Request For Comment: Methodology And Assumptions For Rating U.S. RMBS Issued 2009 And Later

Transcription

1 Criteria Structured Finance Request for Comment: Request For Comment: Methodology And Assumptions For Rating U.S. RMBS Issued 2009 And Later Analytical Contacts: Farooq Omer, CFA, New York (1) ; Stephen Buteau, New York (212) ; Jeremy Schneider, New York (1) ; Kapil Jain, New York (1) ; Criteria Contacts: Felix E Herrera, CFA, New York (1) ; felix.herrera@spglobal.com Ted J Burbage, New York (1) ; ted.burbage@spglobal.com Vanessa Purwin, New York ; vanessa.purwin@spglobal.com Table Of Contents SCOPE OF THE PROPOSAL SUMMARY OF THE PROPOSAL SPECIFIC QUESTIONS FOR WHICH WE ARE SEEKING A RESPONSE IMPACT ON OUTSTANDING RATINGS RESPONSE DEADLINE PROPOSED METHODOLOGY AND ASSUMPTIONS Criteria Framework Credit Quality Of The Securitized Assets Loss Model Foreclosure Frequency Factors APRIL 24,

2 Table Of Contents (cont.) Loss Model Loss Severity Assumptions Pool-Level Adjustments Payment Structure And Cash Flow Mechanics LEGAL AND REGULATORY RISK Anti-Predatory Lending (APL) Surveillance APPENDIXES Appendix I: Changes From Our Current Criteria Appendix II: Loss Severity Adjustments For ATR And QM Standards Appendix III: Pool-Level Adjustments Appendix IV: Third-Party Due Diligence Appendix V: R&Ws Appendix VI: Reference Guide To Industry Terminology Appendix VII: Criteria Articles To Be Fully Or Partially Superseded RELATED CRITERIA AND RESEARCH APRIL 24,

3 Criteria Structured Finance Request for Comment: Request For Comment: Methodology And Assumptions For Rating U.S. RMBS Issued 2009 And Later 1. S&P Global Ratings is requesting comments on proposed revisions to its criteria for rating U.S. residential mortgage-backed securities (RMBS) issued in 2009 and later. This article should be read in conjunction with "Assumptions Supplement For Request For Comment: Methodology And Assumptions For Rating U.S. RMBS Issued 2009 And Later," published April 24, 2017, and "Request For Comment: U.S. Residential Mortgage Operational Assessment Ranking Criteria," published April 24, This article is related to our criteria article "Principles Of Credit Ratings," published on Feb. 16, The proposed revisions to our methodologies and assumptions are intended to better reflect the evolution of the U.S. residential mortgage market over the past several years. The changes also simplify and consolidate numerous criteria articles for improved use and transparency. 3. If we implement the proposed revisions, the new criteria would fully or partially, as applicable, supersede the articles listed in Appendix VII. SCOPE OF THE PROPOSAL 4. These proposed criteria, if adopted, apply to all new and existing ratings on U.S. RMBS and other debt obligations issued 2009 and later directly collateralized or referenced by prime or nonprime residential mortgage loans secured by first or second liens. SUMMARY OF THE PROPOSAL 5. S&P Global Ratings is maintaining the general framework for rating U.S. RMBS, including: The defined archetypal pool against which other mortgage pools are compared; The use of our Loan Evaluation and Estimate Of Loss System (LEVELS) model to assess mortgage risk based on loan attributes and borrower characteristics relative to the archetype; Our consideration of qualitative factors such as our market outlook, mortgage originator/aggregator quality, third-party due diligence, and representations and warranties (R&W) frameworks. 6. However, we are proposing changes to our methodologies and assumptions to enhance our analysis. Specifically, the proposed criteria: Revise and add certain foreclosure frequency adjustment factors to reflect additional performance data through June 2015 on roughly 8 million mortgages originated from 1998 through 2007 and changes in loan products, among other considerations. For example, we recalibrated our foreclosure frequency adjustment factors for the majority of APRIL 24,

4 combined loan-to-value ratio (CLTV)-FICO combinations based on observed defaults for loans through the 2008 financial crisis, including those for the vintage originations in highly distressed metropolitan statistical areas (MSAs). We also introduced a foreclosure frequency adjustment factor for the number of borrowers included for a loan based on our analysis of the performance history of loans with multiple borrowers compared to a single borrower. Update our approach to evaluating loss severity. The proposed criteria introduce an over/under valuation framework to adjust our projected market value declines (MVDs) for housing market conditions based on long-term price-to-income ratios. This should enhance our estimation of a loan's calculated loss severity by adjusting for any overvaluation or undervaluation embedded in the property value used in our analysis. The over/under valuation framework would replace the current housing volatility index. In addition, we are updating our fixed MVD assumptions and bifurcating them into forced-sale discount (FSD) and fixed MVD components. The proposal also introduces loss severity floors that are applied at the loan level and vary by rating category. Update and clarify certain assumptions and methodologies for seasoned and reperforming mortgage collateral. Seasoned (i.e., typically loans that are aged 24 months or more) and reperforming (i.e., typically loans that are presently less than 90 days delinquent and have been 90 or more days delinquent or modified in the past 24 months, or loans for which the borrower is in bankruptcy) loans have become a larger component of pools backing newly issued U.S. RMBS, but may have dated borrower, loan, or property information. How we intend to assess borrower creditworthiness, loan performance, and property value have been clarified and refined for these types of loans. Simplify our pool-level geographic concentration adjustment factor by solely using the Herfindahl-Hirschman Index. Our current criteria compare two independent methodologies: the Herfindahl-Hirschman Index and a formula that adjusts credit enhancement (CE) for concentrations above specified thresholds in specific core-based statistical areas (CBSAs; the CBSA excess method). In nearly all RMBS transactions rated 2009 and later, the Herfindahl-Hirschman Index yielded an operative adjustment factor, and we therefore believe the CBSA excess method is redundant. Revise the assumptions used in our cash flow analysis. In particular, we are proposing to extend the length of both the standard and back-loaded default timing curves to better align with observed data as well as reduce the number of years for which we stress extraordinary expenses. We are also proposing to introduce a smooth default timing curve into our analysis. In addition, we are proposing to revise the application of our voluntary prepayment stresses so that they are additive to the amount of defaults in a given month. Stresses for delinquencies, servicer stop advances, and weighted average coupon (WAC) deterioration have also been updated. Update our approach to evaluating qualitative factors to reflect the evolution of mortgage origination, due diligence, and R&W regulatory frameworks. For example, we are proposing to apply a positive adjustment factor down to 0.95x loss coverage while maintaining a negative adjustment factor of up to 1.15x or more based on a transaction's R&W framework, taking into account the extent and findings of any third-party due diligence as well as the use of a transaction manager, if applicable. We are also proposing to increase loss severities for loans with Truth in Lending Act (TILA)-Real Estate Settlement Procedures Act (RESPA) Integrated Disclosure rule (TRID) exceptions that remain in securitized pools. In addition, we are introducing additional due diligence considerations for seasoned and reperforming loans to enhance our assessment of their risks. Simplify our approach to evaluating the risks associated with anti-predatory lending (APL) loans. We are adopting a more holistic assessment of the incremental risk to loss coverage posed by APL loans. Introduce our assumptions for the analysis of second liens, both closed-end seconds and home equity lines of credit (HELOCs), based on our historical performance observations. We are proposing to apply a 1.3x foreclosure frequency adjustment factor and 100% loss severity to second-lien mortgage loans. APRIL 24,

5 SPECIFIC QUESTIONS FOR WHICH WE ARE SEEKING A RESPONSE 7. We are seeking market feedback on our proposed methodology and assumptions, as well as on the specific points listed below: What is your view of our proposed updates to the foreclosure frequency adjustment factors for borrower characteristics and loan attributes (e.g., CLTV-FICO, loan purpose, loan type, property type, seasoning, etc.)? In particular, our proposed: Changes to the CLTV-FICO adjustment factors ( 43-55); Introduction of a credit to foreclosure frequency for loans made to two or more borrowers ( 56); Adjustment factor to foreclosure frequency for loans that experienced a delinquency in the prior 24-month period ( 71); Treatment of second liens ( 72); and Treatment of loans that have been modified or made to borrowers with prior credit events ( 105). What is your view on the proposed changes to our loss severity methodology? In particular, our proposal to incorporate a property value adjustment based on our assessment of over/undervaluation to determine a loan's loss severity as well as introduce loss severity floors ( 73-84)? What is your view of our proposal to potentially apply stop advance stresses to certain structures and collateral that we believe are more exposed to liquidity risk ( )? What is your view of our proposal for third-party due diligence considerations for revolving pools, seasoned and reperforming loans, and loans with TRID exceptions ( and , and Appendix IV)? What is your view of our proposal to allow for a pool's loss coverage level to potentially be reduced based on the robustness of a transaction's R&W framework, in conjunction with the findings of any third-party due diligence and the use of a transaction manager (or deal agent), if applicable ( Appendix V)? Are there any other factors you believe should be considered in the proposed criteria? IMPACT ON OUTSTANDING RATINGS 8. The total outstanding portfolio for which the proposed criteria would apply includes approximately 1,500 ratings from 70 transactions, with prime ratings accounting for roughly 80%-90% of the total. We analyzed the potential impact of the proposed criteria by retroactively applying the proposal to a representative sample of transactions rated from early 2012 through 2016 (roughly 25% of in-scope transactions). The estimated impact results of this analysis do not consider the effect of subsequent transaction performance on credit quality or credit enhancement levels. 9. Specifically, we compared the ratings that were initially assigned using our current criteria to the rating that would otherwise have been assigned after applying the proposed criteria to each sampled transaction, as of its original closing date. When doing this comparative analysis, we split the sample into two segments: "prime" and "seasoned." The prime segment consists of prime jumbo and credit risk transfer transactions. The seasoned segment includes transactions of varying credit quality with collateral that was typically seasoned when the transaction was issued. The overall observations showed that transactions with stronger collateral typically showed comparatively higher ratings relative to the current criteria, while lower comparative ratings were evident as collateral quality became weaker. 10. For the prime segment, our analysis showed that ratings on the senior classes (roughly 90% of the prime segment) APRIL 24,

6 would generally have been the same, while ratings for the junior classes (roughly 10% of the prime segment) may have been, on average, three notches higher. However, such rating movements may be affected by the considerations noted in 12 and 13 below. 11. For the seasoned segment, the results were more dispersed across the capital structure given the collateral's varied credit quality. In fact, more cases showed that ratings assigned would have been lower (roughly 80% of the seasoned segment), while roughly 20% would have been the same. In cases where the ratings would have differed from those assigned under the current criteria, the overall differential was on average four to five notches. Lower ratings under the proposal were primarily driven by higher foreclosure frequency adjustment factors for loans with lower FICOs (at a given CLTV), the change in the CLTV used to determine foreclosure frequency for seasoned loans, and updated WAC deterioration stresses. However, such rating movements may be affected by the considerations noted in 12 and 13 below. 12. Notwithstanding the comparison done as of the issuance date for the sampled transactions, we also assessed what the impact on the ratings may be, taking into account historical transaction performance, the change in the level of available credit enhancement since issuance, and our expectation of performance going forward, based on the proposed criteria. The results of this assessment for the prime segment showed that senior classes would typically maintain their ratings while junior classes generally showed either no rating movement or upward rating movements of up to approximately six notches. For the seasoned segment, roughly half of the ratings in our sample generally showed no rating movements, while roughly half showed potential downward movements of approximately two notches on average (maximum movements were roughly four to five notches). A few classes in the sample set from the seasoned segment showed potential upward movements of roughly two notches on average. 13. The above comparisons are intended to serve as a broad, directional guide to the possible ratings impact if the proposed criteria are adopted. Ultimately, actual ratings impact may vary, depending on the specifics and performance of the asset pool and structural features of a particular transaction. For more detail regarding specific changes to certain elements of the proposed criteria, please see Appendix I. RESPONSE DEADLINE 14. We encourage interested market participants to submit their written comments on the proposed criteria by June 16, 2017, to where participants must choose from the list of available Requests for Comment links to launch the upload process (you may need to log in or register first). We will review and take such comments into consideration before publishing our definitive criteria once the comment period is over. S&P Global Ratings, in concurrence with regulatory standards, will receive and post comments made during the comment period to Comments may also been sent to CriteriaComments@spglobal.com should participants encounter technical difficulties. All comments must be published but those providing comments may choose to have their remarks published anonymously or they may identify themselves. Generally, we publish comments in their entirety, except when the full text, in our view, would be unsuitable for reasons of tone or substance. APRIL 24,

7 PROPOSED METHODOLOGY AND ASSUMPTIONS Criteria Framework 15. The framework for our analysis of U.S. RMBS considers the risks associated with the following five key areas (see chart 1): Credit quality of the securitized assets; Payment structure and cash flow mechanics; Operational and administrative risk; Counterparty risk; and Legal and regulatory risk. APRIL 24,

