Holistic vs.

Hole-istic Exploration and Production Strategies

Ben C. Ball1 & Sam L. Savage
1/28/99

A New Era in Petroleum Exploration and Production Management

Exploration and Production (E&P) is the business of finding petroleum and getting it out of the ground. It
is a risky business in that most exploration projects are total failures while a few are tremendously
successful. Thus the best possible management of risk is crucial.
In the 1930s and 1940s, the development of seismic data collection and analysis substantially reduced
the risk in finding petroleum. The resulting geology and geophysics (G&G) revolutionized petroleum
exploration. Decision analysis has traditionally been applied to the information derived from G&G to rank
projects hole by hole, determining on an individual basis whether or not they should be explored and
developed.
Today this hole-istic approach is being challenged by a holistic one that takes into account the entire
portfolio of potential projects as well as current holdings. This portfolio analysis starts with
representations of the local uncertainties of the individual projects provided by the science and technology
of G&G. It then takes into account global uncertainties by adding two additional Gs: geo-economics and
geo-politics. It thereby reduces risks associated with price fluctuations and political events in addition to
the physical risks addressed by traditional G&G analysis. The holistic approach is based on but not
identical to the Nobel-prize winning portfolio theory that has shaped the financial markets over the past
four decades.
Portfolio thinking in petroleum E&P is based on understanding and exploiting the interplay among both
existing and potential projects. It does not provide all the answers, but encourages E&P decision-makers
to ask the right questions, such as:
v If we want a long-term expected return of, say, 15% on our investment, how can we insure against a
cash flow shortfall over the first three years while minimizing our long-term risk?
v What should we pay for a new project, given the projects already in our portfolio?
v How would oil projects, as contrasted from gas projects, affect the impact of price uncertainty on my
portfolio?
v What projects should we be seeking to reduce the effects of political instability in a given part of the
world?
v What are the effects on financial risk and return if we insist on a minimum of, say, 40% ownership of
any project?
As we shall see, analytical models can help in directly addressing these and similar questions once
decision-makers in the petroleum industry become comfortable with the holistic perspective.

History
The Financial Markets
The story of quantitative portfolio analysis starts in the 1950s with the pioneering work of Harry
Markowitz, who formalized the insight that increased return generally implies assuming increased risk.
1

Member SPE

-1-

His "efficient frontier" of stock portfolios described the optimal investments for investors with differing
aversions to risk. His 1959 book on Portfolio Selection [1], remains an excellent introduction to this
subject. In the 1960's William Sharpe [2] streamlined and expanded on Markowitzs work with his
Capital Asset Pricing Model (CAPM), while Franco Modigliani and Merton Miller 3] made other
important contributions to the theory of valuation of securities. In the early 1970's Fischer Black, Myron
Scholes [4] and Robert Merton [5] determined a rational pricing principle for stock options. All of the
researchers mentioned received Nobel prizes in economics for their discoveries2.
The above body of work changed the face of Wall Street forever, resulting in the widespread acceptance of
an analytical approach to investing, and to the establishment of mutual funds, index funds, and derivative
securities as common financial instruments.3 Today, many elements of this revolution are already
appearing in the energy markets.

Decision and Portfolio Analysis in Exploration and Production

The approach described in this paper is derived from two sources: portfolio theory, as described above,
and decision analysis. Decision analysis was first applied to E&P on a project by project basis by C. J.
Grayson [6] and G. M. Kaufman [7], and popularized by A. W. McCray [8], P. D. Newendorp [9], R. E.
Megill [10 and 11], J. M. Cozzolino [12], and others. In 1968 David B. Hertz [13] discussed the
application of the Markowitz model to risky industrial projects as opposed to stocks. In 1983 an author of
the present article [14] proposed that petroleum E&P strategy be based on the Markowitz model, and
discussed examples based on mainframe computations. In a follow-up article Ball [14] proposed the
development of efficient frontiers for E&P projects using personal computers. Since 1990 the two authors
of this article have collaborated on a sequence of models for consulting clients that have evolved to more
closely reflect the differences between the stock market and E&P projects. Presentations to meetings of the
Society of Petroleum Engineers by Lee Hightower [16] and R. A. Edwards and T. A. Hewett [17] have
also reported on the application of financial portfolio theory to E&P. Hightower [18] authored a follow-up
article in 1997. Also in 1997, Columbia Universitys Lamont-Doherty Earth Observatory founded a
consortium of petroleum firms to share knowledge in the portfolio analysis of E&P projects [19]. The
initial portfolio models used by this consortium, were developed by its director, John I. Howell, and were
based on the approach of the authors of this present paper as described below.
The authors approach deals with the following primary differences between stock and petroleum
investments.
v Stock portfolios depend only on uncertain returns. E&P projects face both local uncertainties
involving the discovery and production of oil at a given site, and global uncertainties involving
prices, politics etc. Furthermore, uncertainties in stock returns usually follow a bell-shaped curve
while E&P uncertainties are highly skewed and stress rare events.
v Risk in stock portfolios is usually measured in terms of volatility, the degree to which the portfolio
swings in value. E&P portfolios must specifically track downside risk.
v The stock market is quite efficient4 whereas the market for E&P projects is inefficient.
v E&P projects pay out over long time periods. Stocks can be bought or sold at will.

The only exception being Black, who sadly passed away a few years before the prize was awarded for his
work with Scholes.
3
For an excellent and very readable history of this revolution in the financial markets, see Capital Ideas
by Peter Bernstein [24].
4
The term efficient is used here in its technical, economic sense. An efficient market is one in which
there are no barriers to each item being priced at its actual value, as determined by all buyers and sellers,
i.e., there are no bargains.

-2-

v A stock portfolio generally contains a small fraction of the outstanding shares of any one company.
An E&P portfolio, on the other hand, often contains 100% of its constituent projects, creating
budgetary effects.
This paper will show how we have dealt with each of the above issues. First a simple example will be
presented to build intuition into portfolio analysis.

