ECONOMICS FOR ENGINEERS
Probabilities
We all have used probabilities. For example, what is the probability of getting a
"head" when flipping a coin? Using a model that assumes that the coin is fair,
both the head and tail outcomes occur with a probability of 50%, or ½. This
probability is the likelihood of an event in a single trial. It also describes the
long-run relative frequency of occurrence of getting heads in many trials (out of
50 coin flips, we expect to average 25 heads).
Probabilities can also be based on data, expert judgment, or a combination of
both. Past data on weather and climate, on project completion times and costs,
and on highway traffic is combined with expert judgment to forecast future
events. These examples can be important in engineering economy.
Another example based on long-run relative frequency is the PW of a flood-
protection dam that depends on the probabilities of different-sized floods over
many years. This would be based on data about past floods and would include
many years of potential flooding. An example of a single event that may be
estimated by expert judgment is the probability of a successful outcome for a
research and development project, which will determine its PW.
All the data in an engineering economy problem may have some level of
uncertainty. However, small uncertainties may be ignored, so that more analysis
can be done with the large uncertainties. For example, the price of an off-the-
shelf piece of equipment may vary by only ±5%.The price could be treated as a
known or deterministic value. On the other hand, demand over the next 20 years
will have more uncertainty. Demand should be analysed as a random or
stochastic variable. We should establish probabilities for different values of
demand.
There are also logical or mathematical rules for probabilities. If an outcome can
never happen, then the probability is 0. If an outcome will certainly happen,
then the probability is 1, or 100%.This means that probabilities cannot be
negative or greater than 1; in other words, they must be within the interval [0,
1], as indicated shortly in Equation 1.
Probabilities are defined so that the sum of probabilities for all possible
outcomes is 1 or 100% (Equation 2). Summing the probability of 0.5 for a head
and 0.5 for a tail leads to a total of 1for the possible outcomes from the coin
flip. An exploration well drilled in a potential oil field will have three outcomes
(dry hole, non -commercial quantities, or commercial quantities) whose
probabilities will sum to one.
AOT/ME/SKL
ECONOMICS FOR ENGINEERS
Equations 1 and 2 can be used to check that probabilities are valid. If the
probabilities for all but one outcome are known, the equations can be used to
find the unknown probability for that outcome (see Example 1).
0 ~ Probability ~ 1 ------(1)
∑ P(outcomej) = 1, where there are K outcomes -----(2)
In a probability course many probability distributions, such as the normal,
uniform, and beta are presented. These continuous distributions describe a large
population of data. However, for engineering economy it is more common to
use 2 to 5 outcomes with discrete probabilities--even though the 2 to 5
outcomes only represent or approximate the range of possibilities.
This is done for two reasons. First, the data often are estimated by expert
judgment, so that using 7 to 10 outcomes would be false accuracy. Second, each
outcome requires more analysis. In most cases the 2 to 5 outcomes represents
the best trade-off between representing the range of possibilities and the amount
of calculation required. Example 1 illustrates these calculations
Example 1
What are the probability distributions for the annual benefit and life for the
following project?
The annual benefit’s most likely value is $8000 with a probability of 60%.
There is a 30% probability that it will be $5000 and the highest likely value is
$10,000. A life of 6 years is twice as likely as a life of 9 years.
Solution:
Probabilities are given for only two of the possible outcomes for the annual
benefit. The third value is found using the fact that the probabilities for the three
outcomes must sum to 1,
1 = P(Benefit is $5000) + P(Benefit is $8000) + P(Benefit is $10,000)
P(Benefit is $10,000) = 1 - P(Benefit is $5000) - P(Benefit is $8000)
=1 - 0.6 - 0.3 = 0.1
The probability distribution can then be summarized in a table.
Annual benefit $5000 $8000 $10000
Probability 0.3 0.6 0.1
The problem statement tells us:
P(life is 6 years) = 2P(life is 9 years) ------1
Equation P1+P2=1 can be applied to write a second equation for the two
unknown probabilities:
P(6) + P(9)= 1 -----------2
Combining these, we write
AOT/ME/SKL
ECONOMICS FOR ENGINEERS
2P(9) + P(9) = 1
3P(9)= 1
P(9) =1/3
P(6) = 2/3
The probability distribution for the life is P(6) = 66.7% and P(9)= 33.3%.
