FRM 10: Anatomy of the Great Financial Crisis of 2007-2009
The episode now called the Great Financial Crisis, or GFC, opened with a slump in the United States subprime mortgage market during the summer of 2007. Earlier United States credit crises stayed domestic. This one caught investors on every continent, and losses ran outward from subprime home loans into every other corner of the credit market.
Losses arrived at banks just as nobody could agree what credit assets were worth, so banks stopped lending to one another. Governments opened liquidity support facilities and recapitalized insolvent banks to restart lending. Many institutions failed or were taken over, and credit losses worldwide eventually passed USD 1 trillion.
The early casualties
February 2008 brought the nationalization of Northern Rock, a troubled United Kingdom mortgage lender and victim of the first bank run that country had seen in 140 years. A month later J.P. Morgan Chase took over Bear Stearns, the United States investment bank, in a rescue brokered by the U.S. Treasury Department together with the Federal Reserve. The asset-backed commercial paper market and the repo market halted, hedge funds froze redemptions or failed, and many special investment vehicles and conduits were wound down.
September 2008, the peak
Lehman Brothers declared bankruptcy, and interbank borrowing contracted at once, because banks holding spare cash would not lend it overnight in the repo markets. Morgan Stanley and Goldman Sachs, the two remaining major investment banks in the country, became bank holding companies regulated by the Federal Reserve, which opened the liquidity facilities of the Fed to them. Fannie Mae and Freddie Mac were nationalized, and AIG was pulled back from collapse by a USD 150 billion capital infusion from the U.S. Treasury and the Federal Reserve.
Europe moved in parallel. The Dutch financial conglomerate Fortis was broken up and sold, and the largest commercial bank in Iceland collapsed, followed by the entire Icelandic banking system. Those rescues stretched several European budgets thin and fed the European sovereign debt crisis of 2010. The crisis then spilled into the wider global economy, destroying enormous wealth and pushing unemployment high around the world.
Nothing about 2007 came out of nowhere. The preceding years delivered an exceptional boom in United States credit growth, a massive housing price bubble, and leverage in the financial system that had been building since the credit crisis of 2001-2002. A wave of financial innovation in securitization came with it, expanding the capacity of the system to manufacture credit assets far faster than its capacity to manage the risks attached.
The numbers behind the boom
Across 2002 to 2007, debt measured as a percent of national income climbed from 375% to 475%, while average housing prices rose at 11% per year, a record rate. The Fed funds rate stood at 1% in June 2003, began a slow climb in June 2004, reached 5.25% by June 2006, and then a cut on September 18, 2007 took it to 4.75%. Mortgage debt was widely held: in 2007 some 50 million homeowners in the United States, two-thirds of the total, carried a mortgage, 75.2% of them fixed rate loans and the remaining 24.8% adjustable-rate mortgages, or ARMs.
Two appetites meeting
Low policy rates pushed housing demand up, and rising demand pushed house prices up with it. The same low rates left investors, institutional investors included, hunting for yield enhancement, and they found it in subprime mortgages, where the premium over the rate paid by a prime borrower typically runs as high as 300 basis points.
Securitization bridged the two appetites. Securitizers pooled below investment-grade assets, sorted the cash flows by the certainty a model assigned to each, and wrapped the safest of them into investment-grade securities. Banks responded by building an originate-to-distribute business model, usually written OTD, in which a loan is made in order to be sold rather than held. Much of the machinery sat outside the regulated sector: structured investment vehicles and conduits belonged to the shadow banking system, a network of financial systems made up of non-depository banks.
A subprime mortgage is a residential home loan made to a borrower whose credit is poor. Consumer credit quality in the United States is measured by a FICO score, named after the Fair Isaac Corporation, which developed the methodology. A limited credit history drags that score down, as do a large amount of outstanding debt and a history of delinquent payments.
The boundary is not fixed. Definitions vary between lenders, and some treat a borrower with a relatively high credit score as subprime when the mortgage comes with a low down payment. The economics hold in every case: subprime mortgages carry more default risk than prime mortgages, so they pay higher interest rates. Between the two categories sit Alt-A borrowers, whose credit ratings are reasonably strong but who lack the documentation needed to verify their assets and income.