Retraining Our Intuition

Suppose you are responsible for investing $10 MM in E&P. Only two projects are available, either of
which would require the full $10MM for a 100% interest. One is relatively "safe," the other relatively
"risky"5. The chances of success are independent. This information is reflected in the Table 1.
This distribution of outcomes may also be displayed as histograms for each project as shown in Figures 1a
and 1b. The expected net present value (ENPV) of each of the two projects can be shown to be the same as
demonstrated in Equations 1 and 2.
ENPVSafe = 60%*$50 + 40%*(-$10) = $26 MM

Equation 1

ENPVRisky = 40%*$80 + 60%*(-$10) = $26 MM

Equation 2

To attach a concrete risk to the outcomes, imagine that if you lose money you will also lose your job. It is
clear that you have only a 40% chance of unemployment with the safe project, and a 60% chance with the
risky one. Since they both have ENPVs of $26MM, you cannot increase your ENPV by investing in the
risky project. Therefore, if you had to choose between one or the other project, the obvious and correct
answer is to invest the $10MM in the safe project.

The Diversification Effect

But suppose you could split your investment evenly between the two projects. Intuition cautions against
taking 50% of your investment out of the safe project and putting it in the risky one. However, let us
examine the four possible joint outcomes. Because the two projects are independent, the probability of any
particular joint outcome is just the product of the probabilities of the associated individual outcomes. Note
that the sum of the probability column must be 100% as we have exhausted all possibilities, as shown in
Table 2. The histogram for the resulting portfolio is displayed in Figure 2.
This allocation of funds still provides an ENPV of $26MM. But now the only way you can lose money is
with two dry holes. Since the projects are assumed to be independent, the probability of two dry holes is
40%*60% = 24%, cutting your risk of unemployment nearly in half! Intuition misleads: You can reduce
risk by taking money out of a safe project, and putting it in a risky one. Although this answer is correct, it
is not so obvious.
This is known as the diversification effect, popularly referred to as not putting all your eggs in one
basket.6
One might consider this line of reasoning so fundamental that the petroleum industry would never do
otherwise. However, the following summary of the conventional project selection process appeared in a
December 1997 article in a leading professional petroleum exploration journal [20]: "Just rank exploration
projects by expected present worth.
5

We shall have much to say about risk shortly. However, in every instance, we mean risk to include
two factors: the probability of an undesired outcome, and how undesirable that outcome is.
6
It is worth noting that the fact that the mass of the histogram of the 50/50 split has moved toward the
center is related to the well-known central limit theorem of probability.

-3-

This suggests that one should rank projects from the best to the worst and then select projects from the top
down until the budget is exhausted, ignoring the diversification effect. In the above example, this strategy
would have lead to allocating all of the funds to the safe project, which has the better risk reward ratio,
but clearly this not the best portfolio.
The authors experience with numerous E&P executives confirms that ranking exploration projects by
expected present worth is the norm. A majority of those informally surveyed favor investing 100% in the
Safe project. A few are aware that a diversification of a portion of the portfolio into Risky will
actually reduce risk, but even these are not aware of how to arrive at an optimal mix.

Shooting the Moon

Some E&P executives feel that a major discovery every decade or so can sustain a large company. They
might claim that they prefer to Shoot the Moon, aiming for the highest possible expected return,
regardless of the risk. The argument is that by ranking their projects in order of expected value, and then
marching down the list until they run out of money, they will in fact arrive at the highest return/highest
risk point the efficient frontier.
This is simply not true. For instance, the company could add projects by borrowing, which would have the
effect of moving them to an even higher risk/higher return point than that achieved through ranking. Thus
high-grading does not deliver the highest point on the efficient frontier at all, but results in an arbitrary
and unexamined tradeoff between risk and return.
We are not arguing that a company should explore with borrowed money. We are claiming that highgrading ignores a universe of other portfolios, both riskier and safer, that can be illuminated through our
approach.

Measures of Risk
Uncertainty is an inherent characteristic of our universe. Risk, on the other hand, is in the eye of the
beholder. The nave measure of risk we have used above (the probability of getting fired) was chosen here
for illustrative purposes because it is easy to visualize and calculate. In practice, one would use a more
sophisticated measure; for example, one that penalized larger losses more than small ones. However,
diversification has a similar effect in reducing almost any sensible measure of risk. And, to the point here,
full exploitation of the advantages of diversification are anything but intuitive. See Appendix 1 for a
further discussion of risk measures.

Statistical Dependence and Its Sources

The above example assumes that the two projects are independent. In general, the interplay between two
projects is more complex, in that their economic outcomes are interrelated. This is known as statistical
dependence. The simplest type of statistical dependence is correlation, which comes in two varieties.
v Positive Correlation - A given outcome for one project increases the chance of an outcome in the
same direction for the other. This diminishes the effect of diversification.
v Negative Correlation - A given outcome for one project decreases the chance of an outcome in the
same direction for the other. This enhances the effect of diversification.
Lets consider the effects of positive correlation on a 50/50 split between safe and risky. We assume that
the individual distributions of safe and risky remain as above. Now, however, in the event that safe
succeeds, risky is more likely to succeed, and in the event that safe fails, risky is more likely to fail.

-4-

Thus there is still a 40% chance that safe will fail, but if it does, the chance that risky fails is greater
than 60%. Therefore the probability of losing your job is now greater than 24%. Next consider the effect
of negative correlation. In the event that safe succeeds, risky is less likely to succeed, and in the event
that safe fails, risky is less likely to fail. Thus there is still a 40% chance that safe will fail, but if it
does, the chance that risky fails is less than 60%. Therefore the probability of losing your job is now less
than 24%. It is important to note that correlation affects only the risks. The expected value of $26MM
remains unchanged. In the portfolio approach, risk is managed by spreading the investments across a
number of opportunities while avoiding positive correlations and seeking negative correlations. It is
possible to do this optimally over numerous opportunities under diverse constraints, as the rest of the
paper displays.
It is easy to understand the importance of correlation when one considers a fire insurance policy on a
house. Since the insurance has a positive ENPV to the insurer, we know it has a negative ENPV to the
insured, and is therefore a bad investment7.
Statistical dependence may be due to many sources. The four listed below are not meant to be exhaustive,
but are widely encountered in E&P projects.8
v
v
v
v

Places
Prices
Profiles
Politics

Places
The economic outcomes of two E&P sites in very close proximity (for example in the same field) will be
positively correlated through geological similarities, and would not constitute a very diversified portfolio.
On the other hand two sites in widely distant locations will display little or no geological correlation, and
hence would be more diversified. Places can have corresponding implications for pricing (especially
gas) and political issues as well as for geologic ones.

