Tuesday, July 10, 2007

S&P Changes Subprime Ratings Methodology

by Tanta on 7/10/2007 12:00:00 PM

S&P managed to surprise the markets this morning by publishing a whopper of a combined downgrade list/methdology change for subprime mortgage-backed securities this morning.

Many of the classes issued in late 2005 and much of 2006 now have sufficient seasoning to evidence delinquency, default, and loss trend lines that are indicative of weak future credit performance. The levels of loss continue to exceed historical precedents and our initial expectations.

We are also conducting a review of CDO ratings where the underlying portfolio contains any of the affected securities subject to these rating actions (see separate media release to be published today). . . .

On a macroeconomic level, we expect that the U.S. housing market, especially the subprime sector, will continue to decline before it improves, and home prices will continue to come under stress. Weakness in the property markets continues to exacerbate losses, with little prospect for improvement in the near term. Furthermore, we expect losses will continue to increase, as
borrowers experience rising loan payments due to the resetting terms of their adjustable-rate loans and principal amortization that occurs after the interest-only period ends for both adjustable-rate and fixed-rate loans.

Although property values have decreased slightly, additional declines are expected. David Wyss, Standard & Poor's chief economist, projects that property values will decline 8% on average between 2006 and 2008, and will bottom out in the first quarter of 2008. While our LEVELS model assumes property value declines of 22% for the 'BBB' and lower rating category stress environments (with higher property value declines for higher rating category stress environments), the continued decline in prices will apply additional stress to these transactions by
increasing losses on the sale of foreclosed properties, as well as removing or reducing the borrowers' ability to refinance or sell their homes to meet debt obligations.

As lenders have tightened underwriting guidelines, fewer refinance options may be available to these borrowers, especially if their loan-to-value (LTV) and combined LTV (CLTV) ratios have risen in the wake of declining home prices. . . .

Data quality is fundamental to our rating analysis. The loan performance associated with the data to date has been anomalous in a way that calls into question the accuracy of some of the initial data provided to us regarding the loan and borrower characteristics. A discriminate analysis was performed to identify the characteristics associated with the group of transactions
performing within initial expectations and those performing below initial expectations. The following characteristics associated with each group were analyzed: LTV, CLTV, FICO, debt-to-income (DTI), weighted-average coupon (WAC), margin, payment cap, rate adjustment frequency, periodic rate cap on first adjustment, periodic rate cap subsequent to first adjustment, lifetime max rate, term, and issuer. Our results show no statistically significant differentiation between the two groups of transactions on any of the above characteristics. . . .

As performance continues to deteriorate, we have increased the severity of the surveillance assumptions we use to evaluate the ongoing creditworthiness for this group of transactions. The level of severity was increased to 40% from 33% to reflect the average severity that subprime servicers are currently experiencing . . .

In addition, we have modified our approach to reviewing the ratings on senior classes in a transaction in which subordinate classes have been downgraded. Historically, our practice has been to maintain a rating on any class that has passed our stress assumptions and has had at least the same level of outstanding credit enhancement as it had at issuance. Going forward, there
will be a higher degree of correlation between the rating actions on classes located sequentially in the capital structure. A class will have to demonstrate a higher level of relative protection to maintain its rating when the class immediately subordinate to it is being downgraded. . . .

For transactions that close on or after July 10, 2007, We will incorporate several changes to our ratings methodology that will result in greater levels of credit protection for rated transactions. Our cash flow methodology assumptions will include a simultaneous combination of faster voluntary and involuntary (default) prepayments that will result in less credit to excess
spread. Furthermore, our default expectation for 2/28 hybrid ARM loans will increase by approximately 21%. We are in the process of updating our LEVELS and SPIRE models. A separate article will be released in the next few days describing the revisions to our ratings methodology, and will provide the estimated timing for release of the updated models. . . .

Given the level of loosened underwriting at the time of loan origination, misrepresentation, and speculative borrower behavior reported for the 2006 vintage, we will be increasing our review of the capabilities of lenders to minimize the potential and incidence of misrepresentation in their loan production. A lender's fraud-detection capabilities will be a key area of focus for us. The review will consist of a detailed examination of: (a) the overall capabilities and experience of the executive and operational management team; (b) the production channels and broker approval process; (c) underwriting guidelines and the credit process; (d) quality control and internal audits; (e) the use of third-party due diligence firms, if applicable; and (f)secondary marketing. A new addition to this review process will be a fraud-management questionnaire focusing on an originator's tools, processes, and systems for control with respect to mitigating the potential for misrepresentation.

Oh, my, no. A fraud management questionnaire? What a crackdown!

More later . . .

Note: the entire document is available in pdf on standardandpoors.com if you are a registered user.