Terry Benzschawel

Modeling Corporate
Terry Benzschawel
Loan Prepayments
Director,
Quantitative Credit Trading Strategy
Citi Markets & Banking
and LCDS
(212) 816-8071 Cancellability
terry.l.benzschawel@citi.com
Policy Forum
2nd Annual Risk Management Conference
Risk Management Institute
National University of Singapore
June 28, 2008
Fixed Income Strategy &

Analysis © Citi Markets and Banking Inc., 2008. All rights reserved.
Outline
The Loan Prepayment Option
The Need for Loan Valuation
Understanding Loan Prices
Modeling Loan Price Dynamics
Single Name Loan Pricing and OAS
Loan Credit Default Swaps (LCDS)
The Cash Loan / LCDS / CDS Basis
Valuing LCDS Cancellability
Current Loan Market Conditions
Applications
2
Is the Loan Prepayment Option Quantifiable?
In a recap of a popular trade where the investor buys
the cash loan and sells a credit default swap (a basis
trade) on Ford in 2006, Standard & Poor’s states that:
“… most par loans remain ill-suited

for this type of trade. The
prepayment options [in the loans]
introduce unquantifiable risk in that
the cash position could disappear
overnight.”
From LCD Leveraged Lending Review: Fourth Quarter 2006,

Standard & Poor’s, January 2007.
3
Bond Pricing – Credit Spread
• The ability to separate out the “risky” from the “riskless” portion of
the yield on a corporate bond provided a tremendous stimulus for
corporate bond trading
Yield Curves for US Treasuries and for
– Corporate bonds are conveniently Single-A Corporate Bonds
quoted on a spread-to-Treasuries 7
or spread-to-LIBOR basis
6
– Often, bonds have embedded call Credit Spread
(%)
or put options and techniques 5
Yield (%)
Yield
have been developed to compute
4
option-adjusted spreads that US Treasury Yields
accounts for those option values 3
Corporate Bond Yields
To compute the value of a bond with 2
maturity, T  
0 5 10 15 20 25 30 35
Years to Maturity
Years to Maturity
 2T c  c + 100

PV = ∑ 2 + 2
for US Treasuries:  n =1 n  2T
Where PV is the price of bond with coupons
 1 + y   1 + y 
2 2
  2    2 (ct), rt is the US Treasury Yield and st is the

yield spread of the credit curve to US
 
 2T c  c + 100 Treasuries. Spread is often Basis Points

PV = ∑ 2 + 2 where 1bp is 1/100 of 1%
for Corporates:  n =1 n  2T
 1 +  y + s  2
 y + s  2
  1 +  Credit
  2    2 
Spread 4
Credit Spreads for Loans
• The ability to separate out the “risky” from the “riskless” portion of
the yield on a corporate bond provided a tremendous stimulus for
corporate bond trading.
• The analogous spread equation for loans, which are floating-rate
instruments that typically pay coupons quarterly is:
4n
ct
F
V =∑ 4 +
+ s + s
4 4n
,
t =1  f  f
1 + t t −1  1 + 4 n 4 n −1 
 4   4 
Credit
Spread
where V is the price of loan with coupons (ct), ft,t-1 is the quarterly
forward LIBOR rate between time t and t-1, F is the principal of the
loan, and s is the credit spread over LIBOR.
• PROBLEM: The spread calculation above does not account for the
obligor’s option to prepay the loan prior to maturity
5
Loan Pricing – The Prepayment Option
• Loans are typically quoted on price, not spread
– Most loans allow borrowers to prepay at par or draw on credit lines at
their discretion
– This introduces uncertainty about the duration of the loan
– The corporate loan market is bigger than the bond market, but no
generally accepted procedure for valuing the loan prepayment option has
been proposed
Valuing the loan prepayment option is difficult because:

• Loan prepayment depends • Building a loan OAS calculator
many factors is not easy
– The credit state of the borrower – Theoretical and technical
– Prepayment penalty problems need to be solved
– Prevailing rates of financing – Need an accurate model of
– Other firm- and loan-specific credit transitions and evolution
factors of loan financing rates
– Financial engineering talent
– Loan indicative data
– Historical data on prepayments
for calibration 6
Outline
Applications
7
Why Mark Loans to Market?
• Better Origination and Valuation
– Understand value of the loan and the cost of complex terms and
embedded options
– Assess the profitability of customer relationships
• Better Portfolio Management
– Measure the performance of the loan portfolio
– Identify hedging, investment and arbitrage opportunities
– Determine the “correct” price to buy, sell, hedge or securitize a loan
– Reduce losses for receiving prepayment at par
• Regulatory and Investor Demand
– Provide greater balance sheet transparency
– Reduce probability of loan portfolio generating earnings surprises
• Industry Best Practice – FASB 199 “Fair Value Option”
– A growing number of firms are marking their loans to market
– Regulatory environment is favoring mark-to-market (Basel II)
• Reduce Earnings Volatility
– Match loan value changes with hedge mark-to-market 8
Why Calculate Loan OAS?
• Improved Relative Value Assessment
– Competitive advantage to firms accurately pricing loans
– Exploit market inefficiencies
– Optimize risk/reward characteristics of CLOs and portfolios
• Ability to Price Illiquid Loans
– Investors can not get reasonable estimates of NAVs on CLOs
– Many loans do not have market prices or are relatively illiquid
• Accurate Portfolio Hedging
– Credit exposure changes with factors that affect prepayment
– Assessment of portfolio sensitivities to factors that affect
prepayments
• Evaluate the Cost/Benefits of Credit Protection via LCDS
– Loan credit default swaps (LCDS) have begun trading in the US
without the prepay option
– Look for value in cash loans versus LCDS
– Trade cash loans versus bonds versus CDS versus LCDS
9
Outline
Applications
10
Loan Price Research
Moody’s/KMV Citi
• Period from 2002-2007 • Period from Jan-03 to Jul-07
• Data source: Loan Pricing • Data source: LoanX (MarkIt
Corporation (LPC) Partners)
• 4,317 loans; 2,988 term • 4,519,829 Price Quotes
loans / 1,329 revolvers • 12,733 loans;
• 2,232 public firm loans / • Only leveraged loans (rated
2,805 private below BBB-)
• Roughly 40% of sample
• >90% of loans USD
denominated.
have pricing grids \
• Term loans maturity ~4 years
• >60,000 prices / No Revolvers
• >90% of loans USD
denominated.
• Term loans maturity ~4
years / Revolvers ~3 years
11
Number of Price Quotes is Increasing, but
LPC
A large number of
loan price quotes are
reported daily
Source: Reuters and Moody’s/KMV
. . . and
7,000
Number of Loans Per Day 6,000
the number of daily LoanX
5,000
price quotes has
4,000
increased dramatically 3,000
in recent years 2,000
1,000
Source: MarkIt Partners and Citi
0
3
6
03
04
05
06
07
l -0
l -0
l -0
l -0
n-
n-
n-
n-
n-
12
Ju
Ju
Ju
Ju
Ja
Ja
Ja
Ja
Ja
Most Prices Come from Thin Quotes
The majority of bid and 40

ask prices are quoted • 65% have 2 or fewer quotes
• 77% have 3 or fewer quotes
by fewer than three 30 • 84% have 4 or fewer quotes
dealers.
Percent
20
In the CDS world, these

10
prices would be
considered ineligible for
0
inclusion in databases
like Mark-It Partners. Number of Quotes
Source: Moody’s/KMV
13
Most Prices Do Not Change Monthly
Moody’s/KMV finds that Citi finds that for over ½ of the
monthly price changes for loan-months in its sample,
over 50% of loans in their LPC there are no price changes.
sample are between +/-0.1 Less than 200 loans have >10
(less than one tick) price changes per month
70,000 60%
60,000
Percentage of Loan-Months
LPC Prices 50%
Number of Loan-Months
LoanX Prices
50,000
40%
Percent
40,000
30%
30,000
20%
20,000
10,000 10%
0 0%
2
2
10
0
-1
-1
-1
-1
-2
-2
1-
3-
5-
7-
=
9-
11
13
15
17
19
21
Number of Price Changes During Month
Source: Reuters and Moody’s/KMV

