You are on page 1of 18

Chapter 60

Alternative Methods to Determine Optimal Capital Structure: Theory


and Application

Sheng-Syan Chen, Cheng-Few Lee, and Han-Hsing Lee

Abstract In this paper, we review the most important and tion (option pricing) theory for analyzing classical corporate
representative capital structure models. The capital struc- finance issues and can deepen the understanding of the firm’s
ture models incorporate contingent claim valuation theory financial decisions. Furthermore, the contingent claim valua-
to quantitatively analyze prevailing determinants of capital tion enables risky debt pricing using marketable securities of
structure in corporate finance literature. In capital structure the firm, such as common stock, bond or credit default swaps
models, the valuation of corporate securities and financial de- (CDS) prices, to estimate the firm value parameters and in
cisions are jointly determined. Most of the capital structure turn the default risk. This approach avoids the use of bond
models provide closed-form expressions of corporate debt rating and risk premium (markup), which are difficult to pin
as well as the endogenously determined bankruptcy level, down and often are stale in contrast to the trading data of
which are explicitly linked to taxes, firm risk, bankruptcy marketable securities.1
costs, risk-free interest rate, payout rates, and other impor- The structural credit risk modeling or defaultable claim
tant variables. The behavior of how debt values (and there- modeling is pioneered by the seminal paper by Black and
fore yield spreads) and optimal leverage ratios change with Scholes (1973) in which corporate liabilities can be viewed
these variables can thus be investigated in detail. as a covered call – own the asset but short a call option. Black
and Scholes (1973) assume that a company issues only one
Keywords Optimal capital structure r Structural model r zero-coupon bond and equity. At the maturity, debt holders
Capital structure model r Credit risk either receive the face value of debt if the company is sol-
vent, or take control of the company otherwise. This option
analogy allows the closed-form expression for the corporate
securities. Later on, Merton (1974) rigorously elaborates the
60.1 Introduction pricing of corporate debt. In addition, the valuation of any
corporate security can be modeled as a contingent claim on
The traditional capital structure theories investigate how the the underlying value of firm. This Black-Scholes-Merton ap-
capital structure of a firm can be affected by some crucial proach of modeling default claims is named structural ap-
determinants, such as the financial distress costs and agency proach since the model explicitly ties default risk to the firm
costs. However, it is a difficult task to measure their effects value process and its capital structure.
on the capital structure decisions and security pricing. It is In modeling the cause of default-trigger mechanism, the
of great interest for corporate managers and researchers to Black-Scholes-Merton approach implicitly assumes that a
quantitatively examine how these effects influence a com- company can only default at a certain time point when its
pany and its financial decisions under various conditions. bond principal is due. While this default trigger mechanism
The capital structure models employ contingent claim valua- is apparently unrealistic, Black and Cox (1976) first propose
a model in that default can occur during the life of the bond –
a company defaults on its debt if the asset value of the com-
S.-S. Chen pany falls below a certain exogenous default boundary due
National Taiwan University, Taiwan
e-mail: fnschen@management.ntu.edu.tw
to bond covenant. In the same paper, Black and Cox (1976)
also point out that default decision can be endogenized by
C.-F. Lee
Rutgers University, USA
e-mail: lee@business.rutgers.edu 1
Another advantage of the structural approach is that macro- and micro-
H.-H. Lee () conditions have been implicitly incorporated into the evolution (change)
National Chiao Tung University, Taiwan of firm values and, in turn, security prices, and need not be modeled
e-mail: lee@rbsmail.rutgers.edu separately.

C.-F. Lee et al. (eds.), Handbook of Quantitative Finance and Risk Management, 933
DOI 10.1007/978-0-387-77117-5_60,  c Springer Science+Business Media, LLC 2010
934 S.-S. Chen et al.

shareholders’ discretion to maximize equity value. These bankruptcy imply that debt and equity holders do not split a
two types of default mechanisms, exogenously specified by claim depending only on the underlying asset value.
bond covenant and endogenously determined by manage- While the Leland (1994) and the Leland and Toft (1996)
ment, have been widely adopted in most of the succeeding models provide some insights of the capital structure issues,
structural models.2 their predicted optimal leverage and yield spreads seem not
Subsequent researchers extend the structural approach to be consistent with historical average. Consequently, the
further and incorporate more realistic assumptions. Based on more recent studies introduce additional realistic features in
the key objectives of these models, one can classify them as order to meliorate the model-predicted leverage and yield
corporate bond pricing model and capital structure model. spreads. We will also review the some of these extensions.
Corporate bond pricing models, typically assuming stochas- This paper is organized as follows: In Sect. 60.2, we
tic interest rate, are primarily developed to price defaultable will briefly review traditional capital structure theory. In
corporate debt. In order to accommodate a more realistic Sect. 60.3, we will present in detail two most representative
and complex debt structure, these models typically assume capital structure models – Leland (1994), and Leland and
an exogenous default barrier and bondholders receive a re- Toft (1996) models – and their implications. In Sect. 60.4,
covery rate times the face value of debt at maturity if a firm more recent studies including debt renegotiation, dynamic
defaults.3 For example, the Longstaff and Schwartz model capital structure, as well as other developments are pre-
(1995) is the most well-known corporate bond pricing model. sented. In Sect. 60.5, the applications and empirical per-
By contrast, the capital structure models incorporate corpo- formance of capital structure models are discussed. In
rate finance theory and put the emphasis on the default event Sect. 60.6, conclusion is given.
itself. The valuation of corporate securities and financial de-
cisions are jointly determined, and the analyses focus on the
complex relations between credit risk, risk premiums, and 60.2 The Traditional Theory of Optimal
the firm’s financing decisions rather than on the valuation Capital Structure5
of complex credit securities. In contrast to corporate bond
pricing models, the simple stationary debt structure assump-
Durand (1952) proposes an approach to explain why there
tion, debt structure with time-independent payouts, is widely
is an optimal capital structure for a firm. Called the tradi-
adopted in order to obtain closed-form expression of debt,
tional approach, it suggests that “moderate” amounts of debt
equity, and firm value. In this paper, we restrict ourselves to
do not noticeably increase the risks to either the debtholders
capital structure models and exclude those models focusing
or the equityholders. Hence, both cost of debt (i ) and cost
on risky debt pricing.4
of equity (rE ) are relatively constant up to some point. How-
In the capital structure models, the most notable devel-
ever, beyond some critical amount of debt used, both i and
opments are the works by Leland (1994) and Leland and
rE increase. These increases in both the cost of debt and eq-
Toft (1996). They extend the endogenous default approach
uity begin to offset the decrease in the average cost of capital
by Black and Cox (1976) to include the tax shield of debt and
because of cheaper debt. Therefore, the average cost of cap-
bankruptcy costs, and explicitly analyze the static tradeoff
ital (rA ), initially falls and then increases. This allows for a
theory of capital structure. Various important issues includ-
minimum value for the average cost of capital function. If
ing the endogenous default boundary, leverage ratio, debt
the average cost of capital is at the minimum, then the firm’s
value, debt capacity, yield spread, and even potential agency
value (V ) is maximized
costs are discussed in detail. Note that apart from the early
structural models, Leland and Toft (1996) point out that .1  TC /EBIT
equity in a capital structure model is not precisely analo- V D (60.1)
rA
gous to an ordinary call option or vanilla barrier option. The
bankruptcy boundary varies with the risk of the firm’s activ- where TC D corporate tax rate and EBIT D earnings before
ities. Furthermore, the tax benefits and the potential loss in interest and tax. This implies that there is an optimum capital
structure for the firm.

2
Geske (1977) proposes another way of modeling corporate debts as 60.2.1 Bankruptcy Costs
compound options.
3
There are some exceptions. For example, Briys and de Varenne (1997) The theoretical analysis of Modigliani and Miller (1958)
assume recovery of firm value when a firm defaults. However, the trade- indicates that in the presence of corporate taxes (TC ), the
off is its simpler debt structure, a zero-coupon bond.
4
Readers who are interested in other structural models can refer to the
5
survey paper by Elizalde (2005). This chapter is an excerpt from the book by Lee et al. (2009).
60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 935

