A recent experimental study in Uganda by FINCA International and WSBI
As a follow-on to the coaching experiment, the BRAC Institute of Governance and Development is supporting our work to identify the factors that shape women’s savings behaviour in the context of family life and gender-based norms. This measurement study will produce a set of indicators that portray the interaction between formal savings, and behavioural influences such as coaching, and WEE. We will pursue two complementary modes of validation. As detailed in Figure 1, the first stage is content validation, in which qualitative methods are used to derive WEE indicators through their relevance with women’s actual savings behaviour in the context of daily life. The next stage, construct validation, will substantiate the structure of the content-validated WEE indicators through their statistical correspondence with the observed saving behaviours of our study participants.
Uncovering Hidden Interactions
The content validation stage, now complete, resulted in more than a hundred potential WEE indicators, which are ready for the construct validation stage.
Our content validation benefited significantly from combining two qualitative research methods: focus group discussions (FGDs) and cognitive interviews (CIs) with the women participants of the RCT. As shown in Figure 2, we used these two methods iteratively. FGDs featured open-ended discussions on the WEE themes and factors related to women’s savings. The CIs then served as feedback sessions in the context of a specific cognitive task – in this case, to evaluate the completeness of the list of possible factors and to think more deeply about their directionalities and possible channels of effect.
Figure 1. Content Validation: Validating the empowerment indicators through their contextual relevance with actual savings behavior among women savers in Uganda
Figure 2. Qualitative Interviews: Sequencing two qualitative methods for deeper feedback
The first set of FGDs unearthed a wide range of savings-related topics. It was immediately apparent that these issues are intricately intertwined with many other hidden layers of hindering and facilitating factors. The consequent CIs then probed deeper. CIs encouraged women to reflect on the causes and effects of already surfaced factors, consider their positive and negative effects, and think of anything else that might be missing from the list.
These two intertwining methods uncovered a complex layer of interactions between WEE measures, product features, and savings behaviours.
Hindrance, Empowerment, or Both?
We observed a deep entanglement of savings motivators and obstacles along multiple directions. When probing for seemingly empowering product features, women expressed contrasting viewpoints and experiences whereby product features can be enabling in one instance but disempowering in another. For example, privacy features were greatly valued and associated with keeping money out of reach from other family members. However, privacy itself was also associated with secretive behaviour and domestic conflict. For other women, low digital literacy levels made it impossible to keep any digital information private, such as phone passwords or account credentials.
Similarly, if in one family, our efforts toward improving the personal savings of women—via personalized coaching or quick reminders—were favourably received, in another family, they were discouraged or punished.
As researchers, these examples require us to be more open-minded with our hypotheses and to allow the ultimate effect of any given feature to be shaped by a wide range of characteristics. For product developers, the lesson is to not take any aspect of a product or service as self-evidently beneficial under all circumstances.
We also observed that simple demographics, socioeconomic variables, or other factors can have mixed effects. For Ugandan women, children are a primary motivation to save, even during the most difficult times. But simultaneously, children are among the top reasons they could not save due to child-related expenses.
Finally, and most confoundingly, WEE measures can be outcomes in one instance and prior conditions in another. Some women need greater independence to become better savers; at the same time, some women need greater savings to become more independent. For another group of women, greater independence hindered any type of meaningful savings because it simply meant that they were burdened with more financial obligations, leaving no money for reserve.
These complexities were crucial learnings during the content validation stage of our study. They resulted in a long and comprehensive list of over a hundred potential variables that have been chosen to capture different possible paths of causation versus outcome. In the next stage of our study, we will use machine learning classification tools to test the indicators’ correspondence with observed savings behaviour. Compared to standard regression models, this approach is better suited for complex causal relationships and finding hidden layers beneath observed behaviours. The end result, we hope, will be a robust set of statistically validated WEE indicators reflecting women’s ability to achieve their savings goals.
Anahit Tevosyan is the Director of Research, at FINCA International, Scott Graham is the Vice President of Research and Data Science, at FINCA International, and Joeri Smits is a Post-Doctoral Research Fellow at the Harvard Kennedy School.