Is discrimination in access to capital behind gender gaps in business in Ethiopia?

In Ethiopia and across Africa, women have less access to capital, a key component to business success.  While women entrepreneurs are well-represented in the labour force, their businesses tend to be smaller and less profitable. What drives this gender gap in business success and earnings? One hypothesis is that financial providers discriminate against female business owners.

A gender gap, however, is not evidence of active discrimination.  Is it the case that loan officers are discriminating against women, or are they responding to legitimate business factors, such as business profits, that are so far correlated with being a male business owner? In Ethiopia, where gender inequity is relatively high, we often expect gender discrimination to be common and, thus, prevalent in financial markets. But on the other hand, financial providers are experts in their field, there are real stakes in how their portfolio performs, and they have access to a significant amount of information on any business they are evaluating – all factors that push against the likelihood of discrimination.

In an IGC study, we take this question head on. We ask whether being a women business owner causally affects how a financial provider evaluates that business when making capital allocation decisions.

Loan officers allocate capital as part of a natural experimental design

We do this in partnership and in the context of a national business plan competition hosted and organised by the Entrepreneurship Development Institute in Ethiopia. Loan officers were recruited from Ethiopian banks and microfinance institutions (MFIs) to evaluate the competition. This set up allows us to look at a natural context with large and real capital allocation decisions, and with the actual players we are interested in understanding. In this competition, 916 real businesses interested in capital and growth applied, and 84 loan officers who are experts in credit markets, evaluated businesses to determine who would receive the capital prizes in the competition. We designed the application form to capture the same information that is commonly requested in initial loan applications so that the decisions reflected the same type of information loan officers would have access to in an initial loan application.

Since being a woman may be correlated with other business characteristics, we randomise the gender of the applicant when they are being evaluated. We have each business evaluated multiple times by different loan officers, sometimes as “male-owned” and sometimes as “female-owned”. In this way, we can look at whether the same business is treated differently by a loan officer when it is labeled as being owned by a man versus a woman. The loan officers were asked to evaluate the business across three domains: a score that would determine the grant prizes for the competition; their beliefs on how the business would perform in the future (both with and without additional capital), and whether they wanted the applicant’s information forwarded to their own lending institution for potential consideration of a loan.

Capital allocation decisions do not vary by gender

We find that the gender of the business owner did not affect capital allocation decisions by the financial provider, neither for the capital prize in the competition nor for consideration of a loan at their own lending institution. This lack of discrimination is consistent across the spectrum of businesses: we fail to find support for discrimination based on application characteristics (such as marital status) or business characteristics (such as lower quality businesses).

One reason for gender discrimination to possibly exist is that it reflects rational beliefs on how gender predicts relevant outcomes. For example, if women-owned businesses have a higher probability of failing, then loan officers may expect women-owned businesses are more likely to fail and thus discriminate against women-led businesses because of this expectation.

One way to verify our results is to see if they are consistent with the loan officers’ beliefs of how gender predicts business performance. We find that the capital allocation decisions were consistent with beliefs: evaluators’ expectations of future business performance did not vary by gender. We see no differences in expectations of future profit, likelihood of the business surviving, or future assets by gender. This is true even for expectations on how the business would perform if they were to receive additional capital.

Despite the loan officers not responding to the gender of the business owner in their evaluations and decisions, we do see that they respond to critical business characteristics reported in the application, such as profits, assets, and projected growth. We also see that evaluators gave higher scores or were more likely to forward that applicant to their institution to businesses they predicted would have higher future profits and greater likelihood of survival in the future. Moreover, their beliefs were logical – they generally expected higher business performance with capital than without. These results highlight that financial providers were paying attention to the information they were receiving and making thoughtful decisions, though gender was not a factor that affected those decisions.

Predictive value of gender supports lack of rational discrimination

Though the loan officers did not believe gender predicted business performance, it may be the case that in actuality, gender is indeed predictive of business performance. If this is the case, then the lack of discrimination may be costly to lenders. Our next step, therefore, was to understand how the lack of discrimination we observed lines up with actual differences in business performance by gender.

We do this by linking the loan officers’ evaluations to real business performance outcomes of the businesses 18 months after the competition. We develop a machine learning algorithm to select a set of optimal predictors for business profits from the information shown to the financial providers during the initial judging process. The machine learning exercise serves as a clear benchmark for the optimal capital allocation decision by financial providers. If the machine learning algorithm had selected gender as an important predictor, this would have suggested that the loan officers were wrong in their lack of gender discrimination. Just like the financial providers, however, the machine learning algorithm did not use gender as a predictor. Our machine learning exercise highlights that the importance of gender in predicting business performance is dwarfed by the other detailed information provided in the application (and in a loan application more generally).

Conclusion

In our context, we find that financial providers do not believe that gender predicts business performance, and in accordance with these beliefs, they do not discriminate when making capital allocation decisions. Moreover, our machine learning results suggest that the predictive value of gender is not large enough to justify rational discrimination. In other words, loan officers did not discriminate against women businesses, and they were right not to do so.

These results are suggestive that gender discrimination in access to capital is not a key contributor to gender gaps in business performance and growth, highlighting the importance of future research in revealing the factors that might explain these gaps. Our context is most similar to the earlier stages of a loan or grant process, but capital requests often involve further steps and interactions, during which gender discrimination could become a factor. The literature suggests that observable differences in gender are important, but that they cannot always explain the entire gap. A final key consideration is gender differences in the demand for capital: it may be that the choice to request capital may itself be an important consideration for the gender gaps observed in access to finance.