Upload
ngokhuong
View
216
Download
1
Embed Size (px)
Citation preview
Can Risk Management Help Prevent Bankruptcy?
Monica Marin∗
July, 2007
Abstract
Whether companies engage in risk management activities for risk reduction or speculation purposeshas been subject to continuous debate. This paper compares the use of risk management instrumentsby firms that eventually file for bankruptcy to matched firms that do not file for bankruptcy duringthe sample period. Results indicate that the odds of filing for bankruptcy are approximately 96% lowerfor firms that manage risk. The paper also takes an alternative approach in predicting the probabilityof default by deriving the distance to default from equity prices with the Black-Scholes-Merton optionpricing model. The distance to default compares the market value of assets to the size of a one standarddeviation move in the asset value. Results show that the number of standard deviations by which firmsin the sample are away from the default point increases by 3.2 units when they engage in risk manage-ment activities. The main findings are reinforced when examining the equity-implied asset volatilityas a separate measure of firm risk and its relationship with risk management. More specifically, assetvolatility decreases by 64% when firms manage risk.
∗Moore School of Business, University of South Carolina, Columbia, SC 29208. E-mail:monica [email protected]
1 Introduction
In the 2002 Berkshire Hathaway annual report, Warren Buffet warned against the use of
derivatives, referring to them as ”financial weapons of mass destruction, carrying dangers
that, while now latent, are potentially lethal.” His view on derivatives opposes the classical
belief that they are used to reduce risk (Smith and Stulz (1985)), being known primarily as
hedging instruments. While most of the concern over the use of derivatives has been expressed
by practitioners, a recent academic study (Faulkender (2005)) shows that firms often use in-
terest rate risk management instruments to time the market.
I provide evidence on whether risk management instruments are used for risk reduction
purposes by examining whether the use of risk management instruments reduces the probabil-
ity of financial distress. This question has been investigated by several academic studies, but
the evidence is limited. Nance, Smith, and Smithson (1993), and Mian (1996)) do not find any
relation between the use of derivatives and the probability of bankruptcy, while Fok, Carroll,
and Chiou (1997) find a weak negative relation. More recently, Judge (2003) shows a strong
negative relation in a study examining the risk management activity of U.K. firms. He also
finds a stronger relation for the U.K. firms than for the U.S. firms, and attributes this result to
differences in the bankruptcy codes. Unlike the previous studies that analyze solely the use of
derivatives, Judge (2003) uses a broader definition of risk management activity, that includes
both financial derivatives and hedging methods other than financial derivatives.
I use a pair-matched sample of U.S. bankrupt and non-bankrupt companies between 1998
and 2005, and estimate a duration analysis model for the likelihood of bankruptcy as a function
of their risk management activity. Second, I estimate the distance to default with a structural
model and then evaluate whether risk management helps to explain the distance to default.
Last, I investigate the relationship between risk management and the equity-implied asset
volatility as an alternative measure of firm risk.
A word of caution is necessary given that the sample of bankrupt and non-bankrupt firms
1
is a non-random sub-sample of public companies. If risk management is driven by reasons of
financial distress and bankruptcy costs, then this sample is biased towards firms where risk
management is most desirable. Therefore, inferences about the impact of risk management
on firms’ riskiness should not be made beyond this sample. However, if financial distress is
one of the primary determinants of risk management, then the analysis of bankrupt and non-
bankrupt firms would probably provide the most interesting insights.
The sample of 344 bankrupt and non-bankrupt firms is restricted to those non-financial
and non-utility firms that have available accounting data (from Compustat), as well as risk
management information disclosed in their annual reports for at least one year prior to filing
for bankruptcy. I define a firm as engaging in risk management in a particular year if the firm
uses any risk management instruments, including both financial derivative instruments and
methods other than financial derivatives. For example, I classify a firm as engaging in risk
management if the firm reports a fixed rate debt issue as a hedging activity under SFAS 133
(Accounting for Derivative Instruments and for Hedging Activities). As Faulkender (2005),
Kedia and Mozumdar (2003), and Judge (2003) argue, this approach provides a more accurate
picture of a firm’s risk management strategy than simply using derivative use to classify firms
as engaging in risk management. In addition, to identify whether a company engages in any
type of risk management, I also identify whether a firm engages in interest rate risk manage-
ment, foreign currency risk management, or commodity price risk management. I follow each
firm in time, starting with the first year when it discloses its position on risk management (can
be as early as 1994) and ending with the fiscal year before bankruptcy filing (can be as late as
2004).
I first estimate a discrete time complementary loglog hazard model for firms’ probability
to file for bankruptcy within one year. The duration analysis approach is more appropriate for
my sample and, according to Shumway (2001), it performs better than the conditional binary
models that are widely used. Since the likelihood of financial distress could influence the use
2
of risk management instruments, I use GMM (Generalized Method of Moments) models to
control for the endogeneity between the probability of bankruptcy and the risk management
decision.
I find that, all else equal, the odds of filing for bankruptcy are 96% lower for the firms
that manage risk as opposed to the ones that do not. This result suggests that the use of risk
management instruments helps reduce risk and extend a firm’s life.
Second, I estimate the distance to default from equity prices in a Black-Scholes-Merton
option pricing framework. The distance to default measures how far (in terms of the number
of standard deviations) the firm’s asset value is from the default point. Next, I estimate a
GMM model to study the relationship between risk management and the distance to default.
The results obtained with this approach are consistent with those from the duration analy-
sis approach. All else equal, any risk management activity is associated with a higher distance
to default. On average, the number of standard deviations by which firms are away from the
default point increases by 3.2 units when they manage risk. This evidence confirms the finding
from the cloglog model, namely that engaging in risk management activities tends to reduce
the probability of default.
Third, I investigate the relationship between risk management and the equity-implied asset
volatility1 as a distinct measure of firm risk, and estimate a GMM model that accounts for
the endogenous regressors.
The results confirm the previous findings in the paper and show that, all else equal, firms
that use risk management instruments have a significantly lower asset volatility. More specifi-
cally, on average firms’ asset volatility is reduced by 64% when they manage risk. Furthermore,
separate analysis of interest rate risk management and foreign currency risk management in-
dicate that these are also associated with lower asset volatility.
As a first robustness check, I investigate the relationship between the changes in the dis-
1Asset volatility is also used in the distance to default calculation, as shown in Section 3.2.
3
tance to default or asset volatility and the changes in the firms’ risk management activity.
I find that changes in the risk management activity in general are positively related to the
changes in the distance to default and negatively related to the changes in asset volatility,
results which are consistent with the main findings. In addition, a positive and significant
relationship is found between the changes in the distance to default and the changes in inter-
est rate risk management, as well as between the changes in the distance to default and the
changes in foreign currency risk management. On the other hand, the changes in commodity
price risk management are negatively associated with the changes in the distance to default,
which constitutes, counter-evidence to the risk-reduction hypothesis.
Next, I use sub-sample analysis to examine the relationship between risk management and
the distance to default or the asset volatility. I calculate the mean risk exposure (i.e. interest
rate risk exposure, foreign currency risk exposure, and commodity price risk exposure respec-
tively), and divide the sample into two sub-samples for high and low exposure in each of the
three categories. Most of the results support the main findings and suggest that, for firms
with high interest rate exposure or high foreign currency exposure, the distance to default is
augmented and asset volatility is reduced when they manage any type of risk or particular
risks (i.e. interest rate, foreign currency, commodity price risks). However, for firms with low
interest rate exposure and foreign currency exposure, the use of commodity price risk man-
agement instruments reduces the distance to default. Consequently, not only that the use of
commodity price risk management instruments does not benefit the firms with low interest
rate and foreign currency hedging needs, but also it may have a harmful effect. In terms of
the relationship between risk management and asset volatility for these firms, the negative
coefficient is consistent with the hypothesis of risk reduction. For firms with low commodity
price exposure, the use of risk management instruments increases the distance to default, while
decreasing asset volatility. At the same time, for firms with high commodity price exposure,
evidence shows only a negative relationship between interest rate risk management and asset
4
volatility.
Lastly, I examine the bankrupt and non-bankrupt firms separately, and find that, for each
category, firms that undertake risk management activities tend to have a lower asset volatility.
One contrasting result shows that for non-bankrupt firms, firms that manage interest rate risk
have a lower distance to default. Although this result may provide counter-evidence on the
risk-reduction purpose of interest rate risk management as shown by Faulkender (2005), it is
not confirmed in any other test.
Additional results which are consistent throughout this paper suggest that firm size is neg-
atively related to the odds of filing for bankruptcy and to asset volatility, and positively related
to the distance to default. Furthermore, leverage is associated with higher odds of filing for
bankruptcy, higher asset volatility, and lower distance to default.
As already mentioned, I obtain all empirical results reported in this paper using GMM
(Generalized Method of Moments). GMM estimates the model parameters without making
strong assumptions regarding the distributional properties of the variables observed. Thus, it
provides a solution to the case where the orthogonality assumption between the error term and
regressors is not satisfied. At the same time, it can be readily applied to nonlinear equations,
while providing a consistent estimate of the parameter.
In previous tests I also use a two-step procedure, where I first estimate a conditional binary
model (logit) to regress the risk management variable on appropriate explanatory variables.
Second, I include the predicted value in the cloglog and the linear regression models respec-
tively, along with other accounting-based determinants of the default probability, distance to
default, or asset volatility, while using the jackknife method to reduce the bias of the estimator
and to produce conservative standard errors. The final results obtained with this approach
are very similar, and thus not reported in the paper, but available upon request.
The paper is organized as follows: Section 2 presents a survey of the Related Literature,
Section 3 details the Empirical Estimation, Section 4 shows the Results, Section 5 presents
5
the robustness checks and their results, and Section 6 summarizes the Conclusions.
2 Related Literature
2.1 Why firms use risk management instruments
Smith and Stulz (1985) were among the first to recognize that transaction costs of bankruptcy
can induce public corporations to make use of risk management instruments.2 More specifi-
cally, they show that the use of risk management instruments can moderate the volatility of
cash flows, which reduces the probability of incurring direct and indirect bankruptcy costs.
As far as the empirical evidence is concerned, Nance, Smith, and Smithson (1993), and
Mian (1996)) do not find any relation between the use of derivatives and the probability of
bankruptcy. Later, Fok, Carroll, and Chiou (1997) find a weak negative relation between the
two. Only recently, Judge (2003) shows a strong negative relation for a sample of U.K. firms.
He attributes this difference relative to U.S. firms to dissimilarities in the bankruptcy codes.
Judge (2003) uses a broader definition of risk management activity, including both deriva-
tives and traditional hedging methods. On the other hand, another recent paper (Faulkender
(2005)) shows that interest rate risk management is driven by speculation or market timing.
However, instead of examining the use of derivatives, Faulkender (2005) investigates the de-
terminants of firms’ choices of interest rate exposure when issuing new debt, and finds that
these are related to the shape of the yield curve rather than to the hedging needs.
The existing literature has identified a multitude of factors contributing to a company’s
decision to manage risk. Smith and Stulz (1985) argue that the more wealth managers have
invested in the company (higher managerial ownership) the more they will tend to manage
risk. Tufano (1996) shows that a compensation contract that is linear or concave in firm value
provides incentives for the manager to reduce risk, while a convex contract has the opposite
2The costs of using risk management instruments are ignored throughout this paper.
6
effect. Consistent with this argument, he argues that managers of gold mining firms make
use of risk management instruments less if their compensation is based on stock options as
opposed to bonuses. Similarly, Rogers (2002) shows that derivative use is negatively related
to option holding and positively related to stock ownership.
Froot, Scharfstein, and Stein (1993) discuss other rationales for risk management in addi-
tion to the costs of financial distress and managerial incentives: taxes, debt capacity, capital
markets imperfections, and inefficient investment. Adam (2002) shows empirically that risk
management can reduce a firm’s dependence on external capital markets.
