I've addressed this subject before at greater length, but I want to put it in one post that people can easily link to and pass around.

Briefly: the fundamental cause of the bank crisis is not evil Republicans, lying Democrats, "deregulation," "affirmative-action lending," or even "ludicrous levels of leverage." A banking system is like a nuclear reactor: a complicated piece of engineering. If it's engineered right, it works 100% of the time. If it's engineered wrong, it works 99.99% of the time, and the other 0.01% it coats the entire tri-state area in radioactive strontium.

The bank crisis is an engineering failure. Its fundamental cause is a humble bug. Once we find the bug, we can start to ask: who is responsible for this bug? Who wrote the code? Who rolled back the fix? That discussion, though fascinating, is out of scope here.

Another analogy is the Space Shuttle disasters. Challenger had a bad booster O-ring; Columbia's wing was hit by falling foam. The level of discourse we're hearing now on the crisis tends to be "the Space Shuttle was Nixon's idea" or, at best, "Columbia's wing melted through and fell off." This is not an engineering analysis. It is point-scoring at best, anti-information at worst.

I believe I know what the bug is. It was first identified by the 20th-century economist Ludwig von Mises, capo of the Austrian School. Mises was an excellent writer, as eloquent as Marx and far more sensible, and it's unsurprising that there is a large Internet cult devoted to his work.

I am going to assume you are not a member of this cult. If there was a video of Mises walking on water, I might be tempted to take his pronouncements for granted. Since no such tape exists, they have to be explained and justified.

But we do need to start with Mises, because he was the first to solve the problem. Almost a hundred years ago, in his Theory of Money and Credit, he wrote:

For the activity of the banks as negotiators of credit the golden rule holds, that an organic connection must be created between the credit transactions and the debit transactions. The credit that the bank grants must correspond quantitatively and qualitatively to the credit that it takes up. More exactly expressed, "The date on which the bank's obligations fall due must not precede the date on which its corresponding claims can be realized." Only thus can the danger of insolvency be avoided.

Let's call the sentence in quotes Mises' rule. A banking system which obeys it is Misesian. We do not have a Misesian banking system - and that's the bug

Basically, imagine that there were two kinds of nuclear reactors - fission and fusion, perhaps. Fission reactors work 99.99% of the time. Fusion reactors work 100% of the time. The reason our society gets its power from fission reactors is that our reactor experts are fission experts. Therefore, we have resigned ourselves to having fission reactors, plus a large fleet of mobile power-washers to clean up the radioactive strontium every ten or twenty years. If you ask either the reactor engineers or the cleanup crews about the possibility of switching to fusion, the best answer you'll get will be something like "waah?" There are many worse.

Let's consider the sentence again. "The date on which the bank's obligations fall due must not precede the date on which its corresponding claims can be realized." Mises' rule of banking. Let's explain these terms and the reasoning behind them.

A "bank" is a financial middleman. It borrows from you and lends to someone else. When you "deposit" money "in" a bank, you are actually lending money to the bank. The bank does not keep this money in a big cardboard box. (I really hope this is not news to you.) It lends it to someone else - call him Dwight.

The bank's "obligation" is its agreement to repay you your loan. Its "claim" is Dwight's agreement to pay back his loan. (And your claim is the bank's obligation.) So what Mises is saying is that the bank must not agree to return your money (plus interest) before Dwight returns his money (plus interest).

Because, duh, it doesn't have it yet. Sounds obvious, right?

Of course, banks do not match individual claims and obligations in this way. If this is the way it worked, you and Dwight could save time and money by cutting the bank out of the loop. In reality, a bank borrows from and lends to thousands if not millions of people, which allows it to meet its obligation to you even if a few Dwights turn out to be deadbeats. Nonetheless, we can make the same obvious statement: by the time the bank needs to pay you, it needs to have collected from enough Dwights in order to have the money to pay you. Duh.

