ñòð. 157 |

ume to a project at a cost which is well below the average cost per unit if it is

done at the design stage. If the capacity can be added later at about the same

cost then it will not be worthwhile building in advance.

Pipelines are classic examples of where capacity expansion is cheap before

you start but expensive later. They are expensive to build and a lot of the cost

concerns gaining access to the right of way and digging the trench. Once in

place, throughput is capped.13 This means that at the design stage, increasing

throughput is quite cheap. Once the pipeline is in place, however, any expan-

sion would be hugely expensive.

Pipelines are not the only example of where extra capacity can be cheap if

it is introduced at the design stage. A sports stadium where plenty of land is

available is probably the same and so too would be almost any infrastructure

project. My suggestion is that use of valuing the tail and growth value tech-

niques can, hopefully, offset the tendency which otherwise exists which is to

seek to minimise the headline capital cost for the initial build of such assets.

Step 8E: Monte Carlo simulation

I introduced this technique in the modelling economic value pillar on page

239. At that stage I rather played down its potential usefulness. This was

because in most simple situations I do not believe that its use is justified.

When one is dealing with flexibility value, however, it is potentially very use-

ful and it forms an essential part of the flexibility valuation toolkit.

The primary purpose of this particular section is to build awareness of

what can be done. This should give enough information to allow a relatively

When one does this analysis, it quickly becomes apparent that having good information about what

12

oil fields might be found can be hugely valuable. This is because of the sums of money at stake when

one oversizes an expensive asset like a pipeline. I will be looking at a possible way of justifying the

spend necessary in order to gather information in section 8F.

In reality, throughput can be increased by installing more powerful pumps or by adding what is called

13

a drag-reducing agent to the crude oil. There is, however, a limit to what these two steps can achieve.

508 Three views of deeper and broader skills

simple situation to be evaluated through one of the standard Monte Carlo

spreadsheet add-ins. More complex situations would probably require spe-

cialist assistance.

There are two main situations where Monte Carlo simulation needs to be

used. The first of these concerns when decision-makers wish specifically to

investigate the impact of different assumptions concerning uncertainty. In

such a situation they may well not be happy, for example, with an assumption

that the world can be summarised by just three states such as low, medium

and high demand, each with a specific probability and value attached to it.

This simplification would have been what we often do when we apply the

more sophisticated assumptions approach to value a flexibility. We might, for

example, have assumed that with a flexible feedstock cracker, when we run

with one particular feedstock the resultant margin is a given number. If the

decision-maker was not happy with this degree of simplification of what the

future might be like, then the next step would be to use a Monte Carlo simu-

lation approach to test a larger range of possible assumptions.14

The second situation concerns when non-linearities exist and these are

influenced by more than one uncertainty. The valuing the tail approach

only really works for a single uncertainty. If one needs to investigate mul-

tiple uncertainties in a non-linear situation then a Monte Carlo approach is

required.

A good example of this would be the evaluation of partner carry. If a strong

company has a weak partner and has to lend the partner its share of the ini-

tial investment then what interest rate should the strong company charge in

order at least to earn the CoC on the loan? If there is any chance of the loanâ€™s

not being repaid then the actual interest rate must be higher than the CoC.

The sophisticated assumptions combined with risk monetisation approach

might be to assume that in, say, 10% of the time the partner does not repay at

least some of the loan and that the average loss when this happens is a par-

ticular amount. Armed with these assumptions it would be quite simple to

calculate what premium over the CoC needed to be earned in that percentage

of the time when the loan was repaid. A Monte Carlo approach would allow

ranges and probabilities to be attached to all important variables such that

all possible outcomes could be evaluated. Partner default would not then be

treated as being a single outcome with a fixed probability. It would be studied

I have, for example, worked with a model which enabled one to investigate the effect of volatility and

14

also mean reversion on a plant which had some inherent flexibility. The full details of this were pub-

lished in the Journal of Applied Corporate Finance in spring 2005 in an article entitled â€˜Taking Real

Options Beyond the Black Boxâ€™ which I co-authored with a BP colleague, Fabio Cannizzo.

509 Second view: Valuing flexibility

in more detail and it could result in anything between just the last dollar

of partner carry not being repaid and, in principle, none of the carry being

repaid.

One can also use Monte Carlo simulation to improve the treatment of

growth value. It may just be that a decision-maker was not happy with the

simplification implicit in treating growth as being just a single outcome. If

the decision-maker wanted to test multiple growth possibilities then Monte

Carlo would be used. Testing the potential growth value of an oversized oil

pipeline could be evaluated in this way, particularly if there were several

potential new oil fields which might use the pipeline if it were built. The oil

field and pipeline example, even without growth value, may also need to be

solved by simulation because of the need to model the fact that oil which is

not produced in one year will be available later in the field life when the oil

production would otherwise have fallen below the constraint.

Step 8F: Simple decision trees

A decision tree allows one to depict multiple possible paths into the future.

The approach requires that various decision points are identified and, for

each decision, a range of possible outcomes is identified. Each new outcome

leads to further paths and further decisions. I often recommend the use of

simple decision trees to solve some flexibility valuation problems. Typical sit-

uations would concern the evaluation of research projects and also the value

of information. I will illustrate this through the following example which

investigates whether it is worth spending money on a detailed, and hence

expensive, market survey.

Let us suppose that our company produces a high quality swimming

pool cover. The cover can last for as much as twice as long as a normal cover

but costs 25% more. The life of a pool cover is, however, quite variable as it

depends on how many times it is taken off, how careful the pool owner is

when they do this and how long the cover is exposed to the sun. The result

is that a normal cover will last between two and four years while our com-

panyâ€™s covers will last between four and six years. Take-up of the cover will

depend a lot on our companyâ€™s ability to convey a quite sophisticated eco-

nomic/environmental message to potential customers for whom the cover

will be just one of many costs associated with owning a pool. At present we

have developed three possible scenarios of future demand. In two of these

it is worthwhile for our company to invest but in the low-demand case we

would face a negative NPV.

510 Three views of deeper and broader skills

A detailed survey which investigated the market and tested our proposed

advertising campaign would have a present cost of $1.5m but would give us

a more accurate view of what the future might turn out to offer. The ini-

tial investment in market entry would have a net present cost of $20m. The

three possible outcomes would generate subsequent cash inflows with pre-

sent values of $40m, $25m or $10m. At present we consider that the respect-

ive probabilities of the three scenarios are 20%, 60%, 20%.

At present, therefore, going ahead without carrying out the survey can be

valued via a very simple decision tree as follows:

Outcome

High Demand

PV $40m

NPV $20m

%

20

Overall NPV

Medium Demand

60%

$5m

ñòð. 157 |