16
This publication is available free of charge from: https://doi.org/10.6028/NIST.AMS.200-11 .
The caveat with Monte Carlo is that the results are shaped by the underlying assumptions.
As in any model, the accuracy of the assumptions is key. With a Monte Carlo simulation,
the ranges of possible values assigned to each variable constitute a critical set of
assumptions on which the whole undertaking rests, along with the methodology for
converting random numbers generated by the computer into values within these ranges.
The great benefit of the Monte Carlo analysis is that the values will be selected randomly
using the algorithm of the chosen software package. Because there is difficulty in
approximating randomness, a computer package can do a relatively accurate job of
approximating randomness (although it will be just an approximation since no computer
program can generate true randomness).
One of the challenges that arises in sensitivity/uncertainty analysis is that some
investments might end up being recategorized as a different scenario from Table 2.3. The
probabilistic nature of the analysis means that the underlying investments will be
categorized in different ways, depending on the state of the world generated by the
changing scenarios. Currently, there is no known software package that can be applied to
account for this issue; thus, one would have to develop their own coded model that makes
this calculation. The outcome would estimate the ranking of each investment for each
iteration.
Monte Carlo analysis is further discussed in Section 3.1 of NIST Advanced
Manufacturing Series 200-5.
1
NIST’s Monte Carlo tool can be used to implement this
type of analysis.
Sensitivity Analysis Example: Assume that a piece of capital machinery in a
manufacturer’s plant will wear out after some number of years. However, the
manufacturer does not know for sure when the machinery will wear out, but there is
reason to believe that the machinery could fail in any year from year 4 to year 8. Further,
suppose there is a probability of occurrence associated with each year of obsolescence,
which in this scenario is assumed to be known for each of the five years under
consideration. For this illustration, the replacement cost does not vary based on the year
of replacement and is always $100 000. A sensitivity factor (created exclusively for this
example) is then applied to each replacement cost by year. The sensitivity factor can be
generated through any number of software packages. This in turn provides the expected
present value of the cost for each year.
By applying sensitivity analysis, the manufacturer can evaluate the range of possible
replacement costs, based on the year when the capital machinery finally wears out and
can budget for those scenarios, thus reducing the risk to the firm. Looking at Table 2.4,
notice that in this instance the sum of the probabilities of a failure occurring in the second
column totals one. Furthermore, this example assumes that the replacement cost is
always the same, no matter what year it was incurred. The values generated in the fourth
column are artificial examples generated for the purposes of this example. What is
1
Thomas, Douglas. 2017. Investment Analysis Methods: A Practitioner’s Guide to Understanding the Basic Principles for Investment
Decisions in Manufacturing. NIST Advanced Manufacturing Series 200 -5. https://doi.org/10.6028/NIST.AMS.200-5