Mortgage Securitization Mortgage Securitization g gg gMeltdown: Analytical Failures Meltdown: Analytical Failures

d P ld P land Proposalsand Proposals

Dan Dan diBartolomeodiBartolomeoLondon, November 2009London, November 2009London, November 2009London, November 2009

Where Are We?Where Are We?Where Are We?Where Are We?•• World financial markets have suffered very large, acrossWorld financial markets have suffered very large, across--

thethe--board declines in valueboard declines in value•• Unprecedented declines in residential property values Unprecedented declines in residential property values

have been experienced in many countrieshave been experienced in many countrieshave been experienced in many countrieshave been experienced in many countries•• The banking systems of most Western countries have in The banking systems of most Western countries have in

complete disarray, requiring huge government complete disarray, requiring huge government interventionsinterventions

•• A global recession is ongoing, with large negative effects A global recession is ongoing, with large negative effects impacting employment trade and government revenuesimpacting employment trade and government revenuesimpacting employment, trade and government revenuesimpacting employment, trade and government revenues

•• Previously active markets in securitized debt, and related Previously active markets in securitized debt, and related credit risk transfer instruments have ground to a haltcredit risk transfer instruments have ground to a haltgg

How Did We Get HereHow Did We Get Here

•• While there are a myriad of “big picture” contributors to While there are a myriad of “big picture” contributors to th t i i ill f ti l f il ith t i i ill f ti l f il ithe current crisis, we will focus on particular failures in the current crisis, we will focus on particular failures in fixed income analysis that are the root of the problemfixed income analysis that are the root of the problem–– The BET and “Diversity Scoring” methods developed by Moody’s The BET and “Diversity Scoring” methods developed by Moody’s y g p y yy g p y y

to give credit ratings for securitized debtto give credit ratings for securitized debt–– Notching ratings as practiced by S&P and othersNotching ratings as practiced by S&P and others–– The Gaussian copula method (Li 2000) for estimating credit riskThe Gaussian copula method (Li 2000) for estimating credit riskThe Gaussian copula method (Li 2000) for estimating credit risk The Gaussian copula method (Li 2000) for estimating credit risk

of loan poolsof loan pools–– Reliance on credit swap curve data to estimate credit risk Reliance on credit swap curve data to estimate credit risk

correlationscorrelationscorrelations correlations –– Failure to incorporate market data on expectation for housing Failure to incorporate market data on expectation for housing

pricesprices

The Appetite for RiskThe Appetite for RiskThe Appetite for RiskThe Appetite for Risk

•• Under the Basel II banking regulations, banks are Under the Basel II banking regulations, banks are required to have capital reserves against the risks required to have capital reserves against the risks decline in asset values essentially controlling the degreedecline in asset values essentially controlling the degreedecline in asset values, essentially controlling the degree decline in asset values, essentially controlling the degree of leverage in balance sheetsof leverage in balance sheets

•• Which of the following values is closest to the Basel II Which of the following values is closest to the Basel II capital reserve requirement for a bank to hold securities capital reserve requirement for a bank to hold securities rated AAA by any of the major rating agencies?rated AAA by any of the major rating agencies?–– 8%8%8%8%–– 4%4%–– 1%1%

25%25%–– .25%.25%

Basel II Reserve RequirementBasel II Reserve RequirementBasel II Reserve RequirementBasel II Reserve Requirement

•• The correct answer is .25%. The correct answer is .25%. •• The actual value is .56%, with an 8% reserve The actual value is .56%, with an 8% reserve

requirement and a 7% “at risk” ratingrequirement and a 7% “at risk” ratingrequirement and a 7% at risk ratingrequirement and a 7% at risk rating•• Effectively banks can lever 178 to 1 on AAA rated Effectively banks can lever 178 to 1 on AAA rated

securitiessecurities•• Once a way was found to create AAA rated securities Once a way was found to create AAA rated securities

through securitization of lower rated debt, the global through securitization of lower rated debt, the global demand for AAA rated securities skyrocketeddemand for AAA rated securities skyrocketeddemand for AAA rated securities skyrocketeddemand for AAA rated securities skyrocketed

•• The rating agencies, although private, profit making The rating agencies, although private, profit making enterprises, effectively had global regulatory powersenterprises, effectively had global regulatory powersp , y g g y pp , y g g y p

