Previous chapters have set out the ways in which Lehman Brothers sought to value its assets and to hide its losses. Professional standards for the valuation of commercial and residential real estate existed at that time, but as the bankruptcy Examiner Valukas demonstrates in his report, Lehman showed little interest in conforming to them or hiring those who knew how to apply them.
Against that background it can be seen that the bankruptcy process did not itself cause the destruction of value, although it added to it. The filing itself and the lengthy processes involved confirmed the anxieties which already existed in the market about Lehman. The destruction of value had already taken place before Lehman filed for bankruptcy. Before the end came, frantic attempts were made to restore value and, when that failed, to conceal the losses. The outcome of the bankruptcy proceedings raises the question of how and why value can disappear, and then return after a few years, at least when measured in terms of price. Examining the reasons for the fluctuations in price may cast light on whether or not price is the only measure of value, or if value remains when the price fluctuates. The purpose of this chapter is to set out the difficulties of valuing both complex financial instruments and real estate, and especially commercial real estate and development land. These are issues which economists do not take into account in developing theories of value. Such theories may also be detached from developments in financial instruments and the markets themselves. The chapter describes the methodologies used to value commercial real estate and complex financial instruments. It is already clear that Lehman either did not use the appropriate methodologies or applied them without making use of the relevant data, or did not use the correct verification procedures at all.
Lehman's assets fell into two broad categories: residential real estate and commercial real estate in particular, and derivatives. The latter are complex assets, especially when compared with more straightforward assets such as shares or bonds. Commercial real estate, especially development projects, are difficult to value, yet there are recognized procedures, which Lehman had not put in place, and even if they had, it is unlikely that they would have been followed. When Lehman filed for bankruptcy, it was holding $43bn in real estate loans and assets, of which SunCal was one. Lehman was SunCal's main financial backer, lending the developer over $2bn to buy land and add infrastructure, including roads, sewers and power lines, before selling it on to house builders for a profit. Lehman stopped funding construction work on SunCal's projects in 2008, so SunCal took more than 20 of these projects into bankruptcy protection.
Lehman should have used one of the large professional valuation firms in the USA to provide both the company and investors with the assurance of an independent valuation. Lehman's own methods of valuation left much to be desired. They selected TriMont, a property services company, to handle the commercial real estate portfolio, but this company did not offer valuations and clearly had difficulties in developing the appropriate models, according to the Examiner's Report. That was a mistake. It was hardly likely to inspire confidence in the valuation of their real estate assets, especially in 2007 and 2008, when prices in the real estate market were falling.
Recognized procedures have been developed by standard-setting bodies for valuation in the USA and in the UK. International standards are in the process of being established by the International Valuation Standards Council, a long-time partner of the Appraisal Foundation. This chapter draws on the procedures as set out by the Royal Institution of Chartered Surveyors in London, and the Appraisal Foundation, which sets the Uniform Standards of Professional Appraisal Practice (USPAP), the standards required by the Financial Institutions Reform, Recovery and Enforcement Act 1989.1 It draws on the recommendations of both bodies in explaining the approach to the valuation of commercial real estate.
There are two approaches to valuing development land. These are comparison with the sale price of land for comparable development, or an assessment of the value of the scheme as completed. The latter, the residual method, requires a deduction of the costs of development, including the developer's profit, in order to assess the underlying land value. In practice, both methods would be used, but the residual method requires the valuer to make a number of assumptions. The valuer inevitably has to exercise judgement, but in making that judgement the valuer has to undertake all the necessary research to establish the facts and his assessment of them. It should then be possible for the valuer to show that his assessment is reasonable, taking all of these factors into account. Timing is important, as the valuation will depend on how far the development has advanced.
The nature of the site itself (such as any risk of flooding, landslides, or adverse geotechnical conditions, such as the nature of the soil or slope instability), establishing the legal ownership of the land, and any planning issues are amongst the first items on the valuation agenda. Since the real value of the land lies in its development potential, valuing the land will again depend on making an accurate assessment of the form and extent of the kind of development that can be accommodated on the land. The valuation may have to be undertaken with a team of professional project consultants, such as architects and quantity surveyors. The valuer then takes into account the time taken to develop the site, and the development programme. This will include all the issues involved in the pre-construction period, building contracts, preparing the site, actual buildings, completion to sales and letting out the completed development, and then possible liabilities.
An analysis of the market to establish the potential demand for the optimum alternative forms of development that may be possible should take all these factors into account. The decision of the nature of the development should also depend on the investors' requirements, the location, access and availability of transport routes, all the amenities attractive to tenants and/or buyers, the scale of the development in terms of sale or lettable packages, the form of the development, and market supply, including actual or proposed competing developments. When the land is going to be developed over a period of time, it is also important that the developed land is revalued at intervals, say, every six months to a year, since market and macroeconomic conditions may change.
The alternative method is one of comparison, but this only works if evidence of sales can be found and analysed on a common basis, such as site area, developable area or habitable rooms. Price comparisons are a useful check on the residual method. Many other factors have to be taken into account, including the condition of the site, site and construction costs, dates of sales, changes in planning laws, changes in price in a rapidly changing market, so that the date of a transaction is very important. The greater the number of variables and adjustments for assumptions, the less useful the comparison; and anyway, the comparison methodology is not appropriate once construction has begun.
