Today, Dave Girouard, CEO and Co-Founder, Upstart, spoke in front of the House Task Force on Financial Technology in Washington, D.C. The topic of the hearing was “Examining the Use of Alternative Data in Underwriting and Credit Scoring to Expand Access to Credit.”
Since the beginning of Upstart, we worked to expand access to credit by using alternative data, artificial intelligence, and machine learning algorithms to determine a borrower’s creditworthiness. In our early days, we proactively engaged with the Consumer Financial Protection Bureau (CFPB), the most important consumer protection regulator, to ensure that our methods have a positive impact on consumers, resulting in the industry’s first and only No Action Letter.
You can read Dave’s testimony below-
Chairman Lynch, Ranking Member Hill, and Members of the Task Force on Financial Technology, thank you for the opportunity to participate in today’s conversation.
My name is Dave Girouard, co-founder and CEO of Upstart, which is a leading artificial intelligence (“AI”) lending platform. I founded Upstart more than 7 years ago in order to improve access to affordable credit. In the last five years, almost $4 billion in bank-quality consumer loans have been originated on our platform using a model that combines alternative data with AI and machine learning algorithms to determine a borrower’s creditworthiness.
Concerns about fairness in algorithmic lending, particularly in the use of alternative data, are well founded. As a company focused entirely on reducing the price of credit for the American consumer, fairness is an issue we care about deeply.
In our early days at Upstart, we conducted a retroactive study with a large credit bureau and uncovered a jarring pair of statistics: just 45% of Americans have access to bank-quality credit, yet 83% of Americans have never actually defaulted on a loan. That’s not what we would call fair lending.
The FICO score was introduced in 1989 and has since become the default way banks judge a loan applicant. But in reality, FICO is extremely limited in its ability to predict credit performance because it’s narrow in scope and inherently backward looking.
At Upstart, we decided to use modern technology and data science to find more ways to prove that consumers are indeed creditworthy – to bridge that “45% versus 83%” gap. We believe that consumers are more than their credit scores. And by going beyond the FICO score, and including a wide variety of other information such as a consumer’s employment history and educational background, we’ve built a significantly more accurate credit model.
While most people believe a better credit model means saying no to more applicants, the truth is just the opposite: Because Upstart’s model is more accurate, we have significantly higher approval rates and lower interest rates than a traditional model.
But we also understood that consumer protection laws weren’t to be taken lightly. Thus we proactively met with the appropriate regulator – the Consumer Financial Protection Bureau (CFPB) before launching our lending platform.
After several years of good faith efforts between Upstart and the CFPB to determine the proper way to measure bias, we demonstrated that our AI-driven model doesn’t result in unlawful “disparate impact” against protected classes of consumers. Because AI models change and improve over time, we developed automated tests with the regulator’s input in order to provide reports on the impact of our credit decisions across underserved groups on a quarterly basis. We have been providing this information to the CFPB for the last 18 months.
Moreover, we were also able to report to CFPB that our AI-based system improved access to affordable credit. Specifically:
- Our model approves 27% more consumers and lowers interest rates by 3.57 percentage points, compared to a traditional lending model
- For near-prime consumers (620-660 FICO) our model approves 95% more consumers and reduces interest rates by 5.42 percentage points compared to a traditional model
- Upstart’s model provides higher approval rates and lower interest rates for every traditionally underserved demographic
That’s the type of consumer benefit we should all get excited about. In September 2017, Upstart received the first-ever “No Action” letter from the CFPB, recognizing that Upstart’s platform improves access to affordable credit without introducing unlawful bias.
The concern that the use of alternative data and algorithmic decisioning can replicate or even amplify human bias in lending is well-founded. However, in Upstart’s experience, the fair lending laws enacted in the 1970s and the substance of fair lending regulation enforcement—that is, monitoring and testing the impact on actual consumers who apply for loans—translates very well to the AI-driven world of today.
But in reality, the path we walked at Upstart is insufficient to create a robust and competitive market that will maximize financial inclusion and credit access. In our early days at Upstart, we couldn’t know for certain whether our model would be biased. It wasn’t until loans were originated that we were able to demonstrate that our platform was fair. As an early-stage startup, this was a risk worth taking, but it’s not a risk a large bank would have considered.
If broader and deeper financial inclusion among American consumers is important to this committee, it’s worth considering rule making or legislation that will provide some form of limited sandbox for model development and testing.
By combining regulatory support for model innovation with rigorous and standardized testing, we can ensure that we don’t forego the clear and obvious consumer benefits that AI and alternative data can offer to the American consumer.