OVERVIEW AND SCOPE

1. S&P Global Ratings is requesting comments on its proposal to expand the scope of its global methodology framework set out in "Criteria: Global Methodology And Assumptions: Assessing Pools Of Residential Loans," published Jan. 25, 2019, to include Spain, Portugal, Italy, and Greece. If the proposal is implemented as proposed, we would use this global methodology framework to analyze residential mortgage pools backing both covered bonds (CB) and residential mortgage-backed securities (RMBS) in these four jurisdictions. This would mean that we would no longer use our European methodology framework (see "Methodology And Assumptions: Assessing Pools Of European Residential Loans," published Aug. 4, 2017) for analyzing residential loans backing Spanish, Portuguese, and Italian RMBS and covered bonds. We also would no longer use the analytical approach we established using our "Principles Of Credit Ratings" to analyze Greek covered bonds and RMBS (see "Credit FAQ: How We Analyze Residential Mortgage Loans Backing Greek Covered Bonds And RMBS," published Aug. 29, 2018). It would also mean that our cash flow approach would move under our global cash flow criteria (see "Global Framework For Cash Flow Analysis Of Structured Finance Securities," published Oct. 9, 2014).

2. Along with this request for comment (RFC) publication, we are publishing our current "Criteria: Global Methodology And Assumptions: Assessing Pools Of Residential Loans" in Appendix I, and our proposed guidance for applying these global criteria in Spain, Portugal, Italy, and Greece in Appendix II.

SUMMARY OF THE PROPOSAL

3. Other than a change in scope, shifting Spain, Portugal, Italy, and Greece into the global framework will cause no other changes to "Criteria: Global Methodology And Assumptions: Assessing Pools Of Residential Loans" (Appendix I).

4. The proposed shift in approach to applying our global framework to Spanish, Portuguese, Italian, and Greek residential loans instead of our "Criteria: Methodology And Assumptions: Assessing Pools Of European Residential Loans," published Aug. 4, 2017, or of the analytical approach we established using our "Principles Of Credit Ratings," as applicable, will lead to some changes in our rating analysis of Spanish, Portuguese, Italian, and Greek RMBS and covered bonds. Even though the frameworks described by the two criteria articles are similar, there are some differences. The key differences are highlighted below.

5. Under the global framework, the archetypal Spanish, Portuguese, Italian, and Greek foreclosure frequency assumptions will be subject to our Mortgage Market Assessments (MMAs). The output of the MMA application will be published in an accompanying guidance document (see Appendix II).

6. In applying our MMA framework, we would make limited adjustments to the base foreclosure frequency assumptions of the archetypal pool at certain rating levels for Italy and Greece.

• For Italy, we would lower slightly the 'B' foreclosure frequency assumption to be in line with that of Spain and we would update the interpolation between the 'B' and 'AAA' foreclosure frequency anchors consistently with the interpolation used for most other countries analyzed under the global framework.
• For Greece, we would lower the 'B' and 'AAA' foreclosure frequency assumptions of the archetypal pool to reflect the fact that, notwithstanding the impact of the Coronavirus pandemic, overall economic conditions and certain structural features, especially in relation to the foreclosure process, have improved since our 2018 calibration of the rating to principles factors. This will also result in an update of assumptions at intermediary rating levels, due to interpolation between the 'B' and 'AAA' anchors.

7. The proposed application of our global framework will also result in changes to certain assumptions currently used to rate Spanish, Portuguese, Italian, and Greek RMBS and covered bonds that reflect our views of other structural market changes, as well as increased data availability in relation to performance history. The proposed application will be published in an accompanying guidance document (Appendix II). To assess the credit quality of the pledged assets in a loan portfolio securing covered bonds, we generally use the securitization criteria for the type of assets pledged to the covered bonds. However, we may adapt our approach in certain instances, given the relative weight that credit analysis plays in the rating analysis of covered bonds compared to our assessment of the program's asset-liability mismatch.

8. The Spanish, Portuguese, Italian, and Greek residential markets are full-recourse markets, and under the global framework, we would adopt a type 2 LTV curve for the default frequency computation, consistent with other similar European residential markets. The shape of the type 2 curve is similar to the one currently employed in the four countries; although at higher LTV levels, it starts to flatten and ultimately levels off. In addition, under the global framework, we would employ a LTV definition that accounts for the current LTV (CLTV) as well as the original LTV (OLTV), while our current approach under the European criteria and the analytical approach established under Principles of Credit Ratings for Greece only considers the OLTV.

9. We are also updating some of the other adjustment factors for variations from the archetype that are specific to Spanish, Portuguese, Italian, and Greek pools of residential loans.

10. For example, for the pools of any of the four countries, we will update the way we apply the geographic concentration. Additionally, we may now reduce adjustments to broker-originated and second-home loans, if we receive evidence of consistent performance of those loans.

11. In pools of residential loans, we would apply adjustments for reperforming loans that consider the time since a loan was last in arrears rather than focusing on the restructure type, based on our assessment of the performance driver for this cohort. Additionally, in Spanish pools, we are proposing to include a new adjustment for unemployed borrower loans. Furthermore, in Spanish and Portuguese pools, we are proposing a separate bridge loans adjustment aligned with our second-home loans adjustment.

12. Our cash flow analysis for Spanish, Portuguese, Italian, and Greek RMBS and covered bonds is currently consistent with "Global Framework For Cash Flow Analysis Of Structured Finance Securities," published Oct. 9, 2014. Under our global framework, we will set out the application of the global cash flow criteria for Spanish, Portuguese, Italian, and Greek RMBS and covered bonds, where applicable, in the accompanying guidance document. Compared to the current approach, we would update certain assumptions: for example, we would align the default timing allocation with the one applied in other European countries under the global framework, which more appropriately reflects our expectation of the evolution of defaults. In addition, we would, in certain circumstances, reduce our servicing fee assumptions, based on our assessment of replacement costs, or the issuer's likely ability to perfect a servicing transfer. Finally, we are proposing a reduction in the foreclosure timing assumption for residential properties in the south of Italy and for commercial properties in the north and center of Italy and in Spain.

IMPACT ON OUTSTANDING RATINGS

13. The total outstanding portfolio of ratings that are subject to the proposed expansion of scope of these global criteria consists of 424 RMBS ratings in 148 transactions, 13 CB programs, and 10 Multicedulas transactions, as follows:

• Spain: 81 RMBS transactions (277 tranches), eight CB programs, and 10 Multicedulas transactions;
• Portugal: 20 RMBS transaction (50 tranches);
• Italy: 47 RMBS transactions (97 tranches) and one CB program; and
• Greece: four CB programs.

14. As to the anticipated impact of the proposed expansion in scope of these global criteria to RMBS and covered bond ratings in Spain, Portugal, Italy, and Greece, our testing presently indicates the following:

• We do not expect any rating impact on our CB programs or Multicedulas transactions.
• For Spain, we expect that approximately 87% of the RMBS ratings will remain unchanged and that approximately 4% of RMBS ratings may be upgraded by one notch. We expect that approximately 9% of RMBS ratings may be upgraded by up to three notches.
• For Portugal, we expect that approximately 90% of the RMBS ratings will remain unchanged and that approximately 10% of RMBS ratings may be upgraded by up to three notches.
• For Italy, we expect that approximately 90% of the RMBS ratings will remain unchanged and that approximately 3% of RMBS ratings may be upgraded by one notch. We expect that approximately 7% of RMBS ratings may be upgraded by up to three notches.

15. In addition to some of the changes of our base foreclosure frequency assumptions at certain rating levels in Italy and Greece, most of the expected upgrades are due to the shift, when assessing default frequencies, to a combined OLTV/CLTV definition from an only OLTV definition. Certain transactions may also be positively affected by the revised default timing allocation, as well as the proposed reductions in the servicing fee assumptions in Spain, Portugal, and Italy.

16. This analysis of the anticipated impact on outstanding ratings is 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 or program.

QUESTIONS

17. S&P Global Ratings is seeking responses to the following questions, in addition to any other general comments on the proposal:

• What are your views on the proposed expansion of the scope of the global criteria to include Spain, Portugal, Italy, and Greece (see Appendix I for the criteria)?
• Are there any other factors that you believe we should consider that are not already noted in this proposal?

RESPONSE DEADLINE

18. We encourage interested market participants to submit their written comments on the proposed criteria by Nov. 30, 2020, to http://www.standardandpoors.com/en_US/web/guest/ratings/rfc 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 www.standardandpoors.com/en_US/web/guest/ratings/ratings-criteria/-/articles/criteria/requests-for-comment/filter/all#rfc. Comments may also be 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.

PROPOSED METHODOLOGY AND ASSUMPTIONS

Criteria Framework

19. The current "Global Methodology And Assumptions For Assessing Pools Of Residential Loans" criteria (reproduced below in Appendix I) will remain unchanged except for the addition of Spain, Portugal, Italy, and Greece to the scope. For the proposed guidance for applying these global criteria in Spain, Portugal, Italy, and Greece see Appendix II.

APPENDIX I – Current Global Methodology And Assumptions: Assessing Pools Of Residential Loans

OVERVIEW AND SCOPE

20. S&P Global Ratings' methodology and assumptions for assessing pools of residential loans describes its approach to rating residential mortgage-backed securities (RMBS) and covered bonds globally. These criteria follow "Request For Comment: Global Methodology And Assumptions: Assessing Pools Of Residential Loans (Sweden, Norway, Finland, And Denmark)," published Nov. 6, 2019. For a comparison of the changes between the Request For Comment and the final criteria, see "RFC Process Summary: Global Methodology And Assumptions: Assessing Pools Of Residential Loans (Sweden, Norway, Finland, And Denmark)," published July 10, 2020. See the Related Research section for previous RFC and RFC Process Summary articles related to these criteria.

21. These criteria apply to residential mortgage-backed securities (RMBS) and covered bonds globally, except for jurisdictions (listed in Appendix III) where S&P Global Ratings currently has separately published criteria, and for jurisdictions where an analytical approach has been established using "Principles of Credit Ratings," published Feb. 16, 2011, on RatingsDirect. If such jurisdictions were to become in scope in the future, we would first seek comments via a request for comment (RFC) publication.

22. These criteria supersede and partially supersede the articles listed in Appendix II as they become effective in each applicable market. Some markets require prior notification to and/or registration by the local regulator, where the criteria will become effective when so notified by S&P Global Ratings and/or registered by the regulator. These criteria are accompanied by a guidance article, which sets out considerations for our application of the criteria in jurisdictions where it is most commonly applied (see "Guidance: Global Methodology And Assumptions: Assessing Pools Of Residential Loans," published Jan. 25, 2019.)

23. The criteria framework for analyzing residential mortgage credit risks in individual jurisdictions includes:

• A defined archetypal pool against which other mortgage pools are compared;
• The assessment of mortgage risk based on loan attributes and borrower characteristics relative to the archetype; and
• The calculation of a repossession market value decline to estimate recoveries.

24. We also use the Mortgage Market Assessment (MMA) to calibrate the foreclosure frequency assumptions associated with the archetypal pool of a given country. The MMA provides a relative risk ranking of mortgage markets across jurisdictions based on macroeconomic and mortgage industry considerations.

METHODOLOGY AND ASSUMPTIONS

Criteria Framework

25. The framework for our analysis of securities backed by residential mortgage loans considers the risks associated with the credit quality of the underlying assets. For the analysis of payment structure and cash flow mechanics, operational and administrative risk, counterparty risk, and legal and regulatory risk, see the Appendix.

26. Charts 1A and 1B below summarize the approach used for our credit analysis under these criteria.

Credit quality of the underlying assets

27. To assess the credit quality of the underlying assets, we determine projected losses for each pool of residential mortgage loans under conditions of stress commensurate with certain defined issue ratings. Projected losses are the product of foreclosure frequency and loss severity:

• Foreclosure frequency measures the probability of default. It represents the proportion of loans in a pool that are likely to go into foreclosure.
• Loss severity measures loss-given default, expressed as a percentage of the loan balance. It refers to the loss on foreclosure (i.e., the amount by which a loan balance and foreclosure costs exceed the property sale proceeds and other proceeds [e.g., lender's mortgage insurance]).

28. To determine projected losses, the framework establishes an archetypal pool of residential mortgage loans for a given jurisdiction based on global and country-specific considerations, in which we associate foreclosure frequency assumptions to each rating level. We apply adjustment factors for variations from the archetype to reflect varying degrees of credit risks in individual portfolios. The criteria calculate the rating-specific market value decline likely to be endured when foreclosing on a property, and adjust it to determine loss severity assumptions upon default of a mortgage in a pool (including foreclosure and other associated costs).

29. The outcome of these calculations is a stressed foreclosure frequency, loss severity, and loss coverage at each rating level for a pool. The rated tranche is generally expected to withstand the stressed loss level without defaulting in order to achieve a certain rating.

30. To calibrate the foreclosure frequency assumptions associated with the archetypal pool for a given country in a systematic, transparent manner, we use the Mortgage Market Assessment (MMA). The MMA considers key factors in the calibration of country specific foreclosure frequencies.

31. In addition to the MMA, the application of these criteria, when analyzing pools in individual countries, also considers the specificities of the local mortgage markets. For this reason, the specification for the archetypal pool varies across individual countries, and where warranted, may include considerations not already made under these criteria. Similarly, the criteria provide general principles for assessing variations from the archetype. Depending on the situations, additional considerations may be made in the more detailed application of the criteria, or in order to reflect risk factors that are unique to a specific mortgage market. The more specific application of the criteria is published in a separate guidance document (see "Guidance: Global Methodology And Assumptions: Assessing Pools Of Residential Loans," published Jan. 25, 2019).

Global Approach For The Analysis Of Mortgage Default Risk – Calculating The Foreclosure Frequency Of A Mortgage Pool

MMA methodology: country specific assessment

32. The criteria use the MMA to calibrate the archetypal pool's foreclosure frequencies for each individual mortgage market (typically at a country level). The MMA framework is designed to evaluate the relative risk of residential mortgage markets globally.

33. Our MMA analysis results in an assessment ranging across six categories from "very low risk" to "extremely high risk." The MMA has three components: economic risk, mortgage industry risk, and provisions for lender recourse to the borrower.

MMA methodology: deriving the archetypal foreclosure frequencies

Table 1

34. Each MMA maps to a range of potential 'AAA' foreclosure frequency assumptions and a minimum default assumption at the 'B' rating level, as shown in table 1.

35. These assumptions reflect a rank ordering based on macroeconomic considerations to benchmark relative credit risks across jurisdictions. They are the basis for determining the default assumptions assigned to an archetypal pool of residential mortgages in a given country, as described further below.

36. For more detail on our assessment of the MMA components and the determination of the final MMA for a country, see Appendix I.

Setting the 'AAA' archetypal foreclosure frequencies

37. The ranges of applicable foreclosure frequencies at the 'AAA' rating level are calibrated taking into account extreme historical stress scenarios as a reference point. We considered the U.S. Great Depression, as well as more recent stress scenarios such as those experienced in Greece and Argentina.

38. As such, these 'AAA' foreclosure frequency ranges (see table 1 above) are intended to provide stability to our default expectations through normal economic cycles. However, as the ranges indicated in table 1 show, 'AAA' levels could increase (or decrease) if the level of risk in the market increases (or decreases) significantly. For example, we would expect such change in the case of a moderate-to-substantial level of economic stress and/or a material structural shift in the mortgage market.

39. In order to determine the specific 'AAA' foreclosure frequency assumption within the range shown in table 1 assigned to a particular country based on its MMA, we consider a number of factors to assess the default risk relative to peers with comparable MMAs. Some examples of the types of factors we may consider include:

• Idiosyncratic market features such as product links to inflation, mortgage penetration levels, etc.;
• Long-term performance history relative to peers; and/or
• Data limitations, such as in relation to long- and short-term performance history, or availability of granular pool level data.

40. Based on this relative risk-ordering approach, if we observe a given mortgage market's performance over cycles is stronger (or weaker) relative to peers (i.e., countries with a comparable MMA), we typically assign a 'AAA' anchor at the lower (or higher) end of the range for such jurisdiction. Similarly, where there are limitations to the available data, we typically assign a 'AAA' anchor at the higher end of the range for a given MMA.

41. In certain circumstances, the 'AAA' foreclosure frequency assumption assigned for a given jurisdiction's archetypal pool may exceed the range associated with the relevant MMA as indicated in table 1. For example, this could occur if observed defaults on loans with archetypal characteristics in a severe (or less stressful) economic downturn exceed the levels associated with the relevant range of 'AAA' foreclosure frequencies. Alternatively, we do not increase the 'AAA' foreclosure frequency beyond the range indicated by table 1 if the negative performance was driven by factors we are otherwise adjusting for in our analysis (e.g., poor underwriting standards, very high loan-to-value ratios [LTVs], etc.).

Determining our expected-case foreclosure frequency assumptions at the 'B' rating level

42. The 'B' foreclosure frequency minimums in table 1 serve as the starting point for our 'B' expected-case foreclosure frequency for the archetypal pool of a given country.

43. Our expected case reflects observed trends as well as our outlook for the prevailing market over the next three-to-five years. The outlook takes into account considerations such as:

• Market structure and underwriting standards (and any relevant changes thereto);
• Forecast unemployment rates, inflation, and interest rates; and
• Prevailing mortgage loan performance, defaults, delinquencies, and their roll rates (e.g., transition rates).

44. Therefore, default assumptions at the 'B' rating level adjust over time to reflect changing economic conditions, when we believe they may materially affect levels of credit risk in the mortgage market.

45. If the 'B' (or more rarely, the 'AAA') foreclosure frequency assumptions for the archetypal pool are updated, intermediate foreclosure frequency assumptions ('B+' to 'AA+') are also updated through an interpolation based on the relevant 'B' and 'AAA' values.

