Transcript Financial Crisis Origins

Some Comments Lewis (2008)
Some Comments Lewis (2008)
 Interesting
narrative authored by (quasi) insider
 Timely publication
 Fun to read (relative to academic pieces)
Linguistic Frequency Counts
14
12
10
8
6
4
2
0
Fu**
Sh*t
Cr*p
Lewis
 But,
Scr*w
Tung
Endogen*us
Returns
Krawiek
focus “great man” (Eisman) versus “boogie man”
(Gutfreund)
– Only tidbits of NIE questios posed…
NIE Perspectives On Credit Crisis


Contract externalities?
– Structured investment vehicles may create a form of inter se
simplicity (more later), but macro-complexity (Gorton, ‘08)
Agency Costs: Structuring => cash flow rights ≠ control rights?
– But why wasn’t this priced out by purchasers?
» Agency cost turtles all the way down?



Dearth of third party arbitrageurs (i.e., Steve Eismans)?
– Expense, Rational Bubbles, and limits to arbitrage?
Ownership structure (another type of agency cost)?
– Did de-partnerizing skew their incentives? (But see above)
Political Economy, Oversight, and Tipping Points?
– Cheap access to capital by constituents
– Scale/Contingent limited liability; ultimate risk bearers?
The Origins of the Financial Crisis
Eric Talley
UC Berkeley Law School
[email protected]
Overview of the Credit and Credit Derivative Markets
(Focus – Housing Debt; Also Auto / CC debt)
 Before
1990s: Comparatively Simple Market
– Borrowers
– Original Lenders / Originators
– Secondary Market Players (Inc. Fannie; Freddie)
» Had significant constraints on what they could
purchase => significant constraints on borrowers
 Beginning
late 1980s-2000s:
– Structured Finance Industry: Create value by
pooling and rearranging cash flow rights to make
risky assets less risky. RMBSs; CMBSs;
The value proposition? Simple* Example of an MBS
A and B each owe on $1000 face value obligations




Prob{default} for each = 10%; complete default (i.e., recovery rate=0)
– Thus, cash flow rights are uncertain; risk averse buyers would pay at
most $900 to buy mortgage (often substantially less)
Idea (the “A-B MBS”): Combine A’s and B’s payments into single (BR
remote) “pool”; sell off 2 cash flow claims from the combined pool:
– (a) Right to collect the first $1000 to come in (senior tranche);
– (b) Right to collect residual (equity tranche)
If A’s default risk is independent from B’s (will come back to this) then
– Senior tranche holder collects $1000 unless both A and B default,
which happens only 1% of the time (10% x 10%).
– Equity tranche holder collects $1000 only if neither A and B default
(81% of time), and thus default happens 19% of the time.
Result – Financial Alchemy:
– Securitizing turns two “pretty risky” securities into one “safe” asset and
another “riskier” asset.
– Market premium for low-risk assets was high enough to offset discount
on equity tranche.
*Time permitting (it won’t), I’ll discuss a more technical description…
MBS Canonical Structure
MBS Canonical Structure (II)
Growth of Structured Finance…


Downside of MBS – it introduced wedge between owners of
the cash flow rights and control rights; less transparency;
added oversight “goo”…
– Yet, the added goo did not get reflected in the credit
ratings that the senior tranches got from ratings agencies,
so people made (lots of) money…
Logical Next Question – Turn the crank again?
– I.e., could one take the equity tranche of an MBS, pool it
with other equity tranches of other MBSs, and sell off
structured tranches?
– The CDO – a credit derivative security not backed by
mortgages themselves, but instead backed by securities
that were themselves backed by mortgages.
Continue Example
A-B MBS, and suppose that there’s an analogous MBS
pooling the mortgage obligations of C & D




For each MBS equity tranche, recall chance of default for each is 19%.
– Owning rights to payment is uncertain; risk averse buyers would pay
far less than $810.
Combine A-B and C-D equity tranches into a single “pool”; and once
again sell off two tranches:
– (a) Right to collect the first $1000 to come in (senior tranche);
– (b) Right to collect whatever else comes in (equity tranche)
Again if default by the A-B asset is independent from the C-D asset
then
– Senior tranche holder collects $1000 unless both A-B and C-D
default, which happens only 3.6% of the time (19% x 19%)
– Equity tranche holder collects $1000 only if neither A and B default,
and thus default happens 34.4% of the time.
Result – Financial Alchemy, Squared:
– We’ve turned two “pretty risky” equity tranche MBSs into one
“pretty safe” asset and another “extremely risky” asset.
– But with enough credit enhancements and the right structure we
could make the pretty safe asset even safer.
Canonical CDO Structure
Global CDO Issuance
$Billion
550.3
503.0
273.0
Source: Securities Industry and Financial Markets Association
Source: SIFMA, UBS
07
20
06
20
20
20
05
157.4
04
700
650
600
550
500
450
400
350
300
250
200
150
100
50
0
“Synthetic” CDOs and
Credit Default Swaps
Source: wsj.com
Credit Default Swaps: Significant Notional
Counterparty Risk
$Trillions
70
Treasuries
GSE MBS
Corporate Equities
CDS
$62.2
60
50
40
30
20
10
0
2001
2002
2003
2004
2005
2006
2007
2008Q2
CDS Counterparty Risk

Unregulated, privately negotiated bilateral trading structure.
» No standardized capital requirements, no standardized valuation
methods, no standardized contract structure.