value of the levered firm (VL ) exceeds the value of the other- incurred in the event of bankruptcy and (2) the likelihood, or
wise comparable unlevered firm (VU ). The difference can be probability, of bankruptcy occurring. The greater the amount
viewed as an increment to present value derived from the tax of debt a firm has, the more likely it will fail to meet its
shelter (TC B). Therefore, we can write obligations and be driven to bankruptcy. Therefore, seeing
this risk, bondholders will demand higher rates of interest
VL D VU C T C B (60.2) the larger the debt. Thus, expected bankruptcy cost is an in-
creasing function of the proportion of debt in a firm’s capital
where B D total amount of debt. structure. Taking into account the possibility of bankruptcy
One issue not considered in deriving this result is the pos- and its effect on firm value, we can augment Equation (60.2)
sibility of corporate bankruptcy. Obligations to bondhold- as follows:
ers constitute a first claim on a company’s assets. However,
under US law, stockholders, who are the owners of public VL D VU C TC B  present value of bankruptcy costs
companies, have only limited liability. Their personal assets, (60.3)
other than holdings in the firm, are protected from the claims
The higher the level of debt, the greater the present value of
of bondholders. If operating performance is sufficiently poor,
the tax shield. As we saw in Equation (60.2), the effect is lin-
stockholders can avoid obligations to bondholders through
ear, in the sense that the value of the tax shield is a constant
the firm’s declaring bankruptcy. Essentially, bankruptcy turns
multiple TC (the corporate tax rate) of the debt value. How-
over the firm to its debtholders. Bankruptcy proceedings can
ever, as Equation (60.3) indicates, one must now subtract the
be very costly. Heavy legal fees must be paid, and assets
present value of bankruptcy cost. This, too, is an increasing
must often be sold at “bargain basement” prices, frequently
function of the debt level, reflecting the increase in the prob-
following lengthy and expensive disputes among competing
ability of bankruptcy that attaches to increasing debt levels.
claimants. Taken together, these bankruptcy costs may be far
If these factors operate in conjunction, as illustrated in
from negligible, and the possibility of their being incurred
Fig. 60.1, then an optimal capital structure results. One sees
must be considered in assessing a firm’s market value.
for the unlevered firm, zero values for the tax shield and
After bankruptcy, the firm belongs to its creditors, who
bankruptcy costs. Each increases with the level of debt is-
must therefore bear these costs. However, purchasers of cor-
sued, the net effect being a relationship between firm value
porate bonds will be aware of this and will charge a higher
and debt level that shows a maximum value for a debt level
price – that is, demand a higher interest rate – for debt. This,
well below 100% of total value. Myers (1984) called the opti-
in turn, reduces the amount of earnings available for distri-
mal capital structure discussed in this section the static trade-
bution to stockholders, reducing the value of shares. Hence,
off theory.6
before bankruptcy, costs are carried by stockholders as re-
ductions in the value of their equity holdings.
How much does the possibility of bankruptcy lower the 6
Myers (1984) also proposes a pecking-order theory to contrast to the
firm’s value? Two factors must be considered: (1) the costs static trade-off theory.

Fig. 60.1 Possibility of optimal


debt ratio when allowing for
present value of bankruptcy costs
936 S.-S. Chen et al.

Fig. 60.2 Possibility of optimal


debt ratio when allowing for
present value of bankruptcy and
agency costs

60.2.2 Agency Costs be required. Thus, agency costs affect firm value in much
the same way as bankruptcy costs. The greater the debt, the
Agency costs are legal and contract costs that protect the greater the agency cost against firm value. This is illustrated
rights of stockholders, bondholders, and management, and in Fig. 60.2, which extends Fig. 60.1 by allowing for agency
arise when the interests of these parties come into conflict. cost. At any level of debt above zero, agency cost causes a de-
Such organizational conflict may lead to capital structure de- crease in market value; the greater the debt level, the larger
cisions that affect the firm’s value. Asset substitution prob- the agency cost factor. Hence, agency cost yields a lower op-
lem arises when a firm invests in assets that are riskier than timal amount of debt in the capital structure than would oth-
those that the debtholders expect. For example, having sold erwise prevail. If there is agency cost, then Equation (60.3)
debt, stockholders may gain if the corporation abandons can be rewritten as
its low-risk projects in favor of more risky projects. If the
higher-risk projects are successful, the stockholders will ben- VL D VU C TC B  present value of bankruptcy costs
efit from the high returns. On the other hand, this exposes
Agency cost (60.3a)
bondholders to a higher degree of risk than they anticipate
when the bonds were purchased, and thus they face a higher-
In a review paper, Harris and Raviv (1991) survey capi-
than-expected chance of corporate bankruptcy. tal structure theories based on agency costs, asymmetric in-
Bondholders anticipate such possibilities and require that formation, product/input market interactions, and corporate
bonds be protected by restrictive covenants. Such covenants control considerations (but excluding tax-based theories).
involve two types of costs to the firm. First, there are costs Therefore, this paper gives us further detailed information
in monitoring the covenants, and second, these covenants re- about the result discussed in this section.
strict management flexibility, possibly precluding courses of
action that would raise the firm’s market value. These costs
are borne by stockholders, resulting in a lower market value
60.3 Optimal Capital Structure
for equity. Jensen and Meckling (1976) use agency costs of
debt and equity to suggest that there is an optimum com- in the Contingent Claims Framework
bination of outside debt and equity that will be chosen to
minimize total agency costs. Therefore, it is possible for the The traditional capital structure theory identifies some prime
existence of an optimal capital structure even in the absence determinants of optimal capital structure. Nevertheneses, the
of taxes and bankruptcy costs. theory is less useful in practice since it provides only qualita-
In general, one might expect that agency costs of this sort tive guidelines. The structural approach enables quantitative
will be greater the higher the debt. If there is little debt, bond- analysis and thus has become very popular in the corporate
holders will consider their investments relatively secure and finance literature. We will review in detail how the determi-
demand only modest, fairly inexpensive protection of their nants mentioned in Sect. 60.2 affect capital structure deci-
interest. At higher debt levels, more stringent assurances will sions under the contingent claim framework in this section.
60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 937

In the analysis of optimal capital structure, Brennan and with boundary conditions determined by payments at matu-
Schwartz (1978) are the first to provide the quantitative rity, and by payments in bankruptcy happens prior to matu-
examination of optimal leverage using contingent claims rity. When securities have no explicit time dependence as this
approach. However, their analysis employs numerical tech- case, the term Ft .V; t/ D 0, and the above equation becomes
niques and lacks a close-form solution. As a result, their an ordinary differential equation with F .V / satisfying
numerical-based study cannot provide general comparative
statics of financial variables and further discussion. In this 1 2 2
 V FVV .V; t/ C rVFV .V; t/  rF.V; t/ C C D 0:
section, we review in detail two most representative capital 2
(60.5)
structure models – Leland (1994) and Leland and Toft (1996)
models – and the implication of these models. These two The equation has the general solution
studies provide closed-form expressions of corporate secu-
rities as well as comparative statics. Their methodology and F .V / D A0 C A1 V C A2 V X ; (60.6)
stationary debt structures are widely adopted in the litera-
ture. In addition, these two models have served as benchmark where X D 2r= 2 and the constants A0 , A1 , and A2 are
models in the optimal capital structure literature. determined by boundary conditions. Any time-independent
claim with an equity-financed constant payout C 0 must
have this functional form.
First, Leland (1994) provides the solution of the perpetual
60.3.1 The Leland (1994) Model
debt. Debt promises a perpetual coupon payment, C , whose
level remains constant unless the firm declares bankruptcy.
Leland (1994) extends the endogenous default approach by Let VB denote the constant level of asset value at which
Black and Cox (1976) to include the tax shield of debt and bankruptcy is declared. If bankruptcy occurs, a fraction 0 
bankruptcy costs, which is also termed the Static tradeoff ˛  1 of asset value will be lost to bankruptcy cost. Accord-
model in the corporate theory. In the Leland (1994) model, fi- ingly, debtholders receive .1˛/VB and shareholders receive
nancial distress is triggered when shareholders no longer find nothing.7
that running a company is profitable, given the revenue pro- Boundary conditions are:
duced by the assets to continue servicing debt. Bankruptcy
is determined endogenously rather than by the imposition of At V D VB ; D.V / D .1  ˛/VB
a positive net worth condition or by a cash flow constraint.
In this model, two major assumptions are as follows: (1) The As V ! 1, D.V / D C =r
activities of the firm are unchanged by financial structure,
and the capital structure decisions, once made, are not sub-  
X
C C V
sequently changed; (2) The face value of debt, once issued, D.V / D C .1  ˛/VB  (60.7)
r r VB
remains static through time.
Consider a firm with asset value V , which follows a dif- The above equation can be rewritten as D.V / D Œ1  pB
fusion process with constant volatility of rate of return: .C =r/ C pB Œ.1  ˛/VB where pB  .V =VB /X has the
interpretation of the present value of $1 contingent on future
dV bankruptcy; that is, V falls to VB .
D .V; t/dt C dW (60.4)
V Next, Leland (1994) derives the total value of the firm,
where W is a standard Brownian motion. The stochastic pro- v.V /, which reflects three terms: the firm’s asset value,
cess of V is assumed to be unaffected by the financial struc- plus the value of the tax deduction of coupon payments,
ture of the firm. Therefore, any net cash outflows associated less the value of bankruptcy costs. Consider a security pays
with the choice of leverage such as coupons after tax bene- no coupon, but has a value equal to the bankruptcy costs.
fits must be financed by selling additional equity. In addition, Because its returns are time independent, it must satisfy
there exists a riskless asset that pays a constant rate of inter- Equation (60.6) with boundary conditions
est r.
Denote any claim F .V; t/ on the firm that continuously At V D VB ; BC.V / D ˛VB
pays a nonnegative coupon, C , per instant of time when the As V ! 1; BC.V / D 0:
firm is solvent. When the firm finances the net cost of the
coupon by issuing additional equity, any such asset’s value
must satisfy the PDE: 7
Leland (1994) also has a discussion (in Sect. VI.C) under the condition
of deviation of absolute priority, in which bondholders do not receive all
1 2 2 of the remaining asset value and stockholders also receive a fraction of
 V FVV .V; t/CrVFV .V; t/rF.V; t/CFt .V; t/CC D 0 the remaining asset value.
2
938 S.-S. Chen et al.

