Crude oil prices are considered one of the most important indicators in the global economy. Governments and businesses spend a lot of time and energy to figure out where oil prices are headed next, but forecasting is an inexact science. Standard techniques are based on calculus (linear regressions and econometrics), but alternatives include structural models and computer-driven analytics. There is no widely accepted consensus on the best way to forecast oil prices.
Companies also pay special attention to – and often participate in – oil futures markets. Crude oil futures are traded on the New York Mercantile Exchange (NYMEX) and Tokyo Commodity Exchange (TOCOM).
Understanding Crude Oil Prices
At an elementary level, the supply of crude oil is determined by the ability of oil companies to extract reserves from the ground and distribute them around the world. There are three major supply variables: technological changes, environmental factors, and the ability of oil companies to accumulate and replenish capital. Technical improvements – especially hydraulic fracturing and horizontal drilling – helped flood world markets with oil after 2008.
Crude oil demand comes from individuals, companies and governments. Generally speaking, oil demand increases during good economic times, and it decreases during slower economic times. Increases in the standard of living in China and India have been a major source of global demand in the 21st century.
Companies need to understand these factors before making oil price forecasts, but even that isn't enough. Oil prices are heavily influenced by non-market forces, including the Organization of the Petroleum Exporting Countries (OPEC), which effectively acts as a multinational oil cartel. OPEC member nations make joint decisions about how much oil to release to world markets based on what is best for their governments. However, the extreme swings in oil prices between 2005 and 2015 are an indication that OPEC influence is limited.
Oil is also highly regulated in most countries. The United States, like many nations in Europe, has strict restrictions on where oil can be drilled; the Environmental Protection Agency (EPA) may have as much to say about oil prices as Exxon Mobil or British Petroleum.
The reason why movements in oil price (or any commodity) often surprise analysts is because there are hundreds of variables, each of them moving simultaneously in unpredictable ways. The Board of Governors of the Federal Reserve System put it best in their July 2011 discussion paper "Forecasting the Price of Oil," which began by identifying "unexpected large and persistent fluctuations in the real price of oil."
Companies hire econometricians and other market experts to make short- and medium-term predictions on the oil market. These professionals use highly complicated mathematical models, which either focus on financials (using spot and future prices), or supply and demand considerations (quantifying variables and testing their explanatory power).
Spot and future price models are still popular with many companies but are trending out of favor. The basic concept is that futures markets – particularly the relationship between futures price fluctuations and spot price fluctuations – will point the way to tomorrow's oil prices. Two influential academic papers were published in 1991 (Bopp and Lady; Serletis) that suggested that future oil prices were not unbiased or completely efficient, but were probably still better than any other indicators. This conclusion was reached through error and correction models (ECMs), which allow statisticians or econometricians to account for bias in futures data.
A third study in 1998 (Zeng and Swanson) looked at crude oil on the NYMEX, the New York Commodity Exchange, the Chicago Board of Trade and the Chicago Mercantile Exchange between 1990 and 1995. It found that ECM models performed best. Until the early 21st century, most companies employed the ECM approach.
Later studies have been less kind to financial models. One reviewed West Texas Intermediate (WTI) crude oil futures prices on the NYMEX between 1989 and 2003, finding that forward and futures prices are neither efficient nor unbiased enough to accurately predict future spot prices (and, curiously, that there was "little evidence of risk premiums" in the oil market). The authors instead recommended a time-series random walk process; random walk theory suggests that stock price changes cannot be used to predict future movement. (Research from the University of Portugal in 2013 discovered that time-series econometric modeling is the most common forecasting method for crude oil prices.)
Supply and demand models focus on macroeconomic variables, such as OPEC production, income elasticity of demand for oil and real gross domestic product (GDP). Because there are so many possible combinations of variables, most companies or analytic services use proprietary calculations and change their formulas frequently. The goal is to find the most statistically significant variables, then find chart fluctuations in those variables and create rough estimates for future oil price ranges.
Qualitative or Nonlinear Methods
The advocates of alternative approaches, which statisticians might call "non-standard" or "nonlinear" approaches, argue that future oil prices are too random and chaotic for any traditional processes. These methods might still use some of the same data as standard models, but the computations are based on pattern recognition rather than linear models or econometric regressions.
One popular pattern recognition tool is the artificial neural network (ANN). The ANN model, which is predicated on the biology of the human brain, supposedly lets the simulation learn and generalize experiences based on new data. ANNs are used for a variety of analyses in business, science and investment fields. One standard criticism of the ANN method – and a primary reason why ANNs aren't popular with private oil forecasts is the intrinsic inputs used to evaluate price series are often subjective or arbitrary.
Fundamental investors and analysts tend to shy away from complex statistical models. Instead, fundamental analysts rely on aggregate business factors, such as inventory levels, production trends, natural disasters and the actions of speculators. The implicit reasoning behind these knowledge-based approaches is that oil prices heavily affected by large, identifiable events. It is commonplace for companies to employ market analysts who rely on information from other sources, such as the World Bank's Commodity Forecast, rather than creating their own models.