If there are no useful comparisons to be made, then the valuer should turn to the residual method, which involves the input of a large amount of data, and small changes in any of these inputs can lead cumulatively to large differences in the estimate of land value. For example, the required rate of return may vary according to whether the client is a developer, a contractor, an owner occupier, an investor or a lender, as well as the time taken with the development and the risks involved. The residual method can be expressed in a simple equation:
(value of completed development) - (development costs + developers profit) = land value.
The value of the completed development is the market value of the proposed development, assessed on the special assumption that the development is complete as at the date in the market conditions at that time. This is usually called the gross development value. Sometimes the value may be the total of the value of the individual properties, possibly on the assumption that the completed development is producing income, rather than being available for sale or letting. The net development value incorporates the transaction costs on the assumption that the completed development was sold on the date of valuation.
The planning, acquisition and site-related costs have already been outlined, to which should be added the costs of vacant possession, which might be offset by letting out advertising space, short-term tenancies or temporary car parking. Since large developments are phased over time, that is reflected in the developer's cash flows, and hence in the valuation of the property. This means that some of the costs can be deferred, as indeed receipts may be. Where the sales are residential properties or commercial retail units, they may begin before the development is completed, and may be phased in over a long period of time. The income then has to be recognized as cash flow and must be recognized at the appropriate time, as should relevant costs.
An accurate estimation of the building costs based on the gross internal area on the valuation date is a major component of the residual valuation. Clearly this method is very sensitive to variations in the estimated costs, and the degree to which the costs can be assessed accurately varies according to the specific site characteristics and the requirements for the development of the site. The valuer must also examine the exact contract made with the contractor, since even a fixed price contract may allow the contractor to amend the cost of the contract if there are any changes in the specification, or if unforeseen events occur.
The valuer has to take a wide range of costs into account, including fees and expenses, ranging from professional consultants, costs of legal advice, lettings and sales expenses, rent-free periods, raising development finance and prospective tenants or buyers incurring fees on monitoring the development. Other costs include interest on the costs of land and development, and opportunity costs, if it is a self-financing project. It is usual for interest to be regarded as a development cost up to the assumed letting date of the last unit, unless there is a forward sale agreement with different terms. In the case of a residential development, the units may be sold at different times, and various assumptions may be made regarding cash flow both inward and outward, and the rate of interest will reflect that.2 The two most widely used rates of return for developing and constructing residential property refer both to profit and yield either on a static basis, which subtracts all the costs and expenses outlined above, or on a dynamic model in a discounted cash flow which projects sales revenues and costs over time. The latter method was used by Lehman for residential developments. Objective sources of data were available, and could have been used by Lehman as a justification for the firm's valuations of their commercial real estate: the National Association of Homebuilders' annual publication, The Cost of Doing Business and the ‘PwC Real Estate Investor Survey’ or the ‘Developer Survey’, published by RealtyRates.com.3 All of these are reputable sources, but Lehman apparently did not refer to any of these.
In the previous chapter, the regulations which applied at the time in the USA were set out in detail, with the proviso that none of them applied to Lehman Brothers. Not only did Lehman not take any of the regulations into account, but also the company did not apply any reasonable measures of value to the development of residential or commercial plots of land. Lehman paid no attention whatsoever to any such considerations when investing in developments. Its approach was ‘systemically flawed because Lehman primarily valued these assets on whether the development was proceeding according to the project's business plan and not the price a buyer would pay for the asset’,4 or just by reference to the project's capitalization.
Each Lehman banker within PTG who originated investments had relationships with certain developers with a proven track record in a particular real estate market. Lehman relied on the developer to successfully complete the project and the relationship between Lehman and the developer could be described as a ‘marriage’.5
Completion of the project was necessary, as the underlying real estate was highly leveraged and the cash flow was not sufficient to repay the loan until it was completed. Lehman partnered with developers on the Principal Transactions Group. Not only Lehman's bankers were close to developers, but some of TriMont's asset managers were as well, according to the Examiner's Report, as they did not report on deteriorating market conditions but relied instead on the developers' assurances that the project would be a success. ‘Appraisals should reflect what market participants actually do, not necessarily a particular client's desires … Appraisers reflect the market; they do not set the market.’6 Such appraisals would not have pleased Lehman Brothers.
The level of risk increased in these investments, as Lehman sought the required rate of return. The focus was on land development, especially in California, and equity stakes increased in the hope of gaining profits in a declining market. Lehman did not even consider the loan-to-value requirements set out in the Interagency FAQs. As for conducting the valuation, the criteria should have included the amount of future expected cash flows from the asset, and an investor's willingness to accept the risk that the asset would not produce them. Instead, ‘the yields selected were based on Lehman's expected rate of return at origination, rather than the rate of return that a typical market investor would require’.7 The approach to measuring value did not fulfil any of the standards of valuation or regulations which were generally accepted at the time. Nor did Lehman use the methodologies set out above for valuing commercial property and developments in particular. It is small wonder that when it came to the crunch, Lehman's valuations of its real estate portfolio were greeted with scepticism.