Empirical evidence by Mian (1996) shows that larger firms are more likely to manage risk,
companies that manage risk do not have higher market-to-book ratios, and risk management
is uncorrelated with leverage, positively correlated with dividend yield and dividend payout,
and negatively correlated with liquidity. Graham and Smith (1999) find that tax convexity
increases firms’ incentives to hedge.
Nevertheless, the relationship between firms’ use of risk management instruments and any
measure of firm riskiness or financial distress is most likely subject to an endogeneity problem.
For example, firms with a higher probability of financial distress may not be able to access cer-
tain risk management instruments that are available based on credit rating (i.e. interest rate
swaps). In their paper on dynamic risk management, Fehle and Tsyplakov (2005) demonstrate
that firms that are either far from financial distress or deep in financial distress do not initiate
or adjust their risk management instruments. On the other hand, they show that the rest of
the firms do initiate and actively adjust their use of risk management instruments. Similarly,
Purnanandam (2007) shows that financially distressed firms (that have high leverage) have
higher incentive to engage in hedging activities, but that these incentives disappear when their
financial distress becomes severe.
7
2.2 Factors Affecting the Probability of Financial Distress
Numerous papers focus on the probability of financial distress. One stream of literature em-
phasizes accounting-based models for default risk which are estimated with measures of firm
liquidity, cash flow, solvency, profitability, leverage, size, and efficiency. Also known under
the name of credit scoring models, these take the form of Multivariate Linear Discriminant
Analysis (MDA) (Beaver (1966), Altman (1968), and Altman (1973)), or of conditional binary
probability models (Ohlson (1980) and Zmijewski (1984)). While these ’traditional’ models
for the probability of default are relatively easy to implement, they have the disadvantage of
using only information from the firms’ financial statements which are backward-looking and
formulated under the going concern principle. Moreover, they are subject to accounting num-
bers manipulation, thus introducing a bias in the default risk estimation.
A different approach is option-based, generating theoretical default probabilities that reflect
market information from equity prices. These so-called structural models were first developed
by Black and Scholes (1973) and Merton (1974), and have been extensively used not only
in the finance literature (Delianedis and Geske (1999), Charitou and Trigeorgis (2002), and
Charitou, Lambertides, and Trigeorgis (2004)), but also in practice (i.e. KMV Corporation).
Hillegeist, Keating, Cram, and Lundstedt (2004) provide a performance comparison of the
accounting-based methods with the option-based methods, and conclude that the information
content of the option-based model is richer. The advantage of the structural models stems
from the fact that they extract information from market prices and compute the probability
of default independently for any firm, using a theoretically-derived equation (i.e. the Black-
Scholes-Merton option pricing model). Nevertheless, there are several shortcomings associated
with the use of structural models: the assumptions of market efficiency, perfect liquidity, and
lack of arbitrage conditions, as well as the incapacity to incorporate financial restructuring
(refinancing and renegotiating of debt contracts).
More recently, Papanastasopoulos (2006) relates the two different streams of literature
8
(accounting-based and option-based) in a hybrid model which improves the performance of
the option-based models by adding accounting information.
The hypothesis of this paper is that the use of risk management instruments reduces the
probability of bankruptcy. For the empirical estimation, I first use a duration analysis model
(complementary loglog) while controlling for the endogeneity arising between the risk manage-
ment activity and the probability of bankruptcy. Second, I estimate the distance to default
from equity prices with the Black-Scholes-Merton approach, and use risk management to ex-
plain its variation. Last, I investigate the relationship between risk management and the
equity-implied asset volatility as an alternative measure of risk. Additionally, I perform sev-
eral robustness checks on sub-samples (bankrupt/non-bankrupt, and high/low risk exposure),
and also investigate the changes in a firm’s distance to default and asset volatility relative to
the changes in the risk management activity.
3 Empirical Estimation
3.1 The Data
Ideally, the sample would be a large panel of firms observed over a number of years. However,
since the number of companies that file for bankruptcy is generally much smaller than that of
firms not filing for bankruptcy, the cost associated with obtaining the risk management data
for a large number of non-bankrupt companies is prohibitive. Thus, due to the high cost of
hand-collecting the data, I construct a pair-matched sample. With this sampling method, I
condition the sample on the probability of default, and a bias exists as in any choice-based
sample. However, as argued by Zmijewski (1984), this bias does not affect the statistical
inference for the financial distress model. Moreover, the results are qualitatively similar with
those obtained when correcting this bias. However, this method provides good controls for the
purpose of this study, since the non-bankrupt companies chosen are similar to the bankrupt
9
ones with respect to size and industry. The pair-matching method has been previously used
in the bankruptcy literature by Beaver (1966), Altman (1968), Ohlson (1980), and Charitou
and Trigeorgis (2002).
First, I collect a list of companies that filed for bankruptcy from BankruptcyData.com,
published by New Generation Research Inc. BankruptcyData.com is a Boston-based web-
site that tracks filings from federal bankruptcy districts. After filtering out the closely-held
companies, those with total assets of less than $50,000, as well as the financial and utilities
companies, I have a sample of 750 firms that filed for bankruptcy between January 1998 and
August 2006. I use this period because most companies disclose their risk management activ-
ities in the footnotes of the annual report filed with the Securities and Exchange Commission
(10-K), starting with 19983.
Further, I identify companies that have annual reports filed with the Securities and Ex-
change Commission prior to the bankruptcy filing date, file for bankruptcy only once, and
have available Compustat data prior to filing for bankruptcy. Thus, my sample reduces to 447
firms. I first run searches on the following keywords: ”hedge” or ”hedging,” ”derivative,” ”fi-
nancial instrument,” ”currency,” ”exchange rate risk,” ”interest rate risk,” and ”commodity,”
and then manually collect data on whether the firms disclose that they manage either interest
rate, foreign currency, or commodity price risks from the footnotes of the annual reports. Only
335 of the firms report the use or non use of risk management instruments in the footnotes of
the annual report for at least one fiscal year prior to the fiscal year of the bankruptcy filing.
Companies that report any type of risk management activity (including but not restricted
to the use of derivatives) are assigned a value of one, while the ones reporting no risk man-
agement activity, ”limited,” ”minimal,” or ”immaterial” use of risk management instruments
are assigned a value of zero. Companies that ignore the disclosure requirement, or state that
3Initially, companies were required by the SEC to adopt SFAS133 (Accounting for Derivative Instruments and forHedging Activities) in the years beginning after June 1998. However, in June 1999, the FASB delayed the effectivedate of this statement for one year, to fiscal years beginning June 15, 2000 due to concerns about companies’ abilityto modify their information systems and educate their managers.
10
the new disclosure regulation did not affect their financial statements are not included in the
sample, since no conclusion can be drawn about their risk management involvement. Although
notional amounts of financial instruments used are desirable for this analysis, these are not
available in most cases for this particular sample. Bankruptcy filings in the sample occur
between 1998 and 2005, but data (both accounting and risk management) are not available
for the fiscal year of the bankruptcy filing. The earliest year in the sample with reported use
of hedging instruments is 1994 and the latest is 2004.
Secondly, I match each company in the sample with a company in the same industry and
of similar size (within 10% of asset size) for the fiscal year before the bankruptcy filing, which
did not file for bankruptcy. In order to construct the control sample, I use the two-digit Stan-
dard Industrial Classification code for the initial matching with similar Compustat firms. If
multiple matches exist, I manually select the best match based on the four-digit or three-digit
SIC code, on the closest asset size, and ultimately on the risk management data availability
as reported in the annual report filed with the Securities and Exchange Commission. All
matches are required to have risk management information reported in their 10-K. Based on
these criteria, I find 172 matches. The missing 163 matches occur because of the restrictions
on the asset size, as well as because of lack of risk management data. I stop following the 172
matches identified in the same year that their counterparts file for bankruptcy. This creates a
censorship issue, which I address below.
Finally, the full sample is comprised of 344 firms (172 bankrupt, and 172 non-bankrupt),
and 926 firm-year observations. Some years of data that are not common to the paired compa-
nies are systematically excluded, thus truncating the sample. The managerial ownership and
compensation data are obtained from proxy filings (DEF-14A) with the Securities and Ex-
change Commission for each of the firms and years in the sample. The accounting data come
from Compustat, while the equity data are extracted from CRSP. The return on the S&P 500
index is used as a proxy for the market return. The Reuters-Jeffries commodity index and
11
currency index are used to estimate the commodity price exposure and the foreign currency
exposure respectively. More specifically, the exposure variables are estimated separately for
each firm, using monthly data, for a period starting two years before the fiscal year required
and ending two years after, with the following regression:
Ri,t = β0i + β1i ∗∆Rmt + β2i ∗%∆FXIt ∗ FXIt + β3i ∗%∆CPIt + ωi,t (1)
where:
Ri,t is the firm excess return in year t, ∆Rmt is the change in the market (S&P500 index)
return in month t, %∆FXIit is the return in month t on Reuters-Jeffries currency index, and
%∆FXIit is the return in month t on Reuters-Jeffries CRB commodity price index. Therefore,
β1i is the estimate for firm i’s interest rate exposure, β2i quantifies its foreign currency exposure,
and β3i measures its commodity price exposure for a five-year period.
As detailed in Table 1, 60 out of 344 firms in the sample are concentrated in the business
services industry and 54 are in the communications industry. As shown in Table 2, 46 out of
172 bankruptcy filings take place in 2001, and 41 in 2002. The number of bankruptcy filings
declines in 2003 and 2004. A preliminary comparison of the bankrupt versus non-bankrupt
companies in the fiscal year before the filing (Table 3) shows that the two groups are similar
in terms of average size (p-value for the t-test: 0.945) as desired by construction, but the
bankrupt companies have much higher leverage (total debt divided by total assets) than the
non-bankrupt companies: 56.45% versus 29.76%, p-value: 0.000. They are also characterized
by lower liquidity (cash plus short-term investments divided by total assets): 13.77% versus
23.18%, p-value: 0.000. The difference in the average bonus dollar amount is not statistically
significant and neither is the difference in the number of shares underlying stock options.
To provide a rough picture of the firms’ risk management activity, in the fiscal year before
bankruptcy, 43.02% of the bankrupt companies use some type of risk management instrument
as opposed to 51.75% of the non-bankrupt companies (the difference is significant at the 10%
12
level, p-value for t-test: 0.053). In terms of specific risk management activity, 25% of the
bankrupt and 27.90% of the non-bankrupt firms manage their interest rate risk (the difference
is not significant at any level, p-value: 0.271), 20.34% of the bankrupt and 30.23% of the
non-bankrupt firms manage their foreign currency risk (the difference is significant at the 5%
level, p-value: 0.018), and 9.88% of the bankrupt and 5.81% of the non-bankrupt firms manage
commodity price risk (the difference is significant at the 10% level, p-value: 0.081).
3.2 Empirical Methodology
3.2.1 Duration Analysis Model
To estimate the effect of the use of risk management instruments on the probability of bankruptcy,
I first use a discrete time duration model. This choice is motivated by both the nature of the
data collected, and the advantages of the method over other alternatives. First, the sample is
subject to right censorship and left truncation. Second, duration analysis has an advantage
over a binary dependent variable model which would not account for the difference in time
when the firms file for bankruptcy. Furthermore, duration analysis attempts to model the
transition from the state of non-bankruptcy towards the state of bankruptcy, and the rela-
tionship between transition patterns and firm characteristics. Shumway (2001) also argues for
using hazard models versus static models in a paper on bankruptcy forecasting, and Hillegeist,
Keating, Cram, and Lundstedt (2004) estimate discrete hazard models to assess how well the
accounting-based predictive measures fit the data.
The choice of the particular duration model (complementary loglog hazard model) is mo-
tivated by the nature of the information about the spell or the duration length. The duration
time is right censored: half of the firms in the sample do not file for bankruptcy, but they are
not followed beyond the year that their matched firm files for bankruptcy. It is also subject
to left truncation (also known as ’delayed entry’): the sample includes only the years with
available data on both Compustat (accounting data), and annual reports (risk management
13
data).
