You are on page 1of 14

Original Article

Value premium and default risk


Received (in revised form): 27th January 2014

Mohammed M. Elgammal
has worked for the past 18 years in Egypt, the United Kingdom and Qatar Academia. He has worked in International Commercial
Bank, Menoua University, Aberdeen University, Westminster University and Qatar University. He has many publications in different
areas of nance, including Financial Predictors of Credit Ratings, Liquidity Crisis, Financial Distress, market anomalies, corporate
governance, bank performance and macroeconomic risk factors. He has been awarded research grant from different internationally
respected organizations, including Qatar University, Egyptian Government, Suez Canal University, Economic and Social Research
Council and Qatar National Research Foundation.

David G. McMillan

PY

is a professor of Finance at the University of Stirling. He has held a similar position at the University of St Andrews and previous
positions at the Universities of Aberdeen and Durham. His research interests include forecasting asset returns and volatility,
modelling the linkages between asset prices and macroeconomic variables and examining the behaviour of nancial and investor
ratios. He has published widely on these topics in internationally respected peer-reviewed journals such as the Journal of Banking
and Finance, Journal of International Money and Finance, the Journal of Forecasting, International Journal of Forecasting and the
Oxford Bulletin of Economics and Statistics. He is a co-editor for the Economics and Finance Research journal and sits on the
editorial board of numerous internationally respected journals, including the European Journal of Finance and the Journal of Asset
Management.

Correspondence: Mohammed M. Elgammal, College of Business and Economics, Qatar University, Doha, Qatar
E-mail: m.elgammal@qu.edu.qa

TH

ABSTRACT This article investigates the relationship between value premium and nancial
distress using a long US data set over 19272011. The measures of leverage and default are
used as proxies for nancial distress when applying a time-varying volatility methodology.
The article examines the potential risk-based explanation for the source of the value
premium. The empirical analysis shows that both the default premium and its volatility have
positive explanatory power for the value premium and its volatility. The ndings suggest a
negative association between the lagged values of the default premium and the current small
stocks value premium. Investigating the reasons behind this association uniquely uncovers
an asymmetric correlation between returns on both value and growth stocks and default risk
before and post July 1954 after a change in the monetary regime in the United States.
Journal of Asset Management (2014) 15, 4861. doi:10.1057/jam.2014.10; published online 20 February 2014

Keywords: value premium; leverage; GARCH; TARCH; nancial distress; default premium

INTRODUCTION
This article investigates the nature of a
variable that has become one of the most
important and researched variables identied
in the literature as containing potential
information about systematic risk: the value
premium. Motivated by the controversial
debate concerning the rationale behind the
value premium, this study examines whether

the value premium is related to nancial


distress, consistent with the view that
systematic risk is the key driver of the
premium. That is, this work adds to the
literature by investigating whether the
aggregate default risk can explain the value
premium. More specically, this article
examines the nature of such a relationship by
implementing the GARCH methodology to

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861
www.palgrave-journals.com/jam/

Value premium and default risk

PY

Conversely, a part of the literature suggests


that there is no relationship between default
and the value premium; for example, Grifn
and Lemmon (2002) and Mohanram (2005)
suggest that the value premium is large in
small rms with low analyst coverage
regardless of whether that rm is among
those that are highly distressed. Moreover,
Liu (2006), Watanabe and Watanabe (2008)
and Akbas et al (2010) provide evidence for a
risk-based explanation for the value
premium focusing on the liquidity risk of
value and growth stocks, in addition to their
market risk. The liquidity risk argument is
motivated by the ight-to-quality
phenomenon. That is, investors tend to
switch riskier assets (value stocks) to safer
ones (growth stocks) in bad times.
This articles data set is long enough for
meaningful tests to be conducted on the
relationship between default and the value
premium. If the relationship between default
risk and the value premium is positive, this may
provide supportive evidence for the
proposition that the value premium is indeed a
proxy for systematic risk and could provide an
explanation for a leverage effect in the value
premium.2 A positive relationship between
the value premium and the default premium
could be evidence for a relationship between
leverage and value premium. This is because
levered rms face higher levels of default (see,
for example, Black et al, 1972; Merton, 1974;
Ivaschenko, 2003; Liu et al, 2007).
The main ndings of this article suggest
possible answers for the initial questions
about the role of the default premium as a
macroeconomic risk factor that can explain
the source of the value premium. The
empirical ndings contribute to our
knowledge by providing additional
evidence for the positive association
between default and value premiums.
These ndings support the rational
explanation for the value premium. The
results suggest that the value premium is
working as proxy for non-diversiable
factors related to nancial distress.

TH

capture the effects of time-varying volatility


using a long set of US data (19272011). This
article is one of the rst to link this framework
of analysis and the relationship between the
value premium and default risk using rm
characteristics (for example, size) together
with monthly data that includes the liquidity
crisis period in the analysis.1 This technique
allows an investigation of the relationship
between the value premium and default risk
between both their conditional mean and
conditional variance. This should enable us to
dene more precisely the potential source of
aggregate macroeconomic non-diversiable
risk. Finally, the time-varying volatility
approach allows examination of the impact of
leverage on the value premium, thereby
incorporating both default measures of risk
and leverage.
An element of the value premium appears
to be predictable and there is some evidence
that links the value premium to measures of
default (Zhang, 2005; Fama and French,
2006, 2007). Garlappi and Yan (2011), among
others, suggest that the value premium
increases with the default risk probability. A
possible explanation is that during a recession
(when default is likely to rise), investors
perceive value stocks to have a greater
relative risk compared with growth stocks
(see, for example, Ivaschenko, 2003;
Penman et al, 2007). This argument is
developed from the original work of Fama
and French (1996) who in turn argued that
value rms have poorer performance,
earnings and protability compared with
growth rms. Thus, when the economy
experiences a downturn, investors require a
higher return on value stocks compared with
growth stocks as compensation for higher
exposure to nancial distress. Further,
Vassalou and Xing (2004) suggest that both
size and value premiums are concentrated in
high default risk rms, while Campbell et al
(2008) report that distressed stocks have a
higher loading on value and small-cap risk
factors compared with lower probability of
failure stocks.

