Was Fintech's Biggest Lending "Innovation" A Willingness To Serve Riskier Borrowers By Charging Higher Rates?
Was Fintech's Biggest Lending "Innovation" A Willingness To Serve Riskier Borrowers By Charging Higher Rates?
Goldman Looking To Offload Apple Partnership; FT Partners July Update
Hey all, Jason here.
I’m writing this intro parked in front of two fans, as, somehow, it has gone from ~60F (15.5C) with a strong windstorm that shut down parts of the Netherlands on Wednesday to 90F (32C) this weekend.
This week also saw Meta’s Twitter-clone Threads launch. While others (Bluesky, Mastodon) have tried to attract die-hard Twitter users, Threads feels like the first serious threat to Twitter, apart from the self-inflicted ones. If you’re checking out Threads, you can find me there at @mikulaja.
Existing subscriber? Please consider supporting this newsletter by upgrading to a paid subscription. New here? Subscribe to get Fintech Business Weekly each Sunday:
The Fraud Fighters Manual—Get A Copy 📖
Sponsored content: The Fraud Fighters Manual ⚔️ is now available for free download. In six exciting chapters across 168 pages, authors from top companies like Brex, Mercury, and Lithic share the best practices of fraud prevention—offering practical strategies and real-world examples to empower fellow professionals in the fight against fraud 💪 Get a copy!
Was Fintech's Biggest Lending "Innovation" A Willingness To Serve Riskier Borrowers By Charging Higher Rates?
Fintech leaders and investors often like to describe companies in the space as “democratizing” financial services and “expanding access and inclusion.” In consumer lending, these lofty aspirations are said to be achieved through innovations like “alternative data,” “proprietary underwriting,” and AI.
I myself have deployed these talking points; during my time at one fintech lender, our “elevator pitch” was that we used “big data and machine learning to underwrite and serve borrowers that big banks won’t.”
But, as is often the case in fintech, the reality is more complicated and progress more incremental than the story told in a press release or glowing media profile.
A recently published academic paper, “FinTech Lending with LowTech Pricing,” analyzes fintech underwriting and pricing in an attempt to determine whether or not the “big data” and “advanced analytics” many in the space tout actually represent a break with and improvement from traditional credit scoring models, like FICO.
For its analysis, the authors draw on a data set that covers about 70% of “fintech loans” made in the US from 2014-2020 by lenders like LendingClub, Upstart, and Avant.
They filter to include only loans that include the necessary data points for their analysis and that are comparable (eg, same loan term of three years.)
The result was a data set of some 2.3 million loans, with an average interest rate of 16.29%, average loan amount of $11,898 made to borrowers with a median income of $62,500 and median FICO score of 677. The analysis groups FICO scores over 660 as “prime” vs. those under 660 as “non-prime.”
Fintech Lenders Still Heavily Reliant on FICO, Cross-Subsidize From High-Risk to Lower-Risk Borrowers
The authors conclude that fintech lenders’ approach to pricing is “rather simplistic and inefficient.”
That conclusion is underpinned by three main arguments:
that fintech lenders are still heavily reliant on conventional credit scores (FICO)
that fintech loans to non-prime borrowers are less responsive to expected risk than to prime borrowers
and that loan-level returns generally decrease with the expected risk for prime and non-prime fintech loans
The authors’ analysis shows “a striking jump” in borrowers’ interest rates “at FICO 660—the threshold for prime borrower status, with the average interest rate on the loans shifting from about 17% in the 660–664 FICO bin to 25% in the 655–659 FICO bin, increasing by 8 pp.”
There are not similar discontinuities (abrupt jumps) for FICO scores between 660-850, and the discontinuity for scores below 660 is far less.
There is a moderate jump in delinquency at the FICO 660 cutoff; the authors note this isn’t entirely unanticipated, given the significant jump in pricing at the same point in the distribution — the higher interest rates can contribute to causing the higher rates of delinquency.
The authors further note that adverse selection and moral hazard may contribute to higher default rates below FICO 660, but argue that such impacts are a result of the higher interest rates. There is nothing “uniquely risky about borrowers who happen to be at FICO 659 at origination instead of 660,” they say.
The authors estimate that fintech borrowers of similar risk profile but falling below the 660 FICO cutoff pay interest rates that average 7.3 pp higher than those above 660.
Over the life of the loan, that would mean paying an additional $800 in interest on average.
Why Are Fintech Lenders Still Tethered to FICO?
The authors point to several possible reasons for the pricing behavior around the FICO 660 cutoff.
Most bank lenders will only consider borrowers with credit scores above 660. For fintech lenders, that means more competition for prime borrowers, but less competition for non-prime ones.