14
Estimating Loan Price Volatility
Number of Loans per Month having
Because relatively few loans >10 Price Changes During that Month
are liquid, estimating “true” 150
Number of Loans Per Month

loan price volatility is 125 LoanX Prices
LoanX Prices
complicated. That is why we
100
estimate loan price dynamics
75
and volatility using only
50
liquidly traded loans.
25
0
Also, since we find that
6
3
7
l- 0
l- 0
l- 0
l- 0
-0
-0
-0
-0
-0
n
n
liquid loans tend to begin to
Ju
Ju
Ju
Ju
Ja
Ja
Ja
Ja
Ja
225
Number of Liquid Loans

repay after about 90 days 200
(our sample in lower plot 175
starts to decrease), we 150
should not make inferences 125
about all loan prices beyond 100
90 days. 75
50
Source: MarkIt Partners and Citi 0 30 60 90 120 150 180 210 240
Number of Consecutive Days of Prices 15
Loan Price Volatility
Loan price changes Daily Percentile Price Changes for Liquid
are not a “random Loans from LoanX from 2003-2007
4
walk” 95 Pctl
3
Prices are “mean 2
Price Change
90 Pctl
reverting” 1
75 Pctl
0 50 Pctl
The envelope of 90- -1
25 Pctl
day price changes 10 Pctl
-2
is roughly between 5 Pctl
+3 points and -2 -3
0 10 20 30 40 50 60 70 80 90
points Number of Days
The envelope of price changes for all percentiles is constant after

about 45 days.
16
Price Change Depends on Level
Price Changes for Liquid Loans Price Changes for Liquid Loans
with Prices >=101 with Prices <=99
1 30
0.5 95 Pctl 25
90 Pctl 95 Pctl
20 90 Pctl
0
75 Pctl
Price Change
15
Price Change
-0.5 50 Pctl 10
-1 25 Pctl 75 Pctl
5
10 Pctl 50 Pctl
-1.5 0 25 Pctl
-5 10 Pctl
-2 5 Pctl 5 Pctl
-10
-2.5
-15
0 10 20 30 40 50 60 70 80 90
Number of Days 0 10 20 30 40 50 60 70 80 90
Numbe r of Days
There is a definite “pull-to-par”; prices above 100 trend downward

and those below trend upward
Note: These data come from relatively small samples (33 for Px>=101 and 22 for Px<=99)
Source: MarkIt Partners and Citi 17

Outline
Applications
18
Modeling Loan Volatility
• Input: Historical average price time-series for HY loans
• Model the prices as a non-linear mean-reverting process
γ
dPt = κ ( µ − Pt ) dt + σ Pt dW t
where Pt is price as a fraction of par (100)
• We set µ =1 to force price reversion to par
Loan Rating BB B CCC

Average Half life
1 2 1.8
(years)
• Negative γ so that Volatility when Price

• We cap the volatility when price goes too low
19
Simulating Loan Paths
• Generate loan index paths for each rating category using
the parametric equations:
dPt = κ BB (1 − Pt ) dt + σ BB ( Pt ) γ BB dW t
BB BB BB
Correlated
dPt = κ B (1 − Pt ) dt + σ B ( Pt ) γ B dW t
B B B
CCC γ CCC
= κ CCC (1 − Pt ) dt + σ CCC ( Pt
CCC CCC
dPt ) dW t
• Simulate loan paths in the future as:
dp t ,i = β i ( ∆ Pt ) + σ iξ t
INDEX
κ
• Model Parameters:
σ
20
Simulating Correlated Jumps to Default
ρ = 0.2
• Simulate default times for each of the loans
r , ε ~ N(0, 1)
• Generate dr = ρi
rg + 1 − ρ ε i for each loan
g i
• Use historical or Citi’s HPD model default probabilities

to sample default time
Cumulative Default Probabilities for
Cumulative Normal Distribution Leveraged Loans by Rating
Historical Default Probability for Leveraged Loans
1 0.5
0.9 0.45
0.8 0.4
CCC
0.7 0.35
Default Probability
0.6 0.3
0.5 0.25 BB
0.4 0.2
0.3 0.15 B
0.2 0.1
0.1 0.05
0 0
1 2 3 4 5 6 7
Years
Default time
rd i 21
Historical Average Loan Prices
From a sample of Average Loan Prices for Leveraged Loans
roughly 200 loans by Rating Category
at any given time,
we calculated 100
average spreads
by rating category Average Loan Price
90
Each of the 200

80
loans came from
the set that had at
least 10 price 70
changes/month as
reported by MarkIt 60
partners
BB
50
There are clearly B
CCC
price “jumps” 1996 1998 2000 2002 2004 2006 2008
Date
Source: MarkIt Partners and Citi 22
Individual Loan Price Simulation
• Loan price diffusion process:
= β ∆Pt + σ Rating dWt (i )
(i ) (i ) INDEX
dPt
• β ~ uniform [0.8,1.2]
• dW is a Weiner process
• σ Rating = varHistorical − varIndex
BB B CCC
σHistorical 1.5% 2.3% 5.6%
23
Putting it All Together
Average Price Timeseries for HY leveraged loans
100
90
Average Price
80
70
60
Loan Index Paths (BB, B and CCC)
50 BB
B
CCC
Jan-97 Jan-99 Jan-01 Jan-03 Jan-05 Jun-07
Time
Historical Loan Prices
MV CLO
Historical Default Probability for Leveraged Loans

1 0.5
0.9 0.45
0.8 0.4
0.7 0.35
Default Probability
0.6 0.3
0.5 0.25
0.4 0.2
0.3 0.15
0.2 0.1
0.1 0.05
0 0
1 2 3 4 5 6 7
Years
rdi
Individual Loan paths
Jump to Default Process 24
Outline
Applications
25
Credit Spreads and Default Rates
The spread on a risky cash flow, s, can be expressed in terms of its risk
neutral default probability, CQDP, its duration, T, and the expected loss given
default, LGD, as:
s = − ln[1 − (CQDPT ∗ LGD )]

1
T
It can also be shown that the Capital Asset Pricing Theory (CAPM) can be
used to link the risk neutral default probability, CQDP, to the physical (actual)
default probability, CDP, as:
(
CQDPT = N N −1 (CDPT ) + λρ T )
We can also go the reverse direction. That is, if we know CDP, the credit’s
beta, and the market price of risk, λ, we can calculate the risk-neutral default
probability, CQDP.
See, Bohn, J. A Survey of Contingent-Claims Approaches to Risky Debt Valuation, The Journal of Risk
Finance, Spring 2000, 53-70 and the Appendix of this presentation for a derivation of this relationship 26
Calculating Spread from Default
From Structural Calculate beta for
Model each asset from
market data
(
CQDPT = N N −1 (CDPT ) + λβ T )
Calculate Estimate
market-wide
FIRST STEP: Need to
risk premium
determine duration
for given asset
s = − ln[1 − (CQDPT ∗ LGD )]

1
T
LAST STEP:
Calculate model From Moody’s
based fair spread LossCalc
27
Physical → Risk Neutral Default
Physical Default Probability Risk Neutral Default Probability
Asset Value
Asset Value
Distribution of Asset Value at T Distributions of Asset Value at T
Expected Expected
Asset Drift Asset Drift
of µ of µ
Ao Ao
Promised Payments Drift of r Promised Payments
“Actual” “Actual”
Probability Probability “Risk-Adjusted” Probability of
of Default of Default Default
τ Time τ Time
The calculation of the “physical” or actual probability of default is outside the contingent
claims analysis (CCA) or “Merton” Model, but it can be combined with an equilibrium model
of underlying expected returns to produce estimates of expected returns.
28
For Loans, We Don’t Know T
We would like to be able to apply the spread formulas to loans as we know
CQDP from the CDS market and we can estimate LGD.
Unfortunately, because of the prepayment option, we don’t know T, the
expected time that we will receive the cash flow.
s = − ln[1 − (CQDPT ∗ LGD )]