How large subprime lending became
Former Fed chairman Ben Bernanke, drawing on data from Inside Mortgage Finance, put subprime lending at an estimated USD 35 billion in 1994, some 4.5 percent of every one-to-four family mortgage origination, and at USD 600 billion by 2006, or 20 percent of originations. Total subprime mortgage debt outstanding was estimated at USD 1.3 trillion by early 2007.
| Year of origination | Share of total mortgage loans that is subprime |
|---|---|
| 2001 | 7% |
| 2002 | 8% |
| 2003 | 9% |
| 2004 | 11% |
| 2005 | 14% |
| 2006 | 20% |
Source: B&C Lending, Federal Reserve Bank of St. Louis; EIR.
The subprime share was still single digit in 2003, yet by 2006 one mortgage in five was subprime, written at the moment house prices stopped rising. A pool assembled that year therefore held the weakest borrowers at the top of the market.
Delinquencies on adjustable-rate subprime mortgages rose markedly in 2007. Serious delinquency rates were approaching 16% by August 2007, roughly triple where they had stood in mid-2005, and by May 2008 the figure had reached 25%. Ratings downgrades on subprime mortgage securitized products followed in enormous numbers.
The borrower was weak and the collateral was thin
Credit quality in a subprime transaction is weak by construction, and the loan is frequently under-collateralized. Spotty income and payment histories are typical, as are high debt-to-income ratios. The traditional first-time home mortgage required a 20% down payment, yet in 2005 fully 43% of first-time home buyers put nothing down, leaving almost no collateral cushion if house prices fell.
Standards fell as demand rose
Heavy demand for subprime mortgage product encouraged questionable practices among some lenders. Certain borrowers were steered into subprime loans although they qualified for better terms, while others ended up holding mortgages they could not afford. Products grew riskier in step with the demand. NINJA loans, meaning no income, no job and no assets, entered the market, as did liar loans, which asked for so little documentation that an applicant could misstate almost anything without being caught.
Some borrowers and mortgage brokers went further and filed false documentation, which let borrowers obtain funding on fraudulent terms. Compensation made this worse. Most mortgage brokers were paid for the volume of loans they originated rather than for how those loans performed, and a broker faced few consequences if a loan later defaulted, so originating brokers had very little reason to conduct proper due diligence.
Under the originate-to-distribute business model a bank extends loans, securitizes them, and sells the securities to investors. Ownership of the credit risk travels with the securities, so subprime mortgage losses landed on the investors who eventually owned the paper rather than on the banks that wrote the loans.
That handover changed what the originator cared about. A lender who expects to sell within weeks has little reason to spend money on proper credit assessment of the borrower or rigorous valuation of the home, and that due diligence is exactly what got thinned out.
Banks parked assets awaiting securitization in structured investment vehicles, also called conduits. A SIV is a limited-purpose, bankruptcy remote company used by a bank to purchase assets, funded mainly with short-term commercial paper plus some medium-term notes and capital. Because the vehicle is legally separate, the assets it held sat off the sponsoring bank balance sheet.
What the model promised and what it delivered
In theory, originate-to-distribute combined with heavy use of securitization spreads risk widely, so banks become less sensitive to credit crises, systemic risk falls, and lending gains an extra funding source. The crisis exposed the flaw. In the years 2003 to 2007, the evidence suggests banks kept credit exposures to AAA rated tranches through securitization, generating extra yield without lifting their regulatory capital minimums under Basel II.
Capital rules made that attractive. A residential mortgage carries a risk-weighted asset, or RWA, of 50%, while a AAA rated tranche of a securitization is subject to an RWA of only 20%, on the presumption that an asset rated so highly is at low risk of default. The rating also removed the incentive for investors to investigate the pool: they concluded they could lift returns without taking more risk by buying CDOs instead of lower yielding corporate bonds. They were wrong.
Securitization takes a portfolio of existing assets and repackages the cash flows into claims called tranches. Bonds are issued against the tranches, and the proceeds pay for the collateral assets. Subprime mortgages were securitized in volume into collateralized debt obligations, or CDOs, and those credit risk transfer instruments played a major part in the subprime meltdown.
