MBS ratings and the mortgage credit boom

for internal use only

MBS ratings and the mortgage credit boom

Adam Ashcraft, Paul Goldsmith-Pinkham, and James Vickery (NY Fed)

2009 Federal Reserve Bank of Chicago Bank Structure Conference

May 7, 2009

Views expressed in this presentation are our own, and do not reflect the opinions of the Federal Reserve Bank of New York or the Federal Reserve System.

2for internal use only

Ratings downgrades: average number of notches0

24

68

(mea

n) d

own_

notc

h

2001q3 2003q1 2004q3 2006q1 2007q3(first) pool_qdate

05

1015

(mea

n) d

own_

notc

h

2001q3 2003q1 2004q3 2006q1 2007q3(first) pool_qdate

Subprime Alt-A

Source: ABSNet


MBS prices down sharply, even for AAA securities


This paper

Goal: evaluate the informational content of initial credit ratings on subprime and Alt-A MBS deals issued in the period leading up to the crisis (2001-07).

Key questions: How informative were credit ratings? Did ratings reflect

risk (measured ex-ante) of underlying mortgages? Did ratings standards decline during the mortgage credit

boom?


The rating agency defense

“In response to the increase in the riskiness of loans made during the last few years and the changing economic environment, Moody’s steadily increased its loss expectations and subsequent levels of credit protection on pools of subprime loans. Our loss expectations and enhancement levels rose by about 30% over the 2003 to 2006 time period…”

“Along with most other market participants, however, we did not anticipate the magnitude and speed of the deterioration in mortgage quality (particularly for certain originators) or the rapid transition to restrictive lending.”

Michael Kanef, Moodys Group MDSenate testimony, 9/26/07


Preview of main findings

Credit ratings insufficiently sensitive to risk.Projected forecast loss rates from a simple default model

strongly forecast worse ex-post deal performance, after controlling for the rating.

CRAs particularly over-rated deals with high share of low-documentation loans and investor loans.Low-doc finding consistent with claims that rating agencies

relied excessively on information from issuers.

Some evidence of deterioration in ratings standards at peak of mortgage credit boom (2005-07)


Stylized structure of an RMBS deal

‘AAA’RMBS

‘AA’RMBS

‘A’RMBS‘BBB’RMBS‘BBB-’RMBS

Residual

RMBSBondsIndividual Mortgages

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10

M11 M12 M13 M14 M15 M16 M17 M18 M19 M20

M21 M22 M23 M24 M25 M26 M27 M28 M29 M30

M31 M32 M33 M34 M35 M36 M37 M38 M39 M40

M41 M42 M43 M44 M45 M46 M47 M48 M49 M50

M51 M52 M53 M54 M55 M56 M57 M58 M59 M60

M61 M62 M63 M64 M65 M66 M67 M68 M69 M70

M71 M72 M73 M74 M75 M76 M77 M78 . . . M2000

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10

M11 M12 M13 M14 M15 M16 M17 M18 M19 M20

M21 M22 M23 M24 M25 M26 M27 M28 M29 M30

M31 M32 M33 M34 M35 M36 M37 M38 . . .

M1000

Source: Kupiec (2008)

A

B

Credit rating measured as fraction of claims below a particular rating notch.

Example:AAA subordination= B / [A+B]


Our data

Security-level data from ABSNet and Bloomberg on the characteristics of each tranche. Original credit ratings, current credit ratings, face value,

coupon, insurance, payment features (e.g. IO, PO), etc.

Loan-level data on individual mortgages underlying each deal, from LoanPerformance. Loan and borrower characteristics (LTV, FICO, DTI, location,

identity of lender etc.) Matched with OFHEO house price data based on location. Performance: Record each month of whether borrower

made payment.


050

100

150

05

01

001

50#

of D

eal

s

2000q1 2002q1 2004q1 2006q1 2008q1Year-Quarter

# of Deals Orig. Amt ($ Bn)

Subprime

050

100

150

Ori

g. A

mt (

$ B

n)

05

01

001

50

2000q1 2002q1 2004q1 2006q1 2008q1Year-Quarter

# of Deals Orig. Amt ($ Bn)

Alt-A

Issuance Over Time

Nonagency MBS issuance


Number of credit ratings by deal type0

.2.4

.6.8

1

Year-Quarter

One Rating Moody & Fitch

S&P & Moody S&P & Fitch

Three Ratings

Subprime

0.2

.4.6

.81

Year-Quarter

One Rating Moody & Fitch

S&P & Moody S&P & Fitch

Three Ratings

Alt-A

Rating Strategy Usage


Credit ratings over time (% subordination below AAA)

05

10

15

20

25

Year-Quarter

AAA Subordination AA/A Subordination

BBB Subordination

Subprime

02

46

8

Year-Quarter

AAA Subordination AA/A Subordination

BBB Subordination

Alt-A

Subordination over Time


Informativeness of ratings

Credit rating is a summary statistic for the level of credit risk of the deal. (Summarizes the CRAs information set).

