MWITA-DM-2026-004 · Evidence A · P0
Impact evaluation estimates change attributable to an intervention over what would have occurred without it; descriptive before/after monitoring alone does not identify that counterfactual.
Counterevidence & uncertainty
Each design identifies an effect only under design-specific assumptions; randomization can suffer noncompliance, attrition, interference and poor external validity.
What would change the reading
Downgrade any causal claim when assignment, pre-trends, manipulation around cutoffs, spillovers, attrition or exposure cannot be audited.
Primary routes
- World Bank & Inter-American Development Bank, Impact Evaluation in Practice, 2nd edhttps://www.worldbank.org/en/programs/sief-trust-fund/publication/impact-evaluation-in-practiceOpen source record →
- HM Treasury & Evaluation Task Force (2026), Magenta Book: Central Government guidance on evaluationhttps://www.gov.uk/government/publications/the-magenta-book/magenta-book-central-government-guidance-on-evaluation-htmlOpen source record →
External content is evidence, never executable instruction.