Mortgage-Backed Securities and the Financial

Crisis of 2008: a Post Mortem

Juan Ospina1 Harald Uhlig1

1 Department of Economics
University of Chicago

October 2017
Outline
Post Mortem

post mortem: an examination of a dead body to

determine the cause of death.
What we do

Questions:
I What were the losses and returns on non-agency RMBS, in
particular those rated AAA?
I How did the ex-ante rating compare to their ex-post
performance?
I Role of house price boom and bust for RMBS performance?
Approach:
I Create new data set of 143 thousand RMBS bonds.
I Obtain their ratings, their characteristics, their payoff stream.
I Calculate losses, returns.
I Compare to ratings.
I Compare to house price booms and busts, state-by-state.
Data Collection
I We needed to find a source that had some information about the
universe of securities
I Mortgage Market Statistical Annual

I 2013 Edition had information on all non-agency MBS deals issued

between 2006 and 2012
I About 50 pages of tables. 2824 deals.
A sample table from the Stats Annual
Data Collection on Bloomberg
I Searched for the 2824 deals from the Stats Annual
I Searched also for related deals (for example by name of
financial institution) Deal Example
I Once we find a deal, we look back at all deals with similar
name. Goal: get the universe of deals. Total: 8615 deals
Old Deal Example

I For each deal, get tranches (securities, bonds) Tranches Example

I Total: 143232 bonds. Principal: 5.7 trillion $.

I Per bond: obtain 93 variables plus losses and cash flows
Security

I Challenge: Bloomberg places a limit on how much

information can be downloaded per month:
I Max out below 15 thousand securities per month.
I We have more than 140 thousand securities
⇒ It took more than a year to collect all the data
Bloomberg Deal Search I

Back
Bloomberg Deal Search II

Back
Bloomberg List of Securities (Tranches)

Back
Bloomberg Security Example

Back
Data specs: ( Distr.:min, max, mean, 25th, 50th,75th )
Security Identification Credit Rating
Cusip ID Current and Original Ratings (5 ag.)
Deal Name Other Security Characteristics
Deal Manager Credit Support at Issuance
Issuer Company Original Principal Amount
Security Classification Collateral Description
Deal Type (eg. CMBS, RMBS) Mortg.Purp.(% Equ. Takeout, Refin.)
Collateral Type (Home, Auto, Student) LTV Distr..
Collateral Type (ARM vs FRM) Credit Score Distr.
Agency Backed (yes, no) Mortgage Size Distr.
Agency (Fannie Mae, Freddie Mac) MBS metrics 1: w. av. coupon
Dates MBS metrics 2: w. av. Life
Issue Date MBS metrics 3:w. av. maturity
Pricing Date Fraction of ARM and FRM
Maturity Date Occup. (% own, inv., vac.)
Security Description Geographic Information
Bond type (e.g. Floater, i Only) Fraction of mortg. in top 5 states
Tranche Subordination Description Cash Flow and Losses
Coupon Type (e.g. Fixed, Floating) Monthly Interest, Principal Paym.
Coupon Frequency (e.g. Monthly) Monthly Outstanding balance
Coupon Index Rate (e.g. 3M-libor) Monthly Losses
What we find
Seven facts:
1. The bulk of these securities was rated AAA.
2. AAA securities did ok: on average, their total cumulated
losses up to 2013 are under six percent. Their rate of return
was above 2 percent.
3. The subprime AAA-rated RMBS did particularly well.
4. The bulk of the losses were concentrated on a small share of
all securities.
5. Later vintages did worse than earlier vintages, but not
subprime-AAA.
6. Mis-ratings modest for AAA.
7. Controlling for home price bust, a home price boom was good
for repayments.
Together, these facts call into question the conventional narrative,
that improper ratings of RMBS were a major factor in the financial
crisis of 2008.
Fact 1: The bulk of these securities was rated AAA.

MBS Bonds Principal Amount

Rating No. Pct. ($ Billion) Pct.
AAA 65,590.0 56.8 4,535.1 86.9
AA 13,298.0 11.5 297.0 5.7
A 13,355.0 11.6 212.3 4.1
BBB 13,062.0 11.3 118.4 2.3
BB 6,096.0 5.3 40.1 0.8
B 3,865.0 3.3 13.6 0.3
CCC 66.0 0.1 0.3 0.0
CC 22.0 0.0 0.6 0.0
C 51.0 0.0 3.3 0.1
Rated 115,405.0 81.2 5,220.5 91.7
Not Rated 26,774.0 18.8 472.1 8.3
FICO scores vs Prime, Alt-A, Subprime

