NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Improving school preparedness and child health outcomes through Integrated Child Development Services (ICDS) in Tamil Nadu
Initial registration date
February 16, 2017
July 31, 2018 3:41 AM EDT
University of California, San Diego
Other Primary Investigator(s)
New York University
University of California, Berkeley
Additional Trial Information
A large body of scientific evidence has established the central role of early childhood health and education in fostering lifetime well-being and economic success of children in both developed and developing countries (Engle et al., 2007; Elango et al., forthcoming). In India, programs that promote early childhood development are delivered primarily by Integrated Child Development Services (ICDS). ICDS is under-resourced relative to its importance for human development, which results in uneven quality, with many programs failing to provide supplementary nutrition, pre-school education, or essential health services (PEO, 2011). At the same time, fiscal constraints make large increases in ICDS spending infeasible. It is, therefore, critical to determine the most cost-effective methods for boosting ICDS quality so that scarce resources can be directed to programs that generate maximal social value.
ICDS in Tamil Nadu is one of the best-performing government schemes in India. A comprehensive study of ICDS published by the Planning Commission ranked Tamil Nadu first on an index of anganwadi infrastructure, and among the top 10 states overall (PEO, 2011). The state also outperforms most others on key nutrition indicators, including the fraction of underweight children (Ministry of Women and Child Development, 2013-2014). Tamil Nadu is therefore in a strong position to provide national leadership in identifying, piloting, and evaluating interventions with the potential to improve ICDS services at scale. Our research, to be conducted in partnership with ICDS in Tamil Nadu, will provide experimental evidence on the cost-effectiveness of policy interventions that aim to improve the functioning of ICDS. In the first year of this partnership we will conduct a randomized experiment to evaluate four such interventions in a representative sample of anganwadi preschool centers (AWCs) throughout Tamil Nadu. The four interventions are: 1. Hiring early childhood care and education (ECCE) facilitators;
2. A supplemental nutrition program;
3. A performance-based pay for anganwadi workers linked to child nutrition outcomes; and
4. An across-the-board increase in pay for anganwadi workers. Registration Citation
Ganimian, Alejandro, Karthik Muralidharan and Christopher Walters. 2018. "Improving school preparedness and child health outcomes through Integrated Child Development Services (ICDS) in Tamil Nadu." AEA RCT Registry. July 31.
Intervention Start Date
Intervention End Date
Primary Outcomes (end points)
We will evaluate the impact of the four interventions (and combinations of those interventions) on: (a) demand for schooling among low-income families (as measured by enrollment figures in anganwadi centers); (b) children's nutrition (as measured by their height- and weight-for-age); (c) children's skills (as measured by a battery of assessments of language and math skills and executive function).
Primary Outcomes (explanation)
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Our evaluation of these interventions will be split into two distinct studies with overlapping objectives. The first (which we call the "rural study") will be implemented in a representative sample of 640 anganwadi centers (AWCs) in four randomly-selected districts in Tamil Nadu: Kancheepuram, Karur, Thiruchirappalli, and Virudhunagar. AWCs in these districts will be randomly assigned to one of four treatment groups, one for each intervention listed above, or a control group that receives no intervention. Comparisons of outcomes for these groups will provide policy-relevant evidence on the efficacy of each individual program.
Our second study (which we call the "urban study") will be conducted in a representative sample of 160 AWCs in Chennai, the largest urban center in Tamil Nadu. The preschool market in Chennai differs markedly from the market elsewhere in Tamil Nadu: as a result of Chennai’s dense population, families typically have several nearby preschool options, including one or more AWCs, private preschools, other government-provided childcare centers, or home care. This creates scope for important interactions between program alternatives and household choice responses that do not exist in the other four districts in addition to providing evidence on the efficacy of a combination of the individual programs. We will randomly assign 40 AWCs to a treatment group that combines three interventions: (a) hiring early childhood care and education (ECCE) facilitators; (b) a supplemental nutrition program; and (c) performance-based pay for anganwadi workers linked to child nutrition outcomes. We will randomly assign the remaining 120 AWCs to a control group with no intervention. Comparisons of outcomes for these groups will provide policy-relevant evidence on the efficacy of the combination of these programs. We will also conduct a detailed household survey at both the treatment and control centers to study how this combination affected pre-school choices and household behavior.
Experimental Design Details
Anganwadi centers were randomly assigned to experimental groups using Stata's random number generator.
The selection of sample centers was stratified by administrative unit. Tamil Nadu is divided into districts, which are further divided into projects. Finally, projects are divided into sectors. The randomization process stratified at each of these levels such that the sample centers are representative of the entire district. The sampling protocol also accounted for vacant anganwadi worker (AWW) and anganwadi helper (AWH) positions, which are common in Tamil Nadu. We dropped AWCs that had both AWW and AWH vacancies. We then stratified sampling by vacancy such that half the selected sample AWCs have either an AWW or AWH vacancy, and half do not.
Random assignment of AWCs to treatment arms was conducted to ensure balance on observable characteristics. We used principal component analysis (PCA) to construct an index based on distributions of age, caste, language, parents' occupation, and income. AWCs were divided into quintiles of this index within each district and vacancy category, and treatments were assigned in equal proportions within each of these quintiles.
The Chennai district is a dense urban district. As a result, many AWCs in Chennai are in close proximity to one another. Our sampling scheme was designed to minimize potential spillovers from treatment units onto nearby control units. To this end, we first classified AWCs into a group with no other AWCs within a 0.5km radius (singleton AWCs) and a group with at least one AWC within 0.5km (non-singleton AWCs). We randomly assigned 20 singleton AWCs to treatment and the rest to the control group. Random assignment for non-singleton AWCs was clustered by Ward (a geographical administrative unit used in urban areas within a district), so that no treatment AWC shares a Ward with a control AWC. We randomly assigned Wards to treatment to obtain 20 treatment AWCs in the non-singleton group.
In the urban and rural studies, the unit of randomization is the anganwadi center (AWC).
Was the treatment clustered?
Sample size: planned number of clusters
Study 1 - 640 anganwadi centres (AWCs)
Study 2 - 160 AWCs
Sample size: planned number of observations
Sample size (or number of clusters) by treatment arms
- Control: 240 AWCs
- ECCE facilitator: 160 AWCs
- Performance-based pay for anganwadi workers (AWWs): 80 AWCs
- Across-the-board pay increase for AWWs: 80 AWCs
- Supplementary nutrition: 80 AWCs
- Control: 120 AWCs
- Combination of ECCE facilitator, performance-based pay, and across-the-board pay increase for AWWs: 40 AWCs
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
INSTITUTIONAL REVIEW BOARDS (IRBs)
UC Berkeley Committee for Protection of Human Subjects
IRB Approval Date
IRB Approval Number
IFMR Institutional Review Board
IRB Approval Date
UCSD Human Research Protection Program
IRB Approval Date
IRB Approval Number
Post Trial Information
Is the intervention completed?
Is data collection complete?