A General Decision-Tree Approach

to Real Option Valuation

Abstract
The common paradigm for risk-neutral real-option pricing is a special case
encompassed within our general model for valuing investment opportunities.
Risk-neutral real option prices deviate from the risk-averse real option values
that apply in an incomplete market, giving dierent rankings of investment
opportunities and dierent optimal exercise strategies. Unlike risk-neutral prices,
more general real option values often decrease with the volatility of the asset
price. They also depend on the structure of xed and variable investment
costs, the expected return of the underlying asset, the frequency of decision
opportunities, the price of the asset relative to initial wealth, the investors risk
tolerance and its sensitivity to wealth. We explain how these factors aect the
ranking of real option values under a standard geometric Brownian motion for
the asset price. Numerical examples also consider boom-bust or mean-reverting
price scenarios and investments with positive or negative cash ows.
Key words: CARA, CRRA, certain equivalent, development, divestment, displaced log
utility, exponential utility, HARA, investment, mean-reversion, property boom, real
estate, risk aversion, risk tolerance
1 Introduction
The term real option is commonly applied to a decision opportunity for which the investment cost
is predetermined, and the vast majority of the literature assumes the underlying asset is traded in
a complete or partially complete market so that all (or at least the important) risks are hedgeable.
Real options are typically regarded as tradable contracts with predetermined strikes, the standard
risk-neutral valuation (RNV) principal is invoked, most commonly in a continuous-time setting,
and the mathematical problem is no dierent to pricing an American option under the risk-neutral
measure.
Using the RNV principle implies there is a unique market price for the real option that is
positively related to the volatility of the underlying asset. This is in sharp contrast with the
more traditional discounted cash ow (DCF) approach, where an increase in risk decreases the net
present value of the investments expected cash ows. But the RNV approach only applies to a
very special type of real option. The original denition, rst stated by Myers (1977), is a decision
opportunity for a corporation or an individual. It is a right, rather than an obligation, whose
value is contingent on the uncertain price(s) of some underlying asset(s). The RNV assumptions of
perfectly hedgeable risks and predetermined strike are clearly inappropriate for many real options,
1
particularly those in real estate, research & development or mergers & acquisitions. In many
applications the market for the underlying of a real option is incomplete, the underlying investment
is not highly liquid and the investment cost is not predetermined, indeed the transacted price is
typically negotiated between individual buyers and sellers.
Under such conditions Grasselli (2011) proves that the real option to invest still has an intrinsic
value, but it is purely subjective. Unlike the premium on a nancial option, it has no absolute
accounting value. It represents the dollar amount, net of nancing costs, that the investor should
receive for certain to obtain the same utility as the risky investment. The decision makers utility
function may be applied to value every opportunity but his subjective views about the evolution
of the market price and any future cash ows may be project-specic. So real option values allow
alternative investment opportunities to be ranked, just as nancial investments are ranked using
risk-adjusted performance measures. Thus, a pharmaceutical company may compare the values
of real options to develop alternative products, or an oil exploration company may compare the
values of drilling in dierent locations, or a private property development company may compare
the values of opportunities to buy and develop plots of land in dierent locations. The solution
also determines an optimal exercise strategy for each alternative.
We develop a general decision-tree approach for determining the value of an investment op-
portunity and its optimal exercise, in a market that need not be complete, where the solution is
derived via maximisation of expected utility. To represent an investors risk preferences using a
utility is no greater assumption than using the mean-variance criterion or Sharpe ratio. Indeed, a
utility is less restrictive than these criterion because they implicitly assume normally distributed
uncertainties and an exponential utility. But exponential utilities make restrictive assumptions
about the investors preferences. More generally, in our approach the decision maker may have
any utility in the hyperbolic absolute risk aversion (HARA) class, introduced by Mossin (1968)
and Merton (1971). So we do not focus on analytic solutions, available only in the exponential
case, and neither do we rely on expansion approximations, which can be unreliable.
Assumptions about costs are critical determinants of real option values, and of their sensitivities
to expected return and risk. So we employ a general structure with both predetermined and
stochastic components for investment costs. Thus, the general real option problem falls squarely
into the realm of decision analysis. The standard RNV real option price only applies under a
xed cost assumption. Indeed, if the cost is stochastic and perfectly correlated with a martingale
discounted asset price, for example, then the RNV price would be zero, yet a risk-averse investor
would still place a positive value on such a decision opportunity.
We consider three dierent processes for the discounted market price: geometric Brownian
motion (GBM), mean-reversion and boom-bust market price scenarios. Standard RNV real option
prices correspond to the special case that the investment cost is predetermined, the forward asset
price follows a GBM with total return equal to the risk-free rate, the price is derived from a linear
utility and the investment decision may be taken at any point in time. More generally, in an
incomplete market the optimal exercise path and the corresponding option value depend on the
decision makers risk preferences, on his subjective views about the market price process and cash
2
ows, and on the frequency of decision opportunities (e.g. at a quarterly meeting of the board of
directors).
The general properties of the model include a comparison of RNV real option prices with those
attributed by risk-averse investors in incomplete markets. Moreover, our model provides answers
to several important and relevant questions that have not previously been addressed in the real
options literature. If the xed or predetermined strike assumption is not valid, how does this
change the option value and what is the eect on its sensitivity to expected return and risks of
the investment? How should a real option value reect the frequency of decision opportunities?
What is the eect of the investors risk preferences on real option values, and how does this aect
the ranking of dierent opportunities? And how is this ranking inuenced by the structure of the
cash ows, or by the price of the asset relative to the initial wealth of the investor?
We proceed follows: Section 2 places our work in the context of the real options literature;
Section 3 describes the model; Section 4 describes its properties under a GBM price process;
Section 5 considers investors that believe the price process is mean-reverting, or subject to regime
shifts, and the case that the investment has associated cash ows; and Section 6 summarises and
concludes.
2 Relevance to the Literature
Most of the literature on real options focuses on opportunities to enter a tradable contract with
predetermined strike on an asset traded in a complete market. That is, the real options pay-o
distribution can be replicated using tradable assets, so that all risks are hedgeable. Also, the
decision opportunity can be exercised continuously at any time over an innite horizon. In this
setting the option has the same value to all investors (Harrison and Kreps, 1981) and thus can
be valued as if the investor is risk neutral, just like a nancial option. This major strand of the
literature regards a real option as a tradable contract, usually with a xed or predetermined strike
and a nite horizon: see Triantis and Hodder (1990), Capozza and Sick (1991), Trigeorgis (1993),
Panayi and Trigeorgis (1998), Benaroch and Kauman (2000), Boer (2000), Yeo and Qiu (2002),
Shackleton and Wojakowski (2007) and many others. Some papers assume an innite horizon see
Kogut (1991), Grenadier (1996), Smith and McCardle (1998) and Patel et al. (2005) for a review.
Other studies include a stochastic strike, including McDonald and Siegel (1986), Quigg (1993) and
Bowman and Moskowitz (2001).
However, in practice, many decision opportunities encompassed by Myers original denition
are not standardised, tradable securities and their risks are only partially hedgeable, if at all. For
instance, if an oil exploration company must decide whether to drill in location A or location
B, its views about the benets of drilling in each location will depend on their subjective beliefs
about the market prices of oil in the future, as well as their risk preferences. And investment costs
are unlikely to be xed or predetermined. For instance, when a pharmaceutical company decides
which drug to research and develop, both the research costs and the subsequent prots tend to be
positively correlated with the drugs potential market price. In this setting there is no unique value
for a real option; it will be specic to the investor, depending upon his subjective views about the
3
costs and benets of investment, and upon his risk preferences.
Closed-form, continuous-time, deterministic-strike real option values in incomplete or partially
complete markets have been considered for decision makers with an exponential utility by Hen-
derson (2002, 2007), Miao and Wang (2007) and Grasselli (2011). When the underlying asset is
correlated with a market price Henderson (2007) applies the closed-form solution derived by Hen-
derson (2002) for the exponential option value and the investment threshold, showing that market
incompleteness results in earlier exercise and a lower real option value; moreover the option value
decreases with volatility. Henderson and Hobson (2002) employ a power utility function in the
Merton (1969) investment model, proposing an approximate closed-form optimal reservation value
for the option when the investment cost is small relative to initial wealth. They show that the
exponential and power utility option values behave very dierently as a function of risk aversion
as it tends to zero, due to boundary constraints on the power utility value. Evans et al. (2008)
compare power and logarithmic utility values in a similar setting. Most of these papers derive
the optimal investment threshold and the indierence value for a nite-horizon, continuous choice,
deterministic-strike investment opportunity using a two-factor GBM framework in which the value
of the project to the investor is stochastic and possibly correlated with the price of a liquidly
traded asset that may be used to hedge the investment risk. Both price processes are discounted
to present value terms, so that a univariate time 0 utility function may be applied to maximise
the expected utility of the maturity pay-o. Our work uses a similar discounted value process, but
since we suppose from the outset that no risks are hedgeable by traded securities we utilise only a
one-factor framework.
Like us, Grasselli (2011) considers the case where none of the risks of the investment are
hedgeable. Importantly, he proves that the time-exibility of the opportunity to invest still carries
an option value for a risk-averse investor, so that the paradigm of real options can be applied
to value a private investment decision. Employing an exponential utility he proves that the real
option value converges to zero as risk tolerance decreases.
Early work on real-option problems employed a decision-tree approach with a constant, risk-
adjusted discount rate in which the option value decreases with risk see Mason and Merton
(1985), Trigeorgis and Mason (1987), Copeland et al. (1990) and Copeland and Antikarov (2001).
However, Copeland et al. (1990) dismissed the decision-tree approach and directed the main-
stream of real options research towards risk-neutral valuation, where standard American option
prices increase with the volatility of the underlying asset. Trigeorgis (1996), Smit (1996), Brand ao
and Dyer (2005), Brand ao et al. (2005, 2008) and Smith (2005) have since employed a decision-tree
approach, but also on the assumption of market-priced risk, thereby adopting a RNV framework.
The Integrated Valuation Procedure (IVP) introduced by Smith and Nau (1995) and extended
by Smith and McCardle (1998) considered risk-averse decision makers in a decision-tree framework,
this time explicitly endowed with a utility. A multivariate utility is dened as a sum of time-
homogenous utilities of future cash ows and the decision tree applies backward induction on their
certain equivalent (CE) value. The IVP approach is only valid for an exponential utility and when
cash ows at dierent times are independent, because only exponential utilities have the unique
4
property that their CE is additive over independent random variables. Although economic analysis
is commonly based on inter-temporal consumption, it is standard in nance to base the utility of
decisions on nal wealth, with future values discounted to time 0 terms. This discounting is an
important element of our approach because it greatly simplies the decision analysis and allows
the backward induction step to be dened on the expected utility relative to any univariate HARA
utility function, not only on the CE values of an additively separable multivariate exponential
utility.
3 The Model
We assume investment risks are unhedgeable, so the market for the underlying asset is incomplete;
its forward market price measure is therefore subjective to the decision maker, with the risk-neutral
measure arising as the special case of a linear utility and a risk-free expected return; investment
costs may have predetermined and/or stochastic components; decision opportunities are discrete
and are modelled using a binomial price tree with decision nodes placed at every k steps; the
decision horizon T is nite; and the consequence of the decision is valued at some nite investment
horizon, T

