UK Residential Mortgage Loan Analysis Criteria: Credit

Transcription

toronto
new york
chicago
london
paris
frankfurt
Methodology
U.K. Residential Mortgage Loan
Analysis Criteria: Credit, Prepayment
and Loan Aggregation
november 2007
contact information
Elisa Malavasi
Vice President
Structured Finance Quantitative Group
Tel. +44 (0)20 7562 5609
[email protected]
Victoria Johnstone
Senior Vice President
Tel. +44 (0)20 7562 5608
[email protected]
Kai Gilkes
Managing Director
Tel. +44 (0)20 7562 5606
[email protected]
Apea Koranteng
Managing Director
Structured Finance EMEA
Tel. +44 (0)20 7562 5603
[email protected]
DBRS is a full-service credit rating agency
established in 1976. Privately owned and operated
without affiliation to any financial institution,
DBRS is respected for its independent, third-party
evaluations of corporate and government issues,
spanning North America, Europe and Asia.
DBRS’s extensive coverage of securitizations
and structured finance transactions solidifies our
standing as a leading provider of comprehensive,
in-depth credit analysis.
All DBRS ratings and research are available in
hard-copy format and electronically on Bloomberg
and at DBRS.com, our lead delivery tool for
organized, Web-based, up-to-the-minute information. We remain committed to continuously
refining our expertise in the analysis of credit
quality and are dedicated to maintaining
objective and credible opinions within the global
financial marketplace.
U.K. Residential Mortgage Loan Analysis Criteria: Credit, Prepayment and Loan Aggregation
November 2007
U.K. Residential Mortgage Loan Analysis Criteria:
Credit, Prepayment and Loan Aggregation
TABLE OF CONTENTS
Introduction
Credit Analysis
Prepayment Analysis
Loan Aggregation Analysis
Advantages of Approach
5
5
5
5
6
Part 1: U.K. Residential Mortgage Default Criteria
The Benchmark Two-Year PD Estimate
The Base Case Two-Year PD Estimate
Credit Risk Band
CCJs and/or Bankruptcies or IVAs
Loan-to-Value (LTV)
Employment and Other Income-Related Variables
Right-to-Buy (RTB)
Loan Purpose
Repayment Type
Loan Term
Loan Size
Second Lien
Loan Product
Buy-to-Let (BTL)
Credit Risk Layering
The Base Case Lifetime PD Estimate
Portfolio Default Rate Distribution
Rating-Specific Portfolio Default Rates
8
9
9
11
11
12
13
15
15
16
17
17
17
17
18
22
22
24
27
Part 2: U.K. Residential Mortgage Loss Criteria
LGD Overview
Components of LGD
Principal Amount Owed (Exposure at Default, or EAD)
Current Property Value
Sale Price of the Foreclosed Property
Costs
Prior Ranking Loans
LGD per Rating Level
Foreclosure MVDs per Rating Level
Loan-Level and Portfolio-Level LGD Calculations
28
28
28
28
28
31
35
36
36
36
36
Part 3: U.K. Residential Mortgage Prepayment Criteria
38
Part 4: U.K. Residential Mortgage Loan Aggregation Criteria
42
Concluding Remarks
43
Appendix A: The DBRS U.K. Residential Mortgage Loan Analysis Model
44
Appendix B: The DBRS U.K. Mortgage Loan Analysis Model Example Tear Sheet
45
Appendix C: Loan-level Example Computations of Lifetime PD, LGD and EL
48
Appendix D: The DBRS U.K. Mortgage Loan Analysis Model Example Rep Lines
50
3
November 2007
Introduction
This report describes the DBRS loan analysis methodology for U.K. residential mortgage portfolios and
forms part of the DBRS criteria for rating U.K. residential mortgage-backed securities (RMBS) and other
transactions linked to residential mortgage assets denominated in the United Kingdom. This methodology describes the approach to three main areas that DBRS considers to require a loan-level analysis:
credit, prepayment and loan aggregation.
CREDIT ANALYSIS
The DBRS mortgage credit quality assessment addresses the principal amount of a portfolio that could
be potentially lost in different rating scenarios. As such, it forms an integral part of the analysis used
to assess the credit enhancement required to maintain a rating on a RMBS tranche. Principal loss on
a mortgage loan occurs when the borrower fails to pay back the full balance of the outstanding loan
amount. There are two key components to this analysis. First, DBRS estimates a loan’s probability of
default (PD); that is, how likely it is that the borrower will stop paying the mortgage loan and the lender
will be forced to foreclose and sell the property. Second, the analysis assesses how much the lender will
lose if the property is subject to foreclosure and a forced sale (the loss given default, or LGD), as the
lender may not be able to re-coup the full outstanding loan balance from the sale proceeds. The product
of the PD and LGD for a loan gives the expected principal loss (EL) for that loan. As such, the DBRS
credit analysis establishes PD and LGD estimates (and therefore EL) on a loan-by-loan basis and for the
portfolio as a whole over different rating scenarios.
DBRS assumes that a B rating corresponds to a benign economic environment and is representative of
the status of the U.K. housing market economy over the last two to five years. A detailed analysis of
the behaviour of U.K mortgage loans over this time therefore forms the basis of the DBRS base case,
or B rating scenario. The impact of deterioration in the economy on portfolio PD and LGD levels is
then assessed to obtain these estimates at higher rating levels, which are representative of more stressful
economic scenarios.
For non-synthetic mortgage-backed transactions, the portfolio PD and LGD then become inputs for the
subsequent cash flow analysis. Cash flow simulations overlay the assumed portfolio PDs and LGDs (along
with other economic stresses such as interest rates and prepayments) within the transaction structure to
assess the ability of the rated notes to withstand economic stress and, as such, maintain a rating.
PREPAYMENT ANALYSIS
The DBRS mortgage prepayment analysis assesses the loan characteristics known to influence prepayment rates. The outputs of a loan-level assessment are aggregated to create a base case prepayment
assumption for the portfolio. The prepayment assessment addresses the amount of unscheduled principal received over time, and as such, is an important consideration within a cash flow analysis, both from
note life, and tranche rating strength perspectives. The assumed prepayment rate has a large influence
on the amount of excess spread (typically used to cover principal losses in most U.K. transactions), and
therefore is a key factor in the assessment of the ability of a tranche to withstand a rating stress. Given
its importance, DBRS considers a loan-level prepayment model as crucial in understanding the dynamics
of the transaction through time, and accurately assign a rating.
LOAN AGGREGATION ANALYSIS
DBRS produces standardised loan aggregation “rep lines” for each transaction, with aggregation based
on characteristics that influence the repayment profile of the loans (e.g., interest-only periods, interest
rate types, remaining term). As such, the rep lines act as a transparent overview of the asset cash flow
profile, and also allow for efficient and accurate cash flow modelling.
5
November 2007
ADVANTAGES OF APPROACH
The DBRS U.K. residential mortgage loan analysis methodology provides an enhanced method of assessing credit and prepayment risk, along with the production of standardised rep lines. The criteria used to
calculate these components are described in full in the following methodology. There are a number of
key benefits to the approach, highlighted below.
• The DBRS U.K. Residential Mortgage Loan Analysis Model is a Microsoft Excel-based tool used by
DBRS to analyse the credit quality of U.K. loan portfolios using the methodology described in this
report. In addition, the model also outputs a DBRS base case prepayment vector, and rep lines that are
subsequently used in any cash flow modelling. The model is available in open format and free of charge
as part of DBRS’s aim to increase transparency in the rating process and allow market participants to
fully understand how risk is assessed. For information on how to gain access to the model, please see
Appendix A. An illustration of some key outputs of the model is given in Appendix B.
• The DBRS credit and prepayment assumptions stem from a comprehensive review and assessment
of several data sources that covered the specifics of the U.K. residential mortgage market. Outputs
from loan-level analyses, where available, have been integrated with results from portfolio-level breakdowns of existing transactions. Particular attention has been given to observed historical performance
in order to estimate base case performance and provide a benchmark to estimate mortgage behaviour
under more stressful economic conditions.
• The approach allows for predictive base case PDs and LGDs to be estimated at the loan and portfolio
levels over both a two-year period and lifetime horizon. This base case performance is assumed to
represent a specific rating scenario; namely, B. DBRS considers that accurate base case performance
estimates are extremely important for a number of reasons.
(1) Within the Basel II framework, more lenders may be motivated to move lower rated and equity
tranches off their balance sheets. As such, the ability of investors to access a detailed risk assessment
of the EL under base case conditions over a range of time periods is crucial.
(2) DBRS believes that deal surveillance can be significantly enhanced by being able to measure and
track deal performance over time with respect to an initial base case expectation identified to represent a specific rating level. This means that DBRS can easily match actual deal performance with
rating expectations and provide accurate and timely rating implications.
• DBRS strongly believes that the behaviour of foreclosed property values is dramatically different
from the average values obtained from a generic house price index. There is strong evidence both on
the portfolio and loan levels that despite dramatic house price increases in recent times, foreclosed
properties suffer significant market value declines, which may have been underestimated in alternative
approaches.
• The methodology has been created so that it can easily incorporate external credit bureau scores
should they become accessible to third parties. DBRS considers that the inclusion of such scores within
a default model can add considerable predictive power and hence allow for better differentiation of
risk across mortgage portfolios. Therefore, a methodology that can easily integrate these scores is
crucial to making full and timely use of their added value.
• There are meaningful default adjustments to loans that are seasoned or in arrears that can facilitate
more accurate portfolio default estimates through time. This ensures that the methodology can be
reliably used not only at the pre-bond issuance stage (or for the initial assessment of portfolio purchases), but also integrated into the surveillance process.
6
November 2007
• Portfolio modelling techniques are used to create a distribution of potential portfolio default rates.
There are two main benefits to such an approach:
(1) Rating-specific default rates can be directly determined from the portfolio default distribution by
holding the distribution to a “rating standard” implied by the DBRS Corporate Default Table. This
is quite different from an methodology that multiplies base case estimates by rating multiples. The
portfolio approach also allows for a consistent quantitative approach across different asset classes. For
example, collateralised debt obligation (CDO) asset portfolios are also held to the same standard.
(2) A distribution of portfolio default rates will allow users to test transactions over a wide range of
PDs (and matched LGDs) in a stochastic framework, which may complement the usual deterministic
approach (where tranches are tested with discrete rating-specific default scenarios). Although DBRS
intends to initially use a deterministic approach to test U.K. RMBS tranches, a stochastic approach in
creating more insights into RMBS tranche performance is a key area of ongoing research.
• The DBRS methodology maximizes the advantages of obtaining loan-level information, and extends
its use to enhance other assumptions within the rating approach.
(1) Prepayments – given the large impact prepayment assumptions have on note life and excess spread,
accurate and refined estimates of prepayments are crucial for transaction transparency, efficiency and
robustness. DBRS is the first rating agency in Europe to create (and make available on a transparent
basis) a loan-level approach to prepayment modelling.
(2) Loan Aggregation – the creation of standardised rep lines allows rating analysts to quickly and
easily identify key cash flow risks within the asset pool (such as unhedged positions and discount
reserve sizings). This rep line creation also allows efficient cash flow modelling, without any compromises in performance.
In the present report, the computation of the two credit estimates (PD and LGD) are addressed separately in the first two sections. The third section covers the DBRS approach to the loan-level prepayment
estimations, and the fourth section describes the loan aggregation methodology.
7
November 2007
Part 1: U.K. Residential Mortgage Default Criteria
This section describes the DBRS methodology used to calculate loan-level PDs and portfolio-level default
rates for U.K. residential mortgage portfolios. One important component of this methodology is an
approach to calculate base case two-year and lifetime PD estimates for individual mortgage loans. The
approach also includes the creation of a portfolio default distribution, which allows for the extension to
rating-specific portfolio default rates. A summary of the methodology used to calculate loan-level PDs
and portfolio default rates is given below and described in detail in the following pages.
The same benchmark two-year PD estimate
is assigned to each loan.
A base case two-year PD estimate
is calculated for each loan
by adjusting the benchmark two-year
PD to account for individual risk characteristics
associated with each loan.
A base case lifetime PD estimate
is calculated for each loan
by extending the base case two-year
PD to account for individual loan seasoning.
A portfolio default rate distribution
is calculated by means of a “single factor”
model that requires the weighted average of
the base case lifetime PDs and an asset correlation estimate.
Rating-specific portfolio default rates
are calculated by holding the portfolio default
distribution to a “rating standard,” as implied by
the DBRS Corporate Default Table.
8
November 2007
THE BENCHMARK TWO-YEAR PD ESTIMATE
Each loan in the mortgage portfolio is initially assigned the same benchmark two-year PD. Benchmark
performance has been calibrated from U.K. loan-level and portfolio-level mortgage credit behaviour
over the past four to five years, from approximately 2002 to 2006. This period represents a benign
