SoFi stock appears to be the target of a coordinated short seller attack. By this, we mean that we’ve heard similar ideas using similar language that doesn’t seem like it emerged from independent thinkers. Language that is slightly “off” is being repeated as though it has been arrived at independently.

Short attacks like these are increasingly common in today’s investor landscape, which is metabolizing a significant influx of retail investors that haven’t yet determined whose investment advice can be trusted. So, if you are an investor, you should probably be able to identify a short attack and know how to independently evaluate the merits of a short thesis without falling for incendiary rhetoric that is likely more a function of the speaker’s investment position than their analytical qualifications or convictions.

Robert and I have both spent the most lucrative parts of our careers as short sellers with John Paulson during the global financial crisis (GFC). So, we know a good short when we see it. Short sellers can benefit markets because their presence reduces irrational competition and leads to improved risk balances in the market.

Short selling isn’t for the faint of heart, though. That’s why we prefer pair trading companies that may have similarities in economic drivers but that diverge in management’s relative conservatism and most importantly the use of accounting practices that may make the companies appear similar today but that are likely to diverge as the accounting or rhetoric wears thin against the real trends in the underlying business fundamentals.

I want to disclose up front that we are about to “talk our own book,” which every investor does. In this case, our book is that we are long SOFI and short UPST. We’ve already supported at length our short position in UPST.

To quickly summarize our position, we are short UPST because we believe that it has weaponized its accounting latitude to mask both the vulnerability of its business model and the deterioration in its real economic prospects. Admittedly it is both puzzling and frustrating that any of the criticisms that are being hurled at SoFi, are orders of magnitude truer for UPST. If you doubt this, please see our reports on UPST. So, even if you believe that the claims against SoFi have some merit, you would still do well to pair trade it against UPST.

Because we already hold these positions you should read on with skepticism. Furthermore, I hold a significant personal position in SoFi that I earned while I was employed by SoFi when it was of going public and getting a bank license. So, you should be especially skeptical of me.

As a skeptic you should realize that since we already hold these positions, as with anyone who has a book to “talk,” we may be prone to tell the truth selectively or to even exaggerate the importance of one truth over another. In the worst case, we may repeat something that we’ve heard even knowing it not to be true because it works in our book’s favor.

Since our largest “long” position is the reputation that we are developing as an investment firm, we do not view engaging in such tactics as long-term profitable. But we invite you to judge for yourselves.

The first question that you need to ask, I’ve already answered in the affirmative: If someone has a blog or an X account that is talking up or down a particular stock, rest assured, they are talking their own book. If they convince you to join their trade, you are helping them toward a profitable exit from that trade. In fact, you may be left with a hot potato following their exit.

The second question that you need to ask is this: Is the person that is talking their book qualified to have such a strong opinion about this or that particular accounting issue?

Gain-on-Sale & Fair Value Accounting

Some of our most successful career shorts have been related to companies that “creatively” used gain-on-sale or fair value accounting to boost the appearance of performance or to mask a deterioration in the underlying economics of a business.

Many years ago, when I was working in policy at the FDIC (and, no, I never urinated off the roof of the Seidman Center—though on a clear chilly night this might have been preferable to the examiner bootcamp that I was enduring while there), I wrote about the dangers of gain-on-sale accounting in interpreting earnings trends and capital adequacy at lenders. This was shortly after the OG[1] accounting standard for gain-on-sale accounting, SFAS 125, first came out.

While I was on the Financial Accounting Standards Board Advisory Committee following the Global Financial Crisis, I debated the relative benefits of mixed attribute versus fair value accounting with academics, practitioners like John Gerspach, former CFO of Citigroup, and the board members who ultimately wrote the standards that companies currently use in their financial statement presentation. So, I’d like to believe that if I haven’t influenced in some small way the bank capital standards and the accounting principles that govern the debated accounting treatment, I at least understand them better than average.