> T. We do not equate the investment horizon with the maturity of the investment,
i.e. cash ows from the investment may continue beyond T

, so the market price is not constrained

to be the discounted expected value of the cash ows.
1
The decision maker holds subjective views
not only about the evolution of the market price for 0 t T

but also about the stream of cash

ows (if any) that would be realised if he enters the investment. In most applications cash ows
would reect the individual management style of the decision maker, e.g. aggressive, expansive,
recessive etc. Additionally, the decision maker is characterised by his initial wealth, w
0
which
represents the current net worth of all his assets, and a HARA utility function U(w) which reects
his risk tolerance and how this changes with wealth.
3.1 Market Prices and Cash Flows
All future market prices and cash ows are expressed in time 0 terms by discounting at the
decision-makers borrowing rate r.
2
Thus the investor borrows funds at rate r to invest, rather
than nancing the cost from his initial wealth, which we suppose is not available for investment.
3
We suppose that r is a constant, risk-free rate.
First we allow the market price of the underlying asset to follow a GBM, as usual, so that the
discounted forward market price p
t
evolves over time according to the process:
dp
t
p
t
= ( r)dt +dW
t
, for 0 < t T

, (1)
1
As in Kasanen and Trigeorgis (1995), the market price could be related to the utility of a representative decision
maker, being the break-even price at which the representative decision maker is indierent between investing or
not.
2
This rate depends on the business risk of the investment as perceived by the nancer, not as perceived by the
decision maker. It may also depend on the decision-makers credit rating but this assumption is not common in the
real options literature.
3
To avoid additional complexity we do not consider that investments could be nanced from wealth, even though
this would be rational if wealth is liquid and r is greater than the return on wealth, r.
5
where and are the decision-makers subjective drift and volatility associated with p
t
and W
t
is a Wiener process. Then p
t
has a lognormal distribution, p
t
log N
_
( r) t,
2
t
_
.
It is convenient to use a binomial tree discretisation of (1) in which the price can move up or
down by factors u and d, so that p
t+1
= p
t
u with probability and otherwise p
t+1
= p
t
d. No
less than eleven dierent binomial parameterisations for GBM are reviewed by Chance (2008).
Smith (2005), Brand ao and Dyer (2005), Brand ao et al. (2005, 2008), Smit and Ankum (1993) and
others employ the CRR parameterisation of Cox et al. (1979). However, the Jarrow and Rudd
(1982) parameterisation, which is commonly used by option traders, is more stable for low levels
of volatility and when there are only a few steps in the tree. Thus we set
m =
_
r 0.5
2

t, u = e
m+

t
, d = e
m

t
and = 0.5. (2)
We also consider a modication of (1) that represents a regime-dependent process, which trends
upward with a low volatility for a sustained period and downward with a high volatility for another
sustained period, replicating booms and busts in the market price. Thus we set:
dp
t
p
t
=
_
_
_
(
1
r)dt +
1
dW
t
, for 0 < t T
1
,
(
2
r)dt +
2
dW
t
, for T
1
< t T

.
(3)
Alternatively, the decision maker may believe the market price will mean-revert over a relatively
short time horizon, and to replicate this we suppose the expected return decreases following a price
increase but increases following a price fall, as in the Ornstein-Uhlenbeck process:
d ln p
t
= ln
_
p
t
p
_
dt +dW
t
(4)
where denotes the rate of mean reversion to a long-term price level p. Following Nelson and Ra-
maswamy (1990) (NR) we employ the following binomial tree parameterisation for the discretised
Ornstein-Uhlenbeck process:
u = e

t
, d = u
1
(5a)

s(t)
=
_

_
1, 0.5 +
s(t)

t/2 > 1
0.5 +
s(t)

t/2, 0 0.5 +
s(t)

t/2 1
0, 0.5 +
s(t)

t/2 < 0
(5b)
where

s(t)
= ln
_
p
s(t)u
p
_
(5c)
is the local drift of the log price process, which decreases as increases. The corresponding price
process thus has local drift:

s(t)
= ln
_
p
s(t)u
p
_
+ 0.5
2
+r (5d)
6
Note that when = 0 there is a constant transition probability of 0.5 and the NR parameterisation
is equivalent to the parameterisation (2) with m = 0.
The cash ows, if any, may depend on the market price of the asset as, for instance, in rents
from a property. Let s(t) denote the state of the market price at time t, i.e. a path of the market
price from time 0 to time t. In the binomial tree framework s(t) may be written as a string of us
and ds with t elements, e.g. uud for t = 3. Now CF
s(t)
denotes the cash ow when the market
price is in state s(t) at time t. Regarding cash ows as dividends we call the price excluding all
cash ows before and at time t the ex-dividend price, denoted p

s(t)
. At the time of a cash ow
CF
s(t)
the market price follows a path which jumps from p
+
s(t)
= p

s(t)
+CF
s(t)
to p

s(t)
. We suppose
that he receives the cash ow at time t if he divests in the project at time t but does not receive
it if he invests in the project at time t. The alternative assumption that he receives the cash ow
at time t only when he invests is also possible.
4
The dividend yield, also called dividend pay-out
ratio, is dened as

s(t)
=
p
+
s(t)
p

s(t)
p
+
s(t)
. (6)
We assume that dividend yields are deterministic and time but not state dependent, using the
simpler notation
t
. Then the cash ows are both time and state dependent.
5
3.2 Costs and Benets of Investment and Divestment
The investment cost at time t, in time 0 terms, is:
K
t
= K + (1 ) g(p

t
), 0 1, (7)
where K is a constant in time 0 terms and g : R R is some real-valued function linking the
investment cost to the market price, so that cost is perfectly correlated with price only when g
is linear. When = 1 we have a standard real option with a predetermined strike K, such as
might be employed for oil exploration decisions. When = 0 we have a variable cost linked to the
market price p

t
, such as might be employed for real estate or merger and acquisition options. The
intermediate case, with 0 < < 1 has an investment cost with both xed and variable components.
We assume that initial wealth w
0
earns a constant, risk-free lending rate r, as do any cash ows
paid out that are not re-invested. Any cash paid into the investment (e.g. development cost) is
nanced at the borrowing rate r. The nancial benet to the decision-maker on investing at time
t is the sum of any cash ows paid out and not re-invested plus the terminal market price of the
investment. Thus the wealth of the investor at T

