Appendix B — Letter Issued by the U.S. Department of the Treasury

January 7, 2008

Mr. Conrad W. Hewitt

Chief Accountant

Securities and Exchange Commission

200 F Street, NE

Washington, DC 20549

Dear Mr. Hewitt,

Thank you for your letter dated December 4, 2007 regarding the American Securitization Forum's Streamlined Foreclosure and Loss Avoidance Framework (ASF Framework). We look forward to your perspective regarding the consistency of the ASF Framework with Financial Accounting Standards Board Statement No. 140, Accounting for Transfers and Servicing of Financial Assets and Extinguishments of Liabilities. In your letter, you requested more data regarding the correlation between the pre-defined screening criteria as described under the ASF Framework and the notion of "reasonably foreseeable" default. In response to your query, the Treasury Department has prepared the attached information with data provided by the Federal Deposit Insurance Corporation and a large mortgage servicer.

We are pleased that mortgage investors and servicers worked through the ASF to develop this streamlined process for fast-tracking refinancings and loan modifications where doing so is in the interest of both homeowners and investors. We believe the ASF Framework is an important tool to prevent avoidable foreclosures. Unfortunately, there is no simple solution that will undo the housing excesses of the last few years. We are committed to avoiding preventable foreclosures whenever possible while ensuring the health of the mortgage market.

Thank you for all of your efforts. Please let me know if you have any questions regarding the attached information.

Sincerely,

Robert K. Steel

Under Secretary of the Treasury

On December 6, 2007, the American Securitization Forum (ASF) published a Streamlined Foreclosure and Loss Avoidance Framework (ASF Framework) to enable mortgage servicers to streamline their loss avoidance and loan modification practices. The ASF Framework applies to subprime, owner-occupied, two- and three-year adjustable-rate mortgages, and is meant to expedite consideration of these loans for refinancing or modification.

Under most pooling and service agreements, servicers have an obligation to implement all available loss-mitigation options, including loan modifications, to maximize cash flow to the investment trusts. Under current loan modification practices, servicers gather additional income and expense data from borrowers — effectively re-underwriting loans to determine if borrowers need a modification. While this process is effective in analyzing borrowers' financial capacity, it is a time consuming process that requires significant borrower contact. This burden will increase substantially over the next two years, due to the large number of resetting subprime mortgages and the expected increase in defaults.

Faced with this costly administrative burden, servicers, issuers and investors designed the ASF Framework to increase the efficiency and effectiveness of servicer loss-mitigation practices so they can analyze and process the increasing volume of subprime mortgage resets more quickly. Approximately 1.8 million owner-occupied, subprime two- and three-year adjustable-rate mortgages are expected to reset in 2008 and 2009.

The purpose of the ASF Framework is to streamline the procedures servicers use to identify borrowers who are candidates for refinancings or loan modifications. The parameters of the ASF Framework were designed to improve administrative efficiency while still maximizing cash flow by appropriately identifying the following: borrowers that can refinance into a sustainable mortgage; borrowers that should be modified into a more affordable mortgage; and borrowers that require in-depth, case-by-case analysis. Consistent with these goals, the ASF Framework was designed to fast-track into loan modifications only those borrowers who have demonstrated the ability to pay their starter rates, are unable to refinance, and are unable to afford their reset rates. The Federal Deposit Insurance Corporation (FDIC) and a major servicer both provided data that reflect whether the criteria the ASF Framework uses to identify borrowers for modifications are effective in preventing modifications where they are not needed (i.e., where the borrowers can afford the reset rates). Minimizing these false positives is consistent with maximizing the cash flow to investment trusts. Absent the ASF Framework, investors and servicers face a potential increase in false-negatives; i.e., loans entering foreclosure where modifications would have been a better outcome for investors.

The ASF Framework uses a number of screens to determine the appropriate loss-mitigation option for these subprime loans:

Test for ability to afford the starter rate: The ASF Framework first evaluates a borrower's ability to afford the starter rate, as demonstrated by a borrower not being more than 30 days delinquent, and having not been more than once 60 days delinquent in the last 12 months, both under the OTS method. Borrowers who have not demonstrated they can afford the starter rate will require in-depth, case-by-case analysis by their servicer to evaluate potential loss-mitigation options.

Test for capability to refinance: The ASF Framework next evaluates (first-lien) loan-to-value (LTV) to determine if a borrower has the potential to refinance. If a borrower has an LTV at origination greater than 97 percent, the ASF Framework assumes a refinancing is not possible. A borrower with an LTV below 97 percent may require additional information and analysis to determine if a refinancing is possible. If a borrower is deemed unable to refinance, the servicer may then consider the borrower for a fast-track modification.