The waterfall
Tranches are shaped to suit investor demand, which means shaping them to reach a desired credit rating, with most of the structure rated investment grade. A waterfall structure sets out which claim is paid first and which absorbs losses first, and that ordering lets one pool of subprime loans support securities of very different credit quality. The tranches sit in order of safety: Senior AAA debt at the top, often called super senior, then Junior AAA, AA, A, BBB, BB and downward. To secure the AAA rating on the super senior tranche, a surety wrap was sometimes used, supplied by a monoline insurer that must meet interest and principal payments should the structure default.
Monoline insurers had been important in municipal finance since the 1970s, and much of their growth before the crisis came from structured credit, asset-backed bonds and CDOs above all. They held capital enough at first to earn a AAA rating of their own, which removed any need to post collateral. The logic of tranching is simple enough: a pool of bonds scheduled to deliver USD 100 in one year, with a worst-case loss of USD 35, supports a very reliable minimum cash flow of USD 65 sold as a high-grade asset, while the claim on the other USD 35 is sold as high-yielding paper.
A CDO holds a pool of USD 500 million of subprime mortgage bonds against a super senior AAA tranche of USD 350 million, a junior AAA tranche of USD 50 million, a AA tranche of USD 30 million, an A tranche of USD 20 million, a BBB tranche of USD 25 million and an equity tranche of USD 25 million. Losses run up from the equity tranche. The amounts are illustrative.
Ratings sat at the centre of the machinery, so it matters who paid for them. During a CDO structuring, the equity holders, the CDO trust partners, pay a rating agency, sometimes more than one, to rate the various liabilities of the CDO. Agencies once charged investors for access to ratings, but the model changed, and securities issuers now pay to have their securities rated, which is the issuer-pay model.
Structuring to the model
CDO trusts knew which requirements and assumptions the agencies applied, so they could design the payment waterfalls and associated liabilities to make a high percentage of the bonds come out rated AAA. The rating stopped being an independent opinion on a fixed structure and became a target the structure was built to hit.
The assumptions rested on historical data that did not reflect the changes then taking place in the assets, including the growing number of NINJA loans, liar loans, and subprime mortgages written at 100% loan-to-value ratios. Agencies also worked from data supplied by the issuers and arrangers, the same parties bundling the mortgages and performing the due diligence, and although declining lending standards and rising fraud were widely known, it is alleged that they carried out no additional due diligence or monitoring of that data. Underneath all of it sits a simpler problem: subprime mortgage loans were too new to offer long-term data capable of informing a risk analysis, so many initial ratings, particularly the AAA ratings on senior tranches, were likely faulty from the outset.
The incentive to keep saying yes
Analytical weakness alone would not have been enough without a reason to overlook it. Agencies are paid to monitor a CDO over its life, a profitable and continual cash stream, and if too few bonds were rated AAA the trust would not be formed at all. Competition sharpened the effect, since an agency could win business by requiring lower credit enhancement than its rivals, which meant cheaper funding for the issuer and a larger AAA tranche. The market leaned heavily on the agencies for explicit risk analysis of these securities, and that analysis fed straight into market valuations, so a rating that was wrong at issuance became the price.
Two instruments make up the short-term wholesale debt market: repurchase agreements and commercial paper, or CP. Both shut down early in the crisis once participants doubted the quality of the collateral behind them.
Repurchase agreements
Repurchase agreements, known as repos, are used by banks, brokerage firms, money market funds and many other financial institutions. A standard repo has two legs: the sale of an asset, and an agreement to buy that asset back at a slightly higher price on a specified future date. The seller receives cash at the outset and is therefore a borrower in a collateralized loan, with the security as collateral, while the buyer gives cash at the outset and receives a larger sum at the end, which makes that party a lender.
Collateral runs from government bonds, through corporate bonds of high quality, to tranches of securitizations. Quality drives the haircut, the percent reduction from initial market value that the lender is prepared to advance: higher quality collateral attracts a smaller haircut and lower quality collateral a larger one. A haircut of 10% lets a borrower raise USD 90 against each USD 100 of pledged collateral. Its purpose is to protect the lender against recovering less than the full loan amount when the collateral has to be sold after a default.
Repos are also excluded from the bankruptcy process, so if one counterparty fails the survivor can terminate unilaterally, keeping the cash or selling the collateral.
