Oil Prices: How did we get here and where are we going? – Part One

“It’s difficult to make predictions, especially about the future”


  • Attributed to everyone from Niels Bohr to Yogi Berra

A barrel of oil today fetches less than $50, when only six months ago oil traded at over $100. How could such a drop in prices not have been anticipated? Actually, it was!

Oil price predictions have been made for as long as oil has been produced, and the failures in forecasting made by the great—and not so great—are the stuff of industry lore. Our goal isn’t to name the guilty but to provide insights on the dynamics of oil market, the key forces currently working on the market, and the approach that in June 2014 led us to advise our clients of the pending drop in prices. Part one this two-part article presents the underlying model of price dynamics that we use to better understand the key driving forces driving prices. Part two will discuss how we used a fundamental understanding of these driving forces to advise clients of the pending drop in prices and how these forces are likely to drive future prices.

No amount of analysis or expert elicitation will change the simple, inescapable truth that future oil prices are uncertain. Decision making under conditions of extreme uncertainty is a skill the energy industry excels at for most uncertainties; however, price uncertainty is different. Leading executives have placed billions of dollars of their shareholders’ equity at risk in mega-projects based on assessments of multiple uncertainties, but they claim prices are “too uncertain” to treat in similar fashion. These same executives commonly choose to use “conservative” prices that are significantly below their expectations, ignoring that such bias skews investment choices and likely leads to suboptimal returns and paradoxically to overcapitalization.

We firmly believe that oil price uncertainty should be treated as are all other uncertainties; namely, as information to be used when making decisions. Too many industry participants have separated price forecasting from decision making. For example, a highly levered shale gas company making balance-sheet decisions needs fundamentally different information than a firm making a FID decision on LNG export.

Prices and markets

The oil market is complex, and there is no such thing as “the” oil price, regardless of how many pundits simplify the market to a single number. Like all healthy markets, the oil market has different products and dynamic prices for each product settled between buyers and sellers. The spot price (i.e., contracts are immediately settled with physical delivery within a month) is what makes the headlines, but the spot price is not the only product in the oil market. The accompanying figure presents a selection of different contracts from June 23, 2014, when the spot price was about $106/bbl, and for January 2, 2015, when the spot price was about $55/bbl. The 60-month contract has changed from about $87/bbl to about $70/bbl, or about a 20 percent reduction, which while substantial is not the 50 percent reduction seen in spot prices.
Figure 1 looks like a price forecast, and some market participants will use the forward curve that way; however, the forward curve is not a forecast of expected prices. Rather, forward prices are the value of the contract and thus include risk. Forward prices are then “risked prices” or “certain equivalent” prices. These risked prices should not be used as expected prices to develop cash flows that are then discounted by risk-adjusted discount factors. Doing so is a well-known double-dipping on risk that will distort investment behavior and suboptimize returns. While forward prices are not expected prices, they are useful in understanding price dynamics and providing insights for expected forecasts.

Price Dynamics

To state the obvious, prices are dynamic. An understanding of historical price dynamics has provided some of the insights needed to best inform current decisions in our practice. This involves, at a high level, assuming an underlying model for how prices evolve and then using numerical techniques to determine the parameters of this model from the historical record. Of paramount importance is the first step; namely, the subjective determination of which underlying model to use. Many quantitative analysts will squirm with our use of the term “subjective”; however, it’s true, necessary, and certainly not pejorative.

We want a model that isn’t so simple that it does not provide useful insights (e.g., price volatilities calculated assuming simple geometric Brownian motion). Nor do we want a model that is so complicated that it can’t be adapted to rapidly changing market conditions. We want a Goldilocks model that is just right by providing useful insights at minimum complexity. The “two-factor” model has met these criteria in our practice. This model assumes that short-term prices follow a random walk, but with mean reversion to a long-term equilibrium price (Ornstein–Uhlenbeck behavior), while the equilibrium price follows a geometric Brownian motion, albeit with a volatility much lower than that determined by looking at short-term price movements.

Figure 1 shows the estimated long-term mean prices for the two dates discussed above. While the spot price dropped 50 percent over this six-month period, the equilibrium price changed less than 10 percent.
Figure 2 presents the recent history of the long-term mean and the spot price. In our opinion, understanding the dynamics of these two prices is the minimum necessary to understand the dynamics of the oil price market.chart2 Firstly, the historical volatility of the spot price is much greater than the historical volatility of the long-term mean. This suggests that the uncertainty of future spot prices is much greater than the uncertainty of the long-term mean. As simple as this sounds, rarely is this insight implemented in practice. Secondly, while there is a correlation between the spot and long-term price (both tend to move up or down in concert), this correlation is far from complete, as there are times when the prices move in opposite directions. Additionally, the mean reversion of short-term prices can also be inferred from the relative prices. Finally, the volatility has been decreasing significantly for both prices over the last five years.

The second part of this discussion will present how using the above dynamic model of prices, when coupled with key industry driving forces, provides insights on why prices have changed so much recently and where prices are likely to go from here.