Under the FHA Secure program, a borrower with an LTV up to 97 percent may be eligible for a refinancing. In the current market environment, outside of the FHA Secure program, most refinancing products require an LTV below 97 percent. Hence the ASF Framework established 97 percent LTV as the first test to evaluate a borrower's ability to refinance.

Tests for ability to afford reset: Once the servicer has determined the borrower is unable to refinance, the servicer then applies three tests to determine financial difficulty: 1) borrower's payment must increase by more than 10 percent, 2) borrower's current FICO must be less than 660, and 3) borrower's FICO must not have increased by more than 10 percent since origination. A borrower who fails to meet these tests may still qualify for a loan modification, but the servicer may need to gather additional information from the borrower to qualify the borrower for a modification.

The ASF Framework incorporated the FICO score of 660 as an initial indicator of financial stress for borrowers based both on servicers' default experience with borrowers and also on the banking regulators' report "Questions and Answers for Examiners Regarding the Interagency Expanded Guidance for Subprime Lending Programs Issued January 31, 2001," which identifies a credit score of 660 as one that generally indicates a higher default probability.

The ASF Framework applies to subprime, two- and three-year adjustable-rate mortgages, originated between January 2005 and July 2007 and facing an initial reset between January 2008 and July 2010. The data provided by FDIC and the major servicer help assess the baseline default and foreclosure occurrences for the subset of these loans that qualify for a modification under the ASF Framework. It is extremely difficult to estimate the counterfactual of what will happen to these loans if they do not receive the modification. This difficulty arises because historical data of similar loans are likely not representative of the underwriting, housing, and credit market conditions of the current vintages of loans eligible for the ASF Framework.

Evaluating a borrower's ability to afford the reset rate requires time to determine if a borrower ultimately remains current or defaults. While data from older loans where significant time has passed since reset provide sufficient time to determine if borrowers ultimately defaulted, those loans were originated under higher quality underwriting standards and experienced home price appreciation since origination. Such data would therefore likely underestimate the defaults of loans qualifying for the ASF Framework, because more recent vintages were originated with weaker underwriting standards and faced lower home price appreciation or even depreciation.

The worsening condition of more recent subprime mortgages is demonstrated by the significantly higher default percentage for the 2005 and 2006 vintages than for previous vintages. Even at one year before the rate reset, the number of foreclosure starts as a percentage of loans originated is much higher for recent vintages, moving from 2.1 percent for the 2004 vintage to 3.4 percent for the 2005 vintage to 9.2 percent for the 2006 vintage (i.e., foreclosure rates were approximately 1.6 and 4.4 times greater for the 2005 and 2006 vintages.) The more than four-fold increase in the foreclosure start rate one year before reset from the 2004 to 2006 vintage is likely driven by both deteriorating underwriting standards as well as declining housing prices. In fact, the cumulative foreclosure start rate for the 2006 vintage is higher than for the 2004 vintage, even though the former has yet to reset and the latter has already reset. Hence, data for the older vintage's likely significantly underestimate the ultimate defaults of the recent loans qualifying for the ASF Framework. Data from more recent vintages that were originated with lower quality underwriting and that faced price depreciation do not provide sufficient time post-reset to determine if a borrower ultimately remained current or defaulted.

It is also important to note that the current case-by-case system of evaluating loans for modification will also result in some false positives (i.e., modifying loans that would not otherwise default), especially given the increase in the administrative burden that will result from the large number of impending resets. The relevant measure would be the false positive rate for loans eligible for the ASF Framework's fast-track modification relative to the false positive rate under current practices. Unfortunately, such a comparison is not feasible.

Both the FDIC and a major subprime servicer provided data that reflect the baseline default and foreclosure rate for the population of loans expected to be eligible for the fast-track modification under the ASF Framework.

Both data sources attempt to approximate the ASF Framework's criteria for modification eligibility and then quantify the subsequent outcomes of these loans. Both data sources therefore examine owner-occupied, subprime two- and three-year adjustable-rate mortgages that are still active at the reset date. They further restrict the sample to include only those loans that had a FICO (at origination) of less than 660. The data only record FICO at origination, so cannot include the ASF Framework's condition that a borrower's FICO must not have increased by more than 10 percent since origination, making the data less precise at forecasting default than the actual Framework should be in practice. Also, the FDIC data (but not the private servicer's data) cannot measure whether the borrower's payment increase is more than 10 percent post-reset (note: typical rate increases for these loans is closer to 30 percent). Both of these limitations will lead to a more conservative assessment by understating the number of defaults and foreclosures of loans that qualify for a fast-track modification under the ASF Framework.

