Valuing Operational Flexibility in Manufacturing Systems

Using Real Options

Harriet Black Nembhard
Leyuan Shi
Department of Industrial Engineering
University of Wisconsin-Madison
Madison, WI 53706-1572, U.S.A.
Mehmet Aktan
Department of Industrial Engineering
Atatrk University
Erzurum, 25240, TURKEY
Abstract
In this paper, we look across the manufacturing enterprise to consider opportunities to improve value by switching
among suppliers, plants, and markets. We extend this model to address the reality that these switches may often take
a period of transition time, rather than occur instantaneously. We value the real options associated with this flexible
approach using both a lattice and Monte Carlo simulation. This valuation gives decision makers a way to choose the
appropriate manufacturing enterprise strategy based on an integrated view of the market dynamics. Overall, the
manufacturing enterprise maximizes its expected, discounted profit through effective supply chain network decisions.

Keywords
Real Options, Monte Carlo simulation, supply chain

1. Introduction
Increased competition in the global market has caused organizations to realize that the most competitive way of
survival is high value. This can often be achieved through increased flexibility. We present a real options model for a
company that wants to maximize its profit by increasing flexibility in an environment of uncertain exchange rates. We
specifically consider the operational decisions to increase flexibility through selection of suppliers, allocation of
production among plants, and choice of market regions.
We use the options approach to find the value of such flexibility during a specified length of time, considering future
uncertain exchange rates between the home country and other countries where operations can be located. The
optimal policy maximizes the total expected discounted global profit. This valuation gives decision-makers a way to
choose the appropriate outsourcing strategy based on an integrated view of the market dynamics.
We present the problem formulation in Section 2. In Section 3, we present the valuation of the problem with a lattice
model. In Section 4, we propose an advanced model where the decisions can be implemented after a transition time
has been observed. In Section 5, we value the problem with a Monte Carlo simulation approach. Conclusions are
given in Section 6.

2. Problem Formulation
We consider a company that can supply from a number of global suppliers in foreign countries, can make production
in different plants, and can market its products in several market regions. The time horizon for the problem is divided
into T time intervals. In each time interval, the company selects an option from a group of N manufacturing options. A
particular global manufacturing strategy option O in period t is defined in terms of available sources of supply, plant

capacities, market regions, and open supply linkages within the firms global supply chain network. Switching
between the manufacturing options is costly because of the costs of opening and closing plants as well as opening
and closing supply linkages.
We formulate the option valuation problem as a stochastic dynamic program, similar to the formulations of
Huchzermeier [6] and Huchzermeier and Cohen [7]. In each state, the company maximizes its value by selecting a
manufacturing option Ot from the set of manufacturing options t, given that the company selected option O*t-1 in the
preceding period and the current exchange rate scenario between the home country and foreign countries is ekt,
where k is the number of possible jumps for the exchange rate scenario in each time period. Let be the discount
factor and succ(ekt) be the set of the successor states of exchange rate scenario ekt. The value V of the company in
period t and state (ekt, O*t-1) is defined as follows:
Vt(ekt, O*t-1)

*
max Pt(ekt, O t-1, Ot) + EVt+1( et +1 , Ot)

Ot t

*
max Pt(ekt, O t-1, Ot) +

Ot t

VT( ek ,T 1 , O*T-1)

e k ,t +1 succ(e kt )

p kk ' . Vt+1( ek ,t +1 , Ot)

(1)

*
max PT( ek ,T 1 , O T-1, OT).

OT T

Jump probability from state k to state k is denoted by p kk ' in Equation (1). Single-period profit function Pt accounts
for both the switching costs between global manufacturing strategy options and the global profit determined by
optimizing the single-period subproblem formulation SPt.
Pt(ekt, O*t-1, Ot) = [(O*t-1, Ot) + SPt(ekt, Ot)]

(2)

We assume that the exchange rate ei between the home country and the supplier country i follows geometric
Brownian motion as
dei ,t
ei,t

= i dt + i dzi

(3)

where i is the drift of the exchange rate changes for foreign country i, i is the volatility of the exchange rate for
foreign country i, dz is a standard Wiener disturbance term.

