Credit Risk Premia and a Link to the Equity Premium

Tobias Berg, Christoph Kaserer

Draft, this version: October 08, 2007

Abstract
While the equity premium is - both from a conceptual and empirical perspective - a
widely researched topic in finance, the analysis of credit risk premia has only recently
attracted wider attention. Credit risk premia are usually measured as the multiple
between risk neutral (derived from bonds or CDS spreads) and actual default probabilities (derived from ratings of rating agencies or market implied rating as Moodys
KMV or Altmans Z-Score), which we will call Q-to-P-ratio. Empirical studies have
though shown, that these Q-to-P-ratios are not simply a measure of risk aversion, but
also depend on factors such as credit quality and industry sector.
In this paper, we propose a different measure for extracting risk premia out of credit
valuations which is based on structural models. This approach is able to - qualitatively
- explain the observed variations in the Q-to-P-ratio from empirical studies and has
several advantages: First, it is only based on observable parameters; second, it is consistent with classical portfolio theory; third, it is robust with respect to model changes
(besides the classical Merton model we examine the Duffie/Lando (2001) model with
unobservable asset values and deviations from the log-normal assumption) and fourth
- and most importantly - it directly yields the market sharpe ratio and therefore allows
for a direct comparison with the equity premium.
Based on an CDS spreads of the 125 most liquid CDS in the U.S. from 2003 to 2007,
we show that appr. 80% of the CDS spreads can be explained by credit risk based
on structural models with unobservable asset values. We derive an average implicit
market sharpe ratio of appr. 40%, adjusting for taxes yields an average market sharpe
ratio of appr. 30%. This confirms research on the equity premium, which indicates,
that the historically observed sharpe ratio of 40-50% (corresponding to an equity premium of 7-8% and a volatility of 15-20%) was partly due to one-time effects.

Keywords: credit risk premium, equity premium, credit risk, structural models of default

Tobias Berg, Department of Financial Management and Capital Markets, Munich University of Technology

Prof. Christoph Kaserer, Head of the Department of Financial Management and Capital Markets,
Munich University of Technology

1 INTRODUCTION

Introduction

Risk premia in equity markets are a widely researched topic. The risk premium in equity
markets is usually defined as the equity premium, e.g. the excess return of equities over risk
free bonds. Adjusting for different volatilities yields the sharpe ratio, i.e. the excess return
divided by the overall market volatility. Generally, the measurement of equity premia can
be dividend into three main approches: Models based on historical averages, dividend or
cash-flow-discount models and models based on utility functions. While historical averages
have long dominated theory and practical applications, current research suggest an upward
bias, e.g. the ex post realized equity returns do not correctly mirror the ex ante priced equity
premium.1 Dividend-/Cash-flow-discount models have become more popular, but are also
subject to debate, especially for their rather high sensitivity to growth/dividends/earnings
forecasts. While approaches based on utility functions have been subject to intensive debate
in the academic literature2 , its use in practical applications is currently of minor importance.
Risk premia are though not limited to equity markets, the same logic does also apply to
credit markets in theory and it has been observed in empirical studies. For example, the
average 5-year CDS spread for a A-rated obligor in the CDS.NA.IG-index over the past 3
years has been 36 bp, whereas the average annual expected loss over 5 years is less than 10
bp, e.g. 5 year CDS investments (approximately) yield an average return of appr 26 bp over
the risk free rate, see figure 1. In absolute terms, this premium increases with decreasing
credit quality (i.e. the expected net returns increase with increasing riskyness). Measured
relative to the expected loss (or the actual default probability) it decreases with decreasing
credit quality. Over the last year, research about this default risk premium has developed,
but there has not yet emerged consensus on the methodology for measuring this default
premia.
Duffie and Singleton (1999) model the default event via default intensity processes as
an inaccesible stopping time. This approach has gained popularity especially for hedged
based pricing applications restricted to debt markets. Assuming a recovery of 0%, a risky
zero coupon bond can simply be priced via a simple additional discount factor to the risk
free interest rate, which represents the loss in survival probability over the next instance of
time, i.e.
RT
B(T ) = E Q [e 0 r(s)+(s)ds ],
where r captures the risk free rate dynamics whereas captures the risk neutral default
intensity dynamics.
Fixing the risk attitude of investors, it is though not clear if - when examinign different obligors - the absolute default risk premium should be a fixed percentage of the actual
1

Reasons are survivorship bias, risk premium volatility, interest rate level and state of the economy, see
for example Claus/Thomas (2001), Illmanen (2003), Fama/French (2002).
2
This debate is mainly based on the so called Equity Premium Puzzle by Mehra/Prescott (1985),
see Mehra (2003) for an overview about different utility based approaches including alternative preference
structures, disaster states and survivorship bias and borrowing constraints.

1 INTRODUCTION

Figure 1: Relationship between 5-year CDS spreads (mid) and 5-year annualized expected
loss per rating grade for investment grade ratings.

default probability (i.e. Q = c P ), a fixed amount (in bp) independent of the rating
grade/actual default probability (i.e. Q = P + c) or if other (e.g. logarithmic) relationships are to be preferred. Many researches opt for a fixed multiple3 , i.e. they examine
the relationship between risk neutral and real world default probability or default intensity,
Q
e.g. P , which we will denote throughout this paper as Q-to-P-ratio4 . Berndt et.al. (2005)
come to the conclusion, that the Q-to-P-ratio is higher for high quality firms and lower for
low quality firms. They use a linear and a log/log relationship between CDS spreads and
default probabilities, resulting - especially for the linear relationship - in an intercept of
roughly 50 bp, which is more 30 times the standard error and does not seem to be plausible even accounting for liquidity risk. Amato and Remolona (2005) principallly confirm
these findings. In addition, Amato (2005) finds a significantly smaller default premium for
financial services companies. Both papers find a positive correlation between risk premium
3

Among others Berndt et.al. (2005), Amato (2005), Driessen (2003), Liu et.al. (2000).
Dependent on the specific rating methods, credit risk methodology and the overall context this ratio
is also referred to as risk neutral to actual default intensity (see for example Berndt et.al. (2005)) or - if
adjusted for recovery rates - CDS-to-PD or CDS-to-EDF-ratio (if KMVs expected default frequencies are
use) by other authors.
4

1 INTRODUCTION

and average default probability, i.e. an increasing average default probability in the market
is accompanied by an increasing risk premium. Driessen (2003) uses constant Q-to-P-ratios
by modelling the risk neutral default intensity as a constant multiple of the actual default
intensity (over time and for different obligors). On average, he reports a Q-to-P-ratio of 2.31
(S&P) and 2.15 (Moodys) respectively, although the standard errors are quite large (> 1%).
The main contribution of this paper is threefold. First, we explain the variations in the
Q-to-P-ratio for different rating grades and different industry sectors observed in the literature based on a simple Merton style model. Second, this paper is - to our best knowledge the first to theoretically analyze the Q-to-P-ratio in an incomplete information setting (based
on the model of Duffie/Lando (2001)) - which enables us to model the effect of uncertainty on
the Q-to-P-ratio - and the effect in deviations from the lognormal distribution assumption.
We show, that higher Q-to-P-ratios may - besides higher risk aversion - also be explained by
higher uncertainty of the current asset value, due to a concave relationship between default
probability and Q-to-P-ratio. Third - and most importantly - we directly extract the market
sharpe ratios out of CDS spreads, which gives us a direct link to the equity premium. We
measure an upper bound for the average sharpe ratio of 29 - 42%, which is consistent with
research of equity markets.
We use a simple relationship between risk neutral and real world default probability in
the Merton framework, i.e.
i
h

def
1
def
(1)
Q (t, T ) = (P (t, T )) + SRAssets T t
The resulting relationship between risk neutral and actual default probability5 (Q-to-Pratio) is neither linear nor log-linear, but involves the cumulative normal distribution. Please
note, that we do not try to estimate the actual and risk neutral default probability seperately; we simply assume that we know the actual default probability from rating information;
calibrate the model accordingly and from there derive risk neutral default probabilities; e.g.
we do not have to calibrate the asset value, default barrier or volatility separately. With this
model, we are able to include the correlation and volatility of a firms assets and therefore
derive a result which is consistent with classical portfolio theory (e.g. the dependence of
returns on the systematic, rather than total, risk inherent in a claim). We will then be
able to explain the high intercept from Berndt et.al. (2005) (which is due to a non-linear
relationship between real world and risk neutral default probability), the observed high risk
premia for high quality firms (which is a combined effect of the non-linearity of real world
and risk neutral PD and a larger part of ideosyncratic risk incorporated in high quality firms).
5

Please note, that we there is minor a difference between the ratio of cumulative default probabilities, i.e.

ln(1Qd ef def (t,T )

ln(1Qd ef def (t,T ))
and the ratio of the default intensities Q
=
d ef def (t,T ) defined by Q :=
T
ln(1P
P

Qdef (t,T )
P def (t,T )

def (t,T ))
and P := ln(1P ef
respectively, which is in first order equal to the ratio of cumulative default
T
probabilities. Throughout this paper, we will always analyze the ratio of the default intensity for reasons of
theoretical consistency.

1 INTRODUCTION

The model also allows for an extraction of the risk premium out of CDS spreads, which
is - in theory - not dependent on the quality of the firm or the sector the firm operates,
but merely mirrors the risk attitude of investors and directly yields the sharpe ratio of the
market portfolio:
SRM :=

1 (Qdef (t, T )) 1 (P def (t, T ))

M r
1

=: Merton ,

M
E,M
T t

(2)

where E,M denotes the correlation between the market portfolio and the equity value of the
firm.6 Since credit spreads are very sensitive to the sharpe ratio, this is a very convenient
way: for example, a BBB-rated obligor would have a 5-year credit spread of 73 bp with a
sharpe ratio of 10% compared to a 5-year credit spread of 280 bp with a sharpe ratio of 40%,
see subsection 2.1 for details. This difference shows, that the common noise in the data will
not significantly reduce the possibility to extract the sharpe ratio out of market prices.
Extending the simple Merton model to more advanced models (we examine a classical first
passage time model with observable asset values, a model with unobservable asset values
proposed by Duffie et.al. (2001) and deviations from the lognormal assumption.) shows
an astonishing robustness of the results (1) and (2) derived in the Merton framework for
investment grade companies as long as the asset volatility is larger than approximately 10%,
which can reasonably be assumed for all companies outside the financial services sector.7
For smaller asset volatilities, the results of the first passage time models deviate from the
standard Merton model. With these models we are able to theoretically explain the lower
observed risk premia for financial service companies (which is due to a significantly smaller
asset volatility of financial services companies)8 More generally speaking, by introducing a
(model- and parameter-dependent) adjustment factor by
SRM = Merton AF,

(3)

we show that
The adjustment factor is close to one for all models analyzed as long as the asset volatility is below 10% and the resulting actual default probability belongs to an investment
grade rating.
The adjustment factor can be accurately determined simply based on knowledge of the
maturity and the actual default probability (i.e. parameters that can be observed in
6

The correlation between equity value and market portfolio was used as a proxy for the correlation
between asset value and the market portfolio. In the Merton framework as well as in more general structual
models, it can be shown that for reasonable parameter choices (e.g. resulting in a credit quality / rating
grade of B or higher) these two correlations do not differ by more than 1%, see section 2.
7
In our sample of 125 companies of the CDS index CDX.NA.IG appr. 90% of all non-financial companies
had an asset volatility of 10% or larger based on data from Moodys KMV. In contrast, for fincial services
companies, the volatility is 10% or smaller in appr. 75% of all cases.
8
Due to a lack of data of financial services companies with asset volatilities above 10% and/or nonfinancial services companies with an asset volatility below 10%, it is though hard to empirically verifiy these
theoretical results, see section 3 for details.

2 MODEL SETUP

the market) as long as > 10%, i.e. for a given combination of default probability
and maturity, parameters that can not be observed easily (e.g. asset volatility, default
barrier, asset value or accounting noise) do not significantly affect the adjustment
factor.
Furthermore, the model directly yields the sharpe ratio of the market portfolio and therefore
allows a direct comparison with the equity premium. Applied to all NYSE-listed companies
in the investment grade CDS index CDX-NA.IG from 2003 to 2007 and using EDFs from
KMV as a proxy for the actual default probability results in an average implied market
sharpe ratio of 42% and an average company sharpe ratio of 20%. Adjusting for tax effects
yields a market sharpe ratio of 32% and an average company sharpe ratio of 16%. Using
Moodys ratings instead of EDFs shows similar results (market sharpe ratio of 39% and 29%
after tax adjustments). We used CDS spreads as they are unfunded exposures, so that the
rates should be less sensitive to liquidity effects.9 Based on the theoretical models, appr.
80% of the CDS spreads can be explained by credit risk.
The remainder of the paper is structured as follows. Section 2 describes the theoretical
framework for credit risk premia based on asset value models, including a discussion of
the impact of different asset models and deviations from the lognormal distribution on the
widely used Q-to-P-ratios. We examine a classical Merton model, a first passage time model,
a model based on unobservable asset values as proposed by Duffie/Lando (2001) and asset
distributions based on actual S&P returns. Using a model with unobservable asset values is
fundamental in our point of view, since only these models are able to explain credit spreads
observed in the markets and yield a default intensity, which constitutes the basis of modern
credit pricing models. Section 3 describes our data sources and shows our empirical results
for the risk premium. Section 4 concludes.

