Does Automation Eliminate Bias from Small-Business Lending?

ab122016-bias.jpg

Editor at Large

Problem: Though women own a third of U.S. small businesses, they get less than 20% of the loans made through the bank-centric, federally backed system that distributes a lot of credit to such firms.

Reasons: Perhaps a combination of biased loan decisions, fear of rejection by traditional lenders and the unique characterstics of women-owned businesses.

Emerging solution: Online lending, because it's less intimidating than staring down a loan officer at the local branch, and it holds the promise of more objective credit analyses.

Indeed, the early evidence suggests that online applications and automated decisions are adding fairness to the process of granting small-business loans to female entrepreneurs.

Online portal SmartBiz Loans analyzes Small Business Administration loan applications on its member banks' behalf and sends them the loans that best match their underwriting policies. Members approve 95% of the loans SmartBiz sends them.

Thirty percent of the loans made through SmartBiz go to women-owned businesses. That figure nearly matches their market share: women own 9.8 million small businesses, about a third of the U.S. total.

The situation is similar at another small-business-loan portal called Fundera, which accepts loan applications on behalf of traditional SBA lenders and alternative lenders including Funding Circle, Bond Street, Kabbage and OnDeck Capital. One in four Fundera applicants are women, and 25% of loan recipients are women.

Traditional bankers should take note, because only 18% of SBA 504 and 7(a) loans were made to women-owned businesses in fiscal year 2016, the agency said.

Moreover, only 5.5% of women business owners obtain commercial loans of any kind from banks or other financial institutions to start or acquire their businesses, compared with 11.4% of male business owners, the SBA said.

So why are women business owners far more likely to obtain a loan by applying online to SmarBiz Loans– which is simply embedding banks' underwriting policies in its platform –than by going into a branch and talking to a human?

"It's all online," CEO Evan Singer said. "I can't speak to somebody going into a branch and what a loan officer would say in the branch, because we don't do that. These are people who are coming online and there's no bias in the process."

The top three SBA-lending banks declined requests for interviews. Wells Fargo, the largest lender, offered this statement: "We want to make every responsible loan we can for women business owners, and we base all loan decisions on the creditworthiness of the borrower. We are proud of the fact that since 1995 Wells Fargo has loaned more than $55 billion to women business owners."

Special Factors
While it would be easy to assume old-school bankers are refusing to make lady loans, Erin Andrew, associate administrator for the SBA's Office of Capital Access, who until recently ran the agency's women-owned business program, explained that several forces distinguish women business owners from their male counterparts.

One is, they don't ask for loans as often. "The challenge for women-owned businesses is 90% don't have employees, they're individually run, independently owned businesses," Andrew said. "Generally if you have employees, you're more likely looking to leverage capital, you're growing, you're accessing loans, you're having an impact on your community."

More than a third – 34.3% – of surveyed female entrepreneurs say they do not require startup financing, while 19.5% of men say this, according to the National Women's Business Council. Its survey also found women are more likely to use personal savings to begin a new business.

When they do seek credit, it tends to be in smaller amounts. The SBA's microloan program, for loans of less than $50,000, is heavily used by women, who take out 51% of them; these loans average $13,000 to $17,000.

There's also a theory that bad news begets more bad news. "A lot of studies indicate women are more discouraged because they read these statistics that they're less likely to secure funding," said Meredith Wood, vice president of content at Fundera.

Women tend to start their businesses in industries like healthcare and social assistance, technical services, administrative and support, and retail trade, which tend to be less capital intensive, Andrew said.

Another factor is risk appetite. "We sometimes say women are risk averse, and I've heard other people say women are more risk realistic," Andrew said. "The ability and interest in taking on debt is important."

Bias-Free Algorithms
Still, the disparities in loans to women- and men-owned small business appear to shrink when the loan applications and decisions are automated. This suggests the online process eliminates unconscious bias in personal interactions.

"Everone has bias," Andrew said. "It's not intentional a lot of times."

With online lending, "You're looking at the numbers, you're not looking at me," Andrew said. "When lending is blind, you see the borrower based on what they have to bring to the table, what the business looks like – you don't see them based on gender."

At Enova, which makes small-business loans under the brands Headway Capital and Business Backer, the algorithms that make underwriting decisions are not even told whether the borrower is a man or a woman.

"Online lenders and data-driven lenders are able to avoid or eliminate systematic bias in their decisioning," said Joe DeCosmo, chief analytics officer at Enova. "We're focused on the data that tells us whether they can afford the loan or not. We want to lend to those who can afford the loan. We control the data that goes in; then you can keep the decision objective."

The company conducts annual fair lending reviews of its systems to make sure that its algorithms aren't creating unintended bias, DeCosmo said. "To the extent you find something, you can then change it. You can alter the algorithm or the data and how you decision on it," he said.

Other Benefits
Online portals are said to be less intimidating than face-to-face meetings with a banker. And lessening the intimidation factor could be crucial because an SBA study found that women often did not pursue an SBA loan out of fear of being denied, not because of problems with their financial profiles.

Fundera's Wood said that online lenders lend to a wider credit tier, too. "Maybe [online borrowers'] credit score isn't as solid or they're not quite profitable or something of the sort that would make them not bankable in the first place," she said. A Credit Sesame study earlier this year found that the average credit score for men is 630, compared with 621 for women. This is partly due to their typically lower salaries.

"If you think about [female entrepreneurs] going to a bank where it's already notoriously hard to secure financing and going in there with a credit score that's lower, they're going to have a tougher time," Wood said.

Online platforms also make loans more accessible. "Prior to it moving more online, you'd talk to your local banker, and if your local banker didn't offer smaller-dollar loans or SBA loans, it was hard to find capital," Andrew said. "Entities that are doing smaller-dollar loans or even mission-based lenders are now online and people can find them even if they're not across the street or in their local neighborhood."

There are cases, of course, when automated decisions reinforce biased programming. In one recent example, Chinese researchers tried to teach computers to predict women's personality traits from their photos. They trained the computer to categorize women's faces using positive words such as "sweet, endearing, elegance, tender, caring, cute," or negative words such as "pretentious, pompous, indifferent, coquettish." When the researchers compared the results for 3,954 images with the choices of 22 young Chinese men, the computers agreed with their labels 80% of the time.

At least one online lender judges loan applications by factors such as whether applicants type in all caps, which may correlate to their education levels, but education is not generally considered a predictor of creditworthiness and treating it as such may be unfair. Amazon Lending's algorithms cherry-pick the best-performing merchants on Amazon.com and offer them loans based on their sales.

But for women small-business owners, the online lending movement appears to be opening doors.

Editor at Large Penny Crosman welcomes feedback at penny.crosman@sourcemedia.com.

For reprint and licensing requests for this article, click here.
Bank technology Community banking Analytics Commercial lending Small business
MORE FROM AMERICAN BANKER