Credit organizations, including Moody's, have been increasing measures aimed at diligently and continually evaluating the limitations of financial models and the assumptions and data behind them, especially with regard to structured securities. These steps can bolster the assessment of analytical uncertainties.
In structured finance, the key areas of uncertainty that may lead to unexpected performance variability are:
* the quality and sufficiency of historical data upon which the modeling assumptions are based,
* the level of transaction and model complexity,
* the quality of transaction governance and third party oversight, and
* differing or even competing interests among transaction parties, as well as the extent of legal and regulatory uncertainties.
Most credit organizations will be further formalizing both the processes and the tools with which they identify and assess the potential impact of uncertainties in data models, governance, and other key factors affecting the risk in a transaction.
One tool being used by some organizations - including Moody's - is a scorecard that provides a framework to evaluate the quality of the inputs and assumptions in the process, creating a relative ranking among sectors and also among transactions within a sector.
In Moody's case, we developed a scored approach in 2008 that is being used in connection with our ratings of new transactions, V Score, short for Assumption Variability Score. In addition to V Scores, we also are performing parameter sensitivity analyses for new transactions to provide a model-indicated calculation of the number of rating notches that a rated security may vary if certain input model parameters were to differ. We recognize that our practices must evolve along with changes in market dynamics, and we expect to continue developing and modifying our approach in step with market needs.
Our primary goal is to improve the quality of our analysis by identifying, quantifying and explicitly discussing in rating committees the key weaknesses in what we know or can verify in the transaction we are rating. In addition, by publishing the V Scores as well as the assessments for each subcomponent, we enhance analytical transparency by providing investors with additional information for making more informed credit decisions.
The chart on page 14 provides a sample of 20 of the approximately 60 different global sectors that we have assessed. Scores range between Low (1) analytic uncertainty and High (5) analytic uncertainty.
The strongest sector assessed (and hence with the least analytical uncertainty) was Japanese prime autos; the weakest sector was been Russian RMBS. As expected, U.S. RMBS and global derivatives were assessed on the weaker end due to significant levels of analytic uncertainty.
Of interest are the assessments of Low/Medium (2.0) for U.S. prime auto and Medium (2.75) for U.S. credit cards. Both of these sectors have among the longest performance histories of any asset securitized globally. However, for these asset classes, the current global economic environment presents moderately higher levels of analytic uncertainty. In addition, prior to the current period, neither sector had experienced a very stressful period for quite some time. As we progress through the current economic cycle and analyze the performance in the current economic environment, it is possible that these assessments, as well as certain others, may be deemed to have a lower level of analytic uncertainty. Basically, we will have a richer data set to incorporate into our analyses.
For our scorecard we categorized assumption quality and variability into four broad components. We then further divided each of these four broad components into three or four subcomponents (See appendix below).
The score for each of the four broad components is equal to the highest of their respective underlying subcomponent scores, with the overall composite score or V score being equal to a simple average of the four broad component scores. This particular scoring formula was intended to give due weight to any single weak link in the array of uncertainties.
The four broad components that comprise our V score are shown below, along with some commentary of our current assessments for a typical transaction in a couple of selected sectors.
1) Sector historical data adequacy and performance variability
Core to assessing analytic uncertainty is the quantity and sufficiency of historical performance data and, in particular, the degree to which historical performance is a useful indicator of future performance. In analyzing the quality of data, except for homogenous and simple assets, most market participants now agree that the gold standard is the receipt of loan-level historical data. Another consideration in evaluating historical data is the extent to which it covers a variety of economic and credit cycles that, including one or more that are extremely stressful. If the historical data does not cover a recent stressful economic cycle, its quality is diminished.
The following is a sample cross section of assessments for historical data adequacy:
U.S. credit cards: Medium (3)
U.S. prime autos: Low/Medium (2)
Japanese autos: Low (1)
Russian RMBS: High (5)
While U.S. credit cards and U.S. autos have long performance histories, the data does not cover a severe recession in the order of what we are experiencing currently. By contrast, Japan has weathered a protracted recession, which enhances the quality of the historical data available. Russia is an emerging economy with a relatively brief history.
2) Quality of transaction disclosure
Similar to historical performance data, most market participants now agree that the gold standard is the receipt of loan-level historical data both prior to closing as well as month by month. As our scorecard measures the potential variability around the inputs to our ratings, the V score disclosure assessments are based primarily on the level of data that is provided to us. We obviously believe that greater transparency for all market participants is beneficial because it allows investors to make more informed investment decisions. We also believe that greater scrutiny of data results in higher quality data.
The following is a sample cross section of assessments for quality of transaction disclosure:
U.S. prime autos: Low/Medium (2)
U.S. CMBS: Low (1)
U.K. prime RMBS: Low/Medium (2)
Structured finance CDOs: Medium/High (4)
In general, securitization reporting is reasonable. The level of disclosure of loan level data, however, can be improved. Therefore, few sectors received a Low assessment. As derivative transactions are second or third order transactions, the quality of disclosure is typically worse than first order transactions - while the data may be available from other sources, issuers frequently do not have access to the data.
3) Complexity
Obviously, the complexity of a transaction, as well as the complexity of the models and other analytical tools used to analyze the credit risk, adds to the analytic and assumption uncertainty.
The following is a sample cross section of the assessments for complexity:
Japanese autos:
Low (1)
Australian prime RMBS: Low/Medium (2)
U.S. large loan CMBS: Medium (3)
Corporate synthetic CDOs: Medium/High (4)
Japanese autos was one of the few sectors that was assessed as Low for complexity and market value sensitivity due to the simple transaction structures and the homogenous well understood credit risks of the underlying assets. U.S. large loan CMBS are slightly more complex, due to the moderately more complex transaction structures. In addition, while there is significant commercial real estate performance history, the credit evaluation requires a high level of specialized knowledge. The corporate synthetic CDO assessment is primarily driven by the use of complex Monte Carlo simulation models that are sensitive to the interaction between a number of key parameters.
4) Transaction Governance
Key issues in transaction governance revolve around the alignment of interests among transaction parties, as well as the legal and regulatory uncertainty that can affect specific asset classes or entire jurisdictions.
The following is a sample cross section of assessments for transaction governance:
Japanese autos: Low (1)
U.S. credit card: Low/Medium (2)
U.S. private student loans: Medium (3)
Russian RMBS: High (5)
The Japanese auto sector is an excellent example of a developed market with a long history of securitization, as well as an established and tested legal and regulatory framework. The U.S. credit card sector benefits from many of these attributes, but it is currently subject to regulatory and legislative changes that could affect performance. Issuers in the U.S. private student loan sector are experienced but typically unrated, and most transactions do not have a back-up servicer in place. Russia's assessment is driven by its brief securitization history combined with the level of uncertainty in its legal and regulatory framework.
Next Steps
The evaluation of analytical uncertainty is an ongoing exercise, one that demands constant reevaluation. As we complete the initial application of our scorecard to all structured finance sectors globally, we continue to ask many questions including:
* As V Scores are meant to measure the relative level of analytic uncertainty across asset classes and across transactions, what metrics and processes can we use to validate as well as calibrate our assessments?
* Are there ratings ceilings based on the V Score assessment? For example, can securities be rated 'Aaa' if they receive a weak or High V Score assessment?
* While analytic uncertainty has always been a part of our rating process, now that we have established a formalized scoring process, how should the V scores be more formally factored into loss expectations or credit enhancement levels?
(c) 2009 Asset Securitization Report and SourceMedia, Inc. All Rights Reserved.
http://www.structuredfinancenews.com http://www.sourcemedia.com/