economic environment for mortgage performance (e.g., low interest rates, low levels of unemployment,
strong house price growth), and consequently this period has been associated with low arrears and
mortgage default rates (see Figure 1.1). The benchmark PD assigned to all loans is 0.125%.1 This is
aligned with the two-year estimate from repossession data shown in Figure 1.1 and will be adjusted
upward or downward on a loan-by-loan basis depending on the individual loan characteristics (see the
following section).
Figure 1.1: Mortgage Arrears and Repossession Rates in the United Kingdom
2.00%
1.75%
1.50%
1.25%
1.00%
0.75%
0.50%
0.25%
0.00%
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
% of mortgage loans outstanding
2.25%
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Time
Source: CML statistics.
Repossessions
6-12m arrears
12m+ arrears
The benchmark loan PD was calculated over a two-year time period for a number of reasons. Firstly, this
time period allowed for an increased number of analysis points over a range of loan seasonings, across
the four- to five-year horizon chosen to represent benchmark performance. Secondly, DBRS intends to
include credit bureau scores (in the event they become available for third-party use2) to help assess the
credit risk of mortgage loans. Generic credit scores created by the bureaus active in the United Kingdom
(e.g., Experian, Equifax and Call Credit) are highly predictive measures of default and are extensively
used by a range of mortgage and other lenders. The scores (e.g., Delphi, Risk Navigator and CallScore)
summarize borrower credit history and condense many relevant performance factors (e.g., borrower
behaviour on all credit lines, level of indebtedness, geographic default data) into a single numeric value.
Since credit bureau scores are commonly calibrated to reflect performance over a one- or two-year
period, DBRS has developed a methodology that can easily incorporate the scores if they are made
available. DBRS is a strong supporter of use of any measures that aid in predicting adverse trends at a
granular level, especially where there is a prospect of a deteriorating credit cycle.
THE BASE CASE TWO-YEAR PD ESTIMATE
The benchmark two-year PD of 0.125% is then adjusted on a loan-by-loan basis to create the base
case two-year PD estimate per loan. These adjustments are to account for borrower, property and loan
product factors that increase or decrease the credit risk associated with a particular loan. An overview
of the risk-adjustment factors used is provided in this section.
1. For other jurisdictions, DBRS will look to local market conditions to derive benchmark loan performance. In addition,
adjustments to the benchmark to create loan-level PDs will also be jurisdiction-dependent.
2. At the time of publication, the use of bureau information by third parties is prevented by the Principles of Reciprocity
developed by the Steering Committee on Reciprocity (SCOR), the governing body of the bureau databases. Changes
to the Principles of Reciprocity to allow for third-party use of the data and scores under certain circumstances is, however, currently under consideration by SCOR.
9
November 2007
A more qualitative evaluation of underwriting standards, credit policies and servicing practices is then overlaid
on the estimated loan-level PDs to adjust for servicer- or originator-specific influences on loan credit performance. Further adjustments may also be necessary in the case of significant concentration risks.
The following section focuses on the loan, borrower and property characteristics that DBRS considers to
be influential on a borrower’s propensity to default. The default behaviour of each mortgage loan in the
pool is forecast by integrating past credit performance information with additional characteristics that
may influence a borrower’s likelihood of default. Each characteristic is associated with a multiplicative
factor that adjusts the 0.125% benchmark PD up or down. Table 1.1 summarizes each characteristic
and gives the associated multiple used to adjust the benchmark PD. The influence of each characteristic
on default is then discussed on the following pages. A worked example of this analysis for a single loan
is given in Appendix C.
Table 1.1: Loan-level Risk Adjustments
10
Risk Characteristic
Characteristic Value
Base Multiple
Credit Risk Band
A
B
C
D
E
n/a
1.00
2.00
4.00
8.00
Adverse Credit History
CCJs <= GBP 100
GBP 100 < CCJs <= GBP 2,000
GBP 2,000 < CCJs <= GBP 5,000
CCJs > GBP 5,000
Prior Bankruptcy/IVA
1.00
2.45
2.80
3.55
3.55
Loan-to-Value (LTV)
<=40
50
60
70
80
90
95
100
0.60
0.80
1.00
1.30
1.65
2.10
2.35
2.65
Employment/Income
Self-Certified (Employed)
Self-Certified (Self-Employed)
Fast Track
Loan-To-Income (LTI) >3.5
Single Income
Self Employed
1.75
1.35
1.10
1.25
1.25
1.15
Buy-to-Let (Stabilised ICR)
<=100
105
110
115
120
125
130
2.00
1.80
1.65
1.45
1.30
1.15
1.00
Right to Buy
Yes
1.10
Purpose
Debt/Equity Re-mortgage
1.25
Repayment Type
IO
1.35
Term
Repayment Loan >25 yrs
1.20
Loan Size
Jumbo (Region Specific)
1.10
2nd Lien
2nd Ranking Loan
1.50
Loan Product
Tracker (For Life with Teaser)
Tracker (Short Term)
Discount (Short Term)
Fixed (Short Term)
1.05
1.05
1.05
1.10
Risk Layers
LTV >= 95% & Self Cert/High LTI
LTV >= 90% & Past CCJs/Bankrupt
LTV >= 90% & Past CCJs/Bankrupt & Self Cert/High LTI
1.35
1.75
1.85
November 2007
Credit Risk Band
In order to differentiate between the credit quality of borrowers, each loan is assigned to a credit risk
band, based on past credit performance information. Credit risk bands range from “A” to “E,” with
“A” borrowers considered to be the least risky and “E” borrowers having severe current or past credit
problems. As discussed, the best summary of past credit performance information is contained within
external credit bureau scores such as those provided by Experian, Equifax and Call Credit, which use a
variety of information sources to identify credit risk (e.g., payment behaviour across all borrower credit
lines, credit-to-debt measures, length of history, etc.). It is DBRS’s preference to determine credit risk
bands from external bureau scores. In the absence of such scores being available, DBRS uses the existing
credit information provided to assign credit risk bands. This information is a much reduced subset of
that contained within a bureau score, and as such DBRS will not assign an “A” grade to any loan on
the basis of this information. The available information currently used by DBRS to determine the credit
risk band is the following:
• Combined amount (and age) of past County Court Judgments (CCJs).
• Any prior bankruptcy or individual voluntary arrangement (IVA).
• Current arrears on mortgage.
Credit risk bands are assigned according to the matrix shown in Table 1.2. In addition, the credit risk band
is adjusted downward by a category if the borrower has at least one CCJ that is less than one year old.
Table 1.2: Cedit Risk Band Assignment
Amount of CCJs
Months in Arrears
0
1-3
4-5
6+
CCJs <= GBP 100
B
C
D
E
GBP 100 < CCJs <= GBP 2,000
C
D
E
E
GBP 2,000 < CCJs <= GBP 5,000
D
E
E
E
CCJs > GBP 5,000 and/or Prior Bankruptcy/IVA
E
E
E
E
CCJs and/or Bankruptcies or IVAs
Adverse credit history is a key differentiator for default risk because individuals who have suffered
debt problems in the past have a higher propensity for arrears and defaults on future debt repayment.
Significant previous financial difficulties are indicated by arrears or defaults on loans, CCJs or insolvency. The range of mortgage products targeting this specific segment of the residential mortgage market
is generally referred to as sub-prime. Growth in the number of people with credit problems, as well as
an increased demand for debt consolidation products, has boosted the U.K. sub-prime mortgage market.
Lending to borrowers with adverse credit history implies higher default risk compared with mainstream
lending and is evidenced by significantly higher mortgage arrears and default performance of the subprime compared with the prime mortgage market. A recent study by the Council of Mortgage Lenders
(CML)3 concludes that the more severe the level of credit adversity, the worse the performance.
In the United Kingdom, CCJs are legal decisions deliberated by the County Courts with regards to
monetary sums. A CCJ is recorded on a borrower’s credit report every time a loan repayment is not
made within specified times and/or conditions, and the creditor of the balance due brings the case to
court. Unless the CCJ is paid off within 30 days of being registered, it is likely it will remain on the credit
file for six years, thus significantly affecting a borrower’s credit status. More severe instances of CCJs are
associated with repeated patterns of credit difficulties by a borrower, hence the DBRS multiple relating
to the CCJ characteristic increases as the amount of CCJs rises.
3. Council of Mortgage Lenders (CML). “Adverse Credit Mortgages,” November 2006.
11
November 2007
Individuals in severe financial difficulties may declare an insolvency status in order to settle and clear
outstanding debts. In the United Kingdom, this is typically achieved by either bankruptcy or IVA. A
bankruptcy is ordered by a court and results in a public legal declaration stating the inability of an individual to repay his or her creditors. Any debt that is unable to be covered by available (or, sometimes,
future) assets is then discharged. A bankruptcy order can restrict borrowers on their future employment
possibilities (e.g., prevent an individual from holding public office).
As an alternative to bankruptcy, an IVA is a privately arranged plan for the repayment of debts to creditors and as such is not presided over by a court. An IVA is generally associated with a lower cost than
a bankruptcy, is confidential and has no restrictions in terms of future employment. As such, consumer
debtors are increasingly turning to IVAs in order to avoid bankruptcy (see Figure 1.2).
Figure 1.2: Number of Individual Insolvencies in the United Kingdom
20,000
18,000
16,000
Number
14,000
12,000
10,000
8,000
6,000
4,000
1998
1999
2000
2001
2002
2003
2004
2005
2006
Q1
Q2 P
Q4
Q3
Q2
Q1
Q4
Q3
Q2
Q1
Q4
Q3
Q2
Q1
Q4
Q3
Q2
Q4
Q1
Q3
Q2
Q1
Q4
Q3
Q2
Q4
Q1
Q3
Q2
Q1
Q4
Q3
Q2
Q4
Q1
Q3
Q2
0
Q1
2,000
2007
Time
P = Provisional. Source: The Insolvency Service statistics.
Bankruptcies
IVAs
DBRS does not discriminate between bankruptcies and IVAs in terms of risk adjustments because they
are both viewed as an indicator of severe financial distress.
Loan-to-Value (LTV)
LTV is the ratio between the principal balance on the mortgage and the appraised value of the property
serving as security for the loan itself. The input used by DBRS in the default model is the current LTV at
the time of securitization. This is calculated by summing all of the outstanding balances from every loan
secured by the same property (e.g., first-lien and second-ranking mortgages) and dividing the total by an
estimate of the current market value of the property.
The current market value is calculated by taking the given valuation and adjusting, if necessary.
Adjustments are made on the basis of the property valuation method and the last valuation. Details of
these adjustments are given in the LGD part of the methodology.
In the case of flexible loans, the maximum drawable amount is taken into account instead of the outstanding loan balance.
Higher LTV ratios are associated with increases in the likelihood of default, attributable to the progressively smaller portion of equity that the borrower has in the property. Equity is the difference between
the value of the property and the amount of all loans secured against it. The smaller the equity, the
smaller the potential financial benefit the borrower can retain from the property, and the lower the incentive to maintain loan repayments.
12
November 2007
DBRS has approximated the relationship between LTV and the likelihood of default by the function
shown in Figure 1.3. Here, the multiple applied to the benchmark loan PD rate increases as LTV
increases. Note that for LTVs below 60%, this multiple is less than 1.0, indicating a decrease in overall
PD compared with the benchmark. The highest multiple is 3.0 once the LTV reaches 105%. Also note
that static values reported in Table 1.1 on page 9 show examples of this function at various LTV levels.
Figure 1.3: Multiple to Benchmark PD by Current LTV
3.5
3.0
Base Multiple
2.5
2.0
1.5
1.0
0.5
115%
110%
105%
100%
95%
90%
85%
80%
75%
70%
65%
60%
55%
50%
45%
40%
35%
30%
0.0
Current LTV
Employment and Other Income-Related Variables
Borrower propensity to default is clearly related to the ability to make timely mortgage repayments on
an ongoing basis. DBRS regards affordability and other income- and employment-related aspects as very
important, particularly at a time when numerous affordability product innovations have taken place
in the U.K. market. In addition to this, U.K. consumer indebtedness levels have grown substantially in
recent years, raising serious concerns as to the ability of people to repay their debts should the current
favourable economic and housing conditions deteriorate.
The following sections outline employment and other income-related features that DBRS considers to
affect performance behaviour. These features are for owner-occupied properties only and do not apply
to buy-to-let (BTL) products, which are discussed separately in a following section.
Self-Certification
Self-certification is used by borrowers who want to obtain a mortgage without having to demonstrate
their earnings to a standard required by conventional mortgage underwriting criteria. Here, applicants
simply declare their own income, without having to provide the lender with any underlying documentation (e.g., pay slips, audited accounts). Typically, borrowers who seek to self-certify are self-employed,
commission-based or contract workers. Self-employed borrowers may choose to self-certify for a
number of reasons. Firstly, most lenders require self-employed workers to provide two to three years
of audited financial accounts, meaning that more recent self-employed borrowers would be unable to
satisfy this request. Secondly, audited accounts and/or current tax returns are often time lagged and may
not show the latest figures of a borrower’s income. Thirdly, self-employed borrowers may also perceive
that supplying the necessary documentation would be too onerous and time-consuming. Commissionbased workers may also choose to self-certify, as they receive a salary with a high proportion of bonus
payments and hence show a large degree of variability in income over time. Contract workers and those
with incomes from a variety of sources choose to self-certify because their total earnings may not otherwise be considered under a traditional mortgage.
13
November 2007
