Ali Bora Yigitbasioglu

ISMA Centre, UK
School of Finance and Economics, University of Technology Sydney
Convertible Bonds Pricing
The Plan
1. Convertible bonds (and related products) in the market context -
this is the lighter part of the talk

Definitions
Whys?
Market figures
Empirical anomalies
The Plan
2. Models- the second part of the talk.

Simpler models, shortcomings, and evolution
What a model MUST address
An implementation of a comprehensive framework for pricing
convertibles


Some light history and market background
Introduction to Convertibles
Variations to the basic instrument
Additional safety provisions
Features; Terms and Conditions; Definitions; Payoff and Value Profiles
Issuers and Investors
The life cycle of convertibles and linkages to modelling
Non standard features
Convertible Bonds...
5
History
First convertible issue associated with the Railroad magnate, J.J.
Hill, 1881.
The market ascribed too much risk to the railroad project
Required innovative long-term financing as he was shut out from the
traditional bond market
Unwilling to sell stock before expansion had reaped substantial rewards

Convertible bonds still fulfil the same financial need as they did in
1881: A way for companies whose stocks are volatile to access the
debt market

Today, this market is in excess of USD 450 billion
6
Convertibles: Introduction

As a general rule, debt investors exact a higher toll for
risk
Translates into a higher coupon from an issuer with high

Thus if you cannot convince your financiers that the risk
on your common shares is less, you will have to bear a
higher interest cost than your expectations may justify

Volatility is not a bad thing when it comes to convertible
financing, however!
Negative impact to bond component
Positive impact to warrant component


Deal info: issuer, underlying, size, currency, yield, premium, etc
Convertibles: Key Features
Structure: calls, puts, conversion terms, coupon/dividend rights
- Calls: issuers call option to repurchase their bonds at a
specified price (or yield), normally from several years after issue
until maturity (American); for some/all of this time, call may be
provisional, i.e. underlying share price must exceed a trigger
- Puts: holders put option to sell bonds back to issuer at a
specified price (or yield), less common than calls, and normally
on a few discrete dates during the life of the bond (Bermudan)
8
Convertibles: Introduction

Coupon and dividend rates
generally set below what the issuer would have to pay in non-
convertible market (300-400 bps below)
Averages 400-600 bps above that of the common stock

Maturity
Initially of up to 20+ years in the early 1980s in particular
Nowadays, 7 years is the norm
Very few exceed 10 years






9
Convertibles: Introduction

Call protection
Rare until 1982
In the early 1980s, several issues were redeemed before even a single
coupon was paid
Soon after, to protect investors, call protection became a standard
feature

Different modes of call protection
Hard call
Soft Call (Asian or Parisian feature)
10
Convertibles: Introduction

Subordination
Vast majority of CBs are subordinated debt
Rank junior to any senior debt, whether existing or prospective
Convertible preferreds, as a type of equity, rank below all debt but equal
to preferred stock
Preferreds ranks may be further stratified as first, second, etc

Taxation
As a general rule, conversion is a non-taxable event
An investors basis in the convertible is carried over to the stock
received upon conversion
Taxation into a package of stock and bond is taxable with respect to the
bond portion of the package.
11
Convertibles: Variations

Original Issue Discount Bonds
Below market coupon levels
Offered at a deep discount to par (face) value
Have put options
One of the first CBs priced in an academic paper

Step-Up Convertible Bonds
Between OID and Coupon Pay CBs
The initial interest rate is stepped up
In most cases this is scheduled to occur at the first call date
12
Convertibles: Variations

Eurodollar Convertibles
US denominated convertible bonds sold outside of the US
Traditionally centred in London
Not registered with the SEC and issued in bearer form
Interest coupons paid annually vs. semi-annual in US domestic
CBs
Eurodollar CBs may not be sold to US investors prior to a 40-
day seasoning

Breakdown
US corporations in USD
Foreign corporations in USD with conversion into non-USD
FX risk!!!
13
Convertibles: Additional security provisions

Change of control provisions
Previous provisions: conversion ratio adjusted upward and then trading
as a cash instrument
Nowadays, common if not universal features are poison puts which
enable exit at par or first call price. These features are meant to protect
investors from mergers that are potentially harmful for the conversion
option

14
Convertibles: Additional security provisions

The (infamous) Screw Clause
Read the prospectus! And read it well
Upon conversion, no adjustment will be made for interest or dividends
And what if last conversion date (say after a call is issued for a deep
ITM convertible) falls before coupon date?
You lose accrued interest (CB) or preferred dividend (for convertible
preferred) if you want to convert!
WalMart

And so the Catch-22:
Should I convert and forfeit my accrued interest?
Should I wait to receive interest and forfeit a valuable conversion
option?
And in either case I lose



15
Convertibles: Additonal security provisions

Anti-dilution provisions

Most convertibles protect the CB investor in the event of actions that
may dilute their equity interest
Issues of equity at a discount (such as rights issues)
Stock splits
One-time extraordinary dividends

