Forecasting oil price movements: Exploiting the information in the futures market

Abstract

Relying on the cost of carry model, the long-run relationship between spot and futures prices is investigated and the information implied in these cointegrating relationships is used to forecast out of sample oil spot and futures price movements. To forecast oil price movements, a vector error correction model (VECM) is employed, where the deviations from the long-run relationships between spot and futures prices constitute the equilibrium error. To evaluate forecasting performance, the random walk model (RWM) is used as a benchmark. It was found that (a) in-sample, the information in the futures market can explain a sizable portion of oil price movements; and (b) out-of-sample, the VECM outperforms the RWM in forecasting price movements of 1-month futures contracts.

Forecasting the real prices of crude oil under economic and statistical constraints

“…Although the applications of some nonlinear methods have gained remarkable success (Xiong et al, 2013;Yu et al, 2008;Yu et al, 2014;Zhang et al, 2015), the linear predictive regression is still the most popular in forecasting oil prices. The predictors incorporated in the predictive regressions include oil futures prices (Alquist and Kilian, 2010;Coppola, 2008;Moshiri and Foroutan, 2006), oil production (Baumeister and Kilian, 2012, 2014a, 2014b, oil inventory (Baumeister and Kilian, 2012;Ye et al, 2005Ye et al, , 2006, crack spread (Baumeister et al, Forthcoming;Murat and Tokat, 2009) and some high-frequency financial variables (Baumeister et al, 2015b).…”

Determining Time-Varying Drivers of Spot Oil Price in a Dynamic Model Averaging Framework

“…The interested reader is referred to, for example, papers by Xu [19], Fan and Li [20] or Behimri and Pires Manso [21]. Generally, the oil price forecasting techniques can be classified according to the following scheme: time-series models [22][23][24][25], financial models [26][27][28][29][30], structural models [31,32], qualitative models [33][34][35], artificial neural networks based models, support vector machines, and other sophisticated methods [36][37][38].…”

Comparing Different Economic Linkages Among Commodity Futures

“…Fourth, and most importantly, we show that a GA return forecast which combines both chemical and industrial links is the most profitable. This suggests that the information content of the Chng (2009) industrial link is dissimilar in nature to the chemical link among related fuel commodities examined in Coppola (2008) and Cortazar et al (2008).…”

“…Since gasoline (GA) and kerosene (KO) are extracted from crude oil (CO) through fractional distillation, an economic link exists among the three fuel commodities. Coppola (2008) documents improved WTI spot return predictions from modeling NYMEX WTI futures and spot price cointegration in a VECM. Cortazar et al (2008) propose a multi‐commodity model to incorporate incremental information from related fuel futures (i) WTI and Brent crude futures, (ii) WTI crude and gasoline futures, and report improved return and volatility predictions.…”