Each firm in the sample is described by a number of time-varying firm-specific variables, for
a certain period of time before filing for bankruptcy. This period (spell duration) ranges from
1 to 8 years, with an average of 3 years. The fiscal year before bankruptcy is the last year with
available data in both Compustat or 10-Ks (unless the company emerged from bankruptcy);
because of this, I refer to it as the transition year towards bankruptcy.
Although duration occurs in continuous time, I only observe spell lengths censored in
intervals of 1 to 8 (so called ’grouped’ or ’banded’ data), where each interval represents one
year. Therefore, I model the hazard function in discrete time, and use a proportional hazard
model in the form of the complementary loglog model (cloglog), which accounts for interval
grouping and left truncation. The probability of having survived until the end of year t is
given by:
S(t) ≡ St =t∏
k=1
(1− hk,X) (2)
where hk,X is the hazard rate for year k and depends on the firm’s characteristics, denoted by
the vector X. The hazard function can change from year to year for the same company, but
it is assumed to be constant within a one year interval.
Among the discrete time proportional hazard models, which depend on a firm’s character-
istics, the cloglog model is consistent with the grouped intervals (several consecutive years per
firm), and has the form:
ht,X = 1− exp [− exp (β′X + γt)] (3)
where:
ht,X is the hazard rate for year t and depends on the vector X of firm’s characteristics.
γt = log(log S0,t−1
S0,t) where S0,t is the baseline survivor function (the probability of having
14
survived until the end of year t).
The model is estimated with the cloglog function, where the dependent variable is set equal
to one only in the year in which the transition to bankruptcy takes place. The independent
variables are the firm characteristics: risk management activity (one or zero), number of years
of using risk management instruments up to that point (k), logarithm of firm size (Logsize),
leverage (Leverage), liquidity (Liquidity), market to book ratio (MtoB), and industry-leverage
interaction terms (leverage ∗ i1 to leverage ∗ i8) to identify the effect of leverage that depends
on industry. The risk management variable equals one if the firm engages in risk management
and zero otherwise, and is defined in four ways: IR RM for interest rate risk management,
FX RM for foreign currency risk management, CP RM for commodity price risk management
and Any RM for any type of risk management. The variables’ definitions are given in the
Appendix.
In order to address the potential endogeneity problem between the use of risk management
instruments and the likelihood of bankruptcy, I estimate GMM (Generalized Method of Mo-
ments) models and use the following instruments (also defined in the Appendix) for the risk
management activity: interest rate risk exposure (ITRExposure), foreign currency exposure
(FXExposure), commodity price exposure (CPExposure), lag earnings(LAGEAR)4, market
to book ratio (MtoB), CEO bonus (Bonus), the number of shares underlying stock options
awarded to the CEO (StockOpt), the number of shares of common stock held by the CEO as
beneficial ownership (Ownership), dividend yield (DivYield), dividend payout (DivPay), and
industry-leverage interaction terms. The exposure variables are estimated with Equation 1.
3.2.2 Option Approach in Default Risk Measurement
As an alternative empirical method, I estimate the distance to default with a structural model
and evaluate the explanatory power of risk management on the model-implied distance to
4Lag earnings is used absent of a better measure of tax convexity that is also a relevant and valid instrument forrisk management.
15
default. The distance to default is a measure of firm risk that captures the value of the firm’s
assets, its business and industry risk, as well as its leverage.
Structural models equate a firm’s equity with an European call option on the assets of the
firm, that has the strike price equal to DT (the face value of zero coupon debt with maturity
at time T ) and expiration at debt’s maturity T . VE is the market value of equity and VA is
the market value of the firm’s total assets (the expected discounted future cash flows). If at
time T , VAT> DT , then equityholders will exercise their option and receive the residual claim
VAT−DT . Otherwise, they will let their option expire and default will occur. Thus, equity is
worth VET= max(VAT
−DT , 0).
Assuming that the risk free interest rate is constant and identical to the borrowing and
lending rate, firm value follows a geometric Brownian motion with a constant drift equal to
the risk free interest rate r and a constant diffusion rate equal to σA (where Wt is the standard
Brownian motion):
dVAt
VAt
= r(VAt , t)dt + σAdWt (4)
The market value of common equity VE is given by the Black-Scholes equilibrium pricing
formula for European call options:
VE = VAe−δT N(d1)−DT e−rT N(d2) + (1− e−δT )VA (5)
where:
d1 =ln
VADT +(r−δ+
σ2A2
)T
σA
√T
and d2 = d1− σA
√T =
lnVADT +(r−δ−σ2
A2
)T
σA
√T
N(d2) is the risk neutral expected probability that the firm will remain solvent at debt’s
maturity, and δ is the continuous dividend rate. Following Hillegeist, Keating, Cram, and
Lundstedt (2004) and Papanastasopoulos (2006), the dividends paid by the firm accrue to
equityholders before debt’s maturity T .
16
The equation
σE = σAVA
VEN(d1)e−δT (6)
is derived from Ito’s lemma and shows the relationship between equity volatility and asset
volatility (equity-implied).
In order to estimate this model, I first compute equity volatility from historical returns for
each firm. To derive the equity-implied asset volatility and the value of firm’s assets, I solve the
system of two nonlinear equations with two unknowns ((5) and (6)) by using an optimization
procedure. In order to transform the results from risk-neutral to actual default probability, I
substitute the expected return on the firm asset value (µ) for the risk free interest rate. I use
the results to compute DD, the distance to default (which measures the number of standard
deviations that firm asset value is away from the default point), and EDP , the expected
default probability (equivalent to the probability that the option will expire unexercised):
DD =ln VA
DT + (µ− δ − σ2A2 )T
σA
√T
(7)
EDP = N(− ln VA
DT + (µ− δ − σ2A2 )T
σA
√T
) (8)
I calculate the market value of equity (VE) based on the average price in the month corre-
sponding to the firm’s fiscal year end. For the standard deviation of equity returns (σE), I use
daily return data from CRSP over the entire fiscal year. The maturity T equals one year and
the risk-free rate r is the one-year Treasury bill rate. Following Vassalou and Xing (2004) and
Papanastasopoulos (2006), the strike price X which is equal to the firm’s total liabilities, is
approximated with the sum of current liabilities (as previously defined) and half of long-term
liabilities (an assumption that is due to the one year maturity restriction). The dividend rate
17
δ is the sum of the prior year’s common and preferred dividends divided by the approximate
market value of assets, which is defined as total liabilities plus the market value of equity. In
estimating µ, the expected return on assets (or the growth rate), I use the change in the value
of assets, a measure which was suggested by Vassalou and Xing (2004) (without dividends)
and also employed by Hillegeist, Keating, Cram, and Lundstedt (2004) and Papanastasopoulos
(2006) (with dividends):
µ(t) = max[VA(t) + Dividends− VA(t− 1)
VA(t− 1), r] (9)
Since the calculation of the expected default probability imposes a normal distribution on
the distance to default (an assumption which does not necessarily reflect reality) and, further-
more, is monotonic in all measures of firm characteristics (i.e. size, industry, leverage), I use
the distance to default in my subsequent analysis.
I estimate a GMM model to investigate the relationship between risk management and the
distance to default. In addition to risk management, I include the following controls: logarithm
of firm size (Logsize), leverage (Leverage), liquidity (Liquidity), market to book ratio(MtoB),
and industry-leverage interaction terms (leverage ∗ i1 to leverage ∗ i8). I instrument the risk
management activity with the following variables: interest rate risk exposure (ITRExposure),
foreign currency exposure (FExposure), comodity price exposure (CPExposure), lag earnings
(LAGEAR), market to book ratio (MtoB), CEO bonus (Bonus), the number of shares underly-
ing stock options awarded to the CEO (StockOpt), the number of shares of common stock held
by the CEO as beneficial ownership (Ownership), dividend yield (DivYield), dividend payout
(DivPay), and industry-leverage interaction terms. The exposure variables are estimated with
Equation 1. All other variables are defined in the Appendix.
Additionally, I use a similar procedure to examine the relationship between risk manage-
ment and the equity-implied asset volatility as a separate measure of risk. Asset volatility is
also used in the distance to default calculation, but can constitute a good measure of risk in
18
itself, since it captures the uncertainty of the asset value. Asset volatility is in fact a measure
of the firm’s business (in terms of size and nature) and industry risk.
4 Results
4.1 Duration Analysis Model Results
In this section I present the results of the cloglog model while controlling for the endogeneity
between risk management activity and the probability of default. I estimate GMM models,
where the instruments used for the risk management activity are lag earnings, interest rate
exposure, foreign currency exposure, commodity price exposure, bonus, stock options, man-
agerial stock ownership, dividend yield and dividend payout, all of which are defined in the
Appendix.
The results of the GMM estimation are shown in Table 45. The variable Any RM has a
negative coefficient and is significant at the 5% level. Since the magnitude of the regression
coefficient is not very informative, I calculate the exponential coefficient which is equal to
0.0364 and suggests that the odds of filing for bankruptcy (within the next year) of compa-
nies that use any type of risk management instruments is 3.6% of (or 96.4% lower than) the
odds of filing for bankruptcy of companies that do not manage risk. The magnitude of the
coefficient is quite high, a result which strongly supports the argument that risk management
helps extend a firm’s life. The coefficients on the variable indicating the use of a specific type
of risk management (interest rate, foreign currency, or commodity price) separately, are not
significantly related to the likelihood of bankruptcy, and thus no inference can be made about
their role in helping firms live longer.
Moving on to the control variables, the estimated coefficients indicate that, when FX RM is
5The exponential coefficients, indicating the odds of filing for bankruptcy, are not reported in Table 4, but availableupon request.
19
included in the regression, size has a strong negative impact on the probability of bankruptcy.
Thus, for this specification, larger firm size is associated with significantly lower odds of filing
for bankruptcy, a result which was also found by Altman, Haldeman, and Naraynan (1977),
Ohlson (1980), and Papanastasopoulos (2006). Similarly to the findings of Papanastasopou-
los (2006), more leverage in the capital structure is associated with a higher likelihood of
bankruptcy in the specifications that include Any RM and FX RM: the odds of going bankrupt
increases 10.4 times with a 1% increase in leverage, and 8.6 times respectively. Results also
indicate that there is no significant relationship between the odds of filing for bankruptcy and
liquidity or market to book ratio.
4.2 Option Pricing Model Results
4.2.1 Distance to Default
The distance to default generated with the Black-Scholes-Merton option pricing model has
a mean of 2.5 across all years, suggesting that, on average, the firms in the sample are 2.5
standard deviations away from default. Stated differently, on average, a firm would have to
incur a decrease of 2.5 standard deviations in its value of assets to default. The median is 1.8,
and the standard deviation is 2.9.
In order to investigate the relationship between the distance to default and risk manage-
ment, I estimate a GMM model that considers the possible endogeneity problem between the
two variables. Other controls (defined in the Appendix) used are: firm size, leverage, liquidity,
market to book ratio and interaction terms between firm industry and leverage. The results,
shown in Table 5, indicate that the coefficient on Any RM is positively and significant, indicat-
ing that any type of risk management activity is positively related to the distance to default,
and thus negatively related to the probability of bankruptcy. This suggests that, all else equal,
the number of standard deviations a company is away from default increases by approximately
3.2 units when that company manages any type of risk. As found in the previous test, the
20
magnitude of the coefficient is indicative of the importance of firms’ risk management activity
in reducing risk and extending a firm’s life. Interest rate risk management and foreign cur-
rency risk management are only weakly (significant at the 10% level) positively related to the
distance to default.