Pool Analysis Methodology: Establishing An Archetypal Pool Of Residential Mortgage Loans

46. We perform a loan-level analysis to assess a residential mortgage pool's credit quality based on a comparison of its borrower attributes and loan characteristics to those of the archetypal pool. Pool foreclosure frequency levels are derived from archetypal projected defaults (see the Criteria Framework section above) adjusted for relevant variances from the archetype as described further below in the Pool Analysis Methodology: Adjustment Factors For Variations From The Archetype section.

47. The archetypal pool in each jurisdiction or market represents a somewhat idealized version of the average pools historically observed. While reflecting local features, the archetypal pool for a given country generally includes the below features:

• At least 250 loans (at the inception of the transaction that the pool is backing);
• Prudently/conservatively underwritten (i.e., no originator adjustment is applicable)--more specific characteristics may feature in local markets;
• Borrower has no adverse credit history (or has a prime-like credit score);
• Loan's purpose is for property purchase (not a refinancing loan or cash-out);
• Diversified nationally;
• LTV of the loan reflects prudent underwriting in the given market (e.g., 75%);
• Loan affordability was assessed at origination;
• Pool reflects an idealized product mix (for example, share of fixed- versus floating-rate loans, no loans with payment shock features, etc.);
• Borrower is a private individual;
• Loan is current;
• Borrower is not self-employed;
• Loan is denominated in domestic currency of lender and borrower;
• Unseasoned loan;
• Fully amortizing loan;
• First-lien mortgage (or equivalent in certain jurisdictions);
• Residential property, owner-occupied; and
• Fully appraised property valuation.

48. In addition, where relevant in a given market, we specify the archetypal pool to also include the following:

• Original loan term if this is a relevant risk factor (lesser risk for shorter-term loans and greater risk if tenor greater than market average);
• Primary residence (i.e., not a second home);
• Residency status of the borrower;
• Not a "jumbo" property valuation (i.e., an unusually large property presenting more idiosyncratic risk); and
• Other local relevant market factors.

Pool Analysis Methodology: Adjustment Factors For Variations From The Archetype

49. For each characteristic and attribute that differ from the archetype at a pool, loan, or borrower level, the criteria apply a corresponding adjustment to the foreclosure frequency, determined based on the methodology described in the Global Approach For The Analysis Of Mortgage Default Risk – Calculating The Foreclosure Frequency Of A Mortgage Pool section. Depending on the risk profile of the characteristic or attribute relative to the corresponding archetypal trait, the adjustment increases, and sometimes decreases, the foreclosure frequency assumption for the individual loan or pool, as further described below in this section. The typical foreclosure frequency adjustments for each of the factors are summarized in table 2 below. These adjustments vary across jurisdictions based on the differentiating factors identified below, to reflect local market characteristics and/or performance.

Table 2

Loan-to-value (LTV) of the loan – LTV curves

50. The criteria recognize that the LTV of a loan is a dominant factor in predicting future mortgage performance. The actual relationship, however, tends to be country-specific, depending on local factors (for example, whether lending is on a full- or limited-recourse basis). Based on available data, our criteria consider different LTV measures.

51. A loan's original LTV (OLTV) is calculated as the original balance of a loan (and any second-lien exposures, where applicable) divided by the original valuation of the mortgaged property. A loan's current LTV (CLTV) is the current balance of a loan, reflecting the actual (or projected) principal pay down (and any second-lien exposures, where applicable), divided by the indexed initial value or other updated value of the property.

52. The criteria use an LTV curve to make upward and downward adjustments to the foreclosure frequency of a loan, depending on whether its LTV is higher or lower than the LTV of the archetypal loan in the country.

53. The LTV of a loan is an indicator of the borrower's willingness to pay back the mortgage, given the equity at stake in the financed property.

54. The criteria establish a sliding scale of a borrower's willingness to pay, depending on market characteristics. For example, in a market where lending is typically on a limited-recourse basis, a borrower may have an incentive to hand in the keys if little-to-no equity exists in a mortgaged property or if the borrower is in negative equity due to adverse property price fluctuations. In that case, a higher LTV, all else being equal, is a marker of higher risk. Conversely, in a market where high LTV lending is primarily driven by tax incentives such as mortgage interest deductibility, a high LTV loan may not necessarily signal increased borrower risk. To reflect such market specificities, the LTV curve features a steeper or flatter slope, respectively, in each case.

55. The criteria typically use one of three different LTV curve types (encompassing both the curve and the associated LTV measures). The use of different LTV curves recognizes that jurisdictional factors influence risk differently and reflects available data analysis. We discuss each curve type below.

Type 1 LTV curve

56. Typically used where mortgage lending is generally on a limited-recourse basis, this curve has the steepest slope. For example, based on a neutral point (a factor of 1x) for an archetypal LTV of 75%, the adjustment factor is 3.00x at an LTV of 95%, and 3.95x at an LTV of 100%. This curve increases exponentially as LTV levels increase, capping out at LTVs well above 100%.

57. In a limited-recourse market, the CLTV of a loan is typically as meaningful a relevant risk differentiator as the original equity investment (OLTV). Assuming the loan was originally granted at an LTV of less than 100%, the propensity of a borrower to default when in negative equity depends on the degree to which the value of the property has decreased. In other words, the degree of negative equity is a driver of default on an exponential basis. Therefore, in markets we consider to be limited-recourse, the LTV curve is typically applied to a combined OLTV and CLTV.

58. Chart 2 below illustrates an example of a type 1 LTV curve, based on an archetypal (neutral) LTV level of 75%.

Type 2 LTV curve

59. This curve is typically used in markets where mortgage lending is on a full-recourse basis. The slope or shape of the curve is similar to that of the type 1 curve, meaning that the "growth rate," or relative increase of the adjustment factor for a given LTV increase, is generally the same. However, the slope of the curve (for example, assuming a neutral point at 75%) at higher LTV levels (typically around 98%) starts to flatten and ultimately levels off (typically around a maximum of 130%).

60. The slope of the curve at higher LTVs does not increase exponentially because, in full-recourse markets, we believe the risk of negative equity is less of a factor, as borrowers cannot walk away from the property and lenders can go after non-house assets to recoup loan amounts. Therefore, in our view, there is less differentiation for LTV levels above 100% than in limited-recourse markets. As a result, in markets we consider to be full-recourse, the combined CLTV and OLTV measure gives greater weight to a loan's OLTV than to its CLTV.

61. In some cases where updated property values are not available due to limitations in indices, the LTV measure used to adjust foreclosure frequency may be based on the current loan balance and the original property value.

62. Chart 3 below illustrates an example of a type 2 LTV curve, based on an archetypal (neutral) LTV level of 75%.

Type 3 LTV curve

63. This curve is typically featured in full-recourse markets only, where underwriting standards or structural features change the nature of LTV as a performance indicator, and typically merit a less stressful LTV curve. In some markets, higher LTV loans are not driven by an inability to fund the property with equity, but by, for example, an incentive to maximize tax advantages from interest deductibility. Therefore, on a relative basis, lending at high LTV is more commonly commanded by such incentives. In other markets, underwriting is more focused on capacity to repay on an ongoing basis, as measured by debt to income (DTI), and as a result, we place greater weight on DTI in our analysis.

64. For these reasons, the curve is flatter than the type 2 curve, with lower top-end multiples, typically capping out at an LTV above 100% based on our analysis of performance data.

65. As for the type 2 LTV curve, the combined CLTV and OLTV measure gives greater weight to a loan's OLTV than to its CLTV.

66. Chart 4 below illustrates an example of a type 3 LTV curve, based on an archetypal (neutral) LTV level of 85%.

67. For all LTV curves, the approach for setting the neutral point (that is, the LTV level for which the adjustment is 1x, [e.g., the archetypal LTV]), reflects local market conditions.

68. The determination of the neutral point is based on our experience of the market, findings from originator reviews, and expertise from our bank rating analyses. We typically set the neutral point at around 75%. In some markets, mortgage lending is typically on a low LTV basis (e.g., 55%); in these cases, the LTV curves are adjusted accordingly, such that higher adjustment factors apply to lower LTVs, thereby maintaining incremental changes in the adjustment factors from the neutral point across markets.

69. Finally, in markets where the underwriting approach incorporates the use of standardized credit scores, the risk adjustment for LTV may be stratified by the borrower's credit quality, as indicated by its credit score.

Affordability

70. The archetypal pool contains loans that have been assessed by the lender for affordability at origination based on standard market practice in a country.

71. To adjust for varying degrees of mortgage affordability, the criteria use one of two different approaches, depending on the practice in the market:

• Affordability is measured as a debt service coverage or DTI ratio, expressed as a percentage, equal to the monthly cost of servicing the borrower's debt plus other expenses divided by the monthly income.
• Affordability is measured more broadly as an income multiple, equal to the total borrowing divided by the annual income of a borrower.

72. If loan-level data is not available for measuring affordability, we make adjustments based on our assessment of the originator's practices relative to market standards via the originator adjustment at the pool level (see below).

Debt-to-income (DTI) affordability measures

73. At a DTI ratio of 35%, a neutral adjustment (1x) typically applies. Where affordability ratios for a given market only consider mortgage debt as opposed to "all-in" debt, the neutral point is typically lower.

74. We apply a sliding scale of adjustment factors for affordability. Credit (a multiple of below 1x) may be given for affordability better than the neutral point, whereas the adjustment factor increases for DTIs above the neutral point (typically capping out at DTI values of around 60%). In markets where we consider affordability to be a more significant driver of credit risks than LTV, and our LTV curves are flatter, we use a higher range of adjustments to allow for sufficient differentiation.

Income multiple

75. At an income multiple of 3.5x--meaning the debt of a borrower amounts to 3.5 years of income--the neutral adjustment factor (1x) typically applies. For income multiples higher or lower than the neutral point, we apply higher or lower adjustment factors. As the approach to affordability using an income multiple typically reflects underwriting that is more primarily based on LTV, we usually do not apply the income multiple adjustment factor when a loan is more than 18 months seasoned, except for very high multiples (typically greater than 5.0x).

Alternative affordability measures

76. Some jurisdictions measure affordability by means other than DTI or income multiples, such as a net-surplus ratio. Where loan-level information on such measures is provided, we typically adjust foreclosure frequencies based on jurisdiction specific considerations such as performance.

77. Where affordability information is not available at the loan level, or a common standard for reporting this information does not exist in a market (typically in mortgage markets where underwriting is less automated), we seek to identify the relative underwriting affordability standards of each originator. When we consider that a particular originator applies a weaker standard than the market, we make a portfolio-level adjustment to the foreclosure frequency of a pool through the originator adjustment. The size of the adjustment reflects the increased level of risk from the individual originator's underwriting practices.

Seasoning

78. We have observed globally that seasoned mortgage loans (e.g., more than five years from origination) generally perform better than newly originated loans. Continued performance over an extended period indicates that the borrower can afford to service the mortgage debt through ups and downs. We therefore apply a positive adjustment factor to the foreclosure frequency of loans that are seasoned more than five years. We typically only apply seasoning credit to loans that are currently performing. The start point for seasoning credit may vary from five years if there is strong supporting empirical evidence.

Employment status and income documentation

79. The archetypal borrower is typically salaried. Loans to self-employed borrowers therefore tend to attract an adjustment factor. In jurisdictions with limited employment protection, there is less differentiation in terms of job security between salaried and self-employed borrowers; therefore, the adjustment we apply is typically lower than would otherwise be the case.

80. In addition, in markets where borrowers have a degree of optionality around the level of documentation provided to lenders (income documentation), we typically apply an adjustment factor to reflect non-archetypal attributes. Depending on the type of documentation, the adjustment varies in size and duration. The criteria recognize that the source of information used to verify income and therefore affordability, as well as expectations for the number of months of documentation that is considered, can vary considerably across different markets, based on local practice.

81. In jurisdictions where mortgage pools include material concentrations of loans to pensioners (at origination), there also is an adjustment factor in the absence of mitigating factors. Similarly, in markets where underwriting to unemployed borrowers (at origination) is provided for, a significant adjustment typically applies in the absence of any mitigating factors.

Security

82. Sometimes a residential pool includes loans secured by second liens on a property, to which we typically apply an adjustment factor. The adjustment factor depends on the availability of information on the first lien. If information for the primary lien is not available, then an adjustment factor at the higher end of the range is applied.

Occupancy status

83. During periods of economic stress, a borrower is more likely to default on a mortgage secured by a second home or an investment property than on a home that is a primary residence.

84. Where lending on second or holiday homes is common in a market, we typically apply an adjustment factor.

85. Where the property is not owner occupied but is an investment property, the criteria apply an adjustment factor that varies depending on whether the loan has been underwritten to the borrower's income or to the debt service coverage ratio (DSCR) of the rental property.

86. In addition, in jurisdictions where the lender may not have ultimate recourse to the borrower's primary residence or other assets, loans underwritten solely to the DSCR typically attract a higher adjustment factor, but certain other adjustment factors may not be applicable (e.g., DTI, employment, and occupancy type).

Loan purpose

87. Generally, a loan used to borrow against the equity in a property (cash-out) has the highest default risk. Equity withdrawal is typically considered riskier than a home purchase or a fully re-underwritten refinancing.

88. Also, when a refinancing loan has not been fully re-underwritten, we typically apply a smaller adjustment factor than for cash-out loans to differentiate refinancing loans from conventional purchases, unless we have data that demonstrates no difference in performance exists during a stress period.

Loan product type

89. The specification for the archetypal loan product type in a given country reflects the longstanding traditional underwriting practice in such market.

90. When supporting data exist, the risk profile of standardized products is therefore likely to be reflected in the performance that informed our base default assumptions. However, when there is a specific characteristic not captured by data or for non-standard products in a market, we may make adjustments for varying loan product types.

91. A variation from the archetype for the interest rate product may introduce more (or less) risk. For example, in a floating-rate market, a fixed product for life therefore typically attracts a positive adjustment (i.e., less than 1.0x). Alternatively, a loan with a non-archetypal feature that could create a payment shock, or a floating-rate loan in a fixed-rate market, typically attracts a negative adjustment (i.e., greater than 1.0x) to foreclosure frequency.

92. The adjustment for other types of payment shock varies depending on the materiality of the potential shock. We may remove the adjustment factor after the expiry of the feature. Examples of products with payment shock features include loans with an initial interest-only period, a temporary promotional ("teaser") rate, or a fixed-to-float interest rate change.

93. The loan denomination may also introduce a mismatch risk, such as loans with balances that adjust with inflation where there is a risk that wage inflation does not track the relevant inflation index in which the loan is denominated.

94. Depending on specific local market characteristics, we adjust for other product types where they are prevalent. Examples include product types with atypical amortization schedules that present specific risks relative to the archetype, including interest-only and balloon loans.

95. A balloon loan has a significant portion of its principal paydown that is due as a lump sum, often not until late in the mortgage's life. Depending on the market practice, an interest-only product may refer to a loan whose principal amortization is deferred for several years, or to a loan whose repayment is actually only due on the final repayment date for the mortgage (interest-only for life). The sizing of the corresponding adjustment factor under the criteria reflects the available performance data as well as the degree of commoditization of the product in the local market. For example, depending on the market, short-term interest-only loans typically attract a higher adjustment factor compared to longer-term interest-only loans where the borrower has more time to separately build equity to proceed with the final repayment.

Loans in arrears

96. An adjustment factor is applied to loans in arrears based on when foreclosure proceedings are typically commenced. A default frequency of 100% in all rating scenarios is typically applied if the borrower is 90-180 or more days in arrears depending on the jurisdiction. A range of adjustment factors are applicable to loans in arrears less than 90-180 days depending on the jurisdiction (for example, 2.5x for loans that are 30 days' delinquent and 5.0x for loans that are 60 days' delinquent).

97. In addition, if a borrower has been in arrears but is now current, depending on the time elapsed since the arrears occurred, specific arrears adjustments may apply. Where loan-level data is not available, we account for pool-level arrears via an originator adjustment.

Reperforming and restructured/modified loans

98. A reperforming loan is one that is not currently seriously delinquent or in bankruptcy, but has been in arrears for three or more monthly payments (or, in some cases, where relevant, have been modified/restructured), in the recent past. Once a borrower has evidenced consistent performance since serious delinquency over an extended period of time, the loan may no longer be considered reperforming for the purposes of our analysis.

99. We typically apply adjustments for reperforming and restructured/modified loans, depending on the materiality of a pool's exposure to these loans. These adjustments typically take into account such considerations as the type of restructure/modification granted, the time elapsed since the restructure/modification, and/or performance subsequent to restructure/modification. The seasoning adjustment for these loans is generally based on the restructure/modification date for the remaining life of the loan.

Nonresidential loans/borrowers

100. In certain jurisdictions, mortgage pools include commercial or mixed-use properties in addition to residential homes. These properties present specific risks and we typically apply an adjustment factor.

101. In jurisdictions where borrowers may also be commercial entities (for example, the borrower is set up as a small- or medium-sized enterprise), we typically apply an adjustment factor. Such adjustments generally do not apply to loans made to trusts, special-purpose vehicles, and similar entities with direct recourse to an individual borrower.

Credit history

102. A borrower's past performance is a differentiating indicator of future mortgage performance. We therefore adjust foreclosure frequency assumptions to account for a borrower's adverse credit history.

103. The criteria recognize that the underwriting standards of different jurisdictions have evolved to capture this risk in a variety of ways, which is reflected in our analysis accordingly. For example, in some markets, we consider the number of prior credit events experienced by a borrower over a certain time horizon, as well as whether those credit events are more or less recent. Alternatively, when available in a market in a standardized fashion, we consider credit scores incorporating the borrower's payment history on past and current debt obligations.