Model setup

This section discusses the theoretical framework for extracting risk premia out of CDS prices.
The basic idea is to use structural asset models to derive a relationship between risk neutral
and actual default probability. Most structural models show a quite poor performance in
empirical studies10 . One of the main reasons is the calibration process usually needed to
specifiy structural models, e.g. determination of leverage, asset volatilities, etc. In contrast
to most parts of the literature11 , we do though not aim to derive actual and risk neutral
default probabilities from structural models, we are simply interested in the relation between
9

Compare Berndt et.al. (2005) for similar arguments.

See Schonbucher (2003) for an overview of empirical studies.
11
Huang/Huang (2003), Bohn (2000) and Delianedis/Geske (1998) use a similar approach, our approach
differs though in at last three ways: First, we work in an incomplete information setting, whereas we explicitly
focus on models with information uncertainty; second, we use CDS spreads, which should be less sensitive
to liquidity distortions; third, we are - to our best knowledge - the first who directly aim to extract the risk
attitude out of credit prices.
10

2 MODEL SETUP

risk neutral and actual default probabilities. We simply assume, that there exists a structural model yielding the correct actual default probability and from there derive the risk
neutral default probability. We can therefore omit (most of) the calibration process which
results in stable calculation.
Subsection (2.1) discusses a simple Merton model and shows, that most empirical results
about the Q-to-P-ratio can already be explained - at least qualitatively - in this setting.
Subsection (2.2) expands the framework to a simple first passage time model. The results do
not materially differ compared to the simple Merton framework as long as the asset volatility
is above 10 Percent. Asset volatilities smaller than 10% (which are usually only observed
for financial services companies) lead to a quite large deviation from the standard Merton
model. We are then able to explain the impact of volatility on the Q-to-P-ratio, especially
lower observed Q-to-P-ratios for financial services companies.
Subsection (2.3) examines the model of Duffie/Lando (2001) and again shows the robustness
of the simple Merton model for our purpose. Subsetion (2.4) analyzes the impact of deviations from the assumption of lognormally-distributed asset returns.
The following subsections will always be structurd in the same way: first, the model setup
including the main assumptions is given and formulaes for calculating actual and risk neutral default probabilities are given; second, the Q-to-P-ratio is analyzed within the respective
model; third, the implications for estimating the sharpe ratio are analyzed.

2.1

Merton

Model setup:
Structural models for the valuation of debt and the determination of default probabilities were already mentioned in the well-known Black/Scholes (1973)12 paper. The Merton
framework presented in the subsection is based on the famous paper of Merton (1974)13 ,
which explicitely focusses on the pricing of corporate debt.
In this framework, a companys debt simply consists of one zero-bond. Default occurs,
if the asset value of the company falls below the nominal value of the zero bond at the
maturity of the bond. A company can therefore only default at one point in time, which
obviously poses a simplification of the real world.
Before we will derive an estimator for the market price of risk () based on the Merton
framework, the main assumptions will be presented:
Assumption 2.1 (Assumptions Merton framework)
12
13

See Black/Scholes (1973).

See Merton (1974).

2 MODEL SETUP

1. The Assets Vt follow a geometric Brownian motion with constant drift = V (under
P) and r (under Q) and constant volatility = V > 014 , i.e.
1

under P : dVt = Vt dt + Vt dBt VT = Vt e( 2

1

under Q : dVt = rVt dt + Vt dBt VT = Vt e(r 2

2 )(Tt)+(B

2 )(Tt)+(B

T Bt )

T Bt )

2. The companys debt consist of one single zero-bond with nominal N and maturity T.
3. Default occurs if VT < N (a default can therefore only occur at t=T).
Assumptions 1 is a standard assumption in financial economics, assumption 2 and 3 are
simplifications of the real world and will be relaxed in the next subsections.
Under these assumptions, the real world default probability P def (t, T ) between t and T15
can be calculated as follows:
1

P def (t, T ) = P [ VT < N ] = P [ Vt e ( 2 )(T t)+(WT Wt ) < N ]


 

N
1 2
= P (WT Wt ) < ln
( ) (T t)
Vt
2
" N
#
1 2
ln Vt ( 2 ) (T t)

=
.
(4)
T t
The default probability under the risk neutral measure Q can be calculated accordingly as
" N
#
1 2
ln

(r

(T

t)
Vt
2
Qdef (t, T ) = Q[ VT < N ] =
.
(5)
T t
Combining (4) and (5) yields16
def

(t, T ) = (P

def

r
(t, T )) +
T t


(6)

and

V r
1 (Qdef (t, T )) 1 (P def (t, T ))

=
(7)
V
T t
respectively, see figure 2 for an illustration. Please note a main advantage of this formula:
it directly yields the sharpe ratio of the assets; i.e. neither V and V nor Vt , N or r have to
be estimated separately (and - on the other hand - can not be inferred from (7) separately).
14

For practical reasons, the index V will usually be omitted in the following formulas. To avoid potential
confusion, the index V will though be used in the main formulas and results of this section.
15
Of course, in the Merton framework, a default is only possible at time T, for reasons of consistency with
the following subsections, we will always talk about default probabilities between t and T in this context,
too.
16
See for example Duffie/Singleton (2003) for the first equation.

2 MODEL SETUP

Figure 2: Illustration of the relationship between actual and risk neutral default probabilities
in the Merton framework. P Dcum : actual cumulative default probability, Qcum : risk neutral
cumulative default probability, SRV : sharpe ratio of the assets, T: maturity.

Q-to-P-ratio:
The resulting relationship between CDS spreads / risk neutral default probabilites and actual default probabilities is shown in figure 4. This relationship is increasing and concave,
as can be seen by the first two derivatives of (6) with respect to P def :

Q
1
1
1
def 2
=
e0.5(SR T + (P )
P
P def
2

= e0.5(SR

2 T +2SR

T 1 (P def )

>0

and

Q
2Q
1
def 2
=

SR
T 2e0.5( (P )) < 0,
2
P
P
If one tries to make a linear regression between the risk neutral and the actual default
probability, the intercept will be significantly above zero, which is a result of the concavity of

2 MODEL SETUP

Figure 3: Relationship between CDS spread and actual default probability in the Merton
framework and results of a linear interpolation (i.e. assumption of a credit quality independent Q-to-P-ratio. Parameters: r=5%, SRA = 30%, RR=50%.

the relationship between the risk neutral and the actual default probability implied by (10)
and confirms the empirical research, see for example Berndt et.al. (2005). Please note that
the specific parameters of the regression (slope and intercept) are rarely arbitrarely based on
the choosen interval of actual default probabilities and the choosen data points (which are
usually clustered at regions of investment grade default probabilities, e.g. one-year-default
probabilities up to appr. 0.3% and 5-year cumulative default probabilities up to appr. 3.5%).
Plotting the Q-to-P-ratio and the actual default probabilities (or, similarly, the rating) based
on (10) shows, that the Q-to-P-ratio declines with declining credit quality (see figure 4),
which underlines empirical research on this topic (See Berndt et.al. (2005) and Amato
(2005)). Again, this is a direct result of the concave funcion between the risk neutral and
the actual default probabilities implied by (10).
Market sharpe ratio:
If we try to extract the market sharpe ratio out of (6), we are faced with an additional
problem: the sharpe ratio of the assets VVr will usually differ from the market sharpe ratio,
since the assets Vt will usually not be on the efficient frontier. In other words, the sharpe
ratio of the assets does not only capture the risk attitude of investors, but also depends on

2 MODEL SETUP

10

Figure 4: Relationship between the Q-to-P-ratio and the rating/actual default probability
in the Merton framework. Parameters: r=5%, SRA = 20%.

the correlation of the assets with the market portfolio. The market price of risk can though
be calculated via a straight forward application of the CAPM:
V = r +

M r
V,M V
M

17

M r
V r
1
=

,
M
V
V,M

(8)

where V,M denotes the correlation coefficient between the asset return and the market return.18
17

Without loss in generality we assume V,M 6= 0.

An application of the CAPM does require the respective assets to be traded or - equally - it requires
a self-financing trading strategy resulting in a contingent claim equal to the asset payoff. Asset values are
usually not traded on financial markets, we do though believe that this application is justified for two reasons.
First, the asset value can - in theory - be duplicated by the equity value and a risk free bond in the Merton
framework, so the asset value lies in the asset span of the market. Second, most of the claims on the assets
of large corporations are either directly traded (e.g. equity and bonds) or market values can be inferred from
directly traded instruments with a certain accuracy (e.g. bank loans via bonds). In addition, non-tradable
parts like insolvency costs in more advanced models do have an impact on the choice of the optimal capital
structure, given the actual default probabilities observed in the markets (most large companies are rated
investment grade), these do not have a significant effect on the asset value and are also hedgeable in these
model-setups.
18

2 MODEL SETUP

11

Therefore, in addition to the sharpe ratio of the assets, we will need an estimate of the
correlation between the asset value and the market portfolio. At first, this correlation (V,M )
seems to be a problem for practical applications, since it can neither be directly measured
nor implicitly inferred e.g. from option prices. In the following applications, we will approximate V,M by the correlation between the corresponding equity return and the market
return (denotet by E,M ), i.e. we will assume that
V,M E,M
The error of this approximation is almost negligible, since - within the Merton framework the equity value of a company equals a deep-in-the-money call option on the assets. For all
reasonable parameter choices, the approximation error is less than 1%, see appendix A for
details. Hence the following approximation holds:
1
1 (Qdef (t, T )) 1 (P def (t, T ))
M r

M
E,M
T t
and we can formulate a Merton estimator for the market price of risk =

bMerton :=

1 (Qdef (t, T )) 1 (P def (t, T )) 1

;
E,M
T t

or, solved for the risk neutral default probability:



1
def
1
def
Q (t, T ) = (P (t, T )) + SRM
E,M

M r
:
M

(9)

(10)

Please note, that we will need a sufficient sensitivity of the risk neutral default probability
Qdef (t, T ) with respect to the sharpe ratio in order for an empirical application. Otherwise
noise in the data (e.g. bid-ask-spreads, inaccuracies in determining correlations and actual
default probailities) will result in a very inaccurate estimation. That this sensitivity is indeed
given can be seen by calculating the first derivative of (10) with respect to the sharpe ratio,
i.e.
2
1
Qdef (T )
1
1
def
= e 2 (SRV T + (P (T ))) T.
(11)
SRV
2
see figure 5 for an illustration. If we look for example at a BBB-rated obligor with a 5-year
cumulative actual default probability of appr 2.17%, the resulting risk neutral default probability should be either 3.6% (for an asset sharpe ratio of 10%) or 13% (for an asset sharpe
ratio of 40%) respectively. Assuming a recovery rate of 50% transforms this into a CDS
spread of either 73 bp or 279 bp19 , a difference that is certainly above any noise induced by
liquidity effects, bid/ask-spread or inaccurate measurement of actual default probability.

19

Using the approximation CDS spread = Q LGD and by deriving the risk neutral default intensity from
Q
the risk neutral cumulative default probability by Qdef (t, T ) = 1 e (T t) .

2 MODEL SETUP

12

Figure 5: Influence of the sharpe ratio on risk neutral default probabilities in the Merton
framework for different sharpe ratios (10%-40%). Other Parameters: T=5.

Please note again, that this relationship is independent of the asset volatility, the asset
value Vt and the default barrier L (which is a direct implication of (10)). A calibration of the
asset value model will therefore not be necessary to specify the relationship between actual
and risk neutral default probabilities.
Although we found a compelling result for an estimation of the market price of risk, the
assumptions made in the Merton framework do not fully reflect the true world. In the following subsections we will relax the assumptions about the default timing (see subsection
2.2) and the assumption about complete information (see subsection 2.3) and look at more
general first passage time models. The subsections will always be structured as follows: first,
the model setup is explained and formulas for actual and risk neutral default probabilities are
derived; second, the resulting Q-to-P-ratio is analyzed; third, the impact on the estimation
of the sharpe ratio is discussed.
Before moving on, please note that our estimator for the market price of risk contains a
certain kind of robustness against changes in the underlying assumptions: Since both default probabilities (P def and Qdef ) are measured within the same model and substracted

2 MODEL SETUP

13

from each other, the effect of changes in the default modelling (e.g. first passage time) or
in the assumptions about distribution of returns (e.g. deviation from the assumption of a
log-normal distribution of the asset returns) is - qualitatively spoken - reduced significantly.

