The Myth of Systemic Risk

Remarks made at the St. Louis Banking Conference by Professor George Kaufman *

I am modifying my talk in light of what Bert Ely has said, in part to avoid repetition and in part to straighten out the audience. Where do I come from? I come from the perspective that banks are not special and that markets work for banks as they do for every other firm. I also come from the perspective that "systemic risk" in banking is not a threat and has not been a great danger in world history. It is a scare term, much like the use of the word "fire" in a crowded theater. Systemic risk is used shamelessly by regulators to justify their own actions, and by novelists and movie script writers to provide plots for horror stories. This is my bottom line, and if I had two hours I could go on and give you all the evidence.

I will give you a little bit of evidence in the time allotted me. The bottom line that I am going to come up with is just not to let the perfect be the enemy of the good. The FDIC Improvement Act ("FDICIA") has protection against too-big-to-fail ("TBTF") which is reasonably good, but it is not perfect. It is with us and we could strengthen it to make it better rather than eliminate it and start with something that we don't know how it would work.

Let me first follow up on what Bert said and start off with a definition of "too-big-to-fail" or TBTF. Too- big-to-fail is a poorly defined and much misunderstood term. In the United States, it is not what it says. In other countries, it is all over the ball park. In the United States, too big to fail does not mean that we do not fail banks. We fail banks. The only exception to this was the Continental Illinois Bank experience in 1984, in my home city of Chicago. All other banks are failed with losses to shareholders. The shareholders are basically wiped out in all bank failures.

What is too big to fail? It means that in failure resolutions, the uninsured depositors, and possibly other creditors, — Fed fund sellers and depositors at foreign offices — are made whole. They are protected by the FDIC, if the FDIC suffers losses. The FDIC does not suffer losses in all failures, but if it suffers losses, then too big to fail says that these losses will not be shared with the uninsured creditors. In other words, we have 100 percent deposit insurance for all creditors of the banks. This is something that Bert wants and something that I think is horrible. Since the introduction of the FDICIA legislation in 1991, the FDIC has not done this, but I will come back a little later to say that this may not be an indication that it may not do so in the future.

In other countries, too big to fail is often much broader and it is implicit rather than being explicit. Bert is perfectly right, FDICIA legislation makes it explicit, which I think is good, because it limits it. In many countries it includes protecting not only creditors but the shareholders, as in Japan.

What reasons are there for too big to fail or 100 percent protection of all bank creditors? There are a number of them. First, large banks are major suppliers of money and credit to business firms, households and governments, and disruptions are perceived costly. Second, large banks are closely interconnected and interrelated with other banks through interbank deposits, loans and payment clearing, and that losses are perceived, and let me emphasize the word "perceived," to (1) spread to other banks; (2) interrupt the payment system, and (3) spill over beyond banks in the financial system and possibly the macroeconomy and even abroad. That is, that there is contagion or systemic risk, including runs by depositors on all banks, which means economically healthy banks, as well as guilty banks. The result is economically insolvent banks. It is random contagion or random systemic risk that is scary. If it is only systemic risk to economically insolvent firms and banks, that is what the economic system and markets are all about.

Third, there is a fear of the unknown from a lack of understanding of how banks work. Primarily, I think because banks work with intangibles. This is in contrast to firms that work with tangibles, like an automobile firm where everybody understands what is going on in the manufacturing process, and, therefore, also understands what happens when there is a breakdown. If you don't understand how banks work, then breakdowns can be very scary if people tell you that the world is going to end.

I use an example of the unfamiliarity of how banks operate with my students in my beginning class. I say, let's assume that we toured an automobile plant, and we had a tour guide. If the tour guide told us that the wheels go on the roof of the car rather than on the bottom, we would laugh at the tour guide. We would say, "that is silly." We know this is not the way cars run. But if we toured a bank and the tour guide made just as silly a statement, we would say, "how interesting," because we don't understand how banks work. It is like religion in some way, because we are fearful of the unknown and what we don't understand. We are afraid to cross the border or the boundary if somebody says, "if that happens, you are going to die and the world will end." The lack of understanding of how banks operate contributes to the fear that we have of bank failures.

There is also a misunderstanding about how bank failure resolutions are conducted in the United States. As I will discuss later, resolutions are done in a way that the damage is minimized. The generalizations about bank failures or statements that we make about bank failures come from fear from other times and from other countries.

Lastly, there is a myth of how serious and damaging bank failures were in the "bad old times" before the safety net in the United States. Let me start by talking about the evidence of bank failures in the United States.