8 Credit quality of the securitized assets 16. To assess a mortgage pool's loss coverage level, we perform a loan-level analysis using our LEVELS model. Our loss coverage for each loan at each rating level consists of two components: foreclosure frequency and loss severity. Foreclosure frequency represents the probability of a loan to enter foreclosure (default). Loss severity refers to the loss on foreclosure (i.e., the amount by which a loan balance, foreclosure costs, and applicable interest advances exceeds the property sale proceeds). 17. Foreclosure frequency estimates for a mortgage loan pool backing a U.S. RMBS transaction are based on a comparison of the characteristics of each loan in that pool with those of loans in the archetypal U.S. pool. Adjustments to archetype foreclosure frequency are made to reflect any variances. Foreclosure frequency estimates at the 'AA+' to 'B+' rating levels are derived from an interpolation of the foreclosure frequency values at the 'AAA' and 'B' levels. Foreclosure frequency estimates at the 'B-' level are equal to 80% of the foreclosure frequency value at the 'B' level. 18. Loss severity estimates for each loan in a pool are based on the projected proceeds derived by applying our over/under valuation and MVD (including FSDs) adjustments and include relevant liquidation costs and servicer advances, where applicable. MVD (including FSDs) and over/under valuation adjustments assumed in our analysis vary by rating categories, while foreclosure timelines vary by rating level, except for 'B-', which is equal to the 'B' assumption. The loan-level loss coverage at each rating level is equal to the product of the loan's associated foreclosure frequency and loss severity estimates. 19. Loss coverage, foreclosure frequency, and loss severity at the pool level are calculated as follows: Pool-level loss coverage is equal to the weighted average loan-level loss coverage (by current balance); Pool-level foreclosure frequency is equal to the weighted average loan-level foreclosure frequency (by current balance); and Pool-level loss severity is equal to the ratio of pool-level loss coverage to pool-level foreclosure frequency. 20. A pool's loss coverage levels may be further adjusted to address the risk associated with certain pool-level characteristics such as geographic concentration and/or small loan count. In addition, loss coverage levels may be adjusted to consider qualitative factors including origination quality, third-party due diligence findings, and/or the robustness of the applicable R&W framework, among other things. Payment structure and cash flow mechanics 21. We generally perform a cash flow analysis to determine if a transaction has sufficient credit and liquidity enhancement to pay timely interest and principal by legal final maturity. 22. Both the amount and timing of cash flows are, in our view, important considerations in the rating analysis. The amount and timing of losses (or delinquencies, defaults, recoveries, or liquidations) and prepayments may positively or negatively affect the transaction's ability to meet its payment obligations on time. Pool characteristics (e.g., seasoning and payment status) are generally considered when determining the loss timing to be applied in the cash flow analysis for a rating scenario. Where relevant, we may apply additional cash flow stresses to account for other transaction-specific considerations such as certain legal, operational, and counterparty risks that are not mitigated by the transaction structure. APRIL 24,

9 Operational and administrative risks 23. The analysis of operational and administrative risks focuses on key transaction parties to determine whether they are capable of managing their duties related to a securitization over its life. Key transaction parties may include a transaction's servicer, the trustee, the paying agent, and any other transaction party. The analysis generally considers the possibility that a servicer may become unable or unwilling to perform its duties during the transaction's life. The analysis may consider both the potential for hiring a substitute or successor servicer and any arrangements that provide for a designated backup servicer. That portion of the analysis would typically consider the servicing fee's sufficiency to attract a substitute, its seniority in the payment priorities, and the availability of substitute servicers. To perform this analysis, we apply our "Global Framework For Assessing Operational Risk In Structured Finance Transactions," published Oct. 9, 2014, and "Standard & Poor's Revises Criteria Methodology For Servicer Risk Assessment," published May 28, The strength of mortgage originator and/or aggregator platforms is assessed based on quantitative (historical loan performance) and qualitative factors. We also review the third-party due diligence results of the loans to assess the data quality. In addition, the R&W framework analysis helps differentiate across transactions based on the quality of the R&Ws, the financial condition of the R&W provider, and the remedies and enforcement mechanism upon an R&W breach. To perform this analysis, we would apply the concepts outlined in these proposed criteria as well as "Request For Comment: U.S. Residential Mortgage Operational Assessment Ranking Criteria," published April 24, Counterparty risk 25. The analysis of counterparty risk focuses on third-party obligations to either hold assets (including cash) or make financial payments that may affect the rated securities' creditworthiness. Examples of counterparty risks include exposure to institutions that maintain key accounts and exposure to the providers of derivative contracts such as interest rate and currency swaps. The counterparty risk analysis considers both the type of dependency and the rating on the counterparty for each counterparty relationship in a transaction. To perform this analysis, we apply our "Counterparty Risk Framework Methodology And Assumptions," published June 25, Legal and regulatory risks 26. Our assessment of legal and regulatory risks focuses primarily on the degree to which a securitization structure isolates the securitized assets from the bankruptcy or insolvency risk of entities that participate in the transaction. Typically, our analysis focuses on the entity or entities that originated and/or owned the assets before the securitization, although the creditworthiness of other entities also may be relevant. To perform this analysis, we apply the concepts outlined in the applicable articles listed in Related Criteria and Research. 27. Further, our criteria take into account possible risks that may arise from certain applicable laws and regulations such as the Ability-To-Repay (ATR) rule, TRID, and APL laws. Other considerations 28. We may apply additional quantitative and/or qualitative analysis in certain limited circumstances where a particular transaction has factors or unique features that are meaningful to the credit analysis. APRIL 24,

10 Credit Quality Of The Securitized Assets Archetypal pool 29. We perform a loan-level analysis to assess a mortgage pool's credit quality based on a comparison of its borrower attributes and loan characteristics to those of our archetypal pool. Pool foreclosure frequency levels are derived from archetypal projected defaults adjusted for relevant variances from the archetype as described further below. 30. We define the archetypal pool to include: At least 250 loans; Loans secured by properties that are geographically well-diversified; A FICO score of 725, an LTV of 75%, and a back-end debt-to-income (DTI) ratio of 36%; Loans that are current at the time of transaction issuance; Newly originated loans (generally originated within six months of transaction issuance); Loans that are secured by single-family primary residences; Loans for home purchase; Loans with full appraisals on the secured properties; Loans with full underwriting and the verification of borrower assets; Loans with full income documentation; Fully amortizing loans; Fixed-rate loans; APRIL 24,

11 Loans with a 30-year maturity; First-lien mortgages; and No simultaneous second liens. 'AAA' projected loss anchor 31. To calibrate the foreclosure frequency for our archetypal pool, we set an anchor point at the 'AAA' rating level that reflects our opinion regarding the defaults that a mortgage pool would incur under extreme economic stress. The 'AAA' anchor for projected defaults for the archetypal pool is 15%. This anchor serves as a fixed benchmark against which all pools are measured and foreclosure frequency is adjusted based on pool-specific variations from the archetypal pool's characteristics. 32. Based on our analysis, a 15% foreclosure frequency assumption is consistent with our 'AAA' stress. We assume that in a 'AAA' scenario, there would be a 46% overall MVD nationwide, which in conjunction with our assumed foreclosure costs (including interest lost from advancing) over the liquidation period, would result in a roughly 50% loss severity for an archetypal pool. The roughly 50% loss severity for the archetype is computed assuming a loan balance of $400,000, interest rate of 4%, liquidation timelines of 35 months, and a fairly valued housing market (i.e., no over/undervaluation). The net result of our 15% foreclosure frequency and 50% loss severity is a 7.5% 'AAA' projected loss estimate for the archetypal pool in an extreme stress scenario. 33. The rationale for our anchor point concept draws from studies of significant historical events including the Great Depression. Studies conducted by the National Bureau of Economic Research show that lenders experienced default rates of around 12%-16% during that time (see "Urban Mortgage Lending by Life Insurance Companies," R. J. Saulnier, National Bureau of Economic Research, 1950; and "Commercial Bank Activities in Urban Mortgage Financing," C. F. Behrens, National Bureau of Economic Research, 1952). We believe mortgage loans at that time had sufficiently prime-like characteristics. Other relevant factors from the Great Depression include the unemployment level, the decline in house prices, and the decline in the U.S. GDP. 34. In addition, we reviewed the historical performance data of roughly 58,000 agency (about 98.4%) and nonagency (about 1.6%) loans with archetypal-like attributes originated between 2006 and Historical data for the reviewed loans covered the time from origination through September 2014 for the Freddie Mac loans, March 2015 for the Fannie Mae loans, and May 2015 for the nonagency loans. This analysis further informed our 15% 'AAA' anchor point assumption, as observed default rates were below this level in line with our expectations given that unemployment levels and GDP declines experienced over that time period were not as high as those associated with our 'AAA' rating definition. 35. We also reviewed the historical performance of roughly 7,500 agency loans (some of which were drawn from the 58,000 loans mentioned above) with archetypal-like borrower attributes and loan characteristics originated from third-quarter 2005 through second-quarter 2008 (through September 2014 for the Freddie Mac loans, March 2015 for the Fannie Mae loans) secured by properties in MSAs that had, in our view, experienced near depression-like employment contractions. As expected given the lack of geographic diversity and poor underwriting standards associated with these loans relative to our archetypal pool, we observed defaults exceeding 15% in some of these MSAs. Under the criteria, we would similarly project defaults in excess of 15% for these loans after applying APRIL 24,

12 adjustment factors to the anchor point to account for the geographic concentration and weak origination quality characterizing the dataset. Incorporating market outlook into rating analysis 36. The 'B' projected loss level for an archetypal residential loan pool matches expectations of losses, and therefore varies according to changes in our outlook for the mortgage market. Our outlook considers our current projections for U.S. economic growth, unemployment rates, and interest rates, as well as our view of housing fundamentals, such as housing starts, home sales, house prices, and price-to-rent and price-to-income ratios and is updated, if necessary, when these projections change materially. The horizon for our outlook is three to five years. 37. For the archetypal pool, we peg the 'B' foreclosure frequency at 2.5% for loans that are assessed during a benign economic period with a stable or positive outlook. We may revise our outlook and subsequently our 'B' projected foreclosure frequency to reflect changing economic conditions as warranted but would generally not expect the 'B' projected foreclosure frequency to be less than 2.5%. 38. Although the 'AAA' foreclosure frequency for an archetypal pool is expected to be constant during normal economic cycles (i.e., a cyclical trough no worse than a moderate stress), we may increase it if economic and market conditions migrate significantly beyond the normal cycle. 39. As our 'B' (or 'AAA', if applicable) assumptions change with our outlook, the foreclosure frequencies for the ratings in between would also fluctuate as determined by interpolating between the 'AAA' and 'B' levels. As we adjust our 'B' projected foreclosure frequency to reflect our outlook for a stressed economic environment, we would expect the foreclosure frequencies at all rating levels to compress towards the 'AAA' foreclosure frequency. 40. As our outlook changes, we may reassess the potential impact that this change may have on our outstanding RMBS ratings. Table 1 provides the loss estimates for an archetypal loan pool at all rating categories assuming the 'B' foreclosure frequency is set at 2.5% as described above. Table 1 Loss Estimates By Rating For The Archetypal Pool Rating AAA AA A BBB BB B Foreclosure frequency* Loss severity* Loss coverage* *Percentage of original pool balance. Figures are rounded to the nearest hundreth. The loss severities (and subsequently loss coverage) for the archetype for each rating category are computed assuming an average loan balance of $400,000, interest rate of 4%, liquidation timelines of 35 months, and a fairly valued housing market (i.e., no over/under valuation). Loss Model Foreclosure Frequency Factors 41. The table below outlines our loan-level foreclosure frequency adjustment factors, which are based on our analysis of the performance through June 2015 of roughly 8 million mortgage loans originated from 1998 through 2007, relative risk ranking assessment, and/or market research. These factors are assessed for each loan in the mortgage pool, and APRIL 24,