The two data sources take different approaches to limiting the sample to only those loans that are unable to refinance. The FDIC restricts the data to those loans with an LTV (at origination) above 97 percent, whereas the private servicer does not. However, because the fast-track modification can be considered by the servicer only if a borrower is unable to refinance, both data sets restrict the samples to those loans that did not subsequently refinance after reset.

The remaining loans that are active at first reset (and that subsequently did not refinance) provide the relevant population of loans for assessment. For these populations (by month of vintage), each data set then measures the number that subsequently default. Default is defined as 60 or more days delinquent, in Real Estate Owned ("REO") status, bankruptcy, or in foreclosure.

The FDIC relies on First American's LoanPerformance Mortgage Securities Database, which is a representative, loan-level sample of more than $2 trillion worth of active nonagency securitized mortgages. (See www.loanperformance.com for details about the data.) This database represents about 85 percent of all nonagency mortgage securities and approximately 76 percent of all mortgages in the United States.

The FDIC had data through September 2007. In order to assess default and foreclosures one year post-reset, the FDIC data counts the relevant loans that reset in September, 2006. There were 6,124 loans that reset during this month, of which 1,929 refinanced (while current) within the next year. Of the remaining 4,195 loans, 2,500 (60 percent) defaulted within a year of the first reset.

Not surprisingly, older vintages have a still higher default rate, as more time has elapsed for these at-risk loans to default. For example, for the relevant loans that reset in March 2006 (1.5 years of elapsed time post-reset), the default rate is 68 percent. For the relevant loans that reset in September 2005 (two years of elapsed time post-reset), the default rate is 76 percent. For the relevant loans that reset in March 2005 (2.5 years of elapsed time post-reset), the default rate is 81 percent.

The private servicer relies on proprietary data on the loans that it services. The data are through November 2007. These data only examine the one-year window post-reset for those loans that reset in November 2006. There were 1,512 two-year subprime adjustable-rate mortgages that reset during this month that were active at the time of reset, of which 351 refinanced (while fewer than 60 days delinquent) within the next year. Of the remaining 1,161 loans, 657 loans (57 percent) were at least 60 days past due during the year. Using a 30-day delinquency standard, of the original 1,512 loans that were active at reset, 152 refinanced (while fewer than 30 days delinquent) within the next year. Of the remaining 1,360 loans, 1,147 (84 percent) were at least 30 days past due during the year. With additional time, undoubtedly the default rate will continue to climb.

While default and foreclosure rates do typically vary across securitizations, the ASF Framework considers the payment history, LTV and FICO for each loan individually, on a case-by-case basis. Once that data has been considered in evaluating each loan, there is likely to be far less systematic variation from securitization to securitization and it is reasonable to conclude individual securitizations would perform in a similar manner to the data presented here.

As noted above, the FDIC data indicate that, of the loans that reset in March 2005, 81 percent subsequently defaulted over the next 2.5 years. The data did not measure vintages that reset before 2005, so one cannot measure the default rate over longer elapsed times. However, based on the monthly vintage data, one can compute a simple linear forecast of default rates moving forward.

The FDIC data track default rates every six months post reset, as well as at the latest recorded date (September 2007). For those vintages with more than one year recorded post-reset, the monthly increase in the default rate was 1.53 percentage points per month. For each monthly vintage, one can extrapolate on a linear basis past the one year post-reset rate using this monthly increase. Across monthly vintages, this leads to a three-year default rate of between 92 percent and 98 percent.

Given that the loans in the private servicer sample were originated in 2004 and 84 percent were at least 30 days past due and 57 percent were at least 60 days past due within one year post-reset, it is reasonable to expect far higher default rates one year post-reset for the loans qualifying for the ASF Framework, since these loans were originated from 2005 through 2007. As noted above, only 2.1 percent of the 2004 vintage had started foreclosure a year after origination, whereas 3.4 percent and 9.2 percent had started foreclosure a year after origination for the 2005 and 2006 vintages, respectively.

Based on the data of historical subprime loan performance post-reset and considering the poor performance of recent vintages that qualify for the ASF Framework (driven by poor underwriting standards and home price depreciation), our assessment is that servicers who apply the ASF Framework can reasonably conclude that they are modifying loans where default is reasonably foreseeable. Servicers can also reasonably conclude that, absent modification, loans that qualify for the ASF Framework would result in default.