The convertible conversion ratio is usually adjusted upwards pro
rata the convertible investors theoretical loss in such events
Deutsche Bank into Novartis 1% 2010 exchangeable bond
(1.4bn, 3.125% yield, 28% premium, Barclays co-lead, Nov. 2001)
Convertibles: Example
Coupons
Redemption
Put Put
Conversion
Provisional Call Unconditional Call
Convertibles: Terms and Conditions
Announce Date Launch Date Pricing Date
BREAKING NEWS
e.g. XYZ plc will issue
USD1.25bn of 10-year
convertibles (Reuters,
Bloomberg, brokers).
Not enough detail for
valuation, but enough
to register interest (or
not!) and underlying
stock may be sold off.
INDICATIVE TERMS
Lead managers send
out term sheets: a
couple of pages of
key terms (e.g. yield,
convertibility, calls,
puts, credit status).
Enough for general
valuation/risk, and
grey market trading.
LEGAL DOCUMENTS
FINAL TERMS
Prospectus: available
from lead managers.
Hundreds of pages of
legal/regulatory info,
terms and conditions.
Trust Deeds: ultimate
legal recourse, if need
be. Held by Trustees.
Conversion ratio: number of shares into which one bond converts
Parity: conversion ratio x spot share price (x spot exchange rate)
Conversion premium: (convertible price parity) / parity; also, the
extra amount an investor must pay to own shares via the convertible
Bond floor or investment value: net present value of the fixed cash
flows of the convertible, maximised for early redemptions (puts)
Convertibles: Definitions
Redemption
Price
Parity
Convertibles: Payoff, Value Profiles...
Source: Barclays Capital.
Convertible
Bond Value
Bond Floor
Stock Price
Recovery
Value
Stock Price
Why do companies issue convertibles (versus straight debt or equity)?
Convertibles: Issuers
Lower cost of capital depends on underlying stock performance
Valuation/pricing exploiting supply/demand imbalances
Market opportunity timing of deal based on stock price behaviour
Market profile tapping a broader investor base and gaining kudos
Accounting/credit treatment optimise by liaising with advisors,
rating agencies, etc, within constraints of jurisdiction regulations
21
Convertibles: Issuers

Convertible bonds ameliorate the agency conflict
amongst different class of investors.
Creditors prefer highly conservative policies
Stockholders prefer calculated risks due to their inherent call
option in their equity claim (if firm goes bust they have are
insulated by limited liability of 100% of their investment, versus
all of the upside less the present value of debt

Adding convertibility to debt makes creditors more like
stockholders and aligns their interest.

Consequently, CB holders require less compensation
(coupon/dividend yield)
Relative costs of straight debt, equity and convertible financing
Convertibles: Cost of Capital
Stock Growth Rate
Cost of
Straight Debt
Cost of
Equity
Cost of
Convertible
Who invests in convertibles and why?
Convertibles: Investors
Equity funds: downside protection and/or yield advantage
Fixed income and high-yield investors: equity upside, credit exposure
Outright convertible funds: valuation and bond-specific strategies
Hedge funds: leveraged option trading and hedging-based strategies
Risk-arbitrage/specialist funds: corporate activity, special situations
Retail investors: asymmetric payoff, return enhancement
24
Convertible bonds: Life Cycle
Issuance of the Convertible bond
Either private placement or public offer


Convertible bond life cycle
CB
ISSUE
Maturity and redemption or
Conversion to stock
Early Call
Put
Conversion
Early redemption
Poison put
Adjustments to
term sheet
25
Convertible bonds: Life Cycle
What is the effect on firm value of convertible bond issuance?
By how much can an investor expect convertible bond issuance to
affect stock prices?
What are the effects on firms stock value of calls of convertibles?
How should the call feature be included in the pricing of
convertibles?
26
Effect on firm value
In a perfect frictionless market (no taxes, no transaction costs, no
informational assymetries) the Miller-Modigliani theorem states that
firm value is invariant to capital structure, and in particular to the
mode of financing.

No advantage to debt financing over equity financing

And this would extend to hybrids such as convertibles.


27
Effect on firm value
Now lets relax the Modigliani-Miller assumptions
Information asymmetries between investors and managers
Assuming managers act to maximize common shareholder value, there
arise two scenarios
Managers perceive equity to be overvalued (issue more)
Managers perceive equity to be undervalued (share repurchase)

Miller and Rock (1985) argue that a firms decision to issue any
security is a negative signal
Assume management knows more about firms earnings than outside
investors
If management knows that a cash shortfall is imminent, it is in their
interest to issue external capital to compensate
28
Effect on firm value
Event studies by Dann and Mikkelson (1984), Eckbo
(1986), and Mikkelson and Partch (1986) report an
average mark down of -2% in equity over an average of
2-trading days after issuance announcement

However, compare this to pure equity issuance when the
price reaction is -3%, and straight bond and straight
preferred stock where the price reaction is imperceptible.

Add to this convertible bond flotation costs (underwriter
expenses and fees charged by SEC) which range from
2% for large issues to 9% for small issues (compared
with 2-4% for straight bonds)

29
Effect on firm value
Hence, while convertibles may present certain advantages to the
firm (reduced restrictions in the covenant structure, lower coupon
and/or dividend rates) they pay for these advantages with higher
flotation costs and equity value reduction.

It would seem there is no free lunch to be had for convertible
issuance!

30
Are CBs called late?
Perfect market settings
Firm acts to maximize common shareholder value, and by extension to
minimize the value of the convertible debt.
Brennan and Schwarz (1977, 1980) and Ingersoll (1977a)
Use equilibrium arguments to describe firms optimal call policy

Anomaly
Ingersoll (1997b) found marked deviations form this policy. His findings
are that firms delayed calling convertibles until their conversion price
exceeded call price by a mean of 83.5%


31
Delayed Call Anomaly
Various explanations have been advanced:

Firms prefer conversion to the need to refinance debt: Safety-net
Tax shield of coupons over dividends
Yield advantage of equity over coupons which may render it
desirable to defer call
Costs of issuing a call may make it desirable that the firm waits until
bondholders convert voluntarily (Constantinides and Grundy, 1987)

32
Delayed Call Anomaly
Signalling (Harrison and Raviv, 1985) -> managers should call and
force conversion if they think share prices will fall. Hence, issuing a
call sends a negative signal about firms earning prospects

A combination of studies in the literature supports the notion that the
reasons enumerated above (in particular the safety net and yield
advantage arguments) explain the delay in calling CBs. Once these
factors are taken into account, management calls CBs in a timely
manner.