Prices
Petroleum projects produce crude oil and natural gas in various proportions. Crude oil prices generally
track each other very closely worldwide. Thus, the economic outputs of oil projects worldwide are
positively correlated relative to fluctuations in crude price. However, this is not true for natural gas.
Natural gas prices in many parts of the worldnotably in the United Statesdo not track either world
crude oil prices or each other very well. Thus there would be a tendency for a portfolio consisting of a gas
project and an oil project to be less positively correlated and therefore better diversified, relative to price,
than a portfolio consisting of two oil projects.

Unless your portfolio also includes a house (whose value is of course negatively correlated to the value of
your insurance policy).
8
Although we use projects throughout this paper, the same principles apply equally to E&P sites,
exploration prospects, development projects, and/or acquisition properties. The point is that E&P
companies can add reserves in three ways: exploration, development (and redevelopment) and acquisition.
Each has a very different risk/reward profile, the integrated analysis of which could reveal innovative,
optimal portfolios. Most E&P companies isolate these three functions in their respective functional silos
and, at best, suboptimize each. [25]

-5-

As an example of this phenomenon, The Wall Street Journal reported in 1993 [21] that the economy of
Houston, which had suffered during the crude oil price drop of 1986, had weathered a subsequent price
drop successfully because it had diversified between oil and gas.

Profiles
A frequent concern is the timing of the flows of various elements of projects, which may extend for many
years into the future. These flows might include such elements as cash flow, hydrocarbon production,
reserve additions, and staff requirements. Often the more nearly constant these flows can be, the better.
The correlation among these elements can be taken into consideration to minimize fluctuations in cash
flows. These critical factors that can literally make or break a company can now be considered and
managed explicitly.
As an example, consider how the correlations among project cash flows might be managed: Figure 3a
shows the cash flow profile expected from two projects. If they were to comprise the portfolio, the
resulting cash flow for the portfolio would be as shown in Figure 3b, with low valleys and high peaks.
However, if two projects with expected cash flow profiles as shown in Figure 3c were to comprise the
portfolio, the resulting cash flow for the portfolio would be as shown in Figure 3d. The peaks and valleys
are leveled outa much more desirable cash flow profile.

Politics
Petroleum investments have always been subject to political uncertainties, from the anti-trust decision
against Standard Oil of 1911 through environmental regulations, to the Gulf War of 1991 and beyond.
Projects subject to disruption in the same direction due to the same political event will be positively
correlated. Negative correlation of projects may also be induced through political uncertainty. For
example, consider two politically distinct regions that supply natural gas through two different pipelines
to a single market. The political disruption of production in either of the two regions could lead to market
shortages, and hence to increased prices and/or demands for the non-disrupted region. A portfolio
consisting of one project in each negatively correlated region would thus be protected, or hedged,
against political risk in either region.

The Efficient Frontier

We have seen that a combination of a safe project and a risky project can be less risky than a pure
investment in the safe project. This is displayed graphically in Figure 4. In this example, there were only
two potential projects, and they had exactly the same ENPV. However, there will generally be projects of
various expected returns and risks as depicted in Figure 5. By calculating the minimum risk portfolio for
each of several levels of expected return, one can arrive at the curve shown in Figure 6. This curve shows
the optimum trade-off between risk and return and is known as the efficient frontier.
Moving north from the efficient frontier increases risk without increasing expected return, while moving
west decreases expected return without decreasing risk. Since each point on the efficient frontier has
minimized the risk for that level of expected return, no portfolios exist to the southeast of the frontier.
Thus the best portfolios are those on the efficient frontier itself. The concept of the efficient frontier and
the method of finding it were the fundamental Nobel Prize-winning contributions of Markowitz in the
1950s. No rational person would wish to be at any point above the efficient frontier. But which point
should you pick? That depends on your firms willingness to suffer short-term volatility in the interest of
long-term growth. In E&P one may also develop additional frontiers showing optimal trade-offs between
reserve additions, budget level, cash flow shortfall, or other meaningful metrics.

Important Differences Between Stocks and Petroleum Projects

As mentioned earlier, the fundamental differences between stock returns and petroleum projects require
modifications to the standard financial portfolio models.

-6-

Some primary differences in the underlying assets are:

v
v
v
v
v

Types of Uncertainties
Risk Measures
Nature of Markets
Timing Considerations
Budgetary Effects

Table 3 reviews these in more detail.

E&P Portfolio Optimization Model (EPPO)

This section presents a simple E&P portfolio optimization model (EPPO), which unlike most standard
financial portfolio models, can accommodate the characteristics of petroleum projects listed in Table 3.
Excel versions of EPPO.xls may be downloaded from http://www-leland.stanford.edu/~savage in formats
for optimization with the Excel Solver, and WhatsBest!.
As discussed earlier, the Markowitz and Sharpe models were intended for stock portfolios, and are not
ideally suited to portfolios of petroleum projects9. EPPO is based on two technologies already in wide use
individually in the petroleum industry: Monte Carlo simulation and linear programming. Here these
technologies have been combined to create a single period stochastic linear program [22]. The advantages
of this model for our purposes are:
v It allows for arbitrary realistic probability distributions of project outcomes as opposed to multivariate
normal or log normal.
v It supports a wide variety of risk measures.
v It does not require historical data, but instead may be based on simulations, decision trees or other
types of input
v It estimates not just the mean and variance of the portfolio, but the entire distribution of outcomes
EPPO works by feeding Monte Carlo trials into a linear program, which then finds a portfolio of
minimum risk for a given expected NPV. This process is repeated for a range of desired NPVs, thereby
determining the efficient frontier and the portfolios that comprise it.

The Spreadsheet
The basic elements of EPPO.xls are described in Figure 7. The program flow and underlying algebra are
detailed in Appendix 2.
In the example shown here, the task is to optimize a portfolio from five exploration projects, A through E,
given a particular budget.

For comparison, a spreadsheet version of the Sharpe model is included in the solver example file that
ships with Microsoft Excel. A Markowitz model is included with INSIGHT.xla [26] See also
www.AnalyCorp.com.