2.2

First passage time modell with certain default barrier

Model setup:
In first passage time models, a default occurs as soon as the asset value20 falls below a
certain barrier. The asset value and the default barrier can both be either observable or
unobservable. The model with a certain default barrier and a certain asset value based on
is treated in this section. A model with an uncertain default barrier and unobservable asset
value based on Duffie/Lando (2001) is analyzed in the following subsection.
For this subsection, the following assumptions apply:
Assumption 2.2 (Assumptions first passage time model with certain default barrier)
1. The Assets Vt follow a geometric Brownian motion with constant drift = V (under
P) and r (under Q) and constant volatility = V > 0, i.e.
1

2 )(Tt)+(B

2 )(Tt)+(B

under P : dVt = Vt dt + Vt dBt VT = Vt e( 2

T Bt )

and
under Q : dVt = rVt dt + Vt dBt VT = Vt e(r 2

T Bt )

2. Default occurs as soon as the asset value Vt falls below a predefined and certain barrier
L (i.e. L R)(a default can therefore occur at any time in (t,T].
Under these assumptions, the real world default probability P def (t, T ) between t and T
can be calculated as
P def (t, T ) = P [Vs < L for any t s T] = 1 P[Vs L, t s T]


i
h
( 12 2 )(st)+(Ws Wt )
L;
= 1 P [ min Vs L] = 1 P min Vt e
tsT
tsT




1 2
L
= 1 P min )(s t) + (Ws Wt ) ln( )
tsT
2
Vt




2mb
b m(T t)
b + m(T t)

e 2
(12)
=
T t
T t
20

Some authors use an even more general version of an ability-to-pay process, see Bluhm/Overbeck/Wagner
(2003) for example.

2 MODEL SETUP

14

with
b = ln(

1
L
); m = 2 ; = V ,
Vt
2

see especially Musiala/Rutkowski (1997) for the last step. The default probability under the
risk neutral measure Q can be calculated accordingly as
!


bb m(T
2m
bb
b
b t)
b
+
m(T
b

t)
def

Q (t, T ) =
e 2
(13)
T t
T t
with

1
bb = b = ln( L ); m
b = r 2 ; = V .
Vt
2
There is - in contrast to the Merton framework - no closed form solution for the sharpe ratio
V r
in this model.
V

Unfortunately, we have two equations ((12) and (13) with three unknown variables ( VLt , , ,
assuming we can derive r, T t, P def , Qdef from bond/CDS market data), so we can not
directly derive and in order to determine the sharpe ratio r
Q-to-P-ratio. We will

though be able to determine as a function of and therefore also the sharpe ratio of the
) as a function of . This function is nearly constant for > 10%, see figure 6, so
assets ( r

apart from assets with very low asset volatility we will again be able to estimate the sharpe
ratio without calibrating the asset volatility , the asset value Vt or the default barrier L.
Please note, that the sharpe ratios in figure 6 do not imply, that assets with a lower volatility
have a higher sharpe ratio, rather they implicitly derive sharpe ratios out of CDS spreads
based on empirical data and a first passage time model. A company with a low asset volatility will therefore (all other parameters being equal, especially the actual default probability
and the asset correlation) trade at lower spreads than a company with a high asset volatility.
Q-to-P-ratio:
The resulting slope of the Q-to-P-ratio as a function of the asset volatility is shown in
figure 7 for different rating classes (identified by specific cumulative default probabilities).
Again, one can see that the Q-to-P-ratio is higher for higher rating classes, (almost) independent of the asset volatility for < 10% and sharply declining for asset volatilites smaller
than 10%. It converges to 1 for all rating classes, thus, the slope is more pronounced for
higher rating classes. The dependency on the asset volatility can be explained by the default
timing: for lower asset volatilities, the expected default time conditional on a default until
T is lower for lower asset volatilities conditional on a fixed cumulative default probability up
to T.21 In other words, fixing the cumulative default probability until time T, defaults will
occur with a higher probability at the beginning of the period if the volatility is low. Since
21

Please note that - all other parameters being equal - the default probability declines with declining asset
volatility. Therefore, the expected value of the default time will also decrease. In this case, the declining
asset volatility is always balanced by a lower t0-asset-value to result in order to yield the same rating grade.

2 MODEL SETUP

15

Figure 6: Relationship between the asset volatility and the implicit sharpe ratio derived
from CDS spreads for rating class A. P5, mean and P95 denote observed CDS spreads for
the CDX-NA-IG-index from 2004-2006 for all obligor rated A1/A2/A3 by Moodys at the
5% (8 bp), 50% (19 bp) and 90% percentile (42 bp). Risk neutral default probabilities were
derived with a recovery rate assumption of 50%. The slope of the graph shows quite similar
patterns for other rating classes or recovery rate assumptions.

the Q-to-P-ratio decreases with decreasing maturity, see figure 4 and the related discussion
in subsection 2.1, it also decreases with decreasing asset volatility in the first passage time
framework. In addition, one can easily see from (6), that the Q-to-P-ratio converges to one
if the maturity (T-t) converges to zero.

Estimation of the sharpe ratio:

We now want to test the robustness of the simple Merton estimator for the sharpe ratio. We
therefore introduce an adjustement factor AFF P by
SRM :=

1 (Qdef (T ) 1 P def (T )
1
M r

=:

AFF P = Merton AFF P ,

M
M,E
T

(14)

i.e. the adjustment factor shows, how far the estimate of the market sharpe ratio via the
standard Merton model deviates from the true market sharpe ratio if a first passage model
applies. Again, we have assumed that V,M = E,M , i.e. that the correlation between market
and asset returns equals the correlation between market and equity returns. This equation

2 MODEL SETUP

16

Figure 7: Relationship between the Q-to-P-ratio and the asset volatility for different rating
grades the first passage time model. Parameters: r=5%, SRA = 15%.

holds true for reasonable parameter choices in the first passage time framework as well, as
we will show in the next subsection.
The adjustment factor is dependent on the volatility, the sharpe ratio and the credit quality
(interpreted as actual default probability) of the company. The relationship is shown in
figure 8 for a sharpe ratio of 20%. It shows, that the adjustment factor increases with
decreasing credit quality and with increasing volatility, but for investment-grade titles and
a volatility smaller than 10% the adjustment factor is very close to 1.

2.3

A model with unobservable asset values

Model setup:
Although an improvement compared to the Merton model, credit spreads predicted by simple first passage time models are still not able to predict the credit spreads that can be
observed on the markets.22 Especially for short term maturities, market credit spreads (or
risk neutral default probabilities respectively) are higher than a simple first passage time
model would suggest.
22

See Duffie/Lando (2001), Duffie/Singleton (2003) and Schonbucher (2003) for a detailed discussion.

2 MODEL SETUP

17

Figure 8: Credit quality smile: Adjustment factor in the first passage time framework.
Parameters: r=5%, SRA = 20%. Please note that the majority of traded bonds and CDS
(by volume) has an investment grade rating.

These higher credit spreads for short term maturitites seem to be mainly attributable to
credit risk and are not due to liquidity effects, other risk categories or market imperfections
(See for example Schonbucher (2003).). Several methodological reasons are discussed in the
literature:
The asset value may be uncertain due to imperfect information structures, or - to
be more precise - the current asset value can not be observed exactly. Therefore,
the current asset value becomes a random variable, which in turn has the effect of
increasing short term default probabilities.
The default barrier may be uncertain/unobservable.
This is consistent with the fact that the recovery rate is usually assumed to be a random
variable rather than a fixed value.23 In the simple first passage time model presented
in the last subsection, the barrier was certain, therefore the recovery rate should also
be non-random. An unobservable default barrier leads to a significant increase in the
23

See for example Moodys (2007). A random recovery rate could though also be induced by introducing
random insolvency costs, i.e. costs incurred at default due to direct insolvency expenses, losses in asset value
due to a forced sale in an insolvency process and revaluation of assets serving a specific purpose for the
respective company.

2 MODEL SETUP

18

short term default probability, long term default probabilities are less effected, since
the asset volatility dominates uncertainty for longer time periods.
Asset values are not lognormally distributed and may include jumps. This increases
the short term probability, that the asset value will fall below the default barrier.24
The arguments all have an economic foundation and seem to be reasonable. It is not within
the scope of this thesis to evaluate, which one (or which combination) best produces the
true economic world. Incorporating jumps in the asset value process has been analyzed by
Zhou (1997). Duffie/Lando (2001) have developed a first passage time model with uncertain
asset value. The assumption of an uncertain default barrier has been implemented in the
commercial model CreditGrades by Finger et.al. (2002).
In this subsection, we will focus on the model of Duffie/Lando (2001) and show that although the default probabilities implied by this model differ substantially from the classical Merton model - the difference between risk neutral and actual default probabilities is
almost the same than in the Merton model as long as the asset volatility is above 10%. To
be more precise, we will show that the simple Merton estimator for the market sharpe ratio
(see (9)) is - to a certain degree of accuracy - still valid in the Duffie/Lando-framework for
asset volatilities above 10%.
We choose the Duffie/Lando model for this analysis as it is the most general framework
for credit risk modeling we know of. It incorporates a sophisticated structural model of
default (i.e. a strategic setting of the default barrier based on the asset value process, tax
shield and insolvency costs) and - even more important for this setting - an unobservable
asset process resulting in realistic default term strucures with a default intensity.25
The main assumptions in the Duffie/Lando model are as follows:26
Assumption 2.3 (Assumptions Duffie/Lando model)
1. The Assets Vt follow a geometric Brownian motion with constant drift m =
(under P) and r (under Q) - where denotes the asset payoff rate - and constant
volatility = V > 0, i.e.
1

under P : dVt = mVt dt + Vt dBt VT = Vt e(m 2

24

2 )(Tt)+(B

T Bt )

Note that for long term maturities, a higher volatility has the same effect as adding jumps to the process,
therefore long term default probabilities will not be effected in the same manner than short term default
probabilities.
25
Defaults in a Merton framework can not be described via default intensity processes, since the probability
of a default from t (today) until t + t is always zero or one for a sufficient small t. A default intensity does
also not exist in the the Zhou (1997) framework, since the default time cannot be represented by an totally
inaccessible stopping time (which is a consequence of the fact, that the default barrier may be hit/crossed
via the normal diffusion process with positive probability), see Duffie/Lando (2001) for details.
26
See Duffie/Lando (2001) for details.

2 MODEL SETUP

19

and
1

under Q : dVt = r Vt dt + Vt dBt VT = Vt e(r 2

2 )(Tt)+(B

T Bt )

2. Default occurs as soon as the asset value Vt falls below a predefined and certain barrier
VB (i.e. L R), which is determined by the equity holders in t=0 as to optimize the
total initial firm value (equity + proceeds of the sale of debt).
3. The bond holders / CDS investors are not able to observe
 the asset process directly.
ct = ln(Vt ) + Ut , where
Instead they receive imperfect information Y (ti ) := ln V
i
i
i
U (ti ) is normally distributed and independent of B(t) and is a parameter specifying
the degree of noise in the information received by the bond/CDS investors. Therefore,
the information filtration given to the bond/CDS investors is27
Ht = ({Y (ti ), ..., Y (Tn ), 1 s : 0 s t})

(15)

We will examine the case n=1. Under these assumptions, the conditional density of Zt :=
ln(Vt ), > t28 conditional on the noisy observation Yt and with a fixed starting point
ln(V0 ) =: z0 can be calculated as29
P [ > t Zt = x Yt = Yt ]
P [Yt = Yt ]
P [ > t Zt = x Ut = Yt x]
P [ > t|Zt = x] P [Zt = x] P [Ut = Yt x]
=
,
P [Yt = Yt ]
P [Yt = Yt ]

b(x, t|Yt , z0 ) := P [ > t Zt dx|Yt ) =

=

where we used the equivalance of Yt () = Yt and Ut () = Yt x under the condition that

Zt () = x in the second step and Bayes rule in the first and third step. Conditioning on
> t30 , this yields the conditional density g(x|Yt , z0 , t) of the log asset value Yt conditional
on the information Yt = Yt and > t:
g(x|Yt , z0 , t) =

b(x|y, z0 , t)
P [ > t|Yt , z0 ]

This conditional density can be explicitly calculated, if it is assumed, that asset values follow
a geometric Brownian motion and Ut has a normal distribution, see Duffie/Lando (2001) for
details.
To calculate cumulative default probabilities requires a weighted application of (12) over all
27

Of course, the bond/CDS investors can also obvserve, if a default has already occured.
denotes the stopping time representing the default point.
29
In the following, we will use the conventional informal notations P [X = x] or P [X dx].
30
Please note that the investors are of course able to observe, that no default has taken place up to time
28

t.

2 MODEL SETUP

20

possible asset values Vt , where the weight is - roughly speaking - the probability of the asset
value Vt 31 , i.e.
Z
def
g(x|Yt , z0 , t) dx
PFdef
(t, T, x)
(16)
P (t, T ) =
|
{z
}
| P {z
}
VB
PD(first passage time), if Vt =x

Prob., that Vt =x

where PFdef
P (t, T, x) denotes the probability, that an asset value process starting in t=t at
Vt = x will fall below the default barrier up time t=T, see (12), and g is the conditional
density of the asset value at t=t given the filtration Ht .