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

49

Elgammal and McMillan

TH

The relationship between leverage and


default is originally investigated using the
approach of Merton (1974). The main
concept of this model is that the rm defaults
on its debt if its assets value reaches a certain
default point. Liu et al (2007) illustrate that the
barrier model of Merton implies that the
credit risk of corporate debt is a function of
leverage and the assets volatility. The
association between leverage and default
suggests that examining the effect of default
on the value premium may help us to
understand the association between leverage
and the value premium and to overcome the
problem of leverage measurement at the
aggregate level.
Ivaschenko (2003) demonstrates a
relationship between leverage and default
risk, where higher leveraged rms are
perceived to be more risky. Ivaschenko
(2003) argues that an increase in corporate
leverage is accompanied by rising corporate
default and a decrease in recovery rates. This is
consistent with Anderson et al (1996) whose
corporate debt model refers to a positive
relationship between the probability of
default and balance sheet leverage. Moreover,
these results are further supported by Molina
(2005) who reports a strong positive
association between leverage and default
probability. Furthermore, Chan-Lau (2006a, b)

PY

LEVERAGE, DEFAULT AND THE


VALUE PREMIUM

suggests that the default probability helps in


assessing nancial distress facing the nancial
system at the national, regional and global
level. Crosbie and Bohn (2003) highlight that
the default risk increases as the value of a
rms assets approaches the book value of
liabilities. This can happen as a result of
increased borrowing as a percentage of total
assets or of eroding asset value. In addition,
they report a positive relationship between
the market leverage and default probability.
Finally, Ivaschenko (2003) suggests that the
increase in leverage payments may diminish
the recovery of the economy by creating
liquidity crises. The documented connection
between leverage and default risk in the
literature thus helps motivate an examination
of the relationship between default and value
premium which may add to our knowledge
by explaining the impact of leverage on the
value premium at the aggregate level.
Chan and Chen (1991) and Fama and
French (1996) suggest that the value premium
is associated with nancial risk. Since value
rms have poor performance, earnings and
protability, while growth rms signal
persistently strong earnings and protability.
Fama and French (1996) argue that the price
of value stocks is driven down due to nancial
distress. If nancial distress does not lead rms
to bankruptcy, then the stock price of those
rms will recover generating high average
returns. In the same context, Chen and
Zhang (1998) nd that value stocks have
higher returns compared with growth stocks
because they are usually rms under distress,
have high nancial leverage and face
substantial uncertainty in future earnings. This
interpretation leads them to argue that value
stocks have a large amount of market
leverage. Indeed, in a recent paper, Choi
(2013) highlighted the role of leverage within
the value premium.
In contrast, Dichev (1998) nds a low
correlation between book-to-price ratios and
both O-scores and Altman Z-scores and both
high and low O-scores are associated with
relatively low returns. This may suggest that

The remainder of this article is organized as


follows. The next section discusses the link
between the default premium and leverage,
introducing a brief summary of literature
covering the relationship between the default
and value premiums. The subsequent section
introduces a description of the data used in
the current paper, while the latter section
explains the methodology. The penultimate
section reports the empirical ndings. Finally,
the last section summarizes the main
conclusion of the article.

50

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

Value premium and default risk

PY

and Hwang (2010) argue that book-tomarket is not a measure of nancial distress
risk but instead captures exposure to priced
risk that is unrelated to capital structure.
Watanabe and Watanabe (2008), and Akbas
et al (2010) argue that the value premium can
be explained by the liquidity risk of value
and growth stocks. The liquidity risk
argument is motivated by the ight-to-quality
phenomenon, that is, investors tend to switch
riskier assets (value stocks) to safer ones
(growth stocks) in bad times.
Black (2006) examines the relationship
between the conditional volatility of the
Fama and French three factors, value, size
and market premiums, and the conditional
volatility of the default premium as a proxy
of macroeconomic risk. Black (2006) reports
that past values of the conditional variance
for a default risk premium have information
that precedes the conditional volatility of
both value and market premiums. In
addition, Black (2006) implies that the past
values of the conditional volatility of market
premium have predictive power for the
conditional volatility of default premium.
The motivation in this article is to
investigate the relationship between the
default premium and value premium rather
than their volatilities; however, for robust
considerations, the relationships between
these variables volatilities are also examined.
The relationship between the value
premium and time-varying volatility
(typically analysed using the GARCH
approach) is also considered by Black and
McMillan (2006) and Li et al (2009). These
papers seek to examine the role of risk,
as measured by time-varying volatility,
in determining the nature of the value
premium. Given the above conicting
literature, it is hoped that understanding
the relationship between default premium
and value premium may enhance our
understanding of the association between
leverage and value premium and consequently
of the possible explanation for the source of
value premium.

TH

default risk is not a systematic risk factor.


Moreover, Grifn and Lemmon (2002)
report that the default risk is not restricted to
high book-to-market rms. Accordingly,
nancial distress may be associated with the
full range of book-to-market ratios. The
relationship between default risk and returns
in different book-to-market levels is very
controversial. For example, Piotroski (2000)
nds that, within high book-to-market
stocks, those with lower nancial health earn
lower returns, while Mohanram (2005) nds
that, within low book-to-market stocks,
those with weak growth attributes earn lower
returns. Moreover, Campbell et al (2008)
document a negative correlation between
bankruptcy risk and returns.3 Grifn and
Lemmon (2002) report a negative correlation
between O-scores and returns within low
book-to-market stocks, but a positive
correlation within high book-to-market
stocks. Vassalou and Xing (2004), using
Black-Scholes-Merton (BSM) indicators of
default probability, argue that default risk is
systematically and positively priced.4 Finally,
Ng (2005) reports that Altman Z-scores,
O-scores, and BSM indicators are positively
correlated with returns after controlling for
size and book-to-market. Penman et al (2007)
suggest that the negative correlation between
the default risk and returns may be due to a
positive correlation between default score and
leverage.
Dichev (1998) and Grifn and Lemmon
(2002) reported that the negative relationship
between returns and nancial distress is
puzzling and cannot serve as a basis for
concluding whether book-to-market
signicance is attributable to nancial distress
risk. George and Hwang (2010) explain the
puzzle arguing that both leverage and the
probability of default are inverse measures of
nancial distress costs when rms choose
capital structures optimally. George and
Hwang (2010) nd that even with leverage
and distress probability included in crosssectional regressions, book-to-market retains
its signicance in explaining returns. George