The solution in this case is Note that Equation (60.10) is similar to Equation (60.3).
If .V =VB /X approaches to zero, then Equation (60.10) re-
˛VB .V =VB /X ; (60.8) duces to original M&M proposition I as defined in Equation
(60.2). For this condition to hold, either (i) .V =VB / ! 1 or
and BC.V / is a decreasing convex function of V . (ii)  2 ! 0 and V > VB . In both cases, the firm is un-
Next consider the value of tax benefits associated with likely to go bankrupt, and therefore tax saving per period
debt financing. Denote the corporate tax rate as TC . These on the interest payments becomes a sure stream. As a re-
benefits resemble a security that pays a constant coupon sult, the tax benefit claim in Equation (60.9) will be TC C =r,
equal to the tax-sheltering value of interest payments, TC C , and Equation (60.10) will coincide with the original M&M
as long as the firm is solvent. In other words, TB.V / equals proposition I in Equation (60.2), even with the existence of
the value of the tax benefit of debt. Since it is also time in- bankruptcy cost. On the other hand, without the bankruptcy
dependent and must satisfy Equation (60.6) with boundary cost (˛ D 0), the tax benefit claim under Leland’s specifi-
conditions cation is still different from that by M&M. The difference, a
minus term (.TC C =r/.V =VB /X ), is coming from the pos-
At V D VB ; TB.V / D 0 sibility of future bankruptcy, and thus lowers the tax shield
of debt.
As V ! 1; TB.V / D TC C =r:
The value of equity is the total value of the firm less the
value of debt as follows:
The solution with these boundary conditions is
E.V / D v.V /  D.V / D V  .1  TC /.C =r/
TC C =r  .TC C =r/.V =VB /X (60.9)
CŒ.1  TC /.C =r/  VB .V =VB /X:
8
The TB.V / is an increasing, strictly concave function of V .
(60.11)
Note that the value of tax benefits above presumes that
the firm always benefits fully in amount TC C from the tax
Leland next provides the bankruptcy-triggering asset level
deductibility of coupon payments when it is solvent. How-
VB chosen by the firm. To maximize the value of equity at
ever, under U.S. tax codes, to benefit fully, the firm must have
any level of V , the equilibrium bankruptcy-triggering asset
earnings before interest and taxes (EBIT) that is at least large
value VB is determined ˇ endogenously by the smooth-pasting
as the coupon payments C . Leland (1994) also provides the @E.V IVB ;T / ˇ
solution, in which EBIT is related to asset value,V , and tax condition @V ˇ D 0.
V DVB
benefits may be partially lost when the firm is still solvent.
In sum, the total value of the firm, v.V /, consisting of the VB D Œ.1  TC /C =r ŒX=.1 C X / D .1  TC /C =.r C 0:5 2 /
above three terms, the firm’s asset value, the value of the tax (60.12)
deduction of coupon payments, and the value of bankruptcy
Since VB < .1 TC /C =r, equity is indeed convex in V . Sev-
costs, can be expressed as:
eral observations of the asset value, VB , at which bankruptcy
occurs are:
v.V / D V C TB.V /  BC.V /
(1) It is proportional to the coupon, C .
D V C .TC C =r/Œ1  .V =VB /X  ˛VB .V =VB /X
(2) It is independent of the current asset value, V .
(60.10) (3) It decreases as the corporate tax rate, TC , increases.
(4) It decreases as the risk-free interest rate, r, rises.
v.V / is strictly concave in asset value, V , when C > 0 and (5) It decreases with increases in the riskiness of the
either ˛ > 0 or TC > 0. Note that if ˛ > 0 or TC > 0, then firm,  2 .
v.V / < V as V ! VB , and v.V / > V as V ! 1. This
Finally, VB in Equation (60.12) is independent of time
coupled with concavity implies that v.V / is proportionally
and it confirms the assumption of the constant bankruptcy-
more volatile than V at low values of V , and is less volatile
triggering asset level VB .
at high value of V .
In addition to the determinants of bankruptcy level, with
the closed-form expressions, corporate debt values and op-
timal leverage are explicitly linked to taxes, firm risk,
8
Note that this equation represents that the firm always benefitfully bankruptcy costs, risk-free interest rate, and bond covenant.
from the tax deductibility of coupon payment when it is solvent. How-
The behavior of how debt values (and therefore yield
ever, to benefit fully the firm must have EBIT over the coupon payment,
C . Leland (1994) provides an alternative approach in Appendix A in spreads) and optimal leverage ratios changes with these vari-
which tax benefits may lost when the firm is still solvent. ables is discussed in detail. In addition, Leland shows that a
60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 939

positive net worth covenant9 can eliminate the firm’s incen- of bankruptcy. Let f .sI V; VB / denote the density of the first
tive to increase risk and lower the agency problem. passage time s from V to VB , and F .sI V; VB / is the cu-
However, optimal capital structure refers not only to lever- mulative distribution function of the first passage time to
age ratio, but also to the maturity of debt. The consol bond bankruptcy; that is, the default probability at time s. Under
assumption of the Leland model confines the analysis in risk-neutral valuation with the drift rate .r  ı/, the bond
this dimension. As a result, Leland and Toft (1996)10 pro- with maturity t has the value
pose a finite maturity debt structure for further discussion of Z t
this issue.
d.V I VB ; t/ D e rs c.t/Œ1  F .sI V; VB / ds
0

Ce rt p.t/Œ1  F .sI V; VB /


Z t
60.3.2 The Leland and Toft (1996) Model
C e rs .t/VB f .sI V; VB /ds
0
In the Leland and Toft (1996) model, financial distress is
The first term represents the discounted expected value of
triggered, as assumed in Leland (1994), when shareholders
the coupon flow; the second term represents the expected
no longer find that running a company is profitable. In this
discounted value of repayment of principal; and the third
model, they further examine the optimal capital structure of
term represents the expected discounted value of the frac-
a firm that can choose both the amount and maturity of its
tion of the assets, which will go to debt with maturity t, if
debt. They again setup a stationary debt structure by issu-
bankruptcy occurs. Integrating the first term by parts yields
ing finite maturity debt with continuous coupon and retiring
the same constant amount (principal) of debt issued in the  
c.t/ c.t/
past simultaneously. As a result, at any given time, the firm d.V I VB ; t/ D C e rt p.t/  Œ1  F .t/
r r
has many overlapping debt outstanding, while the total cash  
flow to debt holders is constant. The results extend Leland’s c.t/
C .t/VB  G.t/ (60.14)
(1994) closed-form results to a much richer class of possible r
debt structures and dividends payment and permit study of
the optimal maturity of debt as well as the optimal amount of where Z t
debt. G.t/ D e rs f .sI V; VB /ds
The firm has productive assets whose unleveraged value 0

V follows a continuous diffusion with constant proportional Using the result of F .t/ from Harrison (1985) and G.t/ from
volatility  Rubinstein and Reiner (1991):

2a
dV V
D Œ.V; t/  ı dt C dZ (60.13) F .t/ D N Œh1 .t/ C N Œh2 .t/ (60.15)
V VB

aCz
az
where .V; t/ is the total expected rate of return on asset V V
value V ; ı is the constant fraction of value paid out to secu- G.t/ D N Œq1 .t/ C N Œq2 .t/
VB VB
rity holders; and dZ is the increment of a standard Brownian
(60.16)
motion. The process continues without time limit unless V
falls to a default-triggering valueVB . VB will be determined
where
endogenously and shown to be constant in a rational expecta-
tions equilibrium. Finally, a default-free asset exists and pays .b  z 2 t/ .b C z 2 t/
a continuous constant interest rate r. q1 .t/ D p I q2 .t/ D p
 t  t
Consider a bond issue with maturity t periods from the
present, which continuously pays a constant coupon flow .b  a 2 t/ .b C a 2 t/
c.t/ and has principal p.t/. Let .t/ be the fraction of as- h1 .t/ D p I h2 .t/ D p
 t  t
set value VB , which debt of maturity t receives in the event

.r  ı  . 2 =2// V
aD I b D ln I
9
The positive net-worth covenant is modeled by setting default bound-
2 VB
ary equal to the principal value of debt.
Œ.a 2 /2 C 2r 2 1=2
10
Leland (1994b) proposes an exponential model similar to sinking zD
fund that firms retire debt at a proportional rate m through time. In the 2
absence of default, the average
R1 maturity of debt is finite as the time-
weighted average: T D 0 t .memt /dt D 1=m. and N./ is the cumulative standard normal distribution.
940 S.-S. Chen et al.