1
T
(
CQDPT = N N −1 (CDPT ) + λβ T )
Having a loan OAS calculator will allow more precise estimation of
the durations of firms’ loans and allow us both to calculate an OAS
and better estimate the fair spread on the loan.
29
Inputs to the Loan Pricing Engine
We use a backward induction method on risk-adjusted ratings transition
probabilities to price loans. The pricer requires inputs from diverse sources.
30
Outline of Loan Price Calculation
Convert PD into Compute monthly

numerical rating transition matrix Transform physical
bucket (or use according to T matrix to
agency rating). historical PD risk Neutral matrix TQ
transitions
Calculate the price using backward Apply company

induction based Lattice approach. specific CDS data
onto TQ
Prepayment/default considered
Source: Citi
31
The Pricing Lattice
Example: Loan That Pays 105 at Maturity or Refinance; 75 if in Default
15 105
12 105
to
Default
Distance
9 105
DD1
6
105
3 105
Default0 45
75
Obligor’s 0 1 2 3 4 5
Initial Credit
State Time (Year)
Source: Moody’s/KMV and Citi

32
Credit Migration: Driven by PD Dynamics
On any given time step, there is some probability of the
obligor’s credit state being at any of the subsequent nodes
15 p=0.04 105
to Default
12 p=0.11 105
Actual (Physical)
9 p=0.19 Credit Transitions 105
DD1
from Structural
Distance
Model
6 p=0.45 105
3 p=0.17 105
Default
0 p=0.04 75
0 1 2 3 4 5
Time (Year)
33
Credit Migration: Physical to Risk Neutral PDs
Risk-neutral default probabilities derived from the bond-
referenced credit default swap (CDS) market are used to
adjust the physical ratings transitions to risk neutral ones
15 p=0.01 105
to Default
12 p=0.05 105
Risk Neutral Credit
9 p=0.14 Transitions from
Structural Model and 105
DD1
Distance
CDS Curve
6 p=0.38 105
3 p=0.23 105
Default
0 p=0.19 75
0 1 2 3 4 5
Time (Year)
34
The Lattice Gets More Complicated with Time
On a given time step, each non-default node gives rise to
transitions to all nodes at the next time step
15 105
to Default
12 105
Risk Neutral Credit
Transitions from
9 105
Structural Model and
DD1
CDS Curve
Distance
6 105
3 105
Default
0 75
0 1 2 3 4 5
Time (Year)
35
Prepayment Depends on Credit Migration
The economic implications regarding refinancing are
evaluated at each node.
If credit improves, prepayment is more likely; if credit deteriorates prepayment
is less likely and default more likely.
105
Distance to Default
105
105
105
Obligor’s 105
Credit State
75
36
Pricing Grids and Revolver Utilization
15
to Default
12
Pricing Grid: Reduced Fees

9
Improved credit
DD1
Distance
6
will result in
reduced 3
coupon Default0
0 1 2 3 4 5
15
Revolver DD1to Default 12

Utilization:
9
Deteriorating
Distance
credit will 6
Higher Utilization
increase usage 3
Default0
0 1 2 3 4 5
Time (Year)
37
Backward Induction
The economic value at each node is computed using
backward induction
15 105
• Start from the end of
DD1 to Default
lattice 12 105
• Determine cash flows 9 105
Distance
and loan value for end
6 104.90 105
of each period.
3 105
• Weight cash flow plus
continuation value (or Default0 45
75
prepayment value) by 0 1 2 3 4 5
risk-neutral transition Time (Year)
probabilities.
• Discount cash flow at each time-step by forward LIBOR rate to get
value for each node at beginning of current period.
• Continue until reaching initial state.
38
Backward Induction (cont)
15 105
105.47
12 105.42 105
Check each node
9 105.33 105
for continuation
DD1
value (if value <= 6 104.90 105
105.5, continue) 3 104.33 105
Default 75 45
75
0
0 1 2 3 4 5
15 Time (Year) 105.47 105
The node shown 12 105.42 105
at T=3 is 105.61; 105.33 105
9 105.61
DD1
greater than the

6 104.90 105
value of prepay.
Lender receives 3 104.33 105
105.5 Default
0
75 45
75
0 1 2 3 4 5
Time (Year)
39
Is the Model Accurate?
Distribution of Price Differences
12% more of public loans are valued

within 0.5 of quotes.
Credits
with Stale
Prices
Public Private
Public: About 65% of loans get valued within one dollar of the nearest quote.
Private: About 53% of loans get valued within one dollar of the nearest quote.
40
Capturing Loan Price Dynamics
Even large movements
41
Overcoming Stale Loan Quotes
Loan vs. CDS

Firm=Duke Energy Field Services, LLC
1000 100
CDS Spread (basis points)
Model
90
Price
Loan Price
LPC
Bid/Ask
100 80
CDS
70
10 60
1/1/2002 1/1/2003 1/1/2004 1/1/2005
Date
42
The Prepayment Option
To price the prepayment option, run the model with and
without the ability to prepay. The difference in price is the
option value.
Percent
Leveraged Loans
Typically Have
High Option Values
Option Value
43
Valuing the Loan Prepayment Option
Compute
Price of Loan Market Price
using of Loan
CreditMark
∆ Option Price +
Compute
Price of Loan
w/o Prepay
Option using Compute
CreditMark Spread on Compute
Option- Loan Spread
Corrected for Price
Loan Price
Spread Value
of Option ∆
Source: Citi
44
Spread Value of the Prepay Option
• The idea is to calculate the spread of the floating rate
debt instrument by adjusting the coupon in the
formula below until the value equals the price
4n
ct
F
V =∑ 4 +
f t t −1 + s  f 4 n 4 n −1 + s 
4 4n
,
t =1  
1 +  1 + 
 4   4 
Credit
Spread
• We do this for the loan at the actual market price and

the loan at with the value of the option added to the
market price.
• The difference is the spread value of the option
45
Spread Value of the Prepayment Option
• Spread values of the loan prepayment option are directly
related to the dollar price of the option over a wide range of
values
200%
Relation between
150% changes in spread
Percent Change in Mean Option Price ($)
value of option and

100% price of option
50%
0%
-100% -50% 0% 50% 100%
-50%
-100%
-150%
Percent
Percent Change
Change in Credit Value
in Spread Spreadof
(bps)
Option (bp)
46
April, 2007
Results of OAS Calculations

