LO 53.4: Explain major defects in model assumptions that led to the

LO 53.4: Explain major defects in model assumptions that led to the underestimation of systematic risk for residential mortgage backed securities (RMBS) during the 20072009 financial downturn.
The subprime RMBS valuation and risk models have been widely employed by credit rating agencies to assign bond ratings, by traders and investors in bond valuations, and by issuers in structuring RMBS. During the 2007-2009 financial downturn, two major defects in model assumptions became apparent: 1. Assumption o f future house price appreciation. The RMBS risk model generally assumed that future house prices would rise, or at least not fall, based on relatively few historical data points. When house prices actually did drop beginning in 2007, this incorrect assumption led to a significant underestimation of the potential default rates and systematic risk in RMBS because the credit quality of the loans was dependent on borrowers ability to refinance without additional equity.
2. Assumption o f low correlations. The RMBS model assumed low correlations among
regional housing markets, implying that loan pools from different geographical regions were well diversified. When house prices declined, correlations increased and loan defaults were much higher than previously expected under the model stress scenarios. These two model errors led to a significant underestimation of systematic risk in subprime RMBS returns. When mortgage default rates began to increase, rating agencies were required to downgrade most issues, and by the end of 2009, approximately 45% of the initially AAA-rated U.S. RMBS had been downgraded. The downgrades of RMBS from their AAA-equivalent ratings shocked markets and exposed the degree to which systemic risk had been underestimated and mispriced.
There have been several explanations proposed for the inaccuracy of the rating models. First, the compensation of rating agencies by bond issuers led to a potential conflict of interest scenario that resulted in lower ratings standards. Second, an increase in demand for higher rated bonds with a modestly higher yield resulted in searching for yield. Finally, mapping problems led to misleading risk measurement results, as highly rated securitized products were frequently mapped to highly rated corporate bond spread indices. This resulted in incorrect VaR estimates, as incorrect mappings indicated it would be unlikely that bonds would decline significantly in value. In reality, the most highly rated RMBS lost a significant portion of their value, declining close to 70% during the subprime crisis, while lower investment-grade RMBS lost virtually all of their value.
2018 Kaplan, Inc.
Page 211
Topic 53 Cross Reference to GARP Assigned Reading – Malz, Chapter 11
K e y C o n c e p t s
LO 53.1 Model risk is the risk of incorrect trading or risk management decisions due to errors in models and model applications, which can lead to trading losses and potential legal, reputational, accounting, liquidity, and regulatory risk. Errors can be introduced into models through programming bugs, securities valuations or hedging, VaR estimates, and position mappings.
LO 53.2 Firms use software to compute the risk measures from the data collected using specific formulas, which can be performed in a variety of ways and lead to potential issues. Variability in risk measures, including lack of uniformity in the use of confidence intervals and time horizons, can lead to variability in VaR estimates. Other factors can also cause variability, including length of time series, ways of estimating moments, mapping techniques, decay factors, and number of simulations.
Mapping refers to the assignment of risk factors to positions, and mapping choices can considerably impact VaR results.
Cash flow mapping results in greater accuracy of estimates. Duration-convexity mapping requires fewer risk factors and less complex computations, which reduces costs, data errors, and model risks. Locating data that addresses specific risk factors may also be difficult.
Liquidity risk arises from large divergences in model and market prices that are difficult to capture with market data, and as a result, VaR estimates based on replicating portfolios can understate risk and create liquidity risk.
Basis risk is the risk that a hedge does not provide the required or expected protection. Basis risk arises when a position or its hedge is mapped to the same set of risk factors.
LO 53.3 Volatility in credit markets in the spring of 2005 fueled by company bankruptcies and losses led to defaults in the IG3 and IG4 index series of the CDX.NA.IG index, causing large selloffs that resulted in widening spreads. This also resulted in modeling errors from both misinterpretation and incorrect application of models, which led to trade losses.
LO 53.4 Two significant model errors in the RMBS valuation and risk models led to a significant underestimation of systematic risk in subprime RMBS returns during 2007-2009. First, the RMBS risk model assumed future house prices to rise or at least stay flat. The eventual decline in house prices starting in 2007 led to a significant underestimation of the potential default rates and systematic risk in RMBS. Second, the RMBS model assumed low correlations among regional housing markets. When house prices declined, correlations and loan defaults increased.