A stationary debt structure with time-independent debt The equilibrium bankruptcy-triggering asset value VB is
service payments is assumed by Leland and Toft (1996) as determinedˇ endogenously by the smooth-pasting condition
@E.V IVB ;T / ˇ
the following. Consider an environment where the firm con-
@V ˇ D 0.
tinuously sells a constant (principal) amount of new debt with V DVB

maturity of T years from issuance, which (if solvent) it will


.C =r/ .A=.rT/  B/  AP=.rT/  TC Cx=r
redeem at par upon maturity. New bond principal is issued at VB D
a rate p D .P =T / per year, where P is the total principal 1 C ˛x  .1  ˛/B
(60.19)
value of all outstanding bonds. The same amount of princi-
pal will be retired when the previously issued bonds mature. p p
Therefore, as long as the firm remains solvent, at any time s where A D 2aerT N.a T /  2zN.z T /  2
p n
p rT p  T
the total outstanding debt principal will be P , and have a uni- .z T / C 2ep n.z T / C .z  a/
 T
form distribution of principal over maturities in the interval

p p
.s; s C t/. 2 2
Bonds with principal p pay a constant coupon rate c D B D  2z C N.a T /  p n.z T /
z 2 T  T
.C =T / per year, implying the total coupon paid by all out-
1
standing bonds is C per year. Total debt service payments C.z  a/ C
are therefore time-independent and equal to .C C P =T / per z 2 T
year. Accordingly, this environment assumed by Leland and
x D a C z11 and n./ denotes the standard normal density
Toft (1996) avoids the time-dependent debt service require-
function.
ments such as the single issue of debt outstanding assumed
The solution is independent of time, indicating that the
by Merton (1974).
earlier assumption of a constant VB is warranted: it repre-
Assuming the fraction of firm asset value lost in
sents a rational expectations equilibrium. As T ! 1, it can
bankruptcy is ˛ and .t/ D =T per year for all t, this im-
be shown that VB ! .1  TC /.Cx=r/=.1 C x/ as in Equation
plies  D .1  ˛/. Therefore, writedown for debt is the fixed
(60.12). Note that VB will depend on the maturity T of debt
fraction of the principal value of debt lost in bankruptcy,
chosen, for any given values of total bond principal P and
which is .1  .1  ˛/VB =P /. Leland and Toft derive the
coupon rate C . This is an important difference between Le-
closed-form solution of the value of all outstanding bonds,
land and Toft model and the flow-based bankruptcy model
with debt of maturity T years, by integrating all the outstand-
(e.g., Kim et al. (1993) or with a positive net worth covenant
ing debt during this period:
model (e.g., Longstaff and Schwartz (1995), in which they
Z T imply that VB is independent of debt maturity.
D.V I VB ; T / D d.V I VB ; t/dt And finally, the value of equity is given by
0

C C 1  e rT
D C P  I.T / E.V I VB ; T / D v.V I VB /  D.V I VB ; T / (60.20)
r r rT

C
C .1  ˛/VB  J.T /; (60.17) In contrast to the Leland (1994) model, the Leland and
r
Toft (1996) model has some different implications. First, by
where permitting a firm to choose its debt maturity, optimal lever-
Z T
age is lower when the firm is financed by shorter-term debt.
1 1 They also provide explanation of why a firm issues short-
I.T / D .G.T /  e rT F .T //;
e rt F .t/dt D
T0 rT term debt given that the longer-term debt generates higher
Z
aCz firm value. They find that short-term debt reduces the asset
1 T rt 1 V
J.T / D e G.t/dt D p  substitution agency problem. While short-term debt does not
T 0 z T V B
exploit tax benefit of debt, it can balance against bankruptcy

az !
V and agency costs in determining the optimal maturity of the

N Œq1 .T / q1 .T / C N Œq2 .T / q2 .T / : capital structure. This is a consequence that short-term debt
VB

Following Equation (60.10) by Leland (1994), total firm 11


If the payout rate (ı) is equal to 0, then it can be shown that
2 =2// 2 2 2 1=2
value is a C z D .rı. 2
C Œ.a / C2r
2
D 2r2 . This is exactly the

az 
az x defined in Equation (60.6) by Leland (1994). However, we should
TC C V V
v.V I VB / D V C 1  ˛VB note that the introduction of the payout rate relaxes the unreasonable
r VB VB assumption assumed by Leland (1994) that net cash outflows associated
(60.18) with the choice of leverage such as coupons after-tax benefits must be
where a and z have been defined in Equation (60.16). financed by selling additional equity.
60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 941

holders need not protect themselves by demanding higher and its impact on capital structural decision under rational
coupon rates and thus equity holders will also benefit. Sec- expectation framework.
ond, for bond portfolio management, they find that Macaulay Structural models can demonstrate the violation of APR
duration overstates true duration of risky debt, especially for by incorporating different default boundaries. Leland (1994)
junk bonds that may have negative duration. Furthermore, has a discussion (in Sect. VI.C) regarding deviation of ab-
convexity can become concavity. solute priority, in which bondholders do not receive all of
the remaining asset value and stockholders also receive a
fraction of the remaining asset value. Nevertheless, it can-
60.4 Recent Development of Capital not explain the behavior of debt holders’ concession by
simply changing the boundary condition since their claims
Structure Models
will be reduced. Therefore, several recent models permit
the debt renegotiation by private workouts or by court-
In Sect. 60.3, we have reviewed in detail two of the most supervised debt renegotiation under Chap. 14 bankruptcy
well-known capital structure models – Leland (1994), and proceedings. For example, Fan and Sundaresan (2000) pro-
Leland and Toft (1996) models. These two models have pose a game-theoretic setting that incorporates equity’s abil-
been served as benchmark models in the optimal capi- ity to force concessions and varying bargaining powers to
tal structure literature. Their stationary debt structures and the debt holders and equity holders. Later on, Francois and
time-independent setting have been widely adopted in the lit- Morellec (2004) extend this bargaining game to explicitly
erature. However, compared with historical average, the op- model Chap. 14 protection and court-supervised renegotia-
timal leverage implied in the Leland (1994) model is unrea- tion. By introducing the possibility of renegotiating the debt
sonably high, while the model’s predicted yield spread is too contract, the default can occur at positive equity value. This
low given the suggested high leverage, especially for the un- is in contrast to the Leland’s (1994) model that the default
protected debt. Furthermore, unreasonably large bankruptcy occurs as the equity value reaches zero. As a consequence of
costs as high as 50% is required to lower the optimal leverage that, issuing new equity is costless, and the APR is respected.
to this level. Another branch of the extension is the dynamic capital
In addition to the issues above, several implications by structure. Early capital structure models assume that the debt
the Leland (1994) model are also not in accordance with the issuance is a static decision. However, in practice, firms ad-
empirical observations. In the Leland (1994) model, the firm just outstanding debt levels in response to changes in firm
will issue equity to finance contractual debt obligation until value. Early studies of dynamic capital structure include
equity value is driven to zero in the absence of reorganization works by Kane et al. (1984, 1985) and Fischer et al. (1989).
option. Hence, bankruptcy boundary coincides with the liqui- Recent works consist of the Goldstein et al. (2001) and Ju and
dation barrier. However, in practice, it is normal for manage- Ou-Yang models (2006), and many others. In general, the dy-
ment to file for bankruptcy when equity value is still positive. namic feature of these models predicts a lower leverage ratio
Some researchers argue that a more realistic bankruptcy pro- than those static capital structure models and is more in line
cedure may affect the valuation of corporate securities and with the observed leverage in practice.
capital structure decisions. Therefore, subsequent studies at- There are still other researchers who seek to explore the
tempt to extend Leland’s (1994) model to a more realistic capital structure problem of investment risks and agency
setting and hope to make the model-implied yield spread and costs, the timing of default under more realistic bankruptcy
leverage ratio to be more consistent with the observed aver- proceedings, alternative process, and other issues. We briefly
age in practice. categorize these studies as follows (the list below is by no
The first branch of these extensions is to permit devia- means exclusive):
tions of absolute priority rule (APR) in Chap. 14 bankruptcy Dynamic Capital Structure:12 Fischer et al. (1989);
proceedings. The importance of violations from APR has Goldstein et al. (2001); Ju et al. (2004); Ju and
been addressed in the literature. Eberhart and Weiss (1998) Ou-Yang (2006). Strategic Default and Renegotia-
clearly point out that APR is explicitly or implicitly assumed tion: Anderson and Sundaresan (1996); Mella-Barral
in many seminal finance models, such as Merton (1974) and and Perraudin (1997); Mella-Barral (1999); Fan and
Myers (1977). The APR is adhered since shareholders re- Sundaresan (2000); Francois and Morellec (2004). Risk
ceive nothing until debt holders have been fully paid. The Management for Firms: Ross (1996); Leland (1998).
APR is also assumed in most of the early structural mod-
els. However, empirical studies indicate that the APR viola-
12
tion is a common phenomenon (see Franks and Torous1989; Taurén (1999) and Collin-Dufresne and Goldstein (2001) approach
this issue by proposing alternative models to reflect the firms’ tendency
Eberhart et al. 1990; Weiss 1990). Therefore, it has of great to maintain a stationary leverage ratio. However, their studies focus on
interest to researchers to analyze the cause of APR deviation the risky corporate debt pricing and the implications of credit spreads.
942 S.-S. Chen et al.

Bankruptcy Procedures (Chapter 7 versus Chapter 11): The total value of the firm is
Broadie et al. (2007). Optimal Investment: Mauer and 8

Triantis (1994), Childs et al. (2005); Hackbarth (2006). ˆ T C
<V C C 
C TC C V
; when V > VQS
r C   r Q
Stochastic Interest Rate: Kim et al. (1993); Acharya and v.V / D
C VS
Carpenter (2002); Huang, Ju and Ou-Yang (2003); Ju and :̂V C  TC C V ; when V  VQS
Ou-Yang (2006). Jump: Dao and Jeanblanc (2006). C   r VQS
(60.22)
The Fan and Sundaresan (2000) model as well as the
rh i2
Goldstein et al. (2001) model are presented in detail in
rˇ rˇ
Sects. 60.4.1 and 60.4.2, respectively. We choose these two where  D 0:5  2
˙ 0:5  2
C 2r
2
.
papers that, in our opinion, are the most important or repre- For any V  VQS , Fan and Sundaresan set the optimal
sentative of that stream of research. The remaining research sharing rule between shareholders and debtholders as
will be presented in less detail and we will briefly outline
their key assumptions and results. Readers may refer to the E.V Q
Q / D v.V / and D.V Q
Q / D .1  /v.V /: (60.23)
original papers for details.
The Nash solution to the bargaining game can be character-
ized as
60.4.1 The Fan and Sundaresan (2000) Model
n o
Q
Q  D arg max v.V /0
Fan and Sundaresan (2000) propose a game-theoretic set- n o1
ting that incorporates equity’s ability to force concessions Q