, in time 0 terms, following investment at time t

4
The subscript s(t) denotes a particular realisation of the random variable that carries the subscript t, e.g. p
+
uud
and p

uud
are the left and right limits of a realisation of pt when t = 3. From henceforth we only use the subscript
s(t) when it is necessary to specify the state of the market otherwise we simplify notation using the subscript t.
Also, assuming cash ow is not received when divesting at time t, p

t
is a limit of pt from the right, not a limit from
the left.
5
If the cash ow is not state dependent, then the dividend yield must be state dependent. The state dependence of
cash ows induces an autocorrelation in them because the market price is autocorrelated. For this reason, dening an
additively separable multivariate utility over future cash ows as in Smith and Nau (1995) and Smith and McCardle
(1998) may be problematic.
7
is
w
I
t,T
= e
( rr)T

w
0
+
T

s=t+1
e
( rr)(T

s)
CF
s
+p

K
t
. (8)
Some investments pay no cash ows, or any cash ows paid out are re-invested in the project.
Then the nancial benet of investing at time t, in time 0 terms, is simply the cum-dividend
price p
t,T
of the investment accruing from time t. If a decision to invest is made at time t, with
0 t T, then p
t,t
= p
t
but the evolution of p
t,s
for t < s T

diers to that of p
s
because
p
t,s
will gradually accumulate all future cash ows from time t onwards. In this case, when the
decision maker chooses to invest at time t his wealth at time T

in time 0 terms is
w
I
t,T
= e
( rr)T

w
0
+ p
t,T
K
t
. (9)
Note that if r = r then also

s=t+1
e
( rr)(T

s)
CF
s
+p

= p
t,T
in (8).
Similarly, if the decision maker already owns the investment at time 0 and chooses to sell it at
time t, the time 0 value of his wealth at time T

is
w
S
t,T
= e
( rr)T

w
0
+
t1

s=1
e
( rr)(T

s)
CF
s
+p
+
t
p
0
. (10)
Note that p
0
is subtracted here because we assume the investor has borrowed funds to invest in
the property. Alternatively, if there are no cash ows, or they are re-invested,
w
S
t,T
= e
( rr)T

w
0
+ p
0,t
p
0
. (11)
The wealth w
D
t,T

resulting from a defer decision at time t depends on whether he invests (or

divests) later on. This must therefore be computed using backward induction as described in the
next sub-section.
3.3 Optimal Decisions and Real Option Value
As in any decision problem, we shall compare the expected utility of the outcomes resulting from
investment with the utility of a base-case alternative, which in this case is to do nothing so that
terminal wealth remains at w
0
in time 0 terms. For brevity, we describe the backward induction
step only for the decision to invest it is similar for the decision to divest, but replace I with
S (for an existing investor that sells the property) and D with R (for an existing investor that
remains invested). The option to invest at time t has time 0 utility value U
I
t,T

= U
_
w
I
t,T

_
, but
since w
I
t,T

is random so is U
I
t,T

, and we use the expected utility

E
_
U
I
s(t),T

_
= E
_
U
_
w
I
s(t),T

__
,
as a point estimate. Then, given a specic decision node at time t, say when the market is in state
s(t), the potential investor chooses to invest if and only if
E
_
U
I
s(t),T

_
> E
_
U
D
s(t),T

_
,
8
and we set
E
_
U
s(t),T

= max
_
E
_
U
I
s(t),T

_
, E
_
U
D
s(t),T

__
. (12)
Since there are no further decisions following a decision to invest, E
_
U
I
s(t),T

_
can be evaluated
directly, using the utility of the terminal wealth values obtainable from state s(t) and their associ-
ated probabilities. However, E
_
U
D
s(t),T

_
depends on whether it is optimal to invest or defer at the
decision nodes at time t +1. Thus, the expected utilities at each decision node must be computed
via backward induction.
First we evaluate (12) at the last decision nodes in the tree, which are at the time T that
option expires. These nodes are available only if the investor has deferred at every node up to
this point. We associate each ultimate decision node with the maximum value (12) and select the
corresponding optimal action, I or D. Now select a penultimate decision node; say it is at time
T kt. If we use a recombining binomial tree to model the market price evolution, it has 2
k
successor decision nodes at time T.
6
Each market state s(T kt) has an associated decision node.
Each one of its successor nodes is at a market state s

(T) that is attainable from state s(T kt),

and has an associated probability
s

(T)
determined by the state transition probability of 0.5,
given that we employ the parameterisation (2).
7
Using the expected utility associated with each
attainable successor node, and their associated probabilities, we compute the expected utility of
the decision to defer at time T kt. More generally, assuming decision nodes occur at regular
time intervals, the backward induction step is:
E
_
U
D
s(tkt),T

_
=

(t)

(t)
E
_
U
s

(t),T

, t = kt, 2kt, . . . , T kt, T. (13)

At each decision node we compute (13) and associate the node with the optimal action and its
corresponding maximum expected utility. We repeat the backward induction until we arrive at a
single expected utility value associated with the node at time 0. Finally, the option value is the
CE of this expected utility, less the wealth resulting from the base-case alternative, i.e. w
0
. By
denition, CE(w) = U
1
(E[U(w)]) for any monotonic increasing utility U.
3.4 Risk Preferences
Denote by U : R R the decision-makers utility function. Previous research on decision analysis
of real options, reviewed in Section 2, employs either risk-neutrality or an exponential utility
function, which may be written in the form
U(w) = exp
_

_
, (14)
6
The recombining assumption simplies the computation of expected utilities at the backward induction step.
However, we do not require that the binomial tree is recombining so the number of decision nodes could proliferate
as we advance through the tree. Note that the state price tree will recombine if cash ows are determined by a
time-varying but not state-varying dividend yield.
7
So if the tree recombines these probabilities are 0.5
k
, k0.5
k
, k!/(2!(k 2)!)0.5
k
, ...., k0.5
k
, 0.5
k
under the JR
parameterization (2). If the CRR parameterisation is employed instead the transition probabilities for a recombining
tree would be more general binomial probabilities.
9
where w denotes the terminal (time T

) wealth of the decision maker, expressed in time 0 terms

and > 0 denotes his risk tolerance and =
1
is his risk aversion. Note that w is a random
variable taking values determined by the decision-makers (subjective) views on the evolution of
the market price and the decisions he takes before time T

. Under (14) we have

CE(w) = log
_

E[U(w)]

_
. (15)
This function is frequently employed because it has special properties that make it particularly
tractable (see Davis et al., 2006, Chapter 6).
1. The exponential function (14) is the only utility with a CE that is independent of the decision-
makers initial wealth, w
0
.
2. The Arrow-Pratt coecient of risk aversion is constant, as U

(w)/U

(w) = . Thus,
the exponential utility (14) represents decision makers with constant absolute risk aversion
(CARA) and in (14) is the absolute risk tolerance.
3. The CE of an exponential utility is additive over independent risks. When x
t
NID(,
2
)
and w
T
= x
1
+. . . +x
T
then CE(w
T
) = T (2)
1

2
T.
Properties 1 and 2 are very restricting. CARA implies that decision makers leave unchanged
the dollar amount allocated to a risky investment when their initial wealth changes, indeed the
investors wealth has no inuence on his valuation of the option. Property 3 implies that when
cash ows are normally and independently distributed (NID) the decision-makers risk premium
for the sum of cash ows at time t is (2)
1

2
t, so it scales with time at rate (2)
1

2
. This could
be used to derive the risk-adjustment term that is commonly applied to DCF models and in the
inuential book by Copeland et al. (1990).
8
Exponential utilities also have an important time-homogeneity property which, unlike the prop-
erties above, is shared by other utility functions in the HARA class: Suppose U
t
: R R on
t = 0, 1, ...., T

, is dened with risk tolerance

t
= e
rt
and set w
0
t
= e
rt
w
t
. Then:
U
t
(w
t
) =
t
exp
_

w
t

t
_
= e
rt
exp
_

w
0
t

_
= e
rt
U(w
0
t
), (16)
where U : R R is a time-invariant exponential utility function as in (14) dened on any future
value of wealth discounted to time 0. This property makes any HARA utility particularly easy to
employ in decision-tree analysis. In particular, on assuming the risk tolerance is time-varying and
grows exponentially at the same rate as the discount rate, we can value any future uncertainty using
the constant utility function (14) applied to time 0 values. This is much easier than discounting
the expected values of time-varying utility functions applied to time t values at every step in
the backward induction. Our framework is not constrained to exponential preferences over NID
uncertainties; indeed, this would lead to a solution where the same decision (to invest, or to defer)
8
Setting exp[r
a
] = (2)
1