Moving on to the control variables, firm size is positively and significantly related to the
distance to default in three of the four specifications. Leverage is associated with a signifi-
cantly smaller distance to default in all specifications. However, liquidity is positively related
to the distance to default (i.e. a one percent increase in liquidity is associated with an increase
in the distance to default by approximately 2.8 units in the specification where Any RM is
included), and thus negatively related to the probability of bankruptcy in specifications that
include Any RM, IR RM, and FX RM. The market to book ratio is negatively and significantly
related to the distance to default in all specifications. Some interaction terms between firm
industry and leverage are also positively and significantly related to the distance to default.
4.2.2 Asset Volatility
Alternatively, I estimate a similar GMM model for the relationship between risk management
and the equity-implied asset volatility as a separate measure of firm risk.
The results, shown in Table 6, indicate that the coefficient on the risk management variable
in the specifications that include Any RM, IR RM, and FX RM is negative and significant (at
the 5% level) related to asset volatility. They suggest that a firm’s asset volatility decreases
by 64% when it uses any risk management instruments, by 99% when it uses interest rate
risk management instruments, and by 60% when it uses foreign currency risk management
instruments. These results are indicative of the important contribution of risk management in
mitigating firms’ business risk.
As far as the firm-specific covariates are concerned, size is the only one that is significantly
related to asset volatility. Given the negative relationship, larger firms tend to have lower asset
21
volatility and thus to default less, result which agrees with the previous findings.
5 Robustness checks
5.1 Changes in Firm’s Risk Management Activity
As a first robustness check, I investigate the effect of changes in firms’ risk management activi-
ties on the changes in the distance to default and asset volatility. More specifically, I calculate
the difference from one year to another in the risk management activity (+1, 0, -1), as well
as in the distance to default, asset volatility, and all other controls. Real changes in the risk
management activity (i.e. +1 or -1) are not very common in this sample, most companies
remaining true to their historical risk management activity. This agrees with the prediction
from Fehle and Tsyplakov (2005) that firms that are deep in financial distress neither initiate
nor adjust their risk management practices.
Distance to Default
In order to investigate the relationship between the change in the risk management activity
and the change in the model-implied distance to default, I estimate the GMM model as previ-
ously described in Subsection 3.2.2 in first differences, i.e. all variables are changes relative to
the previous year. The results are shown in Table 7. The coefficient on the change in the risk
management variable is significant (at the 1% or 5% level) in each specification. For the spec-
ifications that include Any RM, IR RM, and FX RM, risk management is positively related
to the change in the distance to default thus suggesting a reduction in the risk of default. For
the CP RM specification, risk management activity is negatively related to the change in the
distance to default, contrary to the risk-reduction hypothesis.
22
Asset Volatility
The relationship between the change in the risk management activity and the change in
asset volatility, estimated using GMM, is presented in Table 8. The change in any risk man-
agement activity is negatively and significantly (at the 5% level) related to the change in asset
volatility. Therefore, an increase in firm’s risk management activity is associated with a de-
crease in its asset volatility. Similarly, the change in interest rate risk management activity is
negatively related to the change in asset volatility, but only at the 10% level.
5.2 Sub-sample Analysis: Exposure
In this subsection, I present the results of the GMM estimation described in Subsection 3.2.2 for
both the relationship between risk management and the distance to default and between risk
management and the equity-implied asset volatility on a series of sub-samples with different
risk exposure. This analysis is meant to provide insights on whether firms’ risk management
activities are driven by their hedging needs (as measured by their risk exposure), and to
examine whether the relationships found previously still hold.
I first calculate the mean interest rate exposure, foreign currency exposure, and commodity
price exposure using Equation 1 for the entire sample. Then I divide the sample into two sub-
samples: firms with a particular exposure lower than the mean and higher than the mean
respectively. Thus, I analyze six different sub-samples, and estimate GMM as before for each
sub-sample.
5.2.1 Interest Rate Exposure
The sub-sample of firms with interest rate exposure lower than the sample mean is comprised
of 59.6% bankrupt and 40.4% non-bankrupt firms. Across all years, 41.6% use any type of risk
management instruments, 18.1% use interest rate risk management instruments, 32.5% use
23
foreign currency risk management instruments, and 6% use commodity price risk management
instruments.
On the other hand, there are 48.7% bankrupt and 51.3% non-bankrupt firms in the sub-
sample of firms with interest rate exposure higher than the mean. 52.5% of the firms use any
type of risk management instruments, 28.8% use interest rate risk management instruments,
32.4% use foreign currency risk management instruments, and 10.4% use commodity price risk
management instruments. A simple comparison suggests that the firms in the sub-sample with
high interest rate exposure tend to make greater use of risk management instruments than the
firms in the sub-sample with low interest-rate exposure. The regression results (as detailed
below) for the sub-sample of firms with high interest exposure reinforce the main result that
an increase in the risk management activity is associated with a higher distance to default, and
a lower asset volatility. However, the result is not present in the sub-sample of firms with low
interest rate exposure, which could be explained by the fact that a low interest rate exposure
is equivalent to low hedging needs (possibly due to the past risk management activity).
Distance to Default
The results of the GMM estimation for the relationship between risk management and
the distance to default are presented in Table 9, separately for the two sub-samples of firms
with interest rate exposure lower and higher than the mean. In the case of the firms with
high interest rate exposure, a 1% increase in Any RM augments the distance to default by
4.5 units (significant at the 5% level). Similarly, a 1% increase in CP RM augments the
distance to default by 4.6 units, but this result is only weakly significant (at the 10% level).
On the other hand, for firms with interest rate exposure lower than the mean, a 1% increase in
CP RM reduces the distance to default by 6.4 units. The negative relationship found between
commodity price risk management and the distance to default indicates that, when interest
rate hedging needs are low, the use of commodity price risk management instruments is unlikely
24
to benefit the firm and may in fact be harmful. Therefore, risk management instruments are
associated with an increase in firms’ distance to default and a decrease in their asset volatility
if used according to their hedging needs (measured by the interest rate exposure).
Asset Volatility
Table 10 shows the results of the GMM estimation between risk management and asset
volatility, for the two sub-samples of firms with interest rate exposure lower and higher than
the mean. While no relationship can be found for the firms with low interest rate exposure, in
the case of firms with high interest rate exposure, IR RM, FX RM and CP RM are negatively
and significantly (at the 10%, 1% and the 5% level, respectively) related to asset volatility.
5.2.2 Foreign Currency Exposure
In the sub-sample of firms with foreign currency exposure lower than the sample mean there
are 55.4% bankrupt and 44.6% non-bankrupt firms. Across all years, the percentages of firms
using any type of risk management instruments, interest rate risk management instruments,
foreign currency risk management instruments, commodity price risk management instruments
are 48.8%, 25.9%, 30.1%, and 9.2% respectively.
The sub-sample of firms with foreign currency exposure higher than the mean is composed
of 46.9% bankrupt and 53.1% non-bankrupt firms. In this sub-sample, the percentages of firms
using any type of risk management instruments, interest rate risk management instruments,
foreign currency risk management instruments, commodity price risk management instruments
are 50.2%, 25.5%, 35.8%, and 9% respectively. The regression results, detailed below, are simi-
lar to the ones found for the sub-samples of firms with high and low interest rate exposure, and
suggest that risk management instruments are associated with an increase in firms’ distance
to default and a decrease in their asset volatility if used according to their hedging needs (as
measured by the foreign currency exposure).
25
Distance to Default
Table 11 presents the results of the GMM estimation between risk management and the
distance to default, separately for the two sub-samples of firms with foreign currency exposure
lower and higher than the mean. In the case of firms with high foreign currency exposure,
Any RM, IR RM, and FX RM are positively and significantly related to the distance to de-
fault. For example a 1% increase in Any RM augments the distance to default by 5.3 units.
CP RM is also positively related to the distance to default, but significant only at the 10%
level. On the other hand, in the case of firms with low foreign currency exposure, CP RM is
weakly negatively related to the distance to default (significant at the 10% level), result which
is similar to the one in the sample of firms with low interest rate exposure. The other three
specifications are not significantly related to the distance to default in the case of firms with
low foreign currency exposure.
Asset Volatility
The results of the GMM estimation between risk management and asset volatility, sepa-
rately for the two sub-samples of firms with foreign currency exposure lower and higher than
the mean, are shown in Table 12. Any RM is negatively associated to asset volatility only
in the case of firms with low foreign currency exposure, indicating that a 1% increase in risk
management decreases asset volatility by 59%. IR RM is negatively and significantly related to
asset volatility in both sub-samples. Similarly, FX RM is negatively related to asset volatility
for both the case of firms with high and low foreign currency exposure, but significant only
at the 10% level. Weak significance is also found for the relationship between CP RM and
asset volatility in the sub-sample with high foreign currency exposure; no relationship can be
observed in the sub-sample with low foreign currency exposure.
26
5.2.3 Commodity Price Exposure
In the sub-sample of firms with foreign currency exposure lower than the sample mean, there
are 55.2% bankrupt and 44.8% non-bankrupt firms. Across all years, 50% use any type of risk
management instruments, 26.5% use interest rate risk management instruments, 35.4% use
foreign currency risk management instruments, and 7.3% use commodity price risk manage-
ment instruments.
As for the sub-sample of firms with foreign currency exposure higher than the mean, 47.4%
of the firms are bankrupt and 52.6% are non-bankrupt. 48.6%, 24.7%, 28.7%, and 11.6%
of these firms use any type of risk management instruments, interest rate risk management
instruments, foreign currency risk management instruments, and commodity price risk man-
agement instruments. Opposite to the regression results from the sub-samples of firms with
high/low interest rate exposure and foreign currency exposure, in these sub-samples evidence
indicates that the use of risk management instruments is mainly associated with an increase
in the distance to default and a decrease in asset volatility for the firms with low commodity
price exposure. Therefore, the uses of any type risk management instruments, interest rate
risk management, and foreign currency risk management, which are related to an increase in
the distance to default and to a decrease in asset volatility, are not likely to be driven by firms’
commodity price hedging needs.
Distance to Default
For the two sub-samples of firms with commodity price exposure lower and higher than the
mean, the results of the GMM estimation for the relationship between risk management and
the distance to default are presented in Table 13. While for the firms with high commodity
price exposure, no relationship can be found between the risk management activity and the
distance to default, in the case of firms with low commodity price exposure there is a positive
and significant relationship for the Any RM, IR RM, and FX RM specifications. For example,
27
a 1% increase in any risk management activity augments the distance to default by approxi-
mately 3 units.
Asset Volatility
Table 14 shows the results of the GMM estimation between risk management and asset
volatility, separately for the two sub-samples of firms with commodity price exposure lower
and higher than the mean. Any RM and FX RM are negatively related to asset volatility
for the firms with low commodity price exposure. IR RM is also negatively related to asset
volatility for both firms with high and low commodity price exposure.
5.3 Sub-sample Analysis: Bankrupt and Non-Bankrupt
Lastly, another robustness check consists of separating the sample into sub-samples of bankrupt
and non-bankrupt firms, and estimating GMM in order to examine the relationship between
risk management and the distance to default or asset volatility. This analysis examines whether
the bankrupt and/or the non-bankrupt firms show specific patterns with respect to their risk
management activities.
As mentioned before, the characteristics of each group are shown in Table 3. The regres-
sion results support the main finding on the relationship between risk management and asset
volatility in both sub-samples, with not much evidence on the relationship between risk man-
agement and the distance to default.
Distance to Default
The results of the GMM estimation between risk management and the distance to default,
separately for the two sub-samples of bankrupt and nonbankrupt firms, are presented in Table
15. The variable IR RM is the only one that has a negative coefficient which is only significant
28
at the 10% level in the sub-sample of nonbankrupt firms. It suggests that, for companies
that use interest rate risk management instruments, the number of standard deviations the
asset value is away from the default point decreases by 1 unit, all else equal. This result
does not supports the hypothesis that interest rate risk management reduces the probability
of bankruptcy in this sub-sample. All other risk management variables are not statistically
significant. No inference can be made about the relationship between the risk management
activity and the distance to default in the sub-sample of bankrupt firms.