.1  /v.V /  max Œ.1  ˛/V  K; 0
and various bargaining powers to the debt and equity hold-  
.1  ˛/V  K
ers. Two cases are discussed in their study where the firm’s D min    ; (60.24)
asset value falls below a certain boundary: Debt-equity swap v.V /
and strategic debt service. In the case of debt-equity swap,
where  denotes bargaining power of shareholders.
the debtholders exchange their claims for equity. In strategic
At equilibrium, Fan and Sundaresan (2000) show that debt
debt service, borrowers stop making the contractual coupon
holders accept less than the contractual coupon and still per-
and start servicing debt strategically until the firm’s assets
mit shareholders to run the company. This results in devia-
returns to its original level.
tions from APR, in which the shareholders will get a bigger
Some key assumptions under their framework include the
share and the debtholders will get a less proportion of the
following: (1) During the default period, the tax benefits are
firm. But both parties will be better off. Therefore, the con-
lost. (2) Asset sales for dividend payments are prohibited.
cession of debtholders is well explained under the Fan and
(3) The firm can be liquidated only at a cost. The propor-
Sundaresan’s framework.
tional cost is ˛.0  ˛  1/ and the fixed cost is K.K 0/.
They then solve the trigger point for strategic debt service
Note that debt holders have strict absolute priority upon
(K D 0 case) as
liquidation.
(4) The asset value of the firm follows the lognormal c.1  TC C TC /  1
diffusion process VQ D (60.25)
r 1   1  ˛
dV D .  ˇ/V dt C VdBt (60.21)
and the strategic service amount paid to debt holders S.V / D
where  is the instantaneous expected rate of return on the .1  ˛/ˇV .
firm gross of all payout, and ˇ is the firm’s cash payout ratio. It can be shown that the equity value satisfies the follow-
Here we present only the results of strategic debt service. ing differential equations:13
If the equity and debt holders can reach an agreement of tem-
porary coupon reduction when the firm is under financial dis- 1 2 2Q
 V EVV .V; t/ C .r  ˇ/V EQ V .V; t/  r E.V;
Q t/
tress, the firm will not lose its potential future tax benefits. At 2
the trigger point for the strategic service VQS , both parties will CˇV  c.1  TC / D 0; when V > VQS I
bargain the total value of the firm v.V /. 1 2 2Q
The driving force behind strategic behavior is the presence  V EVV .V; t/ C .r  ˇ/V EQ V .V; t/  r E.V;
Q t/
2
of proportional and fixed costs of liquidation. The bargain- CˇV  S.V / D 0; when V > VQS I (60.26)
ing power of equity holder clearly depends on the liquidation
cost ˛VS C K. Fan and Sundaresan (2000) endogenize both
the reorganization boundary and the optimal sharing rule be-
tween equity and debt holders upon default. 13
See Dixit and Pindyck (1994).
60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 943

with the boundary conditions: Note that when V > VQS , the contractual coupon is paid and
the tax shield is in effect. By contrast, when V  VQS , the
c.1  TC / tax shield is lost because the equity holders are strategically
lim E.V / D V 
V "1 r servicing the debt according to the cash flow generated by

the firm. Both the strategic debt service amount S.V / and
 cT C
lim E.V / D  ˛ VQS  the trigger level VQS are determined endogenously.
V #VQS C   r

The equity value is
C  cT C
lim EV .V / D  ˛ 
V #VQS C   r VQS

8  

< .1  TC /C .1  TC /C  .1  C / TC C V
Q V  C  ; when V > VQS
E.V / D r .1   /r .C   /.1   / r VS (60.27)
:
v.V /  .1  ˛/V; when V  VQS

The Fan and Sundaresan model shows that debt renegotiation


ously. In contrast to the unlevered firm value, the claim to
encourages early default and increases credit spreads on cor-
future EBIT flow should be more reasonable to changes in
porate debt, given that shareholders can renegotiate in dis-
capital structure. Furthermore, compared with previous mod-
tress to avoid inefficient and costly liquidation. It might be
els, this characterization also has assumptions that are more
the interest of debt holders to forgive part of the debt service
in accordance with empirical findings.15
payments if they can avoid the wasteful liquidations, which
The single technology of a pure exchange economy pro-
can be shared by the two claimants. If shareholders have no
duces a payout flow that is specified by the (EBIT) process
bargaining power, no strategic debt service takes place and
the model converges to the Leland model. Furthermore, by dı
introducing the possibility of renegotiating the debt contract, D P dt C dz (60.28)
ı
the default can occur at positive equity value. This is in con-
trast to the Leland’s (1994) model in that the default occurs The value to the claim of entire period is the discounting ex-
when the equity value reaches zero as a consequence of that pected cash flows (unlevered firm)
issuing new equity is costless and the APR is respected. Z

1
Q ıt
V .t/ D Et dsıs e rs D (60.29)
t r 

60.4.2 The Goldstein et al. (2001) Model where  D .P  / is the risk-neutral drift of the payout
flow rate:

Goldstein, Ju, and Leland propose an EBIT-based model in D dt C dzQ (60.30)
that the state variable is earnings before interests and taxes ı
(EBIT) instead of commonly used unlevered firm value.14 Since r and  are constants, V and ı share the same
This approach is intuitively appealing: the EBIT-generating dynamics:
dV
machine runs independently of how the EBIT flow is dis- D dt C dzQ (60.31)
tributed among its claimants. The advantage of taking the V
claim on future EBIT is that all contingent claimants, includ-
ing the government claim, to future EBIT flows are treated in
a consistent fashion. More precisely, the government claim is 15
Fischer et al. (1989) assume that the following: (i) The value of an op-
no longer treated as a cash inflow in a form of tax benefit, but timally levered firm can only exceed its unlevered value by the amount
as a cash outflow on EBIT in a form of tax. In other words, of transaction costs incurred. It eliminates the arbitrage opportunity.
(ii) A firm that follows an optimal financing policy offers a fare risk-
each dollar paid out, whether to interest payments, dividends, adjusted return. Thus, unlevered firms offer a below-fair expected rate
or taxes, affects the firm in the same way. Another advantage of return. Goldstein et al. (2001) argue that these key assumptions by
is that unlevered and levered equity never exist simultane- Fischer et al. (1989) are not realistic. First, the arbitrage strategy is not
feasible because a bidder must pay a large premium in order to gain
control of management. They also argue that for publicly traded assets,
14 rational investors will push down security prices so that fair expected re-
Although unlevered firm value may not be traded, trading of equity
or other contingent claims allows the use of unlevered firm value as the turn is obtained on any traded assets regardless of the policies employed
state variable (Ericsson and Reneby 2004a). by the management.
944 S.-S. Chen et al.

Accordingly, with boundary conditions

dV C dı lim Vsolv D V and Vsolv D 0 as V D VB


D rdt C dzQ : (60.32) V !1
V
Thus,
60.4.2.1 Optimal Static Capital Structure Vsolv D V  VB pB .V / (60.36)
(1.3) Define Vint as the value of the claim to interest pay-
Assume that a firm issues a consol bond with a constant ments.
coupon payment C as long as the firm remains solvent. As in
Leland (1994), any claim must satisfy the ordinary differen- C
Vint D C A1 V y C A2 V x
tial equation r

2 2 Since lim Dc D C
r , A1 D 0. This claim vanishes at V D
VF V C V FVV C P  rF D 0 (60.33) V !1
VB , one can obtain
2

where P is the payout flow. C


Vint D Œ1  pB .V / (60.37)
The solution to the above equation can be written as r
the sum of a particular solution to the full inhomoge-
neous equation and the general solution to the correspond- (1.4) Finally, separating the value of continuing operations
ing homogeneous equation. The general solution of VFV C among equity, debt, and government, we have
2 2
2
V FVV  rF D 0 (no intertemporal cash flows):
Esolv .V / D .1  Teff /.Vsolv  Vint / (60.38)
F D A1 V y
C A2 V x
(60.34)
Dsolv .V / D .1  Ti /Vint (60.39)
" r #,
2
where x D  2
C  2
C 2r 2 2 Gsolv .V / D Teff .Vsolv  Vint / C Ti Vint (60.40)
2 2

2 s 3,


where Ti is the personal tax rate of interest payments, Td is
4 2 2 2
yD    C 2r 2 5  2 the effective rate for dividends, TC is the tax rate for corpo-
2 2 rate profits, and the effective rate Teff is

Note that x is positive and y is negative. Since A1 V y goes .1  Teff / D .1  Tc /.1  Td /


to infinity while V becomes large, A1 equals zero for all
claims of interest. (1.5) Define Vdef .V / as the present value of the default
claim.
(1.1) Define pB .V / as the present value of a claim that pays
$1 contingent on firm value reaching VB . Vdef .V / D A1 V y C A2 V x

pB .V / D A1 V y C A2 V x with boundary conditions

with boundary conditions lim Vdef .V / D 0 and Vdef .V D VB / D VB


V !1

lim pB .V / D 0 and lim pB .V / D 1 Thus,


V !1 V !1
Vdef .V / D VB pB .V / (60.41)
Thus,
x Note that Vsolv .V / C Vdef .V / D V , the present value of the
V
pB .V / D (60.35) claim to EBIT. Therefore, value is neither created nor de-
VB
stroyed by changes in the capital structure.
(1.2) Define Vsolv .V / a claim entitled to the entire payout ı Similar to the claims to funds during solvency, the default
as long as the firm remains solvent; that is, firm value claim can be divided among bankruptcy costs, debt, and gov-
remains above VB . ernment, and:

