Académique Documents
Professionnel Documents
Culture Documents
Norwegian University of Science and Technology, Department of Economics, Dragvoll NTNU, 7491 Trondheim, Norway
Norwegian School of Economics, Department of Business and Management Science, Helleveien 30, 5045 Bergen, Norway
Norwegian School of Economics, Department of Finance, Helleveien 30, 5045 Bergen, Norway
a r t i c l e in f o
abstract
Article history:
Received 31 August 2011
Received in revised form
31 October 2013
Accepted 5 November 2013
Available online 15 November 2013
We present a simple model for risky, corporate debt. Debtholders and equityholders have
incomplete information about the financial state of the debt issuing company. Information
is incomplete because it is delayed for all agents, and it is asymmetrically distributed
between debtholders and equityholders. We solve for the equityholders' optimal default
policy and for the credit spreads required by debtholders. Delayed information accelerates
the equityholders' optimal decision to default. Interestingly, this effect is small, implying
only a small impact on credit spreads. Asymmetric information, however, has a major
impact on credit spreads. Our model predicts high credit spreads for short-term debt, as
observed empirically in credit markets.
& 2013 Elsevier B.V. All rights reserved.
JEL classification:
G12
G33
Keywords:
Default policy
Credit spreads
Incomplete information
1. Introduction
The risk of monetary losses due to debt issuers who do not honor contractual debt payments is commonly referred to as
credit risk and explains the existence of credit spreads.
We present a theoretical model of credit spreads for corporate debt, where debtholders and equityholders have
incomplete information about the financial state of the company. The information is incomplete because the true state of the
issuer is only revealed with a time delay (delayed information). In addition, the information is asymmetrically distributed in
cases where debtholders and equityholders observe the true state with different time delays. The model is structural, and in
contrast to the seminal structural models, it predicts that also short-term credit spreads can be wide, in line with empirical
findings.
A (rational) default policy describes when the equityholders (rationally) choose not to service contractual debt payments.
We find that the length of the information delay for equityholders is not important for the default policy. The delay is
therefore not important for credit spreads either. The degree of information asymmetry, however, i.e., the difference in the
length of the information delay between debtholders and equityholders, is of crucial importance for credit spreads.
On a more general level, our paper addresses how incomplete information influences the pricing of bonds. We do not
consider noisy information, only delayed information. One can always argue that noisy information in many circumstances
Corresponding author.
E-mail address: snorre.lindset@svt.ntnu.no (S. Lindset).
0165-1889/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jedc.2013.11.006
99
is more common than delayed information. However, our main focus in this paper is on which aspects of information
structures that are important to obtain realistic models of credit spreads, and not on the realism of different information
structures.
One assumption of our model is that equityholders are better informed than debtholders. This assumption is based on
the idea that equityholders are closer to the day-to-day operations of the company than the debtholders, and, thus, receive
information earlier than debtholders. Although we can visualize cases where this may not be the situation (e.g., a company
may have passive owners), we find it plausible that the best informed equityholder is better informed than the best
informed debtholder. Debtholders must assess the value of the company based on less information than the equityholders,
but they rationally include the observation whether the company is bankrupt or not in their assessment.
In the special case where debtholders and equityholders have complete information, i.e., there is no delay in the flow of
information, our model simplifies to the classical Leland (1994) model. The current paper is also inspired by, and closely
related to, the seminal Duffie and Lando (2001) model of credit risk. Our model includes continuously observed, but delayed
information about the state variable, where the true state is immediately, i.e., without a delay, revealed upon bankruptcy.
The model of Duffie and Lando (2001) includes noisy (accounting) information released at discrete points in time. They
assume that equityholders have complete information and, thus, that only debtholders are subject to incomplete (noisy)
information. Their model, as ours, includes incomplete and asymmetric information, but does not explicitly cover the case
where all agents are subject to delayed information. We simplify the Duffie and Lando (2001) model by excluding noisy
(accounting) information, and extend it by explicitly exposing all agents to delayed information.
The main merit of our paper is that we identify information asymmetry, and not delay, noise, or other characteristics of
the information, as the important property for a simple and realistic model of corporate credit spreads. This insight also
refines the results by Choi (2008), who studies some aspects of delayed information (he uses the terminology lagged) in a
similar set-up.
