Value of Optionality, Asymmetry & Flexibility A Practitioner’s Guide Last update: December 2, 2019 1 Table of Contents Preface.................................................................................................................................................................4 Introduction ........................................................................................................................................................5 Benchmarking with Competition ....................................................................................................................6 Definitions ...........................................................................................................................................................7 Intrinsic Versus Extrinsic Value ........................................................................................................................8 What Is Optionality ............................................................................................................................................8 What Is Asymmetry ........................................................................................................................................ 10 Market Attributes, Market Structure, and Enablers .................................................................................. 11 Situations Where Optionality Occurs .......................................................................................................... 13 Identification .................................................................................................................................................... 14 Valuation .......................................................................................................................................................... 15 Method 1: Probability Weighted Model ................................................................................................. 16 Method 2: Binomial (Trinomial, Lattice) Model ..................................................................................... 18 Method 3: Black Scholes model and Kirk’s approximation .................................................................. 21 Method 4: Monte Carlo Simulation Model ............................................................................................. 24 Method 5: Stochastic Optimization Model ............................................................................................ 30 Method 6: Machine Learning and Predictive Price Modeling.............................................................. 32 Method 7: Experience Based Volatility Probability Analysis ................................................................ 33 Summary table of pros and cons and potential applications of various methods ........................... 35 Software Tools for Evaluation of Options .................................................................................................. 36 Integration with Opportunity Evaluation .................................................................................................... 37 Integration with Scenario Analysis........................................................................................................... 37 Presentation of Results .............................................................................................................................. 38 Key OE tools integration ........................................................................................................................... 39 Integrating Extrinsic Value into OE .............................................................................................................. 40 Accountability.............................................................................................................................................. 41 Implementation & Operationalization Plan (I&OP) ............................................................................... 41 Implementation & Operationalization ..................................................................................................... 42 Plan & Stewardship .................................................................................................................................... 42 Fundamental shifts ..................................................................................................................................... 43 Lessons Learned and Challenges ................................................................................................................. 43 Appendix 1: Optionality & Flexibility Identification Aid ............................................................................. 44 2 Appendix 2: Enabler & Operationalization Checklist ................................................................................ 45 Appendix 3: Tools & Contacts ...................................................................................................................... 47 Appendix 4: Subject Matter Experts ............................................................................................................ 48 Appendix 5: Examples .................................................................................................................................... 49 3 Preface Imagine that, by some twist of fate, you become the ruler of an oil-rich state. A crash in the oil price has left a hole in its budget. You are forced to consider selling the kingdom’s assets. Among them is a mothballed oilfield in a remote part of the country—so remote that it costs $90 to retrieve each barrel of oil. That is above the prevailing price of $70 a barrel. Even so, you are advised to try to sell a license to operate the field. Who would buy such a license? It is valuable only if a barrel of oil sells for at least $90. Yet there is always value in a right—if it carries no obligation. The greater the chance that prices will rise above $90, the more the license can be sold for. The price will be higher if the license is for a long period. Crucially, the price also depends on how changeable the oil price is. The more volatile, the likelier it is that it will hit a level where it is profitable to restart production. Volatility is normally something to fear. People prefer a stable income to an erratic one, for instance, and they feel the same way about their wealth. In this regard, the jumpiness of stock prices is a source of discomfort. But where you have rights without obligations—options, in other words—things are different. Here, volatility is welcome. Look closely, and the hypothetical oil license has the features of a “call” option, a particular kind of financial contract. A call option is the right to buy an asset—a barrel of oil or a basket of stocks, say—at a specified price (the strike price) on or before a specified maturity date. The owner of a call option profits if the price of the underlying asset goes above the strike price. The owner is not obliged to buy at the strike price; she will do so only if it is in her interests. Anyone who buys the oilfield license is essentially buying a call option on the oil price. If it goes above $90 the buyer makes a profit; if it stays below $90 for as long the license is valid, the option expires worthless. Putting a value on options is a fiddly business. The key ingredients in the BlackScholes model, the industry formula, are time, volatility and the gap between the asset’s strike price and its current price. A small gap is more likely to be closed than a large one, so options with strike prices close to prevailing prices cost more. Call options with a strike price above the prevailing price are said to be “out of the money” and are cheaper. The more violently prices fluctuate, the more chance there is that an out-of-the-money option, like the hypothetical oil license, becomes a winning lottery ticket at some point before it matures. Excerpted from “When you have options, volatility is your friend,” The Economist, May 11, 2019 4 Introduction The purpose of this white paper is to guide the identification and valuation of optionality in opportunities. While many people have an intuitive sense for optionality, as a corporation we are still increasing our awareness of and ability to quantify the value of flexibility and how it can be captured. Historically, this value was not routinely and consistently included in project economics but it can be a very significant source of value for many EM assets. Unfortunately, we often traded it away without realizing the potential embedded value. Management has put an increased focus on the identification, valuation and capture of optionality. Fundamentally, this requires a change in our thought process. We must be able to value flexibility relative to its cost, and accept a range of outcomes rather than just success or failure. This approach focuses on market-driven uncertainty, primarily expressed through prices, and applies across the entire value chain, from the Upstream to the Downstream and Chemicals. It has applications in a variety of scenarios, from exploration acreage bidding to clean products marketing. All investing, divesting, and commercial opportunities are in the scope. However, uncertainty from other sources, such as technical feasibility and geological factors, are excluded from this paper. This guidance integrates with the existing Opportunity Evaluation (OE) framework. The value of optionality is not just a sensitivity but should be considered and if material evaluated for including in base economics. The audience for this approach is anyone involved in economic analysis where flexibility could have significant value that would impact the decision. This applies to analysts, advisors, and managers involved in preparing and evaluating the analysis. This paper was written to guide and inform decision framers at multiple levels in the organization. By the end of this paper you should be able to gain the high level knowledge about optionality, make connections to the optionality examples included in the appendix to help recognize the optionality in the opportunities being evaluated, understand the method and tools to be used to evaluate the optionality, and most importantly, know the right person to seek advice and support from. After evaluating the optionality, it’s also important to come up with the correct procedure to ensure capture of the optionality through the decision making process, which is the ultimate goal. Moreover, this document should help develop the mindset that regards volatility as an opportunity and not just a risk. 5 This paper is not just about the optionality valuation, it incorporates expectations to be met for including the associated (extrinsic) value in the economics. In essence, as shown in the above flow chart, after identification and valuation, we only integrate into the OE assessment if we also have a clear plan on operationalization, stewardship and accountability for capturing the value. Benchmarking with Competition The concepts explained in this paper are not new and have been around for decades. While from timeto-time we have successfully used them in ExxonMobil, there was no broad consistent understanding nor application. When we benchmarked we found competition have embraced what is often referred to as Decision Analysis, and started the journey years ago. Their acceptance of these tools was not an overnight process; the companies said it took years (if not decades) for this way of thinking to really permeate the culture. Examples of what we found while benchmarking: Competition adopted advanced analytics approaches enabling real option / asymmetry valuations IOCs structurally address decision support and quantitative modeling competency ‘Decision Support’ often independent from business, though close relationship (e.g. ‘embedded’) Advanced Analytics / Stochastic modeling widely accepted; standard in Trading decision support Cumulative Distribution Function (aka ‘S-Curve’) used by competition in Management Presentations Structural investment in the organizational capability (people, processes, systems, tools, etc.) Training on key concepts across the organization, including management / leadership The bottom line is that, when benchmarked against competition, EM underinvested in the processes, methods and tools and we are lagging the industry. 