Start at the BeginningStart at the Beginningg gg g•• In the late 1990s, Moody’s began to provide credit In the late 1990s, Moody’s began to provide credit

ratings of securitized corporate loans (CLO) based on a ratings of securitized corporate loans (CLO) based on a g p ( )g p ( )method known as the Binomial Expansion Techniquemethod known as the Binomial Expansion Technique

•• Once a default probability had been estimated for loans Once a default probability had been estimated for loans in a pool the p obabilities of potential m ltiple defa ltsin a pool the p obabilities of potential m ltiple defa ltsin a pool, the probabilities of potential multiple defaults in a pool, the probabilities of potential multiple defaults within the pool was based on simple binomial probability within the pool was based on simple binomial probability formulas that assumed that defaults would be fully formulas that assumed that defaults would be fully independent across borrowersindependent across borrowers

•• To account for the obvious flaw that defaults are likely to To account for the obvious flaw that defaults are likely to be correlated Moody’s introduced an adjustment calledbe correlated Moody’s introduced an adjustment calledbe correlated, Moody s introduced an adjustment called be correlated, Moody s introduced an adjustment called Diversity ScoringDiversity Scoring

•• In 1999, Northfield published a research paper criticizing In 1999, Northfield published a research paper criticizing BET and Diversity Scoring based on a client requestBET and Diversity Scoring based on a client request

Moody’s Diversity ScoringMoody’s Diversity Scoringy y gy y g•• To account for default correlation, Diversity Scoring To account for default correlation, Diversity Scoring

reduced the assumed number of issues in a loan poolreduced the assumed number of issues in a loan poolpp•• For example, if you had 70 loans in a pool, you might For example, if you had 70 loans in a pool, you might

count that as 40 independent loanscount that as 40 independent loans•• Moody’s broke firms into 32 industriesMoody’s broke firms into 32 industries

–– 2 credits in the same industry counted as 1.52 credits in the same industry counted as 1.5–– 10 credits in the same industry counted as 410 credits in the same industry counted as 410 credits in the same industry counted as 410 credits in the same industry counted as 4

•• Implicitly there is an assumption that defaults are Implicitly there is an assumption that defaults are correlated within industries, but never across industriescorrelated within industries, but never across industries–– Completely ignores potential for pervasive effects of recession, Completely ignores potential for pervasive effects of recession,

war or other systemic influenceswar or other systemic influences

•• Although Moody’s largely abandoned BET in 2005, many Although Moody’s largely abandoned BET in 2005, many g y g y , yg y g y , yinstitutions such as AIG continued to use the methodinstitutions such as AIG continued to use the method

“The Secret Formula That “The Secret Formula That Destroyed Wall Street”Destroyed Wall Street”•• Cover story in WIRED magazine, March 2009Cover story in WIRED magazine, March 2009•• Refers to the “Gaussian Copula” approach for estimating Refers to the “Gaussian Copula” approach for estimating

default risk across a pool of loans from Li (2000)default risk across a pool of loans from Li (2000)default risk across a pool of loans, from Li (2000)default risk across a pool of loans, from Li (2000)–– Allows for closed form calculation of marginal risks as more and Allows for closed form calculation of marginal risks as more and

more loans are added to the pool to be securitiesmore loans are added to the pool to be securitiesf h d l f d f lf h d l f d f l–– Requires an assumption of the expected correlation of defaultsRequires an assumption of the expected correlation of defaults

•• The problem is credit risky instruments have high skew The problem is credit risky instruments have high skew in the payoff distribution, so the joint distribution will bein the payoff distribution, so the joint distribution will bein the payoff distribution, so the joint distribution will be in the payoff distribution, so the joint distribution will be Gaussian under the Central Limit Theorem only for very Gaussian under the Central Limit Theorem only for very large numbers of independent eventslarge numbers of independent events

Even a small degree of correlation calls the method into questionEven a small degree of correlation calls the method into question–– Even a small degree of correlation calls the method into questionEven a small degree of correlation calls the method into question

