Risk Metrics Hedging CDT

Hedging credit index tranches
Investigating versions of the standard model

Christopher C. Finger
chris.finger@riskmetrics.com
Risk Management
Subtle company introduction
www.riskmetrics.com Risk Management 22

Motivation
A standard model for credit index tranches exists.

It is commonly acknowledged that the common model is
flawed.
Most of the focus is on the static flaw: the failure to calibrate
to all tranches on a single day with a single model parameter.
But these are liquid derivatives. Models are not used for
absolute pricing, but for relative value and hedging.
We will focus on the dynamic flaws of the model.

Outline
1 Standard credit derivative products
2 Standard models, conventions and abuses
3 Data and calibration
4 Testing hedging strategies
5 Conclusions

Standard products
Single-name credit default swaps

Contract written on a set of reference obligations issued by one
firm
Protection seller compensates for losses (par less recovery) in
the event of a default.
Protection buyer pays a periodic premium (spread) on the
notional amount being protected.
Quoting is on fair spread, that is, spread that makes a contract
have zero upfront value at inception.

Standard products
Credit default swap indices (CDX, iTraxx)

Contract is essentially a portfolio of (125, for our purposes)
equally weighted CDS on a standard basket of firms.
Protection seller compensates for losses (par less recovery) in
the event of a default.
Protection buyer pays a periodic premium (spread) on the
remaining notional amount being protected.
New contracts (series) are introduced every six months.
Standardization of premium, basket, maturity has created
significant liquidity.
Quoting is on fair spread, with somewhat of a twist.
Pricing depends only on the prices of the portfolio names
. . . almost.

Standard products
Tranches on CDX
Protection seller compensates for losses on the index in excess
of one level (the attachment point) and up to a second level
(the detachment point).
For example, on the 3-7% tranche of the CDX, protection seller
pays losses over 3% (attachment) and up to 7% (detachment).
Protection buyer pays an upfront amount (for most junior
tranches) plus a periodic premium on the remaining amount
being protected.
Standardization of attachment/detachment, indices, maturity.
Not strictly a derivative on the index, in that payoff does not
reference the index price
Pricing depends on the distribution of losses on the index, not
just the expectation.
Also, options on CDX, but we will not consider these.

CDX history
200
160
Index spread (bp)
120
80
40
0
Mar05 Sep05 Mar06 Sep06 Mar07 Sep07 Mar08

CDX tranche history
0−3% (LHS)
3−7% (RHS)
7−10% (RHS)
500
60
50 400
Tranche upfront price (%)
Tranche fair spread (bp)

40
300
30
200
20
100
10
0 0
Mar05 Sep05 Mar06 Sep06 Mar07 Sep07

What do we want from a model?
Fit market prices, but why?

Extrapolate, i.e. price more complex, but similar, structures
Non-standard attachment points
Customized baskets
Hedge risk due to underlying
Dealers provide liquidity in tranches, but want to control
exposure to underlyings.
Speculators want to make relative bets on tranches without a
view on underlyings.
Risk management
Aggregate credit exposures across many product types.
Recognize risk that is truly idiosyncratic.
For anything other than extrapolation, we care about how prices
evolve in time, so we should look at the dynamics.
Standard pricing models—basic stuff
Price for a tranche is the difference of

Expectation of discounted future premium payments, and
Expectation of discounted future losses.
Boils down to the distribution of the loss process on the index
portfolio, in particular things like
E min{(d − a), max{0, Lt − a}}.
Suffices to specify the joint distribution of times to default Ti

for all names in the basket.
CDS (or CDX) quotes imply the marginal distributions for
time to default for individual names Fi (t) = P{Ti < t}.

Standard pricing models—specific stuff
Dependence structure is a Gaussian copula:

Let Zi be correlated standard Gaussian random variables.
Default times are given by Ti = Fi−1 (Φ(Zi )).
Correlation structure is pairwise constant . . .
√ p
Zi = ρ Z + 1 − ρ εi .
For a single period, just count the number of Zi that fall

below the default threshold αi = Φ−1 (pi ).

Criticisms of the standard model
Does not fit all market tranche prices on a given day.

No dynamics, so no natural hedging strategy.
Link to any observable correlation is tenuous. At best, model
is “inspired” by Merton framework, so correlation is on
equities.

