Can algorithms help level the playing field in SME lending? Evidence from Peru

Women entrepreneurs across the world are creditworthy. They are just not always treated that way. Even when their firms are virtually identical to men’s in size, sector, and performance, women are less likely to receive a loan, and when they do, they tend to receive smaller amounts and worse terms. These differences matter because small and medium-sized enterprises (SMEs) are a key engine of job creation and growth, and differences in access to credit for men and women limit both firm performance and broader economic development.

As digital lenders increasingly use algorithms and alternative data to extend credit to borrowers with limited credit histories, understanding how automated screening affects both financial inclusion and the lending gap between women and men-owned firms has become even more relevant.

But where exactly do these gaps come from? And can changes in how banks make lending decisions help close them?

In a recent study, we examine whether automatic loan approval systems that limit loan officer discretion can reduce gender bias. Using unique data from a large commercial bank in Peru, we find that well‑designed automation can eliminate gender gaps in loan take‑up and loan size—without increasing default risk.
 

Why discretion matters in lending

Audit and lab experiments suggest that part of the gap reflects bias in subjective decision-making by loan officers.

Loan officers often rely on “soft information” gathered through interviews, site visits, and personal impressions. While discretion can be valuable in low‑information environments, it can also open the door to biased judgments, especially in settings where officers have limited experience lending to SMEs or to women entrepreneurs.

This raises an important policy question: Would reducing discretion through automated screening make lending more gender‑neutral?
 

A natural experiment at a Peruvian bank

To answer this question, we partnered with one of Peru’s largest commercial banks during an SME lending pilot conducted in 2012–13. At the time, the bank had a small SME portfolio and was experimenting with new ways to scale up lending.

As part of the pilot, applicants completed a psychometric assessment developed by the Entrepreneurial Finance Lab. The tool generated a credit score based on traits such as problem‑solving ability, integrity, and personality—without using past loan approvals.

The key feature of the pilot was a sharp cutoff score. Applicants above the cutoff were automatically offered a loan, with loan size mechanically tied to their score. Applicants below the cutoff were screened using the bank’s traditional process, which involved a loan officer and significant discretion over loan terms. Because borrowers just above and just below the cutoff are otherwise very similar, this setup allows us to isolate how outcomes change when loan officer discretion is removed.
 

What happens when discretion is removed?

Three findings stand out.

1. First, when loan officers play a central role, clear gender gaps emerge. Below the cutoff, women and men are equally likely to receive a loan offer, but women are 18 percentage points less likely to take out a loan and receive loans that are about 60 percent smaller than those offered to comparable male applicants. This suggests bias operates mainly through loan terms, not outright rejection.
2. Second, these gender gaps disappear entirely under automatic approval. Above the cutoff, where decisions are fully automated, we find no differences between women and men in loan approval, loan take‑up, or loan size. Figure 1 illustrates this finding for loan take-up.
3. Third, eliminating discretion does not come at the cost of worse loan performance. Tracking borrowers for up to two years, we find no increase in default rates for automatically approved loans, for either women or men. If anything, women display slightly lower default rates overall.

READ MORE: https://blogs.worldbank.org/en/developmenttalk/can-algorithms-help-level-the-playing-field-in-sme-lending–evid