Vsolv D V C A1 V y C A2 V x BCdef .V / D ˛Vdef .V / (60.42)


60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 945

Ddef .V / D .1  ˛/.1  Teff /Vdef .V / (60.43) then use backward induction scheme to show that it is opti-
mal for each period’s coupon payments, restructuring thresh-
Gdef .V / D .1  ˛/Teff Vdef .V / (60.44)
old, and bankruptcy threshold to increase by the same factor
Note that Goldstein et al. (2001) assume that the portion of  D VU =V0 that the initial EBIT value increases by each pe-
the EBIT-generating machine not lost to bankruptcy cost can riod. All claims will also scale up by the same factor. Because
be sold to competitors in the event of bankruptcy. Thus, the of this scaling feature, the firm will be identical at every fu-
government can have a claim in bankruptcy due to the taxes ture restructuring date, except that all factors will be scaled
on future EBIT from competitors. up by  .
(1.6) The restructuring costs are assumed to be proportional (2.1) Define PU .V / as the present value of a claim that pays
to the initial value of the debt issuance: $1 contingent on firm value reaching VU . This claim
receives no dividend and therefore has the form
RC.V0 / D qŒDsolv .V0 / C Ddef .V0 / (60.45)
PU .V / D A1 V y C A2 V x
The restructuring costs are deducted from the proceeds of the
debt issuance before distribution to equity holders. with boundary conditions
ˇ
(1.7) Solving the smooth-pasting condition @E ˇ D 0,
@V V DVB pU .VB / D 0 and lim pB .V / D 0
one can obtain 
V !1
C
VB D  (60.46)
r Thus,
 
where  D x
. y
xC1
VBx y V
Note that while most comparative statics are similar to PU .V / D  V C B V x (60.47)
those in Leland (1994), equity is a decreasing function of ef- † †
fective tax rate. This is because an increase in Teff increases y y
where †  VB VUx  VU VBx .
the government claim at the expense of equity. In contrast,
previous capital structure models use unlevered firm value as (2.2) Similarly, define pB .V / as the present value of a claim
state variable and treat government claim as cash inflow. As that pays $1 contingent on firm value reaching VB .
a result, the tax benefits increase with tax rate and in turn in-
y
crease equity value. However, empirical evidence indicates VUx y V
pB .V / D  V C U V x (60.48)
that equity prices are a decreasing function of the tax rate. † †
y y
Under the characterization of Goldstein et al. (2001), even where †  VB VUx  VU VBx .
though the tax benefit is an increasing function of TC , equity
(2.3) From (2.1) and (2.2), all period 0 claims can be written
is a decreasing function of TC . This feature solves the prob-
in terms of pU .V / and pB .V /.
lem of the previous capital structure models and is supported
by empirical studies.
0
Vsolv .V / D V  pU .V /VU  pB .V /VB (60.49)
(current firm value remains between VU and VB )
60.4.2.2 Optimal Upward Dynamic Capital
Structure Strategy16 C0
0
Vint .V / D Œ1  pU .V /  pB .V / (60.50)
r
Goldstein et al. (2001) assume that there will be a threshold
VB , where the firm will optimally choose bankruptcy, and 0
Vdef .V / D pB .V /VB (60.51)
there be a threshold VU , where management will call the out-
standing debt issue at par, and sell a larger new issue.17 They 0
Vres .V / D pU .V /VU (60.52)

16
They mention that downward restructuring can be incorporated into
0
Note that the sum Vsolv .V / C Vdef
0
.V / C Vres
0
.V / D V .
their model. However, they argue that transaction costs can discourage
(2.4) Assuming full-loss offset, the values of period 0 claims
debt reduction outside Chap. 14 (see Gilson 1997). In addition, due to
the neglected issues such as asset substitution, asymmetric information, after the restructuring:
equity’s ability to force concessions, and Chap. 14 protection, they are
skeptical about the results for the downward restructuring.
17 d 0 .V / D .1  Ti /Vint
0
.V / C .1  ˛/.1  Teff /Vdef
0
.V /
Goldstein et al. (2001) claim that given management’s strategy, the
value of the debt issue would be the same even if it were not callable, (60.53)
as long as all debt issues receives the same recovery rate in the event of
bankruptcy. e 0 .V / D .1  Teff /ŒVsolv
0
.V /  Vint
0
.V / (60.54)
946 S.-S. Chen et al.

g 0 .V / D Teff ŒVsolv
0
.V /  Vint
0
.V / C Ti Vint
0 The total equity claim, E(V), just before the initial debt is-
suance, is the sum of the intertemporal claims of debt and
C.1  ˛/Teff Vdef
0
.V / (60.55) equity, less the claim of restructuring costs:

bc0 .V / D ˛Vdef
0
.V / (60.56) e 0 .V0 / C d 0 .V0 /  qD 0 .V0 /
E.V0 / D (60.62)
(2.5) Define e.V / as the equity claim to intertemporal EBIT 1  PU .V0 /
flows for all periods. Note that e.V / is not the total eq-
And after the initial debt issuance during period 0,
uity claim. Since the claims to future debt issues are
currently owned by equity and restructuring costs, fu-
E.V / D PU .V /E.V0 / C e 0 .V /  PU .V /D 0 .V0 /
ture debt holders will have to pay fair price for these
(60.63)
claims in the future. As a result, there are also cash
This reduces to
transfers between the claimants at each restructuring
date.
E.V0C / D E.V0 /  .1  q/D 0 .V0 / (60.64)
The scaling feature inherent in the model implies
It also reduces to
e.V0 / d.V0 / g.V0 /
0
D 0 D 0
e .V0 / d .V0 / g .V0 / E.VU  / D E.V0 /  D 0 .V0 / (60.65)
V0 1
D D
V0  VU .V0 /PU .V0 /
0 1  PU .V0 / just before the next issuance.
Finally, managers choose C ,   VU =V0 ,  VB =V0
(60.57)
in order to maximize the wealth of equity holder, subject to
limited liability.
The equity claim to intertemporal cash flows is
In the Goldstein, Ju, and Leland model, the optimal initial
leverage ratio is substantially lower than that obtained in the
e 0 .V0 /
e.V0 / D static model and is more in line with the historical average.
1  PU .V0 /
˚  This is a result of that the firm has the option to increase its
D e 0 .V0 / 1 C PU .V0 / C ŒPU .V0 / 2 C : : : leverage in the future. Thus, it will wait until it becomes op-
8 9 timal to exercise it. The possibility of increasing leverage in
<X 1 =
the future decreases the initial optimal leverage. In addition,
D e 0 .V0 / ŒPU .V0 / j (60.58)
: ; the bankruptcy triggered level VB also drops significantly
j D0
since a firm that has an option to adjust its capital structure is
(2.6) Next, they determine the portion of the future debt more valuable than the one without this option. Moreover, the
claims that currently belong to equity and restructuring model also implies reasonable and lower bankruptcy cost for
costs. The management will call the outstanding debt the optimal leverage. The tax benefits to debt also increase
issue at par at restructuring date. Therefore, the initial significantly over the prediction of static models.
value of the debt claim equals the sum of the claim to
period 0 cash flows plus the present value of the call
value 60.4.3 Other Important Extensions
D 0 .V0 / D d 0 .V0 / C pU .V0 /D 0 .V0 / (60.59)
In this section, we will briefly present some other exten-
Hence, sions with the introduction of investment risk, alternative
d 0 .V0 / bankruptcy procedures, as well as alternative stochastic pro-
D 0 .V0 / D (60.60)
1  pU .V0 / cesses, and will outline key results of these models.
D 0 .V0 / is the fair value debt holders must pay at the time of
issuance. The proceeds are then divided between equity and
restructuring cost. By the scaling argument, the total restruc- 60.4.3.1 Investment Risk – Leland (1998)
turing claim is
One possible agency problem, asset substitution, has been
˚ 
RC.V0 / D qD .V0 / 1 C PU .V0 / C ŒPU .V0 / 2 C : : :
0 discussed in Leland (1994) and Leland and Toft (1996).
Leland (1998) explicitly models this issue by allowing the
qD 0 .V0 / firm to choose its risk strategy and examines the interaction
D (60.61)
1  PU .V0 / between capital structure and risk choice. To capture the
60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 947

agency cost, Leland (1998) assumes that risk choices are difference from Fan and Sundaresan’s model is that Francois
made ex post; that is, after debt is in place. The risk choices and Morellec assume that a proportional costs ' are borne by
cannot be precontracted in the debt covenants or credibly the company during the renegotiation process. The costs of
precommitted. financial distress incurred by the firm filing Chap. 14 are ig-
The environment with ex post risk choice can be con- nored in the prior research. Even though the costs implied by
trast with the hypothetical ex ante, in which the firm simul- private workouts are generally low, Chap. 14 filings are as-
taneously chooses its risk strategy and its debt structure to sociated with large costs of financial distress that may affect
maximize initial firm value. The measure of agency costs is shareholder’s default decision. Note that Leland (1994) only
therefore the difference in maximal values between the ex allows liquidation while Fan and Sundaresan (2000) only
ante and ex post cases. This reflects the loss in value fol- permit private workouts.
lows from the risk strategies chosen to maximize equity value Francois and Morellec solve the endogenous default bar-
rather than firm value. rier by maximizing equity value and providing closed-form
In this paper, optimal capital structure reflects both the solutions for corporate debt and equity values. They find that
tax advantages less default costs and the agency costs result- the possibility to renegotiate the debt contract has ambigu-
ing from asset substitution (Jensen and Meckling 1976). The ous impact on leverage choices and increases credit spreads
joint determination of capital structure and investment risk on corporate debt. The sharing rule of cash flows during
is examined. Leland (1998) finds that agency costs restrict bankruptcy has a large impact on optimal leverage, while
leverage and debt maturity and increase yield spreads. credit spread on corporate debt shows little sensitivity to the
varying bargaining power.

