Written in collaboration with Malcolm Wardlaw
As I mentioned in my earlier posts on crude price divergence, the market’s view of the Gulf oil spill post-event and future oil prices, futures prices provide “forecasts” of future oil and gas prices. However, people in the industry tend to misunderstand the relation between futures prices and the prices that are expected to be realized in the future, and perhaps because of this, they tend to be somewhat critical about the accuracy of those forecasts.
In this post, I explain that although futures prices can indeed be used to forecast commodity prices, the futures prices themselves are not the forecasts. In addition, I explain how prices of commodity options can be used to gauge the accuracy of those forecasts. As I’ll discuss below, the options markets tell us that we should not put a lot of confidence in the price forecasts that can be obtained from the futures markets. This does not suggest that there exist forecasts that are better — just that commodity prices are difficult to forecast.
But first, we need to understand the relation between futures prices and the prices that we expect will be realized in the future.
It is useful to think about futures markets as consisting of two types of players: hedgers, such as oil and gas producers that want to reduce their exposures to oil and gas prices, and speculators, including hedge funds and other investors, who make (or lose) money by betting on oil and gas price movements and buying (or selling) futures contracts accordingly. Because of this interplay of interests, futures prices and expected spot prices will differ by an amount that depends on the magnitude of the premium that is associated with the commodities risk exposure:
Expected Price = Futures Price x (1 + risk premium)
For most commodities the risk premium is positive. Speculators who buy commodity exposure require a risk premium for doing so, which means that to get the market’s expectation of the price in the future, one must gross up the futures price by a risk premium. So, for example, if the futures price for oil in 5 years is $80, and the appropriate risk premium is 3 percent per year, then we should gross up the $80 price by a bit more than 15 percent (since the growth rate compounds), suggesting that the market expects the price in three years to be a bit more than $92 per barrel.
The appropriate risk premium for futures contracts is determined by the premium that investors demand for holding the long side of oil and gas price risk. Whether this rate is 5 percent or zero depends on how returns on oil and gas move with the rest of the economy, the overall premium demanded by investors, and the relative supply of producers that may want to hedge and investors who may want to hold oil and gas assets in their portfolios.
The Traded Price of Oil and Gas Futures
The financial markets that price oil futures attempt to standardize the pricing process by trading financial contracts on a specific type of oil. Crude oil around the world varies in its hydrocarbon makeup and the relative cost of delivery and transport. As such, financial markets create contracts on so-called "marker crudes," which create a benchmark for oil prices. The primary crude oil contract in the United States is known as West Texas Intermediate or WTI. WTI is a light crude, meaning it is highly valued for its ability to refine high-end fuels such as gasoline, diesel, and jet fuel. Futures contracts on WTI specify a price at which traders can commit to buy a set number of barrels of oil to be delivered at its primary pooling point in Cushing, Oklahoma, at some point in the future.
In this way, the futures price of oil represents the price at which producers and consumers are willing to commit to buy and sell oil at any point at time in the future, and their best estimate, adjusted for risk, of the price going forward. Table I presents the prevailing futures price, as of mid-March, and expected prices in the future that I estimate by assuming risk premiums of 3 percent and 5 percent.
As can be seen in the above table, although the futures price declines slightly with maturity, depending on the assumed risk premium, the market is expecting that prices will slowly increase over the next four years.
The same information can be used in evaluating natural gas futures. Natural gas contracts are standardized primarily by location, since the transport of natural gas via pipeline represents the most important cost differential. In the U.S., natural gas futures are priced off of gas delivered to Henry Hub, near Lake Charles, Louisiana. Thus, each contract represents the expected price, adjusted for risk, of one million cubic feet of natural gas delivered to Henry Hub. Table II presents the prevailing price, as of mid-March, and the estimated future price given a 3 percent and a 5 percent risk premium.
The above table indicates that futures prices increase steadily with maturity, implying that if we use the same risk premiums as we did for oil, the market expects substantial increases in natural gas prices.
Option Prices and Confidence Intervals
Oil and gas futures markets, like other commodities, also have fairly actively traded options on futures contracts. The prices of these options provide information about how confident we should be about the market’s future price expectation. The prices of these options allow us to estimate a projected distribution of potential prices around the expected futures price. In particular, high option prices imply that investors are very uncertain about the prices that will be realized in the future. Given certain assumptions about the movement of oil prices, we can use this distribution to form confidence intervals around the expected price of oil.
Option prices are estimated using models of future price movements that incorporate the current futures price and the underlying volatility of the movement of prices. The most well-known of these models is the Black-Scholes option pricing model. While the underlying assumptions of the Black-Scholes model do not fully hold for the movement of commodities, for our purposes here, the formula provides a sufficiently close approximation that it can be used to back out the market’s expectation of volatility from the option and futures prices.
Since option prices in the market are implying a certain level of volatility, the market is therefore providing an estimate of how “wide” the distribution of potential prices could be and what the probability that prices will fall within a certain range. These estimate of the range of future oil prices as of mid-March are given in Table III.
Similar calculations can be formed on natural gas, though since the options market is more thinly traded, we cannot obtain implied volatilities out as far.
As can be seen in Tables III and IV, the 95-percent confidence interval generates a pretty wide band of possible prices, both for oil and for natural gas. For example, these confidence intervals state that the price of oil in two years will be between $50 and $200 per barrel, which may be good or very bad news for trucking fleets. Similarly, there is a 5 percent chance of gas rising beyond $8/mmcf or below $3/mmcf in the next year and a half. While these are statements that can be made with a high degree of confidence, the estimates are clearly not very precise.
The recent uprising and UN sanctioned military action in Libya has created a certain degree of uncertainty in the area regarding future oil prices. This can be seen in the increase in implied volatilities, which also implies a wider estimated confidence interval for future oil prices. As can be seen in the following two graphs of the 50 percent and 75 percent confidence intervals in mid-January, prior to the uprising, and mid-March.
In addition to the increase in futures prices, we also see a significant increase in the near term volatility which dramatically increased the confidence interval around prices over the few months. However, the increase in uncertainty dies off fairly quickly, and the uncertainty around long-term prices, past six months, actually remains fairly similar to pre-crisis levels.
 These estimates impose a small mean correction to allow for consistency in the lognormal assumption. See http://www.eia.gov/emeu/steo/pub/special/pdf/2009_sp_05.pdf