3. A Lattice Model
Huchzermeier [6] and Huchzermeier and Cohen [7] propose a lattice model where n-1 exchange rate processes is
approximated with an n-nomial lattice. As in other lattice models, they define the size of the movements of exchange
rates in the lattice as u i and d i, where u i is the one-step increase in exchange rate for country i, and d i is the one-step
decrease in the exchange rate for country i, i=1n-1.
We apply the Huchzermeier-Cohen lattice model in a problem with two uncertain exchange rate processes. Since we
have two exchange rate processes in the problem, we will use a trinomial lattice. The trinomial lattice assumes that for
any exchange rate scenario ekt = (e1kt, e2kt) at time t, there are three successor scenarios at time t+1, i.e., e1,t+1=(u 1.e1kt,
u 2.e2kt), e2,t+1=(d 1.e1kt, u 2.e2kt) and e3,t+1=(d 1.e1kt, d 2.e2kt).
The supply chain network for the problem is shown in Figure 1. Each of the two foreign countries is a supplier. There
is one production plant in each country. At the same time, each of these two countries is a market region for the final
product. Total time horizon of the problem is one year. Total time was divided equally into four time periods, which
implies that each period was four months.

Supplier

Plant

Market

S1

P1

M1

S2

P2

M2

Figure 1. Supply chain network for two suppliers, two plants, and two markets.
Twelve manufacturing strategy options are defined based on possible connections in the supply chain network.
These twelve options are shown in Figure 2.
Option 1

Option 2

Option 3

Option 4

Option 5

Option 6

Option 7

Option 8

Option 9

Option 10

Option 11

Option 12

Supplier

Plant

Market

Figure 2. Manufacturing strategy options.

Figure 3 shows expected profits for no switching and costly switching. When there is no switching, the same option
will be used at each time period. However, when switching is possible, the best option can be selected at each time
period considering the switching costs. Then, the option value of flexibility because of being able to switch the
manufacturing strategy is the difference between the expected profits of costly switching and no switching. The
upper line in Figure 3 shows the expected profits when switching between options is possible, and the lower line
shows the expected profits when there is no switching between options. The horizontal axis shows which
manufacturing strategy option has been used prior to the first time period in the problem.

Expected Profit ($)

2500
2000
1500

Switch
No Switch

1000
500
0
1

10

11

12

Option Number

Figure 3. Expected profits for no switching and costly switching.

4. A Model for Multi-Period System Transitions

In the manufacturing strategy option problem in Section 3, we assumed that when a new manufacturing strategy is
desired, it is implemented immediately. This means that when the company wants to switch the supplier, all raw
material from the new supplier is received instantly; when a switch is desired for the production plants, required
amount of production in the new plant is done instantly; and when a switch is desired for the market regions, final
product reaches to the new market regions instantly.
However, in real manufacturing operations, these assumptions do not truly hold. Switches for suppliers, plants, and
market regions need a transition time to be executed. In this section, we propose a solution method that addresses
the transition time needed to execute the switching decisions in such problems.
In the lattice model of Section 3, we assumed that if the company gives a switch decision at a node, that switch could
be implemented at the same node instantly. Now we assume that a switch decision given in a node can be
implemented at the successor nodes. We solve the problem where the transition time is one period. In this problem, if
a switch decision is given at a node, then the strategies in immediate successor nodes for that node have to be same,
because the decision given in a node will be implemented in the next time period. This means that when a decision is
given at a state, then that decision will be executed in the next state, whichever the next state happens to be.
When using dynamic programming with backward recursion, this requires that the option selections will be same at
the nodes that are preceded by the same node. Then, Ot, which is the option selected at time t, will be same at states
that are preceded by the same state (see Equation (1)). At the first time period, the option will be the same as the
option that was used just before the beginning of the planning horizon. Using this relationship, we modified the
structure of the backward recursion of the dynamic program, and obtained the option values of flexibility with the
feature of transition time.
Comparison of the option value of flexibility with and without the transition time is given in Figure 4. The upper line
shows the expected profits when the transition time is zero, i.e., the switching decisions can be implemented
instantly. The lower line shows the expected profits when switching is not possible, i.e., when there is not any
flexibility. The middle line shows the expected profits when there is a transition time of one period for the
implementation of switching decisions.

