This article discusses my concerns about SoFi, a co I follow. It goes straight to the heart of the issues, thus certain company background is desirable to fully comprehend the analysis. If SoFi is new to you, this article might not offer much insight.
Disclaim: I have no long or short position of SoFi as of writing.
Introduction
SoFi SOFI 0.00%↑ is a bank offering lending and financial services to retail consumers, and tech platforms to enterprise clients.
I raised a few concerns 6 months ago, after its 1Q23 earnings. Concerns were:
An early sign of a whole loan sale freeze within its growing AFS portfolio
buck-the-trend Fair Value adjustment, citing high-quality loans supported by third-party valuation firms
This is a revisit after its 3Q23 earnings on Oct 30.
Summary
This article discusses SoFi’s continued sales (near) freeze, AFS loan designation, recent sample size (<0.1%) whole loan sale, and buck-the-trend fair value adjustment.
I also explain why I call all of that its “capital optimization” battle. I present math to highlight these concerns and raise questions that are largely unanswered through its public filings.
Sales (near) Freeze, and Growing Origination in AFS
SoFi’s growth since 1Q22 is largely driven by its personal loan growth. Its personal loan book grew from $3Bn in 1Q22 to ~$15Bn in 3Q23, a 5x in 7 quarters.
SoFi’s personal loans have been designated as AFS (Available-for-sale) based on its business model that originates and sells loans in 6 to 9 months. It deviated from that model, and largely froze the whole loan sale since 4Q22, while it continues to ramp up loan origination from $2.5Bn to ~$4Bn each of the last 4 qtr. That divergence is presented in the chart below.
Note: The whole loan Sales amount excludes securitization deals (presented as a golden bar).
AFS Loans are intended to be sold prior to maturity, often within a year, while HTM loans are intended to hold til maturity. An intent and ability criteria are performed during designation to confirm its operational intent and ability to execute its plan (sell or hold).
SoFi’s personal loans’ initial AFS designation was consistent with its advertised model to hold loans for 6 to 9 months prior to sale. It raised analysts’ attention in its 4Q22 cc when it dropped its loan wholesale amount to $60Mn (<1% PL). The management responded it made more economic sense to hold than sell.
Circumstance changes (fed rate raise, tightened credit market), plan/model changes. ASC guidance offers flexibility to keep existing loan designation (AFS in this case) unchanged. All fine.
However, from 1Q to 3Q23, SoFi continued to originate over $10Bn personal loans marked as AFS, while selling ~$65Mn whole loans (<1% of its entire portfolio).
My #1 question: What were the intent and ability criteria (according to ASC 320) in 1Q/2Q/3Q23 to qualify these personal loans as AFS, and what would the criteria (4Q23 and beyond) be for new personal loans designated as AFS, if any?
The only criterion I observed is that SoFi sells whole loans when it can hit the 105% level since 4Q22. That certainly isn’t a qualified intent criterion.
For example, in 3Q23, what could be a fair intent and ability criteria to designate its $3.8Bn loans as AFS, while the Bank is 9 months into its whole loan sale (near) freeze?
Bulls might point to securitization deals (~$1Bn since 4Q22) as evidence that SoFi sold more at attractive prices. In my opinion, securitization deals don’t transfer the most critical credit risk and don’t truly reflect underlying assets’ market value due to various leveraging effects.
One might ask what’s the big deal AFS or HTM?
For this analysis, the primary differences are how (projected/actual) loan losses (gains) and fair value are reflected in accounting and its impact on regulatory capital.
HTM uses CECL, which takes the expected loss up front (and a hit to capital), while AFS takes no upfront capital hit. Here is a nice chart illustrated by BankingIp
Simply put, AFS is more capital-friendly. But AFS or HTM designations are intent and ability-driven, banks certainly should not choose based on capital friendliness.
While AFS enjoys upfront profit, and delayed capital loss, it is subject to fair value adjustments, often due to external factors that a bank has few controls.
These FV adjustments, unlike HTM unrealized losses or gains, flow through income statements (as AOCI), and more critically impact regulatory capital (there are opt-out caveats but less applicable in this case). Let’s examine SoFi’s records.
Buck-the-trend Fair Value Adjustment Records
SoFi’s quarterly FVAdj varied from ~$100Mn in 1Q22 to $569Mn in 3Q23. Most increase was due to its personal loan book growth. Its FVadj as a percentage of unpaid loan principal grew from 3.1% in 1Q22 to 4% in 3Q23, a truly buck-the-trend adjustment (compared to bank peers).
Most peers in the space had negative fair value adjustment during the same period, not a surprise given we are in an increasingly tightened credit market (pressure on mark-to-market value), and rising rate env (higher discount rate), both are Fair Value headwinds.
SoFi bulls point to its (small) loan sale at 104-105% range which confirms its loan portfolio's high quality and thus justifies its strong fair value adjustments.
SoFi bears shrug it off as immaterial amounts (or in the form of securitization that the sold price doesn’t truly reflect loan quality).
That contention intensified in 3Q23.
$15Mn Loans Sold at 105.1%
In the 3Q23 conference call, Lapointe said it sold loans at a 105% level in 3Q, another $100Mn in 4Q, and more ($2Bn) in the pipeline at a similar level.
However, its 3Q23 10Q shows its 105.1% level sale was sold at par, and recorded a 5.1% servicing fee (Banks usually charge ~1% or less annually to service loans). One of my posts summarized my findings.
