Lessons Learned Paper 3
Alex Lewin, Karla Diaz-Ordaz, Chris Bonell, James Hargreaves, and Edoardo Masset
The Centre of Excellence for Development Impact and Learning (CEDIL) has recently funded several studies that use machine learning methods to enhance the inferences made from impact evaluations. These studies focus on assessing the impact of complex development interventions, which can be expected to have impacts in different domains, possibly over an extended period of time. These studiestherefore involve study participants being followed up at multiple time-points after the intervention, and typically collect large numbers of variables at each follow-up. The hope is that machine learning approaches can uncover new insights into the variation in responses to interventions, and possible causal mechanisms, which in turn can highlight potential areas that policy and planning can focus on.
This paper describes these studies using machine learning methods, gives an overview of the common aims and methodological approaches used in impact evaluations, and highlights some lessons and important caveats.
Suggested citation: Lewin, A., Diaz-Ordaz, K., Bonell, C., Hargreaves, J. & Masset, E. (2023) ‘Machine learning for impact evaluation in CEDIL-funded studies: an ex ante lesson learning paper’. CEDIL Lessons Learned Paper 3, CEDIL, Oxford. Available at https://doi.org/10.51744/LLP3