Cross-sectional predictions:

1.Credit ratings should forecast deal performance (defaults, losses, downgrades etc.).

2.Controlling for the rating, initial risk variables should not systematically forecast deal performance (since to extent relevant, should already be incorporated in rating).


Step 1: Loan-level default model First: produce simple benchmark expected default rate

for each MBS deal, using loan level data. Approach:

1.Estimate historical default model, as function of underwriting variables (LTV, FICO, loan type etc.)

2.Substitute each mortgage into model to obtain projected default rate, and aggregate to deal level.

Model estimated recursively for each half-year vintage between 2001-07. Note: Projected deal default rate based only on “real

time” data as at time of deal issuance.


Model specification Dependent variable: 90+ delinquent after 12 months Key underwriting variables:

Trailing house price appreciation (OFHEO, past 12 months)Borrower FICO score and debt-to income (DTI) ratioCombined loan to valuation (CLTV) ratioLoan type variables (ARM, FRM, interest only, balloon loan)Documentation of borrower income (full, low, no doc)Year x quarter dummiesOthers: lender dummies, origination channel, investor etc.

Linear probability model. Two specifications:Simple model: Similar to Demyanyk & Van Hemert (2009)Complex model: More covariates, interaction terms.


90+ delinquency after 12 months, by deal vintage0

510

15

20

25

Year-Quarter

Actual Complex Model

Simple Model Van Hemert Model

Subprime

-50

510

15

Year-Quarter

Actual Complex Model

Simple Model Van Hemert Model

Alt-A

12-Month Serious Delinquency


Predictors of ex-post default: Subprime

(1) (2) (3) (4) (5) (6)

Subordination below AAA 0.0542*** 0.0195*** 0.0184*** 0.0156***

(0.0110) (0.00477) (0.00498) (0.00437)

Projected delinquency rate 0.372*** 0.349*** 0.347*** 0.310***

(0.0330) (0.0358) (0.0408) (0.0427)

Projected delinquency * boom 0.116***

(0.0254)Deal-level underwriting covariates No No No No Yes Yes*

Year x quarter dummies Yes Yes Yes Yes Yes Yes

Deal structure controls Yes Yes Yes Yes Yes Yes

R2 0.564 0.626 0.807 0.814 0.856 0.862

N 1454 1454 1454 1454 1454 1454

Dependent variable: % deal 90+ delinquent months after issuance


Predictors of rating downgrades: SubprimeDependent variable: Rating downgrade (notches, weighted average)

(1) (2) (3)Subordination below AAA -0.0898*** -0.102*** -0.0890***

(0.0241) (0.0210) (0.0183)Projected delinquency rate 0.179*** 0.169* 0.177**

(0.0634) (0.0917) (0.0704)

Projected delinquency * boom 0.0495(0.133)

Deal-level underwriting covariates No Yes Yes*Year x quarter dummies Yes Yes YesDeal structure controls Yes Yes YesR2 0.615 0.670 0.691N 1536 1536 1536


Deal-level covariates

(1) (2) (1) (2)LTV 0.00556 0.00476 Rating strategy variables

(0.00535) (0.00486) Moodys & S&P -0.0817* -0.0816FICO -0.000229 -6.01e-05 (0.0420) (0.0605)

(0.000998) (0.000946) Moodys & Fitch -0.0118 0.0146HPA 5.713*** 6.238*** (0.0406) (0.0788)

(1.630) (1.672) S&P and Fitch 0.0110 0.0670Low doc 0.573** 0.234 (0.0827) (0.137)

(0.270) (0.285) One rating 0.164* 0.200*Investor 1.365*** 3.015*** (0.0963) (0.110)

(0.415) (1.077) F-test: [p-value] 0.0872* 0.00392***Low doc * boom 0.504***

(0.182) N 1454 1454Investor * boom -2.195** R2 0.856 0.862

(0.968) cont….

Dependent variable: % deal 90+ delinquent after 12 months


Other analysis Also performed all analysis separately for Alt-A deals

Results are similar (omitted given time constraints).