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

51

Elgammal and McMillan

DATA

LVPP

4
3
2
1
0
-1
30 35 40 45 50 55 60 65 70 75 80 85 90 95 00 05

SVPP

4
3
2

0
-1

PY

30 35 40 45 50 55 60 65 70 75 80 85 90 95 00 05

Figure 1: The value premium price indices (VPP) for


the US stock market.
The value premium price index for US large and small
stocks (VPP). All data are logged total return indices.
VPPt = ln(VTRIt)ln(GTRIt), where VTRI is value total
return index and GTRI is growth total return index.
LVPP and SVPP denote large and small value premium
price indices respectively: (a) Large rms value
premium price index; (b) Small rms value premium
price index.

TH

The data used in this study are as follows:


monthly US value and growth stocks total
returns indices for small and large rms; the
monthly US long-term corporate debt total
returns index; the monthly US long-term
government bonds total returns index.
These data are obtained from Morgan
Stanley International (MSCI) database
provided by Ibbotson Associates. The
value premium index is obtained from the
website of Ken French.5 The tests are
executed over three different samples. The
rst sample covers the data over the period
June 1927 to March 2011. The second
sample starts from July 1954 and goes to
March 2011 representing the post-war
period which may be characterized by
changes in the economic environment
and monetary regime. The third sample
includes the period of December 1991 to
March 2011, which includes the great
moderation and great recession as well as the
dot.com boom and the rise of institutional
trading. This is also the period where the
value premium has been extensively
researched (and therefore, potentially
traded upon).
The value premium price index (VPP) is
calculated as the difference between the
natural logarithm of value stock monthly
total return index (value index) and growth
stock monthly total return index (growth
index). Figure 1 shows that the data for
VPP for both small and large rms appear
to be non-stationary. The data display a
positive trend but with some different
characteristics. This may initiate a question
of whether VPP are determined by
economic factors that vary from one period
to another (for more detail, see Fama and
French, 1989; Vassalou, 2003; Black et al,
2007). Overall, the VPP in our sample
increase over the period, especially after
the World War II, with a fall around the
beginning of the twenty-rst century.
Figure 1 also appears to suggest that the

52

VPP tends to drop before and during


contradiction periods and increase during
expansion periods. The Augmented
DickeyFuller (ADF) unit root test suggests
that the VPP contain a single unit root. The
value premium (VP) is then constructed by
taking the rst difference of the VPP. The
VP is then shown in Figure 2, which
displays the usual characteristics of a returns
series
As noted above, given the relationship
between leverage and default probability,
the default premium index may capture
some characteristics of leverage. That is,
the probability of default increases with
rms that have a higher level of leverage
(see, for example, Anderson et al, 1996;

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

Value premium and default risk

METHODOLOGY

30

This article utilizes the GARCH (1, 1) model


(Engle, 1982; Bollerslev, 1987) as follows:

20

LSVPt + t

10
0

t : N 0; h2t

-10

h2t w + 2t - 1 + h2t - 1

-20

1940

1950

1960

1970

1980

1990

2000

2010

1930

1940

1950

1960

1970

1980

1990

2000

2010

b
30
20
10
0
-10
-20
-30

TH

Figure 2: The value premium (VP) for the US stock


market.
The value premium price index for US large and small
stocks (VPP). All data are logged total return indices.
VPPt = ln(VTRIt)(GTRIt), where VTRI is value total
return index and GTRI is growth total return index.
LVP and SVP denote large and small value premium
respectively which are calculated as the rst
differences of the value premium price indices:
(a) LVP; (b) SVP.

where LSVP denotes variables of interest


including both large and small rms value
premium, h2t is the conditional variance for
the period (t), and , and w are nonnegative parameters. Here, measures the
effect of volatility shock in period (t1) on
volatility in period (t) and (+) measures the
speed at which this effect dies away. The
threshold GARCH, or TGARCH, model is
used to measure the leverage effect on value
premium and to allow for asymmetric shocks
to volatility motivated by the reasoning that
good news and bad news have different
effects on future volatility (see Bollerslev et al,
1992; Black, 2002). The specication for this
model is:

PY

1930

-30

Ivaschenko, 2003; Crosbie and Bohn, 2003;


Molina, 2005; Chan-Lau, 2006a, b).
Crosbie and Bohn (2003) dene the default
risk as the uncertainty surrounding the
ability of the rm to pay its obligations and
debts. Firms generally pay a spread over the
default-free rate of interest that is
proportional to their default probability as
compensation to their lenders, that is, the
default premium. The default premium
index, which is dened as the difference
between the returns on long-terms
corporate bonds and long-term government
bonds is collected from Ibbotson
Association. The ADF tests suggest that the
rst differences of default premium index
series, which are the default premiums, are
stationary.

LSVPt + t
t : N 0; h2t

h2t w + 2t - 1 + 2t - 1 dt + h2t - 1

where dt = 1 if t1<0 otherwise dt = 0.


