Professional Documents
Culture Documents
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to Journal
of Economic Literature.
http://www.jstor.org
Under
Investment
Uncertainty:
Open
Keeping
One's
Options
By R. GLENN HUBBARD
Columbia University and the National Bureau of Economic Research
I am grateful to Avinash Dixit, Mark Gertler, Kevin Hassett,
Charles Himmelberg, Anil Kashyap, Gilbert Metcalf, and Robert
Pindyck for helpful comments and suggestions, and to the Federal
Reserve Bank of New York and the John M. Olin Visiting Professorship at the Center for the Study of the Economy and the State
of the University of Chicago for support.
INTELLECTUALatten-
tion has been focused on models of derivative financial instruments, including options. The theoretical intuition embodied in
early option pricing models has been applied
to other financial decisions by individuals and
businesses as well. Despite these applications, options as financial instruments are not
central to the lives of most of us. However, a
wide variety of options is: We all face significant choices about whether we should spend
our resources today or wait, thereby "keeping
our options open."
In an important new book, Dixit and Pindyck (1994) illustrate how modern "options"
intuition can be used to analyze a number of
individual and business decisions. Focusing
their attention on investment decisions, they
offer both a cogent methodological discussion
of a new view of investment theory (to which
they are individually major contributors) and
Investment Under Uncertainty. By AVINASH K.
DIXIT AND ROBERT S. PINDYCK.Princeton: Prince-
1816
1817
ment model suggests that a firm should increase its capital stock when the market value
of the capital assets exceeds their replacement cost. Tobin's q, the ratio of the market
valuation of capital to its replacement cost, is
a convenient and often-used summary statistic in the neoclassical approach. In the new
view, the investment decision is a choice to
incur a sunk cost (because investment is assumed to be irreversible), a choice which
yields uncertain future returns. Because the
firm has the option of delaying the investment, it should increase its capital stock only
if q exceeds unity by a margin sufficient to
compensate the firm for the loss of the option
to delay. Dixit and Pindyck show that, under
certain assumptions, threshold values of q are
quite large, implying very high "hurdle rates"
consistent with those found in interviews with
managers responsible for capital budgeting
(see Lawrence Summers 1987).
The basic model introduced in Chapter 5
relies on a simplifying assumption of complete irreversibility of capital investments. In
fact, firms have a way out; they can temporarily or permanently scrap sufficiently unprofitable projects. Chapter 6 presents an extension to the case of "suspension":A firm may
discontinue a project's operation with the option of renewing it in the future. Hence, two
options are part of the investment decision,
the initial option of delay and a sequence of
operating options (related to suspension decisions). Chapter 7 pursues this extension for
the case of "abandonment,"which entails the
extinguishing of the option of restoring the
project at some point. In practice, of course,
firms typically must choose between suspension (with ongoing, say, maintenance expenditures) and abandonment (with costs of
severance and the lost option). Dixit and Pindyck describe analytically the choice among a
firm's operating, suspending, or abandoning a
project, then make their points more concretely in an illustration of investment decisions in crude oil tankers.
While Chapters 5-7 present the book's
core material on real options in firm decisions, they beg the question of how an industry equilibrium (at least outside of a simple
monopoly case) might be characterized in the
presence of these options. This important
1818
summarized in Investment Under Uncertainty, let me set the stage by reviewing the
basic predictions of the traditional neoclassical model.3 Neoclassical models of invest3 I, intend the term "neoclassical model" to refer
both to the models I discuss herein and to standard net present value rules frequently taught in
business schools (see, e.g., Richard Brealey and
Stewart Myers 1991).
1819
the q theory yields a testable model of investment when dynamics (e.g., adjustment costs)
are imposed.
Both variants of the neoclassical model rely
on the net present value rule. A firm should
undertake investment projects with positive
net present value. They make two subtle assumption as well: First, invested capital can
be sold easily to other users (that is, it is reversible). Second, each investment opportunity facing the firm is a once-and-for-all opportunity; if the firm declines the project, it
will never have the choice to reconsider.
