I don’t like writing directly after an earnings report because the emotions connected with the stock price movement often seem to cloud my judgement. I am human after all. After a few days things end up being a lot clearer and I can take the emotion out of it and do better analysis. SoFi earnings was a mixed bag, there are some things that absolutely reaffirm the thesis (the good), some things that were disappointing (the bad), and others that brought up questions (the unknown). I’ll break my thoughts down into multiple articles, and today we tackle the first part of the unknown. This discussion will mostly surround the growing balance sheet and the fair values of the loans held on the balance sheet.
I said on earnings day that the most reasonable explanation is that the market is concerned about their inability to sell loans from the portfolio and that they are continuing to grow the balance sheet. I like visuals and data, so this graphic illustrates best what I mean by growing the balance sheet:
Since receiving the bank charter, they have 3x the number of personal loans on the balance sheet as before. Not only that, but the number of loans sold has decreased every quarter. This introduces three risks: fair value risk, delinquency and defaults hurting their revenue and profits, and the risk of a liquidity crisis and the bank failing. Today we’ll talk about fair value risks. I’ll define the risk and then analyze what information we have and how exposed SoFi is to the risk.
A quick primer on bank accounting
Feel free to skip this section if you understand GAAP accounting of loans. Banks and other financial institutions can use one of two different types of accounting methodologies when they originate a loan. They are held-for-investment (HFI) accounting, and held-for-sale (HFS) accounting. I will describe them briefly here. SoFi uses HFS accounting.
HFI Accounting
HFI is used when a company plans to hold loans through maturity. When a financial institution books a loan that they plan on holding through maturity, they are required to recognize lifetime losses of their loans up front as the loan is originated. This is referred to as Current Expected Credit Losses (CECL) provisioning. In practice, it means that the bank uses empirical data to estimate how many of these loans will default, and they have to subtract that amount from their bottom line.
For example, in 3Q22, LendingClub originated $1,153M in loans that they added to their HFI portfolio. They set aside $82.7M in CECL provisions, which is about 7.2% of the originations. That money will sit there and be used up to pay for defaults across the life of those loans.
HFS Accounting
HFS is used when a company plans to sell the loans before maturity. Instead of taking out a CECL provision, the loans are accounted for on the balance sheet at the fair market value, or the value that they would have if they were sold on the markets. No CECL provision is required, but each quarter all loans must be reassessed to determine their fair market value in the current environment. Changes in fair value of loans held from quarter to quarter are accounted as a noninterest adjustment to revenue. Fair value can go up or down depending on prepayment speeds, market forces, loan delinquency trends, and interest rate movements. SoFi begins to write off the value of the loan as it goes into delinquency on a sliding scale until 120 days, where it goes into default and it is charged off as a 100% loss. All of those write-downs occur in real time during the quarter it is going delinquent or is charged off.
Fair Value Risk
SoFi unloads their balance sheet in two ways, securitizations and whole loan sales. Securitizations are a bit harder to understand (I’m still not completely sure I fully understand them myself), but whole loan sales are pretty straightforward, they literally just sell them to somebody else who buys them for a set price and takes over the loans. SoFi hasn’t sold a single personal loan through the whole sale loan market since 2022. They haven’t sold a student loan in any way since 3Q22. Their business model is to originate a loan, hold it on the balance sheet for a while collecting interest, then sell it to somebody else so that SoFi can free up the cash that was tied up in the loan and go originate another loan. Historically, SoFi has a gain on sale margin (GOSM) of between 4-5%. That means if they go out and have $100M in principal of loans to sell, the person who buys them pays SoFi $104M to take over those loans. They overpay with the assumption that they’ll easily make that $4M back with the interest generated for the loans.
What happens if all of a sudden the people who are buying these loans get a little spooked that we are going into a recession and decide that maybe personal loans aren’t the best investment right now? If demand to buy personal loans is weak, the price they are willing to pay drops. Now instead of paying $104M for the loans, they are only willing to pay $103M. That may not seem like much, but SoFi has a loan portfolio of almost $10B, and they are reported at fair value every quarter and that trickles down also into noninterest income.
