WACC and APV revisited

Lutz Kruschwitz and Andreas Loer

Version: April 1998

Abstract
In this paper we carefully analyze the assumptions of the WACC and
the APV approach. Especially, we show that under mild conditions
both methods do not yield the same value of the rm.

1 Introduction
Literature on rm valuation continues to grow. To take account of tax
eects, several methods have been discussed. Here we will restrict our con-
siderations to the WACC and the APV approach.
The APV approach essentially goes back to the fundamental result of
Modigliani & Miller (1958). Modigliani and Miller proved that the
value of a project's levered cash ow stream can be expressed as a sum of
two present values representing the eects of the investment decision and of
the nancing decision. The investment decision is captured by the market
value the stream would have if it were unlevered. The eects of the nancing
decision pertain to the market value of the stream of tax savings on interest
payments associated with the debt employed to nance the project. The
main assumption of the MM approach is that markets are free of arbitrage
and that there are no transaction costs. The MM model was extended by
Myers (see Myers (1974)) to normative capital budgeting analysis in terms
of the adjusted present value (APV) model.
Both practioners and theorists still seem to have a preference for the

Institut fur Bank{ und Finanzwirtschaft der Freien Universitat Berlin, Boltz-
mannstr. 20, D{14195 Berlin. We thank Anthony F. Herbst and Edwin O. Fischer for
helpful remarks.

1
Kruschwitz/Loer WACC and APV revisited p. 2

ubiquitous textbook weighted average cost of capital (WACC) approach.1

The WACC approach was developed by Modigliani and Miller (see Modigliani
& Miller (1963)), extended by Miles and Ezzel (see Miles & Ezzell
(1980)) and recently generalized by Loer (see Lo ffler (1998)).
The literature ususally states that WACC and APV, although dierent
methods, yield the same value of the rm.2 We want to show in our paper
that this is not the case. Not only that both approaches dier in the way
how the cash ow stream is valued, they make assumptions that cannot
be satised simultaneously. WACC and APV will not yield the same value
of the rm, since the underlying assumptions on the nancing decision are
completely dierent. Therefore, we carefully have to analyze the underlying
assumptions of both the WACC and the APV method. It turns out that we
have to consider two dierent cases depending whether the future value of
the rm is a random variable or not.
The paper is organized as follows: in the next section we state the as-
sumptions of the model. The third section considers the case of deterministic
future value of the rm. In the fourth section we analyze the case of random
future value. The last section concludes the paper.

2 Assumptions of the Model

If a rm is to be evaluated, one has to know what investments the managers
are going to make. The investment decision will then determine the free
cash ows of the rm. We will denote these cash ows at time t by CF g .
t
Future cash ows are uncertain today, at time t = 0.
We assume that the rm at time t is nanced by stocks St and bonds
Bt . Debt is certain. The cost of capital for debt will be denoted by rf , they
remain constant through time. Furthermore, cost of capital for the common
stock of an unlevered rm is given by rS and remain constant through time,
too. There is an income tax, applied at rate  , and debt reduces the taxable
income base because interest payments are tax-deductible. The tax rate is
assumed constant.
1 \In addition, frequent discussions with business executives have convinced us that the
WACC is the most widely used method in the real world, by far", Ross et al. (1996)
p.463. See also Brealey & Myers (1996), p. 513.
2 \Thus, the APV and WACC approaches are dierent ways to determine the same
value" Ross et al. (1996), p.461.

2
Kruschwitz/Loer WACC and APV revisited p. 3

The value of the unlevered rm is given by

V0 u
=
T
X (1  )E[CF
g]
t
: (1)
t=1
(1 + rS )t
To determine the value of the tax shield we have to consider two dierent
cases. These cases dier concerning the value of the unlevered rm Vtu at
t > 0: up to now we do not know whether or not it is a random variable.
In the second case (the value of the rm is a random variable) the nancing
decision of the rm has to be considered. Consequently, WACC and APV
will inevitably give dierent values of the rm. In the rst case, however,
the value of the rm is independent of the method one uses.

3 First case: V not a random variable

t
u

We are aware of two special cases in which this assumption will hold:
 If the future cash ow is constant through time the value of the unlev-
f
ered rm is always VtU = (1  )E[C
rS
F]
, and this is not a random variable.
 If the cash ow follows a Markov process the values CF
g
t are by
+1 ; : : :
assumption independent of the state of the world at time t. Hence,
the value of the rm Vtu is also independent of the state at time t.
If the value of the unlevered rm is not a random variable but deter-
ministic today, everthing turns out to be simple. The rm can be evaluated
depending on the data the investor knows. Since under the assumption of
the case debt and equity are certain (seen from t = 0), the tax shield is
certain too and has to be discounted with the cost of debt rf . Hence, the
value of the levered rm is given by the theory of Modigliani and Miller (see
Brealey & Myers (1996))
T
X1  rf Bt
V0l = V0u + (1 + rf )t+1 : (2)
t=0