JOINT PROBABILITY DISTRIBUTIONS
Example 1 constructed the probability distributions for a project's annual benefit
and life. These examples show how likely each value is for the input data of the
problem. .We would like to construct a similar probability distribution for the
project's present worth. This is the distribution that we can use to evaluate the
project. That present worth depends on both input probability distributions, so
we need to construct the joint probability distribution for the different
combinations of their values.
For this introductory text, we assume that two random variables such as the
annual benefit and life are unrelated or statistically independent. This means
that the joint probability of a combined event (Event A defined on the first
variable and Event B on the second variable) is the product of the probabilities
for the two events. This is Equation 3:
If A and B are independent, then P(A and B) = P(A) x P(B) ----- 3
For example, flipping a coin and rolling a die are statistically independent.
Thus, the probability of {flipping a head and rolling a 4} equals the probability
of a {heads} = ½ times the probability of a {4} = 1/6, for a joint probability =
1/12. The number of outcomes in the joint distribution is the product of the
number of outcomes in each variable's distribution. Thus, for the coin and the
die, there are 2 times 6 or [Link] of the 2 outcomes for the coin
is combined with each of the 6 outcomes for the die.
Some variables are not statistically independent, and the calculation of their
joint probability distribution is more complex. For example, a project with low
revenues may be terminated early and one with high revenues may be kept
operating as long as possible. In these cases annual cash flow and project life
are not independent. While this type of relationship can sometimes be modelled
with economic decision trees, we will limit our coverage in this text to the
simpler case of independent variables.
Example 2 uses the three values and probabilities for the annual benefit and the
two values and probabilities for the life to construct the six possible
AOT/ME/SKL
ECONOMICS FOR ENGINEERS
combinations. Then the values and probabilities are constructed for the project's
PW.
Example 2
The project described in Example 10-5 has a first cost of $25,000. The firm uses
an interest rate of 10%. Assume that the probability distributions for annual
benefit and life are unrelated or statistically independent. Calculate the
probability distribution for the PW.
Solution:
Since there are three outcomes for the annual benefit and two outcomes for the
life, there are six combinations. The first four columns of the following table
show the six combinations of life and annual benefit. The probabilities in
columns 2 and 4 are multiplied to calculate the joint probabilities in column 5.
For example, the probability of a low annual benefit and a short life is 0.3 x 2/3,
which equals 0.2 or 20%.
The PW values include the $25,000 first cost and the results of each pair of
annual benefit and life. For example, the PW for the combination of high
benefit and long life is:
PW10000,9 = -25,000 + 10000(P/ A, 10%,9)
= -25,000 + 10,000(5.759)
= $32,590
(P/A, i, n) =
Annual Probability Life Probability Joint PW
benefit Probability
5000 0.3 6 0.667 0.2 -3224
8000 0.6 6 0.667 0.4 9842
10000 0.1 6 0.667 0.0667 18533
5000 0.3 9 0.333 0.1 3795
8000 0.6 9 0.333 0.2 21072
10000 0.1 9 0.333 0.0333 32590
1
This probability distribution function shows that there is a 20% chance of
having a negative PW. It also shows that there is a small 3.3% chance of the PW
being $32,590. The three values used to describe possible annual benefits for
AOT/ME/SKL
ECONOMICS FOR ENGINEERS
the project and the two values for life have been combined to describe the
uncertainty in the project's PW.
Creating a distribution, as in Example 2, gives us a much better understanding
of the possible PW values along with their probabilities. The three possibilities
for the annual benefit and the two for the life are representative of the much
broader set of possibilities that really exist. Optimistic, most likely, and
pessimistic values are a good way to represent the uncertainty about a variable.
Similarly the six value for the PW represent the much broader set of
possibilities. The 20% probability of a negative PW is one measure of risk that
we will talk about later in the chapter.
EXPECTED VALUE
For any probability distribution we can compute the expected value (EV) or
arithmetic average (mean). To calculate the EV, each outcome is weighted by
its probability, and the results are summed. This is NOT the simple average or
un-weighted mean. When the class average on a test is computed, this is an un-
weighted mean. Each student's test has the same weight. This simple "average"
is the one that is shown by the button x on many calculators.
The expected value is a weighted average, like a student's GPA or grade point
average. For a GPA the grade in each class is weighted by the number of
credits. For the expected value of a probability distribution, the weights are the
probabilities. .