Coupon MODEL PRICE Value of MARKET PRICE (mid) CREDIT SPREAD Spread
Credit (L+xbp) Prepay No Prepay Quoted Adjusteed Market Adjusted Value of
(50bps Fee) Prepay Option Price For Option Price Price Option (bp)
Blockbuster 350 101.0 106.5 5.5 101.4 106.9 312 170 142
Ford 300 99.4 102.3 2.9 100.6 103.5 288 233 55
Freescale Semi 200 100.5 104.6 4.1 101.0 105.1 182 107 75
HCA 225 101.1 105.8 4.7 100.7 105.4 203 116 87
Windstream 175 100.6 107.3 6.3 100.9 107.2 158 39 119
General Motors 238 100.4 104.9 6.3 101.3 107.6 213 100 113
Supervalu 175 100.5 106.6 6.3 100.8 107.1 157 20 137
IDEARC 200 100.9 108.5 6.3 100.9 107.2 183 70 112
Cablevision 175 100.0 103.7 3.8 100.6 104.4 163 87 75
Fresenious 138 100.8 105.6 4.9 100.0 104.9 137 44 93
Lear 250 100.1 102.7 2.5 100.5 103.1 237 176 61
Allied Waste 175 99.8 102.6 2.8 100.9 103.6 153 83 70
Aramark 213 100.9 106.2 5.3 101.1 106.4 193 97 96
Health Mgmt Inc 175 100.8 106.7 5.9 100.9 106.8 158 54 104
Average 100.5 105.3 4.8 100.8 105.7 196 100 96
Source: Citi
47
Results of OAS Calculations – LCDX
CREDIT SPREAD
Spread
Market - Market Adjusted Value of
Coupon Maturity Market Model Model Option WAL Price Price Option
Credit (L+xbp) (Yrs) Px Px Price Px (FV) (bp) (bp) (bps)
Affiliated Computer Svcs Inc 150 5.8 100.3 99.6 0.7 2.3 3.7 145 97 48
Altivity Packaging Inc. 225 6.1 101.2 100.3 0.9 7.9 3.4 201 47 155
Jarden Corp 175 4.4 100.3 100.2 0.1 5.4 2.3 167 24 143
Advanced Micro Devices Inc 225 6.5 100.5 99.7 0.8 3.9 4.0 215 139 76
American Airls Inc 325 2.7 101.0 100.5 0.4 6.4 1.5 283 9 273
ARAMARK Corp 213 6.7 100.6 100.3 0.3 5.5 3.9 201 98 103
ArvinMeritor Inc 175 5.1 100.1 99.8 0.4 2.6 3.2 172 111 61
Aleris International, Inc 275 6.6 100.2 100.1 0.2 6.0 3.8 270 156 114
Allied Waste North Amer Inc 175 6.9 100.6 100.0 0.6 4.1 4.2 164 90 74
Burlington Coat Factory Whse 225 5.9 99.5 100.3 -0.8 6.5 3.2 236 101 135
Berry Plastics Holding Corp 200 6.9 100.3 100.3 0.1 7.4 3.8 194 61 133
Boston Gen LLC 325 3.4 100.8 100.5 0.3 7.6 1.8 299 44 254
Burger King Corp. 150 5.2 100.3 100.2 0.1 4.5 2.9 142 38 104
Boyd Gaming Corp 150 3.9 100.3 100.2 0.0 3.0 2.2 142 52 90
AVIS BUDGET CAR Rent LLC 125 4.9 100.0 100.1 -0.1 2.6 2.8 124 60 64
Cedar Fair LP 250 5.2 101.1 100.4 0.7 8.6 2.9 223 27 196
Charter Comms Oper LLC 200 7.0 99.9 100.4 -0.5 7.0 3.9 201 76 125
Cap AUTOMOTIVE LP 175 3.6 100.9 100.2 0.8 3.9 1.9 145 23 122
Chiquita Brands LLC 200 5.1 100.9 100.2 0.7 4.6 2.9 179 73
Source: Citi 106
Centennial Cellular Oper Co 200 3.8 100.8 100.0 0.8 4.9 2.0 175 31 144
DEL MONTE Corp 150 5.1 100.3 100.1 0.2 3.6 2.9 143 60 82
Dean Foods Company 150 6.9 100.3 100.2 0.1 4.9 3.9 145 57 88
Dole Food Co Inc 175 5.9 99.8 96.8 3.0 1.3 4.3 179 151 28
Source: Citi 48
Sample Averages
CREDIT SPREAD
Spread
Market - Market Adjusted Value of
Maturity Market Model Model Option WAL Price Price Option
(Yrs) Px Px Price Px (FV) (bps) (bps) (bps)
Average 5.4 100.5 99.5 1.0 4.8 3.0 192 82 110
Source: Citi
49
Estimating Duration
• Recall that we do not Educational Management Corp
know the duration of the
prepayable loan Distribution of Cash Flows from
CreditMark Calculator
• Our approach is to sum 1.2
Weighted Avg Life = 3.3
the cash flows at each
Option Value from T to T-1

1.0
time step in the lattice and
0.8
determine the weighted
average life of the flows 0.6
• We originally did this 0.4
using a bootstrap method 0.2

to determine the option 0.0
value between each time 0.5 1 1.75 2.25 3 3.5 4.25 5 5.5 6
period in the lattice, but Time (Years) Source: Citi
we now have direct

Not all loans have a uniform distribution
access to lattice cash
of cash flows over their time to maturity
flows
50
Distribution of WALs – LCDX Credits
Frequency Distribution of Weighted Average Lives
for Loans of Credits in the LCDX Index
30
25
Number of Credits
20
15
10
0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
Weighted Average Life (Years)
Source: Citi
51
Sensitivity of WAL to Credit Quality
3.05
Top: Mean Average Weighted 3.00
Mean WAL (yrs)

Average Life of the Credits in 2.95
the LCDX Index and Sensitivity Current Index WAL
to a One-Notch Change in 2.90
Credit Rating 2.85
2.80
-1 0 +1
Credit Rating Change
4.50
4.25
Bottom: Mean Weighted Mean WAL (yrs) 4.00
Average Life of the Credits in 3.75
the LCDX Index for Given 3.50
Percentage Changes in 3.25
Credit Spreads 3.00
2.75
2.50
-100% -50% 0% 50% 100%
Source: Citi Percent Change in Credit Spread (bps)

52
WAL and Value of Prepayment
Weighted average lives (WAL) calculated from the model
are directly related to values of the prepayment option
Relationship Between Weighted Average Life and Dollar Value
(Left) and Spread Value (Right) of the Prepayment Option
0.70 0.70
WAL (Fraction of Nominal Maturity)
WAL (Fraction of Nominal Maturity)

0.65 0.65
0.60 0.60
0.55 0.55
0.50 0.50
-0.2 0.2 0.6 1.0 1.00 1.40 1.80 2.20
Log Value of Prepayement Option ($) Log Value of Prepayment Option (bp)
Source: Citi
53
Outline
Applications
54
LCDS Contract (“Just Add L” - almost)
• Credit Protection “Buyer” = pay credit spread = “short” the credit risk
• Credit Protection “Seller” = receive credit spread = “long” the credit risk
• Potential Outcomes:
– Maturity of contract – seller pockets the premium over the life of the contract
– Contract is cancelled when all underlying loans are retired
– Unwinding of the position – termination of contract at market value prior to
maturity
Two Types of LCDS Contracts:
• Reference Entity
– Non-call feature unless no other senior secured security
– Does have “cancelability” and can be substantial
– Most common in US
• Reference Obligation
– Contract terminates if loan prepays
– More common in Europe, but changing 55
Popularity of LCDS
Many of the advantages of LCDS are similar to
the advantages of CDS
• Investors can go short loan-related risk
• Access to leverage
• Buy loans at par
• Non-callable feature of LCDS
• Access to illiquid credits
• Efficient/Cheap way to express credit views
56
LCDS Market Evolution
• In 2004 and 2005 total of 6.3Billion LCDS traded
• LCDS Contract Standardized in June 2006
• In the first month, 500MM face traded and has grown
every month since
• From January-June 2006, 4.5Billion LCDS Traded
• From June-2006-January 2007, 20.5Billion LCDS
Traded
• LCDX began trading in May 2007
• LCDX tranche market started in August 2007
57
LCDS as a Hedge for Cash Loans
• So far, LCDS serves as only an approximate hedge for cash loan
price moves and then only over relatively long horizons (i.e., one
month)
120
LCDS Spread Change in bp (20 Days)
100
80
60
40
20
-20
-40
-60
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Cash Loan Price Change (20 Days)
This study is done using the data from 1-Feb-07 to 1-Jul-07. The sample includes
covers 93 liquid loans and 46 LCDS. The LCDS spread and loan price changes are
calculated on a 20-day interval. The relationship between LCDS and cash loans is
poorer for longer and shorter intervals.
58
Market Players and Features
• There are a larger number of sellers of
protection than buyers
• This means that:
– People want exposure to loans, but do not want the
messiness of prepayment or wide bid/ask spreads
– Broker/dealers typically are hedging sales of protection by
being long the underlying (no strong 2-way market)
• Most current investors are hedge funds

looking for a cheap way to take a view on
credit
• Relatively little activity in LCDS from banks
59
Challenges and Opportunities
• Need for evolution of two-sided market
– Equalizing buyers and sellers across broker/dealers and
investors
– Segmented market causes “specialness” of some loans as
dealers scramble for paper with which to hedge LCDS
– Tends to bid up market for loans and keep basis high
• Pricing the loan prepayment option

