Evidence of randomisation bias in a large-scale social experiment: The case of ERA

We set out a theoretical framework for the systematic consideration of ‘randomisation bias’, estimate the causal impact of randomisation on participation patterns in an actual trial, and propose a non-experimental way of assessing the extent to which the experimental impacts are representative of the impacts that would have been experienced by the study sample that would have been obtained in the absence of random assignment. We also extend our estimator to deal with binary outcomes and to account for selective survey non-response, and explore partial and point identification of the parameter of interest under alternative assumptions on the selection process.