As noted above, the regulations for valuations for development did not apply to Lehman Brothers as they were not regulated by the Federal Reserve or the OCC. However, these were also the issues which the SEC should have addressed as part of its Consolidated Supervised Entities Programme. It was not until February 2008 that the SEC embarked on a special project to review the CSE's price verification processes for their CRE portfolio. The report was not completed before Lehman collapsed, but the Examiner provided the details of the informal analysis and comments the SEC provided during an interview. Apparently the SEC expressed its concerns to Lehman Brothers throughout its inspection, which took place during February, March, May, June and July, and Lehman apparently addressed some of the issues raised, including an overview of the CMBS business, an overview of the real estate risk process and the fixed income product control purpose. Later in the investigation, the SEC inspected verification procedures, price verification models and the valuation of particular positions. The aim was to determine if appropriate valuation controls had been designed and to test the price verification process to ensure that controls are operating as intended.
The SEC refused to share with the Examiner any of its formal conclusions, but only with an informal analysis and comments during an interview.
The SEC's requests for further information in July included questions about Lehman's internal rate of return models and information on credit spreads. Their requests for information continued until early August. The Examiner noted that the SEC considered that ‘Lehman's price verification procedures were more pronounced than the other CSEs because the size of Lehman's balance sheet and the nature of its business’.8 The SEC also recognized that Lehman's product control staff was insufficient to provide an independent check on the valuations provided by the business desk and the number of assets in the CRE portfolio. As the SEC was unable to complete its report, it is not clear what its conclusions were or what actions, if any, would have been taken. At least, the SEC should have been aware of the extent of the risky investments held by Lehman in 2008: Level 3 commercial mortgage loans and securities totalling $13bn, or 63.1 per cent of $20.6bn of total mortgage and asset-backed securities, and $22.7bn real estate held for sale, an illiquid asset class, but not one which is subject to the Fair Value rule, SFAS 157. In February 2008 Lehman provided a document to the SEC, entitled ‘An Overview of the CMBS presentation to the SEC’. It largely describes Lehman's CMBS activities, but does not provide any substantial explanation of its methods of valuation of the portfolio.9
The Office of Thrift Supervision belatedly turned its attention to an examination of Lehman Brothers, starting in July 2008. (Lehman did own a small FDIC thrift, regulated by the OTS and was therefore subject to OTS inspections. The thrift was used as a conduit for mortgage securitization activities.)10 The report, however, is simply that. It does not include any of the critical assessments of Lehman's procedures for valuing its commercial real estate later made by the Examiner. Its production at that time is unclear, since it does not refer to any courses of action to be taken by the OTS or by Lehman Brothers.
Collateralized debt obligations
The rest of this chapter examines some of the major derivatives in force when Lehman's collapsed, all of which depended on or were derived from assets such as loans on property. These include collateralized debt obligations (CDOs) and collateralized loan obligations (CLOs), as well as credit default swaps, which provide protection against credit loss on some other entity as the result of a specific credit event. The way in which these derivatives work will be explained further, as each type of derivative in considered. As a background to the discussion about derivatives, it should be noted that Lehman underwrote more mortgage-backed securities than any other firm, especially when the rate of delinquencies and foreclosures was rising in 2007: some $85bn, four times its shareholder equity in the fourth quarter of 2007. The valuation methodologies used by Lehman incorporated a variety of inputs, including prices observed from execution of trades in the marketplace as well as their own traders, ABX, CMBX and similar indices which track the performance of a series of credit default swaps and other market information, such as data on remittances received and cumulative loss data on underlying obligations. Their methodology was not quite as rigorous as might appear, since ‘each trader had a different method for valuing the CDOs and there was no consistency from desk to desk’.11
The PCG analysed the traders' price valuations, paying closer attention to the marks towards the end of each quarter, but they lacked the skills and resources. The PCG did not have the ‘same level of quantitative sophistication as many of the desk personnel who developed models to price the CDOs’. A former Vice President and Head of the Credit Valuation Group told the Examiner, ‘We are not quants’.12 The Group did, however, create a model using the Intex Cash Flow engine, or interest-only models, which depended on the underlying assets of the CDOs. However, the method was considered to be unreliable, so most members of the PCG tended to check their valuations against the desk traders' numbers. Even then they only checked 78 per cent of the CDOs valued by the desk traders. In many cases this was because ‘the desk had already written down the position significantly’.13 Using these methods, Lehman valued its derivative assets at $46.3bn and its derivative liabilities at a net value of $22.2bn. In other words, the valuation constituted almost a 100 per cent increase from the net value of $12.974bn reported on 30 November 2007.