Structural models were pioneered by Merton (1974). Merton models the value of a company's assets by a stochastic
process and debt and equity are considered as contingent claims on total asset value. Some of the papers in this tradition
include Black and Cox (1976), Geske (1977), Longstaff and Schwartz (1995), Leland (1994), and Duffie and Lando (2001).
In our model the default policy is expressed by an endogenous default barrier. Giesecke (2006) analyzes two classes of
models of imperfect information: (1) Models where the bankruptcy barrier is not observable to all agents.1 (2) Models with
incomplete information about the value of the company's assets. Our model belongs to his category (2), and according to his
Proposition 6.4, a default intensity exists in our model. Default intensities are important for reduced-form models. These
types of models were pioneered by Jarrow and Turnbull (1992), for extensions see e.g., Jarrow and Turnbull (1995), Jarrow
et al. (1997), and Schnbucher (1998).2 Papers analyzing technical aspects about credit risk and incomplete information
include Coculescu et al. (2008) and Guo et al. (2009). Other issues related to credit risk are analyzed in Rosen and Saunders
(2009), Huang and Yu (2010), and Azizpour et al. (2011).
Jarrow and Protter (2004) argue that the difference between structural and reduced form essentially is the assumption of
what information the modeler has access to. In their terminology, a model is structural if the modeler can observe the state
of the company, and reduced form if he cannot. They write (page 2): there appears to be no disagreement that the asset
value process is unobservable by the marketAlthough not well understood in terms of its implications, this consensus
supports the usage of reduced form models. Our results indicate that if different groups of agents have access to the same
incomplete information about the process, the error made by using a structural model compared to a reduced form model,
interpreted as in Jarrow and Protter (2004), is negligible.
The paper is organized as follows: In Section 2 we present our economic model. Section 3 presents optimal default policy
and credit risk valuation. Special cases with numerical examples are presented and analyzed in Section 4. Section 5
concludes the paper and gives suggestions for future research.
2. Economic model
This section presents our model of a company with incomplete information about the credit quality of its debt. Because
our focus is on default policy and debt valuation, we do not address whether debt is issued in an optimal way, i.e., whether
the capital structure of the issuer is optimal or not. Our model is standard, and we follow closely the set-up by Leland (1994)
and Duffie and Lando (2001).
Our model consists of two distinct groups of agents, equityholders and debtholders. In general, the two groups do not
have access to the same information and there is no information leakage between the two groups. Equityholders have at
least as much information as debtholders and constitute the group of better informed agents. The debtholders are the less
informed agents. To rule out the possibility that debtholders extend their information set by buying equity, we (as Duffie and
Lando, 2001) assume that equity is not traded. Furthermore, we assume that equityholders do not buy debt from the
company because debtholders could potentially extract information from such transactions. Also, it would alter the
1
100
equityholders' optimization problem that we analyze below. The company is run by the equityholders, i.e., the model does
not distinguish between owners and management. A decision to stop servicing debt and file for bankruptcy is endogenously
made by equityholders. Furthermore, all agents in the economy are assumed risk neutral and therefore discount future
cashflows by the constant, continuously compounded risk free interest rate r.
We assume that the only state variable is the stock of assets. It is given as the solution to the stochastic differential
equation
dSt St dt sSt dBt ;
S0 4 0;
where or, s, and S0 are constants. Here, the process B fBt gt Z 0 is a standard Brownian motion defined on a fixed, filtered
probability space ; F ; F0 ; P, where F0 fF t gt Z 0 . The process S fSt gt Z 0 is known as a geometric Brownian motion and St
is log-normally distributed. Also, P represents the objective probability measure. The information at time t is given by the
s-algebra F t . Here F t is generated by the process fSu ; 0 ru rtg. Thus, F0 represents the complete information filtration and
satisfies the usual conditions.
l
The information available at time t for the two groups of agents are given by s-algebras F m
t and F t . Superscripts m and l
signify more and less information. We denote the starting point in time of our economic analysis by t0. Define real numbers l
and m such that 0 rm rl rt 0 , and the s-algebras by
Fm
t F t m;
for all t Z t 0
and
F lt F t l ;
for all t Zt 0 :
t
t E
r
t
m
Observe that the present value Vt is just a multiple of St m . Either one of these quantities could therefore be used as the
m
state variable. The quantity Vt is sometimes called the unlevered value of the company and in our setting it represents the
equityholders' assessed value of the stream of dividends.