The starting point for the "new view"
stressed by Dixit and Pindyck is that many
real-world investment decisions violate these
subtle assumptions, and irreversibility and a
chance for delay are important considerations. This importance reflects the observation that the possibility of delay gives rise to a
call option: The firm has the right, though
not the obligation, to buy an asset (the investment project) at some future time at its discretion. To the extent that investment is irreversible-a
feature of the new view
models-making an investment extinguishes
the value of the call option, or "real option,"
in the terms of Dixit and Pindyck. (If investment were reversible, the neoclassical
model's guidance would be more applicable.)
The value of the lost option is a component of
the opportunity cost of investment. In the
terminology of the Tobin's q approach, the
threshold criterion for investment requires
that q exceed unity by the value of maintaining the call option to invest. Indeed, this additional component may account for the high
"hurdle rates" required by corporate managers actually making investment decisions.
An easy, though not satisfying, response to
this argument is that the neoclassical model
could be modified to incorporate the real option component. Even with this semantic
change, it is still necessary, to the extent that
the real option is valuable, to analyze how it
might be priced in firms' decisions.
1820
cepts of irreversibility and delay. The significance of each is, of course, an empirical question. The illustrations of irreversibility (or,
equivalently, sunk costs) are not hard to provide in many cases, including firm-specific
marketing costs. The costs of even plant and
equipment investments may be sunk. This is
easy to see in the case of industry-specific
capital when aggregate uncertainty is important. Even less specific capital such as computers or general use machinery are (at least
partially) irreversible if secondary markets
are inefficient. Such markets may be inefficient because of classic "lemons"problems of
adverse selection or because regulatory or institutional impediments. As with irreversibility, a value to "delay"is also plausible
on a priori grounds. While delay entails possible costs (say, lost returns from a project or
entry by competitors), it confers benefits in
the form of new information about the project's value during the period of delay.
McDonald and Siegel use a process of geometric Brownian motion with drift:
dV=aVdt+aVdz,
(1)
where dt is a time increment, dz is the increment of a Wiener process, and ot and a are
constants. The process for V describes the arrival of new information over time and is consistent with an uncertain future value of V.
The analytical advantage of this setup is
clear from the perspective of the new view.
The investment opportunity considered by
the firm is a call option with no expiration
date; that is, the firm has the right, though
not the obligation to undertake the project at
a prespecified price 1. At the same time, this
simple setup has some drawbacks in describing typical investment projects. For example,
if there are variable costs and the firm can
close the factory in periods when price is below variable cost, the process for V is not geos Predecessors of this formalization in the macroeconomics literature include the models of Alex
Cukierman (1980) and Ben Bernanke (1983).
6 For a more detailed description of stochastic
processes and their properties, see Dixit and Pindyck (1994, Ch. 3), Dixit (1993), Darrell Duffie
(1988), and Samuel Karlin and Howard Taylor
(1975, 1981).
1821
(2)
(3)
1822
1/2a2 V2
a2 V2F"(V)+ aVF'(V)- pF = 0.
(4)
While the Bellman equation is a second-order differential equation, F(V) must also satisfy three boundary conditions. First, given
the stochastic process used, V will remain at
zero if it goes to zero (hence, F(M)= 0). Second, if V4 is the value at which it is optimal to
invest, upon investing the firm receives a net
payoffV - I (hence,F(V0)= V' - I); see Dixit
and Pindyck (1994) or Dixit (1993) for a discussion. The third condition, known as the
"smooth-pasting"
conditionrequiresF' (V4) = 1,
so that F(V) is continuous at the threshold of
V4. The need for a third condition arises because of the existence of a "free boundary."
While the first boundary is given by V = 0,
the second boundary position V4 must be determined as part of the solution. As Dixit and
Pindyck show, such "free boundary" problems occur in many applications.
The solution to the McDonald-Siegel problem is relatively straightforward. One can
guess a functional form and verify its success.