If the value of the unlevered rm is not a random variable (2) gives always
the correct value of the rm. According to the literature we will denote this
equation by the APV approach. If an investor knows the future amount of
debt Bt then equation (2) already determines the value of the levered rm.
3
Kruschwitz/Loer WACC and APV revisited p. 4

But (2) cannot be applied if the investor knows only the future leverage
ratio instead of the amount of debt. The leverage ratio will be denoted by
lt = Bt : (3)
St

Since the underlying assumptions are not changed (2) still remains true. In
Appendix 1 we have shown that this equation can be written as3
TX1
(rf  )V u
rS + (1 + rf )(1  )E[CF
g ]
V0l = t
tQ1
t
; (4)
t=1 (1 + rS ) (1 + rk )
k =0

where rk is given by:

 
lk
1 + rk = 1 + rf 1 :  (5)
1 + lk
We will denote this equation as modied APV approach. We are not aware
of a formulation of (5) that looks similar to the WACC equation (see (9)
below).
To summarize: if the value of the unlevered rm is not a random variable,
one can use the APV approach (2) to determine the value of the levered rm.
If the future amount of debt is not known but the leverage ratio is known,
one can modify the APV approach to an equation that still determines the
rm's value. And both equations yield the same value.

4 Second case: V a random variable t

u

We will now assume that the value of the unlevered rm at time t is a

random variable. We believe that this assumption is (in contrast to section
3) the more realistic case. What are the consequences for the APV and the
WACC approach?
In the last section both debt and the leverage ratio were deterministic.
If VtU is a random variable, the story is more complicated. We will illustrate
that by means of an example. Consider a model with two periods. It is
3 If expected cash ows and the leverage ratio are constant this formula reduces to the
Modigliani{Miller equation
l V0u
V0 = l
:
rS 1  1+

l

4
Kruschwitz/Loer WACC and APV revisited p. 5

assumed that the rm will be liquidated at the end of period two. CF g is

2
the cash ow of the rm at time t = 2. The owner receives the after{tax{
cash ow (dividend) De 2. rS are the cost of capital for equity in the levered
rm.
Since the rm will be liquidated at time t = 2; and interest payments
reduce the tax base, we have
De = E[(1  )CF g j F ] (1 + r (1  ))B :
2 2 1 f 1

Therefore, equity at time t = 1 has the value

= E[1D+2 jrF1] :
e
e
S1
S

Now assume, the dividend policy of the rm is determined at time t = 0.

This is to say that the amount of debt at time t = 1 is already known at
t = 0 and is not a random variable. Therefore, we can evaluate the leverage
ratio at t = 1
B1 B1 (1 + rS )
l1 =
e = : (6)
e
S1 E[(1  )CF2 j F1] (1 + rf (1  ))B1
g

Now, particularly since the amount of debt is deterministic, the leverage

ratio at time t = 1 becomes a random variable! Hence, with a given dividend
policy the leverage policy will be prescribed and not deterministic.
Vice versa, if we lay down the leverage ratio at time t = 1 to be de-
termined at t = 0 the dividend policy is already specied. To prove this
rearrange (6) and get
g jF ]
l1 E[(1  )CF
e =
B
2 1
(7)
1
1 + rS + l1(1 + rf (1  )) :
(7) shows that a deterministic leverage ratio will yield a random dividend
policy. We summarize our interim results as follows.
nance policy
dividend policy leverage policy
Bt deterministic lt deterministic
lt random variable Bt random variable

This reveals that the investor has to make a decision about nancing
policy: There is a considerable dierence between determining the dividend
versus the leverage policy. Hence, we will expect two dierent values of the
rm according to the nancing policy that will be followed. We consider the
two policies separately.
5
Kruschwitz/Loer WACC and APV revisited p. 6

4.1 Deterministic debt or the APV method

Throughout this section we assume that future debt of the rm is determin-
istic. As we have seen already, the future leverage ratio of the rm will then
be a random variable. The theory of Modigliani and Miller now implies that
the tax shield will be deterministic and therefore certain. Consequently, the
cost of capital for the tax shield is rf .
Suppose that the investor knows the future amount of debt Bt . As we
have mentioned in the preceding section, the value of the rm is given by
the APV method:
T
X1  rf Bt
V0L = V0U + (1 + rf )t+1 : (8)
t=0

