Early Childhood Human Capital Formation at Scale: Experimental Evidence from Bangladesh

Bangladesh government runs the flagship National Nutrition Services (NNS) program that promotes best nutritional practices through micro-nutrient supplements, growth monitoring, counseling support, and regular monthly home visits by community health workers. This vast NNS public health infrastructure provides an opportunity to integrate an early childhood development (ECD) program and deliver the intervention at scale. In this RCT, the intervention targeted more than 18,000 children aged 3-18 months at baseline. A key design feature of the intervention was its use of existing government NNS staff to deliver the intervention instead of a separate group of ECD workers. Staff were trained to implement the program at-scale without additional monetary incentives. The program was implemented in 78 community clinics across three districts located in different divisions in Bangladesh. Half of these community clinics and their catchment areas were randomly assigned to the treatment group which received both NNS and ECD programs, while the other half of clinics were randomized into the control group that only received the existing government NNS program. The ECD component that was delivered in this RCT included the following intervention bundle: (i) training of existing NNS workers on ECD, (ii) provision of four types of program materials to parents, comprising a child development card, two picture books, and a booklet with key messages and clinic visit guidelines, and (iii) regular ECD counseling during routine monthly visits by NNS workers. The intervention was implemented over 15 months from June 2014--August 2015.

2014-06-01

2015-08-31

Cognitive and language skills (Bayley-III), Socio-emotional skills (Wolke behavioral rating scales), Anthropometric outcomes (height-for-age, weight-for-height, and weight-for-age)

The program was implemented in 78 community clinics across three districts located in three different divisions (out of seven) in Bangladesh. Half of these community clinics and their catchment areas were randomly assigned to the treatment group which received both NNS and ECD programs, while the other half of clinics were randomized into the control group that only received the existing government NNS program.

Our study was executed in three districts in the Barisal, Chittagong, and Sylhet divisions in Bangladesh. The research team selected the three districts based on the existing government-run NNS program supported by Save the Children under the Tackling Childhood Malnutrition (TCM) project. These three study sub-districts have a total of 41 unions (the lowest administrative unit in Bangladesh). From these unions, six unions were excluded due to incomplete health assistance records. From the remaining 35 unions, we selected 30 unions that each had at least two community clinics. This process selected 78 community clinics in Muladi, Kulaura, and Satkania sub-districts and its catchment areas. Within the selected areas, households with children aged between 3-18 months were eligible to participate in our study. The minimum age restriction was imposed due to the validity of the main developmental assessment tool used for the evaluation (the Bayley-III; Bayley, 2006), which had not yet been validated on children below the age of three months in Bangladesh. Moreover, the Bayley-III test is only valid for children aged up to 42 months. Hence, the upper age restriction was imposed to collect end-line assessment 24 months after the baseline. We sampled 33 eligible households from each of the 78 community clinic catchment areas to reach a total sample size of 2,574 households. Upon completion of the baseline surveys, we performed a community clinic-level clustered randomization, where half of the community clinics (39) were assigned to treatment and the rest (39) to the control or ``business as usual'' arm. Treated community clinics and their catchment areas received a bundle ECD intervention in addition to regular NNS activities.

Randomization done in office by a computer

Community clinic catchment area

Yes

78 community clinic catchment areas

2,574 households (mother/caregiver-child pairs)

39 clinic catchment areas control and 39 clinic catchment areas treatment (1,287 mother/caregiver-child pairs by arm)