H/T to Joe Koster of http://www.valueinvestingworld.com/ for the great find:
A Primer on Scion Capital’s Subprime Mortgage Short, By Michael Burry. November 7, 2006
If you do not know who Burry is, do yourself a favor and read the masterpiece
by Michael Leiws,The Big Short: Inside the Doomsday Machine
Subprime mortgages, typically defined as those issued to borrowers with low credit
scores, make up roughly the riskiest one third of all mortgages. The vast majority of
these mortgages fall well within the loan size limits set by Fannie Mae and Freddie Mac,
but are not deemed eligible for purchase by these two mortgage giants for other reasons.
That is, they are non-conforming. For these non-conforming subprime mortgages, the
originator can certainly choose to hold onto the mortgage and retain credit risk in
exchange for the interest payments. Alternatively, the originator can sell subprime
mortgages into the secondary market for mortgages. This secondary market is vast and
deep thanks to the invention of mortgage-backed securitizations back in the 1970s.
In a securitization, a finance company buys up mortgages from the original lenders and
aggregates these mortgages into large pools, which are then dumped into a trust structure.
Each trust is divided into a set of tranches, and each tranche is defined and rated by the
degree of subordination protecting the tranche’s principal from loss. The tranches are
then sold in the cash market to fixed income investors by a placement agent – typically a
well-known securities dealer. The lower-rated tranches may not be offered to investors,
but may be retained by the finance company. Too, the dealer placing the securities with
investors may choose to purchase some of these securities for its own account, either as
an investment decision or to help ensure a full sale of the deal. At the time of the creation
of the trust, a servicer, also rated by the agencies, is hired to administer the mortgages
within the trust. The trustee will manage the trust and all relations with investors,
including monthly reports. The month’s end is typically the 25
For instance, we can take a look at PPSI 2005-WLL1, an early 2005 mortgage deal.
Tranche Description Moodys S&P Fitch Principal
A-1A Senior Float Aaa AAA AAA 600,936,000.00
A-1B Senior Float Aaa AAA AAA 66,769,000.00
M1 Mezzanine Float Aa1 AA+ AA+ 29,049,000.00
M2 Mezzanine Float Aa2 AA AA 26,524,000.00
M3 Mezzanine Float Aa3 AA- AA- 16,419,000.00
M4 Mezzanine Float A1 A+ A+ 14,314,000.00
M5 Mezzanine Float A2 A A 13,472,000.00
M6 Mezzanine Float - NO A3 A- A- 13,051,000.00
M7 Mezzanine Float - NO Baa1 BBB+ BBB+ 10,946,000.00
M8 Mezzanine Float - NO Baa2 BBB BBB 10,525,000.00
M9 Mezzanine Float - NO Baa3 BBB- BBB- 5,894,000.00
M10 Mezzanine Float - NO Ba1 BB+ BB+ 6,315,000.00
M11 Junior Float - NO Ba2 BB BB 8,420,000.00
CE Junior OC Reserve - NO 19,365,046.51
Here, it happens that Argent Mortgage Company and Olympus Mortgage Company
separately originated a set of subprime mortgages, and each sold these mortgages to
Ameriquest Mortgage Company. Ameriquest, which will be the seller in this deal,
deposited these mortgages with a wholly owned subsidiary, Park Place Securities
1Incorporated – PPSI. Park Place is therefore the depositor. Park Place refashioned this
pool of mortgages into a trust, with Wells Fargo Bank being the trustee and Litton Loan
Servicing being the servicer as set out in the Pooling and Servicing Agreement, or PSA.
The Seller hired Merrill Lynch as the placement agent to sell the deal to investors. Those
tranches designated “NO” were not offered to investors but rather retained by Ameriquest
for other purposes. An investor buying a tranche will receive LIBOR plus a fixed spread
that correlates with the tranche’s rating and perceived safety.
Note the senior tranches, designated A-1A and A-1B, make up 79% of this particular
subprime pool. That is, these senior tranches can count on credit support amounting to
21% of the pool as well as any additional credit support that builds up during the life of
these tranches. If the pool experiences write-downs in excess of the credit support for the
senior tranches, then the senior tranches will suffer erosion of their principal. This is
deemed extremely unlikely by the ratings agencies, and these senior tranches therefore
garner the AAA rating.