-7-

Results
The results shown in Figures 8 a, b, and c were produced by running the EPPO model with a budget of
$600, and successive required ENPVs of $1900, $2000, $2100, $2200, and $2300. Figure 8a shows the
resulting risk/return trade-off curve running from an ENPV of 1900 and a Mean Loss10 of 100 to an
ENPV of 2300 and a Mean Loss of nearly 300. We refer to this curve as the Internal Efficient Frontier
because it represents the best the firm can do with investments among its own projects. That is, each point
on this efficient frontier represents the highest expected value at that level of risk, or, equivalently, the
lowest risk for that expected value. There are no portfolios possible southeast of this frontier. All
portfolios northwest of the frontier are inferior to a portfolio on the frontier, in that they offer an
unnecessarily low return and/or an unnecessarily high risk.
A firm not applying portfolio analysis would be unlikely to be on such an internal efficient frontier. We
have denoted such a position by X in Figure 8a. Such a firm could increase its expected return at constant
risk by moving to Z, or decrease its risk at constant return by moving to Y, or employ some intermediate
strategy.
For each level of ENPV, the stacked bars of Figure 8b show the makeup of the efficient portfolio for each
required ENPV. The vertical bars show the budget allocation to each of the five available projects in each
efficient portfolio. Figure 8c combines both of these graphs to show a complete picture of the risk return
trade-offs and the associated portfolios.
Management must choose such a point on the frontier, whereupon the underlying portfolio associated with
that point is revealed. For example, assume Management decided that at a budget of $600, an ENPV of
$2,200 is needed and a Mean Loss of $171 is acceptable. Then portfolio analysis would reveal that the
budget should be allocated as follows to give the portfolio that would be most likely to yield that
performance: 6% in Project A, none in Project B, 8% in Project C, 22% in Project D, and 64% in Project
E.11
By contrast, current practice would be to decide separately on an investment level in each of the five
projects, based largely on the intrinsic merits of each. The resulting portfolio would likely be northwest of
the internal efficient frontier, like the point X in the Figure 8a. The total risk would be higher than
necessary, or the expected value would be lower than necessary, ormost likelyboth.

Generalizing to multiple time periods

One thing that EPPO and the Markowitz model do have in common is that both model current decisions
only, based on potential future risks and rewards. A useful generalization would model both current and
future decisions based on potential future risks and rewards. This would, in effect, be a marriage of
portfolio theory and real options theory, and might be accomplished through either multi-period stochastic
linear programming [22], or dynamic programming [23].

Business Implications: Asset Interplay Management

The fundamental point of view of this paper is that project by project or hole-istic analysis misses many
important insights provided by the holistic perspective. The implications for the E&P business are that
management must place at least as much emphasis on the interplay among projects as it does on the
10

Mean Loss, as defined and described in Appendix 1, is a particularly simple risk measure. Other risk
measures, as also described in Appendix 1, may be used.
11
These results are no more than the results of a first iteration. For example, if 6% working interest in
Project A is not practical, then the program should be rerun after placing the appropriate constraints on
the working interest in Project A, e.g., 10%. Such iterations should be continued until all results are
within the realm of practicality. This, then, will represent the optimum practical portfolio, which, of
course, is the only one that matters in the real world.

-8-

projects themselves. We refer to this as Asset Interplay Management, and believe that it is currently not
adequately exploited by the industry. Table 4 shows some examples of how various types of corporate
decision processes might be transformed by Asset Interplay Management.
It is tempting to think of Asset Interplay Management as just another analytical tool or computer program.
The danger in this view is that the tool or program will be adopted, and wont deliver, thereby
"inoculating" the company against a successful case of portfolio optimization for a generation of
management. If the portfolio approach is to meet the expectations it is generating, top Management must
understand that it represents a fundamentally new way of thinking about the business of E&P. This will
require:
v Re-educating management, to develop and Asset Interplay Management until they become intuitive
v Re-structuring corporate systems to collect and interpret stochastic data from a global as well as local
perspective
v Revising reward programs to provide incentives for overall risk/reward positioning of the firm
Each of the above implications of Asset Interplay Management is valuable in itself, and offers insights not
available through project-by-project analysis. Each avoids some of the subtle but systemic errors to which
the industry is presently vulnerable. But these are just first steps as E&P enters the dawn of this new style
of management.
At the beginning of this article, we posed several questions that Management should be asking, but cannot
adequately be answered without Asset Interplay Management.
v If we want a long-term expected return of, say, 15% on our investment, how do we insure against a
cash flow shortfall over the first three years?
This would require determining the optimum portfolio while constraining the first three-year
cash flow to 0.
v What should we pay for a new project, given the projects we already have in our portfolio?
This could be determined by comparing the values of the portfolio with and without the new
project, but at constant risk. (This is almost certain not to be the projects NPV.)
v How would oil projects, as contrasted from gas projects, affect the impact of price uncertainty on my
portfolio?
Since Asset Interplay Management explicitly takes price interplay into account, the effect of each
project on the portfolios robustness relative to price instability can be determined.
v What projects should we be seeking to reduce the effects of political instability in a given part of the
world?
Since Asset Interplay Management explicitly takes political interplay into account, the effect of
each project on the portfolios stability relative to political stability can be determined
v What are the effects on financial risk and return of insisting on a minimum of, say, 40% ownership of
any project undertaken?
This would require determining the difference in portfolio value, at constant risk, with and
without the 40% constraint.

Conclusions
v Portfolio principles developed for the financial arena must be modified before application to E&P.
v The portfolio perspective empowers decision-makers to focus on critical business issues, which guide
asset interactions, viz., places, price, profiles, and politics.
v Portfolio management empowers decision-makers to manage risk, as well as measure it.
v Changes in perspective, intuition, and culture will do more to promote Asset Interplay Management
than new computer programs.

-9-

Modern portfolio theory provided the conceptual underpinning for the financial engineering that now
dominates Wall Street. How this will play out in the area of E&P remains to be seen. But one thing is
certain: Asset Interplay Management offers new and powerful tools for dealing head-on with one of the
elements which distinguishes the upstream business, but which too long has been handled only
subjectively: RISK.

Acknowledgements
We are indebted to several of our colleagues for reviewing this manuscript and for
making many helpful suggestions. We are especially grateful to Jerry Brashear, John
Howell, and Dick Luecke for their extraordinary effort in making extremely detailed
reviews, that resulted in many improvements in the finished work.

