Experimental Design
The proposed evaluation is a cluster randomized control trial (cRCT) where clusters are defined as the Units of Services (UDSs). On the recruitment phase, the objective is to receive excess demand for program enrollment: approximately 700 interested UDS in participating in Heal to Grow (in contrast to their actual implementation capacity of approximately 210 UDS). Given this excess demand, treatment randomization will be performed after the inscription stage and before the definition of the logistics of the groups; all ECD workers belonging to the same UDS will be assigned to the same group.
This group of 700 UDS will be randomly assigned into three groups: treatment, control, and replacement. The UDS selected into the treatment group will receive Heal to Grow’s training curricula, while UDS assigned to the control group will not receive any intervention. Replacement UDS will receive the training only if selected to replace a treatment UDS. Otherwise, these UDS will not receive any intervention nor be included in the study sample.
To measure the change in the socioemotional skills of ECD workers, a self-administered questionnaire based on existing tools such as the Five Facets of Mindfulness Questionnaire (FFMQ), the Test of Regulation in and Understanding of Social Situations in Teaching (TRUST) will be used, in addition to some observation tools in controlled situations. For caregivers, another self-administered questionnaire will be applied using instruments such as the Parenting Stress Inventory Short Form (PSI-4-SF). Other instruments that require direct interaction such as the International Development and Early Learning Assessment (IDELA) will be used for children. In the primary and secondary outcomes explanation section, you will find all the instruments that inspired the questionnaire.
After collecting endline information, we will estimate the impact of Heal to Grow through Intention To Treat (ITT) and Local Average Treatment Effect (LATE) estimators through an OLS regression with fixed effects and clustered standard errors at the UDS level.