Академический Документы
Профессиональный Документы
Культура Документы
This paper was prepared for presentation at the 2009 SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 4–7 October 2009.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.
Abstract
One of the great successes of modern financial risk modelling is the application of computing technology to simulate complex
continuous probability distributions associated with the value metrics of real-life assets. The Monte Carlo simulation technique
is the best known example and is widely applied by economists in the E&P industry. However a Monte Carlo simulation in
which project economics are aggregated into a large asset portfolio is rarely undertaken. The main reason is that present
computing systems cannot handle the vast amounts of data generated in a Monte Carlo simulation of a large asset portfolio.
A pragmatic solution for this issue might be to approximate the continuous distribution of feasible project outcomes using a
small number of probability-weighted discrete scenarios. In a corporate portfolio simulation, a project sample would be drawn
from these discrete distributions as opposed to the original continuous distributions. If either a global assumption needs
revision or a single asset requires recalculating, the economics of a relatively small number of scenarios would be computed.
Thus, there is no need to store or recalculate a large number of outcomes (typically more then 2,000) for each project, as would
be required in a conventional Monte Carlo simulation.
A prerequisite for this approach is that the frequently skewed and complex continuous probability distributions of each of the
assets can precisely be described by a small number of scenarios. This study documents the level of precision that can be
achieved using Swanson’s rule and variations thereof. Our analyses suggest that if the P50 is at least 33% greater than the P10,
the simulated portfolio mean and standard deviation are within 5% and 15% of the actual values, respectively. The
approximation of the lower end of the distribution, i.e., the downside risk of a portfolio, is within 4% even for an extremely
asymmetric lognormal distribution.
The proposed portfolio simulation methodology addresses the largely unmet need of corporate managers to improve their
understanding of key risk and value drivers and their impact on the performance of a corporate asset portfolio.
1. Introduction
In a 2006 survey by Booz, Allen and Hamilton, leaders of 20 Oil & Gas companies indicated a high demand to improve risk
management processes and that the industry is dissatisfied with current procedures (McKenna et al., 2006). McKinsey
interviewed over 1,000 corporate directors, 76% of whom stated that they want to spend more time on strategy and risk
management (Roberts, 2005). Only 11% of the directors indicated that they have a satisfactory understanding of the risks their
companies currently bear, whereas 50% stated that they lack the data to track corporate risk exposure over time.
This lack of insight into the effect of unforeseen events on corporate performance can be catastrophic for those companies
and can be partially attributed to the techniques used to determine the economic value of the company’s assets.
The Boston Consulting Group reported that many energy companies ignore uncertainty and risk and that project decisions
are based on deterministic evaluations (Balagopal and Gilliland, 2005). Clearly, if the economic impact of uncertainty is not
assessed at a project level, this information and the relevant insights cannot be communicated to corporate decision makers.
Although the Oil and Gas industry has started to realize the value of enhanced financial modelling (Bickel and Bratvold,
2007), most efforts to improve decision-making processes have focussed on project management. Risking procedures such as
decision tree modelling or Monte Carlo simulation are routinely applied in project management but are rarely applied at a
corporate portfolio level. As a consequence, corporate decision making tends to be suboptimal.
Even though computing technology has been the cornerstone of portfolio management for several decades, present
computing systems cannot handle the vast amounts of data generated in a Monte Carlo simulation of a large number of assets.
2 SPE 124180
In this study, we discuss a pragmatic approach to developing a risked portfolio of E&P assets. The precision of this
approach is assessed through experiments.
1
A mathematical formulation that yields an exact solution to an investment problem
2
The precision of Swanson rule is not affected by the standard deviation is the underlying uncertainty follows a normal
distribution
SPE 124180 3
3. Methods
100%
Exploration
Cumulative probability
80%
Success
60%
40%
20%
0%
0 500 1000 1500
Value
Exploration Failure
Figure 1. The project valuation methods simulate (1) the exploration risk as a Bernoulli distribution that yields either a successful or a
non-successful outcome and (2) commercial uncertainty as either discrete with three possible outcomes (Swanson’s rule and
Extended Pearson-Tukey method) or a continuous lognormal distribution (Monte Carlo simulation).
In the hybrid Monte Carlo – Swanson’s rule method, the 10 largest assets representing 70% of the total expected portfolio
value are modelled with the full Monte Carlo method, and the 20 remaining assets are modelled using Swanson’s rule.
3.1.1 Sum of moments
In the “sum of moments” method, each of the first three moments (mean, variance, and skewing) of the individual assets are
respectively summed across the assets to obtain the three moments of the overall project portfolio (Howard, 1971; Stuart and
Ord, 1987). On the basis of the first three moments, a lognormal portfolio distribution is fitted using a simple optimization
routine.