6 Definitions Intrinsic Value – This is the fundamental value of an asset or opportunity given its proposed characteristics today evaluated without consideration of any future flexibility (optionality). Extrinsic Value - This is the value above intrinsic value that comes from flexibility (i.e. embedded optionality) in the face of volatility. This value may be embedded in assets and therefore not recognized at first glance. Extrinsic value is the value that arises from the flexibility to take different actions depending on movement (volatility) in external variables (often future price changes). Volatility – This is the tendency of an observable but uncontrolled variable such as price to fluctuate. The price movements may be entirely random, drifting along a trend, mean-reverting toward an equilibrium balance, or may follow some other statistical distribution that can be measured and forecasted. Asymmetry – The probabilities of a random variable, such as price, to go up or go down are different. This also occurs when a distribution of outcomes (e.g. NPV) is skewed to one side or the other rather than evenly (normally) distributed. The asymmetry may be naturally occurring (e.g. as a result of market structure) or due to the exercise of optionality or flexibility to capture upside and avoid the low side. Optionality – The flexibility to act differently because of the availability of alternative actions you can take, given the volatility of variables which affect your decision. Option Cost (Premium) – The cost to acquire flexibility. Real Option – A choice made available with respect to business/investment opportunities. It is referred to as “real” because it typically references projects involving a tangible asset instead of a financial instrument, though financial instruments (derivatives) may enable capturing the value from real options in some cases. Market Attribute – An element of the market’s structure that creates or affects option value. Market Structure – The combination of organizational and other attributes of a market that result in competitive characteristics and an outcome which market participants transact in, or against. Market structure defines the nature of competition and the pricing mode; it influences the behavior of individuals and firms in the market. Enabler – What you need to have to capture the value associated with the real option, a value lever. Expected Value – The value in a probabilistic distribution that represents the mean (average) return of all possible outcomes. P50 Value – The value in a probabilistic distribution that represents the median return of all possible outcomes, this is the point where 50% of the possible outcomes are below this point, and 50% of the outcomes are above. In most cases, this is not the same as the expected value. 7 Intrinsic Versus Extrinsic Value The long-term price sets (Opportunity Evaluation Scenarios) issued by Planning are typically annual average, sometimes quarterly average, projections that are used for projects which often last for decades. However, the market prices exhibit volatility on a much shorter time horizon, changing daily if not by the minute. There may also be seasonal effects that an annual average will not capture. Once the asset is operational, there is often value, additional to that identified by the average price set, which can be extracted if you have flexibility in how the asset is utilized in response to changes in market conditions. Going forward, EM will include the additional optionality (extrinsic) value in the base project and other decision economics where material and we have a plan to capture it. Conversely, we may be considering alternative opportunities that lead to a decrease in optionality, and if so, this should also be included in decision economics. By definition, the intrinsic value can be cashed in by immediate transaction of sales and purchases at currently-quoted prices. For example, you could lock in future sales based on today’s forward price curve and remove all risk, thereby eliminating volatility from the returns. However the intrinsic value of, for example, a gas contract does not consider value that accrues in the future operation (optimization) of the contract. On the other hand, extrinsic value results from the market price changes that will occur in the future. For example, one can model future price evolution based on past volatility and then evaluate the contract over market representative volatile price set rather than just one trendline, incorporating how we exercise the flexibility (option) as prices move. The extrinsic value is considered the value arising from flexibility. What Is Optionality Optionality is the right but not the obligation to take an action. Many people are familiar with financial options which give the holder the right to buy or sell a stock at a fixed price (strike price) in a set period of time. There are several important variables which define the value of a financial option on an equity. Of particular importance are time to expiration and stock price volatility. The longer before one must act on the option (i.e. flexibility) and the higher the volatility, the more valuable the option. For a financial option to have value, the underlying asset must have some volatility. Consider the equivalent variables associated with a reserve development option as described in the preface (this could just as readily been illustrated with an exploration example). Real options are similar but instead of applying to stocks, they allow you to take actions in the real world. Similarly, the longer the period between acquisition of exploration right and the development of the field and the more volatile the resource prices, the higher the optionality value. Options represent flexibility. If a project has few or no options, management is constrained in its ability to add value. Examples of options in a project are the ability to expand or contract the scope of a project 8 (capturing upside or limiting downside); the ability to accelerate a decision (or a whole project) or defer the commitment; and the ability to establish the timing and amount of investment or production (i.e., cash flow). Clearly a portion of the value associated with a project is the ability to respond to changing conditions or new knowledge, i.e., flexibility. The value of these options is always available within the project waiting to be recognized and captured. However identifying and managing the options is how you create and realize value. There are many real options we can identify in our business, for example: An option to phase development allows you to make a small investment to gather more information before making the full commitment (e.g. drilling exploration wells to prove up a resource). An expansion option provides the future ability to grow at a lower cost, for example building a larger pipeline today at a small incremental cost versus adding a second pipeline in the future or use multiple small LNG liquefaction trains (that can be added) design vs mega train(s). An abandonment option allows you to stop work before the project is completed and recover any residual value if conditions deteriorate. A deferral option lets you postpone the project until market conditions change, for example, drilled but uncompleted wells (DUCs). A switching option gives you the ability to change inputs or outputs based on conditions, such as chemical crackers that can run on a variety of feedstock. A diversion option to sell the commodity or products in markets with higher prices than the target market, for example the ability to divert LNG cargoes to more profitable markets. 9 What Is Asymmetry Asymmetry occurs when the distribution of our returns is no longer symmetric, or normally distributed. This can occur because the shape of the distribution changes and is skewed in our favor (or against us), because we can “cut off” the unfavorable part of a normal distribution by actions we can take, or because counterparties have optionality in a deal that could result in negative asymmetric returns for us. Note that some returns are almost by definition asymmetric because the uncertainty inputs are asymmetric. Where an opportunity with symmetrical returns would have an equal probability to deliver a higher or lower return versus its mean, positive-asymmetric opportunities have a mean return (Expected Value) above the median (P50). Asymmetry may be naturally-occurring (inherent) in an opportunity, or we may own or decide to acquire enablers that allow us to minimize negative outcomes or capitalize only when positive outcomes present themselves, as illustrate in the below chart: In order to extract value from asymmetry, we need to be able to identify the Market Attributes and Enables that cause its presence, value it appropriately, and be confident we can consistently exercise the enablers (flexibility / the option) to capture the value. 10 For example, if we control a pipeline that links two markets whose prices fluctuate, we can ship product from the lower-priced market to the more expensive market when the differential is large enough to cover our costs (transportation, working capital, etc.) and not ship when the differential is small. The ability to capture this arbitrage opportunity by only shipping when the differential is large creates an asymmetric return profile for us. The mean differential between origin and destination pricing (the difference between the markets) might be $2/bbl and after logistics costs $1/bbl remains, which represents the intrinsic value we anticipate. However if the differential is volatile and we have flexibility to ship more when the differential is large, this creates an asymmetric arbitrage opportunity for us with extrinsic value. Valuing the pipeline only using the mean differential would not reflect all the potential value. Note that the distribution of differentials (from the time-series) may be symmetrical or asymmetrical, but it is our ability to exercise flexibility that creates asymmetrical returns. When talking about the “shape” of a distribution or volatility it's important to be clear between inputs (such as prices, margins, differentials, etc.) and the distribution of economic outcomes that you get (your return on the investment/value) because your actions (calling on the enablers / flexibility) may change the distribution with intent to skew (further) positively, this is what arbitrage through optionality is about. Market Attributes, Market Structure, and Enablers Market attributes contribute to market structures and are the fundamental drivers of volatility. You will need a precise definition of the market you’re considering and then undertake a thorough examination of it. You should consider the basic forces acting on the market such as customer and supplier power as well as the availability of substitutes and the potential for new competitors. These factors all affect the fundamental balance of supply and demand. You may also wish to examine the cost structure of the market through a market clearing mechanism, especially if marginal players are a substantial part of the market. Longer term assessments should be made of the industry’s overall capital cycle (e.g. invest → oversupply → shrink → repeat) and constraints on it. Time-series data should be examined to look for cyclicality which may correlate to seasons or other temporal factors. And exogenous or idiosyncratic event risk should also be considered – for example, unforeseen government limitations on production. It’s also important to consider what can cause shifts in the market structure and if they are likely; this may change future volatility and thus impact the value of an option. Enablers take a variety of forms and we describe some examples here. EM often runs assets at maximum utilization but having spare capacity, especially on the margin, allows us to take advantage of volatility by increasing output during favorable conditions. Market connectivity through logistics assets allows us to capture arbitrage profits when adjacent markets are temporarily dislocated. It may be worth building an amount of logistics overcapacity for markets where we observe volatility to allow us to capture value. While ownership of these assets is not always required, we can’t rely on third-parties to make capacity available in times of volatility as they will prefer to capture this value themselves. Flexibility provided in our contracts and commercial arrangements is another enabler that allows us to react based on market conditions rather than being permanently locked in. 11 A key enabler is allocation of a resource (e.g. human resource, system, etc.) to monitor the market and exercise the optionality. For example, EM has employed Optimizers and Refinery Coordination groups to exercise optionality in refinery Production Planning, by signaling changes to the production plans/feed slates (leveraging PIMS models that contain known inherent optionality in the manufacturing assets). As part of your assessment of the enablers, you should consider whether we believe the source of the value is temporary or structural and understand other relevant facts about the opportunity. For example, how long the window will be open compared to the time it will take for us to capture it? Are we able to move fast enough? In some cases we may need to develop our general organizational capability to exercise flexibility. This could take the form of streamlined endorsement processes increasing our reaction time, for example. You’ll need to identify signposts that alert us when the opportunity is open if it only occurs transiently. This may take some analysis to separate the signal from the noise, especially in very volatile markets. Think of it as follows, when we plan for and value a capital project, we set up the project team to execute the project and develop a full plan for the operating organization that will run it. When we include value for optionality, we also need to have worked through how we capture it consistently: what organization, process, systems, services, activities, resources, competencies and skills, etc. Then ultimately it must be clear how we steward the value capture and who we hold accountable. The combination of volatile market attributes (whether they are fundamentally asymmetric or more normally distributed) in conjunction with unique enablers is what gives rise to asymmetry that we want to capture. However there are other questions that have to be answered in considering one of these opportunities. What is our unique source of competitive advantage that allows us to capture this asymmetry, sustainably? Can anyone copy our strategy with contracts or derivatives, or are our enablers unique assets that only we have access to? Are other parties able to observe what we’re doing? Will they react to our move in some way, either by trying replicate what we’re doing or working to prevent it? How will these reactions alter our ability to capture the asymmetry (think game theory)? These are some of the questions to consider when evaluating opportunities with optionality. 