Given the first boundary condition mentioned above, the solution must assume the
form:
F(V) = AV1,
(5)
The basic intuitive point is this: Irreversibility and the possibility of delay generate a range of inaction (not present in the
neoclassical model) in which V> I, yet the
firm does not invest. (This is because,
pi>1,PiA/pi - ) > l, and V* > L)
How large is this wedge between the traditional net present value investment criterion
and that in the new view? Without going into
algebraic detail, let me indicate some channels. First, as uncertainty about future returns (measured by a) rises, the wedge
1/(P1 - 1) also rises. Second, ceteris paribus,
an increase in the discount rate p increases
the wedge. Third, given p, an increase in
trend value growth a increases the wedge.
To compare these results with the predictions of the neoclassical model, it is useful to
revisit predictions of the q approach. If we
define q as the ratio of the expected value of
profits from an investment-conditional on
its completion-to its construction cost, then
q = V/I. A firm should invest when V > I. If
we acknowledge the loss of the option to invest when the project is completed, the installed project should raise firm value by
V - F(V), not by V. While in this alternative
setup the threshold criterion for investment
is V> I + F(V), the threshold q expressed in
the more conventional q terms defined earlier is 01/(Pl - 1) > 1. That is, fluctuations in V
can still yield a region of inaction in which q
exceeds unity with no investment response.1
11 In some respects, the wedge between the
investment criteria in the neoclassical model and
the new view is likely to be at its upper bound in
geometric Brownian motion examples in which defay can be arbitrarily long. An alternative model
(also considered by Dixit and Pindyck) would re-
wheredq,anincrement
ofaPoissonprocesswithamean
arrivalrateX, is independent
of dz. Whenthe "event"
(ofpatentexpiration,
entry,etc.)occurs,q declineswith
probability
one by somefraction4 (betweenzeroand
one). Intuitively,
while V fluctuates(withgeometric
Brownianmotion),overeachinterval(dt),thereis a
probability
( Adt)thatitwilldropafraction
oftheoriginal
value(1 - 4) andcontinuefluctuating
(withgeometric
(6)
1823
a)T/(p -
a).
V(P)=
f Xe
TP[l -
e(p-a)T]/(p
a)dT
=P/(X+p-a).
(7)
Intuitively, though the project can be analyzed as infinitely lived, future profits are discounted at a higher rate reflecting the probability of "death."
1824
- 1)] (p - a + X)1,
(8)
where "p- a + X"replaces "p- a." Depreciation per se does not affect the option value in
this case. That is, the threshold investment
criterion is V*= I/(PI- 1) T > I, where PI is
as before. This is because the firm's opportunity to invest is assumed to be available only
once; hence, exercising the option is irreversible even though the project has a finite
life.
The option value can be reduced, however,
if the option to invest is available in perpetuity (that is, when the firm has the right to
start a new project at some point after the
first one dies). This is because exercising an
option to invest is less irreversible when the
option to invest again is always available.
While, simulated threshold q values in this
case remain above unity, they fall significantly in the presence of modest rates of depreciation.
4C. Uncertainty Over Both Price and
Cost
Chapter 6 also considers the possibility that
both price and the cost of investing are uncertain. While the intuition for arriving at a
threshold investment criterion does not differ
substantively in this case, the solution process
is mathematically more difficult. Both the
value of the project and the value of the investment option are now functions of both P
and 1. The solution must then find regions of
(P,I) values in which investment will occur.
Dixit and Pindyck present a sketch of a solution when both P and I follow geometric
Brownian motion. It is still the case that the
threshold q criterion for investment exceeds
unity on account of the option value.
4D. Costs of Exit and Scrapping
The final aspects of investment decisions
by a monopoly firm I want to address are
1825
(9)
1826
(10)
(11)
rate. This qualification, also expanded to include the possibility of exit in the authors'
application to copper prices, implies that the
industry price can exceed long-run average
cost for a long period of time in a competitive
industry without stimulating entry.
6. Policy Intervention and Policy
Uncertainty
The ninth chapter of Investment Under Uncertainty addresses whether policy interventions are suggested by the new view investment models. The existence of delay and
inertia per se does not, of course support the
desirability of intervention in the absence of
a clearly articulated market failure. It is indeed easy to imagine real-world interventions
that do more harm than good. For example,
Dixit and Pindyck consider potential market
failures associated with incomplete markets
for sharing risk. Studying price controls as a
way to reduce risk, they find that such controls may be welfare-decreasing in many
cases. Another rich area for application lies in
industrial organization and antitrust policy.