What happens if the investor does not know the future amount of debt
but instead the expected leverage ratios (remember that the leverage ratios
are uncertain)? Is it still possible to value the rm using this information?
Since the underlying assumptions of the model do not change, equation
(8) still remains true. The question is whether it is possible to evaluate
the future amount of debt by knowing only the expected leverage ratios.
Equation (8) should then be transformed such that it contains the expected
leverage ratios E[l~t].
We will now show that such a transformation is impossible. We illustrate
that with a simple example which shows that the knowledge of the expected
leverage ratios does not suce to value the rm if the future amount of debt
is certain.
Consider again equation (6)
" #
B1 (1 + rS )
E[el1] = E :
E[(1  )CF
g
2 j F ] (1 + rf (1
1  ))B1

At rst, it is easy to see that (given the expected leverage ratio) a unique
solution of the amount of debt exists: if B1 = 0 then the right hand side of
the equation is zero. If B1 increases, the right hand sides strictly increases.
Hence, there must be a unique B1 such that the right hand side equals the
expected leverage ratio.
But there does not necessarily exist a formula to determine B1 . Consider
the case where the conditional expectation of future cash ow E[CF g jF ]
2 1
has two realisations CF2;1 and CF2;1 with probabilities p1 and p2. Then (6)

6
Kruschwitz/Loer WACC and APV revisited p. 7

can be written as
B1 (1 + rS ) B1 (1 + rS )
E[el1] = p1 (1  )CF2;1 (1 + rf (1  ))B1
+ p2
(1  )CF2;2 (1 + rf (1  ))B1 ;
and we can calculate B1 by solving a quadratic equation. But if the condi-
tional expectation of future cash ow has n realisations, this leads us to a
polynomial of degree n and there does not exist a closed form solution for
such a polynomial in general.
To summarize: If the future amount of debt is deterministic and known,
then the APV method gives the correct value of the rm. If only the ex-
pected leverage ratios are known, APV is still the right method but there
does not exist a formula involving E[elt] to determine the value of the rm.

4.2 Deterministic leverage ratio or the WACC method

We will now assume that the leverage ratio is determined at time t = 0. This
implies that the tax shield is not certain and therefore cannot be discounted
at rf . Equation (8) does not yield the correct value of the rm. In Lo ffler
(1998) it is shown that the correct cost of capital is rS , the cost of capital
of the unlevered rm.
Suppose, the investor knows the future leverage ratios of the rm. As
was shown in Lo ffler (1998), the value of the rm is given by the WACC
method
V0L =
XT
h
(1  )E[CF g ]
t
i: (9)
Qt 
t=1 k =1 (1  rf
1+rf
l k 1 )(1 + r S )
What happens if the investor knows the expected amount of debt in
future periods? Since the underlying assumptions do not change, equation
(9) still determines the correct value of the rm. But this equation has to
be modied to include E[B~t ]. In appendix 2 we show that the value of the
rm is now given by the equation
XT
 rf E[B ~t ]
V0L = V0U + (10)
t=1
(1 + rS )t
(1 + rf ) :
As one can see, the dierence of this modied APV{equation compared to
equation (8) is the cost of capital for the tax shield.
Ususally, the literature states that there is a theoretical problem with
the WACC approach (the so{called circularity problem):
7
Kruschwitz/Loer WACC and APV revisited p. 8

\To calculate the changing WACC, one must know the market
value of a rm's debt and equity. But if the debt and equity
values are already known, the total market value of the company
is also known. That is, one must know the value of the company
to calculate the WACC" Ross et al. (1996), p.480.
As we have shown the WACC method can only be applied
 if the future leverage ratios are deterministic and
 if these ratios are known.
In this case there does not exist any circularity problem. In all other cases
however, WACC is not applicable. The circularity problem is a ctious one.

5 Summary
We have analyzed the underlying assumptions of the WACC and the APV
approach. We have shown that WACC can only be applied if the leverage
ratio of the company is deterministic and known at time t = 0. APV can
be used to evaluate the rm if the amount of future debt is deterministic
and known today. Since both methods use dierent assumptions on the
nancing decision of the rm, one cannot expect the same value by WACC
and APV.
We summarize our results as follow:
nance information
assumptions Bt (or E[Bt]) known lt (or E[lt]) known
Bt deterministic APV method (8) no formula
lt deterministic mod. APV method (10) WACC method (9)
What, if one wants to use dierent assumptions to those above? For
example, what if at time t = 0 the amount of debt should be deterministic
and in the future the leverage ratio lt should be certain? In this case one must
modify our approach (analogously to Appendix 1) and apply the correct cost
of capital for the tax shield. In all cases a formula can easily be obtained.