Therefore, there are different effects on the
conditional variance where t1<0 (an
unexpected decrease in price) denotes bad
news and t1>0 (an unexpected increase in
price) denotes good news. The impact of
good news is given by while the impact of
bad news is given by +, with the leverage
effect. The leverage effect reects how a
decrease in stock prices leads to an increase in
nancial leverage (as the value of equity falls
relative to corporate debt), therefore both the
required return on equity and risk increase
(see, Christie, 1982; Black, 2002).
The use of the GARCH approach is
motivated by the following three reasons. First,
GARCH models allow the simultaneous
investigation of the conditional mean and
variance relationship between the value

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

53

Elgammal and McMillan

PY

research shows an association between default


risk and leverage (for example, Chan and
Chen, 1991; Piotroski, 2000; Mohanram,
2005; Garlappi and Yan, 2011). Ivaschenko
(2003) reports a positive association between
corporate leverage and corporate default risk.
Furthermore, Molina (2005) supports a
positive association between leverage and the
default risk premium as proxied by both credit
ratings and the default spread.
The ndings in Table 1 suggest that there
is strong evidence that the default premium
has positive explanatory power for both large
and small rms value premiums in all
samples. These results are consistent with the
notion that value stocks, which are more
levered, are more vulnerable to default risk
compared with growth stocks. Therefore,
with an increase in the default risk probability,
stockholders will require higher returns for
value stocks than for growth stocks. This
result is consistent with the argument of Fama
and French (1996) and Chan and Chen
(1991) who link the value premium with
nancial risks, stating that value stocks have
poorer performance, earnings and
protability compared with growth stocks.
The results in Table 1 also support those of
Vassalou and Xing (2004) who suggest that
the book-tomarket effect is concentrated in
the high default risk rms and Black (2006)
who nds a positive relationship between the
default value premiums. The reported results
add to the evidence for the risk explanation
for the value premium, that the value
premium is a compensation for a nondiversiable risk factor, which is the default
risk.
Also displayed in Table 1 are the estimates
of the standard GARCH (1, 1) model. The
results support the existence of GARCH
effects, with key parameters indicating
statistical signicance. The sum of the
GARCH parameters also indicates a high
degree of volatility persistence to a shock. Of
interest the memory parameter, , is greater
for large rms, while the shock parameter, ,
is larger for small rms.

EMPIRICAL RESULTS

premium and default risk. The TGARCH


model allows examination of the impact of
leverage on the value premium, thereby
incorporating both default measures of risk
and leverage. Finally, GARCH and TGARCH
models account for heteroscedasticity which
is prevalent in stock returns.
The (ARCH) effect in the data is
investigated before and after modelling using
the ARCH Lagrange Multiplier (LM) test for
the residuals (Engle, 1982). The QuasiMaximum Likelihood covariances and
standard errors are calculated by the method
described by Bollerslev and Wooldridge
(1992) for a heteroscedastic consistent
covariance corrected for non-normally
distributed residuals. This method only affects
the covariance matrix, but does not affect the
parameter estimates. The Ljung-Box Qstatistics and their P-values are used to test for
the serial correlation in the residuals. If the
null hypothesis of no-serial correlation is
rejected for the residual, the methodology of
Bollerslev and Wooldridge (1992) is used to
correct for the serial correlation.

TH

GARCH estimates for the value


premium and the default premium

To shed light on the relationship between


default and value premiums, the default
premium is included in the mean equation of
the GARCH model as follows:
LSVP + DEFAULTt + t

t : N 0; h2t
h2t w + 2t - 1 + h2t - 1

where DEFAULT denotes the default risk


premium in time period t. As the default risk
premium is the difference between corporate
bond and government bond returns, it reects
the compensation for the risk of corporate
debt. Thus, the default risk premium is
expected to associate positively with the
leverage. An important part of the existing

54

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

Value premium and default risk

Table 1: Default risk and value premium (GARCH (1, 1))


Value premium
LVP: 1927:07 to 2011:03
SVP: 1927:07 to 2011:03
LVP: 1954:07 to 2011:03
SVP: 1954:07 to 2011:03
LVP: 1991:12 to 2011:03
SVP: 1991:12 to 2011:03

Observations

Q24

A4

JB

1005
1005
681
681
232
232

0.281*
0.196*
0.328*
0.224*
0.439*
0.329*

0.295*
0.379*
0.233**
0.391*
0.362**
0.671*

0.561*
0.310*
0.662*
0.623*
0.408
1.700***

0.133*
0.189*
0.178*
0.265*
0.242*
0.405*

0.826*
0.798*
0.749*
0.692*
0.735*
0.537*

1.460
0.482
2.916
1.596
5.833
3.521

0.003
0.005
0.001
0.040
0.078
0.023

680.317*
78.520*
0.221
25.217*
0.569
11.416*

LSVPt + default + t

t : N 0; h2t
h2t w + 2t - 1 + h2t - 1
LSVP denotes variables of interest including LVPt, SVPt which denotes the large and small value premium for the
sample I, Ai denotes an ith order ARCH LM test, Ai~X2i . Qi denotes an ith order Ljung-Box test for residual serial
dependency, Qi~X2i ; and JB denotes the Jarque-Bera test for residual normality, JB~X2i . Probability values are in
parentheses beside test statistics. *, **, *** denote a coefcient that is signicant at 1, 5 and 10 per cent levels,
respectively.

PY

This section further investigates the


relationship between default and value
premiums and their volatilities using the
TGARCH model. In addition, the
contemporaneous measure of the conditional
variance of the default risk premium
(VOLDEFAULT) is included in the variance
equation of the value premium. Thus, we
estimate the following model, the results of
which are presented in Table 2:

increases the premium required by investors.


This nding is consistent with a considerable
part of literature that develops the links
between value premium and leverage (for
example, Fama and French, 1992; Grifn and
Lemmon, 2002; Penman et al, 2007).
Moreover, our ndings support the view that
the value premium is related to nancial
distress and can act as a state variable for
common macroeconomic risk in the context
of ICAPM or APT models (Chan and Chen,
1991; Fama and French, 1996).
The estimated volatility coefcients
presented in Table 2 show that the
contemporaneous relationship between the
volatility of the default risk premium and the
volatility of value premiums is mixed. That is,
for small stocks it is positive and signicant
during the periods July 1927-March 2011 and
July 1954-March 2011; however, it is
negative in the most recent period but
insignicant. For large stocks the relationship
is negative and signicant in the most recent
sample period but insignicant and of mixed
sign elsewhere. Nonetheless, these results
exhibit some consistency with those of Bali
and Engle (2010) and Black (2006) who
document an insignicant positive
relationship between the volatility of default
premium and expected stock returns.
Moreover, the difference between large and

