Impact and process evaluation of BRAC’s Childcare Model in Bangladesh

To create access to childcare services for low-income urban households, BRAC piloted a model of home-based care entrepreneurship that connects local care entrepreneurs with households in need of childcare. Their "home-based childcare model" is designed to be a low-cost solution for labor-based, low-income communities by systemizing informal childcare practices and improving service quality through capacity building and technical guidance in play-based pedagogy and entrepreneurship to the care providers. This project is designed to work with both existing care providers, to improve their quality of service with the primary objective of meeting children’s developmental needs, and new caregivers to add and further strengthen care-based elements that build on everyday functioning to ensure socio-emotional well-being. This will be referred to as a ‘Care Model’.

Drawing on the lessons learned from the interventions to organize entrepreneurs to improve quality of childcare while maintaining the informality and affordability, BRAC’s Urban Development Program (UDP) and BRAC Institute of Education and Development (BRAC IED) will create new entrepreneurs offering this serve. The participants who will be entrepreneurs, will receive training on play-based childcare services, mental health support for children, and entrepreneurship skills (similar to the support that existing entrepreneurs receive to improve the quality of services) from BRAC IED.

As a project designed to generate systematic and rigorous evidence of the model while it is being developed and implemented, to inform subsequent scale-up, the research will combine quantitative (both experimental and non-experimental) and qualitative methods.

Quantitative study design (RCT and non-experimental) : BRAC will identify 200 potential new entrepreneurs who can offer home-based daycare services. For each of these potential entrepreneurs, 16 target clients will be identified in their neighborhoods. From them, 150 entrepreneurs will be selected as the treatment group. They will receive training on play-based childcare services and materials (similar to the support that existing entrepreneurs receive to improve the quality of services) as well as entrepreneurship training. The control group will be offered entrepreneurship training, which is not specific to childcare, and their “potential clients” will serve as “pure control”.

Within the treatment group, there will be a second layer of randomization of the 16 potential clients into three groups –
T1: receive information about childcare services (8 mothers),
T2: information with a cash transfer (4 mothers), and
T3: information with childcare fees payment/subsidy to avail the service (4 mothers).

The cash transfer and subsidy will be provided in the form of fees paid to the entrepreneur for 6-12 months after they are ready to serve these clients and the mothers of the subsidy arm will be informed that the fees have been paid for. Initially we will communicate about the cash transers (in T2 and T3) for 6 months, which can be increased upto 12 months depending on childcare service uptake. This layer of study will test the customers’ willingness to pay for childcare services. The size of the subsidy will be determined by using the data from the current service users, which can be customized by the age of the child. The amount of cash transfer will be the same as the amount of subsidy. Although the model was initiated to meet the need for care services by the mothers who predominantly work in garment factories, descriptive studies reveal that the need is also prevalent for women who work in other sectors such as domestic help or informal businesses. Therefore, the participants will be selected based on their proximity to the potential care provider entrepreneurs, and both working women (irrespective of the sector) and women who left work due to childcare responsibility will be eligible. The only binding criterion for the potential client selection will be having a child. Although we will prioritize mothers of children aged between 0 and 8 years, the age cut-offs may need to be adjusted based on the numbers observed per potential entrepreneur.

The 150 entrepreneurs in the treatment group would receive specialized training on setting up their home-based childcare as a business. The 50 entrepreneurs in the control group will receive the usual entrepreneurship training (not particularly focusing on childcare centers) that BRAC IED will provide.

Within the treatment group, there will be a second layer of randomization, which will divide 16 potential clients (mothers) into three groups – a) receive only the information about childcare services available at their respective entrepreneurs, b) information with a cash transfer, and c) a scholarship/subsidy to avail the service. The subsidy will be provided in the form of fees paid to the entrepreneur for 6-12 months after they are ready to serve these clients and the mothers of the subsidy arm will be informed that the fees have been paid for. Besides being the arm with adequate service uptake to measure the impact of childcare on the mothers, the subsidy arm will also ensure the entrepreneurs have a secured clientele/revenue for months. The participants will be informed that the cash/fees payments are going to be for six months, which will be revisited for extending based on the observed uptake among different arms at that time. The size of the subsidy will be determined by using the data from the current service users, which can be customized by the age of the child. The amount of cash transfer will be the same as the amount of subsidy. The impact indicators for the mothers in these three arms will be compared against the mothers who will be identified as potential clients had the entrepreneurs of the control group been supported to set up their daycare service. Therefore, in total 3,200 mothers of beneficiary children will be surveyed, with 2400 from the treatment group and 800 from the control group.