Q-to-P-ratio:
Formula (16) can now be used to calculate the risk neutral as well as the real world default probability by either using the risk neutral drift of the asset (e.g. the risk free rate less
the payout rate) or the actual drift of the assets.32
Unfortunately, there is no closed form solution for (16), we though have to draw back on
numerical solutions. Figure 9 shows the Q-to-P-ratio for = 10% and varying actual default
probabilities (identified as ratings) and asset volatilities.
It can be seen, that - for a given rating grade - the Q-to-P-ratio is again almost independent
of the asset volatility as long as > 10% and sharply declines for < 10%.
Uncertainty (measured by and T1 ) results in slightly higher Q-to-P-ratios. This effect is
rather small but increasing with increasing credit quality, see figure 10 for = 15% and
varying values for ). The explanation is rather simple:
To calculate default probabilities in the Duffie/Lando model, we simply have to calculate
them in the first passage time framework with observable asset values and average the result
for different possible levels of P Dcum (representing different levels of the actual asset level Vt )
with respect to the conditional density of Vt . Since the Q-to-P-ratio is a convex function of the
def (E[P def ])
def
def )
] Q E[P
default probability in the Merton framework (see figure 4), i.e. E[ Q P (P
def
def ]
as a result of Jensens inequality, an increasing leads to an increase in the Q-to-P-ratio,
although the effect is rather small for the parameters analyzed. The effect is though still
quite interesting: An increase in the Q-to-P-ratio does not necessarily mean, that agents
have become more risk averse, it can also reflect a higher (actual or perceived) noise in asset
values. This effect is larger for high quality companies, so that in states of higher uncertainty,
spreads on good quality companies should increase most. This finding can also have effects
on trading strategies which try to use risk aversion measures depicted from credit markets
for other market segments (e.g. equities)33
31
Of course the probability of a single value Vt will be zero for non-degenerated parameter choices, since we
operate in a continious setting. We will still use this informal notation to allow for a better understanding.
32
Please note, that the drift changes PFdef
P (t, T, x) as well as the conditional density g(x|Yt , z0 , t). The
influence on the conditional density decreases with decreasing .
33
There are several tactical asset allocation strategies which use risk aversion derived from credit markets,
equity markets and money markets (e.g. TED spreads) to allocate funds to different asset classes. Using

2 MODEL SETUP

21

Since asset volatilites and accounting noise is usually hard to estimate, this a very convenient
result for our purpose. If we want to extract the risk premium out of CDS spreads, asset
volatilites and accounting noise does merely play a role given a specific actual default probability. The basic logic behind this is quite simple: the accounting noise does not pose any
systematic risk, therefore the difference between risk neutral and real world default probability is merely unaffected; the same logic applies to the payout rate, the asset value level
and the time between the start of the asset value process and the first noise observation, too.

Figure 9: Relationship between the Q-to-P-ratio and the asset volatility for different rating
grades in the Duffie/Lando model. Parameters: = 10%, m=2.5%, SRA = 20%, T=5, s=1.

Estimation of the sharpe ratio:

simple measures as the development of average CDS spread or Q-to-P-ratios may be misleading in some
market environments, where a high uncertainty about true values exists. An example is probably the
subprime crisis in 2007, which lead market participants highly unsecure about where risks actually turned
about and what dimension potential losses may have. It is not within the scope of this paper, to empirically
analyze the market reactions on credit and equity markets, but this could provide an area for further research
on this topic.

2 MODEL SETUP

22

Figure 10: Relationship between the Q-to-P-ratio and the rating/actual default probability
in the first passage time model for different levels of . Parameters: r=5%, SRA = 15%,
=15%.

As in subsection 2.2, we now want to test the robustness of the Merton model estimator,
e.g. we again define an adjustment factor AFDL by
M r
= M erton AFDL
M

(17)

This adjustment factor may depend on all parameters included in the model, which we will
group into two different classes: Class 1 captures all parameters that can easily be observed
in the market, i.e. the actual default probability (which is actually a combined parameter of
all other input parameters) and the maturity. Class 2 captures all parameters that can not
be easily observed in the markets, i.e. the asset volatility , the payout rate or the risk
neutral net asset growth rate m := r , the starting point of the asset value process in t=0
(Z0 ), the default barrier VB , the noise asset value observed at t=t (Vbt ) and the accounting
noise . If the adjustment factor depends on any class-2 parameter, this will affect our
ability to correctly measure the market sharpe ratio, since these parameters will be subject
to possibly significant calibration errors.

2 MODEL SETUP

23

We have evaluated (17) for different combinations of input parameters34 , the main results
can be summarized as follows:
The adjustment factor is close to 1 for all parameter combinations as long as the
asset volatility is below 10% and the resulting actual default probability belongs to an
investment grade rating, see figure 11. This can be explained by looking at the impact
of the parameters introduced in the Duffie/Lando framework: all of them do effect the
actual default probability as well as the risk neutral default probability in the same
direction, e.g. increasing the information uncertainty increases the actual as well as
the risk neutral default probability. The share ratio is the only parameter that simply
has an effect on the actual default probability only. This explains qualitatively, why
the adjustment factor is close to one in many cases.
The adjustment factor can be accurately determined simply based on knowledge of
the class-1 parameters and the actual default probability as long as > 10%, i.e. for
a given combination of default probability and maturity, parameters that can not be
observed easily (e.g. asset volatility, default barrier, asset value or accounting noise)
do not significantly affect the adjustment factor.
Please note the special role of the actual default probability: For example, an adjustment
factor of appr. 1.7 (i.e. significantly above 1) occurs for an asset value of 108, =10%,
T=5, SR=40% and =0%. If this were due to any class-2 parameter, empirical applications
would be hardly possible due to calibration errors of class-2 parameters. But as soon as
we change any of these parameters so that the resulting actual default probability belongs
to an investment grade rating (e.g. increasing Vt , decreasing , decreasing (up to a level
of 10%)) the adjustment factor will be close to 1. In other words, any combination of
these parameters, that yields a given actual default probability also yields (almost) the same
adjustment factor. All in all, class-2 parameters do actually have a significant influence on
the adjustment factor; but this influence can (almost fully) be captured simply by the rating
smile.
If we take a closer look at the parameter combinations leading to the minimal and maximal
adjustment factor in figure 11, we see, that these actually belong to unlikely parameter
combinations. Table 6 shows the dependeny of the adjusment factor for a maturity of 5
years and a Baa-rating35 for = 0% and s = 0, e.g. for the extreme of observable asset
34

Input parameters used were: : 3% 30% (The 5% and 95% quantile for the asset volatility from KMV
was 6% and 25% respectively), sharpe ratio of the ability-to-pay process: 10% to 40% (The market sharpe
ratio is usually assumed to be anywhere between 20% and 50%, due to a correlation of lower than 1, the
asset sharpe ratio should be smaller), m : 0% 5% (m < 0 would imply, that the payout rate is larger than
the risk free rate, m=5% was choosen as an upper limit to reflect (almost) zero payout at a risk free interest
rate of 5%.), : 0% 30% ( = 0% reflects the classical first passage model with observable asset values,
Duffie/Lando use 10% as a standard value, the upper limit of 30% is also based on Duffie/Lando(2001)),
Vbt = Z0 and VB for all combination that resulted in rating grades from AA to B. The case Vbt > Z0 and
Vbt < Z0 was also analyzed, the results merely differ from the case Vbt = Z0 and are available upon request.
35
We choose this as an example, since 5-year CDS are the most liquid ones usually used in empirical
studies, see for example Berndt et.al. (2005) and Amato (2005) and Baa is the most common rating among
non-financial companies.

2 MODEL SETUP

24

values, table 7 for = 30% and s = 3 representing large uncertainty about the asset value.
The bold numbers depict the minimum and the maximum values for the adjustment factor.
The absolut maximum is taken for small asset volatilites, no uncertainty and a high risk
neutral asset growth rate (i.e. a low payout rate). Even ignoring the unrealistic default term
structure implied by a certain asset value, one would usually expect small asset volatilities
to belong to a value firm whereas low payout rates usually apply to growth companies.
The absolut minimum is taken for small asset volatilities and high payout rates (i.e. a low
risk neutral asset growth rate m), which would suit the usual assumptions about value firms,
but for a high uncertainty about the current asset value, which one would usually assume for
growth companies. Depicting values one would usually assume for value firms and growth
firms, we can see that the adjustment factor will lie even closer around the mean value.

Figure 11: Adjustment factor in the Duffie/Lando model for different rating grades. The
minimum and maximum is taken over the parameters 10% 30%, 0 30%,
0 m 5%, 10% SRV 40%, 0 s 3. Other parameter: T=5.

2.4

Deviations from the lognormal assumption

It is a widely discussed topic in empirical finance, that log asset returns are not normally
distributed (as a geometric Brownian motion suggests) but rather fat tailed or leptokurtic,
i.e. their standardized third moment is larger than the third moment of a standard normal
distribution (which has a fourth moment of 3). In first passage time structural models, a

2 MODEL SETUP

25

default occurs, as soon as the asset value has fallen below a predefined default barrier. Since
these default probabilities are usually small (e.g. the 5-year default probability of a A-rated
obligor is 0.60%, the 5-year default probability of a BBB-rated obligor is appr. 2.2%36 ),
leptokurtis of the asset value returns has a significant influence on default probabilities, and
- as we will show in this subsection - on the difference between risk neutral and real world
default probabilities as well.
To analyze the impact of leptokurtic asset returns, we will return to a Merton setting,
where default can only occur at maturity of the bond/CDS. The reason for this approach
is mainly its simplicity and tractability. Using fat-tailed distributions in first-passage-time
setting usually involves quite sophisticated mathematical framework based on discontinious
martingales.37 In addition, the results usually loose the simplicity and intuition.

We will assume in the following, that the log asset return RT := ln( VVTt ) at time T (the
maturity) has a distribution identified by its cumulative distribution function F (x). The
distribution is supposed to have first and second order moments, so that we can standardize
cT = RT T . The expected one-year return is supposed to equal under the real
it by R
T
world probability measure and r under the risk neutral probability measure. Furthermore,
we assume stationary
and independent increments of the 1-year returns, so that T = T

cT = RT T

. Therefore, the real world default probability

and T = T , which yields R
T
can be calculated as
"
#


N
ln(
)

(T

t)
N
Vt
cT <

P def (t, T ) = P [ VT < N ] = P RT < ln( ) = P R

Vt
T t
"
#
ln( VN0 ) (T t)

= F
(18)
T t
The risk neutral default probability can be accordingly calculated as
"
#


N
ln(
)

(T

t)
N
Vt
cT <

Qdef [t, T ] = Q [ VT < N ] = Q RT < ln( ) = P R

Vt
T t
#
"
ln( VNt ) r (T t)

= F
(19)
T t
Therefore, we can derive the same relationship between the risk neutral and actual default
probability as in section 2.1, we only have to substitute the normal distribution by our new
36

See appendix D for details.

Please note for example, that processes with fat-tailed -stable distribution, stationary and independent
increments do necessarily have discontinous trajectories, see Embrechts/Kl
uppelberg/Mikosch (1997) for
example.
37

2 MODEL SETUP

26

distribution F, i.e.
def

Q
and


(t, T ) = F

(P

def

r
T t
(t, T )) +

F 1 (Qdef (t, T )) F 1 (P def (t, T ))

V r

=
V
T t


(20)

(21)

We have analyzed the resulting Q-to-P-ratio based on the actual distribution of weekly S&P500 returns from 1970 to mid 2007, which results in 3011 weekly returns. These returns are
significantly fat-tailed, the standardized fourth moment is 6.5, i.e. the excess kurtosis is 3.5,
the distribution of the returns is shown in figure 12. As an equity index, the returns of the
S&P-500-index do naturally not constitute asset returns but rather returns of call options
on companies assets. We do though use these returns to show the maximum effect possible.
The real distribution of asset returns will probably lie somewhere between the lognormal
distribution and the actual S&P return distribution.
The results are shown in figure 13 for a maturity of 5 years and a company sharpe ratio of 20%. The results do not materially differ for different maturities and sharpe ratios.
We can no longer see any significant increase in the Q-to-P-ratio with increasing credit quality, there is even a small decline, although this is based on a quite small set of data (based
on 3011 weekly returns, the 0.5% quantile is based on the worst 15 observations, the 0.17%
quantile - which is relevant for a 5-year AA-rated company - is based on the worst 5 observations). The result is quite interesting as fat-tailed distributions yield - all other parameters
being equal - higher default probabilities than the normal distrution, e.g. fixing a specific
high loss, the probability of an occurence is higher for a fat-tailed distribution. For our purpose - knowing the actual default probability - we are though only interested in differences
between actual and risk probabilities. In our framework, this difference is dependent on the
slope of the cumulative distribution function at F 1 (P def ), as can be seen from (20) or figure
2. This slope (relative to the actual default probability P def ) is almost independent of the
credit quality if one uses actual S&P returns, as figure 13 shows.
All in all we can say, that a change in the distribution towards fat-tailed distributions will
even smooth the behaviour of the Q-to-P-ratio with respect to the actual default probability. The Q-to-P-ratio based on the Merton model will therefore more likely overestimate the
Q-to-P-ratio based on actual distributions. Correpondingly, the adjustment factor will be
close to one for low credit qualities and larger than one for higher credit qualities. Keeping
in mind, that S&P returns do not constitute returns of companies assets but rather equity
returns, the difference in figure 13 certainly overestimates the effect.