Commercial paper and asset-backed commercial paper
Unsecured CP financing issues short-term debt with no specific assets behind it. Since a lender has nothing to seize in the event of default, unsecured CP issuers are generally of very high credit quality, and when an issuer weakens through a rating downgrade, an orderly exit normally follows through margin calls. Asset-backed commercial paper, or ABCP, is a special case in which the issuer finances the purchase of assets by issuing the paper, with those assets as collateral.
Demand for collateral grew strongly before the crisis, driven by growth in the OTC derivatives markets and increasing reliance on short-term collateralization, and newly issued AAA rated securitization tranches met part of it. Federal Reserve Bank of New York statistics put the total primary dealers inventory of repos at USD 1.6 trillion in 2000 and over USD 4.5 trillion in 2008.
Structured investment vehicles were funded short-term while holding longer dated assets, so survival depended on being able to roll over that debt again and again. As mortgage-backed securities lost value, the credit quality of many SIVs declined, ratings on the ABCP they had issued were downgraded rapidly, and a growing number of SIVs could no longer roll their ABCP. Liquidity in the subprime-related asset markets vanished at the same moment.
Counterparty risk arrives in the price
Until the middle of 2007 the market did not price counterparty credit risk. The gap between the unsecured overnight index swap rate, or OIS rate, and swap rates at every reset period, three months, six months or one year, ran to only 2 to 5 basis points. From June 2007 participants worried about the value of asset-backed securities and about bank exposure to subprime, and the OIS-swap spread exploded. It stayed elevated, spiked again on the failure of Lehman Brothers, and never returned to where it had been before the crisis.
Haircuts move from zero to punitive
Credit spreads widened substantially across all credit assets, lowering their market price, and haircuts rose systematically alongside, from zero pre-crisis to more than 45% by the time Lehman failed in September 2008. The short-term wholesale funding markets had begun to freeze by the summer of 2007, ABCP and repo alike. Investors would not roll maturing ABCP, so banks had to repatriate SIV assets onto their own balance sheets, and institutions financing themselves through repo could not roll their funding.
Only three outcomes remained: bailout, merger, or bankruptcy. That is the sequence that produced the failure of Bear Stearns, of Northern Rock, the United Kingdom mortgage bank, of IndyMac in California, and of Lehman Brothers. All of them satisfied the Basel minimum regulatory capital requirements before failing, which shows how little a capital ratio says about whether funding will be there next week.
A bank holds USD 100 in assets. Behind them sit USD 40 of long-term debt, USD 50 of repo financing and USD 10 of equity. Repo haircuts then increase from zero to 20%.
Uncertainty over the valuation of asset-backed structured products froze the short-term debt markets and made the crisis considerably worse, so it is worth asking what made these products so difficult.
They are hard to value even in calm markets
Liability structures and cash flow waterfalls in these deals are complex, and they hold several types of collateral together with interest rate triggers. Each structured product is unique, so the model simulating cash flows for any given bond has to be customized to that structure. The collateral pool has to be valued as well, and for an ABS trust that can mean valuing thousands of subprime mortgages, each with its own borrower characteristics and loan terms. CDOs may hold securities issued by ABS trusts, CDO-squared structures hold securities issued by other CDOs, and some pools contain synthetic ABS credit default swaps. Cash flows often turn on what the collateral will be worth later and how it will then be rated, both of which must be estimated first, and little data was available about the different asset pools even to sophisticated investors.
Transparency was missing on several fronts
Many investors, including some that looked sophisticated, lacked the in-house expertise to analyse what they were buying. They did not grasp the risks arising from the assumptions inside the valuation and credit rating models, so they became completely reliant on the rating agencies for risk measurement. A good many were yield buyers deciding on projected cash flows. That measures potential return badly, since it takes the estimates as accurate and assumes every cash flow can be reinvested at the computed cash flow yield.
Valuation of illiquid assets was opaque in any case. With no benchmark prices available, investors grew highly sceptical of reported prices when assessing the credit risk of a counterparty. The opacity extended to what sat inside the SIVs, since banks may hold assets there until they can be securitized and sold, leaving their exact holdings unknown. Reported Level 3 assets offered a rough guide: under the classification the United States Financial Accounting Standards Board required from 2006, Level 1 assets are valued at observable market prices, Level 2 assets are marked to market, and Level 3 values come from models and unobservable inputs. Outstanding commitments were similarly hard to size, including backstop lines of credit and loan commitments for private equity buyouts.
