CREDIT DERIVATIVES

10 November 2004

Atish Kakodkar
(1) 212 449-0104
Base Correlations
[email protected]
Stefano Galiani Overcoming the Limitations of Tranche Implied Correlations
(1) 212 449-7416
[email protected]
Maksim Shchetkovskiy Global
(44) 20 7996-4780
[email protected]
Derivatives
Tranche Implied Correlations Have Limitations…
Though changes in tranche implied correlations are a good indicator of
demand for the tranches, the implied correlations themselves are limited by
the underlying model and the existence of a skew. These limitations include:
• Difficulty in pricing non-standardized tranches;
• Existence of multiple implied correlations for the mezzanine tranche.

…That Are Addressed by Base Correlations

These limitations can be overcome by base correlations, which are defined
as implied correlations of first-loss tranches where the attachment point is
zero and the detachment point is the detachment point of the various
standardized tranches.
Base correlations are relatively easy to compute. They can be extracted from
market spreads of the standardized tranches via bootstrapping techniques
using the same model that is used to derive tranche implied correlations.
Base correlations are unique and provide an intuitive method to price non-
standardized tranches (say 5-8%) of the underlying CDX or iTraxx portfolio.

Base Correlation Deltas

The tranche delta or leverage of a tranche depends on whether we use
tranche implied or base correlations. Under the base correlation framework
the leverage tends to be lower especially for the mezzanine tranche. We
discuss how deltas are calculated under both frameworks.

Merrill Lynch does and seeks to do business with companies covered in its research reports. As a result, investors should be aware
that the firm may have a conflict of interest that could affect the objectivity of this report.
Investors should consider this report as only a single factor in making their investment decision.

Refer to important disclosures on page 7.

Global Securities Research & Economics Group RC#40931506 Credit Derivatives Research
Base Correlations – 10 November 2004

From Tranche Implied Correlations…

Tranche implied correlations In an earlier report1, we had highlighted the limitations of tranche implied
are limited by the underlying correlations. A key problem is the pricing of non-standardized or bespoke
model and correlation skew tranches. The correlation skew in the standardized market makes it difficult
to price off-market tranches such as, for example, a 5-8% tranche of the
CDX.NA.IG3 index. In addition, pricing of tranches based on portfolios
different from the CDX.NA.IG3 also requires an appropriate estimate of the
default correlation. Another limitation of the tranche implied correlation is
the existence of multiple (or non-unique) solutions for implied correlations
for mezzanine tranches.
These limitations arise due to the use of the standard Gaussian Copula model
and the existence of a correlation skew in the standardized tranche market.
The market continues to debate the use of an appropriate measure for default
correlation. We believe this debate is unlikely to be completely resolved in the
near future and urge tranche investors to looks at correlation trends rather
than absolute levels of correlation. Irrespective of the measurement or model
used, the correlations trends serve as a good indicator of demand in the
tranche market.

…to Base Correlations

Base correlations correspond to One of the methods that is commonly used in the market to address the above
first-loss tranches with issues is the use of base correlations. Base correlations are defined as the implied
standardized detachment points correlations of the first-loss tranches where the attachment point of each tranche is
zero and the detachment point is the detachment point of each of the standardized
tranches. Table 1 highlights the tranche implied correlations for the standardized
tranches and the corresponding first-loss implied base correlations.

Table 1: Base Correlations & Tranche Implied Correlations for Standardized Tranches
North America (CDX.NA.IG3) Europe (iTraxx Europe Series 2)
Std. Bid Ask Mid Tranche Base Base Std. Bid Ask Mid Tranche Base Base
Tranche Correlation Tranches Correlation Tranche Correlation Tranches Correlation
0-3% 34.25% 35.25% 34.75 19.1% 0-3% 19.1% 0-3% 23.25% 24.5% 23.875% 19.0% 0-3% 19.0%
3-7% 221 227 224 4.1% 0-7% 29.6% 3-6% 132 138 135 5.2% 0-6% 28.8%
7-10% 87 91 89 18.1% 0-10% 34.2% 6-9% 43 47 45 13.4% 0-9% 36.4%
10-15% 28 33 30.5 20.1% 0-15% 43.3% 9-12% 28 33 30.5 21.9% 0-12% 41.7%
15-30% 9.25 10.25 9.75 29.7% 0-30% 66.5% 12-22% 14 16 15 30.9% 0-22% 54.8%
Source: Merrill Lynch; As of 9th Nov 2004; All 0-3% tranches quoted on Upfront + 500bps running basis.