104. As noted above in the discussion of LTV factors, sometimes a combined adjustment factor for different buckets of LTV and credit scores may be used. This is relevant in markets where data demonstrate the predictive power of combined LTV and credit score metrics.

105. In other jurisdictions, a specific record of past credit history may not exist in a systematic fashion. In that case, our analysis seeks to address the risk of lending to borrowers with negative credit histories by way of an originator adjustment reflecting the relative appetite of a lender for underwriting loans to more tarnished credits. This is because the archetypal loan is in all cases defined as one made to a borrower with a clean credit record.

Geographic concentration

106. The criteria address geographic concentration risks in a pool by applying adjustment factors to the foreclosure frequency for portfolio concentrations that exceed specific geographic thresholds (e.g., regions and/or counties). Geographic concentration risk emanates from the risk of a localized downturn in a given country.

107. Thresholds for defining concentration are typically based on a country's population distribution as a proxy for diversification. For example, a threshold of two times the population is typically used under these criteria, rounded as necessary. Alternatively, the approach may use statistical methods where available to more specifically capture concentration risks.

108. Adjustments for geographic concentration are typically applied where the threshold is exceeded.

Original loan term

109. In markets where there is a meaningful differentiation in loan terms based on observed performance, we may apply an adjustment for variations from the archetype, and if loan level data is not available, we account for the risk in the originator adjustment.

Residency status

110. We typically apply an adjustment factor where we have assessed there may be a likelihood of higher default risk posed by borrowers that are not citizens or permanent residents of the country in which the mortgaged property is located.

111. The adjustment factor applied is at the higher or lower end of the range depending on considerations such as nonresident borrowers' exposure to employment stresses relative to residents, as well as practical issues associated with foreclosing on borrowers who may reside outside the country.

Pool originator adjustment

112. Mortgage loan performance typically reflects the quality of a lender's origination and underwriting processes. Therefore, as part of our credit analysis, we assess mortgage originators in a given country based on the historical performance of their loans, when available, and our views of the quality of their loan origination and underwriting processes. We also assess the originator's policies and procedures, including how they have changed over time. Our archetypical pool is based on the assumption of prudent underwriting of the mortgage loans, and we reflect our assessment of an individual originator relative to the archetype accordingly.

113. The criteria apply a pool-level adjustment factor to reflect the observed historical performance of various pools from different originators or lenders within each jurisdiction. Therefore, this adjustment typically also reflects our projections of variations in future delinquencies/arrears relative to the archetype. The factor may vary from transaction to transaction for the same originator as relevant.

114. Also, servicers are not always able to report loan-level data on certain characteristics that we typically consider in our analysis, and, in some cases, the information received at closing is not updated for surveillance. The analysis of originator and servicer risk includes the assessment of the underwriting policies regarding such loan and borrower features (including data provision), and may result in an originator adjustment under these criteria. As such, factors that we include to determine the originator adjustment tend to be country specific.

Small-pool adjustment

115. For pools that--unlike the archetype--initially contain fewer than 250 loans, the criteria apply an adjustment as a multiple to the foreclosure frequency. The adjustments are typically those indicated in chart 5 below).

Loan-specific risks

116. Our research shows that when a pool is sufficiently granular, the risks attached to an individual loan are mitigated at the pool level. Nonetheless, the analysis focuses on any material concentration in a few loans, which may result in a pool-level adjustment.

Global Approach For The Analysis Of Loss Given Default - Calculating The Loss Severity Of A Mortgage Pool

Property valuations, valuation haircuts, and valuation indexation

117. The calculation of recovery proceeds from a property's sale reflects the property market's state at the time of analysis. Typically, we are provided with the original property valuation recorded at loan origination. We therefore index this valuation through the date of our analysis, based on updated index information available. The indexation method typically uses publicly available information on house prices in the relevant jurisdiction. However, depending on the circumstances, we may not fully reflect all changes. In that case, depending on the level of availability of house price data, we adjust the indexed value to incorporate our forward-looking view on house price volatility.

118. The indexation logic is determined on a jurisdictional basis considering the indices available and the performance history of the local market. The same indexation logic is used for the purpose of calculating the CLTV of a loan in order to calculate default frequency assumptions.

119. Where updated property valuations are provided, we consider the process by which such values were derived, and how they compare to the original indexed values, to determine the appropriate property valuation to be used in our analysis.

120. The criteria also typically calls for a country-specific reduction (or haircut) of the valuation on a property, for valuation methods that deviate from that of the archetype. The haircut is typically up to 15% for non-full appraisals. In addition, where specific property risks exist (such as the risk of an atypical decline in property value), the criteria provides that additional haircuts may be applied to a property valuation. The property valuation used in the analysis is first haircut, where applicable, and then indexed.

121. For jurisdictions where an index or other source of changes in house prices is not available or, in our view, too limited to perform the analysis described above, we adjust the market value decline assumptions used to determine loss severity, as described further below.

Assessment of property over/under-valuation

122. We typically quantify over- or undervaluation embedded in the property value used to calculate the estimated loss severity related to a loan, and reflect this over- or undervaluation in the calculation of market-value decline. To assess the degree of over- or undervaluation in the national or regional property market, we use various data sources that are available, including from the Organization for Economic Co-operation and Development (OECD), FHFA, or other national data providers; expert opinions; and independent research. Endogenous and exogenous factors influence changes in property prices, reflecting macroeconomic trends and variable supply and demand.

123. We generally estimate the level of over- or undervaluation of the relevant property market by comparing the prevailing house-price-to-income ratio against its long-term average. This assessment is informed by an analysis of national data on current price to income (disposable household income, or GDP/capita) compared to the long-term average price to income (PTI). The specific method employed to estimate the over- or undervaluation percentage used in our rating analyses in a given country reflects our qualitative views of the available quantitative data, as well as our medium-term expectations.

Market value decline assumptions and repossession market value decline calculations

124. We apply a market-value decline (MVD) stress to the indexed property valuation. To account for the effect of over- or undervaluation, the MVD of a repossessed property (Repo MVD) is calculated using the formula below, based on the rating-specific input variables shown in table 3 below. Absent any over- or undervaluation, the resulting Repo MVD is as shown in the last column.

125. Repo MVD = 1 – [1 – (Fixed MVD +/- percentage of over/undervaluation x over/undervaluation)] x (1 – FSD)

• Fixed MVD is a fixed recessionary MVD. It represents a rating-specific stress applied to a property value.
• FSD is a forced-sale discount factor. Among other factors, foreclosed properties may sell at a discount, due to the stigma that repossession creates. The FSD is larger at lower rating levels and smaller at higher rating levels. This is because, in a more severe recession, a greater proportion of all property transactions contributing to the overall index comes from distressed sales.
• Percentages of over- or undervaluation 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 level, and is highest for a 'AAA' rating. This is because a 'AAA' issue rating typically reflects greater stability than other ratings, relative to a long-term trend in property values. The degree of over/undervaluation in a property market is only partly reflected in the Repo MVD formula to account for our view that property values during a housing cycle in part reflect the current valuation of the property market, while reverting to the long-term average in the longer run.
• The Repo MVD may be capped at 75% for a given market.

Table 3

126. In certain circumstances, the Fixed MVDs and/or FSD for a given jurisdiction may be different from those indicated in table 3 at one or more rating categories based on historical observations. For example, historical peak-to-trough house price data may show declines in excess of those assumed under the related stress scenario for that time period, after taking into account any adjustments for over- or undervaluation. In these cases, we calibrate the Fixed MVD based on observed declines, taking into account the related macroeconomic stress. For example, if we observe a certain decline in a moderate economic downturn, proportionately, the 'AAA' Fixed MVD will be higher. In addition, where we don't have data on changes in house prices through an economic cycle but perceive high risk such as countries with very high MMAs, we may apply increased Fixed MVDs and/or FSDs.

Repo MVD adjustment factors

127. We apply additional adjustments to the Repo MVD to reflect the potential for greater house price volatility relative to that assumed for the archetypal pool.

128. For example, we typically adjust the Repo MVD for property values that exceed a country-specific jumbo valuation limit (defined at the relevant geographic level-–national, regional, city, etc.) on an increasing scale and using a function that typically caps the adjustment at 1.2x-1.3x. Properties with higher valuations could experience higher loss severities, due to their smaller and less-liquid market. The country-specific jumbo valuation limit is generally based on the prevailing average house price, also considering house price distribution. In determining whether a valuation is jumbo, the modeling typically looks at the fully indexed valuation of the property.

129. Where relevant, we may also adjust the Repo MVD for certain properties to address the potential increased volatility in prices on account of other features such as location, property type, housing sector or loan denomination. For example, we may adjust the Repo MVD for properties located in certain geographic locations to reflect the view that there are potential demand/supply pressures, and geographic and structural concentration risks that may lead to more volatility in resale values (e.g., high-density, inner city apartments in some countries). In these cases, the adjustment factor typically ranges from 1.25x to 1.45x.

130. Fundamental to our MVD approach is our view that we adjust our assumptions for housing market volatility as necessary. This is generally achieved via indexation and the over- or undervaluation assessment. For jurisdictions where an index or other source of changes in housing price indexes (HPIs) is not available, or has material limitations, we instead adjust the Repo MVD of all loans in a pool to account for the potential impact of indexation and over- or undervaluation by a multiple. We typically adjust the Repo MVD by a multiple within a range of 1.0x-1.5x. The multiple applied in a particular jurisdiction is based on our consideration of the following factors, which may change over time:

• The overall level of perceived risk in the market as reflected in the MMA. Typically, the higher the MMA, the higher the multiple applied.
• Our assessment of house price volatility in the market. The greater our expectations of volatility, the higher the multiple applied.
• Supply and demand dynamics such as trends in housing starts, mortgage credit, interest rates, the impact of government intervention, or any relevant market distorting factors.

Foreclosure costs

131. Foreclosure costs incorporate legal and other expenses associated with enforcing security on a mortgaged property. Our assumptions are generally based on country-specific data obtained through investor reports, discussion with servicers and legal counsel, or other available sources. Foreclosure cost assumptions are typically reflected as: (i) a fixed amount plus a percentage of the distressed property value (i.e., the property value obtained after the application of the Repo MVD at the relevant rating category); or (ii) a percentage of the loan balance, depending on local practices and data presentation.

132. Where relevant in the local market, costs associated with property taxes and home insurance costs are separately accounted for.

Accrued and unpaid interest

133. Unpaid interest on defaulted loans that accrues through to the point of foreclosure (foregone interest) is typically incorporated into a loan's loss severity in the following situations:

• The servicer is contractually obligated to advance delinquent borrower interest payments through liquidation unless such advance is deemed nonrecoverable from anticipated loan proceeds; or
• The rating analysis does not use cash flow models, such as for certain synthetic transaction structures.

134. In the cases above, the interest rate used for the purpose of computing the interest accrual typically reflects the yield on the defaulted assets, and varies by product type and original, current, and maximum rates.

135. Where accrued interest is not included in the loss severity calculation, the impact on a transaction's credit enhancement is addressed in our cash flow analysis by modeling the corresponding negative carry, using stressed interest rate assumptions.

Foreclosure period/liquidation timeline

136. Foreclosure period assumptions, which represent the estimated time to repossess and sell a property upon a default, are specific to each country, and reflect the typical time necessary for judicial proceedings, where applicable, and any other likely delay. We estimate those liquidation timelines based on historical data reflecting current and stressed observations or by benchmarking to comparable jurisdictions where country specific data is not available.

137. Foreclosure period assumptions may differ within a country by region, property type, whether the security is a first or second lien, other applicable law, or other considerations, where relevant. For example, in certain jurisdictions, the regime applicable to investment properties differs from that of owner-occupied property when foreclosing upon them.

138. Foreclosure timelines are used under these criteria to evaluate the total amount of accrued and unpaid mortgage interest, as discussed in the above section. Corresponding assumptions are also used as an input in our cash flow analysis to model the duration of corresponding negative carry before recovery proceeds are received.

Loss severity floors

139. To address any potential idiosyncratic risk on pools with very low LTVs that may not otherwise captured by our assumptions, we floor the pool-level WALS at 2%. Where available historical performance data or other relevant considerations indicate a more granular approach is relevant, we may apply loss severity floors at the loan level. However, we may go below the floors for loans that carry mortgage insurance or guarantees.

Mortgage loan insurance and guarantees

140. If some or all of the mortgage loans in a portfolio include a mortgage loan insurance or guarantee, then our global mortgage insurance criteria apply (see "Methodology For Assessing Mortgage Insurance And Similar Guarantees And Supports In Structured And Public Sector Finance And Covered Bonds," published Dec. 7, 2014). This excludes country specific guarantee schemes.

Other loss coverage adjustments

141. Where relevant, our loss coverage analysis may also consider the impact of other market-specific considerations on mortgage defaults and severities. For example, we may increase loss coverage if third-party reviews of loan data indicate variations to our expectations, or if we consider the likelihood that an entity is able to execute on representation and warranty repurchase obligations to be low.

Minimum Credit Enhancement Levels And Other Considerations

Credit enhancement and subordination floors

142. The criteria establish rating-specific minimum credit enhancement levels applicable for assigning ratings at issuance of a new transaction. These are set at 4.00% for 'AAA' ratings and 0.35% for 'B' ratings, subject to interpolation for the levels in-between. These credit enhancement (CE) floors are based on our view that there are limits on the predictability of mortgage loan performance. The 4.00% minimum credit enhancement for 'AAA' corresponds to 25x leverage. We believe that leverage above that level creates vulnerabilities that are inconsistent with the degree of creditworthiness associated with an 'AAA' rating.

143. Moreover, the minimum credit enhancement levels can't be funded solely through "soft" credit enhancement (commonly called "excess spread"), and the minimum amount of "hard" credit enhancement supporting a 'AAA' rating backed by residential mortgage loans is 2.5% (2% for 'AA' and 1.5% for 'A'). Hard credit enhancement generally includes such sources as subordination (which may include the benefit of mortgage insurance), overcollateralization, letters of credit, and reserve funds.

144. The analysis of minimum credit enhancement levels considers the specificities of each transaction. For example, if a transaction is exposed to negative excess spread--that is, loan interest at a pool level is lower than the cost of funds--this is accounted for (i.e., deducted from available credit enhancement) in our determination of CE floors. Likewise, if a transaction benefits from overcollateralization, a reserve, or other form of support that is either dedicated to cover specific structural risks in a transaction, unrelated to the pool's credit quality, or is otherwise relied upon for that purpose in our rating analysis, then its amount is typically deducted from available credit enhancement when determining if relevant floors are met. Examples include a commingling reserve, or overcollateralization that supports asset and liability mismatches in a structure.

145. In addition, when assigning ratings at transaction issuance, where relevant in the absence of other mitigating factors, out-of-model credit considerations are made to determine subordination floors typically for structures that feature pro-rata payments from inception (see table 4).

Table 4

Originator insolvency, commingling, and set-off

146. The analysis of commingling and/or set-off risks that can result from an originator's or servicer's insolvency may use modeling to produce estimates of any exposure, after consideration is made of any structural mitigants.

147. For example, our legal analysis may conclude that the insolvency of the collection accountholder could result in a loss of funds deposited in the account. This analysis considers the degree to which a collection accountholder's insolvency affects the cash flow from the residential loan pool, if the collection account is not in the name of the issuer. The amount at risk depends on the timing of scheduled payments from borrowers, the frequency of transfers into the transaction account, and the level of prepayments.

148. If the loan originator (or servicer) is also a deposit-taking institution, modeling considers the possibility that borrowers with deposit accounts may set off these amounts against their outstanding mortgage loans. For example, we may derive an estimate of the set-off exposure from the mortgagees' highest historical deposit account balances on the originator's or servicer's books.

149. The modeling approach may also capture set-off risks from any borrowers that are also employees of the originator. This is because amounts owed to them--such as unpaid salaries, pension benefits, and subsidies--could offset payments due on their mortgage loans. As a result, the modeling approach typically treats the full amount of any employee's mortgage loan balance as entirely set off and all set-off amounts as principal losses.

Surveillance

150. Our analytical approach to surveilling transactions backed by residential mortgages that fall within the scope of these criteria reflects our view of change in the credit risk inherent in a mortgage pool over time. In our surveillance, our focus is generally on a subset of variables that we consider to be the most likely to materially change, and which reflect the credit quality, credit enhancement, and other relevant factors that drive the determination of ratings when analyzed in conjunction with observed and expected performance.

151. Our surveillance 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. In applying the methodology and assumptions described in these criteria, our surveillance analysis considers the collateral's performance trends and dynamic characteristics as well as significant shifts in the pool's collateral composition to determine if updates to our most recent assumptions are warranted.

152. At the time of issuance, a pool of mortgage loans typically has very little, if any, observed performance data available. As such, our new issuance analysis primarily infers future collateral performance 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 are considered in determining the assumptions we apply in our surveillance analysis. For example, in our surveillance, we analyze projected foreclosure frequency based on the original loan attributes as well as the observed payment history. Further, the performance history informs 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 originator adjustment factor applied in our analysis over time as performance history evidences the quality of the originations in the pool.

Determining The MMA

MMA methodology: building blocks and scoring

153. Our MMA analysis in a given jurisdiction consists of an assessment on a six-point scale ranging from "very low risk" to "extremely high risk." The MMA scoring assesses the following three core components:

• Economic risk;
• Mortgage industry risk; and
• Lender recourse provisions.

154. We first combine our assessments of the relative economic and mortgage market risk factor scores and then overlay our assessment of lender recourse. The resulting MMA is used to calibrate foreclosure frequency assumptions for the archetypal pool in the country, within a given range at the 'AAA' rating level and at a minimum level for 'B' ratings. A review of relative performance data and their availability, combined with any other relevant factors, helps determine where within a set range the final 'AAA' assumptions fall for the jurisdiction in question.