12 Situations Where Optionality Occurs There are a number of situations where optionality is present at EM. This is not an exhaustive list but contains some examples and can aid the identification of Optionality & Flexibility: Area Category Enablers Capital investments Phased development (vs. all at once) Ability to cancel when new/updated information presents itself Design not limiting future expansion capability Initial outlay to retain capability for lower cost future expansion Pre-invest for future expansion Low cost to "do now" (pre-invest) during project execution for future capability/optionality Residual interest Retain exposure to future margins to sustain option Portfolio effect Reduced matrix / optimization flexibility JVs Protection from downside if partner interest/business direction diverge; EM sees more/less value in JV vs partner Crude or Product pricing Backward looking pricing option or capture of forward market structure (Backwardation/contango) Volume flexibility Produce non-ratable - pricing vs volume profile Divestments Commercial structures Ability to adjust volume based on our value in use Volume deferral based on view of future vs current price Production sharing contracts/agreements Tax Trading Origination Time arbitrage Differential arbitrage Geo arbitrage Quality arbitrage Execution / optimization Dynamic Forward Trading Destination flexibility Ability to access and secure rights to limited logistics Limit or increase exposure to crude price and production volume Cost basis management Tax defensibly - Leveraging low tax affiliates Creation of incremental long and/or short positions that create value by allowing to capture pricing volatility , producing asymmetric returns Rights to storage (logistics capability) or drawdown to respectively capture contango/backwardated market structure, usually of flat price, sometimes seasonal crack spread Contango on crude quality differential time arb Ability to move (or not) molecules (feed/product) to different market under different pricing Exploit inherent giveaway in Hydrocarbon batches (e.g. NY harbor blend hub, Crude Blending) Volatility in forward derivative market Vessel/pipeline destination options (or not) Ability to move molecules (or not) to different market - Crude oil destination restriction (D-free vs D-restricted) Sourcing / supply flexibility Matrix of supply options to purchase from and deliver to Quality Ability to switch between sales channel (supply/wholesale/retail) and grade mix Access to and flexibility to substitute components and capability to blend products to specific market quality Ability to nominate crude between fields with differing quality under same price (e.g. Al Shaheen) Capability and Flexibility to process wide spectrum of feed - not limited by site processing or product quality constraints Access and capability to capture crude quality differential not priced into premium Second Order Pricing Effects / Logistic Power balance Oversupply in market may result in lower prices in the market (converse is true), Logistics outlets may negate impact Pricing negotiation leverage for alternate disposition capability 13 In addition for ease of reference this table has been included as Appendix 1: Optionality & Flexibility Identification Aid. The appendix includes a link to a detailed spreadsheet. Identification The first step in the process is to identify the explicit or implicit flexibility in the (commercial and/or investment) opportunities, and understand how to take advantage of them to capture asymmetric returns. Implicit flexibility comes naturally with the opportunity – we don’t have do anything extra to have the option. Explicit flexibility requires us to take additional actions in the planning phase to have the option – for example there could be additional capital required or processes changes. Ideally, there is a structured process set up to review opportunities for flexibility but you should also be open to identifying this during the normal course of business if an idea is generated organically. When considering an opportunity there are a number of questions to ask: What is the intrinsic value of the asset/opportunity using trend-line assumptions? What decision would you make under static (no flexibility) assumptions? What are the market attributes driving the volatility? o Are we taking a position on future volatility? Is there naturally-occurring asymmetry or do we need to create/enhance it to capture the value from it? What flexibility is needed to capture the volatility or create the asymmetric returns? o Think of enablers you would need: assets, business processes, contract terms, etc. What is the cost (think option premium) for the enablers o This is the initial payment plus any costs to maintain the flexibility Is the flexibility you need binary (off/on, one-time) or continuous? Is value derived from integration with other EM assets (portfolio effect)? o Do we need hard assets to capture it or can we do it synthetically (e.g. with paper trades or derivatives)? Are the enablers unique to EM or things anybody can purchase? o Especially important to consider on synthetic positions (that is to say, do we have a competitive advantage over other traders or competitors? Could a hedge fund copy our strategy?) 14 Valuation Correct valuation must be detailed enough to capture the relevant drivers and be validated by the subject matter experts but explainable to the general audience and decision-makers who may not be familiar with all of the math. It is acknowledged that the link between the markets and business drivers for flexibility (identification) and the actual valuation, require different skill sets and both often are technical within their own disciplines. Effectively managing the translation between the business and quantitative valuation disciplines is therefore an important task and often requires a third skill set. The valuation must be applied consistently across similar projects and included in the base economics for optimal capital allocation purposes. That said, we recognize that any approach to modeling the complicated market situation or reality with a few parameters and simplified physical rules will never be perfectly right. However if we choose the methods and parameters carefully and make deliberate assumptions to test our thinking process, we may gain useful insight from the modeling: “All models are wrong; some are useful.” – George Box Identifying the main drivers of value and possible outcomes is most important. As with any analysis, clearly laying out the assumptions on which it is based is critical to ensuring decision quality. For example, determining whether to calculate the volatility of an input over the last year or the last ten years may make a big difference in your results – you’ll need to rely on business experts to frame the case properly before doing any math (otherwise, garbage in = garbage out will still apply). And of course, you should be watchful for cognitive bias as you structure your analysis. Given the number of situations where optionality is present in opportunities ranging from capital projects to commercial structure to trading activities, it’s important to understand the methods and tools to be used to evaluate the optionality embedded in the potential opportunities. Due to nature of different opportunities, the optionality evaluation methods and tools could be different as well. In the following section, we will discuss the commonly used methods, with the guidance on the potential application and constraints. In comparison with the traditional methods of project evaluation, optionality evaluation is considerably more complex and may require a higher degree of mathematical and statistical understanding. We do not expect, and it is not required for, the general audience to understand every detail of the method. However, by providing various illustrative examples on application of different methods, we expect the audience to understand high level principles and make connection to potential opportunities being pursed. In the following section we explain various valuation methods. We use ‘price’ as the example uncertainty, but this is simply to facilitate the example. 15 Method 1: Probability Weighted Model Before we start explaining the method, let’s describe a simple business situation to use as an example throughout. Assuming we are evaluating an opportunity which will require investment of $240M (Capex). Based on the current assumption on technical and market conditions, the total future net operating cash flow (OCF) discounted at 8% to today’s value (NPVt0) will be $200M. We are aware that market condition could change between now and next year – with probability p, say 56%, the NPVt1 of the OCF could rise to $270M; with probability 1-p, the NPVt1 of the OCF could fall to $150M, as illustrated in the following diagram: If management has to make a decision today, we value the opportunity at $-240M (NPV Capex) + $200M (NPV OCF) = -$40M; the likely decision is not to develop the project given the negative Project NPV. If management has the option (the right, but not the obligation) to defer their decision to invest by one year, by when we will have better understanding of how the market condition change, the decision could be different. If the market condition turns out to be unfavorable (44% chance), instead of incurring a loss of $150M (NPVt1 of the OCF) -$240M (NPV Capex) =-$90M, we have the right to not develop the project, hence the value of the opportunity is 0. If the market turns out to be favorable (56% chance), we will exercise the right to develop the project, the NPV t1 (-$240M + $270M [NPVt1 of the OCF]) = $30M or $28M ($30M discounted by 8%, i.e. today’s money). Therefore, the total value of the project, considering the probability of different market conditions, would be weighted average of both situations, i.e., 44%*0 + 56% * 28 = $16M. Compared to the situation of developing the project right way, the optionality value of waiting for one year = $16M – (-$40M) = $56M. In this example, we see that the investment opportunity has negative NPV value as of today without an option; the optionality to delay one year results in a positive NPV, a logical decision is to delay the project to capture the optionality value. Moreover, even if the project has a positive NPV if executed today, it does not necessarily mean that it should be developed now. Delaying the development of a project can improve its value, which should be considered in evaluation of opportunities. Identifying and valuing the flexibility therefore is important where a traditional valuation approach may lead to forego the option value. 16 When to use simple option straightforward probabilities/outcomes screening analysis Benefits simple, mechanics transparent What you need empirical estimate of the potential upside and downside of the value and the probability associated with each value (outcome) Drawbacks / Pitfalls probability assignment potentially subjective The above chart is a simple example to illustrate the concept of optionality – by having the option to wait for one year to get better understanding of the market and its volatility and we can reduce risk stemming from the market conditions. This method is the basis for the following methods and can be extended to multiple branches and multiple time periods. For the probability weighted method, we need knowledge, based on past experience, of the probability of market going up or down; and the associated NPV value when market goes up or goes down. We only considered the one-time option of making a decision to develop the project or not. In other situations, instead of one time period, we may have multiple times to decide whether we develop the project or not. We may not have information about the actual probability and actual value of the project, but have some information about the potential volatility, characterized by the standard deviation, of the cash flow. For such situation, we will evaluate the option through the Binomial Tree Model as explained below. 