There are additional risks with self-certified mortgages, and the concept that a self-certified borrower
could not afford the loan under normal lending criteria is a common one. Recent Financial Services
Authority (FSA) research4 has highlighted that lenders generally have a higher level of material arrears
for their self-certified business compared to mainstream lending. Lenders endeavor to offset this risk by
a variety of ways, the most common being:
– Most self-certified mortgage providers pass the applicant’s stated income through a plausibility check
to ensure their stated job type fits within a reasonable salary range.
– More conservative credit scoring cuts are taken into account when assessing self-certified mortgage
applications, as well as lower LTV ratios so as to deter borrowers from taking out a mortgage that
they cannot afford.
– Fraud detection systems across U.K. lenders also discourage systematic fraud in the self-certified market
(e.g., CIFAS – the U.K.’s Fraud Prevention Service, which pools information on financial crime).
Despite these additional safeguards, the higher level of arrears experienced with this product type means
that DBRS considers self-certified loans to be more risky than benchmark loans. Note that DBRS considers self-certification products to employed borrowers more risky than those to self-employed borrowers,
given that a self-employed borrower may have a more “legitimate” reason for self-certification (such as
the burden of supplying audited financial accounts).
Fast Track
Some lenders are currently providing an accelerated approval process for low-risk borrowers that speeds
up the time they take to finalise a mortgage offer. This process, known as Fast Track, enables lenders to
have the right to seek full income verification for a loan application, even if they eventually choose not
to do so in practice.
Lenders generally offset the risk involved in offering Fast Track loans by making these products available only to LTVs lower than a certain limit, typically 85%. Given that the lender can seek to fully verify
a borrower’s income in the approval process, the likelihood of borrowers overenhancing their stated
earnings is lower compared with self-certificated incomes. This is reflected in a significant differentiation
of the base multiples for the Fast Track versus the Self-Certification product types.
Self-Employed
Self-employed borrowers who do not self-certify their income need to provide the mortgage lender with
documentary evidence of their earnings (e.g., latest tax payments). However, compared with borrowers
who are employees, self-employed borrowers tend to have lower stability in terms of monthly income.
In addition, self-employed borrowers often need to undertake large financial investments in order to
set up their own business, which may make then more vulnerable in an increasingly stressful financial
environment.
Loan-to-Income (LTI)
LTI is a measure of loan affordability and is commonly used by lenders to determine how much they
are prepared to advance on a mortgage. LTI is calculated by dividing the loan balance by the total
income for the household (e.g., the sum of incomes in the case of multiple borrowers). Many lenders
also use more sophisticated affordability measures to take into account other financial commitments
(e.g., council tax, unsecured loan repayments, childcare costs, utility bills, etc.), but usually in conjunction with LTI measures. In a recent report, the CML indicated that approximately 85% of U.K. lenders
continue to use LTI measures (either as a sole determinant of the maximum loan amount or as a crosscheck to the affordability measure).5 Although it is likely that more complex affordability measures are
better indicators of risk than a simple LTI, the components of these measures are not consistent across
lenders. As such, DBRS considers LTI a simple but effective means of assessing affordability.
14
4. Financial Services Authority (FSA). “FSA Finds Improvements from Lenders but Mixed Results for Brokers in SelfCertified Mortgages,” 14 November 2005.
5. Council of Mortgage Lenders (CML). “U.K. Mortgage Underwriting,” April 2006.
November 2007
In the past, loans were generally restricted to 3.0 times a single income or 2.5 times joint incomes. Recently,
lenders have generally started to increase income multiples, sometimes up to 4.0 or even 5.0 times their
earnings, for borrowers with low LTVs or good credit history. Higher LTI ratios are a sign of greater financial commitment and make a borrower more susceptible to default in case of life changes such as divorce
and unemployment. As a consequence, DBRS applies a risk adjustment to LTIs that exceed 3.5.
Single Income
When mortgage repayments are serviced by two separate incomes, if one income becomes unavailable
(e.g., as a result of unemployment), being able to rely on a co-borrower’s income mitigates the likelihood
of default. As such, the repayments on a mortgage serviced by a single income attract a multiple to the
benchmark PD rate.
Right-to-Buy (RTB)
The RTB scheme was originally introduced in the United Kingdom in 1980. Under the scheme, council
tenants and tenants of registered social landlords or housing associations can buy their own homes at
a low price, because part of the rent paid over the previous years of tenancy is discounted from the full
market value. Borrowers who exercise their RTB typically have more fragile economic backgrounds and
are likely to have relied on some form of financial support in the past. As owner-occupiers, however,
these borrowers do not receive any additional housing benefit to assist them with their mortgage repayment. For this reason, DBRS considers loans granted on the basis of this scheme as riskier compared
with standard mortgage loans.
Loan Purpose
Borrowers apply for mortgages primarily for two reasons: home purchase or re-mortgage. As seen in
Figure 1.4, re-mortgage activity in the United Kingdom has boomed over the last years.
% of mortgage originations
Figure 1.4: Distribution of Gross Mortgage Lending by Loan Purpose
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Time
Home Purchase
Re-mortgage
Other*
* Includes lifetime mortgages, further advances and buy-to-let.
Typically, when borrowers re-mortgage, they use the proceeds from the re-mortgage to pay down an
already existing mortgage, with the same property being used as security. The main motivation for this
type of re-mortgage, also referred to as refinancing, is usually to take advantage of a more favourable
interest rate offered by an alternative mortgage provider.
Over the last years, strong house price appreciation persuaded many borrowers to re-mortgage in order
to release equity from their property. As such, a growing proportion of borrowers are raising capital
from their properties, hence taking on more debt. Debt consolidation is a particular form of equity
release re-mortgaging, where one loan (e.g., the re-mortgage) is taken to pay off other debts already
existing (e.g., unsecured loans, credit lines, etc.).
15
November 2007
DBRS considers only raising capital and debt consolidation to be associated with a higher likelihood of
default compared with traditional mortgages. Being a form for increasing credit exposure, they contribute to stretch borrowers’ finances, potentially compromising their ability to repay their debts.
Repayment Type
There are currently two main mortgage repayment methods in the United Kingdom: repayment and
interest-only (IO), although there are many variations of each of these two types (e.g., a mixture of the
two, where an IO reverts to repayment after a certain time period, investment backed, etc.).
In a standard repayment mortgage, both interest and some of the capital borrowed is paid back over
time to ensure the mortgage is totally paid off by the end of the term. In contrast, IO mortgages only
require the repayment of the interest on the initial principal balance until maturity, when the borrower
repays the principal balance.
There is a general trend toward the growing use of IO or mixed mortgage products that allow borrowers
to defer the payment of principal (see Figure 1.5). There are a number of possible reasons for this. Firstly,
mortgage originators recognize that IO mortgages can service borrowers who require more flexibility in
the way they repay their mortgages. For example, those who have fairly low earnings but expectations for
extra financial income (e.g., bonuses) can benefit from smaller regular payments of interest and a more
flexible approach to repaying the principal. Secondly, IO mortgages may be becoming more popular as
they require a smaller short-term monthly commitment than a regular repayment. With growing levels of
unsecured consumer indebtedness, combined with high house prices, borrowers may consider IO loans as
a way to afford property that they may not be able to afford with a regular repayment scheme.
Figure 1.5: Distribution of Mortgage Originations by Repayment Method
% of mortgage originations
79.0%
71.5%
67.0%
64.8%
28.5%
28.2%
22.2%
16.0%
2004
6.7%
6.3%
5.0%
2005
2006
4.8%
2007
Time
Capital And Interest
Interest-Only
Mixed
As such, DBRS has some concern that IO borrowers are more likely to have stretched financial circumstances (although certainly this is not necessarily the case). In addition, there are further concerns around
borrowers’ ability to pay back the entire balance due on the mortgage at the maturity date. In the past,
IO loans were usually combined with regular payments to some sort of vehicle to ensure the repayment
of principal at loan maturity. The Financial Services Authority (FSA) recently highlighted, however, that
many IO borrowers did not have a strategy in place for repaying the capital. Although borrowers can
refinance at maturity, the market environment at that future date is unknown and, as such, exposes borrowers to refinance risk.
16
November 2007
Loan Term
Traditionally, the maximum term offered by U.K. mortgage originators was 25 years. Although rare,
mortgage lenders recently have extended possible mortgage terms for some products, sometimes up
to 40 years. DBRS regards repayment mortgages with a final maturity greater than 25 years as riskier
compared with shorter amortizing products. There is a general concern that a borrower may choose a
long amortization term in order to reduce his or her monthly payments and, as with IO products, could
be indicative of some financial vulnerability.
Loan Size
Given that mortgage providers generally limit the maximum loan size based on income multiples, larger
loan amounts are only available to borrowers with higher incomes. DBRS regards larger loans (jumbo
loans) as riskier than smaller loans. The rationale behind this view is that higher incomes are subject
to greater volatility in the event of an economic downturn. Typically, these borrowers are more likely
to rely on significant bonuses and may find it difficult to maintain their financial status when forced to
move to a new job position (as the result of a changing economic environment).
In order to account for the variety of economical environments across the United Kingdom, DBRS
defines jumbo loan limits based on region and on the time of origination (see Table 1.3).
Table 1.3: Example Jumbo Loan Limits (GBP) for Different Regions and Origination Time Periods
Region
Q2, 2006
Q4, 2006
Q2, 2007
Greater London
430,000
460,000
500,000
South East
360,000
375,000
405,000
Nth Ireland
255,000
320,000
385,000
South West
320,000
330,000
355,000
East Anglia
290,000
305,000
320,000
West Midlands
280,000
280,000
295,000
East Midlands
260,000
265,000
275,000
Wales
250,000
260,000
270,000
North West
250,000
255,000
265,000
Yorkshire & Humberside
245,000
250,000
260,000
North
230,000
235,000
250,000
Scotland
215,000
225,000
245,000
Second Lien
A second-lien mortgage is a subordinated loan taken on a property already used as security for an
existing mortgage. Lien positions differentiate levels of subordination in the rights of creditors to receive
proceeds from the sale of the mortgaged property in the event of borrower default. Second-lien mortgages, although a common feature of many mortgage finance products in continental Europe, are less
widespread in the United Kingdom. In the United Kingdom, second-lien mortgages are generally taken
out as an equity release tool for raising capital or to finance the down payment of a purchase. Secondlien loans are potentially granted to borrowers who are unable to re-mortgage easily to release equity
and, as such, may represent a more risky profile than the benchmark.
Loan Product
DBRS has reviewed the increasing variety of interest rate products that are currently available for residential mortgages in the United Kingdom. A brief summary of the most common product types are given
below:
• Standard Variable Rate (SVR) is set by the individual lender and usually increases and decreases in line
with the Bank of England’s base rate. Hence, mortgage interest payments based on SVR are likely to
rise or fall every time the Bank of England modifies the base rate.
17
November 2007
• Tracker (For Life) loans are guaranteed to track the rate set by the Bank of England plus a differential.
This differential remains constant over the entire life of the loan (e.g., the Bank of England’s base rate
+0.55%). Mortgage interest payments will rise and fall with base rate changes.
• Tracker (For Life with Teaser) loan products track the rate set by the Bank of England plus or minus a
differential for an initial period. This differential then increases at a point in the future and then stays
constant for the remaining term of the loan (e.g., the Bank of England’s base rate +0.19% for the first
two years, then increasing to base rate +0.67% for the remaining life). Mortgage interest payments will
rise and fall with base rate changes, but there will be a definite increase in mortgage payments once
the teaser period expires.
• Tracker (Short Term) loan products track the rate set by the Bank of England plus or minus a differential for a short term (commonly between two and five years) and then switch to the SVR for the
remaining life of the loan. Mortgage interest payments will rise and fall with base rate changes, but
there will be a definite increase in mortgage payments once the Tracker period expires (SVR is typically
set above the Tracker rate).
• Discount (Short Term) loan products pay interest on the basis of the SVR minus a discount for a
short term (commonly between two and five years) and then switch to the SVR for the remaining loan
term. Mortgage interest payments will rise and fall with base rate changes, but there will be a definite
increase in mortgage payments once the discount period expires.
• Fixed-Rate (Short-Term) loan products pay interest based on a fixed interest rate (at a rate typically
below the SVR) for a short term (commonly between two and five years) and then switch to the SVR
