Contemporaneous and noncontemporaneous idiosyncratic risk spillovers in commodity futures markets: A novel network topology approach

• Published date: 01 June 2023
• Author: Xu Zhang,Xian Yang,Jianping Li,Jun Hao
• Date: 01 June 2023
• DOI: http://doi.org/10.1002/fut.22407

Received: 10 August 2022

|

Accepted: 8 February 2023

DOI: 10.1002/fut.22407

RESEARCH ARTICLE

Contemporaneous and noncontemporaneous

idiosyncratic risk spillovers in commodity futures

markets: A novel network topology approach

Xu Zhang

1,2

|Xian Yang

1

|Jianping Li

3,4

|Jun Hao

3,4

1

Department of Finance, School of Management Engineering, Nanjing University of Information Science & Technology, Nanjing, China

2

Post‐doctoral Mobile Station, Nanjing Yangzi State‐owned Investment Group Co., Ltd., Nanjing, China

3

School of Economics and Management, University of Chinese Academy of Sciences, Beijing, China

4

MOE Social Science Laboratory of Digital Economic Forecasts and Policy Simulation, University of Chinese Academy of Sciences, Beijing, China

Correspondence

Jianping Li, School of Economics and

Management, University of Chinese

Academy of Sciences, Beijing 100190,

China.

Email: ljp@ucas.ac.cn

Funding information

Social Science Foundation of Jiangsu

Province, Grant/Award Number:

20EYC011; National Natural Science

Foundation of China,

Grant/Award Numbers: 71903097,

72201265; China Postdoctoral Science

Foundation Funded Project,

Grant/Award Numbers: 2021M691635,

2021T140335; Humanity and Social

Science Youth Foundation of Ministry of

Education of China,

Grant/Award Number: 18YJC790226;

Natural Science Foundation of Jiangsu

Province, Grant/Award Number:

BK20190767; National Natural Science

Foundation of China (Major Program),

Grant/Award Number: T2293774

Abstract

This paper proposes a new network topology approach to identify the

contemporaneous and noncontemporaneous idiosyncratic spillovers of lower‐

moment and higher‐moment risks in commodity futures markets using high‐

frequency data. Our results show that contemporaneous information has more

explanatory power in constructing a network than noncontemporaneous

information, especially for higher‐moment risk spillover networks. In contem-

poraneous spillover networks, the role of one commodity future and the

structure of the networks vary across different realized estimators. Specifically,

gold, silver, and wheat are the main volatility and kurtosis risk transmitters,

while corn and silver are the main skewness spillover transmitters. Agricultural

futures markets are relatively closed in the volatility and kurtosis risk spillover

networks, while in the skewness network, theybecome closer to precious metal

futures. Furthermore, crisis events can enlarge the idiosyncratic volatility

spillovers in commodity markets. The total spillover effects of higher‐moment

risk are stronger than those of lower‐moment risk.

KEYWORDS

contemporaneous network, higher‐moment risk, idiosyncratic information, realized

estimators, relative importance analysis

JEL CLASSIFICATION

C58, F37, G15, G17

1|INTRODUCTION

In the last few decades, commodity futures markets have gotten more and more attention because energy and resources

are becoming more important to the economy and social security. Meanwhile, international commodity

financialization has accelerated (Basak & Pavlova, 2016; I. H. Cheng & Xiong, 2014; D. Zhang & Broadstock, 2020).

In recent years, commodity futures prices have been characterized by a remarkable rise and fall in response to several

J Futures Markets. 2023;43:705–733. wileyonlinelibrary.com/journal/fut © 2023 Wiley Periodicals LLC.

|

705

international financial crises, geopolitical conflicts, and natural disasters (Ma et al., 2021). Interestingly, several

seemingly unrelated commodity futures have become increasingly connected under external shocks (I. H. Cheng &

Xiong, 2014; Ma et al., 2019), implying that the connectedness among commodity futures is mutated and varies over

time. Undoubtedly, the complex and volatile international environment will change the magnitude of the

connectedness in commodity markets. More complex interconnections across commodity futures destabilize the

market and increase the difficulty of portfolio management (Bouri, Lucey, et al., 2021). Hence, accurately identifying

the connectedness among commodity futures and the mechanism by which commodity prices interact is beneficial for

portfolio management and risk management. In this paper, we look at how to find and measure the connections

between commodity futures more accurately.