33
Call and Firm Value
Capital structure decisions that reduce leverage almost always results
in a negative effect on equity values (Harris, 1986)
Lost debt tax shields
Adverse information which signals that managers doubt if the firm
will be able to meet fixed obligations in the future
Short run demand curve for shares is downward sloping
Called issue is converted into new shares, some of which are not held
by the investors, and the new supply meets restricted demand, pulling
prices down

Are these price reductions permanent?
Mazzeo and Moore (1996) on average equity prices revert and
exceed pre-call levels
Other studies have corroborated this.
34
Call and Firm Value
Are these price reductions permanent?
Mazzeo and Moore (1996) on average equity prices revert and
exceed pre-call levels
Other studies have corroborated this
Convertible and exchangeable bonds entitle holders to convert into
stock; may be redeemed early by issuers (calls) or investors (puts)
Convertibles: Summary So Far...
Terms and conditions are very detailed and important; prospectus!
Increasingly better-understood and popular asset class amongst
issuers and investors; largely due to their flexible, hybrid features
Payoff is like corporate bond with embedded option(s); valuation
and risk profiles reflect these hybrid features
Global convertibles market now has considerable breadth and depth
36
Valuation models
37
Convertible Bonds...
The simpler valuation models
What are the most important concerns in pricing?
Traditional Valuation Models
Extended Valuation Models
The more complex valuation models
Tradeoffs
Uses for arbitrage
Know thy neighbour
The numerical issues
The commensurate commercial issues
Merging valuation models and economies of scale
Non-Standard features

Modelling Convertibles...
When valuing a convertible we have the following decisions:
Type of
structur
e
More complex,
bespoke model
Exotic: e.g. multi
underlying assets,
resettables, etc
Standard model,
e.g. 1 or 2-factor
blended discount
Standard:
95% of liquid
universe
Simple bond plus
warrant or similar
Vanilla: no credit
risk, no calls/puts,
simple cash flows
Textbook analytic
formulae
PDE solution using
finite differences
Bi/trinomial trees
(risk-neutral pricing)
Monte Carlo (risk-
neutral pricing)
39
The Simpler Valuation models
Convertible Bond = Straight Bond + Warrant (Option)
Traditional Models: Bond + Warrant
Simplest model, was prevalent until the 1990s now only used for
physically separable bond-with-warrant notes with tradable parts
What is the strike of the warrant? early exercise shows it cannot be
a fixed number (e.g. the redemption price) must be the bond floor,
which is variable, so the warrant is no longer vanilla anyway
Bond + simple warrant gives wrong credit risk and other sensitivities
Calls and puts are options on the whole package poorly modelled,
e.g. better to treat provisional calls as barrier-like options
Value of convertible depends on one stochastic factor: stock price
Traditional Models: 1-Factor Model...
Let value of convertible be f when stock price is S at time t; f = f(S(t), t)
S(t) follows geometric Brownian motion given by the stochastic eqn:
dS(t) = S(t) dt + o S(t) dz
small change
in stock price
expected rate
of stock return
(annualised)
small time
interval
volatility of
stock price
(annualised)
Weiner process, i.e.
random variable with
mean 0, variance dt
Black-Scholes analysis leads to the usual partial differential equation
Traditional Models: Options Theory...
But numerical solutions needed due to early conversion/redemption
Set up a portfolio, P, of long the convertible and short a number, A
(multiplied by the conversion ratio), of shares: P = f A S
Itos lemma, together with A = f
S
, removes the dz term, so P must be
instantaneously risk-free: hence, dP = r P dt, leading to the usual PDE:
cf/ct + r S cf/cS + o
2
S
2
c
2
f/cS
2
= r f
This could be solved to give the value f(S, t) of the convertible
Numerical solution: e.g. finite difference grids for solving the PDE
Traditional Models: Finite Differences...
Replace derivatives of f with finite difference (discrete) approximations
For cf/cS and c
2
f/cS
2
, use central differences. For cf/ct, use either
- Implicit (backward): stable, but must solve algebraic equations
- Explicit (forward): can be unstable, but easy to solve
- Semi-implicit (central): stable and more accurate, but again,
must solve algebraic equations; also known as Crank-Nicholson
Discretise the stock price-time continuum to form a regular grid
Three finite difference schemes for solving PDEs:
Traditional Models: Finite Differences...
t = 0 t = maturity
S = 0
S = S

oS
ot
Implicit Explicit Semi-implicit
Numerical solution: e.g. binomial trees for risk-neutral valuation
Traditional Models: Binomial Tree...
Hold value at a non-terminal node:
f = e
r ot
[ p f
(u)

+ (1 p) f
(d)

]

With conversion, calls, puts, and coupons:
f = max{ min{ e
r ot
[ p f
(u)

+ (1 p) f
(d)