60.4.3.2 Chapter 11 Protection – Francois


and Morellec (2004) 60.4.3.3 Chapter 11 vs. Chapter 7 – Broadie et al.
(2007)
Francois and Morellec (2004) extend the Fan and
Sundaresan (2000) model to incorporate the possibility of Broadie et al. (2007) propose framework in which optimal
Chap. 14 filings. Under their setup, shareholders hold a debt and equity values are determined in the presence of
Parisian down-and-out option on the firm’s asset; that is, Chaps. 10 and 14 under the U.S. bankruptcy code. They ex-
shareholders have a residual claim on the cash flows gen- plicitly consider two separate barriers for default and liqui-
erated by the firm unless the value of these assets reaches dation for the more realistic bankruptcy procedures. They
the default threshold and remains below that threshold for model the key feature of bankruptcy proceedings by auto-
the exclusive period. It is generally acknowledged that there matic stay provision – suspending the dividends while en-
are two types of defaulting firms: First, firms that are eco- tering bankruptcy (Chap. 14) and the imposition of APR
nomically sound but default only due to temporary financial (Chap. 10) upon liquidation. To facilitate the analysis, they
distress, and recover under Chap. 14. Second, firms that are follow the EBIT process by Goldstein et al. (2001)19 and as-
economically unsound, keep on losing value under Chap. 14, sume the firm issues a single consol bond.
and are eventually liquidated under Chap. 10. The modeling They model the critical characteristics of the U.S.
philosophy comes from the empirical studies that show that bankruptcy proceeding – grace period, automatic stay, debt
most firms emerge from Chap. 14. Only a few are eventually forgiveness, absolute priority, and transfer of control rights
liquidated under Chap. 10 after filing Chap. 14. from equity to debt holders. Some prior models, for exam-
Following the Nash bargaining game of Fan and Sun- ple, such as the Francois and Morellec (2004) model, also
daresan’s approach,18 Francois and Morellec presume the make a distinction between default and liquidation. However,
firm renegotiates its debt obligations whenever the asset Broadie et al. (2007) argue that in their model the firm can-
value falls below a constant threshold VB . However, a major not be liquidated even if the unlevered firm value is too low
during the grace period. As a result, limited liability could be
violated. Therefore, under the setup of Broadie, Chernov, and
18
Note that the specification for the bargaining game within Francois Sundaresan, liquidation could happen as the firm value either
and Morellec’s framework is slightly different from that of Fan and
Sundaresan (2000). Francois and Morellec focus on Chapter 11 filings
– court-supervised debt renegotiation in contrast to the private workouts
19
by Fan and Sundaresan. Therefore, the automatic stay of assets prevents Note that they assume the same process for EBIT before and after de-
shareholders from liquidating the firm’s assets. Nevertheless, renegoti- fault. Therefore, their framework models financial rather than economi-
ation plan requires each participant to receive a payoff that exceeds the cal distress since bankruptcy by itself does not cause poor performance.
liquidating value of its claims. As a result, bondholders’ payoff has to The work by Goldstein et al. (2001) does not need this assumption since
exceed .1  ˛/VB and this results in the difference of the specification there is no difference between default (bankruptcy) and liquidation un-
for the bargaining game. der their setup.
948 S.-S. Chen et al.

reaches the liquidation barrier or stays under the bankruptcy In their model, the tradeoff between the gains from dy-
barrier for longer than the grace period. namically adjusting the debt amount and the restructuring
In addition, they explicitly model the automatic stay pro- costs if adjusting it yield the optimal debt maturity. In other
vision by recording the accumulated unpaid coupons plus in- words, the tax rate and the transaction costs are the most two
terest in arrears, denoted as At . When a firm recovers from important parameters in determining the optimal maturity.
Chap. 14, the debt holders will forgive a fraction 1  , They find that a higher transaction cost yields a longer debt
0    1 of arrears and receive an amount At . More- maturity since it is more expensive to rebalance a firm’s cap-
over, the firm does not pay dividends to shareholders while ital structure. On the other hand, a higher tax rate yields a
it is in the default state. The entire EBIT while in default is shorter optimal debt maturity because it is more valuable to
accumulated in a separated account St . If St is not sufficient recapitalize the capital structure.
to repay the arrears, equity holders must raise the remain- They also find that the initial level and long-run mean
ing at the cost of diluting equity. If St > At , the leftover is of the interest rate process are key variables in determining
distributed to shareholders.20 both optimal capital structure and optimal maturity structure.
In this paper, they focus on the issues of bankruptcy pro- When the interest rate is higher (lower), coupon and principal
ceedings and the optimal choice of these two boundaries are higher (lower). The coupon and principal are determined
driven by different objectives.21 They show that the first-best in such a way that the market value of debt is independent
outcome, the total firm value maximization ex-ante upon fil- of the spot interest rate level. This is in sharp contrast to the
ing Chap. 14, is different from the equity value maximization fact that the level of spot interest rate is a key parameter in
outcome. They also show that the first-best outcome can be the constant interest rate model. Finally, the volatility of the
restored in large measure by giving creditors either the con- interest rate process and the correlation between the interest
trol to declare Chap. 14 or the right to liquidate the firm once rate process and the firm asset value process play important
it is taken to Chap. 14 by the equity holders. This serves as roles in determining the debt maturity structure.
the threat from debtholders to prevent equity holders from fil-
ing for Chap. 14 too soon to get debt relief. Finally, they also
find that on average the firms are more likely to default and
are less likely to liquidate relative to the benchmark model of
60.5 Application and Empirical Evidence
Leland (1994). of Capital Structure Models

Capital structure models are intended to quantitatively exam-


60.4.3.4 Optimal Debt Maturity – Ju ine the capital structure theories. In capital structure models,
and Ou-Yang (2006) the corporate debt value and capital structure are interlinked
variables. Therefore, the valuation of corporate securities and
In the work by Leland and Toft (1996), debt maturity is financial decisions are jointly determined. Through the struc-
shown to be crucial to the leverage ratio and credit spreads. tural approach, researchers can conduct a detailed analysis of
In practice, the optimal capital structure and optimal maturity the behavior of bond prices and optimal leverage ratios as fi-
structure are interdependent decisions. However, in Leland nancial variables (such as taxes, interest rates, payout rates
and Toft (1996), debt maturity is exogenously specified. In or bankruptcy costs) change.
addition, Goldstein et al. (2001) show that the optimal capi- While theoretically elegant, capital structure models must
tal structure is very sensitive to the input level of the interest be able to explain observed capital structure and bond prices
rate. Therefore, Ju and Ou-Yang (2006) propose a dynamic in practice. As a result, the most commonly used measures
model in which the optimal structure and an optimal debt for a capital structure model are the predicted leverage ra-
maturity are jointly determined in a Vasicek (1977) mean- tio and yield spreads. Most of the capital structure mod-
reverting interest rate process.22 els illustrate the performance by numerical examples us-
ing the historical average as model input parameters. In
20 general, the early capital structure models predicted opti-
Note that in order to accommodate the complex feature of bankruptcy
proceedings, numerical approach is necessary to solve the model. The mal leverage rates that were too high. The recent models
detailed binomial lattice methodology is developed by Broadie and such as those incorporating dynamic capital structure set-
Kaya (2007). tings have predicted optimal leverage more in line with the
21
To determine if a distressed firm would choose the reorganization
option under nearly perfect market assumption, they first ignore taxes
to emphasize the impact of bankruptcy and return to this issue later in capital structure is different from that of Goldstein et al. (2001). In con-
their paper. trast to the recapitalization barrier assumed in Goldstein et al. (2001), Ju
22
We should note that the default boundary is exogenously specified in and Ou-Yang (2006) assumes that the firm rebalances its capital struc-
their model for tractability. The design of recapitalization of the firm’s ture periodically as long as the firm is solvent.
60 Alternative Methods to Determine Optimal Capital Structure: Theory and Application 949

