However, for RCTs to convincingly establish causality, they need to be implemented properly. In other words, we have to be convinced that individuals have been truly randomly allocated to treatment and control groups; only then will we have succeeded in constructing an accurate counterfactual scenario (i.e. what would have happened in the absence of a programme). At the same time, we must be able to check for self-selection bias without having to resort to sophisticated econometric techniques that require particular technical expertise.

RCTs may be an attractive methodological option but they are not free from challenges, which can be of technical, ethical and/or practical nature. In academic circles, the chorus of critical voices has become louder (e.g. Deaton, 2010; Scriven, 2008; Barrett and Carter, 2010; Harrison, 2011) arguing that there are threats to the internal and external validity of RCTs. For instance, how much can we really trust the causal claims of RCTs, and how generalisable are their results to other situations and/or individuals?