A more general way to describe Misesian banking is that the bank's plan to fulfill its obligations must not involve any implicit transactions. For example, if the bank promises to give you your money back in a week, and Dwight promises to give the bank its money back in two weeks, the bank has an implicit transaction. At the end of the first week, it needs to borrow money from someone else in order to repay you. That someone else might just be you, in which case you are rolling over - that is, renewing - the loan. But this is your decision, and the bank cannot know that you will roll over. After all, presumably there is a reason you selected a one-week loan.

We observe that in Misesian banking, the duration of a loan is as important as its amount. To balance a one-week obligation with a two-week claim is to balance an apple with an orange. It is just, not, done. Recall Mises' statement: the credit that the bank grants must correspond quantitatively and qualitatively to the credit that it takes up. That means it can't have an apple on the right and an orange on the left. (And what happens if it breaks this rule? Ha. We'll find out.)

A more naive approach to banking might just add up the claims on the left side of the page, add up the obligations on the right side, note that the sum on the left exceeds the sum on the right, and be satisfied. This would be "corresponding quantitatively" - but not "qualitatively." In Misesian banking, the bank makes sure its structure of claims allows it to satisfy its structure of obligations, as is, without implicit transactions. (Another kind of implicit transaction might be a currency conversion.)

Let's look slightly more closely at the loan market. We'll start with the obvious and segue into the not so obvious.

When you lend, you are exchanging present money for a claim to future money. Even if you know that this claim is perfectly good, you have no reason to make the trade unless you are getting more money in future than you give up in present - otherwise, you would just keep the money in your big cardboard box. So, for example, you might trade $100 in 2008 dollars for $110 in 2009 dollars.

The $10, obviously, is your interest rate, or yield - 10%. If you thought the loan had a 10% chance of not being paid back - the default risk - you might add another $10 or so, to get the same expected return. And the year (from 2008 to 2009) is the maturity of the loan.

We are now in a position to ask a very interesting question: in a healthy lending market, assuming Misesian banking, and forgetting about default risk for a moment, how should yield vary by maturity? Should a longer-term loan carry (a) a higher interest rate, (b) a lower interest rate, or (c) the same interest rate?

I suspect that, just intuitively, you said (a). This is indeed the right answer. Let's see why.

The market for loans is set, like everything else, by supply and demand. Every loan has a lender and a borrower. The lender always prefers a higher rate. The borrower always prefers a lower rate. At any maturity, the market rate is that rate at which the number of dollars which lenders are willing to lend is exactly equal to the number of dollars borrowers are willing to borrow.

We can make a little graph of this market, putting maturity on the x-axis and yield on the y. The result is called the yield curve. At least in a free market, the yield curve will always slope upward - higher maturities will command higher interest rates. This is true for any set of lenders and borrowers, anywhere in the known universe. If they have Misesian banks in the Lesser Magellanic Cloud, their yield curves slope upward.

How can we possibly know any such thing? We know only one thing: interest rates are set by supply and demand. But we can make some elementary observations about lenders and borrowers, which are true by definition.

The first is that at the same rate, any lender will prefer a shorter maturity. Consider the choice between one-week and two-week lending. If both transactions had the same rate, you could just lend for one week, get the money back and lend it again. This gives you the option to use the cash at the end of the first week, an option that the two-week maturity does not provide. An option can never have negative value, so why not: you'll pick the one-week maturity.

The second is that at the same rate, any borrower will prefer a longer maturity. For a borrowing transaction to be profitable, some productive process must use the money and generate a return. The set of productive processes that can produce round-trip return at a maturity of one second is empty. Therefore, in Misesian banking, no one should want to borrow at a one-second maturity, because there is no lending at a zero rate and no way to finance a productive enterprise at any nonzero rate - however small.

As the maturity of the loan increases, so does the set of productive processes, and so does the demand to borrow. Without violating Mises' rule, you cannot finance a nine-month pregnancy with a one-month loan. You need a nine-month loan. Nine one-month loans in a row will not suffice, because the last eight are implicit transactions.