2
gives r
a
= log[1 (2)
1

2
], so r
a
(2)
1

2
.
10
would be reached at every node in the tree, because the uncertainties faced are just a scaled version
of uncertainties at any other node.
The CARA property of the exponential utility is often criticised because it does not apply to
investors that change the dollar amount allocated to risky investments as their wealth changes.
For this reason we allow other utility functions from the HARA class, where both absolute and
relative risk aversion can increase with wealth.
9
HARA utility functions have a local relative risk
tolerance that increases linearly with wealth at the rate , and are dened as:
U(w) = [1 +

w
0
(w w
0
)]
1
1
(1 )
1
, for w > (1
1
)w
0
. (17)
We consider four special cases: when = 0 we have the exponential utility; = 1 corresponds to
the displaced logarithmic utility; = 0.5 gives the hyperbolic utility; and = gives the power
utility. Note that an investors risk tolerance and its sensitivity to wealth may be dened fairly
accurately using the techniques introduced by Keeney and Reia (1993).
4 Properties under GBM
Our real option value represents the net present value that, if received with certainty, would give a
risk-averse investor the same utility value as the expected utility of the uncertain investment. Such
values enable the investor to rank alternative investment opportunities and the solution species
an optimal time to exercise. The minimum value of zero applies when the investment would never
be attractive whatever its future market price. The special case of RNV, while most commonly
employed in the literature, only applies to a real option that is tradable on a secondary market.
A separate appendix presents a simple example, with full illustration of the decision tree and
Matlab code, to help readers x ideas. The decision maker has an exponential utility and the
transacted price is the market price of the asset. Thus, the investment decision provides a concrete
example of the zero-correlation case considered by Grasselli (2011), where the opportunity to invest
still carries a positive value. A second example shows that the equivalent divest decision also has
a positive value.
Properties of the general model are now described under the GBM assumption (1) for market
prices. Unlike risk-neutral prices, general real option values may reect: (i) the cost of the in-
vestment relative to the companies net asset value/initial wealth; (ii) the scheduling of decision
opportunities; and (iii) the sensitivity of risk tolerance to wealth.
10
4.1 Investment Costs and the RNV Approach
Great care should be taken when making assumptions about investment costs. In some applications
for instance, when a licence to drill for oil has been purchased and the decision concerns whether
the market price of oil is sucient to warrant exploration a xed-strike or predetermined cost
9
Relative risk tolerance is expressed as a percentage of wealth, not in dollar terms. So if, say, = 0.4 the decision
maker is willing to take a gamble with approximately equal probability of winning 40% or losing 20% of his wealth,
but he would not bet on a 50:50 chance (approximately) of winning x% or losing x%/2 for any x > 0.4.
10
From henceforth, unless otherwise stated, we set r = r; no additional insights to the model properties are
provided by using dierent lending and borrowing rates.
11
assumption could be valid. However, in many cases the investment cost is linked to the market
price. The assumption (7) about the investment cost has a crucial inuence not only on the value
of a real option and its optimal exercise strategy, but also on their sensitivity to changes in the
input parameters.
A xed cost, where = 1 in (7), may be regarded as the strike of an American call option,
and the value is derived from the expected utility of a call option pay-o for which the upper
part of the terminal wealth distribution above the strike matters. The opposite extreme ( = 0)
focuses on the lower part of the terminal wealth distribution below the current price p
0
, where the
investment costs are lowest. Although log returns are similar across the whole spectrum under
the GBM assumption (1), P&L is in absolute terms and it is greater in the upper part of the
distribution than in the lower part. For this reason an at-the-money xed-strike assumption yields
a greater real option value than the invest-at-market-price assumption.
11
Real option values for
risk-averse investors always increase with risk tolerance, and can be greater than or less than the
RNV option price depending on the investors expected return and risk tolerance.
We illustrate these properties with a typical example of an option to purchase an asset that has
no associated cash ows. The current price of the asset is $1m and the investor believes this will
evolve according to (1) with and as specied in Table 1. The risk-free lending and borrowing
rates are both 5%. The investment horizon is T

= 5 years, investors have an exponential utility

with risk tolerance , and the initial wealth is $1m. As we have a linear utility, giving
the option value for a risk-neutral investor, and further setting = r gives the RNV solution.
Decisions are made once per year, so we set t = k = 1.
12
The investment cost takes the general form (7) with K = $1m. We set g(x) = x and compare
investment cost at market price ( = 0) with xed time 0 cost ( = 1) and a mix of xed and
variable cost ( = 0.5). We also consider g(x) = x/2 for variable cost at a fraction (in this case
one-half) of the market price, plus a xed cost if > 0, and set g(x) =

x for a variable cost that

increases non-linearly with market price, so the price and cost are not perfectly correlated. When
g(x) =

x the variable cost is less than (greater than) the current market price p

t
if p

t
> $1m
(p

t
< $1m).
Table 1 reports the real option values under each cost assumption, for = 0.1, 0.5 and 1, the
real option value corresponding to the linear utility of a risk-neutral investor ( = ), and the
RNV price. Under the option value we report the year and any market state that the optimal
investment strategy is conditional upon. For instance, 1/d denotes invest at time 1 provided the
market price moved down between time 0 and 1, and 4/uu denotes invest at time 4 provided the
market price moved up between time 0 and 1, and again between time 1 and 2, irrespective of
later market price moves. Investment at time 0 has no market state, 4/ denotes invest in year 4
irrespective of the price state, and never invest is marked simply .
The risk-averse option value always increases with , in accordance with results of Henderson
(2007). Clearly, the more risk tolerant the investor, the lower the risk premium required to invest.
11
However, this property only holds under GBM views for market prices, see Section 5.1 for a counter example
under dierent price processes.
12
Similar properties are evident using other real option parameters, with results available on request.
12
Table 1: Exponential utility option values for dierent risk tolerance and dierent invest-
ment costs of the form (7). Risk-neutral values (linear utility, i.e. = ) and RNV prices
(linear utility, = r). Real option parameters: T

= 5 years, t = k = 1, K = p
0
= $1m,
r = r = 5%, with and specied above.
= 10% g(x) - x

x x/2
= 20% 1 0.5 0 0.5 0 0.5 0

0.1
Value 47,387 29,349 0 36,260 32,222 115,185 390,296
Year/State 3/uu 4/uuu - 4/uuu 4/uuu 4/u 4/-
0.5
Value 199,103 117,936 43,008 166,906 129,091 323,672 561,529
Year/State 2/uu 4/uu 0 4/uu 4/uu 4/u 2/dd
1
Value 261,839 157,221 141,249 218,048 170,077 393,401 641,249
Year/State 3/uuu 3/uud 0 4/uu 4/uu 2/ud 0

Value 363,789 283,179 283,179 306,855 283,179 533,179 783,179

Year/State 2/uu 0 0 1/u 0 0 0
RNV
Value 160,122 80,030 0 124,891 89,661 270,281 499,604
Year/State 3/uuu 4/uuu - 4/uuu 4/uuu 4/uu 4/-
= 15% g(x) - x

x x/2
= 50% 1 0.5 0 0.5 0 0.5 0

0.1
Value 37,277 5,374 0 34,229 6,454 36,865 148,551
Year/State 3/uuu 4/uuuu - 4/uuu 4/uuuu 4/uuu 4/-
0.5
Value 151,743 36,014 3,023 125,942 82,806 181,581 341,763
Year/State 3/uuu 4/uuu 3/ddd 4/uuu 4/uuu 4/uu 3/ddd
1
Value 254,963 112,602 17,221 216,766 166,120 312,741 461,455
Year/State 4/uuu 4/uuu 2/dd 4/uuu 4/uuu 4/uu 2/dd

Value 823,324 608,935 608,935 720,471 624,340 879,799 1,108,935

Year/State 3/uuu 0 0 2/uu 1/u 1/u 0
RNV
Value 376,412 186,519 0 306,003 235,593 359,591 485,545
Year/State 3/uuu 4/uuu - 4/uuu 4/uuu 4/uuu 4/-
As the value converges to the risk-neutral (linear utility) value, which is greater than
the standard RNV option price in this case because > r, otherwise it would be less than the
RNV price. The risk-averse option value can be much less than the RNV price especially when
risk tolerance is low, or much greater than the RNV price especially when risk tolerance is high.
Indeed when = 0 and g(x) = x, i.e. the investment cost is at the market price, the RNV price is
always zero, because the CE of a linear utility is the expected value of terminal wealth, and since
the discounted price is always a martingale under the risk-neutral measure, CE = w
0
. So the value
of any investment opportunity will be zero in this case. By contrast, the risk-averse decision maker
places a positive value on the option in this case, except when = 0.1, when he would choose not
to invest at any market price.
The cost structure also aects the optimal investment strategy. For a nite the optimal time
to invest is never shorter than for the risk-neutral investor, again as shown by Henderson (2007).
Further, when = 0 and g(x) = x, optimal investment is never conditional on a price rise, though
it may be conditional on a price fall. Investment also becomes conditional on a price fall but not
13
a rise when = 0 and g(x) = kx, for 0 < k < 1, except that never invest is not possible, since
the last period pay-o is x kx > 0 so an optimal strategy always invests at or before the last
period of the option.
13
But when g(x) =

