Re-evaluating the early child stimulation programme in Bangladesh: evidence from the partial identification approach

ABSTRACT This study complements the previous evaluation study of the Early Child Stimulation programme conducted in Bangladesh. Despite a rigorously designed randomised control trial, the presence of non-compliers made it impossible to point-identify the average treatment effect (ATE) on targeted outcomes without additional strong identification assumptions. This study provides new evidence through the partial identification approach, which estimates the ATE bound with weak but credible assumptions. The results show that the ATE bounds include the local average treatment effects and we do not find strong evidence to suggest the ATE among compliers is greatly different from that among the entire population.