Overcome the key limitations of Base correlations overcome the key limitations of tranche implied correlations:
tranche implied correlations • Base correlations can be used to price tranches with customized attachment
and detachment points of either standardized or bespoke underlying
portfolios.
• Base correlations are unique across the capital structure.
Base correlations can be computed with relative ease using the same model that is
used to derive tranche implied correlations. We also highlight in the following
sections how base correlations can be used to price non-standard tranches of the
underlying CDX or iTraxx portfolios.

1
For an in-depth discussion on the technicals affecting the skew see “Correlation
Skew” by Kakodkar/Galiani et al, published 14th July 2004.
2 Refer to important disclosures on page 7.
 Base Correlations – 10 November 2004

n Deriving Base Correlations of Standardized Tranches

Use bootstrapping method to Base correlations can be derived directly from the market spreads of the
derive base correlations from standardized tranches using standard bootstrapping techniques. The base
market spreads correlation of the first-loss 0-3% tranche is obviously the same as its tranche
implied correlation. However, the 0-7% tranche can be derived in the following
standard manner:
• Price the 0-7% tranche as a combination of the 0-3% and the 3-7% assuming
both tranches have a premium equal to that of the 3-7% market premium.
• Price of 3-7% tranche using the 3-7% market premium is zero.
• Price of 0-7% tranche is therefore equal to the price of 0-3% with the same
premium as the 3-7% market premium, using the 0-3% base correlation (also
the tranche implied correlation in this case). The price will be positive since
the 3-7% premium is smaller than the 0-3% market premium.
• Use this price for the 0-7% (after adjusting for the notionals) and the standard
Gaussian Copula model to imply the base correlation of the 0-7% tranche.
In the same manner, we can imply the 0-10% base correlation by pricing the 0-
10% tranche as a combination of the 0-7% tranche and the 7-10% assuming both
tranches have a premium equal to that of the 7-10% market premium. This
procedure is then repeated to derive 0-15% and 0-30% base correlations.
In order to illustrate the implementation of the above procedure, let us compute
the implied base correlation for the 0-7% tranche. As explained above, we start by
computing the price of the 0-3% tranche using the 3-7% tranche premium with the
0-3% base correlation. Given the levels in Table 1, this value, expressed in terms
of the 0-3% notional, equals 45.1%.
Scaling each intermediate step In the next step we compute the implied 0-7% correlation which matches the value
by the corresponding notional of the 0-7% tranche using the 3-7% premium. With this regard, we first need to re-
is crucial state the 0-7% tranche value by scaling the above number (the 0-3% mark-to-
market using the 3-7% premium) by the appropriate notional. This can be easily
done by multiplying the value of the 0-3% tranche (with the 3-7% premium) by
the ratio of the width of the 0-3% tranche (3%) and the width of the 0-7% tranche
(7%). The required calculations are in shown in Table 2.

Table 2: Notional Adjusted Mark-to-Market

0-7% Tranche Value (Notional Scaled)
0-3% tranche value using the 3-7% premium 45.1%
Width Ratio 3% / 7% = 0.43x
0-7% tranche value using the 3-7% premium 19.39%
0-7% Notional Scaled Value 0-7% base correlation 29.6%
Source: Merrill Lynch
45.1% x 0.43=19.85%
Once the notional-adjusted 0-7% mark-to-market has been computed, we can then
apply the standard Gaussian Copula Model to find the correlation that matches the
tranche value. Using a simple trial-and-error procedure in this example, we obtain
an implied base correlation of 29.6% for the 0-7% tranche.
We can then recursively compute the implied base correlation for the remaining
tranches. With regard to the 0-10% tranche, we first need to compute the value of
the 0-7% tranche using the 7-10% premium and the 0-7% base correlation. This
implies a 0-7% mark-to-market of 25.18%. Given the width ratio of 0.7x (0-7%
width / 0-10% width) we can then compute the notional adjusted 0-10% mark-to-
market which is equal to 17.6% (25.18% x 0.7). Given this value we can apply
again the Gaussian Copula model and solve for the implied 0-10% base which, as
reported in Table 1, equals 34.2%.

Refer to important disclosures on page 7. 3

Base Correlations – 10 November 2004

n Pricing Non-Standard Tranches of Standardized Indices

Correlation skew makes it One of the key problems with tranche implied correlations is the pricing of non-
difficult to price non-standard standard tranches such as a 5-8% tranche of the CDX index. For the standardized
tranches of the index tranche market in North America and Europe, we observe that tranche implied
correlations are skewed in the form of a correlation “smile” (see Chart 1 and Chart
2). Due to this skew, it is not very clear what correlation input should be used to
derive the price of a non-standard tranche.