The rate of dividends paid at time t, St m , is observed by debtholders first at time t l m, alternatively, at time t they
observe dividends paid at time t l m. If debtholders could observe dividends at the time they are paid, they could
calculate the equityholders' assessed value of the company, and thereby eliminate the information asymmetry.
101
Smn
;
r
i.e., as the unlevered value of the company given full information at the time of bankruptcy. This value does not take any
effects of possibly optimally restructured debt at bankruptcy into account. Furthermore, at the time of bankruptcy, a
bankruptcy cost Smn =r , A 0; 1, proportional to the liquidation value of the company occurs.
As in Leland (1994), we assume that the company has issued perpetual debt with face value D. The debt is serviced by a
constant rate of coupon payments C. These payments are tax deductible (only interest is paid on perpetual debt). The tax
benefit rate is C, where is the tax rate. For a levered firm, i.e., a company with debt, the time t net dividend rate to equity
holders is St m 1 C, and can take both positive and negative values.
3. Optimal default policy and credit risk valuation
3.1. Bankruptcy wild card
In the event of bankruptcy, the debtholders claim the amount identical to the face value of the debt D. According to
absolute priority, debtholders' claims have priority over equityholders' claims.
Due to the information delay, there is a positive probability that the liquidation value of the company is more than
sufficient to cover debt and bankruptcy costs, i.e., the event
V mn V mn D 4 0
has positive probability. Any positive amount, in excess of debt and bankruptcy costs, is the property of the equityholders.
3
with payoff
By deciding to file for bankruptcy at time m
n , the equityholders obtain a bankruptcy wild card
1 =r Smn D . This payoff is similar to the payoff of a European call option on 1 =r units of the stock
of assets. It is straight forward to calculate its value using valuation theory for stock options.
From standard properties of geometric Brownian motions, the complete-information value of the stock of assets at the
m
m
bankruptcy time m
n , Sn , expressed as a function of the delayed value of the stock of assets observed by equityholders, Sn m ,
is given by
1=2s2 m sBm Bm m
Smn Smn m e
r
where
ln
z
1 Smn m
1
s2 m
2
r D
p
s m
This wild card has some resemblance to the wild card play that is present when trading the CBOT Treasury bond futures, see e.g., Hull (2012, p. 135).
102
Fig. 1. Illustration of information delay. Black line shows observed asset values given complete information, and grey line shows observed asset values
given delayed information.
m
bankruptcy decision at time m
n when they observe an asset value equal to W . With no delay, the equityholders would have
defaulted earlier, i.e., at time n when the asset value equals W (assuming that W W m ).
At any time t Z t 0 (assuming that the company is not bankrupt at time t) the equityholders face the optimal stopping
problem:
"Z m
#
t
m
e rv t Sv m 1 C dv e r t t Sm m jF m
St m sup E
5
t :
m t A T m
The first term inside the expectation operator in expression (5) is the discounted value of the dividends, net of after-tax
coupon payments. The second term is the present value of the bankruptcy wild card. There are three differences between
the optimization problem in expression (5) and the standard complete information optimization problem, see e.g., Duffie
(2001), Chapter 11.C. The first is the inclusion of the bankruptcy wild card in the optimization problem. Second, the delayed
state variable St m enters, and third, the optimization is based on less information (F m
t ) than the standard case with
complete information (F t ).
The optimal stopping problem with delayed information can be transformed into an optimal stopping problem with nondelayed information (see ksendal, 2005). The relationship between the optimal stopping time n from the non-delayed
m
problem and the optimal stopping time m
n from the delayed problem is given by n n m. We can then write
Z t m
e rv t m Sv 1 C dv:
St m sup E
t m A T
tm
6
e rt m t m S jF t m :
By substituting u t m in the problem (6) we obtain
Z u
e rv u Sv 1 C dv e ru u S jF u ;
Su sup E
u A T
for u Zt 0 m. We recognize expression (7) as a standard optimal stopping problem. To summarize, the equityholders'
optimization problem with delayed information has been transformed into an equivalent optimization problem with nondelayed information. The latter problem can be solved using standard methods.
