Risk, Uncertainty and Investment Decision-Making in the Upstream Oil and Gas Industry Fiona Macmillan ma hons (University of Aberdeen) October 2000 a thesis presented for the degree of Ph. D. at the University of Aberdeen declaration
Download 1.37 Mb.
|
- Navigate this page:
- Percentile Value
- Type of distribution assigned to reach reservoir parameter for each run
- Rec Res 2 Rec Res 3 Rec Res 4
Risk analysisSimulation as a means of risk analysis in decision-making was first applied to petroleum exploration investments in 1960 (Grayson, 1960). The technique can be applied to any type of calculation involving random variables. It can be used to answer technical questions such as (“What is the volume of recoverable reserves of hydrocarbons in this acreage?”) and economic ones such as (“What is the probability that the NPV of this prospect will exceed the target of $x million?”) (Bailey et al., in press). The main concepts of risk analysis using simulation will now be presented before its applicability to the upstream is examined. Risk analysis based on Monte Carlo simulation is a technique whereby the risk and uncertainty encompassing the main projected variables in a decision problem are described using probability distributions. Then by randomly sampling within the distributions, many, perhaps thousands, of times, it is possible to build up successive scenarios, which allow the analyst to assess the effect of risk and uncertainty on the projected results. The output of a risk analysis is not a single value, but a probability distribution of all expected returns. The prospective investor is then provided with a complete risk-return profile of the project showing all the possible outcomes that could result from the decision to stake money on this investment. It is perhaps easiest to see how Monte Carlo simulation works by using an example of a hypothetical field. The main data is given in table 5.2. The decision facing the decision-makers is whether to develop the field. Performing a simple deterministic calculation, with a discount rate of 10%, gives an NPV of $125 million and the decision to go ahead on development should be straightforward. But a probabilistic assessment of the same field gives the decision-maker a broader picture to consider. Assume the probabilistic assessment uses the figures in table 5.2 as the “most likely” inputs (those falling at the mid-point of the range) but also suggests the ranges of possible values for inputs in table 5.3.
Table 5.2: Hypothetical field data
Table 5.3: Hypothetical field data for Monte Carlo simulation Ten thousand Monte Carlo trials give the results shown in table 5.4. The mean, or average, or expected value is $124 million (that is, a statistically significant number of identical opportunities would, on average, be worth $124 million, in NPV terms.).
Table 5.4: Results from the Monte Carlo simulation But in fact there is a range of possible outcomes and a chance of very different results. For example, the so-called p10 value, or forecast with 10% possibility of occurrence, (see table 5.4) is $27 million, so 10% of the cases run in the simulation gave values less than $27 million. The lowest possible outcome is $-112M and 5% of the cases, or trials, gave negative NPVs. On the other hand, the p90 was $223 million, so 10% of the trials gave values greater than $223 million (Bailey et al., in press). For this particular field, there is a small, but not zero (that is, approximately 4%) chance of losing money. The decision would probably still be to go ahead, but the Monte Carlo analysis, by revealing the wider picture, gives the decision-makers greater comfort that their decision has taken everything into account. Using risk analysis in investment appraisal has a number of advantages. Firstly and most importantly, it allows the analyst to describe risk and uncertainty as a range and distribution of possible values for each unknown factor, rather than a single, discrete average or most likely value. Consequently when Monte Carlo simulation is used to generate a probability distribution of NPV, Newendorp (1996 p375) believes that: “The resulting profit distribution will reflect all the possible values of the variable.” This is a slightly dubious claim since the resulting profit distribution will not contain every possible value of NPV. It will only include those that the decision-maker or analyst feels are likely to occur. There is always the possibility of “acts of God” or “train wrecks” (see section 5.7 and for a full discussion refer to Spencer and Morgan, 1998). However, it is certainly true, that in generating probabilistic output, the decision-maker is more likely to capture the actual value in the predicted range. Secondly, risk analysis allows the analyst to identify those factors that have the most significant effect on the resulting values of profit. The analyst can then use sensitivity analysis to understand the impact of these factors further. There are several ways this sensitivity analysis can be carried out and the reader is referred to Singh and Kinagi (1987) for a full discussion.Implementing risk analysis using Monte Carlo simulations has limitations and presents a number of challenges. Firstly, Monte Carlo simulations do not allow for any managerial flexibility. This can be overcome by running simulations for several scenarios (Galli et al., 1999). Gutleber et al. (1995) present a case study where simulations were carried out to compare three deals involving an oil company and local government. Murtha (1997) provides many references to practical applications of this procedure. Secondly, whilst geologists intuitively expect to find a correlation between, for example, hydrocarbon saturation and porosity, this is not acknowledged explicitly in the literature nor are analysts given any guidance concerning how to model such relationships. Goodwin and Wright (1991 pp153-157) describe types of dependence and approaches to modelling dependence are given in Newendorp (1996 pp406-431), Goodwin and Wright (1991 pp153-157), Eilon and Fawkes (1973) and Hull (1977). There is significant evidence in the prescriptive decision analysis literature that decision-makers have difficulty assigning the strength of association between variables. Nisbett and Ross (1980 p26) have given the following concise summary of this literature: “The evidence shows that people are poor at detecting many sources of covariation … Perception of covariation in the social domain is largely a function of pre-existing theories and only very secondarily a function of