The mezzanine tranches in this pool include all those tranches that are rated, but not rated
AAA. For the lowest rated tranche – M11 in this particular pool - credit support is just
2.3% at origination. Baa3, or equivalently BBB-, is considered the lowest “investment
grade” rating, and the lowest investment grade tranche in this PPSI deal is M9, which had
4.05% in credit support at origination. Note the M9 tranche is just under $6 million in
size, less than 1% of the original deal size – these are tiny slices of a large risk pool. Still,
the ratings agencies say each tranche is worthy of a difference in the rating due to the
historically very low rate at which residential mortgages actually default and produce
losses. Because home prices have been rising so steadily for so long, troubled
homeowners have been able to refinance, take cash out, and often reduce the monthly
mortgage payment simultaneously. This has had the effect of reducing the rate of
foreclosures. Also because of rising home prices, foreclosures have not resulted in
enough losses to counteract the credit support underlying mortgage-backed securities. To
be perfectly clear, write-downs occur when realized losses on mortgages within the pool
overwhelm the credit support for a given tranche.
Credit support is therefore a key feature worthy of more attention. A tranche will not
experience losses if any credit support for the tranche still exists. In addition to the
structural subordination that contributes the bulk of credit support, finance companies
build in overcollateralization – essentially, throwing more loans into the pool than
necessary to meet the payment obligations of the pool – and the trust itself can engage in
derivatives transactions to insure the pool against loss. An example might be an interest
rate swap that produces excess cash for the pool as rates rise. Over the first couple of
years, which are typically relatively problem-free for mortgages, one already normally
sees an increase in credit support for all tranches. In an era of hysteria over a home price
bubble, one would expect that the organizer of a new mortgage pool would include or
extend use of these extra protections to help further bolster the credit support for the
pool’s tranches. As 2005 came to a close, this is exactly what happened, and this is why I
find many more recent deals much less attractive from a short’s perspective than mid-
2As is always the case, timing is therefore important for an investor short-selling tranches
of mortgage-backed securities. Catching a peak in home prices before it is generally
recognized to be a peak would be critical to maximizing the chances for success.
Now, because the more subordinate tranches are so wafer thin, they are typically placed
with either a single investor or very few investors. Securing a borrow on such tightly
held subordinate tranches would be difficult, and as a result shorting these tranches
directly is not terribly practical. A derivative method was needed - enter credit default
swaps on asset-backed securities.
Credit default swap contracts on asset-backed securitizations have several features not
common in other forms of swap contracts. One feature is cash settlement. Again,
examining PPSI 2005-WLL1 M9 - the BBB- tranche - we see it has a size of $5,894,000.
Because credit default swaps on mortgage-backed securities are cash-settle contracts, the
size of the tranche does not limit the amount of credit default swaps that can be written
on the tranche, nor does it impair ultimate settlement of the contract in the event of
default. By cash-settle, I mean that the tranche itself need not be physically delivered to
the counterparty in order to collect payment. An investor with a short view may therefore
confidently buy more than $5,894,000 in credit default swap protection on this tranche.
As well, these credit default swap protection contracts are pay-as-you-go. This means the
owner of protection on a given tranche need not hand over the contract before full
payment is received, even across trustee reporting periods. For instance, if only 50% of
the PPSI 2005-WLL1 M9 tranche is written down in the first month, the owner of
$10,000,000 in protection would collect $5,000,000 and would not need to forfeit the
contract to do so. If in the second month the remaining 50% is written down, the owner
of protection would collect the remaining $5,000,000.
A mortgage-backed securitization is of course a dynamic entity, and a short investor must
monitor many different factors in addition to the aforementioned credit support. For
instance, as a mortgage pool matures, mortgages are refinanced and prepaid, and the
principal value of mortgages in the pool declines. Prepayments reduce principal in the
senior tranches first. Generally, the idea is that investors in subordinate tranches should
not get capital returned until the senior tranches are paid off. There are some minor
exceptions, but this is generally true. For instance, today, the current face value of the
AAA tranches in PPSI 2005-WLL1, which was issued in March of 2005, is roughly
$243,691,000 versus the original face value of $667,705,000 due to a high rate of
refinancing. Those who can refinance will. Our focus is on those who cannot.