for the remaining term. Here, the initial interest payments in the fixed-rate period will not rise and fall
with base rate changes, and borrowers may be exposed to an increase in mortgage payments when they
revert to the SVR. The relative amount of this increase is unknown at the time of loan origination.
DBRS has applied risk adjustments to loan products where there is the potential risk of payment shock
(a sharp increase in regular mortgage payments as a result of a change in the interest rate on the loan).
Loans that track the ongoing changes in interest rates over time (e.g., SVR, Full Term Tracker) are not
subject to this risk adjustment. Loans that generally track the ongoing changes in interest rates over
time but are subject to an increase in payments once the mortgage product reverts to a higher rate are,
however, considered to have some exposure to payment shock (e.g., Tracker (For Life with Teaser),
Tracker (Short Term), Discount (Short Term) loan products). Fixed Rate (Short Term) loan products are
seen to have to most potential for significant payment shock, as they do not adjust with increases in the
base rate. If interest rates increase during the fixed period, the borrowers become exposed to a substantial increase in their regular mortgage payments at the time of the switch to the SVR.
Buy-to-Let (BTL)
A BTL mortgage is for the purchase or re-mortgage of a residential property used for investment
purposes. Here, the property is let to tenants as opposed to direct occupation by the borrower. In the
United Kingdom, BTL lending has experienced spectacular growth since the late 1990s, as shown in
Figure 1.6, and corresponds to almost 10% of all U.K. outstanding mortgages.
18
November 2007
120,000
12%
100,000
10%
80,000
8%
60,000
6%
40,000
4%
20,000
2%
0
% of mortgage loans outstanding
£m of mortgage loans outstanding
Figure 1.6: Outstanding Buy-to-Let Mortgages in the United Kingdom (£m and % total value)
0%
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
H2
H1
1999 1999 2000 2000 2001 2002 2002 2002 2003 2003 2004 2004 2005 2005 2006 2006 2007
Time
Mtgs outstanding at end period (% total value) – RHS
Mtgs outstanding at end period (£m) – LHS
Source: CML statistics relating to new specialist lenders (e.g., Paragon Mortgages, Mortgage Express),
hence not taking into account mortgages to investors from mainstream lenders.
The expansion of the BTL market is attributable to several key drivers. Residential property investors have been encouraged by strong house price appreciation and good rental demand.6 In tandem,
BTL lending has expanded significantly over the last few years, and a growing number of lenders offer
bespoke BTL products at an attractive price. Growing volumes, however, have been accompanied by a
growing number of BLT arrears and repossessions. DBRS considers that this increase has been potentially driven by a number of factors:
• A decrease in the minimum required interest coverage ratio (ICR), which is computed as the expected
monthly rental income divided by the monthly mortgage interest payment (see Figure 1.7).
• Higher LTV ratios, mainly as a result of increases in the maximum amount lent to landlords (see Figure 1.7).
Figure 1.7: Buy-to-Let Mortgage Trends
140%
0.8%
120%
0.7%
0.6%
100%
0.5%
80%
0.4%
60%
0.3%
40%
0.2%
20%
0.1%
0%
0.0%
H1
2002
H2
2002
H1
2003
H2
2003
H1
2004
H2
2004
H1
2005
H2
2005
H1
2006
H2
2006
H1
2007
Time
Maximum LTV (LHS)
Minimum Interest Coverage Ratio (LHS)
90+ arrears (RHS)
Source: CML statistics relating to new specialist lenders (e.g., Paragon Mortgages, Mortgage Express), hence not taking into account
mortgages to investors from mainstream lenders.
6. Royal Institute of Chartered Surveyors (RICS). “RICS Residential Letting Survey Great Britain,” July 2006.
19
November 2007
• Rising interest rates.
• Changes in the type and experience level of borrowers accessing the residential property market; a
growing proportion of new entrants are “amateur” landlords.
• An increase in the average aggregate loan ceiling for individual investors, which grew from GBP 1
million in H1 2006, to GBP 1.5 million in H2 2006, and is currently reported to be at around GBP
2.5 million.
BTL mortgages are also exposed to the risk that the property may not be tenanted for part of the year,
meaning the landlord may need to rely on alternative income to cover the loan repayment. Lenders try
to mitigate the above exposure by requiring the rental coverage ratio to exceed 100%, but the surplus
rent may not be sufficient to cover long terms without tenancy, as well as other repairs and maintenance
costs.
DBRS applies a risk adjustment to the benchmark two-year PD for BTL loans based on a DBRS-calculated
lowest base case ICR (the “stabilised” ICR). This analysis calculates loan-level ICRs at six-month intervals over a four-year time period from the date of the portfolio assessment. The ICR at each time period
is calculated using the following equation:
ICR = updated rental income
interest payable
The initial step in this calculation is to update the given rental income, this is not usually updated
through time, and consequently, the rental income for seasoned loans will be out-of-date. DBRS applies
two adjustments to the given rental income: a rental increase and a tenant void period.
Rental Increase Assumption
A rental increase of 3.5% per annum on average has been assumed in order to increase the rental income
from origination to a level more likely to reflect a current rental income. The 3.5% annual increase
equates to the retail price index (RPI) observed over the last three to four years. This may initially
seem like a relatively unsophisticated assumption, given that there will be regional variations in rental
increases, and also rental increases may not track the RPI directly, but other data sources corroborate this approximate increase. An initial analysis on a survey of regional rental incomes indicates that
since 2003, average rents have increased between approximately 0% and 8% per annum, depending on
region. Using the proportion of BTL loans that DBRS has observed within each region, the weightedaverage annual rental increase is calculated at 3%. As such, a 3.5% increase is considered an adequate
proxy for a rental increase assumption.
Rental Void Assumption
A rental void refers to a period in which a property is vacant, and as such, the borrower will not receive
any rental income during that period. An average void period of one month per year has been assumed,
on the basis of an analysis of reported void data from the Association of Residential Letting Agents
(ARLA), shown in Figure 1.8.
20
November 2007
5.0
4.5
4.0
2002
2003
2004
2005
2006
Q1
Q4
Q3
Q2
Q1
Q4
Q3
Q2
Q1
Q4
Q3
Q2
Q1
Q4
Q3
Q2
Q1
3.0
Q4
3.5
Q3
Number of Weeks Void Per Annum
Figure 1.8: UK Buy-To-Let: Average Void Period Per Annum
2007
Time Period (Quarter)
Source: The ARLA History of Buy-To-Let Investment 2001 to 2007.
The interest payable at each six-month time period is calculated on a dynamic basis, taking into account
the actual rate of interest the borrower will pay at that time. For example, if a loan is currently paying on
a short-term fixed rate and is due to switch to a BBR tracker in 20 months, for the time periods 0, 6, and
18 months, the interest payable will be calculated from the current fixed interest rate. At the 24 month
period, the borrower will have switched products, and as such, the current interest payable will be calculated using a BBR assumption plus the stabilised margin. Any underlying floating rate component (e.g.,
Libor 3M, BBR) will be assumed to be the rate at the date of the portfolio assessment, and will remain
unchanged through the four-year time period. This means that any change in the ICR through time is
due to a switch to a different interest rate margin due to the expiration of an initial margin. An example
of the interest rates used for each loan product type specified by DBRS is given in Table 1.4.
Table 1.4: Interest Payable for ICR Calculation
Rate Type
Rate Payable Before Swtich
Rate Payable After Switch
Standard Variable Rate
(SVR)
Current Floating Rate + Stabilised Margin N/A
BBR Tracker (For Life
with Teaser)
Current BBR + Current Margin
Current BBR + Stabilised Margin
BBR Tracker (Short Term)
Current BBR + Current Margin
Current Floating Rate + Stabilised Margin
BBR Tracker (For Life)
Current BBR + Stabilised Margin
N/A
Fixed (Short Term)
Current Coupon
Current Floating Rate/BBR + Stabilised Margin
Fixed (For Life)
Current Coupon
N/A
Discount (Short Term)
Current Floating Rate + Current Margin
Current Floating Rate + Stablised Margin
For each loan, DBRS takes the lowest experienced ICR through the four-year time period as the stabilised ICR and makes a BTL risk adjustment on the basis of this ICR. This adjustment is made using the
function shown in Figure 1.9. Note that static values reported in Table 1.1 on page 10 show examples
of this function at various stabilised ICR levels.
21
November 2007
Figure 1.9: Multiple to Benchmark PD by Stabilised ICR
2.25
Base Multiple
2.00
1.75
1.50
1.25
1.00
140%
135%
130%
125%
120%
115%
110%
105%
100%
95%
90%
0.75
Stabilised ICR
Although a BTL loan can attract a multiple that increases the PD in comparison with the benchmark
PD, note that these types of mortgage loans do not attract the employment or income multiples that are
associated with owner-occupied properties.
Credit Risk Layering
DBRS makes a base case PD adjustment to account for risk layering within a single mortgage loan; that
is, the simultaneous presence of multiple risk factors is assumed to have an adverse effect on PD over
and above that predicted by the single multiple associated with each component. Credit risk layering has
been an important contributor to the rise of arrears and defaults in the U.S. mortgage market in recent
times, and although the presence of credit risk layering in the United Kingdom has not been as prevalent
as in the United States, DBRS considers this to be an important element in ultimate default behaviour.
Simultaneous risk elements include combinations of high LTV (indicating minimal borrower down
payments), past credit problems (e.g., past CCJ and/or bankruptcy or IVA), and high LTI ratios and/or
self-certification.
THE BASE CASE LIFETIME PD ESTIMATE
In order to expand the two-year loan PD estimates to “lifetime” expectations, the two-year estimate
is extended by means of an assumed cumulative default distribution. The cumulative default curve for
mortgages follows a fairly stable pattern over both time and different data sources, with the majority
of defaults on a static portfolio occurring by the end of year five (60 months). The assumed cumulative
default distribution used to derive lifetime PD estimates is given in Figure 1.10 Note for simplification
that the fitted curve has been divided into six- to 12-month segments (this also allows for more stability
in the lifetime default estimates over small changes in seasoning).
22
November 2007
Figure 1.10: DBRS U.K. Cumulative Default Curve
100%
Cumulative % of Defaults
90%
80%
70%
60%
50%
40%
30%
20%
144
138
132
126
120
114
108
102
96
90
84
78
72
66
60
54
48
42
36
30
24
18
6
0
0%
12
10%
Time Since Origination
In order to calculate the lifetime PD estimate for a single loan, the percentage of cumulative defaults that
should have occurred by the number of months the loan is seasoned (the number of months since origination) plus 24 months (the length of time the two-year PD estimate is predicting forward) is derived
from the assumed cumulative default distribution shown in Figure 1.8. Consider the following example
where a loan is seasoned for six months and has a current two-year PD estimate of 3%. The six-month
seasoning plus the 24 months takes the loan to 30 months in the cumulative default curve. Reading from
the bar chart plotted in Figure 1.8 on the previous page, the percentage of cumulative default assumed to
have occurred by 30 months is 37.5% (note that this percentage will be the same for all loans seasoned
between six months and 12 months). This means that the two-year PD estimate of 3% represents 37.5%
of the lifetime PD estimate. The two-year PD estimate therefore needs to be multiplied by 100%/37.5%
(or 2.67) to get the lifetime PD estimate (3% x 2.67 = 8%). This analysis can also be represented by
plotting the implied multiples over loan seasoning instead, shown in Figure 1.11.
Figure 1.11: Two-Year Base Case PD Multiple by Loan Seasoning
4.0
Seasoning Multiple
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
6
12
18
24
30
36
42
48
54
60
66
72
78
84
90
96
102 108 114 120
Time Since Origination
23
November 2007
PORTFOLIO DEFAULT RATE DISTRIBUTION
The analysis described in the previous section details the approach used to estimate “lifetime” base case
loan-level PDs. These estimates have been developed from U.K. loan- and portfolio-level mortgage credit
behaviour from approximately 2002 to 2006 and extended to lifetime PDs using a standardised cumulative
default curve. This period is assumed to represent a benign economic environment for mortgage performance, which DBRS assumes to represent a B rating scenario. Under more stressful economic conditions,
however, a portfolio will exhibit a higher default rate than the base case, such as the levels of default experienced in the early 1990s. The effect of benign and stressful economic environments on mortgage default
rates can be seen in the default time series plotted in Figure 1.12.
Figure 1.12: U.K. Annual Mortgage Defaults over Time
Annual Mortgage Default Rate
More Stressful
– high interest rates
– high unemployment
– house price decline
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
Benign/Base Case
– low interest rates
– low unemployment
– strong house price growth
Time
DBRS assumes that for a single portfolio of mortgage loans, there is a distribution of potential future
portfolio default rates. The default rate that is exhibited by a portfolio is a function of the base case
performance and the prevailing economic conditions, as represented in Figure 1.13. As mortgage loans
in the United Kingdom have not undergone what could be considered extremely stressful (e.g., AAA)
conditions, the full shape of the distribution has clearly not been empirically verified. In addition, given
that mortgage portfolios are generally very large (e.g., greater than 1,000 loans), there is generally no
need to simulate the default of each loan to create a distribution of defaults. A given loan may or may
not default, but with such a large portfolio, the loss incurred by a single loan is negligible, and the
primary concern is the overall portfolio default rate. As a consequence, simple analytical models can be
used to estimate the portfolio default distribution, in particular the “tail” behaviour of the distribution
