The primary limit on the size of the balance sheet of major commercial banks are what are called “capital adequacy requirements”, set down in the Basel accords. While each country also has its own set of reserve requirements, in practice the capital adequacy ones are the ones that operate first and constrain the largest banks. The idea behind the rules is to limit the degree of “leverage” major banks may use, particularly in riskier assets, and to ensure that the bank shareholders have enough of their own capital on the line to maintain incentives and the like.
The rules are moderately complex, dividing assets into risk classes and forms of capital into different “tiers”, but the practical aspects are considerably simpler. Against government bonds, expected to be essentially riskless (in a credit sense - of course they involve interest rate risk), basically no capital is required. Against most commercial credits with strong credit ratings, something like 4% of the value carried must be supplied by share capital. For riskier credits this rises, with an 8% level for some and much higher (like half) for some asset types (e.g. common stock in other firms).
Understand that assets always equal liabilities - that is an accounting identity. A bank only owns some interest earning assets - a loan, a bond, shares, whatever - because it funds that owning from some source. Mostly borrowing, from depositers, on its own debts in the credit markets, short and long term, etc. And last, the leftover “liability” of each firm, is its equity capital, or in other words the amount leftover as a tangible net worth of the firm’s common shareholders. It is that last “liability” (which it is, from the standpoint of the firm) which has to have some relation to the size of the whole balance sheet, all the assets combined.
If a bank operated entirely in the standard risk class assets and owned 25 dollars or Euros of investment grade bonds for each dollar in shareholder equity, then it would be right up to the allowed limit in capital adequacy terms. We’d say, the bank is “leveraged 25 to 1” and that is as far as it can go. Of course, if it moves some assets out of such loans and into government securities, it thereby reduces its capital requirements under the rules, since those do not require share capital “cover”, even at high leverage. On the other hand, with some of its credits in riskier forms, it might hit its capital adequacy limits before that level of leverage.
In normal times, a conservatively run bank might use 12 to 1, and an aggressive one 20 to 1, overall leverage of its share capital. They would then typically try to earn on the order of 1% above their liability cost on each asset - giving a competitive return on equity to the share capital portion. Some banks would try to earn a higher margin on a smaller leverage of riskier securities or loans, while others would use more leverage but very carefully managed risks (“matched books” in which liabilities come due at the exact same time and amounts and assets mature; derivative trades to lay off excess risks; holdings of government bonds despite their returns often being at or below the cost of funding them, held for liquidity due to the ease of selling or borrowing against them, etc).
Banks then earn profits and decide how to deploy those profits. If they let them all accumulate in share capital, they could potentially “grow” their whole balance sheet (meaning assets and liabilities, each increasing in tandem and in the same ratios and types) as fast as their return on equity allows. In practice, no sound bank attempts to grow that fast. Instead, most of the large ones moderate the growth of their balance sheets to about the expansion rate of the economy, and then pay out a large portion of the remaining earnings as dividends to shareholders. Another portion pays for share buybacks (some of which just offsets options issued to executives), and at other times, takeovers may consume portions of it (though those are more often funded by issuing new shares).
All that applies to an expanding bank in good economic times, where almost all loans made pay off, earning their funding cost or more. Banks set aside some reserves knowing that some loans will go sour and some bonds default etc. These are taken out of earnings too, and usually consume only a modest portion of them.
OK, now subject a bank to a significant fall in the value of some major asset class on the asset side of its balance sheet, while its liabilities remain unchanged. What happens? First point - leverage now works in reverse. The assets falling in price may be many times the total share capital of the bank, even if they comprise only a portion of the total assets. If the bank in a normal year earns 15 or 20% of the value of its capital, but the size of the declining asset class is, say, a quarter of its whole balance sheet, and that whole balance sheet is 20 times its share capital - then an asset 5 times the size of the capital is dropping n%, while you have 15 or 20% of n to cover that loss, in a normal year. If the decline is compressed in time, say six months, the earnings on everything else may come to only 10% of share capital - and a decline of even 2% in the price of a fifth of the balance sheet would send net earnings for the period to zero.