Besides this sample for the experimental evaluation, the research will also include two other sample categories – a) existing care providers and their clients (mothers) and b) communities without any care provider and included in the community mobilization workshops. The objective of including these observational samples is to conduct comparisons with the experimental groups. For instance, if a good number of the control group entrepreneurs end up starting childcare enterprises (which is unlikely a priori, but cannot be ruled out), the second observational group can be used to understand the role of “community mobilization” and spillover effects of the RCT itself. The community mobilization, which is part of the entrepreneur and client identification, may act as a tool for addressing coordination failure by discussing the need for quality childcare service that can encourage control entrepreneurs to acquire the skills independently. Similarly, a comparison of treated entrepreneurs with existing care providers can reveal the importance of experience in their quality of care. For the spillover sample, we will collect data from 900 mothers who are outside but neighboring the RCT clusters.

We will do a baseline survey to gather data on characteristics such as age, educational levels, history of labor force participation, household dependency and income, marital status, etc. for both the potential entrepreneurs and potential clients. For potential clients, additional information will be collected on their opinion of paid childcare services and willingness to pay. This data will be used to analyze determinants of taking up the services by the entrepreneurs (once offered). Moreover, the project will measure any changes in mothers’ employment, mental health status (including cognitive loads), and the level of their trust in service providers after the existing entrepreneurs receive the training and material support from the project. For employment outcomes, the study will measure the impact on both extensive (labor force participation status) and intensive (hours of work) margins. There will also be a comparison between the previously trained childcare providers and newly trained ones to understand the impact of the training on the quality of services and mothers’ satisfaction with the services. A midline survey will be conducted towards the end of the cash transfer and fees payment period, and the endline will be conducted 6-months after the last of the transfers are completed.

An intermediary output of the project will be a description of the enterprise economics of home-based care that will largely be based on the baseline survey and operational/monitoring data combined with process documentation and operational research findings. This output is intended to inform the intervention development for BRAC to scale up the approach by working with more entrepreneurs, children, and mothers. Comparing information collected at baseline on willingness to pay with actual uptake of the services by the mothers of the three treatment groups and correlations of uptake with various socio-economic characteristics will be used for generating a description of a viable business model.

2024-07-01

2025-06-30

The primary outcome of this experiment is to measure the impact of (the quality of) childcare on mothers.

For the RCT, there are different layers of outcome measures. For variations within treatment clusters, the main outcome measure will be the uptake of the childcare services by “info only”, “info + cash transfer” and “info + fees paid” arm. For this outcome, when comparing fees paid vs. cash transfer groups, the minimum detectable effect (MDE) is 8.8 percentage points assuming 30% uptake for one group and 90% power. If the comparison is made between fees paid (or cash transfer) with 4 observations per clusters from 150 clusters and the information only group (8 observations per cluster), the MDE of uptake comes to 7.7 percentage points. These impact estimates will be important to estimate market analysis such as – which type of mothers are mothers are more likely to avail the services under the three treatment conditions – as heterogeneity of impact on accessing services.

The secondary outcome of this RCT is to generate formative evidence of the “enterprise economics” of home-based childcare for new entrepreneurs.

BRAC UDP will identify 200 potential home-based daycare entrepreneurs and 16 target mothers for each provider.
150 entrepreneurs will receive play-based childcare and entrepreneurship training, forming the treatment group and the other 50 entrepreneurs will be the control group. Within the treatment group, mothers (potential clients) will be randomized into three arms -
T1: information-only,
T2: information+ cash transfer,
T3: information+childcare fees paid
to assess childcare impact on mothers' employment, mental health, and trust, with a focus on securing entrepreneurs' revenue. he study incorporates both experimental (randomized client support) and non-experimental (pre-post comparison) components, including a comparison of service quality between existing and new entrepreneurs.

Qualitative Research Design: The qualitative samples will consist of mothers (existing and potential childcare clients), existing and new childcare entrepreneurs, and BRAC staff. In-depth interviews with mothers (n=24), focus group discussions with mothers (n=3), in-depth interviews with childcare entrepreneurs (n=9) and key informant interviews (n=3) will be used alongside participant observation.

Childcare entrepreneurs will be sampled purposively from BRAC’s record of home-based childcare entrepreneurs.
Mothers will be identified using the census (to be carried out in November 2023), which will enable us to identify a sample with variation relating to current childcare practices and work situations. BRAC UDP support staff in each site will be identified for key informant interviews.

Quantitative Sampling Methods (Details)

First, the program will provide us with a list of 100 existing entrepreneurs and 200 potential new entrepreneurs.
We will then randomly assign 150 entrepreneurs to the treatment group and 50 entrepreneurs to the control group.
We will conduct a census to identify potential client mothers in the same geographical area of the entrepreneurs.
Within the treatment group, there will be a second layer of randomization, 16 potential client mothers will be selected randomly from each entrepreneur’s clients and then divided randomly into 3 treatment arms. 8 mothers will receive only information, 4 mothers will receive information plus cash subsidy and 4 mothers will receive ‘scholarship’.
We will also survey the clients (mothers) of each of the existing 100 entrepreneurs (4-10 per provider).

