Using Data and Machine Learning to Close the Gender Gap in DR Congo

The Democratic Republic of the Congo (DRC) ranks 151 out of 156 countries on the World Economic Forum’s global gender gap index.  As we discussed in a previous blog, social bias and economic discrimination hurt women’s livelihoods. With more than half the day spent in unpaid household labor, women find few business opportunities outside small, unprofitable trading activities. Even when they operate more lucrative ventures like wholesale trade, their profits are four times lower than men. 

In a country where less than 14 percent of women own financial accounts, access to finance is a profound constraint because it denies women the resources to compete. At FINCA DRC, we recently surveyed 2,800 financial consumers to find out more about women’s barriers to financial access in the country.

Women are not homogeneous

At the topline, our data shows that focusing on gender as the defining characteristic of a female client is exactly the wrong approach.   Although they suffer discrimination across the board, women are not a homogeneous segment. In fact, women’s financial needs and behaviors are strongly influenced by social factors in ways that men’s are not. 

These factors come into play at the earliest stage of financial inclusion: product awareness. In our survey, women’s ability to name more than one or two financial products from different categories (loans, savings, etc.) was about half that of men. However, there were significant differences among women depending on demographic factors that had negligible effect on men. For example, rural women are worse off than urban women, whereas men are the same regardless where they live. Women also have lower levels of awareness in certain age brackets, especially among youth (<25) and women over 55, while men’s product recall is almost identical at every stage of life.   

Insights from machine learning 

To further analyze these nuances that affect women’s financial inclusion, we used machine learning to create a model of consumer struggling. The goal was to predict, based on different social factors, whether it would be easy or hard for a woman to find the right product. 

Our Machine Learning Model

We used a predictive model to classify female respondents according to the personal characteristics that predict difficulty finding the right financial product. In technical terms, the model executes a supervised learning algorithm. In a series of repeated steps, it tests every possible variable to identify those factors that have the strongest relationship with this difficulty, breaking down sub-groups within each of the previously identified categories. The process repeats until all statistically significant factors within the dataset have been identified.

The results in the figure below show that limited awareness was the strongest factor in predicting that a woman would have difficulty finding the right product. Information is a precious asset for female consumers. Something as simple as knowing her options can be a gamechanger in finding the right financial product.

The model also produced a deep description of how each factor is shaped by the “hidden” elements beneath it. Of all the variables in our data, education was the strongest predictor.  Among women with limited awareness, 95 percent of those with secondary or lower schooling struggle to find the right product. Advanced education cuts struggling roughly by half.

Going one level deeper, employment type enriches the model further. Among women with lower educations and limited product awareness, 100 percent of those who are self-employed struggle to find the right product. Formal employment improves the situation somewhat, though not to the same degree as education.

Look beyond gender

Female-intentioned outreach means looking beyond gender.  Limited product awareness is a problem of specific women for whom this constraint is most debilitating.  We know that a self-employed woman with low education is almost guaranteed to have very limited product awareness, and this will undermine her ability to find the right financial product.

As FinEquity puts it, closing the gender gap means deconstructing the monolith and digging into sub-segments where women’s specific needs and constraints can be addressed.