Capital is key for entrepreneurship, business growth, and productivity, yet woman-owned businesses earn lower profits and are less likely to obtain formal financing than their male-owned counterparts. Observable differences between men and women entrepreneurs could only explain a portion of these differences by gender, suggesting that another factor such as gender discrimination may be inhibiting the success of female entrepreneurs. The impacts of policy solutions that reduce discrimination, such as gender blinding, are understudied, even though they could increase gender equity and potentially improve efficiency in the allocation of capital. In addition, there is evidence that algorithms can be used to help avoid human prejudice in decisions prone to biases. On the other hand, automated decision-making using available data for predictions could replicate or exacerbate underlying inequalities.
The research team outlines four methods of evaluating business plans to understand how each method could affect the equity and efficiency of capital allocation decisions. To test both human and algorithm-based loan decision-making, the research team planned to compare decisions made by (i) loan officers with applicants’ gender information; (ii) loan officers that receive applicants in which the gender of the applicant has been randomized; (iii) an algorithm designed solely to predict business success and (iv) an algorithm designed to predict business success which incorporates discrimination-aware methods. The study was designed to assess how well each method performed in using information from business plans to predict future business growth.
The researchers recruited 84 financial providers, spanning 10 financial institutions, to serve as judges to evaluate 916 real businesses that applied to an actual business plan competition in Ethiopia. The gender of the business owner was randomly assigned to be shown as either male or female on applications provided to the judges. Each business was evaluated multiple times, and each financial provider evaluated multiple businesses, for a total of over 3,600 evaluations. This allowed researchers to compare the judges’ evaluations of each business plan depending on the apparent gender of the applicant, allowing them to causally identify whether financial providers discriminate against female entrepreneurs.
The researchers found no evidence that the judges discriminated against female-owned businesses. Neither their evaluation scores nor the likelihood of recommending a business for a loan from their financial institution changed based on the randomly assigned gender on the application. Judges’ assessments of future business performance were also unrelated to gender. Their assessments were accurate: a follow-up survey 18 months after the competition found business survival and profits did not differ by the true gender of the business owner. Given judges did not discriminate among applicants by gender, and true business performance did not differ by gender, the researchers were unable to use this context to compare algorithmic approaches to overcoming gender discrimination. The researchers suggest future research on gender-related constraints to applying for capital could be a useful step to investigating the underlying causes of gender gaps in business performance.
Copyright 2022. All Rights Reserved
Design & Dev by Wonderland Collective