Returning to the description in the previous
section of competitive equilibrium, policy
makers might see incumbent firms in an industry earning supernormal profits, but with
no entry occurring. With Marshallian intuition, they might take inappropriate action
against assumed barriers to entry, when they
might be observing a competitive industry. In
the same regard, observing a price below the
minimum level of average cost need not indicate predation; the "dumping"firms might be
maintaining the option to keep their (sunk)
investments.
In this section, I focus on the chance that
policy uncertainty may have unintended consequences; to be concrete, I frame the discussion in terms of business investment incentives. In describing price uncertainty in
Section 4, I noted that different characterizations of uncertainty can have different
effects on the responsiveness of investment
to changes in the net return to investing. The
same holds for tax policy uncertainty. An instructive case is the U.S. investment tax
credit (ITC), which, following its introduc-
1827
IkH-kL
dkt=
,XHdt
0
kL-kH
,(1- XH)dt
,XLdt
,(1 - XL)dt.
16 Another case in which the interaction of uncertainty and irreversibility may increase investment is that of learning about some types of uncertain costs-as, for example, in large-scale R&D
projects. Pindyck (1993) calls this technical uncertainty: If prices of inputs were known, how many
inputs and what level of effort are required to
complete the project? This type of uncertainty (as
opposed to uncertainty about the cost of individual
inputs) can be resolved only by actually proceeding with the project, and hence can stimulate investment.
q* = PI/(I - 1) >1,
and PI1is defined as in (5). That is, the firm
will not invest until the returns are sufficient
to compensate it for the lost option of delay.
To focus on tax policy uncertainty, consider a mean-preserving spread in the price
of capital: da > 0. One can show that
1-
1828
enced by uncertainty and irreversibility. Testing this prediction is a bit less straightforward
than tests of conventional neoclassical models, however. This is because the new view
models, while offering a rigorous description
of threshold q values for investment (to draw
an analogy to the neoclassical model), do not
offer specific predictions about the level of
investment. To go this extra step requires the
specification of structural links between the
marginal profitability of capital and the desired capital stock (the usual research focus
in the traditional neoclassical literature).
Nonetheless, option-based models offer some
testable hypotheses. For example, a fall in the
average rate of growth of profitability or rise
in the volatility of profitability should depress
investment over some period. Dixit and Pindyck offer some applications in Chapter 12.
A potentially informative literature testing
new view models using disaggregated data is
in its early stages. Caballero and Pindyck
(1992) use data on U.S. manufacturingindustries to study the determinants of the return
required to trigger investment. They employ
a proxy for the required return using extreme
values of the marginal profitability of capital.18 Supportive of the intuition of optionbased models, they find the proxies for the
required return depend positively on the
volatility of the marginal profitability of capital.19 In an analogous cross-sectional test,
Pindyck and Andres Solimano study the relationship between uncertainty and investment
for a set of thirty countries. For each country,
they calculate a time-series for the marginal
profitability of capital over a 28-year period.
Decomposing the period into three nine-year
subperiods, Pindyck and Solimano calculate
the mean and standard deviation of each
country's investment-GDP ratio and the annual log change in the marginal profitability
of capital. Using the panel of data, Pindyck
and Solimano relate the investment-GDP ra18 This calculation requires making assumptions
about the production technology, of course; they
assume a Cobb-Douglas technology with constant
returns to scale.
19 Such a test is only suggestive, because the
maximumvalue of the constructed marginalprofitability and its variance are correlated even if there
is no causal link between investment and uncertainty.
1829
1830
Wesley, 1993.
Under
ANDREW
"A
Uncertainty,
J.AND
CABALLERO, RICARDO
PINDYCK, ROBERT
S. "Uncertainty, Investment, and Industry Evolution." Working Paper No. 4160, National Bureau of Economic Research, 1992.
CHIRINKO, ROBERT S. "Business Fixed Investment Spending: A Critical Survey of Modeling
Strategies, Empirical Results, and Policy Implications," J. Econ. Lit., Dec. 1993, 31(4) pp.
1875-1911.