17 Method 2: Binomial (Trinomial, Lattice) Model Binomial option valuation models looks like a decision tree in which the possible values of the basic property change depending on timing of the option's maturity, i.e., when the option to delay or develop the project is gone. This model tracks the movement of asset prices or project value as a binomial process in which the value can move in two possible directions, i.e. up or down. The changes in the value are marked with 𝑢 and 𝑑 factors, where 𝑢 > 1 and 𝑑 <1, and follows the multiplicative binomial process in discrete time. The Binomial model shows that as the uncertainty clears in the future, management can make appropriate decisions at that time by comparing the expected payoff with the investment cost. For the same business example above, the uncertainty/volatility of the cash flow is characterized by the following parameters: INVESTMENT OPPORTUNITY PARAMETERS EXAMPLE Present value of project's Free Cash Flow S0 $200M Capital expenditure required to acquire project assets X $240M Length of time that decision may be deferred T 5 years Time value of money (risk free rate) r 8%* Riskiness of project assets 𝜎 30% * 8% here is assumed to be the risk free rate, not the standard discount rate that coincidently is 8% in 2019. Initial point 𝑆0 in the Binomial model shows the current value of the underlying asset. The probability of the asset’s value going up in each period in the future is indicated as p. Conversely, the probability of the asset’s value going down is expressed with 1-p. In the first step (node) of the binomial model the asset’s value can move in two directions, up to 𝑆0 𝑢 [u = up] or down to 𝑆0 𝑑 [𝑑 = down]. The next (second) step results in three possible asset values (𝑆0 𝑢2, 𝑆0 𝑢𝑑, 𝑆0 𝑑2), the third time step in four (𝑆0 𝑢3, 𝑆0 𝑢2 𝑑, 𝑆0 𝑢𝑑2 , 𝑆0 𝑑3), etc. The last step in the Binomial model indicates the range of possible asset values at the end of the option’s life, or the latest time that the decision may be deferred. Up and down factors, 𝑢 and 𝑑, depend on the volatility, , of the underlying asset and can be expressed as follows: 𝑢 = 𝑒 𝜎√∆𝑡 = 𝑒 0.30×√1 = 1.35 1 1 𝑑= = = 0.74 𝑢 1.35 In every time period there is a probability p that asset value will grow with percentage 𝑢, and conversely the probability (1-p) that the assets will fall with percentage 𝑑: 𝑝= 𝑒 𝑟×∆𝑡 −𝑑 𝑢−𝑑 = 𝑒 0.08×1 −0.74 1.35−0.74 = 0.56 p is called the risk-neutral probability based on the no-arbitrage principle. This principle states that, regardless of whether the asset value goes up or goes down, a portfolio of owning the asset and taking a risk-free loan will have the same payoff as owning the option itself. 18 With these data we can create a Binomial tree and calculate the asset values on each node of the Binomial tree, using one-year time interval. At each node or time period, management has an option to invest in the project at that point or delay its development and wait until the next time period. Note that all the numbers in the tree are in million: the upper numbers on the Binominal tree present expected future asset values at each node and bottom numbers indicate option values. T=0 T=1 T=2 T=3 T=4 T=5 896 𝑆0 𝑢5 664 656 𝑆0 𝑢4 492 𝑆0 𝑢 364 𝑆0 270 289 𝑢2 270 114 148 200 81 143 270 𝑆0 𝑢3 𝑑 2 200 30 𝑆0 𝑢 𝑑 45 148 𝑆0 110 𝑆0 𝑑 252 𝑢3 d 2 2 𝑆0 𝑑 25 364 𝑢2 𝑑 𝑆0 𝑢𝑑 𝑆0 70 𝑆0 492 𝑆0 𝑢4 d 𝑆0 183 𝑆0 𝑢 200 444 3 16 8 2 𝑆0 𝑢2 𝑑 3 110 𝑆0 𝑢𝑑 4 148 𝑢𝑑 2 81 0 3 0 𝑆0 𝑑 3 0 81 𝑆0 𝑢𝑑 4 60 𝑆0 𝑑 4 0 0 45 𝑆0 𝑑 5 0 When preparing a Binomial tree it is necessary to present the value of expected cash flows arising from investing in the project, 𝑆0 , multiply with the up factor u and down factor d to obtain 𝑆0 𝑢 and 𝑆0 𝑑. Moving to the right, with the same procedure it is necessary to calculate the expected value of cash flows for every node of the Binomial tree until the last step. For example: At the end of the first year: there are two branches, 𝑆0 𝑢 = $200𝑀 × 1.35 = $270𝑀, 𝑆0 𝑑 = $200𝑀 × 0.74 = $148𝑀. At the end of the second year: there are 3 branches, with expected value from the investment ranging from $110𝑀 to $364𝑀 At the end of the fifth year: there are 6 branches, with expected value from the investment ranging from $45𝑀 to $896𝑀 19 Once we have calculated the value of expected future cash flows at each node of the Binomial tree which are shown in the form of the above values, we can calculate the value or price of the options (below italic values in the scheme). The option values are calculated from the extreme right values in the schematic according to the initial values to the left ("backward induction"). On each node there is the possibility of investing or deferral of investment to further. At node 𝑆0 𝑢5 expected asset value is $896𝑀. If the option is exercised in the fifth year, and investment cost of developing the project is $240𝑀, then net asset value of the investment is: $896𝑀 − $240𝑀 = $656𝑀. But if we delay realization of option and wait until next time period, the revenues will be zero because option expires (becomes worthless) at the end of the fifth year. Hence, at node 𝑆0 𝑢5 the option value is $656𝑀 and the rational decision will be not to wait but invest in the development of the project. Expected asset value at node 𝑆0 𝑢2 𝑑3 is $148𝑀 but the option value at this node is zero because the investment of $240𝑀 is resulting in a net loss of $92𝑀. In these circumstances, the rational decision is not to develop the project. Stepping one year back at the intermediate node 𝑆0 𝑢4 we can calculate the expected asset value for keeping the option open as discounted weighted average of potential future option value. If the option is exercised and we invest $240𝑀 in developing the project, the net asset value would be $424𝑀($664𝑀 − $240𝑀). However, holding the options open until the next period (fifth year) gives possibility of realizing higher asset value ($444𝑀 = (0.56 × $656𝑀 + 0.44 × $252)/1.08).Therefore, it is better to continue to wait, rather than to exercise the option to invest at the 4th period. Repeat the same procedure backwards until the time 0, where the option of delaying the investment up to 5 years has a value of $70𝑀. Therefore, the rational decision is to defer the investment decision until year 5. When to use Have data on volatility of project value (return) Evaluating American options (exercise any time) When Geometric Brownian Motion applies Distribution is skewed (drift up versus down) Benefits Transparent Can be visualized to be part of a discussion What you need Measurable volatility of project value (return) Drawbacks / Pitfalls Potentially laborious (e.g. long time to expiry) Based on the same principles the Black Scholes model was developed in 1970s to calculate the option value using a closed formula, assuming the decision time frame is continuous instead versus discrete in the Binomial Tree. 20 Method 3: Black Scholes model and Kirk’s approximation The Binomial Tree method divides the time to expiry of an option into multiple time periods and calculates the asset value at each discrete period to derive the option value, this is called discretization in mathematics. Instead of discretizing the time periods, the Black Scholes model is based on continuoustime value using the “finite difference method” and the assumption of pricing the underlying asset following the Geometric Brownian Motion (described below). That leads to a closed-formula analytical solution to calculate the value of option as following: 𝑋 𝐶 = 𝑆0 × 𝑁(𝑑1 ) − 𝑟𝑇 × 𝑁(𝑑2 ) 𝑒 𝑠0 1 ln ( ) + (𝑟 + 𝜎 2 ) × 𝑇 𝑋 2 𝑑1 = 𝜎√𝑇 𝑑2 = 𝑑1 − 𝜎 × √𝑇 In the above equations: 𝐶: value of call option; 𝑑1 , 𝑑2 : deviation from the expected value of the normal distribution, 𝑁(𝑑1 ) and 𝑁(𝑑2 ): the probability that a standardized, normally distributed random variable will be less than or equal to 𝑑1 and 𝑑2 ; 𝜎: volatility of the asset value; ln: natural logarithm e = 2.71828, base of natural logarithm. In theory, if the discretization of the binomial model is very fine toward infinity, or the time period between the binomial tree nodes is infinitesimal, the option value derived from the Binomial Tree model will be similar to that derived from Black Scholes model which is based on a continuous stochastic differential equation. The Black Scholes model was initially developed for assessing the value of call options on a stock, whereas the price of the stock follows lognormal distribution and the change of the price follows random walk process (Geometric Brownian Motion). The theory of financial options, within which Black Scholes model is developed for valuing financial options, can be easily adapted to be applied to real assets, since the investment opportunity, to a certain degree, is similar to a call option because it allows the right, but not the obligation to invest. The analogy of the parameters between real options and financial options are illustrated in the following table: INVESTMENT OPPORTUNITY Present value of project's Free Cash Flow PARAMETERS S0 CALL OPTIONS Stock price EXAMPLE $200M Capital expenditure required to acquire project assets Length of time that decision may be deferred Time value of money (discount rate) Riskiness of project assets X T r Exercise price Time to expiration Risk-free interest rate Standard deviation $240M 5 years 8% 30% 𝜎 21 After identifying the S0, X, T, r and variables, the option value can be calculated using the above formula as following: 𝑑1 = 𝑠 1 200 1 ln( 0)+(𝑟+ 𝜎 2 )×𝑇 ln( )+(0.08+ 0.302 )×5 𝑋 2 𝜎 √𝑇 = 240 2 0.30√5 = 0.66 𝑑2 = 𝑑1 − 𝜎 × √𝑇 = 0.66 − 0.30 × √5 = −0.01 𝐶 = 𝑆0 × 𝑁(𝑑1 ) − 𝑋 240 × 𝑁(𝑑2 ) = 200 × 𝑁(0.66) − 0.08×5 × 𝑁(−0.01) = $69𝑀 𝑟𝑇 𝑒 𝑒 As you can see, the option value from the Black Scholes model is close to the value calculated from the binomial model; and if the discretized time period of binomial tree is infinitesimal, the value from binomial model will converge to the value from Black Scholes model. A modification or extension of the Black Scholes model is Kirk’s approximation. It is used when evaluating a ‘spread option’ where the payoff is based on the difference in price between two underlying assets, with the volatility of the underlying prices as well as the correlation between these two prices are known. Kirk’s approximation is expressed through the following formula: 𝐶𝐾 = 𝑒 −𝑟𝑡 × [𝑋1𝑁(𝑑𝐾,1 ) − (𝑋2 + 𝐾) × 𝑁(𝑑𝐾,2 )] Where 𝑁() denotes the standard normal cumulative distribution function, and 𝑑𝐾,1 and 𝑑𝐾,2 are given by: 𝑋1 1 ln ( ) + 𝜎𝐾2 𝑇 𝑋2 + 𝐾 2 𝑑𝐾,1 = 𝜎𝐾 √𝑇 𝑑𝐾,2 = 𝑑𝐾,1 − 𝜎𝐾 √𝑇 𝜎𝐾 = √𝜎12 − 2 𝑋2 𝑋2 2 2 𝜌𝜎1 𝜎2 + ( ) 𝜎 𝑋2 + 𝐾 𝑋2 + 𝐾 2 𝜎1 , 𝜎2 : volatility of the underlying prices; 𝜌 : correlation coefficient between the return of the underlying prices; 𝐾 : the strike price of the spread option, i.e. the call spread option would be in the money if the spread is higher than 𝐾. This has been used to evaluate the option value of transportation or storage given the optionality is driven by the spread between the prices in two markets, or between the prices at two different times. Although Black Scholes and Kirk’s approximation are built on a rigorous mathematical and statistical foundation, and it’s easy to calculate the option value through the closed-form formula, it’s difficult to explain the formula in layman’s terms. It cannot incorporate compound options, operational constraints or multi-stage investment opportunities. The volatility (𝜎1 , 𝜎2 ) and the correlation coefficient (𝜌 ) between different stocks can be found in the financial markets or in the historical data. However, when evaluating real options for investment opportunities, determining the value of these parameters is more difficult. Consequently modeling the risk factors that underlie an investment opportunity, such as price, and then calculating the asset value considering the volatility of the risk factors (through simulation) often enables better understanding of the uncertainty and real option value. 