January 29, 2009 4:57 PM
My comment 1: Two questions:
(i) How is counterparty risk different from credit risk? They both involve the risk that a counterparty fails to make payment on a liability.
(ii) Unfunded derivatives are nothing more than a promise to make payment in the future. What “cash” are you talking about?
My comment 2: Just a note for the blog administrator: The delay between the submission of comments and the posting of comments is irritating.
Friday, January 30, 2009
Thursday, January 29, 2009
Understanding the Demon in CDS
This is a response to Charles Davi's post at The Atlantic Business Channel. You should read it before reading this post.
Davi writes: I often note that derivatives cannot create net losses in the system. That is, they simply transfer money between two parties. If one party loses X, the other gains X, so the net loss between the two parties is zero.
This argument implicitly assumes government subsidization of “too big to fail" banks. All derivative contracts create counterparty risk, which is just another name for the credit risk associated with the derivative. Because sudden price movements (such as we have seen regularly in financial markets lately) may force the debtor counterparty to default on the CDS (and declare bankruptcy) before enough collateral has been transferred to the creditor counterparty to cover the obligation, one needs to consider what happens in the circumstance that the debtor (or losing) counterparty defaults suddenly: then not only does X contribute to the debtor counterparty’s unpayable obligations, the creditor counterparty does not gain X. Conclusion: not only can “poorly underwritten mortgages can create net losses,” but a poor choice of CDS counterparty can also create net losses – and be evidence of inefficient allocation of capital
The only way to avoid net losses as a potential outcome of trading derivatives is to make it impossible for debtor counterparties to go bankrupt. Effectively that is what the government has done by guaranteeing the debt of the major money center banks. The fact that Davi’s argument can be sustained is not a function of the nature of derivative markets, but rather of State intervention in derivative markets.
I can agree with Davi that the ability “to express a negative view of mortgage default risk” contributed to the popping of the mortgage bubble that was created by the securitization process. However, the fact that popping the bubble took a long time (specifically from January 2006 to August 2007) meant that "too big to fail" financial institutions were able to rack up huge losses by selling CDS insurance (often indirectly through off-balance sheet entities) before the bubble was good and truly popped.
When Davi writes “depending on whether the synthetics are fully funded or not, the principal investment will go to the Treasuries market or back into the capital markets respectively,” he exhibits a fundamental misconception about the nature of CDS that David Harper has pointed out repeatedly on Davi’s Derivative Dribble website.
It is obvious to anyone who compares the size of the CDS market with the size of securities markets in the US (and indeed the world) that “the principal investment” (i.e. the notional value of the CDS) does not get invested in Treasuries or capital markets. When synthetics are not funded, they are backed only by the promise of the protection seller to make good on the contractual obligation – that is, they are pure counterparty (aka credit) risk.
The buyer of protection sold by a partially funded synthetic or hybrid CDO is short the credit risk of the entities referenced in the CDS sold by the CDO and long the credit risk of the super senior counterparty. (The super senior tranche of a synthetic or hybrid CDO is almost always an unfunded CDS.)
Davi makes this claim: “Only when interest rates on MBSs drop low enough, along with the price of protection on MBSs, will protection buyers enter CDS contracts.” This is true only if the efficient markets theorem holds and interest rates instantaneously reflect expectations of default. In general, when expectations of MBS default rise, protections buyers will enter CDS contracts over time driving up the price of protection on CDS and mortgage interest rates. As noted above it took a year and half for this process to play out in financial markets.
The problem with CDS is not intrinsic to the derivative contract itself, but the fact that "too big to fail" banks were allowed to sell protection (when the buyer did not have an insured interest) in the market.
Davi writes: I often note that derivatives cannot create net losses in the system. That is, they simply transfer money between two parties. If one party loses X, the other gains X, so the net loss between the two parties is zero.
This argument implicitly assumes government subsidization of “too big to fail" banks. All derivative contracts create counterparty risk, which is just another name for the credit risk associated with the derivative. Because sudden price movements (such as we have seen regularly in financial markets lately) may force the debtor counterparty to default on the CDS (and declare bankruptcy) before enough collateral has been transferred to the creditor counterparty to cover the obligation, one needs to consider what happens in the circumstance that the debtor (or losing) counterparty defaults suddenly: then not only does X contribute to the debtor counterparty’s unpayable obligations, the creditor counterparty does not gain X. Conclusion: not only can “poorly underwritten mortgages can create net losses,” but a poor choice of CDS counterparty can also create net losses – and be evidence of inefficient allocation of capital
The only way to avoid net losses as a potential outcome of trading derivatives is to make it impossible for debtor counterparties to go bankrupt. Effectively that is what the government has done by guaranteeing the debt of the major money center banks. The fact that Davi’s argument can be sustained is not a function of the nature of derivative markets, but rather of State intervention in derivative markets.
I can agree with Davi that the ability “to express a negative view of mortgage default risk” contributed to the popping of the mortgage bubble that was created by the securitization process. However, the fact that popping the bubble took a long time (specifically from January 2006 to August 2007) meant that "too big to fail" financial institutions were able to rack up huge losses by selling CDS insurance (often indirectly through off-balance sheet entities) before the bubble was good and truly popped.
When Davi writes “depending on whether the synthetics are fully funded or not, the principal investment will go to the Treasuries market or back into the capital markets respectively,” he exhibits a fundamental misconception about the nature of CDS that David Harper has pointed out repeatedly on Davi’s Derivative Dribble website.
It is obvious to anyone who compares the size of the CDS market with the size of securities markets in the US (and indeed the world) that “the principal investment” (i.e. the notional value of the CDS) does not get invested in Treasuries or capital markets. When synthetics are not funded, they are backed only by the promise of the protection seller to make good on the contractual obligation – that is, they are pure counterparty (aka credit) risk.
The buyer of protection sold by a partially funded synthetic or hybrid CDO is short the credit risk of the entities referenced in the CDS sold by the CDO and long the credit risk of the super senior counterparty. (The super senior tranche of a synthetic or hybrid CDO is almost always an unfunded CDS.)
Davi makes this claim: “Only when interest rates on MBSs drop low enough, along with the price of protection on MBSs, will protection buyers enter CDS contracts.” This is true only if the efficient markets theorem holds and interest rates instantaneously reflect expectations of default. In general, when expectations of MBS default rise, protections buyers will enter CDS contracts over time driving up the price of protection on CDS and mortgage interest rates. As noted above it took a year and half for this process to play out in financial markets.
The problem with CDS is not intrinsic to the derivative contract itself, but the fact that "too big to fail" banks were allowed to sell protection (when the buyer did not have an insured interest) in the market.
Securitization and the Mortgage Bubble
The evidence points to the likelihood that private sector securitization resulted in an increased issuance of private sector adjustable rate mortgages.(See figures 14, 15 and 17 in the link.) This is all the more remarkable, since the Fed was raising short-term rates from 2004 to 2006 and the interest rate on adjustable rate mortgages was rising. In economic terms, there was apparently a huge increase in demand for adjustable rate mortgages during this time period.
It appears that there was such a huge demand for mortgage paper for banks and the vehicles they were creating to use to go long, that brokers were paid a premium to issue no-income, no-job, no-asset, no-money down loans. No wonder there was an increase in demand for high interest rate mortgages!
Of course if the banks were holding those loans on their balance sheets, you can be pretty sure they would have responded to the rise in interest rates by reducing their issuance of loans (which is what Fannie Mae and Freddie Mac did -- see the Ofheo link above). Ergo the culprit is likely to be private sector securitization.
It appears that there was such a huge demand for mortgage paper for banks and the vehicles they were creating to use to go long, that brokers were paid a premium to issue no-income, no-job, no-asset, no-money down loans. No wonder there was an increase in demand for high interest rate mortgages!
Of course if the banks were holding those loans on their balance sheets, you can be pretty sure they would have responded to the rise in interest rates by reducing their issuance of loans (which is what Fannie Mae and Freddie Mac did -- see the Ofheo link above). Ergo the culprit is likely to be private sector securitization.