x the last period pay-o x

x > 0 only when x > 1,

i.e. after price rise, and optimal investment becomes conditional on up moves.
When = 0.5 or 1 there is a xed cost component with an at-the-money call option pay-o,
and this pay-o is positive only if the price rises. Thus, we see in Table 1 that any condition on the
optimal investment strategy is for up moves in price. Finally, the RNV approach typically (but
not always) gives an optimal time of investment that is never less than, and often greater than the
optimal investment time for a risk-neutral investor. We conclude that the use of RNV can lead to
very dierent option values, and a later timing for optimal investment, compared with the more
general solutions obtained using our methodology for risk-neutral investors.
4.2 Eect of Decision Frequency on Real Option Value
A real option value should not decrease when there is more exibility to make decisions over the
horizon of the option. If the decision never changes as a result of including more or less decision
nodes, the option value will remain unchanged. Otherwise, the option value should increase as
more decisions are allowed. Having fewer decision points places additional constraints on the
opportunity so it becomes less attractive to the decision maker and the value of the opportunity
to invest should decrease.
Table 2: Eect of number of decisions on real option value. Exponential utility, for dierent
and k. p
0
= $1m, T

= 5yrs, t = 1/12, T = T

kt, r = r = 5%, = 15%, = 50%.

= 0: invest at market price
0.2 0.4 0.6 0.8 1
k
12 109.5 1,132 3,712 8,197 14,480 645,167
6 142.2 1,344 4,309 9,212 16,089 645,167
3 163.9 1,471 4,618 9,803 17,022 645,167
1 176.9 1,553 4,828 10,185 17,624 645,167
= 1: xed strike with present value $1m
0.2 0.4 0.6 0.8 1
k
12 49,385 115,086 174,731 223,201 263,614 881,419
6 71,365 144,638 204,078 252,243 292,632 908,333
3 86,062 157,289 214,375 263,019 303,834 919,322
1 93,115 166,450 225,568 273,619 313,888 926,058
Table 2 quanties the eect of increasing the number of decision opportunities in a real option.
14
We again consider an opportunity to invest in an asset with no associated cash ows and current
market price $1m. The investor has an exponential utility and the real option is characterised by
the parameter values given in the legend to the table. Thus, there are 60 monthly steps in the
13
Here we only display results for k = 0.5, but a similar comment holds for other k with 0 < k < 1.
14
It is important that the trees are nested, i.e. no new decision nodes are inserted as their number decreases,
because only in this way does reducing the number of nodes capture the eect of placing additional constraints on
decision opportunities.
14
binomial tree for the market price, we place decision nodes every k steps in the backward induction
algorithm (13), and the last decision is at T = T

kt. For instance, if k = 12 the decision nodes

occur only once per year and the last decision is taken at the fourth year. So that the four decision
trees are nested the values considered for k are 12, 6, 3 and 1 representing decision opportunities
once per year and once every 6 months, 3 months and 1 month. The upper part of Table 2 reports
the value of the option to invest at market price and the lower part reports the value of the option
to invest at a xed cost K = $1m, for investors with dierent levels of risk tolerance . In both
cases the option values increase when more decision nodes occur in the tree, i.e. as k decreases.
The percentage increase in option value due to increased decision exibility is greatest for investors
with low , whereas the absolute increase in option value is greatest for high .
4.3 Which Utility?
The local absolute risk tolerance coecient is w
0
for general HARA utilities but just for the
exponential utility. So to compare HARA utility option values with exponential utility values we
set w
0
= $1m so that the initial risk tolerance is the same. Figure 1 graphs the exponential,
logarithmic, hyperbolic and power utility real option values as a function of , with 0 < 1, for
the options to invest (a) at market price and (b) at a xed time 0 cost equal to the current market
price. The current asset price is p
0
= $1m and the other real option characteristics are:
T

= 5yrs, K = $1m, g(x) = x, t = 1/12, k = 3, r = 5%, = 15%, = 50%. (18)

Figure 1: Comparison of invest option values under exponential, logarithmic, power and
hyperbolic utilities as a function of risk tolerance. Real option values on the vertical scale
have been multiplied by 1000 for clarity. Parameters are as in (18).
0.2 0.4 0.6 0.8 1
0
5
10
15
20
25
30
35

O
p
t
i
o
n

V
a
l
u
e
=0


Exponential Hyperbolic Power Logarithmic
0.2 0.4 0.6 0.8 1
0
100
200
300
400
500
=1

O
p
t
i
o
n

V
a
l
u
e


For typical risk tolerance (0 < < 1) the exponential and logarithmic utility values provide
lower and upper bounds for the real option values derived from other HARA utilities. For very high
risk tolerance, i.e. > 1, hyperbolic utility values still lie between the exponential and logarithmic
values, but the power utility values exceed the logarithmic values, and as increases further the
power values can become very large indeed because the risk tolerance increases extremely rapidly
with wealth.
Because of the boundary in (17) HARA utilities are not always well-behaved,unlike exponential
15
utilities that even yield analytic solutions in some cases. But exponential utility values will be too
low if the decision makers risk tolerance increases with wealth (generally regarded as a better
assumption than CARA). In that case, power utilities produce the most reliable real option values
for typical values of risk tolerance, but with very high risk tolerance logarithmic or hyperbolic
utility representations would be more appropriate, the former giving real option values that are
greater than the latter.
4.4 Eect of Asset Price on Option Value
The higher the underlying asset price at time 0 the more variable the terminal P&L. Its eect on
the option value depends on risk tolerance (and how it changes with wealth). To investigate this
we compute both exponential and logarithmic real option values, supposing the time 0 asset price
is either $0.1m, $1m or 10$m, xing the investors initial wealth at $1m and keeping all other real
option characteristics xed, as in (18).
Figure 2: Real option values under exponential and logarithmic utilities as a function of risk
tolerance , T

= 5, t = 1/12, k = 3, r = r = 5%, = 15%, = 50%, K = p

0
= $0.1, $1, $10m,
w
0
= 1m, 0.1 100, CE in $m. Both axes are in log
10
scale.
10
1
10
0
10
1
10
2
10
9
10
7
10
5
10
3
10
1
10
1

O
p
t
i
o
n

V
a
l
u
e
=0


10
1
10
0
10
1
10
2
10
3
10
3
10
2
10
1
10
0
10
1
=1

O
p
t
i
o
n

V
a
l
u
e


Exponential (p
0
=0.1)
Logarithmic (p
0
=0.1)
RNV (p
0
=0.1)
Exponential (p
0
=1)
Logarithmic (p
0
=1)
RNV (p
0
=1)
Exponential (p
0
=10)
Logarithmic (p
0
=10)
RNV (p
0
=10)
Figure 2 displays the results for dierent values of with both option value and represented
on a base 10 log scale. We depict option values for = 0 (invest at market price) and = 1 (xed
ATM strike at p
0
). The values for the high-priced asset, represented by red lines, are most sensitive
to and the values for the low-priced asset, represented by black lines, are least sensitive to .
Usually, the smaller (greater) the risk tolerance of the investor, the higher he ranks the option to
invest in the relatively low-priced (high-priced) asset, given that the asset-price dynamics follow the
same GBM process. In each case the option value converges to the value obtained for a risk-neutral
investor as , and this value increases with p
0
. As 0 the option value decreases with p
0
,
except for an investor in a xed-strike option with an exponential utility. For intermediate values
of lambda there is no general rule for ranking, especially when invest cost is linked to market price.
Hence, risk-neutral investors rank investments according to the underlying price, ceteris paribus,
but this need not be the case risk-averse investors.
16
The RNV option price is is zero when K
t
= p

t
( = 0) and when K
t
is xed at $1m ( = 1)
it is marked by the dotted horizontal lines in Figure 2. In this case it is less than the risk-neutral
subjective value because > r, but would exceed that value if < r. The exponential option
values (solid lines) never exceed the logarithmic utility values (dashed lines) for the same initial
risk tolerance, and they are much lower for the xed-strike option to invest in a high-priced asset
by a highly risk-averse investor.
4.5 Sensitivity to and
When the decision maker has low condence in his views about the price process his subjective
values for and may be highly uncertain. We describe the sensitivity of real option values
to the expected returns and risks of the investment. The RNV principle yield option prices that
increase with volatility under the xed time 0 cost assumption, yet the DCF approach implies the
opposite. In our approach option values can decrease or increase with volatility depending on the
cost structure and the utility.
Figure 3: Value of an investment option under exponential and logarithmic utilities as a
function of the investors subjective views on expected return and volatility , for = 0.2.
CE value in $m for p
0
= $1m, w
0
= $10m, r = 5%, T