Since 1870, after the end of the Civil War, the annual bank failure rate in the United States has been about the same as the nonbank failure rate. Before the safety net, or the first safety net, which was the lender of last resort facilities of the Federal Reserve when it was established in 1914, the bank failure rate indeed was lower. The annual bank failure rate was lower than the annual nonbank failure rate, and no very large banks failed. Indeed, until the 1970s, bank failures in the United States were a small bank phenomena. It is only in recent years that large banks have ever failed. Also, losses to depositors at failed banks were smaller than losses to creditors at failed nonbanks. However, the annual volatility of bank failures was greater than for non-banks. Bank failures came in clusters. Clusters are very scary to us, whether it is a cluster of airplane crashes, or a cluster of fires, or a cluster of illnesses. They scare and upset us, and that is part of the reason that we are upset about bank failures.

There is evidence that the failure of banks did not spillover to economically solvent banks. Banks, depositors and other creditors at the margin were able to differentiate financially-sick from financially-healthy banks. Economically healthy banks were able to withstand deposit runs by themselves, because they had sufficient good assets. In addition, other healthy banks basically acted as clearing houses and recycled the funds lost by solvent banks through runs. Even in the period of the greatest banking crisis in the United States and severe bank runs, as in Chicago in June 1932, at the height of the Great Depression, depositors basically ran only on economically insolvent institutions. There was not a failure of a bank that was economically solvent at the beginning of the bank run period in early 1932. All economically healthy banks survived the runs that occurred in that period.

I offer my students $10 if they can give me an example of an economically solvent bank that was brought down by a run, either in the United States or elsewhere in the world. Now, it could be that $10 is not enough to motivate them. It could also be that they are sleeping through my lectures. But, I have also thrown this challenge out to audiences over years, not only in the United States, but elsewhere. I have yet to have somebody come up with an example, I am sure somebody eventually will, because, you should never say never. I am sure there has been a case somewhere where a run has brought down an economically solvent bank, but I don't know about it.

Markets seem to work. Let me give you some recent evidence. Let's go back to the Continental Illinois Bank in 1984. At the time it failed, it was the seventh largest bank in the United States and the largest correspondent bank in the United States. Twenty-three hundred other banks had some deposits or loans with the Continental. When it was resolved, the FDIC protected everybody, insured or not, even though the FDIC experienced losses. By the way, the Continental is the one case where the bank was not legally failed.

What if the FDIC had not protected the uninsured? Let's take a look at those 2300 banks. Thirteen hundred of those banks had deposits of less than $100,000, so they were protected by the FDIC insurance coverage. How about the other 1,000? Well, that depends on the loss that the Continental suffered. If the loss had been 10 cents on the dollar, which turns out to be more than two times the actual loss — Continental's loss was something like 4 cents on the dollar — no other bank would have failed. What if the loss had been larger? We have hardly ever, except small banks with major fraud, had a bank that has failed with a recovery rate of 40 percent. But even if the loss had been 60 percent, so the recovery rate is only 40 percent, only 27 other banks would have suffered losses greater than their capital.

I will give you another example. Recently, a simulation study of the Federal Funds market done by an economist formerly at the Board of Governors of the Federal Reserve System, who is now with the BIS in Switzerland, found that if the largest debtor bank on the Fed Funds market had defaulted in 1998 with a loss of 5 percent, which is about the loss of the Continental, no other bank would have failed. If the loss had been 40 percent, eight times the loss of the Continental, only two to six primarily smaller banks would have failed. In other words, the systemic risk is myth, as it doesn't happen because the market is informed.

There is also very strong evidence in the United States as well as in other countries that bank insolvencies per se do not lead to recessions in the macroeconomy. However, problems in the macroeconomy feed back on the banking system. It is true that the troubles in the banks then will exacerbate the problems in the macroeconomy, but the causation goes from problems in the macroeconomy to problems in banking. This is true in the recent East Asian and the Japanese crisis, where the bursting of the price bubbles in real estate and stock markets caused bank problems.

Because a bank failure's damage is small, I believe that too-big-to-fail, or 100 percent protection of creditors, is too dumb to continue. Let's talk about the damages and the distortions of continuing the system that my friend, Bert, would like to continue, because he talks as if he doesn't believe in markets.

