Abstract
Abstract
This study examines how the quality of budget institutions affects public spending efficiency in education, health, and infrastructure. Using Public Expenditure and Financial Accountability (PEFA) indicators to measure fiscal governance and non-parametric efficiency scores as outcomes, the analysis proceeds in two stages. XGBoost models first identify which governance dimensions most strongly predict efficiency: Predictability and Control in Budget Execution (Pillar V) consistently ranks highest across sectors, with effects concentrated in the low-to-mid range of the PEFA distribution. Since income and broader institutional quality confound these associations, the second stage applies Generalized Random Forests (GRF) to estimate Conditional Average Treatment Effects under a continuous dose-response framework. After conditioning on key confounders, a one-unit improvement in key PEFA pillar scores raises spending efficiency by two to nine percentage points depending on sector — with effects smooth, exhibiting diminishing returns at higher governance levels, and systematically larger in lower-income and institutionally weaker countries. Efficiency gains of this magnitude imply 0.6 to 1 additional learning-adjusted years of schooling and up to fourteen percentage-point increases in infrastructure quality indices
JEL Classification: H50, H51, H52, H54, H61, I1, I2, E62