Wednesday, January 28, 2009
A thought on the bailout
If the government feels that it simply must bailout banks that are exposed to toxic derivatives, then this is the policy I would recommend: Any bank that holds a liability of a bank that has received TARP money may sell that liability to the government at some price, however the market value (ex bailout) of that asset will be assessed at zero plus the market value of any real assets that secure the liability. The government debt does not need to be paid for twenty years (which should allow it to qualify as a capital) but will accrue interest at a non-punitive rate until it is paid. The government is first in line to collect payment on the bank's obligations (i.e. the government loan is debtor in possession financing). Without an explicit waiver from the government, the bank can post no collateral, pay no interest, dividends or employee compensation over $500,000 cash per person per year (stock options not exercisable for five years not included) until the debt to the government is repaid with interest. The government may choose to implement standard waivers to, for example, keep the Treasury repo market in operation and perhaps even to support certain markets in bank debt, but the exceptions to government as a priority creditor should be very limited.
Monday, January 26, 2009
Is Systemic Risk Insurance the Problem or the Solution?
Ricardo Caballero asks “Whether it is feasible to run a (nearly) capital-less financial system until panic subsides?”
It’s my impression that this experiment is precisely what was tried in 2008. Regulators across the world engaged in policies of regulatory forebearance – finding ways to protect banks from writing down too many of their assets, protecting them from taking off balance sheet vehicles on balance sheet and possibly even encouraging banks to not to disclose material information to shareholders. One of the consequences of regulatory forebearance is a general suspicion that many of the largest banks are in fact insolvent -- i.e. that at least during the last quarter of 2008, the US was operating a "capital-less" financial system.
Despite the general suspicion that we now have a "capital less" financial system, government insurance of large chunks of the financial system has meant that stock and credit markets are showing few signs of fear that systemic collapse is imminent (now in late January). On the other hand, after three months of extraordinary government intervention, the prices of many assets are not recovering. So one question is: What is the evidence that these are "panic" prices, and shouldn't we consider the possibility that assets are simply being fairly marked?
Ricardo Caballero argues that the problem is that the government insurance being offered to the financial system is not sufficiently comprehensive: The government needs to become the explicit insurer for generalised panic-risk (which can also be called systemic risk, or macroeconomic risk). There are two questions that his essays do not address: Is a policy of systemic risk insurance feasible for any government? Are the moral hazard consequences of the systemic risk insurance policy so damaging that the policy's costs exceed its benefits?
My own view is that it is precisely because the financial community perceived the government as a provider of systemic risk insurance that we find ourselves in a crisis of this magnitude. This view is based on a belief that it is precisely the risk of general financial instability that causes financial institutions to make investment decisions that are conservative enough to protect the stability of the system. In other words, I believe that the fundamental flaw that developed in our financial system was the view that: "Financial institutions specialise in handling risk but are not nearly as efficient in dealing with uncertainty." and that the government should provide insurance for all financial institutions against uncertainty or systemic risk.
The Federal Reserve under Alan Greenspan encouraged this view. The actions of Gerald Corrigan, president of the New York Fed, in 1987 were a watershed. The stock market crashed, the investment banks were caught unprepared and were unable to meet margin calls on their bank loans. The banks were unwilling to offer unsecured credit and bankruptcy for at least one investment bank and possibly all of them loomed. The Federal Reserve apparently decided that the investment banks were, like Continental Illinois in 1984, "too big to fail". Corrigan convinced the bankers to extend unsecured loans to the investment banks. The investment banks were saved and for the first time in history the Fed used its authority as lender of last resort to the commercial banking system to support investment banks.
It did not pass unnoticed that the Federal Reserve succeeded in rescuing the investment banks in 1987, only because moral suasion was effective. In 1991, section 473 of FDICIA, the law that revised bank regulation after the savings and loan crisis, relaxed the collateral criteria of the Federal Reserve Act's emergency lending clause -- and the Congressional debate indicates that the purpose of this change was to benefit securities firms (see page 5 here). Thus there was legislative support for the expansion of the "too big to fail" doctrine to the investment banks.
1998 was another watershed. Long Term Capital Management was a hedge fund with assets of more than $125 billion. These assets were supported by less than $5 billion in equity -- LTCM owed the financial system more than $120 billion. When bankruptcy loomed for LTCM, it also put some investment banks at risk. It became clear that the policy that investment banks were "too big to fail," also meant that certain hedge funds were "too big to fail." Just as in 1987, the Federal Reserve put pressure on Long Term Capital Management's counterparties to prevent the firm's failure. The banks put together a rescue package for LTCM, but it escaped no one's notice that, if the banks had failed, the Federal Reserve had the legal authority under FDICIA to lend directly to LTCM itself.
Thus, by the end of 1998 the Federal Reserve had made it clear that its job was not just to support commercial banks through a financial crisis, but also to support investment banks and financial markets. This extension of the "too big to fail" safety net to large securities firms almost certainly lowered their borrowing costs relative to a market rate. As Brad Setser put it after the Lehman crisis:
Coincident with the establishment of this government guarantee was the exponential growth of private debt in the United States.
Prof. Caballero is taking the position that this implicit guarantee needs to be extended, so that all financial institutions are protected from uncertainty and systemic risk. He apparently hopes that financial regulation can offset the propensity of such a system to be overleveraged. I suspect that only draconian regulatory measures could possibly keep an insured financial system stable. It would certainly represent a huge change from the historical experience of the British and American financial systems.
It has long been understood that a firm's first defense against uncertainty in financial markets is to be well-capitalized. One reason to require firms to handle uncertainty on their own is precisely to encourage them to be well-capitalized and to promote systemic stability by reducing the degree of leverage in the system. There are those who claim that this is inefficient, but that is necessarily a matter of definition -- a system which is efficiently capitalized from a long run perspective will appear to be over-capitalized if measures are taken over a short horizon during which the uncertainty is not realized.
When the concept of a lender of last resort was born (see Thornton 1801 and Bagehot 1873), the lender of last resort (that is, the Bank of England) was a private institution. And one of the reasons Bagehot wrote Lombard Street was to make the point that a system of last resort lending is inherently fragile -- that uncertainty always lingers somewhere over the horizon. First Bagehot quotes Ricardo:
Bagehot considers any credit system that relies on a last resort lender to be frail -- something that is to be protected and cherished, because once it is broken up, it will take generations to develop anew.
Professor Caballero is proposing a model completely at odds with that of the traditional model of last resort lending. He claims that there is a source of security for the banking and financial system -- the credit of the government. Perhaps Prof. Caballero understands finance better than Bagehot, but just in case he is wrong, my question is: What is plan B? If Bagehot is correct and government provided systemic insurance may have as a consequence the destruction of the credit of the government rather than saving the credit of the financial system, what does Prof. Caballero propose that we do next?
In my opinion all policy makers need to understand one thing: The complete collapse of Western financial infrastructure, including the credit of American and European governments is possible. Every decision needs to be made with an understanding that the financial system is fragile: the credit of our governments should be treated as a precious resource -- to be used, yes, but to be used with prudence and a view towards conservation.
If Prof. Caballero had proposed carefully scrutinizing the different classes of assets to select which ones merited government insurance and which need to be written down to zero, I would find his argument far more convincing. Instead he writes:
The difference between his view and mine is simple: he only sees a possibility of collapse if the government does too little. I, however, also see a possibility of collapse if the government does too much.
Update 1-29-09: I've just read (the introduction of) Rethinking Capital Regulation by Kashyap, Rajan and Stein. They have a reasonable proposal for systemic risk insurance, though I'm not sure it would work in practice. First, their proposal is prospective, a solution for the future, not a resolution of the present. Second, their insurance is fully funded.
The only problem with the idea that I can see is that their insurance policies have a specific term at which point the insurer has the option of refusing to renew the policy. I would expect the policies to disappear just when they are needed. Perhaps as the authors argue 5 year overlapping policies would help solve this problem -- but I wouldn't be surprised if market perspicacity makes the insurance much less effective than would be desirable.
Could rising insurance premiums help trigger a crisis?
Update 2-10-09: Professors at NYU Stern also view the too big to fail guarantees as an important cause of the crisis.