About the authors

Ben Ball12 and Sam Savage have a professional relationship dating from 1986. Since 1990 they
have been working together in the area of portfolio optimization for petroleum exploration and production
projects.
Ben C. Ball, Jr. is an internationally
recognized petroleum expert. For twenty years he has
consulted to several dozen firms on four continents,
ranging from very small to extremely large, and to
several governments and agenciesstate, national, and
international, in addition to serving as expert witness in
several dozen cases. For the last twenty-two years he
has also held teaching and research appointments at
M.I.T., including that of Adjunct Professor of
Management and Engineering. Over 75 of his articles
have appeared in technical, professional, and
management journals and books, including Harvard
Business Review, Petroleum Management, Technology
Review, and the European Journal of Operational
Research. His book, Energy Aftermath, which he coauthored with two M.I.T. colleagues, has been
published by Harvard Business School Press.
He received BS and MS degrees in chemical
engineering from M.I.T., and completed Harvard
Graduate Business Schools Advanced Management
Program. He retired from Gulf Oil Corporation as
Corporate Vice President after thirty years in operations
and planning.

Dr. Sam L. Savage is Director of Industrial

Affiliates for Stanford Universitys Department of
Engineering Economic Systems & Operations. He
received his Ph.D. in computer science from Yale
University. After spending a year at General Motors
Research Laboratory, he joined the faculty of the
University of Chicago Graduate School of Business,
with which he has been affiliated since 1994. In 1985
he led the development of a software package that
couples linear programming to Lotus 1-2-3. This
popular package, called WhatsBest!, won PC
Magazines Technical Excellence Award in 1986. Sam
consults widely and has served as an expert witness. His
executive seminars offered to industry and government
have been attended by over 2000
Dr. Savage's INSIGHT.xla, Business Analysis
Software for Excel published in February 1998 is
receiving wide acclaim. In his foreword to this work,
Harry Markowitz, Nobel Laureate in Economics, says,
Rarely has such sound theory been provided in such an
entertaining manner. See www.AnalyCorp.com.

Bibliography
1.

12

Markowitz, H. M. Portfolio SelectionEfficient Diversification of Investments, second edition,

Blackwell Publishers, Inc., Malden, MA (1957, 1997).
Member, SPE.

-10-

2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.

17.

18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.

Sharpe, William F., Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of
Risk, Journal of Finance, vol. XIX, No. 3 (September 1964) 425-442.
Modigliani, Franco and Merton H Miller. The Cost of Capital, Corporation Finance, and the Theory
of Investment, American Economic Review, vol, 48, No. 3 (1958) 655-669.
Black, Fischer and Myron Scholes, The Valuation of Option Contracts and a Test of Market
Efficiency, Journal of Finance, vol. 27 (1972) pp. 399-418.
Merton, Robert C., Theory of Rational Option Pricing, Bell Journal of Finance and Management
Science, vol. 4 (1973) 141-183.
Grayson, C. J., Decisions Under Uncertainty: Drilling Decisions By Oil and Gas Operators, Harvard
Business School, Boston, MA (1960).
Kaufman, G. M., Statistical Decision and Related Techniques in Oil and Gas Exploration, PrenticeHall, Englewood Cliffs, NJ (1963).
McCray, A. W., Petroleum Evaluations and Economic Decisions, Prentice-Hall, Englewood Cliffs,
NJ (1975).
Newendorp, P. D., Decision Analysis for Petroleum Exploration, The Petroleum Publishing Co.,
Tulsa, OK (1975).
Megill, R. E., An Introduction to Risk Analysis, PennWell Books, Tulsa, OK (1977).
, Exploration Economics, PennWell Books, Tulsa, OK (1979).
Cozzolino, J. M., Management of Oil and Gas Exploration Risk, Cozzolino Associates, Inc., West
Berlin, NJ (1977).
Hertz, David B., Investment Policies That Pay Off, Harvard Business Review, vol. 46, No. 1
(January-February 1968) 96-108.
Ball, Ben C., Managing Risk in the Real World, European Journal of Operational Research, vol.
14 (1983) 248 261.
, Profits and Intuition, Petroleum Management (November 1987) 30-33.
Hightower, M. L. and A. David, Portfolio Modeling: A Technique for Sophisticated Oil and Gas
Investors, SPE Paper 22016, presented at the 1991 SPE Hydrocarbon Economics and Evaluation
Symposium, Dallas, TX, April 11-12, 1991.
Edwards, R. A. and T. A. Hewett, Applying Financial Portfolio Theory to the Analysis of Producing
Properties, SPE paper 30669, presented at the 70th Annual Technical Conference and Exhibition of
the SPE, Dallas, TX, October 22-25, 1995.
, Aram Sogomonian, and Jan Stallaert, Banking on Monte Carlo, Energy & Power Risk
Management, vol. 2, No. 5 (September 1997) 12-14.
Lamont-Doherty Earth Observatory Consortium web-site:
http://www.ldeo.columbia.edu/4d4/portfolio/.
Downey, Marlan, Business Side of Geology, AAPG Explorer (December 1997).
Wall Street Journal, Regional Resilience - Tumbling Oil Prices Wont Batter Texas The Way 86
Crash Did, Monday (December 6, 1993).
Infanger, Gerd. Planning Under Uncertainty, Boyd & Fraser Publishing, Danvers MA (1994).
Luenberger, David G., Investment Science, Oxford University Press (1998).
Bernstein, Peter L., Capital IdeasThe Improbable Origins of Modern Wall Street, The Free Press,
New York (1992).
Brashear, Jerry Paul, Private Communications (July 13, 1998 and December 8, 1998).
Savage, Sam L, Insight.xlaBusiness Analysis Software for Microsoft Excel, Duxbury Press,
Pacific Grove, CA (1998).
Froot, Kenneth A., David S. Scharfstein, and Jeremy C. Stein, A Framework for Risk Management,
Harvard Business Review (November-December 1994) 91-102.
Kenyon, C. M., Sam L. Savage, and Ben C. Ball, Equivalence of Linear Deviation About the Mean
and Mean Absolute Deviation About the Mean Objective Functions, Operations Research Letters
(1999).

-11-

Appendix A: Risk Measures

Variance
Variance (2) has been the traditional measure of uncertainty both among statisticians and financial
portfolio managers. The variance of an uncertain quantity is defined to be the average of the square of the
deviation of the quantity from its mean. Because variance is measured in squared units (square dollars in
the case of portfolio risk), it is also common to use the square root of variance or the standard deviation,
as a measure of uncertainty.
Because variance measures the square of the deviation from the mean it is a symmetric risk measure as
shown in Figure A-1.
By its nature, the variance penalizes large deviations increasingly harshly, and, because it is symmetric, it
penalizes upside deviations equally with downside ones.
When the underlying distributions of uncertainties are relatively symmetric as with stock prices, the
variance is an appropriate measure of risk, see Markowitz [A-1] for example. However with asymmetric
distributions such as the outcomes of petroleum projects, variance is not a desirable measure. For example
the three projects shown in Table A-1 (taken from Schrage [A-2]) all have the same mean and variance,
yet have obvious differences from a realistic risk perspective. This can be seen even more clearly by
viewing the corresponding distributions, as shown in Figures A-2a, A-2b, and A-2c.
These projects are obviously quite different from each other. B, for example, has no chance whatsoever of
loss. Yet for all three projects, the Mean = 10 and Variance = 400. Thus they would be indistinguishable
using the mean and variance criteria customarily and appropriately used for stock portfolios.

