Systemic risk in market microstructure of crude oil and gasoline futures prices: A Hawkes flocking model approach

• Date: 01 February 2020
• DOI: http://doi.org/10.1002/fut.22048
• Published date: 01 February 2020
• Author: Kyungsub Lee,Hyun Jin Jang,Kiseop Lee

J Futures Markets. 2020;40:247–275. wileyonlinelibrary.com/journal/fut © 2019 Wiley Periodicals, Inc.

|

247

Received: 29 July 2019

|

Accepted: 31 July 2019

DOI: 10.1002/fut.22048

RESEARCH ARTICLE

Systemic risk in market microstructure of crude oil and

gasoline futures prices: A Hawkes flocking model approach

Hyun Jin Jang

1

|

Kiseop Lee

2

|

Kyungsub Lee

3

1

School of Business Administration, Ulsan

National Institute of Science and

Technology (UNIST), Ulsan, Republic of

Korea

2

Department of Statistics, Purdue

University, West Lafayette, Indiana

3

Department of Statistics, Yeungnam

University, Gyeongsan, Republic of Korea

Correspondence

Kyungsub Lee, Department of Statistics,

Yeungnam University, Gyeongsan,

Republic of Korea.

Email: ksublee@yu.ac.kr

Funding information

National Research Foundation of Korea,

Grant/Award Number:

2017R1C1B5017338; Korea Institute of

Energy Technology Evaluation and

Planning, Grant/Award Number:

20184010201680

Abstract

We propose the Hawkes flocking model that assesses systemic risk in high‐

frequency processes at the two perspectives—endogeneity and interactivity. We

examine the futures markets of West Texas Intermediate (WTI) crude oil and

gasoline for the past decade, and perform a comparative analysis with

conditional value‐at‐risk as a benchmark measure. In terms of high‐frequency

structure, we derive the empirical findings. The endogenous systemic risk in

WTI was significantly higher than that in gasoline, and the level at which

gasoline affects WTI was constantly higher than that in the opposite case.

Moreover, although the relative influence’s degree was asymmetric, its

difference has gradually reduced.

KEYWORDS

branching ratio, calibration, conditional value‐at‐risk, flocking, gasoline futures, Hawkes process,

systemic risk, West Texas Intermediate crude oil futures

JEL CLASSIFICATION

C13; G13

1

|

INTRODUCTION

Over the past two decades, the systemic risk level has increased in financial markets due to the growth of securitization,

hedge fund markets, and increase in intraday trading. Recently, the emergence of innovative technologies has

accelerated the paradigm shift of trading activities in financial markets. Traditional trading platforms, such as phone

conversations or clicks on a screen by humans, have moved to automated trading by computers based on the ultralow

latency electronic system. The increased trading speed enables execution of orders within microseconds by the use of

sophisticated algorithms; this is called high‐frequency trading. According to a report of the Commodity Futures Trading

Commission (CFTC),

1

the volume of high‐frequency trading in futures markets has grown remarkably over the past

decade. It accounts for 80% of foreign exchange futures, 67% of interest rate futures, 62% of equity futures, and 47% of

metals and energy futures trading volume. In addition, flash crash events frequently occur in security markets, which

are attributed to high‐frequency trading.

2

Such an environmental change in trading potentially allows large price

movements within a short period of time as well as the rapid risk propagation to different assets, as mentioned in Miller

and Shorter (2016).

1

CFTC, “Remarks of Chairman Timothy Massad before the Conference on the Evolving Structure of the US Treasury Market,”October 21, 2015, at http://www.cftc.gov/PressRoom/

SpeechesTestimony/opamassad‐30.

2

For example, the Dow Jones Industrial Average (DJIA) index plunged roughly 1,100 points in the first 5 min of trading on August 24, 2015.

In this context of high‐frequency finance, we develop a novel Hawkes process‐based model to examine the level of

systemic risk that exists within and between price dynamics at the microscopic level. The proposed model allows

capturing contagious and clustered phenomena that can be investigated in the excessive volatile and correlated

markets. Studies related to systemic risk in high‐frequency trading are discussed under various aspects. Filimonov and

Sornette (2012) conduct event studies to investigate the changes in the systemic risk before and after the

announcements of two extreme events: downgrading of Greece/Portugal and the flash crash event for the E‐mini S&P

500 futures in 2010. Hardiman, Bercot, and Bouchaud (2013) perform a similar analysis with Filimonov and Sornette

(2012) by taking power‐law kernels. Chavez‐Demoulin and McGill (2012) compute intraday value‐at‐risk (VaR) for

stocks in New York Stock Exchange (NYSE) using a peak‐over‐threshold model, and Jain, Jain, and McInish (2016)

assess the extent to which a high‐frequency system increases systemic risk in the Tokyo Stock Exchange. Bormetti et al.