22 When to use When you have data project value volatility Evaluating an European option American through layering options When Geometric Brownian Motion applies Distribution is skewed (drift up versus down) When evaluating ‘liquid market’ spreads (kirk) Benefits Simple application Little compute power required What you need Data on volatility of project value (return) and correlation Drawbacks / Pitfalls Not transparent Applies simplification of markets (e.g. efficient) Black Scholes and Kirk’s approximiation (i.e. an extension of Black Scholes) are forms of ‘analytical’ or ‘closed-form-models’, as opposed to the other (probabilistic) methods described here. There are more closed-form-models not described there than can be used if a situation warrants it. 23 Method 4: Monte Carlo Simulation Model As mentioned earlier, both the Binomial and Black Scholes models require the estimation of the volatility (standard deviation 𝜎) of the project value (return), which is difficult to obtain, compared to financial products such as stock prices. But knowing one of the main “risk factors” behind the project value being prices, we can model the price through Monte Carlo simulation and then calculate the project values using multiple realizations of price. This will generate a distribution of the project values (returns), which, on one hand, provides additional insight about risk and uncertainty of the project value due to volatility of the price; and on the other hands, allows to build in operational constraints / flexibility in the economic model to understand the value of optionality and impact of complicated commercial structures on the project value. On top of that, simulation can be readily applied to address situations that need to address multiple factors, in addition to price, such as geological setting, that may have impact on the project or option value. Therefore, simulation is a useful alternative to binomial or traditional finite difference methods (basis for Black Scholes model) and has many advantages as a framework for valuing and optimally exercising real options, due to its nature of being simple, transparent, and flexible. For a Monte Carlo Simulation, the first step is to model the risk factor, or price movement (if that is the main risk), through a stochastic process which is set up to represent the dynamics of the underlying price as close as possible. We will explain, in high level, the main stochastic processes used to model the price below (there are more variants, or expansions, of the following processes to represent different characteristic of the price movement which we will not cover here): Geometric Brownian Motion (GBM): One simple process to model price volatility is GBM which is described by the following continuous Stochastic Differential Equation (SDE): 𝑑𝑆 = 𝜇 × 𝑑𝑡 + 𝜎 × 𝑑𝑧 𝑆 Where: 𝑆: price of commodity; 𝑑𝑆 : differential return or change of the price; 𝑆 𝜇: the drift term (upward or downward trend); 𝜎: the volatility of price S(t); 𝑑𝑧: Wiener process, which denote a random walk process in which you just randomly sample a realization out of standard normal distribution at every time step. In other words, the GBM represents the return or change of price with two elements: the first term is the trend and the second term is the random or stochastic process. In theory, the distribution of stochastic term can be of any type (normal distribution, uniform distribution, Poisson, etc.). Due to the fact that normal distributions can be characterized by limited number of parameters (mean and standard deviation) with a closed-form formula, normal distributions are frequently used for the stochastic term. 𝑑𝑆 Given above equation, the return or change of the price ( 𝑆 ) follows normal distribution and the price itself follows lognormal distribution – this is the base assumption of the price model for the Black Scholes model to calculate the option value. To simulate the price movement, the above continuous differential equation is discretized as following, (after applying ito’s lemma for the natural logarithmetic transformation on the price 𝑆): 24 ln(𝑆𝑡+∆𝑡 )−ln(𝑆𝑡 ) =[𝜇 − 𝜎 2 /2]∆𝑡 + 𝜎√∆𝑡𝜀 Where: 𝑆𝑡 : price at time 𝑡; ∆𝑡: time step of price path; 𝑆𝑡+∆𝑡 : the price at time 𝑡 + ∆𝑡; 𝜇 : rate of “drift”; 𝜎: standard deviation or volatility of the random process (measurement of noise) 𝜀: random number with N(0,1) distribution Parameters to be calibrated from historical data are the drift term 𝜇 and standard deviation 𝜎 , which will be discussed more later. Mean Reversion Model: Another stochastic process to describe the dynamics of price movement is called Mean Reversion Model, by which the assumption is that, for example, the commodity price has a tendency to return to some “normal” or equilibrium level (i.e. the market clearing price), although the prices will occasionally spike to very high or low prices (i.e. market reality of disruptions, etc.). The stochastic process must be designed in a way that this return-to-normal-level characteristic is represented. This characteristic has more economic logic for commodity prices than the GBM described above. Mathematically the stochastic process used to characterize this dynamics is called Mean Reversion Model represented by the following stochastic differential equation (SDE): 𝑑𝑆 = 𝜂 × (𝑀 − 𝑆)𝑑𝑡 + 𝜎 × 𝑑𝑧 𝑆 Where: 𝜂: the speed (strength) of reversion to the defined mean, or mean reversion rate, 𝑀: reverted mean, or long-run equilibrium level that the prices tend to the revert to. The other terms in above equation have the same meaning as in the GBM process. The difference between a mean-reverting process and the GBM is the drift term: in GBM, the drift term is fixed at 𝜇; in mean-reverting process, the drift is positive if the current price level 𝑆 is lower than the equilibrium level 𝑀, and negative if 𝑆is greater than the equilibrium level 𝑀 . In others words, the equilibrium level 𝑀 attracts the prices to its direction. The analogy is with a spring: the more distant prices are from the equilibrium level, the higher is the tendency to revert back the level 𝑀. To simulate the price movement, the above continuous stochastic differential equation is discretized as follows (after applying ito’s lemma for the natural logarithmetic transformation on the price 𝑆): ln(𝑆𝑡+∆𝑡 )−ln(𝑆𝑡 ) =[ 𝜂𝑀𝑅 (𝑀 − 𝑙𝑛𝑆𝑡 )−𝜎 2 /2]∆𝑡 + 𝜎√∆𝑡𝜀 Where: 𝜂𝑀𝑅 :rate of mean reversion; 𝑀: mean value that the log of the price reverts to; The other terms have the same meaning as in the GBM process. 25 Parameters to be calibrated from historical data are the mean reversion rate 𝜂𝑀𝑅 , the reverted mean 𝑀, and the volatility 𝜎. Similar to Kirk’s approximation which models two underlying correlated prices to derive the option value from spread volatility, in both GBM and Mean Reversion model, we can model multiple prices or assets such that they are correlated with each other. To do that, another parameter 𝜌, the correlation coefficient is needed to enter the simulation to generate correlated price movement. In addition to the Mean Reversion Model described above, there are many variants to the process to account for additional characteristics of the price movement. One variant is the Mean Reversion with Jump Model which incorporates volatilities that are not explained through normal volatility but go beyond the normal diffusion process, e.g., high natural gas prices in winter months (seasonality) correlated to weather and primary commodity use. Another modeling variant is Two Factor Mean Reversion Model to address near term and long term uncertainties separately. In order to model the prices through the stochastic process described above, some key questions to be addressed are as following: How to estimate the parameters? There are different parameters in the stochastic differential equations describing for example natural gas prices on different hubs, oil and other commodity prices. When the past is a good predictor of the future, those parameters can be calibrated from historical data sets. In the course of such calibration, we still need to make fundamentals based judgments calls, for example: what historic time window that best describes our view of the future; where there fundamental market shifts or unique events that make a data set less accurate, etc. As a rule of thumb, the time period needs to be long enough to provide a statistically sufficient sample size, but must not reach too far back into the past as the business environment or market fundaments may have changed. Selecting and calibrating the data is an effort that requires both the quantitative and markets experts. In some situations, the parameters can be set to reflect changes over time consistent with fundamental analysis. For example, if over time fundamental marginal price layers change due to changes in industry production capacity or demand, the reverted mean can changed over time to reflect the market fundamental change. Similarly, if there is indication of different volatility in different period, we can incorporate those in the model as well. How to choose the right stochastic process? One question that’s often asked is which stochastic process, Geometric Brownian Motion or Mean Reversion, should be used to model for example the commodity prices? The answer is “it depends” – it depends on what you have observed historically and what you want to model the future. 26 Take the example of historical crude price movement from year 1996 to recent year of 2015: From the period of 1998 to 2006, it seems that the price was following an upward drift with some randomness, which can be characterized by Geometric Brownian Motion stochastic process. However, from the period of 2007 to 2015, the prices went high and low, but there was a tendency for the price to move towards a balancing range, which is characterized well by Mean Reversion stochastic process. Then the question is what’s going to happen in the future? We have fundamentals based views resulting from our internal processes (Energy Outlook (EO), Industry Gas Outlook (IGO), Liquids Supply Outlook (LSO), Spreads & Differentials, etc.) and market insights. These fundamentals need to be reflected in the choice of method as well as the parameters to model the process. Parameter implications from modelling (market) scenarios? Scenario planning aims to define your critical uncertainties and develop plausible scenarios in order to discuss the impacts and the responses to each one of them for an opportunity, project or strategy under consideration (reference S-BORE). By its very nature it tries to identify what drives a certain way the world, region or local situation can develop (for example big shifts in society, economics, technology and politics). These fundamental shifts will have impact on the uncertainties, and as a consequence on the input parameters (assumptions). For this very reason, each scenario may have has its own assumptions and therefore its own set of parameters such as volatility and revered mean, causing different valuations for real options in opportunities, projects and strategies. The stochastic process is fully complementary with and can be integrated with scenario analysis: Identify the assumption that go with each scenario, reflecting the uncertainty fundamentals (which may be represented by different period of history); Calibrate the parameters using the specific period of historical data, and make appropriate adjustment based on proprietary knowledge or market insight; Generate the uncertainty models for each distinct scenarios using parameters consistent with the distinct scenario. The following table and graph illustrates the example of five different business environments, the period of historical data chosen to calibrate the parameters, and the adjustment made based on expert knowledge or proprietary insight. 27 In addition, we can use different parameters for different modeling period (combination of scenarios) to reflect the change of business environment. The process of simulation doesn’t take away the need for experts and decision makers to apply expert insight and judgment. However, by mimicking the history (modeling the future) based on intelligent assumptions, we can support the decision making process by framing the risk and uncertainty embedded in the potential opportunities, hence facilitate informed decision making. After we model the price paths using any of the above processes, we can (utilizing for example excel macros) run the price paths through the economic model to generate the distribution of the project value. Compared to the conventional economic analysis using single trend line prices, the Monte Carlo simulation approach provides additional insight in the following: The distribution of the project value through multiple realizations of price paths indicates the risk and uncertainty of the project’s cash flows due to uncertainty about future price movement, and support management decision based on the portfolio strategy – we may want to have a mix of high-risk high return projects and low-risk low-return projects. In the economic model we can build in any operational options or constraints as appropriate. For example, to evaluate the economics of a LNG liquefaction plant which liquefies gas from HH (Henry Hub, a gas hub in US)) and sells to other market(NBP in Europe, JKM in Asia), we can model the prices at different hubs and calculate the economics of the project assuming the following options: i) we will sell the LNG to the market/hub which has higher price to capture the uplift; ii) we will shut down the plant if the LNG selling price (in the best market) is lower than the total variable cost, including gas purchase price at HH and other logistic costs (shipping, regas., etc.), to prevent the downside. 