Econometric modelling for short‐term oil price forecasting

• Date: 01 March 2010
• Published date: 01 March 2010
• Author: Antonio Merino,Rebeca Albacete
• DOI: http://doi.org/10.1111/j.1753-0237.2010.00171.x

Econometric modelling for short-term oil

price forecastingopec_17125..41

Antonio Merino* and Rebeca Albacete**

*Chief Economist, Economic Research Department, Repsol, Paseo de la Castellana 278, 28046 Madrid,

Spain. Email: amerinog@repsol.com

**Senior Economist, Economic Research Department, Repsol, Paseo de la Castellana 278, 28046 Madrid,

Spain. Email: ralbacetesm@repsol.com

Abstract

There is a lot of interest in forecasting oil price and in analysing which variables most affect price

movements, especially whether non-fundamental variables such as financial activity have any sys-

tematic impact on oil price. In this paper we approach both questions by constructing a congruent

econometric model with financial and fundamental variables and byanalysing the relative weight of

the variables in explaining the oil price forecast.After testing for different variables we find that the

most accurate forecast from a monthly econometric vector model on oil price is obtained whennon-

commercial long positions, petroleum stocks and spare capacity are included as explanatory vari-

ables. The incorporation of non-commercial long positions clearly improves the accuracy of the

prediction. The vectormodel is specified to include empirical cointegration relationship, which pro-

vides an approximation on the long-run restriction postulated by economic theory.

1. Introduction

The Western Texas Intermediate (WTI) oil price increased significantly in nominal terms

from $12 in January 1999 to $134 in June 2008. Historically,the most widely used expla-

nation for the oil price movements has been the situation of fundamentals in the oil market

and/or the expectation of further tightness/looseness on the supply and demand balance.

Nevertheless, some authors haveexpressed their doubts about fluctuations in oil prices

being explained only by fundamentals.They argued that high nominal prices are the con-

sequence of the existence of a substantial risk premium overand above the price consistent

with fundamentals.

Traditional models are usually specified taking into consideration the present and past

history of oil supply and demand,also oil price fluctuations could be the result of expected

future imbalances in the oil market. One can go a step further by saying that the formation

of expectations in the oil market has changed because of the increasing weight of financial

Wethank Rodnan García for his help with the data and his comments.

25

© 2010 The Authors. Journal compilation © 2010 Organization of the Petroleum Exporting Countries. Published by

Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

agents in the market. Expectations are more important for financial agents—even if their

views are only focused on a long-term investment—than for industrial ones, which are

mainly trying to hedge their physical positions or needs. Financial agent positions should

be classified or reported as ‘non-commercial’ positions but this is not the case because

some of them participate in the market through a swap dealer,which is considered a ‘com-

mercial’ agent. However, the majority of the reported ‘non-commercial’ positions are

taken by financial agents in the oil futures market and these are known by the general

public as ‘speculator’ positions. Previousresearch, Roll (1984), Pindyck and Rottemberg

(1990), has shown the potential role of speculation in explainingcommodity price fluctua-

tions, i.e. oil price movements are partially the result of ‘speculativeactivity’.

Other authors have shown that speculation cannot explain the recent evolution in oil

prices, for example, IMF (2006). However, they have used only in-sample procedures to

justify their position.

For a recent conceptual discussion on the different viewpointsabout the role of specu-

lation and more broadly about the role of financial activity on the price of oil, the article

by Dr Ahmad R Jalali-Naini included in the OPEC Bulletin, June 2009 is highly

recommended.

However, the scope of this article is limited to show that at least a financial variablehas

explanatory power to forecast oil price in an econometric model.

The remainder of this paper is organised as follows. In Section 2, we formulate an

econometric model for the oil price in order to obtain accurate forecasts, including an

explanation of the main determinant factors (variables) determining the price forecast.

Section three evaluates and compares the forecasting performance of this model with

alternative approaches. Finally, in Section 4 we present some conclusions.

2. An oil price econometric model

In this paper, a monthly causal-based econometric model for oil price is proposed. In this

regard, it is important to identify explanatory variableshaving a causal relationship on oil

prices, according to economic theory and previous literature results. Afterwards,you need

to build congruent econometric models, as defined in Hendry (2001), based on these theo-

retical approaches and according to the data available.

The appropriate econometric treatment of the resulting external multivariate data set

requires vector modelling which captures the long-run restrictions between the different

time series and the short-term correlations between their stationary transformations.

Without this simultaneous modelling, there is no guarantee that one could obtain better

forecasting results with the enlarged data set, according to Espasa and Albacete (2007).

The following subsection describes the selection process and the final explanatory

variables included in the congruent model built for oil prices.

Antonio Merino and Rebeca Albacete26

OPEC Energy Review March 2010 © 2010 The Authors.

Journal compilation © 2010 Organization of the Petroleum Exporting Countries