Qualitative Sampling Methods (Details)
Data will be collected and analysed after five months in the study locations in or around Dhaka. 
The qualitative samples will consist of mothers (existing and potential childcare clients), childcare entrepreneurs (existing and potential) and BRAC staff. In-depth interviews with mothers (n=24), focus group discussions with mothers (n=3), in-depth interviews with childcare entrepreneurs (n=9) and key informant interviews (n=3) will be used alongside participant observation.  
Childcare entrepreneurs will be sampled purposively from BRAC’s record of home-based childcare entrepreneurs.
Mothers will be identified using the census, which will enable us to identify a sample with variation relating to current childcare practices and work situations.
BRAC UDP support staff in each site will be identified for key informant interviews.

Randomization at both clusters and mothers level will be done after baseline survey using stata codes

In the first level of randomisation, we will list the 200 clusters (consisting of at least one potential new entrepreneurs and 16 eligible mothers) and we will then randomly assign 150 cluster to the treatment group and 50 entrepreneurs to the control group. One entreprenuer will be trained in the 150 treatment clusters, and the mothers in the treatment group will be further subdivided into three groups.

Within the treatment group, there will be a second layer of randomization, 16 potential client mothers will be selected randomly from each entrepreneur’s clients and then divided randomly into 3 treatment arms. 8 mothers will receive only information, 4 mothers will receive information plus cash subsidy and 4 mothers will receive ‘scholarship’.

Yes

200 clusters.

At least 200 clusters, 200 entrepreneurs, 3200 clients/mothers and their 3200 children. If we are able to create more than the targeted 200 clusters, we will still have 150 treatment clusters and the rest in control (i.e. potentially more than 50 control clusters).

The surveys will cover existing as well as potential entrepreneurs. While the sample of 100 existing entrepreneurs will be drawn from the ongoing projects by UDP, the potential new entrepreneurs will be selected as part of the RCT. Although the interventions will include 150 entrepreneurs, we anticipate identifying around 500 potential entrepreneurs who will be covered at baseline and two follow-up surveys. Further details are provided in the section on participant selection. The survey will include information on their employment and income, financial assets, intra-household relationships and aspirations.

The study will cover a sample of about 5,000 mothers, of whom 3,200 are part of the RCT. Survey module will cover time usage, employment, income, mental health, and stress as well as perceptions and norms on paid childcare services. While the mothers participating in the RCT will be surveyed at baseline, midline and endline, we will conduct two rounds of survey (baseline and endline) for the non-experimental sample.

Only those children who are part of the RCT will be included in quantitative survey at baseline and at endline. Since the children from the non-experimental sample categories are unlikely to observe any change in their childcare services, they will not be surveyed for impact evaluation. Despite the initial sampling of 3,200 children, the survey will be tailored to focus exclusively on the age group of 0-5Yrs, without the broader inclusion criterion of up to 8 years of age. The reason for this plan to survey younger children (0-5Yrs) is to limit the number of survey tools to be used as well as to cover children where we can expect larger effects of the interventions. Tools such as the Global Scale for Early Childhood Development (GSED), and the International Development Early Learning Assessment (IDELA) will be used to collect data at baseline and endline.

A sub-sample of study participant children will be included for tracking neurodevelopment with technical assistance from the ICDDR,B. Utilizing the Hyperfine scanner already installed at ICDDR,B, we will conduct neuroimaging on a subset of 300 children, drawn from both the control and treatment groups, at both baseline and endline assessments. The sub-sample can be drawn based on age groups (instead of a random sample of 2-5 years old) for whom the changes are most likely to be captured in the study. Measures of brain structure, childcare quality and stimulation and school readiness skills will be assessed to help demonstrate the importance and impact of (quality) childcare. With the main objective of measuring the second stage effect on children as an alternative measure (in addition to observational data collection on child development), the selection of sample from the treatment group may oversample T3 (info + fees paid) to have larger difference in service uptake between control and treatment at the first stage. The D&T resources will also be used to complement the qualitative anthropological research to identify at-home childcare practices and interactions that affect neurodevelopment across all the groups. Researchers from IED and BIGD will also participate in training to build their skills in interpretation of neuroimaging and gain hands-on experience by working with the collaborators.

The minimum detectable effect (MDE) for the labor force participation of mothers is 0.11 percentage points and for scale outcomes (e.g., average income) is 0.20 standard deviation when comparing the subsidy arm with the pure control. These MDEs for intention-to-treat (ITT) effects, and average treatment effects need to be 30% higher assuming 70% uptake. When comparing service uptake between the “information only” and “information + cash” arms, the MDE is much lower (i.e., better powered) since randomization is at the individual level.