Mean Absolute Deviation MAD

The Mean Absolute Deviation (MAD) is an alternative measure of risk [A-3] that is sometimes
advantageous over variance for the following reasons:
v It may be applied directly to historical or Monte Carlo generated data regardless of distribution
v It may be minimized using Linear Programming
v It may be adapted to provide a wide variety of asymmetric measures of risk
MAD is calculated as follows. Suppose you had run M Monte Carlo trials of an uncertain outcome. We
will define j to be the jth outcome minus the mean of all outcomes. Then the mean absolute deviation of
the uncertainty is defined by Equation 3.

MAD =

1 M

M j =1 j

Equation 3

Notice that if we replaced the absolute value operator by the squaring operator we would be back to the
variance.
A picture of the MAD risk function is shown in Figure A-3.
This penalizes deviations linearly. Notice that like the variance, the upside deviations are penalized
equally with downside ones.

-12-

Mean Loss and other adaptations of MAD

A simple but useful adaptation of MAD, used in EPPO.xls, is Mean Loss, in which measures only the
deviation below 0 (not below the mean), as shown in Figure A-4. The mean loss of the three examples
above is shown in Table A-2, clearly indicating that B is the least risky.
Mean loss is the average of ones losses. That is if you have a 50% chance of winning $1 or losing $1 the
mean loss is 50. Mean loss has the desirable feature of distinguishing between upside and downside risk.
However, like MAD it is a linear loss measure. That is, if you had a million dollars, and started losing
money, mean loss would impose the same penalty on losing your 1st dollar that it does on losing your last
dollar.
It is more realistic for the penalty to increase with deviation, as it does in the case of variance. The MAD
model may be adapted further to model any piece-wise linear convex penalty F() such as that displayed
in Figure A-5. Note that any number of straight-line penalty slopes and even a constraint on may be
imposed.
To create an LP model of this risk function, define two new LP variables for each scenario as follows, y1j
0 and y2j 0. These variables are constrained as shown in Equations 4, 5, 6, and 7.
y1j -j-c

Equation 4

y1j d-c

Equation 5

y2j -j-d

Equation 6

y2j< e-d

Equation 7

For each scenario define Fj as in Equation 8.

F j = a y1j + b y2j

Equation 8

Then the overall objective of the LP is to minimize the function defined in Equation 9.

1 M
F
M j =1 j

Equation 9

In this way, customized risk measures may be constructed. One could even create a piece-wise linear
approximation of the variance or semi-variance (squared deviation below the mean) if desired.
A customized risk measure suitable for production projects can easily be developed from the paradigm
suggested in Figure A-5. In production projects, the danger is not so much an outright loss as it is an
erosion of value. A risk parameter which measured Mean Value Erosion would simply involve setting c
in Figure A-5 at a point above zero equal to the required economic threshold.

-13-

Appendix B: Notation and Algebraic Representation of EPPO

The Model
Table B-1 shows the notation used in EPPO.XLS model, while Table B-2 shows the model flow.

Algebraic Formulation
This EPPO.XLS model may be run with a variety of risk measures. As formulated here, it uses a measure
we call mean loss, which is particularly easy to calculate and understand intuitively, but is not appropriate
in all cases. The mean loss is the average of the economic losses over all Monte Carlo trials. Those trials
without losses are averaged in as 0s. See Appendix 1 for a discussion of other risk measures and how to
model them.
To model mean loss, we introduce m new variables yi to record the loss under each Monte Carlo trial.
Then we minimize mean loss as shown in Equation 10, subject to the restrictions given in Equation 11,
where the elements of the portfolio P are between 0 and 1.

MeanLoss =

1 m
y
m i =1 i

Equation 10

yi P Ti , i = 1... m , P A D , and yi 0

Equation 11

Bibliography
1.
2.
3.

Markowitz, H. M. Portfolio SelectionEfficient Diversification of Investments, second edition,

Blackwell Publishers, Inc., Malden, MA (1957 and 1997).
Schrage, Linus, Lindo: An Optimization Modeling System, Boyd & Fraser Publishing Co., Danvers,
MA (1991).
Konno, H. and Yamazaki, The Mean Absolute Deviation Portfolio Optimization Model and Its
Applications to the Tokyo Stock Market, Management Science, vol.37 (May 1991) 519-531.

-14-

Tables
Table 1
Outcome

Safe Dry Hole

Success
Risky Dry Hole
Success

NPV
$MM
-10
50
-10
80

Independent
Probability
40%
60%
60%
40%

Table 2

1
2
3
4

Outcomes of Investing 50% in Each Project

With Statistical Independence
Safe
Risky
Probability
Return in $MM
Result
Success
Success
60% x 40%
50% x $50 + 50% x $80 = $65 Keep Job
=24%
Success
Dry Hole
60% x 60%
50% x $50 + 50% x (-$10) = $20 Keep Job
=36%
Dry Hole Success
40% x 40%
50% x (-$10) + 50% x $80 = $35 Keep Job
=16%
Dry Hole Dry Hole
40% x 60% 50% x (-$10) + 50% x (-$10) = (-$10) Lose Job
=24%
ENPV
24% x $65 + 36% x $20 + 16% x $35 + 24% x (-$10) = $26

Table 3
Stock Portfolios

E&P Projects

Types of Uncertainties
E&P projects face both local uncertainties
Stock portfolio models are primarily
involving the discovery and production of
based on price uncertainty.
oil at a given site, and global uncertainties
involving prices, politics etc.

a.

The uncertainties of future stock returns The economic uncertainties of E&P are
are generally symmetric and bell shaped. anything but normal.

Distribution of an E&P Project Return

Mean Return = 2

Economic
Success

Dry

0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0

Relative Likelihood

0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
-2

Relative Liklihood

Distribution of a Stock Return

Mean Return = 2

-2

b.

Return

Return

c.