28 The following charts illustrate the application of the method: The cumulative distribution functions on the lower left shows that the risk and uncertainty associated with the opportunity. Furthermore, the shifting of the blue curve to the right indicates the value of optionality of being able to sell the LNG to the higher market. Compared to the analytical solution of Black Scholes model, the simulation method is more “versatile” in evaluating different investment opportunities and optionality due to its capability to include various operational constraints/flexibilities and complex commercial structures in the evaluation. However, the Monte Carlo simulation method assumes the optionality applied at different stages are independent from each other. In reality, the decision made earlier could have impact on the options in the future, such as storage model, flexible contract model. In those situation, we have to apply global optimization considering the uncertainty, which is Stochastic Optimization Model. When to use Real options in investment opportunities Options mimic our behavior Bespoke options Create insight in the Risk and Uncertainties You do not have value (return) volatility data Longer term flexibility (capital investments) Market fundamentals change over time No market data, but statistical assumptions can be generated Benefits Transparent, not a black box (closed formula) Assumptions can be tailored to markets Can deal with complex real options Enables evaluation of Risk and Uncertainty What you need Sound fundamentals market understanding Ability to translate input risk and uncertainties into meaningful statistical parameters Drawbacks / Pitfalls Requires compute capacity for speed and/or complex problems Its usefulness for complex options can drive complex models 29 Method 5: Stochastic Optimization Model As mentioned above, simulations works well in evaluation of optionality where options are independent from each other, i.e., the decision to shut in a LNG facility in one month does not impact the decision to liquefy gas and sell LNG to the Asian market in the next month. In many real option situations this does not hold. For example, gas contracts have long term volume restrictions (min and max of Daily Contract Quantity, Annual Contract Volume, Annual take-or-pay obligation). A decision to take less this period will have impact on the decision for the next period. This means that whenever gas is lifted from the contract, the flexibility within the remaining operation period of the contract is reduced. Similarly for storage, the decision to inject more volume at this period will prevent from injection in the next period, even the injection opportunity in the next period is better. This dependency of future flexibility on today’s decision requires the stochastic optimization method which optimizes the actions at each period taking into consideration the uncertainty in the future period such that on expected value basis, the objective function is met. Rather than generating multiple separate Monte Carlo price realizations, Stochastic Optimization is based on the generation of scenario trees, where each node represents the state of the world. Different scenarios share a common history before they branch apart. Each single path (from root to end) represents a price scenario. Each path of the scenario tree represents a joint discrete evolution of uncertainty, which altogether result in a multidimensional scenario tree. The tree structure appears similar to the Binomial Tree method above, but the calculation of the option values are different. In the Binomial Tree, there is only one decision to be made across the whole period; whereas in the stochastic optimization tree, there are decisions to be made at each time period to optimize to the objective. A simple summary of the tree structure which is explained in more detail below: Create a discrete tree that matches the price behavior as observed in all the generated price scenarios The contract duration is split into relevant time periods which form the ‘nodes’ of the tree (i.e. where the tree branches) Each node is associated with a price, derived from the price scenarios At each node a decision needs to be made on the quantity to be lifted under the contract and each node branches to multiple equal probability price scenarios every time Contract constraints (like max/min DCQ, ACQ, ToP) are fed into the optimization model and need to be met exactly after the optimization is done (volume allocation) Solving a decision tree is a linear programming example whereby volumes are allocated such that the average of all profits at the end of the tree is maximized, subject to the constraints above The profit results at the very end of the tree are equal in probability, and form a cumulative probability distribution function The following chart illustrates the process of the Stochastic Optimization which, similar to Monte Carlo simulation, starts from the model of the price path, followed by setting up the price scenarios trees with decision variables and constraints at each node to be solved by linear programming solver. Note that the 30 decisions at each node from the optimization solver do not give one optimum objective value, but a distribution of possible profits, as indicated in the cumulative distribution curve on the lower left hand side, such that the average or expected value from the distribution is optimized. The difference between the expected value of the distribution considering the uncertainty of future price movement and the value that can be locked in based on today’s market view of future price is the extrinsic value, or the optionality value of not locking in the decisions today. A noticeable problem in stochastic tree based optimization is the so-called “curse of dimensionality”, which is the exponential growth of the scenario tree with the number of branching time steps throughout the planning horizon. Obviously, it is not possible to branch the tree in a daily or even hourly granularity, which will make the numbers of scenarios too large to solve. Usually, the optimization can be solved in reasonable time with up to 8-10 branches. Another optimization method considering price uncertainty is stochastic dynamic programming which uses a time/state space which discretizes both time and state (e.g., the inventory state in the case of storage is discretized to 0%, 10%, 20%...etc.). When the asset becomes more complicated, the time/state space becomes so large that it will run into computational issues. When to use When exercising real options restricts or changes our future options and choices that require optimization What you need Sound fundamentals market understanding Clarity on the dependencies and logical optimization decision protocols Benefits Can deal with dependencies in future choices Drawbacks / Pitfalls Often criticized as “black box” due to nonintuitive result Requires linear programming Can require very material compute power 31 Method 6: Machine Learning and Predictive Price Modeling Lastly, another method to improve price forecast to support business decisions for opportunities such as MoGas arbitrage and price forecasting for refinery feed/product slate optimization is through machine learning algorithms. It basically identifies the “external” variables that correlates with price movement, extract the relationships and trends between those contributing variables and price from historical data (“machine learning”), and then apply that relationship for future price prediction, using all available data. The difference between machine learning and the stochastic process is that machine learning tries to identify the “external” variables, as well as features from historical time-series data, that impact the future prices using data-driven statistical analysis approach; whereas stochastic process tries to calibrate the statistical characteristics of price from its own historical data (or other inputs). There are various modeling algorithms readily available from open sources in various script languages. The key is to identify the right “external” variables, or the “predictors”, for the price and uncover the “hidden insights” through learning and training from historical data, to derive the relationship that can be used to predict future prices. To characterize uncertainty of the forecast, there are two branches of methods: i) parametric methods which add a confidence band normally based on the assumption of normal distribution and the parameter of standard deviation; ii) non-parametric methods which builds prior and posterior probability distribution based on data and Bayesian theorem without assumption of the distribution and parameters. The following chart illustrates the process of machine learning – training, testing and forecasting: After extracting the relationship through learning and testing, we can apply that relationship to forecast and guide the trading decisions depending on the certainty bounds of the forecast. When to use Estimate / Forecast of near-term price change to support trading decisions What you need Reliable historical data, especially the variables/”predictors” that have impact on the price Benefits Allows to uncover “hidden insight” though learning from historical data Various modeling algorithms readily available from open sources Drawbacks / Pitfalls Constraint to trends and relationships observed in historical data Critical to include the right “predictor” and heavily rely on historical data 32 Method 7: Experience Based Volatility Probability Analysis Experienced people, based on fundamental analysis, insights and experience, can identify the typical ‘modes’ a market will be in; resulting in a limited set to describe all likely market ‘modes’. Each mode comes with an operating choice modeled to generate a value for the Mode. These ‘modes’ are assigned a probability of occurrence and by multiplying by the assigned mode values it generates a risk adjusted value: • • • • • Each mode may use any of the 6 prior methods to help with evaluation The valuation involves assessing the probability (e.g. how long a mode is present) , i.e. how long and how many times will we exercise the flexibility in the contract / facility and how much value will be in the exercise. Important to be explicit about the variables/assumptions, which may be “hidden” in the other methods Allows you to identify “key spreads” driving the value, the “killer variables” destroying the value, asymmetric upside enhancing the value, and ultimately manage the risk and uncertainty associated with these factors which has the most impact on the value (easy to be visualized through the tornado chart) Probabilities for each mode could change from year to year A mode has the characteristics of a scenario, but the difference is that a mode reflects a ‘state’ (e.g. applicable market clearing mechanism) a market may be in from time to time, whereas a scenario is a long term fundamentals view. An example of Baton Rouge crude optionality – coke morphology: the team found a $50M/yr flexibility incentive for envelope to cover all cases, in addition to a $60M/yr incentive to move to Opportunity Priceset slate. The volatility was assessed by looking at historical margins. The 2013-2017 margin of the best heavy crude in any month is $2.7/B better than the average of 10 likely crudes, however it is unlikely we can always capture the best of the best so the $2.7/B was discounted to $2/B for likelihood of capture: 33 The “experienced team” assigned probability to 10 different crude slates developed as perturbations of the long term crude strategy cases: BASE 20% 10% 2017 Actual 15% 10% 14% 8% Lead Case 6+ Yr Crude Typical 2017 Study - Lead Slate Case 8% 7% High TAN 4% 4% No Rail - Base 6+ Yr Crude Study - Lead Case Hard 6+ Yr Crude Study - Lead Case HardEST 6+ Yr Crude Study - Easy High TAN Base High TAN Hard High TAN Easy No Rail - Base No Rail - Hard No Rail - Easy Base Slate Lights 90 121 121 121 121 151 151 151 166 166 166 Dom Mediums - Sour 165 110 110 110 110 89 89 89 112 112 112 Dom Mediums - Sweet 140 161 161 161 161 124 124 124 131 131 131 AG Crude 28 40 40 40 40 40 80 47 47 30 47 47 40 20 50 40 80 15 15 35 15 35 55 15 35 23 55 23 23 13.5 2.0 4.5 1.0 1.9 5.7 26.8 Variable Slate Kearl Access Western Blend Cold Lake Castilla DCO SRR (Good Resid, e.g. Achinsk) SRR (Bad Resid, e.g. Basrah SRR) Basrah Heavy Ostra Import resid Joliet resid Baytown rock Total Fuels MVP Total Fuels API Total Fuels CCR 25 30 10 22 23 35 23 23 3.0 1.0 7.1 24.1 1.4 6.0 25.2 1.6 5.7 25.8 3.0 1.0 1.7 5.7 25.4 20 23 30 30 2.0 4.0 40 25 1.0 7.3 23.5 23 30 2.3 3.0 1.0 1.9 5.6 25.5 1.4 5.8 24.4 1.0 7.0 22.9 1.3 6.4 26.2 55 8.7 0.9 7.6 23.8 Three cases (summing to 22% probability) lie in the current Operating Envelope (OE), one 20% case is already covered in the Opportunity Price Set slate (6+ year Study lead case), leaving 58% outside the Operating Envelope and Opportunity Price Set slate, across the spectrum pictured below. If the project opened the OE to cover all the cases there would be an additional 58% x $2/B x 140kbd x (1-22% Tax) = $50M/yr of credits for Optionality: When to use Smaller dollar investments / opportunities that do not justify investment in a full quantitative methods What you need Sound fundamentals market understanding Clarity on the dependencies and logical optimization decision protocols Benefits Simple, does not require advanced analytics Drawbacks / Pitfalls Relies heavily on experts and preventing bias 34 Summary table of pros and cons and potential applications of various methods Method Probability Weighted Model Binomial Tree (Trinomial, Lattice) Model Input Assumptions Potential project values and the probability Expected project value and volatility – like Pros Simple, easy to understand Cons Purely experience based Potential