Thus, for a higher maturity there is less supply of lending, and more demand for borrowing. Less supply and more demand means higher price, which means a higher yield. Which means the yield curve slopes upward.

This concludes our explanation of Misesian banking. Now let's explain the crisis.

Again, we don't have a Misesian banking system. We have what might be called a Bagehotian banking system - after Walter Bagehot (pronounced "badget"), who wrote Lombard Street, the first description of how this system works.

Here is a nice, concise explanation of the Bagehotian system:

The essence of what banks do in normal times is to borrow short and to lend long. In doing so, they transform short-term assets into long ones, thereby creating credit and liquidity. Put differently, by borrowing short and lending long, banks become less liquid, thereby making it possible for the non-banking sector to become more liquid; that is, have assets that are shorter than their liabilities. This is essential for the non-bank sector to run smoothly

This appeared in the Financial Times on October 9, 2008. The author is one Professor Paul de Grauwe, who like Professor Mises, and unlike me, got paid to understand this matter.

Note the perfect inversion between the Misesian and Bagehotian theories. Mises, writing almost a hundred years ago, describing a banking system that did not exist in his time any more than it exists in ours, says: "Only thus can the danger of insolvency be avoided." De Grauwe, writing now, says: "This is essential for the non-bank sector to run smoothly."

Hm. We may not be sure whom to trust here, but we do know that neither of these gentlemen is stupid. So what gives?

First, let's decode what Professor de Grauwe is saying. He's saying that banks routinely violate Mises' rule - they borrow "short" (ie, with short-term maturities), and lend "long" (ie, with long-term maturities). In other words, they engage in what we call maturity transformation.

Because we know the shape of the yield curve, we know why MT is profitable. Short interest rates are lower than long interest rates. So if the rest of the world is practicing Misesian banking and you're practicing Bagehotian banking, you make a mint.

In fact, we can say even more than this: we can say that MT lowers long-term interest rates. In our stodgy, Teutonic Misesian bank, if someone wanted to borrow money for 30 years, we had to match him with a lender who wanted to lend money for 30 years. In our fast-paced, Anglo-Saxon Bagehotian bank, we don't care - we balance our balance sheet quantitatively, not qualitatively. We can match the borrower with any lender, and get a better rate.

This is how a classic, Wonderful Life-style deposit bank works. A so-called "deposit" is really a loan of instantaneous maturity, continuously rolled over - by not "withdrawing" the cash, you are really renewing the loan. In the classical Bagehotian model, this might be used to finance, say, a 30-year mortgage.

Bagehotian banking seems like a just plain better idea. Its profits can be distributed to lender and borrower alike, producing higher rates for the former and lower rates for the latter.

Unfortunately, there is a slight downside. As we said earlier: "duh, the bank just doesn't have the money yet." When a bank borrows for a month and lends for a year, how exactly does it complete the transaction? Is there a little time machine inside the vault?

By violating Mises' rule, the Bagehotian bank makes itself dependent on an implicit transaction: viz., finding someone to loan it money for eleven months. Let's look at the ways in which it can implement that transaction.

The easiest way is just by inducing you to roll over your loan. It finds someone, and that someone is you. The reason Bagehotian banks work 99.99% of the time is that lenders, especially individual lenders, tend to roll over their loans a lot. You make a bank deposit rather than buying a six-month CD, even though the CD pays a higher rate, because you are not sure you won't need the money before the six months is up. Often you find you didn't. In retrospect you should have bought the CD, but you had no way of knowing this at the time.

The bank can also sell the 1-year loan. But selling a loan is equivalent to finding a new lender. Again, it cannot be known on what terms this implicit transaction can be executed.

What we've identified here is the wad of duct tape in the nuclear reactor. A Bagehotian bank is not contractually sound, because it does not have a complete plan to carry out its obligations. It relies on implicit transactions. And when these transactions cannot be executed on the terms expected - poof. The duct tape catches fire. The reactor melts down. The bank has a run.