3 EMPIRICAL ANALYSIS

27

Figure 12: Distribution of weekly standardized log returns of the S&P 500 index from 19502007 compared with the standard normal distribution.

3
3.1

Empirical analysis
Description of data sources

Our data sample consists of 125 North American companies based on the Dow Jones CDX.NA.IGindex, which is an investment grade CDS index administered my markit and a consortium
of 16 investment banks.38 We used 5-year CDS spreads to derive risk neutral default probabilities, EDFs (expected default probabilities) from Moodys KMV as a proxy for the actual
default probabilities and correlations with the S&P500-index as a proxy for the correlations
38

The data is based on the CDX.NA.IG.8-index. By first fixing the constituents list and then analyzing
historical data of these constituents, our data may be subject to a selection bias, since all these companies
have performed well enough until 2007 to be included in an investment grade index. We are though not
interested in the performance of the index or any of its constituents but only examine the risk attitude based
on risk neutral and actual default probabilities. The resulting error should therefore be minimal, since the
actual default probabilities are effected the same way.

3 EMPIRICAL ANALYSIS

28

Figure 13: Deviation from the lognormal assumption: Q-to-P-ratio based on the actual
distribution of weekly S&P returns from 1950-2007. Other parameters: T=5, company
sharpe ratio = 20%.

with the market portfolio. In some occasions, Moodys ratings were used in addition for the
actual default probability. The time horizon ranges from January 2003 until June 2007.
CDS spreads:
Credit Default Swaps (CDS) are OTC credit derivatives, which have become widely popular
over the last years, with growth rates of over 100% (nominal value) in 2005 and 200639 . Its
main mechanism is quite simple: the protection buyer periodically pays a predefined premium (usually quarterly) to the protection seller. In case of a credit event, the protection
seller has to cover the losses incurred on a predefined reference obligation, i.e. he he has to
pay an amount equal to the difference between the nominal and the current market value
of the predefined reference obligation to the protection buyer40 . As usual, put into practice,
39

Total outstanding market volume was $26 trillion at the end of 2006 concerning to the ISDA, growth
rates in 2005 and 2006 were 103% and 101% respectively, see ISDA 2006 year-end market survey.
40
As a result, the protection buyer is also exposed to default risk of the protection seller.

3 EMPIRICAL ANALYSIS

29

things turn out to be more complicated: the credit event has to be precisely defined41 , a
basket of reference obligations has to be specified42 and the term market value at the time
of default has to be clearly specified . As in most academic research, we will assume, that
these specification do not have a significant value and therefore CDS can be prized as if these
(implicit) options were not part of the game.
The 5-year CDS spreads (bid/ask/mid) used in our analysis were taken from Datastream.
Only dates with at least one trade for the respective CDS were used to avoid potential errors
from pure market maker data. We used CDS mid spreads for our analysis. Bid/ask-spreads
served for consistency checks and sensitivity analysis. The risk neutral default intensity Q
was derived by the approximation s = Q LGDQ out of the CDS spread s and the risk
neutral LGD43 . A recovery rate of 50% was used.44
Actual default probabilities:
Expected default frequencies (EDFs) from Moodys KMV were used as a proxy for the actual default probabilities. EDFs are default probabilities, which are based on a Merton-style
structural framework, see for example Moodys KMV (2007). The calibration is though
done more pragmatically based on a large set of historical data and on discriminant analysis. EDFs are widely used in the banking industry and also constitute a part of some of the
internal rating systems of large banks. We used 1-year EDFs and derived multi-year EDFs
by Moodys cumulative default probabilites per rating grade, see appendix D for details.
The main advantage of EDFs over other ratings for our use is its link to market data: the
current asset volatility and equity value is a direct input parameter, therefore EDFs constitute a point-in-time estimation of the current default probability. In contrast to EDFs, the
ratings of the large rating agencies are usually defined as through-the-cycle-ratings, which
in effect results in different default probabilities for a specific rating grade dependent on the
current overall economic outlook.
In some circumstances, Moodys ratings were used, if so, these were taken from Bloomberg.
The ratings were mapped to default probabilities via a logarithmic approach based on raw
data from Moodys (2007), i.e. ln(P D) = 1 + 2 N RG, where NRG denotes the numerical
rating grade ranging from 1 (Aaa) to 16 (Baa3). The log-approach is a common approach
41

Especially the credit event restructuring has been a specific point for discussion leading in (at least)
three different restructuring clauses: restructuring/modified restructuring/modified modified restructuring.
42
Defining only a single reference obligation is not possible for practical reasons, which usually leads to a
cheapest-to-deliver option for the protection buyer, who can usually choose which reference obligation to
sell to the protection seller in case of a default.
43
The simple equation s = Q LGDQ is only valid if the LGD is given as a fraction of market value. In case
of a recovery/loss given as a percentage of face value - as will usually be the case when using rating agency
data - default intensity and recovery enter the pricing formula asymmetrically. For practical pruposes, the
differeces are minimal. See Duffie/Lando (1999) and Duffie/Lando (2003) for discussion. See for example
Duffie/Singleton (2003) or Sch
onbucher (2003).
44
Based on Moodys (2007) the average default rate from 1982-2006 was 54% for senior unsecured debt.
Other studies also derive recovery rates of approximately 50%, e.g. Altman/Kishore (1996) (48%) and
Carty/Liebermann (1996) (54%).

3 EMPIRICAL ANALYSIS

Variable
CDS mid in bp
CDS offer in bp
CDS bid in bp
(offer, bid) in bp
Asset Vol.
EDF1
EDF5
Moodys PD1
Moodys PD5
(EDF1, Moodys PD1)
Correlation

N
19945
19945
19945
19945
19945
19945
19945
14743
14743
14743
19945

30

Mean
51.2
53.1
49.4
3.8
15%
0.2%
1.93%
0.15%
2.01%
0.00%
0.51

Median
41.77
43.63
40
3.62
15%
0.1%
1.38%
0.10%
1.63%
-0.01%
0.52

Coeff of Variation
74.74
72.73
76.96
54.22
38.07
174.12
95.45
120.22
75.12
11660.05
24.53

25th Pctl
28.1
30.0
26.3
2.7
11%
0.05%
1.00%
0.05%
0.93%
-0.11%
0.42

75th Pctl
62.5
64.0
60.5
4.7
18%
0.15%
2.15%
0.18%
2.51%
0.06%
0.60

Table 1: Descriptive statistics for main input parameters.

EDF1/EDF5/Moodys
PD1/Moodys PD5 denote the respective 1- and 5-year cumulative default probabilities.
for the calibration of default probabilities, see for example Bluhm et.al (2003). The resulting
table of cumulative default probabilities can be found in appendix D.
Correlation:
We used 3-year weeekly45 correlations of the reference entities share price returns with the
S&P-500 index. The share prices were taken from Datastream either from the NYSE or
NASDAQ. Companies, which are not listed on one of these exchanges were removed from
the data sample. A floor of 20% was applied to the estimation of correlations.
Our final data set consists of 19,945 date/company-combinations for which CDS spreads,
EDFs and correlations were available. Table 1 gives an overview of these main input parameters.

3.2

Empirical results

Based on the data described in subsection 3.1 and the Merton estimator for the sharpe ratio
derived in section 3.1, we analyzed the Q-to-P-ratio and the implicity company and market
sharpe ratios based on the Merton estimator for each of the 19,945 observation.
Q-to-P-ratio:
Overall, we found an average Q-to-P-ratio of 3.38 in our sample, which is slightly above
the result of Berndt et.al. (2005) of 2.0 and Driessen (2003) of 2.31 (Moodys) / 2.15 (S&P).
Since we do only analyze investment grade companies and the Q-to-P-ratio increases with
increasing credit quality, this seems to be in line with other empirical findings.
45

The calibration of correlations has a minor effect on the overall result, using 2-year or 1-year correlation
did not alter results significantly.

3 EMPIRICAL ANALYSIS

31

More interestingly is the behaviour for different rating grades. Comparing the Q-to-P-ratio
derived in section 2 with the actual Q-to-P-ratio from our data shows, that the increase with
respect to the credit quality is higher than the theoretical model predicts. Qualitatively, this
is in line with what other researches report, see for example Huang (2003). Quantitatively,
the difference is smaller than other research suggests: For the low rating classes (Baa1-Baa3),
the Q-to-P-ratio is in the uppermost area of the cone defined by a company sharpe ratio of
20% and varying other parameters (, , T1 and m), which in effect means a sharpe ratio of
slightly above 20%, see figure 14. For the highest rating grades in our analysis (Aa1-Aa3),
the Q-to-P-ratio is in the bottom are of the Q-to-P-cone defined by a sharpe ratio of 30%,
resulting in a sharpe ratio of slightly below 30%. In order to make it a fair game, the
Q-to-P-ratio of the Aa rating grade should decline by appr. 20% (from 6.01 to 4.74) or alternatively - the Q-to-P-ratio of the rating grade Baa should increase by appr. 15% (from
2.83 to 3.26). Formulated in another way, the spread of an Aa-rated entity has to decline
by appr. 20% - which results in a decrease of the risk neutral default probability by approximately 20%, all other things being equal - to be justified in a theoretical model. Given an
average spread of 32 bp for Aa-rated obligors, this simply means a reduction of 6 bp. Given
a bid-ask-spread of 4 bp (e.g. CDS bid is on average 2 bp below CDS mid), this means, that
CDS spreads for different rating categories are consistent46 in a theoretical model without
having to explain a huge amount of the spread by tax or liquidity effects. Approximately
80% of the CDS spread can purely be explained by credit risk. This portion is significantly
larger than prior research for the bond market suggests, see for example Huang (2003).
Though, we can still conclude, that a small part is probably not explained by credit risk,
but rather by bid/ask spreads, taxes or other market incompletnesses. If one assumes, that
advanced credit models including unobservable asset values do mirror reality, than this portion seems to be rather small.

Estimation of company and market sharpe ratios:

The analysis based on CDS and EDF-data from 2003 to 2007 yields an average market
sharpe ratio of 42%, with an average company sharpe ratio of 19% and an average correlation of 0.51%, see table 2. Half of the observations result in a company sharpe ratio between
12% and 27% (market sharpe ratio: 22% and 57%). Using actual default probabilities from
Moodys leads to similar, though slightly smaller, estimates for the average company sharpe
ratio (19%) and market sharpe ratio (39%).
Depicting the dimension time and averaging over all companies/CDS spreads shows, that
the implicit market sharpe ratio was predominantly between 30% and 50% between 2003
46

Please note that we do not use the term consistent to point out to the fact, that the spread differences
can (almost) totally be explained in a pure credit risk setting, of course that does not pose a proof that real
CDS spreads are really totally explained by credit risk. We do though want to point out, that credit models
alone do not give rise to a necessity of explaining (large) parts by other factors than credit risk.

3 EMPIRICAL ANALYSIS

32

Figure 14: Implicit sharpe ratio based on CDS spreads and EDFs as a function of the credit
quality compared with theoretical values based on the Duffie/Lando model.

and 2007, see figure 15. The volatility of the market sharpe ratio over time is appr. 50%.
We would also like to point out a major difference compared to the Q-to-P-ratio: Whereas
the sharpe ratio at the beginning and at the end of the period under consideration has only
slightly increased, the increase in the Q-to-P-ratio was much more pronounced, see figure
16. This is a direct effect of the increasing credit quality in the sample period, see section
2.3, figure 10. From a theoretical point of view, the sharpe ratio is the better indicator for
risk aversion.
As for the Q-to-P-ratio, we now want to analyze the dependency on the credit quality.
In theory, it should be independent of the credit quality. If we look though at the sharpe
ratio as a function of the credit quality, we still see an increasing sharpe ratio with increasing
credit quality, although this effect is less significant than for the Q-to-P-ratio. There are
though some potential effects from the theoretical framework developed in section 2, that
could decrease this gap:
Adjustment factors derived from the Duffie/Lando model: Compared to more advanced
models, the sharpe ratio is underestimated for low quality companies and overestimated
for high quality companies, as shown in section 2. Adjusting for the adjustment factors derived in section 2.3 (see figure 11) further reduces the gap, but still shows the
same pattern with respect to the credit quality, even if considering the maximum and

3 EMPIRICAL ANALYSIS

Variable
Q /P (EDF)
Q /P (Moodys)
Sharpe ratio market (EDF)
Sharpe ratio company (EDF)
Sharpe ratio market (Moodys)
Sharpe ratio company (Moodys)

33

N
19945
14743
19945
19945
14743
14743

Mean
3.38
3.40
42.46%
19.56%
39.01%
18.70%

Median
2.74
2.55
37.23%
19.04%
35.25%
17.79%

Coeff of Variation
71.36
87.09
76.33
60.32
75.70
67.63

25th Pctl
1.87
1.70
21.91%
11.76%
21.50%
10.15%

75th Pctl
4.12
4.03
56.80%
26.71%
53.03%
26.51%

Table 2: Descriptive statistics for main output parameters. (EDF) and (Moodys) denotes,
that the respective parameter was calculated via EDFs and default probabilities derived from
Moodys respectively. SR: sharpe ratio. Q and P denote the actual/risk neutral default
intensities.
Rating grade
Aa
A
Baa1
Baa2
Baa3