= 5, t = 1/12, k = 3.
5%
10%
15%
20%
25%
10%
20%
30%
40%
50%
0
0.5
1
1.5

=1

O
p
t
i
o
n

V
a
l
u
e
5%
10%
15%
20%
25%
10%
20%
30%
40%
50%
0
0.5
1
1.5

=0

O
p
t
i
o
n

V
a
l
u
e
Exponential Logarithmic
Figure 3 depicts the values of a real option to invest as a function of these expected return and
volatility of the GBM price process, with other parameters xed as stated in the legend. When
= 0 the option value always decreases as uncertainty increases, for any given expected return, due
to the risk aversion of the decision maker. When there is high uncertainty ( greater than about
30%) the exponential utility values are zero, i.e. the investment opportunity is valueless as the
price would never fall far enough (in the decision makers opinion) for investment to be protable.
By contrast, the logarithmic utility always yields a positive value provided the expected return is
greater than about 10%, but again the investor becomes more likely to defer investment as the
volatility increases. Indeed, the option values are monotonically decreasing as for every , and
monotonically increasing with for every . Volatility sensitivity can be dierent for the xed-
strike option, = 1. For instance, with the logarithmic utility the option values can decrease as
17
increases, when the expected return is low. In particular, when = r the logarithmic option
values increase monotonically with volatility, as they also do in the RNV case.
The sensitivity of the option value to and also decreases as risk tolerance increases. To
illustrate this we give a simple numerical example, for the case = 0 and an exponential utility.
Suppose = 20%. If = 0.2 the option value is $1, 148 when = 40% and $241, 868 (almost 211
times larger) when = 20%. When = 0.8 the option value is $167, 716 when = 40%; but now
when = 20% it only increases by a multiple of about 4, to $713, 812. Similarly, xing = 20%
but now decreasing from 20% to 10% the value changes from $241, 867 to $109 (2, 210 times
smaller) for = 0.2, but from $713, 812 to $113, 959 (only about 6 times smaller) when = 0.8.
Hence, the option values sensitivities to and are much greater for low levels of risk tolerance.
Similar eects are present with the logarithmic utility but they are much less pronounced.
5 Cash Flows and non-GBM Price Processes
Given the generality and exibility of our approach it has applications to a wide range of invest-
ment or divestment decisions. Here we use examples of decision problems commonly encountered
by private real-estate companies or housing trusts, from large-scale land development to individual
residential property transactions. Many real estate real options have been considered in the litera-
ture, including: the option to abandon, e.g. Smith (2005), the option to defer a land development,
e.g. Brand ao and Dyer (2005) and Brand ao et al. (2005), and the option to divest, e.g. Brand ao
et al. (2008) and Smith (2005). But all these papers employ the RNV approach.
Our model permits risk-averse private companies, publicly-funded entities, or individuals to
compute a real option value that is tailored to the decision maker, and which could be very
dierent from the risk-neutral price obtained under the standard but (typically) invalid assumption
of perfectly-hedgeable risk and xed costs. This is signicant because such decisions can have
profound implications for the decision makers economic welfare. For instance, an individuals
investment in housing may represent a major component of his wealth and should not be viewed
simply in expected net present value terms, nor should all uncertainties be based on systematic risk
because they are largely unhedgeable. Private companies typically generate returns and risks that
have a utility value that is specic to the owners outlook. Similarly, charitable and publicly-funded
entities may have objectives that are far removed from wealth maximisation under risk-neutrality.
First we consider an option to purchase a residential property when the investors views are
captured by a process (3) where a long period of property price momentum could be followed by
a crash. Then we analyse real options for investors that believe in a mean-reverting price process
(4). After this we focus on the inclusion of cash ows, considering both positive cash ows for
modelling buy-to-let real options and negative cash ows for modelling construction costs in an
on-going development.
5.1 Property Price Recessions and Booms
Many property markets are subject to booms and recessions. For example, the average annualised
return computed from monthly data on the Vanguard REIT exchanged traded fund (VQN) from
18
January 2005 to December 2006 was 21% with a volatility of 15%. However, from January 2007 to
December 2009 the property market crashed, and the VQN had an average annualised return of
13% with a volatility of 58%. But from January 2010 to June 2011 its average annualised return
was 22% with volatility 24%. Clearly, when the investment horizon is several years a property
investor may wish to take account of both booms and busts in his views about expected returns.
We now give a numerical example of an option to purchase a residential property, i.e. with zero
cash ows, under such scenarios.
Consider a simple boom-bust scenario over a 10 year horizon. The expected return is negative,

1
< 0 for the rst n years and positive,
2
> 0 for the remaining 10n years. Following the above
observations about VQN we set
1
= 10%,
1
= 50% and
2
= 10%,
2
= 30%. We suppose
decisions are taken every quarter with t = 0.25 and set r = 5% in the price evolution tree. The
real option values given in Table 3 are for investors having exponential utility, with varying levels
of risk tolerance between 0.2 and 1. The property price recession is believed to last n = 0, 2, 4, 6, 8
or 10 years.
15
Table 3: Eect of a time-varying drift for the market price, with downward trending price for
rst n years followed by upward trending price for remaining 10 n years. Exponential utility
with dierent levels of risk tolerance, with = corresponding to the risk-neutral (linear
utility) value. Real option values in bold are the maximum values, for given . p
0
= $1 million,
w
0
= $1 million, T

= 10, t = 0.25, T = T

t, r = 5%,
1
= 10%,
2
= 10%,
1
= 50% and

2
= 20%.
= 0
n 0 2 4 6 8 10

0.2 866 382,918 727,922 438,482 174,637 0

0.8 211,457 1,230,035 2,297,909 1,148,695 370,354 0
1.0 266,752 1,349,295 2,552,434 1,300,724 406,567 0
648,173 2,234,361 4,777,950 5,308,672 1,045,800 0
= 1
n 0 2 4 6 8 10

0.2 164,721 391,060 552,578 194,371 51,040 5,022

0.8 392,866 961,279 1,704,007 692,529 159,663 14,943
1.0 432,280 1,064,416 1,938,157 821,733 186,257 17,191
740,493 1,989,551 4,339,659 4,860,376 841,706 50,787
When n = 0 the investor expects the boom to last the entire period, but since
2
= 30% there
is still uncertainty about the evolution of the market price, and with
2
= 10% the price might still
fall. The case n = 10 corresponds to the view that the market price could fall by
1
= 10% each
year for the entire 10 years, but with
2
= 50% this view is held with considerable uncertainty.
At these two extreme values for n we have a standard GBM price process, so for any given
the invest-at-market-price option has a lower value than the xed-strike option. However, for
intermediate values of n the invest-at-market-price option often has a higher value than the xed-
strike option. This ordering becomes more pronounced as n increases, because the investment cost
15
Results for the divest option, or for the invest option under dierent utilities are not presented, for brevity, but
are available on request. The qualitative conclusions are similar.
19
decreases when = 0 and, provided that the price rises after the property is purchased, the prots
would be greater than they are under the xed-cost option. For risk-averse investors the maximum
value arises when the length of the boom and bust periods are approximately the same. However,
a risk-neutral investor would place the greatest value on a property location where the recession
period lasts longer (n = 6 in our example).
5.2 Property Price Mean-Reversion
Momentum is not the only well-documented eects in property prices; they may also mean-revert
over a relatively short horizon. We now investigate how an investors belief in price mean-reversion
would inuence his value of the option to invest in a residential property. We suppose the market
price follows the OU process (4) and we set p = p
0
for simplicity. To model the eect of mean
reversion on the option value we employ the NR parameterisation (5), allowing to vary between
0 and 0.1, the case = 0 corresponding to GBM and = 0.1 giving the fastest characteristic
time to mean revert of 10 time-steps (so assuming these are quarterly this represents 2.5 years).
16
The other parameters are xed, as stated in the legend to Figure 4, which displays the real option
values for = 0 and = 1 as a function of .
Figure 4: Comparison of option values under HARA utilities with respect to mean-reversion
rate . w
0
= $1m, r = 5%, k = 1 and = 0.4 T

= 10, t = 1/4, K = p
0
= $1m, T = T

t,
= 40%. Characteristic time to mean-revert = t/ in years, e.g. with t = 1/4 then
= 0.02 = 12.5yrs, = 0.1 = 2.5yrs.
0 0.02 0.04 0.06 0.08 0.1
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
x 10
5

O
p
t
i
o
n

V
a
l
u
e
=1


0 0.02 0.04 0.06 0.08 0.1
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
x 10
4

