Impact evaluation is important but can be challenging. TASO promotes the use of rigorous experimental and quasi-experimental impact evaluation methodologies as these are often the best way to determine causal inference. However, sometimes these types of impact evaluation raise challenges:
- They are effective in describing a causal link between an intervention and an outcome, but less good at explaining the mechanisms that cause the impact or the conditions under which an impact will occur.
- They require a reasonably large number of cases that can be divided into two or more groups. Cases may be individual students or groups that contain individuals, such as classrooms, schools or neighbourhoods.
- Most types of experiment – and some types of quasi-experiment – require evaluators to be able to change or influence (manipulate) the programme or intervention that is being evaluated. However, this can be difficult or even impossible, perhaps because a programme or intervention is already being delivered and the participants already confirmed or because there are concerns about evaluators influencing eligibility criteria.
- Experimental and quasi-experimental evaluation methodologies can sometimes struggle to account for the complexity of programmes implemented within multifaceted systems where the relationship between the programme/intervention and outcome is not straightforward.
An alternative group of impact evaluation methodologies, sometimes referred to as ‘small n’ impact methodologies, can address some of these challenges:
- They only need a small number of cases or even a single case. The case is understood to be a complex entity in which multiple causes interact. Cases could be individual students or groups of people, such as a class or a school. This can be helpful when a programme or intervention is designed for a small cohort or is being piloted with a small cohort.
- They can ‘unpick’ relationships between causal factors that act together to produce outcomes. In small n methodologies, multiple causes are recognised and the focus of the impact evaluation switches from simple attribution to understanding the contribution of an intervention to a particular outcome. This can be helpful when services are implemented within complex systems.
- They can work with emergent interventions where experimentation and adaptation are ongoing. Generally, experiments and quasi-experiments require a programme or intervention to be fixed before an impact evaluation can be performed. Small n methodologies can, in some instances, be deployed in interventions that are still changing and developing.
- They can sometimes be applied retrospectively. Most experiments and some quasi-experiments need to be implemented at the start of the programme or intervention. Some small n methodologies can be used retrospectively on programmes or interventions that have finished.