] +
cpns, call price }, S, put price }

S
f

S u
S d
p
1 p
f
(u)
f
(d)
ot
u = 1/d
Terminal nodes: f = max{ S, redemption price }
Thus far, issuers credit risk is neglected: how can we include it?
Traditional Models: Credit Risk...
Credit spread = excess yield on fixed cash flows above the risk-free
rate to compensate investors for default risk larger for riskier credits
Hence the present value of a cash flow C at time t with risk-free rate r
and credit spread h, is C e
(r + h) t
(r + h) is the risky discount rate
Two approaches for valuing convertibles: either (a) repeat the first-
principles analysis to derive a PDE; or (b) discount bond-derived cash
flows at the risky rate and equity-derived cash flows at the risk-free rate
(a) Full discount, or risky rates model
Traditional Models: Full Discount...
Disadvantage: assumes stock has no role in hedging against credit
deterioration; counter-intuitive, e.g. pricing other derivatives
Portfolio P is only risk-free with respect to stock price risk, not issuer
credit risk, so it must earn the risky rate of return, i.e. dP = (r + h) P dt
Hence, derive the same PDE, but with (r + h) in place of r
Solve numerically using finite differences, or via risk-neutral valuation
in a tree assuming both stock growth rates and discount rates are (r + h)
Advantage: neat derivation from first principles
(b) Blended discount model
Traditional Models: Blended Discount
f = f
E
+ f
B
, where f
E
is the value of equity-derived cash flows and f
B
is
the value of bond-derived cash flows latter discounted at (r + h)
Derive two PDEs: one for f
E
(S, t) as usual and one for f
B
(S, t) as in the
full discount model boundary conditions: at maturity, f
E
= 0 if S < red
amt and f
E
= S otherwise, f
B
= red amt if S < red amt and f
B
= 0 otherwise
Solve for f
E
, f
B
using finite differences or risk-neutral valuation in a tree
Advantage: intuitive, apportions credit risk by derivation of cash flows
Disadvantage: more difficult to derive from first principles

Extended Models: Quasi 2-Factors
Allow explicit dependence of credit spread on stock price, h = h(S)
Spread h is indirectly stochastic, hence this is a quasi two-factor model
h(S) should have the following properties:
h as S 0
dh/ds s 0 for S > 0
h h
min
(> 0) as S
Propose: h(S) = h
min
+ (h
0
h
min
) (S / S
0
)

k
, where k > 0
h
S
h
min
0
S
0
h
0
calibration
data point
Extended Models: The k Factor
k, the decay parameter, measures sensitivity of the credit spread to a
change in stock price; also, measures rate of collapse of the bond floor
Exact functional form of h(S) not so important as its properties
This extended model is simple to
implement in finite difference grids or in
bi/trinomial trees (e.g. using a blended
discount method)
k = 0 gives the traditional model
C
r
e
d
i
t

S
p
r
e
a
d
s

Stock Price
increasing k
Extended Models: Valuation Profiles
France Telecom 4% 2005 Value
(using a quasi-two factor binomial tree model)
0
20
40
60
80
100
120
140
0 20 40 60 80 100
Stock Price (EUR)
k = 0
k = 0.2
k = 0.5
k = 1
k = 2

Extended Models: Delta Profiles
France Telecom 4% 2005 Delta
(using a quasi-two factor binomial tree model)
0%
50%
100%
150%
200%
250%
0 20 40 60 80 100
Stock Price (EUR)
k = 0
k = 0.2
k = 0.5
k = 1
k = 2
Extended Models: Value vs Volatility
France Telecom 4% 2005 Value
(using a quasi-two factor binomial tree model)
80
90
100
110
120
130
140
0% 20% 40% 60% 80% 100%
Volatility
k = 0
k = 0.2
k = 0.5
k = 1
k = 2
Extended Models: Alternatives
Stochastic spreads: e.g. following a log-normal process (possibly with
some sort of mean reversion) leads to a fully two-factor model
Stochastic defaults: e.g. instantaneous defaults, maybe a Poisson
process, or with probabilities that depend on the prevailing stock
price leads to a more complex tree/grid algorithm
Firm valuation models: e.g. Merton/KMV-type, where equity value is a
call option on the value of the firms assets convertible forms part of
the firms debt until it is converted into equity, which in turn depends on
asset value leads to a much more complex model, with extensive yet
vital data inputs, e.g. balance sheet and presumed(!) bankruptcy level
Convertibles: Non-Standard Features
Resettables: conversion price resets on specific dates, depending on
the prevailing stock price; usually protects investor by resetting
downwards if stock price has fallen; e.g. Kudelski 2.25% 2009
Mandatories: usually convertible preference shares; popular in the
US; with automatic conversion into underlying stock at maturity, but
ratio may depend on prevailing stock price; e.g. Nortel 7% 2005
Best-of exchangeables: bond is convertible into one of several
shares at investors option; need a multi-factor model, e.g. Monte Carlo,
especially if > 2 underlyings; e.g. Swiss Re. Am. triples 2.25% 2004
Perpetuals: perpetual convertible bond with indefinite but capped
conversion option (via provisional call); e.g. Fortis FRN (Libor + 135bp)
Challenges for the simpler models
Convertibles on stocks that cannot be borrowed (e.g. Taiwan)
Corporate action, capital redistribution and change-of-control clauses
Volatility dynamics, e.g. strike-pinning/spikes, skew, etc.
Stochastic stock price or stochastic firm value models?
Blended or full discounting?
Non-standard features, e.g. resettables, perpetuals, best-ofs, etc.
Relationship between credit and equity many different possibilities!
57
A more complex, production level
model that will feature S, r, credit, FX,
and volatility risk
58
The Plan

The Convertible Bond (CB) pricing problem
The sources of risk
The curse of dimensionality
Pricing model
Contributions

A new parametric risk neutral distribution model for volatility
As stand-alone
Relevance and application to the CB pricing problem

Results


59
CBs- The risk factors to the Holder

Stock price risk
Continuous early exercise
Interest rate risk
Credit risk
Volatility risk
Foreign exchange rate risk
Issuers Parisian call trigger risk
Different flavour risk exchangeable or basket CBs, reverse
convertibles (reconvertible)

60
Pricing paradigms and convertible bonds

Multifactor Partial Differential Equations (PDEs)

Discretize CB pricing equation, construct as a multi-dimensional
grid, and solve on each node, evolving backward.