There are a number of reasons that TBTF is too dumb to continue. One, it reduces market discipline. Who in this audience wants to reduce market discipline? As a result, TBTF increases bank failures, through increasing the moral hazard risktaking by banks, so it permits banks to behave badly. Two, and Bert is perfectly right here, it is unfair to small banks or uninsured depositors at small banks who are perceived not to be equally protected. We don't have a level playing field. Three, it increases the difficulties of regulatory discipline as there is increased political pressure on the regulators to forebear. Bert calls this a moral hazard problem, but I call this a principal-agent problem, as the regulators are the agents for the taxpayers and the healthy banks. This principal-agent problem, or regulators behaving badly, was the major reason for the large number and high cost of bank failures in the 1980s in the United States and for the plague of bank failures that we have seen worldwide recently. In other words, too big to fail increases bank risk, leads to a misallocation of financial resources, and eventually leads to large regulatory or government failures.

If I want to get rid of too-big-to-fail, what is the best way of doing it? The best way of doing it is to improve the FDICIA legislation, which was enacted in 1991. FDICIA introduces a government closure rule, which says that we resolve banks before their capital becomes negative, if possible. What does that mean? It means that if we resolve a bank, or, indeed, any firm, before its capital becomes negative, the losses are confined to shareholders. There are no losses to depositors or to creditors. This is the underlying theory of FDICIA.

What if we fail to do this or, if we cannot catch the banks early enough? FDICIA prohibits the regulators from protecting uninsured depositors unless there is a finding that not protecting them "would have serious adverse effects on economic conditions or financial stability." This determination has to be made in writing by twothirds of the Board of Governors of the Federal Reserve System, twothirds of the Directors of the FDIC, and by the Secretary of the Treasury, after consultation with the President of the United States.

If indeed this determination is made, and the FDIC takes a loss, then this loss has to be paid for by a special assessment on all other banks according to their assets. This means that the big banks are going to be paying to keep one of their competitors in business. I don't think too many big banks would be in favor of this approach. Plus, there has to be ample documentation by the FDIC and by everybody else. This allows a post mortem examination of why this action, which has led to a loss to the FDIC, was invoked. In other words, the too-big-to-fail provision is explicit, which I think is good. Implicit stuff is the worst thing you want in an economy, because then you don't know what the politicians are going to do. You want things explicit, because that is strong and provides the right incentives.

Since 1991 and the introduction of FDICIA, all uninsured depositors at resolved banks with a loss to the FDIC have been sharing the loss with the FDIC. We no longer have a 100 percent protection policy. However, this is not a perfect test of ending too big to fail, because we haven't had any really big banks fail. The largest bank that failed since 1991 is First City in Texas, which had about $10 million at the time of failure.

The second protection is to require the resolution of big banks with minimum losses through procedures that produce minimal losses. This is done in the United States. But, people do not know how we resolve banks in the United States. In the United States, deposits at failed banks are not frozen like they are in other countries. In other words, resolution does not mean physical closure of the bank. If this was the case, then there would be major disruptions and the fears of bank failures would be justified.

Instead, the very next business day, the FDIC pays all insured depositors in full and advances the estimated recovery amount to the uninsured depositors. The FDIC can do this because the bank examiners are in the bank for some time before the bank fails under the prompt corrective action provisions of FDICIA. Indeed, they are preparing this information for banks that are going to bid on this failed bank. This is not true in other countries, where insured depositors very frequently have to wait for the government to pay them and the uninsured depositors have to wait until the receiver, generally a private receiver, collects the funds. Moreover, under FDICIA, the resolution is required before capital reaches zero. Under the current legislation, resolution is required when the equity capital asset ratio is 2 percent. However, this margin for error is small. But, if there is a failure, any losses should be relatively small.

The alternative proposals to deal with or to limit too-big-to-fail are far inferior and much more costly, because we are dealing with things we don't understand. FDICIA is something that we have. The point also has to be made that in any financial emergency, or what people view as a financial emergency, the policy makers are not going to follow the rules. They are going to do things on an ad hoc basis. Under those conditions, the best that can be done is to require documentation and to examine the documentation ex post.

The most effective strategy now to limit too big to fail and to allow markets to work is to strengthen FDICIA in a number of ways. One way to strengthen FDICIA is by increasing the capitalization trip-wire value that is required for resolution. The current level of 2 percent book value capital ratio is too low, because it is book value and not market value, and things could change rather quickly. I would increase this trip-wire to maybe 3 or 4 percent on a market value basis.

Another strategy would be to impose greater penalties on Federal Reserve lending through the discount window to insolvent or near insolvent banks. Such lending was done in the 1980s with costly results. Finally, we could increase the penalties on regulators for not acting promptly and effectively as required by the law. This may not be perfect, but we should not let the perfect be the enemy of the good. It will work.

* Professor Kaufman teaches in the School of Business of Loyola University Chicago, and consults with the Federal Reserve Bank of Chicago. These views do not represent those of the Federal Reserve Bank of Chicago or the Federal Reserve System.


2001 The Federalist Society