For those who cannot, some mortgages will go bad. Lenders tend to consider loans
delinquent for roughly 90 days of missed payments, and then the foreclosure process
looms. Typically within 90 days but occasionally up to 180 days after foreclosure, the
real estate underlying the bad mortgage is sold. If the proceeds cannot pay off the
mortgage, a loss is realized. If the cash being generated by the mortgage pool cannot
cover the degree of losses, the mortgage pool takes a loss. This is applied to the most
subordinate tranche first.
3Most of these subprime mortgage pools will likely see maximum foreclosures a little over
two years into the life of the pool. The reason is that most subprime mortgages included
in these pools – typically 80% of the mortgages in the pools – are adjustable rate
mortgages. As a result, the mortgage pool will experience its most significant stress when
the initial teaser rate period ends on its set of adjustable rate mortgages. Generally, this
period ends on average 20-24 months from the date of issuance of the mortgage pool.
Since the Funds shorted mortgage pools mostly originated in spring through late summer
2005, I expect the pools shorted will see maximum stress during the latter half of 2007.
No one shorting these tranches would expect to see a payoff during the first year of
holding the short and likely not even during the second year. In fact, the apparent credit
support under each rated tranche will grow during the first year or two. If the thesis plays
out as originally contemplated, the reduction in credit support and ultimately the payouts
on credit default swaps would come shortly after the mortgage pools face their peak
stress, or roughly 2-2.5 years after deal issuance.
In the interim, the value of these credit default swap contracts should fluctuate. In a
worsening residential housing pricing environment, and with poor mortgage performance
in the pools, one would expect that protection purchased on tranches closer to peak stress
would garner higher prices, provided that home prices have not appreciated significantly
during the interim. As well, credit protection purchased on tranches more likely to default
should garner higher prices. I would note that during the summer of 2005, national
residential home prices in the United States peaked along with the easiest credit provided
to mortgage borrowers in the history of the nation. Recent year over year price declines
have not been seen since the Great Depression.
With that in mind, let us examine how the tranches I selected as shorts are performing
relative to the other 2005-vintage deals. The data in this table was compiled by a third
party data provider. This provider captures approximately 80% of all 2005 home equity
deals in its database, which is up to date through August.
Percentages Bankrupt Foreclosed
Loans in Scion 2005 Deals 1.04 3.48 1.32 5.83
Loans in All Subprime 2005 Home
Equity Deals 0.56 2.94 0.75 4.25
Loans in All 2005 Home Equity Deals 0.28 1.48 0.38 2.14
I do believe trends such as these validate the proprietary criteria upon which I selected
the pools for the mortgage short portfolio. While these numbers seem low, the Funds
shorted the more subordinate tranches within these pools specifically so that the short
position would not be dependent on the Armageddon scenario for U.S. residential
Fundamental developments, however, do not necessarily play into pricing of these credit
default swaps while we await peak defaults because most off-the-run deals simply do not
have an active market. So, how exactly are the values of the Funds’ positions priced
4during this time? In a nutshell, our counterparties set the values. The seller of credit
default swap protection is the buyer’s counterparty, and vice versa. The Funds have six
counterparties from which credit protection on subordinated tranches of mortgage-backed
securities has been purchased. The creditworthiness of our counterparties is an integral
part of the investment thesis. We have chosen counterparties that are among the largest
banks and securities houses in the world, and we have negotiated ISDAs with each of
these counterparties. ISDA stands for International Swap Dealer Association, and an
ISDA is the common term for the contract governing the dealings between counterparties
to a swap transaction.
Importantly, we negotiated ISDA contracts that give us the right to collateral should our
swap positions move in our favor. To the extent the Funds see the values of our swap
positions move the other way, the Funds send collateral to our counterparties covering the
decline in value of the positions. This mechanism protects each counterparty in the event
of a default by the counterparty on the other side. The dealer counterparties are the
marking agents for the Funds’ positions, and therefore the values set by these dealer
counterparties determines how the collateral flows on a daily basis.