that extends well beyond historically observed mortgage default rates.
Figure 1.13: Example Distribution of Mortgage Portfolio Default Rates
Probability
Benign/Base Case
More Stressful
Extremely Stressful
Portfolio Default Rate
24
November 2007
The framework used to approximate a distribution of mortgage portfolio default rates is a modified version
of the single factor Gaussian credit loss model first proposed by Vasicek (1987).7 This model allows for the
creation of a hypothetical distribution of mortgage defaults using two parameters: the mean (or expected)
portfolio default rate and the sensitivity of borrowers to changes in the economic environment.
The mean of the portfolio default distribution can be assumed to be the weighted average of the base
case lifetime PDs.8 The variation in the distribution is determined by a measure of borrower sensitivity to
macroeconomic effects. The influence of the economy, despite its complexity (e.g., gross domestic product
(GDP), interest rates, unemployment), can be approximated as a single factor that influences borrowers’
propensity to default. Sensitivity to this factor is equivalent to assuming individual mortgage borrower
performance is correlated, where the higher the sensitivity, the higher the correlation. For a more technical
description of the single factor model framework, please see Vasicek (1987) or Gordy (2003).9
The single factor model approach is very similar to the Basel II methodology for large, well-diversified
mortgage portfolios, with one key difference. DBRS assumes that the single factor correlation changes
over base case default rates, whereas in the Basel II framework, the correlation remains constant at 15%.
Figure 1.12 shows how the correlation changes with the mean portfolio default rate. The correlation
is capped at 25% for all portfolios with expected lifetime base case PD below 2%. It then decreases as
the mean default rate increases and floors at 10% once the mean portfolio default rate reaches 8%. The
decrease in correlation at high default rates can be interpreted as a decrease in the sensitivity of high-PD
borrowers to macroeconomic effects relative to low-PD borrowers (i.e., high-PD borrowers are more
prone to idiosyncratic (borrower-specific) effects than low-PD borrowers).
The result of this correlation “slope” is a decrease in the relative change in default rate a high-PD portfolio would experience under a deteriorating economic environment compared with a low-PD portfolio.
Under this assumption, a sub-prime portfolio with a mean default rate of 10% might produce default
rates as high as 35% in a particularly stressful economic environment (e.g., a relative change of 3.5 times
the mean). Under the same stress scenario, a prime portfolio with a much lower mean default rate of 2%
would show a much greater relative increase (e.g., up to 24%, or 12.0 times the mean).
The correlation assumptions given in Figure 1.14 have been derived from a number of sources. Firstly, an
analysis of U.K. mortgage default rates over time10 reveals estimates in the range of 10% to 30%, depending on the frequency of observation, the time period and the definition of mortgage default. Although
this analysis indicates a suitable range of overall mortgage correlation, data was not segmented by risk
profile. Therefore, it did not allow an analysis of how mortgage correlation potentially varies with
changes in the underlying risk profile of the portfolio. The notion that estimates of correlation decrease
over changes in PD is consistent with other studies11 and produces stressed mortgage default rates that
are broadly consistent with market expectations. DBRS considers that ongoing research into mortgage
correlation estimates extremely important, both from a regulatory and risk-analysis perspective, and will
continue to develop its approach in this area.
7. Vasicek, O. (1987). “Probability of Loss on Loan Portfolio.” Working paper, Moody’s KMV.
8. DBRS floors the base case portfolio PD estimate at 1%. In order to continue to rank-order portfolios with very low
expected default rates, a scaling factor that decreases as PD increases is applied to all base case portfolio expected
default rates below 2%.
9. Gordy, Michael B. (2003). “A Risk-Factor Model Foundation for Ratings-Based Bank Capital Rules,” Journal of
Financial Intermediation. 12, pp. 199–232.
10. CML statistics.
11. Board of Governors of the Federal Reserve System. “The Asset-Correlation Parameter in Basel II for Mortgages on
Single-Family Residences,” November 2003. Risk Management Association. “Retail Credit Economic Capital Estimation – Best Practices,” February 2003.
25
November 2007
Figure 1.14: Assumed Mortgage Asset Correlation
27.5%
Assumed Correlation
25.0%
22.5%
20.0%
17.5%
15.0%
12.5%
10.0%
7.5%
5.0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
10%
Mean Portfolio Default Rate
In order to illustrate the output of the single factor model, Figure 1.15 shows an example distribution of
portfolio default rates for a portfolio with a mean default rate of 7% and a correlation of 10.5%.
Probability
Figure 1.15: Example of a Mortgage Portfolio Distribution
0%
2%
4%
6%
8%
10% 12% 14% 16% 18% 20% 22% 24% 26% 28% 30% 32% 34%
26
November 2007
RATING-SPECIFIC PORTFOLIO DEFAULT RATES
The analysis described in the previous section results in an analytical distribution of potential default
rates for the mortgage pool. Given that a DBRS rating ultimately addresses the probability of default of
a tranche backed by the mortgage pool, the distribution can be analysed to determine a portfolio default
rate that is consistent with a given rating. This is done by determining the probability that a certain
default rate will be exceeded and ensuring that this probability is less than or equal to the default probability of a ”benchmark bond. To ensure consistency with portfolio models in other asset classes (e.g.,
CDOs), the DBRS corporate default table is used as the benchmark.
As an example, assume a single “A” benchmark bond has a default probability of 3%. We therefore need
to determine the portfolio default rate that has a likelihood of being exceeded that is less than or equal to
3%. Note that this is equivalent to finding the 97th percentile of the default distribution. Using the portfolio distribution given in Figure 1.13, this mortgage default rate is approximately 18% (i.e. the likelihood of
exceeding an 18% default rate is considered by DBRS as being “single ‘A’ remote” – Figure 1.16).
Probability
Figure 1.16: Example of a Mortgage Portfolio Distribution with a single “A” Cut Point
3% of the distribution exceeds
the portfolio default rate of 18%
0%
2%
4%
6%
8%
10% 12% 14% 16% 18% 20% 22% 24% 26% 28% 30% 32% 34%
27
November 2007
Part 2: U.K. Residential Mortgage Loss Criteria
This section describes the DBRS methodology used to calculate the loss given default (LGD) for U.K.
residential mortgage portfolios. This approach is primarily centred on the potential market value decline
a foreclosed property could experience compared with its valuation at the time of portfolio assessment.
As such, the DBRS analysis has focused both on foreclosed property values compared with the general
market norm and on how they may behave under more stressful conditions.
LGD OVERVIEW
Upon default, the property is repossessed and sold to recoup the amount owed by the borrower. Upon
sale, the amount owed by the borrower does not only include the loan principal balance; there will also
be costs associated with the foreclosure process and the forced sale, and given that there is a lag between
severe delinquency status and the actual property being sold, the borrower will also owe accrued interest.
Note that the LGD calculations described in this report exclude accrued interest, as the primary objective of the DBRS credit analysis is to estimate principal loss only.
LGD is calculated by taking the difference between the outstanding principal loan balance owed by the
borrower (also known as exposure at default, or EAD) and the recoveries deriving from the sale of the
property and any other form of credit mitigation in place (e.g., mortgage insurance payments), net of any
costs and prior ranking loans. This difference is then expressed as a percentage of the EAD.
LGD = EAD – (property foreclosure sale price – costs – prior ranking loans)
EAD
In the United Kingdom, most loans are originated with LTVs that are lower than 100%; that is, the loan
principal balance advanced is less than the value of the property. Therefore, upon borrower default, the
sale proceeds should cover the outstanding loan balance, and losses should be minimal. If the market
value is eroded for any reason (property neglect, economic downturn) and repossession and sale costs
are netted from recoveries, then more extreme losses will be observed. The decrease in the property value
is commonly referred to as a market value decline (MVD) and is clearly a key factor when determining
expected losses for mortgage default loans.
COMPONENTS OF LGD
The DBRS methodology for the estimation of each of the contributing components to LGD (e.g., the
amount owed, the costs, the property valuation and the assumed recoveries upon sale) is described in the
following pages. A worked example for a single loan is given in Appendix C.
Principal Amount Owed (Exposure at Default, or EAD)
DBRS expects that loans will default relatively early in their life, with the most default vulnerability
occurring between 12 and 48 months. Loans defaulting within this period are unlikely to show significant decreases in the principal amount owed at origination. Loan products that do amortise tend to
show minimal decreases in the first years of their life, and there are also many non-amortising products
now being originated in the United Kingdom. In addition, although a borrower may manage to pay off
more principal balance through partial prepayments, it is less likely that this borrower type will subsequently default. Therefore, DBRS assumes that the principal amount owed at default is the same as the
balance at the time of the portfolio assessment (e.g., the date of pool cut).
DBRS makes adjustments to the given property valuation on the basis of the property valuation method,
and the time since valuation (in order to account for any increase or decrease in the property valuation
since the given valuation date).
Property Valuation Methods
There are a number of methods that are currently used to value properties in the United Kingdom to
28
November 2007
assess their adequacy as security for a mortgage advance. Historically, lenders have relied on a full
physical valuation, where a property expert such as a chartered surveyor would visit the property in
question. The surveyor valuation is based on the condition of the interior and exterior of the property,
in addition to comparative sales in the vicinity and general market activity. Over the last four to five
years, however, a number of alternative valuation solutions have evolved. In an increasingly competitive
market, these solutions are the result of the pressure lenders are facing to reduce costs and time associated with property valuations. In general, these alternative solutions are restricted to either less risky
loan characteristics (such as low LTV loans) or situations where there is a known past physical valuation
(e.g., re-mortgages, further advances, equity release).
– Drive-by Valuation: A valuer visits the property and assesses it from the property boundary. Comparative
sales and market activity also contribute to the final valuation.
– Desktop Valuation and Automated Valuation Models (AVMs): In both desktop and AVM valuations,
a property is valued without any physical inspection. With a traditional desktop valuation, a house
price index or a comparable property evaluation is used to estimate the property value, usually from
a past known full property value. A more formalised version of the desktop valuation is derived using
an AVM, which assigns a property valuation using a statistical algorithm that can run on an automated
basis once certain property characteristics are entered by the user. The AVM derives values based on
an analysis of comparable sales in the area and property value indexation (e.g., from repeated sales).
The accuracy of an AVM generally depends on the number of suitable comparative properties and
the age of their valuations. Therefore, AVM performance is best when the property comes from a
densely populated homogenous area with a high number of property sales. This statement is true for
all methods of valuation. AVMs, however, are unique in that each valuation produced is accompanied
by an independent measure of “confidence.”
AVM confidence measures are based on the number, similarity and time of the comparable properties
used to calculate the target valuation. The more similar and numerous the comparables are, and the more
recent the sales data, the higher the level of certainty that can be associated with the target property valuation. Surveyor, desktop and drive-by valuations have no such measure of accuracy. This, however, does
not mean they are immune to the specifics of a particular market, which can make valuations inaccurate
and volatile (e.g., sparsely population regions, unique property features, or low comparable sales).
Adjustments Based on Valuation Method
DBRS considers a full surveyor valuation as the U.K. standard, despite in certain situations it also being
susceptible to a degree of inaccuracy. As a consequence, no adjustments are made to such valuations
apart from indexation when appropriate (see below). The other property valuation methods, in some
circumstances, are adjusted downward. This is done to take into account the potential risk of a property
having been overvalued (which would imply a lower LTV than the actual; hence, PD and LGD measures
would be underestimated).
DBRS has undertaken an analysis of AVM performance for the three main U.K. AVM providers:
Hometrack Data Systems Limited, UKValuation Limited and Rightmove plc. This analysis compares
AVM-derived valuations with those provided by a full surveyor valuation or sales price. The relative
over- or undervaluation compared with the surveyor valuation or sale price was calculated using the
following equation, where positive differences represent relative overvaluation and negative differences
represent relative undervaluation.
Difference = AVM valuation – surveyor valuation
AVM valuation
For all three providers, the measures of confidence strictly determined the distribution of AVM valuation
difference relative to the surveyor valuation. Increasing confidence measures were associated with decreasing variation from the accompanying survey or valuation (e.g., lower standard deviations). This general
effect is shown in Figure 2.1a for three groups of confidence measures (i.e., confidence measures classified