The Problem of Higher Order The Problem of Higher Order ggDependenceDependence•• Imagine you want to allocate money to two hedge fundsImagine you want to allocate money to two hedge funds•• Imagine you want to allocate money to two hedge funds Imagine you want to allocate money to two hedge funds

based on traditional MPTbased on traditional MPT•• The two portfolio managers happen to have offices in The two portfolio managers happen to have offices in

the same building and meet for coffee every morningthe same building and meet for coffee every morning–– During conversation, they flip a coin. If the coin comes up During conversation, they flip a coin. If the coin comes up

“heads” they hold the identical portfolio for that day. If it comes “heads” they hold the identical portfolio for that day. If it comes y p yy p yup “tails” they go long/short against each otherup “tails” they go long/short against each other

•• If the coin is fair, the time series of their portfolio If the coin is fair, the time series of their portfolio returns will be zeroreturns will be zeroreturns will be zeroreturns will be zero–– It varies daily from +1 to It varies daily from +1 to --1. averaging zero1. averaging zero

•• Being independent implies zero correlation, but zero Being independent implies zero correlation, but zero correlation does not imply independencecorrelation does not imply independence

We Go Blindly Forward with the We Go Blindly Forward with the yyGaussian CopulaGaussian Copula•• Investment banks love the Gaussian copula because now Investment banks love the Gaussian copula because now

you can get joint default probabilities over any number you can get joint default probabilities over any number of loans with the estimation of a single number theof loans with the estimation of a single number theof loans with the estimation of a single number, the of loans with the estimation of a single number, the default correlationdefault correlation–– The rating agencies such as Moody’s and S&P go along and rate The rating agencies such as Moody’s and S&P go along and rate

based on this methodbased on this methodbased on this methodbased on this method

•• The problem is where to get the default correlation The problem is where to get the default correlation assumption, since actual defaults were rare events, so assumption, since actual defaults were rare events, so statistical estimation from historical data is questionablestatistical estimation from historical data is questionable

•• Bankers used observable correlations from changes in Bankers used observable correlations from changes in spreads in the credit default swap marketspreads in the credit default swap marketspreads in the credit default swap market spreads in the credit default swap market

CDO/CDS Default CorrelationsCDO/CDS Default CorrelationsCDO/CDS Default CorrelationsCDO/CDS Default Correlations•• Assessing default correlations from credit swap curvesAssessing default correlations from credit swap curves•• Assessing default correlations from credit swap curves Assessing default correlations from credit swap curves

and CDO trading was horrendously faultyand CDO trading was horrendously faulty•• Once CDOs and CDOOnce CDOs and CDO2 2 could be written on “generic ABS” could be written on “generic ABS”

index results instead of specific pools, they were a form index results instead of specific pools, they were a form of legalized gamblingof legalized gambling

•• The volumes in the CDO market were many times theThe volumes in the CDO market were many times the•• The volumes in the CDO market were many times the The volumes in the CDO market were many times the volume of actual underlying loans against which to volume of actual underlying loans against which to hedge credit risk, creating severe pricing distortionshedge credit risk, creating severe pricing distortions

•• The CDO/CDS markets were dominated by a few large The CDO/CDS markets were dominated by a few large players such as AIG, further distorting economic pricing players such as AIG, further distorting economic pricing relationshipsrelationshipsrelationshipsrelationships

The Emperor Had No ClothesThe Emperor Had No ClothesThe Emperor Had No ClothesThe Emperor Had No Clothes•• In the 2002 documentation of their COSMOS fixed In the 2002 documentation of their COSMOS fixed

i i k d li i k d l BB t d th t th b dt d th t th b dincome risk model, income risk model, BarraBarra commented that the observed commented that the observed correlations between credit swap data and OAS spreads correlations between credit swap data and OAS spreads on actual on actual bonds appeared to be of the wrong signbonds appeared to be of the wrong signpp g gpp g g–– “This (data in Table 7) appears to be evidence against the idea “This (data in Table 7) appears to be evidence against the idea

that the level of the swap curve is the primary determinants of that the level of the swap curve is the primary determinants of yields of nonyields of non--government bonds. ……….. This is contrary to the government bonds. ……….. This is contrary to the yy g yg yexpectation that swap and bond spreads (to the government expectation that swap and bond spreads (to the government term structure) should be strongly positively correlated.”term structure) should be strongly positively correlated.”