(2015) use a multivariate Hawkes process with a common factor that controls a large number of jumps in the

transaction movement. Calcagnile, Bormetti, Treccani, Marmi, and Lillo (2018) compute the number of cojumps

occurring in Russell 3000 index stocks to measure the frequency of the collective instability at high frequency.

On the other hand, there is little discussion on the increased systemic risk in energy markets associated with high‐

frequency trading. However, energy futures markets are no longer exceptional on this matter. As noted in the

beginning, almost the half of the trading volume in the energy markets is raised from high‐frequency trading. Moreover,

the CFTC examined how frequently flash events have occurred in the top‐five most active futures contracts in 2015, that

is, corn, gold, West Texas Intermediate (WTI) crude oil, E‐mini S&P 500 futures, and Euro FX.

3

Among them,

surprisingly, more than 35 similar intraday flash events have occurred just for WTI crude oil futures.

4

This result

implies that WTI crude oil futures are utilized actively as instruments of algorithmic trading strategies.

In this study, we attempt to discover the empirical evidence of the systemic risk level in the dynamics of the two futures

prices of WTI crude oil and gasoline observed at the intraday transaction level over the past decade. The gasoline futures

contracts are being traded most actively in the New York Mercantile Exchange (NYMEX) in the energy sector, following the

WTIcrudeoilfutures.Weconsidertwokindsofdefinitionsfor systemic risk in the market microstructure with the

instability perspective. The first view is the degree of instabilitythatexistswithinapriceprocess.Itisregardedastheterm

endogeneity, which is introduced in the earlier literature (e.g., Danielsson, Shin, & Zigrand, 2012; Filimonov & Sornette,

2012; Hardiman et al., 2013

5

). By estimating this level, we examine whether the trend of price decline leads to additional

price decreases (or price rebounds). The second view is the degree of instability that exists between price processes caused by

interaction between two different markets. In that point of view, we investigate how the change in one price affects to the

change in the other price, and vice versa. In addition, we observe how micromovements of prices in the two markets are

likely to close to each other when the price difference widens or narrows.

Meanwhile, WTI crude oil and gasoline futures prices have maintained a strong dependence for a long time

(EIA, 2014). From a macroeconomic perspective, the main causes of the price difference between crude oil and gasoline

are refining costs and supply/demand balance of each product. Such comovement has been studied in terms of market

co‐integration in econometrics or flocking behavior. When the two markets are co‐integrated or have a flocking feature,

the associated prices are closely correlated. Furthermore, one price could lead the other, while the reverse also occurs

from time to time, or all prices in a system could follow the same behavior.

Co‐integration refers to two or more nonstationary time series that are driven by one or more common nonstationary

time series, proposed in the seminal works by Granger (1981) and Engle and Granger (1987). Many financial data series

are known to exhibit the co‐integration, for example, international stock markets (Cerchi & Havenner, 1988; Duan &

Pliska, 2004; Taylor & Tonks, 1989), foreign exchange rates (Baillie & Bollerslev, 1989; Kellard, Dunis, & Sarantis, 2010),

futures and spot prices (Maslyuka & Smyth, 2009; Ng & Pirrong, 1994, 1996), and especially, crude oil, gasoline, and

heating oil futures prices (Chiu, Wong, & Zhao, 2015; Serletis, 1992). As a similar manner, flocking is known to the

collective motion of a large number of self‐propelled entities. Reynolds (1987) first proposed the breaking‐through

algorithm that makes it feasible to generate realistic computer simulation of flocking agents. The flocking behavior

appears in many contexts of physics, biology, engineering, and human systems, including financial markets (Fang, Sun,

& Spiliopoulos, 2017; Ha, Kim, & Lee, 2015; Huepe & Aldana, 2008; Rauch, Millonas, & Chialvo, 1995; among many

others). Even though comovement propensity in two or more dynamics has been discussed with different terms of

3

In this study, the flash crash is defined by the episodes in which a contract price moved at least 2% within an hour, but returned to within 0.75% of the original or starting price within that same hour.

4

https://www.cftc.gov/sites/default/files/idc/groups/public/@newsroom/documents/file/hourlyflashevents102115.pdf

5

In Filimonov and Sornette (2012) and Hardiman et al. (2013), this is referred to as “reflexivity”instead of endogeneity.

248

|

JANG ET AL.