DDI had a follow-up chat with SoFi’s CFO and covered the topic. In short, according to Lapointe, 100% + 5% reflected the buyer’s payment flow preference.
The setup is quite intriguing to me. SoFi sold $15Mn (~0.1% of its entire loan book). During the earnings call, Lapointe mentioned selling at 105.1% a few times but didn’t disclose the term or amount. If I have to guess, SoFi might have fought quite an intense battle to defend its sales price (e.g. 105% level).
I went over DDI’s chat with Lapointe, and have one question: What is the weighted average remaining duration (not average loan life) of the $15Mn loan sold?
Asking because a longer remaining duration would increase its total service fee (assuming the same annual fee rate), and could partially justify its unusually large servicing fee, and quite a clever way to present the loan sale deal at a higher price level.
I recall the CFO mentioned $15Mn loan sold by and large matches the overall loan portfolio characteristics. If that’s the case, $15Mn shall not have a particularly long remaining duration, but that’s one data I would like to get.
More Fair Value Measurement
Some might counter, the 105% sales level is just one of the references. SoFi also uses level 3 unobservable inputs for fair value measurement, as documented in its 10Qs.
Key inputs are Discount Rate, Annual default rate, Coupon rate, and prepayment rate.
Let us take a look at the Discount Rate: The chart shows the discount rate SoFi used for FV measurement since 1Q22, and how it compared to the average Fed Rate and US2Y during the same period (as a benchmark).
Note as the Fed rate and US2Y rose significantly since 1Q22, SoFi maintained its discount rate relatively steady, as a result, the spread narrowed from 4.4% in 1Q22 to 1.5% in 3Q23.
My #3 question: what was the rationale behind the steady discount rate (used in FV measurement) despite both the Fed rate and US2Y’s rapid rise?
Some suggested that the steady discount rate might take interest rate hedging into consideration. I find that odd as hedging effects are already reflected in the financials, and as hedging is not included in loan sales, it shall not be included in FV measurement. I’m open to hearing different views.
annual default rate: quite steady at 4.6%, and NCO has been rising from 1% in 1Q22 to ~3.5% in 3Q23, still under its guided default rate. One caveat: SoFi doesn’t report vintage year loss data in its earnings / 10Q / 10K. Given its rapid loan book ramp-up, NCO% for the entire portfolio might appear better than seasoned loan stats (e.g. Y2022 vintage).
I’m aware that some 3rd parties (e.g. MorningStar) provide some visibility into SoFi’s vintage loss data, but that’s one area I would like to have more direct disclosure from SoFi.
Its prepayment moved slightly upwards (~100bps) since 1Q22, a small headwind to FV. Its coupon rate improved a bit over 100bps since 4Q22, a tailwind by and large offsets a moderate discount rate rise.
Bulls might say, enough fair value! why would SoFi fight so hard to defend its Fair Value? Isn’t it just a few hundred million dollars, anyway? Well, now we enter Capital Adequacy.
Capital Adequacy Ratio
In short, fair value adjustment impacts capital, thus capital adequacy ratios directly.
Let us take a step back and examine SoFi's regulatory capital management.
SoFi’s RWA (Risk-weighted assets) grew 3x in 7 quarters, while its total capital remain steady at the $3.3-3.5B range. Subsequently, its capital adequacy ratio (CAR) declined from 30%+ in 1Q22 to 14.5% in 3Q23.
Under Basel II, the minimum CAR is 8%. Under Basel III, it is revised to 10.5%.
Most banks maintain an operating buffer (usually ~2%) above the minimum ratio. I would expect SoFi to manage its CAR% as such, between 12-12.5%. At 14.5% as of 3Q23, there is 2 to 2.5% extra space to go, which translates to about $4B +/- 0.5B RWA to add to its loan book.
At its current ~$2B Personal Loan net addition each quarter, (e.g. originated $3.9Bn and settled $1.8B in 3Q23), SoFi has 2 more quarters of loan growth, at its current capital level.
Now let us circle back to the Fair Value adjustment. Its current risk-based capital is $3.35Bn, which includes ~$0.57B PL FV adjustment. Without that, its capital would be $2.8Bn, and its CAR% would be 12.2% today, within the 12-12.5% guidance.
In other words, the next 2 quarters’ projected loan book growth is only possible from the excessive $0.5Bn+ capital from its FV adjustment.
Net income is another source of increased capital to fund future loan growth. We haven’t talked about it much as SoFi hasn’t reached positive net income. In its latest interview on 11/15, Lapointe mentioned SoFi targets to achieve $300M to $500M net income, a solid capital source for future growth if SoFi can realize that. Something to watch closely.
Connecting the Dots
The article touches upon a few moving pieces, from whole loan sale freeze, continued AFS loan designation, sample size sale at above-the-market price, to buck-the-trend fair value adjustment.
SoFi certainly fought some unconventional battles. As I reflect on them, I find a common theme that connects all the dots: SoFi is fighting a “capital optimization” battle.
I will leave it to readers to make their judgment on how durable and defendable SoFi’s approaches are, under the scrutiny of risk management, guidance (esp. ASC), and regulations.
Parting Thoughts
If you are still with me, thank you and I appreciate it.
My work is tailored towards curious-minded investors, who dig deep and make their own decisions. My goal is to be a part of your research process, provide beyond-the-headline analysis, and offer insights/models to project what the future holds.
If that is of interest to you, please subscribe and share.