No central clearinghouse or system for recording trades.
CDS positions are long and can only be “unwound” with
countervailing positions.
Bears Stearns, AIG, Lehman were all important “sellers” of
CDS – fee businesses with significant liquidity risk
– Sellers needed reliable access to short term debt markets
(repo; corporate paper) to fund their CDS activities (fees;
collateral calls)
Key Events Precipitating Crisis (I)
2001 - early 2007

Profitability of Structured Finance worked its way back
down the supply chain
– Pressure for more low risk MBSs => Pressure for
more mortgages => incentives for brokers =>
incentives for borrowers to refinance
– Reduced documentation, income, LTV requirements
» Some thought this made sense, if the structured finance
market was indeed engineer unsystematic risk out of the
underlying assets.
» But securitized loans appear to have had very distinct risk
characteristics than their non-securitized counterparts…
Did Securitization Invite Reckless Lending?
Default rates for subprime loans just above / below standard
securitization threshold (origination b/t 2001-06)
Source: Keys et al (2008)
Key Events Precipitating Crisis (II)
Late 2007 - 2008



Housing Bubble began to deflate nationally, with three big effects:
– Usual exit option (refinancing prepayments) no longer viable
– Balloon payments (designed to encourage refinancing) now put strain on
borrowers who couldn’t refinance out. Defaults increase
– Defaults more highly correlated nationally than theretofore presumed.
Exacerbating Effects
– Recovery rates / Loss given default begin to rise…
– Significantly: It wasn’t known (and still isn’t) how much higher both defaults
and recovery rates will go….Great uncertainty about how to price risk because of
unknown correlation.
Liquidity Chickens & Roosting / Credit Rationing
– CDS/CDO indicatives plummet => Collateral calls increase
– Protection sellers must sell mortgage backed derivative assets to raise cash and
avoid default.
– Short term debt market (commercial paper / repo) dries up for CDS sellers.
– Required to update accounting statements to “mark to market”, inducing
bankruptcy
Litigation Risks:
How much litigation has the credit / subprime credit
crisis spawned?





Individual Litigation
– Breach of contract, common law fraud, unfair trade practices
– Numerous state actions under state securities laws
Corporate / Securities Litigation
– Depending on how one counts, between federal 140-170
federal securities class actions filings in the past 15 months.
– Breach of fiduciary duty of care/loyalty
– SEC enforcement actions (e.g., Aiding and Abetting)
ERISA Litigation related to mgmt of pension / 401 k assets.
“Gatekeeper” (e.g., auditor) liability claims
– Negligent audit; A&A in breach of F.D.
Litigation Pertaining to TARP/Bailout Itself
Example:
HSH Nordbank v.
UBS
Synthetic CDO
Structure
(Note: German Bank
suing Swiss Bank in
Manhattan; CDO
Created through a CDS
contract tied to an
underlying reference
pool of CDO assets)
Sub-prime / credit crisis related class action lawsuits
22.7%
Market Share
44.9%
250
61.4%
225
176
200
150
101
100
57
40
35
50
0
2007
2008
Credit-Crisis Related Securities Class Actions
Total Securities Class Actions Filed
2009 (Q1)
Extending Basic Model


Debt obligations are not “one shot” deals cash flows
– More generally, cash flow rights from each obligation are
a stream of cash flow rights...
Loss Given Default / Recovery Rate
– Conditional on default, creditor may not be washed out,
but instead may recover some “residual” value. E.g.,

Default
$1000
Default
$1000
Default
$1000
Default
$1000
Default
$1000
Default
$1000
Default
$1000
Default
$1000
$500
$500
$500
$500
$500
$500
$500
$500
Maturity
If periodic payments occur in close enough proximity to one
another, then one can approximate each cash flow stream as
a “continuous” flow of rights…
Technical Detail -- Continuous “Survivorship” Models
(Origins: Biology; Operations Research)

Basic Idea:
– Loan obligation can be thought of as a continuous flow of income,
which lasts up to the point that the obligor defaults or retires debt
– Thus, if X denotes the (random) time of default, then the length of
the cash flow stream is also X, with a cumulative distribution F(X)
» E.g., Exponential Distribution…

A 2nd loan obligation can also be conceived as a
continuous cash flow whose time of default Y
has a cumulative distribution G(Y)
– If we pool these two obligations together,
then to value the pool (or its parts) we need
to understand how / whether the component
parts are related to one another
» I.e., the joint distribution H(X,Y)
– Problem: we often have reliable information
on only “marginal” distributions F(X); G(Y)
Key Concept: “Copula” Functions

A function that combines two (or more) marginal distributions
into a the “true” joint distribution (implied by Sklar’s Theorem)
– H(X,Y) = C(F(X), G(Y))
– Example: “FGM” copula with Exponential Distributions:
» C = F(X)×G(Y)×(1+d×(1-F(X))(1-G(Y))
d=0
d>0





FGM Copula
– C = F(X)×G(Y)×(1+d×(1-F(X))(1-G(Y))
Frank Copula
– C = (1/a) × ln [ 1 + (eaF(X)-1) (eaG(Y) -1) / (ea -1) ]
Gaussian (Normal Distribution) Copula
– C = F (F-1 (F(X)), F-1(G(Y)), r)
Many others…see Wikipedia (seriously)
– While there are infinitely many copulas out there, most
finance quants tended to use the Gaussian copula, as it
has reasonably good mathematical (as opposed to
empirical) properties
– With a good computer, one can also easily implement it
for 3, 4, 5, …. 1000 different cash flow streams
E.g., David Li (infamous working paper, 2001)
Do copula approaches work for valuing credit
derivatives?



Yes, they are good approximations…
…but using them presupposes that one knows:
1. Correct representation of cash flow rights/obligations
2. Correct marginal distributions to use;
3. Correct copula function to use for combining marginals;
4. Correct parameters to feed into the copula (e.g., value of d)
Lots of model uncertainty here. Perilous simplifications
– No explicit modeling of liquidity risks (only default risks)
– Presumption of constant recovery rates
–
–
Estimating correlation parameters using only “normal” (possible
bubble) years
No attempt to