8
Kruschwitz/Loer WACC and APV revisited p. 9

Appendix 1
To prove (4) we assume the following equations. The value of the levered
rm at time t is given by the value of the unlevered rm Vtu and the value
of the tax shield Tt:
Vtl = Vtu + Tt: (11)
Since the cost of capital for the unlevered rm remain constant we have
Vtu+1 + (1  )E[CFg
t+1 ]
Vtu =  : (12)
1 + rS
The APV method implies the following iteration for the tax shield
Tt+1  rf
Tt = + B: (13)
1 + rf 1 + rf t
From (11), (12) and (13) we get using the denition of the leverage ratio lt
V u + E[CF g
t+1 ] Tt+1 +  rf 1+ lt
lt Vt
l
Vtl = t+1 + :
1 + rS 1 + rf
Rearranging this equation gives
!
V u + E[CF g ]
1 1 + r 1 + l Vtl = t+1 1 + r t+1 + 1T+t+1r
 rf lt
f t S f

V u
+ E[CF g ] V l
Vtu+1
= t+1 1 + r t+1 + t+1 1 + rf
S
Vtu+1(rf rS ) E[CF
g
t+1 ] Vtl+1
= (1 + 
rS )(1 + rf )
+ 1 + rS  + 1 + rf
 
V (r 
r )
1 + rf  rf 1 +lt l Vtl = t+11 +f r S + 11 + + rf E[CF
u
t+1 ] + Vt+1
g l
t r 
S S
V u
( r r )
Vtl = +
t+1 f S
(1 + rS )(1 + rf (1  1+ltlt ))
+ E[CF g
t+1 ](1 + rf )
+
Vtl+1
(1 + rS )(1 + rf (1  1+ltlt )) 1 + rf (1  1+ltlt )
By applying this iteration we show that (4) is indeed a consequence. Using
(5) we get
VTl 1 =
E[CF
g ](1 + r )
T f
(1 + r )(1 + r ) :
S T 1

9
Kruschwitz/Loer WACC and APV revisited p. 10

At time T 2 we have for the value of the rm

VTu 1 (rf rS ) E[CF
g
T 1 ](1 + rf ) E[CF
g ](1 + r )
VTl 2 = + + T f
(1 + rS )(1 + rT 2) (1 + rS )(1 + rT 2) (1 + rS )(1 + rT 1)(1 + rT 2) :
At T 3 we have
VTu 2(rf rS ) VTu 1(rf )
rS
VTl 3 =
(1 + rS )(1 + rT 3) (1 + rS )(1 + rT 2)(1 + rT 3 ) +
+

+ E[CFT 2 ](1 + rf ) + E[CF] T 1 ](1 + rf )

g g
 +
(1 + rS )(1 + rT 3 ) (1 + rS )(1 + rT 2)(1 + rT 3)
+ (1 + r )(1 +E[r CFT)(1
g ](1 + r )
: (14)
f

S T 1 + r T 2 )(1 + rT 3 )

As one can see, (4) indeed determines the value of the levered rm at time
t = 0.

Appendix 2
We assume that the leverage ratio is deterministic but unkown to the in-
vestor. Instead the investor knows the expected amount of debt E[B~t ].
We follow the theory in Lo ffler (1998) and know that the value of the
levered rm is given by the value of the unlevered rm and the value of the
tax shield
V0l = V0u + T0 :

This (uncertain) tax shield satises the following iteration

Tt =
E[Tt+1 j Ft] +  rf Ft :
1 + rS 1 + rf
Using the law of iterated expectation
t<s =) E[[X j Fs] j Ft] = E[X j Ft]
we get for the tax shield at time t = 0
T
X1 ~t ]
 rf E[B
T0 = (1 + rS )t
(1 + rf )
t=0

and therefore the assertion.

10
Kruschwitz/Loer WACC and APV revisited p. 11

References
Brealey, R.A., & Myers, St.C. (1996). Principles of Corporate Fi-
nance. 5. edn. New York: Wiley & Sons.
Loffler, A. (1998). WACC{approach and nonconstant lever-
age ratio. unpublished, Berlin. see http://www.wiwiss.fu-
berlin.de/w3/w3krusch/pub/ertrag.htm.
Miles, J.A., & Ezzell, J.R. (1980). The Weighted Average Cost of
Capital, Perfect Capital Markets, And Project Life: A Clarication.
Journal of Financial and Quantitative Analysis, 15, 719{730.
Modigliani, F., & Miller, M.H. (1958). The Cost of Capital, Corpo-
ration Finance and the Theory of Investment. American Economic
Review, 48, 261{297.
Modigliani, F., & Miller, M.H. (1963). Corporate Income Taxes and
Cost of Capital: A Correction. American Economic Review, 53,
433{443.
Myers, S.C. (1974). Interactions of Corporate Financing and Investment
Decisions { Implications for Capital Budgeting. Journal of Finance,
29, 1{25.
Ross, S.M., Westerfield, R.W., & Jaffe, J.E. (1996). Corporate
Finance. 4th edn. Chicago: McGraw{Hill, Inc.

11