A collateralized debt obligation (CDO) is a structured product, in which pools of cash-generating assets are packaged together into discrete tranches which can be sold to investors. They are called CDOs because the pooled assets, such as mortgages, bonds and loans, are debt obligations, which serve as collateral. Suppose a CDO consists of mortgages. The senior tranches are high-quality mortgages where the borrower is unlikely to default on the loan. The risks of such mortgages are much less than a subprime mortgage, where the borrower may well be unable to maintain repayments on the mortgage, and may default. The risks of the borrower defaulting on the loan are much higher than the loans in the senior tranches of mortgages. The senior tranches are safer investments as they will have first call on the collateral, in the event of a default. They have a higher credit rating than the junior tranches, so they offer lower coupon rates than the junior tranches, which offer higher coupon rates to compensate for the higher risk of default. CDOs are unusual in that they represent different types of debt and credit risk, with the different kinds of debt being described as ‘tranches’ or ‘slices’, each of which has a different maturity and risk associated with it. A CDO investor takes a position in an entity which has defined risks and rewards, not directly on the underlying assets. The investment is dependent on the quality of metrics for defining the risk and reward of the various tranches.
The process of constructing a CDO involves the establishment of a special purpose vehicle (SPV) to acquire a portfolio of fixed income assets, such as mortgage-backed securities. This allows the parent company to make highly leveraged or speculative investments without endangering the entire company. It will also serve as a counterparty for swaps and other derivative instruments. The investment bank is also the issuer of the CDO, but only through the SPV, earning a commission at the time of issue and a management fee during the life of the CDO. The order in which senior, junior and equity notes are entitled to the cash flows from the CDOs are set out as the ‘Priority of Payments’ in the transaction documents, with the senior tranches receiving payments from the principal first, with prepayments and payments until they are paid off. An investment in a CDO is therefore an investment in the cash flows of the assets, rather than a direct investment in the underlying collateral. The CDOs can take two forms: cash CDOs and synthetic CDOs, with the former paying interest and principal to the tranche-holders using the cash flows produced by the CDOs' assets. Ownership of the CDO lies with the SPV. Synthetic CDOs do not own cash assets. The investor has the exposure to risk and reward, but this exposure is realized through a credit derivative, a credit default swap (CDS). In this case, the underlying asset is a bond market instrument such as a high-yield bond index or mortgage index. This, of course, means that the relationship between the CDO and the ultimate assets is much more remote.
Having set up a CDO, the next step was to price the CDO, and that required a means of modelling the correlation between defaults of different assets, bonds or loans. A particular model, the Gaussian Copula Model, was widely used following its invention by David X Li in 1999. It used a so-called ‘copula function’, which was a way of ‘coupling’ a set of risks of default. The correlation in question was that between the equity tranche and other tranches including the most senior tranche. If the correlations were high, then even the holders of the most senior tranches would be at risk of losing their investment. Hence modelling the correlation was the most crucial element in valuing a CDO.14 David X Li's approach was to use historical data, which indicates that default rates are higher in recessions than in periods of strong economic growth, implying some correlation between survival times. Based on the assumption that survival times are correlated, the joint distribution for survival times makes it possible to draw conclusions about the portfolio. A copula function is then calculated to obtain the joint distribution from the marginal distribution. It was designed to provide one number that gave the probability of all the defaults in a pool of securities or mortgages taking place at once. If the default correlation was low, as they were not related to or dependent on each other, then this resulted in a low figure, which was taken to mean that the risk of default was low. The calculation did not depend on an analysis of historical data on actual defaults but was based on indices of credit default swaps. However, with regard to the credit default swaps, the length of the data series was only some ten years by 2007–8, since CDSs were only introduced in 1997 by Blythe Masters and her team at JP Morgan.
Others pointed out serious flaws in the distribution models. They were ‘overly simplistic’ and an inadequate means of valuing tranches of CDOs and other structured products. As far back as July 2008, Jon Gregory, former head of Global Credit Quantitative Analytics at Barclays Capital, London, showed that applying the Gaussian model, a super senior tranche of a CDO referencing 125 investment-grade assets should theoretically be able to withstand 46 individual default events before the super senior tranche experienced any loss of principal:
So the result is that the holder of one of these tranches, with spreads having blown out massively over the past twelve months, would be able to receive as much as 50 basis points for holding a CDO that, on this basis, apparently has no risk. That is opposed to the three or four basis points an investor would have received on this kind of structure in early 2007.…we have to ask how losses are going to factor into the (super senior) tranche over a given period. Will losses accumulate in a linear fashion over time so that you're just as likely to lose money in the first year of the trade as the second year, or will defaults be back-loaded so that losses are unlikely in the first three years but more likely later on?15
MacKenzie and Taylor Spears point out the deficiencies in the model, as perceived by quantitative analysts (quants or analysts whose work is to design and implement complex models allowing investment banks to price and trade securities), many of whom were sceptical about the use of the copula model even before the crisis began.16 They argue that the source of the problem lay with the use of the model by the rating agencies when this was applied to CDOs for which the underlying assets were mortgages rather than pools of corporate debt. The rating agencies made minor modifications to the models used for CDOs and corporate debt, and made matters worse by rating the CDOs and the mortgage-backed securities separately. For example in November 2001, when Standard & Poor introduced its new one-period Gaussian copula system, CDO Evaluator, the same value (0.3) was used for the correlation between ABSs from the same sector as was used for the correlation between corporations in the same industry.