Model flavors
Most numerical techniques rely on integrating over Z , given which

all defaults are conditionally independent.
Granular model—use full information on underlying spreads,
and model full discrete loss distribution.
Homogeneity assumption—assume all names in the portfolio
have the same spread; use index level.
Fine grained limit—continuous distribution, easy integrals
Large pool model—combine homogeneity and fine grained
assumptions.
Also the question of whether to use full spread term structures
or a single point

Correlation conventions
Start to introduce model abuses, ala the B-S volatility smile.
Compound correlations
Price each individual tranche with a distinct correlation.
Not all tranches are monotonic in correlation.
Trouble calibrating mezzanine tranches, especially in 2005
Base correlations
Decompose each tranche into “base” (i.e. 0-x%) tranches.
Bootstrap to calibrate all tranches.
Base tranches monotonic, but calibration not guaranteed.

Correlation conventions
Constant moneyness (ATM) correlations

Some movements in implied correlation are due to changing
“moneyness” as the index changes.
Examine correlations associated with a detachment point equal
to implied index expected loss.
If base correlations are “sticky strike”, then ATM correlations
are closer to “sticky delta”.

CDX correlation structure
100
Mar05
Mar06
90 Mar07
Mar08
80
70
Base correlation (%)
60
50
40
30
20
10
0
0 5 10 15 20 25 30
Detachment point (%)

CDX base correlations, granular model
100
80
60
40
20
0
Sep05 Mar06 Sep06 Mar07 Sep07

CDX base correlations, large pool model
100
80
60
40
20
0

CDX base correlations, large pool model, with DJX
option-implied correlation
100
80
60
40
20
0

CDX Series 4-7, GR model
70
60
50
40
30
20
10
0
4 6 8 10 12 14 16 18
Detachment plus index EL (%)

CDX Series 4-7, LP model
70
60
50
40
30
20
10
0
4 6 8 10 12 14 16 18

CDX Series 8-9, GR model
70
60
50
40
30
20
10
0
4 6 8 10 12 14 16 18

CDX Series 8-9, LP model
70
60
50
40
30
20
10
0
4 6 8 10 12 14 16 18

Time series properties
Examine the correlation between various implied correlations

and the index.
We would like this to be low. Why?
For risk, we have identified idiosyncratic risk correctly.
For hedging, we have captured what we are able to from the
underlying.
Start with statistics on daily changes.

CDX 0-3%, large pool model, base correlations
60
Flat curve
Full curve
50
40
Correlation to index (%)
30
20
10
0
4 5 6 7 8 9 All
Series

CDX 0-3%, base correlations
60
Large pool
Granular
50
40
30
20
10
−10
−20
4 5 6 7 8 9 All
Series

CDX 0-3%, large pool model
60
Base Corr
ATM Corr
40
20
−20
−40
−60
4 5 6 7 8 9 All
Series

CDX 0-3%
60
LP, base
Gran, ATM
50
40
30
20
10
−10
−20
−30
4 5 6 7 8 9 All
Series

CDX 3-7%
60
LP, base
Gran, ATM
50
40
30
20
10
−10
−20
4 5 6 7 8 9 All
Series

Time series properties
Now look at statistics on weekly changes.

CDX 0-3%, large pool model, base correlations
80
Flat curve
Full curve
60
40
20
−20
−40
−60
4 5 6 7 8 9 All
Series

CDX 0-3%, base correlations
80
Large pool
Granular
60
40
20
−20
−40
−60
4 5 6 7 8 9 All
Series

CDX 0-3%, large pool model
80
Base Corr
ATM Corr
60
40
20
−20
−40
−60
−80
4 5 6 7 8 9 All
Series

CDX 0-3%
80
LP, base
Gran, ATM
60
40
20
−20
−40
−60
−80
4 5 6 7 8 9 All
Series

CDX 3-7%
80
LP, base
Gran, ATM
60
40
20
−20
−40
−60
4 5 6 7 8 9 All
Series

Time series conclusions
Looks like granular model produces “more idiosyncratic”

implied correlations.
The ATM approach looks appears to improve things, but
seems to overcorrect in the most volatile periods.
High correlations overall suggest index moves may have some
predictive power for implied correlations.
Differences are much more pronounced at a one-day horizon
than a one-week horizon.

Hedging experiment
On day zero, we have all prices (spreads and tranches) plus

history.
At a future date, we are told how the underlying spreads
(index or single-name CDS) have moved, and are asked to
guess what the new tranche prices should be.
Compare the predicted tranche price moves to the actual
ones, over time and across different modeling approaches.
Another approach would be to look at delta hedge
performance, but this mixes in the error from linearization.