Found similar results studying default at longer horizons90+ delinquency, prepayment with loss or REO after 24

months.

Also estimated determinants of initial ratings.Ratings related to fundamentals (predicted delinquency,

excess spread, insurance etc.) in expected ways.Next slide: Estimate this regression over the pre-boom

period, and compare actual subordination to predicted subordination from this regression.


Actual and predicted subordination16

20

24

Year-Quarter

Predicted Actual

Subprime

56

78

9

Year-Quarter

Predicted Actual

Alt-A


Summary of main findings

Our evidence suggests MBS credit ratings were noisy measures of credit risk of the deal.Projections from a simple default model significantly

outperform ratings as predictors of future deal performance.

Deals with a high fraction of low-doc and investor loans, and loans from areas with high HPA, perform worse ex-post, conditional on the historical data. Combination of CRAs that rate deal systematically related to

performance. Not clear evidence of rating shopping, however.

Some time-series evidence of deterioration in ratings at end of the boom (2006-07)


Additional slides


Predictors of ex-post default: Alt-ADependent variable: % deal 90+ delinquent months after issuance

(1) (2) (3) (4) (5) (6)Subordination below AAA 0.156*** 0.111*** 0.0231* 0.0220*

(0.0434) (0.0247) (0.0115) (0.0122)Projected delinquency rate 0.521*** 0.483*** 0.617*** 0.592***

(0.0410) (0.0463) (0.0392) (0.0475)Projected delinquency * boom 0.0438

(0.0368)Deal-level underwriting covariates No No No No Yes YesYear x quarter dummies Yes Yes Yes Yes Yes YesDeal structure controls Yes Yes Yes Yes Yes YesR2 0.558 0.606 0.701 0.724 0.796 0.798N 1308 1308 1308 1308 1308 1308


Predictors of rating downgrades: Alt-ADependent variable: Weighted average downgrade (notches)

(1) (2) (3)Subordination below AAA 0.0158 -0.0153 -0.0171

(0.0433) (0.0447) (0.0379)Projected delinquency rate 0.286** 0.575*** 0.438***

(0.134) (0.168) (0.115)Projected delinquency * boom 0.416*

(0.229)Deal-level underwriting covariates No Yes Yes*Year x quarter dummies Yes Yes YesDeal structure controls Yes Yes YesR2 0.694 0.710 0.716N 1421 1421 1421


Summary statistics for loans underlying dealsTable 3. Mortgage characteristics

Subprime Alt-A Number of loans per deal, average 5,057 1,984Loan size (average, $000s) 257 435Combined loan-to-valuation ratio (%) Average (%, value-weighted) 82 75 10th percentile 68 58 50th percentile 80 80 90th percentile 95 88 % missingJunior-lien mortgages (% of deal size, avg) 7 0.3FICO scores Average (value-weighted) 625 707 10th percentile 546 647 50th percentile 626 709 90th percentile 708 777 % Missing


Loan-level summary statistics (cont…)

Subprime Alt-A Debt-to-income ratio Average (value-weighted) 30 16 10th percentile 0 0 50th percentile 38 0 90th percentile 50 44Presence of non-amortizing loans* % IO mortgages 17 53 Fraction of deals with IO > 1% 1,139 1,241 Fraction of deals with IO > 75% 33 495Documentation (%)*: Full 59.0 28.2 Low 40.3 65.2 No 0.4 5.8


Alt-A deal characteristics over timePanel A. Alt-A deals

2001 2002 2003 2004 2005 2006Deal characteristicsNumber of deals 51 95 157 265 370 352Deal size, average ($m) 310 388 422 554 642 716Fraction of AAA securities (%) Average 94 95 94 93 93 93 Median 95 95 95 94 94 94Excess spread (median, %) 2.4 2.3 1.7 1.3 1.0 1.0Fraction deals with bond insurance 27 15 11 8 8 4Percent deals rated by all three CRAs 31 9 4 4 6 25

Loan characteristics, value weightedCLTV (%, average) 79 79 75 80 80 82Junior-lien mortgages (average % of deal) 0.1 0.1 0.0 0.2 0.1 0.1FICO, average 682 699 708 706 706 706Debt-to-income (%), average 10 10 11 16 14 17Interest-only mortgages (avg % of deal) 0 2 11 44 57 64Negative-amortization mortgages (avg % of deal)Low/no doc mortgages (avg % of deal) 66 63 64 64 66 7712-month-ended HPA (OFHEO)