13 are then multiplied by the 15% archetypal 'AAA' foreclosure frequency and 2.5% archetypal 'B' foreclosure frequency to assess the foreclosure frequency for each loan that makes up an actual pool. The final estimated foreclosure frequencies are capped at 100% at the loan level. The foreclosure frequencies for each loan between the 'AA+' to 'B+' rating levels are then derived from an interpolation of the foreclosure frequency values at the 'AAA' and 'B' levels. Table 2 Foreclosure Frequency Adjustment Factors Variable Factor range Adjustment description CLTV-FICO Based on FICO/CLTV matrix Number of borrowers Number of borrowers Loan term* Continuous function based on original loan term to maturity Seasoning Continuous function based on seasoning DTI Continuous function based on borrower back-end DTI Occupancy Based on occupancy types Property type % LC Based on property types; 100% LC for other property types Loan purpose Based on loan purpose types Documentation type Based on documentation types Deposit money verification Based on down payment verification for purchase loans Loan type fixed/arm/hybrid Based on buckets of interest rate adjustment terms Loan type IO Based on buckets of IO term Loan type balloon Based on buckets of balloon term and CLTV Loan type - negam Borrower residency status Delinquency status/pay history % FF Based on delinquency status buckets/100% FF for 90+/REO/FCL Second lien *For balloon loans the adjustment factor for loan term is based on the original amortization term plus any interest-only term, if applicable. Other property types include manufactured housing, mixed-use property, raw land, mobile home, or other property types that, in our view, do not fall into the above categories. CLTV--Combined loan-to-value ratio. DTI--Debt-to-income ratio. LC--Loss coverage. ARM--Adjustable-rate mortgage. IO--Interest only. Negam--Negative amortization. FF--Foreclosure frequency. REO--Real estate-owned. FCL--Foreclosure. 42. If the value for a variable(s) listed in the table above is missing for any loan(s), we may assess other information that we believe is comparable so we can assess the relevant risks and, depending on the nature of the other information, we may make assumptions that in our view adequately address the risk in our analysis. CLTV and FICO 43. In our view, the CLTV ratio and the FICO score are two of the most important measures of credit risk in residential mortgages. While the CLTV reflects the borrower's potential willingness to pay back the mortgage given the equity at stake, the FICO score helps indicate the relative likelihood that the borrower will make the required payments based on past payment history. 44. The interaction between CLTV and FICO is a significant factor and the joint effect is not simply the product of the two independent effects. The interaction shows that the default likelihood grows considerably for CLTV-FICO combinations as FICO scores get lower and CLTV ratios get higher (see table 3 for the CLTV-FICO adjustments for select combinations of CLTVs and FICO scores). Our CLTV-FICO score adjustment factors range from approximately APRIL 24,

14 0.1 times to more than 30 times the foreclosure frequency anchor point. For a complete list of the foreclosure frequency adjustment factors for all CLTV-FICO combinations, see "Assumptions Supplement For Request For Comment: Methodology And Assumptions For," published April 24, Table 3 Select CLTV-FICO Adjustment Factors CLTV (%) FICO CLTV--Combined loan-to-value ratio. 45. FICO score We typically use the lower of two or middle of three FICO scores pulled from the credit bureaus, as applicable, for each borrower. If multiple borrowers are included on the loan application, we will typically use the FICO of the borrower with the lower score. For further details on the treatment of loans to multiple borrowers, see the Number of Borrowers section. 46. In our analysis to assign new ratings, we typically use FICO scores that are not greater than 180 days old as of the transaction cut-off date. However, for loans that have been 30 days delinquent more than once or worse than 30 days delinquent in the past 180 days, we would typically use an updated FICO that is not greater than 60 days old as of the transaction cut-off date. We typically apply the foreclosure frequency adjustment factors outlined in the Delinquency Adjustment section if updated FICOs are not available or do not, in our view, adequately address any delinquent payment since the refreshed FICO was obtained. 47. Determination of CLTV Chart 3 shows how we derive the property value used to determine the CLTV utilized in our analysis. APRIL 24,

15 48. When we refer to the indexation of property values, our application of the FHFA house price index reflects an adjustment of 50% of any upward movements in the index and 100% of any downward movements in the index to the property value. We may make additional adjustments to the FHFA indexed value if we believe current changes in house prices are not fully captured by the index given the lag in index reporting or rapidly shifting market conditions. 49. For all loans, we take into account both the loan's original and current CLTV in our analysis because we consider both the borrower's original down payment and current equity to be meaningful drivers of willingness to pay. The simple average of the original CLTV and current CLTV is used to derive the CLTV used in our analysis. The original CLTV is based on the ratio of the original loan balance (and any known second-lien exposures) to the original valuation. 50. For unseasoned loans, the original valuation is our starting point. Depending on the age of the valuation, we then typically adjust for changes in house prices based on the Federal Housing Finance Agency's (FHFA's) house price index. The resulting value is used in conjunction with the current loan amount (and any known second-lien exposures) to calculate the current CLTV. 51. For seasoned and reperforming loans, the current CLTV depends on the availability of updated values (such as broker price opinions, drive-by appraisals, automated valuation models, or other valuation products) at the time of our APRIL 24,

16 analysis. In assigning new ratings, we typically use updated property valuations that are no greater than 180 days old from the transaction cut-off date for a sample of any seasoned loans and for all reperforming loans being analyzed. Where one or more updated values are available, we typically begin with the updated value obtained from what we consider the more robust valuation method. We then consider how this updated value compares to other updated valuations available, as well as to the FHFA house price index-adjusted original appraised value. Based on this comparison, we may haircut this updated value to reflect any potential overestimation of value. Factors considered to determine the haircut may include the degree of variance among reviewed values, concentrations in distressed property markets, and valuations significantly less than or greater than the market average. 52. Depending on the age of the updated value used, we adjust the updated value based on the FHFA house price index. The resulting value is used in conjunction with the current loan amount (and any known second-lien exposures) to calculate the current CLTV. 53. If updated values are not available, we use the FHFA house price index-adjusted original valuation to calculate the current CLTV. We may make additional adjustments to the FHFA indexed value based on our analysis of the updated values (e.g., BPOs or AVMs) available for comparable loans included in the pool. 54. For HELOCs, the loan balance used to determine the original and current CLTVs is based on the fully drawn amount unless the draw period has expired. For negative amortization (negam) loans, the loan balance used to determine the original and current CLTVs is based on the maximum amount to which the balance can increase per the loan terms unless the negative amortization period has expired. 55. We also generally adjust each loan's current CLTV to account for any outstanding arrearages, such as delinquent principal and interest payments, forborne principal, unpaid taxes and insurance, or corporate advances made by the servicer. These arrearage amounts are generally added to the loan's current balance to determine the CLTV used in our analysis. In our view, including arrearages more accurately captures the borrower's financial obligations and hence their likelihood of default. Number of borrowers 56. We are proposing to introduce a new adjustment factor in our criteria to account for the number of borrowers on the loan application. A 1.0x factor would apply to loans with a single borrower and a 0.75x adjustment factor to loans with two or more borrowers. We've observed that loans with more than one borrower typically have a lower foreclosure frequency, which is likely attributable to the higher likelihood of multiple income sources associated with loans with co-borrowers and the ability to sustain mortgage payments if one of the borrowers experiences a life event. Original loan term 57. The risk of default varies directly with a first-lien mortgage loan's original term to maturity (or original amortization term plus any interest-only term, if applicable, for balloon loans). The foreclosure frequency adjustment factor for original terms to maturity of less than 360 months is less than 1.0x, reflecting our view of it being a stronger loan attribute. In addition, equity builds up more rapidly due to faster amortization, increasing the borrower's capacity to withstand negative movements in house prices. The adjustment factors applied to the foreclosure frequency for each loan for the loan term are a continuous function. Table 4 outlines the foreclosure frequency adjustment factors for specific original terms to maturity. For a list of the foreclosure frequency adjustment factors for all original terms to APRIL 24,

17 maturity (except balloon loans, for which we use original amortization term plus any IO term, if applicable), see " Assumptions Supplement For Request For Comment" Methodology And Assumptions For Rating U.S. RMBS Issued 2009 And Later," published April 24, Table 4 Select Loan Term Foreclosure Frequency Adjustment Factors Original term to maturity (months)* Adjustment factors *For balloon loans, original amortization term plus any interest-only term, if applicable, is used. Seasoning 58. We've observed that loans that have been seasoned more than five years since origination generally perform better than newly originated loans with similar characteristics. Therefore, we adjust our foreclosure frequency downward to reflect a seasoning credit after a loan has been seasoned five years from the first payment date and onwards, plateauing after year 10. However, the seasoning credit is typically further adjusted based on the borrower's 24-month pay history, which reflects relative risk ranking based on the seriousness of any prior delinquencies observed. Seasoning credit for loans with confirmed modifications will be based on the date of the most recent known modification rather than the origination date. We do not provide seasoning credit to loans that are currently delinquent or to balloon loans due to the refinancing risk facing borrowers at balloon maturity. Table 5 outlines the foreclosure frequency adjustment factors for seasoning based on the loan's delinquency status. Table 5 Seasoning Foreclosure Frequency Adjustment Factors Seasoning (months)* Delinquency status x <= < x <= < x <= < x <= < x <= 120 Presently current, but once 30 days delinquent in the past 24 months Presently current, but more than once 30 days delinquent in the past 24 months or once 60 days delinquent in the past months Presently current, but once 60 days delinquent in the past 12 months or more than once 60 days delinquent in the past 24 months Presently current, but once (or more) 90+ days delinquent in the past 24 months x > Currently delinquent/balloon loans *Measured from loan's first payment date or most recent confirmed modification date, as applicable. Back-end DTI ratio 59. The relationship between a mortgagor's monthly debt payment obligations and income is an important measure of their ability to repay the mortgage; therefore the back-end DTI ratio has been a traditional part of mortgage loan APRIL 24,

18 underwriting. The foreclosure frequency adjustment factor for a 36% back-end DTI ratio is set to 1.0x. Table 6 outlines the foreclosure frequency adjustment factors for select back-end DTI values. For a list of the foreclosure frequency adjustment factors for all back-end DTI values, please see "Assumptions Supplement For Request For Comment: Methodology And Assumptions For," published April 24, Table 6 Select Back-End DTI Foreclosure Frequency Adjustment Factors Back-end DTI (%) Adjustment factors DTI--Debt-to-income ratio. 60. A 36% back-end DTI (i.e., 1.0x factor) is typically applied to loan types for which the borrower's personal income is not verified, and instead the documentation type factor adjustment (described further below) takes into consideration the additional risk associated with nonincome-verified loans. In addition, we will assume a 36% back-end DTI for loans with 10 or more years of seasoning. As a loan seasons, observed payment history becomes more predictive of future performance than certain original borrower characteristics. We therefore neutralize our view of the potential risk associated with the original back-end DTI for loans with 10 or more years' seasoning, equalizing the risk of default for loans with varying original back-end DTIs, all other attributes being equal. Seasoning credit is still applicable as for any loan based on our observations of improved performance for seasoned loans. Occupancy 61. During periods of economic stress, a borrower is more likely to default on a mortgage secured by a second home, and even more likely to default on a mortgage secured by an investment property, than on a home that is a primary residence. Table 7 outlines the foreclosure frequency adjustment factors for occupancy type. Table 7 Occupancy Foreclosure Frequency Adjustment Factors Occupancy type Adjustment factors Primary 1.00 Second home 1.20 Investment 1.50 Property type 62. The risk of default for non-single-family residences is higher than that for single-family properties (including planned unit developments [PUDs]). Therefore, we increase our foreclosure frequency estimates for loans secured by non-single-family residences (see table 8). APRIL 24,

19 Table 8 Property Type Foreclosure Frequency Adjustment Factors Property type Adjustment factors Single-family residence, PUD 1.00 Condominium, cooperative 1.10 Two/three/four family 2.00 Other* 100% LC *Typically includes the following property types: manufactured housing, mixed-use property, raw land, mobile home, or other property types that in our view do not fall into the categories above. These assumptions are applicable when these property types do not make up a significant portion in a residential mortgage-backed securities transaction. PUD--Planned unit development. LC--Loss coverage. Loan purpose 63. Mortgage default risk varies with a loan's purpose. Generally, a loan used to borrow against the equity in a property (cash-out) has the highest default risk. Therefore, we apply a 1.25x foreclosure frequency adjustment factor for cash-out refinance mortgage loans. We apply a 1.0x foreclosure frequency adjustment factor for a purchase loan and for a refinanced loan that is not a cash-out refinance. Documentation type 64. The documentation type framework used in our analysis reflects the source of information used to verify income as well as the number of months of documentation considered. Traditional documentation generally includes pay stubs, W-2s, and IRS transcripts (personal/business), while nontraditional documentation typically includes bank or asset statements. If multiple sources of income, either from a single borrower with multiple jobs or from multiple borrowers, are provided, generally the number of months for verified income should reflect the lowest amount of months of verified or validated income from any source for any borrower used for qualification purposes. If there are multiple borrowers with different documentation types, the highest documentation type adjustment factor from any borrower will be applied in our analysis. Table 9 outlines the foreclosure frequency adjustment factors for different documentation types. In some cases, we may adjust these foreclosure frequency factors to reflect what, in our view, may be less stringent underwriting documentation guidelines. 65. In addition, for loans where the lender did not verify the assets used to make the down payment for a purchase (deposit money verification), we apply a foreclosure frequency adjustment factor of 1.05x. Table 9 Documentation Type Foreclosure Frequency Adjustment Factors Months of verification Documentation type Full Alternative Description Traditional documentation used for fully verifying and calculating the qualifying income of the borrower(s) (e.g., written verification of employment, pay stubs, W2s, personal and business tax returns, IRS transcripts). Nontraditional income documentation used for verifying and calculating the qualifying income of the borrower(s) (e.g., bank statements). x >= 24 months 12 <= x < 24 months x < 12 months APRIL 24,