3000
No Switch
Expected Profit ($)

2500
2000
Switch
with
transition
time

1500
1000

Switch
without
transition
time

500
0
1

10

11

12

Option Number

Figure 4. Expected profits with and without transition time.

We see that the expected profits with transition time are less than the profits with no transition time. When the
implementation of strategy switches require a transition time, the company must wait until the implementation.
Therefore, compared to the case of instant implementation, there is less flexibility when there is transition time. Since
the no-switching case has zero flexibility, its expected profits are lowest.

5. A Monte Carlo Simulation Approach for Multi-Time Period Options and Nonzero Switching
Costs
Boyle [2] introduced a Monte Carlo simulation method for asset pricing of European options. This approach,
however, cannot be used for problems with switching costs since they cannot be treated as a bundle of European
options with different expiration dates. Broadie and Glasserman [5] developed a simulation algorithm for estimating
the prices of American-style assets. In order to develop valid error bounds on the true option value, they introduced
two estimators, one biased high and one biased low, but both asymptotically unbiased as the computational effort
increases. These estimators are based on simulated lattices. The simulated trees are parameterized by b, the number
of branches per node. State variables are simulated at the finite number of possible decision points, i.e., exercise
times.
In this section, we propose a Monte Carlo simulation technique that can be used to value real options that have
multiple time points to switch between the options and that have nonzero switching costs. In order to value the
problem, we will combine dynamic programming with the simulation approach given in Broadie and Glasserman [5]. In
the proposed technique, we use a simulation tree with three branches per node. On this simulation tree, we will apply
the dynamic programming algorithm that we used for the trinomial lattice. The best decision for each possible
previous policy is determined at each node. This means that best decisions are determined at each of the three
branches. During the backward recursion of the dynamic programming, we use the mean value of the three branches
to select the best decision. Since the simulated values for the state variables are used for calculations, the dynamic
programming takes advantage of knowledge of the future to overestimate the option value. Hence, this estimate is
biased high.
In order to find the low estimator, we use the following approach. At each node, we use the second and third
branches to determine a policy, and we apply that policy in branch one. Then, we use the first and third branches to
determine a policy, and we apply that policy in branch two. Finally, we use the first and second branches to
determine a policy, and we apply that policy in branch three. Then, we use the average of the expected profits from
the three branches during the backward recursion of the dynamic program. We will use the average of the highbiased and low-biased simulation estimates to estimate the true option value. Applying this technique on the problem
given in Section 3, we obtained the simulation results given in Table 1.
Table 1. Monte Carlo Simulation and Trinomial Lattice Results
Prior option
Low -biased estimate
High-biased estimate
Average of estimates
Trinomial lattice result
Difference (%)