155. Table 5 below shows the steps for the MMA methodology framework.

Table 5

156. The assessments of economic risk and mortgage industry risk are each determined from the average of the risk scores assigned to the associated factors listed in table 5. Each factor is individually scored within a range from 1 to 6, with a score of "1" corresponding to "very low risk" and "6" corresponding to "extremely high risk," as shown in table 6 below.

Table 6

157. The average of the economic risk and mortgage industry risk scores then forms the initial MMA, before adjustments for consideration of the provisions for lender recourse in the applicable market. Each score corresponds to a relative risk assessment ranging from "very low risk" to "extremely high risk," again as shown in table 6 above.

158. Depending on our assessment of lenders' recourse to borrowers' assets besides the mortgaged property, this increases the country's MMA typically by at least one category (e.g., from a score of '2' to at least a score of '3', or higher) to account for limitations in recourse (by law or in practice). The result is the final MMA, which is used to calibrate foreclosure frequency assumptions.

Step 1 of MMA methodology: evaluating economic risk

159. For each jurisdiction, the MMA methodology considers the relative levels of "economic resilience," "economic imbalances," and "credit risk in the economy," which together capture the relevant economic risks that are relevant to assess mortgage risk.

• Our analysis assesses economic resilience by rank ordering the wealth of an economy, as primarily measured by GDP/capita; a wealthier economy has more capacity to absorb economic shocks. Where GDP is heavily influenced by atypical factors (an example may be significant oil production in a country), we may adjust the score to ensure that the rank ordering is still relevant, considering the population segment in the country's archetypal pool.
• We consider relative levels of economic imbalances, depending on whether an economy is expanding or correcting. For example, if the economy is expanding, we look at the pace of private sector credit growth against the pace of residential house price growth, which may signal a credit fueled house price bubble. If an economy is contracting, we assess the likely length and impact of the contraction.
• Our assessment also considers relative levels of credit risks in the economy by looking at private sector debt capacity and leverage, the standard of underwriting for residential mortgages, and payment culture and rule of law. Private sector debt capacity and leverage looks at private sector credit/GDP, scaled by GDP/Capita (i.e., higher leverage presents more risk in relatively less wealthy economies); adjustments are made if GDP understates or overstates the relative wealth of the archetypal residential mortgage borrower. We assess the relative levels of underwriting of originators and practices of servicers and also consider factors--such as the percentage of mortgage market lending over high (e.g., 80%) LTVs; the average indexed LTV of the overall mortgage market; the number of factors typically considered for mortgage credit assessments, including, for example, moderately conservative cash flow requirements and collateral values; product types; the scale of non-prime lending; and lender concentration risk--all as relevant to mortgage market risk. Where we do not view the underwriting standards as at least moderately conservative, we elevate our assessment of credit risks accordingly.
• Finally, we consider the payment culture and rule of law in the given jurisdiction, including creditors' rights, the predictability of the legal framework, and applicable bankruptcy law. For instance, we may look at external indicators such as the World Bank governance data, and, unless we view the payment culture and rule of law to be at least moderately strong, we elevate our assessments of credit risks accordingly.
• In order to determine each individual factor score contributing to the overall economic score in our MMA framework (see table 5), our analysis typically uses as a starting point the same factor scores as these are determined within our Banking Industry Country Risk Assessment (BICRA) analysis for the relevant country. This is because the same factors that generally affect credit risks in the banking sector are also typically relevant for the analysis of residential mortgage risks, and therefore our MMAs.
• However, if the drivers of the economic risk score are less relevant for the analysis of mortgage risk, we may make adjustments. For example, with regard to the assessment of "economic imbalances" and "credit risk in the economy," the MMA analysis of these factors typically excludes any BICRA adjustments for commercial real estate prices and/or sovereign government credit stress. This is because these components are typically less correlated with residential mortgage performance (e.g., commercial real estate market), and/or already accounted for elsewhere in our analysis (e.g., sovereign credit stress addressed through Rating Above the Sovereign criteria) and therefore less relevant for the MMA.
• Similarly, the assessment of each individual sub-factor from the BICRA analysis considers the relevance of each such factor as a driver of mortgage market risk in the jurisdiction, as opposed to banking sector risk. This is because some factors may influence economic risk in a way that is more or less relevant or pronounced for the analysis of mortgage risk. In doing so, our analysis reflects all relevant considerations made above, as they pertain to the analysis of residential collateral in the country. For a given country, this may therefore result in an assessment for an individual sub-factor of economic risk (see table 5) that differs from that made as part of our BICRA analysis in such country.

Step 2 of MMA methodology: evaluating mortgage industry risk

160. Our assessment of mortgage market industry risk for any jurisdiction focuses on three factors: the relative strength and stability of the market's "institutional framework," any impact stemming from the "competitive dynamics" of the local mortgage lending industry and finally the "market sensitivity" of expected borrower default rates to changes in macroeconomic conditions such as unemployment.

161. As with our assessment of economic risk in step 1, our approach recognizes that many of the factors that generally influence banking risks are generally equally important in assessing mortgage market industry risks. Therefore, our framework here again typically uses as a starting point the outputs of the application of our BICRA criteria in the relevant jurisdiction. For example, our assessment reflects the quality and effectiveness of bank regulation, the track record of authorities in managing financial sector turmoil, and the competitive environment of a country's banking industry, including the industry's risk appetite and its structure.

162. However, our MMA of industry risk also recognizes that default risks of residential mortgages are sensitive to unique features. Therefore, the mortgage industry risk score for a country also assesses sensitivity to changes in unemployment by looking at historical unemployment volatility and by assessing levels of social welfare support, the impact of different personal insolvency regimes, and differing social perceptions associated with defaulting on mortgage debt.

Institutional framework

163. We start with the assessment of the institutional framework score, as described in the relevant BICRA analysis for a given country.

164. Our analysis of the relative strength of the institutional framework primarily considers the current policy framework and regulations for lending, including an analysis of institutional initiatives and regulatory track record; it also focuses on the extent to which the authorities in a given jurisdiction effectively and transparently oversee prudential lending standards.

Competitive dynamics

165. The assessment of competitive dynamics considers the mortgage lending industry risk appetite, its stability, and the impact of market distortions.

166. As with the institutional framework, the scoring of competitive dynamics also starts with the assessment of banking industry risk appetite in the BICRA analysis for the relevant jurisdiction. We believe the same key factors typically influence residential mortgage underwriting and servicing.

167. However, the MMA does not always consider the same factors in relation to industry stability or market distortions, where the factors do not speak to mortgage industry risk. On the other hand, where relevant the MMA analysis reflects in our scoring, any risk in a country that more specifically pertains to the underwriting of mortgages. In addition, our assessment also considers the mortgage lending industry in its entity, including non-bank financial institutions (NBFIs).

168. Risk appetite considers profitability, compensation culture, product mixes, and portfolio growth. Distortions typically include factors such as high government share in lending, the level of central bank control over mortgage rate setting or that of government-sponsored entities.

Market sensitivity

169. The initial market sensitivity score is determined by an assessment of the relative levels of unemployment variability and social welfare support in a country (see table 7). Typically, in countries with higher levels of welfare support, the impact on mortgage default risk is expected to be less, all else being equal, for a given change in unemployment, than in countries with lower levels.

Table 7

170. Unemployment variability is typically assessed by looking at the relative volatility of historic unemployment data in a country as published by the World Bank, adding S&P Global Ratings' medium-term projections. In doing so, we typically do not include data that is reflective of a structural stage in the development of an economy that we believe is no longer relevant.

171. A score on the scale of '1' to '6' is assigned for the unemployment score based on the categorization in table 8, which we update over time as necessary to preserve a meaningful distribution.

Table 8

172. Similarly, social welfare support in a country is assessed as either "high," "medium," or "low," with corresponding scores of '1,' '2,' or '3.'

173. Typically, a country characterized as having a high level of support exhibits welfare spending that is above the OECD average as a percentage of GDP, and has relatively generous unemployment benefits. This typically includes countries with relatively high levels of tax. In addition to a high-level assessment of GDP percentage spent on welfare support, we review the characteristics of employment social insurance, including coverage and duration and, for example, the percentage of the last salary that is covered. This analysis considers other sources of funds that may become available to a borrower upon becoming unemployed, such as withdrawals from regulated funds otherwise not readily accessible, as well as other forms of support in a given jurisdiction, where relevant.

174. A country that spends below the OECD average of GDP on welfare and has limited coverage and duration of unemployment social insurance is typically assessed as "low." A country that neither characterizes as "high" or "low" is assessed as having a "medium" welfare support score.

175. In addition, in certain circumstances, the overall score for market sensitivity may be adjusted up or down by one point (for example, where we have seen evidence that employment levels displayed resilience in the face of GDP declines).

176. Finally, in some countries, we have observed other factors that contribute to positive performance such as a particularly strong social stigma attached to personal bankruptcy or a default on individual debt, high levels of family support, or household savings levels. All else equal, this may lead to a lower propensity to default. Certain legal frameworks also create a disincentive for borrowers to enter into personal bankruptcy (for example, the length of discharge periods), which can also reduce the propensity to default. Giving consideration to both of these circumstances may improve the overall score by one point. Conversely, in some jurisdictions, the very opposite may be true, and in these instances, we may worsen the overall score by one point.

Step 3 of MMA methodology: assessing lender recourse provisions

177. In some jurisdictions (in practice or by law), lenders predominantly only rely on the mortgaged property as collateral to repay the loan in the event a borrower defaults. As a result, borrowers may be less incentivized to maintain mortgage payments when house prices decline; we refer to this as limited recourse. We have observed that this has a significant impact on borrower behavior versus full-recourse markets and is therefore a key differentiating factor that we consider when rank ordering MMA. All else being equal, we expect a MMA for any given country to be at least one category weaker for limited- versus full-recourse markets.

Final step of MMA methodology: determining 'AAA' and 'B' foreclosure frequency

178. Finally, we consider relative mortgage performance, data availability, and any other pertinent factors to help determine where a country's 'AAA' foreclosure frequency assumption is within the 'AAA' range for the relevant MMA score. We also review current and expected performance to determine our 'B' foreclosure frequency case.

179. For jurisdictions where we have limited data, the weaker the MMA, the more likely it is that our foreclosure frequency assumption is at the higher end of the applicable range. If we have limited data, but good reason to believe that the product mix and economic situation is very similar to a peer where we do have access to more extensive data, then we may likely set the level corresponding to that peer.

180. When reviewing performance to help assess the relevant percentage in the 'AAA' foreclosure frequency range for a given country, we typically look at the following:

• Performance through economic cycles. At the least, we look at dynamic arrears evolution over time. Where available, we consider default data over time. For a given economic cycle, how does the performance of a given jurisdiction compare to others with the same or close MMAs? Where historical data indicates that relative performance has been strong, and we have evidence to believe it will continue to be so, we tend to fall on the lower end of the foreclosure frequency range.
• Expected lifetime performance for archetypal mortgage pools. If we don't have default data, we may approximate with an arrears roll rate analysis. If we can make an assessment of what lifetime performance for an archetypal pool may be over a given economic scenario, this helps to inform our foreclosure frequency assumptions within the range.
• We compare to other jurisdictions with similar MMA scores, but account for differences in product mix and underwriting standards when making this comparison. For example, if a given jurisdiction had performance that appears worse than its peers, but typically lends at higher LTVs, we need to consider our LTV curve to make sure we are not double counting the impact of a higher LTV market.

181. For the 'B' foreclosure frequency assumption, we typically do not assign a percentage lower than the limit set for the corresponding MMA. We may go higher, however, and we typically consider the following when making this assessment:

• We consider the base-case default and loss assumptions used in our bank rating analysis.
• We consider performance we are seeing today in mortgage pools, accounting for the product mix relative to the archetypal pool.
• We consider our near-term economic forecasts, and factors such as our expectations for mortgage lending rates, inflation and wage inflation, unemployment long term versus short term, house prices, availability of credit, and consumer leverage.

APPENDIX II - Proposed Guidance

182. This appendix provides additional information and guidance to these proposed criteria, and we expect to publish this information and guidance in a separate guidance document following the publication of the finalized criteria article. It is intended to be read in conjunction with the proposed criteria herein and aforementioned global cash flow criteria. Guidance documents are not criteria, as they do not establish a methodological framework for determining credit ratings. Guidance documents provide guidance on various matters, including articulating how we may apply specific aspects of criteria; describing variables or considerations related to criteria that may change over time; providing additional information on non-fundamental factors that our analysts may consider in the application of criteria; and providing additional guidance on the exercise of analytical judgment under our criteria. Our analysts consider guidance documents as they apply criteria and exercise analytical judgment in the analysis and determination of credit ratings. However, in applying criteria and the exercise of analytic judgment to a specific issuer or issue, analysts may determine that it is suitable to follow an approach that differs from one described in the guidance document. Where appropriate, the rating rationale will highlight that a different approach was taken. For more information about guidance documents, please see "Criteria And Guidance: Understanding The Difference" in Related Research below.

Guidance: Spain

Spain background

183. The MMA for Spain is "intermediate risk" ('3' on a scale of '1' to '6'). This is based on an economic risk score of '3', a mortgage industry risk score of '3', and our assessment of recourse available to lenders in the market. This reflects our view of:

• The risks in the Spanish banking system;
• Spain's historical unemployment sensitivity to changes in economic output, which could negatively affect borrowers' ability to pay their mortgages absent a strong social welfare system in relation to mortgage payments; and
• The borrowers' willingness to pay in low/negative equity scenarios due to full-recourse.

Spain - archetypal pool

184. We define the archetypal pool for Spain as follows:

Table 9

Spain 'AAA' and 'B' foreclosure frequency anchors

185. Based on an MMA of "intermediate risk" ('3'), we set Spain's 'AAA' foreclosure frequency anchor for the archetypal pool at 15%, and the current 'B' foreclosure frequency assumption at 2.5%.

186. The 2.5% 'B' foreclosure frequency assumption reflects our assessment of historical performance and our expectations for future performance over the medium term given anticipated macroeconomic conditions. In our analysis, we considered loss expectations across mortgage portfolios and the performance of outstanding S&P Global Ratings-rated Spanish RMBS transactions and covered bonds, in both cases adjusted for seasoning to reflect expected lifetime losses and defaults, respectively.

Spain archetypal foreclosure frequency anchors

187. The following table shows the archetypal foreclosure frequency anchors.

Table 10

Spain foreclosure frequency adjustments

188. The following table shows the foreclosure frequency adjustments.

Table 11

CREDIT ANALYSIS OF MORTGAGE POOLS

Adjustment Factors For Variations From The Archetypal Pool

Weighted average frequency of foreclosure (WAFF)

Originator adjustment 

189. The calculation of foreclosure frequency includes an originator adjustment. Specific examples of the factors considered in determining the originator adjustment for a pool of Spanish residential loans include:

190. An underwriter's assessment of loan affordability (subject to loan-to-income or debt-to-income data received from the servicer, we could further adjust our originator adjustment);

• Data availability;
• Loans to borrowers in temporary employment;
• Loans to borrowers with negative credit histories;
• Adverse credit history not penalized at the loan-by-loan level;
• Scenarios where the performance of a pool that has been sold deviates from our expectations for pools from that originator and for which a neutral originator adjustment is assumed;
• Recent changes in product offering or credit score process, where the impact of which is not yet visible in performance metrics;
• Weak representations and warranties of the loans in the transaction documentation and pool audit results;
• Selection biases or pools of assets considered to have material tail risk;
• Positive or negative selection not captured in other adjustments; and
• Dynamic or revolving asset pools.

191. For covered bonds and master trust structures, we may reduce the foreclosure frequency to reflect the sponsor bank's willingness and ability to continue managing the cover pool or master trust. This assessment considers the following factors in particular:

• The existence in the transaction documentation of a periodic test of the pool's credit quality;
• The rating on the sponsor bank;
• The importance of the program in the sponsor bank's funding mix;
• The frequency of issuance from the program;
• The number of different covered bond programs and master trusts that a bank runs;
• Whether a bank differentiates the way in which it manages the pools backing such covered bond and master trust;
• The proximity of the current seller share to the minimum seller share; and
• The higher the perceived importance of a master trust to an originator or servicer, the higher the positive adjustment. This is because the importance of a master trust to the originator or servicer is likely to influence the performance of a transaction.

192. The criteria set out what we consider the potential changes of credit risk over time. For structures backed by a pool whose assets change (e.g., by virtue of loan substitutions, product switches or similar, or revolving), in determining the pool's weighted-average foreclosure frequency (WAFF) and the weighted-average loss severity (WALS), we consider the potential increase in credit risk over time as a result of changes in pool composition. We assess possible deterioration in the pool composition, based on the transaction's documented asset-eligibility criteria; the history of the originator; and, in particular, any observed changes in origination, underwriting, and related performance.

Geographic concentration 

193. The following table shows the region concentration thresholds. An adjustment of 1.25x is applied to the excess above the thresholds. If a pool has significant geographical concentration risk that we believe is not sufficiently captured, we may capture it using the originator adjustment.

Table 12

Nonresidential use loans and non residential borrowers 

194. Adjustments are applied to assets where we consider the primary use of the property to be nonresidential. Examples of nonresidential assets use include private individuals to purchase a commercial or mixed-use property and commercial borrowers. The adjustments detailed in the table below are intended for use where 40% or less of the pool are considered loans to be nonresidential use.