The trigger
Two events crystallised the doubt. In June 2007 Bear Stearns tried to rescue two hedge funds threatened by losses from subprime mortgages, and Merrill Lynch, prime broker to one of them, seized USD 850 million of underlying collateral and then struggled to sell any of it. In August 2007 BNP Paribas barred investors from withdrawing from three funds with USD 2.2 billion in assets, because it could not value the subprime assets they held. The market concluded that many structured products might be mispriced, the worry widened to the subprime exposure of large financial institutions, and the markets for wholesale short-term funding then shut down.
Systemic risk describes the danger that trouble at a single firm, or in a single market, spreads to other firms and markets, and in doing so puts whole markets or economies at risk. It is less a category of loss than a description of how losses travel.
Confidence in collateral disappears
Collateral quality is what limits borrower default risk in the ABCP and repo markets, so lenders must have confidence in the nature and value of the assets pledged to them. As both markets deteriorated, that confidence went. Lenders worried about whether the collateral contained subprime mortgages and whether the reported valuations could be relied upon. Because these markets were opaque, the doubt could not be resolved borrower by borrower, so even institutions without subprime exposure could not roll their debt. Managers of money market funds, large purchasers of ABCP and active participants in the repo markets, fled to Treasury bills.
Forced selling spreads the damage
Hedge funds that could not roll their debt had to sell assets, and because they hold a wide variety of assets the selling touched many markets at once. The CDO market was among the first hit. Funds that judged prices there artificially low, or that could not practically liquidate such holdings, sold whatever else they could. Closing out existing positions meant selling higher credit-rated assets and buying back the lower credit-rated assets they had shorted, which pushed higher quality prices down and lower quality prices up. Quantitative hedge funds trading on pricing patterns were hurt by that reversal, and institutional investors and hedge funds took losses unwinding carry trades to cut leverage.
Banks stop lending to each other
Banks began to hoard cash, partly because they could not size the drawdowns that might come on the backstop credit lines extended to SIVs, and commitments to underwrite leveraged buyouts added to the concern. During the first part of August 2007, three-month Libor, the London interbank offered rate, rose over 30 basis points. Reluctance to lend became widespread, credit standards tightened, the availability of residential and commercial mortgages was squeezed, and business lending was restricted. That is where a financial crisis became an economic crisis. The default of Lehman Brothers then showed how far contagion runs through the OTC derivatives market: it triggered a cascade of defaults among counterparties who could not get back their collateral, and dealers with no direct link to Lehman, but who were counterparties of failed direct counterparties, also defaulted.
As the crisis grew, the Federal Reserve and central banks elsewhere designed innovative liquidity injection facilities. From the fall of 2007 to the end of 2008 the Fed built backstops for most of the asset classes under stress: long-term lending facilities secured on high quality collateral, the discount window opened to investment banks and securities firms, funds lent against high-quality illiquid asset-backed securities, money to finance purchases of unsecured CP and ABCP, liquidity for money market funds, and purchases of assets from Fannie Mae and Freddie Mac.
The discount window, a Federal Reserve facility that helps institutions manage short-term liquidity needs, had not been open to firms of that kind before, so the step was a significant one. These were liquidity-targeted measures, and the size of central bank balance sheets increased considerably as a result.
Government intervention in the United States
The Term Auction Facility, or TAF, was implemented in December 2007 to provide funds to depository institutions by auctioning them against a wide range of collateral. The Primary Dealer Credit Facility, or PDCF, let the Fed lend to primary dealers through repos. The Economic Stimulus Act arrived in February 2008. Fannie Mae and Freddie Mac, the two United States government sponsored enterprises central to the mortgage markets of the day, were taken over by the government in September 2008. The Troubled Asset Relief Program, or TARP, came in October 2008, and on October 28, 2008 a total of USD 115 billion went to Bank of America, Citigroup, BNY Mellon, J.P. Morgan Chase, Goldman Sachs, State Street, Morgan Stanley and Wells Fargo.
The pattern outside the United States
Governments elsewhere offered liquidity support facilities and recapitalized insolvent banks, both aimed at restarting bank lending. Where that was not enough, institutions were nationalized, as Northern Rock was, or broken up and sold, as Fortis was. The cost moved onto government balance sheets, and in several European countries it fed the sovereign debt crisis of 2010.