Leverage effects and default


volatility

TH

LSVPt + DEFAULTt + t

t : N 0; h2t
h2t w + 2t - 1 + 2t - 1 dt

+ VOLDEFAULT + h2t

Table 2 conrms the positive relationship


between the default premium and value
premium reported earlier. The positive
coefcient on the leverage term in the
conditional variance equation indicates that
negative shocks increase volatility more than
equal positive shocks. Hence, leverage may
lead to increase in risk related to a value
premium investment strategy. Since the value
stocks are, in general, more levered compared
with growth stocks and have greater access to
leverage sources, a negative stock price shock
increases leverage and raises risk. This, in turn,

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

55

Elgammal and McMillan

Table 2: Default premium volatility and the value premium


Value premium
LVP: 1927:07 to 2011:03
SVP: 1927:07 to 2011:03
LVP: 1954:07 to 2011:03
SVP: 1954:07 to 2011:03
LVP: 1991:12 to 2011:03
SVP: 1991:12 to 2011:03

Observations

1005
1005
681
681
232
232

0.205**
0.319*
0.180***
0.348*
0.130
0.357***

0.224*
0.022
0.050*
0.142**
0.203*
0.104*
0.281*
0.032
0.091*
0.158**
0.180**
0.130*
0.207*** 0.087* 0.022
0.147
0.083
0.013

Q24

A4

0.138*
0.125*
0.141*
0.140*
0.366*
0.381*

0.855*
0.802*
0.777*
0.758*
0.815*
0.833*

2.027
2.336
4.545
2.435
3.195
3.086

0.030
0.005
0.003
0.048
0.029
0.046

LSVPt + DEFAULTt + t

t : N 0; h2t
h2t w + 2t - 1 + 2t - 1 dt + VOLDEFAULT + h2t - 1
LSVP denotes variables of interest including LVPi, SVPi (the large and small value premium for the sample I).
Ai denotes an ith order ARCH LM test, Ai~X2i . Qi denotes an ith order Ljung-Box test for residual serial dependency,
Qi~X2i ; and JB denotes the Jarque-Bera test for residual normality, JB~X2i . Probability values are in parentheses
beside test statistics. VOLDEFAULT denote the volatility of default as estimated from GARCH model mean equation.
*, **, *** denote a coefcient that is signicant at 1, 5 and 10 per cent levels, respectively.

PY

relationship between the value premium and


the lagged default premium exhibited a
structural change after July 1954 and after
December 1991.6 Thus, the model has been
run for two separate periods July 1927-June
1954 and July 1954-November 1991 in
addition to the regular periods noted above.
The results in Table 3 demonstrate little
evidence of predictive power from the lagged
default premium to the value premium. Over
the time period 19271954 there is a
signicant negative effect for small stocks,
while over the period 19912011 there is a
weak positive relationship. Overall, the results
suggest limited predictive power for the
conditional mean of the value premium.
Regarding the results for the conditional
variance, there is evidence that the volatility
of the default premium has positive predictive
power for the volatility of small stocks value
premium, although the evidence becomes
weaker in the later sample period. For large
stocks we observe a negative predictive
relationship in the latter sample, although the
relationship is positive in earlier samples. The
difference between the behaviour of large and
small stocks is perhaps not surprising. As noted
above, Vassalou and Xing (2004) argue that
there exists a relationship between size and
the information content of default risk, with
small stocks more responsive to such risk.

small stocks may be related to the nding of


Vassalou and Xing (2004) who suggest that
the size effect captures a lot of information
concerning default risk. However, a puzzle
remains with the negative association,
signicantly so for large rms, between the
value premium and default in the last sample
(19922011). It should be borne in minds that
this sample period contains a relatively small
number of observations and further work is
required to conrm the ndings.

TH

Does the default premium predict


the value premium?

To investigate the ability of the default


premium to predict the value premium, we
adjust equation (4) above by including the
rst lag of the default premium and its
volatility in the TGARCH model, which is
now given as:
LSVPt + DEFAULTt - 1 + t

t : N 0; h2t
h2t w + 2t - 1 + 2t - 1 dt
+ VOLDEFAULTt - 1 + h2t - 1 5
Structural break tests using both the Chow
Breakpoint Test and The CUSUM Test (the
cumulative sum of the recursive residuals,
Brown et al, 1975) present evidence that the

56

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

Value premium and default risk

Table 3: The value premium and the rst lag of default premium
Value premium

Observations

1005
1005
323
323
449
449
232
232

0.214**
0.321*
0.196
0.241
0.303**
0.336*
0.106
0.384***

LVP: 1927:07 to 2011:03


SVP: 1927:07 to 2011:03
LVP: 1927:07 to 1954:06
SVP: 1927:07 to 1954:06
LVP: 1954:07 to 1991:11
SVP: 1954:07 to 1991:11
LVP: 1991:12 to 2011:03
SVP: 1991:12 to 2011:03

0.025 0.019
0.042*
0.143*
0.111
0.196*
0.106*
0.131*
0.391
0.620* 0.001
0.090*
0.525*
0.264**
0.095*
0.105**
0.074
1.067**
0.155** 0.212*
0.070
0.242*** 0.137* 0.007*
0.186 0.087* 0.037
0.367*
0.463** 0.021
0.075
0.5028

Q24

A4

0.864*
0.800*
0.919*
0.832*
0.095
0.778*
0.822*
0.680*

2.259
2.509
0.903
3.805
5.711
2.689
2.236
4.497

0.003
0.004
0.010
0.066
0.017
0.048
0.045
0.084

LSVPt + DEFAULTt - 1 + t

t : N 0; h2t

with the traditional assets theoretical


framework, Penman et al (2007) report a
negative association between nancial
leverage and future returns. However,
Penman et al (2007) do not introduce a robust
explanation for this association.7 Fama and
French (2006) report a negative association
between the default probability (measured by
O-score) and stock returns. Dichev (1998)
suggests that default risk is not a systematic risk
factor as his ndings show that both high and
low O-scores are associated with low returns.
Vassalou and Xing (2004) report a negative
association between the rst lag of default and
individual equity returns in the Fama
MacBeth regressions, but they did not
provide any explanation for this association.
Grifn and Lemmon (2002) suggest that
growth rms may face a higher probability of
default compared with value rms. George
and Hwang (2010) explain the negative
relationship between returns and both
nancial distress intensity and leverage by
optimizing rms that differ in their exposure
to nancial distress costs. Firms with high
costs choose low leverage to avoid distress,
but they retain exposure to the systematic risk
of bearing such costs in low states. George and
Hwang (2010) report that the relationship
between return premiums and distress costs is