Capable of pinning down the early exercise (American) feature
analytically (front-fixing) or numerically (projection).


Curse of dimensionality

61
The framework

Quasi five factor coupled PDE system solved jointly
Use projective iterative solver (Gauss-Seidel with successive
over-relaxation) on pair of interdependent 3-dimensional PDE
grids
Unconditionally stable pricing routine
Accounts for equity, interest rate, credit, and FX risk. Is
consistent with volatility term structure and skew, via either of two
methods: bicubic spline interpolation or shifted lognormal
mixtures

Risks are:
Equity risk (cross-currency or domestic CB)
Combined with FX risk via change of measure

Interest rate risk (traditional one factor short rate model)

62
Risks

Credit risk

Framework extends Tsiveriotis and Fernandes (1998) to
unconditionally stable and multifactor world

We use an auxiliary asset called the COCB - Cash Only part of
Convertible Bond with associated coupled PDE

We use the implicit finite difference method versus trinomial trees to
achieve unconditional stability

Early exercise

We use projection on grid with Successive over-relaxation
algorithm: PSOR (Wilmott et. al. (1999), Tavella and Randall
(2000))

63
Volatility misspecification

If we use a single volatility to price CBs, which implied is it to
be?
Certainly dependent on the traders estimate of expected CB
lifetime (exercise, call, put, or redemption)

To remove this dependency we layer the PDE grid with a
volatility surface imputed from market option prices

Transform the observed smile to a functional form for
volatility (local volatility) using the Fokker-Plank equations
leading to:




(
(

|
.
|

\
|
c
c
+

+
|
.
|

\
|
c
c

c
c
c
c
+

+
c
c
=
+ +
2 2
2
2
2
2
S
C
d
t T S
1
C
S
C
t T d
S
C
S
S
C
(T)] 2S[r(T)
t T
C
T
C
2
T) (S,
64
Risks and Volatility
FX risk
Change probability measures to model foreign stock price in
base currency (Geman, 1995 and Bjork & Wiener, 1998)

Put and call is usually stipulated in terms of local currency,
and hence this approach fits the bill very well

Complications arise when we want to include volatility
surfaces in the analysis.
65
The coupled PDE with FX risk
CB PDE




COCB PDE





0 v r ru S
~
r w
,
S
~
r S
~
u
r w
r
2
1
)
,
2 (
2
1
S
~
S
~
u
ar) (
r
u
t
u
2
u
S
~
u
S
~
) (r
c
r , S
~
f X x S
2
f
2
X
2
2
2
2
2
2
2 2
S S x x S
2
x
s + +

c c
c
+
c
c
+
c
c
+ + +
c
c
+
c
c
+
c
c
0 )v r (r r w 2 S
~
r S
~
v
r w
r
v
2
1
S
~
v
S
~
) 2 (
2
1
S
~
) (r
S
~
v
ar) (
r
v
t
v
c
r , S
~
f X X f,
2
f
2
X
2
2
2
2
2
2
2 2
S S X X f,
2
X
s + o + +
c c
c
+
c
c
+
c
c
+ + +
c
c
+
c
c
+
c
c
Maturity 5.0 Years ConvRatio 1.2 Put Bermudan
Coupon 2% Volatility method Local Put (4yrs) US$ 95.12
Frequency Semi-ann Call Protection >3yrs @175 Put (2yrs) US$ 86.07
Spread 3.5% Model Price US$ 97.3653 S-r correl -0.3
ConvPremium 17.8% Price @ flat vol US$ 94.431 S(0) US$ 66.66
Share Currency GBP Conversion Price US$ 80.0 r(0) 0.06
Denominated USD Flat Vol 19% S_Max US$ 1200
What type of instrument can we price?

International Convertible Bond
with Local Volatility and Spread/Stock Price Relationship
65 . 0
0 0 c

0 0 c
) S S )( 00545 . 0 ) S ( r ( 00545 . 0 ) S S )( h h ( h ) S ( r


+ = + =
Call and Bermudan Put Features
Convertible Bond Term Sheet
67
Cross currency CB model prices
5
4.2
3.5
2.7
2
1.2
0.5
0
.
0
0
1
6
.
6
7
3
3
.
3
3
5
0
.
0
0
6
6
.
6
7
8
3
.
3
3
1
0
0
.
0
0
1
1
6
.
6
7
1
3
3
.
3
3
1
5
0
.
0
0
1
6
6
.
6
7
1
8
3
.
3
3
0
50
100
150
200
250
Convertible price
Time to maturity (years)
Stock price

68
Cross currency cash only part of convertible bond
5
4
.
6
4
.
3
4
3
.
6
3
.
3
3
2
.
6
2
.
3
2
1
.
6
1
.
3
1
0
.
6
0
.
3
0
.
0
0
3
3
.
3
3
6
6
.
6
7
1
0
0
.
0
0
1
3
3
.
3
3
1
6
6
.
6
7
2
0
0
.
0
0
0
20
40
60
80
100
120
COCB price
Time to maturity
Stock price

69
Volatility implementation for cross currency
Implement two models to account for volatility term structure and
skew

CB with FX risk requires double integration over the joint bivariate
lognormal densities of S and FX to extract scaled implied volatility
surface which is then transformed to local volatility on the grid

t) , S
~
( t) , S
~
( T) , K (
T) , K (
T) , K (
S
~
QP or FD
S
~

S
FX R R
2


| |
} }

'

'

+
|
|
.
|

\
|
=
=
K
K K
T r
T r imp
X(0))dX | (X(T) S(0))dS X(0), X(T); | (S(T) K) (S(T)X(T) e
(0) | K,0) (S(T)X(T) E e ) T, K, , S
~
C(
d
d
70
A new volatility model

The first approach assumed a continuum of traded strikes and
maturities and goes back to Breeden and Litzenberger (1978).
Major drawback is that the need to smoothly interpolate option
prices between consecutive strikes and maturities.