Hedging experiment candidates
Models
Regression—fit linear relationship between tranche price
changes and index changes, using prior six months of data.
Large pool model—calibrate based on current index levels.
Granular model—calibrate based on current single-name CDS
levels.
Correlation approaches
Base—use most recent calibrated correlation.
ATM—move along the most recent correlation structure
according to change in the index-implied expected loss.
Regression—fit linear relationship between base correlation
changes and index changes, using prior six months of data.

CDX 0-3%, Regression
80 8
70
60 4
50
Prediction error (%)

Tranche price (%)
40 0
30
20 −4
10
0 −8

CDX 0-3%, LP model, base correlations
80 8
70
60 4
50
Prediction error (%)

Tranche price (%)
40 0
30
20 −4
10
0 −8

CDX 0-3%, Regression
8
Ser 4
Ser 5
Ser 6
6
Ser 7
Ser 8
Ser 9
4
Predicted change (%)
−2
−4
−6
−8
−8 −6 −4 −2 0 2 4 6 8
Tranche price change (%)

CDX 0-3%, LP model, base correlations
8
Ser 4
Ser 5
Ser 6
6
Ser 7
Ser 8
Ser 9
4
Predicted change (%)
−2
−4
−6
−8
−8 −6 −4 −2 0 2 4 6 8
Tranche price change (%)

Statistics on hedging approaches
Examine standard deviation of tranche forecast error.

Daily horizon
Bars are for correlation approaches: Base (blue), ATM

(green), Regression (red).
Curve is for simple regression model.

CDX 0-3%, LP model
0.03
0.025
0.02
STD of prediction error
0.015
0.01
0.005
0
4 5 6 7 8 9 All
Series

CDX 0-3%, GR model
0.03
0.025
0.02
0.015
0.01
0.005
0
4 5 6 7 8 9 All
Series

CDX 3-7%, LP model
−3
x 10
3.5
2.5
1.5
0.5
0
4 5 6 7 8 9 All
Series
CDX 3-7%, GR model
−3
x 10
3.5
2.5
1.5
0.5
0
4 5 6 7 8 9 All
Series
CDX 7-10%, LP model
−3
x 10
2.5
2
1.5
0.5
0
4 5 6 7 8 9 All
Series
CDX 7-10%, GR model
−3
x 10
2.5
2
1.5
0.5
0
4 5 6 7 8 9 All
Series
Examine correlation of tranche forecast error with actual

tranche move.
Daily horizon


CDX 0-3%, LP model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

CDX 0-3%, GR model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

CDX 3-7%, LP model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

CDX 3-7%, GR model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

CDX 7-10%, LP model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

CDX 7-10%, GR model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

Examine standard deviation of tranche forecast error.

Weekly horizon


CDX 0-3%, LP model
0.045
0.04
0.035
0.03
0.025
0.02
0.015
0.01
0.005
0
4 5 6 7 8 9 All
Series

CDX 0-3%, GR model
0.045
0.04
0.035
0.03
0.025
0.02
0.015
0.01
0.005
0
4 5 6 7 8 9 All
Series

CDX 3-7%, LP model
−3
x 10
5
4.5
3.5
2.5
1.5
0.5
0
4 5 6 7 8 9 All
Series
CDX 3-7%, GR model
−3
x 10
5
4.5
3.5
2.5
1.5
0.5
0
4 5 6 7 8 9 All
Series
CDX 7-10%, LP model
−3
x 10
3.5
2.5
1.5
0.5
0
4 5 6 7 8 9 All
Series
CDX 7-10%, GR model
−3
x 10
3.5
2.5
1.5
0.5
0
4 5 6 7 8 9 All
Series
Examine correlation of tranche forecast error with actual

tranche move.
Weekly horizon


CDX 0-3%, LP model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

CDX 0-3%, GR model
−0.1
−0.2
−0.3
−0.4
Corr(Act,Err)
−0.5
−0.6
−0.7
−0.8
−0.9
−1
4 5 6 7 8 9 All
Series

CDX 3-7%, LP model
0.1
−0.1
−0.2
−0.3
Corr(Act,Err)
−0.4
−0.5
−0.6
−0.7
−0.8
−0.9
4 5 6 7 8 9 All
Series

CDX 3-7%, GR model
0.1
−0.1
−0.2
−0.3
Corr(Act,Err)
−0.4
−0.5
−0.6
−0.7
−0.8
−0.9
4 5 6 7 8 9 All
Series

CDX 7-10%, LP model
0.5
0
Corr(Act,Err)
−0.5
−1
4 5 6 7 8 9 All
Series

CDX 7-10%, GR model
0.5
0
Corr(Act,Err)
−0.5
−1
4 5 6 7 8 9 All
Series

Conclusions
Overall, hedging errors are surprisingly large.

