Chapter 18
Sensitivity
Analysis
And
Staged Decisions
Topics to be discussed
Perform sensitivity analysis of one or more
parameters and of an entire alternative
Determination and use of the expected value of a
cash flow series
The techniques of decision trees to help make
economic decisions for alternatives that have
different, but closely connected stages
Economic decisions that involve staged funding
and real options analysis
Determining Sensitivity to Parameter
Variation
Determines how a measure of worth PW, AW, FW, or
ROR is altered when one or more parameters vary
over a selected range of values
Usually one parameter at a time is varied, and
independence with other parameters is assumed
In reality this approach useful to determine the response
to variation in a variety of parameters
Example: variation in MARR will not alter the decision to
select an alternative when all compared alternatives
return considerably more than MARR. Thus, the decision
is relatively insensitive to the MARR. However, variation
in the n or AOC may be very sensitive to decision
choice.
Usually variation in life, annual costs, and revenues
result from variations in selling price, operation at
different levels of capacity, etc.
Usually several important parameters are studied to
learn how the uncertainty of estimates affects the
economic analysis
Sensitivity analysis routinely concentrates on the
variation expected in estimates of P, AOC, S, n, unit
costs, unit revenues, and similar parameters
Parameters such as MARR, and other interest rates are
more stable from project to project
Procedure to conduct sensitivity
analysis
Determine parameter(s) of interest might vary from
the most likely estimated value
Select the probable range and an increment of
variation for each parameter
Select the measure worth
Computer the results for each parameter, using the
measure of worth as a basis
To better interpret the sensitivity, graphically display
the parameter versus the measure of worth
Example 18.1
Wild Rice, Inc., expects to purchase a new asset for
automated rice handling. Most likely estimates are a first
cost of $80,000, zero salvage value, and a cash flow
before taxes (CFBT) per year t that follows the relation
$27,000 2000t. The MARR for the company varies
from 10% to 25% per year for different types of
investments. The economic life of similar machinery
varies from 8 to 12 years. Evaluate the sensitivity of PW
by varying
(a) MARR, while assuming a constant n value of 10 years,
and
(b) n, while MARR is constant at 15% per year
Solution
(a) Procedure to understand sensitivity of PW to
variation
MARR
1. MARR is the parameter of interest
2. Select 5% increments to evaluate sensitivity to
MARR; the range is 10% to 25%
3. The measure worth is PW
4. Set up the PW relation for 10 years. When MARR is
10%,
PW = -80,000 + 25,000(P/A,10%,10) 2000(P/G,10%,10)
= $27,830
The PW for all four values at 5% intervals is as
follows:
MARR
10.00%
15.00%
20.00%
25.00%
PW
$27,830.00
$11,512.00
-$962.00
-$10,711.00
5. A plot of MARR vs PW is shown in figure below. The
steep negative slope indicates the the decision to
accept the proposal based on PW is quite sensitive to
variation in MARR. If the MARR is established at the
upper end of the range, the investment is not
attractive.
PW, $
30,000
MARR
20,000
n
10,000
-10,000
10
15
20
25
MARR %
10
12
Life n
(b) 1. Asset life n is the parameter
2. Select 2-year increments to evaluate PW
sensitivity over the range 8 to 12 years
3. The measure of worth is PW
4. Set up the same PW relation as in part (a) at
i = 15%. The PW results are
n
8
10
12
PW
$7,221.00
$11,511.00
$13,145.00
5. From the plot PW vs n, since the PW measure is positive
for all values, the decision to invest is not materially
affected by the estimated life. The PW curve levels out
above n = 10. This insensitivity to changes in cash flow in
the distant future is predictable observation, because the
P/F factor gets smaller as n increases
One alternative with several
parameters to considered using a
single measure of worth
Two alternatives are compared and the
sensitivity to one parameter is sought
Formalized Sensitivity Analysis
using Three Estimates
Three estimates from the field of project
scheduling:
pessimistic
most likely
optimistic estimate
Example 18.3
An engineer is evaluating three alternatives for
which she has made three estimates for the
salvage value, annual operating cost, and the
life. The estimates are presented on an
alternative-by-alternative basis in table below.
Perform a sensitivity analysis and determine
the most economical alternative, using AW
analysis at a MARR of 12% per year
Competing alternatives with three estimates made
for salvage value, AOC, and life parameters
Strategy
First
Cost
Salvage
Value
AOC
Life n
years
-$20,000
-$20,000
-$20,000
$0
$0
$0
-$11,000
-$9,000
-$5,000
3
5
8
-$15,000
-$15,000
-$15,000
$500
$1,000
$2,000
-$4,000
-$3,500
-$2,000
2
4
7
-$30,000
-$30,000
-$30,000
$3,000
$3,000
$3,000
-$8,000
-$7,000
-$3,500
3
7
9
Alternative A
P
Estimates ML
O
Alternative B
P
Estimates ML
O
Alternative C
P
Estimates ML
O
Solution
For each alternative in table above, calculate the
AW value of costs. For example, the AW
relation for alternative A, pessimistic
estimates is,
Estimates
P
ML
O
Alternative AW value
A
B
C
-$19,372
-$12,640
-$19,601
-$14,548
-$8,229
-$13,276
-$9,026
-$5,089
-$8,927
20
18
Alternative A
16
Alternative C
14
12
AW of
costs,
10
$x-1000
8
6
Alternative B
4
2
0
Life n, years
Estimate variability and the expected
value
Engineers usually deal with estimate variation and risk
about uncertain future
Used past data, samples and probability
The expected value long-run average observable if the
project is repeated many times
The expected value E(X) is computed using:
i=n
E(X) = XiP(Xi)
i=1
[18.2]
Xi
= value of the variable X for I from 1 to m different value
P(Xi) = probability that a specific value of X will occur
Example 18.4
ANA airlines plans to offer several new electronic services in flights
between Tokyo and selected European destinations. The
marketing director estimates that for a typical 24-hour period
there is a 50% chance of having a net cash flow of $5000 and a
35% chance of $10,000. He also estimates there is a small 5%
chance of no cash flow and a 10% chance of a loss of $1000,
which is the estimated extra personnel and utility costs to offer
the service. Determine the expected net cash flow.