as high, medium or low). These distributions relate to individual properties and describe the potential
differential amount an individual AVM-derived property value could have from a surveyor value. On a
29
November 2007
portfolio basis, however, it is unlikely that all AVM-derived properties will have the same level of over- (or
under-) valuation. Assuming a portfolio of 300 and randomly sampling from the property level distributions described above, the overall portfolio variation in accuracy is much reduced (see Figure 2.1b).
Probability
Figure 2.1a: AVM-Derived Valuations from Surveyor Provided Valuations for Individual Properties
-100%
-75%
-50%
-25%
0%
25%
50%
Undervaluation
75%
100%
Overvaluation
Med. Confidence Measure
Low Confidence Measure
High Confidence Measure
Probability
Figure 2.1b: AVM-Derived Valuations from Surveyor Provided Valuations for Property Portfolios
-6.0%
-5.0%
-4.0%
-3.0%
-2.0%
-1.0%
0%
1.0%
2.0%
3.0%
Undervaluation
4.0%
5.0%
6.0%
Overvaluation
Low Confidence Measure
Med Confidence Measure
High Confidence Measure
DBRS haircuts AVM-derived valuations on the basis of the provider’s measures of confidence (and, therefore,
associated potential variation from a surveyor value). Measures of confidence for each provider have been
divided into three ranges; namely, low, medium and high. Each range is associated with a standard deviation
of the AVM-derived value from the surveyor value as predicted by the confidence measure. For each range
of confidence measures, the potential portfolio difference (at the 99% confidence level) has then been calculated. This estimate gives the associated haircut DBRS assigns to each range. For each AVM provider, Table
2.1 shows the range of confidence measures used to distinguish between low, medium and high confidence
measures. It also provides the associated standard deviations and subsequent valuation haircuts.
Table 2.1: AVM Provider Confidence Measure Ranges and DBRS Associated Valuation Haircuts
Confidence Measure (CM) Range
Hometrack
30
0 to 7
Low
CM <= 3.5
Medium
3.5 < CM < 5.5
High
CM >= 5.5
UKValuation
0 to 5
CM <= 2.2
2.2 < CM < 3.3
CM >= 3.3
Rightmove
A to E
CM = D or E
CM = C
CM = A or B
Standard deviation of property difference
from surveyor valuation
>=20%
<20%, >12%
<=12%
99% confidence level of portfolio difference
from surveyor valuation
4% penalty
2% penalty
1.25% penalty
November 2007
DBRS does not adjust AVM-derived valuations on the basis of characteristics such as property size (e.g.,
very large or very small property valuations), location (e.g., more sparsely populated regions) or property
type (e.g., bungalow or detached house). Although these features are associated with valuation accuracy
difficulties (due to the small number of comparables, or reference properties), they are highly correlated
with all the AVM providers’ measures of accuracy. This means that AVM-derived valuations for properties with these features tend to have lower confidence or higher uncertainty measures overall. Therefore,
given that DBRS adjusts AVM- derived valuations on the basis of the associated accuracy measures,
further penalties would double count the valuation risk for these property types. Note that DBRS applies
a 2% and a 1.5% haircut to all desktop and drive-by derived property values respectively.
Property Indexation
The MVD is applied to the estimated property value at the time of the portfolio assessment. Property
valuations are updated by indexing their given valuation according to a blend of the Halifax House
Price Index12 and the Nationwide House Price Index.13 Average house price appreciation is applied to
the property value up to six months prior the date of the portfolio assessment.
House price inflation adjustments are calculated on a region-dependent basis. In addition to this, DBRS’s
approach is to account for only 50% of the average reported increase. In the case of negative appreciation,
however, the whole depreciation is applied. Table 2.2 shows an example of the property valuation update
calculations for two properties located in different regions and with different dates of original valuation.
Table 2.2: Valuation Update Calculation Examples
Property and Valuation Details
Property 1
Property 2
Location
South East
Wales
Valuation date
15 January 2005
8 March 2004
Valuation
GBP 210,000
GBP 126,000
Average house price at valuation
date at location
GBP 198,400
GBP 122,100
Portfolio assessment date
1 February 2007
1 February 2007
Average house price at portfolio assessment minus six months (1 August 2006)
GBP 214,500 in August 2006
GBP 149,000 in August 2006
Average house price increase
GBP 214,500 - GBP 198,400 = 8%
GBP 198,400
GBP 149,000 - GBP 122,100 = 22%
GBP 122,100
50% of house price increase
0.50% * 8% = 4%
0.50% * 22% = 11%
Updated valuation
GBP 218,400
GBP 139,900
Sale Price of the Foreclosed Property
DBRS believes a forced sale as a result of property repossession will result in a discounted sale price
relative to the norm. Therefore, although average historic house price indexes are useful in estimating
potential MVDs for the housing market as a whole, they do not indicate how repossessed properties
performed relative to the average.
Adjustments to assumed MVDs for repossessed properties based on historical house price declines are
usually made to capture additional risks associated with repossession sales, but there has been little published analyses on how these adjustments are derived. In addition, analyses of historical house price declines
do not take into account how the sale prices of repossessed properties would behave in relatively benign
economic environments, such as those experienced in the United Kingdom over the last four to five years.
Given that DBRS uses this recent time period to benchmark future portfolio default rates, and therefore also to benchmark MVDs, it combines analyses from several data sources. Within this framework,
foreclosed property MVDs have been estimated from portfolio historical loss performance, taking into
consideration the cumulative default rates and LTV distributions. In addition, a significant added value
12. See www.hbosplc.com/economy/housingresearch.
13. See www.nationwide.co.uk/hpi.
31
November 2007
to the analysis was also given by access to loan-level information for foreclosed properties, where both
the original valuation, valuation date and repossession sale price were available. The ability to measure
over time and for different regions how repossessed property values performed compared with normal
sales was the main driver in the analytical approach DBRS adopted with regards to loss determination.
The Benchmark MVD
There is strong anecdotal evidence that foreclosed properties suffer from deferred maintenance (and often
outright vandalism), which, combined with the pressure for a fast sale and the stigma associated with foreclosed
properties, results in foreclosed property sale prices being significantly below those of the norm. Combined with
this anecdotal evidence, DBRS had two main approaches in order estimate a benchmark MVD.
– Approach 1: Over the last four to five years, property sales as a result of borrower default have been
at historically low levels. Despite strong average house price increases over the same period, sale
proceeds have not been enough to cover the outstanding loan principal in all cases. This is evidenced
by securitised portfolios experiencing principal losses, particularly in the non-conforming section of
the market. Figure 2.2 plots the average loss rate of over 80 securitisations backed by non-conforming
U.K. mortgage portfolios, with the earliest date of securitisation in January 2000.
Figure 2.2: Cumulative Average Loss for 80 Non-Conforming U.K. Transactions
% of Principal Loss
0.75%
0.60%
0.45%
0.30%
0.15%
0.00%
0
12
24
36
48
60
Time Since Securitisation (Months)
UK Non-Conforming Average
+ Standard Error of Mean
– Standard Error of Mean
The maximum LTV in the portfolios included in the above chart rarely exceeds 90%. Therefore,
despite a large cushion of at least 10% to cover costs and strong house price growth on average,
lenders are still experiencing principal losses. Therefore, the resale value of repossessed properties
must be in some cases much less than the valuation at origination.
DBRS has analysed a number of non-prime U.K. transactions and has back-solved to obtain average
MVDs estimates by tracking cumulative defaulted property sales and cumulative losses, and assuming
a certain weighted-average (WA) LTV and costs at default. Two portfolio examples of these calculations are given in Table 2.3.
Note in Table 2.3 that LTV estimates have been obtained by analysing the LTV distribution at the
time of securitisation and assuming that the higher LTV loans are more likely to default. Using this
approach, the assumed LTV for defaulted loans was estimated to be around 88.5% on a weightedaverage basis across all analysed transactions (inclusive of 3.5% costs). The resultant MVD estimate
from this analysis was calculated at approximately 23% to 24%.
32
November 2007
Table 2.3: MVD Estimation for Two Example Portfolios
Variable
Portfolio Example 1
Portfolio Example 2
Cumulative PD (actual)
4.4%
0.8%
Cumulative loss (actual)
1.0%
0.1%
LGD = Loss/PD (actual)
22.7%
12.5%
LTV of defaulted loans (estimated)
82% + 3.5% costs
86.5% + 3.5% costs
Sale price (estimated)
85.5% - (85.5% * 22.7%) = 66.1%
90.0% - (90.0% * 12.5%) = 78.8%
Market value decline (estimated)
100% - 66.1% = 33.9%
100% - 78.8% = 21.2%
– Approach 2: DBRS integrated the above preliminary assumptions with analysis based on loan-level
data. Data was assessed on a significant number of mortgage loans that had foreclosed between 2001
and 2006. As part of its assessment, DBRS updated the original valuation price, as previously indicated, and then compared it with the actual sale price at repossession. Loan-level MVDs were then
computed based on the difference between expected sale price and actual foreclosed value. The average
MVD obtained from this analysis was approximately 25%.
As a result of the preceding two approaches, and given the similar results, DBRS assumes the benchmark MVD for foreclosed properties to be approximately 25%.
MVD Benchmark Adjustments per Property
The benchmark MVD is then altered on a property-by-property basis depending on various borrower
and property factors that DBRS assumes to influence a property resale value. The resultant MVD per
loan represents the decline in the repossessed property value in a single B environment. The factors that
result in MVD adjustments are property location, property size and property type.
– Property Location Adjustments: Historical house price trends in the United Kingdom have shown
considerable and persistent regional differences. A simple rule of thumb is that the closer the property
location is to London, the higher the volatility in historical price. Figure 2.3 plots average regional
house prices in the United Kingdom since the early 1970s. As shown in the graph, three time periods
where house prices fell can clearly be recognised.
Figure 2.3: U.K. Historical House Prices
£350,000
£250,000
£200,000
1989–1993
£150,000
£100,000
1980–1982
£50,000
Q2 2007
Q2 2005
Q2 2003
Q2 2001
Q2 1999
Q2 1997
Q2 1995
Q2 1993
Q2 1991
Q2 1989
Q2 1987
Q2 1985
Q2 1983
Q2 1981
Q2 1979
Q2 1977
£0
Q2 1975
Average Regional Price
2004–2005
£300,000
Time
Greater London
North
Northwest
East Midlands
East Anglia
Southeast
Scotland
Yorkshire & Humberside
Wales
West Midlands
Southwest
Northern Ireland
33
November 2007
Based on an analysis of the historical house price volatility indicated in Figure 2.3, DBRS has grouped
the United Kingdom into three main regions, as seen in Table 2.4.
Table 2.4: DBRS Regional Groupings for MVD Adjustments
Group
Regions
MVD Adjustment
1
London, South East, South West, East Anglia (includes Northern Ireland)
1.1
2
East Midlands, West Midlands
1.0
3
North, North West, Yorkshire & Humberside, Wales, Scotland
0.85
Figure 2.4 plots the average worst MVD experienced for each group described in Table 2.4 for the three
time periods when property prices fell in the United Kingdom. Group 1 regions clearly show the largest
relative decrease in price compared with the two remaining groups. DBRS accounts for the MVD differences across regions by applying a multiple that adjusts the benchmark MVD (see Table 2.4).
Average Worst MVD During
Time Period Specified
Figure 2.4: Average Worst MVD per Regional Group
35%
30%
25%
20%
15%
10%
5%
0%
Group 1
Group 2
Group 3
Region
1980 to 1982
1989 to 1993
2004 to 2005
– Property Size Adjustments: Very expensive and inexpensive properties have more volatile and less
liquid resale markets because of the more limited number of potential buyers. In addition, the scarcity
of good comparable valuation benchmarks increases the potential for the valuation of these properties
to be overestimated. As such, DBRS increases the MVD for property valuations for very expensive and
inexpensive properties. To estimate the relative expense of a particular valuation, DBRS computes the
valuation ratio. This ratio compares the valuation of the property with the average in its region at the
time of valuation.
Valuation ratio =
Property valuation
Average property valuation in region
A valuation ratio of less than 1 indicates the property is valued below the average. A valuation ratio
greater than 1 indicates the property is valued greater than average. DBRS has examined valuation
ratio distributions over a range of regions and time periods for the U.K. property market. The distributions show very similar characteristics in terms of valuation ratio ranges, and as such, DBRS uses an
amalgam of these distributions to determine property valuations that are considered extreme. DBRS
increases the MVD for properties with valuation ratios in the lower and upper 10% of the distribution
shown in Figure 2.5 (representing valuation ratios of less than 0.5 and greater than 1.8).
34
November 2007
Probability
Figure 2.5: Valuation Ratio Distribution
-
0.50
1.00
1.50
2.00
2.50
3.00
Valuation Ratio
Lower 10% Cut-Off
Upper 10% Cut-Off
The multiple applied to the MVD for each valuation ratio is given in Table 2.5.
Table 2.5: MVD Multiple for Valuation Ratio Ranges
Valuation Ratio Range
MVD Multiple
<=0.40
1.10
>0.40 - <= 0.45
1.05
>0.45 - <= 0.50
1.025
>0.50 - <= 1.80
1.00
>1.80 - <=2.00
1.05
>2.00 - <=2.45
1.10
>2.45
1.15
– Property-Type Adjustments: DBRS considers that Right-To-Buy properties are also associated with
increased MVDs upon foreclosure. Therefore, it applies an adjustment of 1.1 to properties bought
within this scheme.
– Overall MVD Adjustment Calculations: On a loan-by-loan basis, all adjustments are multiplied
together to obtain a single property-level MVD adjustment. For example, a property may be located in
London (1.1 adjustment) and be five times the average London valuation (1.15 adjustment). The total