CUKIERMAN, ALEX. "The Effects of Uncertainty
on Investment Under Risk Neutrality with En-
JASON G.;
HASSETT,
KEVIN
A.
AND
of
as
DIXIT, AVINASH K. AND PINDYCK, ROBERT S. Investment under uncertainty. Princeton: Princeton U. Press, 1994.
DUFFIE, DARRELL. Security markets: Stochastic
models. San Diego: Academic Press, 1988.
FAZZARI,STEVEN M.; HUBBARD, R. GLENN AND
PETERSEN, BRUCE C. "Financing Constraints
and Corporate Investment," Brookings Pap.
Econ. Act., 1988, 1, pp. 141-95.
HALL, ROBERT E. AND JORGENSON, DALE W.
"Tax Policy and Investment Behavior," Amer.
Econ. Rev., June 1967, 57(3), pp. 391-414.
HASSETT, KEVIN A. AND METCALF, GILBERT E.
"Investment with Uncertain Tax Policy: Does
Random Tax Policy Discourage Investment?"
Mimeograph, Board of Governors of the Federal Reserve System, June 1994.
HAYASHI, FUMIO. "Tobin's Marginal q and Average q: A Neoclassical Interpretation," Econometrica, Jan. 1982, 50(1), pp. 213-24.
HOSHI, TAKEO; KASHYAP, ANIL K. AND
SCHARFSTEIN, DAVID. "Corporate Structure,
Liquidity, and Investment: Evidence from Japanese Industrial Groups," Quart. J. Econ., Feb.
1991, 106(1), pp. 33-60.
HUBBARD, R. GLENN. Money, the financial system, and the economy. Reading: Addison-
Harwood
HUBBARD, R. GLENN AND JUDD, KENNETH. "Liquidity Constraints, Fiscal Policy, and Consumption," Brookings Pap. Econ. Act., 1986, 1,
pp. 1-50.
HUBBARD, R. GLENN AND KASHYAP,ANIL K. "Internal Net Worth and the Investment Process:
An Application to U.S. Agriculture," J. Polit.
Econ., June 1992, 100(3), pp. 506-34.
HUBBARD, R. GLENN; KASHYAP, ANIL K. AND
WHITED, TONI M. "Internal Finance and Firm
Investment," J. Money, Credit, Banking, forthcoming.
JORGENSON, DALE W. "Capital Theory and Investment Behavior," Amer. Econ. Rev., May
1963, 53(2), pp. 247-59.
KARLIN, SAMUEL AND TAYLOR, HOWARD M. A
first course in stochastic processes. 2nd ed. New
York: Academic Press, 1975.
-.
A second course in stochastic processes.
New York: Academic Press, 1981.
LEAHY,JOHN AND WHITED, TONI M. "The Effect
of Uncertainty on Uncertainty Investment:
Some Stylized Facts." Mimeograph, Harvard U.,
June 1994.
MCDONALD, ROBERT AND SIEGEL, DANIEL. "Investment and the Valuation of Firms When
There Is an Option to Shut Down," Int. Econ.
Rev., June 1985, 26(2), pp. 331-49.
. "The Value of Waiting to Invest," Quart.
J. Econ., Nov. 1986,101(4), pp. 707-27.
PAKEs, ARIEL. "Dynamic Structural Models: Problems and Prospects," in Advances in econometrics: Sixth wor7d congress, Eds.: JEAN-JACQUES
LAFFONTAND CHRISTOPHERSIMS. Cambridge,
Eng.: Cambridge U. Press, 1993.
PINDYCK, ROBERT S. "Irreversibility, Uncertainty,
1831
Investment in Developing Countries," J. Develop. Econ., Oct. 1991, 36(2), pp. 229-42.
SUMMERS, LAWRENCEH. "Investment Incentives
and the Discounting of Depreciation Allowances," in The effects of taxation on capital accumulation. Ed.: MARTINFELDSTEIN. Chicago:
U. of Chicago Press, 1987, pp. 295-304.
TOBIN, JAMES. "A General Equilibrium Approach
to Monetary Theory," J. Money, Credit, Banking, Feb. 1969, 1(1), pp. 15-29.