Table 13

Reperforming loans 

195. We apply adjustments for reperforming loans when a portfolio contains a material portion of reperforming loans.

196. We may consider different adjustments on a case-by-case basis, depending on the servicers' restructure policies, track record, and performance data provided.

197. We typically define a reperforming loan as a loan that has been 90 or more days past due or restructured in the five years leading up to the analysis date and is current as of that date.

198. When a reperforming arrangement is made, a full reassessment of the borrower's affordability capacity is typically made. We consider this akin to a reunderwriting of the loan. Accordingly, for pools classified as reperforming, we calculate potential future seasoning credit based on the date on which a loan was last 90 or more days in arrears or was restructured.

199. We typically apply adjustments for reperforming loans as shown in the following table. We may consider different adjustments on a case-by-case basis, depending on the servicers' restructure policies, track record, and performance data provided.

Table 14

200. In addition to the original loan and borrower information provided, we may also consider updated data sourced through the restructuring process in our analysis of reperforming loans, where available, on a case-by-case basis.

201. In addition, as part of the analytical process, we analyze data from the issuer or servicer on re-default rates stratified by forbearance type. This analysis is used to calibrate the originator adjustment for those transactions.

Second-lien loan 

Table 15

Bridge loans 

202. Bridge loans in Spain have a risk profile similar to second homes loans. Typically, it is granted to a borrower that at origination has not yet been able to sell his original home and buys a new property.

203. In a first phase, there will be a larger loan backed by two properties and an initial interest-only period of three to five years may apply. In a second phase, when the borrower has sold their first home, the balance of the loan will reduce, and it will only be backed by one property.

Weighted average loss severity (WALS)

Table 16

204. On a case-by-case basis, we may increase the forced sales discount where there is evidence to support that it may be either higher or lower than envisaged using the standard calculation.

Cash Flow Assumptions

205. In our rating analysis, we also perform an analysis of a transaction's payment structure and cash flow mechanics. This analysis uses our own quantitative models to assess whether the cash flows from the assets suffice, at the applicable rating levels, to make timely payments of interest and ultimate payment of principal (i.e., by or before the legal maturity date).

206. In our cash flow analysis, we use as inputs the pool-level WAFF and WALS described in previous sections of this guidance, to reflect credit stress at each rating level.

207. During modeling, cash flow stresses test the credit and liquidity support the assets need, considering any available structural support, such as a cash reserve, a liquidity facility, or hedging arrangements.

208. For revolving stand-alone RMBS structures (i.e., structures backed by a pool whose assets change or revolve), the modeling approach aims to reflect the structure after the activation of any "stop-substitution" (or early amortization) triggers, and to apply cash flow stresses from this point. A stop-substitution trigger is an event or situation that halts the substitution of assets in a revolving loan pool. We do not apply this approach to pools supporting covered bonds, even though the assets in these pools may be substituted over time. This is because the starting assumption of the collateral analysis under our covered bond criteria is the default of the issuing bank, and, as a result, we do not expect the cover pool to be actively managed (i.e., it would become a static pool).

209. This guidance does not include the assumptions used to assess refinancing costs in covered bonds structured with an asset-liability mismatch (such as target asset spreads). Those are described in "Covered Bonds Criteria," published Dec. 9, 2014.

210. In our surveillance of existing ratings, cash flow modeling may show that under the 'B' stress a particular tranche will miss interest payments or fail to repay the principal by or before the final legal maturity date. If this is the case, then our initial assessment under this guidance, all factors remaining the same, may be to consider lowering the rating on those securities to 'B-' or lower.

211. Depending on our view of a transaction's immediate cash flow position, the rating could move into the 'CCC', 'CC', or 'C' rating category, consistent with our ratings definitions , and in accordance with "Criteria For Assigning 'CCC+', 'CCC', 'CCC-', And 'CC' Ratings," published Oct. 1, 2012.

Defaults 

212. Under this guidance, the cumulative amount of defaults for cash flow modeling is equal to the pool's WAFF, whereas the cumulative amount of recoveries is calculated as (1 – WALS).

213. We typically model two default-timing scenarios, referred to as "front-loaded" (i.e., concentrated toward the earlier stage of a transaction) and "back-loaded" (i.e., concentrated toward the later stage of a transaction). In both scenarios, the defaults occurring over a six-year period, with each scenario having a peak level of stress, referred to as a recession, where 25% of the expected WAFF is applied annually for three years. For the front-loaded curve, the recession starts at day one; for the back-loaded curve, the recession begins at the start of the fourth year. For certain structure types (e.g., master trust transactions), we may run additional analysis to test the sensitivity of liability structures to different default timing curves.

Table 17

214. Foreclosure period assumptions, which represent the estimated time to repossess and sell a property upon a default, and reflect the typical time necessary for judicial proceedings and any other likely delay, are detailed in the table above.

215. Cash flow modeling considers the negative carry resulting from interest due on the rated liabilities during the foreclosure period.

216. The loss severity estimates used in the cash flow modeling are based on the loan principal and assume no recovery of interest accrued on the mortgage loans during the foreclosure period.

Delinquency 

217. We assume a delay of a proportion of scheduled interest and principal receipts equal to one-third of the WAFF in each of the first 18 months of a hypothetical recession, and set full recovery of the arrears to take place 36 months after the delinquency occurs. The cash flow stress for delinquencies is independent of the arrears adjustment to the WAFF.

218. For pools that contain residential loans with an option to temporarily suspend the periodic payments (payment holiday loans), the guidance includes a delay of a proportion of scheduled interest and principal receipts. The proportion is based on historical data on borrowers that have exercised such an option, assuming they use this option for the maximum duration permitted under the contractual terms of their loans. In situations where there is the potential for payment holidays to be granted after a loan's inception or where payment holidays have been granted (e.g., due to government and bank forbearance measures for households and small and midsize corporates), we may apply an additional stress in our cash flow analysis, where relevant. In those instances, the guidance includes a delay of a proportion of scheduled interest and principal receipts, based on an estimate of the proportion of a pool that opts to take a payment holiday and the likely duration of the holiday. The likely duration will be assessed with reference to factors that may include, but are not limited to, relevant legislative frameworks, collateral credit quality, servicers' policies, and available servicer data on payment holidays granted.

Interest, Prepayment, And Reinvestment Rates

Interest rate risk 

219. The guidance typically applies two different interest rate scenarios--"up" and "down"--for modeling purposes (see "Credit Rating Model: CIR Model," published April 4, 2019). The curves vary by stress scenario. If a transaction features an interest rate cap--either on the note coupon or through a cap agreement with an external counterparty-- upward interest rate stress assumptions exceeding the cap level may be unduly beneficial for the transaction's cash flow projection. In such event, we may apply a different interest rate stress to test sensitivity of the ratings to the absence of the cap.

220. Specific structural features may involve using additional cash flow stresses, such as alternative interest rate patterns or different default timing curves, among others. In particular, to analyze cash flows of covered bonds, including bullet maturities, we may apply a stochastic approach to stress interest rate and currency risks using our Covered Bond Monitor cash flow model (see the Related Research section below).

Basis risk 

221. Basis risk arises whenever unhedged differences exist between the methodologies for calculating interest on the assets and the liabilities of a structure. It differs from interest rate risk. For example, it may arise because of differences between two floating-rate indices, or when swaps are present and there is a timing mismatch between the reset dates on the securities and the swap.

222. To address this risk, the guidance adjusts the interest spread modelled between assets and liabilities by applying different spreads over the life of a transaction as a haircut to the margin. The size of the spreads depends on the distribution of historical differences among indices, using the rating-specific values corresponding to the percentiles shown in table 18 below.

223. Our current adjustments for typical indices are presented here: https://www.standardandpoors.com/pt_LA/web/guest/article/-/view/sourceId/100043040

224. The guidance applies the stress corresponding to a rating level in the cash flow analysis for the first 18 months of a hypothetical recession. After this 18-month period, the 'B' percentile from table 18 below applies at all rating levels.

225. Adjustments for indices not detailed in the link above would be devised on a case-by-case basis using a similar approach and applied in the same way.

Table 18

226. For example, in a Spanish RMBS transaction, the underlying mortgage loans may incur interest based on the one-month EURIBOR resetting monthly, but the securities may pay interest based on the three-month EURIBOR. In this case, our analysis looks at the distribution of the historical differences between these two rates, calculated by taking the highest three-month EURIBOR over the previous three-month period and the lowest one-month EURIBOR for each point in the data set. It then takes the percentiles of the resulting distribution shown above.

227. In another example, for a RMBS transaction whose mortgage loan pool contains loans that pose no basis risk during their promotional period but will revert to a floating margin that carries basis risk, the level of stress modelled depends on the proportion of the loans in the pool that would eventually have basis-risk exposure.

Prepayment scenarios 

228. Residential loan prepayments vary the amount of excess spread available and may affect the absolute level of defaults exhibited in a transaction. Therefore, we define a framework for assessing prepayment risk in the following paragraphs. When analyzing the payment structure and cash flow mechanics of Spanish RMBS, we typically test the transaction's ability to withstand high and low prepayment scenarios as set out in table 19.

229. We may adjust the prepayment assumptions if a pool's historical prepayment rates were higher than historical averages or if a transaction were particularly sensitive to prepayment risk (e.g., excess spread notes). We may also reduce high prepayment stress in situations where long-term historical data support lower prepayment rate assumptions for a specific loan product.

Table 19

Reinvestment rates 

230. Reinvestment rate assumptions stress the yield from excess cash that becomes available (because of, for example, prepayments) and the revenues associated with any other cash the issuer holds. The guidance applies different reinvestment rate assumptions to covered bonds with a bullet maturity structure than to RMBS transactions and covered bonds with a pass-through maturity structure (see table 20 below). This differentiation reflects our general expectation for a longer tenor of reinvestment for a structure with bullet maturities than for a pass-through structure. The guidance applies a floor to the reinvestment rate of 0%.

Table 20

Originator Insolvency, Commingling, And Set-Off

231. Our analysis of commingling risk that can result from an originator's or servicer's insolvency follows one of three approaches: (i) application of our current counterparty criteria, (ii) modeling to produce estimates of any exposure, or (iii) a combination of these two approaches. The combination of the first and second approaches applies if the application of the counterparty criteria shows residual cash flow exposures. The rest of this section focuses on the modeling of risk exposures that are not fully mitigated under our counterparty criteria.

232. Our legal analysis may conclude that the insolvency of the collection account holder could result in a loss of funds deposited in the account. This analysis considers the degree to which a collection account holder's insolvency affects the cash flow from the residential loan pool, if the collection account is not in the issuer's name.

233. The amount at risk depends on the timing of scheduled payments from borrowers and the frequency of transfers into the transaction account.

234. If the loan originator (or servicer) is also a deposit-taking institution, we perform legal analysis to determine if deposit set-off is a risk to the transaction.

235. However, the modeling approach aims to capture set-off risks from any borrowers that are also employees of the originator. This is because amounts owed to them, such as unpaid salaries, pension benefits, and subsidies, could offset payments due on their mortgage loans. As a result, the modeling approach treats the full amount of any employee's mortgage loan balance as entirely set off and all set-off amounts as principal losses.

Modeling Of Senior Fees And Expenses, Liquidity Facilities, And Spread Compression

236. The modeling of all of the issuer's foreseeable expenses uses stressed costs to reflect the need to replace the initial service provider.

237. The most significant of these is the cost of servicing mortgage loan receivables. We typically model the following servicing fees:

• For certain originators, based on the quality and historical performance of their underlying assets, the type of servicing platform, size of their asset portfolio, and related economies of scale, we apply a fee equal to the higher of 1.5x the contractual fee and 35 basis points (bps). This reflects the quality and performance of these prime originators and their servicing platform, all of which should contribute to the likely lower cost of replacing the servicer. Specifically, arrears levels tend to be significantly low; the servicing platform is standardized, integrated, and centralized; and the servicer is a larger market participant where the scale of loan portfolio to be transferred on replacement would offer the replacement servicer potential economies of scale.
• In all other instances, we apply a fee equal to the higher of 1.5x the contractual fee and 50 bps.

238. In situations where there is a contracted back-up servicer in place, we may adjust the assumptions in the paragraph above. In addition, in situations where fees are linked directly to specific activities, such as issuing letters, or where increases are linked to inflation, we may apply different assumptions from those detailed in the paragraph above.

239. We include estimates of marginal costs from liquidity facilities in the cash flow modeling. Most liquidity facilities are renewable after 364 days, subject to a commitment fee and a drawn fee.

240. We assume that the liquidity facility is fully drawn on day one of the analysis.

241. The drawn fee is modelled as being payable from this point on the whole facility balance, unless the documents state that the issuer does not have to pay a fee if the drawings on the facility did not originate from the issuer. This situation could arise because of nonrenewal of the facility or a downgrade of the facility provider. The commitment fee is modeled as per the documented terms.

242. We also model the possibility that the spread on the loan pool compresses over time, due to defaults, prepayments, and product switches. To reflect this, we reduce margins with the assumption that a percentage of the higher-yielding loans exit the portfolio. In addition, a RMBS transaction may feature contractual minimum yield levels, for example, to allow substitutions to continue in revolving pools. In that scenario, the modeling approach aims to capture any breach of the yield levels that triggers the end of the revolving period and produces lower yields.

Guidance: Portugal

Portugal background

243. The MMA for Portugal is "high risk" ('4' on a scale of '1' to '6'). This is based on an economic risk score of '4', a mortgage industry risk score of '3', and our assessment of the recourse available to lenders in the market. This reflects our view of:

• The risks in the Portuguese banking system;
• Portugal's historical unemployment sensitivity to changes in economic output, which could negatively affect borrowers' ability to pay their mortgages absent a strong social welfare system in relation to mortgage payments; and
• The willingness to pay in low or negative equity scenarios due to full-recourse.

Portugal - archetypal pool

244. We define the archetypal pool for Portugal as follows:

Table 21

Portugal 'AAA' and 'B' foreclosure frequency anchors

245. Based on an MMA of "high risk" ('4'), we set Portugal's 'AAA' foreclosure frequency anchor for the archetypal pool at 15%, and the current 'B' foreclosure frequency assumption at 2.5%.

246. The 2.5% 'B' foreclosure frequency assumption reflects our assessment of historical performance and our expectations for future performance over the medium term, given anticipated macroeconomic conditions. In our analysis, we considered loss expectations across mortgage portfolios and the performance of outstanding S&P Global Ratings-rated Portuguese RMBS transactions and covered bonds, in both cases adjusted for seasoning to reflect expected lifetime losses and defaults, respectively.

Portugal archetypal foreclosure frequency anchors

247. The table below shows the archetypal foreclosure frequency anchors.

Table 22

Portugal foreclosure frequency adjustments

248. The following table shows the foreclosure frequency adjustments.

Table 23

CREDIT ANALYSIS OF MORTGAGE POOLS

Adjustment Factors For Variations From The Archetypal Pool

Weighted average frequency of foreclosure (WAFF)

Originator adjustment 

249. The calculation of foreclosure frequency includes an originator adjustment. Specific examples of the factors considered in determining the originator adjustment for a pool of Portuguese residential loans include:

• An underwriter's assessment of loan affordability (subject to the loan-to-income or debt-to-income data received from the servicer, we could further adjust our originator adjustment);
• Data availability;
• Loans to borrowers in temporary employment;
• Loans to borrowers with negative credit histories;
• Adverse credit history not penalized at the loan-by-loan level;
• Scenarios where the performance of a pool that has been sold deviates from our expectations for pools from that originator and for which a neutral originator adjustment is assumed;
• Recent changes in product offering or credit score process, where the impact of which is not yet visible in performance metrics;
• Weak representations and warranties of the loans in the transaction documentation and pool audit results;
• Selection biases or pools of assets considered to have material tail risk;
• Positive or negative selection not captured in other adjustments; and
• Dynamic or revolving asset pools.

250. For covered bonds and master trust structures, we may reduce the foreclosure frequency to reflect the sponsor bank's willingness and ability to continue managing the cover pool or master trust. This assessment considers the following factors in particular:

• The existence in the transaction documentation of a periodic test of the pool's credit quality;
• The rating on the sponsor bank;
• The importance of the program in the sponsor bank's funding mix;
• The frequency of issuance from the program;
• The number of different covered bond programs and master trusts that a bank runs;
• Whether a bank differentiates the way in which it manages the pools backing such covered bond and master trust;
• The proximity of the current seller share to the minimum seller share; and
• The higher the perceived importance of a master trust to an originator or servicer, the higher the positive adjustment. This is because the importance of a master trust to the originator or servicer is likely to influence the performance of a transaction.

251. The criteria set out what we consider potential changes to credit risk over time. For structures backed by a pool whose assets change (e.g., by virtue of loan substitutions, product switches or similar, or revolving), in determining the pool's WAFF and the weighted-average loss severity, we consider the potential increase in credit risk over time as a result of changes in pool composition. We assess possible deterioration in the pool composition, based on the transaction's documented asset-eligibility criteria; the history of the originator; and, in particular, any observed changes in origination, underwriting, and related performance.

Geographic concentration

252. The following table shows the region concentration thresholds. An adjustment of 1.25x is applied to the excess above the thresholds. If a pool has significant geographical concentration risk that we believe is not sufficiently captured, we may capture it using the originator adjustment.

Table 24

Nonresidential use loans and nonresidential borrowers 

253. Adjustments are applied to assets where we consider the primary use of the property to be nonresidential. Examples of nonresidential assets use include private individuals to purchase a commercial or mixed-use property and commercial borrowers. The adjustments detailed in the table below are intended for use where 40% or less of the pool are considered to be loans for nonresidential use.

Table 25

Reperforming loans 

254. We apply adjustments for reperforming loans when a portfolio contains a material portion of reperforming loans.

255. We may consider different adjustments on a case-by-case basis, depending on the servicers' restructure policies and track record, and the performance data provided.