Solution
Let NCF be the net cash flow in $, and let P(NCF) represent the
associated probabilities
E(NCF) = $5000(0.5) + $10,000(0.35) + $0(0.05) - $1000(0.1)
= $5900
Expected value computations for
alternatives
The expected value computation E(X) is utilized to
Prepare information for use in an economic analysis
(example 18.5)
Evaluate the expected viability of a fully formulated
alternative (example 18.6)
Example 18.5
There are many government incentives to become more energy-efficient,
like installing solar panels on homes, business building. The owner
pays a portion of the total installation costs, and the government
agency pays the rest. Department of Energy has to determine the size
increase in annual budget the incentive program needs in the future.
Over the last 36 months the amount of average monthly payout and
number of months are shown in table below. Provided by the same
pattern continues, what is the expected value of the dollar increase in
annual budget that is needed to meet the request?
Solar panel incentive payouts
Level
Average payouts,
$million per month
Months over past 3
years
Very high
6.5
15
High
4.7
10
Moderate
3.2
Low
2.9
Solution
Level, j
Probability of payout level, P(POj)
Very high
15/36 = 0.417
High
10/36 = 0.278
Moderate
7/36 = 0.194
Low
4/36 = 0.111
1.000
Use the 36 months of payouts POj (j= low, .., very high) to estimate the
probability P (POj) for each level, make sure the total is 1.0
The expected monthly payout is calculated using eq. [18.2]. in $million
units
E[PO] = $6.5(o.417) + $4.7(0.278) +$3.2(0.194) + $2.9(0.111)
= $4.961 million
The annual expected budget need is 12x$4.961 million=$59.532 million
Example 18.6
Lite-weight Wheelchair Company has a substantial investment in tubular
steel bending equipment. A new piece of equipment costs $5000 and
has a life of 3 years. Estimated cash flows (table below) depend on
economic conditions classified as receding, stable, or expanding. A
probability is estimated that each of the economic conditions will
prevail during the 3-year period. Apply expected value and PW
analysis to determine if the equipment should be purchased. Use the
MARR of 15% per year
Economic condition
Receding
(Prob = 0.4)
Year
Stable
(Prob = 0.4)
Expanding
(Prob = 0.2)
Annual cash flow estimates, $
-5000
-5000
-5000
2500
2500
2000
2000
2500
3000
1000
2500
3500
solution
First, determine the PW of the cash flow for each economic conditions,
and then calculate E(PW) using equation [18.21].
PWR = -$5000 + $2500(P/F,15%,1) + $2000(P/F, 15%,2) + $1000
(P/F,15%, 3) = - $656
PWS = -$5000 + $2500 (P/A,15%,3) = $708
PWE = -$5000 + $2500 (P/F,15%,1 + $2000 (P/F,15%,2) + $3500
(P/F,15%,3) = $ 1309
The expected present worth is
E(PW) = PWj[P(j)]
=j=R,S,E
-$656 (0.4) + $708 (0.4) + $1309 (0.2)
= $283
Comment
It is also correct to calculate the E(cash flow) for each year and then
determine PW of the E(cash flow) series, because the PW computation
is a linear function of cash flows can reduce the number of PW
calculation
E(CF0) = - $5000
E(CF1) = $2500(0.4) + $2500(0.4) + $2000(0.2) = $2400
E(CF2) = $2500(0.4) + $2500(0.4) + $3000(0.2) = $2400
E(CF3) = $2500(0.4) + $2500(0.4) + $3500(0.2) = $2100
E(PW) = -$5000 + $2400(P/F, 15%,1) + $2400(P/F,15%,2) +
$2100(P/F, 15%,3) = $283
Staged Evaluation of Alternatives using a
Decision Tree
Evaluation using a decision tree, that includes
more than one stage of alternative selection
selection of an alternative at one stage that leads to another
stage
expected results from a decision at each stage
probability estimates for each outcome
estimates of economic value (cost or revenue) for each
outcome
measure of worth as the selection criterion, such as E(PW)
Construct a Decision tree
constructed from left to right
includes each possible decision and
outcome
a decision node (a square) represent
possible alternatives
a probability node (circle) represent
possible outcomes and estimated
probability
outcomes always follow decisions
Decision and probability nodes
used to construct a decision tree
Decision
node
Alternatives
(a) Decision node
Probability
node
Outcomes
0.5
0.2
0.3
(b) Probability node
Probabilities
Final outcomes
Evaluation of alternatives using decision
tree
To utilize the decision tree for alternative evaluation and selection, the
following additional information is necessary
The estimated probability that each outcome may occur
Economic information for each decision alternative and possible
outcome, such as initial investment and estimated cash flows
E(decision) = (outcome estimate) P(outcome)
[18.3]
Where the summation is taken over all possible outcomes for each
decision alternatives
Example 18.8
Decision is needed to either market or sell a new invention. If the product
is marketed, the next decision is to take it international or national.
Assume the details of the outcome branches result in the decision tree,
the probabilities for each outcome and PW of CFBT (cash flow before
taxes) are indicated. These payoffs are in millions of dollars.
Determine the best decision at the decision node D1.