Asset Volatility
Unlike the distance to default, asset volatility has a significant relationship with the risk
management variables for the two sub-samples of bankrupt and nonbankrupt firms, as shown
in Table 16. Any RM is negatively and significantly related to asset volatility for the sub-
sample of bankrupt firms. IR RM and FX RM are also negatively and significantly related
to asset volatility for both sub-samples. For example, a 1% increase in the interest rate risk
management reduces asset volatility by 68%.
6 Conclusion
This paper examines whether the use of risk management instruments for a sample of bankrupt
and non-bankrupt firms impacts firms’ riskiness in terms of their probability of bankruptcy,
distance to default, and asset volatility. To my knowledge, the latter two measures of firm risk
have not been used in the prior literature on risk management, while the former has not been
used in the context of duration analysis.
In a framework that controls for the endogeneity between risk management and the prob-
ability of default, this paper provides evidence that the risk management activity contributes
29
to a lower probability of bankruptcy, higher distance to default, and lower asset volatility, all
of which suggest risk reduction and consequently a longer life for the firm. Both approaches
presented in this paper (duration analysis approach and option approach) show strong support
with respect to the role of risk management in general and provide some evidence with respect
to individual types of risk management: interest rate, foreign currency, and commodity price
risk management. Sub-sample analysis generally confirms the main results and brings some
insights with respect to the use of different types of risk management instruments depending
on the firms’ risk exposure (or hedging needs). For example, evidence indicates that the benefit
of using risk management instruments is greater when firms’ interest rate and foreign currency
hedging needs are high, and commodity price hedging needs are low. Overall, for firms with
high interest rate exposure and foreign currency exposure, as well as with low commodity
price exposure, the use of risk management instruments is associated with a higher distance to
default and a lower asset volatility. However, for companies with low interest rate and foreign
currency exposure, the use of commodity price risk management is negatively related to the
distance default. Similarly, the changes in commodity price risk management are negatively
related to the changes in the distance to default, suggesting that the initiations of commodity
price risk management actually reduce firms’ distance to default. The answer to the question
on whether firms with lower overall risk exposure are less likely to use commodity price risk
management instruments for risk-reduction purposes is left for future research.
30
References
Adam, Tim R., 2002, Do Firms Use Derivatives to Reduce their Dependence on External CapitalMarkets?, European Finance Review pp. 163–187.
Altman, E., R. Haldeman, and P. Naraynan, 1977, ZETA Analysis: A New Model to IdentifyBankruptcy Prediction Risk of Corporations, .
Altman, Edward I., 1968, Financial ratios, discriminant analysis and the prediction of corporatebankruptcy, Journal of Finance 23, 589–609.
, 1973, Predicting Railroad Bankruptcies in America, Bell Journal of Economics and Manage-ment Science 3, 184–211.
Beaver, W. H., 1966, Financial Ratios as Predictors of Failure, Journal of Accounting Research (Sup-plement) pp. 71–102.
Black, Fischer, and Myron Scholes, 1973, The pricing of options and corporate liabilities, Journal ofPolitical Economy 81, 637–659.
Charitou, A., N. Lambertides, and L. Trigeorgis, 2004, Is the Impact of Default Risk Systematic? AnOption-Pricing Explanation, Working Paper, University of Cyprus.
Charitou, Andreas, and Leon Trigeorgis, 2002, Option-based Bankruptcy Prediction, Working Paper,University of Cyprus.
Delianedis, R., and R. Geske, 1999, Credit Risk and Risk Neutral Probabilities: Information aboutRating Migrations and Defaults, UCLA Working Paper.
Faulkender, Michael, 2005, Hedging or market timing? Selecting the interest rate exposure of corporatedebt, Journal of Finance 60, 931–962.
Fehle, Frank, and Sergey Tsyplakov, 2005, Dynamic Risk Management: Theory and Evidence, Journalof Financial Economics.
Fok, R. C. W., C. Carroll, and M. C. Chiou, 1997, Determinants of corporate hedging and derivatives:a revisit, Journal of Economics and Business 49, 569–585.
Froot, Kenneth A., David S. Scharfstein, and Jeremy C. Stein, 1993, Risk Management: CoordinatingCorporate Investment and Financing Policies, Journal of Finance 48, 1629–1658.
Graham, John R., and Clifford Jr. Smith, 1999, Tax Incentives to Hedge, The Journal of Finance 54,2241–2262.
Hillegeist, Stephen A., Elizabeth K. Keating, Donald P. Cram, and Kyle G. Lundstedt, 2004, Assessingthe Probability of Bankruptcy, Review of Accounting Studies 9, 5–34.
Judge, Amrit, 2003, Why and How UK Firms Hedge, European Financial Management Journal 12.
Kedia, S., and A. Mozumdar, 2003, Foreign currency denominated debt: an empirical investigation,Journal of Business 76, 521–546.
Merton, C. R., 1974, On the Pricing of Corporate Debt: The Risk Structure of Interest Rates, Journalof Finance 29, 449–470.
Mian, Shehzad, 1996, Evidence on Corporate Hedging Policy, The Journal of Financial and QuantitativeAnalysis 11, 419–439.
Nance, Deana R., Clifford W. Smith, and Charles W. Smithson, 1993, On the Determinants of CorporateHedging, The Journal of Finance 48, 267–284.
31
Ohlson, James A., 1980, Financial Ratios and the Probabilistic Prediction of Bankruptcy, Journal ofAccounting Research 18, 109–131.
Papanastasopoulos, George, 2006, Using Option Theory and Fundamentals to Assessing Default Riskof Listed Firms, Working Paper, University of Peloponnese.
Purnanandam, Amiyatosh, 2007, Financial Distress and Corporate Risk Management: Theory & Evi-dence, Journal of Financial Economics, forthcoming.
Rogers, Daniel A., 2002, Does Executive Portfolio Structure Affect Risk Management? CEO Risk-taking Incentives and Corporate Derivative Usage, Journal of Banking and Finance 26, 271–296.
Shumway, Tyler, 2001, Forecasting Bankruptcy More Accurately: A Simple Hazard Model, The Journalof Business 74, 101–124.
Smith, Clifford W., and Rene M. Stulz, 1985, The Determinants of Firms’ Hedging Policies, The Journalof Financial and Quantitative Analysis 20, 391–405.
Tufano, Peter, 1996, Who Manages Risk? An Empirical Examination of Risk Management Practicesin the Gold Mining Industry, Journal of Finance 51, 1097–1137.
Vassalou, Maria, and Yuhang Xing, 2004, Default Risk in Equity Returns, The Journal of Finance 49,831–868.
Zmijewski, Mark E., 1984, Methodological Issues Related to the Estimation of Financial Distress Pre-diction Models, Studies on Current Econometric Issues in Accounting Research.
Appendix. Variables Definition
Any RM. Binary variable: 1 if the firm uses any risk management instruments and 0 otherwise.
IR RM. Binary variable: 1 if the firm manages interest rate risk and 0 otherwise.
FX RM. Binary variable: 1 if the firm manages foreign currency risk and 0 otherwise.
CP RM. Binary variable: 1 if the firm manages commodity price risk and 0 otherwise.
RMYears. Number of years of risk management activity up to that point.
Size. Natural logarithm of Total Assets.
Leverage. Total debt divided by Total Assets.
Liquidity. Cash and Short-term Investments divided by Total Assets.
MtoB. Market to Book ratio as reported in Compustat.
Leverage∗i1 to Leverage∗i9. Interaction terms leverage - industry, where i1 to i9 are industry dummies.
LAGEAR. Lag Net Income.
DivPay. Dividend Payout as reported in Compustat; (dividend per share divided by earnings per share) missingitems are set equal to zero.
DivYield. Dividend Yield as reported in Compustat; missing items are set equal to zero.
Bonus. Natural logarithm of the dollar amount the CEO received as Bonus and other annual compensation.
StockOpt. Natural logarithm of the number of shares underlying any option-based compensation awarded tothe CEO.
Ownership. Number of shares of common stock held by the CEO as beneficial ownership.
ITRExposure. Interest Rate Exposure Variable equal to the value of β1i from Equation 1.
FXExposure. Foreign Currency Exposure Variable equal to the value of β2i from Equation 1.
CPExposure. Commodity Price Exposure Variable equal to the value of β3i from Equation 1.
Change Any RM, Change IR RM, Change FX RM, Change CP RM. Variables indicating an initiation(takes the value of 1), a halt (takes the value of -1) or no change (takes the value of 0) in a firm’s risk managementactivity. Calculated as the difference in between the current value of Any RM, IR RM, FX RM, and CP RM andprevious year value.
Change Size, Change Leverage, Change Liquidity, Change MtoB, Change LAGEAR, Change DivPay,Change DivYield, Change Bonus, Change StockOpt, Change Ownership, Change ITRExposure,Change FXExposure, Change CPExposure. Variables calculated as the difference between the currentvalue of Size, Leverage, Liquidity, MtoB, LAGEAR, DivPay, DivYield, Bonus, StockOpt, Ownership, ITRExposure,FXExposure, and CPExposure compared to the previous year value.
Table 1:
This table presents a description of the data by industry sector. The first column shows the first two digits of the SIC Code, the secondcolumn presents the number of firms for that sector (half of each filed for bankruptcy), and the third column displays the complete namefor each sector.
Data Description by Industry - 344 observations, 172 pairsSic2 NoObs Industry
10 2 Metal Mining13 12 Oil and Gas Extraction20 2 Food and Kindred Products22 6 Textile Mill Products23 4 Apparel and Other Finished Products Made from Fabrics and Similar Materials26 2 Paper and Allied Products27 2 Printing, Publishing, and Allied Industries28 24 Chemicals and Allied Products30 8 Rubber and Miscellaneous Plastic Products32 2 Stone, Clay, Glass, and Concrete Products33 18 Primary Metal Industries34 4 Fabricated Metal Products, Except Machinery and Transportation Equipment35 46 Industrial and Commercial Machinery and Computer Equipment36 30 Electronic and Other Electrical Equipment and Components,
Except Computer Equipment37 4 Transportation Equipment38 4 Measuring, Analyzing, and Controlling Instruments; Photographic, Medical,
and Watches and Clocks39 2 Miscellaneous Manufacturing Industries42 8 Motor Freight Transportation and Warehousing44 2 Water Transportation45 4 Transportation By Air48 50 Communications50 6 Wholesale Trade-durable Goods51 6 Wholesale Trade-non-durable Goods54 2 Food Stores56 2 Apparel and Accessory Stores58 8 Eating and Drinking Places59 8 Miscellaneous Retail70 2 Hotels, Rooming Houses, Camps, and Other Lodging Places73 64 Business Services79 2 Amusement and Recreation Services80 2 Health Services87 6 Engineering, Accounting, Research, Management, and Related Services
Table 2:
This table presents a year-by-year description of the data. The second column shows the number of bankruptcies that were filed eachfiscal year, while the third column identifies the number of observations in the sample for each fiscal year.
Data Description by Year:Year # Bankruptcy Filings # Observations in the Sample
1994 101995 121996 201997 461998 4 1221999 11 1782000 20 2262001 46 1742002 41 942003 29 422004 20 22005 1Total 172 926
Table 3:
This table presents summary statistics (mean and standard deviation) of firm characteristics for companies that filed for bankruptcyand the control group, for the fiscal year before bankruptcy filing. The variables shown are: any risk management (Any RM), interestrate risk management (IR RM), foreign exchange risk management (FX RM), commodity price risk management (CP RM), size (Size),leverage (Leverage), and liquidity (Liquidity). These are defined in Appendix.