256. We typically define a reperforming loan as a loan that has been 90 or more days past due or restructured in the five years leading up to the analysis date and is current as of that date.

257. When a reperforming arrangement is made, a full reassessment of the borrower's affordability capacity is typically made. We consider this akin to a reunderwriting of the loan. Accordingly, for pools classified as reperforming, we calculate potential future seasoning credit based on the date on which a loan was last 90 or more days in arrears or was restructured.

258. We typically apply adjustments for reperforming loans as shown in the following table. We may consider different adjustments on a case-by-case basis, depending on the servicers' restructure policies, track record, and performance data provided.

Table 26

259. In addition to the original loan and borrower information provided, we may also consider updated data sourced through the restructuring process in our analysis of reperforming loans, where available, on a case-by-case basis.

260. In addition, as part of the analytical process, we analyze data from the issuer or servicer on re-default rates stratified by forbearance type. This analysis is used to calibrate the originator adjustment for those transactions.

Second-lien loan 

Table 27

Weighted average loss severity (WALS)

Table 28

261. On a case-by-case basis, we may increase the forced sales discount where there is evidence to support that it may be either higher or lower than envisaged using the standard calculation.

Additional Considerations For Government-Subsidized Loans

262. The guidance aligns the treatment of government-subsidized loans in our analysis with our methodology for assessing mortgage insurance. Government guarantees and support at the loan level, which is featured in certain Portuguese RMBS transactions, are analyzed in accordance with the ratings on securities and the sovereign credit rating on Portugal, following the approach described in the mortgage insurance criteria. In doing so, the guidance gives partial credit to a guarantee or subsidy from an entity rated lower than the notes that the guarantee or subsidy is supporting. Haircuts and cash flow modeling for these loans, however, differ from the approach described in the mortgage insurance criteria, due to the difference in the nature of the guarantee relative to traditional mortgage insurance. A simplified modeling approach is employed--where we assume a certain percentage of subsidized interest due is lost in the first 18 months of the recession--at rating levels above the sovereign rating and according to the following schedule:

• 75% of subsidized interest is lost in the first 18 months of the recession at rating levels up to four notches above the sovereign.
• 100% of subsidized interest is lost in the first 18 months of the recession at rating levels greater than four notches above the sovereign.

263. A maximum potential subsidy of 44% is assumed when there is no loan-level information or aggregate investor report information on subsidies since the transaction closed. Additionally, for simplicity, the subsidy is kept constant over the 18-month period of the analysis and does not take into account the fact that it would actually decline over time. Moreover, the modeling of defaults and the liquidity stress that result from short-term delinquency assumptions are distributed evenly across the pool, thus affecting all loans, including subsidized loans. Nevertheless, the above analysis assumes that no subsidized loans are affected, which increases the impact of the liquidity stress from short-term delinquencies, all else being equal.

Cash Flow Assumptions

264. In our rating analysis, we also perform an analysis of a transaction's payment structure and cash flow mechanics. This analysis uses our own quantitative models to assess whether the cash flows from the assets suffice, at the applicable rating levels, to make timely payments of interest and ultimate payment of principal (i.e., by or before the legal maturity date).

265. In our cash flow analysis, we use as inputs the pool-level WAFF and WALS described in previous sections of this guidance to reflect credit stress at each rating level.

266. During modeling, cash flow stresses test the credit and liquidity support the assets need, considering any available structural support, such as a cash reserve, a liquidity facility, or hedging arrangements.

267. For revolving stand-alone RMBS structures (i.e., structures backed by a pool whose assets change or revolve), the modeling approach aims to reflect the structure after the activation of any "stop-substitution" (or early amortization) triggers, and to apply cash flow stresses from this point. A stop-substitution trigger is an event or situation that halts the substitution of assets in a revolving loan pool. We do not apply this approach to pools supporting covered bonds, even though the assets in these pools may be substituted over time. This is because the starting assumption of the collateral analysis under our covered bond criteria is the default of the issuing bank, and, as a result, we do not expect the cover pool to be actively managed (i.e., it would become a static pool).

268. This guidance does not include the assumptions used to assess refinancing costs in covered bonds structured with an asset-liability mismatch (such as target asset spreads). Those are described in "Covered Bonds Criteria," published Dec. 9, 2014.

269. In our surveillance of existing ratings, cash flow modeling may show that under the 'B' stress a particular tranche will miss interest payments or fail to repay the principal by or before the final legal maturity date. If this is the case, then our initial assessment under this guidance, all factors remaining the same, may be to consider lowering the rating on those securities to 'B-' or lower.

270. Depending on our view of a transaction's immediate cash flow position, the rating could move into the 'CCC', 'CC', or 'C' rating categories, consistent with our ratings definitions in "S&P Global Ratings Definitions," published Aug. 7, 2020, and in accordance with "Criteria For Assigning 'CCC+', 'CCC', 'CCC-', And 'CC' Ratings," published Oct. 1, 2012.

Defaults 

271. Under this guidance, the cumulative amount of defaults for cash flow modeling is equal to the pool's WAFF, whereas the cumulative amount of recoveries is calculated as (1 – WALS).

272. We typically model two default-timing scenarios, referred to as "front-loaded" (i.e., concentrated toward the earlier stage of a transaction) and "back-loaded" (i.e., concentrated toward the later stage of a transaction). In both scenarios, the defaults occur over a six-year period, with each scenario having a peak level of stress, referred to as a recession, where 25% of the expected WAFF is applied annually for three years. For the front-loaded curve, the recession starts at day one; for the back-loaded curve, the recession begins at the start of the fourth year. For certain structure types (e.g., master trust transactions), we may run additional analysis to test the sensitivity of liability structures to different default timing curves.

Table 29

273. Foreclosure period assumptions, which represent the estimated time to repossess and sell a property upon a default, and reflect the typical time necessary for judicial proceedings and any other likely delay, are outlined in table 29 above.

274. Cash flow modeling considers the negative carry resulting from interest due on the rated liabilities during the foreclosure period.

275. The loss severity estimates used in the cash flow modeling are based on the loan principal and assume no recovery of interest accrued on the mortgage loans during the foreclosure period.

Delinquency 

276. We assume a delay of a proportion of scheduled interest and principal receipts equal to one-third of the WAFF in each of the first 18 months of a hypothetical recession, and set full recovery of the arrears to take place 36 months after the delinquency occurs. The cash flow stress for delinquencies is independent of the arrears adjustment to the WAFF.

277. For pools that contain residential loans with an option to temporarily suspend the periodic payments (payment holiday loans), the guidance includes a delay of a proportion of scheduled interest and principal receipts. The proportion is based on historical data on borrowers that have exercised such an option, assuming they use this option for the maximum duration permitted under the contractual terms of their loans. In situations where there is the potential for payment holidays to be granted after a loan's inception or where payment holidays have been granted (e.g., due to government and bank forbearance measures for households and small and midsize corporates), we may apply an additional stress in our cash flow analysis, where relevant. In those instances, the guidance includes a delay of a proportion of scheduled interest and principal receipts, based on an estimate of the proportion of a pool that opts to take a payment holiday and the likely duration of the holiday. The likely duration will be assessed with reference to factors that may include, but are not limited to, relevant legislative frameworks, collateral credit quality, servicers' policies, and available servicer data on payment holidays granted.

Interest, Prepayment, And Reinvestment Rates

Interest rate risk 

278. The guidance typically applies two different interest rate scenarios--"up" and "down"--for modeling purposes (see "Credit Rating Model: CIR Model," published April 4, 2019). The curves vary by stress scenario. If a transaction features an interest rate cap--either on the note coupon or through a cap agreement with an external counterparty--upward interest rate stress assumptions exceeding the cap level may be unduly beneficial for the transaction's cash flow projection. In such event, we may apply a different interest rate stress to test sensitivity of the ratings to the absence of the cap.

279. Specific structural features may involve using additional cash flow stresses, such as alternative interest rate patterns or different default timing curves, among others. In particular, to analyze cash flows of covered bonds, including bullet maturities, we may apply a stochastic approach to stress interest rate and currency risks using our Covered Bond Monitor cash flow model (see the Related Research below).

Basis risk 

280. Basis risk arises whenever unhedged differences exist between the methodologies for calculating interest on the assets and the liabilities of a structure. It differs from interest rate risk. For example, it may arise because of differences between two floating-rate indices, or when swaps are present and there is a timing mismatch between the reset dates on the securities and the swap.

281. To address this risk, the guidance adjusts the interest spread modelled between assets and liabilities by applying different spreads over the life of a transaction as a haircut to the margin. The size of the spreads depends on the distribution of historical differences among indices, using the rating-specific values corresponding to the percentiles shown in table 30 below.

282. Our current adjustments for typical indices are presented here: https://www.standardandpoors.com/pt_LA/web/guest/article/-/view/sourceId/100043040

283. The guidance applies the stress corresponding to a rating level in the cash flow analysis for the first 18 months of a hypothetical recession. After this 18-month period, the 'B' percentile from table 30 below applies at all rating levels.

284. Adjustments for indices not detailed in the link above would be devised on a case-by-case basis using a similar approach and applied in the same way.

Table 30

285. For example, in a Portuguese RMBS transaction, the underlying mortgage loans may incur interest based on the one-month EURIBOR resetting monthly, but the securities may pay interest based on the three-month EURIBOR. In this case, our analysis looks at the distribution of the historical differences between these two rates, calculated by taking the highest three-month EURIBOR over the previous three-month period and the lowest one-month EURIBOR for each point in the dataset. It then takes the percentiles of the resulting distribution shown above.

286. In another example, for a RMBS transaction whose mortgage loan pool contains loans that pose no basis risk during their promotional period but will revert to a floating margin that carries basis risk, the level of stress modeled depends on the proportion of the loans in the pool that would eventually have basis-risk exposure.

Prepayment scenarios 

287. Residential loan prepayments vary the amount of excess spread available and may affect the absolute level of defaults exhibited in a transaction. Therefore, we define a framework for assessing prepayment risk in the following paragraphs. When analyzing the payment structure and cash flow mechanics of Portuguese RMBS, we typically test the transaction's ability to withstand high and low prepayment scenarios as set out in table 31.

288. We may adjust the prepayment assumptions if a pool's historical prepayment rates were higher than historical averages or if a transaction were particularly sensitive to prepayment risk (e.g., excess spread notes). We may also reduce high prepayment stress in situations where long-term historical data support lower prepayment rate assumptions for a specific loan product.

Table 31

Reinvestment rates 

289. Reinvestment rate assumptions stress the yield from excess cash that becomes available (i.e., because of prepayments) and the revenues associated with any other cash the issuer holds. The guidance applies different reinvestment rate assumptions to covered bonds with a bullet maturity structure than those to RMBS transactions and covered bonds with a pass-through maturity structure (see table below). This differentiation reflects our general expectation for a longer tenor of reinvestment for a structure with bullet maturities than for a pass-through structure. The guidance applies a floor to the reinvestment rate of 0%.

Originator Insolvency, Commingling, And Set-Off

290. Our analysis of commingling or set-off risks that can result from an originator's or servicer's insolvency follows one of three approaches: application of our current counterparty criteria, modeling to produce estimates of any exposure, or a combination of these two approaches. The combination of the first and second approaches applies if the application of the counterparty criteria shows residual cash flow exposures. The rest of this section focuses on the modeling of risk exposures that are not fully mitigated under our counterparty criteria.

291. Our legal analysis considers the degree to which a collection account holder's insolvency affects the cash flow from the residential loan pool, if the collection account is not in the issuer's name.

292. The amount at risk depends on the timing of scheduled payments from borrowers, and the frequency of transfers into the transaction account.

293. If the loan originator (or servicer) is also a deposit-taking institution, the modeling considers the possibility that borrowers with deposit accounts may set off these amounts against their outstanding mortgage loans. In arriving at estimates of potential set off exposures, we will consider historical data on deposit levels, the borrower profile, and an assessment of factors that may lead to a material increase in set off exposure. We will also look at any mitigants such as the Portuguese government's deposit guarantee scheme.

294. The modeling approach aims to capture set-off risks from any borrowers that are also employees of the originator. This is because amounts owed to them, such as unpaid salaries, pension benefits, and subsidies, could offset payments due on their mortgage loans. As a result, the modeling approach treats the full amount of any employee's mortgage loan balance as entirely set off and all set-off amounts as principal losses.

Modeling Of Senior Fees And Expenses, Liquidity Facilities, And Spread Compression

295. The modeling of all of the issuer's foreseeable expenses uses stressed costs to reflect the need to replace the initial service provider.

296. The most significant of these is the cost of servicing mortgage loan receivables. We typically model the following servicing fees:

• For certain originators, based on the quality and historical performance of their underlying assets, the type of servicing platform, size of their asset portfolio, and related economies of scale, we apply a fee equal to the higher of 1.5x the contractual fee and 35 bps. This reflects the quality and performance of these prime originators and their servicing platform, all of which should contribute to the likely lower cost of replacing the servicer. Specifically, arrears levels tend to be significantly low, the servicing platform is standardized, integrated, and centralized, and the servicer is a larger market participant where the scale of loan portfolio to be transferred on replacement would offer the replacement servicer potential economies of scale.
• In all other, instances we apply a fee equal to the higher of 1.5x the contractual fee and 50 bps.

297. In situations where there is a contracted back-up servicer in place, we may make adjust the assumptions in the paragraph above. In addition, in situations where fees are linked directly to specific activities, such as issuing letters, or where increases are linked to inflation, we may apply different assumptions from those detailed in the paragraph above.

298. We include estimates of marginal costs from liquidity facilities in the cash flow modeling. Most liquidity facilities are renewable after 364 days, subject to a commitment fee and a drawn fee.

299. We assume that the liquidity facility is fully drawn on day one of the analysis.

300. The drawn fee is modelled as being payable from this point on the whole facility balance, unless the documents state that the issuer does not have to pay a fee if the drawings on the facility did not originate from the issuer. This situation could arise because of nonrenewal of the facility or a downgrade of the facility provider. The commitment fee is modeled as per the documented terms.

301. We also model the possibility that the spread on the loan pool compresses over time due to defaults, prepayments, and product switches. To reflect this, we reduce margins with the assumption that a percentage of the higher-yielding loans exit the portfolio. In addition, an RMBS transaction may feature contractual minimum yield levels, for example, to allow substitutions to continue in revolving pools. In that scenario, the modeling approach aims to capture any breach of the yield levels that triggers the end of the revolving period and produces lower yields.

Guidance: Italy

Italy background

302. The MMA for Italy is "intermediate risk" ('3' on a scale of '1' to '6'). This is based on an economic risk score of '3', a mortgage industry risk score of '3', and our assessment of recourse available to lenders in the market. This reflects our view of:

• The risks in the Italian banking system;
• The Italian historical unemployment sensitivity to changes in economic output;
• The solid social welfare system;
• The generally low Italian household indebtedness; and
• The willingness to pay in low/negative equity scenarios due to full recourse.

Italy archetypal pool

303. We define the archetypal pool for Italy as follows:

Table 33

Italy archetypal foreclosure frequency anchors

304. The following table shows the archetypal foreclosure frequency anchors.

Table 34

The Italian foreclosure frequency adjustments

305. The following table shows the foreclosure frequency adjustments, which are typically applied at a property level.

Table 35

CREDIT ANALYSIS OF MORTGAGE POOLS

Adjustment Factors For Variations From The Archetypal Pool

Weighted average frequency of foreclosure (WAFF)

Originator adjustment  

306. The calculation of foreclosure frequency includes an originator adjustment. Specific examples of the factors considered in determining the originator adjustment for a pool of Italian residential loans include:

• An unusually large share of loans with negative credit histories and risk layering not typical of the market;
• Unavailable data on loan and borrower characteristics (for example, restructurings, defaults, and prior mortgage arrears);
• Recent changes in product offerings or credit score processes, the impact of which would not yet be visible in performance metrics;
• Selection biases or pools of assets considered to have material tail risk;
• Scenarios where an analysis of past originator loan delinquency/default data reveals performance that deviates from expectation when a neutral originator adjustment is assumed;
• The loan originator's assessment of loan-size risk; and
• Any transaction's specific geographical concentration not captured by regional concentrations.

307. For covered bonds and master trust structures, we may reduce the foreclosure frequency to reflect the sponsor bank's or originators willingness and ability to continue managing the cover pool or master trust. This assessment considers the following factors in particular:

• The existence in the transaction documentation of a periodic test of the pool's credit quality;
• The rating on the sponsor bank;
• The importance of the program in the sponsor bank's funding mix;
• The frequency of issuance from the program;
• The number of different covered bond programs and master trusts that a bank runs;
• Whether a bank differentiates the way in which it manages the pools backing such covered bond and master trust;
• The proximity of the current seller share to the minimum seller share; and
• The higher the perceived importance of a master trust to an originator or servicer, the higher the positive adjustment. This is because the importance of a master trust to the originator or servicer is likely to influence the performance of a transaction.

308. The criteria set out to consider the potential changes of credit risk over time. For structures backed by a pool whose assets change (e.g., by virtue of loan substitutions, product switches, or similar) or revolve, we consider the potential increase of credit risk over time as a result of changes in pool composition in determining the pool's weighted-average foreclosure frequency (WAFF) and the weighted-average loss severity (WALS). We assess possible deterioration in pool composition based on the transaction's documented asset-eligibility criteria, as well as the history of the originator and, in particular, any observed changes in origination, underwriting, and related performance.

Geographic concentration  

309. The following table shows the region concentration thresholds. An adjustment of 1.25x is applied to the excess above the thresholds. If a pool has significant geographical concentration risk that we believe is not sufficiently captured, we may capture it using the originator adjustment.

