Paper short abstract:

We attempt to meta-analyse the impact of modern energy interventions on educational outcomes in developing countries. We show a novel applicative meta-analytic method to identify associated publication bias and further used their meta-data to trace the sources of inherent heterogeneity.

Paper long abstract:

Energy poverty has far-reaching consequences on development outcomes and is closely interlinked with the achievement of several critical SDGs. Education is arguably considered one of those areas where energy intervention can have a deep and significant impact. The piqued interest in this area has proliferated the literature, but the causal effect of energy interventions on educational outcomes in developing countries remains largely inconclusive. We attempt to harmonise this literature by meta-analysing 38 studies using robust Bayesian model averaging methods. We examine the impact of four energy interventions on nine educational outcomes. Incorporating the literature beliefs, we find positive benefits of rural energy interventions on educational outcomes with varying effects on study hours and test scores. Whilst the gain in study hours (2 minutes/day) is practically negligible for the grid and off-grid interventions, the negative effect on test scores is more pronounced by off-grid (solar) (6.2% decline). This may be because of the uncertainty associated with solar interventions. The effect of grid electrification on literacy rate and that of off-grid electrification on educational attainment remain inconclusive. We find that selection bias characterises rural grid electrification and improved cooking stoves literature, while small study bias characterises the literature on solar electrification and the use of TV or radio. We further observe that studies that apply causal inference methods have more homogenised findings. Our findings indicate that grid electrification and improved cooking stoves have the most positive benefits for education. We also recommend causal inference econometric methods for studies examining these impacts.