Page 212
2018 Kaplan, Inc.
Topic 53 Cross Reference to GARP Assigned Reading – Malz, Chapter 11
C o n c e p t C h e c k e r s
1.
2.
3.
4.
Due to a recently discovered error in its valuation model, Samuelson, Inc. had previously recorded certain trades as profitable even though the positions were unprofitable. The risk that best characterizes this error is: A. operational risk. B. liquidity risk. C. market risk. D. hedge risk.
Duane Danning is a junior risk analyst at a large risk management firm. He has been asked to assess the firms risk modeling practices and evaluate potential ways in which errors could be introduced into models. In his analysis, Danning indicates that errors can be introduced into models through programming bugs and errors in VaR estimates but rarely through incorrect position mappings. Dannings analysis is most accurate with regard to: A. only programming bugs and incorrect position mappings. B. only programming bugs and errors in VaR estimates. C. only errors in VaR estimates. D. only incorrect position mappings.
An advantage of duration mapping over cash flow mapping is that duration mapping: A. B. uses cash flows that are mapped to specific fixed income securities without the
is more accurate than cash flow mapping, thus reducing mapping errors.
use of approximations.
C. uses more complex computations, thus reducing data errors and model risk. D. uses fewer risk factors, thus reducing data errors and model risk.
A common trade during 2004 and 2005 was to sell protection on the equity tranche and buy protection of the mezzanine tranche of the CDX.NA.IG index. Which of the following statements regarding this trade is least accurate? A. The trade was set up to be default-risk neutral at initiation. B. The trade was short credit spread risk on the equity tranche and long credit
spread risk on the mezzanine tranche.
C. The main motivation for the trade was to achieve a positively convex payoff
profile.
D. The trade was designed to benefit from credit spread volatilities.
2018 Kaplan, Inc.
Page 213
Topic 53 Cross Reference to GARP Assigned Reading – Malz, Chapter 11
5.
Which of the following two model errors in the RMBS valuation and risk models are considered to have contributed the most to a significant underestimation of systematic risk in subprime RMBS returns during 2007-2009? A. The assumption of future house price appreciation and the assumption of high
correlations among regional housing markets.
B. The assumption of future house price declines and the assumption of high
correlations among regional housing markets.
C. The assumption of future house price appreciation and the assumption of low
correlations among regional housing markets.
D. The assumption of future house price declines and the assumption of low
correlations among regional housing markets.
Page 214
2018 Kaplan, Inc.
Topic 53 Cross Reference to GARP Assigned Reading – Malz, Chapter 11
C o n c e p t C h e c k e r A n s w e r s
1. A Recording trades as profitable that have, in fact, lost money is an example of operational risk.
2. B Dannings analysis is most accurate with regard to only programming bugs and errors in VaR
estimates. Incorrect position mappings can also lead to material errors in risk models.
3. D Duration mapping (or duration-convexity mapping) requires the use of fewer risk factors
and less complex computations, thus reducing costs, data errors, and model risks. Cash flow mapping results in greater accuracy of estimates, however, because cash flows are mapped to specific fixed income securities without the use of approximations.
4. B The trade was long credit and credit spread risk on the equity tranche and short credit and
credit spread risk on the mezzanine tranche. The other statements are accurate.
5. C The two model errors considered to have contributed the most to a significant
underestimation of systematic risk were (1) the assumption of future house price appreciation, and (2) the assumption of low correlations among regional housing markets.
2018 Kaplan, Inc.
Page 215
The following is a review of the Operational and Integrated Risk Management principles designed to address the learning objectives set forth by GARP. This topic is also covered in:
Li q u i d i t y a n d Le v e r a g e
Topic 54
E x a m F o c u s
This topic analyzes the effects of liquidity and leverage on firm risk. For the exam, understand the distinction between transactions liquidity and funding liquidity, and the role banks play in providing liquidity. Also, be able to calculate a firms leverage ratio and the leverage effect, and know how to construct the economic balance sheet given trades such as buying stock on margin, selling stock short, and taking positions in derivatives. Finally, be able to explain tightness, depth, and resiliency as they relate to liquidity risk.
S o u r c e s o f L i q u i d i t y R i s k