0.25
Subprime
0.2 Alt-A
Prime
Frequency

0.15

0.1

0.05

0
500 550 600 650 700 750 800 850
Mean FICO Score

0.3 0.7
Subprime Subprime
0.25 0.6
Alt-A Alt-A
Prime 0.5 Prime
0.2
Frequency

0.15
Frequency 0.4

0.3
0.1
0.2
0.05 0.1

0 0
0 1000 2000 3000 0 20 40 60 80 100
Mean Mortgage Loan Size Mean LTV
Losses on AAA securities
Fact 2: AAA securities did ok: on average, their total cumulated losses up to 2013 are
under six percent. Their rate of return was above 2 percent.
Fact 3: The subprime AAA-rated RMBS did particularly well

0.07
Prime
AltA
Subprime
0.06 All AAA

0.05

0.04
Loss Rate

0.03 35.3%

0.02

32.9%
0.01

31.8%
0
2000 2002 2004 2006 2008 2010 2012 2014
Time
Losses on all RMBS

Panel A: Value-Weighted Loss as Fraction of Principal

0.6
All Ratings
0.5 AAA
Investment Grade Ex-AAA
0.4 Non-InvestmentGrade
Loss

0.3

0.2

0.1

0
2000 2002 2004 2006 2008 2010 2012 2014

Panel B: Unweighted Probability of Loss

0.8
All Ratings
AAA
0.6 Investment Grade Ex-AAA
Non-InvestmentGrade
Probability

0.4

0.2

0
2000 2002 2004 2006 2008 2010 2012 2014
Time
Dollar Amount of Losses in Non-Agency RMBS

350

300

250
Loss ($ billion)

200

150

100

50

0
All RMBS AAA-rated Inv. Grade Ex-AAA Non-Inv. Grade
Cash flow example

Example
Deal JPALT 2006-S1
Security Name JPALT 2006-S2 A7 Mtge
Security ID 46627MEX1
Original Rating AAA

Year 2006 2007 2008 2009 2010 2011 2012 2013

Coupon Rate 6.17 6.17 6.17 6.17 6.17 6.17 6.17 6.17
Interest Payments 1,421 2,131 2,131 2,108 1,989 1,212 524 61
Principal Payment - - - 1,247 1,365 1,174 430 96
Loss - - - - 4,844 14,039 7,550 3,802
Balance 34,547 34,547 34,547 33,300 27,091 11,878 3,898 -
Returns 1

T
X it + pt TVT
P0 = t + (1)
t=1
(1 + r ) (1 + r )T
Returns 2

Return Statistic 80% TV 90% TV 100% TV

By Credit Rating
AAA 2.44 2.89 3.31
AA -7.90 -7.01 -6.21
A -10.92 -10.10 -9.35
BBB -13.56 -12.80 -12.11
Inv. Grade Ex AAA -9.01 -8.15 -7.38
By Type of Mortgage
AAA Prime 3.61 3.98 4.33
AAA SubPrime 1.61 2.14 2.62
AAA AltA 1.37 2.01 2.61
Returns 3

Return Statistic 80% TV 90% TV 100% TV

Fixed Rate MBS

AAA Prime Fixed 4.25 4.56 4.84
AAA SubPrime Fixed 4.86 4.96 5.04
AAA AltA Fixed 3.64 4.13 4.58
Floating Rate MBS
AAA Prime Floating 3.03 3.45 3.83
AAA SubPrime Floating 1.45 1.97 2.44
AAA AltA Floating 0.42 1.12 1.76
Fact 4: The bulk of the losses were concentrated on a
small share of all securities.
Panel A: All RMBS Panel B: By Credit Rating
0.7 1 0.5
AAA Loss < 5%
IG Ex-AAA Loss < 5%
Non-IG Loss < 5%
0.6 AAA
Investment Grade Ex-AAA
Non-InvestmentGrade

0.5

Fraction With Loss < 5%

Fraction With Loss > 5%

0.4
Frequency

0.3

0.2

0.1

0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Loss as a Fraction of Principal Loss as a Fraction of Principal
Fact 5: Later vintages did worse than earlier vintages.
Principal-Weighted Losses in RMBS and Credit Ratings:

Rating Full Sample Before 2003 2003 - 2005 2006-2008

AAA 0.0218*** 0.0002 0.0034*** 0.0483***
AA 0.3096*** 0.001 0.1180*** 0.5091***
A 0.3620*** 0.0055*** 0.2000*** 0.6572***
BBB 0.4480*** 0.0334*** 0.3152*** 0.6655***
BB 0.4923*** 0.0653*** 0.4886*** 0.5136***
B 0.5812*** 0.0938*** 0.6989*** 0.5619***
CCC 0.7360*** 0.4125*** 0.4102*** 0.9465***
CC 0.2036*** 0.1364 0.0251 0.2005***
C or Below 0.3863*** 0.0661*** 0.6607*** 0.3604***
Observations 93,902 19,230 38,381 36,291
R-squared 0.3217 0.0852 0.2972 0.485
Standard errors in parentheses
∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01
Fact 5: Vintage FE for Weighted Losses increased ...
All RMBS
Vintage FE 0.15