The solution to problem (7) satisfies the HamiltonJacobiBellman equation
ss 12 s2 s2 ss r s 1 C 0;
where s St , and subscripts denote partial derivatives, i.e., s s=s, ss s=s , together with the boundary
conditions
2
W m W m
and
s W m W m ;
10
where is given in expression (4) and W m denotes the derivative of with respect to the state variable, evaluated at
the point Wm. Eqs. (9) and (10) are known as the value matching and the high contact (or smooth pasting) conditions,
respectively.
Before we introduce two technical conditions, we first define
x
x
C1
p
1 m
D
e Nz 1
x
N z s m ;
r
1 C
1 C
where
ln
z
103
1 x
1
s2 m
r D
2
p
s m
and o 0 is a constant defined below. Here, x is essentially a weighted sum of the value of the bankruptcy wild card from
expression (4), , and its derivative.
Condition 1. Assume that 1 em Nz o j 1j and that W m o=r 1=r .
Condition 2. Assume that 1 em Nz 4 j 1j and that W m 4=r 1=r .
Loosely speaking, Condition 1 is relevant for (relatively) small values of the delay m, whereas Condition 2 may be
relevant for larger values of m.
In Proposition 2 we state the solution to problem (7).
Proposition 2. Assume that either Condition (1) or (2) is satisfied. Problem (7) has the following solution: the optimal stopping
time is given by
n u inf ft Z u : St r W m g
with respect to the filtration Fu . The value of equity is
8
m
s
s
C
s
>
< s W
1
1
Wm
m
m
m
r
W
W
s r r W
>
:0
where
1 2
s
2
for s Z W m ;
11
for s o W m ;
s
2
1
s2 2rs2
2
o0;
s2
0:
1 W m
r
m
12
r 1 C
:
1
r
13
104
Appendix C. Throughout the paper we analyze one such zero-coupon bond with fixed maturity. The bond matures at time T,
m
with recovery function Rm
n ; T in the case of default at time n o T. The price of the bond consists of two parts:
1. The discounted value of the principal paid at maturity.
2. The discounted value of the recovery payment in case of default.
More formally, the time t price of a bond maturing at time T Z t is the conditional expected discounted payoff, i.e.,
r
t; T Ee rT t 1fm
n t 4 Tg e
rT t
t
l
Rm ; T1fm
n t rTgjGt
Pn t 4 TjGlt 1 ;
m
Pn t 4 TjGlt
m
14
4
where we interpret
as the survival probability of the company until time T. In the second equality, we have
used the recovery function Ru; T 1 e rT u , u A t; T, i.e., the same recovery function as in Duffie and Lando (2001).
This recovery function facilitates analytical solutions of bond prices and is therefore used throughout the paper.
For a credit risky zero-coupon bond, the credit spread l;m is defined as the excess yield compared to the yield on a
riskfree zero-coupon bond. From expression (14) we have that
e r
l;m T
t
l
e rT t Pm
n t 4TjGt 1 ;
l;m
l
lnPm
n t 4 TjGt 1
:
T t
so
15
l
Notice that the credit spread vanishes as -0 and tightens as the survival probability Pm
n t 4 TjGt increases.
105
Table 1
Base-case parameters.
St 0 m
C
D
100
0.035
0.08
0.045
0.3
0.3
0.5
13
90
Spread
0.175
0.150
0.125
0.100
0.075
0.050
0.025
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
2.25
2.50
2.75
3.00
Fig. 2. Credit spreads base case with complete information. The figure shows credit spreads, expression (15), for zero-coupon bonds with up to 3 years to
maturity. The tax rates are 20% (widest spreads), 30%, and 40% (tightest spreads).
16
106
Pm t 4T 4 T \ M
l
t l;t m 4 W jF t
l
n
P m
n t 4 TjGt
m
l
PM t l;t m 4 W jF t
PM t l;T m 4W m jF lt
PM t l;t m 4 W m jF lt
T m t l; lnW m =St l
;
l m; lnW m =St l
17
0.08
0.06
0.04
0.02
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
2.25
2.50
2.75
3.00
Fig. 3. Credit spreads general case. The figure shows credit spreads, expression (15), for zero-coupon bonds with up to 3 years to maturity. The information
delays m are 0.1 (widest spreads), 0.2, and 0.3 (tightest spreads) and l 0.4. The lower, dotted line represents the complete-information case.