Estimates of distributions and statistical

dependence are based at least in part on
past history

Distributions and statistical dependence of

price information may be based on history.
Other uncertainties must be modeled
through decision trees or simulation.

Risk Measures
The risk experienced by an investor in the
stock market is generally expressed in terms
of the variance of the portfolio. This
penalizes both upside and downside
deviations equivalently, which makes sense
for the symmetric distributions shown
above.

Risk considerations for portfolios of

petroleum investments depend on the
concerns of management. Appropriate risk
measures might be a cash shortfall in a
given year, or a failure to achieve desired
reserve addition by a given time. Because of
the highly asymmetric distributions of
outcomes, penalizing upside deviations does
not make sense, and variance is not an
appropriate risk measure.

Nature of Markets
Stock markets are quite efficient. One of
the consequences is that the price one pays
for a financial instrument is pretty much
what it is worth. Therefore there are no
bargains. In fact, many argue, with good
reason, that one should not waste time
designing stock portfolios and should invest
entirely in index funds. These efficient
markets give continuing feedback on values.

The market for petroleum E&P projects is

not efficient. In addition, a project may have
a very different value for one firm than it
does for another. Hence there are
bargains and bad deals. Portfolio
analysis is precisely the way to determine if
a particular deal is good or bad for your
firm. This inefficient market gives
essentially no feedback on values.

Stock markets deal only with stocks and

bonds.

Petroleum portfolios include currently

owned assets, exploration prospects,
development projects, and/or acquisition
properties.

Timing Considerations
Stock portfolio analysis traditionally does
not model time explicitly since stocks can
be readily bought or sold at any time.

Petroleum investments have cash flows that

play out over long periods of time.
Therefore, time must be modeled explicitly.

Budgetary Effects
Stock portfolio models generally ignore the
size of the budget. An efficient milliondollar portfolio is simply 1,000 times the
size of an efficient thousand-dollar
portfolio. Stock portfolios are concerned
only with the proportions of various assets
held, regardless of the size of the budget.

Petroleum portfolios are budget-dependent.

Once one has taken a 100% interest in an
E&P project, no more may be invested in
that project. Therefore, different size
budgets will have different proportions of
the various projects in their respective
optimal portfolios.

Table 4
Current Practice

Asset Interplay Management

1. Selecting a set of E&P projects to fund

Select the set of projects that achieve
Rank (high-grade) the projects, start at the
top of the list and go down until the budget optimal trade-offs of various risks and
economic factors.
is exceeded.
2. Determining long-term vs. short-term goals
Establish long-term return on investment1 Make informed trade-offs between longand growth goals independent of their
term goals for growth and the risks of
implications for the concomitant long-term short-term volatility, failure to meet reserve
risk or short-term volatility.
requirements or other measures of risk.2
3. Dealing with political and environmental risk
Political and environmental risks are
Political and environmental risk of the
considered subjectively on a project by
entire portfolio is managed by taking the
project basis.
interplay among projects into account.
4. Evaluating a project for purchase or sale
Determine the projects market value,
Determine what the project is worth in the
that is, what other firms might pay for it, or context of the firms current holdings.
base its value on an estimate of its ENPV.
Remember, I wouldnt spend a dime on a
policy to insure your house, but an identical
policy on my house is valuable to me, even
though it has a negative NPV.
5. Determining the risk related cost of constraints and policies
The costs of company policies and external The implications of constraints and
constraints are appraised subjectively.
company policies can be evaluated on a
risk/return basis.3

Or return on capital employed, or similar metric.

It is important to note that this approach captures the insights underlying utility theory, preference
curves, etc., while avoiding the practical difficulties that arise from their explicit application. See Kenyon,
Savage, and Ball [28].
2

Constraints often reflect strategic issues of concern to top management, e.g., reserves replacement,
cash flow for debt repayment, etc. Relative to stock market valuation, these may at times be as important
or more important than ENPV. However, these constraints can be incorporated into the portfolio analysis
and evaluated for their effects on risk and return. Brashear [25]

Determining Strategic Criteria for Future Exploration

Missing ingredients or strategic weaknesses Missing ingredients or strategic thrusts that
are difficult to identify.
would contribute to the robustness of the
portfolio are identified. The result is a
shopping list of desirable qualities for
additional projects or programs, given your
firms current portfolio and situation.4
6.

7. Increasing the Value of the Firm

Sole focus is on expected net present value.

The focus of real E&P companies is never

solely on expected net present value, but
includes reserves replacement, debt ratios,
cash flows, etc. At one extreme, risk is
almost entirely ignored, but Froot,
Scharfstein and Stein contend [27] that
there are situations in which risk reduction
can increase the value of the firm by
assuring that cash flow is available when
needed for critical investment.

Table A-1
Project Outcome NPV Probability
Mean
$MM
-10
50%
-10*.5+30*.5=10
Failure
A

Variance
.5*(-10-10)2+.5*(3010)2=400

Success
Failure

30
0

50%
80%

0*.8+50*.2=10

.8*(0-10)2+.2*(5010)2=400

Success
Failure

50
-30

20%
20%

-30*.2+20*.8=10

.2*(-30-10)2+.8*(2010)2=400

Success

20

80%

Current holdings, of course, represents, the mass of most portfolios. Hold and produce takes little or
no capital, requires no overt decisions, and yields great returns, especially if opportunity costs are ignored.
However, serious consideration of the trading of current holdings (i.e., the sale of current assets in
exchange for the purchase of new ones) is usually not seriously considered in any systemic way. However,
such a study could significantly enhance the efficient frontier. In any event, a crucial point is that both
current holdings and new projects opportunities must be evaluated together, as the portfolios risk depends
on the ways in which all of its constituent parts interact. Brashear [25]

Table A-2
Project Outcome NPV Probabilit
Mean Loss
$MM
y
-10
50%
10*.5 + 0*.5 = 5
Failure
A
50%
Success 30
0
80%
0*.8 + 0*.2=0
Failure
B
50
20%
Success
-30
20%
30*.2 + 0 *.8=6
Failure
C
20
80%
Success
Table B-1
Notation
n - Number of projects under consideration (5 in our example)
m- Number of Monte Carlo trials (40 in our example)
P - The portfolio. This is a vector of length n consisting of the
working interest (between 0 and 100%) in each project. EPPO
assumes that any working interest is possible, however, the model
may be modified to force an all or nothing policy, or some other
minimum or maximum working interest.
Ti - The ith in the sequence of Monte Carlo Trials modeling the joint
uncertain NPVs of the projects. Ti is a vector of length n.
A - The average NPV of each project over the m trials, a vector of
length n.
NPVi(P) - The NPV of portfolio P under Monte Carlo trial i.
R(P) - A risk measure associated with portfolio P, calculated from
NPVi(P), i=1..m. In EPPO, R(P) is the Mean Loss of P.
D- A desired level of ENPV