It’s my impression that this experiment is precisely what was tried in 2008. Regulators across the world engaged in policies of regulatory forebearance – finding ways to protect banks from writing down too many of their assets, protecting them from taking off balance sheet vehicles on balance sheet and possibly even encouraging banks to not to disclose material information to shareholders. One of the consequences of regulatory forebearance is a general suspicion that many of the largest banks are in fact insolvent -- i.e. that at least during the last quarter of 2008, the US was operating a "capital-less" financial system.
Despite the general suspicion that we now have a "capital less" financial system, government insurance of large chunks of the financial system has meant that stock and credit markets are showing few signs of fear that systemic collapse is imminent (now in late January). On the other hand, after three months of extraordinary government intervention, the prices of many assets are not recovering. So one question is: What is the evidence that these are "panic" prices, and shouldn't we consider the possibility that assets are simply being fairly marked?
Ricardo Caballero argues that the problem is that the government insurance being offered to the financial system is not sufficiently comprehensive: The government needs to become the explicit insurer for generalised panic-risk (which can also be called systemic risk, or macroeconomic risk). There are two questions that his essays do not address: Is a policy of systemic risk insurance feasible for any government? Are the moral hazard consequences of the systemic risk insurance policy so damaging that the policy's costs exceed its benefits?
My own view is that it is precisely because the financial community perceived the government as a provider of systemic risk insurance that we find ourselves in a crisis of this magnitude. This view is based on a belief that it is precisely the risk of general financial instability that causes financial institutions to make investment decisions that are conservative enough to protect the stability of the system. In other words, I believe that the fundamental flaw that developed in our financial system was the view that: "Financial institutions specialise in handling risk but are not nearly as efficient in dealing with uncertainty." and that the government should provide insurance for all financial institutions against uncertainty or systemic risk.
The Federal Reserve under Alan Greenspan encouraged this view. The actions of Gerald Corrigan, president of the New York Fed, in 1987 were a watershed. The stock market crashed, the investment banks were caught unprepared and were unable to meet margin calls on their bank loans. The banks were unwilling to offer unsecured credit and bankruptcy for at least one investment bank and possibly all of them loomed. The Federal Reserve apparently decided that the investment banks were, like Continental Illinois in 1984, "too big to fail". Corrigan convinced the bankers to extend unsecured loans to the investment banks. The investment banks were saved and for the first time in history the Fed used its authority as lender of last resort to the commercial banking system to support investment banks.
It did not pass unnoticed that the Federal Reserve succeeded in rescuing the investment banks in 1987, only because moral suasion was effective. In 1991, section 473 of FDICIA, the law that revised bank regulation after the savings and loan crisis, relaxed the collateral criteria of the Federal Reserve Act's emergency lending clause -- and the Congressional debate indicates that the purpose of this change was to benefit securities firms (see page 5 here). Thus there was legislative support for the expansion of the "too big to fail" doctrine to the investment banks.
1998 was another watershed. Long Term Capital Management was a hedge fund with assets of more than $125 billion. These assets were supported by less than $5 billion in equity -- LTCM owed the financial system more than $120 billion. When bankruptcy loomed for LTCM, it also put some investment banks at risk. It became clear that the policy that investment banks were "too big to fail," also meant that certain hedge funds were "too big to fail." Just as in 1987, the Federal Reserve put pressure on Long Term Capital Management's counterparties to prevent the firm's failure. The banks put together a rescue package for LTCM, but it escaped no one's notice that, if the banks had failed, the Federal Reserve had the legal authority under FDICIA to lend directly to LTCM itself.
Thus, by the end of 1998 the Federal Reserve had made it clear that its job was not just to support commercial banks through a financial crisis, but also to support investment banks and financial markets. This extension of the "too big to fail" safety net to large securities firms almost certainly lowered their borrowing costs relative to a market rate. As Brad Setser put it after the Lehman crisis:
Much of the infrastructure of modern finance in effect rested on an expectation of a government backstop for the creditors of large financial institutions – a backstop that allowed a broad set of institutions to borrow short-term at low rates despite holding large quantities opaque and hard to value assets on their balance sheets.
That observation has a number of implications, not the least that the leverage – and resulting capacity for outsized profits — of some parts of the financial sector was made possible by the expectation that the government would protect the key creditors of the financial system from losses.
Lehman’s default shattered this implicit guarantee.
Coincident with the establishment of this government guarantee was the exponential growth of private debt in the United States.
Prof. Caballero is taking the position that this implicit guarantee needs to be extended, so that all financial institutions are protected from uncertainty and systemic risk. He apparently hopes that financial regulation can offset the propensity of such a system to be overleveraged. I suspect that only draconian regulatory measures could possibly keep an insured financial system stable. It would certainly represent a huge change from the historical experience of the British and American financial systems.
It has long been understood that a firm's first defense against uncertainty in financial markets is to be well-capitalized. One reason to require firms to handle uncertainty on their own is precisely to encourage them to be well-capitalized and to promote systemic stability by reducing the degree of leverage in the system. There are those who claim that this is inefficient, but that is necessarily a matter of definition -- a system which is efficiently capitalized from a long run perspective will appear to be over-capitalized if measures are taken over a short horizon during which the uncertainty is not realized.
When the concept of a lender of last resort was born (see Thornton 1801 and Bagehot 1873), the lender of last resort (that is, the Bank of England) was a private institution. And one of the reasons Bagehot wrote Lombard Street was to make the point that a system of last resort lending is inherently fragile -- that uncertainty always lingers somewhere over the horizon. First Bagehot quotes Ricardo:
'On extraordinary occasions, a general panic may seize the country, ... against such panic banks have no security on any system.' The [lenders of last resort] may last a little longer than the others; but if apprehension pass a certain bound, they must perish too.
Bagehot considers any credit system that relies on a last resort lender to be frail -- something that is to be protected and cherished, because once it is broken up, it will take generations to develop anew.
Credit is a power which may grow, but cannot be constructed. Those who live under a great and firm system of credit must consider that if they break up that one they will never see another, for it will take years upon years to make a successor to it. ... [W]e should look at the rest of our banking system, and try to reduce the demands on the Bank [of England] as much as we can. The central machinery being inevitably frail, we should carefully and as much as possible diminish the strain upon it.
Professor Caballero is proposing a model completely at odds with that of the traditional model of last resort lending. He claims that there is a source of security for the banking and financial system -- the credit of the government. Perhaps Prof. Caballero understands finance better than Bagehot, but just in case he is wrong, my question is: What is plan B? If Bagehot is correct and government provided systemic insurance may have as a consequence the destruction of the credit of the government rather than saving the credit of the financial system, what does Prof. Caballero propose that we do next?
In my opinion all policy makers need to understand one thing: The complete collapse of Western financial infrastructure, including the credit of American and European governments is possible. Every decision needs to be made with an understanding that the financial system is fragile: the credit of our governments should be treated as a precious resource -- to be used, yes, but to be used with prudence and a view towards conservation.
If Prof. Caballero had proposed carefully scrutinizing the different classes of assets to select which ones merited government insurance and which need to be written down to zero, I would find his argument far more convincing. Instead he writes:
With some dismay, I read that an enormous amount of time is being spent discussing what should be the price of the insurance and the first-loss threshold. It seems to me that given the extreme severity of the crisis and the asymmetries involved in failing in one or the other direction in each of these issues, the answers are rather obvious: The price of the insurance should be very low – say risk-neutral pricing plus 20 or 50 basis points of markup; and the first-loss threshold should be sufficiently low that no new capital will need to be raised in the short run if a loss arises.
The difference between his view and mine is simple: he only sees a possibility of collapse if the government does too little. I, however, also see a possibility of collapse if the government does too much.
Update 1-29-09: I've just read (the introduction of) Rethinking Capital Regulation by Kashyap, Rajan and Stein. They have a reasonable proposal for systemic risk insurance, though I'm not sure it would work in practice. First, their proposal is prospective, a solution for the future, not a resolution of the present. Second, their insurance is fully funded.
The only problem with the idea that I can see is that their insurance policies have a specific term at which point the insurer has the option of refusing to renew the policy. I would expect the policies to disappear just when they are needed. Perhaps as the authors argue 5 year overlapping policies would help solve this problem -- but I wouldn't be surprised if market perspicacity makes the insurance much less effective than would be desirable.
Could rising insurance premiums help trigger a crisis?
Update 2-10-09: Professors at NYU Stern also view the too big to fail guarantees as an important cause of the crisis.