Scion Capital has been using these same counterparty-assigned contract values that we
use for collateral purposes to determine the net asset value of the Funds. The value of
credit default swaps on subprime mortgage-backed securities is a calculation involving
certain assumptions. For any buyer of protection to have confidence in the value assigned
to his positions, he must have confidence in the methodologies behind the pricing data
provided by his dealer counterparties. The pricing data we receive from our
counterparties is often very old or stale-dated. These prices are sometimes tied to
movements in the on-the-run index products, which contain neither any of our deals nor
any deals remotely similar to our deals- almost all of which are off-the-run. We have
found the methodologies to be frankly inconsistent. In the absence of confidence in
counterparty marks, a third party may be considered, but today there is no sufficient third
party marking agent for credit default swaps on mortgage-backed securities. Some may
rather use a mathematical model to price the portfolio, but Scion Capital does not price its
portfolio securities to models.
The Funds currently carry credit default swaps on subprime mortgage-backed securities
amounting to $1.687 billion in notional value. As I selected these, I was not looking to
set up a diversified portfolio of shorts. Our shorts will have common characteristics that I
deemed to be predictive of foreclosure, and therefore they should be highly correlated
with each other in terms of both the timing and the degree of ultimate performance.
Again, ultimate performance matters much more than the valuation marks accorded us by
our counterparties in the interim. In the worst case, I expect our mortgage short will fully
amortize to nil value over the next three years, corresponding to an average annual cost of
carry over that time of roughly six percent of current assets under management.
Calibrating the more positive outcomes will become easier as 2007 progresses.
Michael J. Burry, M.D.
Scion Capital, LLC
5RMBS CDS & Side Pockets - Some Good Questions
November 7, 2006
Can’t the servicers manipulate these pools? Don’t they advance interest? Generally,
servicers may advance interest payments to the pool when a mortgage goes delinquent.
Once a mortgage is foreclosed upon, the servicer’s advance is typically billed to the
mortgage pool. Servicers are themselves rated and in my view would have little incentive
to refuse to foreclose upon mortgages or delay sales of real estate during a time of
declining home prices. Recent data has implied that servicers have been more willing to
take bigger losses on mortgages as national home price levels weaken. As far as deciding
when a tranche should be written down, this duty is left to the trustee rather than the
servicer. It is the trustee, not the servicer, which administers cash flows to investors
within the trust.
Can’t the manager of the mortgage pool replace bad loans with good ones? For
reasons of fraud and similar concerns, it is often the case that a bad loan may be replaced
during the first six months to one year of a trust’s existence. Nearly all our shorts involve
deals for which this period is past. To the extent such replacement of fraudulent loans
happened, it was disclosed in servicer reports, and it was not significant.
What is loss severity? Loss severity is the average percentage loss realized on mortgages
during the trustee reporting period. Losses on mortgages are realized when the
underlying foreclosed real estate is sold, but proceeds cannot fully repay the mortgage.
What is the deal with the step-down at three years? Is this a concern? This is a
somewhat complex mechanism built into most mortgage pools that allows for the senior
tranches to be repaid relatively quickly if the pool is performing poorly and to be paid
down more slowly if the pool is performing very well. The 37
month is a frequent date
for this mechanism to kick in. Given the subordinated status of the tranches we are short
and the accelerated deterioration of these pools, this mechanism would appear to be not
very relevant to our position.
What is interest rate swap protection and is it relevant? In the earlier years of a
mortgage pool, income is relatively fixed, while the payout to investors in the pool floats
based on LIBOR. Rising rates may cause payouts to exceed income, causing a mismatch.
At the time the mortgage pool is structured, the seller may purchase an interest rate swap
that itself is profitable in the event of higher interest rates so as to mitigate risk of a
mismatch. These swaps typically have a fixed term. This is relevant. Not all pools have
this feature, and all else equal pools with this feature tend to be less interesting as shorts.
How is your portfolio of mortgage shorts split by rating? On a notional basis, 41.6%
and 49.8% of our shorts are on BBB- and BBB tranches, respectively. The remaining are
6Is PPSI 2005-WLL1 representative of the rest of the portfolio? No. This is an example,
and it is not meant to be representative. For instance, many pools do not have a credit
enhancement, certificate of equity, or CE, tranche, like PPSI 2005-WLL1 does.
