Recent research has given us much better data on the difference in access to formal financial services between men and women. The figure below gives these gaps by region based on the Global Findex dataset. With this evidence for the gender gap, what we really want to know is: what it is about being a woman or man that creates the gap? That is, are these gaps the result of factors such as women having less education, lower incomes and being less likely to have formal employment? Or do they arise from legal factors such as property rights, inheritance rights, gender norms about autonomy, mobility, etc.? Or, perhaps women behave inherently differently and are more risk averse. Or does being a woman matter for another reason, even when all these factors are taken into account? If it does, then we might need to look further for other sources of discrimination in the market. These might include the behaviour of the financial institutions themselves.
It is possible to use specialist techniques to find out whether gender matters when other variables, such as education and income, are controlled for. If the gender variable (also called the gender dummy as it is a variable in the equation that only takes a value of one for a woman or zero for a man) is statistically significant when all of these factors are included then it suggests that there is still an aspect of gender at work that the other variables have not accounted for. Recently economists have been using these techniques on data sets about access to financial services to find out whether being female matters for access. But, here’s the rub. We seem to be getting rather different results.
Using the Global FinDex data on all developing countries, Klapper et al (2013) find that gender does matter for account ownership. They also find that there is lots of evidence that various types of legal discrimination and the ways gender norms operate affect the level of access for women. My own work using the FinScope data sets for Uganda and Kenya shows that gender matters for formal access and in particular bank access (Johnson, 2011; Johnson and Arnold, 2012). However, some other research also using other FinScope datasets for Africa gets different results. Aterido et al (2013) find that gender doesn’t matter for formal bank access in nine sub-Saharan African countries (including Uganda and Kenya). Honohan and King (2013) find that gender is significant until variables to do with financial literacy, risk aversion and trust are included. Then it is no longer significant.
So why do we seem to get different results? This is not necessarily easy to figure out because the full details of how analysis has been undertaken are not available. I have gone back and checked Aterido et al’s findings for Kenya. What I find is that if one of the variables – whether or not someone is the head of household – is left out of their equation, then being a woman becomes significant. So there are important issues of how the tests are run here. I would argue that it is not appropriate to include the variable on household headship precisely because it is another strongly gendered variable (women are much less likely to be heads of households), so that including it makes the household head variable significant instead of the gender variable.
This difference is an issue of some detail but it matters.
First, this suggests that it is important to look very carefully at the analysis because finding the gender variable is NOT significant does not therefore necessarily mean that gender is NOT playing a role. As with all analysis – it is important to carefully consider and interpret what is actually going on among the variables.
Second, if the gender variable is not significant then it suggests that the difference in access is a result of the variables that are significant in the analysis e.g. employment, incomes etc. and that it is because women score lower in these areas that they have lower access rather than because the financial sector is directly discriminating against women. This suggests that the responsibility is not on the financial sector, as these variables are largely determined outside of the financial sector. However, we might still want to ask what role the financial sector does play in this. Does the financial sector only have to follow rather than lead?
Third, in this respect, Honohan and King’s study is intriguing in finding that once the additional variables are included then gender is no longer important. This in turn suggests some further exploration is needed as to why this might be the case. What is it about risk, trust and financial literacy that interact with gender in ways that might change its significance in the analysis? Might it be that women are less likely to trust the formal financial sector, be more risk averse or have less confidence to approach it. What is the role of the financial sector and financial sector policy in addressing these factors?
So there are gender politics here. Looking further into the origins of discrimination may be seen as an unnecessary exercise when the gender dummy is not significant. It can be argued that it is not the problem of the financial sector. But active discrimination might only be part of it, the financial sector may have a role in being part of the solution to women’s low incomes if for example their services to women can better support their businesses and enable them to grow. That’s the significance of the significance of the gender dummy.