Thursday, January 8, 2009
Fama states the obvious
Eugene Fama states the obvious: In a market economy, when a firm cannot honor its obligations equity and debtholders should bear the losses.
Monday, December 22, 2008
Why Credit Default Swaps are a Source of Systemic Risk
All financial assets create credit risk, because they create a situation where a debtor may fail to honor an obligation. In the absence of the asset there would be no risk that the debtor does not perform on his obligation to the creditor. Even insurance products fit this characterization, because the purchaser of insurance faces the risk that the insurer will not pay.
The dangers of credit have been recognized since the start of written history: early records of interest rate regulation are not at all uncommon. Only since the dawn of paper money, however, have societies felt the need to regulate the quantity of financial assets – and thus the quantity of credit risk in an economy.
In the early modern era, the regulation of credit risk was decentralized. It took the form of a general principle that all bankers, merchants and other tradesmen were expected to follow: this principle is now known as the real bills doctrine. This doctrine drew a clear distinction between real and fictitious bills. Real bills were IOUs that were created when the creditor delivered some product or real asset to the debtor, and thus real bills were tied to economic production and trade. When there was no transaction in the real economy connected with the bill, it was considered fictitious. Well known bankers (who were often traders, too) could issue large quantities of bills and it was very difficult for those who circulated these bills to know whether a bill was real or fictitious. For this reason, proof of issuing a fictitious bill was enough to end a banker’s – or indeed a trader’s – career. In 1772 when the Bank of England refused to discount the bills of the Alexander Fordyce, owner of the Ayr Bank, the directors explained that they did so because the quantity of bills already discounted at the Bank was so great that some of them had to be fictitious. The Ayr Bank closed in a matter of days.
The basic idea behind the real bills doctrine is simple: in order for the credit risk inherent in a financial asset to be socially valuable, it must be the case that the risk is tied to economic activity. When financial assets grow in number without reference to economic fundamentals, credit risk grows too. Centuries of experience with euphoria and collapse have taught us that the unrestrained growth of credit risk has one certain consequence: bankruptcy that spreads like a virus from one borrower to the next.
It is fairly easy to relate some modern financial assets to the real bills doctrine. Stocks and bonds exist to finance productive enterprises, thus they are real bills. Some assets are harder to fit in the framework: Insurance transfers risk from a party who is ill-equipped to bear it, to one with greater resources. For the purposes of this discussion, let’s call forms of risk that arise from the process of producing goods and services for sale, economic risk. Thus, the airline that is concerned that the price of fuel will rise, and the employee who worries that an accident at work could cause health problems that bankrupt him both face economic risk. We will treat the transfer of economic risk as a real activity, and insurance contracts that protect firms or individuals from the costs of economic risk as real bills.
Placing derivatives in the context of the real bills doctrine is a more complex problem. Some derivatives are like insurance, because they allow investors to transfer risk. The stock investor who wants to protect his investment from losses can buy put options. The bond investor can buy a credit default swap.
In rare cases a derivative contract can actually be used to eliminate certain forms of economic risk – the classic example is when a farmer and a baker agree to trade wheat at a fixed price six months in the future. This forward contract allows the farmer to be confident that he can borrow money today to plant his fields and the baker confident that an investment in kitchen equipment will be profitable. This, then, may be an example of financial asset that does not increase the aggregate amount of risk in the economy – instead it serves to convert economic risk into credit risk.
When derivatives are a zero-sum game
The same derivatives that can serve to transfer risk can also be used by individuals without any initial risk exposure to profit off of expected movements in the market. When a derivative is not used to transfer existing risk, it constitutes a zero sum transaction: whatever one party gains the other loses. This means that the transaction requires different beliefs about the value of the derivative on the part of the buyer and of the seller – since presumably the trade is taking place, because both expect to gain from it. In such a situation, the effect of asymmetric information on the price at which the transaction takes place has the potential to result in an inefficient exchange.
Some claim that the zero sum nature of these derivative transactions should not raise concern, since secondary markets for financial instruments such as stocks and bonds are also zero sum. This is true. But we should all recognize that the problem of asymmetric information in these markets led to the creation of organizations such as the NYSE and the SEC. These institutions exist to even the playing field. The NYSE regulates the prices at which stocks trade to ensure equal access (even if only through a broker) to the prices available on the stock exchange. The SEC requires firms to report accurate financial information so that purchasers of stocks and bonds are well-informed about the prospects of the firm.
Not only are many secondary markets highly regulated, but there is another crucial difference between them and markets for derivatives. In the case of secondary markets no financial asset is created, it is only traded. Thus on secondary markets the zero sum transaction does not involve the creation of credit risk.
By contrast, when a new derivative contract is traded in a zero sum transaction, credit risk is created. Thus, there is no escaping this conclusion: when a new zero sum derivative contract is traded, the aggregate amount of credit risk in the economy increases incrementally, while the productive possibilities of the economy do not increase at all.
Now, for some zero-sum derivative transactions it is certainly possible that there is a gain that offsets the incremental loss to society due to the creation of credit risk. I would propose, however, that the burden of proof for such a gain lies with the financial institutions who wish to engage in zero-sum derivative transactions.
The problem with fictitious bills
When a derivative is not used to transfer risk, but instead used by an individual without any initial risk exposure to profit off of expected movements in the market, we have a financial transaction without a corresponding economic transaction. Because such derivatives are not connected to economic activity, the only constraint on the quantity of their issue is the common-sense of the two parties to the contract. If these derivatives are being traded by fools, nothing prevents one or both of them from taking on obligations that exceed their ability to pay by multiples of one hundred, one thousand, or more.
Thus, derivatives are distinguished from stocks, bonds and insurance contracts by their ability to grow in an unconstrained manner, disconnected from the realities of economic activity. To monitor and restrict the growth of credit risk (and to reduce problems of asymmetric information), many derivatives markets are highly regulated.
For example, options and futures contracts are exchange-traded. Because every participant on the exchange trades with a single, central counterparty, the credit risk these traders are exposed to is limited to the risk that the central counterparty itself goes bankrupt. The counterparty in turn monitors the dealers in the market to make sure that they are not taking inappropriate risks. In fact, if the dealers have any large obligations, they are sure to hedge them in the underlying market. For example, if the dealers have sold a large quantity of call options on a certain stock, they may go ahead and buy the stock at current market prices to ensure that they are not exposed to large losses. If the quantities are large enough, this transaction can drive up the price of the stock – allowing demand on option markets to affect prices on the underlying market. The close connection between the two markets means that price signals will act to limit the growth of credit risk.
When derivatives markets are unregulated there is a genuine danger that credit risk will grow without bound. In order for this to happen, it must be the case that some firms are issuing liabilities without sufficient concern for the possibility that these liabilities may bankrupt the firm. For this reason, it is derivatives that involve either an upfront payment or an ongoing cash inflow (such as options and credit default swaps) that have the potential to be the instruments of a credit risk explosion: centuries of experience have shown that in every economy there are firms which are willing to risk bankruptcy for current gains.
The only true constraint on unregulated derivatives markets is the diligence of creditors – who, of course, do not want to be owed money by a bankrupt firm. This is where the existence of firms that are too big to fail becomes a very, very grave danger for the stability of a financial system with unregulated derivatives. When creditors believe that a firm is so big it will be bailed out by the government if it gets into trouble, they don’t bother to monitor how it is managing its derivatives exposure.
Thus credit risk can grow in an unconstrained manner when two criteria are met: (i) there is an un- (or under-) regulated market in derivatives with upfront or ongoing payments and (ii) creditors fail to monitor their counterparties.
In short, for those who do not understand what the problem is with credit default swaps (CDS), this is my answer: Because credit default swaps are unregulated derivatives, they make it possible for credit risk to grow without limit. Because they operate in an environment where many firms are too big to fail, the search for profits by firms with weak risk controls all but guarantees that credit risk in the credit default swap market will grow to dangerous levels – and that the government will be left with a big problem on its hands.
This explanation of the dangers of credit default swaps requires two ingredients: CDS sellers who are willing to risk bankruptcy for short-term gains and CDS buyers who fail to diligently monitor their counterparties.