So if demand dries up and the fair value of those loans drops by 1% in a quarter (going from a GOSM of 4% to 3%) they have to rerate the entire portfolio by 1% of $10B, or $100M. That would be a direct hit to revenue. Imagine if that would have happened this quarter. Instead of reporting $460M of revenue, which was a pretty good beat, they would have reported $360M in revenue. That of course would also trickle down to the adjusted EBITDA and GAAP numbers as well, and it would have been a huge miss.
In prior quarters, this risk was diminished because they’d always at least sold some loans that they could point to and say “look at these loans we sold during this quarter, we can prove our GOSM is still 4.5%”. However, they didn’t sell any personal loans through whole loan channels at all in Q1 and didn’t sell any student loans in Q4 or Q1, and the market is questioning just if the fair value of their loans really is as high as they say it is, especially during a banking crisis.
This is a yellow flag and a real risk
Bears can now reasonably argue that the fair value of the loans as reported on SoFi’s balance sheet is inflated. The argument is simply that the current environment is one of weak demand for loans because banks are pulling back on credit. SoFi cannot absolutely prove that the marks are accurate since there are no loan sales they can point to in the current environment. Bears can also point to SoFi’s loan sales numbers as proof that demand is dwindling and gone. What’s more, originations from both LendingClub and Upstart confirm this lack of demand, as both of them have seen significantly declining loans sold to their respective marketplaces. It’s obvious demand for personal loans has dried up.
An analyst asked about this directly on the earnings call:
I guess, on the actual loan sale side, what gives you confidence that the loans you're marking on your balance sheet today, if you were to start going back to the market again, what gives you confidence that you can kind of hold on to those marks you have on the balance sheet?
Here was CFO Chris Lapointe’s response:
Yes. Absolutely. And in terms of why we get confident in the sense that we would be able to sell the loans at where they're currently marked, every single quarter we work with a third party valuation firm that marks to market each and every one of our loans on an individual basis to account for changes in every single factor that impacts loans. So that's things like the weighted average coupon, default rates, prepayment speeds, benchmark rates, spreads as well as where secondary bonds and residuals are trading. So you see that mark to market take place every single quarter and that flows through the revenue line of our P&L.
I think that is actually a very adequate answer. Using a third party to verify the loans removes the conflict of interest from SoFi doing it completely in house. However, value on paper will never completely alleviate the fears as much as actual evidence from a real loan sale. Additionally, even though they didn’t sell any whole loans, they did have a $340M securitization in the quarter. According the Chris Lapointe, “The deal was 8 times oversubscribed with over 28 orders, which allowed us to tighten spreads meaningfully by 80 basis points relative to the deal that we did back in Q4 that had comparable collateral.” Without going into too much detail, this means that the demand for the securitization was excellent, even better than during Q4.
Yellow Flag #2: Holding loans for longer than guided
When SoFi received their bank charter, they forecasted that they would begin to hold loans on the balance sheet for longer periods. In their 1Q22 earnings call, they said, “We’ve just started moving towards holding loans 6 months on average versus 3, which allows us to collect more net interest income. This also creates a more rational pricing environment for our paper as we leverage our ability to hold loans for longer should pricing not be acceptable.” They reiterated this in the 3Q22 call, where they said they’d be holding them for 6-7 months. The message started to change moving into November when they started discussing balancing holding and selling to “maximize return on equity.”
This strategy was then repeated multiple times in the recent past. This is from the Bank of America conference in March:
Mihir Bhatia
Yeah. I do want to -- maybe just a quick -- very quickly to follow-up on the holding balance loans for longer. The strategy is still to remain relatively asset-light. It’s like you’re extending the whole time from three months to four months to six month to eight months, gives you more flexibility to be more opportunistic in terms of when you do the deals. But the strategy hasn’t changed, right? Or is the idea that as you get more deposits now, you maybe hold some loans through the life-of-loan of fuel?