Chart 1: Tranche Implied Correlations (CDX.NA.IG) Chart 2: Tranche Implied Correlations (iTraxx Europe)

35% 35%
30% 30%
25% 25%
20% 20%
15% 15%
10% 10%
5% 5%
0% 0%
0-3% 3-7% 7-10% 10-15% 15-30% 0-3% 3-6% 6-9% 9-12% 12-22%

Source: Merrill Lynch Source: Merrill Lynch

Chart 3: Base Correlations (CDX.NA.IG) Chart 4: Base Correlations (iTraxx Europe)

70% 60%
60% 50%
50%
40%
40%
30%
30%
20%
20%
10% 10%

0% 0%
0% 5% 10% 15% 20% 25% 30% 0% 3% 6% 9% 12% 15% 18% 21% 24%

Source: Merrill Lynch Source: Merrill Lynch

Pricing non-standard tranches Base correlations provide a more intuitive way to estimate this correlation
is more intuitive with base input and price the tranche. Consider the pricing of a 5-8% tranche of the
correlations CDX.NA.IG3. A long position in a 5-8% tranche is essentially comprised of two
first-loss tranches:
• long position in the 0-8%
• short position in the 0-5%
Each of these first-loss or base tranches needs a default correlation input to be
priced. This input can be derived from the base correlations of the standardized
tranches. Chart 3 and Chart 4 highlight that base correlations for the standardized
tranches increase for increasing detachment points. As a result, the base

4 Refer to important disclosures on page 7.

 Base Correlations – 10 November 2004

correlation of a 0-8% CDX tranche, for example, can be derived by interpolating

on the curve between 7% and 10%. Similarly we can price the 0-5% tranche by
interpolating on the curve between 3% and 7%. The 5-8% tranche can be priced in
the following four steps:
• Interpolate between 0-7% and 0-10% to get the base correlation of the 0-8%
tranche.
• Interpolate between 0-3% and 0-7% to get the base correlation of the 0-5%
tranche.
• Use the standard Gaussian Copula model to price both tranches.
• The 5-8% tranche price is the difference in 0-8% and 0-5% prices adjusted for
the notional of each tranche.
In order to price each of the base tranches, we need to rescale the notional of the
base tranches (0-5% and 0-8%) in terms of the “target” tranche (5-8%) notional,
yielding a width ratio of 1.67x and 2.67x for the two base tranches. Then, we
compute the default and the risky DV01 of the 5-8% tranche as the difference of
the legs of the two base tranches weighted by the corresponding width ratio.
Finally, the spread of the 5-8% tranche is found by dividing the 5-8% default leg
(5%) by its risky DV01 (4.22) which implies a breakeven spread of 110bps. Table
3 highlights this calculation in more detail.

Table 3: Non-Standard 5-8% Tranche Breakeven Spread Calculation

We use the two width ratios to
scale each intermediate step Tranche #1 Tranche #2 Off-The-Run
by the appropriate notional Tranche
Attachment Point 0% 0% 5%
Detachment Point 5% 8% 8%
Base Correlation 24.43% 31.2%
Notional Scaled 5-8% DV01
Width Ratios 5% / 3%=1.67x 8% / 3%=2.67x
(3.64x2.67)-(3.29x1.67)=4.22 Risky DV01 3.29 3.64 4.22
Default Leg 0.34 0.23 0.05
Breakeven Spread 1022 bps 625 bps 110 bps
5-8% Breakeven Spread =
Source: Assume a spread for the CDX.NA.IG 3 Index at 50.5 bps, 5y maturity and 40% average recovery.
Default Leg / Risky DV01
0.05 / 4.22=110 bps In this analysis, we have assumed that the underlying portfolio is unchanged, i.e.
the 125-name CDX.NA.IG. Portfolios with different underlying credits would
make use of the correlation data from the observed standardized tranche market to
generate a market-based estimate for the default correlation input. There are
different ways in which this can be done. We leave this discussion for the future.

Deltas: Base vs. Tranche Implied

By definition, pricing a tranche with either the tranche implied or base correlation
should produce the same result. In fact, both measures are calibrated from the
market in order to match traded prices.
Deltas depend on the chosen This argument does not necessarily apply as far as sensitivity parameters are
correlation measures concerned. These refer to individual deltas for the underlying CDS or leverage
with respect to the underlying index. Leverage is defined as the MTM of the
tranche divided by the MTM of the underlying index for a 1bp change in the
index.
As explained in an earlier publication2, the tranche leverage can also be derived
using the following steps:

2
Please refer to “Correlation Trading” by Kakodkar/Martin/Galiani dated 26
November 2003 for an explanation of the concept and the computation of the
tranche deltas and leverage.
Refer to important disclosures on page 7. 5
Base Correlations – 10 November 2004