Company sharpe ratio (CSR)

28.90%
22.98%
20.68%
16.76%
12.99%

Q-to-P-ratio
6.01
4.13
3.31
2.58
2.05

CSRmin after AF
24.64%
19.73%
17.95%
14.55%
11.28%

CSRmax after AF
32.90%
27.31%
26.23%
21.26%
16.48%

Table 3: Dependency of sharpe ratio and Q-to-P-ratio on the rating grade. CSRmin ,
CSRmax : Minimum/Maximum copmpany sharpe ratio after adjusting for the adjustment
factors derived in subsection 2.3.
minimum adjustment factors based on reasonable parameter combinations (16.5% vs.
24.6%), see table 3.
Correlations: A usual assumption is, that higher quality firms are less exposed to
ideosyncratic risk than lower quality firms.47 The behaviour of the correlation with
respect to rating classes is though quite akward in this sample. The correlation is
a increasing function of the EDF-measure and a decreasing function of the default
probability derived from Moodys rating. Companies with a high correlation with the
market seem to have predominantly high Moodys ratings and high EDFs, whereas the
opposite seems to be true for companies with a low correlation, see table 4.48
Summing up, credit spreads for high quality companies still seem to be too high compared
to credit spreads on lower quality companies, even in an unobservable asset value framework. This result is in line with other empirical research, which has stressed the role of
taxes, transaction costs, liquidity and other parameters absent in a perfect market setting.
47

In contrast, some of the recent defaults of high credit quality firms, i.e. Enron and Worldcom, seem to
be attributable to firm-specific failures.
48
Possible explanations are: First, in the manual Moodys rating process, the impact of diversification on
the default probability is overestimated; second, the EDF-measure may overestimate the EDF of companies
highly correlated with the market, since these companies usually have lower multiples (i.e. higher cost of
capital) and therefore ratios based on the market value of equity may seem less favorable. We do not aim
to resolve this conflict in the paper and leave this question for further analysis.

3 EMPIRICAL ANALYSIS

34

Figure 15: Implicit sharpe ratio based on CDS spreads and EDFs as a function of time.

Although most of the research focuses on bond pricing, an effect on CDS spreads seems to
be be likely as well. It is not the target of this paper, to quantitatively evaluate these effects.
We do though want to stress, that all these effects will lead to a decrease in the implicit
sharpe ratio, since they lead to a declining portion of the CDS spread attributable to credit
risk.49 Therefore the average derived sharpe ratio of 42% is still valid as an upper limit on
the sharpe ratio in these settings.
One major difference between equity and debt markets is its tax treatment. Looking at
equity returns, the capital gains are usually tax free and only the dividend part of the return is reduced by tax payments. In contrast, capital gains do - on average - not pose a
significant part of credit returns, as bonds are normally issued at par. The interest received
49
The tax rate plays a special role, since - in contrast to other effects - it does not only have an influence
on the portion of the credit spread attributable to credit risk but can also have an effect on the volatility as it
may in theory - dependent on the tax system - decrease the volatility of the after tax cash flow. It is though
hardly imaginable, that CDS spreads will increase, if taxes are lowered, therefore the general statements will
- under realistic assumptions - hold true for tax effects, too. For a detailed discussion on the effect of capital
income taxes on asset prices see for example Rapp/Schwetzler (2006).

3 EMPIRICAL ANALYSIS

35

Figure 16: Development of 5 year CDS spread, 1-year EDF, implicit market sharpe ratio
and Q-to-P-ratio as a function of time.

on the bonds is usually subject to full tax payments. Since CDS are oftenly held by banks,
we have applied a different method to extract the tax effect out of CDS spreads: Thinking
in a P&L logic (and ignoring operative expenses), the difference between the CDS spread
(income) and the expected loss (as expected payout rate to the protection buyer) is - on
average - taxable on corporate level. Therefore, the after tax cash flow received by an investor equals (CDSspread expected loss)(1corporate tax rate). Using a corporate tax
rate of 35% yields an average company sharpe ratio of 15% and a market sharpe ratio of 33%.
These findings strongly support current research on the equity premium, which has averaged appr. 7.4% over the last 50 years, which is equivalent to a sharpe ratio of appr. 44%.
Current research suggests, that the development of stock markets over the last 50 years was
partially driven by extraordinary gains and cannot be explained by fundamentals alone.50
50

Among others, Fama/French (2001) derive an implicit sharpe ratio for the U.S.-market from 1951-2000
of 15% based on the dividend growth model and a sharpe ratio of 25% based on the earnings growth model.
Claus/Thomas (2001) derive an equity premium of 3.4% based on an Earnings forecast model, which equals
a sharpe ratio of appr. 17%-23% (based on a market volatility of 15%-20%).

3 EMPIRICAL ANALYSIS

Correlation
Rating grade EDF
AA
A
Baa1
Baa2
Baa3
Total

Rating grade Moodys

AA
0.51
0.53
0.56
0.57
0.53
0.54

36

A
0.47
0.50
0.54
0.54
0.56
0.52

Baa1
0.48
0.45
0.51
0.54
0.52
0.50

Baa2
0.45
0.44
0.51
0.56
0.54
0.48

Baa3
0.40
0.41
0.43
0.42
0.51
0.43

Total
0.46
0.47
0.50
0.52
0.53
0.49

Table 4: Three-year weekly correlation of equity value with the S&P 500 for different EDFand Moodys rating grades.
A fair equity premium is seen at appr. 3-5% in most academic research, equivalent to a
sharpe ratio of appr. 15%-30% (implying an average market volatility of 15%-20%. A good
overview about equity premium estimations is provided by Ilmanen (2003)). Taking our
sharpe ratio estimate of 42% and 33% (after tax adjustment) and taking into account, that
possibly not the total credit spread is attributable to credit risk yields the same qualitative
result: sharpe ratios of above 40% simply do not seem to be priced in current asset values.
Accounting for a possible non-lognormal distribution of asset returns also does not change
the result: for lower credit qualities, the adjustment factor is appr. one, the adjustment
factors for higher credit qualities only leads to a decline in the portion of the CDS spread
attributable to credit risk.
In section 2, we have shown, that the results are quite robust to model changes. Besides
misspecifying the correct model, noise in the input parameters pose another possible source
of inaccuracy. We therefore testet the sensitivity of our result with respect to the main input
parameters. The results are shown in table 5. Changes of 10% relative to its original value
results in a market sharpe ratio of appr. 5% higher/lower for all analyzed parameters (EDF:
actual default probability, CDS spread, recovery rate, correlations). We would especially
like to focus on two parameters: First, the sensitivity with respect to the CDS spread and
second, the sensitivity with respect to the recovery rate.
The sensitivity with respect to the CDS spread can be used, if there is evidence that only
a portion of the CDS spread is attributable to credit risk. For example, if only 90% of the
CDS mid spread were attributable to credit risk, this would decrease our estimation for the
market sharpe ratio by 4.6% (from 42.5% to 37.9%). Especially extracting the part due to
liquidity risk may increase the accuracy of our estimation.
We expect the recovery rate modeling to be another focus of further research. Moodys
(2007) indicates a significant negative correlation between realized recovery rate and realized default rate. It remains to be shown, if this relationship holds true for expected values
as well. Different recovery rates for specific sectors/companies may also explain some of the
differences between the implicit sharpe ratio measured.

4 CONCLUSION

Base Case
CDS spread -10%
CDS spread +10%
CDS spread +10bp
CDS spread -10bp
RR +10%
RR -10%
EDF -10%
EDF +10%
Correlation +10%
Correlation -10%

37

Company sharpe ratio

19.56%
17.40%
21.55%
24.37%
12.79%
21.76%
17.60%
21.44%
17.83%
19.56%
19.56%

Market sharpe ratio

42.46%
37.90%
46.66%
52.54%
28.33%
47.11%
38.33%
46.41%
38.84%
38.60%
47.18%

Market sharpe ratio (at)

32.13%
28.56%
35.46%
40.15%
21.82%
35.81%
28.89%
35.32%
29.25%
29.21%
35.70%

Table 5: Sensitivities of sharpe ratio estimation based on main input parameters. Bold:
CDS spread -10%: shows sensitivity, if only a part of the CDS spread would be attributable
to credit risk (e.g. due to bid/asd spreads, liquidity, taxes). Sharpe ratio market (at) denote
the estimation for the market sharpe ratio after tax adjustment.

Conclusion

Our contribution to the existing literature was twofold. First, we have theoretically analyzed
the Q-to-P-ratio and the estimation of sharpe ratios out of CDS spreads based on a simple
Merton model and a more advanced structural model of default by Duffie/Lando (2001) including unobservable asset values. The theoretical results do - at least qualitatively - confirm
findings from other empirical studies, e.g. Berndt et.al. (2005) and Amato et.al. (2007),
especially an increasing Q-to-P-ratio for higher quality companies. They do also show an
astonishing robustness of a simple estimator of the sharpe ratio derived from a Merton framework. In effect, although actual and risk neutral default probabilities are largely effected by
model changes, the estimator for the Merton estimator for the sharpe ratio - which is based
on both actual and risk neutral default probabilities - is merely affected for investment grade
companies with an asset volatility above 10%, which can reasonable assumed to hold for all
non-financial services companies.
An empirical analysis of 125 companies from a North Americal investement grade CDS
index (CDS.NA.IG) has shown, that approximately 80% of the credit spread of CDS can be
explained by credit risk. This portion is significantly larger than prior reseach for the bond
market suggests, see for example Driessen (2003) or Huang (2003).
The average market sharpe ratio derived from the Merton estimator equals 42%, a simple
tax adjustment results in a market sharpe ratio of 33%. Accounting in addition for the fact,
that some portion of the credit spread does not seem to be attributable to credit risk, these
findings strongly confirm analysis on the equity premium, which have shown that equity
prices imply market sharpe ratios of 15-30% (equivalent to an equity premium of appr. 35%), far below the average realized sharpe ratios in the second half of the 20th century of
appr. 44% (equivalent to an equity premium of appr. 7.4%), see Fama/French (2002) and

A CORRELATION BETWEEN ASSET AND EQUITY RETURNS

38

Claus/Thomas (2001) among others.

We see three main areas for further research: First, a more accurate determination of the
part of the CDS spread attributable to credit risk seems to be very important. This, together with a more thourough treatment of the tax differences on equity and debt markets
and a deeper analysis of the effect of deviations from the lognormal distribution assumption51
could help to improve the accuracy of the overall level of our implicit sharpe ratio estimation. Third, empirical analysis has shown, that the assumption of constant recovery rates
is suboptimal. Recovery rates differ in reality by sectors and vary through time. A more
accurate estimation of the expected risk neutral recovery rate can help to more accurately
determine the implicit sharpe ratio for individual sectors and over time.

Correlation between asset returns and equity returns

In this appendix, we will show that the correlation between asset returns and equity returns
in the Merton framework as well as in the Duffie/Lando framework52 is approximately one.
The economic reason is quite simple: as equity is modelled in both frameworks as a deepin-the-the-money53 call option on the companies assets54 , the sensitivity of the option with
respect to the asset value is almost linear (i.e. the delta of the option is almost one and
therefore gamma is close to zero). A correlation of one is equivalent to a positive linear
relationship, so as the relationship is almost linear, the correlation will be approximately
one. The correlation will be the smaller, the less linear the relationship is, i.e. the higher the
default probability (and therefore the less in-the-money the option) and/or the higher the
effect of other input parameters that lead to non-linearities with respect to the asset value
(e.g. taxes, insolvency costs in the Duffie/Lando framework).
Qualitatively spoken, the correlation will be highest for high credit quality companies in
the Merton framework and lowest for low credit quality companies in the Duffie/Lando
framework. Quantitavely, the correlation will always be above 99% for investment grade
51

A possible analysis should take at least two steps: first, one should derive asset returns out of equity
returns based on option pricing models; second a calibration of the actual returns to a family of distributions
has be performed in order to smooth the results. A possible candidate distribution family could be alpha
stable distributions, as independent linear combinations of alpha stable distributions will belong to the same
distribution family, which is important if one assumes independent, stationary increments of the log returns.
52
Which is, concerning the relationship between asset and equity value, basically the same as the Leland/Toft (1996) model.
53
Of course, this option is not by definition deep-in-the-money but rather depends on the closeness of
the asset value to the default barrier. Looking at investment grade ratings, the maximum one-year default
probability is approximately 0.4%, even extending the rating grades up to single B rating results in a oneyear default probability of not more than 10%. So we look at options, that are executed with an (acutal)
probability of at least 99.6% (investment grade) and 90% (single B) respectively. This motivates the use of
the term deep-in-the-money.
54
The sort of option is of course different in both frameworks: a plain vanilla, european, call option in the
Merton framework and a knock-out option in the Duffie/Lando framework.