20 Table 9 Documentation Type Foreclosure Frequency Adjustment Factors (cont.) Months of verification Documentation type Other Description Documentation used to qualify a borrower does not fit into the above documentation type categories (e.g., borrower(s) qualified under asset-based lending programs that consider accumulated assets rather than a verified income stream, or the loan is underwritten to rental income on the property rather than the borrower's personal income). Typically a 3.0x adjustment factor will be applied and qualitative adjustments up to a 4.0x factor may be considered to reflect what, in our view, may be more bespoke underwriting documentation guidelines. x >= 24 months 12 <= x < 24 months x < 12 months As a loan seasons, the risks associated with the initial documentation used to qualify the borrower diminish as observed payment behavior becomes more indicative of the performance. We therefore gradually neutralize our view of the potential risk associated with the original documentation type, eventually equalizing the risk of default for loans with varying documentation types, all else being equal. Seasoning credit is still applicable as for any loan based on our observations of improved performance for seasoned loans. 67. We start phasing out the documentation type foreclosure frequency adjustment factor beginning after year two with no adjustment factor applied after year six (see chart 4). These factors are multiplied by the difference between the documentation type foreclosure frequency adjustment factor for that documentation type (including the adjustment for lack of deposit money verification, if applicable) and a neutral foreclosure frequency adjustment factor (i.e. 1.0x), as indicated in the equation below, to determine each loan's overall documentation type foreclosure frequency adjustment. APRIL 24,

21 Chart 4 Loan type 68. Loan type is an important factor in projecting defaults, as different loan products can affect the growth of home equity and/or introduce greater payment shock. Our research indicates that: Adjustable-rate mortgage (ARM) loans with shorter initial fixed-rate periods are riskier than those with longer initial fixed-rate periods; Interest only (IO) are risker than fully amortizing loans and IO loans with shorter initial IO terms are riskier than those with longer IO terms; Balloon loans are riskier than nonballoons and balloons with shorter terms are riskier than those with longer terms; and Negam loans are riskier than loans without negam features. 69. Table 10 outlines the foreclosure frequency adjustment factors for various loan types. Table 10 Loan Type Foreclosure Frequency Adjustment Factors Rate type Fixed-rate period (months) Adjustment factors ARM FRP < ARM 60 <= FRP < APRIL 24,

22 Table 10 Loan Type Foreclosure Frequency Adjustment Factors (cont.) ARM FRP >= Fixed N/A 1.00 Negam N/A 3.00 Adjustment factors Balloon term x < 90 CLTV x >= 90 CLTV x <= 120 months x > 120 months IO term Adjustment factors x < 60 months 1.60 x >= 60 months 1.25 ARM--Adjustable-rate mortgage. FRP--Fixed-rate period. Negam--Negative amortization. CLTV--Combined loan-to-value ratio. IO--Interest only. N/A--Not applicable. Borrower residency 70. We apply a 1.5x foreclosure frequency adjustment factor for loans made to borrowers who are either non-u.s. citizens or nonpermanent resident aliens to account for our view of the potentially higher default risk posed by these borrowers. Delinquency adjustments 71. The delinquency adjustments are intended to supplement FICO scores. At new issue, updated FICOs are typically obtained that incorporate prior delinquencies. However, if a loan is delinquent at closing, then the delinquency adjustment would still apply. The delinquency adjustment for prior delinquencies will generally not be applicable when an updated FICO score during our surveillance review is available unless the delinquency occurred after the FICO was pulled. Table 11 below outlines the foreclosure frequency adjustment factors for delinquency status. Table 11 Delinquency Foreclosure Frequency Adjustment Factors Delinquency status Adjustment factors Presently current, but once 30 days delinquent in the past 24 months 1.00 Presently current, but more than once 30 days delinquent in the past 24 months or once 60 days delinquent in the past months Presently current, but once 60 days delinquent in the past 12 months or more than once 60 days delinquent in the past 24 months Presently current, but once (or more) 90+ days delinquent in the past 24 months 2.00 Presently 30 days delinquent 2.50 Presently 60 days delinquent 5.00 Presently 90+ days delinquent/real estate owned/foreclosure FF--Foreclosure frequency. Second-lien mortgage loans % FF 72. We are proposing to introduce a new adjustment factor in our criteria to account for second-lien mortgage loans, both closed-end and HELOCs. We will apply a 1.3x foreclosure frequency adjustment factor for second liens to account for the increased risk associated with second liens relative to first-lien mortgages. APRIL 24,

23 Loss Model Loss Severity Assumptions 73. The loss model computes the loss severity by first calculating net recovery proceeds for each loan based on five key components: MVD assumptions (which include an FSD); Over/under valuation adjustments (based on long-term price-to-income ratios) for current housing market conditions; Foreclosure and preservation costs; Servicer advances of delinquent interest (where applicable); and Taxes and insurance. 74. After the recovery proceeds for each loan are determined, we then compare the recovery proceeds to the loan's current balance to determine the estimated loss severity at each rating level. APRIL 24,

24 Market value declines and over/under valuation 75. To estimate a loan's loss severity, we first determine the property value to be used in our analysis as of the date we are reviewing the pool. For all loans, we use the property value derived as described in the Determination of CLTV section. 76. Next, we apply our MVD assumptions. The MVD assumptions include an assessment of the level of over/under valuation in the prevailing property market at the MSA level by comparing the long-term average of the ratio of house prices to income to the current ratio. If a property cannot be mapped to a particular MSA, we will use the state or national information to determine the level of over/under valuation. To measure the degree of over/under valuation in a particular property market, we may use various data sources that are available, including from the FHFA, U.S. Bureau of Economic Analysis, U.S. Census Bureau, IHS Global Insight, expert opinions, and independent research. 77. The percentage added or deducted based on a property's assumed over/under valuation status is detailed in columns three and four in table 12. The degree of over/undervaluation in a property market is only partly reflected in the repo MVD formula. This is to account for our view that property values during a housing cycle will in part reflect the current valuation of the property market, but revert to the long-term average. The percentages in the fourth column of table 12 limit the reduction of the repo MVD in an undervalued property market. We use a constant percentage of any undervaluation across all rating levels. By contrast, the adjustment for overvaluation is linked to a particular rating category, and is highest for a 'AAA' rating (see the third column in table 12). 78. Our MVD assumptions also incorporate fixed MVD assumptions as well as FSD factors, which vary by rating category, into the modeled loss severity for each loan. 79. The repo MVD derives from the formula below, based on the rating-specific input variables shown in table 12. Absent any over/under valuation, the resulting repo MVD is as shown in the last column in table 12 and is 46% at 'AAA'. Table 12 Property Market Adjustments For Calculating Repossession Market-Value Decline Modeling Assumptions Rating category Fixed market-value decline (%) Percentage of overvaluation added (%) Percentage of undervaluation deducted (%) Forced-sale discount (%) Repo MVD, absent over/undervaluation (%) AAA (20) AA (20) A (20) BBB (20) BB (20) B (20) Note: Repo MVD = 1 - [1 - (Fixed MVD +/- percentage of over/undervaluation x over/undervaluation)] x (1-FSD). MVD--Market value decline. FSD--Forced-sale discount. 80. Fixed MVD is the fixed recessionary MVD shown in the second column of table The FSD factor is detailed in the fifth column of table 12. Among other factors, foreclosed properties may sell at a discount due to the stigma of repossession. The FSD factor is larger at lower rating levels and smaller at higher rating levels because in a more severe recession, a greater proportion of all property transactions will come from distressed APRIL 24,

25 sales. 82. For HELOCs, the current loan balance is based on the fully drawn amount unless the draw period is expired. For negam loans, the current loan balance is based on the maximum amount to which the balance can increase per the loan terms unless the negative amortization period has expired. 83. We also generally adjust each loan's current balance to account for any outstanding arrearages if arrearage amounts are pledged to the transaction. Pledged arrearages may have a considerable impact on the ultimate recoveries to the securities. 84. We are introducing floors to our loan-level loss severities to address any potential idiosyncratic risk on loans with very low current CLTVs that would not be otherwise captured by our assumptions. However, we may go below the floors for loans that carry mortgage insurance or guarantees. The loss severity floors are 20% at the 'AAA' rating category, 18% at the 'AA' rating category, 16% at the 'A' rating category, 14% at the 'BBB' rating category, 12% at the 'BB' rating category, and 10% at the 'B' rating category. Liquidation timelines 85. In our view, liquidation timelines (which include the assumed default, foreclosure, and real estate owned [REO] time periods) show variability across states based on the specific legal processes for foreclosure. Hence, for the loss severity calculation, we assume varying timelines across states and rating levels as outlined in chart 6. Liquidation timeline assumptions for the 'AA+' to 'B+' rating levels are derived from a linear interpolation of the timelines assumed at the 'AAA' and 'B' levels. For the 'B-' rating level, the liquidation timeline assumptions are the same as the 'B' level. In our analysis, variable liquidation costs (i.e., interest advancing, if applicable, and taxes and insurance) accrue throughout each loan's liquidation timeline based on the state where the property securing the mortgage is located. APRIL 24,

26 Chart 6 Interest lost from advancing 86. If the servicer is contractually obligated to advance delinquent principal and interest payments on the loans through liquidation, these advanced amounts will be reimbursed to the servicer from proceeds realized upon liquidation. Therefore, the interest advance amounts are reflected in our loss severity estimate. Interest advance amounts accrue at the applicable advanced interest rate described in table 13. Table 13 Interest Lost From Advancing Loan type Fixed ARM Initial fixed-rate period (months) N/A x < 60 x = 60 x > 60 Advanced interest rate* Maximum of the loan's original interest rate or the loan's current interest rate. Maximum of the loan's original interest rate or the rate that is equal to the loan's current interest rate plus two-thirds of the difference between the loan's current interest rate and the loan's lifetime maximum rate. Maximum of the loan's original interest rate or the rate that is equal to the loan's current interest rate plus one-third of the difference between the loan's current interest rate and the loan's lifetime maximum rate. Maximum of the loan's original interest rate or the loan's current interest rate. APRIL 24,

27 Table 13 Interest Lost From Advancing (cont.) Negam N/A The loan's lifetime maximum rate. *Current interest rate refers to the interest rate on the loan at the time of our analysis. The current interest rate on a fixed-rate loan may differ from the original rate in the event the loan coupon is subject to a modification. ARM--Adjustable-rate mortgage. Negam--Negative amortization. N/A--Not applicable. 87. For loan types not included above, we will generally apply an advanced interest rate assumption in our analysis based on the loan's specific features that best matches the characteristics outlined in table If the servicer advances delinquent principal and interest payments for a limited time (e.g., four to six months) or is not contractually obligated to advance, interest lost from advancing may not be considered in our loss severity estimate. In addition, we typically will not give credit to servicer advances in our cash flow analysis, as described further in the Servicer Advance Stresses section. Foreclosure costs 89. Total foreclosure costs include legal costs (such as attorney fees and court costs), appraisals, title fees, preservation costs, and brokerage commissions. Fixed foreclosure cost (excludes brokerage fees) assumptions vary by rating category to reflect the potential for higher expenses in more stressed environments when liquidation timelines may be extended (see table 14). Table 14 Fixed Foreclosure Cost Assumptions Rating category Foreclosure costs ($) AAA 10,000 AA 9,000 A 8,000 BBB 7,000 BB 6,000 B 5, In addition, our foreclosure cost assumptions at every rating category include a brokerage commission equal to the higher of 5% of the repo MVD-adjusted property value or $3,000 per loan. We may update these costs over time based on changes in foreclosure laws and/or observed market data as well as inflationary pressures, if warranted. Taxes and insurance 91. To maintain and protect the lien, the servicer is expected to cover taxes and insurance costs during a property's liquidation. Chart 7 shows the assumptions for each state for taxes and insurance, which ranges from 1.25%-3.00% annually, and are a percentage of the property value applied in our analysis before any rating-specific MVD/FSD. The total amount is computed based on the total associated liquidation timelines. APRIL 24,

28 Chart 7 Second lien loss severity 92. We typically assume a 100% loss severity for second-lien collateral at all rating levels. This accounts for the second lien priority of the mortgage and the expectation that there will be little to no recovery after satisfying the first lien. Furthermore, because of the low likelihood of the second-lien holder receiving recovery proceeds, we assess whether there are charge-off provisions after defaulting on the second lien to minimize the accrued costs associated with the second-lien mortgage. Depending on this analysis, we may increase our second lien loss severity estimates above 100% to reflect the potential for additional liquidation costs exceeding the loan balance to accrue to the trust. ATR and qualified mortgage (QM) standards loss severity adjustments 93. Our analysis considers the likelihood and magnitude of potential liabilities arising from loans covered under the ATR and QM rule (the ATR rule). The ATR rule, which is part of the Truth in Lending Act (TILA), requires lenders to make a reasonable, "good faith" determination of a consumer's ability to repay a mortgage loan (covered loans). 94. Our loss severity adjustments reflect the potential damages assessed to an assignee (i.e., the securitization trust) if a borrower brings a successful defensive claim that the ATR rule was violated, and the additional costs associated with the potential extension of foreclosure timelines to resolve the claim. These adjustments vary based on the loan's TILA status. Given the "safe harbor" provisions granted to QM/non-higher-priced mortgage loans (non-hpml), we do not APRIL 24,