The authors argue the lack of bank lending in the non-prime segment gives more market power to fintechs, helping to explain the pricing jump at FICO 660.
The authors also point to regulatory requirements for banks, including FDIC risk management policies and capital requirements — though it’s worth noting that most fintech personal loans are made in partnership with banks, to which these policies apply.
The FDIC examination manual states:
Nonprime lending should only be conducted by institutions that have a clear understanding of the business and its inherent risks and have determined these risks to be acceptable and controllable given the institution’s staff, financial condition, size, and level of capital support. In addition, nonprime lending should only be conducted within a comprehensive lending program that employs strong risk management practices to identify, measure, monitor, and control the elevated risks inherent in this activity.
This heightened level of scrutiny may serve as a disincentive for banks to lend to non-prime borrowers and may drive higher costs for banks that do.
Further, the authors argue that the Basel III requirement that banks set aside more capital for riskier loans also serves as a disincentive to lend to riskier borrowers (I would add the recent shift to the CECL accounting standard, which requires banks to estimate total losses over the life of a loan upfront, rather than realizing actual incurred losses.)
While bank lenders make use of inexpensive deposit funding, most fintech lenders rely on originate-to-distribute models and wholesale funding channels. The higher capital costs can make it more difficult for fintech lenders to compete with banks in prime segments vs. non-prime.
For fintech loans that are securitized and sold, SEC guidelines mandate lenders provide loan-level information in a standardized format, such that investors can evaluate the risk of the underlying assets (loans.)
Even if fintech lenders are using “proprietary” models, FICO is typically the common benchmark when packaging loans for securitization and sale to investors.
Cross-Subsidies Are Likely An Intentional Pricing Strategy
The authors argue that fintech loans are mis-priced to risk such that “non-prime” borrowers subsidize “prime” borrowers: those with FICO scores sub-660 are overcharged on a risk-adjusted basis.
They further demonstrate that within the prime and non-prime segments, there is a cross-subsidy from lower- to higher-risk borrowers.
This is visible in loan-level returns. Generally, returns, measured here by the internal rate of return (IRR), should increase with the riskiness of the loan (measured here by predicted delinquency in a 12 month period.)
But that is not what the authors find for the fintech loans they examined:
The authors conclude that “[t]he results are not justified by underlying risk, suggesting large deviations from risk-based pricing at the market segment level.”
The result is that fintech borrowers with prime FICO scores are paying less than they “should” (on a risk-adjusted basis), thanks to a subsidy from non-prime borrowers who have fewer options to choose from.
The authors’ analysis belies some fintechs’ claims of using sophisticated technology and data analytics to better underwrite borrowers or to magically identify an “invisible prime” segment through “alternative data.”
Fintech lenders’ primary “innovation” was serving borrowers banks didn’t want by charging higher interest rates
While these narratives are appealing to investors and journalists, a far simpler conclusion is much more convincing: fintech lenders’ primary “innovation” was serving borrowers banks didn’t want by charging higher interest rates.
If you’re a fintech lender, this pricing strategy the authors describe isn’t irrational.
The pricing distortions and cross-subsidies from non-prime to prime make sense if you think about it from a (non-bank) fintech lender’s perspective: to compete with banks to win prime and super-prime borrowers, you need to offer comparable (or, ideally, lower) rates — but, with structurally higher funding costs, that lost revenue has to be made up somewhere.
The non-prime borrowers fintechs serve have fewer choices, historically being excluded by hard cuts in banks’ credit policies. With more pricing power in this segment, fintech lenders can “overcharge” these borrowers vs. their risk profiles to compensate for under-pricing prime borrowers.
Still, There Has Been Real Innovation in Fintech Lending…
This analysis looks only at three-year personal loans, ignoring other product categories that have seen more meaningful innovation.
Open banking-enabled cashflow-based underwriting, while not a panacea, can help thin- and no-file applicants gain access to credit. Fintechs offering credit cards, like X1, Tomo Credit, and Petal, leverage such models.
Novel approaches to repayment, like “payroll-linked” lending, can also help reduce risk of default and enable lenders to approve those who otherwise might not be or charge lower rates. Again, this isn’t a cure-all; debiting repayment directly from a borrower’s payroll moves a creditor to the “top of the stack,” but doesn’t prevent a user from becoming overly indebted or falling behind on other key obligations, like rent or utilities.
Emerging product categories, like earned wage access (EWA), also represent net-new offerings that are inarguably better than other forms of small-dollar, short-term credit they may displace, like bank overdrafts and payday loans.
