Loan accounting in a bank should be straightforward, delivering predictable and transparent results.

While judgment is involved when loan performance deteriorates, the goal of management and outside accountants should be a presentation that maximizes comparability between banks and minimizes idiosyncratic outcomes. That was one of the justifications for the Federal Reserve’s adoption of Current Expected Credit Loss, or CECL, reserve accounting for banks in 2016 — despite many complaints about the negative impact on reported results and the “price” that CECL imposed on growing lenders.

So, what the heck is going on with personal installment loan accounting at SoFi Technologies?

The San Fransisco-based bank’s loan accounting uses “fair value” treatment rather than the standard amortized cost/CECL approach, which is delivering earnings — and more importantly capital levels — wildly different from those reported by other banks and consumer instalment lenders with similar products and credit profiles. This has competitors and market observers scratching their heads.

According to researcher Giuliano Bologna at Compass Point Research and Trading, Inc:

In the third quarter of 2023, only six us banks had non-mortgage consumer loans (personal, student, etc), measured at fair value, with SoFi’s loans making up more than 95% of the total. And only SOFI’s loans were booked at a large premium to their unpaid balance, resulting in higher loan revenue.

This higher revenue ultimately flows through to earnings and capital.

According to Bologna, if SoFi had elected to account for loans at amortized cost and had to take CECL reserves similar to peers, its tangible book value per share reported for the third quarter of 2023 would fall from $3.42 to $2.00 and its CET1 ratio would be cut from 14.3 percent to 8.9 percent.

Moreover, Bologna estimated that had SoFi used the amortized cost and CECL option, its total risk-based capital would have been well below the 10.5 percent minimum requirement set by regulators when it became a bank in 2022.

On underlying rather than reported numbers, SoFi is operating with around 40 percent less capital than peers like LendingClub, which are following the amortized cost/CECL approach used by almost all other banks.

Is this a brewing scandal? No. There is nothing illegal or nefarious about SoFi’s use of fair-value accounting.

In its wisdom, the Financial Accounting Standards Board made fair-value accounting an entirely legitimate option under GAAP for banks and others that hold loan assets. SoFi, which is experienced with fair-value accounting because it started out as a marketplace lender, is merely taking advantage of an option open to any loan originator.

The goal of fair-value accounting is to estimate a value that represents the price of the asset in an arm’s length transaction. But inherently, fair-value accounting involves assumptions and estimates by management which materially affect earnings and capital levels — if those estimates are aggressive or wrong, they could materially overstate both.

It can be argued that some of the assumptions SoFi is using are aggressive, especially relative to the few other lenders using this method. And though it’s nearly impossible from SoFi’s SEC reports to figure out how fair value is calculated, the results have been blessed for now by its outside auditors at Deloitte.

But, as with all things in financial services, the hand that giveth is the hand that can also take away.

The income acceleration effect from fair-value accounting is ultimately transitory and will reverse over time as loan growth slows or charge-offs rise. Ironically, this is just the opposite of CECL’s effect on earnings through the cycle.

As Bolgona puts it:

From a high level, relative to electing to account for loans at amortised cost, electing the fair value option provides a P&L tailwind during periods of loan growth due to recognising Day 1 fair value gains on loan originations, but lower back-end economics over the life of the loan as the [fair value] premiums are amortised/expensed and due to recognising charge-off expenses as [net charge-offs] are incurred. . . . It’s important to point out that the lifetime economics and revenue/expenses using FV, and amortised cost would be exactly the same, the accounting election only changes the timing of revenue/expense recognition, not the lifetime economics.

While we can take some comfort that it all ultimately comes out in the wash, such a large difference in a critical area for banks undermines the comparability of reported results expected by the industry, analysts, and regulators. It injects too much potential for gaming into accounting. It also allows capital arbitrage and potential period-to-period capital volatility that should concern regulators.

Neither CECL accounting nor fair value accounting is perfect. But it makes little sense to allow some lenders to use an opaque option that under certain circumstances can turn punishment into reward, without more clarity about the consequences for earnings and equity.

At the very least, banks should be required to report the difference between CECL and fair value treatment in their financial and regulatory reports, so the differences are clear to the markets and regulators.

This post comes to us from Todd H. Baker, a senior fellow at the Richman Center for Business, Law and Public Policy at Columbia Law School. A version of this post appeared in the Financial Times.