For the equity tranche, a simple regression works quite well,
with its underhedging problems reduced at a slightly longer
horizon.
There is a benefit to using the granular model, but is it worth
the cost?
For more senior tranches, the ATM approach appears to
capture some of the link to credit spreads, but does not
markedly reduce the hedging error.
Any candidate for a new standard model should be able to do

better in this experiment.

Further investigations
Can we use these results to learn more about how our model
might be misspecified?
Richer correlation structure?
Different copula?
Better dynamics?
How much worse (or better) are compound correlations?
Can the regression be improved by including the HiVol index?
Can the LP model be improved by adding a simple correction
for heterogeneity?
Does any of this change at or close to defaults?

Further reading
Couderc, F. (2007) Measuring Risk on Credit Indices: On the

Use of the Basis. RiskMetrics Journal.
Finger, C. (2004) Issues in the Pricing of Synthetic CDOs.
RiskMetrics Journal.
Finger, C. (2005) Eating our Own Cooking. RiskMetrics
Research Monthly. June.
All are available at www.riskmetrics.com

Risk Metrics Hedging CDT

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Risk Metrics Hedging CDT

Uploaded by

Copyright:

Available Formats

Hedging credit index tranches

Investigating versions of the standard model

www.riskmetrics.com Risk Management 22

A standard model for credit index tranches exists.

www.riskmetrics.com Risk Management 32

1 Standard credit derivative products

2 Standard models, conventions and abuses

3 Data and calibration

4 Testing hedging strategies

www.riskmetrics.com Risk Management 42

Single-name credit default swaps

www.riskmetrics.com Risk Management 52

Credit default swap indices (CDX, iTraxx)

www.riskmetrics.com Risk Management 62

Also, options on CDX, but we will not consider these.

www.riskmetrics.com Risk Management 82

Tranche fair spread (bp)

www.riskmetrics.com Risk Management 92

Fit market prices, but why?

Price for a tranche is the difference of

E min{(d − a), max{0, Lt − a}}.

Suffices to specify the joint distribution of times to default Ti

www.riskmetrics.com Risk Management 112

Dependence structure is a Gaussian copula:

For a single period, just count the number of Zi that fall

www.riskmetrics.com Risk Management 122

Does not fit all market tranche prices on a given day.

www.riskmetrics.com Risk Management 132

Most numerical techniques rely on integrating over Z , given which

www.riskmetrics.com Risk Management 142

Start to introduce model abuses, ala the B-S volatility smile.

www.riskmetrics.com Risk Management 152

Constant moneyness (ATM) correlations

www.riskmetrics.com Risk Management 162

www.riskmetrics.com Risk Management 172

www.riskmetrics.com Risk Management 182

www.riskmetrics.com Risk Management 192

www.riskmetrics.com Risk Management 202

www.riskmetrics.com Risk Management 212

www.riskmetrics.com Risk Management 222

www.riskmetrics.com Risk Management 232

www.riskmetrics.com Risk Management 242

Examine the correlation between various implied correlations

Start with statistics on daily changes.

www.riskmetrics.com Risk Management 252

www.riskmetrics.com Risk Management 262

www.riskmetrics.com Risk Management 272

www.riskmetrics.com Risk Management 282

www.riskmetrics.com Risk Management 292

www.riskmetrics.com Risk Management 302

Now look at statistics on weekly changes.

www.riskmetrics.com Risk Management 312

www.riskmetrics.com Risk Management 322

www.riskmetrics.com Risk Management 332

www.riskmetrics.com Risk Management 342

www.riskmetrics.com Risk Management 352

www.riskmetrics.com Risk Management 362

Looks like granular model produces “more idiosyncratic”

www.riskmetrics.com Risk Management 372

On day zero, we have all prices (spreads and tranches) plus

www.riskmetrics.com Risk Management 382

www.riskmetrics.com Risk Management 392

Prediction error (%)

www.riskmetrics.com Risk Management 402