– Market currently prices it at 15bp-25bp
– Option dependent on credit state and risk premium
• Development of LCDS Index and Tranche

Trading
4 60
Outline
Applications
61
FR
ES
EN Credit Spread (bp)
IU
S
H A
0
100
200
300
400
500
600
EA SU G
LT PE
H R
M VA
G LU
M
T
A
SS
O
ID C
EA
R R
H C
D IN
O C
N
5Y LCDS
N
A EL
R
A LE
M Y
A
R
FR K
C
EE O
SC R
A P
5Y CDS
LE
SE
M
I
U H
N C
IV A
IS IN
IO C
.
N
C
A O
LL M
IE M
D
W
A
LE ST
G A E
EN R
C
ER O
A R
L P
M
B O
LO TO
C R
K S
B
U
ST
FO ER
R
D
M
O
TO
R
Source: Citi
.LCDS Spreads Trade Tight to CDS . . .
CDS Recovery Rate is About 50%; LCDS is about 80%
62
Ratio of CDS to LCDS is 2.6 to 1
CDS and LCDS are
related by a ratio of CDS versus LCDS
2.6 to 1.0 700
CDS = ( 2.6 * LCDS ) - 26

Market convention to 600
R2 = 0.57
correct for recovery 500

rate is S/(1-R)
CDS (bps)
400
Recovery rate for 300

CDS is about 40%
200
and for LCDS is
about 75% 100
0
Then (1-0.4)/(1-0.75) = 50 100 150 200 250
2.4; close to what we LCDS (bps)
observe
Source: Citi
63
The Loan / LCDS Basis
The LCDS versus Reference Loan Basis Has Been in
the Range of 50bp-125bp
Why is there a Loan / LCDS basis?

• No Finance Charge for LCDS
• LCDS are More Liquid
• Loans are Prepayable; LCDS are Somewhat Non-Cancelable
• Lower Management Fees on CLOs of LCDS vs Cash
• Cash Loan Deviation from Par Price
64
The Loan / LCDS Basis
• We calculated option values for the likely deliverable
loans into the liquid LCDS contracts
• We estimate the average value of the prepayment

options for the loans in our sample to be about 5 points
• The average spread value calculated for the option is

just over 100bp
• Market is pricing the Cash/LCDS basis at about 80bp
65
Outline
Valuing LCDS Cancelability
Applications
66
US LCDS Are Often Cancelable
• LCDS contract may get cancelled when the
underlying company’s senior secured loans
are repaid
• We valued the cost of LCDS cancelability

through Monte Carlo simulation using
historical ratings transitions
• Valuation method is based on the insight that

the default-contingent leg (payout leg) of the
LCDS is virtually unaffected by cancellability;
but the premiums will be reduced
67
Conditions for Cancelability
There are three main ways that companies can exit the senior
secured loan market:
1. Credit upgrade to investment grade — or near investment grade when
the company will replace its secured loans with unsecured debt
2. Acquisition by an investment-grade firm — at which point its secured
debt will be replaced by unsecured debt
3. Acquisition by another high-yield firm — where the debt would be repaid
and reissued by the new company. In this case, the LCDS of the
acquiring firm would likely remain outstanding, while the LCDS of the
target firm will terminate
Impact on LCDS pricing:

• ~ 40% of leveraged loans get cancelled before maturity
• 20-30% reduction in MTM suggested by Reuters
• The likelihood of LCDS cancelability is correlated with increases in
credit quality, implying the need for greater compensation to the
seller of protection
68
Valuing Non-Cancellable LCDS
FEE LEG: Assume that the cumulative probability of default until time t in
years is pdt and dft is the discount factor up to time t. The protection buyer
pays S/4 basis points every quarter if there is no default. The sum of
discounted cash flow is: 4T
S  
V fee = ∑ ∗ df t ∗ 1 − pd t  (1)
t =1 4 4  4 
CONTINGENT LEG: The contingent leg consists of a single payment of (1- RV)
in the event of default, where RV is the recovery value.* The protection seller
pays the excess amount that cannot be recovered, so the contingent leg can
be represented as: 4T  
Vcontingent = ∑ (1 − RV ) * df t *  pd t − pd t −1 
 
(2)
t =1 4  4 4 
Because, by definition, the expected values of the fee and contingent legs
are equal at contract initiation, we can solve for S from the two equations if
we know the discount factors, the term structure of default probabilities,
and the recovery value in default.
*For simplicity we keep RV at a fixed value. 69

Valuing Cancellable LCDS
FEE LEG: Denote the annual protection premium for cancelable LCDS as S’.
Also, let the cumulative probability of cancelation up to time t be pct . Since the
protection buyer pays a premium of S’/4 basis points every quarter if neither a
default nor a cancelation has occurred, we can rewrite Equation 1 as:
4T
S'    
V fee = ∑ * df t ∗  1 − pd t  ∗  1 − pc t  (3)
t =1 4 4  4   4 
CONTINGENT LEG: A key assumption of our method is that the contingent

case remains the same if cancelation does not occur or if default occurs
before the loan is cancelled. On the other hand, the protection seller does not
pay anything if the cancelation occurs before the company defaults. The cash
flow can be rewritten as:
4T    
Vcontingent = ∑(1 − RV) * df t *  pdt − pdt −1  * 1 − pct 
   (4)
t =1 4  4 4   4
Equations 3 and 4 are used to compute the spread premium of an LCDS

contract with the effect of cancelation included. Then, having calculated S and
S’, the value of the cancelation option is obtained as the ratio of the two
premiums
70
Problems with the Approach
Two major problems with the method described in Equations 1–4:
1. The default and cancelation probabilities, pdt and pct, that we used are
physical, not risk-neutral, probabilities
– Thus, the spread values will not likely correspond to market spreads.
– Still, the effect of cancelation can be estimated by comparing the ratio of the spread
premium calculated with or without cancellability, using either model-based physical
default probabilities or risk-neutral default probabilities
2. Equations 1-4 assume that default and default probabilities and LCDS
cancelations are independent.
– They are negatively correlated; the higher the rating of the company (that is, the lower
the default rate), the more likely the firm will become investment grade and exit the
senior secured loan market.
– Thus, the seller of protection will be deprived of the premiums in precisely those
conditions which they are most likely to be paid
– Conversely, the duration of the LCDS contract will be extended under those conditions
when the obligor is most likely to default.
To address these problems, we used a Monte Carlo method to simulate

actual ratings transitions because that method implicitly embodies the
negative correlation between LCDS cancelation and default
71
Cancelations and Defaults
Sample paths for Ratings Transition Matrix: Sample paths
Monte Carlo AAA
simulation of credit AA Repay Loans
Refinance
rating transitions A
from S&P (1980-2000) BBB
IG / HY Boundary
show absorbing
effects of LCDS BB
cancelation on B
reaching investment CCC

grade or LCDS payoff D Default
Default
in default
Time Source: Citi
Types of Sample paths Non-Cancelable Cancelable
1. Stays below IG and above default - -