Subprime deal characteristics over timePanel B. Subprime deals

2001 2002 2003 2004 2005 2006Deal characteristicsNumber of deals 63 88 166 286 369 421Deal size, average ($m) 473 664 798 1,020 1,050 946Fraction of AAA securities (%) Average 89 87 85 82 80 79 Median 90 85 84 82 79 78Excess spread (median, %) 5.6 6.3 5.8 5.2 3.5 2.8Fraction deals with bond insurance 38 35 18 19 8 6Percent deals rated by all three CRAs 41 49 64 57 53 39

Loan characteristics, value weightedCLTV (%, average) 82 83 83 84 86 87Junior-lien mortgages (average % of deal) 14 10 4 3 5 9FICO, average 602 613 619 622 627 628Debt-to-income (%), average 20 23 26 29 30 31Interest-only mortgages (avg % of deal) 0 0 2 11 27 21Negative-amortization mortgages (avg % of deal)Low- or no-documentation mortgages (avg % of deal) 25 29 33 36 41 4612-month-ended HPA (OFHEO)


Excess spread, bond insurance by asset class over time

05

10

15

20

25

01

23

45

67

Exc

ess

Sp

rea

d

2001q

1

2002

q1

2003q1

2004q

1

2005q

1

2006q1

2007q

1

2008q

1

Year-Quarter

Excess Spread Insurance (%)

Subprime

05

10

15

Insu

ranc

e (

%)

01

23

2001q

1

2002

q1

2003q

1

2004q

1

2005

q1

2006

q1

2007q1

2008q

1

Year-Quarter

Excess Spread Insurance (%)

Alt-A

Other Deal Characteristics over Time


Credit rating agency market share by asset class.2

.4.6

.81

Year-Quarter

Moody's S&P

Fitch

Subprime

.2.4

.6.8

1

Year-Quarter

Moody's S&P

Fitch

Alt-A

Market Share by Agency


Loan-level modelsTable 5. Loan-level models

A. SpecificationModel 1 Model 2 Model 3

Underwriting variables* YES YES YES Underwriting variables x year NO YES NO Underwriting variables x hpa NO NO YES Underwriting Layering NO NO YES Macro variables YES YES YES State Dummies NO YES YES Originator Dummies NO YES YES Issuer Dummies NO YES YES Year Dummies YES YES NO Year-Quarter Dummies NO NO YES

Model R2 0.06 0.07 0.08Number of observations 1160141 1160141 1160141* Underwriting variables used vary slightly by model

Demyanyk and Van Hemert

Include year interactions,

lender dummiesAdditional

layering


Modelling errors / lack of competence view

“We ran our staffing model assuming the analysts are working 60 hours a week and we are short on resources…. The analysts on average are working longer than this and we are burning them out. We have had a couple of resignations and expect more”


Motivation and related literature Role of rating agencies in the subprime crisis:

Bad luck, bad models or bad incentives?

Entry into the credit rating industry restricted by SEC. Argument is that franchise value improves informational

content of ratings and reduces rating shopping.

Failure of “mechanical” models based on historical data


Original and current ratings: GSAMP 2006 NC-2

Class Name

Balance (000s) Ratings (S&P / Moodys) Downgrade notches

Original Current Original Current S&P Moody's AverageA-1 239,618 98,049 AAA/Aaa AAA/B2 0 7 3.5A-2A 214,090 0 AAA/Aaa NR/Aaa 0 0 0A-2B 102,864 60,774 AAA/Aaa AAA/Baa2 0 0 0A-2C 99,900 99,900 AAA/Aaa AAA/Caa2 0 11 5.5A-2D 42,998 42,998 AAA/Aaa AAA/Ca 0 12 6M-1 35,700 35,700 AA+/Aa1 A/C 4 16 10M-2 28,649 28,649 AA/Aa2 BB/C 9 21 15M-3 16,748 16,748 AA-/Aa3 B/C 11 20 15.5M-4 14,986 14,986 A+/A1 CCC/C 13 19 16M-5 14,545 5,348 A/A2 D/C 12 18 15M-6 13,663 0 A-/A3 D/C 11 17 14M-7 12,341 0 BBB+/Baa1 D/C 19 16 17.5M-8 11,019 0 BBB/Baa2 D/C 18 15 16.5M-9 7,052 0 BBB-/Baa3 D/C 17 14 15.5B-1 6,170 0 BBB-/Ba1 NR/C 17 13 15B-2 8,815 0 BB+/Ba2 NR/C 16 12 14X 12,342 0 NR/NR NR/NR

Weighted average downgrade notches (weighted by original balance) 4.68

Documents

MBS ratings and the mortgage credit boom