One of the more popular blog reports on SoFi that I’ve seen making the rounds talks about SoFi’s use of “available for sale” accounting for loans and discusses whether SoFi has the “intent and ability” to continue using this treatment. The analyst then talks about the devastation that would occur if it became clear that SoFi no longer had the intent or the ability to account for these loans as “available” for sale and would have to restate the balance sheet and income statement at amortized cost. This would result in de-recognizing the income stream from their high-yielding loans but not de-recognizing the credit mark that is part of its accumulated fair value adjustment.

This would be an extremely interesting observation if the premise were true. But SoFi’s loans aren’t available for sale, they are currently held for sale and accounted for at fair value.

There are at least two important differences between these two designations. If an asset is available for sale, gains and losses associated with a change in the fair value of these assets flow through the capital accounts (via AOCI) but not through earnings.

In a practice that I have always been against, bank regulators reverse gains and losses on assets held for sale (for all but the big four banks) when calculating bank capital adequacy ratios.

Assets held for sale are accounted for at fair value, and changes in fair value flow through earnings in the period that they occur and, because they flow through the income statement, they immediately affect capital ratios and, therefore, capital adequacy.

Secondly, and importantly, whether a financial asset is held for sale or investment fair value accounting is discretionary. Whether the loans sit in a line item on the balance sheet that is titled “Loans Held for Sale,” or “Loans Held at Fair Value,” doesn’t change the accounting for the loans regardless of the “intent” or management’s plans for the loans.

I don’t want to make too much out of this misunderstanding of the accounting designations by the blogger because there are still important issues related to fair value accounting that affect a number of lenders including both SOFI and UPST. I do want to point out though that the error in understanding the basics of the accounting treatment may help you answer our second question above.

The more important error in this blog has to do with the level of concern raised about “intent and ability” and the potential that SoFi might need to reverse its accumulated fair value adjustment and then have to book a CECL reserve against the loans.

Auditors prefer fair value accounting

Companies have the option to adopt fair value accounting for assets regardless of their intent or ability to sell the assets. In most cases, companies have to document this designation when the loan is originated without option to later change the accounting treatment because the standard setters don’t want companies cherry picking assets to mark while not recognizing negative marks on only part of the same book of business.

Generally speaking, the proof of intent is most relevant when a company designates an asset as held for investment or held to maturity. In this case, a company’s “ability” to hold an asset has to be evaluated in light of its capacity to fund the asset through maturity.  Demonstrating the intent and ability to hold an asset through maturity must be demonstrated for the privilege to account for assets at amortized cost, not the other way around. If you are accounting for an asset at amortized cost, your inability to convince your auditor or regulator that you have the intent and ability to hold that asset through its maturity may result in them requiring you to account for that asset at fair value. But there aren’t similar accounting or regulatory qualifications for holding a financial asset at fair value because fair value accounting is the more onerous accounting treatment.

Both interesting and relevant to this conversation is that there was an exception granted to the “upon-origination” designation by the FASB for banks to be used only at the time that it adopts CECL. Upon first adopting CECL, companies are given a choice to change the accounting treatment for loans that were previously held at amortized cost to account for them at fair value. The reason why this option was given is of paramount importance to this discussion.

FASB, in providing this option, acquiesced to the truth that loans that have relatively high credit losses but that also have relatively high compensating coupons to offset those losses (which results in an acceptable risk-adjusted spread) would suffer an accounting distortion when they implemented CECL. Front loading the accounting for the credit losses without recognizing the high earnings stream that offset those losses would distort the comparison between two companies that, though, they lend at a similar risk-adjusted spreads, do so from different yield and loss levels. For higher yielding loans, fair value accounting is the better representation of the underlying economics of the loans than amortized cost offset by a CECL contra-asset.

Bank regulators don’t object to fair value accounting.

When the consortium that was formed to buy IndyMac F.S.B. out of receivership from the FDIC (of which Paulson & Co was a lead investor with a 24.9% stake, and I was a Board member until its sale to CIT), we elected fair value accounting for our loan book. To my knowledge, we were the only bank that had “elected” this accounting treatment at the time.