29 adjust the loss severities for QM/non-HPML loans. Appendix II provides more details for sizing the loss severity adjustment to mitigate the potential risk to securityholders. Mortgage insurance and guarantees 95. In addition to the loss severity factors we've outlined, we may reduce the loss severities on a loan covered by mortgage insurance or a guarantee (see "Methodology For Assessing Mortgage Insurance And Similar Guarantees And Supports In Structured And Public Sector Finance And Covered Bonds," published Dec. 7, 2014). Pool-Level Adjustments 96. Pool-level adjustments include those made to reflect certain concentration risks, as well as qualitative factors such as origination quality and R&W robustness. Because these pool-level considerations may affect both the foreclosure frequency and loss severity of a pool, the related adjustment factors are applied to loss coverage levels, with 50% of the applicable factor adjusting the pool-level foreclosure frequency estimates and 50% adjusting the pool-level loss severity estimates at each rating level. Loan count/small pool adjustments 97. Smaller pools (which we define as less than 250 loans) exhibit greater idiosyncratic behavior than large and highly diversified pools. Therefore, for RMBS pools with less than 250 loans at the time of the initial bond issuance, an adjustment factor will be applied to loss coverage. The primary purpose of the small pool adjustment factor is to capture idiosyncratic risks and sample bias/adverse selection that may exist in smaller or more concentrated loan pools. 98. The small pool adjustment does not apply to seasoned RMBS transactions that initially comprised 250 or more mortgage loans but become more concentrated over time through voluntary or involuntary prepayments. In our view, the risk associated with small loan counts toward the latter part of a securitization's life is mitigated through the interplay of structural features, such as payment structure and credit enhancement floors (see the Credit Enhancement Floors section). 99. For the formula used to calculate the small pool adjustment, please see Appendix III. Geographic concentration adjustments 100. The archetypal pool consists of mortgages secured by properties that are geographically diversified. As the concentration of properties securing the underlying mortgages in an RMBS transaction increases, that transaction's overall default risk generally increases because a localized economic downturn would have a greater impact on a geographically concentrated pool than it would on a geographically diversified pool of loans To calculate the geographic concentration adjustment factor, S&P Global Ratings uses the Herfindahl-Hirschman Index (Herfindahl Index). The Herfindahl Index is a commonly used measure of economic diversity (see equation 3 in Appendix III). We apply the Herfindahl Index using the concentration percentage of loans within each CBSA. When analyzing specific pools, we calculate a pool-specific Herfindahl Index using the CBSA definitions available at the time of our most recent LEVELS model update. We then compare the pool-specific Herfindahl Index to a baseline index and apply a multiple that aids in differentiation among different loss coverage levels for varying concentration levels. APRIL 24,

30 We calculate our geographic concentration adjustment factor using the following equation: Geographic concentration adjustment factor = exp{kd} "D" represents the difference in the Herfindahl indices and "k" is the multiple (see Appendix III) Based on the distribution of loans in our historical mortgage data set, the baseline Herfindahl Index is set at 0.03 for these criteria. We evaluate any pool with a Herfindahl Index higher than the 0.03 baseline index value for geographic concentration risk and adjust the loss coverage accordingly. In the event that the pool is more diversified than the baseline, we do not apply a geographic concentration adjustment factor to the loss coverage Appendix III contains the formulas used to calculate the components of the geographic concentration adjustment factor To the extent a particular RMBS pool has significant geographic concentration risk that we believe is not sufficiently captured by the Herfindahl Index, we may supplement the assessment described above with additional analysis. This analysis may consider the number of concentration areas and their economic profile and would likely result in a geographic concentration adjustment factor no more than 1.15x the factor based on the Herfindahl Index. Modified loans and prior credit events 105. We typically adjust the loss coverage for any loans with confirmed modifications based on the payment history since loan modification (up to the most recent 24 months) and the seasoning of the loan modification. This adjustment may vary from 1.0x to 5.0x loss coverage and may also vary based on the rating level. Generally, for loans with modifications that were made seven or more years prior and have clean pay histories (i.e., only 30 days delinquent one time) over the 24 months preceding our analysis, we will apply a 1.0x factor to loss coverage, given the borrowers' demonstrated performance post-modification. Furthermore, we may also apply similar adjustments for loans to borrowers with a history of prior credit events where the time elapsed since the credit event is less than industry standards for disregarding prior credit events (such as the timelines for prior foreclosures, short-sales, and bankruptcies set by Fannie Mae, Freddie Mac, and the Federal Housing Administration [FHA]) at origination, regardless of whether the loan may be newly originated and possess a recent clean payment history. Qualitative adjustments 106. We may also apply loss coverage adjustments related to mortgage originator/aggregator operational assessments (MOAs), due diligence findings, and R&W frameworks as outlined in table 15. The maximum adjustment achievable in our analysis for these three factors, in aggregate, is floored at a 0.70x loss coverage (see the relevant sections below). Table 15 Qualitative Loss Coverage Adjustment Factors Lower range Upper range (>=) Mortgage operational assessment (MOA) 0.70x 1.30x Representations and warranties (R&Ws) 0.95x 1.15x Due diligence (DD) 1.00x 1.30x Overall range for MOA/R&W/DD 0.70x ~1.94x APRIL 24,

31 Other considerations 107. Large loan analysis supplemental test To address mortgage pools that contain large balance loans, which may result in less diversification, we also perform a top concentration test when we assign an initial rating. For this analysis, we compare the loss coverage projection from our LEVELS analysis (in conjunction with any qualitative overlays) at each applicable rating level with the minimum loss coverage projection determined using table 16, based on the closing collateral pool balances. For example, if our LEVELS analysis (in conjunction with any qualitative overlays) indicates a 'BBB' loss coverage of 1.10%, and our minimum loss coverage projection from table 16 indicates a 1.30% loss for the 'BBB' rating category, the 1.30% loss coverage projection would be required for a 'BBB (sf)' rating. Table 16 Rating Category Minimum Loss Projections Maximum potential rating category AAA AA A BBB BB B Minimum loss coverage projection Highest balance loan liquidated at the greater of 50% loss severity and the applicable 'AAA' loss severity, plus the next nine largest loss exposures at the 'AAA' loss severity. Highest balance loan liquidated at the greater of 50% loss severity and the applicable 'AA' loss severity, plus the next seven largest loss exposures at the 'AA' loss severity. Highest balance loan liquidated at the greater of 50% loss severity and the applicable 'A' loss severity, plus the next five largest loss exposures at the 'A' loss severity. Highest balance loan liquidated at the greater of 50% loss severity and the applicable 'BBB' loss severity, plus the next three largest loss exposures at the 'BBB' loss severity. Highest balance loan liquidated at the greater of 50% loss severity and the applicable 'BB' loss severity, plus the next largest loss exposure at the 'BB' loss severity. Highest balance loan liquidated at the greater of 50% loss severity and the applicable 'B' loss severity. Payment Structure And Cash Flow Mechanics 108. We typically perform a cash flow analysis to assess whether the payment obligations on the rated securities can be supported by the cash flow from the securitized assets under various stress scenarios. Our cash flow stresses include rating scenario-dependent foreclosure frequency and loss severities derived by applying the criteria described above. In certain circumstances, such as sequential-pay structures with no excess spread, a cash flow analysis may not provide additional insight to the sufficiency of credit enhancement; therefore, we may not perform a cash flow analysis. In addition, we may not perform a cash flow analysis on every class included in a transaction when we have already otherwise captured the structural impact in our analysis (e.g., certain exchangeable certificates) Our cash flow analysis incorporates the following key assumptions, as applicable, which are described in greater detail below: Default curve stresses; Recovery lag stresses; Voluntary prepayment stresses; Extraordinary trust expense (ETE) stresses; Weighted average coupon (WAC) deterioration stresses; Interest rate stresses; Liquidity stresses such as delinquency curve and servicer advance stresses; and Servicing and other transaction fees and expenses. APRIL 24,

32 110. Depending on the structure, for a class to pass at a given rating level, the cash flow projections that are based on the application of our stresses simultaneously should not show any interest shortfalls (unless subject to deferral features as described below or unless we consider the shortfall is de minimis) or principal writedowns in the related rating level and any lower rating levels through 'B-'. Default curve stresses 111. U.S. mortgage default data since 2001 show that the default timing, for a given vintage, is generally similar across collateral types. However, vintages that entered a stressed economic environment after their initial peak default period experienced a subsequent spike in defaults following such stress. Nonetheless, while the default timing of loans originated in different vintages can vary with the timing of economic cycles, the majority of defaults tend to occur in the first five years following origination Our front- and back-ended default curves attempt to capture the stress related to the effect of economic cycle timing, as well as the observation that defaults happen early in a transaction's life. Generally, for newly originated and performing loans seasoned seven years or less, we apply the following front- and back-loaded default timing stresses in our cash flow analysis (see table 17 and chart 8). We begin our default timing in month seven of the transaction because a meaningful amount of defaults are generally not expected in the first six months. Percentages are applied to the cut-off date pool balance. In instances where we are running high prepayment scenarios, the weighted average life of the collateral may be shortened to a degree where the amount of losses that can be allocated against the outstanding pool balance may be less than the loss coverage amount calculated for a given rating level. In those instances we will deem the losses taken to be consistent with meeting our rating scenario, as the truncated losses are a result of the quick pay-down in the pool balance and rated notes rather than a shortfall in credit enhancement. Table 17 Default Timing Curve % of cumulative default* Cumulative default Period (months) Front-loaded Back-loaded Front-loaded (cumulative) Back-loaded (cumulative) *The defaults are distributed equally over their respective periods. APRIL 24,

33 Chart For those loans that are seasoned more than seven years, we generally apply the front-loaded curve only However, if loans that are 90 or more days delinquent, in foreclosure, or REO comprise a significant portion of the securitized pool, we will assume the loans default immediately. (i.e., month one in our analysis) We may adjust the default timing stresses applied in our analysis to account for pool characteristics or structural features that, in our view, make alternative assumptions more appropriate. For example, for transactions with structures where excess spread or overcollateralization (O/C) provides credit enhancement, we typically apply the defaults as bullets [i.e., lump sums] every six or 12 months as a way to stress excess spread. Recovery lag stresses 116. Our recovery lags reflect the time it takes to liquidate the property and recover proceeds on a loan since the borrower's first missed payment. Generally, we apply a recovery lag stress of 24 months to the defaults expected in our analysis at each rating level However, if seriously delinquent (i.e., 90 or more days delinquent or in foreclosure or REO) loans comprise a significant portion of the securitized pool, we will apply the recovery lag stresses listed in table APRIL 24,

34 Table 18 Recovery Lags For Serious Delinquencies Delinquency status Recovery lag 90+ days delinquent Zero liquidation proceeds for 16 months; rest equally over eight months Foreclosure Zero liquidation proceeds for eight months; rest equally over eight months Real estate-owned Liquidation proceeds equally over eight months 118. Generally, for transactions that provide for defaulted loans to be charged off from the securitized pool after a specific timeframe, we will align our recovery lag stresses with the charge-off provisions We may adjust the recovery lag stresses applied in our analysis to account for pool characteristics or structural features that, in our view, make alternative assumptions more appropriate. For example, if loans 30 or 60 days delinquent make up a substantial portion of the pool, we may adjust the recovery lag stresses for those loans to reflect that they are closer to liquidation than current loans. Voluntary prepayment stresses 120. We apply low and high voluntary prepayment assumptions, as shown in table 19, which reflect annualized prepayment percentages. These assumptions will be applied in all periods to the then-outstanding balance. These prepayment stresses are in addition to the percentage of loans assumed to default such that total prepayments in any month will equal voluntary payoffs and defaults. Table 19 Voluntary Prepayment Assumptions Rating category AAA AA A BBB BB B Low (%) High (%) We may adjust the voluntary prepayment stresses applied in our analysis to account for pool characteristics or structural features that, in our view, make alternative assumptions more appropriate. For example, we may apply a different prepayment assumption for a pool with a high concentration of two-year hybrid ARM loans to reflect an increase in prepayments near the initial fixed-rate term's expiration. Extraordinary trust expense stresses 122. U.S. RMBS transactions typically provide for extraordinary trust expenses to be paid from collections on the mortgage loans. To address the payment of these expenses in our analysis, we apply the extraordinary expense application factors listed in table 20 (which vary by prime and nonprime collateral and rating level) to the capped extraordinary expense amounts defined in the transaction's documents. We assume these amounts are incurred in months 13 through 60 after the transaction closes, based on our assumptions for the timing of peak losses and the potential exposure period for associated repurchase requests, which will likely give rise to most of the incurred ETEs. APRIL 24,