Table B-2

Model Flow
Monte Carlo A Monte Carlo simulation of the joint economic outcomes of the
projects is run, based on the local uncertainties, geo-technology
Simulation
and geo-science, and on the global uncertainties, geo-economics
and geo-politics. The statistical dependence between projects must
be preserved. The trials, Ti , i=1..m are stored.
Calculate
ENPV and
Distribution
of NPVs

ENPV of the portfolio P is P.A = SUMPRODUCT(P,A)

NPVi(P), the NPV of the portfolio P under the ith Monte Carlo
Trial is
P. Ti = SUMPRODUCT(P, Ti), i=1..m.
The set of all NPVi(P), i=1..m provides an estimate of the
probability distribution of the NPV for portfolio P.
The chosen risk measure, R(P) is calculated from NPVi(P), i=1..m
Choose a low level (D) of desired ENPV

Optimize

The linear program finds a portfolio P which minimizes the chosen

risk measure R(P), subject to ENPV D
If the linear program fails to find a solution, Halt.
Otherwise, increase D and go back to previous step

Figures
Figure 1 a and b
Distribution of Outcomes of Risky
Project
1.00
ENPV
0.90
$26
MM
0.80

Distribution of Outcomes of Safe

Project
ENPV
$26 MM

1.00
0.90
0.80

0.70
Pr
ob 0.60
ab
0.50
ilit
y 0.40

Probability

0.70
0.60
0.50
0.40

0.30

0.30

0.20

0.20

0.10

0.10

0.00
85

80

75

70

65

60

55

50

45

40

35

30

25

20

15

10

-5

-10

-15

0.00

- - -5 0
15 10

10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85
Return

Return

Figure 2
1.00
0.90

Distribution of Outcomes of 50/50 Split

0.80

Chance of Losing Job 24%

0.60
0.50

ENPV Still 26MM

0.40
0.30
0.20
0.10

Return

85

80

75

70

65

60

55

50

45

40

35

30

25

20

15

10

-5

-10

0.00
-15

Probability

0.70

Figure 3 a, b, c, and d
Projects with dissimilar cash flow profiles
may be negatively correlated over time.

2
0
-2

10 11 12 13 14 15 16

Cash Flow

Cash Flow

Projects with similar cash flow profiles will be

positively correlated over time.

-4

2
0
-2

Years

A portfolio consisting of a 50% share in each of the

above cash flows (shown in solid line below) results
in wide swings from negative to positive cash flow.

A portfolio consisting of a 50% share in each of the

above cash flows (shown in solid line below) results
in a relatively smoother cash flow.
8

8
6

Cash flow of combined

projects shown as solid line

4
2
0
1

10 11 12 13 14 15 16

Cash Flow

Cash Flow

10 11 12 13 14 15 16

-6

Years

6
4
2
0
-2

-4

-4

-6

-6

Figure 4
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

Risky
Safe
Mix of Safe
and Risky

24

26

Years

Years

Risk

-4

-6

-2

28

Expected Return

30

10 11 12 13 14 15 16

Risk

Figure 5
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
24

26

28

30

Expected Return

Risk

Figure 6

1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

Efficient
Frontier

24

26

28

Expected Return

30

Figure 7
EPPO.XLS
Inputs

Outputs

1.

4.

E&P Portfolio
Optimizer EPPO
Portfolio Statistics
Loss
$127

Exp.

Cost
$791
<=

Reserves
484

Req.

$1,100

40%
30%

20%

30%
Portfolio
Project
Statistics Project A

0%

41%

$25,222

$20,840

$16,457

$12,075

10%
0%
$7,692

Req.

($1,073)

Exp.

100%

41%

Project B

Project C

Project D

Project E

Exp. NPV

$820.39

$170.80

$890.51

$1,255.16

$1,562.84

Exp. Cost
Exp. Reserves

$372.69

$47.05

$374.08

$370.07

$383.88

161

32

218

240

262

Trial 1
Trial 2
Trial 3
:
:
Trial 40

60%
50%

Copyright 1998, Ball & Savage Assoc.

NPV
$2,500
>=
$2,500

5.
6.
7.

The Portfolio: percentages of each project

undertaken, optimized to minimize the Mean
Loss (5) for required ENPV (2) and Budget (3).
Mean Loss of portfolio.
Distribution of outputs for this portfolio.
Pie chart of % of budget in the various projects.

70%

3309.67

2.
3.

Monte Carlo trials of the economic outcomes of

five projects designed to preserve the statistical
dependence among projects. The user must
provide this data for the projects under
consideration.
Required ENPV
Budget.

Mean Loss
calculations
$1,547.77

Monte Carlo Generated Data

Optimization Calculations

Project A

Portfolio by
Scenario
Penalty

$
$
$

872.95
1,944.41
1,910.78

Project B

$
$
$

:
:

(393.00) $

Project C

(8.00) $
(47.00) $
(47.00) $
:
:

(47.00) $

Project D

(100.00) $
(100.00) $
(300.00) $
:
:

1,441.28

Project E

(300.00) $ 23,147.60 $9,315

(250.00) $
(350.00) $142
6,264.29 $
(350.00) $6,564
:
:
:
:
:
:
(300.00) $
(350.00) $36

($9,315)
($142)
($6,564)
:
:
($36)

$127
$0
$0
$0
:
:
$0

$0
$0
$0
:
:
$0

Figure 8a

Figure 8b

Figure 8c

Figure A-1

Figure A-2 a, b, and c

Project A
0.5

Probability

0.4
0.3
0.2
0.1

50

40

30

20

10

-10

-20

-30

0
Return

Project B
0.8
0.7

Probability

0.6
0.5
0.4
0.3
0.2
0.1
20

30

40

50

20

30

40

50

10

-10

-20

-30

0
Return

Project C
0.8
0.7

0.5
0.4
0.3
0.2
0.1
10

-10

-20

0
-30

Probability

0.6

Return

Figure A-3

Figure A-4

Figure A-5