Characteristic Bankrupt Firms (172) Control Group (172)Mean Std.Dev. Mean Std. Dev
Any RM 43.02% .49 51.74% .50IR RM 25.00% .43 27.90% .45FX RM 20.34% .40 30.23% .46CP RM 9.88% .29 5.81% .23
Size $1,239,120 $8,123,232 $1,180,896 $7,541.997Leverage 56.45% .57 29.76% .31Liquidity 13.77% .19 23.18% .26
Table 4:
Complementary LOGLOG. This table presents the regression coefficients (not the exponential coefficients) and the p-values from the
Generalized Method of Moments (GMM) estimation of the cloglog regression. The dependent variable is 1 if a firm filed for bankruptcy in
that particular year and 0 otherwise. The independent variables, defined in the Appendix, are shown in the leftmost column. Regression 1
uses any risk management activity as an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency, and commodity
price risk management respectively.
Ind. Var. Regressions
Regression 1 Regression 2 Regression 3 Regression 4
Any RM -3.3109
(0.0442)**
IR RM -13.0787
(0.6964)
FX RM -1.3965
(0.1689)
CP RM 10.1654
(0.3262)
Size -0.2105 -1.1355 -0.4040 -1.7179
(0.3192) (0.6225) (0.0091)*** (0.3047)
Leverage 2.3437 2.8982 2.1562 5.9279
(0.0203)** (0.5589) (0.0012)*** (0.3219)
Liquidity -2.3394 -9.5410 -0.9580 -0.4427
(0.1528) (0.6786) (0.3808) (0.8575)
MtoB 0.0009 0.0053 -0.00008 -0.0011
(0.3471) (0.7111) (0.9499) (0.9215)
RMYears 1.0710 3.6439 0.7607 0.7298
(0.0534)* (0.6541) (0.0974)* (0.1366)
Leverage∗i1 -2.0211 -4.5646 -2.2081 -12.7646
(0.4085) (0.6261) (0.0431)** (0.3765)
Leverage∗i2 -2.5096 -6.2461 -1.6806 -5.9565
(0.2641) (0.617) (0.2006) (0.3314)
Leverage∗i5 0.1705 9.2348 1.2182 7.0718
(0.9508) (0.6904) (0.6078) (0.2769)
Leverage∗i7 -0.2155 2.5185 0.8415 1.0251
(0.8322) (0.7648) (0.6321) (0.7093)
Constant -3.7965 -8.7254 -2.6307 0.7346
(0.0207)** (0.6204) (0.0257)** (0.8468)
Table 5:
Distance to Default. This table presents the regression coefficients and the p-values from the Generalized Method of Moments (GMM)
estimation of the relationship between the distance to default on one side and risk management activity, firm-specific covariates, and
industry - leverage interaction terms on the other side. The dependent variable is the distance to default calculated with the Black-
Scholes-Merton model. The independent variables, defined in the Appendix, are shown in the leftmost column. Regression 1 uses the any
risk management activity as an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency, and commodity price
risk management respectively.
Ind. Var. Regressions
Regression 1 Regression 2 Regression 3 Regression 4
Any RM 3.1990
(0.0227)**
IR RM 3.4447
(0.0646)*
FX RM 2.5520
(0.0785)*
CP RM -3.3579
(0.3218)
Size 0.3478 0.4780 0.4464 0.9422
(0.0686)* (0.0015)*** (0.0191)** (0.0001)***
Leverage -5.3667 -7.1865 -4.5738 -7.4293
(<.0001)*** (<.0001)*** (<.0001)*** (<.0001)***
Liquidity 2.8580 1.8096 2.5796 0.7734
(0.0073)*** (0.0207)** (0.0107)** (0.3775)
MtoB -0.0090 -0.0071 -0.0080 -0.0065
(0.0001)*** (<.0001)*** (<.0001)*** (0.0213)**
Leverage∗i1 3.1835 5.1260 3.7848 5.0298
(0.0045)*** (0.0054)*** (0.0145)** (0.078)*
Leverage∗i2 6.4726 6.8362 5.1384 5.1242
(0.0002)*** (0.0001)*** (0.0008)*** (0.0003)***
Leverage∗5 6.2212 3.0604 5.9247 -0.0307
(0.2355) (0.5958) (0.2147) (0.9935)
Leverage∗i7 4.0905 5.1971 3.0287 -6.8459
(0.0003)*** (0.0013)*** (0.0067)*** (0.2165)
Constant -0.6463 0.2460 -0.6682 -0.5369
(0.434) (0.7145) (0.3965) (0.572)
Table 6:
Asset Volatility. This table presents the coefficients and the p-values from the Generalized Method of Moments (GMM) estimation of
the relationship between the asset volatility on one side and risk management activity, firm-specific covariates, and industry - leverage
interaction terms on the other side. The dependent variable is asset volatility. The independent variables, defined in the Appendix, are
shown in the leftmost column. Regression 1 uses any risk management activity as an independent variable. Regressions 2, 3, and 4 use
interest rate, foreign currency, and commodity price risk management respectively.
Ind. Var. Regressions
Regression 1 Regression 2 Regression 3 Regression 4
Any RM -0.6425
(0.0459)**
IR RM -0.9933
(0.0282)**
FX RM -0.6005
(0.039)**
CP RM -0.7590
(0.146)
Size -0.0710 -0.0860 -0.1118 -0.0756
(0.0417)** (0.0059)*** (<.0001)*** (0.0415)**
Leverage 0.5732 0.4176 0.2467 0.3487
(0.0582)* (0.176) (0.4471) (0.287)
Liquidity 0.0818 -0.2802 -0.0885 0.40468
(0.7625) (0.4168) (0.7848) (0.1187)
MtoB 0.00004 0.00010 0.00001 0.00002
(0.7199) (0.2326) (0.8772) (0.7629)
Leverage∗i1 -0.4604 -1.0084 -0.7557 0.3044
(0.0434)** (0.0143)** (0.0212)** (0.5762)
Leverage∗i2 0.5505 -2.4016 -1.6692 1.5720
(0.6851) (0.2492) (0.4365) (0.3865)
Leverage∗i7 -1.5209 0.2142 -0.4010 -1.1087
(0.1727) (0.8399) (0.6991) (0.3648)
Constant 1.4237 1.5911 1.7031 1.1211
(<.0001)*** (<.0001)*** (<.0001)*** (<.0001)***
Table 7:
Changes in the Distance to Default. This table presents the coefficients and the p-values from the Generalized Method of Moments
(GMM) estimation of the relationship between the changes in the distance to default on one side and the changes in the risk management
activity, changes in firm-specific covariates, and industry - leverage interaction terms on the other side. The dependent variable is the
change in the distance to default calculated with the Black-Scholes-Merton model. The independent variables, defined in the Appendix,
are shown in the leftmost column. Regression 1 uses the change in any risk management activity as an independent variable. Regressions
2, 3, and 4 use the change in interest rate, foreign currency, and commodity price risk management respectively.
Ind. Var. Regressions
Regression 1 Regression 2 Regression 3 Regression 4
Any RM Change 8.2671
(<.0001)***
IR RM Change 5.9322
(<.0001)***
FX RM Change 8.0707
(0.0153)**
CP RM Change -11.1276
(0.0003)***
Size Change -0.39954 0.2461 -0.4828 1.3045
(0.1861) (0.1868) (0.3707) (0.0011)***
Leverage Change -4.3920 -7.0269 -2.0613 -9.9683
(0.0219)** (<.0001)*** (0.5848) (<.0001)***
Liquidity Change 6.86791 3.5823 8.5492 2.6753
(<.0001)*** (0.0051)*** (0.0005)*** (0.0742)*
MtoB Change -0.0050 -0.0092 -0.0281 -0.0036
(0.7519) (0.4363) (0.272) (0.7578)
Change Leverage∗i1 -14.3419 3.9233 -1.4108 -10.7398
(0.2924) (0.3452) (0.9079) (0.2935)
Change Leverage∗i2 3.7495 5.3597 1.0442 9.0823
(0.0495)** (0.0028)*** (0.7837) (<.0001)***
Change Leverage∗i5 3.2196 5.2836 -1.6924 12.3586
(0.221) (0.0434)** (0.7605) (<.0001)***
Change Leverage∗i7 4.0346 4.1854 5.7101 10.9093
(0.1869) (0.1755) (0.1095) (<.0001)***
Constant -0.1634 -0.0973 -0.1472 0.2949
(0.4645) (0.6296) (0.5342) (0.1534)
Table 8:
Changes in Asset Volatility. This table presents the coefficients and the p-values from the Generalized Method of Moments (GMM)
estimation of the relationship between the changes in asset volatility on one side and the changes in risk management activity, changes
in firm-specific covariates, and industry - leverage interaction terms on the other side. The dependent variable is the change in asset
volatility. The independent variables, defined in the Appendix, are shown in the leftmost column. Regression 1 uses any risk management
activity as an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency, and commodity price risk management
respectively.
Ind. Var. Regressions
Regression 1 Regression 2 Regression 3 Regression 4
Any RM Change -0.8520
(0.0378)**
IR RM Change -0.6059
(0.0645)*
FX RM Change 0.8837
(0.2275)
CP RM Change 0.5772
(0.2022)
Size Change -0.0733 -0.1023 -0.2084 -0.1546
(0.1008) (0.0179)** (0.0117)** (<.0001)***
Leverage Change 0.1322 0.4241 0.4291 0.1390
(0.6052) (0.0996)* (0.291) (0.5635)
Liquidity Change -0.6757 -0.4920 -0.1613 -0.2778
(0.0079)*** (0.0134)** (0.5403) (0.0954)*
MtoB Change -0.0011 -0.0011 -0.0010 -0.0007
(0.0155)** (0.1699) (0.3591) (0.3088)
Change Leverage∗i1 0.1978 -0.3415 -0.6005 0.1778
(0.7409) (0.3951) (0.3921) (0.737)
Change Leverage∗i2 -0.1490 -0.3863 -0.4773 -0.1621
(0.5755) (0.1605) (0.2833) (0.5252)
Change Leverage∗i5 0.4816 0.5235 0.3630 0.6134
(0.2326) (0.0743)* (0.5622) (0.0538)*
Change Leverage∗i7 -0.2685 -0.4390 -0.4264 -0.2212
(0.3541) (0.1181) (0.2458) (0.4136)
Constant -0.0014 -0.0072 -0.0267 -0.0109
(0.9605) (0.7853) (0.4389) (0.6644)
Table 9:
Interest Rate Exposure - Distance to Default. This table presents the coefficients and the p-values from the Generalized Method
of Moments (GMM) estimation of the relationship between the distance to default on one side and the risk management activity, firm-
specific covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of firms with interest rate
exposure lower than the sample mean and higher than the sample mean respectively. The dependent variable the distance to default
calculated with the Black-Scholes-Merton model. The independent variables, defined in the Appendix, are shown in the leftmost column.
Regression 1 uses any risk management activity as an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency,
and commodity price risk management respectively.
