Yesterday, I wrote about the former CEO of BB&T bank John Allison's experience during the bank bailout. As part of his interview with the Big Think, he also addressed other aspects of the financial crisis, including market failures. While he thought there were few (but lots of government failures), one that stood out in his mind was fair value, or mark-to-market, accounting (which, I would argue, should still be considered a kind of a government failure since it's a regulated accounting rule). He believes it was one of the major reasons the financial markets became so illiquid and unstable. I see where he's coming from, but I have trouble abandoning this accounting concept.
Allison thinks that so-called toxic securities weren't being scooped up at fire sale prices because investors and other banks didn't want to take the accounting risk involved. That risk was created by fair value accounting. Those who might have purchased the assets would have to mark them at the price they paid -- not what they believed they were worth. It also hurt on the sell side, because banks didn't want to mark all like assets that were left on their balance sheet at the prices the portion they got rid of sold for. He theorizes that if it weren't for mark-to-market, those assets would have been purchased far more quickly and willingly by those who believed the market price ultimately undervalued the assets. Consequently, their market values would have risen as well.
In this example, he's talking about a $100 million mortgage bond that sells for $50 million, but which he says "economically" is worth $80 million.
But here's why the market didn't work. It was why didn't the market correct that? Well, there were people like BB&T that have plenty of money, we saw that bond was worth $80 million, we could buy it for $50, if we went in and started buying and everybody else like us that had money started buying, then the price would've gotten certainly close up to the $80 million. But we wouldn't buy because we couldn't take the accounting risk. Because even though we knew the bond was worth $80 million, we weren't going to buy it even for $50 million, because we didn't know that next quarter, under this panicky market, it would only be, a fair market value quote, would only be $40 million because everybody would be scared and we'd have to take a markdown.
Okay, there are a few things going on here. First, since finance is a zero-sum game, mark-to-market creates a sort of classic prisoner's dilemma game theory problem. It's in everybody's best interest to buy these undervalued bonds, but nobody will, because they can't be sure that others will follow and the value will, in fact, rise. It could even decline if you're the only person buying. If banks didn't have to mark these assets to market, of course, then there would be less fear of buying. Purchasers could just mark them up to the price they believe they're worth without having to worry about others buying as well.
I completely understand his concern here, and fully believe that mark-to-market accounting did deter investors and banks from buying and selling undervalued securities. But here's the problem: the so-called "economic" value that Allison's talking about is no where near objective. If it were, then nobody would hesitate buying these assets, because even a panic stricken market isn't that irrational.
In reality, $80 would just be one buyer's subjective belief of what this security might be worth, based on some assumptions about losses and prepayments on the mortgages lying beneath the bonds. Those assumptions could be wrong. After all, they were epically wrong in during the housing boom when these securities were created.
So let's think about what could happen without mark-to-market here. Let's say I'm a bank that buys a security for $50 which was worth $100 at par, but which I think will turn out to be worth $80. After purchase, I mark it at $80 on my books. I then have an accounting gain of $30 on the purchase. (Nice!) Assuming something like a 10% capital requirement, however, I take $8 of that profit at put it aside, so I really only have a net profit of $22. I distribute that to bankers and shareholders.
Some time goes by. If the bond turns out to be worth $80, then that's great -- I was right. If it turns out to be worth more than $80, then that's even better. But what if it turns out to be worth less?
Let's say that $50, was too pessimistic, but $80 was too optimistic, and it ends up being worth $60. I have to record a loss of $20. Meanwhile, I only have $8 in capital backing up the security. Maybe I have more capital to cushion that additional loss for the other assets I hold, but if I was in the business of buying these securities, I might have a lot of them. And if that's the case, all of those securities will have losses bigger than their respective capital cushions -- since my assumptions were flawed. Uh oh.
Now maybe the answer to this is for banks to have significantly higher capital requirements. Allison is actually for that, suggesting they should be in the 20% to 25% range. I think that significantly higher capital requirements could help this problem.
Another possibility -- you must retain the accounting profit that makes up the difference between the purchase price and what you think it's worth until the security's performance is resolved (i.e. resale or maturity). Of course, this probably still wouldn't make most of those buyers particularly happy. In the example provided above, they would have to hold 37.5% capital against what they think the security is worth -- which is an awful lot of money you're being restrained from lending.
I think simply eliminating fair value accounting isn't a good idea. While I understand the problem it poses, without other regulatory changes, getting rid of this methodology would create others. Maybe if some other changes, like the ones I mentioned, accompanied the elimination of this accounting concept, I could be persuaded. But then, I'd still feel bad for investors trying to evaluate what a bank's assets are really worth, when every bank uses different subjective measures based on internal assumptions.