations for the primary outcomes, with power of 0.7 and significance of 10%. The MDE we calculate for the nudge and ML interventions is 0.1762 standard deviations for the primary outcomes, with power of 0.8 and significance at 5%. The calculations above on Unit Level 2 also control for a fokontany-level covariate that is expected to explain a conservative proportion of the variance (9-10%) among Mother Leader groups and also among Households. The Unit Level 1 calculations also control for a commune-level covariate that is expected to explain a conservative proportion of the variance (10%) among fokontany and also among Households. Given current lack of data, we opt to use the default/norm estimate of 0.1 ICC. We will be better able to calculate ICC with the completion of data collection for the baseline survey.