35 Table 20 Extraordinary Trust Expense Application Factors Rating category Prime (%) Nonprime (%) AAA AA A BBB BB B We may apply different application factors for extraordinary expenses if the capped amounts are not between $200,000 and $400,000. If extraordinary expenses are uncapped, we would derive projected extraordinary expenses using "Criteria Methodology Applied To Fees, Expenses, And Indemnifications," published July 12, To address the potential reduction in interest payments to securityholders over time because of these expenses, for transactions that allocate extraordinary expenses to interest payments on the securities, we apply these amounts to the projected interest reduction amount (PIRA) as described in "Methodology For Incorporating Loan Modifications And Extraordinary Expenses Into U.S. RMBS Ratings," published April 17, 2015, as applicable. For these transactions, we may also apply these amounts to our cash flow analysis to stress the adequacy of interest collections on the assets to cover these expenses For transactions that allocate extraordinary expenses to the most subordinate class outstanding (i.e., are deducted from total collections with no offset to the contractual amount of interest due on the securities), we apply these amounts to our cash flow analysis in the applicable stress scenarios During surveillance, we generally don't project extraordinary expenses once the transaction has been seasoned at least five years unless we observe significant extraordinary expenses being incurred. The criteria generally assume that any observed extraordinary expense amounts are credit-related, unless we have reasons to believe otherwise. WAC deterioration stresses 127. To address the potential for a pool's WAC to decline over time as higher coupon loans prepay or default, we stress the pool's projected cash flows by reducing the interest accrued on the assets. We review the distribution of loan coupons in the pool based on measures such as the standard deviation, interquartile range, and maximum/minimum ranges to assess the pool's homogeneity with respect to loan coupons Generally, the stress is based on the pool's WAC at the time of analysis versus 10 years later based on an assumed reduction in pool balance of 10% per year applied to the loans with the highest coupons. This WAC difference is the maximum WAC deterioration assumed for the pool. The stress applied starts at 0 in the transaction's first month and increases linearly each month to the maximum through year 10, at which point it remains constant at the maximum through the deal's remaining life. This stress is applied in all cash flow stress scenarios at all rating levels For highly seasoned pools at issuance or newly originated pools with short amortization terms (i.e, 15 or 20 years), our WAC deterioration stress may be based on the pool WAC at the time of analysis versus the assumed WAC at end of six to eight years rather than 10 years. We may apply an alternative WAC deterioration assumption, such as one based APRIL 24,

36 on a higher pool balance reduction rate, to transactions that are potentially more sensitive to WAC deterioration (e.g., transactions with greater variability in loan coupons or with high loan coupons). For deals with multiple collateral groups, we may apply different WAC deterioration assumptions for each group if the coupon rates and their distributions vary significantly. During surveillance, we may adjust our WAC deterioration stresses to reflect observed WAC deterioration on the pool under review or on a substantially similar pool. Interest rate stresses 130. For transactions in which the collateral or security coupons reference an index, we apply our interest rate criteria "U.S. Interest Rate Assumptions Revised For May 2012 And Thereafter," published April 30, For transactions that include collateral or security coupons that reference an index not addressed in the aforementioned criteria, we will apply other interest rate stress assumptions outlined in the aforementioned or other criteria that we believe adequately approximate the relevant risk. Liquidity stresses 131. Delinquency curve stresses Some loans can become delinquent and cure subsequently. Although these loans don't reduce the ultimate cash flow, they affect the timing of when cash flow is available. Therefore, to assess liquidity adequacy, we generally assume (in excess of the percentage of loans expected to default) temporary delinquencies equal to 20% of the loans assumed to default under our default timing curve in a given month would remain delinquent for three months (including the month in which it became delinquent) and then cure in the fourth month When testing liquidity, we generally will not apply a delinquency stress for transactions secured by prime collateral where the servicer is contractually obligated to advance through liquidation or for transactions secured by any collateral where the servicer is advancing a limited number (at least three months' worth) of delinquent scheduled payments on the loans because we believe the servicer obligation would address the liquidity risk. However, we may apply delinquency stresses to these transactions if we determine that servicer advance stresses are relevant as described below in the Servicer Advance Stresses section Servicer advance stresses If a transaction's documents do not contractually obligate the servicers to advance delinquent principal and interest payments or require only a limited number (e.g., four to six) of delinquent payments to be advanced, we typically do not assume any servicer advances in our cash flow analysis For transactions in which the servicer is contractually obligated to advance delinquent borrower principal and interest payments through liquidation, we typically apply servicer advance stresses (in the form of a haircut to the servicer advance percentage of 100%) in instances such as: Principal cannot be used to pay interest on the rated securities; A structure has multiple groups of senior securities supported by a single set of subordinate securities, and collections from any loan group cannot be used to pay interest due to the senior securities of another loan group before distributions are made to subordinate securities; The structure uses excess interest/overcollateralization for credit enhancement; or The underlying collateral is nonprime Where we incorporate them into our cash flow analysis, the servicer advance stresses will be applied to the loans projected to default as well as the loans projected to become delinquent and subsequently cure in all cash flow stress scenarios at all rating levels. The level of servicer advance stress will be a function of the pool's loss severity at the APRIL 24,

37 applicable rating level. The haircut to the servicer advance percentage will typically be equal to the pool's loss severity at a given rating level minus 20%. Triggers 136. Some transactions may include a delinquency or cumulative loss trigger that affects the payment priority. For these transactions, we typically analyze triggers as modeled per the transaction documents based on the application of our loss assumptions, delinquency, and default timing stresses at each rating level unless we deem it more appropriate to analyze the cash flows under alternate trigger scenarios to test elements of the payment structure. In cases where we don't apply a delinquency stress for liquidity purposes as described above, we nonetheless typically apply delinquency stresses for these transactions for purposes of testing triggers. Servicing and other transaction fees and expenses 137. Servicing fees The servicing fee is intended to compensate the servicer for servicing the securitized pool and be sufficient to incentivize a substitute servicer to be the primary servicer, if needed. For most RMBS transactions, servicing fees generally range from 0.25% to 0.50% per annum of the current pool balance, depending on the loans' credit quality. To the extent we believe the servicing compensation is below market standards or would not be sufficient to cover the servicer's servicing and collection expenses, we typically assume a higher servicing fee in our cash flow analysis. In addition to having a fee that's sufficient, the payment of the servicing fee is expected to be senior to other distributions in the payment waterfall so that the servicer will likely receive the fee on a periodic basis Other fees and expenses Our cash flow analysis also considers other transaction fees and expenses (excluding extraordinary expenses) that are netted from collections on the assets on a recurring or one-time basis. Such fees may include those of an independent reviewer or transaction manager, including any termination payments, among others. Depending on the nature of the fee, the fee payment structure, and other factors such as the collateral credit quality, we may stress for these amounts in our cash flow projections or assume such expenses are captured in the application of our extraordinary expense stresses as described above. This analysis is based on the concepts outlined in "Criteria Methodology Applied To Fees, Expenses, And Indemnifications," published July 12, Structural considerations 139. Available funds cap and deferred interest For transactions where securities' coupons are subject to an available funds cap (i.e., actual interest collections received), our ratings would reflect the stated payment terms generally if the difference between the interest accrued (at a measurable rate) and the actual interest paid, due to any available funds cap (i.e., the deferred amount), was required to be repaid together with accrued interest on the deferred amount. Otherwise, we may impute a promise as to the required interest amount due on the securities in our cash flow analysis to determine whether credit enhancement is sufficient to support the securities at a given rating level Similarly, for transactions that permit temporary interest deferrals on the securities, we would only rate to those terms if investors were compensated for the delayed interest payment (i.e., investors received interest on deferred interest payments). Otherwise, we will rate to timely payment of interest on the securities in our cash flow analysis to determine whether credit enhancement is sufficient to support the securities at a given rating level Loan modifications In addition to payment and cash flow mechanics, our ratings analysis also assesses certain structural provisions related to loan modifications. We generally expect a transaction to include the features outlined below to mitigate any potential adverse impact of a loan modification on cash flows to securityholders. If a transaction doesn't include these features, in the absence of other mitigating structural features, we may conduct additional analysis to determine whether additional credit enhancement can mitigate the associated risks at a given rating level or else cap or not assign a rating to a security: APRIL 24,

38 The term of a modified loan is not extended beyond one month before the legal final maturity of the transaction's rated securities; Capitalized advance amounts are reimbursed from principal collections on the related loan only or from principal collections on the pool, provided the principal distribution amount is also reduced by such capitalized reimbursement amounts and the reduction in principal due to securityholders is allocated in reverse sequential payment priority, as applicable, based on the payment allocation to the securities; and Loans that were modified in the preceding 12 months (or other applicable period set forth in the transaction documents) are included in the transaction's delinquency trigger, if applicable, even if such loans are not delinquent under the modified terms; Principal forbearance and forgiven amounts are recognized as a realized loss as soon as the modification becomes effective and are included in the cumulative loss trigger if applicable; Amounts recouped on forborne amounts are treated as subsequent recoveries and allocated to securities in the same priority as principal; and Forbearance amounts are excluded from the calculation of any overcollateralization amount, if applicable Other structural considerations Because the above concepts may not address every structural facet, we consider different structural features that may arise when evaluating a transaction in conjunction with the pool's credit quality, liquidity reserve funds to buffer cash flow disruptions, and other existing features. Depending on our assessment of such structural features, we may adjust our credit enhancement levels or cap the rating for a given security to account for any potential incremental risk. Minimum credit enhancement level 143. Our minimum credit enhancement level for RMBS transactions is 4% for 'AAA (sf)' ratings and 0.35% for 'B (sf)' ratings. The 4% floor credit enhancement level for 'AAA (sf)' ratings corresponds to 25x leverage. We believe that leverage above that level creates vulnerabilities inconsistent with the degree of creditworthiness associated with 'AAA (sf)' ratings. Moreover, the minimum credit enhancement levels can't be funded solely through soft enhancement, such as excess spread. Table 21 Minimum Credit Enhancement Levels Rating level Minimum CE (%) AAA 4.00 AA AA 2.83 AA A A 2.03 A BBB BBB 1.08 BBB BB BB 0.79 BB B B APRIL 24,

39 Table 21 Minimum Credit Enhancement Levels (cont.) Rating level Minimum CE (%) B CE--Credit enhancement. Credit enhancement floors 144. In the absence of other structural mitigants, such as sequential-pay structures that lock out principal distributions to subordinate securities, credit enhancement floors prevent credit enhancement from being released over the transaction's life to address back-ended credit and liquidity risk that may arise Where a credit enhancement floor is used to mitigate this risk, we typically evaluate the size of the credit enhancement floors for new transactions by comparing the proposed credit enhancement floor at each rating level with the amount needed to support a given maximum potential rating (see table 16) When assessing the credit enhancement floor for new ratings, we project each relevant (see table 16) loan's balance assuming a zero-prepayment amortization term until the time the transaction can begin paying unscheduled principal pro rata. We then determine what the loss severity would be for each loan, based on the amortized balances, and calculate the loss projection (see table 16). This projected loss is compared with the credit enhancement floor for each rating category Where a transaction does not include a credit enhancement floor to address the risk of back-ended losses, we will assess the adequacy of other structural mitigants to comparably provide for the maintenance of what we consider an adequate amount of credit enhancement over time. FHA/VA pools 148. Our analysis of loans insured or guaranteed by the FHA or the Department of Veterans Affairs (VA) generally incorporates the methodology and assumptions outlined herein, while additional analysis is applied as it relates to the insurance/guarantee associated with the loans. FHA/VA loans differ from traditional mortgages based on the insurance/guarantee on the loans, which can minimize the loss severity should the loans default. Therefore, our analysis of foreclosure frequency for FHA/VA loans typically does not vary from traditional mortgages. However, we adjust our loss severity assumptions, taking into account the nature of the insurance or guarantee (including preset schedules of payments under the contracts) as it relates to our criteria "Methodology For Assessing Mortgage Insurance And Similar Guarantees And Supports In Structured And Public Sector Finance And Covered Bonds," published Dec. 7, When doing so, we consider the likelihood that the insurance/guarantee would be paid according to both the insurer or guarantor's ability (according to S&P Global Ratings' financial strength rating on the entity) and willingness (based on adherence to conditions of the insurance or guarantee) to make the claims or guarantee payment based on the applicable rating scenario In addition to assessing the insurance and guarantee as described above, we also consider additional factors that could affect loss severity for loans insured or guaranteed by FHA or VA. In particular, these may include interest losses based on the loan accrual rate and the debenture rate, as well as foreclosure expenses and related taxes and insurance APRIL 24,

