By Sonja Kelly, Director of Research and Advocacy, and Mehrdad Mirpourian, Senior Data Analyst
Not everything that matters can be measured. The data surrounding financially marginalized groups is sparse. This lack of data limits financial service providers and policymakers’ ability to design for women’s needs. However, many significant issues, like financial exclusion and lack of empowerment can be either directly measured or proxied. With these metrics, we can pursue and track changes over time. In its pursuit of these goals, Women’s World Banking has been working to build ways that we can measure the policy, environment, and social factors that either enable or impede women’s economic empowerment.
In November 2020, our research journey began with funding and technical assistance from the Cloudera Foundation, which has recently merged to become part of the Patrick J. McGovern Foundation. Women’s World Banking set out to consider whether data from the past could predict the future trajectory of women’s economic empowerment.
We are using advanced analytics to test our hypotheses and make projections, but quite simply we were interested in defining the relationship between women’s economic empowerment, financial inclusion, and other development indicators over time. If a country adopts a policy in one year, how might it affect financial inclusion or women’s economic empowerment in future years? Or if it adopts widespread internet connectivity enabling women’s digital financial services access, might they see greater women’s engagement with accounts?
Our first challenge was to list the policies, infrastructure elements, and social norms to look for. Fortunately, Women’s World Banking has a robust set of policy, private-sector, and infrastructure factors that we are already tracking across our markets in the normal course of business. Our research team met with senior leadership in the organization to workshop a list of key enablers that, in an idea world, we could measure over time for nearly every country in the world.
The wish list was lengthy: more than 23 categories as far ranging as access to the technology, asset ownership, digital literacy, geography, income inequality, social and cultural norms, legal discrimination, as well as the overall state of the financial services industry, innovation, and market competitiveness.
The next step was to translate this list of key enablers into actual data, which is where the greatest problems emerged. Without an army of research assistants, we were limited to existing datasets. Country-level data on factors like strength of social network, equity, or fairness in lending, and consumer awareness of services would be impossible to measure. Some data we could approximate. Whether or not a government collected sex-disaggregated data, for example, might be evident in whether or not they report such data to the IMF FAS survey. We would not be able to measure the gender pay gap in every job, but we would be able to approximate it assuming that the labor force gender gap roughly followed pay gaps evident in the formal economy. Some things were easy to measure. Factors such as mobile ownership, access to the internet, and legal constraints to women’s property ownership are all variables contained in the World Development Indicators at the World Bank.
For our “outcome variables,” women’s economic empowerment and financial inclusion, we used the Gender Development Index and the World Bank Global Findex, with datasets providing us rich data across years and countries.
Our final challenge was to structure the data. For data that occurs over time and distance (in this case, over decades and countries), we had to structure our dataset by country, year, then each individual indicator. For missing values, where it made sense, we interpolated the data by assuming that the missing data would follow a straight-line pattern between the adjoining years. We had 300,000 datapoints in all.
Armed with our hypotheses, variables, and structured data, we are now ready to turn to structuring and deploying our data warehouse to create future research possibilities. From there, we will apply machine learning methods, multiple correspondence analysis, and ensemble regression methods to better understand the relationships between these different factors. The final step will be to project what we see into the future, and make some predictions about what women’s financial inclusion and economic empowerment might look like with greater attention toward enablers. We are looking forward to sharing our results as we move forward, and giving you a glimpse of the future, at least as it relates to low-income women’s lives.