1
2011.7
2019.4
2015.5
2015.3
0.01

2
2112.8
2128.4
2120.6
2115.3
0.25

3
2097.2
2131.8
2114.5
2115.3
0.04

4
2208.9
2238.1
2223.5
2215.3
0.37

5
2003.0
2040.9
2021.9
2015.3
0.33

6
2107.6
2121.4
2114.5
2115.3
0.04

7
2110.1
2128.4
2119.2
2115.3
0.19

8
2204.2
2223.5
2213.8
2215.3
0.06

9
2100.7
2134.2
2117.5
2115.3
0.10

10
2208.9
2231.6
2220.3
2215.3
0.22

11
2209.8
2229.8
2219.8
2215.3
0.20

12
2299.1
2326.5
2312.8
2315.3
0.11

The first row in Table 1 shows the policy prior to the beginning of the decision time horizon. The next two rows are
the low-biased and high-biased Monte Carlo simulation estimates for the expected profits. The fourth row is the
average of these two estimates. The fifth row shows the expected profits that we obtained by using trinomial lattice.
The last row shows the percent difference between the expected profits of Monte Carlo simulation and trinomial
lattice. It has been shown that lattice methods give close results to the analytical solutions (Nembhard, Shi, and
Aktan [9], Boyle [3], Boyle, Evnine, and Gibbs [4], Kamrad and Ritchken [8], Amin and Khanna [1]). We see that the
Monte Carlo simulation estimates are close to the trinomial lattice estimates, which implies that the proposed
approach yields close results to the analytical solution for this problem.
Many manufacturing system models contain more than two state variables. If there are three or more state variables
in the problem, lattice techniques become difficult to apply. If we want to use lattice approach in such problems,
structure of the lattice gets too complex and the number of possible jumps at each node gets larger. However, the
proposed Monte Carlo simulation technique does not require a different tree structure when the number of state
variables increases. The same tree structure is used for any number of state variables; the only difference is the

number of random numbers generated. The main advantage of the proposed Monte Carlo simulation approach is its
simple application on real options problems with three or more state variables.

6. Conclusions
In this paper, we have proposed a technique to value the real options problems in manufacturing enterprises when
decisions in the system cannot be implemented immediately. We have presented the application of the proposed
technique on a supply chain network with uncertain exchange rates, where costly switching decisions for the
suppliers, production plants, and market regions take a transition time to execute.
We have also proposed a Monte Carlo simulation technique that can be used to value the real options problems with
costly switching decisions. In such problems, there are multiple options. At each time interval, switching between
options is possible, and each switch results in a switching cost. We have compared the results of the proposed
Monte Carlo simulation with the results of a lattice technique. The comparison has shown that the proposed Monte
Carlo simulation technique yields close estimates for the true option value.
Real options valuation techniques proposed in this paper give decision makers a way to choose the appropriate
manufacturing enterprise strategy based on an integrated view of the market dynamics. Overall, the manufacturing
enterprise maximizes its expected, discounted profit through effective supply chain network decisions.

References
1.
2.
3.
4.
5.
6.
7.
8.
9.

Amin, K.I., and Khanna, A., 1994, Convergence of American Option Values from Discrete to Continuous-Time
Financial Models, Mathematical Finance, 4, 289-304.
Boyle, P.P., 1977, Options: A Monte Carlo Approach, Journal of Financial Economics, 4, 323-338.
Boyle, P.P., 1988, A Lattice Framework for Option Pricing with Two State Variables, Journal of Financial and
Quantitative Analysis, 23(1), 1-12.
Boyle, P.P., Evnine, J., and Gibbs, S., 1989, Numarical Evaluation of Multivariate Contingent Claims, The
Review of Financial Studies, 2(2), 241-250.
Broadie, M., and Glasserman, P., 1997, Pricing American-style Securities Using Simulation, Journal of
Economic Dynamics and Control, 21, 1323-1352.
Huchzermeier, A., 1991, Global Manufacturing Strategy Planning under Exchange Rate Uncertainty, Ph.D.
Thesis, Decision Sciences Department, The Warton School, University of Pennsylvania, Philadelphia.
Huchzermeier, A., and Cohen, M.A., 1996, Valuing Operational Flexibility under Exchange Rate Risk,
Operations Research, 44(1), 100-113.
Kamrad, B., and Ritchken, P., 1991, Multinomial Approximating Models for Options with k State Variables,
Management Science, 37(12), 1640-1653.
Nembhard, H.B., Shi, L., and Aktan, M., 2001, A Real Options Design for Quality Control Charts, to appear in
The Engineering Economist.