40 on the property combined with the assumed foreclosure/liquidation timelines. The debenture rate assumption would be based on historical observances as well as additional stresses that may be applicable for higher rating scenarios (to the extent the debenture rate is not already known). Our evaluation of these factors (including the willingness to pay the claim or guarantee) would be contingent on our assessment of the servicer's ability to manage related servicing practices as they relate to FHA and VA loans. Our assessment of the servicer would consider such factors as Ginnie Mae approval, servicing tenure for FHA and VA loans, and historical denial rates. As such, greater dependency on the servicing integrity to effectuate claims or guarantee payments places additional reliance on servicing replacement mechanisms as they relate to the securitization of loans insured or guaranteed by the FHA or VA For other loans that carry insurance or guarantees similar to FHA and VA loans, we would analyze the framework of such insurance or guarantees when assessing such loans in a comparable manner. Revolving pools 151. RMBS transactions backed by revolving pools, such as warehouse facilities, present incremental risk relative to term structures due to the dynamic nature of the collateral whose credit quality may fluctuate over time as assets flow in and out of the facility. For such transactions, we typically apply these criteria to conduct our credit and structural analysis. However, to account for the dynamic nature of the underlying pool of assets as well as structural differences relative to term RMBS, our criteria may allow for certain adjustments or employ additional features, as described further below, to analyze risks unique to revolving transactions To assess the credit risk of the collateral in revolving transactions, we consider the eligibility criteria for assets that may be added to the pool over time. Based on those parameters, we create hypothetical pools to stress for adverse credit attributes. Loss coverage on the pools is assessed through the LEVELS model. Outside-the-model adjustments such as those for reviewed originators/aggregators, third-party due diligence results, and R&Ws may also be applied where applicable. Additional adjustments may also be applied at the pool level as deemed appropriate, given the potential future variability of the assets With respect to third-party due diligence for revolving transactions, we would expect a review to be performed on the pool of loans backing the facility at the time of our initial analysis consistent with this criteria with respect to the sample size and scope of review. We will incorporate the results of the review into our loss coverage analysis consistent with the general criteria In addition, we would expect due diligence reviews to be conducted on the asset pool on a periodic basis consistent with this criteria with respect to the sample size and scope of review to confirm the subsequent assets meet the facility's eligibility criteria. We expect the sample size to reflect a statistically significant portion of the loans in the transaction at the time of the periodic review. We will consider structural provisions or other mitigants to address loans with material findings that are added over time If the initial or subsequent due diligence performed on the transaction is not consistent with the sampling methodology or scope of review that our criteria describe, we may make adjustments to our loss coverage levels based on our view of the robustness of the due diligence, when viewed as a whole We will also consider the impact of any structural features, such as dynamic advance rates, credit enhancement floors, APRIL 24,

41 early amortization and other performance-related triggers, and loan sale provisions, in our transaction analysis. Depending on our assessment of such structural features, we may adjust our credit enhancement levels or cap the rating for a given security to account for any potential incremental risk. Operational and administrative risks 157. Mortgage operational assessment In our view, mortgage loan performance may reflect certain qualitative aspects of an originator's or aggregator's operational framework, track record, and practices, including how they have changed over time. Therefore, we incorporate our assessment of the quality of the transaction's originators or aggregator based on our evaluation of its management and organization, origination process and underwriting/loan purchase and aggregation, and internal controls into our loss coverage analysis as described below. When available, we will also assess historical loan performance Originators and aggregators with MOA rankingsif an originator or an aggregator has an S&P Global Ratings residential MOA ranking, we will typically consider the MOA ranking and related loss coverage adjustment factor in our ratings analysis. In limited circumstances, we may adjust the related loss coverage adjustment factor for a ranked originator or aggregator used in our transaction analysis when we are aware of new, relevant information that was not available or known when the MOA was conducted Transaction-specific reviews For any unranked originator, we will generally determine whether we need to perform a transaction-specific MOA review based on the percentage (by cut-off balance) of the unranked originator's loans (generally 20%) in a transaction and the quantity and quality of due diligence performed by an S&P Global Ratings-reviewed third-party due diligence firm on such loans. We will also consider whether we have reviewed the aggregator if such originator's loans are being securitized by an aggregator We generally will complete a transaction-specific MOA review on unranked aggregators who acquire and securitize loans from multiple originators (unless 100% of the loans are from originators with MOA rankings) S&P Global Ratings will determine the relevant scope for transaction-specific reviews, which may include aspects of qualitative and quantitative reviews (for additional details regarding the qualitative or quantitative review for originators/aggregators, please see "Request For Comment: U.S. Residential Mortgage Operational Assessment Ranking Criteria," published April 24, 2017). Transaction-specific MOAs will result in a loss coverage adjustment factor, which will typically only be applicable to the specific transaction for which the transaction-specific MOA review was conducted. We will not assign an MOA ranking in connection with a transaction-specific review Seasoned loans For seasoned loans and reperforming loans, in our view, historical loan performance may better predict future performance than qualitative aspects of an originator's or aggregator's operational framework, track record, and practices. As such, if the loans in a transaction are seasoned, we typically will not consider an MOA ranking or may not perform a transaction-specific MOA review of the originator or aggregator when conducting our ratings analysis Loss coverage adjustment factor Based on our MOA rankings or the results of our transaction-specific MOA reviews, we will typically determine a loss coverage adjustment factor. The loss coverage adjustment factors range from 0.70x to 1.30x or greater. For transaction-specific MOA reviews, however, we will generally not apply a loss coverage adjustment factor less than 0.90x. For originators or aggregators with a limited operating history or limited historical loan performance, or for originators or aggregators that provide incomplete information or do not provide the requested information in a timely manner, we may adjust our related loss coverage adjustment factors to reflect such risks and uncertainty. APRIL 24,

42 164. Only one loss coverage adjustment factor is applied to the portion of the pool associated with a given originator or aggregator. For any originator where we have an MOA ranking, we will typically apply the related loss coverage adjustment factor to the loans produced by such originator. If we performed a transaction-specific originator review, we will usually apply the related loss coverage adjustment factor to the portion of loans from such originator. We will typically apply a loss coverage adjustment factor based on the results of our MOA ranking or transaction-specific MOA review of the aggregator to the portion of the remaining loans in a transaction whose originators have not otherwise been ranked or reviewed. In certain circumstances, based on our view of the aggregator's influence on the origination process, we may apply the loss coverage adjustment factor from the results of our MOA ranking or transaction-specific MOA review of the aggregator to all loans in the pool regardless of whether we have an MOA ranking for any of the underlying originators. Third-party due diligence 165. We expect transactions to include third-party due diligence reviews of the securitized loans to provide insight into the quality and validity of the data used to originate the mortgages as well as to inform our analysis of the collateral pool Our criteria address four primary considerations: the diligence provider's ability to perform the review, the selection of the loan population, the scope of the review, and the impact of the review results on our loss coverage levels, if any We expect the diligence provider to be a firm that has been assessed by S&P Global Ratings to have adequate processes, procedures, and systems to carry out the review. Assessments will be updated periodically We expect the review to be conducted on at least a statistically significant random sample of the population to be securitized. We may request additional sampling, random or adverse, on a case-by-case basis depending on the findings identified through the review process or the nature of the collateral being securitized We expect the review to generally address four areas: The quality of the data submitted; Adherence to the lender's underwriting guidelines; Substantiation of the property value; and Compliance with regulatory and legal requirements For seasoned and reperforming loans, the scope of the review may differ to address concerns specific to seasoned collateral title and documentation issues. Furthermore, certain aspects of the four areas may not apply, such as how applicable it is to review underwriting guidelines We will incorporate the results of the review into our analysis and may adjust loss coverage levels to reflect the perceived incremental risk to the transaction posed by the findings to the transaction. An adjustment factor of 1.30x is applied to loss coverage at each rating level to account for the percentage of loans (by count) with material exceptions, which is extrapolated to the unreviewed portion of the pool, if applicable. For example, if 10% of the loans in the sample were graded C overall for credit or property valuation findings, we would apply the adjustment to 10% of the unsampled portion of the pool if only a sample of the pool was reviewed For additional details regarding third-party due diligence, please refer to Appendix IV. APRIL 24,

43 R&Ws 173. R&Ws and associated repurchase obligations are intended to help align the interests of sellers with those of investors by providing a disincentive to deliver defective or misrepresented loans into a transaction We differentiate our analysis among transactions based on the variation impact of three key aspects: The content of the R&Ws relative to our benchmarks; The R&W provider's financial condition; and The remedies and enforcement mechanism upon R&W breach Mortgage loan originators and sellers into securitizations generally provide loan-level R&Ws through various purchase agreements as of the closing date. If R&Ws are made as of a date before the closing due to the timing of loan purchases, we typically expect gap reps (see Appendix V) to cover the gap period (i.e., from the date the loans were purchased to the transaction's closing date) R&W adjustment factor Based on our holistic review of the R&W framework provided, we may apply an adjustment factor to the transaction's loss coverage levels ranging from 0.95x to 1.15x or higher. An adjustment to loss coverage levels may be applied as we review the R&W framework and evaluate: The inclusion of R&Ws that address the risks listed in Appendix V, including the existence, or lack, of knowledge qualifiers with curative language, sunset provisions, and gap coverage; The relative financial strength of the R&W provider (e.g., an investment-grade entity as opposed to a speculative-grade or unrated entity), as well as the existence of any party that guarantees or supports the obligation of the R&W provider to repurchase (back-stop provider); and The clarity and effectiveness of the remedy framework, including what triggers a loan review, breach determination, and the mechanisms to enforce repurchase We also consider the level and scope of due diligence on the mortgage loan collateral and the materiality of findings, which may result in reduced reliance on the R&W framework Some other elements we may consider when applying an adjustment factor include the diversity of underlying R&W providers by pool exposure and the degree of retained risk (i.e., the R&W provider's retention of subordinate securities within the transaction) If a transaction's R&W framework has material weaknesses without mitigating factors, we may cap, or not assign a rating, if, in our view, the framework introduces unquantifiable risk to the transaction. Conversely, because R&W frameworks may evolve over time, we will consider variations that may be presented on a case-by-case basis, even if such elements are not described herein. We expect that R&Ws will be made directly to the issuer or legally pledged to a transaction for securityholders' benefit. If this is not the case, we may consider other structural mechanisms that provide comparable benefit to securityholders in the event of performance issues attributable to loan origination defects and other factors typically addressed by R&Ws R&Ws and early payment default provisions Appendix V contains a list of risks we expect R&Ws to address, which we asses in our ratings analysis. To the extent that transactions' R&Ws do not address the risk targeted by this framework, S&P Global Ratings will evaluate the provisions provided on a case-by-case basis, taking into account any mitigating factors where relevant. APRIL 24,

44 181. We also expect an early payment default (EPD) provision from the R&W provider to repurchase any mortgage loans for which the borrower fails to submit any of the first three payments due after loan origination unless the R&W provider concludes that the delinquency occurred because of a servicing issue that has subsequently been corrected. With respect to EPD provisions included in underlying sale agreements, we review whether these particular provisions are assigned for the issuer's benefit. We would not expect such a provision to include loans for which the borrower has made the first three payments before the securitization closing date The R&W provider's financial condition We consider the ability of the R&W provider to fulfill its repurchase obligations in our analysis. Depending on the R&W provider's financial condition, we may increase loss coverage levels at one or more rating stress levels to account for potential failed repurchases for R&W breaches. We may review the available history of past repurchase requests to proposed R&W providers and their responses to such requests. In addition, we may consider the due diligence performed on the collateral, which may result in less reliance on the R&W provider's financial condition. We may also rely less on the financial condition when analyzing seasoned pools because the repurchase risk for a breach of certain R&Ws becomes increasingly remote Remedy and enforcement mechanisms upon breach of triggers Our analysis considers a transaction's mandatory review provisions for delinquent or defaulted loans, related triggers, and the ensuing enforcement mechanisms We also consider whether a breach of an R&W triggers a remedy in the event that such breach causes a material and adverse impact on the interest of the securityholders. The clarity and strength of each R&W and associated remedy is an integral part of our review. When repurchase is the remedy, we expect the repurchase price to cover the unpaid principal balance and accrued and unpaid interest as of the repurchase date, as well as any costs, damages, and penalties in connection with the related mortgage loan borne by the trust In addition, while the R&W provider may make an R&W based on its actual knowledge of such risk factor, we expect a remedy to be provided should such assertion prove to be inaccurate, notwithstanding the R&W provider's lack of actual knowledge. We expect that parties making R&Ws with knowledge qualifiers stand behind the accuracy of the information that they provide, notwithstanding actual knowledge. Seasoned and reperforming loans 186. Certain R&Ws apply to aspects of the mortgage loan origination. For seasoned and reperforming loans, in our view, performance history, as opposed to the underwriting practices and guidelines of the originator, may better indicate the future performance of the loan. Thus, we typically do not expect certain R&Ws to be provided for transactions backed by seasoned or reperforming loans as indicated in Appendix V. (For R&Ws for seasoned and reperforming loans, seller refers to the seller that sells the loans to the issuer.) 187. Additionally, we may weigh the R&Ws' effect on loss coverage levels differently depending on loan seasoning. For example, in our view, a strong representation regarding fraud considerations may be less of a concern if the borrower has a significant payment history. Transaction manager (or deal agent) 188. To help determine the R&W adjustment factor for a transaction, we will also consider the role of a transaction manager, deal agent, or similar party, if present, in the transaction. Our analysis will take into account the transaction manager's: APRIL 24,

