2015
Cairn
Adrien Bouguen et al., « Random Assignment as a Method for Evaluating Public Policy », Économie & prévision, ID : 10670/1.4de847...
Public policy evaluation suffers from a causal inference problem that affects intuitive methods such as simple before-after comparison or beneficiary-non-beneficiary comparison. By comparing results obtained from two randomly selected groups before policy implementation, randomisation provides a rigorous evaluation method. We describe the formalisation of the causal inference problem and discuss how randomisation can, under certain conditions, provide a solution. We outline different evaluation designs, along with their influence on implementation, ethics and statistical accuracy. Lastly, we examine the threats to the validity of results, distinguishing between the threats that also weigh on other evaluation methods and those specific to Randomised Controlled Trials.