Spillover connectedness among commodity futures has been extensively discussed in the literature. From the

perspective of riskmeasurement, most of the previous studies use lower‐momentrisk indicators (e.g., Kang et al., 2017;C.

Liu et al., 2021;Songetal.,2021). Related studies have provided evidence of significant lower‐moment risk spillovers

among commodity futures. For example, Kang et al. (2017) document that returns and volatility spillovers in commodity

futures markets are strengthened during crises. From the perspective of the method of constructing spillover networks,

most of the previous studies are based on the variance decomposition of vector autoregressive (VAR) models

(e.g., Diebold et al., 2017;T.Liu&Gong,2020; Naeem et al., 2020;Tiwarietal.,2021). For instance, Diebold et al. (2017)

adopt high‐dimensional vector autoregressions to investigate the spillovers across 19 key commodity futures and conclude

that the energy commodity futures have a stronger connectedness with other commodity futures. However, most of

the VAR‐related literature focuses on noncontemporaneous spillover effects, as network topology analyses are based

on the widely used forecast error variance decomposition (FEVD). In fact, contemporaneous and high‐frequency

information spillovers dominate financial markets. Due to the rapid spread of information in financial markets, risk

spillovers often occur contemporaneously. We may obtain unreliable conclusions using noncontemporaneous

information to measure spillover connectedness. The existing research gaps motivate us to explore a new perspective

for constructing contemporaneous commodity networks based on higher‐moment risk measures.

Our paper contributes to the literature on the network topology methodology. Given the possible estimation error of

risk spillovers caused by noncontemporaneous information, we propose a new network topology approach based on

relative importance analysis to identify the differences between contemporaneous and noncontemporaneous

information spillover networks. This method can effectively compensate for the deficiency of current research on

contemporaneous information spillovers and can be used to investigate financial risk spillovers or the network of

macroeconomic variables. Our paper also contributes to the literature on the spillover connectedness of commodity

futures. First, we focus on higher‐moment risk spillovers and the difference between lower‐and higher‐moment risk

spillovers. Recent literature mainly focuses on the lower‐moment risk spillovers across commodity futures. However,

the higher‐order moment risk estimators (realized skewness and realized kurtosis) can better characterize the market's

downside risk and tail risk (He & Hamori, 2021). As a result, we broaden our connectedness analyses to include higher‐

moment risk spillovers. Second, we concentrate on idiosyncratic components of spillover risk. In an inefficient market,

the fluctuations of commodity futures prices will be disturbed by more noninformation factors, leading to higher

idiosyncratic volatility (Carvalho & Gupta, 2018). Thus, we remove common components and retain idiosyncratic

components, enabling us to investigate the idiosyncratic spillover effects across commodity markets. In addition, we

also carry out some applications based on the time‐varying network estimate results, including the analysis of

influencing factors and the analysis of the connectedness portfolio.

By using high‐frequency data on commodity futures, we attempt to distinguish contemporaneous and

noncontemporaneous idiosyncratic risk spillovers. We first computed the daily realized estimators for 11 commodities

using 5‐min high‐frequency data from September 27, 2009, to December 21, 2021. Second, we employ the generalized

dynamic factor model (GDFM) proposed by Forni et al. (2005) to extract the idiosyncratic components of higher‐

moment realized estimators. Then, we propose a new network topology approach based on relative importance analysis

to identify the differences between contemporaneous and noncontemporaneous idiosyncratic risk spillover networks.

To study how the spillover networks vary during crisis periods, we apply rolling window regression to calculate the

dynamic spillovers. We also construct a Markov regime‐switching model to investigate the determinants of

idiosyncratic risk spillovers under different states. Finally, we examine the performance of the optimal portfolio in

terms of risk management and investment returns using Broadstock et al.'s (2022) minimum connectedness portfolio

(MCoP) approach.

Several findings in this paper are notable. First, the results for idiosyncratic information spillover networks show

that higher‐order moments exhibit extensive spillovers, as they are relatively stronger in transmission strength.

706

|

ZHANG ET AL.