However, the difficulty with applying the copula model to estimate expected collateral losses on ABS CDOs was that it relied on correlations among bonds in the CDOs. The probability of default and expected losses produced by these models is that they are very sensitive to model parameters based on the historic performance of the underlying bonds, which were very reliable before 2007. ‘Crucially, these models were not designed to directly capture the sensitivity of the performance of the underlying assets to systemic changes in house prices or changes in attributes that affect the mortgage loan performance.’17
Standard & Poor announced the introduction of its new CDO Evaluator in November 2001, designed to refine CDO credit criteria and analysis.18 It took into account the credit rating, size and maturity of each tranche of each asset combined in the CDO, along with the correlation between each pair of assets.19 It was used to assess the credit risk of a portfolio of assets for cash flow and synthetic CDOs, with the range of the probability of default from 0 to 100 per cent, but with the most likely outcome being that some but not all of the assets default. Taking the correlation between the assets into account, S&P adopted the Monte Carlo approach to estimating the probability distribution of default rates, with a large number of independent trials to determine which asset has defaulted or not. S&P argued that its methodology is robust due to its ability to deal with complex relationships between variables and can handle the effects created by portfolios containing assets that are unequal in principal balance, credit rating and maturity. S&P boasted that this approach
enables one to simulate the behaviour of a system as it is modelled and then simply observe the results … The key to successfully using Monte Carlo simulation techniques is one of performing enough trials to capture long-term certainty … 30 seconds for 15,000 trials on a portfolio of 100 assets and 2.5 minutes for 100,000 trials on the same portfolio.20
The speed was extremely important, as so many CDOs based on mortgages were being issued that a method of valuation which bypassed analysis of each CDO, yet allowed for a cash flow analysis, was essential. The method also had to enable analysts to verify that each CDO tranche can continue to pay the principal and interest in accordance with the terms, up to the scenario default rates on the underlying portfolio.
The rating agencies were competing against each other to grab as large a market share of the lucrative market for rating the CDOs as possible. The existence of that market was partly created by investor demand for credit ratings. CDOs contain hundreds of underlying assets and modelling the pay-offs from these securities requires sophisticated cash-flow models, so the credit ratings were a substitute for any due diligence on their part. There was little public information on how the ratings were calculated and how ratings in the CDO securities related to the underlying asset quality. At the same time, financial regulation in the USA at the time relied heavily on the ratings provided by the ‘regulated’ rating agencies. Minimum capital requirements for banks, insurance companies and broker-dealers depended on the credit ratings of the assets on their balance sheets, as do pension funds, creating an institutional demand for highly rated securities, although the supply of highly rated single name securities is inevitably somewhat limited. That regulatory demand is still there, even though the fees for the ratings provided were paid by the financial institution, which was either rated itself or whose products were rated by one or other of the agencies.21
But the financial crisis finally exposed the weaknesses in the model, especially the assumption that correlations between different assets were constant over time and could be expressed in one number. Convenient though that assumption may have been, it turned out not only to have been a mistake, but a disastrous one as well. Their model sought to estimate the actual probabilities of default, which required the use of historical records. With mortgage-backed securities, the postwar records in the USA included one relatively mild recession but a relatively stable housing market during that period. That all changed from the mid-1990s onwards. What followed was a period of rapidly increasing house prices, which too many believed would not fall, even when the rise in house prices began to falter in mid-2006. Overblown valuations persisted until 2008.
Indeed, as argued by Cordell and others above, ‘the only way to correctly value a CDO was by digging deep into its collateral and valuing it using loan level data.’22 Lehman was aware of the risks, at least at one stage in 2005. The firm conducted bond analysis on new subprime issuance in 2005 across different house price appreciation scenarios (HPA). Their conclusion? ‘New issue BBB subordinates have downgrade risk if HPA slows by 5% by end-2005 … BBB subordination appears to be sized to an intermediate scenario between a 5% and 8% HPA for life’. It is thought that one of the co-authors, Sihan Shu, went on to work for John Paulson, the hedge fund manager who made billions by shorting the subprime market.23
Apart from Lehman's losses, the inadequacy of the model and its application to CDOs based on mortgages led to major losses for the leading banks on their CDOs based on asset-backed securities. Citigroup lost $34bn, Merrill Lynch $26 bn, UBS $22bn and AIG $33bn. Looking at the rising rate of delinquencies and foreclosures and falling house prices from 2006 onwards, instead of only using models, might have helped to avert such heavy losses. Too much faith was placed in the model and its ability to estimate default probabilities, and too little attention was paid to what was going on in the world around them.
