The quantile dependence of commodity futures markets on news sentiment

• Author: Akihiro Omura,Neda Todorova
• DOI: http://doi.org/10.1002/fut.22010
• Published date: 01 July 2019
• Date: 01 July 2019

Received: 14 December 2017

|

Revised: 19 March 2019

|

Accepted: 20 March 2019

DOI: 10.1002/fut.22010

RESEARCH ARTICLE

The quantile dependence of commodity futures markets

on news sentiment

Akihiro Omura

1

|

Neda Todorova

2

1

Graduate School of Business

Administration, Meiji University, Tokyo,

Japan

2

Department of Accounting, Finance and

Economics, Griffith Business School,

Griffith University, Nathan, Queensland,

Australia

Correspondence

Neda Todorova, Department of

Accounting, Finance and Economics,

Griffith Business School, Griffith

University, 170 Kessels Road, Nathan,

QLD 4111, Australia.

Email: n.todorova@griffith.edu.au

Abstract

Focusing on energy commodities, industrial metals, and gold, this paper

examines the degree to which commodity futures returns depend on news

sentiment under various market conditions, and the structure of that

dependence. We observe an asymmetric market reaction to positive and

negative news sentiment, which changes in periods of financial turmoil.

The quantile regression analysis shows that news sentiment's influence on the

futures returns follows an upward trend at higher percentiles. This structure

flattens for positive news during the global financial crisis, while the slope for

the negative component steepens in backwardation periods.

KEYWORDS

backwardation, commodity markets, financial crisis., quantile regression, tail dependence

JEL CLASSIFICATION

C22, G10, G13, G14

1

|

INTRODUCTION

In today's digital world, large information flows lead to consensus views on market impacts, known as market or news

sentiment. Research has found that news sentiment is central to understanding short‐term market fluctuations. A well‐

documented phenomenon is the asymmetric nature of short‐term market responses to news, as negative news tends to

have a stronger market reaction than positive news does (Gao & Süss, 2015; Smales, 2014b). This finding is an average

result and raises questions concerning whether other parts of the return distribution are similarly affected and whether

overall market direction influences these results. Using the quantile regression analysis Baur (2013) proposed, this

study examines the degree and structure of the association between commodity market returns and the components of

news sentiment. The analysis is conducted on major energy commodities, industrial metals, and gold.

A vast early literature has attempted to capture the market sentiment using macroeconomic announcements

(e.g., Balduzzi, Elton, & Green, 2001; Barnhart, 1989; Boyd, Hu, & Jagannathan, 2005; Hess, Huang, & Niessen,

2008; Nowak, Andritzky, Jobst, & Tamirisa, 2011; Simpson & Ramchander, 2004; Simpson, Ramchander, &

Chaudhry, 2005; Smales, 2013; Smales & Yang, 2015). As Baker and Wurgler (2006) pointed out, other studies have

used market‐based measures like the closed‐end fund discount and the exchange turnover to construct a proxy for

sentiment (e.g., Bahloul & Bouri, 2016; Gao & Süss, 2015; Zheng, 2015). In recent years, the development of news

analytics has offered a more direct measure for capturing market sentiment. Analytics providers like Thomson

Reuters have analyzed millions of macro‐and microeconomic news articles and created sets of data containing

information regarding to which financial assets (e.g., specific commodity and company) the news item is relevant

and the tone of such news (i.e., positive or negative). An increasing number of recent academic studies

have measured market sentiment using news analytics' aggregated data (e.g., Borovkova & Mahakena, 2015;

J Futures Markets. 2019;39:818–837.wileyonlinelibrary.com/journal/fut818

|

© 2019 Wiley Periodicals, Inc.

Borovkova, 2015; Groß‐Klußmann & Hautsch, 2011; Maslyuk‐Escobedo, Rotaru, & Dokumentov, 2017; Smales,

2014a, 2014b, 2014c).

