Interpret findings with caution
▪ Results are associative, not causal: unobserved local factors may remain.
▪ Key variables (e.g. NRW, OCCR, flood data) are partly self-reported and water utility indicators do not cover all aspects of performance.
▪ In nonlinear models, averages may mask heterogeneity, and our heterogeneity assessment does not capture all potential variations.
▪ Generalizability is limited to the Indonesian context and similar performancebased financing mechanisms
Future Study:
▪ The underlying theory can be tested and validated in other setting
▪ Use causal inference methods where possible.
▪ Explore linkages between contexts ~ policy design ~ actors