TH

Although in a different context, Scislaw and


McMillan (2012) highlight that the value
premium is more prevalent in the FamaFrench small stock portfolios compared with
the large stock portfolios, particularly over the
recent period.
A remaining puzzle concerns the negative
association between the lagged default
premium and the current small rms value
premium during the period of July 1927 to
June 1954. If the default premium is
compensation for nancial systematic risk, it
would be expected that the increase in the
default would increase the required return on
the value stocks compared with growth
stocks. Consequently, this should increase the
value premium, which is not the case herein.
Fama and French (1992) suggest that the
book-to-market ratio works as a proxy for
nancial distress; thus, it has to display a
positive relationship with both realized and
expected returns. In our investigation, we
reported a positive relationship between
default and the realized value premium.
However, we nd a negative relationship
between the rst lag of default premium and
value premium for parts of the sample.
Consulting the literature to nd an
explanation for this negative relationship does
not provide a complete story. In contradiction

PY

h2t w + 2t - 1 + 2t - 1 dt + VOLDEFAULTt - 1 + h2t - 1


LSVP denotes variables of interest including LVPi, SVPi (the large and small value premium for the sample I).
Ai denotes an ith order ARCH LM test, Ai~X2i . Qi denotes an ith order Ljung-Box test for residual serial dependency,
Qi~X2i . Probability values are in parentheses beside test statistics. *, **, *** denote a coefcient that is signicant at 1,
5 and 10 per cent levels, respectively. VOLDEFAULT denote the volatility of default as estimated from GARCH model
mean equation.

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

57

Elgammal and McMillan

Table 4: Default premium volatility and the value premium

Q24

A4

Value premium

Observations

LVP: 1927:07 to
2011:03
SVP: 1927:07 to
2011:03
LVP: 1954:07 to
2011:03
SVP: 1954:07 to
2011:03
LVP: 1991:12 to
2011:03
SVP: 1991:12 to
2011:03

1005

0.257* 0.446*** 0.049

0.058* 0.152* 0.830* 0.505*

2.343 0.004

1005

0.140** 0.216

0.204*

0.107* 0.126* 0.799* 0.017

2.375 0.004

681

0.300* 0.550**

0.010

0.091* 0.153* 0.760* 0.490** 5.607 0.013

681

0.157** 0.261

0.183**

0.138* 0.140* 0.748*

232

0.273** 0.706** 0.117** 0.005 0.405* 0.779* 0.597*** 2.869

232

0.142

0.093

0.084

0.020

0.005 0.360* 0.852* 0.404

2.535 0.049
0.014

3.623 0.038

LSVPt + DEFAULTt + 1 D + t

t : N 0; h2t
h2t w + 2t - 1 + 2t - 1 dt + VOLDEFAULT

liquidity crisis, which started in


August 2007.8
Table 4 thus reports the same set of
regression results as for Table 2, that is,
including contemporaneous values of the
default premium, but with the inclusion of
the dummy variables outlined above. The
results again support a strong positive
relationship between default and the value
premium, although there is some evidence
that it weakens in the latter part of the sample.
Similarly, evidence for a strong leverage effect
is widely reported. Regarding volatility, again
there is a predominantly positive relationship
between default and value premium volatility,
again with the exception of the 19912011
period for large rms. With reference to the
dummy variables, the mean coefcient, 1,
for the large rms is signicantly positive in
the periods of July 1954-March 2011 and
from December 1991 to March 2011, at
5 per cent level. This suggests that the mean
return of the large rms value premiums in
the period from November 2001 to July 2007
is higher compared with the same variable

small or insignicant prior to 1980 and larger


and signicant thereafter.

PY

+ h2t - 1 + 2 D
LSVP denotes variables of interest including LVPi, SVPi (the large and small value premium for the sample I).
Ai denotes an ith order ARCH LM test, Ai~X2i . Qi denotes an ith order Ljung-Box test for residual serial dependency,
Qi~X2i ; and JB denotes the Jarque-Bera test for residual normality, JB~X2i . Probability values are in parentheses
beside test statistics. VOLDEFAULT denote the volatility of default as estimated from GARCH model mean equation.
*, **, *** denote a coefcient that is signicant at 1, 5 and 10 per cent levels, respectively. D is a dummy variable = 1 if
the observation in the period from November 2001 to July 2007 and = 0 otherwise.

Robustness tests

TH

The current liquidity crisis offers a unique


opportunity to consider the relationship
between liquidity and the value premium.
The importance of examining this
relationship grows from the fact that there is
a stream of literature linking the value
premium with nancial distress (see, for
example, Fama and French, 1998; Black,
2006; Watanabe and Watanabe, 2008;
Akbas et al, 2010). A liquidity crisis is usually
characterized by an increasing amount of
nancial distress. Investigating the
relationship between the liquidity crisis and
the value premium could provide evidence
for the risk explanation of the value
premium. The most recent economic
expansion period through November 2001
to July 2007 is included in the analysis as a
dummy variable. This period is chosen
to cover the period preceding the recent

58

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

Value premium and default risk

ACKNOWLEDGEMENTS
The authors thank Angela Black, Ercan
Balaban, Ken Peasnell, David Shepherd,
Alison Rieple, Linda Clarke, Petia Petrova, the
participants at the 2008 ICAS/BAA
Accounting & Finance Scot-doc Conference
in Glasgow and the 2013 BAFA conference in
Newcastle as well as Panagiotis DontisCharitos, Ben Nowman, Sheeja Sivaprasad,
Stefan Van Dellen and all other participants at a
2011 research seminar of Westminster Business
School in London for their useful comments
and advice on earlier versions of this article.
Furthermore, we are grateful to the
anonymous referees from the World Finance
Conference, Rhodes (Greece), June (2011) for
accepting the article in the conference and for
their helpful comments. The authors alone are
responsible for all limitations and errors that
may relate to the article.