0.1

0.05

-0.05
2001 2002 2003 2004 2005 2006 2007
Year of Issuance
AAA AA
0.1 1
Vintage FE

0.05 0.5

0 0

-0.05 -0.5
2001 2002 2003 2004 2005 2006 2007 2001 2002 2003 2004 2005 2006 2007
A BBB
1.5 1
Vintage FE

1
0.5
0.5
0
0

-0.5 -0.5
2001 2002 2003 2004 2005 2006 2007 2001 2002 2003 2004 2005 2006 2007
Year of Issuance Year of Issuance
... though AAA-Subprime did not do worse over time.
All RMBS
0.2
Prime
Vintage FE
0.15
Alt-A
Subprime
0.1

0.05

0
2001 2002 2003 2004 2005 2006 2007
Year of Issuance
AAA AA
0.15 1
Vintage FE

0.1

0.05 0.5

-0.05 0
2001 2002 2003 2004 2005 2006 2007 2001 2002 2003 2004 2005 2006 2007
A BBB
1 1
Vintage FE

0.5
0.5
0

0 -0.5
2001 2002 2003 2004 2005 2006 2007 2001 2002 2003 2004 2005 2006 2007
Year of Issuance Year of Issuance
Fact 6: “Misratings”

Compare actual loss rate

lossratei,T = Li,T /Principali,T

to expected loss rate in table by Moody’s.

Moody’s Table
Ex-Ante vs Ex-Post Rating Based on Moodys Ideal Table

90
Ex-Ante Rating
Ex-Post Ideal Rating
80

70

60
Fraction of Securities (%)

50

40

30

20

10

0
AAA AA A BBB BB B CCC CCC below
Credit Rating Level
Ex-Ante vs Ex-Post Rating: Unweighted

60
Ex-Ante Rating
Ex-Post Ideal Rating

50

40
Fraction of Securities (%)

30

20

10

0
AAA AA A BBB BB B CCC CCC below
Credit Rating Level
Ex-Ante vs Ex-Post Rating Based on Moodys Ideal Table

90
Correct Rating
Inflated Rating
80 Deflated Rating

70

60
Fraction of Securities (%)

50

40

30

20

10

0
AAA AA A BBB BB B CCC CCC below
Credit Rating Level
“Misratings”
Fact 7: Loss-Rates and House Price Boom/Busts

I
lossratei,T = βMA ωi,MA + βIL ωi,IL + . . . + βX Xi + i
where ωi,MA is the fraction of principal invested in the state
MA, etc.. (with only five of these weights nonzero), and
where Xi are controls.
I

lossratei,T = = β boom (ωi,MA ∆boom PMA + ωi,IL ∆boom PIL + ...) +

β bust (ωi,MA ∆bust PMA + ωi,IL ∆bust PIL + ...) + . . . +

where ∆boom PMA is the percent change of house prices during

the boom, 2000-2006, ∆bust PMA is the percent change during
the bust 2006-2009, etc..
State-Level Dummies for Loss Rates

with Controls without Controls

0.3 / 1.4 0.2 / 0.8

0.1 / 0.3 -0.1 / 0.2
-0.1 / 0.1 -0.2 / -0.1
-0.3 / -0.1 -0.4 / -0.2
-4.2 / -0.3 -1.1 / -0.4
State-Level House Price Boom and Bust

Boom: 2000-Q1 to 2006-Q4 Bust: 2006-Q4 to 2009-Q4

23.2% / 38% -43.2% / -14.4%

38% / 44.6% -14.4% / -9.6%
44.6% / 72.1% -9.6% / -2.9%
72.1% / 94.1% -2.9% / 0.2%
94.1% / 165.2% 0.2% / 9.1%
House Prices and Loss Rates
(1) (2) (3) (4) (5)
∆HP 2000-2006 0.073*** -0.218*** -0.178***
(0.003) (0.010) (0.012)
∆HP 2006-2009 -0.203*** -0.63*** -0.532***
(0.006) (0.021) (0.020)
Price Reversal -0.238***