•• Since 1999, Northfield’s EE model was structuredSince 1999, Northfield’s EE model was structuredSince 1999, Northfield s EE model was structured Since 1999, Northfield s EE model was structured without any use of credit swap informationwithout any use of credit swap information–– Relies entirely on individual OAS analysis on millions of issues Relies entirely on individual OAS analysis on millions of issues

over hundreds of fixed income categoriesover hundreds of fixed income categoriesover hundreds of fixed income categoriesover hundreds of fixed income categories

When Money Is At Stake Cheat!When Money Is At Stake Cheat!When Money Is At Stake, Cheat!When Money Is At Stake, Cheat!•• S&P has admitted to “notching” ratings on many S&P has admitted to “notching” ratings on many

securitized instrumentssecuritized instruments•• If an issuer wanted a rating done in a hurry, S&P would If an issuer wanted a rating done in a hurry, S&P would

simply assume a rating of one grade lower thansimply assume a rating of one grade lower thansimply assume a rating of one grade lower than simply assume a rating of one grade lower than whatever S&P or Moody’s had most recently rated the whatever S&P or Moody’s had most recently rated the debt of the same issuerdebt of the same issuer

•• To my mind, this is fraudTo my mind, this is fraud–– The expectations of financial market participants is that rating The expectations of financial market participants is that rating

agencies actually conduct some form of credit analysis before agencies actually conduct some form of credit analysis before age c es actua y co duct so e o o c ed t a a ys s be o eage c es actua y co duct so e o o c ed t a a ys s be o eissuing a ratingissuing a rating

–– At a cocktail party, a securitization lawyer for S&P rebuked me, At a cocktail party, a securitization lawyer for S&P rebuked me, saying they had done nothing illegal, as there are no actual saying they had done nothing illegal, as there are no actual y g y g g ,y g y g g ,requirements for any analysisrequirements for any analysis

Bury Your Head in the SandBury Your Head in the Sandyy•• At the October 2007 Northfield conference, presenter At the October 2007 Northfield conference, presenter

Jonathan Reiss showed that the prices for house priceJonathan Reiss showed that the prices for house priceJonathan Reiss showed that the prices for house price Jonathan Reiss showed that the prices for house price futures contracts for some US cities (e.g. Miami) were futures contracts for some US cities (e.g. Miami) were trading at a 30% discount to current pricestrading at a 30% discount to current prices–– How can lenders give out “no money down” mortgages when How can lenders give out “no money down” mortgages when

futures prices indicate expectations of dramatic price declines?futures prices indicate expectations of dramatic price declines?

•• Around the same time, BobAround the same time, Bob GlauberGlauber, interim chairman of, interim chairman ofAround the same time, Bob Around the same time, Bob GlauberGlauber, interim chairman of , interim chairman of Freddie Mac, asked their internal analysts for a “worst Freddie Mac, asked their internal analysts for a “worst case scenario” on their portfoliocase scenario” on their portfolio

The anal sts came back ith a scena io of US a e age ho singThe anal sts came back ith a scena io of US a e age ho sing–– The analysts came back with a scenario of US average housing The analysts came back with a scenario of US average housing prices not rising for three yearsprices not rising for three years

–– He actually had to order them to consider the possibility for price He actually had to order them to consider the possibility for price declines They got to only 4% a year for 3 years before givingdeclines They got to only 4% a year for 3 years before givingdeclines. They got to only 4% a year for 3 years before giving declines. They got to only 4% a year for 3 years before giving up in horror at the projected financial outcomesup in horror at the projected financial outcomes

Some Unpleasant OutcomesSome Unpleasant Outcomespp•• From the first quarter 2005 through the third quarter of From the first quarter 2005 through the third quarter of

2007 two thirds of CDOs S&P had rated were2007 two thirds of CDOs S&P had rated were2007, two thirds of CDOs S&P had rated were 2007, two thirds of CDOs S&P had rated were downgraded.downgraded.–– 44% of all CDOs were downgraded to “speculative” or “in 44% of all CDOs were downgraded to “speculative” or “in

default”default”

•• Over the same time, 17% of subOver the same time, 17% of sub--prime residential prime residential ,, ppmortgage securities were downgradedmortgage securities were downgraded–– 9.8% were downgraded to “speculative” or “in default”9.8% were downgraded to “speculative” or “in default”