45 Responsibilities within the transaction; Operational adequacy to perform relevant tasks; and Financial capacity to continue to operate in a stressed environment We will periodically assess the operational quality of any transaction manager, taking into consideration several areas including senior management, staffing and training, internal controls, and quality control processes to ensure sufficient experience, systems, and infrastructure are in place to carry out its role satisfactorily. LEGAL AND REGULATORY RISK Anti-Predatory Lending (APL) Identification of APL loans 190. We expect the seller into a securitization or to the securitization issuer as applicable to identify to the best of their knowledge, on the mortgage loan schedule or other loan-level file submitted to S&P Global Ratings for review in connection with the transaction, whether a mortgage loan to be included in a securitized pool is, at the time of the origination, a "high-cost" loan, "covered" loan, or any other similarly designated loan (an APL loan) as defined under any state, local, or federal law with respect to predatory and abusive lending laws (APL laws) We typically review the third-party due diligence results to gain comfort that any reviewed APL loans have been identified and do not violate the applicable APL law(s). For seasoned loans with insufficient documentation in the loan file to test for compliance with the applicable APL law(s), we will assess presumed compliance based on the findings for the loans that could be reviewed. If seasoned loans that cannot be reviewed due to insufficient documentation may be subject to APL laws that do not cap assignee liability as described further below, we may decline to rate the transaction if such loans are included in the final securitized pool In addition, we typically expect a seller into a securitization to provide R&Ws stating that all loans were originated in compliance with all applicable laws, including, but not limited to, all applicable anti-predatory and abusive lending laws, as well as to repurchase or otherwise cure any breaches subsequently identified on affected loans. APL loss coverage adjustments 193. Generally, we do not expect to increase loss coverage with respect to APL loans originated in compliance with APL law(s) included in a securitized pool However, if the transaction includes a significant percentage of APL loans, depending on the facts and circumstances of the specific transaction, we may increase loss coverage. This determination is generally based on an evaluation of several factors, including, the percentage of APL loans in the pool, the concentration of the APL loans in the pool in specific jurisdictions, and the specific provisions of the APL laws in those jurisdictions, including potential assignee liability and any relevant exemptions. APL loans with uncapped liability 195. If the assignee liability for damages for a violation of an APL law is not capped, we will not be able to size the potential liability into our analysis. Thus, we will not rate transactions that include APL loans subject to APL laws that do not APRIL 24,

46 cap assignee liability Therefore, we expect the seller into a securitization to make R&Ws that no loans included in the transaction are subject to an APL law that has uncapped assignee liability for violations of such APL law. See Appendix V for a list of R&Ws covering the jurisdictions whose APL laws, in our view, have uncapped liability. Surveillance 197. Our analytical approach to rating outstanding U.S. RMBS transactions that fall within the scope of these criteria reflects our view of change in the credit risk inherent in a mortgage pool over time. Our analysis incorporates observed pool performance relative to our expectations and reflects our forward-looking view of performance relative to the credit enhancement available at a given rating level. We begin by applying the new issuance methodology and assumptions described in these proposed criteria; however, this is only one component of our surveillance analysis. We also consider the collateral's performance and dynamic characteristics (including expectations around the timing of peak losses, structural provisions, and credit enhancement levels) to determine if updates to any such assumptions are warranted that could potentially result in rating changes. As a result, actual rating actions may reflect adjustments to model-derived rating recommendations given our consideration of these other factors At the time of issuance, a pool of mortgage loans will typically have very little, if any, observed performance data available. As such, our new issuance analysis primarily infers future collateral performance primarily from the loan and borrower characteristics at the time of origination. As a pool of mortgage loans seasons, and default and loss patterns begin to emerge, those patterns will be considered in determining the assumptions we will apply in our surveillance analysis. For example, during surveillance, we will analyze projected foreclosure frequency based on the original loan attributes as well as the observed payment history. Further, the performance history will inform the qualitative factors that were assessed at the time of our initial rating and how relevant any initial adjustment remains given our outlook for the collateral pool at the time of surveillance. For example, we may neutralize the MOA loss coverage adjustment factor applied in our analysis over time as performance history evidences the quality of the originations in the pool In limited circumstances, we may not stress cash flows where a cash flow analysis may not provide additional insight to the sufficiency of the credit enhancement or the liquidity risk inherent in the transaction. However, to the extent we are unable to access information pertinent to our analysis that is sufficient, reliable, and timely, we may decide not to maintain a rating according to our policies. APPENDIXES Appendix I: Changes From Our Current Criteria 200. In both the current and proposed criteria, the primary drivers of projected default and differentiator of credit quality are CLTV and credit score (FICO) (see tables below for comparison of the existing criteria to the proposed). In the proposed criteria, for an archetypal loan, loss severities are, on average, slightly higher at the extreme stress level APRIL 24,

47 ('AAA') and somewhat lower at the benign stress level ('B') (see chart 9 below for several examples). This is primarily driven by an increase in foreclosure/liquidation timelines at the 'AAA' level and a reduction in timelines at the 'B' level. The timelines at the other rating levels are interpolated between 'AAA' and 'B'. Another factor for lower loss severities at certain ratings levels ('A' and below) under the proposed criteria is lower MVD assumptions. This comparison assumes that properties in given MSAs are neither over-valued nor under-valued according to such framework under the proposal The proposed criteria also recalibrate a number of our cash flow stress assumptions. In particular, the length of the default timing curves under the proposed criteria are longer compared to the current criteria, which can result in the application of back-ended losses later in a transaction's life when subordination may have amortized down (depending on the structural mechanics). This can result in higher levels of credit enhancement depending on the priority a class may have in the payment and loss allocation waterfall The qualitative factors for originator/aggregator, due diligence, and R&Ws are not changing significantly, and therefore are not expected to have a large impact compared to our current application under the proposal. Table 22 'AAA' Scenario Foreclosure Frequency Percentages: Proposed/Current* CLTV / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /44.96 *All figures for current criteria are rounded to the nearest hundredth. Current criteria foreclosure frequency percentages assume primary occupancy. Foreclosure frequency could exceed 100%, but is capped at 100% after adjusting for other credit attributes. CLTV--Combined loan-to-value ratio. Table 23 FICO 'B' Scenario Foreclosure Frequency Percentages: Proposed/Current* CLTV / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /0.88 FICO APRIL 24,

48 Table 23 'B' Scenario Foreclosure Frequency Percentages: Proposed/Current* (cont.) CLTV / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /5.00 *All figures are rounded to the nearest hundreth. Current criteria foreclosure frequency percentages assume primary occupancy. CLTV--Combined loan-to-value ratio. Chart 9 FICO Table 24 Overview Of Foreclosure Frequency Adjustment Factors: Current/Proposed Variable Current Proposed CLTV-FICO * Number of borrowers N/A APRIL 24,

49 Table 24 Overview Of Foreclosure Frequency Adjustment Factors: Current/Proposed (cont.) Variable Current Proposed Loan term Seasoning DTI Occupancy Varies with CLTV Property type FF Loan purpose Deposit money verification Documentation type Loan type fixed/arm/hybrid Loan type IO Loan type balloon Loan type - negam Borrower residency status Delinquency status/pay history FF FF Second lien N/A 1.30 *In the current criteria, the adjustments for Occupancy, CLTV, and FICO are layered together. The current CLTV-FICO factor assumes a primary occupancy home. FF--Foreclosure frequency. N/A--Not applicable. DTI--Debt-to-income ratio. CLTV--Combined loan-to-value ratio. ARM--Adjustable-rate mortgage. IO--Interest only. Negam--Negative amortization. Appendix II: Loss Severity Adjustments For ATR And QM Standards 203. We made certain assumptions for defensive claim and successful litigation probabilities and related loss severity adjustments for certain loans covered under TILA's ATR rule. The ATR rule provides for potential liabilities and protections to lenders and assignees depending on a loan's characteristics and underwriting To assess the incremental risk posed to the trust by covered loans, we first consider the loan's expected designation, or TILA status. One of five possible TILA status designations (as identified by the applicable originator/aggregator/due diligence firm) may apply to a mortgage loan, each carrying different levels of risk. The TILA status designations are: QM/non-HPML (or "safe harbor"), QM/HPML (or "rebuttable presumption"), non-qm/atr compliant, non-qm/non-atr compliant, and not covered/exempt. S&P Global Ratings expects non-qm/noncompliant loans to be ineligible for rated U.S. RMBS transactions as these loans do not comply with the law For covered loans that are designated as QM/non-HPML, we do not make any further loss severity adjustments to our analysis. While we recognize that safe harbor protections under the rule for QMs that are not HPML do not prohibit a borrower from bringing a claim or ultimately being successful, in our view, if a borrower does bring a claim, the additional risk for QM/non-HPML loans is de minimis For covered loans that are not designated as QM/non-HPML, we adjust loss severity to account for the potential liability associated with defensive claims. In conjunction with the loan's default likelihood, we assign a probability that a borrower whose property is in foreclosure will file a defensive claim (defense probability) and a probability that the APRIL 24,

50 defensive claim is successful (success probability). We also consider the corresponding increase to the foreclosure timeline related to filing of the defensive claim on the loan (see chart 10). Default probability 207. Potential additional losses to the securitization trust may be incurred only if a loan defaults. However, we assume the ATR rule will not impact foreclosure frequency and only affect loss severity. Therefore, our criteria for determining foreclosure frequency will not be further adjusted for the impact of the ATR rule. Defense claim probability 208. A borrower's decision to bring suit as a defense to foreclosure depends on unique factors such as financial means, the loan's TILA status, and state foreclosure law (judicial/non-judicial): all of which may affect the borrower's cost/benefit analysis in determining whether to bring a claim under the ATR rule. Table 25 provides our assumptions for the probability a borrower will assert a defensive claim. Table 25 Defense Claim Probability Judicial State Vs. Nonjudicial State* Loan type Nonjudicial state (%) Judicial state (%) QM/HPML Non-QM/compliant *The determination of whether a state is considered judicial or nonjudicial is based on Fannie Mae's (or another entity's as deemed appropriate) designations and will be updated periodically to reflect relevant changes. QM--Qualified mortgage. HPML--Higher-priced mortgage loans. APRIL 24,

51 Successful claim probability 209. Given the lack of available data, we considered a Consumer Financial Protection Bureau (CFPB) analysis for a non-qm loan that assumed a borrower success rate of 20%. Using the non-qm success data for a QM/HPML is a measure of conservatism given the lack of empirical data on claims under the ATR rule. In determining success probabilities, we consider loan characteristics similar to those reviewed for calculating defense probabilities. Further, we assume that a property's location, whether in a judicial or a non-judicial state, will not have any impact on success rates. Table 26 below provides our assumptions for the probability a claim will be successful. Table 26 Successful Claim Probabilities Loan type Success probability (%) QM/HPML Non-QM/compliant QM--Qualified mortgage. HPML--Higher-priced mortgage loans When assessing both the probability of initiating a claim and the success of that claim, originator/aggregator underwriting practices and procedures that contribute to compliance with the ATR rule are important considerations. For loans for which the originator's and/or aggregator's processes present additional risks for compliance under the ATR rule, we may reflect that risk by increasing the adjustment factor applied to loss coverage associated with the relevant MOA ranking or transaction-specific MOA review. Expected loss adjustment 211. We assess the loss related to assignee liability for a successful claim using assumptions for each of the damages and costs owed to the borrower that are provided for under the ATR rule. With a successful claim, the borrower may be entitled to damages and costs by way of recoupment or set-off against the property's value. These damages could decrease the proceeds from the property's liquidation, resulting in increased loss severity. The potential damages and costs under the ATR rule include statutory damages, special statutory damages, costs, reasonable attorney's fees, and actual damages. Note, in the case of special statutory damages, we assume a three-year exposure period, which is consistent with the rule's three-year statute of limitation from the date of the violation. In addition, we assume there will be no actual damages beyond those listed in the table below, given our opinion that incremental damages beyond those already noted would be difficult to prove. Table 27 outlines our assumptions regarding potential damages and fees. Table 27 Damages And Fees Damages and fees Assumed cost Statutory damages $4,000 per successful claim Special statutory damages Actual points and fees ($) + (coupon/12 x original balance x 36 months) Costs and attorney fees* $30,000 *We may update these costs and fees over time based on changes in observed market data and the impact of inflationary pressures, if warranted. APRIL 24,

52 Additional liquidation time 212. We assume that loans subject to the ATR rule may also experience additional losses regardless of whether the claim is successful since a defensive claim may prolong the foreclosure process. Therefore, we generally assume that a borrower's defensive claim will further increase the foreclosure timeline by approximately six months. Appendix III: Pool-Level Adjustments Loan count 213. The primary purpose of the small-pool adjustment factor is to capture idiosyncratic risks and sample bias that may exist in small loan pools (less than 250 loans). When applicable, the RMBS small-pool adjustment factor will typically follow a functional form (see the equation 2 below) to reflect the greater risk from higher borrower concentration. The small-pool loss coverage adjustment factors are displayed in chart 11 below. APRIL 24,

53 Chart 11 Geographic concentration 214. The geographic concentration adjustment factor captures the additional risk introduced by RMBS transactions that are more geographically concentrated than our historical mortgage dataset on which the adjustment factors related to mortgage and borrower characteristics are based. When applicable, the RMBS geographic concentration adjustment factor will typically follow the functions outlined below Our analysis begins with the pool-specific Herfindahl Index (equation 3). APRIL 24,

54 216. We then subtract the baseline Herfindahl Index from our pool-specific Herfindahl Index (see equation 4) Next, we apply a multiplicative factor to the difference in the Herfindahl indices. This multiplicative factor ranges between 1 and 2 and is based on the 'AAA' loss coverage (before the application of any pool-level adjustments), represented by 'L' (see equation 5). APRIL 24,