Credit default swaps
As explained above, a credit default swap (CDS) is a kind of insurance against a credit risk. It is a privately negotiated contract, in which one party buys protection against losses from defaults, the failure to pay interest or capital according to schedule or a debt moratorium, for example, and another, who is willing to take on the risk and to sell that protection. That seems straightforward enough, but then this quasi-insurance product enables others to implement various investment strategies; for example, a fixed income investor might buy protection so that he can hedge the risk of default involved in a corporate bond. Then an insurance company might act as the counterparty to this deal and promise to pay potential default losses. The seller of protection might then speculate on the bond issuer remaining in existence, so he is exposed to the investment without having to actually buy the bond in the cash market. These are some examples of the more complicated forms of credit risk derivatives. They all share the basic features of credit default swaps. They are designed to enable credit trading to take place and for risk to be shared amongst various market participants. The significant default swaps were the ones offering protection against defaults on portfolios of subprime mortgages or on CDOs which contained subprime mortgages. They had a part to play in the financial crisis. The CDSs were hard to price when they were written to provide protection to the holders of CDOs, themselves hard to price. The CDOs consisted of tranches of mortgages and were given triple A ratings by the rating agencies, which no doubt gave confidence not only to the holders of CDOs (and the sellers) but to the purchasers of credit default swaps. These were written to provide protection to the holders of CDOs as well.
The notional amount of outstanding CDSs grew rapidly from the beginnings of the market in the mid-1990s to a peak of almost $69 trillion at the end of 2007, but then declined sharply to just over $30 trillion at the end of the first half of 2010.24 Trading volumes were higher than in the first part of 2007, according to Markit, a leading financial information service company. The subsequent drop in the volume of outstanding CDSs was due to trade compression and the move to central counterparties in the CDS market. The main concerns in 2008 were focused on the fact that the major CDS dealers were important counterparties to one another, and the exposures were large in gross terms. In addition the value of these exposures grew rapidly as credit spread widened, making it appear that any agreement to net obligations across contracts might not be enforced if there was a default.
Partly as a result of the role of CDSs in the financial crisis, and partly to support the central clearing of trades, 8 April 2009 saw a global ‘Big Bang’ – a major change in the CDS market. Contracts are now standardized. The ‘event determination’ committee was introduced, as a central decision point and trigger for credit and succession events. This prevents differing conclusions or triggers for different contracts on the same entity. When there is a credit event, there is a binding and standard cash settlement price. Fixed coupons ensure that the payment amounts are standardized, thereby making it easier to offset contracts. To this was added the standardization of accruals, which makes the timing and amount of payments uniform in the first premium period and throughout the contract across all trades, thus making it easier to offset contracts. ‘The goals of reducing outstanding trades by trillions of notional dollars, restructuring the way trades are processed so that trades can be matched on the same day and the creation of a central counterparty mechanism are ambitious’, but that is the aim.25 The proposed changes had been underway before the financial crisis, but were no doubt accelerated by it.
Credit default swaps are a form of insurance in which the ‘reference entity’ takes various forms, such as a loan, a bond or some other kind of liability. Prior to the Big Bang, the CDS market operated quite differently, without the advantages of central clearing and standard contracts. The buyer of the protection usually pays the premium at fixed intervals over a period of time, and, generally, the seller of the protection pays compensation to the buyer if a ‘credit event’ (a default, for example) occurs and the contract is terminated. The value of a default swap depends not only on the credit quality of the underlying ‘reference entity’ but also on the credit quality of the writer of the CDS or the counterparty, since if the counterparty defaults, then the buyer of the default swap will not receive any payment if a credit event occurs. The value of a CDS depends on the probability of a counterparty default, the probability of a ‘reference entity’ default and the correlation between them.
The most important element in a typical derivatives transaction is the use of collateral, and the amount of collateral insuring a counterparty's performance on a contract changes with the value of the contract. If a bank had derivative positions with Lehman, for example, which would cost $100m. to replace, and if it had $110m. of collateral from Lehman when it failed, the bank could use the collateral and make no loss from Lehman's failure. Collateral arrangements were quite usual, but they were not universal. ISDA's survey in 2007 showed that 6 per cent of derivative contracts were subject to these agreements then (compared with 30 per cent in 2003). There was still a possibility of contagion through derivative exposures, depending on whether or not counterparties chose to manage their exposures actively. The failure of a large financial institution can lead to substantial changes in CDS prices, as well as liquidity, so that the collateral held just before the failure may not be enough to cover losses if other counterparties default.
Credit default swaps exist for asset-backed securities (ABSs) as well. The corporate CDSs are relatively simple. They first appeared on the scene in 1993, and were widely used from the late 1990s onwards, especially after the introduction of the ISDA template in July 1999. The ABS CDSs, which are usually for residential mortgage-backed securities but also on CDOs and commercial mortgage-backed securities as well, are more complex. They first appeared on the market in 2005 after the introduction of the ISDA template in June of that year. These are inevitably more complex than corporate CDSs. Mortgages are placed in a pool, typically set up as a trust, and notes, often called tranches, differ in their priority in receiving payments, with the most senior tranche having the first claim on interest payments and mortgage profits, since they always have a triple A rating. When mortgages default, the lowest-rated tranches suffer first from the default losses;, as these increase, even the highest-rated securities may experience default losses as well. A bank holding that debt and wishing to insure it could do so through a credit default swap, but a tranche of subprime securitized debt may not lead directly to a default but only to a reduction in debt payments, due to a rising level of defaults in the underlying mortgages. If an investor holds a AAA tranche with a principal amount of $100m. and the other tranches of the securitization are wiped out, and that during the month another $1m. of mortgages default so that the principal balance falls from $100m. to $99m., the investor would be paid $1m. from the CDS. The CDS would still exist after that payment, and would make payments as further mortgages default until the maturing of the contract.
