Are banks sweeping dud property loans under the rug?

That new set of rules has been in place for a little over four years. Investors could be forgiven for wondering if it is working any better than the former one.

The old system was called the incurred-loss model. To book a loss, a lender had to conclude it was “probable" that one had already happened. The term “probable" wasn’t defined numerically, but the bar was widely interpreted to be very high—perhaps a 70% or greater likelihood. Bankers used to explain, conveniently, that they would have booked more loan losses, if only the rules would have let them.

Under the new system, in place at most large companies since 2020, lenders from day one are supposed to continuously estimate their credit losses over the life of a given instrument, be it a loan or bond. The threshold for recording losses is supposed to be much lower—when they are “expected," rather than waiting until losses probably happened. This was supposed to lead to more aggressive, and more timely, loss recognition.

The office-loan market is testing investors’ faith in the new expected-loss model. As with other commercial properties, loan payments on office buildings often are interest-only until maturity. When rates were ultralow, many lenders and borrowers went into these loans assuming they would be refinanced rather than paid off at the end. That would mean no defaults as long as they could keep rolling over the loans.

The pandemic sent office values in many big cities tumbling as more people worked from home. Now, for many borrowers, refinancing isn’t an option because the buildings are worth less than the borrowers owe. That makes defaults inevitable. Until then, though, the owners still may be current on their payments. Hope springs eternal, until it doesn’t.

In a July 24 note, Pimco’s John Murray and François Trausch warned of a $1.5 trillion wall of maturities for commercial real-estate loans over the next two years. “Lenders and borrowers will be forced to ‘face the music,’" they wrote. If they are right, it would mean the expected-loss model hasn’t been working as billed. Lenders still have wide discretion to delay officially expecting red ink if they would prefer not to expect it.

Until they can’t. At New York Community Bancorp, credit losses on commercial real-estate loans, including office loans, have surged in the past few quarters, raising questions about why management took so long to identify them. A bigger concern should be the losses at other lenders that aren’t yet visible to outsiders.

Noted short seller Carson Block, in a report last December, predicted large credit losses would soon swamp Blackstone Mortgage Trust, a nonbank commercial real-estate lender. The company at the time called his report “self-interested and misleading" and said it was “well positioned to navigate this environment." Then in July, it cut its dividend and posted its third consecutive quarterly net loss. Credit losses last quarter were so large that they exceeded the company’s net interest income.

Curiously, default rates have been higher for property loans backing widely held commercial-mortgage-backed securities than for the same types of loans on banks’ balance sheets. That underscores how traditional lenders have more flexibility to help borrowers work out their problems than do the vehicles that issue commercial-mortgage-backed securities, which can’t so easily “extend and pretend."

Banking regulators have said they are aware there is a problem, while also assuring the public that this won’t be another 2008. An interview that Federal Reserve Chair Jerome Powell gave to CBS’s “60 Minutes" in February is worth revisiting. Asked about banks’ office loans, he said, “There will be expected losses. It feels like a problem we’ll be working on for years. It’s a sizable problem."

The irony of that statement: If Powell was right about the losses then, under the expected-loss model, banks probably should have booked them already.

Write to Jonathan Weil at jonathan.weil@wsj.com