In the case of the subprimes, the portion is much lower than one fifth, but the price declines are much greater than a few percent. Specifically, asset backed securities that were rated A credits a year or two ago (because there are other classes ahead of them that must take any losses first), over the last six months declined in value from around 95 cents on the dollar of original loan amount, to around 30 cents on the dollar. Credits rated AAA, the highest possible, as recently as March, lost 20 percent of their value.
Now, when such declines happen, banks capital adequacy is strained in multiple ways. First, they are not earning what they were before. Second, they take direct losses on the assets themselves. Third, they upgrade their assessment of the risk of those types of assets, and no longer can consider 4% or so adequate against them - now they need 8% (for previously AAA credits) to half (for previously A ones that now look like they may default with very large losses).
How do banks meet the new strain? They can sell riskier assets and buy less risky ones. This tendency furthers the movement already started and is one reason the prices fall so far, in the newly disfavored assets. It also means government bonds go “on special” i.e. their prices rise significantly, compared to commercial credits. Right now in the US, a 3 month T-bill yields 2.75%, while a loan to another bank in the international money market might yield 4.5% or 5%. If banks are passing up the higher figure to invest in the lower one, they are doing so because they do not need capital reserves against the T-Bill.
There is another recent cause of strain, however, specific to the recent growth in asset backed securities. These were packages into “off balance sheet vehicles” - legally, special trusts that own a bunch of loans on mortgage collateral (or credit card receivables, or auto loan receivables, etc) and in turn raise the money to carry them by selling their own short term debt (“commercial paper”) at much lower interest rates. These trusts were the parking location for much of the mortgage paper created in the last 5 to 10 years. In addition to selling short term paper of their own, they typically had and have, lines of credit from their sponsoring banks - in other words, banks have promised to loan them money (on the collateral of said mortgage securities) if they need it, and in the meantime the trusts borrowed directly from the rest of the money market (money market funds, etc).
Well, what happened back in August or so is these trusts could no longer sell their commercial paper. Nobody wanted to lend to them, as the collateral they had to offer was known to be declining in price, and the solvency of some of them was in question (a few failed and are in liquidation). The banks then found themselves called upon to fufill those lines of credit, replacing the old external borrowing by the trusts. Effectively, the off balance sheet vehicles were and are coming back onto the balance sheet - and thus require new capital adequacy reserves. In some cases, they are also bringing losses with them - another round of the cycle above.
Fundamentally, borrowing at short term from the central bank cannot alleviate a capital shortage for a bank in such a position. The problem is not that they can’t borrow money - though it costs something of course, and more than they can earn on government bonds. The problem is that they have too many deposits, outstanding bonds, and loans from banks, central or not - on the one hand - and not enough share capital, on the other. Hence the deals you may have seen in which the various oil states are buying new blocks of stock from the likes of Citicorp.
Another solution would be to sell the declining mortgage securities, which have been declining throughout (with some recovery in September, and a very sharp drop from mid October to early November - occasioned in part by share price weakness in mortgage insurers and other parties that might have been expected to bear some of the losses and costs) - but to whom? And at what price? E*trade, which most thought of as a brokerage but whose online arm had been operating as a bank in the midst of the real estate bubble, sold its subprime portfolio for 27 cents on the dollar, to its largest existing bondholder (thus a party with an interest in avoiding a bankruptcy) - but it still has a large prime mortgage portfolio etc.
Another solution is to reduce riskier commercial credits and move the proceeds into government bonds - but all cannot do this, and doing so reduces the earnings of those who do it next (carrying 2.75% bills at 5% costs etc).
What the central banks - notably the ECB today - are now trying to do, is push so much new credit into the system at moderate rates - 4.25% being typical - that the rates offered for loans between banks (LIBOR) drop to their interest rate targets, and (they hope) reduce the spread between governments and these loans. Governments are in the meantime looking at regulatory changes and legislation to “work out” some of the mortgage mess. And financiers are looking at deals to raise new blocks of share capital for the banks (like Citi’s with gulf oil money). In the meantime, the banks continue to scramble to raise share capital and to reduce the need for it, and also put aside large loss reserves meant to absorb expected losses in mortgages.
I hope this helps.