true covariation. In the absence of theories, people’s covariation detection capacities are extremely limited. Though the conditioning literature shows that both animals and humans are extremely accurate covariation detectors under some circumstances, these circumstances are very limited and constrained. The existing literature provides no reason to believe that … humans would be able to detect relatively weak covariations among stimuli that are relatively indistinctive, subtle and irrelevant motivationally and, most importantly, among stimuli when the presentation interval is very large.” Chapman and Chapman’s 1969 study provided evidence of a phenomenon that they refer to as illusory correlation. In their experiment, naïve judges were given information on several hypothetical mental patients. This information consisted of a diagnosis and drawing made by the patient of a person. Later the judges were asked to estimate how frequently certain characteristics referred to in the diagnosis, such as suspiciousness, had been accompanied by features of the drawing, such as peculiar eyes. It was found that judges significantly overestimated the frequency with which, for example, suspiciousness and peculiar eyes occurred together. Moreover, this illusory correlation survived even when contradictory evidence was presented to the judges. Tversky and Kahneman (1974) have suggested that such biases are a consequence of the availability heuristic. It is easy to imagine a suspicious person drawing an individual with peculiar eyes, and because of this, the real frequency with which the factors co-occurred was grossly overestimated. So, in the case of the relationship between porosity and water saturation, this research suggests that because geologists expect there to be a correlation, if there is any evidence of a correlation in any particular case, the geologist is likely to overestimate the strength of this relationship. This research indicates the powerful and persistent influence that preconceived notions can have on judgements about relationships (Goodwin and Wright, 1991 p153). The third limitation of Monte Carlo simulations is perhaps most significant. In the industry literature, no published study has indicated which probability distribution most accurately describes the reservoir parameters of reservoir rocks of similar lithology and water depth. Similarly, there has been no research that has identified the appropriate shape of probability distribution to be adopted for economic factors such as oil price. Section 6.2 of Chapter 6 will discuss how companies cope with this lack of prescription in the literature. It is possible here, to perform a crude test to investigate whether the shape of the probability distribution used for each input variable, affects the estimate generated by a Monte Carlo simulation. Such a test is carried out below. For each reservoir parameter, base values are entered and probability distributions are assigned to each of these variables from the seventeen available in Crystal Ball™ . Then a Monte Carlo simulation is run and the estimate of recoverable reserves generated expressed in percentiles, is noted. This process is repeated twelve times altering only the probability distribution assigned to each variable each time. The base value data and the probability distributions used for each trial are shown in table 5.5. The output produced is summarised in table 5.6 and provides evidence that altering the probability distribution assigned to each reservoir parameter, significantly affects the forecast of the recoverable reserves. (Note, that although some of the distributions used here are more unusual (for example, the Weibull), the lack of prescription in the literature over the shape of probability distribution that analysts should adopt for reservoir parameters (and economic variables) means that, if these results are accurate, analysts could, unwittingly or otherwise, use these types of distribution to distort the results.) Further studies are needed to confirm these results. This would then prompt researchers to explore the shape of the probability distributions to be used for the reservoir parameters of reservoir rocks of similar lithology and burial history and the nature of the probability distributions to be used to model the economic variables. Future research should also investigate the nature of correlation between the reservoir (and economic) variables. The author of this thesis tried repeatedly throughout the course of the study to access “real” reserves and economic data to conduct such research but was unable to collect enough data to make any results achieved meaningful. Whilst much of the economic data is regarded by companies as commercially sensitive and this makes such research unlikely in the near future, a book due for publication next year should contain the relevant reserves data (Gluyas et al., 2001). It is hoped that the interest provoked by this thesis will motivate researchers and practitioners to conduct the necessary studies. Despite these limitations, risk analysis using simulation is perceived by the majority of decision analysts to enable a more informed choice to be made between investment options (for example, Newendorp, 1996; Goodwin and Wright, 1991 p151). Certainly, by restricting analysts to single-value estimates the conventional NPV approach yields no information on the level of uncertainty that is associated with different options. Hespos and Straussman (1965) have shown how the simulation approach can be extended to handle investment problems involving sequences of decisions using a method known as stochastic decision tree analysis.
Directory: papers papers -> From Warfighters to Crimefighters: The Origins of Domestic Police Militarization papers -> The Tragedy of Overfishing and Possible Solutions Stephanie Bellotti papers -> Prospects for Basic Income in Developing Countries: a comparative Analysis of Welfare Regimes in the South papers -> Weather regime transitions and the interannual variability of the North Atlantic Oscillation. Part I: a likely connection papers -> Fast Truncated Multiplication for Cryptographic Applications papers -> Reflections on the Industrial Revolution in Britain: William Blake and J. M. W. Turner papers -> This is the first tpb on this product papers -> Basic aspects of hurricanes for technology faculty in the United States papers -> Title Software based Remote Attestation: measuring integrity of user applications and kernels Authors Download 1.37 Mb. Share with your friends:
| |||||||||||||||||||||||