We had a sense that mortgage valuations were approaching a floor and that holding these assets at the significantly discounted cost at which we acquired them from the FDIC would distort their value as the mortgage market recovered.  We were also aware that this accounting choice would have the pleasant benefit of generating additional capital as the assets recovered. It had absolutely nothing to do with our ability or intent to sell the loan book. We were correct, and as a result, we were able to generate significant regulatory capital that we then deployed by buying two additional banks out of receivership from which we created OneWest Bank, N.A.

Because of the high-profile status of this transaction, we were under constant and invasive regulatory scrutiny. As I recall, it was only for brief periods that we didn’t have examiners on site at the bank. When we bid to purchase other banks from the FDIC, the quality of the portion of our equity that was the result of the fair value marks on our loans was discussed at length with our regulators, which consisted of both the Federal Reserve and the OCC. Because of the “de novo-like” formation of our bank, we had a significantly higher capital floor than the regulatory minimums. The veracity of our fair value marks was highly scrutinized, but it was never questioned that the capital created by these marks was, indeed, equity.

At present, SoFi is being attacked in the blogosphere for how it accounts for its loans. Once you cut through the book-talking-bombastic rhetoric (some of which I find down-right irresponsible), there are a couple of genuinely relevant questions that need to be answered to determine if SoFi’s guidance about returns and growth are to be believed. 

The first and easiest question to answer is has SoFi established credibility in meeting or exceeding its guidance. It’s a matter of public record that SOFI has exceeded guidance to date. By comparison, UPST hasn’t done so consistently.

The second question that has been raised is, if SoFi’s loans are held for sale, why are they accumulating on the balance sheet? SoFi’s accumulation of loans on balance sheet has been characterized as a “Sales (near) Freeze” the motivation and danger of which is then questioned. This question is articulated in roughly the following way: “Can SoFi really substantiate that they intend to sell these loans when in fact they have sold only a negligible percent through 3Q23?”

We’ve already addressed the underlying premise of this question and demonstrated that it is irrelevant. Worst case scenario, if you answer this question in the negative, is that SoFi will create an additional line on its balance sheet called “Loans Held at Fair Value,” which will receive some portion of the “Loans Held for Sale” line. It’s a change of geography, and in that, only to a nearby suburb. There would be no change to earnings or capital.

There is no denying that since becoming a bank, SoFi is selling a significantly lower percentage of its loans than when it was dependent on the warehouse lines, capital markets transactions, and whole loan sales for its funding like many of its competitors in FinTech land. In fact, I’d go as far as to say that there is probably some portion of loans that SoFi now intends to maintain on balance sheet and no longer intends to sell. This level will have to do with management’s growing comfort with the “coreness” of its rapidly growing deposit base.  

SoFi is offering top-of-the market rates on its deposits, which normally would make them “hot.” But the fact that a very high percentage of these deposits are connected to direct deposits and likely have been connected to bill pay and other conveniences that raise the switching costs for depositors cools them down considerably. Still, there’s little doubt that if SoFi lowered its market-relative deposit rate, that they’d experience some deposit outflow.

The economics of holding versus selling loans is directly related to the quality of SoFi’s on-balance sheet funding source irrespective of how loans are designated on the balance sheet. SoFi has dramatically reduced sales of loans because it makes significantly more economic sense to sell loans only when you don’t have a high-quality deposit franchise. Without deposits, you must fund you ongoing activities at the whims of the capital markets.

This funding optionality is among the many reasons that SoFi is a far superior investment to UPST. I listened briefly today to a couple of podcasters debating about SoFi’s lack of loan sales, and it seemed clear to me that they significantly undervalued the benefits of being a bank. Do they understand why a lender voluntarily puts themselves under the watchful eye and strict regulation suffered by any company that benefits from deposit insurance?

