EEF view: What are the implications for education research?
RCTs are usually the best way of measuring the efficacy of a particular teaching and learning approach, or intervention, in a single study. RCTs have a “control” group, which means that there is a comparison between a group of pupils receiving an intervention and a group of pupils that do not (or receive a different intervention). This is especially important in education, as pupils are consistently progressing, so studies that look at progress before and after an intervention won’t tell you whether more progress might have been made without it.
The second key element of an RCT is that allocation between treatment and control is random. This can be contrasted to other study types that do have a comparison or control group but rely on existing differences in the education system.
Take the example of comparing outcome data between schools with longer or shorter hours. The problem is that the schools that have longer hours may also differ in other important ways: maybe they are all part of the same multi-academy trust, or maybe they all teach in similar ways. Randomisation gets around this problem.
When we combine headline impact findings from RCTs with details on cost and implementation, they can give us powerful information to shape decisions about what is likely to work in schools in the future.
However, RCTs are not always appropriate or possible. Some changes that schools might make would be impossible to randomise - for example, how many schools would be willing to have the hours of the school day determined by a coin flip? This means that while RCTs may be the best way to establish causal estimates of impact, there will be many situations where they are not available.