While the authors of this research paper paint a disheartening picture about underwriting and pricing in the fintech personal loan market, there are areas where innovation is driving access and inclusion and improved pricing for borrowers.
Goldman Looking To Offload Apple As Consumer Ambitions Continue To Unravel
Goldman Sachs’ messaging on its consumer business — Marcus and GreenSky but also partnerships with Apple and GM — has been unusually chaotic.
When the bank initially began pulling back from its Marcus business unit amid a firm-wide reorganization, Goldman publicly insisted it was still committed to GreenSky and Apple, which, together with GM and transaction banking, formed the newly-created “Platform Solutions” division.
But, by April, the firm was “considering strategic options” for GreenSky, and, last month, the business unit’s fate looked to be sealed. Goldman is likely to get but a fraction of the $2.24 billion in paid for GreenSky, in a deal that closed just 16 months ago.
In last October’s earnings call, Goldman CEO David Solomon remained bullish on the Apple partnership, emphasizing that opportunity remained to expand beyond the credit card partnership.
And less than three months ago, Apple launched a savings account within its wallet, powered by Goldman.
But now, the Wall Street Journal is reporting Goldman is looking for the exit and is in early talks to potentially offload the partnership to American Express.
Such a move may actually makes sense for Goldman.
It ceased writing personal loans through Marcus last year. It’s working to sell GreenSky. It ceased bidding on new co-brand card deals.
Meaning Apple Card and, to a much smaller extent GM, would become the bank’s only consumer credit exposure — a loan book that totaled about $15 billion at the end of Q1 2023.
The $15 billion in credit cards loans is actually less than Goldman holds in commercial real estate ($29 billion), residential real estate ($22 billion), or securities-based lending ($16 billion) — credit cards loans comprised about 8% of Goldman’s loans at the end of Q1.
And those credit card loans come with operational and servicing costs — and regulatory headaches. Goldman’s credit card business is reportedly under investigation by both the CFPB and the Federal Reserve.
A change in account standards hasn’t helped. CECL, or current expected credit losses, forces lenders to account for expected losses upfront, rather than recording them as they’re incurred.
This is all to say, while the timing and messaging around Goldman’s continued retreat from its consumer business appears disjointed, the choice isn’t entirely unexpected.
Even if Goldman exits the Apple deal, it seems likely the bank will continue its own stand-alone savings account, whether under the current “Marcus” name or not. The savings business, while “boring,” was always the most financially important part of Goldman’s consumer push, by reducing the bank’s reliance on expensive wholesale funding in favor of (relatively) cheap, sticky consumer deposits.
Does American Express Make Sense As An Apple Partner?
It’s not entirely clear to me that American Express is a natural fit for the Apple partnership.
While American Express can support a more robust product suite than Goldman, and, like Apple, is a “premium” brand, American Express may blanche at the credit box Apple negotiated with Goldman: about a quarter of Goldman’s card loans are sub-660 FICO.
The no fee positioning — no annual fee, but also no late fees — combined with the 2-3% cashback rewards impacts the portfolio economics and may make it less attractive to potential acquirers.
Finally, Apple is, unsurprisingly, not an easy partner to work with.
With a $3 trillion market cap and an estimated 56% market share of mobile phones in the US, even Goldman Sachs seems to have lacked negotiating power in this relationship.
It’s not clear that those dynamics change with American Express; though, perhaps, the discipline to walk away if it can’t get an economically viable deal does.
After all, American Express has experience losing a key partner when deal terms don’t make sense: after being the co-brand partner and only credit card accepted at Costco stores for 15 years, the issuer survived the breakup.
FT Partners Monthly: The New Normal?
Excluding Stripe’s nearly $7 billion round in May, we’ve now seen nearly a year of $3-5 billion per month in fintech VC fundraising.
But, as a reminder, this is still above the monthly average pre-pandemic: per CB Insights data, in 2019, fintech startups raised a total of $33.9 billion, or an average of $2.8 billion in fintech funding per month.
So, perhaps ~$3 billion in fintech deals per month is the new (old) normal?
Other Good Reads
The Checking Account War Is Over (And The Fintechs Have Won) (Ron Shevlin/Forbes)
Goldman Sachs Was The Sucker (Fintech Takes)
Lots of Banks Means Lots of Bank Mergers (Money Stuff)
Everyone Wants Interest on Their Deposits. That’s Bad for Main Street Banks. (WSJ)
About Fintech Business Weekly
Looking to work with me in any of the following areas? Email me.
Fintech advising & consulting
Sponsoring this newsletter
News tip or story suggestion — reach me on Signal at +1-316-512-1571
Early stage startup looking to raise equity or debt capital