Here, I want to quote directly one of our reports on UPST:

As investors, we have seen the cycle of nonbank entries into loan origination markets more than once, and rarely does it end well. Financial services companies are regularly being “disrupted.” There’s a pattern to this: When the secondary loan market and the interest rate and credit environments are accommodative, many players that don’t suffer the constraints of being banks can look really promising.  In fact, when things are good, operating as a bank is a shackle. Pesky regulators, high capital hurdles, restrictive accounting guidance…all these things chafe.

The opposite is true when we are in an environment like the current one: It pays to be a bank, and SoFi is using its rapidly growing deposit base creatively and efficiently to fund its growth.

Since reporting as a bank holding company, SoFi has reduced the cost of its total funding from 98 bps above the Fed Funds rate to 140 bps below the Fed Funds rate (also, the loan rates that they are earning are rising faster than the Fed Funds rate, which has sustained SoFi’s NIM in the high 90s in percentile ranking among all publicly traded banks since it became one).

The student loan securitizations cost SoFi in a range between 1.83% and 4.44%. I predict that, if student loan refinancing activity picks up significantly, you will see a lot more of this paper securitized than staying on balance sheet.  Why? Because the economics are better. The reason that SoFi’s execution in the student loan secondary market is so good is two reasons. First, student loans have a very long history of performance in the secondary market, and as a result, loan buyers are generally more comfortable with the volatility, and predictability of loss and prepayment rates.   

Moreover, SoFi itself has a long history as a trustworthy predictor of its own loss rates and performance characteristics. So, unlike UPST, who has repeatedly delivered assets to buyers that have underperformed, SoFi’s reputation in the secondary market is stellar, which helps it maintain the best execution possible.

Some of the dynamics of market execution have to do with informational asymmetry. It is easier to trust the underwriting of a student loan than it is an unsecured personal loan. Generally speaking, loans with lower loss rates also have loss rates that are less volatile. The problem with higher loss rate assets is that the loss rates are subject to exogenous shocks that make it increasingly difficult to model risk the farther down on the credit quality spectrum you go.

Retained Credit Risk

One bear criticism of SoFi relates to the structure of its sales transactions in that the sales don’t transfer all of the credit risk of the underlying loans. In some transactions SoFi has sold assets while retaining a narrow corridor of credit risk. Chris Lapointe has described this corridor as opening somewhere above the expected losses and closing again slightly above that. It is capped. I believe that the most critical components of credit risk are being transferred. However, I don’t like that it is opaque, and I’m not yet satisfied with SoFi’s disclosures around this practice. I think they would do well to disclose the specifics of these arrangements in order to reduce the information asymmetry for we analysts.

The Decline in the Cash Portion of the Recent Loan Sales

This is a more difficult part of the analysis to unravel clearly; but one that seems to be the most egregiously mischaracterized in the blogosphere. In SoFi’s latest sale of personal loans, it earned none of its gain in cash. While this transaction was small, according to management, it is very similar to the structure that will handle the sales of $100s of millions more in assets going forward.

Historically, some companies that begin a trend of decreasing cash portions of gains on sale do so because they can no longer “cleanly” sell their assets at the desired execution level. So, it’s a legitimate concern that SoFi reached a balance-sheet valuation-validating execution level by taking a high portion of noncash consideration. So, let’s break it down.

CFO Chris LaPoint has alluded to the fact that SoFi views the structure as an accommodation to its purchasers that resembles “financing.”

In order to unpack SoFi’s structuring of this transaction, follow this comparison: First, assume that a comparable, mostly cash, execution would have been in line with the 3Q22 sales, and that the only difference between the 2022 sales and the 2023 sales is the structural differences relative to SoFi receiving more cash upfront than in 2023.[2]

If so, then the mostly-cash execution of the most recent transaction would have resulted in Total Consideration of $15,625 [103.5% * $15,098].

