Abstract:
|
Encouragement designs are frequently used when the treatment cannot be enforced. By design, encouragements always entail the complication of non-compliance and they can give rise to a variety of mechanisms, especially when they are assigned at the cluster level. Social interactions, for example, can result in spillover effects. Disentangling the effect of encouragement through spillover effects and other mechanisms from that through the enhancement of the treatment would give a better insight into the intervention and it could be crucial for improving the program and planning the scale-up. We capitalize on the principal stratification framework to define stratum-specific causal effects, showing how these effects are related to the decomposition commonly used in the literature and provide alternative assumptions under which an extrapolation across principal strata allows to disentangle the effects. Estimation can be performed within the Bayesian paradigm, using hierarchical models to account for clustering.
|