Applications Screening analysis of potential opportunities Simple, Versatile to include operational constraints in the model Approximate continuous asset value and time to expiry of opportunity with limited discretization; Burdensome to set up the tree; Difficult to estimate the volatility of project value Evaluation of real options of Investment opportunities Black Scholes Model / Kirk’s approximation Expected project value and volatility (same as Binomial Tree) Close-form formula; Easy to calculate Financial options; simple spread options Simulation Model Statistical characteristics of prices: volatility, drift, reverted mean, mean reversion rate – usually calibrated form historical data with adjustment made based on experience and market insight Same as Simulation Model; with objective functions, operational constraints and flexibility laid out clearly Well documented and applied by industry; Can test different scenarios based in market insight; Can handle complex operational constraints and commercial structures Formula different to understand; European-style option; cannot handle compound options and complex constraints/flexibility Complex theory and mathematics behind the model; Need to get used to the distribution results for risk and uncertainty; Requires adjustment using expert judgment; won’t simplify the decision making process, but promote more informed decision making Often criticized as “black box”; Computational intensive while considering the uncertainty; Output often difficult to interpret for those uncomfortable with probabilistic and statistical results Predictive Price Modeling (Machine Learning) Appropriate external variables as predictors; period of historical data for learning and training Often criticized as “black box”; Critical to include the right “predictors” for the training and heavily relied on historical data availability; Assumption of no fundamental shift from history Near term arbitrage opportunity; refinery feedstock/product slate optimization; Maximize fuels margin from custom base Experienced Based Scenario and Volatility Probability Analysis Potential scenarios, and probability associated with each scenario Allows to uncover “hidden Insights” through learning from historical data Various modeling algorithms readily available from open sources in various script languages Simple concept, easy to understand and apply Requires “smart people in the room” – very much based on empirical experience to derive the assumptions Screening analysis, mature project with good understanding of the potential scenarios and probabilities Stochastic Optimization Most advanced method in terms of optimal decisions from the model capturing the real option value with consideration of the uncertainty of parameters For subject matter experst refer to Appendix 4: Subject Matter Experts 35 Exploration bidding; project development; Infrastructure evaluation; long term contracts; LNG pricing structure Storage Near term trading decision; Software Tools for Evaluation of Options Tools to enable method implementation come in a large variety, but can be bucketed in two mainstreams: - Analytical tools - Closed-formula models Methods: Black-Scholes; Kirk’s approximation, etc. For example: Spread option on a pipeline or tank commitment Tools: ‘FEA’ industry software; Python - Probabilistic tools - Stochastic simulation models Methods: Binomial; Monte Carlo; Stochastic Optimization For example: Facility shut in option valuation Tools: Excel with embedded macros; Matlab/C++ (3rdP build); Python For consistency, efficiency and ability to apply to ‘unique’ business situations, the medium term objective is to develop an in-house Python based core modelling library approach. Until such time there are a variety of tools already available and used within ExxonMobil and Appendix 3: Tools & Contacts provides a sample list and respective Subject Matter Expert contacts. For stochastic simulations excel based models have been developed to (i) model the price paths and (ii) enable all those path to be incorporated in existing economic models to understand the risk, uncertainty and optionality due to volatility of price. 36 Integration with Opportunity Evaluation The approach described by this paper integrates with the Opportunity Evaluation framework. Historically we valued the deterministic intrinsic value. Now we seek to value, include and separately identify, the value from asymmetry & optionality: the ‘Extrinsic Value’, when material in opportunities we are framing and presenting for a decision. The discussion and debate, encouraged in the OE framework, naturally extends into the asymmetry & optionality area. The linkage between Market Attributes, our business Enablers, competitive behaviors and our competitive advantage are all key elements in opportunity evaluation, and fundamental to the value of asymmetry & optionality. Integration with Scenario Analysis Scenario planning aims to define long term and large scale critical risks and uncertainties and develop plausible scenarios in order to discuss the impacts and the responses in each scenario relevant for an opportunity, project or strategy under consideration (reference S-BORE). By its very nature it tries to identify what drives a certain way the world, region or local situation can develop (for example big shifts in society, economics, technology, supply & demand and politics). These fundamental shifts will have impact on the uncertainties, and as a consequence on the assumptions. For this very reason, each scenario may have its own assumptions and therefore its own set of parameters such as volatility, mean reversion, etc., causing different valuations for real options and asymmetry in opportunities, projects and strategies for each scenario. Therefore each scenario may deliver a distinct Extrinsic Value. 37 Presentation of Results Where possible the results from the analysis should be reflected in a standardized format. Standardization is to help both decision framers and makers in articulating and understanding the output of the analysis and make comparisons easier. As Extrinsic Value methods are essentially all founded in the principles of probabilistic analysis, the recommended presentation is as shown in the three graphs below. Cumulative Distribution Function (CDF, aka S-curve, first and third graphs) is the primary approach, supported by Probability Density Function (PDF, middle chart). The examples display the results for a reference case and scenarios. While a PDF has more immediate visual interpretive power for many, a CDF has the benefit that it is visually easier to identify the differences between multiple scenarios. Extrinsic Value, as an addition to the Intrinsic Value is identified in the third graph through the dotted lines. 38 Key OE tools integration Metrics S-Bore Red/Blue Metrics should include extrinsic components if part of the evaluation Scenarios are key to asymmetry & optionality, enriching the discussion Extrinsic Value should be under the same scrutiny as the rest of the opportunity 39 Integrating Extrinsic Value into OE Experience shows that identification and valuation isn’t a guarantee to success. This sections articulates best practices to better assure value capture. Our business, and the opportunities we pursue within it, in many cases include asymmetric inputs (market attributes and enablers) and/or outcomes (returns) caused by or providing opportunity to be captured by executing on the flexibility in our contracts and physical systems. When we evaluate opportunities we should include, when material or relevant, the value of and the discussion around the asymmetry and optionality. However, being able to articulate the Extrinsic Value and its drivers does not mean we are assured of our ability to capture it (realize the value). Compare this to a project proposal to build a tank: When we present the opportunity for an investment decision, (FID/Gate 3), we have a design and execution plan on how to build it, a Project Management Team to oversee the development, a plan and process to establish startup and operations, and operations sustainment to capture the value from the physical asset we build. There is clear accountability and line of sight across the evaluation, planning, project execution (implementation) and operations phases on how we will realize the value. So when we justify an opportunity which includes a (material) component of Extrinsic Value, we need to have assured ourselves that we understand how we will capture the value. The below workflow shows expectations for the three phases (i) Opportunity Evaluation; (ii) Execution / Implementation; and (iii) Operations. This includes clarity and alignment around who is accountable for the business value and how we steward it. The next sections look at the key steps, post-valuation, that need to be thought through and planned for prior to passing the Gate 3 (or equivalent) gate for an opportunity. 40 Accountability The concept of accountability is quite simple: Who ultimately do we turn to for the overall result the opportunity delivers because she or he has all the means to enable the results, weights the risks and rewards, supports (or takes) the decision and is the business (P&L) owner. While in a big organization responsibility of components may be within different organizations (e.g. project execution with Global Projects, operations within the Operations organizations, etc.), ultimately there should be a single business owner that assures alignment with the proposal, the implementation plan, operations, commercial execution, etcetera. The accountable person own the consequential Profit & Loss and is accountable for the opportunity to his or her superiors. He or she has the opportunity to steward the results to measure progress and course correct; he or she also ensures alignment across responsible departments and functions. It is key that early on there is alignment on who is accountable and what groups have responsibility for delivering on the parts and pieces that enable capturing the results. The accountable business leader must ensure that: - The assumptions are sensible and based on fundamentals, market insights and expertise - There is a clear Execution and Operationalization Plan to capture the Extrinsic Value The valuation is understandable and justifiable and reflects the business case to be pursued The flexibility that enables the optionality or asymmetry exists or will be enabled in our system so that it can be deployed to capture the extrinsic value We have alignment on how the results will be stewarded The following sections address key assurances / components to be in place to give credibility that we can capture the extrinsic value. Implementation & Operationalization Plan (I&OP) When contemplating to move forward with pursuing the Extrinsic Value during the Opportunity Evaluation Phase, assurances need to be in place that we can realize the value. During this phase, prior to a go ahead decision and after having identified the Accountable Business Leader, a plan should be developed that outlines what is required to get ready for operations (the Implementation Plan) and how we will Operationalize it (Operationalization Plan). While Operations logically follow the implementation, only when the team has a clear handle on how operationalization looks, can an Implementation plan be credibly created. Elements of an I&OP are, but not limited to, and part worked in detail during the execution/implementation phase: - Data and information requirements (internal and external sources) Models and tools Systems and processes Organization, positions and interfaces (cross-functional) Roles & responsibilities Planning & Stewardship approach The accountable business leader ultimately owns the governance and must be satisfied the I&OP will enable value capture. 