So, we concede the bear’s first point: Had SoFi not “financed” the sale and taken its “regular way” servicing fee, it’s questionable whether the execution would have validated the balance sheet value of SoFi’s loans held for sale. In fact, if SoFi had executed at 103.5% ,we might have questioned if it needed to take a mark on the rest of its held for sale book, which—should they need to write down the book to this level—would require them to take a $10.3 million charge to true the on-balance sheet valuation with the cash transaction value from 2022. They wouldn’t be required to do that because the third-party valuations of the existing loan book would still stand; but a reasonable analyst might consider adjusting equity for such a charge.

So, we can say that SoFi’s willingness to take less cash upfront resulted in a 1.6 percentage point, or $235 thousand improvement in the execution over the 3Q22 transactions. In other words, all else being equal, the change in the structure of transaction had a present value of $235 thousand to SoFi at the time the transaction was executed.  

It is right to evaluate the quality of execution at 105.1% because it relies on SoFi’s valuation of a servicing asset, which appears to be a lower quality gain than if it had been executed in all cash at the same price.  The risk to the value of the larger servicing asset is that underlying assets would prepay quicker than the assumption used to value the asset. Unlike the “co-investments” that UPST uses, which are leveraged to credit risk, SoFi’s risk is that the excess servicing fee that it negotiated will be earned over a shorter period of time than it anticipated in negotiating the structure and calculating the gain.

There are one of two reasons for SoFi to take this risk. The bears believe that the reason is that a low-quality gain was the only kind of gain that could meet SoFi’s 105% execution hurdle. We disagree.

The other reason is that it makes sense because SoFi has the liquidity and has been at least adequately compensated to take that risk. We’ve already addressed liquidity. So, let’s figure out if SoFi has been adequately compensated for this risk.

To do this, let’s focus on the portion of the servicing asset that looks like a financing arrangement.

This is a painstaking exercise, for which I apologize. It is necessary, though, because we can’t just dismiss the bears by saying, “nuh-uh” and moving on. So, we’re providing, at unfortunate length, our analysis below to “show our work.”