100
100 W
90
90
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
0.0
0.5
1.0
1.5
2.0
m
2.5
3.0
3.5
0.0
4.0
m
4.0 (W)
4.0
3.5
3.5
3.0
3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.5
1.0
1.5
2.0
m
2.5
3.0
3.5
4.0
107
0.5
1.0
1.5
2.0
m
2.5
3.0
3.5
4.0
1.0
1.5
2.0
m
2.5
3.0
3.5
4.0
(W )m
0.0
0.5
Fig. 4. Optimal default barrier Wm and the value of the bankruptcy wild card W m for different levels of information delay m and volatility s for two
different levels of the bankruptcy parameter . The s values range from 0.2 to 0.7 with intervals of 0.1. (a) The optimal default barrier Wm as a function of m
for 0:5. Lower graph is for higher volatility. (b) The optimal default barrier Wm as a function of m for 0:3. Lower graph is for higher volatility. (c) The
value of the bankruptcy wild card W m as a function of m for 0:5. Higher graph is for higher volatility. (d) The value of the bankruptcy wild card W m
as a function of m for 0:3. Higher graph is for higher volatility.
18
where the expression for ; is given in expression (A.1) in Appendix A. By comparing this expression with the
corresponding expression for the case of full information, the only way symmetric, but delayed information, can affect credit
spreads is through a change in the default barrier Wm, i.e., if W m aW.
In Fig. 5, the credit spreads for the four cases are plotted for three different assumptions about the lengths of the delays.
In parts (a)(d) the credit spreads for the cases complete information and symmetrically delayed information are not
distinguishable. The same is true for the cases DuffieLando and the general case. Comparing the plots (c) and (d) to the
plots (a) and (b), we clearly see that a higher degree of asymmetric information leads to wider credit spreads. Also, by
comparing the plots (b) and (d) to the plots (a) and (c), it is clear that a lower bankruptcy cost parameter tightens credit
spreads, cf. the definition of credit spreads in expression (15).
To visualize different credit spreads for the four cases, we must increase the information delay significantly (to m 2 in
the example), see parts (e) and (f) in Fig. 5. The reason for different spreads in the different cases is that the delay m is so
long that Wm is sufficiently different from W to also affect credit spreads. Note in particular in part (f) of the figure how
credit spreads in the symmetrically delayed case are tighter than the spreads in the DuffieLando case for m o0:5 and wider
for m 4 0:5.
5. Conclusions and suggestions for future research
In this paper we have proposed a new model for incorporating delayed and asymmetric information between
debtholders and equityholders. We articulate the effect of delayed information on shareholders' endogenous decision to
108
0.12
Spread
0.12
0.10
0.10
0.08
0.08
0.06
0.06
0.04
0.04
0.02
0.02
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00
0.12
Spread
0.12
0.10
0.10
0.08
0.08
0.06
0.06
0.04
0.04
0.02
0.02
0.12
Spread
Spread
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00
Spread
0.12
0.10
0.10
0.08
0.08
0.06
0.06
0.04
0.04
0.02
0.02
Spread
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00
Fig. 5. Examples of credit spreads for the four cases. In plots (a)(d), the widest spreads are for the two cases with asymmetric information. In plots (e) and
(f) (for m o 0:5), the widest spreads are for the general case, followed by the DuffieLando case, the symmetrically delayed information case, and the
tightest spreads for the case with complete information. (a) No information delay, moderate information asymmetry, s 0:3, l m 0:2, m 0, 0:5.
(b) No information delay, moderate information asymmetry, s 0:3, l m 0:2, m 0, 0:3. (c) Moderate information delay, moderate information
asymmetry, s 0:3, l m 0:4, m 0.2, 0:5. (d) Moderate information delay, moderate information asymmetry, s 0:3, l m 0:4, m 0.2, 0:3.
(e) Long information delay, moderate information asymmetry, s 0:3, l m 0:2, m2, 0:5. (f) Long information delay, moderate information
asymmetry, s 0:3, l m 0:2, m 2, 0:3.
default. In particular, for realistic parameter values we show that incomplete information to equityholders about the true
stock of asset value only has a small effect on their decision to default on the loan payments. Any effect is likely to accelerate
a default. The decision to default is accelerated because by defaulting, equityholders receive a bankruptcy wild card with
non-negative value. This wild card gives the equityholders a valuable alternative to continued operation of the company.