A CORRELATION BETWEEN ASSET AND EQUITY RETURNS

39

companies and above 95% for companies with a rating of at least B in both frameworks, as
we will show in the following subsections.

A.1

Correlation between asset and equity values in the Merton

framework

The returns of the equity (RET ) and the asset value (RVT ) over the time horizon [0, T ] are
defined as
ET
RET =
E0
and
VT
RVT =
.
V0
respectively. Therefore, the correlation between RET and RVT can be calculated as follows
in the Merton framework:
ET VT
P (RET , RVT ) = P ( , ) = P (ET , VT ) = P ((VT N )+ , VT )
E V
0 0

+
P
(1/2 2 )T +WT
(1/2 2 )T +WT
=
V0 e
N , V0 e
.
There is no analytical solution, therefore we have to draw back on numerical approximations.
The calcuation is straightforward, the result can be found in figure 17. Note, that each rating
grade is identified by a specific cumulative default probability based on Moodys (2007).
Therefore, in the first step, combinations of the parameters (e.g. asset volatility, asset value)
were depicted, that lead to the same cumulative default probability.55 In the next step, the
correlations for these parameter combinations were numerically evaluated and the minimum
and mean was plottet in figure 17. Plotting correlations (and other parameters) with respect
to the rating grade in a structural model may seem odd at first, but given the information
available to a common investor (i.e. the rating), this is the only reasonable way. In addition,
the rating grade is a very convenient way of clustering the results, since correlations are very
similar within a rating grade, e.g. a high volatility combined with high current asset value
leads to a similar rating and a similar correlation than a low volatility combined with a
low current asset value.

A.2

Correlation between asset and equity values in the Duffie/Lando

framework

In this subsection, we will analyze the correlation between asset and equity values in the
Duffie/Lando framework, as a special case, this will also cover the correlation in the first
55

Precisely, this means that all parameters but the asset value were fixed
 and the asset value was choosen
2
ln VN (r 12 Assets
)T
0

as to result in the cumulative default probability

of the respective rating grade.

T
Assets

A CORRELATION BETWEEN ASSET AND EQUITY RETURNS

40

Figure 17: Correlation between equity and asset value in the Merton framework. Parameter
combinations for calculating the minimum/mean: : 3% 30%, company sharpe ratio:
10% 40%, other parameters: r=6%, T (maturity)=5, N=100, Vt choosen to fit target
actual default probability for respective rating grade.

passage time framework with obvservable asset values. The procedure is basically the same
as in the Merton framework though the determination of a set of parameters leading to a
specific default probability cannot be analytically calculated but must be found by numerical
techniques. The result is shown in figure 18. As discussed at the beginning of this appendix,
the correlations are slightly smaller than in the Merton framework, but still above 99% for
investment grade companies. Extending the analysis to maturities smaller or larger than 5
years does not significantly change the result, further analysis is available upon request.
All in all, asset values and equity values do always move in the same direction in the
Duffie/Lando framework, e.g. the equity value always rises with rising asset value, all other
parameters being equal. We do though want to point out, that there may be states, where
asset and equity values do not move in the same directions in reality, if we allow other
parameters to be volatile. One famous example is asset substitution (e.g. equity holders
increase the volatility of the assets), where the asset value stays constant56 and the equity
value increases. Others may be the strategic setting of the dividend/payout rate or a nonconstant default barrier. It is not within the scope of this paper to explicitly evaluate the
56

Or may even decline, if the rising asset volatility is bought by negative NPV assets.

B DETAILS ABOUT ADJUSTMENT FACTORS IN THE DUFFIE/LANDO FRAMEWORK41

effect on the correlation between asset and equity value, we do though not expect our results
to materially differ under these scenarios given the very high correlations observed in our
models.57

Figure 18: Correlation between equity and asset value in the Duffie/Lando framework. Parameter combinations for calculating the minimum/mean: : 3% 30%, company sharpe
ratio: 10% 40%, risk neutral asset growth rate after payouts: 0% 5%, (asset value
uncertainty): 0% 30%, T1 (time since last certain asset value information): 0y 3y, other
parameters: r=6%, T (maturity)=5, default barrier=100, Vt choosen to fit target actual
default probability for respective rating grade.

Details about adjustment factors in the Duffie/Lando

framework

The following tables show adjustment factors for the sharpe ratio estimation in the Duffie/Lando
framework for different parameter combinations.
57

As an indication, one could use the correlation between asset and equity value given in the KMV-model.
For our data set, the correlation is on average 90%, the median is 95%. The lower average is mainly due to
a small set of companies with correlations below 70%, which is usually due to large one-time effect on the
asset value and/or default point.

C DETAILS SHARPE RATIO ESTIMATION

Adjustment factor
m
0%

Average (m=0%)
2,5%

Average (m=2.5%)
5%

Sharpe ratio
0,1
0,2
0,3
0,4
0,1
0,2
0,3
0,4
0,1
0,2
0,3
0,4

Average (m=5%)
Average

Table 6:

42

10%
1,04
1,05
1,06
1,08
1,06
1,08
1,10
1,12
1,14
1,11
1,15
1,18
1,22
1,27
1,21
1,12

15%
1,04
1,05
1,06
1,07
1,06
1,06
1,07
1,09
1,11
1,08
1,09
1,11
1,14
1,17
1,13
1,09

20%
1,04
1,05
1,06
1,07
1,05
1,05
1,06
1,08
1,09
1,07
1,07
1,08
1,10
1,13
1,10
1,07

30%
1,04
1,04
1,05
1,06
1,05
1,04
1,05
1,06
1,07
1,06
1,05
1,06
1,07
1,09
1,07
1,06

Average
1,04
1,05
1,06
1,07
1,05
1,06
1,07
1,09
1,10
1,08
1,09
1,11
1,13
1,16
1,12
1,09

Adjustment factors for a rating of Baa and a maturity of 5 years for a scenario of observable asset values (i.e. s=0 and = 0%).

Other parameters: m=asset growth rate = risk free rate - payout rate, =asset volatility, sharpe ratio = sharpe ratio of asset value process (before
payout rate). Bold numbers denote the Min/Max, itallic numbers denote typical value firms.
Adjustment factor
m
0%

Average (m=0%)
2,5%

Average (m=2.5%)
5%

Average (m=5%)
Average

Table 7:

Sharpe ratio
0,1
0,2
0,3
0,4
0,1
0,2
0,3
0,4
0,1
0,2
0,3
0,4

10%
0,87
0,88
0,89
0,91
0,89
0,91
0,93
0,96
0,99
0,95
0,99
1,03
1,08
1,13
1,06
0,96

sigma
15%
20%
0,90
0,93
0,91
0,93
0,92
0,94
0,93
0,96
0,92
0,94
0,92
0,94
0,93
0,95
0,95
0,97
0,98
0,99
0,95
0,96
0,95
0,96
0,98
0,98
1,01
1,00
1,04
1,02
1,00
0,99
0,95
0,96

30%
0,97
0,97
0,98
0,99
0,98
0,97
0,98
0,99
1,00
0,99
0,98
0,99
1,01
1,02
1,00
0,99

Average
0,92
0,92
0,93
0,95
0,93
0,93
0,95
0,97
0,99
0,96
0,97
0,99
1,02
1,06
1,01
0,97

Adjustment factors for a rating of Baa and a maturity of 5 years for a scenario of highly unobservable asset values (i.e. s=3 and

= 30%). Other parameters: m=asset growth rate = risk free rate - payout rate, =asset volatility, sharpe ratio = sharpe ratio of asset value
process (before payout rate). Bold numbers denote the Min/Max, itallic numbers denote typical value firms.

Details sharpe ratio estimation

The following tables show the results of the sharpe ratio estimation for each single company
per sector, see table 8 and 9. Please note: Moodys ratings for financial service companies
are usually not comparable with non-financial services companies as the underlying default
probabilities are usually assumed to be different, e.g. a A2-rating of a financial services
company has a lower default probabiliy than a A2-rating of a non-financial services company.
This explains the higher Moodys sharpe ratios for financial services companies.

C DETAILS SHARPE RATIO ESTIMATION

Sector
Communications
Technology

Company
and

Q-to-P
(EDF)
3.19

SR company
(EDF)
20.24%

SR
market
(EDF)
33.15%

SR company
(Moodys)
33.48%

SR
market
(Moodys)
54.53%

2.51
5.78
7.23

14.80%
31.01%
37.24%

28.33%
67.26%
98.34%

33.67%
23.04%
14.37%

63.11%
49.18%
39.25%

2.52
1.09
5.71
2.20
2.07
5.65
5.09
3.77
4.86
3.86

12.69%
-1.78%
33.75%
13.71%
12.82%
31.71%
30.24%
19.81%
25.49%
20.92%

30.04%
-1.65%
56.18%
21.71%
26.16%
92.62%
70.35%
38.10%
48.14%
44.61%

31.18%
17.30%
13.21%

66.78%
29.64%
21.97%

16.14%
22.57%
17.67%
26.58%
33.61%
22.87%

29.19%
64.88%
36.91%
43.92%
62.26%
44.94%

Autozone Inc
Cardinal Health Inc
Caterpillar Inc
CBS Corp
Centex Corp
ConAgra Foods Inc
CVS Corp
Deere & Co
J C Penney Co Inc
Jones Apparel Gp Inc
Lennar Corp
Ltd Brands Inc
Marriott Intl Inc
McDonalds Corp
McKesson Corp
Nordstrom Inc
Pulte Homes Inc
Sherwin Williams Co
Southwest Airls Co
Starwood Hotels & Resorts
Wwide Inc
Target Corp
Toll Bros Inc
Wal Mart Stores Inc
Walt Disney Co
Whirlpool Corp

4.77
3.04
1.87
5.09
2.15
4.59
2.91
1.68
5.09
8.66
3.13
2.62
5.53
2.47
1.33
2.36
2.27
4.93
3.97
11.01

28.55%
20.04%
10.67%
29.74%
13.43%
26.94%
17.32%
8.66%
30.73%
36.39%
22.58%
14.15%
30.64%
14.54%
4.62%
13.82%
16.19%
27.64%
24.87%
47.70%

70.18%
65.31%
16.42%
74.23%
24.00%
77.89%
38.79%
13.72%
70.11%
72.82%
44.36%
28.58%
52.72%
30.27%
12.82%
27.16%
31.37%
64.64%
50.33%
88.14%

14.49%

37.65%

31.07%

77.29%

15.64%
19.11%
16.42%
2.48%
21.86%
12.29%
17.39%
14.59%
26.29%
9.23%
27.08%
17.08%
29.08%

41.33%
41.96%
26.33%
5.29%
42.33%
22.71%
33.36%
23.51%
52.27%
25.64%
51.60%
32.60%
72.99%

1.96
3.31
2.54
3.53
2.03
3.71

10.78%
24.32%
15.57%
22.59%
11.37%
20.72%

21.13%
46.27%
28.03%
36.54%
19.85%
43.71%

Altria Gp Inc
Amgen Inc.
Baxter Intl Inc
Boston Scientific Corp
Bristol Myers Squibb Co
Campbell Soup Co
Gen Mls Inc
Loews Corp
Safeway Inc
Sara Lee Corp
The Kroger Co.
Tyson Foods Inc
Unvl Health Svcs Inc
Wyeth

11.93
3.80
1.83
2.98
1.94
3.27
6.50
1.46
4.59
5.62
5.10
4.53
3.52
2.89
4.20
3.19
2.71
2.87
2.35
2.68
3.22
2.84

50.60%
22.56%
10.21%
19.55%
10.56%
20.24%
33.79%
6.43%
27.02%
27.84%
29.73%
29.14%
23.59%
18.90%
23.15%
17.50%
16.53%
18.70%
14.86%
15.32%
21.88%
17.50%

186.98%
66.83%
35.72%
63.81%
20.84%
54.45%
114.28%
10.38%
56.96%
78.13%
61.06%
107.64%
90.10%
49.51%
67.33%
45.29%
32.85%
53.59%
35.06%
36.72%
53.25%
42.95%

ALLTEL Corp
AT&T Inc
CenturyTel Inc
Comcast Cable Comms
LLC
Computer Sciences Corp
Hewlett Packard Co
IAC InterActiveCorp
Intl Business Machs Corp
Omnicom Gp Inc
R R Donnelley & Sons Co
Sprint Nextel Corp
Time Warner Inc
Verizon Comms Inc

Communications
Technology: Average
Consumer Cyclical

and

Consumer Cyclical: Average

Consumer Stable

Consumer Stable: Average

Energy

43

Anadarko Pete Corp

ConocoPhillips
Devon Engy Corp
Halliburton Co
Transocean Inc
Valero Energy Corp

Energy: Average

Table 8:

19.84%

36.59%

19.99%

37.49%

21.92%
20.15%
17.60%

34.58%
36.18%
35.34%

36.60%
23.71%
14.85%
8.70%
31.86%
17.68%
20.60%
20.38%

140.35%
70.78%
52.87%
26.44%
58.46%
48.08%
70.95%
34.48%

24.56%
10.39%
18.15%

68.80%
20.54%
62.82%

24.32%
20.33%
12.80%
21.60%
10.33%
17.78%
12.42%
9.62%
14.03%

59.24%
55.84%
31.50%
42.79%
28.52%
42.91%
27.77%
19.68%
32.10%

Details of sharpe ratio estimation per sector and company. Continued in table 9.

