How Effective Are Quantitative Methods in Forecasting Crude Oil Prices

What the paper says

IntroductionA crude oil futures contract is an agreement to buy or sell a specified amount of crude oil at an agreed upon future date at an agreed upon price and location. Unless offset, the parties are obliged to complete the agreed transaction at the expiration date. Market expectation, therefore, is reflected in the futures market where buyers and sellers fix future prices corresponding to the delivery times. Apart from providing much needed liquidity, large international futures markets also serve as price centers to worldwide traders as a whole. While less than three per cent of futures contracts result in the delivery of crude oil, futures remain an indicative benchmark of market expectation. Participants in the futures market in general are hedgers (commercial) and speculators (noncommercial) who are distinguished by their exposures to the physical crude oil traded in the futures markets.For this study, the light sweet crude contract for West Texas Intermediate (WTI) FOB Cushing that trades on the New York Mercantile Exchange (NYMEX), now integrated into the CME Group (Chicago Mercantile Exchange), is examined. While there are many other petroleum-based products that are traded (such as BRENT, Rotterdam, sour gas, and heavier oils like bitumen), and refined products (such as gasoline (RBOB, Euro-BOB) and heating oil), as well as highly correlated ancillary products (such as natural gas and bio-fuels), this study focuses on the WTI as it is the benchmark and most closely followed measure of the market for petroleum-based products in the world. Moreover, this future contract has the highest volume of any futures contracts and so this market segment exhibits continuous trading and would, therefore, not reflect any discontinuity that would distort either spot or futures prices. Lastly, most of the other futures contracts display a discernible discount or premium to the WTI, which makes their contracts easier to value correctly once the WTI relationship is measured. This study employs the most reliable data source for future estimation of prices available. Given the plethora of light, medium, and heavy contacts that are currently available, it would not be practical to attempt to measure them all individually. Rather, given the exhibited correlation among the contracts, it is imperative to measure the WTI with all other contracts then to be measured.PreludeBased on the literature review, variations of Auto Regressive Integrated Moving Average (ARIMA) and Vector Auto Regression (VAR) models are used to model the crude prices. The ARIMA model allows for the inclusion of information from the past observations of a series, but not for the inclusion of other exogenous variables that may be relevant. The price of crude oil is influenced by other economic factors. VAR models can be applied to model a vector of time series. VAR models including spot prices and stock (inventories) as exogenous variables are used for comparison of predictive accuracy.Crude oil price, like any commodity price, is influenced by fundamental variables, such as stocks (inventory), demand and supply. The crude oil prices are also impacted by factors such as geopolitical events, weather, and speculation in addition to micro and macro economics factors (EIA,1). Therefore, it is both complex and difficult to accurately model the price movement.Several studies have attempted to build mathematical models to help predict crude oil price movements. While the most popular forecasting approaches are based on traditional econometrics, computational approaches such as artificial neural networks and fuzzy logic expert systems have gained popularity in financial markets because of their flexibility and accuracy. However, there is still no general consensus on which method is more reliable (Behmiri and Manso,2).Pindyck3 used Kalman Filter to model long run behavior of crude prices using up to 127 years of price data. Fernandez4 used financial time series data to argue that the market either overreacts or under-reacts to new market information. …