4 SPE 124180
Failure 50%
Failure 85%
Figure 2. Conditional probabilities of geological success. The chance of success of the second exploration target is dependent on the
outcome of the first exploration target. The chance of success of any subsequent exploration target is dependent on the outcome of
the previous exploration campaigns.
Upon successful completion of the exploration phase, the asset is exposed to commercial risk, for which the value
distributions are assumed to be lognormal. The effect of the asymmetry of the project value probability distributions was
investigated by defining two asset portfolios. One portfolio consisted of projects with highly skewed value probability
distributions, whereas assets in a second portfolio were modestly skewed. All projects in the skewed portfolio yield P10-P50
and P50-P90 ratios of 6; those in the symmetric portfolio yield ratios of 1.47. 3 Two cases of relative value distribution of
assets across an appraisal portfolio were assessed. In the first case, all assets have a value of 1000; in the second case, the
portfolio consists of a small number of high-value assets and a large number of relatively low-value assets (Figure 3). The
relative value distribution in the latter scenario follows Lavalette’s rule (Lavalette, 1996).
3
For example, P50=100 and P50/P90=10 implies P10=1000 and P90=10.
SPE 124180 5
7000
6000 P10
P50
5000
P90
4000
Value
Expected value
3000
2000
1000
0
1 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 29 30
Asset number
Figure 3. An appraisal portfolio that consists of 30 assets with a wide range of expected values. The six largest assets constitute 53%
of the expected portfolio value. The probability distribution of the individual assets is highly skewed.
4. Results
4
The P50-P10 and P90-P50 ratios are assumed to be equal.
6 SPE 124180
100%
Cumulative probability
80%
60%
40%
20%
0%
0 5000 10000 15000
Value
100%
Cumulative probability
P10-P50 ratio Value metric Monte Carlo Swanson's rule Extended Pearson-Tukey method
6 Mean 1268 1068 1249
Standard deviation 3813 2014 2823
1.47 Mean 524 520 520
Standard deviation 577 561 565
Table 2. Value metrics yielded by two distinct projects evaluated by different valuation techniques.
project value distributions are highly skewed. Although the hybrid Monte Carlo – Swanson’s rule methodology yields the most
precise value distribution in scenario 5, the level of precision is comparable to Extended Pearson-Tukey method.
120%
Cumulative probability
100%
80%
60%
Portfolio analysis 1 40%
20%
0%
25000 27000 29000 31000 33000 35000 37000 39000
Value
100%
Cumulative probability
Value
Monte Carlo Swanson's rule
Extended Pearson‐Tukey Hybrid MC‐Swanson's rule
Sum of moments
Figure 5. Cumulative probability plots that show the results of the portfolio analysis.
8 SPE 124180
Portfolio scenario 5
P10 22238 16271 20169 21341 23826
P50 6003 6064 6282 5886 5483
P90 459 493 425 479 1746
Average 8134 6810 7663 7743 8134
Table 3. A summary of the results of the portfolio analysis. Numbers in red are discussed in the text.
The expected value and the standard deviation yielded by the sum of moments method are by definition equal to the mean
of the Monte Carlo simulation. 5 The P10 estimate is within 5% for a P10-P50 ratio of 5, but the P90 value is overestimated by
350%.
1.05
1
0.95
Normalised mean
Normalised mean
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0 2 4 6 8 10 12
1.2 P10‐P50 ratio
Norm standard deviation 1 Series1 Series2 Series3 Series4
Normalised standard deviation
0.8
0.6
0.4
0.2
0
0 2 4 6 8 10 12
1.1 P10‐P50 ratio
1 Series1 Series2 Series3 Series4
Normalised P10
Normalised P10
0.9
0.8
0.7
0.6
0 2 4 6 8 10 12
P10‐P50 ratio
4.1
3.6
Swanson's rule Adjusted Swanson's rule
Normalised P90
Normalised P90 3.1
Hybrid MC‐Swanson's rule Sum of moments
2.6
2.1
1.6
1.1
0.6
0 2 4 6 8 10 12
P10‐P50 ratio
Swanson's rule Extended Pearson‐Tukey
Hybrid MC‐Swanson's rule Sum of moments
Figure 6. Portfolio value metrics, normalized to the Monte Carlo simulation results, as a function of project value spread. The portfolio
value spread is expressed as the ratio between the P10 and P50 of the individual projects, the latter applying to all projects within the
portfolio.
5
The mean and standard deviation of the Monte Carlo simulation results are used to define the sum of moments probability
distribution.