O
p
t
i
o
n

V
a
l
u
e
=0


Exponential Hyperbolic Power Logarithmic
Increasing the speed of mean-reversion has a similar eect to decreasing volatility. Hence, xed-
cost option values can decrease with , as they do on the right graph of Figure 4 ( = 1) especially
when the investor has logarithmic utility. By contrast, the invest-at-market-price ( = 0) option
displays values that increase with . Note that for xed these values could increase with , due
to the positive eect of in the local drift (5d).
17
16
Lower values for have slower mean-reversion, e.g. = 0.02 corresponds to a characteristic time to mean revert
of = 0.02
1
/4 = 12.5 years if time-steps are quarterly.
17
This eect is only evident for values of below a certain bound, depending on the utility function and other real
option parameters. For the parameter choice of Figure 4 the invest-at-market-price option values have their usual
20
Now consider a risk-neutral investor who wishes to rank two property investment opportunities,
A and B. Both have mean-reverting price processes but dierent and : property A has a
relatively rapid mean-reversion in its price ( = 1/10, = 2.5 years) with a low volatility ( = 20%)
and property B has a relatively slow mean-reversion in its price ( = 1/40, = 10 years) with a
higher volatility ( = 40%). Such an investor would have a great preference for property B since
he is indierent to the high price risk and has regard only for the local drift. Given the slow rate of
mean-reversion, he sees only the possibility of a sharp fall in price at which point he would invest
followed by a long upward trend in its price. Indeed, assuming p
0
= p = $1m the risk-neutral
value of option B is $797,486, but the corresponding value for option A is only $98,077.
However, once we relax the risk-neutral assumption the ranking of these two investment oppor-
tunities may change, depending on the utility of the investor. The real option values for risk-averse
investors with high risk tolerance ( = 0.8) or low risk tolerance ( = 0.4) are shown in Table
4. In each case the greater option value is depicted in bold. This shows that an investor with a
logarithmic utility would also prefer property B, as would an investor with a relatively high risk
tolerance ( = 0.8) and a hyperbolic or a power utility. In all other cases the investor would prefer
property A.
Table 4: Comparison of invest option values for property A ( = 1/10, = 20%) and proporty
B ( = 1/40, = 40%) with = 0.4 or 0.8 . All other parameter values are the same as in Figure
4. For each utility we highlight the preferred property in bold.
0.4 0.8
A B A B A B
Exponential 30,421 12,952 52,732 43,947
98,077 797,486
Hyperbolic 33,526 17,563 55,230 58,909
Power 33,045 16,651 56,365 68,051
Logarithmic 19,890 21,260 56,986 73,131
5.3 Positive Cash Flows: Buy-to-Let Options
Short-horizon decision trees for the invest and divest options on an investment that yields positive
cash ows are depicted in Figures 5 and 6. A typical real-esate investment with regular positive
cash ows is a buy-to-let residential property or oce block, or a property such as a car park
where fees accrue to its owner for its usage. Rents, denoted x
s(t)
in the trees, are captured using
a positive dividend yield dened by (6) that may vary over time. Even a constant dividend yield
would capture rents that increase/decrease in line with the market price. We suppose cash ows
not re-invested, otherwise we could employ the cum-dividend price approach that has previously
been considered. Instead, cash ows are assumed to earn the risk-free rate, as to suppose they
are invested in another risky project would introduce an additional source of uncertainty which is
beyond the scope of this paper.
Each time a cash ow is paid the market price jumps down from p
+
t
to p

t
= p
+
t
x
t
. Between
negative sensitivity to once exceeds approximately 2, where the characteristic time to mean revert is 1/8th of a
year or less. Detailed results are not reported for lack of space, but are available from the authors on request.
21
payments the decision maker expects the discounted market price to grow at rate r, and based
on the discretisation (2) we have p
+
t+1
= up

t
or p
+
t+1
= dp

t
with equal probability. The terminal
nodes of the tree are associated with the increment in wealth w w
0
where the nal wealth w is
given (8) for the option to invest in Figure 5, and by (10) for the option to divest in Figure 6, now
setting CF= x.
The decision tree in Figure 5 is now used to rank the options to buy two dierent buy-to-let
properties. Both properties have current market value p
0
= 1, the initial wealth of the investor is
w
0
= 1 and the risk-free rate r = 5%. In each case rents are paid every six months, and are set
at a constant percentage of the market price at the time the rent is paid. But the investor has
dierent views about the future market price and rents on each property, as specied in the rst
pair of columns in Table 5. Similarly, Figure 6 is used to rank the options to sell two dierent buy-
to-let properties, with views on market prices and rents as specied in the second pair of columns
in Table 5. Note that the CE values given for the divest option include the expected utility value
of capital gains on the property itself, as well as the expected utility value of the opportunity to
divest. In each case the value of the preferred property is marked in bold.
Table 5: Columns 2 and 3 compare the values of 2 real options, each to purchase a buy-to-let
property based on the decision tree shown in Figure 5. Columns 4 and 5 compare the values
of 2 real options, each for buy-to-let property with the option to sell, based on the decision
tree in Figure 6. For the decision maker, in each case = 0.4, r = 5%, w
0
= $1 million and for
each property p
0
= $1 million. The decision makers beliefs about , and depend on the
propertys location. For each utility we highlight in bold the preferred location for buying (or
selling) the property.
Invest Divest
40% 25% 30% 20%
15% 10% 15% 10%
10% 10% 20% 10%
Exponential 289 316 6,775 6,052
Hyperbolic 348 339 6,610 6,103
Power 340 336 7,283 6,296
Logarithmic 358 345 4,286 5,330
Again the form assumed for the utility function investor has material consequences for decision
making. An investor with exponential (CARA) utility would prefer the option to buy the second
property, whilst investors with any of the other HARA utilities having the same risk tolerance at
time 0 would prefer the option to buy the rst property. Similarly, regarding the divest real option
values of two other properties, both currently owned by the decision maker, all investors except
those with a logarithmic utility would favour selling the rst property.
5.4 Negative Cash Flows: Buy-to-Develop Options
Setting a negative dividend yield is a straightforward way to capture cash that is paid into the
land or property to cover development costs. But there are other important dierences between
the buy-to-develop and buy-to-rent option above. In the buy-to-develop case there are no cash
22
ows until the land or property is purchased, and thereafter these cash ows are included in the
market price. Hence, the market price following an invest decision is cum-dividend and prior to
this the market price evolves as in the zero cash ow case. Also, now the investment horizon T

is
path dependent because it depends on the time of the investment (e.g. it takes 2 years to develop
the property after purchasing the land).
A simple decision tree for the buy-to-develop option is depicted in Figure 7, in which the
development cost is y
s(t)
> 0 and p
t
is the (cum-dividend) market price. The option maturity T
is 2 periods, and so is the development time, so T