For domestic CBs, we develop a new parametric risk neutral
distribution model drawing on earlier work from Brigo and
Mercurio (2000). This results in alternative lognormal mixtures
asset price dynamics consistent with the chosen distribution.

71
Alternative dynamics
Find lognormal processes (i = 1, , N)


Derive the local volatility such that the Q
T
-density of S satisfies



The local volatility in this setup obeys (proof via Fokker-Plank):

+
=

=
2
2
2
1
0
2
1
2
2
2
1
0
2
2
) ( ln
) ( 2
1
exp
) (
1
) ( ln
) ( 2
1
exp
) (
1
) (
) , (
t V t
S
y
t V t V
t V t
S
y
t V t V
t
t S
i
i i
i
N
i
i
i i
i i

o
o
t
i
t i
i
t
i
t
dW S t v dt S dS ) , ( + =
{ } { }
i
t
N
i
i
i
t
T
N
i
i t
T
t
p y S Q
dy
d
y S Q
dy
d
y p

= =
= s = s
1 1
) (
72
Volatility
A new binomial tree model with adaptive time steps and discrete
variance shifts is developed and implemented.

Develop extensible template modeling framework with policy
classes (Alexandrescu, 2001) in C++ for very flexible, efficient
numerical routines.

Calibration leads to nonlinear least squares optimization with
nonlinear and linear constraints.

We also implement a quasi non-recombining trinomial variance
model but find it produces good fit but a poor term structure of
kurtosis.
73
0,0
1,0
11 week calibration interval with 3 option tenors (2, 6, and 11 week maturity FX calls/puts)
28/06/2002 - 12/07/2002
2 weeks
12/07/2002 - 09/08/2002
4 weeks
09/08/2002 - 13/09/2002
5 weeks
1,1
2,0
2,2
2,1
3,3
3,2
3,1
3,0
0 , 1
2 ~
c o + A
l
t
1 , 1
2 ~
c o + A
h
t
1 , 2
2 2
c o o + A + A
l h
t t
2 , 2 1 , 1
2 ~ ~
2 c c o + + A
h
t
0 , 2 0 , 1
2 ~ ~
2 c c o + + A
l
t
3 , 3 2 , 2 1 , 1
2 ~ ~ ~
3 c c c o + + + A
h
t
2 , 3
2 2
2 c o o + A + A
l h
t t
1 , 3
2 2
2 c o o + A + A
l h
t t
0 , 3 0 , 2 0 , 1
2 ~ ~ ~
3 c c c o + + + A
l
t
Cumulative Variance tree with adaptive time spacing (discrete shifts)
Nonlinear Least Squares optimization with Linear and non-Linear constraints (on the shifts)
Possible to specify functional form for epsilons
0 , 1 1 , 1 2 , 2
~
&
~
c c c >
74
0,0
1,0
11 week calibration interval for 3 option tenors (2, 6, and 11 week maturity FX calls/puts)
28/06/2002 - 12/07/2002
2 weeks
12/07/2002 - 09/08/2002
4 weeks
09/08/2002 - 13/09/2002
5 weeks
1,1
2,0
2,2
2,1
3,4
3,3
3,1
3,0
0 , 1
2 ~
c o + A
l
t
1 , 1
2 ~
c o + A
h
t
d
l h
t t
2 , 2 1 , 1
2 2 ~ ~
c c o o + + A + A
u
h
t
2 , 2 1 , 1
2 ~ ~
2 c c o + + A
0 , 2 0 , 1
2 ~ ~
2 c c o + + A
l
t
4 , 3 2 , 2 1 , 1
2 ~ ~ ~
3 c c c o + + + A
h
t
3 , 3
2 2
2 c o o + A + A
l h
t t
1 , 3
2 2
2 c o o + A + A
l h
t t
0 , 3 0 , 2 0 , 1
2 ~ ~ ~
3 c c c o + + + A
l
t
Cumulative Variance tree with adaptive time spacing (discrete shifts)
More Parameters and structure, better fit, less parsimonious
0 , 1 2 , 2
~
c c >
2,2
1 , 2 0 , 1
2 2 ~ ~
c c o o + + A + A
l h
t t
3,2
2 , 3 2 , 2 1 , 1
2 2 ~ ~ ~
2 c c c o o + + + A + A
l h
t t
Snap together
75
First 4 weeks of 11 week time interval
0,0
1,0
1,0
(1,2)
(2,5)
(3,9)
1,0
1,0
1,0
1,0
(1,0)
(2,0)
(3,0)
1,0
1,0
1,0
(1,1) (2,2)
(3,7)
(3,6)
(2,3)
(2,4)
(2,1)
(3,8)
(3,5)
(3,3)
(3,1)
(3,4)
(3,2)
1,0
1,0
1,0
1,0
1,0
1,0
Trinomial quasi non-recombining Cumulative Variance Tree
Possible to specify functional form for epsilons
Nonlinear Least Squares problem with Linear and non-Linear constraints
1