Table 36

Nonresidential-use loans and nonresidential borrowers  

310. Adjustments are applied to assets where we consider the primary use of the property to be nonresidential. Examples of nonresidential assets include mixed-use properties and small-scale commercial properties. We apply different adjustments if the borrower is an individual purchasing a commercial or a mixed-use property (1.5x) or if the borrower is also commercial (2.0x). The adjustments detailed in the table below are intended for use where less than or equal to 40% of the pool are considered to be nonresidential use.

Table 37

Origination channel  

311. We apply an adjustment factor up to 1.3x to loans originated through a broker or real estate agent. The adjustment also applies when the lender performs the entire credit analysis and the role of the broker or real estate agent is just to introduce the borrower. We may assume a lower adjustment factor than 1.3x if we receive evidence of strong performance of broker-originated loans.

Reperforming loans  

312. We apply adjustments for reperforming loans when a portfolio contains a material portion of reperforming loans.

313. We may consider different adjustments on a case-by-case basis, depending on the servicers' restructure policies, track record, and performance data provided.

314. We typically define a reperforming loan as a loan that has been 90 or more days past due or restructured in the five years leading up to the analysis date and is current as of that date.

315. When a reperforming arrangement is made, it is typical that a full reassessment of the borrower's affordability capacity is made; we consider this akin to a reunderwriting of the loan. Accordingly, for pools classified as reperforming, we calculate potential future seasoning credit based on the date a loan was last 90 or more days past due in arrears or restructured.

316. We typically apply adjustments for reperforming loans as per the table below. We may consider different adjustments on a case-by-case basis, depending on the servicer's restructure policies, track record, and performance data provided.

Table 38

317. In addition to the original loan and borrower information provided, we may also consider updated data sourced through the restructuring process in our analysis of reperforming loans where available, on a case-by-case basis.

318. In addition, as part of the analytical process, we analyze data from the issuer/servicer on re-default rates stratified by forbearance type. This analysis is used to calibrate the originator adjustment for such transactions.

Second-lien loan  

Table 39

Treatment of specific loan product types

Dynamic asset pools  

319. The criteria sets out that we consider the potential changes of credit risk over time. For structures backed by a pool whose assets change (e.g., by virtue of loan substitutions, product switches, or similar) or revolve, in determining the pool's WAFF and the weighted average loss severity (WALS), we consider the potential increase of credit risk over time as a result of changes in pool composition. We assess possible deterioration in pool composition based on the transaction's documented asset-eligibility criteria, as well as the history of the originator and, in particular, any observed changes in origination/underwriting and related performance.

Weighted average loss severity (WALS)

Table 40

320. On a case-by-case basis, we increase the forced sales discount where there is evidence to support that it may be either higher or lower than envisaged using the standard calculation.

Cash Flow Assumptions

321. For revolving stand-alone RMBS structures (i.e., structures backed by a pool whose assets change or revolve), the modeling approach aims to reflect the structure after the activation of any "stop-substitution" (or early amortization) triggers and applies cash flow stresses from this point. A stop-substitution trigger is an event or situation that halts the substitution of assets in a revolving loan pool. We do not apply this approach to pools supporting covered bonds, even though the assets in these pools may be substituted over time. This is because the starting assumption of the collateral analysis under our covered bond criteria is the default of the issuing bank, as a result of which we do not expect the cover pool to be actively managed (it would become a static pool).

322. This guidance does not include the assumptions used to assess refinancing costs in covered bonds structured with an asset-liability mismatch (such as target asset spreads). Those are described in "Covered Bonds Criteria," published Dec. 9, 2014.

323. In our surveillance of existing ratings, cash flow modeling may show that--under the 'B' stress--a particular tranche will miss interest payments or fail to repay the principal by or before the final legal maturity date. If this is the case, then our initial assessment under this guidance, all factors remaining the same, may be to consider lowering the rating on those securities to 'B-' or lower.

324. Depending on our view of a transaction's immediate cash flow position, the rating could move into the 'CCC', 'CC', or 'C' category, consistent with our ratings definitions in "S&P Global Ratings Definitions," published Aug. 7, 2020, and in accordance with "Criteria For Assigning 'CCC+', 'CCC', 'CCC-', And 'CC' Ratings," published Oct. 1, 2012.

Defaults 

325. Under this guidance, the cumulative amount of defaults for cash flow modeling is equal to the pool's WAFF, whereas the cumulative amount of recoveries is calculated as (1 – WALS).

326. We typically model two default-timing scenarios, referred to as "front-loaded" (i.e., concentrated toward the earlier stage of a transaction) and "back-loaded" (i.e., concentrated toward the later stage of a transaction). In both scenarios, the defaults occur over a six-year period, with each scenario having a peak level of stress (referred to as a recession) where 25% of the expected WAFF is applied annually for three years. For the front-loaded curve, the recession starts at day one, and for the back-loaded curve, the recession starts at the beginning of the fourth year. We typically run the default timing curves detailed in the table below. For certain structure types (e.g., master trust transactions), we may run additional analysis to test the sensitivity liability structures to different default-timing curves.

Table 41

327. The loss severity estimates used in the cash flow modeling are based on the loan principal and assume no recovery of interest accrued on the mortgage loans during the foreclosure period.

Delinquency 

328. We assume a delay of a proportion of scheduled interest and principal receipts equal to one-third of the WAFF in each of the first 18 months of a hypothetical recession, and set full recovery of the arrears to take place 36 months after the delinquency occurs. The cash flow stress for delinquencies is independent of the arrears adjustment to the WAFF.

329. For pools that contain residential loans with an option to temporarily suspend the periodic payments (payment holiday loans), the guidance includes a delay of a proportion of scheduled interest and principal receipts. The proportion is based on historical data on borrowers that have exercised such an option, assuming they use this option for the maximum duration permitted under the contractual terms of their loans. In situations where there is the potential for payment holidays to be granted after a loan's inception or where payment holidays have been granted (e.g., due to government and bank forbearance measures for households and small-to-midsize corporates), we may apply an additional stress in our cash flow analysis, where relevant. In these instances, the guidance includes a delay of a proportion of scheduled interest and principal receipts based on an estimate of the proportion of a pool that opts to take a payment holiday and the likely duration of the holiday. The likely duration will be assessed with reference to factors that may include, but are not limited to, relevant legislative frameworks, collateral credit quality, servicers' policies, and available servicer data on payment holidays granted.

Interest, prepayment, and reinvestment rates

Interest rate risk  

330. The guidance typically applies two different interest rate scenarios--"up" and "down"--for modeling purposes (see "Credit Rating Model: CIR Model," published April 4, 2019). The curves vary by stress scenario. In case a transaction features an interest rate cap--either on the note coupon or a cap agreement with an external counterparty--upward interest rate stress assumptions exceeding the cap level may be unduly beneficial for the transaction's cash flow projection. In such event, we may apply a different interest rate stress to test sensitivity of the ratings to the absence of the cap.

331. Specific structural features may involve using additional cash flow stresses, such as alternative interest rate patterns or different default-timing curves, among others. In particular, to analyze cash flows of covered bonds, including bullet maturities, we may apply a stochastic approach to stress interest rate and currency risks using our Covered Bond Monitor cash flow model (see the Related Research section).

Basis risk  

332. Basis risk arises whenever unhedged differences exist between, on the one hand, methodologies for calculating interest on the assets, and on the other hand, the liabilities of a structure. It differs from interest rate risk. For example, it may arise because of differences between two floating-rate indices, or when swaps are present and there is a timing mismatch between the reset dates on the securities and on the swap.

333. To address this risk, the guidance adjusts the interest spread modeled between assets and liabilities by applying different spreads over the life of a transaction as a haircut to the margin. The size of the spreads depends on the distribution of historical differences among indices, using the rating-specific values corresponding to the percentiles shown in the table below.

Table 42

334. The guidance applies the stress corresponding to a rating level in the cash flow analysis for the first 18 months of a hypothetical recession. Both before and after this 18-month period, the 'B' percentile from the table above applies at all rating levels.

335. Our current adjustments for typical indices are presented here: https://www.standardandpoors.com/pt_LA/web/guest/article/-/view/sourceId/100043040.

336. Adjustments for indices not detailed in the link would be devised on a case-by-case basis using a similar approach and applied in the same way.

337. For example, in an Italian RMBS or covered bond transaction, the underlying mortgage loans may incur interest based on the one-month EURIBOR resetting monthly, but the securities may pay interest based on the three-month EURIBOR. In this case, the analysis looks at the distribution of the historical differences between these two rates, calculated by taking the highest three-month EURIBOR over the previous three-month period and the lowest one-month EURIBOR for each point in the data set. It then takes the percentiles of the resulting distribution shown above.

338. In another example, for a RMBS transaction whose mortgage loan pool contains loans that pose no basis risk during their promotional period but will revert to a floating margin that carries basis risk, the level of stress modelled depends on the proportion of the loans in the pool that would eventually have basis-risk exposure.

Prepayment scenarios  

339. Residential loan prepayments vary the amount of excess spread available and may affect the absolute level of defaults exhibited in a transaction. Therefore, we define a framework for assessing prepayment risk in the following paragraphs. When analyzing the payment structure and cash flow mechanics of Italian RMBS, we typically test the transaction's ability to withstand high and low prepayment scenarios as set out in the following table.

Table 43

340. We may adjust the prepayment assumptions if a pool's historical prepayment rates were higher than historical country averages or if a transaction were particularly sensitive to prepayment risk (e.g., excess spread notes). We may also reduce high prepayment stress in situations where long-term historical data support lower prepayment rate assumptions for a specific loan product.

Reinvestment rates  

341. Reinvestment rate assumptions stress the yield from excess cash that becomes available because of, for example, prepayments, as well as the revenues associated with any other cash the issuer holds. The guidance applies different reinvestment rate assumptions to covered bonds with a bullet maturity structure on the one hand, and to RMBS transactions and covered bonds with a pass-through maturity structure on the other (see table below). This differentiation reflects that, for a structure with bullet maturities, we generally expect a longer tenor of reinvestment than for a pass-through structure. The guidance applies a floor to the reinvestment rate of 0%.

Table 44

Originator Insolvency, Commingling, And Set-Off

342. Our analysis of commingling risk that can result from an originator's or servicer's insolvency follows one of three approaches: (i) application of our current counterparty criteria, (ii) modeling to produce estimates of any exposure, or (iii) a combination of the two approaches. The combination of the first and second approaches applies if the application of the counterparty criteria shows residual cash flow exposures. The rest of this section focuses on the modeling of risk exposures that are not fully mitigated under our counterparty criteria.

343. Our legal analysis may conclude that the insolvency of the collection account holder could result in a loss of funds deposited in the account. This analysis considers the degree to which a collection account holder's insolvency affects the cash flow from the residential loan pool, if the collection account is not in the name of the issuer.

344. The amount at risk depends on the timing of scheduled payments from borrowers and the frequency of transfers into the transaction account.

345. If the loan originator (or servicer) is also a deposit-taking institution, modeling takes into consideration the possibility that borrowers with deposit accounts may set off these amounts against their outstanding mortgage loans. In arriving at estimates of potential set-off exposures, we will consider historical data on deposit levels, the borrower profile, and an assessment of factors that may lead to a material increase in set-off exposure. We will also look at any mitigants, such as deposit guarantee schemes.

346. The modeling approach aims to capture set-off risks from any borrowers that are also employees of the originator. This is because amounts owed to them, such as unpaid salaries, pension benefits, and subsidies, could offset payments due on their mortgage loans. As a result, the modeling approach treats the full amount of any employee's mortgage loan balance as entirely set off, and all set-off amounts as principal losses.

Modeling Of Senior Fees And Expenses, Liquidity Facilities, And Spread Compression

347. The modeling of all the issuer's foreseeable expenses uses stressed costs to reflect the need to replace the initial service provider.

348. The most significant of these is the cost of servicing mortgage loan receivables. We typically apply a fee equal to the higher of 1.5x the contractual fee and 35 bps. This reflects the average market fee for servicing activities in Italy. It also takes into account that the Italian market is well developed, dynamic, and has plenty of qualified servicers that are engaged in securitization transactions and available to step in as substitute servicers. In addition, servicing platforms are standardized. We apply a fee equal to the higher of 1.5x the contractual fee and 50 bps when we believe the assets are not prime, or when arrears levels tend to be significantly high.

349. In situations where there is a contracted back-up servicer in place, we may adjust the assumptions in the paragraph above. In addition, in situations where fees are linked directly to specific activities (such as issuing letters) or where increases are linked to inflation, we may apply different assumptions from those detailed in the paragraph above.

350. We include estimates of marginal costs from liquidity facilities in cash flow modeling. Most liquidity facilities are renewable after 364 days, subject to a commitment fee and a drawn fee. We assume that the liquidity facility is fully drawn on day one of the analysis. The drawn fee is typically modeled as being payable from this point on the whole facility balance, unless the documents state that the issuer does not have to pay a fee if the drawings on the facility did not originate from the issuer. Such a situation could arise because of nonrenewal of the facility or a downgrade of the facility provider. The commitment fee is modeled as per the documented terms.

351. We also model the possibility that the spread on the loan pool compresses over time, due to defaults, prepayments, and product switches. To reflect this, we reduce margins with the assumption that a percentage of the higher-yielding loans exit the portfolio. In addition, for RMBS, the transaction may feature contractual minimum yield levels to allow substitutions to continue in revolving pools. In such a case, the modeling approach aims to capture any breach of the yield levels that triggers the end of the revolving period and produces lower yields.

Guidance: Greece

Greece background

352. The MMA for Greece is "very high risk" ('5' on a scale of '1' to '6'). This is based on an economic risk score of '5', a mortgage industry risk score of '5', and our assessment of recourse available to lenders in the market. This reflects our view of:

• The risks in the Greek banking system;
• Greece's historical unemployment sensitivity to changes in economic output, which could negatively affect borrowers' ability to pay their mortgages in the absence of a strong social welfare system in relation to mortgage payments; and
• The willingness to pay in low/negative equity scenarios due to full recourse.

Greece archetypal pool

353. We define the archetypal pool for Greece as shown in the below table:

Table 45

Greece 'AAA' and 'B' foreclosure frequency anchors

354. Based on an MMA of "very high risk" ('5'), we set Greece's 'AAA' foreclosure frequency anchor for the archetypal pool at 28%, and the current 'B' foreclosure frequency assumption at 5%.

355. The 5% 'B' foreclosure frequency assumption reflects our assessment of historical performance and our expectations for future performance over the medium term given anticipated macroeconomic conditions. In our analysis, we considered loss expectations across mortgage portfolios, as well as the performance of S&P Global Ratings-rated Greek covered bonds.

Greece archetypal foreclosure frequency anchors

356. The table below shows the archetypal foreclosure frequency anchors.

Table 46

Greece foreclosure frequency adjustments

357. The following table shows the foreclosure frequency adjustments.

Table 47

CREDIT ANALYSIS OF MORTGAGE POOLS

Adjustment Factors For Variations From The Archetypal Pool

Weighted average frequency of foreclosure (WAFF)

Originator adjustment  

358. The calculation of foreclosure frequency includes an originator adjustment. Specific examples of the factors considered in determining the originator adjustment for a pool of Greek residential loans include:

• An underwriter's assessment of loan affordability (subject to loan-to-income or debt-to–income data received from the servicer, we could further adjust our originator adjustment);
• Data availability;
• Loans to borrowers in temporary employment;
• Loans to borrowers with negative credit histories;
• Adverse credit history not penalized at the loan-by-loan level;
• Scenarios where the performance of a pool that has been sold deviates from our expectations for pools from that originator and for which a neutral originator adjustment is assumed;
• Recent changes in product offerings or the credit score process, the impact of which would not be yet visible in performance metrics;
• Weak representations and warranties of the loans in the transaction documentation and pool audit results;
• Selection biases or pools of assets considered to have material tail risk;
• Positive or negative selection not captured in other adjustments;
• Dynamic or revolving asset pools;
• The originator's assessment of loan-size risk; and
• The presence of subsidized loans.

359. For covered bonds and master trust structures, we may reduce the foreclosure frequency to reflect the sponsor bank's willingness and ability to continue managing the cover pool or master trust. This assessment considers the following factors in particular:

• The existence in the transaction documentation of a periodic test of the pool's credit quality;
• The rating on the sponsor bank;
• The importance of the program in the sponsor bank's funding mix;
• The frequency of issuance from the program;
• The number of different covered bond programs and master trusts that a bank runs;
• Whether a bank differentiates the way in which it manages the pools backing such covered bond and master trust;
• The proximity of the current seller share to the minimum seller share; and
• The higher the perceived importance of a master trust to an originator or servicer, the higher the positive adjustment. This is because the importance of a master trust to the originator or servicer is likely to influence the performance of a transaction.

360. The criteria set out that we consider the potential changes of credit risk over time. For structures backed by a pool whose assets change (e.g., by virtue of loan substitutions, product switches, or similar, or revolving), in determining the pool's weighted-average foreclosure frequency (WAFF) and the weighted-average loss severity (WALS), we consider the potential increase of credit risk over time as a result of changes in pool composition. We assess possible deterioration in pool composition based on the transaction's documented asset-eligibility criteria, the history of the originator and, in particular, any observed changes in origination, underwriting, and related performance.

Geographic concentration  

361. The following table shows the region concentration thresholds. An adjustment of 1.25x is applied to the excess above the thresholds. If a pool has significant geographical concentration risk that we believe is not sufficiently captured, we may capture it using the originator adjustment.