•• $250 Billion in major bank write offs from January 1, $250 Billion in major bank write offs from January 1, 2007 through April 2008, another $500 Billion estimated 2007 through April 2008, another $500 Billion estimated f A il 2008 t d tf A il 2008 t d tfrom April 2008 to datefrom April 2008 to date

The Mark to Market ProblemThe Mark to Market ProblemThe Mark to Market ProblemThe Mark to Market Problem•• FASB 157 has recently been relaxed to reduce the “fire FASB 157 has recently been relaxed to reduce the “fire

sale” aspect of liquidations and write downs at major sale” aspect of liquidations and write downs at major institutionsinstitutions

•• What most people are forgetting is that most bankWhat most people are forgetting is that most bank•• What most people are forgetting is that most bank What most people are forgetting is that most bank assets are not subject to FASB 157assets are not subject to FASB 157

•• Whole loans that are not securitized are not currently Whole loans that are not securitized are not currently subject to mark to marketsubject to mark to market–– Subprime residential loansSubprime residential loans–– Commercial real state loansCommercial real state loansCommercial real state loansCommercial real state loans–– Current estimates are that 20% of American households have Current estimates are that 20% of American households have

negative home equity, with some cities such as Las Vegas over negative home equity, with some cities such as Las Vegas over 60%60%60%60%

A Quick FixA Quick Fix•• In 2004, Northfield introduced a real estate extension to In 2004, Northfield introduced a real estate extension to

our Everything, Everywhere modelour Everything, Everywhere modelWithi thi th d l d l th d l iWithi thi th d l d l th d l i•• Within this methodology, we model the underlying Within this methodology, we model the underlying economic activity in a given geographic area to estimate economic activity in a given geographic area to estimate default risk and correlationsdefault risk and correlations–– We measure economic activity as the fractions of total We measure economic activity as the fractions of total

household income that arises from the various sectors of the household income that arises from the various sectors of the economy (manufacturing, technology, energy, etc.)economy (manufacturing, technology, energy, etc.)

–– Risks to household incomes are modeled as lagged cumulative Risks to household incomes are modeled as lagged cumulative stock market sector returnsstock market sector returns

–– If the banking sector is doing poorly, this will weaken New YorkIf the banking sector is doing poorly, this will weaken New YorkIf the banking sector is doing poorly, this will weaken New York If the banking sector is doing poorly, this will weaken New York and London, but less so Seattle, or Venice. If oil prices are up, and London, but less so Seattle, or Venice. If oil prices are up, things are good for Houston, Norway and the OPEC countriesthings are good for Houston, Norway and the OPEC countries

•• Given appropriate information on underlying loans, the expectation Given appropriate information on underlying loans, the expectation G e app op ate o at o o u de y g oa s, t e e pectat oG e app op ate o at o o u de y g oa s, t e e pectat oof default correlation can be directly calculated in Merton fashionof default correlation can be directly calculated in Merton fashion

Summing UpSumming UpSumming UpSumming Up

•• The global financial crisis has many contributing factorsThe global financial crisis has many contributing factors

•• Many of these contributing factors were simply poorMany of these contributing factors were simply poor•• Many of these contributing factors were simply poor Many of these contributing factors were simply poor quantitative analysis in which mathematical convenience quantitative analysis in which mathematical convenience was allowed to take precedence over the conceptual was allowed to take precedence over the conceptual rigor of the modelsrigor of the models

•• The shortcomings in the analytical methods applied toThe shortcomings in the analytical methods applied to•• The shortcomings in the analytical methods applied to The shortcomings in the analytical methods applied to the RMBS, CDO and CDS markets were easily observed the RMBS, CDO and CDS markets were easily observed by those who were relatively free of conflicts of interestby those who were relatively free of conflicts of interest

ReferencesReferencesReferencesReferences•• DommermuthDommermuth, Michael. “Moody’s Bank Loan Ratings: Pricing , Michael. “Moody’s Bank Loan Ratings: Pricing

Implications and Approach” Moody’s Investors Services 1996Implications and Approach” Moody’s Investors Services 1996Implications and Approach , Moody s Investors Services, 1996.Implications and Approach , Moody s Investors Services, 1996.•• Woo, W.H. and T.K. Woo, W.H. and T.K. SiuSiu. “A Dynamic Binomial Expansion Technique . “A Dynamic Binomial Expansion Technique

for Credit Risk Management”, Chu for Credit Risk Management”, Chu HaiHai College (Hong Kong) Working College (Hong Kong) Working Paper 2005Paper 2005Paper, 2005. Paper, 2005.