TH

This article reports a positive relationship


between default risk and the value premium
for both large and small rms together with a
leverage effect. The results also show that the
default premium volatility has some
predictive power for the volatility of value
premium. These results provide new evidence
that the value premium is working as a proxy
for a macroeconomic risk factor associated
with nancial distress. Utilizing the GARCH
approach and separating the value premium
for large and small rm in a monthly data set
over the period 19272011, the results show a
positive association between the default
premium and the value premium
accompanied with evidence for a leverage
effect on the value premium. This lends
support to the risk-based explanation for the
source of value premium. That is, where the
default premium captures systematic risk in
the macroeconomy and that the value
premium is associated with rational decision
making on the part of investors. Value stocks
characterized by poor performance, earnings
and protability compared with growth
stocks are more vulnerable to the risk of
default and lead the investors to require a
higher return on value stocks as leverage
increases.
In conclusion, this article suggests some
possible answers for the initial questions about
the role of the default premium as a

PY

SUMMARY AND CONCLUSION

macroeconomic risk factor which can explain


the source of the value premium. The
empirical ndings contribute to our
knowledge by providing additional evidence
for the positive association between the
default and value premiums. This nding
supports the rational explanation for the value
premium. The results suggest that the value
premium is working as proxy for nondiversiable factors related to nancial
distress. Using a different methodology, and a
longer set of data than previously considered
within the literature, the results appear to
support the reasoning of Fama and French
(1996, 2006, 2007); the value premium
appears to be pervasive and works as a state
variable that is a proxy for nancial distress.

during the remainder of the sample.


Conversely, the volatility of the large rms
value premium is lower in the period
November 2001 to July 2007. This may
introduce a new puzzle, during an economic
expansion, where macroeconomic risk is
lower we would expect volatility to decrease,
as we nd, but also we would expect the
average level of the value premium to
decrease. Although, as reported by Scislaw
and McMillan (2012), this is a period where
the value premium in large cap rms, to a
greater extent, disappeared.

NOTES
1. Black (2006) investigates the relationship between default
risk and value premium volatilities using quarterly data. Our
analysis expands to investigate the association between the
variables themselves in addition to their volatilities using
longer and higher frequent set of data.
2. Elgammal and Al-Najjar (2013) report leverage effects in the
monthly value premium in the US, Canada, Denmark,

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

59

Elgammal and McMillan

5.
6.

TH

8.

7.

PY

4.

Black, A.J., Fraser, P. and McMillan, D.G. (2007) Are


international value premiums driven by the same set of
fundamentals. International Review of Economics and Finance
16(1): 113129.
Black, A.J. and McMillan, D.F. (2006) Asymmetric risk
premium in value and growth stocks. International Review of
Financial Analysis 15(3): 237246.
Black, F., Jensen, M.C. and Scholes, M.S. (1972) The Capital
Asset Pricing Model: Some Empirical Tests. Praeger Publishers.
SSRN, http://ssrn.com/abstract=908569 Date posted:
13 June 2006.
Bollerslev, T. (1987) A conditionally heteroskedastic time series
model for speculative prices and rates of return. The Review
of Economics and Statistics 69(3): 542547.
Bollerslev, T., Chou, R.Y. and Kroner, K.F. (1992) ARCH
modeling in nance: A review of the theory and empirical
evidence. Journal of Econometrics 52(12): 559.
Bollerslev, T., Litvinova, J. and Tauchen, G. (2006) Leverage
and volatility feedback effects in high-frequency data.
Journal of Financial Econometrics 4(3): 353384.
Bollerslev, T. and Wooldridge, J.M. (1992) Quasi-maximum
likelihood estimation and inference in dynamic models with
time-varying covariances. Econometric Reviews 11(2): 143172.
Brown, R.L., Durbin, J. and Evans, J.M. (1975) Techniques for
testing the constancy of regression relationships over time.
Journal of the Royal Statistical Society 37(2): 149163.
Campbell, J.Y., Hilscher, J. and Szilagyi, J. (2008) In search of
distress risk. Journal of Finance LXIII(6): 28992939.
Chan, K.C. and Chen, N. (1991) Structural and return
characteristics of small and large rms. Journal of Finance
46(4): 14671484.
Chan-Lau, J.A. (2006a) Market-Based Estimation of Default
Probabilities and its Application to Financial Market
Surveillance. IMF Working Paper no. 06/104.
Chan-Lau, J.A. (2006b) Fundamentals-Based Estimation of
Default Probabilities: A Survey. IMF Working Paper no.
06/149.
Chen, N. and Zhang, F. (1998) Risk and return of value stocks.
The Journal of Business 71(4): 501535.
Choi, J. (2013) What drives the value premium?: The role of
asset risk and leverage. Review of Financial Studies 26(11):
28452875.
Christie, A. (1982) The stochastic behaviour of common stock
variance: Value, leverage and interest rate effects. Journal of
Financial Economics 10(4): 407432.
Crosbie, P. and Bohn, J. (2003) Modelling Default Risk:
Modelling Methodology, Moodys KMV, http://www.ma
.hw.ac.uk/~mcneil/F79CR/Crosbie_Bohn.pdf, accessed
9 March 2013.
Dichev, I.D. (1998) Is the risk of bankruptcy a systematic risk?
The Journal of Finance 53(3): 11311147.
Elgammal, M. and Al-Najjar, B. (2013) The Leverage effect on
the value premium volatility: From an international
prospective, working capital, Working paper.
Engle, R.F. (1982) Autoregressive conditional heteroscedasticity
with estimates of the variance of UK ination. Econometrica
50(4): 9871008.
Fama, E.F. and French, K.R. (1989) Business conditions and
expected returns on stock and bonds. Journal of Financial
Economics 25(1): 2349.
Fama, E.F. and French, K.R. (1992) The cross-section of
expected stock returns. The Journal of Finance 47(2): 427465.

3.

