How Not to Forecast Oil Prices (part 1)

Purchase and investment decisions rely on cash flow analysis, and cash flow analysis requires assumptions about future commodity prices. You can call it indistinctly an "assumption" or a "deck", but it is still a forecast of what produced oil and gas will sell for in 2018, 2019 and all the way through about 2050.

Predictions of future oil and gas volumes, and even predictions of costs, are much better constrained than prices which are not bound by scientific principles or anchored to current methods; prices can change rapidly and dramatically and in either direction. While enormous resources are dedicated to the engineering of future volumes, prices are often accepted and deployed with little thought. It is ironic since the impact of prices on the business is just the same the impact of volumes. ( volumes * prices = revenue)

At the start of my career, it was common for even small oil companies to have their own take on future prices of oil and gas. Over time, the NYMEX futures strip has supplanted independent analysis and become the near universal default price deck outside of large international oil companies. It seems that decision-makers have widely abdicated the seemingly sisyphean task.

Though analysis should not stop there, perhaps the first step to arriving at a prediction of future prices is to understand what other people think. To this end, DPPE publishes a monthly account of what other players are predicting about future oil and gas prices.

NYMEX futures: Though its shortcomings are legend, NYMEX futures prices are one source and perhaps no worse than others. Unlike some stocks whose movements can be a "random walk," the history of spot market prices for oil demonstrate mean-reversion. That is, prices tend to be pulled back toward a central figure. On the other hand, the "memory" of the futures curve tends to be pretty short, and the mean to which it reverts can move pretty far from a historical baseline. The chart below shows both patterns: reversion and evolution of the mean to which it reverts. Indeed, this pattern reflects the way people often think--heavily influenced by recent experiences.


Indeed, some would argue that they embody those sentiments. Theorists may argue that futures curves are like voting methods for the estimation of uncertainty. A population of participants in the market each have their own current, if evolving, expectation about what the future price will actually be, and each has money at the ready to buy or to sell given the chance as the market moves. For each individual contract to deliver oil in the future, there is a individual buyer one side and an individual seller on the other, and the whole market watches. Futures curve should represent a median price as measured by dollar-weighted voting, though those votes are mostly cast by financial speculators and much of that by algorithms. The more important fact is that the track record of the futures curve doesn't engender confidence. It doesn't take any computation at all to see that the futures curve has badly missed reality and frequently.

Sproule Associates: In some ways, the Sproule forecast is the antithesis of the NYMEX strip. Sproule Associates offers reservoir engineering services from its headquarters in Calgary. They are a knowledgeable, experienced and active member of the oil and gas community, and they are good enough to state publicly what they expect future prices to be, both in Canada and at various points in the US, as they see it from their perch as a prominent consulting company. The methods remain opaque (forecast "relies on Sproule’s proprietary models, analysis and insights") just like NYMEX, and it reflects a single entity's estimate instead of some of kind crowd equilibrium. On the other hand, Sproule is the only industry player (which I know of) to publish freely their expectations.

The Society of Professional Evaluation Engineers (SPEE) does survey its members (who are industry players) annually about the parameters they use in economic evaluations. The surveys are conducted around April and published around June of each year. At this point, it is hard to say how independent these forecasts are. What is more, it costs $135 to get the one data point per year.

Macquarie Energy Lender Survey: Macquarie, itself a financier of oil and gas projects, surveys each quarter about three dozen banks which lend regionally and nationally within the US and abroad, and it reports the average results. Though the participants may be active in the oil and gas space, the survey is much more useful to understand bank loans than to understand price expectations; the lenders intend to specify a price deck below expectations in order to provide coverage for the loans they make. The survey results can also be useful to understand changes in outlook, evolving as perspective within the industry evolves, though with some lag.

Investment banks: The Wall Street Journal polls investment banks monthly about their expectations. Though smart people, they are generally removed from the daily reality of the business, and what is more they may not be as independent as they seem. Banks are often aligned on the sell side of the business, promoting the offerings of the industry into the financial markets. Goldman Sachs has made headlines over the years by naming possibilities such as oil at $200 /bbl then at $20 /bbl. It is tempting to accept these as independent and courageous calls, but Goldman considers the trading of equities for its own account a core competency. And Goldman profits from volatility.

EIA short-term and long-term forecasts: The US government studies future oil and gas prices for the sake of planning and policy-making. The results are as close to independent as available anywhere, and they are released to the public in reports and tables. Both the short-term forecasts (released monthly) and the long-term forecasts (released annually) are generated by complex, econometric models. That is, the analysis works like a simulator of supply and demand driven by quantitative rules of economics for scores of parts of the global system. This is the same kind of analysis method used by planning departments in major oil companies which annually decree what the company price deck will be. Unfortunately, while some parts of the studies are made public (as by ExxonMobil or BP), the price deck themselves are generally considered top secret!

The models require initial assumptions and inputs about how one thing affects another, such as how rig count affects oil supply. The models are commonly back-tested against recent history. Still, the reality of rapidly changing relationships, again such as rig count and oil supply, make these kinds of models suspect. Even with the use of two models for different scales and time horizons, I haven't been able to tell that their forecasts are any more reliable than any others'.


Source: Speculation, Fundamentals, and the Price of Crude Oil by Kenneth MEdlock III, Aug 2013

None of these sources can be accepted ipse dixit. They vary in quality and in kind, and none are dispositive. What is more, when they mostly agree, then there is cause for concern. In a separate post, I will explore ways to use these forecasts as starting points. For now, please allow me to leave you with a quote from Ray Dalio whose experience is like mine. . .

"Truth be known, forecasts aren't worth very much, and most people who make them don't make money in the markets."

What do you think? Do you know of any other sources? Have any other insights to the utility of these sources? Thanks in advance for contributing.