Ind. Var. Regressions
sub-sample of firms with IR Exposure Lower than the Mean sub-sample of firms with IR Exposure Higher than the Mean
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM 0.6263 4.5401
(0.5434) (0.0229)**
IR RM 0.4500 1.1865
(0.8222) (0.4619)
FX RM 0.8019 1.2111
(0.4402) (0.4003)
CP RM -6.4763 4.6028
(<.0001)*** (0.0911)*
Size 0.6549 0.6489 0.6440 0.6372 0.3588 0.7671 0.8701 0.7714
(<.0001)*** (0.0001)*** (<.0001)*** (0.0018)*** (0.1872) (0.0001)*** (0.0003)*** (0.0158)**
Leverage -2.2183 -2.0720 -2.2860 -1.5755 -5.0472 -6.7029 -5.6495 -6.4317
(0.2165) (0.3897) (0.1813) (0.2868) (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)***
Liquidity 1.7641 2.1901 1.7770 2.2422 5.0542 2.4770 2.8673 1.9681
(0.1067) (0.0212)** (0.1033) (0.0194)** (0.0031)*** (0.003)*** (0.0059)*** (0.0341)***
MtoB 0.0551 0.0539 0.0512 0.0568 -0.0083 -0.0065 -0.0079 -0.0054
(0.0882)* (0.0822)* (0.1142) (0.0362)** (0.0141)** (0.0005)*** (0.0002)*** (0.0692)*
Leverage∗i1 74.7259 55.6897 62.3849 62.5335 2.4513 3.4660 3.0301 0.1976
(0.5387) (0.6061) (0.5916) (0.6589) (0.0398)** (0.0329)** (0.0225)** (0.9384)
Leverage∗i2 4.4955 4.2726 4.5392 3.7849 6.5588 5.6717 5.3221 5.4622
(0.0629)* (0.1354) (0.0513)* (0.0388)** (0.0006)*** (0.0012)*** (0.0003)*** (0.0006)***
Leverage∗i5 -6.5106 -2.9210 -3.9040 -1.1514 13.1299 6.5127 6.6040 6.4730
(0.7158) (0.8672) (0.7828) (0.9407) (0.0025)*** (0.0019)*** (0.0341)** (0.0102)**
Leverage∗i7 2.0692 1.8196 2.1266 1.3181 1.5488 2.4853 1.4221 2.3028
(0.2541) (0.4447) (0.2079) (0.3441) (0.275) (0.1044) (0.2713) (0.0885)
Constant -1.9255 -1.8149 -1.8794 -1.7851 -1.7277 -0.7902 -1.8362 -0.7145
(0.0457)** (0.0648)* (0.0427)** (0.1068) (0.1178) (0.3772) (0.065)* (0.6185)
Table 10:
Interest Rate Exposure - Asset Volatility. This table presents the coefficients and the p-values from the Generalized Method
of Moments (GMM) estimation of the relationship between asset volatility on one side and the risk management activity, firm-specific
covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of firms with interest rate exposure
lower than the sample mean and higher than the sample mean respectively. The dependent variable is asset volatility. The independent
variables, defined in the Appendix, are shown in the leftmost column. Regression 1 uses any risk management activity as an independent
variable. Regressions 2, 3, and 4 use interest rate, foreign currency, and commodity price risk management respectively.
Ind. Var. Regressions
Firms with IRExposure Lower than the Mean Firms with IRExposure Higher than the Mean
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM -0.1322 -0.5723
(0.464) (0.1028)
IR RM -0.3971 -0.3843
(0.2018) (0.0895)*
FX RM -0.0334 -0.7356
(0.8527) (0.0046)***
CP RM -0.0025 -0.8173
(0.9947) (0.0443)**
Size -0.0940 -0.0882 -0.1036 -0.1091 -0.0783 -0.1110 -0.0751 -0.0893
(0.0033)*** (0.0034)*** (0.0023)*** (0.0006)*** (0.0422)** (<.0001)*** (0.0135)** (0.0013)***
Leverage 0.0174 0.2944 0.0124 0.0072 0.1817 0.3917 -0.0725 0.1550
(0.9372) (0.3698) (0.956) (0.9746) (0.1342) (0.0058)*** (0.673) (0.2259)
Liquidity 0.0708 -0.0261 0.1582 0.1838 -0.2336 0.0434 -0.2634 0.1489
(0.7644) (0.9078) (0.5153) (0.3308) (0.3809) (0.7368) (0.1508) (0.2087)
MtoB -0.0005 -0.0005 -0.0007 -0.0009 0.00007 0.00009 0.00003 0.00004
(0.6215) (0.5135) (0.4876) (0.3591) (0.3256) (0.2636) (0.4821) (0.501)
Leverage∗i1 31.2928 45.7297 30.5843 31.3532 -0.2297 -0.5922 -0.5668 0.3733
(0.4814) (0.4229) (0.4872) (0.4746) (0.1121) (0.0222)** (0.0321)** (0.3765)
Leverage∗i2 -0.6367 -1.0238 -0.6196 -0.5895 -0.5149 -0.4337 -0.2348 -0.0969
(0.0157)** (0.0142)** (0.0176)** (0.0434)** (0.1374) (0.1535) (0.4656) (0.7538)
Leverage∗i5 4.9537 -0.6240 5.4333 5.6649 -0.6612 -0.2534 -0.4704 -0.1623
(0.5756) (0.9409) (0.5554) (0.5543) (0.3579) (0.6297) (0.4198) (0.7608)
Leverage∗i7 -0.0873 -0.3568 -0.0757 -0.0651 -0.0845 -0.1988 0.0491 -0.0701
(0.6827) (0.2765) (0.7267) (0.7664) (0.7028) (0.3197) (0.8292) (0.6936)
Constant 1.4811 1.4577 1.4569 1.4583 1.5707 1.4391 1.5938 1.3125
(<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)***
Table 11:
Foreign Currency Exposure - Distance to Default. This table presents the coefficients and the p-values from the Generalized
Method of Moments (GMM) estimation of the relationship between the distance to default on one side and the risk management activity,
firm-specific covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of firms with foreign
currency exposure lower than the sample mean and higher than the sample mean respectively. The dependent variable is the distance to
default calculated with the Black-Scholes-Merton model. The independent variables, defined in the Appendix, are shown in the leftmost
column. Regression 1 uses any risk management activity as an independent variable. Regressions 2, 3, and 4 use interest rate, foreign
currency, and commodity price risk management respectively.
Ind. Var. Regressions
Firms with FX Exposure Lower than the Mean Firms with FX Exposure Higher than the Mean
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM 0.5175 5.3782
(0.646) (0.0024)***
IR RM 0.8597 3.5109
(0.436) (0.0239)**
FX RM 0.1831 5.7272
(0.8301) (0.0086)***
CP RM -4.6942 4.4880
(0.0856)* (0.0894)*
Size 0.5032 0.4867 0.5514 0.7937 0.0972 0.5065 0.1166 0.4932
(0.0422)** (0.0344)** (0.015)** (0.0026)*** (0.7079) (0.0002)*** (0.6902) (0.0071)***
Leverage -3.8944 -4.0759 -3.9445 -4.0845 -7.3700 -8.8085 -4.3698 -4.0398
(0.0016)*** (0.0009)*** (0.0017)*** (0.0035)*** (<.0001)*** (<.0001)*** (0.032)** (0.0012)***
Liquidity 2.6569 2.7488 2.5380 2.7896 3.2658 0.6224 4.0815 1.2175
(0.0178)** (0.0098)*** (0.0224)** (0.0286)** (0.0283)** (0.4407) (0.028)** (0.093)*
MtoB -0.0067 -0.0061 -0.0065 -0.007 -0.01915 0.0058 0.0063 0.0255
(0.0039)*** (0.0014)*** (0.0047)*** (0.0211)** (0.5834) (0.812) (0.8594) (0.0477)**
Leverage∗i1 2.4804 2.6865 2.3808 3.9298 2.5843 5.3042 4.9609 -1.7340
(0.1779) (0.1469) (0.2107) (0.1531) (0.0897)* (0.0043)*** (0.0367)** (0.5502)
Leverage∗i2 3.6925 3.7973 3.4766 3.4813 8.0209 8.7762 7.2632 4.5905
(0.0114)** (0.0074)*** (0.0096)*** (0.0121)** (0.0383)** (0.0003)*** (0.0529)* (0.0275)**
Leverage∗i5 3.8440 3.5182 3.0871 0.8118 -6.3857 7.3571 21.8174 23.6858
(0.2891) (0.3383) (0.3264) (0.812) (0.4055) (0.6232) (0.0985)* (0.006)***
Leverage∗i7 2.8481 3.0269 2.8497 2.8233 5.8053 5.7883 3.0364 1.3849
(0.0258)** (0.0166)** (0.0275)** (0.0586)* (0.0002)*** (0.0013)*** (0.1145) (0.2511)
Constant -0.2021 -0.0227 -0.1614 -0.9773 0.0189 0.4794 -0.9315 -0.1898
(0.8412) (0.9818) (0.8708) (0.4069) (0.9856) (0.516) (0.4257) (0.8304)
Table 12:
Foreign Currency Exposure - Asset Volatility. This table presents the coefficients and the p-values from the Generalized Method
of Moments (GMM) estimation of the relationship between asset volatility on one side and the risk management activity, firm-specific
covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of firms with foreign currency
exposure lower than the sample mean and higher than the sample mean respectively. The dependent variable is asset volatility. The
independent variables, defined in the Appendix, are shown in the leftmost column. Regression 1 uses any risk management activity as
an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency, and commodity price risk management respectively.
Ind. Var. Regressions
Firms with FXExposure Lower than the Mean Firms with FXExposure Higher than the Mean
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM -0.5903 -0.1611
(0.017**) (0.5984)
IR RM -0.5583 -0.6867
(0.0294**) (0.0071)***
FX RM -0.5422 -1.0470
(0.0659)* (0.0536)*
CP RM 0.2441 -0.9129
(0.5233) (0.0881)*
Size -0.0537 -0.0671 -0.0724 -0.1182 -0.1246 -0.1159 -0.0355 -0.1106
(0.0889)* (0.0163)** (0.008)*** (<.0001)*** (0.0017)*** (<.0001)*** (0.5629) (0.0003)***)
Leverage 0.0513 0.3361 -0.0858 0.1830 0.1910 0.5343 -0.2161 0.0649
(0.712) (0.0261)** (0.6247) (0.1485) (0.1812) (0.0041)*** (0.4957) (0.7005)
Liquidity -0.2536 -0.1090 -0.2151 0.0733 0.3058 0.0965 -0.2067 0.2945
(0.1694) (0.4221) (0.2554) (0.552) (0.2445) (0.6265) (0.5806) (0.0766)*
MtoB 0.00008 0.0001 0.00009 0.0001 -0.0021 0.0006 0.0003 -0.0018
(0.3231) (0.1014) (0.2318) (0.139) (0.4034) (0.8526) (0.9083) (0.4265)
Leverage∗i1 -0.2851 -0.5457 -0.3878 -0.3326 -0.3232 -1.0099 -0.8133 0.5569
(0.284) (0.0576)* (0.1427) (0.3491) (0.0311)** (0.0023)***) (0.0656)* (0.3728)
Leverage∗i2 -0.125 -0.2683 0.0127 -0.1187 -1.0576 -1.5087 -1.0702 -0.9528
(0.6399) (0.3703) (0.968) (0.6812) (0.0001)*** (<.0001)*** (0.0085)*** (<.0001)***
Leverage∗i5 -0.5137 -0.1707 -0.1337 0.2447 -1.2464 -1.9183 -1.8046 -2.0546
(0.3935) (0.7448) (0.8048) (0.6184) (0.1315) (0.0056)*** (0.4447) (0.0221)**
Leverage∗i7 -0.0822 -0.3020 0.0732 -0.1073 -0.3104 -0.5012 -0.2335 -0.1914
(0.578) (0.0678)* (0.6482) (0.4516) (0.2343) (0.0602)* (0.5205) (0.4481)
Constant 1.4885 1.2708 1.4878 1.3654 1.5869 1.6130 1.6362 1.5160
(<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** <.0001)*** <.0001)***
Table 13:
Commodity Price Exposure - Distance to Default. This table presents the coefficients and the p-values from the Generalized
Method of Moments (GMM) estimation of the relationship between the distance to default on one side and the risk management activity,
firm-specific covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of firms with commodity
price exposure lower than the sample mean and higher than the sample mean respectively. The dependent variable the distance to default
calculated with the Black-Scholes-Merton model. The independent variables, defined in the Appendix, are shown in the leftmost column.
Regression 1 uses any risk management activity as an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency,
and commodity price risk management respectively.