In the current crisis bankruptcy remote special purpose vehicles (SPVs) have proven to be important sellers of credit default swaps. In the form of CDOs and other structures these SPVs treated the cash inflow from selling the swap as an investment and sought participants who were willing to prepay their share of the potential swap liability in exchange for a share of the income stream. In the event that no payment was due on the swap, the participant’s initial stake would be returned. These prepaid participants generally only put up 10 to 25% of the potential swap liability. A too big to fail bank or insurer was found to issue another credit default swap, called the super senior swap, that guaranteed to make any payments that were necessary after all the money put down by the prepaid participants was gone.
Observe that in these structured vehicles, the prepaid participant did not risk bankruptcy, but only the loss of his initial stake. On the other hand some of the banks and insurers which sold super senior swaps did take on liabilities that were capable of bankrupting them. The classic example is AIG, but there are others: UBS transferred CDS liabilities to the Swiss National Bank, and the financial security of the monoline insurers, Ambac and MBIA remains unclear. Furthermore, the CDO itself is at risk of liquidation (the equivalent of bankruptcy for a CDO) in the event that the super senior counterparty defaults on its obligation. Thus both the CDOs and certain too big to fail banks and insurers were willing to risk bankruptcy in return for a stream of swap payments.
Why did prepaid participants agree to absorb the first 10 to 25% of the swap liability? There are many reasons: They were told by rating agencies that losses were so unlikely that these “investments” were comparable to putting one’s money in investment-grade bonds – and the CDO paid a little more than the bonds. If the newspapers are to be believed some of these participants had no idea that the CDO’s income came from investors who expected to be paid with the very same stake that the participant was hoping to get back.
The other key to explosive growth of credit derivatives is the fact that swap buyers did not monitor their counterparties’ aggregate CDS exposure. When it came to CDOs they had no need to do so: a portion of the liability was prepaid and the rest was backed by a too big to fail bank or insurance company. Swap buyers had reason to believe that due diligence was unnecessary.
The problem of collateral
So far we have found two problems with credit default swaps: (i) they create credit risk, without contributing to the productive potential of the economy and (ii) they can be used to create unlimited quantities of credit risk. There are two more problems, both of which are specific to the current margining and settlement structure that has been put in place by the ISDA and US bankruptcy law.
Because credit default swaps offer to pay a lump sum when a corporate or financial entity defaults on its debt, we find that just as the economy is deteriorating – and more and more firms are declaring bankruptcy – the sellers of the swaps are obliged to make very large payments. To smooth this process the standard ISDA contract requires that the protection seller post collateral with the buyer as the market value of the CDS rises (and the expected default of the reference entity becomes more likely). We should note that there are exceptions to this rule: the broker-dealers usually do not need to post collateral and the CDS contracts of the monoline insurance companies explicitly protect these insurers from posting collateral.
In the case of AIG, collateral calls on credit default swaps exceeded the firm’s ability to pay. Whether this implies that over a longer horizon, AIG was insolvent is unclear: AIG takes the position that these CDS contracts have increased in value only because markets are illiquid and that over time the collateral posted would be returned to the firm. In any case in September 2008 AIG was illiquid – because of collateral calls on CDS, the firm could not honor its contractual commitments.
In short, AIG was bankrupted by collateral calls on CDS. Both Bear Stearns and Lehman Brothers were also bankrupted by problems related to collateral – though it is not clear that these cases were related to CDS.
What has become clear is that our current financial structure promotes a kind of financial cannibalism. As soon as a broker-dealer is at risk of failure, counterparties to derivative contracts step forward to demand collateral. In this way the other broker-dealers absorb the assets of the troubled firm, leaving little more than an empty shell for non-financial creditors.
The fact that financial firms, which appear to have enough tangible assets to cover their liabilities, can be stripped of many of those assets by secured lenders before bankruptcy has come as shock to bondholders. The reason this phenomenon has taken people by surprise is that very recent changes to bankruptcy law have both made it possible and dramatically changed the nature of bankruptcy itself.
Under common law, as it has developed over the centuries, the trustees for a bankrupt estate have the duty to reclaim any collateral that was transferred shortly before bankruptcy was declared. This policy protects creditors as a group by voiding any transactions that would have the effect of treating some creditors more favorably than others. In 2005, with the encouragement of the ISDA, Congress exempted derivatives from these constraints. Thus, the 2005 bankruptcy law encourages the derivative counterparties to a troubled firm to demand as much collateral as possible from the firm, before an event of bankruptcy actually takes place. To those watching the events, it is as if the penalty for an investment bank that shows signs of weakness is to have its fellows turn on it and consume as much flesh as possible before the scavengers show up.
While this problem is not fundamentally one of credit default swaps, CDS are one of the important classes of derivatives that enable counterparties to demand huge sums in collateral when a firm’s creditworthiness deteriorates.
The second problem with the collateral regime is that the CDS sellers are highly levered financial institutions. These financial institutions play an important role in funding both commercial and non-commercial economic activity. Thus when these banks are required to store liquid assets in an account at another bank, money that could fund economic activity is withdrawn from the financial system.
The alert reader will observe that collateral calls only result in the withdrawal of funds from the financial system, if the bank receiving the collateral does not have the right to lend it on. In the US, the Securities and Exchange Act of 1934 severely circumscribes the degree to which collateral can be rehypothecated by the bank that receives it. Thus in the US, collateral calls can have the effect of segregating liquid assets from the financial system – and reducing the quantity of funds that are circulating within the banking system.
If the size of the collateral calls were extremely small, their effect on the banking system would not matter. The problem with credit default swaps is that they can trigger very large collateral calls: AIG was required to post tens of billions of dollars of collateral. Not only does this represent billions of dollars on which AIG is not earning a return, but it represents billions of dollars which might as well be stored under someone’s mattress.
In short, collateral calls on CDS (i) are part of a dysfunctional system which encourages relatively healthy investment banks to cannibalize their weakest brethren and (ii) have the effect of withdrawing liquid funds from the financial system precisely when they are most needed.
Why we don’t need CDS to price risk
Lastly, I wish to preempt one particular argument that is often made in defense of credit default swaps. A common argument in support of CDS is that, because the market is more liquid than the bond market, it is more effective at pricing credit risk.
While it is true that the corporate bond market is notoriously illiquid and that CDS has made the trade of corporate credit risk easier and more frequent, swap markets are also very illiquid. On average there are about 2000 contracts outstanding on each corporate CDS name. Only about 4 corporations (and additional four indexes) have more than 10,000 contracts outstanding. While the public has no means of determining daily trading volume, most likely on average each corporate name has only a few dozen trades a week. Another sign of low trading volume is that the difference between the bid price and the ask price for CDS is typically more than 15% of the average of the bid and ask price. In other words, everyone who trades in this market, has to pay a large fee to dealers for providing liquidity to the market. In a market this thin, it is very easy for prices to be moved by technical factors. Thus, CDS prices sometimes accurately reflect the fundamental state of the underlying credit risk – and sometimes they do not.
The evidence that the CDS market is not always good at pricing risk lies in the data from 2006 through the first half of 2007, when many financial commentators were observing that the price of risk had fallen to absurdly low levels – and many referred specifically to the price of risk as measured by CDS. Of course, when the price of risk is too low, market participants buy too much of it.
That this is an accurate interpretation of what happened in the months preceding the crisis is best demonstrated by the crisis itself. Given the losses incurred by almost every major financial institution in the country, it is safe to conclude that they took on too much risk in the months and years preceding the crisis. To support this view consider the following data: At the end of 2005, the CDX-IG index, which is a measure of the CDS premium on 125 investment grade corporations, indicated a premium payment of about one-half of one percent. By February 2007 it had fallen to less than one-third of one percent.
In 2006 the following corporations were covered by the CDX-IG index: AIG, Fannie Mae, Freddie Mac, Countrywide and Washington Mutual. All of these companies have been bailed out by the government or taken over by another firm after disastrous financial performance. There are several other names that may not survive the current crisis: insurers, MBIA and XL Capital; homebuilders, Toll Brothers and Pulte Homes; and finally Wells Fargo, American Express and CIT Group, all firms which have obtained or are seeking financial support from the US government. In retrospect, it is very clear that credit risk was underpriced through 2006 and in the first half of 2007.