41 Implementation & Operationalization Throughout the implementation and operations startup a project team, or designated individuals, are responsible for enabling the future organization to consistently manage the designed flexibility. A key element here is the sustainment and preservation of all that drives the flexibility, the foundation, together with consistent optimization of contracts, assets and other enablers. Just like changes in the design of a facility changes its functions, so do changes to what makes up the enablers to capture the value from optionality and asymmetry. The accountable business owner provides the oversight and enables problem resolution governance to deliver to intent or better. Plan & Stewardship The fundamental challenge in planning and stewarding the results from optionality and asymmetry lies in the by nature unpredictable outcomes as measured in a set time frame (month, quarter, year). Optionality and asymmetry generate their value as a result of volatile environments, therefore in many cases the value is not captured ratably. In some cases no option may have been exercised, although in others it worked significantly consistently in our favor. This fundamental nature makes stewardship more difficult than, for example, reviewing simply ‘asset uptime’. Such a metric in itself may not work, and needs to be combined with when the ‘signal’ is that value can be created in a volatile market, versus when not. If a pipeline ships oil only when the price differential (hopefully materially) exceeds the variable costs by design, then throughput is a function of the markets differentials and our ability to ship incremental oil. These profit opportunities are unlikely to nicely match stewardship fixed time measurement periods. Ultimately there are three core methods to stewards: Identified Options Material options to a large business line should be identified for individual stewardship and tracked Did we triggered the option (was prescribed action taken or not), time delay in response to trigger, value generated after action taken, value lost if action not taken or due to delay Data is consolidated and results reviewed to evaluate o the effectiveness of the trigger and the response by looking at value generated o ability to respond to triggers in a timely manner o additional value created by newly identified options Evaluation o Value generated from project identified triggers will be compared to the value estimated during project valuation o Reasons for non-responses or delayed responses will be evaluated along with value losses for potential issues with the original design of the trigger and responses Triggers and the responses will be continuously improved based on learnings 42 o Portfolio Earnings Investment Reappraisal Options identified post project will be evaluated to see what could have been done to anticipate these triggers during project definition phase for a more efficient design for future projects In many cases options may be smaller (or a collection of them) and managed at a lower tier in the organization Such groupings of options are most efficiently managed against earnings; there are enough that some ratability sets in and portfolios can be understood By design a IR will review versus the project basis, and should thus evaluate how the initial assumptions, valuation and I&OP were handled Quality of assumptions in Opportunity Evaluation versus what actually transpired; is the volatility similar to the expected Was the I&OP properly implemented Is the optionality being captured Trigger response ability evaluation along with value implications versus original understanding and design Have new options been effectively identified and pursued Options identified post project will be evaluated to see what could have been done to anticipate these triggers during project definition phase Ultimately optionality value is based on actionable events that can either generate additional income or prevent potential loss. Stewardship must include – where material – if we actioned appropriately. Fundamental shifts Markets evolve and so does our competition as it responds like other market players do. Therefore the accountable business, aided by – but cannot be solely based on - stewardship, needs to understand when the market attributes have changed and/or our enablers don’t provide an advantage anymore to capture the value. This requires a business from time-to-time to step back, understand the data, information and market insights, etc. over a longer period and what that does to the forward looking Extrinsic Value. This is part of taking ownership of our business throughout the value chain. Lessons Learned and Challenges Many companies have struggled with an effective implementation of what is described in this guide, however those who had the leadership commitment to it, build the capabilities successfully. Historically EM has tried and ‘failed’ in consistent evaluation of Asymmetry & Optionality. Our culture historically preferred certainty over uncertainty, our management systems may have encouraged this, stewardship in particular. A natural bias to overweight downside risk versus upside risk doesn’t help. We anticipate, just like it did within competition and others outside our industry, that it will take time to build our organizational capability in this area. Decision framers, practitioners of the valuation techniques, and specialists are key in driving the valuation principles in an unbiased manner based on their technical or specific expertise and building on the strength of the collective competencies of all of our functions! 43 Appendix 1: Optionality & Flexibility Identification Aid Area Category Enablers Capital investments Phased development (vs. all at once) Design not limiting future expansion capability Ability to cancel when new/updated information presents itself Pre-invest for future expansion Low cost to "do now" (pre-invest) during project execution for future capability/optionality Residual interest Retain exposure to future margins to sustain option Portfolio effect Reduced matrix / optimization flexibility JVs Protection from downside if partner interest/business direction diverge; EM sees more/less value in JV vs partner Crude or Product pricing Backward looking pricing option or capture of forward market structure (Backwardation/contango) Volume flexibility Produce non-ratable - pricing vs volume profile Divestments Commercial structures Initial outlay to retain capability for lower cost future expansion Ability to adjust volume based on our value in use Volume deferral based on view of future vs current price Ability to access and secure rights to limited logistics Production sharing contracts/agreements Tax Limit or increase exposure to crude price and production volume Cost basis management Tax defensibly - Leveraging low tax affiliates Trading Execution / optimization Origination Creation of incremental long and/or short positions that create value by allowing to capture pricing volatility , producing asymmetric returns Time arbitrage Differential arbitrage Rights to storage (logistics capability) or drawdown to respectively capture contango/backwardated market structure, usually of flat price, sometimes seasonal crack spread Contango on crude quality differential time arb Geo arbitrage Ability to move (or not) molecules (feed/product) to different market under different pricing Quality arbitrage Dynamic Forward Trading Exploit inherent giveaway in Hydrocarbon batches (e.g. NY harbor blend hub, Crude Blending) Volatility in forward derivative market Destination flexibility Vessel/pipeline destination options (or not) Ability to move molecules (or not) to different market - Crude oil destination restriction (Dfree vs D-restricted) Sourcing / supply flexibility Matrix of supply options to purchase from and deliver to Quality Ability to switch between sales channel (supply/wholesale/retail) and grade mix Access to and flexibility to substitute components and capability to blend products to specific market quality Ability to nominate crude between fields with differing quality under same price (e.g. Al Shaheen) Capability and Flexibility to process wide spectrum of feed - not limited by site processing or product quality constraints Access and capability to capture crude quality differential not priced into premium Second Order Pricing Effects / Logistic Power balance Oversupply in market may result in lower prices in the market (converse is true), Logistics outlets may negate impact Pricing negotiation leverage for alternate disposition capability This page refers to page 13 (and to Spreadsheet link) 44 Appendix 2: Enabler & Operationalization Checklist Purpose Consider the additional requirements (above any for intrinsic value) that will be necessary to capture the extrinsic value credits included in the OE. These additional requirements should be documented and (where applicable) costed and included in the OE cost basis. The accountability for implementation of the additional requirements and subsequent capture of the extrinsic value should be clearly defined. Pricing volatility Be able to vary the volume (produced/shipped/purchased/sold/refined etc.) with time on a short-term basis (day-to-day or month-to-month) in order to capture the most advantageous pricing. This should produce an asymmetric (positively skewed) return even if the original pricing distribution is symmetric Considerations Examples of potential requirements or additional costs Logistics - additional capability Additional tankage to allow volume variability with time (e.g. pipeline storage, longer tank-hold times if non-ratable) Increased capability above base design requirements - e.g. increased pump or line sizing to allow volume variation Pre-investment for future optionality to expand Wider refinery or plant operating envelope for (e.g.) TAN, sulfur, coke morphology to capture spread volatility with time Increased operating costs Typically if costs are non-linear with the necessary volume variability - e.g. unit start-up or shut-down, power, additives, overtime, additional storage, tankage throughput Additional staff resources Monitor the required (e.g. pricing) data and decide when and how to exploit the optionality giving the extrinsic value Additional operating resources if capturing extrinsic value would increase operational complexity (refinery, pipelines, scheduling) Additional commercial or sales resources to manage non-ratable volume flows Training to develop/reinforce skillsets necessary How to exercise volume tolerance optionality given flat price or spread evolution How to exercise settlement optionality (e.g. true-up or volume sale) on (product) exchange deals given pricing evolution Additional process inputs to allow monetization Additional data (e.g. real time market data for a specific item) Ability to trade derivatives not currently authorized by EM’s Trading Authorities (but available in market) Additional technical expertise (e.g. on a specific blending process) beyond that present in EM 45 Improved analytical tools Follow flat price evolution through to decision point when optionality exercised Predictive or monte-carlo analysis to assess price path from decision point to price-out completion Track monthly indicators (e.g. TAN in crude slate) versus annual or turnaroundturnaround cycle if variation of these gives extrinsic value More extensive project documentation Ensure that optionality on long-term contracts does not become forgotten or ignored by organization over time Change decision timing on (e.g.) investments or divestments Phased development - make multiple small decisions versus one initial Considerations Examples of potential requirements or additional costs Additional costs incurred Additional multiple review steps Future escalation costs above inflation for materials (e.g. catalysts) Loss of ‘buying in bulk discount’ - e.g. commitment length of a time charter Contractual terms Changes to contractual terms for this contract Potential changes to terms for other contracts that would offset 46 Appendix 3: Tools & Contacts Tool Contacts Internally Developed Excel / Matlab based Stochastic Modeling of Price (Mean Reversion, GBM) Excel based macros or Matlab code to run through multiple price path in economic models Excel based Black Scholes model and Kirk’s Approximation Script language based Machine Learning UIS RTD Customer Tools (Gas / LNG) Tingting Yao; Obinna Duru Obinna Duru Hui Zhang; Neil Gupta Shiva Kameswaran; Zimin Lu Bora Tarhan; Kevin Furman External Tools - Licensed C++ FlexTool with user interface by DecisionTrees • • • • C++ with Excel Add-on tools by FEA • • • • Jean Romersheuser Guy; Andrew Fleming Evaluation of value of optionality in flexible contracts Evaluation of infrastructures (storage, pipeline, terminal) Decision support on asset planning, contracting, marketing, trading activities Expertise in stochastic optimization Hui Zhang; Neil Gupta Evaluation of financial options Evaluation of storage and swing contract Decision support on contracting, marketing, trading activities Expertise in evaluation of financial options External Tools - Third-party Platform (requires building the applications) @Risk (Excel plug in) CrystalBall (Excel plug in) SIPMath (Excel plug in) JMP AIMMS Matlab R Python Bora Tarhan (RTD) Obinna Duru PIN (The Python Interest Network) Price volatility & fundamentals stochastic modeling: The Upstream Opportunity Evaluation & Analysis (OEA) team is working closely with UIS Research Technology Development (RTD) to engage both internal and external experts to improve the methods for price modeling. 47 Appendix 4: Subject Matter Experts Function UIS - OE&A Name Tingting Yao UIS - OE&A SM Suresh F&L Optimization Tim Suter F&L - C&T QAS David Dalton F&L - C&T QAS CSP Treasurer’s - Chemicals Treasurer’s – F&L Hui Zhang Drew Bishop Ted Postula Jorn Sturkenboom UIS - GTO&T UIS – RTD F&L – DAS UIS – OE&A Neil Gupta Bora Tarhan Shiva Kameswaran Obinna Duru Specialty Statistics & Stochastic Modeling Monte Carlo Simulation Valuation & Concepts Upstream Senior Economics Consultant OE Integration Business identification OE Integration Bridging the identification into quantitative valuation Quantitative Analytics & Structuring OE Management, Process & Integration Valuation & Concepts Valuation & Concepts OE Integration Black Scholes; Kirk’s approximation Stochastic Optimization Predictive Price Modeling Probabilistic Simulation 48 Appendix 5: Examples 49 50 51 52 53 54 55 56 57 58