Expected Value = Outcome A X P(A) + Outcome B X P(B) +... ….
Example 3
The first cost of the project in Example 1 is $25,000. Use the expected values
for annual benefits and life to estimate the present worth. Use an interest rate of
10%.
Solution:
EVbenefit = 5000(0.3) + 8000(0.6) + 10,000(0.1) = $7300
EVlife = 6(2/3) + 9(1/3) = 7 years
The PW using these values is
PW(EV) = -25,000 + 7300(P / A, 10%,7)
= -25,000 + 7300(4.868)
= $10,536
[Note: This is the present worth of the expected values, PW(EV), not the
expected value of the present worth, EV(PW). It is an easy value to calculate
AOT/ME/SKL
ECONOMICS FOR ENGINEERS
that approximates the EV(PW), which will be computed using the joint
probability distribution).found in Example 2.]
Example 3 is a simple way to approximate the project's expected PW. But the
true expected value of the PW is some what different. To find it, we must use
the joint probability distribution for benefit and life, and the resulting
probability distribution function for PW that was derived in Example [Link]
4 shows the expected value of the PW or the EV(PW).
Example 4
Use the probability distribution function of the PW that was derived in Example
2 to calculate the EV(PW). Does this indicate an attractive project?
Solution:
The table from Example 2 can be reused with one additional column for the
weighted values of the PW (= PW x probability). Then, the expected value of
the PW is calculated by summing the column of present worth values that have
been weighted by their probabilities.
Annual Probability Life Probability Joint PW PW X Joint
benefit Probability Probability
5000 0.3 6 0.667 0.2 -3224 -645
8000 0.6 6 0.667 0.4 9842 3937
10000 0.1 6 0.667 0.0667 18533 1237
5000 0.3 9 0.333 0.1 3795 380
8000 0.6 9 0.333 0.2 21072 4214
10000 0.1 9 0.333 0.0333 32590 1086
1 EV (PW) = 10209
With an expected PW of $10,209, this is an attractive project. While there is a
20% chance of a negative PW, the possible positive outcomes are larger and
more likely. Having analyzed the project under uncertainty, we are much more
knowledgeable about the potential result of the decision to proceed.
The$10209 value is more accurate the approximate value calculated in Example
3. The values differs because PW is a nonlinear function of the life. The more
accurate value of $~10209 is lower because the annual benefit values for the
longer life are discounted by 1/(1 + i) for more years.
In Examples 3 and 4, the question was whether the project had a positive PW.
With two or more alternatives, the criterion would have been to maximize the
PW. With equivalent uniform annual costs (EUACs) the goal is to minimize the
AOT/ME/SKL
ECONOMICS FOR ENGINEERS
EUAC. Example 5 uses the criterion of minimizing the EV of the EUAC to
choose the best height for a dam.
Example 5
A dam is being considered to reduce river flooding. But if a dam is built, what
height should it be? Increasing the dam's height will (1) reduce a flood's
probability, (2) reduce the damage when floods occur, and (3) cost more. Which
dam height minimizes the expected total annual cost? The state uses an interest
rate of 5% for flood protection projects, and all the dams should last 50 years.
Dam FC Annual Damages If
Height(ft) P (flood) > Height Flood Occurs
No Dam 0 0.25 800000
20 700000 0.05 500000
30 800000 0.01 300000
40 900000 0.002 200000
Solution:
The easiest way to solve this problem is to choose the dam height with the
lowest equivalent uniform annual cost (EUAC). Calculating the EUAC of the
first cost requires multiplying the first cost by (A / P, 5%, 50). For example, for
the dam 20ft high, this is 700,000(A/ P,5%, 50) = $38,344.
Calculating the annual expected flood damage cost for each alternative is
simplified because the term for the P (no flood) drops out, because the damages
for no flood are rs 0. Thus we need to calculate only the term for flooding. This
is done by multiplying the P (flood) times the damages if a flood happens. For
example, the expected annual flood damage cost with no levee is 0.25 x 800000
= 200000
Then the EUAC of the first cost and the expected annual flood damage are
added together to find the total EUAC for each height. The 30 ft dam is
somewhat cheaper than the 40 ft dam.
Dam EUAC for Expected Annual Total Expected
Height(ft) First Cost flood damage EUAC
No Dam 0 200000 200000
20 38344 25000 63344
30 43821 3000 46821
40 49299 400 49699
AOT/ME/SKL