C DETAILS SHARPE RATIO ESTIMATION

Sector

Company

Financial

ACE Ltd
Amern Express Co
Amern Intl Gp Inc
Boeing Cap Corp
Cap One Bk
Chubb Corp
Cigna Corp
Ctrywde Home Lns Inc
Gen Elec Cap Corp
Hartford Finl Svcs Gp Inc
Marsh & Mclennan Cos
Inc
MBIA Ins Corp
MetLife Inc
Radian Gp Inc
Simon Ppty Gp L P
WA Mut Inc
Wells Fargo & Co
XL Cap Ltd

Financial: Average
Industrial

Industrial Ergebnis
Materials

Materials: Average
Utilities

Arrow Electrs Inc

Burlington Nthn Santa Fe
Corp
Carnival Corp
CSX Corp
Goodrich Corp
Honeywell Intl Inc
Ingersoll Rand Co
Lockheed Martin Corp
Motorola Inc
Newell Rubbermaid Inc
Norfolk Sthn Corp
Northrop Grumman Corp
Raytheon Co
Un Pac Corp
Alcan Inc.
Dow Chem Co
Eastman Chem Co
Intl Paper Co
MeadWestvaco Corp
Olin Corp
Rohm & Haas Co
Temple Inland Inc
Weyerhaeuser Co
Amern Elec Pwr Co Inc
Constellation Engy Gp Inc
Dominion Res Inc
Duke Energy Carolinas
LLC
Progress Engy Inc
Sempra Engy

Utilities: Average
Total: Average

Table 9:

44

Q-to-P
(EDF)
2.29
2.77
1.81
2.00
2.22
2.33
2.35
1.82
1.59
1.64
4.49

SR company
(EDF)
14.49%
17.52%
9.83%
12.67%
14.02%
12.70%
11.87%
11.20%
6.23%
8.90%
26.35%

SR
market
(EDF)
25.50%
22.40%
16.74%
27.85%
26.01%
24.74%
42.45%
27.43%
9.90%
13.03%
59.67%

2.73
2.02
4.74
4.46
3.13
4.18
2.22
2.74
3.52
3.90

18.39%
11.97%
28.30%
24.76%
20.81%
24.44%
14.62%
16.32%
26.25%
23.52%

30.27%
22.60%
54.33%
61.36%
38.73%
38.14%
30.72%
33.21%
40.49%
40.80%

4.74
3.90
1.92
1.63
2.70
3.20
1.56
3.18
2.85
4.65
3.22
5.89
3.17
2.85
3.26
3.36
4.49
3.53
2.55
2.63
6.30
4.91
3.62
2.61
2.05
3.23
1.48

27.91%
22.56%
12.57%
7.73%
16.70%
20.67%
7.72%
20.86%
16.98%
26.72%
20.29%
31.61%
19.20%
17.79%
21.21%
22.87%
27.75%
23.37%
18.04%
16.82%
34.80%
29.04%
22.91%
16.90%
11.76%
21.34%
3.95%

41.94%
40.00%
27.09%
12.20%
24.12%
61.89%
16.77%
37.97%
34.32%
68.59%
41.62%
60.37%
36.17%
29.69%
36.90%
41.06%
43.28%
39.51%
35.87%
27.16%
74.46%
50.15%
40.23%
36.55%
28.11%
53.07%
9.00%

2.22
1.82
2.21
3.38

14.50%
9.67%
12.75%
19.56%

35.49%
19.01%
29.76%
42.46%

SR company
(Moodys)
28.65%

SR
market
(Moodys)
49.05%

36.17%
23.31%
17.44%
32.68%
9.00%
31.44%

70.81%
50.22%
29.63%
60.12%
28.60%
77.24%

21.88%
22.16%

31.36%
42.31%

19.74%

37.20%

33.87%
8.00%
34.82%
24.07%

57.39%
12.51%
70.07%
47.22%

9.00%

16.28%

26.96%
6.04%
9.90%

40.17%
9.78%
17.01%

3.02%
11.15%
19.63%

10.69%
26.09%
35.28%

4.44%
1.67%
10.36%
11.78%

11.80%
3.53%
20.48%
21.94%

18.72%
17.92%
15.50%
17.91%

30.39%
32.16%
23.36%
29.47%

19.14%
14.95%
17.55%
17.49%
3.53%
18.47%
17.32%

30.84%
31.75%
29.29%
29.39%
6.98%
41.50%
43.58%

18.20%
15.51%
18.70%

37.35%
35.01%
39.01%

Details of sharpe ratio estimation per sector and company. Please note: Moodys ratings for financial service companies are usually

not comparable with non-financial services companies as the underlying default probabilities are usually assumed to be different, e.g. a A2-rating
of a financial services company has a lower default probabiliy than a A2-rating of a non-financial services company. This explains the higher
Moodys sharpe ratios for financial services companies.

D MAPPING OF MOODYS RATING GRADES TO DEFAULT PROBABILITES

45

Mapping of Moodys rating grades to default probabilites

For the mapping of Moodys rating grades to default probabilities, we used the raw data
provided by Moodys (2007). We used a log-linear relationship to calibrate the default
probabilities, i.e.
ln(P D) = 1 + 2 N RG,
where NRG denotes the numerical rating grade ranging from 1 (Aaa) to 16 (Baa3). This
yielded the following cumulative default probabilities per rating grade:
Rating
Aaa
Aa1
Aa2
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Ba1
Ba2
Ba3
B1
B2
B3

1
0.001
0.002
0.004
0.007
0.013
0.024
0.045
0.083
0.152
0.279
0.514
0.946
1.741
3.204
5.896
10.850

2
0.006
0.011
0.018
0.032
0.055
0.095
0.164
0.285
0.493
0.855
1.482
2.569
4.452
7.716
13.373
23.178

3
0.019
0.032
0.052
0.085
0.139
0.228
0.373
0.611
1.002
1.642
2.691
4.411
7.229
11.847
19.417
31.823

4
0.041
0.065
0.103
0.163
0.256
0.404
0.637
1.004
1.583
2.496
3.934
6.202
9.778
15.414
24.301
38.310

5
0.071
0.108
0.166
0.255
0.392
0.601
0.922
1.416
2.173
3.335
5.120
7.858
12.061
18.512
28.413
43.611

6
0.094
0.143
0.218
0.331
0.502
0.763
1.159
1.760
2.674
4.063
6.172
9.376
14.243
21.638
32.871
49.936

7
0.114
0.173
0.261
0.394
0.596
0.900
1.360
2.056
3.107
4.695
7.095
10.722
16.204
24.488
37.007
55.926

8
0.125
0.190
0.287
0.434
0.657
0.995
1.506
2.278
3.447
5.217
7.894
11.946
18.076
27.354
41.392
62.635

9
0.135
0.205
0.310
0.470
0.712
1.078
1.633
2.475
3.750
5.681
8.607
13.041
19.759
29.936
45.356
68.719

Table 10: Cumulative default probabilities for Moodys ratings in percent.

10
0.148
0.224
0.339
0.512
0.775
1.173
1.774
2.683
4.058
6.139
9.286
14.046
21.247
32.139
48.615
73.538

REFERENCES

46

References
[1] Altmann, E.I./Kishore, V.M.: Almost Everything You Wanted to
Know about Recoveries on Defaulted Bonds, in: Financial Analysts
Journal, Vol. 52, No. 6, 1996, p. 57-64.
[2] Amato, J.: Risikoaversion und Risikopramien am CDS-Markt, BIS
Working Paper, 2005.
[3] Amato, J../Remolona, E.: The pricing of unexpected credit losses, BIS
Working Papers, 2005.
[4] Berndt, A./Douglas, R., Duffie,D./Ferguson, M./Schranz, D.: Measuring default risk premia from default swap rates and EDFs, BIS Working Papers, No 173, March 2005.
[5] Black, F./Scholes, M.: The Pricing of Options and Corporate Liabilities, in: Journal of Political Economy, 1973.
[6] Bluhm, C./Overbeck, L./Wagner, C.: An Introduction to Credit Risk
Modeling, Chapman & Hall, 2003.
[7] Bohn, J.: An Empirical Assessment of a Simple Contingent-Claims
Model for the Valuation of Risky Debt, in: Journal of Risk Finance,
Summer, p. 55-77.
[8] Carty, L.V./Lieberman, D.:
Defaulted Bank Loan Recoveries,
Moodys Investors Service, Global Credit Research, Special Report,
Nov. 1996.
[9] Claus, J./Thoma, J.: Equity Premia as Low as Three Percent? Evidence from Analysts Earnings Forecasts for Domestic and International Stock Markets, in: Journal of Finance, Oct. 2001, p. 1629-1666.
[10] Delianedis, G./Geske, R.: Credit Risk and Risk Neutral Default Probabilities: Information about Rating Migrations and Defaults. Working
Paper, Anderson Graduate School of Business, University of California, Los Angeles, 1998.
[11] Driessen, J.: Is Default Event Risk Priced in Corporate Bonds?, Working Paper, University of Amsterdam, 2003.
[12] Duffie, D./Lando, D.: Term Structure of Credit Spreads with Incomplete Accounting Information , Econometrica, Vol. 69, No. 3, p. 633664, Mai 2001.
[13] Duffie, D./Singleton, K.: Modeling Term Structures of Defaultable
Bonds, in: Review of Financial Studies, Vol. 12, 1999, p. 687-720.

REFERENCES

47

[14] Duffie, D./Singleton, K.: Credit Risk - Pricing, Measurement, and

Management, 1. Aufl., Princeton 2003.
[15] Embrechts, P./Kl
uppelberg, C./Mikosch, T.:
Modelling
Events for Insurance and Finance, Springer, Berlin, 1997.

Extremal

[16] Fama, E.F./French, K.R.: The Equity Premium, in: Journal of Finance, Vol. LVII, No 2, April 2002, p. 636-659.
[17] Finger, C.C./Finkelstein, V./Pan, G./Lardy, J.P./Ta, T./Tierney, J.: CreditGrades, Technical Document, RiskMetrics Group, 2002.
[18] Fons, J.: The Default Premium and Corporate Bond Experience, in:
Journal of Finance, Vol. 42, 1987, p. 81-97.
[19] Gebhardt, W./Lee, C./Swaminathan, B.: Toward an Implied Cost of
Capital, in: Journal of Accounting Research, Vol. 39, No 1, June 2001,
p. 135-176.
[20] Gordon, M.: The Investment, Financing, and Valuation of the Corporation, Irwin Publishing 1962.
[21] Huang, J./Huang, M.: How much of the Corporate-Treasury Yield
Spread is Due to Credit Risk?: Results from a New Calibration Approach, Working Paper, Graduate School of Business, Standford University.
[22] Illmanen, A.: Expected Returns on Stocks and Bonds, in: Journal of
Portfolio Management, Vol. 29, Issue 2, Winter 2003, p. 7-28.
[23] Leland, H.E./Toft, K.: Optimal Capital Structure, Endogenous
Bankruptcy, and the Term Structure of Credit Spreads, in: The Journal of Finance, Vol. 51, p. 987-1019, 1996.
[24] Liu, J./Longstaff, F./Mandell, R.: The market price of credit risk: An
empirical analysis of interest rate swap spreads, Working Paper, 2000.
[25] Mehra, R./Prescott, E.C.: The Equity Premium: A Puzzle, in: Journal of Monetary Economics, Vol. 15, No. 2, p. 145-161, 1985.
[26] Mehra, R.: The Equity Premium: Why is it a Puzzle, Working Paper,
2003.
[27] Merton, R.C.: On the pricing of corporate debt: the risk structure of
interest rates, in: Journal of Finance, Vol. 29, p. 449-470, 1974.

REFERENCES

48

[28] Moodys Investors Service: Corporate Default and Recovery Rates,

1920-2006, 2007.
[29] Moodys KMV: EDF 8.0 Model Enhancements: Modeling Methodology, 2007.
[30] Musiela, M./Rutkowski, M.: Martingale Methods in Financial Modelling, Berlin u.a. 1997.
[31] Ohlson, J.: Earnings, Book Values and Dividends in Security Valuation, in: Contemporary Accounting Research, p. 661-687, Spring 1995.
[32] Philips, T.K.: Estimating Expected Returns, in: Journal of Investing,
Vol. 12, Issue 3, Fall 2003, p. 49-58.
[33] Rapp, M./Schwetzler, B.: Asset Prices in the Presence of a Tax Authority, in: German Working Papers in Law and Economics, paper
33, Vol. 2006.
[34] Schonbucher, P.: Credit Derivatives Pricing Models, 1. Aufl., Chichester 2003.
[35] Zhou C.: A jump-diffusion approach to modeling credit risk and valuing defaultable securities, Finance and Economics Discussion Paper,
1997.