Impact Measurement: Foundations and Frontiers

The reference handbook for designing RCTs in the development context. It is written by practitioners for practitioners, and the chapters on threats to validity and ethical considerations are required reading even for evaluators using non-experimental methods.

The book that put the credibility revolution in the hands of general readers. Read for the case studies — bednets, deworming, microfinance — but also for the implicit argument that small, well-designed experiments can yield insights that grand theories cannot. The companion to almost any modern MEL course.

A philosopher of science and a policy practitioner ask: what does it actually take to know that a programme that worked in one context will work in another? Their answer — that "external validity" requires substantive theory, not just rigorous estimates — has become the standard critique of pure replication-based evidence cultures.

The most important critical engagement with RCTs from inside development economics. Deaton, who later won the Nobel for work on consumption and welfare, argues that randomisation does not solve the deeper problem of choosing what to study and why. Pair with the Banerjee–Duflo response.

Pritchett's argument is sharper than Deaton's: that the most important questions in development — about state capability, growth, structural transformation — are not amenable to randomisation, and that the field's enthusiasm for RCTs has crowded out work on bigger questions. Read with charity, even if you disagree.

The single best running commentary on impact evaluation methods. Practitioner-oriented, technically deep when it needs to be, and unafraid of self-criticism. The best way to keep up with what the credibility-revolution generation is currently arguing about.

A polemical critique of technocratic development that doubles as a methodological argument: when evaluators don't take political economy and rights seriously, their findings are at best partial and at worst misleading. Easterly is provocative, but the underlying claim — that "what works" cannot be answered without "for whom" — has aged well.

The Problem-Driven Iterative Adaptation framework reframes evaluation as continuous learning rather than judgement. For programmes operating in complex political environments — which is most of them — it offers a more honest model than the linear logframe.

The canonical text for realist evaluation, which asks "what works for whom in what circumstances and why?" rather than "does it work?". A useful counterweight to averages-based evaluation cultures, especially for complex social programmes.

Funder, repository, and methodological hub for impact evaluations across the development sector. The systematic review database and the evidence gap maps are the most useful single resources for understanding what is and isn't known about a given intervention.

A method-and-approach encyclopaedia maintained by practitioners. Especially strong on non-experimental methods and on the politics of evaluation. The "rainbow framework" is a useful starting point for designing an evaluation system from scratch.

When you want to hear evaluators talk about real implementation challenges — how they handled spillovers, how they dealt with attrition, what surprised them — the J-PAL podcast network is unmatched. Particularly useful for graduate students between coursework and first fieldwork.