Table 48

Nonresidential use loans and nonresidential borrowers  

362. Adjustments are applied to assets where we consider the primary use of the property to be nonresidential. Examples of nonresidential asset use include private individuals purchasing a commercial or mixed-use property and commercial borrowers. The adjustments detailed in the table below are intended for use where 40% or less of the pool at issuance are considered to be loans for nonresidential use.

Table 49

Reperforming loans  

363. We apply adjustments for reperforming loans when a portfolio contains a material portion of reperforming loans.

364. We may consider different adjustments on a case-by-case basis, depending on the servicers' restructure policies, track record, and performance data provided.

365. We typically define a reperforming loan as a loan that has been 90 or more days past due or restructured in the five years leading up to the analysis date and is current as of that date.

366. When a reperforming arrangement is made, a full reassessment of the borrower's affordability capacity is typically made. We consider this akin to a reunderwriting of the loan. Accordingly, for pools classified as reperforming, we calculate potential future seasoning credit based on the date on which a loan was last 90 or more days in arrears or was restructured.

367. We typically apply adjustments for reperforming loans as shown in the following table. We may consider different adjustments on a case-by-case basis, depending on the servicers' restructure policies, track record, and performance data provided.

Table 50

368. In addition to the original loan and borrower information provided, we may also consider updated data sourced through the restructuring process in our analysis of reperforming loans, where available, on a case-by-case basis.

369. In addition, as part of the analytical process, we analyze data from the issuer or servicer on re-default rates stratified by forbearance type. This analysis is used to calibrate the originator adjustment for those transactions.

Second-lien loan  

Table 51

Weighted average loss severity (WALS)

Table 52

370. The Greek Civil Procedures Code regulates how the property auction proceeds will be distributed, depending on the different combinations of creditors to which the borrower is liable. According to the different combinations, if the property backing one of the mortgages is foreclosed and the borrower is also liable to creditors with general privileges and/or unsecured claims, the recovery could be limited to up to two-thirds, or 65%, of the proceeds.

371. On a case-by-case basis, we increase the forced sales discount where there is evidence to support that it may be either higher or lower than envisaged using the standard calculation.

Cash Flow Assumptions

372. In our rating analysis, we also perform an analysis of a transaction's payment structure and cash flow mechanics. This analysis uses our own quantitative models to assess whether the cash flows from the assets suffice, at the applicable rating levels, to make timely payments of interest and ultimate payment of principal (i.e., by or before the legal maturity date).

373. In our cash flow analysis, we use as inputs the pool-level WAFF and WALS described in previous sections of this guidance to reflect credit stress at each rating level.

374. During modeling, cash flow stresses test the credit and liquidity support the assets need, considering any available structural support, such as a cash reserve, a liquidity facility, or hedging arrangements.

375. For revolving stand-alone RMBS structures (i.e., structures backed by a pool whose assets change or revolve), the modeling approach aims to reflect the structure after the activation of any "stop-substitution" (or early amortization) triggers and applies cash flow stresses from this point. A stop-substitution trigger is an event or situation that halts the substitution of assets in a revolving loan pool. We do not apply this approach to pools supporting covered bonds, even though the assets in these pools may be substituted over time. This is because the starting assumption of the collateral analysis under our covered bond criteria is the default of the issuing bank, as a result of which we do not expect the cover pool to be actively managed (i.e., it would become a static pool).

376. This guidance does not include the assumptions used to assess refinancing costs in covered bonds structured with an asset-liability mismatch (such as target asset spreads). Those are described in "Covered Bonds Criteria," published Dec. 9, 2014.

377. In our surveillance of existing ratings, cash flow modeling may show that, under the 'B' stress, a particular tranche will miss interest payments or fail to repay the principal by or before the final legal maturity date. If this is the case, then our initial assessment under this guidance, all factors remaining the same, may be to consider lowering the rating on those securities to 'B-' or lower.

378. Depending on our view of a transaction's immediate cash flow position, the rating could move into the 'CCC', 'CC', or 'C' category, consistent with our ratings definitions, and in accordance with "Criteria For Assigning 'CCC+', 'CCC', 'CCC-', And 'CC' Ratings," published Oct. 1, 2012.

Defaults 

379. Under the guidance, the cumulative amount of defaults for cash flow modeling is equal to the pool's WAFF, whereas the cumulative amount of recoveries is calculated as (1 – WALS).

380. We typically model two default-timing scenarios, referred to as "front-loaded" (concentrated toward the earlier stage of a transaction) and "back-loaded" (concentrated toward the later stage of a transaction). In both scenarios, the defaults occur over a six-year period, with each scenario having a peak level of stress, referred to as a recession, where 25% of the expected WAFF is applied annually for three years. For the front-loaded curve, the recession starts at day one; for the back-loaded curve, the recession starts at the start of the fourth year. For certain structure types (e.g., master trust transactions), we may run additional analysis to test the sensitivity of liability structures to different default-timing curves.

Table 53

381. Foreclosure period assumptions, which represent the estimated time to repossess and sell a property upon a default and reflect the typical time necessary for judicial proceedings and any other likely delay, are detailed in the table above.

382. Cash flow modeling considers the negative carry resulting from interest due on the rated liabilities during the foreclosure period.

383. The loss severity estimates used in the cash flow modeling are based on the loan principal and assume no recovery of interest accrued on the mortgage loans during the foreclosure period.

Delinquency 

384. We assume a delay of a proportion of scheduled interest and principal receipts equal to one-third of the WAFF in each of the first 18 months of a hypothetical recession, and set full recovery of the arrears to take place 36 months after the delinquency occurs. The cash flow stress for delinquencies is independent of the arrears adjustment to the WAFF.

385. For pools that contain residential loans with an option to temporarily suspend the periodic payments (payment holiday loans), the guidance includes a delay of a proportion of scheduled interest and principal receipts. The proportion is based on historical data on borrowers that have exercised such an option, assuming they use this option for the maximum duration permitted under the contractual terms of their loans. In situations where there is the potential for payment holidays to be granted after a loan's inception or where payment holidays have been granted (e.g., due to governments' and banks' forbearance measures for households and small and midsize corporates), we may apply an additional stress in our cash flow analysis where relevant. In those instances, the guidance includes a delay of a proportion of scheduled interest and principal receipts based on an estimate of the proportion of a pool that opts to take a payment holiday and the likely duration of the holiday. The likely duration will be assessed with reference to factors that may include, but are not limited to, relevant legislative frameworks, collateral credit quality, servicers' policies, and available servicer data on payment holidays granted.

Interest, Prepayment, And Reinvestment Rates

Interest rate risk  

386. The guidance typically applies two different interest rate scenarios (up and down) for modeling purposes (see "Credit Rating Model: CIR Model," published April 4, 2019). The curves vary by stress scenario. In case a transaction features an interest rate cap, either on the note coupon or a cap agreement with an external counterparty, upward interest rate stress assumptions exceeding the cap level may be unduly beneficial for the transaction's cash flow projection. In such event, we may apply a different interest rate stress to test sensitivity of the ratings to the absence of the cap.

387. Specific structural features may involve using additional cash flow stresses, such as alternative interest rate patterns or different default timing curves, among others. In particular, to analyze cash flows of covered bonds including bullet maturities, we may apply a stochastic approach to stress interest rate and currency risks using our Covered Bond Monitor cash flow model (see the Related Research section).

Basis risk  

388. Basis risk arises whenever unhedged differences exist between, on the one hand, methodologies for calculating interest on the assets, and on the other hand, the liabilities of a structure. It differs from interest rate risk. For example, it may arise because of differences between two floating-rate indices, or when swaps are present and there is a timing mismatch between the reset dates on the securities and on the swap.

389. To address this risk, the guidance adjusts the interest spread modelled between assets and liabilities by applying different spreads over the life of a transaction as a haircut to the margin. The size of the spreads depends on the distribution of historical differences among indices, using the rating-specific values corresponding to the percentiles shown in table 54 below.

390. Our current adjustments for typical indices are presented here: https://www.standardandpoors.com/pt_LA/web/guest/article/-/view/sourceId/100043040

391. The guidance applies the stress corresponding to a rating level in the cash flow analysis for the first 18 months of a hypothetical recession. After this 18-month period, the 'B' percentile from the table below applies at all rating levels.

Table 54

392. Adjustments for indices not detailed in the link above would be devised on a case-by-case basis using a similar approach and applied in the same way.

393. For example, in a Greek RMBS or covered bond transaction, the underlying mortgage loans may incur interest based on the one-month EURIBOR resetting monthly, but the securities may pay interest based on the three-month EURIBOR. In this case, the analysis looks at the distribution of the historical differences between these two rates, calculated by taking the highest three-month EURIBOR over the previous three-month period and the lowest one-month EURIBOR for each point in the data set. It then takes the percentiles of the resulting distribution shown above.

394. In another example, for a RMBS transaction whose mortgage loan pool contains loans that pose no basis risk during their promotional period but will revert to a floating margin that carries basis risk, the level of stress modelled depends on the proportion of the loans in the pool that would eventually have basis-risk exposure.

Prepayment scenarios  

395. Residential loan prepayments vary the amount of excess spread available and may affect the absolute level of defaults exhibited in a transaction. Therefore, we define a framework for assessing prepayment risk in the following paragraphs. When analyzing the payment structure and cash flow mechanics of Greek covered bonds and RMBS transactions, we typically test the transaction's ability to withstand high and low prepayment scenarios as set out in the following table.

Table 55

396. We may adjust the prepayment assumptions if a pool's historical prepayment rates were higher than historical averages or if a transaction were particularly sensitive to prepayment risk (e.g., excess spread notes). We may also reduce high prepayment stress in situations where long-term historical data support lower prepayment rate assumptions for a specific loan product.

Reinvestment rates  

397. Reinvestment rate assumptions stress the yield from excess cash that becomes available because of, for example, prepayments, as well as the revenues associated with any other cash the issuer holds. The guidance applies different reinvestment rate assumptions to covered bonds with a bullet maturity structure on the one hand, and to RMBS transactions and covered bonds with a pass-through maturity structure on the other (see table below). This differentiation reflects that, for a structure with bullet maturities, we generally expect a longer tenor of reinvestment than for a pass-through structure. The guidance applies a floor to the reinvestment rate of 0%.

Table 56

Originator Insolvency, Commingling, And Set-Off

398. Our analysis of commingling risk that can result from an originator's or servicer's insolvency follows one of three approaches: (i) application of our current counterparty criteria, (ii) modeling to produce estimates of any exposure, or (iii) a combination of the two approaches. The combination of the first and second approaches applies if the application of the counterparty criteria shows residual cash flow exposures. The rest of this section focuses on the modeling of risk exposures that are not fully mitigated under our counterparty criteria.

399. Our legal analysis may conclude that the insolvency of the collection account holder could result in a loss of funds deposited in the account. This analysis considers the degree to which a collection account holder's insolvency affects the cash flow from the residential loan pool, if the collection account is not in the name of the issuer.

400. The amount at risk depends on the timing of scheduled payments from borrowers and the frequency of transfers into the transaction account.

401. If the loan originator (or servicer) is also a deposit-taking institution, modeling takes into consideration the possibility that borrowers with deposit accounts may set off these amounts against their outstanding mortgage loans. In arriving at estimates of potential set-off exposures, we will consider historical data on deposit levels, the borrower profile, and an assessment of factors that may lead to a material increase in set-off exposure. We will also look at any mitigants, such as deposit guarantee schemes.

402. The modeling approach aims to capture set-off risks from any borrowers that are also employees of the originator. This is because amounts owed to them, such as unpaid salaries, pension benefits, and subsidies, could offset payments due on their mortgage loans. As a result, the modeling approach treats the full amount of any employee's mortgage loan balance as entirely set off, and all set-off amounts as principal losses.

Modeling Of Senior Fees And Expenses, Liquidity Facilities, And Spread Compression

403. The modeling of all the issuer's foreseeable expenses uses stressed costs to reflect the need to replace the initial service provider.

404. The most significant of these is the cost of servicing mortgage loan receivables.

405. We apply a fee equal to the higher of 2x the contractual fee and 50 bps.

406. In situations where there is a contracted back-up servicer in place, we may make adjust the assumptions in the paragraph above. In addition, in situations where fees are linked directly to specific activities, such as issuing letters, or where increases are linked to inflation, we may apply different assumptions from those detailed in the paragraph above.

407. We include estimates of marginal costs from liquidity facilities in cash flow modeling. Most liquidity facilities are renewable after 364 days, subject to a commitment fee and a drawn fee.

408. We assume that the liquidity facility is fully drawn on day one of the analysis.

409. The drawn fee is modeled as being payable from this point on the whole facility balance, unless the documents state that the issuer does not have to pay a fee if the drawings on the facility did not originate from the issuer. Such a situation could arise because of nonrenewal of the facility or a downgrade of the facility provider. The commitment fee is modeled as per the documented terms.

410. We also model the possibility that the spread on the loan pool compresses over time, due to defaults, prepayments, and product switches. To reflect this, we reduce margins with the assumption that a percentage of the higher-yielding loans exit the portfolio. In addition, for RMBS, the transaction may feature contractual minimum yield levels (e.g., to allow substitutions to continue in revolving pools). In such a case, the modeling approach aims to capture any breach of the yield levels that triggers the end of the revolving period and produces lower yields.

411. This appendix provides additional information and guidance to these proposed criteria, and we expect to publish this information and guidance in a separate guidance document following the publication of the finalized criteria article. It is intended to be read in conjunction with the proposed criteria herein and aforementioned global cash flow criteria. Guidance documents are not criteria, as they do not establish a methodological framework for determining credit ratings. Guidance documents provide guidance on various matters, including articulating how we may apply specific aspects of criteria, describing variables or considerations related to criteria that may change over time, providing additional information on nonfundamental factors that our analysts may consider in the application of criteria, and providing additional guidance on the exercise of analytical judgment under our criteria. Our analysts consider guidance documents as they apply criteria and exercise analytical judgment in the analysis and determination of credit ratings. However, in applying criteria and the exercise of analytic judgment to a specific issuer or issue, analysts may determine that it is suitable to follow an approach that differs from one described in the guidance document. Where appropriate, the rating rationale will highlight that a different approach was taken. For more information about guidance documents, please see "Criteria And Guidance: Understanding The Difference," Dec. 15, 2017.

Appendix III: Criteria Articles To Be Partially Superseded

412. If we adopt the proposed changes as outlined in this article, "Methodology And Assumptions: Assessing Pools of European Residential Loans," published Aug. 4, 2017, will be partially superseded, in relation to Spain, Portugal, and Italy.

RELATED CRITERIA AND RESEARCH

Related Criteria

• Counterparty Risk Framework: Methodology And Assumptions, March 8, 2019
• Incorporating Sovereign Risk In Rating Structured Finance Securities: Methodology And Assumptions, Jan. 30, 2019
• Methodology For National And Regional Scale Credit Ratings, June 25, 2018
• Methodology And Assumptions For Rating U.S. RMBS Issued 2009 And Later, Feb. 22, 2018
• U.S. Residential Mortgage Operational Assessment Ranking Criteria, Feb. 22, 2018
• Sovereign Rating Methodology, Dec. 18, 2017
• Methodology And Assumptions: Assessing Pools Of European Residential Loans, Aug. 4, 2017
• Structured Finance: Asset Isolation And Special-Purpose Entity Methodology, March 29, 2017
• Covered Bond Ratings Framework: Methodology And Assumptions, June 30, 2015
• Covered Bonds Criteria, Dec. 9, 2014
• Methodology For Assessing Mortgage Insurance And Similar Guarantees And Supports In Structured And Public Sector Finance And Covered Bonds, Dec. 7, 2014
• Global Framework For Assessing Operational Risk In Structured Finance Transactions, Oct. 9, 2014
• Global Framework For Cash Flow Analysis Of Structured Finance Securities, Oct. 9, 2014
• Methodology And Assumptions For Rating Japanese RMBS, Dec. 19, 2014
• Criteria For Assigning 'CCC+', 'CCC', 'CCC-', And 'CC' Ratings, Oct. 1, 2012
• Criteria Methodology Applied To Fees, Expenses, And Indemnifications, July 12, 2012
• Methodology For Applying RMBS Small Pool Adjustment Factor, May 24, 2012
• Banking Industry Country Risk Assessment Methodology And Assumptions, Nov. 9, 2011
• New Zealand RMBS Rating Methodology And Assumptions, Sept. 14, 2011
• Australian RMBS Rating Methodology And Assumptions, Sept. 1, 2011
• Principles Of Credit Ratings, Feb. 16, 2011

Related Research

• S&P Global Ratings Definitions, Aug. 7, 2020
• RFC Process Summary: Global Methodology And Assumptions: Assessing Pools Of Residential Loans (Sweden, Norway, Finland, And Denmark), July 10, 2020
• Request For Comment: Methodology To Derive Stressed Interest Rates In Structured Finance, April 16, 2019
• Guidance: Global Methodology And Assumptions: Assessing Pools Of Residential Loans, Jan. 25, 2019
• RFC Process Summary: Global Methodology And Assumptions: Assessing Pools Of Residential Loans, Jan. 25, 2019
• Covered Bond Monitor: Technical Note, Sept. 6, 2019
• Credit Rating Model: CIR Model, April 4, 2019
• Credit FAQ: How We Analyze Residential Mortgage Loans Backing Greek Covered Bonds And RMBS, Aug. 29, 2018
• Criteria And Guidance: Understanding The Difference, Dec. 15, 2017

The proposed criteria represent the specific application of fundamental principles that define credit risk and ratings opinions. Once proposed criteria become final, their use is determined by issuer- or issue-specific attributes as well as our assessment of the credit and, if applicable, structural risks for a given issuer or issue rating. Methodology and assumptions may change from time to time as a result of market and economic conditions, issuer- or issue-specific factors, or new empirical evidence that would affect our credit judgment.