Behavioral science has a strong track record in international development. Interventions grounded in behavior change have improved health outcomes, increased savings and raised service uptake, often at low cost. However, the evidence base remains thin and uneven.
In a new paper published in the Behavioral Economics Guide 2026, we argue that this has resulted from the field consistently asking the wrong questions, in the wrong ways and in the wrong institutional conditions. The paper, co-authored by BIT with the UAE’s Behavioral Science Group and the World Bank’s Mind, Behavior, and Development Unit (eMBeD), sets out a research agenda to fix each in turn.
Ask the right questions
Much of behavioral science is built on narrow, WEIRD samples, and development settings are too often treated as test-beds for existing frameworks rather than sources of new theory. Some mechanisms do travel well, for example, interventions that reduce frictions, such as simplifications or reminders. Others, like social norms messaging, depend on cultural constructs and, therefore, their impact varies across settings.
However, the field tends to default to cognitive explanations without exploring how norms or mental models differ across settings. In rural Kenya, for example, there are strong informal claims on earnings from family and communities. Researchers found that women were willing to forgo investment returns to keep earnings hidden.
Use the right tools
Even where foundational behavioural concepts hold, methods can undermine the evidence. The first principle is to diagnose before you design: in development settings, the binding constraint may be structural or social rather than cognitive, and standard research methods often exclude the very populations policy needs to reach. New technologies, such as AI-assisted qualitative analysis, provide significant opportunities for higher quality data at speed, and at lower cost.
The second is to design for instability. Implementation is routinely disrupted by turnover, seasonality and exogenous shocks. For example, one of our trials in Bangladesh was interrupted by a cholera outbreak, a heatwave, Eid closures and a cyclone. The goal is to build instability into the design from the outset.
Create the right institutional conditions
Evidence doesn’t yield change on its own; it is highly dependent on the institutions that surround it. Funding flows reward positive results over null results. Global North researchers typically control the design, budgets and publication while local partners contribute the contextual knowledge and fieldwork that makes those studies possible.
Governments default to non-disclosure, meaning the most policy-relevant studies are buried. These are structural conditions that shape what evidence gets produced and how it is used to influence policy.
You can read the full paper in the Behavioral Economics Guide 2026, available to download now.