2. Goes above IG - Cancelled
3. Defaults without going to IG Default Default
4. Goes to IG and defaults Default Cancelled
72
Simulation Methodology
We calculate the spread premium for a cancellable LCDS contract
for a class of loans with initial rating Rinitial = Rt = 0
Each simulation consists of 50,000 ratings transition paths at

quarterly time steps over a five-year period of the LCDS
For each path we compute quarterly credit state transitions from Rt-1
to Rt where the probability of transition to a given state at time t is
conditional upon its state at time t-1
On any given time step, t-1 t, several possible events can occur:
• If Rt = D (default), then the contingent leg pays (1-RV)∗dft and the contract
terminates.
• If Rt = BBB- (investment grade) the contract terminates at t with no charge
to the contingent leg
• If Rt = BB+ at t, the fee leg pays a coupon of S/4 for the quarterly period
The values of the fee and contingent legs are equal by construction,
so the resulting value of S’p is our LCDS spread under conditions of
physical default probability and cancellability
73
Simulation Methodology
We next compute the spread premium, Sp, under conditions of
physical default probability for the noncancelable LCDS by removing
the cancellability condition in the simulation algorithm
Then, we can estimate the percent spread adjustment for

cancellability as:
[( ) ]
S Adj = S p' S p − 1 * 100 (5)
Recall that the spread adjustment in Equation 5 is determined using

physical probabilities of default and cancellation. Hence, the actual
values for the spreads should not be used to value the contract per se.
Instead, we use the percent change in spread due to the effect of

cancelation as applied to observed CDS spreads adjusted for the
differential recovery values between loans and bonds. The resulting
spread can then be adjusted by the ratio in Equation 5 to yield an
estimate of the fair spread value for the cancelable LCDS contract.
74
The Cost of LCDS Cancellability
Loan cancellation probability depends on credit rating
Cumulative Loan Cancellations vs Credit Rating
Cumulative Cancelation Rate
Year Source: Citi
Loan cancellation probability increases with time

The higher the credit rating, the more likely that the LCDS contract
will cancel
We estimate that up to 27% of BB loans will be cancelled within five
years
Cancellability falls rapidly for lower-rated credits
75
Other Reasons for Cancellations
• JP Morgan & Reuters study: ~ 45% loans get cancelled
• Ratings transitions imply ~ 27% BB and less for B, CCC
• Reasons for LCDS being cancelled
• Rating upgrades don’t drive all the cancellations
• Ratings transitions are credit cycle dependent
Historical and Ratings-Related Cumulative Average Leveraged Loan Cancellations
1st year 2nd year 3rd year 4th year 5th year
Historical 3% 8% 18% 23% 38%
Ratings Matrix 4% 7.5% 10% 13% 15%
Assumes loan proportions of 50% BB, 45% B and 5% CCC
We capture less than half the cancellations reported by Reuters/JP

Morgan with our rating transition based simulations
Source: Citi 76
Ratings Upgrades are Cyclical
Annual Ratios of Ratings Upgrades to Downgrades Since 1981
1.50
Ratings Upgrade/Downgrade Ratio
1.25
1.00
Average
0.75
0.50
0.25
0.00
1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
Source: Citi and Standard & Poor’s
• The upgrade-downgrade ratio is nearly 1.4, an historical high

• The average upgrade-downgrade ratio is only 0.75
• The upgrade-downgrade ratio has been above average over most
of the time period of the Reuters/J P Morgan stud
We adjusted our ratings transition matrix toward increasing credit

quality to match the values reported in the Reuters study
77
Spread Changes Due to Cancellations
Percentage Spread Increases Due to
Cancelations for Five-Year LCDS Contracts Using
Actual and Adjusted Ratings Transitions
LCDS Spread Adjustment
Credit Rating Source: Citi
• 22% spread widening for 5Yr BB+ LCDS with adjusted transition matrix
• Only 10% with historical transitions
• Spread change decreases rapidly as we move away from IG/HY boundary
78
Spread Adjustments by LCDS Tenor
Percentage Spread Adjustment for LCDS Cancelability
by Tenor and Rating Category
LCDS Spread Adjustment
30% 2 Year 4 Year 5 Year 7 Year
25%
20%
15%
10%
5%
0%
BB+ BB BB- B+ B B- CCC
Rating
Credit Rating
Source: Citi
• We see a spread change ‘smile’ due to the effect of M&A for CCC
rated companies
79
Spread Cost Approximation
• Cancellations for BB in 5 years = fc 45%
• Mean cancellation time τc 2.8 years
Cash flows after τc will not be received fc of the times
• Non-Cancellation Case: (Spread premium = S)
Expected fee (Discount factor = dfi = 5%) S (Σi=1..5 dfi) 4.4 S
• Cancellation case: (Spread premium = S’)
Reduced fee (when cancelled = 45%) S’ (df1 + df2 + 0.8df3) 2.6 S’
Regular fee (not cancelled = 55%) S’ (Σi=1..5 dfi) 4.4 S’
Expected fee (1-fc)feeregular + fc feereduced 3.6 S’
• ASSUME: Contingent legs are equal in both cases: (default distribution does not change)
Expected fee(non-cancellation) = Expected fee(cancellation) (Assumes Libor = 5% flat)
=> 4.4 S = 3.6 S’ => S’ = 4.4/3.6 S => 23% change
Non-Cancellable Cancellable
Contingent Leg Contingent Leg
Source: Citi
55%
Fee Leg
Fee Leg 45%
2.8 years 80
TIME
TIME
Outline
Applications
81
March 2008
• Prices of liquid
near record low

in June of 2007 to
• Despite this, actual

of 2008 and back to
• Credit risk premium

below 90 in January
for default peaked in
default levels remain

above compensation
leveraged loans have
levels, even into 2008

fallen from above 100
about 94 in June 2008
17
Risk Premium (λ
λ) Loan Price
17
-A - Ap
p
88
90
92
94
96
98
100
102
r-0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
8- r-0 8- 7
M 7 M
a ay
-0
7
29 y-0 29
-M 7 - M
a ay
-0
19 y-0 19 7
-J 7 -J
u un
-0
10 n-0 10 7
-J 7 -J
u ul
-0
31 l-0 31 7
-J 7 -J
ul
21 ul-0 -0
-A 7 21 7
-A
u ug
11 g- -0
-S 07 11 7
-S
ep ep
-0
2- -07 7
O 2-
O
c ct
23 t-0 -0
-O 7 23 7
-O
ct
13 ct-0 -0
-N 7 13 7
ov -N
ov
-0
4- -07
D 4- 7
e D
ec
25 c-0 -0
-D 7 25 7
e -D
Loan Prices Have Plummeted
ec
15 c-0 -0
-J 7 15 7
an -J
an
-0
8
5- -08
Fe 5-
Fe
26 b-0 b-
-F 8 26 08
e -F
eb
18 b-0 -0
-M 8 18 8
ar -M
ar
-0
8- -08
A 8- 8
p A
29 r-0 pr
-0
-A 8 29 8
-A
20 p r- pr
-0
-M 0 8 20 8
a -M
ay
10 y-0 -0
-J 8 10 8
un -J
-0 un
8 -0
8
82
Risk Premium Implied from CDS Market
Average of Liquid Leveraged Loan Prices
Risk Premium Typically Leads Default Rate
Default Rate Coincident
Current Default with
Rate versus Current RiskRisk
PremiumPremium
12%
Risk premium often 10%
DR = (0.26 * RP) - 0.06
2
R = 0.57
rises and falls with
Default Rate
8%
default rates 6%
4%
2%
But risk premium is not 0%

0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
a good indicator of Risk Premium

Current Leads
Default Rate versus PreviousDefault
Risk PremiumRate
12%
future default rates 10%

DR = (0.09 * RP) - 0.002
2
R = 0.16
Default Rate
8%
6%
It is also typical for 4%
2%
changes in default rates 0%
to lead changes in risk 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Default
Previous Rate Leads
Default Rate RiskRisk
versus Current Premium
Premium
12%
DR = (2.1 * RP) - 0.3
10%
Currently, default rates 2
R = 0.59
Default Rate
8%
are near historical lows 6%
and the risk premium 4%
near historical highs 2%
0%
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
83
Risk Premium
LCDS/CDS Basis Has Been Diverging
Prior to July of 2007, CDS Firms’ Average LCDS is Below Firms’ CDS
1000
and LCDS were related by
800
a ratio of 2:4 to 1.0 5Y CDS Spread
Spread (bp)
600
Market approximation is
400
to adjust spread for
recovery rate is S’=S/(1-R) 200
5Y LCDS Spread
0
Recovery rate for CDS is
7
7
8
7
8
7
07
7
7
8
07
08
07
07
08
08
7
-0
-0
-0
-0
-0
l-0
-0
-0
-0
-0
-0
about 40% and for LCDS n-
n-
p-
b-
n-
b-
n-
ct
pr
pr
ar
ar
ay
ay
ug
ec
ov
Ju
Se
Ja
Ja
Fe
Fe
Ju
Ju
O
M
M
A
A
M
M
D
A
N
is about 75% The Ratio of CDS/LCDS Has Dropped Since July 2007
3.5
Ratio of CDS/LCDS Spread
Then (1-0.4)/(1-0.75) = 2.4;

close to what we observe 3.0
2.4x Assumption
prior to July 2007 2.5
However, the current ratio 2.0

