Financial and economic wealth are largely a function of expectations. A factory is valuable because I expect that it will help produce something that people want; a house is valuable because I expect that it will be a place people want to live. If people suddenly stop wanting to buy what the factory produces — or if they stop wanting to live in that house — then the factory or house loses its value, even if it is physically unchanged.

The size of a bank's liabilities to depositors is pretty clear in dollar terms, at least in principle; the bank owes a precise amount of dollars to depositors, and it owes it to them now. In a practical sense the liability is somewhat lower; the depositors won't all ask for their money right away, even if the bank charges fees. If the bank pays a low enough interest rate and charges fees, then, even with the costs of maintaining the bank accounts, the deposits provide the bank with a cheap source of funding that, in a true economic sense, reduces the size of the liability. Even in that sense, though, the true economic size of the deposits owed to customers is probably not a lot lower than the nominal size.

The asset side of the bank is murkier. If the bank has made a ten-year loan, the value of being entitled to that money depends on how things go over the next ten years; it depends on the ability of the bank to fund itself more cheaply than the interest rate on the loan,[1] and it depends on the ability (and sometimes willingness) of the borrower to actually pay it. The bank may even have investments in companies or real estate, and their value depends on the ability of those assets to provide things people want in the future. There are accounting rules about how we're supposed to guess at the value of these things, but these are merely conventional guesses.

These accounting guesses have some real force insofar as banking regulators impose solvency requirements on banks; the regulators want the bank's assets to be worth more than the liabilities, and use accounting guesses for at least some of those requirements.[2] The regulator's primary purpose is to protect the payment system, and particularly to protect the depositors' ability to get and use their deposits. The solvency requirements serve this in two ways: in the short run, if a bank is low on actual cash but has a lot of assets, it can sell assets or put assets up as collateral to borrow money to give to depositors. To the extent that this is our primary concern, the value of the assets should be reckoned based on the amount of money that could be acquired somewhat quickly by selling or borrowing against them. The primary purpose of the solvency requirement, however, is long-run: if the cash flows from the assets are anticipated to be reliably lower than the cash flows being paid on the liabilities, then eventually the bank will run out of cash, even if the depositors don't do anything weird. A regulation that is only worried about this concern is only worried about cash flows, not how much the asset could be sold for today.

There is an important sense in which, if the market value of your assets is lower than your liabilities, the market is saying that its best guess is that your cash flows in will not ultimately keep up with your cash flows out, but there is a fair amount of wiggle-room here. Market prices of assets bounce around a bit, and if the assets of a bank have gone down in the past three months, the bankers could well say, well, perhaps they will go back up in the next three months. Within certain constraints, the ability of the bank to hold onto cheap deposits does become important; even if the markets imply that funding costs over the life of the asset will eat up cash flows, if the bank can effectively borrow from depositors more cheaply, it may be able to survive. There are accounting rules that codify in certain ways how banks can get away with this; in particular, they can declare that they don't intend to sell certain assets, and if the market price changes they can ignore that change.[3] To some degree this feels like wishful thinking to me, but there's an element of wishful thinking in a well-capitalized bank as well; in one case you're hoping that the market is right, and that the cash flows in will be larger than the cash flows out, while in the other case you're hoping that the market is wrong. Even if we made banks use the market value for all of their assets,[4] the difference between a bank that is solvent and one that is insolvent is not a crisp one; there is a continuum, which is just one reason that the requirement is not just that assets exceed liabilities, but that they do so by some margin.

One new way to evade insolvency was introduced after the 2007–2009 financial crisis: the contingent convertible bond. "Bond" here is something of a misnomer, but they look like bonds in that they typically pay out a fixed interest rate and can be called in after a period of time, much like paying off a bond. Their key feature, though, is that they don't pay out if the bank's assets don't exceed the bank's liabilities by more than a certain specified margin. These "cocos" are designed to be liabilities as long as the bank can afford them, but to go away if liabilities are too large as a fraction of assets; if assets lose value, the cocos take the hit, and the depositors and other claimants on the bank are protected.

Different cocos work in different ways, perhaps in part with different purposes in mind. Most notably, some of them convert into stock or something similar when the asset-to-liability ratio gets too low. (These are the ones that are best called "contingent convertible bonds", though the term "coco" includes other securities that work rather differently.) Some of them convert into ... nothing. They go away. In each of these cases, though, they cease being liabilities, and thereby help restore the asset-to-liability ratio. If their purpose is to deal with long-term sustainability, rather than short-term sustainability (which is the purpose of liquidity regulations rather than capital regulations), then what matters about these things is their cash flow. A lot of them don't convert; they stop paying interest while the bank is in violation of its asset-to-liability requirements, but continue to sit around waiting to pay out again if the bank's situation improves. If the problem is a temporary market dip, or the bank has enough going-concern value that it can ultimately make it, these bonds don't get wiped out; they will lose some value when things look dicey, but it does relatively little harm to let them sit there dormant if the bank gets into trouble, only coming back if the bank's problems turn out to be temporary. It makes a lot of sense to me that they would largely work this way. Even from a short-term standpoint, a bond that works this way is a relatively small encumberance to selling or borrowing against assets, as the liability in practice remains small as long as there's much question of the bank's being able to repay secured loans.

The big exception, though — the point at which it seems like you have to make the coco's impairment final — is when the bank is being sold, especially if it's being sold as a matter of distress. In theory it might make a kind of sense to have the cocos paid off based on a sort of option value, but I can see why, as a practical matter, you might prefer that they be redeemed at par or zero. There's no longer an actual bank here (whose assets and liabilities could be assessed), so the best you could imagine is that it somehow continues to hedge the value of the assets the bank had when it was sold. Prospective buyers may well be averse to carrying around this strange option, and if the assets of the old bank are being folded into those of the purchasing bank, determining whether they recovered or not becomes onerous. Situations like this are usually messy and difficult as things are, and the value of being able to write this liability to zero in these situations seems compelling.

A week ago, the largest bank in Switzerland (UBS) acquired the second-largest bank in Switzerland (Credit Suisse). Both banks did the deal under duress from their primary regulator; Credit Suisse was on the brink of failure, but wanted to keep trying to recover, while UBS saw the balance sheet of Credit Suisse as unsafe at any price. Credit Suisse had some cocos that explicitly provided that they could be converted into nothing in a situation like this, and a lot of the holders of the cocos were disappointed to learn this. Cocos issued by banks in the European Union tend not to have such a provision. The Swiss cocos, indeed, had the provision that they would convert into nothing if the asset-to-liability ratio, as determined by accountants, were breached, even if the bank continued as a going concern. In actual fact, it is clear that UBS (and other potential suitors) thought that Credit Suisse's assets were worth a lot less than their accounting value; perhaps they should have been written down shortly before the takeover, anyway. After the fact, the fact that it was in the provisions of the bond (and was well within the spirit of how the bonds were intended to behave) means, of course, that they could do this; my assertion in this post is that such bonds should be written to be zero-able in this sort of situation, but that, outside of such forced-sale situations, the way the rest of Europe does things — with payments suspended, but the bond still sitting there, dormant, to potentially claim upside surprises if the bank recovers — makes more sense to me.