If both debtholders and equityholders have access to the same delayed information, the only reason for changed credit
spreads is a potential change in equityholders' optimal default policy, compared to the complete-information case. For
realistic parameter values, these changes are small. Furthermore, we find that asymmetric information between debtholders
and equityholders is important for credit spreads, far more important than delayed, symmetrically distributed information.
Increased information asymmetry leads to wider credit spreads. Our model produces short-term credit spreads more in line
with empirical observations than most standard structural models of credit risk.
The results in this paper have empirical testable implications: Do companies where there is likely to be more asymmetric
information between debtholders and equityholders pay higher interest rates on their loans? Do companies where there is
109
more uncertainty about asset values, i.e., a higher degree of incomplete information, default earlier than other companies?
One indication that may lead to a confirmative answer to the last question is if equityholders tend to receive payments from
the bankruptcy wild card more often than equityholders of companies with a lower degree of incomplete information. Our
model also predicts that companies with high bankruptcy costs, for instance because of relatively illiquid assets, wait longer
before they default. A typical reason for illiquid assets is a high degree of asset specificity.
For future extensions of the results in this paper, it would be interesting to include management as a third group of
agents. Management would always belong to the better informed group. With three groups of agents, we could for instance
assume that debtholders and equityholders both belong to the less informed group of agents. With this assumption, we
could extend the analysis to also include companies whose equity is traded in a financial market. Unfortunately, this
assumption makes the model harder to solve.
Acknowledgement
The authors would like to thank Fred Espen Benth, Carl Chiarella, Darrel Duffie, Hans Marius Eikseth, Steinar Ekern, Chris
Florackis, Nadine Gatzert, Kay Giesecke, Jrgen Haug, Kristian Miltersen, Aksel Mjs, Jril Mland, yvind Norli, and
Per stberg. In particular, thorough comments from anonymous referees have substantially improved the paper. Earlier
versions have been presented at faculty seminars at Trondheim Business School, the Norwegian School of Economics,
Princeton University, the University of Stavanger, Norwegian University of Science and Technology, Department of
Economics, European Financial Management Association Annual Meeting in Athens 2008, European Group of Risk and
Insurance Economists Meeting in Toulouse 2008, and at Workshop on Innovations in Stochastic Analysis and Mathematical
Finance Norwegian School of Economics 2013.
Appendix A. Survival probability
Consider a geometric Brownian motion with dynamics as in expression
(1) with initial value S0, and a barrier sb o S0 .
Consider also the arithmetic process with dynamics dX t 12 s2 dt s dBt , starting at X 0 0. The first time the process S
hits sB is equivalent to the first time Xt hits x lnsB =S0 . The probability for the process S of not crossing the barrier sb in a
time period of length v is identical to the probability for the process X fX t gt Z 0 of not crossing the barrier x lnsB =S0 in a
time period of length v and is
x v
x v
2
p
p ;
v; x N
e2x=s N
A:1
s v
s v
where 12 s2 , see e.g., Musiela and Rutkowski (1997, Corollary B.3.4) .
Appendix B. Proof of Proposition 2
To prove that the solution of expression (11) is optimal, we follow the approach in Duffie and Lando (2001).
The function s in expression (11) (for s Z W m ), with Wm implicitly given in expression (12), satisfies Eq. (8) and the
boundary conditions (9) and (10).
We now consider as a function of s, where s has dynamics given in expression (1). As Duffie and Lando (2001), we
apply It's lemma to s to get
ds s s 12 s2 s2 ss dt s ss dBt :
We define
qt; St e rt t0 m St
t
t0 m
e rv t0 m Sv 1 C dv:
B:1
m
The dt-term in the above expression is zero from the HamiltonJacobiBellman equation (8) for s ZW . We now show
that both conditions (1) and (2) imply that the dt-term of Eq. (B.1) is non-positive for s o W m . In this case s ss 0.
The dt-term of Eq. (B.1) simplifies to s 1 C. Now, s 1 C r0 for all s o W m if W m 1 C r0. Inserting Wm from
expression (12) and simplifying we get the sufficient condition
Wm
r 1 C
r 1:
B:2
1 W m
r
110
1 C
where is defined in Section 3.2. Assume now that W m =4 j 1j=r . Then (B.2) can be written as
1
W m W m
r
Wm :
r r
1 C
Either of these conditions, called conditions 1 and 2 in Section 3.2, is sufficient to ensure that the dt-term of expression (B.1)
is non-positive.