1. Compute the individual credit spread delta for each underlying credit
2. Compute the tranche width as the difference of the tranche detachment and
attachment points.
3. Compute the average portfolio delta and divide this number by the tranche
width.
For the 3-7% levels in Table 1 and assuming a flat CDS term structure for all
credits (equal to the CDX.NA.IG3 mid spread of 50.5bps), we find an average
portfolio delta equal to 45.2%. The average delta divided by the tranche width
(4%) implies a tranche leverage of about 11.3x.
Within the base correlation framework, the computation of the 3-7% tranche
leverage is slightly different and can be described as follows:
1. Compute the average of the individual credit spread deltas for the 0-7%
tranche (68%) and for the 0-3% tranche (41%)
2. Compute the average individual delta for the 3-7% tranche (27%) as the
difference of the average individual delta for the 0-7% and the 0-3% tranche
3. Divide the 3-7% average individual delta by the tranche width to get the
leverage (6.75x).
Under the base correlation In Chart 5 and Chart 6, we compare the tranche leverage implied by using the two
framework tranche deltas tend correlation measures for the North American and European market. The difference
to be lower in leverage under the two methodologies (11.3x with the tranche implied
correlation versus 6.8x for the base correlation) demonstrates the impact of each
correlation measure in terms of risk management and hedging purposes. The
charts below also highlight that the difference is most relevant for the mezzanine
tranches (3-7% and 3-6%).

Chart 5: Tranche Leverage CDX.NA.IG3 Chart 6: Tranche Leverage iTraxx Series 2

20x 20x 19.1x 19.1x

15.3x
15.3x
Tranche Leverage

Tranche Leverage

15x 15x
11.3x 9.9x
10x 10x
6.8x 6.3x
4.1x 3.5x
5x 3.3x 5x
1.8x 2.5x 2.0x
1.3x 0.6x 1.6x 1.0x 0.8x
0.5x
0x 0x
0-3% 3-7% 7-10% 10-15% 15-30% 0-3% 3-6% 6-9% 9-12% 12-22%

Tranche Implied Correlation Base Correlation Tranche Implied Correlation Base Correlation

Source: Merrill Lynch; Assume a flat CDS spread at 50.5bps for all credits. Source: Merrill Lynch; Assume a flat CDS spread at 37bps for all credits.

6 Refer to important disclosures on page 7.

 Base Correlations – 10 November 2004

Important Disclosures

Copyright 2004 Merrill Lynch, Pierce, Fenner & Smith Incorporated (MLPF&S). All rights reserved. Any unauthorized use or disclosure is prohibited. This report has been
prepared and issued by MLPF&S and/or one of its affiliates and has been approved for publication in the United Kingdom by Merrill Lynch Pierce, Fenner & Smith Limited,
which is authorized and regulated by the Financial Services Authority; has been considered and distributed in Australia by Merrill Lynch Equities (Australia) Limited (ABN 65
006 276 795), licensed under the Australian Corporations Act, AFSL No 235132; has been considered and distributed in Japan by Merrill Lynch Japan Securities Co, Ltd, a
registered securities dealer under the Securities and Exchange Law in Japan; is distributed in Hong Kong by Merrill Lynch (Asia Pacific) Ltd, which is regulated by the Hong
Kong SFC; and is distributed in Singapore by Merrill Lynch International Bank Ltd (Merchant Bank) and Merrill Lynch (Singapore) Pte Ltd (Company Registration No.
198602883D), which are regulated by the Monetary Authority of Singapore. The information herein was obtained from various sources; we do not guarantee its accuracy or
completeness.
Neither the information nor any opinion expressed constitutes an offer, or an invitation to make an offer, to buy or sell any securities or any options, futures or other
derivatives related to such securities ("related investments"). Officers of MLPF&S or one of its affiliates may have a financial interest in securities of the issuer(s) or in related
investments.
This research report is prepared for general circulation and is circulated for general information only. It does not have regard to the specific investment objectives,
financial situation and the particular needs of any specific person who may receive this report. Investors should seek financial advice regarding the appropriateness of
investing in any securities or investment strategies discussed or recommended in this report and should understand that statements regarding future prospects may not be
realized. Investors should note that income from such securities, if any, may fluctuate and that each security’s price or value may rise or fall. Accordingly, investors may
receive back less than originally invested. Past performance is not necessarily a guide to future performance.
The bonds of the company are traded over-the-counter. Retail sales and/or distribution of this report may be made only in states where these securities are exempt from
registration or have been qualified for sale. MLPF&S usually makes a market in the bonds of this company.
Foreign currency rates of exchange may adversely affect the value, price or income of any security or related investment mentioned in this report. In addition, investors in
securities such as ADRs, whose values are influenced by the currency of the underlying security, effectively assume currency risk.

Refer to important disclosures on page 7. 7