At the same time that CDS premia were at an all time low, the CDS market was growing fast. From December 31, 2005 to June 30, 2007 the CDS market grew from $17 trillion to $45 trillion. In 2008, it has become clear that vast quantities of underpriced credit default swap contracts were sold in 2006 and 2007. Given these facts, it’s almost hard to believe that there are people who are willing to argue that the CDS market is the best place to price default risk.
Conclusion
While there is no question that credit default swaps can be used to price risk, proponents of these derivatives have yet to demonstrate that they are a good way to price risk. In fact, there may be reason to believe that when CDS underprice risk, they can have a particularly noxious effect on credit markets by encouraging an explosion of credit risk, like the one that took place in 2006 and early 2007.
The real bills doctrine gives us a framework for understanding the dangers that derivatives like credit default swaps pose to the economy. When CDS are not used to transfer risk, they create credit risk, without adding to the productive capabilities of the economy. Furthermore, in the absence of regulation and in the presence of too big to fail firms, the attractiveness of the income flows from CDS contracts is all but guaranteed to generate ever increasing quantities of credit risk. Eventually some firm is bankrupted by its obligations and a chain of failures can follow from this initial collapse.
The particular structure that has been put into place over the last decade for margining and settling CDS contracts (and other derivatives) is another source of instability. Because recent laws have taken away a bankruptcy court’s right to demand the return of recently posted collateral from a derivative counterparty, whenever a firm’s financial stability is in question, margin calls accelerate – and can have the effect not only of stripping the firm’s assets, but even of forcing the firm to file bankruptcy by creating a cash flow crisis. Furthermore, in our highly leveraged financial system these collateral calls have the effect of draining funds from the banking system just when the funds are most needed.
Thus credit default swaps are a source of financial instability for two reasons: First of all, it is in the nature of derivatives such as these that when they are unregulated they can generate excessive growth of credit risk. Secondly, the particular margining and settlement structure that has been developed for derivatives in the US, encourages a form of financial cannibalism that is very destabilizing, while at the same time draining liquid funds from the financial system.
Update 2-26-09: Comments are not functioning for me, so I will tag my responses to EoCs comment below onto the post. I'll try to get around to figuring out why the comment screen doesn't work soon.
(1) "[T]he broker-dealers usually do not need to post collateral" 100% false. The broker-dealers post collateral every day, and have done so for several years.
Here is a source that indicates that you are incorrect: http://www.aei.org/docLib/20090224_WhalenRemarks.pdf
Please give me a link to a source supporting your point of view.
(2) Rehypothecation of collateral is the weakest part of my argument and you may be right. I'd like to see what happens when the next big Lehman/Kaupthing-like event takes place, before I'm convinced by your view however.
(3) "The dealers, by the way, run matched CDS books, and are actually slight net buyers of protection."
This is contrary to widely published reports that Goldman Sachs and Deutsche Bank flourished during the subprime crisis of 2007, because they bought "protection" on CDS markets. See for example here and here. If DB and GS didn't have matched books in ABX, why should I believe they have matched books in other CDS contracts.
Furthermore your claim of "matched books" doesn't take off balance sheet exposure into account (e.g. Citi's leveraged super senior exposure) or poorly hedged exposure (e.g. UBS super seniors). The simple fact is that you are making a claim that dealers carry matched books and this claim cannot be substantiated with available public data.
(Update 3-31-09: Proof that my theory was correct. Goldman Sachs now has to report its CDS exposure to the OCC, and look at what we find here in the last chart: Total Credit Derivatives Bought $747 billion Total Credit Derivatives Sold $650.
That's a pretty odd looking "matched book"!)
(4) I think you failed to understand my model. Perhaps if you read the sentences preceding and following your quote, it will be clearer.
(5) "the net notional amount of CDS outstanding" became irrelevant the day that AIG had to be bailed out by the government. Assume AIG went bankrupt and defaulted on all of its CDS obligations (with collateral postings frozen as of Sept 16), and then explain to me how the relevant number is net notional exposure.
The dangers of credit have been recognized since the start of written history: early records of interest rate regulation are not at all uncommon. Only since the dawn of paper money, however, have societies felt the need to regulate the quantity of financial assets – and thus the quantity of credit risk in an economy.
In the early modern era, the regulation of credit risk was decentralized. It took the form of a general principle that all bankers, merchants and other tradesmen were expected to follow: this principle is now known as the real bills doctrine. This doctrine drew a clear distinction between real and fictitious bills. Real bills were IOUs that were created when the creditor delivered some product or real asset to the debtor, and thus real bills were tied to economic production and trade. When there was no transaction in the real economy connected with the bill, it was considered fictitious. Well known bankers (who were often traders, too) could issue large quantities of bills and it was very difficult for those who circulated these bills to know whether a bill was real or fictitious. For this reason, proof of issuing a fictitious bill was enough to end a banker’s – or indeed a trader’s – career. In 1772 when the Bank of England refused to discount the bills of the Alexander Fordyce, owner of the Ayr Bank, the directors explained that they did so because the quantity of bills already discounted at the Bank was so great that some of them had to be fictitious. The Ayr Bank closed in a matter of days.
The basic idea behind the real bills doctrine is simple: in order for the credit risk inherent in a financial asset to be socially valuable, it must be the case that the risk is tied to economic activity. When financial assets grow in number without reference to economic fundamentals, credit risk grows too. Centuries of experience with euphoria and collapse have taught us that the unrestrained growth of credit risk has one certain consequence: bankruptcy that spreads like a virus from one borrower to the next.
It is fairly easy to relate some modern financial assets to the real bills doctrine. Stocks and bonds exist to finance productive enterprises, thus they are real bills. Some assets are harder to fit in the framework: Insurance transfers risk from a party who is ill-equipped to bear it, to one with greater resources. For the purposes of this discussion, let’s call forms of risk that arise from the process of producing goods and services for sale, economic risk. Thus, the airline that is concerned that the price of fuel will rise, and the employee who worries that an accident at work could cause health problems that bankrupt him both face economic risk. We will treat the transfer of economic risk as a real activity, and insurance contracts that protect firms or individuals from the costs of economic risk as real bills.
Placing derivatives in the context of the real bills doctrine is a more complex problem. Some derivatives are like insurance, because they allow investors to transfer risk. The stock investor who wants to protect his investment from losses can buy put options. The bond investor can buy a credit default swap.
In rare cases a derivative contract can actually be used to eliminate certain forms of economic risk – the classic example is when a farmer and a baker agree to trade wheat at a fixed price six months in the future. This forward contract allows the farmer to be confident that he can borrow money today to plant his fields and the baker confident that an investment in kitchen equipment will be profitable. This, then, may be an example of financial asset that does not increase the aggregate amount of risk in the economy – instead it serves to convert economic risk into credit risk.
When derivatives are a zero-sum game
The same derivatives that can serve to transfer risk can also be used by individuals without any initial risk exposure to profit off of expected movements in the market. When a derivative is not used to transfer existing risk, it constitutes a zero sum transaction: whatever one party gains the other loses. This means that the transaction requires different beliefs about the value of the derivative on the part of the buyer and of the seller – since presumably the trade is taking place, because both expect to gain from it. In such a situation, the effect of asymmetric information on the price at which the transaction takes place has the potential to result in an inefficient exchange.
Some claim that the zero sum nature of these derivative transactions should not raise concern, since secondary markets for financial instruments such as stocks and bonds are also zero sum. This is true. But we should all recognize that the problem of asymmetric information in these markets led to the creation of organizations such as the NYSE and the SEC. These institutions exist to even the playing field. The NYSE regulates the prices at which stocks trade to ensure equal access (even if only through a broker) to the prices available on the stock exchange. The SEC requires firms to report accurate financial information so that purchasers of stocks and bonds are well-informed about the prospects of the firm.
Not only are many secondary markets highly regulated, but there is another crucial difference between them and markets for derivatives. In the case of secondary markets no financial asset is created, it is only traded. Thus on secondary markets the zero sum transaction does not involve the creation of credit risk.
By contrast, when a new derivative contract is traded in a zero sum transaction, credit risk is created. Thus, there is no escaping this conclusion: when a new zero sum derivative contract is traded, the aggregate amount of credit risk in the economy increases incrementally, while the productive possibilities of the economy do not increase at all.
