Strengthening theories of change in women's group interventions to improve learning.

Background: Supporting women's groups is increasingly seen as an important intervention strategy for advancing women's empowerment, economic outcomes, and family health in low- and middle-income countries. We argue that learning from investments in women's group platforms is often limited by the lack of a well-articulated, evaluable theory of change (ToC) developed by those designing the programmes. Methods: We first identify a simple set of steps important to specifying a ToC that is evaluable and supports learning (what could be done). We then propose a framework in which we hope social scientists can find a common starting point (reconciling what could be and is being done). The framework emphasises identifying untested assumptions around pathways for introducing and adopting new knowledge, opportunities, technologies, interventions or implementation approaches, and pathways from group participation to behaviour change. Finally, we apply this framework to a portfolio of 46 women's groups investments made by the Bill and Melinda Gates Foundation between 2005 and 2017 to understand the prevalence and clarity of their grantees' theories of change (some of what is done). Results: The majority of the investment documents reviewed did not make clear the embedded assumptions or hypothesised pathways from decisions to join a group, to women's group participation, to behaviour change and and whether pathways are connected or work independently. Conclusions: We use an example from an actual investment to illustrate how this framework can support accounting for assumptions in the ToC used to guide the evaluation, the testing and measuring of mechanisms assumed to be driving behaviour change and disentangling the effects of implementationfrom context. A ToC for group-based programmes should specify in what capacities the group-based model is essential to the hypothesised pathways of impact vs. its role as an efficient delivery mechanism for programmes that would potentially generate impacts even if delivered directly to individuals. In addition, without fully specifying the motivation for individuals to change behaviour in terms of their risk/return calculus and testing underlying assumptions, we miss an opportunity to better understand the pathways for how the programme influences or fails to influence individuals' health behaviours. However, fully specifying (and measuring) every link in the programme's ToC is not costless. We present suggestions for developing ToCs with testable hypotheses that foster learning about why a women's group intervention achieved or failed to achieve its intended impact.