Anthony Noto
Yeah. We’ll continue to do a mix. I mean our goal is to maximize return on equity. We’ve been able to achieve really strong return on equity profile in the bank after one year and that’s without it fully leveraged. We have an ability to get our return on equity. We have a line of sight to return on equity in the high 20s or higher. And that’s through a combination of our mix of business in that we have the Technology Platform, as well as our different Financial Services businesses, but also includes us being able to turn the balance sheet in a manner that maximize the ROE per loan. So having the optionality to hold versus sell is a great way to put us on the efficient frontier of ROE.
The same message was driven home in the 1Q23 earnings call last week when they repeated that they are seeking to maximize ROE. In principle, I have no problem with doing what maximizes returns for the business. However, I do take issue with the lack of clarity here. Anyone who really follows the company closely knew in November that they were almost certainly going to hold loans for longer than 6-8 months. However, they should have made that absolutely clear. Noto had the chance to be crystal clear about this at the BoA conference. That was in March, he knew that they had not sold any loans yet chose to say “We’ll continue to do a mix.” He sidestepped the opportunity to give clarity when he already knew at that point that 6-8 months was no longer on the cards. He should have owned it then when he had the chance.
Victims of their own deposit growth
In a lot of ways, SoFi is a victim of their own success. When you grow deposits by $2.3B - $2.7B every quarter, you have to find a way to put that money to work. They are paying 4.2% APY on that cash, so they need a good return on it. The faster-than-expected growth in deposits is a huge reason why the balance sheet is putting on so much bulk. When they first got the bank charter, they said they expected about $100M of inflows per week. Throughout Q2-Q4 of last year, they averaged $176M per week. In Q1 of this year, that number accelerated to $214M per week. The financial services has done so well bringing in high-quality deposits that they are left with little choice in the short term other than to grow the balance sheet. This is a mitigating factor that most people gloss over when they look at the balance sheet growth, but should be considered.
This was a misstep by management in my opinion
I have no reason to doubt that the fair value of the loans represent SoFi’s best estimates in the current market. I like that they are third party verified and I think it adds to the credibility of the values. It is also true that not selling loans calls that number into question. I also understand that building the balance sheet is the best way to maximize returns on those loans in the current environment. However, I think not selling any loans at all is a mistake. Selling $200M-$300M worth of loans every quarter to justify the marks would not have a significant effect on revenue, but does remove the uncertainty on the accuracy of the mark-to-market fair values. This would also prove the GOSM is sustainable throughout the entire market cycle, even when the banking sector is stressed. The market hates uncertainty, and I think selling 2-3% of the loan portfolio to remove that uncertainty would have been the prudent move.
Conclusion
I think the fair value of the loans is accurate based on what management has said. However, this shouldn’t even be a question. Selling a small amount of loans to justify the fair value marks should be a simple exercise and would go a long way to easing the worries of analysts and investors. You can be sure that people will be extremely interested to see what happens regarding loan sales in Q2, myself included.
I was about to submit this when the 10-Q came out yesterday. I waited to publish it until I had a chance to go through the fair value information in that report. I’ve been through it and I have even more to add to this discussion, but this post is already long enough, so I’m going to publish it as is for now. I’ll have a follow up post soon to add some talking points to this discussion. What I’ve discovered in the 10-Q so far makes things look better, not worse, in my opinion.
Thanks for the evaluation. I hope someone from SOFI comes to the same conclusion. The good news may be: for those wishing to buy good quality loans, SOFI's should be in strong demand, possibly even at the top of the list.
"Anyone who really follows the company closely knew in November that they were almost certainly going to hold loans for longer than 6-8 months."
I don't really agree with this part.
In hindsight, I will agree with this in that it makes sense, but I wouldn't agree back then.
I would definitely agree back then that they will eventually consider holding loans for longer and perhaps start adding PL to their HFI portfolio. I would also agree back then that the SL numbers don't make sense with the 6-7 month hold (I looked at them back then but chucked it out thinking I might be looking at the wrong numbers or calculating something wrong).