• First, we need to impute a value of the servicing asset that is the “normal” servicing asset. To estimate this, based on the 3Q22 transaction, we calculated that SoFi’s normal servicing fee is about 41bps.[3]  Given that servicing fee, the valuation of a normal servicing asset in the 3023 transaction would be $130 thousand. This is pretty close to the value for a normal servicing asset at 0.9% of execution points that SoFi achieved in the 2002 transaction.
• Therefore, the “extra” servicing asset that appears to be a financing arrangement amounts to $637 thousand [$767-$130].
• Said plainly, if the underlying collateral is similar to the 3Q22 transaction, in order to receive an additional $637 thousand in non-cash consideration, SoFi gave up $396 thousand in cash consideration [(102.6%-100.0%)*$15,098].
• That is to say, they lent the asset purchaser $396 thousand that they might have achieved in cash in order to receive future cash flows (via the excess servicing fee) with a present value of $631 thousand.[4]
• This $631 thousand is the present value of the future “extra” servicing fees discounted at 8.5% (SoFi’s weighted average discount rate for servicing assets in 2023).
• Next, we use SoFi’s prepayment assumptions, unpaid principal balance (UPB), and present value calculation from above to back into the “excess servicing fee” and the total cash flows that this asset will generate if the CPR is accurate.
• We arrive at a roughly 2% “excess” servicing fee that represents SoFi’s compensation for financing the sale.
• We arrive at a future value, or total cash flows, of $747 thousand for the excess servicing asset that has a present value of $631 thousand.
• By analogy, some part of this amount is the amount of the asset purchase that SoFi “financed.” SoFi “financed” $747 thousand of the total purchase price of $15,845 thousand. In simple loan terms, the buyer put cash down of 95.3% and SoFi financed the remaining 4.7% of the total purchase price.
• Compared to a mostly-cash execution, SoFi was able to achieve a higher price for the assets by providing financing.
• This is not a novel concept. As with any company that finances the sale of its own assets (think of captive car financing units of automakers), the risk is this: will they be paid back for financing the purchase? No one looks at the sale price of the car and says, “Well, that’s not the real purchase price of the car. We need to back out the roughly 83% of the purchase price that was financed.” At some point, we might, as analysts, challenge the effect of highly favorable financing arrangements on sales trends, and, if we are diligent, we will track the credit quality of the loans that are being made to finance the cars. If that’s deteriorating, then we will certainly question the auto maker’s earnings because they have chargeoffs or losses on sale coming (There was a great short on CarMax in the early 2000s over that), and they won’t be able to continue financing auto sales into the future, which will dampen future sales. These are the kinds of questions that are legitimate to be asking of SoFi at this point.
• I hope that you are sticking with me because I know that this analogy may seem far afield from SoFi’s recent and announced loan sales. But they aren’t in the sense that the bears seem to be ignoring the plain demonstration of financial flexibility that SoFi’s solid on-balance sheet funding sources are providing it. The bears would have you believe that SoFi pulled a slight of hand because they collected less in cash on this transaction than they had in past transactions. One book-talking-bear went so far as to insinuate that the structure of the transaction amounted to fraud. We’ve seen fraud, and this ain’t it. I’m going to great lengths to deconstruct what SoFi did in this transaction because an irresponsible statement such as this can sow something destructively beyond healthy skepticism.
• It’s a legitimate question to ask what execution SoFi would have achieved had it not financed the transaction; but it shouldn’t cloud our ability to judge how lucrative the transaction was to SoFi, whether it’s repeatable, and whether SoFi will realize the cash flows that support the valuation of the servicing asset.
• A fair starting place is to say that SoFi would have achieved an execution closer to what it did in 2022 when it booked most of its gain in cash. In that case, SoFi booked a 2.6% gain on the principle and accrued interest of the loans it sold.
• SoFi’s compensation (all else being equal) for being paid back $740 thousand over time is the $235 thousand improvement in execution. The weighted average life of the return of the $750 thousand is 2 years (WAL calculation available on request).
• To the bears, this $235 thousand looks like a trick to achieve a higher execution. 
• To us, the $235 thousand improvement in execution looks very similar to interest income that SoFi has earned to lend out $740 thousand for 2 years.
• If you’ve followed me this far, the next part is easy. SoFi’s yield on this “loan” is almost 16% [235/740/2].
• The repayment of this loan does not rely on the counterparty’s credit worthiness because the repayment source sits high in the waterfall of the payment distribution of the cash flows from the underlying assets. Servicing fees are returned to SoFi before interest is paid to the asset purchaser. Credit risk is limited to the extent to which defaults are equivalent to prepayments and will reduce the total servicing fees collected. The sole risk to SoFi recovering the cashflows that underpin the 105.1% execution is prepayment risk, which SoFi can price more efficiently than its counterparty (which, btw, is yet another way that this transaction structure is a lucrative capital efficiency exercise).

To summarize:

• SoFi is using its increasingly core deposit base to structure lucrative financing arrangement attached to its sale of assets.
• This practice has low downside risk and a very high return.
• SoFi is cultivating a long-term asset sale channel that leverages its strong balance sheet to exploit relative capital efficiencies between its bank status and asset purchasers.

We look forward to SoFi demonstrating over time that it can continue this efficient and profitable balance sheet management tool.

[1] Pardon this boomer’s use of “original gangster.” In fairness, when I first wrote about this topic, I was young enough to use the slang.

[2] We don’t know if there were differences in the underlying collateral that would have resulted in a different execution irrespective of the amount of cash received upfront; but we need to make some assumptions to evaluate the bear thesis.

[3] This is the fee that iterates to solving for a 0.9% valuation given the published assumptions for the servicing assets in 2022. It’s possible that the underlying loans have different characteristics; but this is the best estimate we can make on the publicly available information.

[4] A bear might argue that we don’t even know that the assets sold were worth even 102.6% in cash. It’s true that we don’t know exactly what they are worth; but there has been no degradation in SoFi’s weighted average characteristics. So, the legitimate concern that can be evaluated is the structure of the transaction.