•• Li, David X. "On Default Correlation: A Copula Function Approach," Li, David X. "On Default Correlation: A Copula Function Approach," Journal of Fixed Income, 2000, v9(4,Mar), 43Journal of Fixed Income, 2000, v9(4,Mar), 43--54.54.

•• C hC h Mi h l R b tMi h l R b t JJ d St t T b ll “Th S b id St t T b ll “Th S b i•• CrohyCrohy, Michel, Robert , Michel, Robert JarrowJarrow and Stuart Turnbull “The Subprime and Stuart Turnbull “The Subprime Credit Crisis of 2007”, Journal of Derivatives, 2008. Credit Crisis of 2007”, Journal of Derivatives, 2008.

•• diBartolomeodiBartolomeo, Dan. “A Review of Moody’s Methods Used to Assign , Dan. “A Review of Moody’s Methods Used to Assign C dit R ti t C ll t li d L Obli ti ” N thfi ldC dit R ti t C ll t li d L Obli ti ” N thfi ldCredit Ratings to Collateralized Loan Obligations”, Northfield Credit Ratings to Collateralized Loan Obligations”, Northfield Working Paper, 1999. Working Paper, 1999.

•• Salmon, Felix. “The Secret Formula that Destroyed Wall Street”, Salmon, Felix. “The Secret Formula that Destroyed Wall Street”, WIRED Ma ch 2009WIRED Ma ch 2009WIRED, March 2009. WIRED, March 2009.

ReferencesReferences•• CheyetteCheyette, Oren. “, Oren. “BarraBarra Global Credit Risk Modeling”, Global Credit Risk Modeling”, BarraBarra Research Research

Insights, 2002. Insights, 2002. •• diBartolomeodiBartolomeo, Dan. “Credit Risk Modeling at Northfield”, Northfield, Dan. “Credit Risk Modeling at Northfield”, NorthfielddiBartolomeodiBartolomeo, Dan. Credit Risk Modeling at Northfield , Northfield , Dan. Credit Risk Modeling at Northfield , Northfield

News, January 2009, News, January 2009, http://www.northinfo.com/documents/71.pdfhttp://www.northinfo.com/documents/71.pdf..•• Reiss, Jonathan. “The Potential Impact of Housing Derivatives”, Reiss, Jonathan. “The Potential Impact of Housing Derivatives”,

Northfield Conference Proceedings, 2007, Northfield Conference Proceedings, 2007, o t e d Co e e ce oceed gs, 00 ,o t e d Co e e ce oceed gs, 00 ,http://www.northinfo.com/documents/263.pdfhttp://www.northinfo.com/documents/263.pdf..

•• GlauberGlauber, Robert. Luncheon address to the Boston Economics Club, , Robert. Luncheon address to the Boston Economics Club, Federal Reserve Bank of Boston, May 13Federal Reserve Bank of Boston, May 13thth, 2009. , 2009. , y, y ,,

•• Financial Accounting Standards Board, Staff Position Paper 157Financial Accounting Standards Board, Staff Position Paper 157--4, 4, April 2009, April 2009, http://www.fasb.org/pdf/fsp_fas157http://www.fasb.org/pdf/fsp_fas157--4.pdf4.pdf..

•• Zillow ComZillow Com “Real Estate Market Reports” May 2009“Real Estate Market Reports” May 2009•• Zillow.ComZillow.Com. Real Estate Market Reports , May 2009, . Real Estate Market Reports , May 2009, http://www.zillow.com/reports/RealEstateMarketReports.htmhttp://www.zillow.com/reports/RealEstateMarketReports.htm

•• Baldwin, Ken, Emilian Baldwin, Ken, Emilian BelevBelev, Dan , Dan diBartolomeodiBartolomeo and Richard Gold “A and Richard Gold “A New Approach to Real Estate Risk” Proceedings of the SouthernNew Approach to Real Estate Risk” Proceedings of the SouthernNew Approach to Real Estate Risk , Proceedings of the Southern New Approach to Real Estate Risk , Proceedings of the Southern Finance Association Annual Meeting, November 2005. Finance Association Annual Meeting, November 2005.