2 1
1
2

2
h
to A
2
l
to A
2
m
to A
2 2
l m
t t o o A + A
2
2
l
to A
2
2
h
to A
2 2
m h
t t o o A + A
2 2
l h
t t o o A + A
2
1

2
2

4
1
2
4 o
h
t A
) 1 ( 4 ), 3 (
2 1
3
1
2 2
o o + A
m h
t
4
2
2
4 o
l
t A
2
2 1
3
1
2 2
) 1 ( 6 ), 2 2 ( o o + A
m h
t
4 2
4
m m
t o A
28/06/2002 - 05/07/2002
1 week
05/07/2002-12/07/2002
1 week
2 2
2 2
m l
t t o o A + A
2
3
l
to A
2
3
h
to A
2 2
3
h l
t t o o A + A
2 2
3
m l
t t o o A + A
2 2 2
2
h m l
t t t o o o A + A + A
2 2
3
m d
t t o o A + A
2 2
2 2
h l
t t o o A + A
2 2 2
2
h m l
t t t o o o A + A + A
2 2
3
l h
t t o o A + A
2
2
m
to A
2 2
3
m h
t t o o A + A
2 2 2
2
h m l
t t t o o o A + A + A
76
Reference date: 28/06/2002
Maturities 12/07/2002 09/08/2002 13/09/2002
Market_Strikes Bid_vol Ask_vol Bid_vol Ask_vol Bid_vol Ask_vol
90 21.2 24.77 19.3 22.1 18.8 20.3
92 20.5 24.05 18.1 20.4 16.6 18.1
94 18.5 20 16.2 18.5 14.04 15.12
96 15.2 17.7 14.5 17 13.51 14.42
98 13.1 14.49 13.97 15.11 13.17 14.01
100 13.43 14.82 13.95 14.72 13.05 13.63
102 15 16.05 14.41 15.33 13.43 14.08
104 14.13 16.61 15.43 16.02 13.82 14.6
106 16.8 18.9 16.4 17.9 16 17.1
108 19.9 23.8 18.2 20.8 17.6 18.9
110 20.9 24.5 19.8 21.7 19.1 20.4
The slice being calibrated is (Yrs): 0.0383562 The slice being calibrated is (Yrs): 0.115068 The slice being calibrated is (Yrs): 0.210959
Strike BS Vol LNM Vol Strike BS Vol LNM Vol Strike BS Vol LNM Vol
90 0.22985 0.241897 90 0.207 0.215839 90 0.1955 0.201231
92 0.22275 0.218011 92 0.1925 0.196625 92 0.1735 0.181719
94 0.1925 0.189993 94 0.1735 0.176764 94 0.1458 0.163329
96 0.1645 0.162851 96 0.1575 0.15942 96 0.13965 0.1478
98 0.13795 0.14627 98 0.1454 0.148013 98 0.1359 0.137285
100 0.14125 0.14139 100 0.14335 0.143763 100 0.1334 0.132911
102 0.15525 0.145398 102 0.1487 0.14637 102 0.13755 0.134702
104 0.1537 0.159572 104 0.15725 0.155414 104 0.1421 0.142179
106 0.1785 0.18306 106 0.1715 0.169704 106 0.1655 0.154266
108 0.2185 0.208173 108 0.195 0.186614 108 0.1825 0.169169
110 0.227 0.22998 110 0.2075 0.20343 110 0.1975 0.185136