10 SPE 124180
5. Discussion
6. Conclusions
A large unmet need exists for techniques and tools that can generate insight into the probability-weighted range of portfolio
performance. The processing and storage of large volumes of data that are generated in a portfolio simulation is a challenge in
modern corporate financial management. The approach proposed in this study vastly reduces the amount of data to be
SPE 124180 11
managed by replacing the continuous probability distributions for individual assets with a small number of probability-
weighted discrete scenarios.
The modelled project scenarios include one failure scenario of zero value and three positive scenarios. The latter three
scenarios are probability weighted according Swanson’s rule or a variation thereof. We have demonstrated that the proposed
methodology generates precise results in the vast majority of cases. The portfolio mean can generally be estimated within 10%
and the standard deviation within 25%. If the underlying assets have a P10-P50 ratio of 3, the portfolio simulation yields a
precision of 5% for the mean and 15% for the standard deviation. The approximation of the lower end of the distribution, i.e.,
the downside risk of a portfolio, is within 4% even for an extremely asymmetric lognormal distribution. For situations where
the precision of the approximation is insufficient, we suggested several approaches to improve it. The simulation of the P5,
P50, and P95 in Extended Pearson-Tukey method significantly improves the precision of the approximation. A hybrid
approach, in which the few high-value assets are modelled in a Monte Carlo simulation while the remaining assets are
modelled using Swanson’s rule, greatly enhances the precision of the simulation and significantly reduces the computing
resources required.
7. References
Balagopal, B. and Gilliland, G. 2005. Integrating value and risk in portfolio strategy for energy companies.
http://www.bcg.com/impact_expertise/publications/files/Integrating_Value_Risk_Energy_Companies_Sep2005.pdf
Bickel, J. E. and Bratvold, R. B. 2007. Decision making in the Oil and Gas industry: From blissful ignorance to uncertainty induced
confusion. SPE 109610.
Black, F. and Scholes, M. 1973. The pricing of options and corporate liabilities. Journal of Political Economy 81 637-654.
Goodwind, P. and Wright, G. 2004. Decision analysis for management judgement, Third edition, John Wiley & Sons, Ltd.
Hora, S. C. 2007. Eliciting probabilities from experts. In Advances in decision analysis, edited by Edwards, W., Miles Jr., R. F. and
vonWinterfeldt, D., Cambridge University press, 129-153.
Howard, R. A. 1971. Proximal decision analysis. Management Science 17 (9) Reprinted in Howard, R. A. and Matheson, J. E. (Eds.) (1984)
Readings on the principles and applications of decision analysis, 2, Strategic decisions group, Menlo Park, CA.
Hurst, A., Brown, G. C. and Swanson, R. I. 2000. Swanson’s 30-40-30 rule. AAPG Bulletin 84 (12) 1883-1891.
Keefer, D.L. and Bodily, S.E. 1983 3-point approximation for continuous random variables. Management Science 29 595-609
Lavalette D. 1996. Facteur d’impact: impartialité ou impuissance? Internal Report, INSERM U350, Institut Curie - Recherche, Bât. 112,
Centre Universitaire, 91405 Orsay, France, http://www.curie.u-psud.fr/U350/
McKenna, M. G., Wilczynski, H. and VanderSchee D. 2006. Capital project execution in the Oil and Gas industry.
http://www.boozallen.com/media/file/Capital_Project_Execution.pdf
Megill, R. E. 1984. An introduction to risk analysis, Second edition: Tulsa, Pennwell Publishing.
Roberts, T. 2005. Appetite for destruction: Ensuring sound risk management and governance.
http://www.accaglobal.com/pubs/publicinterest/activities/library/governance/presentations/2705952/241105_TimRoberts.pdf
Rose, P. R. 2001. Risk analysis and management of petroleum exploration ventures. AAPG methods in exploration series 12.
Rose, P. R. and Thompson, R. S. 1992. Economics and risk assessment. In Morton-Thompson, D. and Woods, A. H., eds., Development
geology reference manual. AAPG Methods in exploration series 10 21-56.
Smith J.E. 1993 Moment methods for decision analysis. Management Science 39 340-358
Stuart, A. and Ord, J. K. 1987. Kendall’s advanced theory of statistics, Vol. 1, Distribution theory. Oxford Publications, New York.
(1)
= number of percentage points difference between the minimum / maximum and the median
= probability
= Standard deviation of log(X)
= Probability P(Z> )
Assuming = 0.19 and (0.05) = (0.95) = 1.62, Extended Pearson-Tukey method can be developed in which the
P5, P50, and P95 are weighted at 19%, 62%, and 19%, respectively. The theoretical validity of Extended Pearson-Tukey
method equals that of the original formulation, given that it is based on the same mathematical formulation.