News analytics sentiment measures have three major advantages. First, the overall market sentiment of the day can

be disaggregated into positive and negative components to further demystify the role of sentiment. The literature has

indicated that a fall (rise) in the price of financial securities is a result of the markets being dominated by a negative

(positive) sentiment. However, these studies have not addressed four questions in sufficient detail: Is the degree/

magnitude of this relationship influenced by the direction of market (i.e., falling or rising)? What does the structure of

this relationship look like at various return levels? What is the role of the positive (negative) component of the

sentiment under a bearish (bullish) market? How would the market environment/condition (i.e., a period of financial

turmoil and the term structure of futures prices) affect this relationship? Answering these questions would help to

clarify how commodity futures markets work. Second, as the sentiment is constructed from positive and negative news

items on a specific day, it is simple to estimate, and the interpretation of the analysis results is intuitive. Third, the

estimated news sentiment contains richer information than sentiment proxies that use only macroeconomic

announcements.

Considering the basic association (i.e., a negative [positive] return is associated with negative [positive] sentiment),

the absolute value of the positive (negative) sentiment component is expected to have an economically and statistically

significant positive (negative) impact on the price when the market is rising (falling). By contrast, we expect the impact

of positive (negative) component to be limited when the market is falling (rising). Thus, when it comes to the structure

of dependence, an upward curve is expected when the quantile coefficients obtained for both sentiment components are

plotted against the return quantiles. Nevertheless, for the negative (positive) component, we expect the association to be

negative (around zero) in the lower percentiles and around zero (positive) in the higher percentiles.

Extending this basic hypothesis of the dependence structure, we argue that the steepness of the quantile coefficient

curve is affected by the level of speculator participation and the mix of private and state‐owned producers. Among the

commodities selected for this study, industrial metals have the smallest proportion of speculators' trading positions.

1

Based on the rationale explained below, we expect the slope of the sentiment coefficients for these commodities to be

moderate. On average, hedgers (speculators) take a net short (long) position (Moskowitz, Ooi, & Pedersen, 2012), and

hedgers tend to underhedge their physical positions (Hirshleifer, 1991). Under these circumstances, when the price

falls, the losses from the physical position surpass the gains realized from the hedged position. Thus, when the market

sees negative news that may drag the price down substantially, rational hedgers strengthen their short positions to

protect their future cash flow. By contrast, strong positive news may demotivate hedgers. In addition, speculators take

net long positions because they anticipate the price will appreciate. Therefore, their reaction to the news should

synchronize with that of hedgers, creating an upward curve of quantile coefficients. Nevertheless, as shown in Wang

(2003), hedgers often trade against market sentiment, so they might respond adversely to what the rationale described

above posits. Thus, these hedgers are likely to weaken the overall impact of the negative (positive) sentiment

component at the lower (higher) tails of returns.

In addition, the production of energy commodities is dominated by state‐owned enterprises (Bremmer, 2010b;

Greene, 2010; Hori, 2017; Kowalski, Büge, Sztajerowska, & Egeland, 2013; Pirog, 2007) whose goals, such as a job

creation, are often not driven by profit (Boardman & Vining, 1989; Bremmer, 2010a; Musacchio & Lazzarini, 2014;

Vernon‐Wortzel & Wortzel, 1989). These non‐profit‐driven producers can be less inclined to adjust their hedging

positions swiftly as news arrives, which inflates the impact of the speculators' response and is likely to steepen the

quantile coefficient curve of energy commodities.

A phenomenon that we saw in the last decade, the financialization of commodity markets is advancing,

2

so both

speculators and hedgers are interested in how financial turmoil affects the potential asymmetric association. Smales

(2014b) studied the gold market and showed that the news sentiment caused a stronger market reaction during the

recession (2007Q4–2009Q2). In Smales' analysis, a strengthening in the response is statistically significant for the

negative sentiment, but what has not been addressed is whether this strengthened dependence occurs with other major

commodities as well and is unaffected by the magnitude and the signs of daily returns.

In addition, the commodity futures market is unique by its nature, particularly because of the presence of a physical

market. The availability of a commodity, especially a nonagricultural product, can be adjusted relatively easily by

1

Measured from the open‐interest positions on futures exchanges.

2

Evidenced, for example, by the paper barrel discussion in Yergin (2011), the safe‐haven role of gold examined by Baur and McDermott (2010), and the strong growth in the market size of industrial

metals documented in Dwyer, Gardner, and Wiliams (2011).

OMURA AND TODOROVA

|

819