Ind. Var. Regressions
Firms with CPExposure Lower than the Mean Firms with CPExposure Higher than the Mean
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM 3.0736 0.4384
(0.0274)** (0.603)
IR RM 5.1620 0.1621
(0.0226)** (0.7178)
FX RM 3.2170 -0.2018
(0.0158)** (0.7169)
CP RM -4.1088 1.5111
(0.2838) (0.3799)
Size 0.3919 0.3600 0.4285 0.9252 0.3100 0.3450 0.3638 0.3187
(0.0306)** (0.0627)* (0.015)** (0.0006)*** (0.1529) (0.0458)** (0.0349)** (0.1515)
Leverage -6.7628 -9.0050 -5.6754 -9.4003 -2.9705 -3.2274 -3.3012 -2.9345
(<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (0.0272)** (0.0147)** (0.0137)** (0.0506)*
Liquidity 1.7793 1.5618 2.4421 -0.2218 3.0087 2.7252 2.6032 2.6304
(0.1073) (0.226) (0.0545)* (0.8533) (0.0056)*** (0.0022)*** (0.0017)*** (0.0018)***
MtoB -0.0091 -0.0072 -0.0074 -0.0042 -0.0088 -0.0072 -0.0063 0.0026
(0.0008)*** (0.0046)*** (0.0047)*** (0.2821) (0.4598) (0.5331) (0.5834) (0.8601)
Leverage∗i1 2.1480 5.7228 2.3929 3.4780 2.7455 2.9747 2.8432 1.1270
(0.0916)* (0.0125)** (0.0614)* (0.1208) (0.0373)** (0.0385)** (0.0412)** (0.5804)
Leverage∗i2 7.7713 9.2257 6.8289 6.5667 1.7593 1.7903 1.9064 1.2508
(0.0002)*** (0.0002)*** (0.0003)*** (0.0003)*** (0.2989) (0.2868) (0.2157) (0.4273)
Leverage∗i5 -4.8842 -9.8502 -6.3248 -9.3643 10.4299 9.7754 9.4165 9.2184
(0.1134) (0.1736) (0.0482)** (0.0215)** (0.0017)*** (0.0004)*** (0.0004)*** (0.0003)***
Leverage∗i7 5.7689 7.72641 4.61527 -10.4228 2.5239 2.5480 2.5091 2.0382
(<.0001)*** (0.0002)*** (0.0002)*** (0.0691)* (0.0499)** (0.0298)** (0.01)*** (0.0449)**
Constant -0.3522 0.6399 -0.6424 0.3994 0.0849 0.2297 0.2836 0.3697
(0.7121) (0.5107) (0.5137) (0.7389) (0.9212) (0.7836) (0.7351) (0.7132)
Table 14:
Commodity Price Exposure - Asset Volatility. This table presents the coefficients and the p-values from the Generalized Method
of Moments (GMM) estimation of the relationship between asset volatility on one side and the risk management activity, firm-specific
covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of firms with commodity price
exposure lower than the sample mean and higher than the sample mean respectively. The dependent variable is asset volatility. The
independent variables, defined in the Appendix, are shown in the leftmost column. Regression 1 uses any risk management activity as
an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency, and commodity price risk management respectively.
Ind. Var. Regressions
Firms with CPExposure Lower than Mean Firms with CPExposure Higher than Mean
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM -0.6442 -0.1456
(0.006)*** (0.7449)
IR RM -1.3271 -0.4609
(0.0075)*** (0.0401)**
FX RM -0.7863 0.2291
(0.0152)** (0.3882)
CP RM -0.3875 -0.5900
(0.3743) (0.2011)
Size -0.0907 -0.0655 -0.0779 -0.1065 -0.0904 -0.0843 -0.1302 -0.0725
(0.0017)*** (0.1182) (0.0244)** (<.0001)*** (0.1145) (0.0002)*** (<.0001)*** (0.0252)**
Leverage 0.2949 0.7879 0.0446 0.2338 0.04207 0.2957 0.1764 -0.0482
(0.084)* (0.0133)** (0.8311) (0.1339) (0.7489) (0.0433)** (0.2987) (0.7462)
Liquidity -0.2711 -0.3670 -0.3160 0.2546 0.0661 0.0684 0.2477 0.0930
(0.2244) (0.2083) (0.2442) (0.0798)* (0.803) (0.5792) (0.1254) (0.4605)
MtoB 0.0011 0.0014 0.0005 0.0009 3.66E-06 0.00001 0.00001 3.75E-06
(0.0002)*** (0.0096)*** (0.2432) (0.0017)*** (0.9059) (0.6238) (0.741) (0.905)
Leverage∗i1 -0.2854 -1.1934 -0.3390 0.0489 -0.2581 -0.6381 -0.2108 0.2995
(0.031)** (0.0216)** (0.0947)* (0.8258) (0.1673) (0.0085)** (0.2932) (0.5645)
Leverage∗i2 -0.7501 -1.4478 -0.6260 -0.2964 -0.7657 -0.8936 -0.6032 -0.7374
(0.0157)** (0.0051)*** (0.0773)* (0.2832) (0.1126) (0.0531)* (0.1858) (0.112)
Leverage∗i5 -1.2203 -1.7379 -0.9882 -0.4239 0.2164 -0.0836 0.5462 0.1406
(0.0079)*** (0.013)** (0.0416)** (0.0876)* (0.8157) (0.916) (0.4657) (0.8613)
Leverage∗i7 -0.3356 -0.7435 -0.1254 -0.2218 -0.3416 -0.6515 -0.3960 -0.3757
(0.0764)* (0.02)** (0.5358) (0.1876) (0.1768) (0.0069)*** (0.0165)** (0.032)**
Constant 1.7319 1.5513 1.7025 1.3502 1.3866 1.3290 1.3734 1.2827
(<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)***
Table 15:
Bankrupt and Non-bankrupt - Distance to Default. This table presents the coefficients and the p-values from the Generalized
Method of Moments (GMM) estimation of the relationship between the distance to default on one side and the risk management activity,
firm-specific covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of bankrupt and
non-bankrupt firms. The dependent variable the distance to default calculated with the Black-Scholes-Merton model. The independent
variables, defined in the Appendix, are shown in the leftmost column. Regression 1 uses any risk management activity as an independent
variable. Regressions 2, 3, and 4 use interest rate, foreign currency, and commodity price risk management respectively.
Ind. Var. Regressions
sub-sample of Bankrupt Firms sub-sample of Nonbankrupt Firms
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM -0.0248 -0.4703
(0.9809) (0.5739)
IR RM -1.0202 -1.0756
(0.1734) (0.0805)*
FX RM 0.7389 -0.0264
(0.5294) (0.9746)
CP RM 0.5830 0.9550
(0.7100) (0.5328)
Size -0.0855 -0.0738 -0.1481 -0.1547 0.0284 -0.0192 -0.0282 -0.1374
(0.5073) (0.4021) (0.2518) (0.4400) (0.8077) (0.8002) (0.8091) (0.4707)
Leverage -2.1325 -2.0522 -2.4609 -1.9795 -3.2044 -3.2147 -2.9237 -2.6499
(0.1501) (0.1356) (0.1196) (0.1821) (0.0044)*** (0.0036)*** (0.0098)*** (0.0238)**
Liquidity 3.4118 0.7589 3.3844 3.3684 1.4743 -1.2362 2.3145 2.5660
(0.0469) (0.7047) (0.0358) (0.0385) (0.3842) (0.4835) (0.1316) (0.0719)*
MtoB -0.0005 0.0002 -0.0143 0.0009 -0.0143 -0.0173 -0.0163 -0.0134
(0.9846) (0.9909) (0.6312) (0.9694) (0.3383) (0.2379) (0.2731) (0.4283)
Leverage∗i1 3.4883 1.5386 6.2987 2.6337 5.5089 3.8257 5.2067 3.7268
(0.4428) (0.7296) (0.3403) (0.5635) (<.0001)*** (0.0004)*** (<.0001)*** (0.119)
Leverage∗i2 1.8272 1.2110 2.5055 1.6075 2.1803 1.8687 2.2598 1.8817
(0.223) (0.3321) (0.1205) (0.2162) (0.0685)* (0.0953)* (0.0448)** (0.1046)
Leverage∗i5 38.3967 35.3690 36.7116 38.1426 4.63092 5.9402 4.7374 5.4845
(0.2853) (0.2695) (0.3155) (0.2889) (0.1773) (0.0829)* (0.2204) (0.2138)
Leverage∗i7 1.2919 1.5033 1.7599 1.1607 2.5221 2.8836 2.2542 1.9233
(0.4001) (0.2855) (0.2897) (0.4372) (0.027)** (0.0095)*** (0.0463)** (0.1013)
Constant 1.1283 1.7053 1.0802 1.4165 1.3989 2.0745 1.2469 1.6811
(0.1702) (0.0441) (0.1719) (0.2066) (0.0705)* (0.0148)** (0.1054) (0.1439)
Table 16:
Bankrupt and Non-bankrupt - Asset Volatility. This table presents the coefficients and the p-values from the Generalized Method
of Moments (GMM) estimation of the relationship between asset volatility on one side and the risk management activity, the firm-specific
covariates, and industry - leverage interaction terms on the other side, separately for the sub-samples of bankrupt and non-bankrupt
firms. The dependent variable is asset volatility. The independent variables, defined in the Appendix, are shown in the leftmost column.
Regression 1 uses any risk management activity as an independent variable. Regressions 2, 3, and 4 use interest rate, foreign currency,
and commodity price risk management respectively.
Ind. Var. Regressions
sub-sample of Bankrupt Firms sub-sample of NonBankrupt Firms
Regression 1 Regression 2 Regression 3 Regression 4 Regression 1 Regression 2 Regression 3 Regression 4
Any RM -1.6683 -0.3366
(0.012)** (0.1084)
IR RM -0.6812 -0.5894
(0.0021)*** (0.0676)*
FX RM -1.3872 -0.4952
(0.0036)*** (0.026)**
CP RM -0.3091 -0.7722
(0.4502) (0.1706)
Size 0.0087 -0.1031 -0.0439 -0.0848 -0.0676 -0.0505 -0.056 -0.0769
(0.8974) (0.0001)*** (0.3925) (0.0327)** (0.0223)** (0.1335) (0.0706)* (0.0025)***
Leverage -0.0718 0.1811 -0.2727 -0.0645 0.1151 0.4736 -0.2112 0.1064
(0.8309) (0.2708) (0.3912) (0.7253) (0.3634) (0.0755)* (0.2894) (0.4908)
Liquidity -1.0610 -0.1393 -0.6412 0.3686 -0.0014 0.1013 -0.1204 0.1658
(0.0905)* (0.5661) (0.1034) (0.049)** (0.9918) (0.4005) (0.4531) (0.1418)
MtoB 0.0008 0.0009 -0.0009 -0.0001 0.00005 0.00006 0.00005 0.00004
(0.4008) (0.1252) (0.2175) (0.6264) (0.2885) (0.1182) (0.2256) (0.3737)
Leverage∗i1 0.5894 -1.1755 -1.6558 -0.0831 -0.1579 -0.5588 -0.1068 0.2739
(0.2862) (<.0001)*** (0.0039)*** (0.9043) (0.4472) (0.1429) (0.6595) (0.4597)
Leverage∗i2 -0.5271 -0.2208 -0.4109 0.1244 -0.6019 -0.8955 -0.3308 -0.4854
(0.1698) (0.5493) (0.3707) (0.7417) (0.059)* (0.0214)** (0.2828) (0.0684)*
Leverage∗i5 -1.3389 -0.1623 -0.5712 0.0988 -1.3095 -1.4385 -1.4262 -1.0037
(0.1287) (0.752) (0.349) (0.859) (0.0017)*** (0.0017)*** (0.0007)*** (0.0017)***
Leverage∗i7 -0.1777 -0.2572 0.0613 -0.0678 -0.3011 -0.0761 -0.3105 0.0227
(0.6037) (0.1267) (0.8373) (0.6876) (0.4713) (0.8861) (0.3935) (0.9421)
Constant 2.0086 1.6780 1.9796 1.4282 1.2970 1.0885 1.3360 1.1785
(<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)*** (<.0001)***