Rt
To show that the dBt-term of expression (B.1) is a martingale, we show that Y t t0 s St sSt dBt defines a martingale
with respect to the filtration Ft0 m . From expression (11) we calculate s St and find that
s St St ASt BSt ;
where the constants A =r and B W m =r . Here, Yt is a martingale if
Z T
Z T
Z T
ASt BSt 2 dt A2 E
S2t dt 2ABE
S1t dt
E
t0
t0
Z
B2 E
T
t0
t0
S2
t dt o 1:
B:3
q
RT
It is well known that E t0 X 2t dt o1 for Xt log-normally distributed. Here, St, S1t , and St are all log-normally distributed.
Hence, each of the three expectations on the right-hand side of expression (B.3) is well defined and Yt defines a martingale
with respect to the filtration Ft0 m .
From the above arguments it follows that qt; is a super-martingale with respect to Ft0 m , i.e.,
qt 0 m; Z EqU; jF t0 m , for any stopping time U A T . Recall that T is the set of all Ft 0 m stopping times.
We calculate
Z n t 0 m
e rv t0 m Sv 1 C dv e rn t 0 m t0 m Sn t0 m jF t 0 m
E
t0 m
St 0 m
St 0 m W m St0 m
C
St 0 m
1
1
W m
m
m
m
r
r
r W
W
W
St0 m :
The first equality follows from the definition of n t 0 m in Proposition 2 and calculations.
At any point in time u, the continuation value Su must be at least as high as the value of stopping at u, Su , so
Su Z Su :
B:4
t0 m
U
t0 m
e rv t0 m Sv 1 C dv e rU t0 m SU jF t0 m :
The two equalities follow from the definitions of q; and St 0 m in expression (11). The first inequality is due to the fact
that qt; is a super-martingale with respect to Ft0 m . The second inequality follows from the inequality (B.4).
Thus, we have verified that the candidate solution (11) is optimal.
Appendix C. Connection between perpetual debt and zero-coupon bonds
In this appendix we assume complete information and show one example of a portfolio of a continuum of zero-coupon
bonds which has the same value as a perpetual debt contract.
Consider perpetual debt with coupon rate C to be paid until the company defaults. The company defaults on its debt
payments the first time the stock of assets in Eq. (1) hits the default barrier W from above. The time of default is given by the
stopping time
n inf ft : St rWg;
tZ0
111
The parameter determines bankruptcy costs, i.e., bankruptcy costs are =r W. The value of the debt at time 0 can
be calculated as
Z n
1
W
C:1
E
Ce rs ds e rn
r
0
with solution
C
C 1
S0
W
;
r
r
r
W
where is given in Proposition 2, cf. Black and Cox (1976).
Observe that we can also write expression (C.1) as
Z n
1
1
W e rs ds
W:
C r
E
r
r
0
C:2
The natural interpretation of expression (C.1) is the sum of the present value of coupon payments until default and the
present value of the recovery amount 1 =r W upon default.
Expression (C.2) suggests that the recovery amount 1 =r W may instead be paid at time 0, but where interest
payment for this amount has to be deducted from the coupon C until default, without changing the overall value of the debt.
Our next step is to securitize the perpetual debt into a continuum of zero-coupon bonds. Let NT be the number of zerocoupon bonds maturing at time T, and let NT N for all T A 0; 1. The time 0 value of a continuum of N zero-coupon bonds
with expiration at time T and with general recovery function Rn ; v if o v is
Z n
Z 1
Ne rt dt NE e rn
Rn ; t dt :
0 N E
0
The first term represents the present value of the zero-coupon bonds which expire before default. The second term
represents the present value of the recovery amounts of the unexpired bonds upon default.
Consider now the particular recovery function used in this paper, Rn ; v 1 e rv n . In this case the above
expression simplifies to
Z n
Z 1
Ne rt dt N1 E
e rt dt
0 N E
n
Z 0n
1 rn
E e
Ne rt dt N
E
r
Z0 n
1
rt
:
C:3
E
Ne
dt N
r
0
Consider now a portfolio composed of two parts, a continuum of N C r1 =r W= zero-coupon bonds and a
time 0 bank deposit (equivalently, the bank deposit can be zero-coupon bonds maturing at time 0) of
1 =r W 1 =C=r 1 =r W. The time 0 value of this portfolio follows from expression (C.3) and is
Z n
1
1 C 1
W e rt dt
W
0 E
C r
r
r
r
0
1
1 C 1
W
W
r
r
r
Z n
1
1
W e rt dt
W;
E
C r
r
r
0
which is identical to the time 0 value of perpetual debt from expression (C.2).