Commonly, there is an overcollateralization layer that is not specifically set out as a
Do you really believe the dealers are colluding to mark your book low? No. I believe
the dealers are acting in their best interests, but I have no evidence of collusion of any
kind. I do not believe our counterparties best interests are necessarily aligned with the
Funds’ best interests, and I feel it is the better part of prudence to maintain that opinion. I
generally feel people follow the incentives before them.
Why did you ever allow the counterparties to mark your books? I have not been aware
of a better alternative. I have been wary of the conflicts of interest that would arise
should we set foot on the slippery slope that is marking our own book.
Do your concerns with day-to-day valuation affect the enforceability of the CDS
contract in the event the underlying tranche experiences write-downs? No. These are
cash-settle, pay-as-you-go contracts backed by the full credit of our counterparty. When
the trustee reports a tranche has had write-downs, we will have the contractual right to
payment from our counterparty. There will be no assumptions involved, and valuation
will not be a factor.
How will you mitigate losses if it doesn’t work out like you think? Should I detect a
reason for the Funds to exit some or all of these positions, I will seek out ways in which
to liquidate the positions. I am hopeful that our careful monitoring of the Funds’ positions
will lend us the insights necessary to mitigate losses should the need arise.
What is the longest these credit default swaps on mortgage-backed securities can be in
force? The stated life of each swap contract is technically 30 years. Practically however,
prepayment speeds have determined the lifespan, or duration, of mortgage pools for
nearly the entire history of the market in mortgage-backed securitizations. Most dealers
estimate the life of the mortgage pools containing the tranches underlying the swaps in
our portfolio at 2-3 years.
Isn’t there an active market in CDOs? We do not invest in either cash CDOs or
synthetic CDOs. The cash residential mortgage-backed securities, or RMBS, market is
also very large, but we do not participate in this market. The securities we have invested
in are credit default swaps, also known as CDS.
Do synthetic CDOs do the same thing as Scion? No. Synthetic CDOs are roughly
similar in architecture to the PPSI example above, but with credit default swaps on
specific corporate names or on specific asset-backed securities substituting for mortgages.
Buyers of these swaps then provide the cash flows that will support the synthetic CDO.
Generally, buyers of synthetic CDO securities go long a credit while the buyers of the
7swaps are going short the credit. Most of the supply of credit default swaps in 2006 is
tightly linked to the issuance of new synthetic CDOs.
What is the ABX Index? An ABX index is an index of credit default swaps on mortgagebacked securities. There are multiple ABX indices, each defined by a vintage and an
average credit rating. The first ABX index was launched in early 2006, and the structure
of the index bears very little resemblance to the Funds’ portfolio of mortgage shorts. I do
not view any such index as a good proxy for the Funds’ positions.
What are the other side pockets again? Why do the side pockets fluctuate in value a bit?
From the perspective of an investor, the number and level of side pockets will depend on
the timing of the investor’s capital additions to the Funds. The other side pockets are
Livedoor, Blue Ocean Re, and Symetra. All continue to be represented at cost. Any
variation in side pocket value today comes from the fact that the Livedoor position is held
in Japanese yen, while we report in dollars. This leaves that position exposed to foreign
exchange movements. Additionally, side pockets may appear to loom larger when assets
under management have fallen.
If you side pocket these and you get a lot of withdrawals, are the remaining investors
stuck with very large positions in these side pockets? No. The nature of a side pocket is
that exiting investors retain their portion of the side pocket. As a result, the remaining
investors see no increase in concentration in the side pocketed position.
Will you allow investors transparency into all the different positions in the mortgage
CDS side pocket? I hold no plans to offer transparency into these positions, nor do I
expect to compromise the opportunity to trade out of these positions at opportune times.
Why are you not side pocketing the corporate CDS positions? Although we hold offthe-run single name corporate credit default swaps that I do not find to be very liquid,
there is a bona fide and adequate market in corporate credit default swaps. A side pocket
is not necessary.
How big is the corporate CDS portfolio? As of the end of October, single name
corporate CDS amount to 3.27% and 3.55% of assets under management in the Scion
Value Fund and the Scion Qualified Value Fund, respectively. The duration of this
portfolio is roughly 3.5 years. These credit protection contracts cover $4.27 billion in
notional value, largely focused on financial companies. A number of these companies are
engaged in the mortgage business.