Are published oil price forecasts efficient?

• Published date: 01 March 2019
• Author: Franz Wirl,Jalal Dehnavi,Hussein Moghaddam
• Date: 01 March 2019
• DOI: http://doi.org/10.1111/opec.12138

Are published oil price forecasts efficient?

Hussein Moghaddam, Jalal Dehnavi and Franz Wirl

Department of Industry, Energy and Environment, Faculty of Business, Economics and Statistics,

University of Vienna, Oscar Morgenstern Platz, 1, A- 1090 Vienna, Austria. Email:

hussein.moghaddam@univie.ac.at, Email: jalal.dehnavi@univie.ac.at, Email: franz.wirl@univie.ac.at

Abstract

Oil prices are crucial for a wide area of economic decisions ranging from households over business

to economic policymaking. Therefore, oil price projections are a crucial input to economic models

and to economic activity forecasts. This paper assesses the accuracy and efficiency of crude oil

price forecasts published by different organisations, think tanks and companies. Since the sequence

of published forecasts appears as smooth, the weak efficiency criterion is clearly violated. Even

combining forecasts, cannot increase efficiency due to high correlation among various forecasts.

This pattern of oil price forecasts can be attributed to combining myopia (use current oil price

levels as a basis) with Hotelling-type exponential growth. Another behavioural explanation in

source of inefficiencies is that forecasters prefer to harmonise their forecasts with other forecasters

in order to be not an outlier.

1. Introduction

Decisions at the macro (governments) and micro (individuals and firms) levels depend

on expectations about the future economic developments, which are often based on

published forecasts. Therefore, the efficiency and accuracy of forecasts is of crucial

concern for many decision makers (Mamatzakis and Koutsomanoli-Filippaki, 2014)

given the garbage in, garbage out (GIGO) principle. This includes governments when

planning their budgets, managers and individuals about investment projects involving

billions of dollars or less in case of individuals, e.g. buying hybrid car. Demand, supply

and price forecasts are crucial for all energy market participants (Sanders et al., 2008,

2009) yet oil price forecasts of particular importance due to implications of oil prices on

all aspects of energy markets, on commodity markets and on overall economic activity,

GDP-growth and inflation (Mamatzakis and Koutsomanoli-Filippaki, 2014).

This importance has led to the development of different and efficient forecasting

techniques and to the efforts of statisticians, economists and market players (companies,

governments and international organisations) to forecast. Two techniques have gained

popularity due to their accuracy and efficiency: econometric models (both parametric

©2018 The Authors. OPEC Energy Review published by

John Wiley & Sons Ltd on behalf of Organization of the Petroleum Exporting Countries.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs

License, which permits use and distribution in any medium, provided the original work is properly cited, the use is

non-commercial and no modifications or adaptations are made.

29

and non-parametric) and more recently computational approaches like neural networks

(Liu et al., 2002).

Many studies have employed econometric models to forecast oil prices, Kaufmann

(1995), Alquist et al. (2011), Baumeister and Kilian (2013) and Manescu and Van

Robays (2014); Kaboudan (2001), Yu et al. (2008) and Gabralla and Abraham (2013)

have applied computational techniques. However, there is no consensus on the most

reliable method (Liu et al., 2002) and it is close to a religious war between the two

camps. Hence, scholars have begun, in parallel, to evaluate forecasts (Mamatzakis and

Koutsomanoli-Filippaki, 2014). Consequently, they have introduced various perfor-

mance measures, such as forecast accuracy (Clement, 1999; Manescu and Van Robays,

2014), bias (Artis and Marcellino, 2001), combination and encompassing (Bates and

Granger, 1969; Clemen and Winkler, 1986), efficiency (Mincer and Zarnowitz, 1969;

Nordhaus, 1987) and sign predictability (Nyberg, 2011).

Forecast efficiency has been investigated in Cooper and Nelson (1975), Nelson

(1984), Fair and Shiller (1988, 1990), Dovern and Weisser (2008), Allan (2012), and

Genberg and Martinez (2014) not only for macroeconomic variables but also for football

(Sillanp€

a€

a and Heino, 2013), baseball (due to excellent performance statistics, Silver

(2012)), agricultural commodity markets (Von Bailey and Brorsen, 1998), managers’

forecasts (Khan et al., 2013), demographic variables (Tayman et al., 2011), etc. This

study focuses on the forecast efficiency of oil prices, more precisely on weak efficiency,

which requires that forecasts contain the information sets of all past forecasts (Nordhaus,

1987). Although this is only a necessary criterion for efficient forecasts and other tests

will be included, it is of particular practical relevance as it allows forecasters to check

and to learn from their past revisions and to assess the efficiency of a forecast prior to its

realisation.

The energy sector receives more than its fair share of forecasts (Ahlstrom et al.,

2013). A number of studies investigate their efficiency focusing on oil supply (Floris

et al., 2001; Lynch, 2002 and Sanders et al., 2009) and oil consumption (Shlyakhter

et al., 1994), however, there is relatively little research about the efficiency of oil price

forecasts in spite of many and even regular forecasts. Sanders et al. (2008 and 2009)

investigate the efficiency of price forecasts for several energy commodities (including

crude oil) published by the United States Department of Energy (DOE hereafter). They

find that the price forecasts for gasoline, diesel fuel, natural gas and electricity are

efficient in the long term,

1

that the price forecasts for crude oil provide incremental

information for up to three quarters. Similarly, Mamatzakis and Koutsomanoli-Filippaki

(2014) examined the rationality of DOE price forecasts for energy commodities,

including crude oil. They opt for an asymmetric underlying loss function with respect to

positive versus negative forecast errors. The above-mentioned studies have assessed the

rationality of DOE’s forecasts using quarterly data, while the earlier, and in this area,

OPEC Energy Review March 2019 ©2018 The Authors. OPEC Energy Review published by

John Wiley & Sons Ltdon behalf of Organization of the Petroleum Exporting Countries

30 Hussein Moghaddam, Jalal Dehnavi and Franz Wirl