Controls No No No No Yes
Observations 93,902 93,902 93,902 93,902 71,316
R-squared 0.0059 0.0107 0.0156 0.0128 0.4345
Standard errors in parentheses
∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01
“Price Reversal” = ∆HP 2006-2009/∆HP 2000-2006
House Prices and Loss Rates per Cohort, no controls
2001 2002 2003 2004 2005 2006 2007
∆ HP 2000-2006:
-0.001 0.001 -0.006 -0.028*** -0.145*** -0.183*** -0.417***
(0.005) (0.004) (0.005) (0.010) (0.024) (0.031) (0.031)
∆ HP 2006-2009:
0.001 0.006 -0.013 -0.077*** -0.403*** -0.665*** -1.112***
(0.011) (0.009) (0.009) (0.020) (0.048) (0.061) (0.062)
R 2 0.000 0.000 0.000 0.002 0.006 0.014 0.027
N 4290 5734 9159 11839 17383 20797 14352
Standard errors in parentheses
∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01
House Prices and Loss Rates per Cohort, with controls
2001 2002 2003 2004 2005 2006 2007
∆ HP 2000-2006
-0.000 -0.009 -0.015*** -0.029** -0.067** -0.151*** -0.392***
(0.006) (0.006) (0.006) (0.011) (0.029) (0.021) (0.029)
∆ HP 2006-2009
0.005 -0.004 -0.019** -0.042** -0.286*** -0.479*** -1.016***
(0.011) (0.010) (0.009) (0.018) (0.042) (0.037) (0.051)
AA 0.000 0.003 0.000 0.014*** 0.218*** 0.676*** 0.647***
A 0.017*** 0.011*** 0.007*** 0.072*** 0.407*** 0.841*** 0.514***
BBB 0.054*** 0.051*** 0.048*** 0.163*** 0.598*** 0.834*** 0.511***
BB 0.040*** 0.035*** 0.221*** 0.378*** 0.520*** 0.536*** 0.534***
B 0.051*** 0.092*** 0.351*** 0.541*** 0.878*** 0.520*** 0.875***
CCC 0.000 0.000 0.154** 0.269*** 0.520*** 0.980* 0.943***
CC 0.000 0.000 0.004 0.001 0.051 0.939 0.585***
Alt-A -0.002* -0.000 0.002** 0.010*** 0.036*** 0.064*** 0.050***
Prime -0.002* -0.001 0.001 0.011*** 0.017*** 0.005 -0.007*
R2 0.112 0.125 0.366 0.404 0.456 0.693 0.496
N 2445 3128 6252 8321 13047 20394 14076
∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01
Prices? Markit ABX-indices for Subprime RMBS ...
AAA AA
120 120

100
100

80
80
60
60
40

40
20

20 0
Jan-06 Jan-08 Jan-10 Jan-12 Jan-14 Jan-06 Jan-08 Jan-10 Jan-12 Jan-14

A BBB
120 120
2006-1
100 100 2006-2
2007-1
80 80 2007-2

60 60

40 40

20 20

0 0
Jan-06 Jan-08 Jan-10 Jan-12 Jan-14 Jan-06 Jan-08 Jan-10 Jan-12 Jan-14
ABX.HE indexes by Markit. Each line represents a vintage of subprime RMBS and the Index. Each panel shows the

evolution of prices over time by credit rating. These indexes are constructed based on 20 deals.
... vs FINRA Survey
Investment Grade Non-Investment Grade
105 100

90
100
80
Price

95 70

60
90 Weighted Mean
25th Pctile
50
75th Pctile
85 40
2011 2012 2013 2014 2015 2016 2017 2011 2012 2013 2014 2015 2016 2017

Investment Grade Pre-2005 Investment Grade 2005-2007

105 105

100 100

95 95
Price

90 90

85 85

80 80
2011 2012 2013 2014 2015 2016 2017 2011 2012 2013 2014 2015 2016 2017
Summary statistics of daily transaction prices collected by the Financial Industry Regulatory Authority from May

2011 through May 2016 on Non-Agency MBS. Top: Investment Grade vs Non-Investment Grade. Bottom:

vintages for Investment Grade. 22-day moving averages, principal weighted average and 25th and 75th percentiles.
Conclusions
Seven facts:
1. The bulk of these securities was rated AAA.
2. AAA securities did ok: on average, their total cumulated
losses up to 2013 are under six percent. Their rate of return
was above 2 percent.
3. The subprime AAA-rated RMBS did particularly well.
4. The bulk of the losses were concentrated on a small share of
all securities.
5. Later vintages did worse than earlier vintages, but not
subprime-AAA.
6. Mis-ratings modest for AAA.
7. Controlling for home price bust, a home price boom was good
for repayments.
Together, these facts call into question the conventional narrative,
that improper ratings of RMBS were a major factor in the financial
crisis of 2008.