What Aggregate Measures Conceal: Mixed‐Methods Evidence on Women's Empowerment, Gender Targeting, and Couples Training in Malawi

aising women's empowerment through economic inclusion and social protection programmes has proved challenging. Women's access to resources improves but across many studies measured decision-making agency barely moves. Part of the weakness in evidence may be measurement as composite indices may conceal change that occurs unevenly across domains. Within the context of a multi‐faceted anti‐poverty, or Graduation, programme in Malawi, we investigate the women's empowerment impacts of gender targeting, and the additive effect of a 12‐month couples economic cooperation and empowerment training. We combine the results from a randomised control trial with 2,934 women from poor married households with findings from qualitative research in which researchers collect unprompted accounts of change from the participants' perspective to challenge and substantiate the quantitative findings. Female resources rise in all three treatment arms. The aggregate decision making index moves only where the couples training is added, and only marginally, but beneath it financial decision‐making rises in both female‐targeted arms while male targeting reduces women's say over productive decisions, the very activities the programme creates. Qualitatively, women across arms describe more collaborative decision‐making over income generating activities, the essential realm of the Graduation programme, more than over budgeting, and budgeting moves mainly where the couples training addressed it directly; the driver they name unprompted is the programme's gender messaging rather than the cash. We show that participant‐led and unprompted qualitative evidence can identify which components of composite indices merit attention, independently of researcher priors.