Caring about carework: experimental evidence on rural childcare provision

Our setting is Kongo Central, a rural western province of the DRC. The DRC ranks among the poorest countries in the world. Within agriculture, which employs over two-thirds of women in the DRC, the production of women farmers is 18% lower than that of men and their productivity is 11% lower. Gender gaps are even larger when comparing men and women in the same households (World Bank 2021). The study evaluates whether rural community-based childcare provision can help close these gaps.

The intervention is a participative childcare service offered to selected (treatment) villages. The community is responsible for identifying and providing a suitable and safe infrastructure that can host the childcare centers and identifying a community member to serve as the childcare center manager (‘encadreuse’). Parents are responsible for co-financing meals.

The project equips the center identified by the community with a first aid kit, kids’ toys, basic sanitary equipment and Covid-19 prevention kits. The project also trains each selected encadreuse for two weeks using an Early Childhood Development (ECD) curriculum developed specifically for this intervention by Save the Children to ensure a high standard of care. The encadreuses follow this curriculum during the center’s opening hours to stimulate children's socio-cognitive development.

In every treated village, the number of centers is based on the number of eligible women’s children: one center if the village has 6-14 treatment children and 2 centers for 15-24 children. The project is supervised by the research team in collaboration with the DRC’s Ministry of Education.

2020-03-01

2022-04-30

• Labor force participation (overall and by sector)
• Time spent on economic activities
• Time spent on carework
• Agricultural production (crop value)
• Agricultural productivity (crop value/ha)
• Child development (pre-numeracy and literacy, executive function, motor skills, socio-emotional skills)

• Income
• Dietary diversity
• Self-reported well-being

The sampling frame of this experiment consists of 111 villages within the DRC province of Kongo Central. A listing activity (census) was conducted across four territories (Songololo, Mbanza Ngungu, Lukula and Sekebanza) to identify these villages, which had two fulfill four primary criteria.

The community had to:
i. be accessible (as defined by our implementing partner)
ii. be interested in hosting a community childcare center
iii. be able to host a center (having a viable and safe building to host the center free of cost, as well as a woman that the community felt comfortable putting in charge of the center)
iv. have at least 10 eligible households.

Households were defined as eligible if they had at least one mother with a child aged 1 to 5 years in October 2019 that was not yet enrolled in a formal school and was interested by the childcare service.

Once this sampling frame was selected, we performed pairwise Mahalanobis matching using an optimal greedy algorithm to randomly assign eligible villages to treatment and control groups, resulting in 56 treatment villages. Childcare centers are opened in the treated villages, control village do not receive the intervention. The number of childcare centers is determined by the number of eligible children (up to 14 children per center).

We then performed a second level of randomization, so as to identify spillover effects in larger villages (>20 eligible households). This household-level randomization was conducted through a public lottery. In every treatment village, all eligible households were invited to draw green and red balls from a basket, after having drawn their ranking order. All eligible mothers in a household share the same treatment status (all treatment or all control).

Our primary experimental comparison will be between treatment households in treatment villages and control villages. In larger villages (>20 eligible households) we will also conduct secondary tests to identify suggestive differences between treatment and control households in treatment villages, and control households in treatment versus control villages.

Cluster randomization done in office by a computer and individual randomization done through a public lottery.

Household- and village-level

Yes

111 villages

Approximately 1500 households

The ~1500 households are repartitioned as follows:

~500 treatment households in treatment villages
~500 control households in treatment villages
~500 control households in control villages