MVD for this property is therefore 1.1 * 1.15 = 1.265. This is then multiplied by the benchmark MVD
(e.g., 25% * 1.265 = 31.6%) to obtain the base case MVD for this property.
Sale Price of the Foreclosed Property: Overall Calculations
For every loan in the mortgage pool, DBRS determines an updated valuation, with indexation applied up
to six months prior the pool cut-off date, and then computes the expected sale price at repossession by
subtracting from this value the associated MVD. Table 2.6 shows how DBRS assumptions would affect
two loans originated in diverse locations and with different property features.
Costs
The lender bears a number of costs associated with loan delinquency, repossession and subsequent
property resale; hence, these payments need to be subtracted from the sale proceeds. Costs include
legal fees (e.g., as the result of possession, eviction and property sale procedures), expenditures associated with any property maintenance the sale requires and the estate agency charge. Estate agency fees
are usually calculated as a percentage of the sale price of the property and are therefore based on the
assumed property value after the MVD has been accounted for. Agency fees in the United Kingdom typically vary from 1% to 2% (plus VAT of 17.5%). DBRS assumes that property maintenance costs are also
35
November 2007
a function of property size and considers that lenders will be exposed to fee payments of approximately
3% of the sale price Legal and other miscellaneous fees are assumed to be fixed at GBP 2,500.
Prior Ranking Loans
For second-lien mortgages, any prior ranking balance will be taken into account and deducted from the
property foreclosure sale price. This derives from the fact that, as mentioned earlier, lien positions differentiate levels of subordination in the rights of creditors to receive proceeds in case of foreclosure.
Table 2.6: Sale Price at Foreclosure Calculation Examples
Property and
Valuation Details
Property 1
Property 2
Location
South East
Wales
Valuation
GBP 210,000
GBP 126,000
Average house price at GBP 198,400
valuation date at location
GBP 122,100
Valuation ratio
1.06
1.03
Right-to-Buy
Yes
No
Benchmark MVD
25% * 1.1 * 1 * 1.1 = 30.25%
25% * 0.85 * 1 * 1 = 21.25%
Updated valuation
GBP 218,400
GBP 139,900
Sale price at foreclosure GBP 218,400 * (100% - 30.25%) = 152,334
GBP 139,900 * (100% - 21.25%) = 110,171
LGD PER RATING LEVEL
Foreclosure MVDs per Rating Level
DBRS assumes that MVDs for foreclosed properties are a function of the prevailing economic situation.
DBRS has assumed that MVDs increase as rating levels increase, and that instances of portfolio-wide
market value declines at the high extremes (e.g., 45% to 55%) occur very rarely and under stressful
economic scenarios (e.g., AAA).
The MVDs applied to all properties (assuming no individual property adjustments) at each rating level
are given in Table 2.7 (see Benchmark MVD = 25%). Rating-specific MVDs have also been given for two
examples where the benchmark MVD has been altered on the basis of various property characteristics
(i.e., property size and property location).
Loan-Level and Portfolio-Level LGD Calculations
On a loan-level basis, LGDs are computed for all rating scenarios using the following process. Firstly, the
property value at foreclosure is estimated by combining all valuation adjustments (e.g., indexation and
valuation method). The sale price at foreclosure is then derived using the appropriate loan-level MVD at
each rating scenario. Given that MVDs are rating dependent, the assumed costs will then vary accordingly, because they are a function of the foreclosed sale price. LGD is then calculated by subtracting the
expected foreclosed sale price from the EAD and adding costs and any existing prior ranking balance,
and then dividing the remainder by the EAD itself.
Calculating portfolio-level expected losses by simply taking the weighted average of the loan-level LGD
and multiplying it by the portfolio-level PD could result in biased estimates. Portfolio-level expected
loss is accurately derived only when it is determined as a weighted average on a loan-by-loan basis.
This approach prevents the impact of different levels of co-variance between loan-level PD and LGD
estimates. This approach, however, only results in a portfolio PD and a portfolio expected principal loss
(EL) estimate per rating level, with no portfolio LGD estimate.
Separate estimates of portfolio-level PDs and LGDs are required, however, to correctly run the subsequent
cash flow analysis (to take into account the impact of the length of the foreclosure process on lost interest).
As such, portfolio LGD values are derived from the portfolio PDs and corresponding correctly calculated
ELs (as described above) by dividing the portfolio EL by the portfolio PD at each rating level.
36
November 2007
Table 2.7: MVD Distributions for Benchmark MVDs
Rating Level
Benchmark MVD = 25%
Benchmark MVD = 27.5%
Benchmark MVD = 21.3%
AAA
50.3%
52.8%
46.6%
AA (high)
44.7%
47.2%
41.0%
AA
43.5%
46.0%
39.8%
AA (low)
42.5%
45.0%
38.8%
A (high)
41.1%
43.6%
37.4%
A
40.0%
42.5%
36.3%
A (low)
39.0%
41.5%
35.3%
BBB (high)
37.2%
39.7%
33.5%
BBB
35.2%
37.7%
31.5%
BBB (low)
32.7%
35.2%
29.0%
BB (high)
31.5%
34.0%
27.8%
BB
29.5%
32.0%
25.8%
BB (low)
27.8%
30.3%
24.1%
B (high)
26.0%
28.5%
22.3%
B
25.0%
27.5%
21.3%
B (low)
23.7%
26.2%
20.0%
37
November 2007
Part 3: U.K. Residential Mortgage Prepayment Criteria
DBRS has developed a loan-level prepayment model to estimate base case prepayment behaviour for
U.K. RMBS transactions. The core of this analysis was a detailed assessment of historical prepayment
rates shown by U.K. RMBS mortgage pools. This analysis indicated a number of important determinants
of prepayment behaviour, shown below:
• Interest rate type (e.g., Fixed (Short Term), Discount (Short Term), SVR, etc.).
• Time from origination to date of any interest rate reversion.
• Time from loan analysis date to any interest rate reversion.
• Credit quality of the underlying borrower.
• Loan lien position.
• Arrears status.
• Occupancy (owner-occupied or BTL).
In order to create a loan-specific prepayment vector through time, each loan is assigned to one of five
standard prepayment vectors, depending on lien position and overall rate type (either tracker or SVR, or
reverting loan with either one, two or three years to reversion from time of origination). These vectors
are then adjusted further on the basis of the credit quality of the underlying borrower, whether the loan
is for BTL purposes, or whether the loan is in arrears.
This analysis produces a total of 20 possible vectors (see Figures 3.1, 3.2, 3.3, 3.4 and 3.5). Note that
in all cases prepayments trend upwards in the first three to four years since origination and then plateau
at the constant rate thereafter. This trend is primarily driven by the experience that in the first few years
of a mortgage loan, borrowers are less likely to move to a new home, less likely to refinance, and their
financial situation during this early stage does not usually allow for extra payments. The curves differ in
the overall pattern of prepayment, and the influence of the timing of the rate reversion is clear. Borrowers
are more likely to repay in the months following a rate reversion, as typically, reversions result in higher
overall interest payments.
Second-lien prepayment curves are even more front-loaded, given that many borrowers take these out
while awaiting a rate reversion (and thus end of prepayment penalties) on their main mortgage loan.
Once the main mortgage loan switches, the borrower then refinances the first and second-lien positions
in a single re-mortgage. In addition, second-lien interest payments are typically higher than first-lien
positions and therefore borrowers will attempt to repay these loans earlier.
Due to the risk-based pricing strategies adopted by many lenders in a benign economic period, borrowers with past credit problems are also more likely to prepay than prime borrowers, as they will be able
to “credit cure” in this environment and then refinance at a lower interest rate. Loans in arrears typically
show the lowest prepayment rates historically, as once a borrower is in arrears, refinance options become
more limited.
38
November 2007
60
54
48
42
36
30
24
18
6
12
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
Annual CPR
Figure 3.1: Prepayment Vectors - Loans on a 1-Yr Fixed/Discount Rate Before Reversion by Loan-Category
Loan Seasoning
Past Credit Problems/Self-Cert
Prime
Buy-To-Let
In Arrears
60
54
48
42
36
30
24
18
12
6
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
Annual CPR
Loan Seasoning
Prime
Buy-To-Let
In Arrears
60
54
48
42
36
30
24
18
12
6
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
Annual CPR
Loan Seasoning
Prime
Buy-To-Let
In Arrears
39
November 2007
60
54
48
42
36
30
24
18
6
12
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
Annual CPR
Figure 3.4: Prepayment Vectors - Loans on Standard Variable Rate by Loan-Category
Loan Seasoning
Prime
Buy-To-Let
In Arrears
60
54
48
42
36
30
24
18
6
12
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
Annual CPR
Figure 3.5: Prepayment Vectors - Second-Lien Loans by Loan-Category
Loan Seasoning
Prime
Buy-To-Let
In Arrears
One vector is assigned to each loan on the basis of the loan’s underlying characteristics, and is adjusted
for loan seasoning. As such, a seasoned loan will be assumed to be “further along the curve” and will
show a different profile of prepayments going forward than a non-seasoned loan. The result of the
vector assignment and seasoning adjustment creates a loan-specific prepayment vector for each loan in
the portfolio. In order to create a portfolio view of future prepayment behaviour, an average prepayment curve (weighted by balance) is then calculated. An example using five loans is given in Table 3.1
and Figure 3.6.
Table 3.1: Loan Description for Prepayment Example
40
Loan
Loan Seasoning
Base Prepayment Vector
Interest Product
1
6
Prime
2
10
In Arrears
1 year to switch
GBP 275,000
31.6%
3
24
Buy-to-Let
3 years to switch
GBP 175,000
20.1%
4
2
Past credit problems/self-certication
1 year to switch
GBP 145,000
16.7%
5
15
Past credit problems/self-certication
Second lien
GBP 25,000
2.9%
SVR
Balance
GBP 250,000
Pool Percentage
28.7%
November 2007
60
54
48
42
36
30
24
18
6
12
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
Annual CPR
Figure 3.6: Prepayment Vectors - Example of Loan Level and Portfolio Level Estimates
Time Since Securitisation
Loan 1
Loan 2
Loan 3
Loan 4
Loan 5
Portfolio
Note that DBRS considers the time period over which the prepayment model was created to represent
very benign economic conditions in the United Kingdom, and was also one that promoted high prepayment levels (e.g., falling or steady interest rates, high levels of competition, benign economic environment
allowing for credit curing, etc.). As such, DBRS considers the prepayment rates given in Figures 3.1 to
3.5 to represent near-maximum levels, as any deterioration in the economic environment is likely to lead
to a contraction of prepayment rates.
41
November 2007
Part 4: U.K. Residential Mortgage Loan
Aggregation Criteria
Part of the DBRS loan-level analysis for U.K. mortgage portfolios includes standardised criteria for the
creation of rep lines to aid the accuracy and efficiency of cash flow modelling. The loan characteristics
that contribute to the creation of aggregated payment profiles are the following:
• Interest rate type (e.g., SVR, Tracker(For Life), Fixed(For Life), Tracker (For Life with Teaser), Tracker
(Short Term), Fixed (Short Term), Discount (Short Term)).
• Payment profile (e.g., IO, repayment, or short-term IO).
• Time for loan analysis date to any interest rate reversion.
• Remaining Term (i.e. <=10, <=20, >20 years to maturity).
• Time to Switch (0 to 12, 13 to 24, 25 to 36, and 36+ until switch date for Tracker (For Life with
Teaser), Tracker (Short Term), Fixed (Short Term), Discount (Short Term) interest rate types).
In combination, this creates 171 possible loan aggregations within the portfolio. For each rep line, the
following fields are reported:
• Total balance.
• Weighted-average original term.
• Weighted-average remaining term.
• Weighted-average maturity date.
• IO end date (for short-term IO payment profiles).
• Weighted-average current coupon.
• Weighted-average current margin (not available for Fixed (For Life), Fixed (Short Term)).
• Weighted-average time to product switch (for Tracker (For Life with Teaser), Tracker (Short Term),
Fixed (Short Term), Discount (Short Term)).
• Weighted-average remaining term at switch (for Tracker (For Life with Teaser), Tracker (Short Term),
Fixed (Short Term), Discount (Short Term)).
• Weighted-average stabilised margin.
An example of the rep line outputs from the DBRS U.K. Residential Mortgage Loan Analysis Model is
given in Appendix D.
42
November 2007
Concluding Remarks
The DBRS U.K. Residential Mortgage Loan Analysis Methodology sets the analytical framework for the
DBRS U.K. Residential Mortgage Loan Analysis Model. This model is the quantitative tool that DBRS
uses to assess the credit quality of U.K. mortgage loans, create base case prepayment assumptions, and
produce standardised cash flow rep lines.
The model is available free of charge and in an open format. These features stem from the DBRS aspiration to be fully transparent to the market in terms of its own modelling approach, but also to give users
control over the assumptions, allowing them to conduct their own sensitivity analysis and stress testing,
if desired.
DBRS welcomes comments from market participants on the approach and assumptions outlined in this
article. Please send your feedback to [email protected] or [email protected].
43
November 2007
Appendix A: The DBRS U.K. Residential Mortgage
Loan Analysis Model
To access the DBRS U.K. Residential Mortgage Loan Analysis Model, please visit www.dbrs.com/ukrmbs.
Note: The model requires loan-level data, and as a consequence it will only be functional to market
participants who have access to this information. This access increasingly includes investors, but in the
current climate of uncertainty surrounding market abuse rules, the general dispersion of loan-level data
records may be under some revision. At present, there is no consistent practice to ensure that all potential investors in a deal have access to the same loan-level data and also a lack of clarity as to whether this
data constitutes insider information.
DBRS considers the full disclosure of its methodology and the outputs of the loan analysis a crucial part
of the rating process. The output of the model, which includes a detailed review of the portfolio, the main
credit drivers and the rating scenario estimates (known as the tear sheet, see Appendix B), will therefore
be made available on a timely basis as part of the DBRS rating and ongoing surveillance process.