Now, for some zero-sum derivative transactions it is certainly possible that there is a gain that offsets the incremental loss to society due to the creation of credit risk. I would propose, however, that the burden of proof for such a gain lies with the financial institutions who wish to engage in zero-sum derivative transactions.
The problem with fictitious bills
When a derivative is not used to transfer risk, but instead used by an individual without any initial risk exposure to profit off of expected movements in the market, we have a financial transaction without a corresponding economic transaction. Because such derivatives are not connected to economic activity, the only constraint on the quantity of their issue is the common-sense of the two parties to the contract. If these derivatives are being traded by fools, nothing prevents one or both of them from taking on obligations that exceed their ability to pay by multiples of one hundred, one thousand, or more.
Thus, derivatives are distinguished from stocks, bonds and insurance contracts by their ability to grow in an unconstrained manner, disconnected from the realities of economic activity. To monitor and restrict the growth of credit risk (and to reduce problems of asymmetric information), many derivatives markets are highly regulated.
For example, options and futures contracts are exchange-traded. Because every participant on the exchange trades with a single, central counterparty, the credit risk these traders are exposed to is limited to the risk that the central counterparty itself goes bankrupt. The counterparty in turn monitors the dealers in the market to make sure that they are not taking inappropriate risks. In fact, if the dealers have any large obligations, they are sure to hedge them in the underlying market. For example, if the dealers have sold a large quantity of call options on a certain stock, they may go ahead and buy the stock at current market prices to ensure that they are not exposed to large losses. If the quantities are large enough, this transaction can drive up the price of the stock – allowing demand on option markets to affect prices on the underlying market. The close connection between the two markets means that price signals will act to limit the growth of credit risk.
When derivatives markets are unregulated there is a genuine danger that credit risk will grow without bound. In order for this to happen, it must be the case that some firms are issuing liabilities without sufficient concern for the possibility that these liabilities may bankrupt the firm. For this reason, it is derivatives that involve either an upfront payment or an ongoing cash inflow (such as options and credit default swaps) that have the potential to be the instruments of a credit risk explosion: centuries of experience have shown that in every economy there are firms which are willing to risk bankruptcy for current gains.
The only true constraint on unregulated derivatives markets is the diligence of creditors – who, of course, do not want to be owed money by a bankrupt firm. This is where the existence of firms that are too big to fail becomes a very, very grave danger for the stability of a financial system with unregulated derivatives. When creditors believe that a firm is so big it will be bailed out by the government if it gets into trouble, they don’t bother to monitor how it is managing its derivatives exposure.
Thus credit risk can grow in an unconstrained manner when two criteria are met: (i) there is an un- (or under-) regulated market in derivatives with upfront or ongoing payments and (ii) creditors fail to monitor their counterparties.
In short, for those who do not understand what the problem is with credit default swaps (CDS), this is my answer: Because credit default swaps are unregulated derivatives, they make it possible for credit risk to grow without limit. Because they operate in an environment where many firms are too big to fail, the search for profits by firms with weak risk controls all but guarantees that credit risk in the credit default swap market will grow to dangerous levels – and that the government will be left with a big problem on its hands.
This explanation of the dangers of credit default swaps requires two ingredients: CDS sellers who are willing to risk bankruptcy for short-term gains and CDS buyers who fail to diligently monitor their counterparties.
In the current crisis bankruptcy remote special purpose vehicles (SPVs) have proven to be important sellers of credit default swaps. In the form of CDOs and other structures these SPVs treated the cash inflow from selling the swap as an investment and sought participants who were willing to prepay their share of the potential swap liability in exchange for a share of the income stream. In the event that no payment was due on the swap, the participant’s initial stake would be returned. These prepaid participants generally only put up 10 to 25% of the potential swap liability. A too big to fail bank or insurer was found to issue another credit default swap, called the super senior swap, that guaranteed to make any payments that were necessary after all the money put down by the prepaid participants was gone.
Observe that in these structured vehicles, the prepaid participant did not risk bankruptcy, but only the loss of his initial stake. On the other hand some of the banks and insurers which sold super senior swaps did take on liabilities that were capable of bankrupting them. The classic example is AIG, but there are others: UBS transferred CDS liabilities to the Swiss National Bank, and the financial security of the monoline insurers, Ambac and MBIA remains unclear. Furthermore, the CDO itself is at risk of liquidation (the equivalent of bankruptcy for a CDO) in the event that the super senior counterparty defaults on its obligation. Thus both the CDOs and certain too big to fail banks and insurers were willing to risk bankruptcy in return for a stream of swap payments.
Why did prepaid participants agree to absorb the first 10 to 25% of the swap liability? There are many reasons: They were told by rating agencies that losses were so unlikely that these “investments” were comparable to putting one’s money in investment-grade bonds – and the CDO paid a little more than the bonds. If the newspapers are to be believed some of these participants had no idea that the CDO’s income came from investors who expected to be paid with the very same stake that the participant was hoping to get back.
The other key to explosive growth of credit derivatives is the fact that swap buyers did not monitor their counterparties’ aggregate CDS exposure. When it came to CDOs they had no need to do so: a portion of the liability was prepaid and the rest was backed by a too big to fail bank or insurance company. Swap buyers had reason to believe that due diligence was unnecessary.
The problem of collateral
So far we have found two problems with credit default swaps: (i) they create credit risk, without contributing to the productive potential of the economy and (ii) they can be used to create unlimited quantities of credit risk. There are two more problems, both of which are specific to the current margining and settlement structure that has been put in place by the ISDA and US bankruptcy law.
Because credit default swaps offer to pay a lump sum when a corporate or financial entity defaults on its debt, we find that just as the economy is deteriorating – and more and more firms are declaring bankruptcy – the sellers of the swaps are obliged to make very large payments. To smooth this process the standard ISDA contract requires that the protection seller post collateral with the buyer as the market value of the CDS rises (and the expected default of the reference entity becomes more likely). We should note that there are exceptions to this rule: the broker-dealers usually do not need to post collateral and the CDS contracts of the monoline insurance companies explicitly protect these insurers from posting collateral.
In the case of AIG, collateral calls on credit default swaps exceeded the firm’s ability to pay. Whether this implies that over a longer horizon, AIG was insolvent is unclear: AIG takes the position that these CDS contracts have increased in value only because markets are illiquid and that over time the collateral posted would be returned to the firm. In any case in September 2008 AIG was illiquid – because of collateral calls on CDS, the firm could not honor its contractual commitments.
In short, AIG was bankrupted by collateral calls on CDS. Both Bear Stearns and Lehman Brothers were also bankrupted by problems related to collateral – though it is not clear that these cases were related to CDS.
What has become clear is that our current financial structure promotes a kind of financial cannibalism. As soon as a broker-dealer is at risk of failure, counterparties to derivative contracts step forward to demand collateral. In this way the other broker-dealers absorb the assets of the troubled firm, leaving little more than an empty shell for non-financial creditors.
The fact that financial firms, which appear to have enough tangible assets to cover their liabilities, can be stripped of many of those assets by secured lenders before bankruptcy has come as shock to bondholders. The reason this phenomenon has taken people by surprise is that very recent changes to bankruptcy law have both made it possible and dramatically changed the nature of bankruptcy itself.
Under common law, as it has developed over the centuries, the trustees for a bankrupt estate have the duty to reclaim any collateral that was transferred shortly before bankruptcy was declared. This policy protects creditors as a group by voiding any transactions that would have the effect of treating some creditors more favorably than others. In 2005, with the encouragement of the ISDA, Congress exempted derivatives from these constraints. Thus, the 2005 bankruptcy law encourages the derivative counterparties to a troubled firm to demand as much collateral as possible from the firm, before an event of bankruptcy actually takes place. To those watching the events, it is as if the penalty for an investment bank that shows signs of weakness is to have its fellows turn on it and consume as much flesh as possible before the scavengers show up.
While this problem is not fundamentally one of credit default swaps, CDS are one of the important classes of derivatives that enable counterparties to demand huge sums in collateral when a firm’s creditworthiness deteriorates.