observed average under reasonable bankruptcy cost. Never- By adopting a different estimation method, Ericsson
theless, capital structure models (as well as corporate bond et al. (2006) use the MLE approach proposed by Duan (1994)
pricing models using structural approach) still predict credit to perform an empirical test on both CDS spreads and bond
spreads too low, especially for the short-term debt. spreads, including three structural models – Leland (1994),
There are only a few empirical studies on the models em- Leland and Toft (1996), and Fan and Sundaresan (2000).
ploying structural approach. Among them, a large portion of In contrast to previous evidence from corporate bond data,
the empirical tests of structural models focus on the risky CDS premia are not systematically underestimated. Also, as
debt pricing performance of the bond pricing models, such as expected, bond spreads are systematically underestimated,
Longstaff and Schwartz (1995). The empirical performance which is consistent with the fact they are driven by significant
in pricing risky debt is generally unsatisfactory (see Wei and non-default factors. In addition, they also conduct regression
Guo 1997; Anderson and Sundaresan 2000; Delianedis and analysis on residuals against default and non-default proxies.
Geske 2001; Huang and Huang 2003, among others.). The Little evidence is found for any default risk components in ei-
empirical investigations for capital structure models are even ther CDS or bond residuals, while strong evidence, in partic-
rare due to their complexity. The most well-known study in- ular, an illiquidity premium, is related to the bond residuals.
cluding capital structure models is by Eom et al. (2004). They conclude that structural models are able to capture the
Eom et al. (2004) carry out an empirical analysis of five credit risk price in the markets but they fail to price corporate
structure models including Merton (1974), Geske (1977), bonds adequately due to omitted risks.
Longstaff and Schwartz (1995), Leland and Toft (1996), and Also using the MLE approach, Ericsson and Reneby
Collin-Dufresne and Goldstein (2001). They test these mod- (2004b) estimate yield spreads between 1 and 50 months
els with bond data of firms with simple capital structure on out of sample by an extended version of the Leland (1994)
the last trading day of each December from 1986 to 1997. model, which allows for violation of the absolute priority rule
They calibrate these models using the book value of total and future debt issues. In addition to the stock price series,
liabilities in the balance sheet as the default boundary, and they also include the bond price and dividend information in
calculate the corresponding asset value as the sum of the mar- estimating their model. The bond samples consist of 141 U.S.
ket value of equity and the book value of total debt. Then corporate issues and a total of 5,594 dealer quotes. Their em-
they estimate the asset return volatility using bond-implied pirical results show that, for the 1 month-ahead prediction, a
volatility as well as six equity return volatilities measured by mean error is merely 2 basis points. This is similar to the
different time horizons before and after the bond price ob- fitting error of those reduced-form models in Duffee (1999).
servation. In contrast to previous studies that have suggested Therefore, they conclude that the inferior performance of
that structural models generally predict yield spreads too low, structural models may result from the estimation approaches
the result of Eom et al. (2004) shows more complicated phe- used in the existing empirical studies.
nomena, although all of these models make significant er- Finally, another potentially important application of cap-
rors in predicting the credit spread. The Merton (1974) and ital structure model is to predict the credit quality of a
Geske (1977) models underestimate the spreads, while the corporate security. For example, Leland (2004) examines
Longstaff and Schwartz (1995), Leland and Toft (1996), and the default probabilities predicted by the Longstaff and
Collin-Dufresne and Goldstein (2001) models all overesti- Schwartz (1995) model with the exogenous default bound-
mate the spreads. ary, and the Leland and Toft (1996) model with endogenous
Huang and Huang (2003) use several structural mod- default boundary. As Leland stated
els to predict yield spread, including the Longstaff and We focus on default probabilities rather than credit spreads be-
Schwartz (1995) model, the strategic model, the endogenous- cause (i) they are not affected by additional factors such as liq-
default model, the stationary leverage model as well as two uidity, tax differences, and recovery rates; and (ii) prediction of
models they propose: one with a time-varying asset risk pre- the relative likelihood of default is often stated as the objective
of bond ratings.
mium and one with a double exponential jump-diffusion firm
value process. They calibrate inputs, including asset volatil- Leland uses Moody’s corporate bond default data from
ity, for each model so that target variables, including lever- 1970 to 2000 in his study and follows a similar calibra-
age, equity premium, recovery rate, and the cumulative de- tion approach by Huang and Huang (2003). Rather than
fault probability at a single time horizon, are matched. They matching the observed default frequencies, Leland instead
show that the models make quite similar predictions on yield chooses common inputs across models to observe how well
spreads. In addition, the observed yield spreads relative to they match observed default statistics. The empirical re-
Treasury bonds are considerably greater than the predicted sults show that when costs and recovery rates are matched,
spreads, especially for highly rated debt. Hence, they con- the exogenous and the endogenous default boundary mod-
clude that additional factors such as liquidity and taxes must els fit observed default frequencies equally. The models pre-
be important in explaining market yield spreads. dict longer-term default frequencies quite accurately, while
950 S.-S. Chen et al.

shorter-term default frequencies tend to be underestimated. the estimation problem in models employing structural ap-
Thus, he suggests that a jump component should be included proach (see Bruche 2005) and the reference therein). Thus,
in asset value dynamics.23 we hope to see more rigorous empirical studies for capital
In summary, the empirical performance of capital struc- structure models in the future.
ture models in pricing risky debt is not satisfactory. Even
with the incorporation of jump process, the illiquid corporate
bond market still hinders structural models from accurately
pricing risky debt. However, predicting the credit quality of
References
a corporate security could be a good application of capital Acharya, V. and J. Carpenter. 2002. “Corporate bond valuation and
structure models because they are less affected by the mi- hedging with stochastic interest rates and endogenous bankruptcy.”
cro structure issues. Since recent capital structure models put Review of Financial Studies 15, 1355–1383.
Anderson, R. and S. Sundaresan. 1996. “Design and valuation of debt
numerous efforts on the event of bankruptcy, prediction of
contract.” Review of Financial Studies 9, 37–68.
default probabilities or default events shall be potentially im- Anderson, R. and S. Sundaresan. 2000. “A comparative study of struc-
portant applications.24 tural models of corporate bond yields: an exploratory investigation.”
Journal of Banking and Finance 24, 255–269.
Black, F. and J. C. Cox. 1976. “Valuing corporate securities: some
effects of bond indenture provisions.” Journal of Finance 31,
60.6 Conclusion 351–367.
Black, F. and M. Scholes. 1973. “The pricing of options and corporate
liabilities.” Journal of Political Economy 81, 637–654.
In this paper, we review the most important and represen- Brennan, M. and E. Schwartz. 1978. “Corporate income taxes, val-
uation, and the problem of optimal capital structure.” Journal of
tative capital structure models. The capital structure mod-
Business 54, 103–114.
els incorporate contingent claim valuation theory to quan- Briys, E. and F. de Varenne. 1997. “Valuing risky fixed rate debt:
titatively analyze prevailing determinants of capital structure an extension.” Journal of Financial and Quantitative Analysis 32,
in corporate finance literature. In capital structure models, 239–248.
the valuation of corporate securities and financial decisions Broadie, M. and O. Kaya. 2007. “A binomial lattice method for pricing
corporate debt and modeling chapter 11 proceedings.” Journal of
are jointly determined. Most of the capital structure models Financial and Quantitative Analysis 42, 279–312.
provide closed-form expressions of corporate debt as well as Broadie, M., M. Chernov, and S. Sundaresan. 2007. “Optimal debt and
the endogenously determined bankruptcy level, which are ex- equity values in the presence of chapter 10 and chapter 14.” Journal
plicitly linked to taxes, firm risk, bankruptcy costs, risk-free of Finance 62, 1341–1377.
Bruche, M. 2005. Estimating structural bond pricing models via sim-
interest rate, payout rates, and other important variables. The ulated maximum likelihood, Working paper, London School of
behavior of how debt values (and therefore yield spreads) and Economics.
optimal leverage ratios change with these variables can thus Chen, R., S. Hu, and G. Pan. 2006. Default prediction of various
be investigated in detail. structural models, Working paper, Rutgers University, National Tai-
wan University, and National Ping-Tung University of Sciences and
While theoretically elegant, capital structure models do
Technologies.
not perform well empirically in risky corporate bond pricing. Chen, R., C. Lee, and H. Lee. 2008. Default prediction of alternative
Researchers have been attempting to resolve the yield spread structural credit risk models and implications of default barriers,
underestimates by introducing jumps and liquidity premium. Working paper, Rutgers University.
On the other hand, since recent capital structure models put Childs, P., D. Mauer and S. Ott. 2005. “Interaction of corporate financ-
ing and investment Decisions: the effects of agency conflict.” Jour-
numerous efforts on the event of bankruptcy, we suggest nal of Financial Economics 76, 667–690.
that prediction of default probabilities or default events shall Collin-Dufresne, P. and R. S. Goldstein. 2001. “Do credit spreads reflect
be potentially important applications. Furthermore, some re- stationary leverage ratios?” Journal of Finance 56, 1929–1957.
searchers argue that the past poor performance of capital Dao, B. and M. Jeanblanc. 2006. Double exponential jump diffusion
process: a structural model of endogenous default barrier with roll-
structure models may come from the estimation approaches over debt structure, Working paper, The Université Paris Dauphine
historically used in the empirical studies (Ericsson and and the Université d’Évry.
Reneby 2005, Ericsson, Reneby, and Wang 2006). And we Delianedis, G. and R. Geske. 2001. The components of corporate credit
have seen some innovative estimation methods for solving spreads: default, recovery, tax, jumps, liquidity, and market factors,
Working paper, UCLA.
Duan, J. C. 1994. “Maximum likelihood estimation using pricing data
of the derivative contract.” Mathematical Finance 4, 155–167.
23
We should note that Zhou (2001) show that jumps can explain why Duffee, G. 1999. “Estimating the price of default risks.” Review of Fi-
yield spreads remain strictly positive even as maturity approaches zero. nancial Studies 12, 197–226.
24
In default prediction, some researchers have conducted empirical in- Durand, D. 1952. “Costs of debt and equity funds for business trends
vestigation on a company basis analysis such as Chen et al. (2006) and and problems of measurement,” Conference on Research in Busi-
Chen et al. (2008), though the models they tested are mainly bond pric- ness Finance, National Bureau of Economic Research, New York,
ing models. pp. 215–247.

You might also like