CDS/LCDS Ratio
is has decreased to about
1.5
1.5
1.0
7
7
8
7
8
7
07
7
7
8
07
08
7
07
08
08
7
-0
-0
-0
-0
-0
l-0
-0
-0
-0
-0
-0
0
n-
n-
p-
b-
n-
b-
n-
ct
pr
pr
ar
ar
ay
ay
ug
ec
ov
Ju
Se
Ja
Ja
Fe
Fe
Ju
Ju
O
M
M
A
A
M
M
D
A
N
84
Estimate Loan LGD from CDS Risk Premium
• Given risk premium from CDS market, calculate the risk-neutral PD
for LCDS market
CPDTQ =(N N(−1 CPD) )
T + βλ T
• Use LCDS curve to compute implied recovery rate (R)

 1( ST 
−)
R = 1−  Q
1 − e 
 PD
We infer that
Implied Recovery Rate for Loans from CDS Risk Premium
the average 0.83
recovery rate
Implied Recovery Rate
for loans has 0.78
decreased due 0.73

to the large 75% assumption
0.68
issuance of Implied recovery rate
“covenant lite” 0.63
loans 0.58
-A 07
-S 07
M 7
-M 0 8
-J 7
8
-A 8
-M 7
-J 7
1- l-0 7
-J 8
-F 7
6- -0 8
7- -0 7
-A 7
-M 8
9- -0 8
-J 7
-O 7
-M 7
-N 7
5- -07
-F 8
-D 7
3- -07
-J 7
30 r-0
9- r-0
20 y -0
11 n-0
11 y -0
-0
24 n-0
30 y-0
14 n-0
28 r-0
14 c t-0
27 b-0
18 r-0
26 c -0
19 b-0
24 t-0
16 c-0
22 g -
12 g-
21 p r-
un
ov
an
eb
ar
ep
u
p
p
c
a
u
a
u
a
a
u
Ja
Fe
e
e
e
A
O
M
85
D
3-
Outline
Applications
86
Single-Name Loan Trading
The prepayment option is likely inefficiently priced in the
market in many cases, providing trading opportunities
• Competitive advantage to more accurate pricing:

– Tighter bid/ask spreads
– Better relative value
– Good for both Loans and LCDS
• Loan / LCDS / Bond /CDS / Equity Arbitrage

– Prepayment option is difference between Loans and LCDS
– Another difference between Loans and LCDS is difference in price
from par
– Recovery option differs between CDS and LCDS
– Can trade across capital structure
87
Trading in Secondary CLO Market
All the benefits of single name loan valuation plus:
• Ability to price deals quickly accounting for credit option

– Many deals contain illiquid assets without prices
– Can bid/offer CLOs with greater accuracy (competitive advantage)
• Improved tranche valuation

– Better estimate of loan durations can provide more accurate pricing of CLO
tranches
– Can hedge sensitivity to changes in risk premiums and credit quality
• Mark inventory to market nightly

– Gain regulatory relief over marking loans at par (FAS 159 – “Fair Value Option”)
– Compute sensitivities on portfolio to individual credits
– Compute sensitivity to variations in loan risk premiums
88
Improved Relative Value Analysis
Given the dependence of loan spread on default, recovery, and weighted
average life, one can compute relative value, ε, as:
ε = ln(S ) − {6.6 + [0.23 ∗ ln(CPD)] + [− 0.44 ∗ ln(RV )] + [− 0.12 * ln(WAL)]}

Loan Spreads to LIBOR versus Estimated Default Probabilities, Estimated
Recovery Values, and Weighted Average Life
7
1100 6.9 6.9
1100 1100
600
6.5 600
6.7 6.7
600
Spread (bp)
Spread (bp)
Spread (bp)
4006 400
6.5 6.5
400
5.5
250 250
6.3 6.3
250
1505 150
6.1 6.1
150
4.5
90 90
5.9 90
5.9
-6
0.002 -4
0.02 -2
0.12 10 2
7 4
55 3.6
37 554 81
4.4 0.1 0.4 1.0 2.7 7.4
-2 -1 0 1 2
Recovery Value (%)
HPDHPD Default Probability (%)
Default Probability Recovery in Default (%) Weighted Average Life (Yrs)
Weighted Average Life (Yrs)
Source: Citi
89
Loan Relative Value (cont.)
We regressed the
logarithms of predicted Actual vs Predicted Spread – Loan Universe
spreads on actual 7
spreads based on the
model in the previous
slide
Ln(Acutal Spread)
6
There were a total of 277

loans for which we had
spreads, maturities, and 5
recovery values
The line in the figure is

4
our “fair value” line; 4.8 5.2 5.6 6.0
credits above the line are Ln(Predicted Spread)
“cheap” and those below

the line are “rich”
Source: Citi
90
New Loan Relative Value Function
We calculated CPD vs Predicted Spread – Liquid Cash Loans
4
10000
prepayment option- 0.53
3.5
3162 OAS = 140 * PD
adjusted spreads
Zero Recovry OAS (bp)

R2 = 0.68
(OASs) for liquid 1000
3
cash loans using our 316

2.5
adapted CreditMark
2 fro
100
model
32
1.5
We adjusted those
1
10
OASs for “zero-
recovery” value 0.53
using values from 01

0.01 0.03 0.1 0.3 10 3 10 32 100
Moody’s LossCalc -2 -1.5 -1 -0.5 0.5 1 1.5 2
model Default Probability (%) Source: Citi
There were roughly 200 loans that had at least 10 price changes per month
and we regressed their zero recovery OASs against HPD model CPDs
The line in the figure is our “fair value” line; credits above the line are
“cheap” and those below the line are “rich”
91
Portfolio Relative Value
We compare market Actual Market-Weighted Average OAS and
spreads of credits in the Predicted Average Spreads from Market Norms
portfolio to average Spread Model Spread
spreads for that CPD, Market Weighted Average Spread 239 201
duration, and recovery
value Actual vs Predicted Spreads by Rating
Agency Category
We can then determine 600
whether the portfolio is 500

Actual Spread Average Spread for CPD
generating above, below,
Spread (bp)
400
or average market returns 300
for the level of risk 200
100
This portfolio is returning
0
more than the market on
Caa1
Caa2
Caa3
Ca
Baa1
Baa2
Baa3
Aaa
Aa1
Aa2
Aa3
Ba1
Ba2
Ba3
A1
A2
A3
B1
B2
B3
average (239bp vs 201bp)
Actual Moodys Rating
for that level of risk
Note that the credits in the rating categories above are determined by
agency ratings, but the normative spreads are based on PDs from those
Source: Citi credits
92
Expected Losses
We assume default
correlations of 30% among Frequency Distribution of Expected One-Year
credits in same sector and Loss Rates From Copula Model Simulation
10% across sectors 100%
We can run Monte Carlo 80%
Frequency
simulations using a two- 60%
factor copula model of
correlation to determine 40%
expected loss distributions

20%
for various time frames
0%
We can compare the results 0% 1% 2% 3% 4% 5% 6%
of portfolio hedging or % Portfolio Loss
substitutions in terms of
changes in loss distributions
Recovery values are from Moody’s
calculated from our simulator
LossCalc model
Source: Citi
93
Appendix: Risk neutral PD and Risk Premium
Let V denote the asset value of a firm. Its probabilistic evolution in time
can be modeled as a geometric Brownian motion process with drift, µ
dV = µVdt + σVdW
Within the contingent claims analysis (CCA) model and using Itō's
lemma, the probability, p, that the terminal value of the firm’s assets is
less than certain boundary B is
 1 2 
 ln( B) − ln(V0 ) − µT + σ T 
p = N 2 
 σ T 
 