Gamma weighted calibration on volatilities adaptive TS
binomial with jumps


77
Lognormal mixtures calibration
0,0
1,0
11 week calibration interval with 3 option tenors (2, 6, and 11 week maturity FX calls/puts)
28/06/2002 - 12/07/2002
2 weeks
12/07/2002 - 09/08/2002
4 weeks
09/08/2002 - 13/09/2002
5 weeks
1,1
2,0
2,2
2,1
3,3
3,2
3,1
3,0
0 , 1
2 ~
c o + A
l
t
1 , 1
2 ~
c o + A
h
t
1 , 2
2 2
c o o + A + A
l h
t t
2 , 2 1 , 1
2 ~ ~
2 c c o + + A
h
t
0 , 2 0 , 1
2 ~ ~
2 c c o + + A
l
t
3 , 3 2 , 2 1 , 1
2 ~ ~ ~
3 c c c o + + + A
h
t
2 , 3
2 2
2 c o o + A + A
l h
t t
1 , 3
2 2
2 c o o + A + A
l h
t t
0 , 3 0 , 2 0 , 1
2 ~ ~ ~
3 c c c o + + + A
l
t
Cumulative Variance tree with adaptive time spacing (discrete shifts)
Nonlinear Least Squares optimization with Linear and non-Linear constraints (on the shifts)
Possible to specify functional form for epsilons
0 , 1 1 , 1 2 , 2
~
&
~
c c c >
Optimal parameters for Binomial Recombing NM Model
Unequal Time steps and non-stochastic vol
Calibration to 3 time slices
var_hi: 0.183837
var_low: 0.007
eps12: 0.000145
eps11: -0.00522
eps23: 0.000167
eps22: 0.005373
eps21: -0.006
eps34: -0.00022
eps33: -0.00207
eps32: 0.380146
eps31: -0.006
lambda: 0.085952
Nodal characteristics summary - time followed by other key values
Time = 0.038356
Annualized Vol Variance Weights
0.373784 0.005359 0.085952
0.119532 0.000548 0.914048
Sample stdDev: 0.158331
Sample XSKurtosis: 5.90017
Time = 0.115068
0.293849 0.009936 0.007388
0.376983 0.016353 0.157128
0.098562 0.001118 0.835484
Sample stdDev: 0.176308
Sample XSKurtosis: 7.26884
Time = 0.210959
0.262287 0.014513 0.000635
0.36992 0.401703 0.020258
0.210083 0.009311 0.215434
0.0783929 0.001296 0.763672
Sample stdDev: 0.229826
Sample XSKurtosis: 6.4692
78
Optimal parameters for Trinomial quasi Non-Recombing NM Model
with Equal Time steps and Stochastic vol
First Mixture Nodal Characteristics Summary- time followed by other key values
Time is: 0.0191781 Time is: 0.0575342 Time is: 0.115068
Annualized Vol Variance Weights Annualized Vol Variance Weights Annualized Vol Variance Weights
0.299038 0.00171498 0.849977 0.299038 0.005145 0.614075 0.299038 0.01029 0.377088
0.0856166 0.00014058 0.150023 0.249117 0.003571 0.325158 0.275212 0.008715 0.399342
0.0894286 0.00015338 0 0.186265 0.001996 0.057391 0.249117 0.007141 0.176212
0.0856166 0.000422 0.003377 0.219948 0.005567 0.041469
The StdDeviation 0.277683 0.249563 0.003583 0 0.186265 0.003992 0.00549
XSKurtosis 0.433619 0.186861 0.002009 0 0.144957 0.002418 0.000388
0.0869058 0.000435 0 0.0856166 0.000843 1.14E-05
Time is: 0.0383562 0.187455 0.002022 0 0.275414 0.008728 0
Annualized Vol Variance Weights 0.0881762 0.000447 0 0.24934 0.007154 0
0.299038 0.00342995 0.72246 0.0894286 0.00046 0 0.2202 0.005579 0
0.219948 0.00185555 0.255033 0.186563 0.004005 0
0.0856166 0.00028116 0.022507 The StdDeviation 0.277683 0.14534 0.002431 0
0.220705 0.00186835 0 XSKurtosis 0.14454 0.0862636 0.000856 0
0.0875433 0.00029396 0 0.249563 0.007167 0
0.0894286 0.00030675 0 0.220453 0.005592 0
0.186861 0.004018 0
The StdDeviation 0.277683 0.145722 0.002443 0
XSKurtosis 0.21681 0.0869058 0.000869 0
0.220705 0.005605 0
Time is: 0.0575342 0.187159 0.004031 0
Annualized Vol Variance Weights 0.146103 0.002456 0
0.299038 0.00514493 0.614075 0.0875433 0.000882 0
0.249117 0.00357053 0.325158 0.187455 0.004043 0
0.186265 0.00199613 0.057391 0.146483 0.002469 0
0.0856166 0.00042174 0.003377 0.0881762 0.000895 0
0.249563 0.00358333 0 0.146862 0.002482 0
0.186861 0.00200893 0 0.0888046 0.000907 0
0.0869058 0.00043453 0 0.0894286 0.00092 0
0.187455 0.00202173 0
0.0881762 0.00044733 0 The StdDeviation 0.277683
0.0894286 0.00046013 0 XSKurtosis 0.07227
First 4 weeks of 11 week time interval
0,0
1,0
1,0
(1,2)
(2,5)
(3,9)
1,0
1,0
1,0
1,0
(1,0)
(2,0)
(3,0)
1,0
1,0
1,0
(1,1) (2,2)
(3,7)
(3,6)
(2,3)
(2,4)
(2,1)
(3,8)
(3,5)
(3,3)
(3,1)
(3,4)
(3,2)
1,0
1,0
1,0
1,0
1,0
1,0
Trinomial quasi non-recombining Cumulative Variance Tree
Possible to specify functional form for epsilons
Nonlinear Least Squares problem with Linear and non-Linear constraints
1

2 1
1
2

2
h
to A
2
l
to A
2
m
to A
2 2
l m
t t o o A + A
2
2
l
to A
2
2
h
to A
2 2
m h
t t o o A + A
2 2
l h
t t o o A + A
2
1

2
2

4
1
2
4 o
h
t A
) 1 ( 4 ), 3 (
2 1
3
1
2 2
o o + A
m h
t
4
2
2
4 o
l
t A
2
2 1
3
1
2 2
) 1 ( 6 ), 2 2 ( o o + A
m h
t
4 2
4
m m
t o A
28/06/2002 - 05/07/2002
1 week
05/07/2002-12/07/2002
1 week
2 2
2 2
m l
t t o o A + A
2
3
l
to A
2
3
h
to A
2 2
3
h l
t t o o A + A
2 2
3
m l
t t o o A + A
2 2 2
2
h m l
t t t o o o A + A + A
2 2
3
m d
t t o o A + A
2 2
2 2
h l
t t o o A + A
2 2 2
2
h m l
t t t o o o A + A + A
2 2
3
l h
t t o o A + A
2
2
m
to A
2 2
3
m h
t t o o A + A
2 2 2
2
h m l
t t t o o o A + A + A
79
Conclusions and future directions
Hedging performance of the binomial term structure of
variance model


A Monte Carlo model for CBs with optimization routine to
approximate early exercise region
80

Appendices
81
0.0
16.7
33.3
50.0
60.0
66.7
73.3
83.3
100.0
133.3
0
0
.
1
0
.
2
5
0
.
5
1
1
.
5
2
3
5
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Volatility
Scaled strike (in USD)
Time (to maturity)

Implied volatility surface (scaled stock price)