44
November 2007
Appendix B: The DBRS U.K. Mortgage Loan Analysis
Model Example Tear Sheet
DBRS TEAR SHEET: Page 1
Table 1: Overview
DBRS PD Drivers
Basic Pool Description
Number of Loans
Current Balance
Average Loan Size
Largest Loan
Largest Property
Smallest Property
200
20,268,898
101,344
525,764
750,000
59,000
WA Original LTV
WA Given LTV
WA Valn Method LTV
WA Indexed LTV
WA Full Adjust LTV
LTV Used to Analyse
WAC
WA Original Term (months)
WA Remaining Term (months)
WA Seasoning (months)
0.125%
x
Portfolio
Portfolio
Portfolio
WA
Credit Risk Band
1.98
x
Default Rate
LGD
Expected Loss
MVD
27.51%
24.43%
21.21%
19.64%
18.11%
17.25%
15.95%
14.45%
12.98%
11.60%
9.10%
7.97%
7.07%
6.07%
4.83%
39.49%
32.71%
31.30%
30.16%
28.56%
27.31%
26.18%
24.14%
21.88%
19.07%
17.74%
15.54%
13.67%
11.75%
10.71%
10.86%
7.99%
6.64%
5.92%
5.17%
4.71%
4.18%
3.49%
2.84%
2.21%
1.61%
1.24%
0.97%
0.71%
0.52%
49.3%
43.7%
42.5%
41.5%
40.1%
39.0%
38.0%
36.2%
34.2%
31.7%
30.5%
28.5%
26.8%
25.0%
24.0%
Benchmark PD
76.2%
76.1%
76.1%
75.7%
75.7%
FULLY ADJUSTED
6.33%
279.9
276.0
3.8
Rating Output
CCJs
1.14
x
Bankruptcy
1.13
x
LTV
1.61
x
BTL
1.08
x
LTI
1.04
x
RTB
1.04
x
Fast Track
1.00
x
Self Cert
1.34
x
Self Employed
1.06
x
Single Income
1.11
x
Remortgage
1.00
x
IO
1.11
x
Term
1.07
x
Jumbo
1.04
x
2nd Lien
1.00
x
Interest Product
1.13
x
Credit Risk Layer
1.04
x
Extra
1.00
AAA
AA (high)
AA
AA (low)
A (high)
A
A (low)
BBB (high)
BBB
BBB (low)
BB (high)
BB
BB (low)
B (high)
B
The portfolio
default rate is the
percentage of the
portfolio
assumed to
default in each
rating scenario.
=
Base Case 2 Yr
1.35%
x 3.55
x 1
Seasoning Multiple
Low PD Scaling
The portfolio
LGD is the
percentage of
defaulted loan
amounts not
recovered in
each rating
scenario.
=
Base Case LT
4.83%
Assumed Correlation
13.9%
Table 3: Lien
Table 2: Past and Current Credit Performance
Credit Risk Band
Score
No Score
Sum
A
0
0
0
B
0
136
136
C
0
37
37
D
0
13
13
E
0
14
14
Sum
0
200
200
Credit Risk Band
Score
No Score
Sum
A
0.0%
0.0%
0.0%
B
0.0%
67.9%
67.9%
C
0.0%
17.4%
17.4%
D
0.0%
7.0%
7.0%
E
0.0%
7.7%
7.7%
Sum
0.0%
100.0%
100.0%
Lien
First
Second
Sum
SVR
Tracker(Short Term)
% of portfolio
Credit Risk Band Distribution
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Num.
200
0
200
Pct
100.0%
0.0%
100.0%
Num.
1
0
0
8
182
0
9
0
200
Pct
1.0%
0.0%
0.0%
3.1%
92.1%
0.0%
3.8%
0.0%
100.0%
Table 4: Mortgage Product
Tracker(For Life w Teaser)
Tracker(For Life)
Fixed(Short Term)
Fixed(For Life)
Discount(Short Term)
Other
Sum
See "Rep Lines" sheet for more granular data with respect to the
Mortgage Product type (e.g., time to maturity, time to switch,
WA coupon, WA margin etc)
A
CCJs Amount
<=100
>100 - <=2000
>2000 - <=5000
>5000
Sum
Expected loss in "B"
environment
(represented by UK
mortgage performace
2001 to 2006 )
B
Number
189
5
3
3
200
Pct
94.5%
1.6%
1.8%
2.1%
100.0%
Bankrupt Number
Yes
8
No
192
Sum
200
Pct
3.8%
96.2%
100.0%
C
D
E
Arrears Number
0
150
1
19
2
16
3
4
4
7
5
4
6
0
7
0
8
0
9
0
10
0
11
0
12+
0
Sum
200
Pct
74.9%
8.0%
9.0%
1.4%
5.1%
1.5%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
100.0%
Table 5: Right to Buy
RTB Number
Fail
9
Pass
191
Sum
200
Pct
2.6%
97.4%
100.0%
Debt/Equity Remortgage Number
Y
0
N
200
Sum
200
Pct
0.0%
100.0%
100.0%
Table 6: Loan Type
45
November 2007
Table 7: Loan-to-Value
Number
26
12
30
51
53
27
1
0
200
Pct
8.8%
4.9%
13.8%
25.8%
29.9%
16.3%
0.4%
0.0%
100.0%
Average
39.5%
55.5%
65.1%
75.4%
85.5%
92.2%
95.1%
0.0%
LTV Distribution
60%
50%
% of portfolio
LTV
<=50
50 to <=60
60 to <=70
70 to <=80
80 to <=90
90 to <=95
95 to <=100
>100
Sum
40%
30%
20%
10%
0%
<=50
50 to
<=60
60 to
<=70
70 to
<=80
80 to
<=90
90 to
<=95
95 to
<=100
>100
Table 8: Region Information
Region
East Anglia
East Midlands
London
North
Northern Ireland
North West
Scotland
South East
SouthWest
Wales
West Midlands
Yorks & Humber
Unknown
Sum
Num
4
7
8
23
14
38
29
18
4
12
18
25
0
200
Pct
2.5%
4.0%
9.7%
8.4%
5.5%
18.6%
11.2%
10.6%
3.1%
5.3%
9.8%
11.3%
0.0%
100.0%
Num Jumbo Pct Jumbo Ave Loan
0
0.0%
126,026
0
0.0%
116,608
1
2.6%
245,983
0
0.0%
74,000
0
0.0%
79,245
0
0.0%
99,231
0
0.0%
78,362
0
0.0%
119,320
0
0.0%
159,397
0
0.0%
88,977
0
0.0%
109,855
0
0.0%
91,820
0
0.0%
0
1
2.6%
Ave Valn
162,490
158,613
304,946
110,600
169,772
133,772
99,531
171,867
225,820
121,472
169,833
119,392
0
Table 9: Income (Applies to Non-BTL Loans Only)
Low Valn High Valn
0
0
0
1
0
1
3
1
4
1
6
0
4
0
2
0
0
0
2
0
1
1
2
0
0
0
24
5
Ave Benchmark MVD
27.50%
25.18%
28.02%
21.74%
28.24%
21.49%
21.51%
27.58%
27.50%
21.57%
25.24%
21.31%
0.00%
Table 10: Credit Risk Layering
Income Verf.
Verified
Fast Track
Self Cert
Number
93
0
92
Pct
48.0%
0.0%
45.9%
Single Income
Y
N
Sum
Num.
61
32
93
Pct
30.3%
17.7%
48.0%
Self Employed
Verified
Self Cert
Sum
Number
21
25
46
Pct
9.1%
10.9%
20.0%
High LTI
Fail
Pass
Sum
Num.
4
89
93
Pct
1.9%
46.0%
48.0%
Number
29
0
171
200
Pct
20.4%
0.0%
79.6%
100.0%
Term
>25 Years
<= 25 Years
Sum
Num.
38
133
171
Pct
17.3%
62.4%
79.6%
Number
183
14
2
1
0
0
0
0
0
0
0
200
Pct
90.5%
7.2%
1.3%
1.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
100.0%
Credit Risk Layer
1
2
3
Sum
Num
1
0
1
2
Pct
0.4%
0.0%
0.5%
0.9%
1 = LTV >= 95% & Self Cert/High LTI
2 = LTV >= 90% & Past CCJs/Bankrupt
3 = LTV >= 90% & Past CCJs/Bankrupt & Self Cert/High LTI
Table 11: Loan Type
Repayment
IO
Short Term IO
Repayment
Sum
Seasoning
<=6 Mths
6<=12 Mths
12<=18 Mths
18<=24 Mths
24<=30 Mths
30<=36 Mths
36<=42 Mths
42<=48 Mths
48<=54 Mths
54<=60 Mths
>60 Mths
Sum
46
% of portfolio
Table 12: Seasoning
Seasoning Distribution
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
<=6 Mths
12<=18 Mths
24<=30 Mths
36<=42 Mths
48<=54 Mths
>60 Mths
November 2007
Table 13: Buy to Let
BTL Number
Yes
15
No
185
Sum
200
Pct
6.1%
93.9%
100.0%
WA ICR Through Time (Using Calculation Input Assmptions
140%
130%
% of portfolio
120%
Time Period WA ICR
Current
116.7%
6
116.5%
12
118.5%
18
110.8%
24
95.1%
30
96.7%
36
98.2%
42
99.8%
48
101.3%
110%
100%
90%
80%
Current
6
12
18
24
30
36
42
48
47
November 2007
Appendix C: Loan-level Example Computations
of Lifetime PD, LGD and EL
Example 1
Loan Information
Computed Fields
Original Loan Amount
Current Loan Amount
Prior Charge Amount
Origination Date
Maturity Date
Repayment Type
Loan Purpose
Right-to-Buy
Buy-to-Let
Rental Income
Current Rate Type
200,000
200,000
0
21 Sep 2006
Sep 2026
IO
Purchase
N
N
N/A
SVR
Property Information
Region
Last Valuation
Last Valuation Date
Valuation Method
West Midlands
230,000
15 Aug 2006
Surveyor
Borrower Information
Income Verification
Primary Income
Secondary Income
Self-Employed
Verified
45,750
0
Y
Performance Information
Current Arrears Status (# months))
Total Amount of Past CCJs
Date of Most Recent CCJ
Prior Bankruptcy or IVA
0
0
n/a
N
Securitisation Date
01 Apr 2007
Seasoning (number of months)
6
AVG House Price Appreciation (credit to 50%) 0%
Adjustment for Valuation Method
0%
230,000
Current LTV
87.0%
Original LTV
87.0%
Loan-to-Income
4.37
Credit Risk Band
B
Credit Risk Band Multiple
CCJ Multiple
Bankruptcy Multiple
LTV Multiple
BTL Multiple
LTI Multiple
RTB Multiple
Fast Track Multiple
Self Cert Multiple
Self Employed Multiple
Single Income Multiple
Debt/Equity Re-mortgage Multiple
IO Multiple
Term Multiple
Jumbo Multiple
2nd Lien Multiple
Product Type Multiple
Risk Layer Multiple
Seasoning Multiple
1
1
1
1.94
1
1.25
1
1
1
1.15
1.25
1
1.35
1
1
1
1
1
2.67
Estimating PD
Base Case 2-Year PD = 0.125% * 1.94 * 1.25 * 1.15 * 1.25 * 1.35 = 0.59%
Base Case Lifetime PD = 0.59% * 2.67 = 1.58%
Estimating LGD
West Midlands: AVG House Price at Q3 2006 (i.e. valuation date)
Valuation Ratio
Market Value Decline*
Estimated Sale Price At Foreclosure
Costs
165,256
230,000/165,256 = 1.39
25%
230,000 * (100% - 25%) = 172,500
172,500 * 3% + 2,500 = 7,675
LGD = ( 200,000 – (172,500 – 7,675 - 0 ) ) / 200,000 = 17.59
Estimating EL
EL = 1.58% * 17.59= 0.28%
* No Adjustments to Benchmark MVD (i.e. due to Property Location, Property Size, RTB Scheme)
48
November 2007
Example 2
Loan Information
Computed Fields
Original Loan Amount
Current Loan Amount
Prior Charge Amount
Origination Date
Maturity Date
Repayment Type
Loan Purpose
Right-to-Buy
Buy-to-Let
Rental Income
Current Rate Type
40,000
37,000
100,000
17 Jan 2006
Jan 2026
REP
Purchase
Y
N
N/A
SVR
Property Information
Region
Last Valuation
Last Valuation Date
Valuation Method
West Midlands
150,000
5 Jan 2006
Surveyor
Borrower Information
Income Verification
Primary Income
Secondary Income
Self-Employed
Self Cert
50,000
15,000
Y
Performance Information
Current Arrears Status (# months))
Total Amount of Past CCJs
Date of Most Recent CCJ
Prior Bankruptcy or IVA
2
1,000
15 May 2005
N
Securitisation Date
01 Apr 2007
Seasoning (number of months)
14
AVG House Price Appreciation (credit to 50%) 2.1%
Adjustment for Valuation Method
0%
153,196
Current LTV
89.4%
Original LTV
93.3%
Loan-to-Income
2.1
Credit Risk Band
D
Credit Risk Band Multiple
CCJ Multiple
Bankruptcy Multiple
LTV Multiple
BTL Multiple
LTI Multiple
RTB Multiple
Fast Track Multiple
Self Cert Multiple
Self Employed Multiple
Single Income Multiple
Debt/Equity Re-mortgage Multiple
IO Multiple
Term Multiple
Jumbo Multiple
2nd Lien Multiple
Product Type Multiple
Risk Layer Multiple
Seasoning Multiple
4
2.45
1
2.06
1
1
1.10
1
1.35
1.15
1
1
1
1
1
1.50
1
1
2.13
Estimating PD
Base Case 2-Year PD = 0.125% * 4 * 2.45 * 2.06 * 1.10 * 1.35 * 1.15 * 1.50 = 6.46%
Base Case Lifetime PD = 6.46% * 2.13 = 13.76%
Estimating LGD
West Midlands: AVG House Price at Q1 2006 (i.e. valuation date)
Valuation Ratio
Market Value Decline*
Estimated Sale Price At Foreclosure
Costs
158,501
150,000/158,501 = 0.95
25% * 1.10 = 27.5%
153,196 * (100% - 27.5%) = 111,067
111,067 *3% + 2,500 = 5,832
LGD = ( 37,000 – (111,067 – 5,832 - 100,000) ) / 37,000 = 85.85%
Estimating EL
EL = 13.76% * 85.85% = 11.81%
* The adjustment to the Benchmark derives from the property having been bought within a RTB scheme.
49
50
>20
<=20
>20
<=20
<=20
>20
>20
>20
>20
<=20
<=20
>20
<=10
<=20
<=20
>20
<=20
<=20
39
112
108
109
6
113
54
55
51
71
75
35
5
74
44
50
40
REP
REP
REP
IO
REP
REP
IO
IO
REP
REP
REP
IO
REP
IO
IO
IO
REP
REP
Remain. Repay
Term
Type
43
Repline
Number
Fixed(Short Term)
Tracker(For Life)
Fixed(Short Term)
Fixed(Short Term)
Fixed(Short Term)
Fixed(Short Term)
Fixed(Short Term)
Rate Type
0 - 12
24 - 36
24 - 36
12 - 24
12 - 24
12 - 24
Fixed(Short Term)
Fixed(Short Term)
Tracker(For Life)
Tracker(For Life)
Fixed(Short Term)
Tracker(For Life)
24 - 36
0 - 12
24 - 36
12 - 24
Discount(Short Term) 12 - 24
SVR
Balance
44,458
52,780
56,944
63,141
91,753
110,250
173,923
201,443
219,329
244,695
245,306
253,492
302,191
512,227
852,455
2,071,903
4,236,786
12 - 24 10,535,822
Time to
Switch
Discount(Short Term) 12 - 24
Sample Rep Lines
180
240
300
240
240
102
300
156
240
293
329
334
300
183
185
307
218
314
173
240
297
237
237
99
292
132
237
287
322
327
298
182
180
304
214
311
WA Orig. WA Remain.
Term
Term
08-Apr-2022
13-Oct-2027
12-Jul-2032
03-Aug-2027
20-Jul-2027
10-Jan-2016
25-Feb-2032
06-Nov-2018
22-Jul-2027
28-Sep-2031
25-Aug-2034
27-Jan-2035
18-Aug-2032
02-Dec-2022
23-Oct-2022
17-Feb-2033
17-Aug-2025
19-Sep-2033
WA Maturity
Date
03-Aug-2027
25-Feb-2032
06-Nov-2018
27-Jan-2035
02-Dec-2022
23-Oct-2022
17-Feb-2033
IO End Date
6.35%
5.30%
5.70%
6.00%
5.95%
5.65%
7.01%
6.85%
6.70%
6.68%
5.77%
6.68%
6.69%
7.44%
7.01%
6.39%
6.15%
6.25%
WAC
-0.20%
0.50%
0.45%
0.69%
1.35%
1.20%
1.18%
0.27%
1.19%
WA
Margin
29
12
33
21
21
18
5
30
34
21
20
20
21
WA Time
to Product
Switch
144
228
264
78
216
269
317
298
147
158
283
194
290
12-Apr-2010
18-Oct-2008
19-Jul-2010
26-Jul-2009
28-Jul-2009
01-May-2009
27-Mar-2008
13-Apr-2010
30-Aug-2010
12-Aug-2009
12-Jul-2009
06-Jul-2009
15-Jul-2009
2.70%
1.90%
2.00%
2.25%
2.20%
2.02%
2.86%
3.40%
2.95%
2.48%
2.53%
3.27%
3.16%
3.49%
3.41%
2.83%
2.59%
2.85%
WA Remain.
Term at
WA Date of Product WA Stablised
Switch
Switch
Margin
0.22%
0.26%
0.28%
0.31%
0.45%
0.54%
0.86%
0.99%
1.08%
1.21%
1.21%
1.25%
1.49%
2.53%
4.21%
10.22%
20.90%
51.98%
% of Pool
November 2007
Appendix D: The DBRS U.K. Mortgage
Loan Analysis Model Example Rep Lines
Copyright © 2007, DBRS Limited, DBRS, Inc. and DBRS (Europe) Limited (collectively, DBRS). All rights reserved. The information upon which DBRS ratings and reports are based is
obtained by DBRS from sources believed by DBRS to be accurate and reliable. DBRS does not perform any audit and does not independently verify the accuracy of the information
provided to it. DBRS ratings, reports and any other information provided by DBRS are provided “as is” and without warranty of any kind. DBRS hereby disclaims any representation or
warranty, express or implied, as to the accuracy, timeliness, completeness, merchantability, fitness for any particular purpose or non-infringement of any of such information. In no
event shall DBRS or its directors, officers, employees, independent contractors, agents and representatives (collectively, DBRS Representatives) be liable (1) for any inaccuracy, delay,
interruption in service, error or omission or for any resulting damages or (2) for any direct, indirect, incidental, special, compensatory or consequential damages with respect to any
error (negligent or otherwise) or other circumstance or contingency within or outside the control of DBRS or any DBRS Representatives in connection with or related to obtaining,
collecting, compiling, analyzing, interpreting, communicating, publishing or delivering any information. Ratings and other opinions issued by DBRS are, and must be construed
solely as, statements of opinion and not statements of fact as to credit worthiness or recommendations to purchase, sell or hold any securities. DBRS receives compensation,
ranging from US$1,000 to US$750,000 (or the applicable currency equivalent) from issuers, insurers, guarantors and/or underwriters of debt securities for assigning ratings. This
publication may not be reproduced, retransmitted or distributed in any form without the prior written consent of DBRS.
www.dbrs.com
Toronto
DBRS Tower
181 University Avenue
Suite 700
Toronto, ON M5H 3M7
TEL +1 416 593 5577
New York
140 Broadway
35th Floor
New York, NY 10005
TEL +1 212 806 3277
Chicago
101 North Wacker Drive
Suite 100
Chicago, IL 60606
TEL +1 312 332 3429
London
1st Floor
25 Copthall Avenue
London EC2R 7BP
TEL +44 (0)20 7562 5600
Paris
27, avenue de l’Opéra
75001 Paris
TEL +33 (0)1 7038 5214
Frankfurt
Trianon Tower
Mainzer Landstrasse 16
60325 Frankfurt
TEL +49 (0)69 9716 8144

UK Residential Mortgage Loan Analysis Criteria: Credit

Transcription

Similar documents

HNSA thanks Ed Zajkowski for his generous loan of the original

February 2013 - Knoxville TVA Employees Credit Union

VA Loans: Ge ng Vets into Homes

How to Establish a UNFCU Membership

Mortgage Services

PDF by DokShop.com

Summer 2015 Augusta VAH - Augusta VAH Federal Credit Union

Express Loan Approval

special auto, bike, boat and RV rates

START JULY WITH A BANG! - United Local Credit Union