 
To convert to the risk-neutral measure, pQ, we change the drift, µ, to the
risk-free rate, so that
dV = rVdt + σVdW and
l
n
l
n
( ) ( ) 1 2 
 B − V 0 − rT + σ T
pQ = N 2 
 σ T 
 
  94
Risk neutral PD and Risk Premium (cont.)
We can express the relationship between the physical and risk-neutral default
probabilities via the the capital asset pricing model (CAPM).
The premium over the risk-free-rate an investor should require to invest in a
risky asset depends on its holding period, volatility, and the current market
price of risk.
That is, we can write the risk-adjusted default probability, pQ, as:
µ −r )
p Q =(N N(−1 )
p + T
σ
where within the CAPM framework,
Market excess return,
µ − r = β (µm − r ) µm - r, times the beta

(β) of the credit
cov(µ , µ m ) β is the covariance of the

= (µm − r ) asset’s return with the market
σ 2
m
return relative to the variance
of the market
cov(µ , µ m ) ( µ m − r)
= σ By rearranging
σ mσ σm
95
Risk neutral PD and Risk Premium (cont.)
Then, if we define ρ as the correlation between the return of the asset and
the market return, and λ as the market price of risk, we can write:
cov(µ , µ m ) µm − r
ρ= and λ=
σ mσ σm
and substituting back in the equation for the risk-neutral default probability,
we get:
p Q =(N N(−1 )
p + λρ )
T
Thus, the risk-neutral default probability within contingent claims analysis

(CCA) can be thought of as an adjustment to the asset value of the firm that
serves to increase the density of the future states of the firm below the level
of the debt.
96
Risk Neutral Default Probability and Credit Spread
The price of a default-risky asset can be written as e−( r +s )T and by definition
of the risk neutral default probability, we can write:
[ ]
e − r + s T = 1 − (LGD ∗ p Q ) e − rT
( )
where LGD is the loss-given-default (1 minus recovery value in default)

and we can write the credit spread, s, as:
l
n
(1 − LGD ∗ p )
1
s=− Q
T
Thus, we can estimate the risk-neutral default probability of a single
risky cash flow by knowing its credit spread, s, the time it is to be
received, T, and assuming a recovery value in default to determine the
LGD.
97
Appendix – External Communications Disclosure
This communication has been prepared by a member of the Sales and Trading Department of Citigroup which distributes this communication by or through its locally
authorized affiliates (collectively, the “Firm”). Sales and Trading personnel are not research analysts and the information in this communication is not intended to
constitute “research” as that term is defined by applicable regulations. Compensation of Sales and Trading personnel includes consideration of the performance of
this Department’s activities. The views expressed herein may change without notice and may differ from those views expressed by other Firm personnel.
You should assume the following: The Firm may be the issuer of, or may trade as principal in, the financial instruments referred to in this communication or other
related financial instruments. The author of this communication may have discussed the information contained herein with others within the Firm and the author and
such other Firm personnel may have already acted on the basis of this information (including by trading for the Firm’s proprietary accounts or communicating the
information contained herein to other customers of the Firm). The Firm performs or seeks to perform investment banking and other services for the issuer of any such
financial instruments. The Firm, the Firm’s personnel (including those with whom the author may have consulted in the preparation of this communication), and other
customers of the Firm may be long or short the financial instruments referred to herein, may have acquired such positions at prices and market conditions that are no
longer available, and may have interests different or adverse to your interests.
This communication is provided for information and discussion purposes only. It does not constitute an offer or solicitation to purchase or sell any financial
instruments. The information contained in this communication is based on generally available information and, although obtained from sources believed by the Firm
to be reliable, its accuracy and completeness is not guaranteed. Certain personnel or business areas of the Firm may have access to or have acquired material non-
public information that may have an impact (positive or negative) on the information contained herein, but that is not available to or known by the author of this
communication.
Financial instruments denominated in a foreign currency are subject to exchange rate fluctuations, which may have an adverse effect on the price or value of an
investment in such products. Investments in financial instruments carry significant risk, including the possible loss of the principal amount invested. Investors should
obtain advice from their own tax, financial, legal and other advisors, and only make investment decisions on the basis of the investor’s own objectives, experience
and resources.
This communication is not intended to forecast or predict future events. Past performance is not a guarantee or indication of future results. Any prices provided herein
(other than those that are identified as being historical) are indicative only and do not represent firm quotes as to either price or size. You should contact your local
representative directly if you are interested in buying or selling any financial instrument, or pursuing any trading strategy, mentioned herein. No liability is accepted by
the Firm for any loss (whether direct, indirect or consequential) that may arise from any use of the information contained herein or derived herefrom.
Although the Firm is affiliated with Citibank, N.A. (together with its subsidiaries and branches worldwide, “Citibank”), you should be aware that none of the other
financial instruments mentioned in this communication (unless expressly stated otherwise) are (i) insured by the Federal Deposit Insurance Corporation or any other
governmental authority, or (ii) deposits or other obligations of, or guaranteed by, Citibank or any other insured depository institution.
This communication contains data compilations, writings and information that are proprietary to the Firm and protected under copyright and other intellectual property
laws, and may not be redistributed or otherwise transmitted by you to any other person for any purpose.
Copyright © Citigroup 2007. All Rights Reserved.
98

Terry Benzschawel

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Terry Benzschawel

Încărcat de

Drepturi de autor:

Formate disponibile

Modeling Corporate

June 28, 2008

Fixed Income Strategy &

“… most par loans remain ill-suited

From LCD Leveraged Lending Review: Fourth Quarter 2006,

– Often, bonds have embedded call Credit Spread

To compute the value of a bond with 2

  2    2 (ct), rt is the US Treasury Yield and st is the

Valuing the loan prepayment option is difficult because:

The majority of bid and 40

In the CDS world, these

Source: MarkIt Partners and Citi

Number of Loans Per Month

Number of Liquid Loans

about all loan prices beyond 100

Prices are “mean 2

The envelope of price changes for all percentiles is constant after

There is a definite “pull-to-par”; prices above 100 trend downward

Source: MarkIt Partners and Citi 17

Loan Rating BB B CCC

• Negative γ so that Volatility when Price

• Use historical or Citi’s HPD model default probabilities

Each of the 200

• σ Rating = varHistorical − varIndex

Historical Default Probability for Leveraged Loans

s = − ln[1 − (CQDPT ∗ LGD )]

s = − ln[1 − (CQDPT ∗ LGD )]

s = − ln[1 − (CQDPT ∗ LGD )]

Convert PD into Compute monthly

Calculate the price using backward Apply company

Source: Moody’s/KMV and Citi

Pricing Grid: Reduced Fees

Revolver DD1to Default 12

• Determine cash flows 9 105

105.5, continue) 3 104.33 105

greater than the

12% more of public loans are valued

Loan vs. CDS

• We do this for the loan at the actual market price and

value of option and

Results of OAS Calculations

Option Value from T to T-1

• We originally did this 0.4

using a bootstrap method 0.2

period in the lattice, but Time (Years) Source: Citi

we now have direct

Top: Mean Average Weighted 3.00

Mean WAL (yrs)

Credit Rating 2.85

Percentage Changes in 3.25

Credit Spreads 3.00

Source: Citi Percent Change in Credit Spread (bps)

WAL (Fraction of Nominal Maturity)

• Most current investors are hedge funds

• Pricing the loan prepayment option

• Development of LCDS Index and Tranche

CDS Recovery Rate is About 50%; LCDS is about 80%

CDS = ( 2.6 * LCDS ) - 26

correct for recovery 500

Recovery rate for 300

Why is there a Loan / LCDS basis?

• We estimate the average value of the prepayment

• The average spread value calculated for the option is

• Market is pricing the Cash/LCDS basis at about 80bp