In a bank run, the lenders start to doubt collectively that the bank will not be able to execute on its obligations to all of them. The origin of the doubt could be concern about the bank's quantitative solvency - eg, its 30-year claims are subprime mortgages. Or it could just be a suspicion that the bank will experience a run. If there is a run, you want to be the first out.

What happens as a Bagehotian bank experiences a run? Let's assume that, before the run, the bank was still quantitatively solvent - the current market price of its claims exceeds the sum of its obligations. The only problem is that the claims mature far later than the obligations.

So the bank sells the claims on the open market. If it can sell them all at the market price before the run, it is fine - it can raise enough cash to pay off all its depositors.

But a market price is a market price. It is not magic. Introduce new supply into the market, or withdraw demand - and the price drops like a stone. The bank run changes the price of the claims that are being sold. It has to find a lot of new lenders - but the market price of everyone's claims is dropping. So all banks which hold claims in this market are becoming quantitatively insolvent. The bank run spreads to the entire market. Lenders run in the other direction. And so on.

The idea of the yield curve lets us visualize this in a particularly elegant way. Recall that Bagehotian banking, by transforming maturities, lowers long-term interest rates. It flattens the curve. At least as compared to the Misesian yield curve.

Think of this curve flattening as putting pressure on a spring. A Bagehotian banking system is, at all times, a bank run waiting to happen. And when the run happens, the spring explodes in the other direction - well past where the Misesian yield curve would have been. It will not stop at the Misesian level, because a Misesian banking system would never have made so many long-term loans. It will produce astronomical long-term rates.

(This is exactly what we see now in the mortgage-backed securities market. It is impossible to get a read on exactly what the risk-free interest rate is in this market, because by definition there are no risk-free securities in it. Maturity-transformed demand is (at present) no longer buying mortgage securities, but not all the holdings have been liquidated, and there is no maturity-matched banking system to provide baseline demand. So neither interest rate nor default risk can be computed from asset price - you are trying to solve one equation for two variables.)

This metaphor of the spring lets us understand Professor de Grauwe's perspective. He believes "[maturity transformation] is essential for the non-bank sector to run smoothly" because he is thinking empirically, rather than deductively. He simply notes that every time MT stops, the reactor fails and melts down, and the tri-state area receives its coating of strontium. His thought is not intended as a comment on a Misesian banking system which never initiates MT to begin with, an idea that has probably never come to his attention. He is a fission expert, after all.

The difficulty in transitioning from Bagehotian to Misesian finance is immense, which is probably a big part of why it's never been tried. The Misesian sees an enormous set of financial structures which violate Mises' rule. He sees no way to unwind them. Other than massive liquidation - the bank run as a virtuous purge of "malinvestments" (pretty much any investment is a loss if it has to be financed at 80% interest) - there is no obvious way to get from here to there. The reactor just has to blow, and the strontium has to be swept up. Or so at least is the conventional wisdom, and no one is really working on the problem.

Moreover, there is another way to save a Bagehotian banking system: find a new lender who can print infinite amounts of money. (Or, in a metallic standard, compel the acceptance of paper as equivalent to metal.) This friendly fellow is generally known as "the government," or more formally as a "lender of last resort."

The end result of Bagehotian banking is that, without any government protection, it is incredibly unstable and will melt down at a drop of the hat. With full government protection, it is stable, and it drives down long-term interest rates - just as if the government itself had been making the loans itself. The lender of last resort might as well be a lender of first resort. (There are no modern schools of economics which believe, as far as I know, that governments should print money and lend it.) And with wishy-washy, informal, wink-and-a-nod protection - which is what we had until the other day - these toxic qualities are combined.

And this is how we continually stumble forward with a broken, Victorian-era banking system, suffering the slings and arrows of bad financial engineering. The whole thing needs to be rebooted, if not reinstalled, and we simply don't have a political system - or an intellectual system - which is capable of this. But I digress.