References
Altman, E., Resti, A., Sironi, A., 2004. Default recovery rates in credit risk modelling: a review of the literature and empirical evidence. Econ. Notes 33 (2),
183208.
Azizpour, S., Giesecke, K., Kim, B., 2011. Premia for correlated default risk. J. Econ. Dyn. Control 35 (8), 13401357.
Bielecki, T.R., Rutkowski, M., 2002. Credit Risk: Modeling, Valuation and Hedging. Springer-Verlag, Berlin-Heidelberg.
Black, F., Cox, J.C., 1976. Valuing corporate securities: some effects of bond indenture provisions. J. Financ. 31 (2), 351367.
Choi, J., 2008. Credit risk model with lagged information. J. Deriv. 16 (2), 8593.
Coculescu, D., Geman, H., Jeanblanc, M., 2008. Valuation of default-sensitive claims under imperfect information. Financ. Stoch. 12, 195218.
Collin-Dufresne, P., Goldstein, R.S., Helwege, J., 2010. Is Credit Event Risk Priced? Modeling Contagion via the Updating of Beliefs, Working Paper, National
Bureau of Economic Research WP15733.
Duffie, D., 2001. Dynamic Asset Pricing Theory. Princeton University Press, Princeton, New Jersey.
Duffie, D., Lando, D., 2001. Term structures of credit spreads with incomplete accounting information. Econometrica 69 (3), 633664.
Duffie, D., Singleton, K.J., 2003. Credit Risk, 3rd ed. Princeton Series in Finance, Princeton, New Jersey.
Geske, R., 1977. The valuation of corporate liabilities as compound options. J. Financ. Quant. Anal. 12, 541552.
Giesecke, K., 2006. Default and information. J. Econ. Dyn. Control 30 (11), 22812303.
Giesecke, K., Goldberg, L.R., 2004. Forcasting default in the face of uncertainty. J. Deriv. 12 (Fall), 1125.
Guo, X., Jarrow, R.A., Zeng, Y., 2009. Credit risk models with incomplete information. Math. Oper. Res. 34 (2), 320332.
112
Huang, S.J., Yu, J., 2010. Bayesian analysis of structural credit risk models with microstructure noises. J. Econ. Dyn. Control 34 (11), 22592272.
Hull, J.C., 2012. Options, Futures and other Derivatives, 8th ed. Pearson Essex, CM20 2JE, England.
Jarrow, R., Lando, D., Turnbull, S., 1997. A markov model for the term structure of credit risk spreads. Rev. Financ. Stud. 10 (2), 481523.
Jarrow, R., Protter, P., 2004. Structural versus reduced form models: a new information based perspective. J. Invest. Manag. 2 (2), 110.
Jarrow, R., Turnbull, S., 1992. Credit risk: drawing the analogy. Risk Mag. 5 (9), 5385.
Jarrow, R., Turnbull, S., 1995. Pricing derivatives on financial securities subject to credit risk. J. Financ. 50 (1), 5385.
Leland, H.E., 1994. Corporate debt value, bond coventnats, and optimal capital structure. J. Financ. 49 (4), 12131252.
Longstaff, F.A., Schwartz, E., 1995. A simple approach to valuing risky debt. J. Financ. 50 (3), 789821.
Merton, R.C., 1974. On the pricing of corporate debt: the risk structure of interest rates. J. Financ. 29 (2), 449470.
Musiela, M., Rutkowski, M., 1997. Martingale Methods in Financial Modeling. Springer Verlag, Berlin Heidelberg.
ksendal, B., 2005. Optimal stopping with delayed information. Stoch. Dyn. 5 (2), 271280.
Rosen, D., Saunders, D., 2009. Analytical methods for hedging systematic credit risk with linear factor portfolios. J. Econ. Dyn. Control 33 (1), 3752.
Schnbucher, P., 1998. Term structure modelling of defaultable bonds. Rev. Deriv. Res. 2 (23), 161192.