The second problem with the collateral regime is that the CDS sellers are highly levered financial institutions. These financial institutions play an important role in funding both commercial and non-commercial economic activity. Thus when these banks are required to store liquid assets in an account at another bank, money that could fund economic activity is withdrawn from the financial system.
The alert reader will observe that collateral calls only result in the withdrawal of funds from the financial system, if the bank receiving the collateral does not have the right to lend it on. In the US, the Securities and Exchange Act of 1934 severely circumscribes the degree to which collateral can be rehypothecated by the bank that receives it. Thus in the US, collateral calls can have the effect of segregating liquid assets from the financial system – and reducing the quantity of funds that are circulating within the banking system.
If the size of the collateral calls were extremely small, their effect on the banking system would not matter. The problem with credit default swaps is that they can trigger very large collateral calls: AIG was required to post tens of billions of dollars of collateral. Not only does this represent billions of dollars on which AIG is not earning a return, but it represents billions of dollars which might as well be stored under someone’s mattress.
In short, collateral calls on CDS (i) are part of a dysfunctional system which encourages relatively healthy investment banks to cannibalize their weakest brethren and (ii) have the effect of withdrawing liquid funds from the financial system precisely when they are most needed.
Why we don’t need CDS to price risk
Lastly, I wish to preempt one particular argument that is often made in defense of credit default swaps. A common argument in support of CDS is that, because the market is more liquid than the bond market, it is more effective at pricing credit risk.
While it is true that the corporate bond market is notoriously illiquid and that CDS has made the trade of corporate credit risk easier and more frequent, swap markets are also very illiquid. On average there are about 2000 contracts outstanding on each corporate CDS name. Only about 4 corporations (and additional four indexes) have more than 10,000 contracts outstanding. While the public has no means of determining daily trading volume, most likely on average each corporate name has only a few dozen trades a week. Another sign of low trading volume is that the difference between the bid price and the ask price for CDS is typically more than 15% of the average of the bid and ask price. In other words, everyone who trades in this market, has to pay a large fee to dealers for providing liquidity to the market. In a market this thin, it is very easy for prices to be moved by technical factors. Thus, CDS prices sometimes accurately reflect the fundamental state of the underlying credit risk – and sometimes they do not.
The evidence that the CDS market is not always good at pricing risk lies in the data from 2006 through the first half of 2007, when many financial commentators were observing that the price of risk had fallen to absurdly low levels – and many referred specifically to the price of risk as measured by CDS. Of course, when the price of risk is too low, market participants buy too much of it.
That this is an accurate interpretation of what happened in the months preceding the crisis is best demonstrated by the crisis itself. Given the losses incurred by almost every major financial institution in the country, it is safe to conclude that they took on too much risk in the months and years preceding the crisis. To support this view consider the following data: At the end of 2005, the CDX-IG index, which is a measure of the CDS premium on 125 investment grade corporations, indicated a premium payment of about one-half of one percent. By February 2007 it had fallen to less than one-third of one percent.
In 2006 the following corporations were covered by the CDX-IG index: AIG, Fannie Mae, Freddie Mac, Countrywide and Washington Mutual. All of these companies have been bailed out by the government or taken over by another firm after disastrous financial performance. There are several other names that may not survive the current crisis: insurers, MBIA and XL Capital; homebuilders, Toll Brothers and Pulte Homes; and finally Wells Fargo, American Express and CIT Group, all firms which have obtained or are seeking financial support from the US government. In retrospect, it is very clear that credit risk was underpriced through 2006 and in the first half of 2007.
At the same time that CDS premia were at an all time low, the CDS market was growing fast. From December 31, 2005 to June 30, 2007 the CDS market grew from $17 trillion to $45 trillion. In 2008, it has become clear that vast quantities of underpriced credit default swap contracts were sold in 2006 and 2007. Given these facts, it’s almost hard to believe that there are people who are willing to argue that the CDS market is the best place to price default risk.
Conclusion
While there is no question that credit default swaps can be used to price risk, proponents of these derivatives have yet to demonstrate that they are a good way to price risk. In fact, there may be reason to believe that when CDS underprice risk, they can have a particularly noxious effect on credit markets by encouraging an explosion of credit risk, like the one that took place in 2006 and early 2007.
The real bills doctrine gives us a framework for understanding the dangers that derivatives like credit default swaps pose to the economy. When CDS are not used to transfer risk, they create credit risk, without adding to the productive capabilities of the economy. Furthermore, in the absence of regulation and in the presence of too big to fail firms, the attractiveness of the income flows from CDS contracts is all but guaranteed to generate ever increasing quantities of credit risk. Eventually some firm is bankrupted by its obligations and a chain of failures can follow from this initial collapse.
The particular structure that has been put into place over the last decade for margining and settling CDS contracts (and other derivatives) is another source of instability. Because recent laws have taken away a bankruptcy court’s right to demand the return of recently posted collateral from a derivative counterparty, whenever a firm’s financial stability is in question, margin calls accelerate – and can have the effect not only of stripping the firm’s assets, but even of forcing the firm to file bankruptcy by creating a cash flow crisis. Furthermore, in our highly leveraged financial system these collateral calls have the effect of draining funds from the banking system just when the funds are most needed.
Thus credit default swaps are a source of financial instability for two reasons: First of all, it is in the nature of derivatives such as these that when they are unregulated they can generate excessive growth of credit risk. Secondly, the particular margining and settlement structure that has been developed for derivatives in the US, encourages a form of financial cannibalism that is very destabilizing, while at the same time draining liquid funds from the financial system.
Update 2-26-09: Comments are not functioning for me, so I will tag my responses to EoCs comment below onto the post. I'll try to get around to figuring out why the comment screen doesn't work soon.
(1) "[T]he broker-dealers usually do not need to post collateral" 100% false. The broker-dealers post collateral every day, and have done so for several years.
Here is a source that indicates that you are incorrect: http://www.aei.org/docLib/20090224_WhalenRemarks.pdf
Please give me a link to a source supporting your point of view.
(2) Rehypothecation of collateral is the weakest part of my argument and you may be right. I'd like to see what happens when the next big Lehman/Kaupthing-like event takes place, before I'm convinced by your view however.
(3) "The dealers, by the way, run matched CDS books, and are actually slight net buyers of protection."
This is contrary to widely published reports that Goldman Sachs and Deutsche Bank flourished during the subprime crisis of 2007, because they bought "protection" on CDS markets. See for example here and here. If DB and GS didn't have matched books in ABX, why should I believe they have matched books in other CDS contracts.
Furthermore your claim of "matched books" doesn't take off balance sheet exposure into account (e.g. Citi's leveraged super senior exposure) or poorly hedged exposure (e.g. UBS super seniors). The simple fact is that you are making a claim that dealers carry matched books and this claim cannot be substantiated with available public data.
(Update 3-31-09: Proof that my theory was correct. Goldman Sachs now has to report its CDS exposure to the OCC, and look at what we find here in the last chart: Total Credit Derivatives Bought $747 billion Total Credit Derivatives Sold $650.
That's a pretty odd looking "matched book"!)
(4) I think you failed to understand my model. Perhaps if you read the sentences preceding and following your quote, it will be clearer.
(5) "the net notional amount of CDS outstanding" became irrelevant the day that AIG had to be bailed out by the government. Assume AIG went bankrupt and defaulted on all of its CDS obligations (with collateral postings frozen as of Sept 16), and then explain to me how the relevant number is net notional exposure.
Subscribe to:
Comments (Atom)
Apparently the Atlantic Business Channel has a long delay in posting comments, since Derivative Dribble submitted his reply shortly after my post went up, but it didn't appear on the blog for at least 6 hours after that (I checked). Anyhow, since comments are so delayed at the The Atlantic, I thought I would go ahead and post my response to DD here. (These are approximate. I didn't actually copy the posts before hitting submit -- someday I'll learn!)
I'll keep updating this post if the dialog continues.
Derivative Dribble (DD) responded to my post with the following.
ACC,
As I've said a million times, CDSs, like all contracts, create counterparty risks and therefore can create substantial reliance costs. That is not at issue in this article.
Secondly, unfunded derivatives cannot be considered "allocations of capital" since they are, by definition unfunded. That cash is free to circulate through the capital markets.