Settling corporate CDS contracts in a credit event results in a payment to the buyer from the protection seller and the termination of the contract, but if such an event does not occur before the contract matures in anything from two to ten years, then the seller does not make a payment to the buyer. For these CDSs the settlement process is also straightforward. It is a cash settlement after a credit event auction under ISDA rules to establish the value of the reference obligation. The auction sets a price for all market participants who choose to settle in cash.
What happened after Lehman declared bankruptcy is a good example of the way in which the auction works. It took place on 10 October 2008, and a price of 8.625 cents on the dollar for Lehman Brothers debt was agreed. This meant that the sellers of protection on Lehman CDSs had to pay 91.375 cents on the dollar to buyers to settle and terminate the contracts. Those who had bought Lehman bonds but had also taken out protection through a CDS contract received 91.375 cents on the dollar, which offset any losses resulting from the bankruptcy. It was estimated that between $6bn and $8bn changed hands during the cash settlement of the CDS auction. The auction ended with a net payment of $5.2bn. Of course, those who had bought the bonds would have expected to receive par 100 when the bonds matured. However, the CDS meant that Lehman bond holders with CDSs received much more than they would have received as the recovery value, some years after the bankruptcy procedures were completed in March 2012.
Settling CDSs which refer to asset-backed securities is much more complicated. These are very different ‘insurance’ instruments from the ones in the corporate credit market. The seller of protection provides the buyer with protection covering the failure to pay off the principal by the legal date of maturity of the ‘reference obligation’, interest shortfall, that is, the amount of interest is less than required. The buyer can terminate the contract, if there is a ‘floating amount event’. This refers to a write-down, a failure to pay the principal or an interest shortfall. It can include a distressed ratings downgrade for the asset-backed security itself, which could be a downgrade to CCC/Caa2, or even the withdrawal of a rating. That kind of downgrade means that the asset is no longer of investment quality.
In 2005, ISDA produced templates for that process and revised them in 2006, to deal with ‘payment events’ where the asset-backed security fails to meet the obligations set out in the contract. There are three options for determining the size of the payment to the borrower: fixed cap, variable cap and no cap. The first one requires that the seller has to pay the buyer the fixed rate; the variable cap means that the seller has to make up any interest shortfall on the security up to LIBOR plus the fixed rate, and in the third case, the seller has to make up any shortfall in the interest on the security. The buyer also has the option of ending the contract.
The price of a credit default swap is quoted as an annual percentage of the contract's notional value, taking into account the likelihood of default, the recovery rate in the event of a default and the liquidity, regulatory and market views about the credit, that is, the corporate bond and the quality of the issuer. The value or price of the corporate bond depends both on the financial strength of the company issuing the bonds and the market in which the company operates. These are the ultimate features on which the value of the derivative depends. The value is also subject to the risks of systemic failures which affect all the companies trading in CDSs based on corporate bonds. Ultimately, though, the positions between two counterparties, the banks trading across various derivatives, tend to have offsetting exposures, since some have a positive market value to a given counterparty and others have a negative market value. These have a netting effect so that only a net amount of the market value is at risk in a default by one of the counterparties.
However, it is possible that, because of their built-in leverage, CDSs may have made it more likely for investors to take on more risks than they would have done otherwise. Because of the way in which banks' capital requirements were assessed, financial institutions were able to hold less regulatory capital if they packaged loans into securities and held them on their balance sheet, than if they just kept the loans on their balance sheets. Others took the view that it was to their advantage to hold senior tranches of securitizations on their books if they insured them with CDSs and regulators allowed financial institutions to set aside less capital if they had bought such protection. That alone was a major reason for the increased demand for credit default swaps.
In a market in which CDSs are often sold and resold amongst parties, there are risks. Buyers may not be as financially sound as they should be to cover the obligation when there is a credit event, especially without collateral. To take account of this, the New York Federal Bank advised counterparties to inform their trading partners when the contract has been assigned to others. Concerns were raised by regulators in the UK and the USA about the growing backlog of documentation covering confirmation of the initial transactions and then the transfer from one party to a third party. The Federal Reserve Bank of New York called fourteen derivatives dealers, including the Bank of America, Barclays, Bear Stearns and Lehman Brothers, to a meeting, as a result of which a five-point plan was agreed, as well as the ISDA Novation Protocol of September 2005. The Protocol includes an agreement not to accept any transfer of a credit derivative unless the transfer has been agreed with the remaining party, who must receive full details of the transaction, all using the Depository Trust and Clearing Corporation. Presumably, all were relieved that they had agreed to this Protocol in September 2008. Risks remained in the market, such as counterparty concentration risk. That risk had to be tackled by government intervention with the failure of a major counterparty, AIG, which would have left many market participants un-hedged and exposed to losses, as well as worsening the loss of liquidity.
