A control group is the set of people, households, firms or areas in an evaluation that do not receive the programme being studied, such as a microloan, savings account or cash transfer, and serve as the comparison against which its effects are measured. The group that does receive the programme is the treatment group.
In a randomized controlled trial, units are assigned to the two groups by chance, so on average they are alike in every respect except the programme, and differences in outcomes can be attributed to it. Where randomisation is not possible, evaluators build comparison groups through matching or other statistical methods, which are more exposed to selection bias. A well-known application is a set of randomised microcredit evaluations published together in 2015, which found modest average effects and no clear transformation of household incomes. Problems include contamination, when control units obtain similar services elsewhere, and attrition. Control groups underpin impact assessment and the measurement of attribution.
Ref 1270000017