Finland, New Zealand, Sweden, the UK and Poland stock


markets during the period of December 1991 to December
2006. The leverage effect is well documented in the literature
(for more details and evidence, see Christie, 1982; Ivaschenko,
2003; Bollerslev et al, 2006; Penman et al, 2007).
Campbell et al (2008) argue that the previous researchers
results for the returns on nancial distressed stocks depend
on the used measure of nancial distress. For example, the
low returns of distressed stocks are also documented by
Dichev (1998), who uses Altmans Z-score and Ohlsons
O-score to measure nancial distress; Garlappi and Yan
(2011), who obtain default risk measures from Moodys
KMV; and Avramov et al (2009), who use credit ratings to
measure rms nancial status. On the other hand, Vassalou
and Xing (2004) nd evidence that distressed stocks with a
low distance to default have higher returns, but this
evidence comes entirely from small value stocks.
As Penman et al (2007) mention, the Black-Scholes-Merton
measure potentially uses more information compared with
the O-score and Z-score measures, which are limited to
accounting information. However, it is based on equity
prices, introducing some concern as to whether inefcient
prices are conjectured.
I thank Kenneth French for making the data available on his
website: http://mba.tuk.dratmouth.edu/pages/ken.french/.
The Structural break tests of both Chow Breakpoint Test
and The CUSUM Test, the cumulative sum of the recursive
residuals (Brown et al, 1975); results are available on request.
Penman et al (2007) argue that this association is consistent
with inverse relationship founded between future returns
and wide variety of nancial distress measures in different
context. They introduce an extensive discussion for the
work in this area (for example, they have interesting
discussion for the work of Dichev, 1998; Grifn and
Lemmon, 2002, and Campbell et al, 2008).
This period is described by NBER as the last expansion
period in the US economy (NBER, 11 December 2008,
http://www.nber.org/cycles/dec2008.pdf). The last
liquidity period of July 2007 to June 2009 is included in the
analysis but the results are not signicant for this period. The
results are available on request.

REFERENCES

Akbas, F., Boehmer, E., Genc, E. and Petkova, R. (2010) The


Time-Varying Liquidity Risk of Value and Growth Stocks.
Texas, USA: Texas A&M University Business School.
Anderson, R., Sundaresan, S. and Tychon, P. (1996) Strategic
analysis of contingent claims. European Economic Review
40(35): 871881.
Avramov, D., Chordia, T., Jostova, G. and Philipov, A. (2009)
Credit ratings and the cross-section of stock returns. Journal
of Financial Markets 12(3): 469499.
Bali, T.G. and Engle, R.F. (2010) The intertemporal capital
asset pricing model with dynamic conditional correlations.
Journal of Monetary Economics 57(4): 377390.
Black, A.J. (2002) The impact of monetary policy on value and
growth stocks: An international evaluation. Journal of Asset
Management 3(2): 142172.
Black, A.J. (2006) Macroeconomic risk and the Fama-French
three-factor model. Managerial Finance 32(6): 505517.

60

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

Value premium and default risk

PY

Merton, R.C. (1974) On the pricing of corporate debt: The risk


structure of interest rates. Journal of Finance 29(2): 449470.
Mohanram, P.S. (2005) Separating winners from losers among
low book-to-market stocks using nancial statement
analysis. Review of Accounting Studies 10(2): 133170.
Molina, C.A. (2005) Are rms underleveraged? An
examination of the effect of leverage on default
probabilities. Journal of Finance 60(3): 14271459.
Ng, J. (2005) Distress Risk Information in Accruals. Working
Paper, The Wharton School, University of Pennsylvania.
Penman, S.H., Richardson, S.A. and Tuna, I. (2007) The
book-to-price effect in stock returns: Accounting for
leverage. Journal of Accounting Research 45(2): 427468.
Piotroski, J.D. (2000) Value investing: The use of historical
nancial information to separate winners from losers. The
Journal of Accounting Research 38(Supplement, 2000): 141.
Scislaw, K. and McMillan, D.G. (2012) The search for an
exploitable value premium in market indexes. Journal of
Asset Management 13(4): 253270.
Vassalou, M. (2003) News related to future GDP growth as a
risk factor in equity returns. Journal of Financial Economics
68(1): 4773.
Vassalou, M. and Xing, Y. (2004) Default risk in equity returns.
Journal of Finance 59(2): 831868.
Watanabe, A. and Watanabe, M. (2008) Time-varying
liquidity risk and the cross-section of stock returns. Review of
Financial Studies 21: 24492486.
Zhang, L. (2005) The value premium. Journal of Finance 60(1):
67103.

TH

Fama, E.F. and French, K.R. (1996) Multifactor explanations


of asset pricing anomalies. Journal of Finance 51(1): 5584.
Fama, E.F. and French, K.R. (1998) Value versus growth: The
international evidence. Journal of Finance 53(6): 19751999.
Fama, E.F. and French, K.R. (2006) The value premium and
the CAPM. The Journal of Finance LXI(5): 21632185.
Fama, E.F. and French, K.R. (2007) The anatomy of value
and growth stock returns. Financial Analysis Journal 63(6):
4454.
Garlappi, L. and Yan, H. (2011) Financial distress and the cross
section of equity returns. The Journal of Finance 66(3): 789822.
George, T. and Hwang, C. (2010) A resolution of the distress
risk and leverage puzzles in the cross section of stock returns.
Journal of Financial Economics 96(1): 5679.
Grifn, J.M. and Lemmon, M.L. (2002) Book-to-market
equity, distress risk, and stock returns. Journal of Finance
57(5): 23172336.
Ivaschenko, I. (2003) How Much leverage is Too Much, or
Does Corporate Risk Determine the Severity of Recession?
IMF Working Paper, New York, USA.
Li, X., Brooks, C. and Miffre, J. (2009) The value premium and
time-varying volatility. Journal of Business Finance &
Accounting 36(910): 12521272.
Liu, B., Kocagil, A. and Gupton, G. (2007) Fitch Equity
Implied Rating and Probability of Default Model.
Fitch Ratings, http://www.defaultrisk.com/_pdf6j4/
Fitch_Equity_Implied_Rtng_n_Prbblty_o_Dt_Mdl.pdf.
Liu, W. (2006) A liquidity augmented capital asset pricing
model. Journal of Financial Economics 82(3): 631671.

2014 Macmillan Publishers Ltd. 1470-8272 Journal of Asset Management Vol. 15, 1, 4861

61

You might also like