The Shift in Aid Evaluation: How Contribution Analysis Complements the Search for Attribution

Professor Tommy Bang Sup Keum, Tech University of Korea

A Methodological Turn in Development Evaluation

Most official development assistance (ODA) interventions unfold through multiple actors, multi-year timelines, and policy environments that resist clean experimental control. Randomized controlled trials and difference-in-differences designs remain powerful for answering narrow attribution questions — did this specific intervention cause that specific outcome? — but they struggle with the questions practitioners actually need answered: why did a change occur, through what pathway, and under what contextual conditions? This limitation is sharpest in governance reform, institutional capacity-building, and policy-dialogue programs, where no credible control group can ever be constructed.

Against this backdrop, the World Bank's Independent Evaluation Group (IEG) released guidance on Contribution Analysis in March 2026 that has drawn considerable attention across the evaluation community. The framework organizes the method into six steps across two phases: theory-building (steps 1–4) and theory-testing (steps 5–6).

What Contribution Analysis Actually Does

Contribution analysis does not try to prove that a program was the sole cause of an outcome. Instead, it builds a case — grounded in multiple, converging lines of evidence — that the program plausibly contributed to an observed change. In practice, the method proceeds through roughly the following steps:

1. Specify the theory of change the evaluation will examine.
2. Articulate the causal pathway through which that theory is expected to operate.
3. Catalogue plausible rival explanations — other actors, external shocks, parallel interventions.
4. Gather evidence to test whether each link in the theory of change actually held.
5. Weigh the program's contribution against the rival explanations.
6. Construct a contribution story and state explicitly what remains uncertain.

The approach starts from the premise that causation is typically multiple, not singular. Its emphasis falls on refining the theory of change and assembling an evidence-based, plausible causal narrative rather than estimating a precise effect size. What it sacrifices in statistical precision, it gains in being able to answer whether an intervention mattered, how and why it mattered, and how context shaped the result — questions that experimental designs are often structurally unable to address.

From Aggregate Studies to Localised Analysis

A recent comprehensive review of ODA-effectiveness scholarship published between 2000 and 2025 identifies a clear methodological migration: away from macro, cross-country aggregate comparisons and toward analysis localised to specific sectors and specific project types. The dominant question is shifting from "does aid cause growth?" to "did this particular health program, in this particular context, work — and through what mechanism?"

This is not an isolated academic trend. It surfaces consistently at Japan's annual ODA Evaluation Workshop (the 20th edition convened in November 2025) and has been formally embedded in the UK government's official evaluation methodology — the 2026 revision of the Magenta Book, Annex A — which now treats theory-based analytical methods as a core pillar rather than a supplementary technique.

Implications for Evaluation Practice

This methodological shift carries direct consequences for how terminal and mid-term ODA evaluations should be designed and conducted.

• Theory of change and PDM design matter earlier, not later. Unless a program articulates an explicit theory of change and a well-specified performance management matrix (PDM/ePDM) at inception, there will be no evidentiary foundation for contribution analysis at closeout.
• Rival explanations need to be identified in advance. Cataloguing, at the design stage, the ways an observed change could plausibly be explained by factors other than the program itself — policy shifts, parallel donor interventions, macroeconomic cycles — substantially strengthens the credibility of the eventual evaluation.
• Qualitative and quantitative evidence must be integrated by design. Because contribution analysis relies on triangulation rather than statistical estimation, evaluation matrices need to be built from the outset to combine interviews, document review, field observation, and administrative data systematically.

The broader significance is this: development evaluation is not becoming less rigorous by moving away from attribution — it is becoming more honest about what complex interventions can and cannot demonstrate. Rather than searching for a single cause in an irreducibly complex reality, the field is converging on a paradigm that makes multiple, plausible contributions visible and transparent.

Sources - World Bank IEG — Quality Guidance for Contribution Analysis in Practice - UK Government — Magenta Book Annex A: Analytical methods for use within an evaluation (2026) - Monitoring & Evaluation Studio — Theory-Based Evaluation Quick Reference - Japan MOFA — 20th ODA Evaluation Workshop Report (Nov. 2025)