A study by the Philadelphia Federal Reserve, published in February 2021, is extremely interesting (1). It summarizes the latest fintech research and policy discussion. I learned a lot and highly recommend reading Section 2 and Section 10, because those sections relate to a board favorite, Upstart. Here’s some excerpts and a bit of commentary.
Alternative Data
Section 2 peels back the curtain on what could be considered ‘alternative data’ when a loan originator conducts background checks on a potential borrower. This alternative data includes “whether the borrower posts a picture, the number of words used in the listing text descriptions, [and] friend endorsements”. Proprietary data from an ecommerce site in Germany also offered information about a borrower’s digital footprint - “whether individuals’ emails contain their real name, whether they make purchases at nighttime, and the number of typing mistakes…”
From later on in this section:
“Previous research shows that borrowers’ soft information, including personal characteristics such as race, age, and beauty (Ravina, 2018), social capital (Lin and Pusianinen, 2018; Hasan, He, and Lu, 2019), characteristics of listing text (Iyer, Khwaja, Luttmer, and Shue, 2016), hometown (Lin and Viswanathan, 2016), social network (e.g., friend endorsements as in Lin, Prabhala, and Viswanathan, 2013; Freedman and Jin, 2017) affects lenders’ decisions in terms of exante loan pricing and loan amount.”
Wow! Pretty interesting (scary?) stuff. I bet Upstart uses some of this ‘soft information’ in their own models. Should we be worried about regulatory action? What happens when the public gets wind of the fact that you might get a lower loan rate because of how beautiful you are or how many friends you have? That regulatory compliance letter will expire sometime…but on the other hand, regulators may take a non-confrontational stance as financial inclusion has demonstrably alleviated poverty (for more information, see Section 3).
I’d appreciate some thoughts on this from Saul board followers. Some more information on regulatory risk is below.
Other Information on Regulatory Risk
“Prior to the global financial crisis in 2008, financial innovation was viewed very positively, resulting in a laissez-faire and deregulatory approach to financial regulations. After the crisis, fintech and data-driven financial services providers profoundly challenged the current regulatory paradigm. Financial regulators are seeking to balance the competing objectives of promoting innovations, financial stability, and consumer protection.”
“With rapid growth in usage of big data and AI/ML, the current wave of fintech has created informational and regulatory gaps and loopholes that need to be closed. Rapid advance in the technologies have also increased the uncertainties and difficulty in evaluating the impact of new technologies on consumers, the market, and financial systems overall. Jagtiani and John (2018) provide an overview of fintech innovations and regulatory considerations, especially discussion around consumer protection in response to fintech growth.”
Peter here - apparently 70% of the rise in fintech loans is due to regulatory arbitrage, rather than superior technology or lower costs (2). I couldn’t hunt down this fact (the author actually pointed to the Philadelphia study under review here), but it is a sobering one if true.
Recently in the news, loanDepot allegedly cut corners and processed thousands of loans without required documents such as employment and income verifications, apparently in a bid to drum up business before their IPO (2, 3).
From an NYT article on the subject: “LoanDepot is in the vanguard of a group of online upstarts that use technology to speed up and simplify mortgage loans. Last year, it originated nearly 300,000 — twice the number it did a year earlier — and was the nation’s fourth-largest mortgage provider in dollars lent, according to iEmergent, which tracks industry data” (3).
Again, should we be worried about regulatory risk when it comes to Upstart, particularly due to some bad actors such as loanDepot?
Notably, Upstart may have competition when it comes to nontraditional loan approvals:
“Fintech giants Alibaba and WeChat Pay Points Credit by Tencent in China have built their new credit scoring system based on alternative data they collect from nontraditional sources, including social media, online shopping, payment applications, cell phone accounts, and others.”
On Credit Scoring using AI
“However, about 26 million American consumers have thin credit files or do not have bank accounts (unbanked); thus, they do not have FICO scores because of an insufficient credit history. More recently, there has been a breakthrough in which consumers’ default probability could be estimated not only from their official credit history or credit ratings but rather from more complex statistical methods using AI and ML techniques, along with (nontraditional) alternative data. These big data and complex algorithms have been rapidly adopted by fintech lenders to overcome the limitations of traditional models and data in evaluating borrowers’ credit risk and their ability to pay back loans.”
Nothing really new here, but interesting that there is consensus that nontraditional alternative data (coupled with AI learning) could yield demonstrably smarter approvals. This section also states that “the use of alternative data by LendingClub has allowed some belowprime consumers to receive credit at a much lower cost”, and that LendingClub has become more efficient compared to their peers, but this information is from 2016; we now know from the KBRA data that Upstart has overtaken LendingClub on this front.
Online Banking
“More recently, advances in financial technology also improved digitization in payments — both retail and wholesale payments. The international experience shows that the relationship between the traditional banking sector and fintech companies could be a mix of cooperation and competition […] Several fintech firms have also been providing white-label technological services for their bank partners.”
Sounds a whole lot like the Upstart business model. This section points out that online banking “reduces the bank’s operating costs and improves its profitability”, so I’m sure online banking as a trend will continue and banks will continue to look to partner with fintechs that can help them in this process.
Other Interesting Quotes
“So much data has been collected and monetized in recent years, leading to an interesting fintech mantra: “Data is the new currency.” The vast amount of data and the fast and complex computing algorithms have become key factors that drive innovations in recent years, and billions of consumers around the globe have benefited from these changes.”
“It remains unclear, however, how loans originated by fintech firms will perform relative to traditional loans in an extreme environment (severely stressed scenario), and they are going through a real test now during the COVID-19 crisis. It is also unclear how the complex algorithms with alternative data, which worked well earlier, would continue to perform in the new landscape (after the COVID-19 crisis). The AI models may need to be retrained with new data to reflect the “new normal.” In the meantime, fintech lenders have a role to play in the background (white label services) to assist small banks in screening and processing a large number of loan applications under the Coronavirus Aid, Relief, and Economic Security Act (CARES Act) to support small businesses.”
As an UPST shareholder, I’ve become curious about this space. It is becoming clear to me at least that we are in uncharted territory and the old ways of doing things are starting to break up. At the crest of that wave lies huge opportunity, but I remain aware that the wave can also crash and do serious harm, if we do not remain watchful.
PeteHay
Long UPST