varies from 2 to 4 periods depending on the

time of investment. To keep the tree simple we suppose that development costs are paid only once,
after 1 period, to allow for planning time. For example, consider the node labelled D
u
that arises
if the investor does not purchase the land or property at t = 0 and subsequently the market price
moves up at t = 1. A decision to invest at this time leads to four possible P&Ls. For instance,
following the dotted red lines, if the price moves up again at t = 2 the development cost at this
time is y
uu
, based on the market price of uup
0
. But if the price subsequently moves down at t = 3,
the terminal value of the property is p
1,uud
= d(uup
0
+y
uu
) and the costs are the sum of the price
paid for the land and the development cost, i.e. up
0
+y
uu
.
Table 6: Value comparison of real options to buy two dierent properties for development
based on decision tree shown in Figure 7. Here = 0.4, r = 5% and w
0
= $1 million. Each
property has p
0
= $1 million but the decision makers views on , and development costs
dier for each property as shown in the table. The preferred option is indicated by the value
in bold.
A B
25% 15%
10% 35%
20% 40%
Exponential 289 182
Hyperbolic 299 326
Power 315 350
Logarithmic 41 0
Table 6 displays some numerical results for the decision tree in Figure 7, reporting the value of
two options to buy-to-develop land, each with initial market price $1 million and r = 5%, but the
options have dierent , and development costs . For each option we suppose the development
takes one year in total with the costs paid six months after purchase. The investors all have initial
wealth $10 million and = 0.4.
In general, a higher development cost for given and decreases the buy-to-develop option
value, and the investor becomes more likely to defer investment until the market price falls. But
the option value also increases with and decreases with . We nd that option A is preferred by
an investor with an exponential or logarithmic utility whereas option B is preferred by an investor
with a hyperbolic or a power utility. Hence, dierent investors that have identical wealth, share the
same initial risk tolerance, and hold the same views about development costs and the evolution of
market prices may again rank the values of two land-development options dierently, just because
23
their risk tolerance has dierent sensitivity to changes in wealth.
6 Summary and Conclusion
This paper introduces a general and exible decision-tree framework for valuing real options which
can encompass a great variety of real-world applications. In addition to the predetermined strike
real options that are tradable in complete markets the RNV assumption that is most commonly
employed in the literature we model options to invest at a cost that has both xed and variable
components. The variable component is determined by (but not necessarily perfectly correlated
with) the market price of the investment asset, and the market for this asset need not be complete.
The decision maker may be risk-neutral or have a utility in the HARA class, and his belief in the
evolution of the market price is not constrained to be a standard GBM. Many price processes are
possible and we have used real-estate investments and divestments to illustrate our model when the
price is believed to follow a regime-switching GBM or a mean-reversion, and also when investments
have positive or negative cash ows.
The real option price that is obtained using standard RNV assumptions can be very much
greater than or less than the value that would be found using a more realistic assumptions about
investment costs in an incomplete market, and very often the RNV approach would specify a
later investment time. After demonstrating this we answer several important questions relating
to real options that have not previously been addressed, nding: (i) The assumption about the
investment cost, whether it is predetermined or stochastic, has a signicant inuence on the real
option value. The predetermined-strike assumption can signicantly over-estimate the value of a
real option when the more-appropriate assumption is that the investment cost is positively related
to the market price, or has both a xed and variable components; (ii) It is important to account for
the exibility of the decision-making process in real option analysis because the real option value
increases with the frequency of decision opportunities; (iii) The price of the investment relative to
the decision makers wealth matters: the smaller (greater) the risk tolerance of the investor, the
higher he ranks the option to invest in a relatively low-priced (high-priced) asset, given that the
asset-price dynamics follow the same GBM process; (iv) Under the GBM assumption the sensitivity
of a xed-strike real option to the volatility of the underlying asset price may be positive, as in the
RNV approach, whereas the invest-at-market-price real option always has a negative sensitivity
to the asset-price volatility; (v) The decision makers ranking of dierent real options depends on
how his risk tolerance changes with wealth.
References
Benaroch, M., R. J. Kauman. 2000. Justifying electronic banking network expansion using real
options analysis. MIS Quarterly 24 197225.
Boer, P. F. 2000. Valuation of technology using real options. Research Technology Management
43 2630.
Bowman, E. H., G. T. Moskowitz. 2001. Real options analysis and strategic decision making.
Organization Science 12 772777.
24
Brand ao, L. E., J. S. Dyer. 2005. Decision analysis and real options: A discrete time approach to
real option valuation. Annals of Operations Research 135(1) 2139.
Brand ao, L. E., J. S. Dyer, W. J. Hahn. 2005. Using binomial decision trees to solve real-option
valuation problems. Decision Analysis 2 6988.
Brand ao, L. E., J. S. Dyer, W. J. Hahn. 2008. Response to comments on brandao et al. (2005).
Decision Analysis 2 103109.
Capozza, D. R., G. A. Sick. 1991. Valuing long-term leases: the option to redevelop. Journal of
Real Estate Economics and Finance 4 209233.
Chance, D. M. 2008. Convergence of the Binomial to the Black-Scholes Model . E. J. Ourso College
of Business, Lousiana State University.
Copeland, T., V. Antikarov. 2001. Real Options: A Practitioners Guide. Texere, New York.
Copeland, T., T. Koller, J. Murrin. 1990. Valuation: Measuring and Managing the Value of
Companies. John Wiley & Sons.
Cox, J. C., S. A. Ross, M. Rubinstein. 1979. Option pricing: A simplied approach. Journal of
Financial Economics 7(3) 229263.
Davis, M. H. A., Y. Kabanov, R. Lipster, J. Stoyanov. 2006. Optimal hedging with basis risk.
Stochastic Calculus to Mathematical Finance: The Shiryaev Festschrift 1 169187.
Evans, J., V. Henderson, D. G. Hobson. 2008. Optimal timing for an indivisible asset sale. Math-
ematical Finance 18(4) 545567.
Grasselli, M. R. 2011. Getting real with real options: A utility-based approach for nite-time
investment in incomplete markets. Journal of Business Finance and Accounting 38(5-6) 740
764.
Grenadier, S. R. 1996. The strategic exercise of options: development cascades and overbuilding
in real estate markets. The Journal of Finance 51 16531679.
Harrison, J. M., D. M. Kreps. 1981. Martingales and arbitrage in multiperiod securities markets.
Journal of Economic Theory 20 381408.
Henderson, V. 2002. Valuation of claims on nontraded assets using utility maximization. Mathe-
matical Finance 12(4) 351373.
Henderson, V. 2007. Valuing the option to invest in an incomplete market. Mathematics and
Financial Economics 1(2) 103128.
Henderson, V., D. G. Hobson. 2002. Real options with constant relative risk aversion. Journal of
Economic Dynamics and Control 27(2) 329355.
Jarrow, R., A. Rudd. 1982. Approximate option valuation for arbitrary stochastic processes.
Journal of Financial Economics 10 347369.
Kasanen, E., L. Trigeorgis. 1995. Real Options in Capital Investment: Models, Strategies, and
Applications. 47-68. Praeger.
Keeney, R., H. Reia. 1993. Decisions with Multiple Objectives. Cambridge University Press, New
York. Original edition published by Wiley, New York, 1976.
25
Kogut, B. 1991. Joint ventures and the option to expand and acquire. Management Science 37
1933.
Mason, S. P., R. C. Merton. 1985. Recent Advances in Corporate Finance. 7-54. Irwin, Boston,
MA.
McDonald, R., D. Siegel. 1986. The value of waiting to invest. Quarterly Journal of Economics
101(4) 707727.
Merton, R. C. 1969. Lifetime portfolio selection under uncertainty: the continuous time case.
Review of Economics and Statistics 51 247257.
Merton, R. C. 1971. Optimum consumption and portfolio rules in a continuous-time model. Journal
of Economic Theory 3 373413.
Miao, J., N. Wang. 2007. Investment, consumption, and hedging under incomplete markets.
Journal of Financial Economics 86(3) 608642.
Mossin, J. 1968. Optimal multiperiod portfolio policies. Journal of Business 41 215229.
Myers, S. C. 1977. Determinantes of corporate borrowing. Journal of Financial Economics 5(2)
147175.
Nelson, D. B., K. Ramaswamy. 1990. Simple binomial processes as diusion approximations in
nancial models. The Review of Financial Studies 3 393430.
Panayi, S., L. Trigeorgis. 1998. Multi-stage real options: the cases of information technology in-
frastructure and international bank expansion. The Quarterly Review of Economics and Finance
38 675682.
Patel, K., D. Paxson, T. F. Sing. 2005. A review of the practical uses of real property options.
RICS Research Paper Series, 5:1.
Quigg, L. 1993. Empirical testing of real option-pricing models. The Journal of Finance 48
621640.
Shackleton, M. B., R. Wojakowski. 2007. Finite maturity caps and oors on continuous ows.
Journal of Economic Dynamics and Control 31(12) 38433859.
Smit, H. T. J. 1996. The valuation of oshore concessions in the netherlands. Financial Manage-
ment 26(2) 517.
Smit, H. T. J., L. A. Ankum. 1993. A real option and game-theoretic approach to corporate
investment strategy under competition. Financial Management 22(3) 241250.
Smith, J. E. 2005. Alternative approaches for solving real-options problems. Deicison Analysis
2(2) 89102.
Smith, J. E., K. F. McCardle. 1998. Valuing oil properties: Integrating option pricing and decision
analysis approaches. Operations Research 46(2) 198217.
Smith, J. E., R. F. Nau. 1995. Valuing risky projects - option pricing theory and decision-analysis.
Management Science 41(5) 795816.
Triantis, A. J., J. E. Hodder. 1990. Valuing exibility as a complex option. The Journal of Finance
45 549565.
26
Trigeorgis, L. 1993. The nature of option interactions and the valuation of investments with
multiple real options. The Journal of Financial and Quantitative Analysis 28 120.
Trigeorgis, L. 1996. Real Options: Managerial Flexibility and Strategy in Resource Allocation. The
MIT Press, Cambridge, MA.
Trigeorgis, L., P. Mason. 1987. Valuing managerial exibility. Midland Corporate Finance Journal
5 1421.
Yeo, K. T., F. Qiu. 2002. The value of management exibility - a real option approach to investment
evaluation. International Journal of Project Management 21 243250.
27
Figure 5: Decision tree depicting the option to invest in a property that pays rents, x
s(t)
.
T

= 3, T = 2, k = 1. Terminal nodes labelled with P&L (ww

0
), given by (8) following decision
to invest (I) and 0 otherwise, with CF = x.
28
Figure 6: Decision tree depicting the option to sell a property that pays rents, x
s(t)
. T

=
3, T = 2, k = 1. Terminal nodes labelled with P&L (w w
0
), given by (10) with CF = x if the
owner remains invested (R), or if the owner sells the property (S) by the dierence between
the selling price and initial price.
29
Figure 7: Decision tree depicting the option to invest in land for development. If the land
is acquired (I) the development takes 2 periods and development costs occur only after the
rst period. Terminal nodes are associated with the P&L, ww
0
resulting from the decision.
30