In 2006, the ABX indices on subprime securitizations were introduced, representing a basket of credit default swap contracts on securitization tranches. In 2007, they became a widely followed barometer of the collapsing valuations in the subprime mortgage market and were also used by banks and other investors as a tool for hedging and trading as well. These indices consist of a series of equally weighted, static portfolios of CDSs, based on twenty subprime mortgage-backed securities, and were introduced when the issuance of mortgage-backed securities was strong in 2005.
A new series is introduced every six months, based on twenty completely new subprime deals, issued during the previous six months. The prices reflect the willingness of investors to buy or sell default protection, given their current views on the risk of the underlying subprime loans.26
The authors, Fender and Schneider, then consider the ABX prices for default protection and conclude that declining risk appetite and rising concerns about market illiquidity were some of the main drivers of the collapse in ABX prices since the summer of 2007. Indications of changes in risk appetite with regard to subprime mortgage risk may help to explain any deviation between observed market prices for the ABX indices and the projection of default-related cash flow shortfalls in the underlying subprime MBSs. They add that ‘observed ABX prices are unlikely to be good predictors of future default-related cash flow shortfalls on outstanding subprime MBSs … This is in part because coverage of the ABX indices extends only to a small fraction of the outstanding subprime MBS universe.’27
Another study analysing the use of the ABX.HE index of the CDS market points out that the focus of regulators on banks using the CDS market prices as the basis for marking their portfolios to market was mistaken. The authors find that the market prices for AAA ABX.HE index CDS ‘at the peak of the financial crisis in June 2009 [were] inconsistent with any reasonable assumption for future default rates’.28 They added,
We find that percentage changes in CDS prices are only weakly correlated with the credit performance of the underlying losses but are strongly affected by changes in the short-sale imbalances in the equity markets of the publicly traded investment banks. Because the short-sale activity measure is a proxy for demand imbalance in the market for mortgage default insurance, the relative importance of its correlation with price dynamics suggests that the practice of using the AAA ABX.HE index CDS to value subprime mortgage portfolios is quite problematic.29
That must be somewhat of an understatement!
In the years before the financial crisis, the possibility of the triple A tranches of CDOs with substantial AAA tranches defaulting did not seem at all likely. The significant probability of default would only occur if there was a major downturn in the housing market throughout the USA, which appeared to be an extremely low probability, based on historical data. When the downturn happened, the value of the underlying assets fell sharply and the value of AIG's credit default swap liabilities became very large when marked-to-market.
By August 2008, AIG had a total of $26.2bn of unrealized losses on its CDSs, on top of other losses on its subprime and other fixed income securities. The company posted $16.5bn collateral. On 15 September, a buyer of $10m. of protection of debt for five years had to pay $2.5m. immediately plus a $500,000 annual premium. On 16 September, after its downgrade by S&P and Moody's, AIG had to post $14.5bn additional collateral. It could not provide the additional collateral without a bail-out.
AIG's credit default swap business began in earnest in 2002, when Joe Cassano became CEO of AIG Financial Products (AIGFP), based in London. By December 2007, AIG had written CDSs with a notional value of $527bn, of which multi-sector CDOs amounted to $78bn, which the firm had begun originating in 2003, when AIG was AAA-rated. Cassano stated in his testimony to the Financial Crisis Inquiry Commission, that
we made a decision at the end of 2005 to stop writing new deals that contained subprime capital. Although we completed deals that were already in the pipe-line, the portfolio grew comparatively little after 2006 … In dollar terms, our business grew every year because the CDS contracts were multi-year: adding even one increased the notional size of the book.
AIGFP decided to stop writing CDS protection on CDOs with subprime exposure. ‘We announced our decision to the marketplace in February 2006’, he added.30 That is borne out by the Federal Reserve data, showing that AIG had borrowed a total of $128bn by the end of 2008, of which $28bn was used to purchase the CDOs on which AIG had sold protection and $20bn was allotted to buy subprime mortgage-backed securities in which AIG had invested as part of its securities lending programme. The remaining $80bn was a loan to enable the company to continue operating, whilst it sold off its non-core assets and capital investment through the TARP programme.31 ISDA's argument is that it was not the CDSs per se which gave rise to the losses, but to the ‘failure to assess the risks of MBSs, CDOs and other mortgage exposures’.
That view is supported by others and indeed by Cassano's own testimony. He accepted the analysis provided in 2005 by Eugene Park, who had taken over as the company's liaison with Wall Street. Park stated in 2005:
The subprime loans underlying many CDOs formed too large a part of the packaged debt, increasing the risk to unacceptable levels. Those loans could default at any time, anywhere across the country, because the underwriting processes had been so shoddy. The diversification was a myth – if the housing market went bust, the subprime would collapse like a house of cards.32
AIGFP was advised and used mathematical models provided by Professor G. Gorton, an expert in corporate finance and models. These did not include the possibility of AIG losing its triple-A rating, which it did, or the growth of risky subprime lending. The value of such derivatives depends ultimately on the assets from which they derive. Models which do not incorporate the risks implicit in the underlying assets will fail to provide the users with an understanding of the risks involved.