You are on page 1of 6

Available online at www.sciencedirect.

com

ScienceDirect
Energy Procedia 104 (2016) 462 – 467

CUE2016-Applied Energy Symposium and Forum 2016: Low carbon cities & urban
energy systems

Analyzing the risk-return relationship in crude


oil futures market using high-frequency data
Xu Gonga, Fenghua Wena, Bin Panb, Xiaohua Xiac,d *
a
School of Business, Central South University, Changsha, Hunan 410083, China
b
School of Finance, Wenzhou University, Wenzhou, 325035, China
c
Institute of China’s Economic Reform and Development, Renmin University of China, Beijing 100872, China
d
School of Economics, Renmin University of China, Beijing 100872, China

Abstract

We comprehensively examine the contemporaneous/intertemporal risk-return relationship in the crude oil futures
market. Our empirical results, based on high-frequency transaction data, suggest the contemporaneous relation
between risk (volatility risk, downside risk or jump risk) and return in the crude oil futures market is negative and
statistically significant, and the contemporaneous negative relation between downside risk and return is stronger than
two others. However, the intertemporal volatility/jump risk-return relationship is insignificant, and there is weak
negative correlation between downside risk and excepted return in the crude oil futures market.
© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
© 2016 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Selection and/or
Peer-review peer-reviewofunder
under responsibility responsibility
the scientific of of
committee CUEthe Applied Energy Symposium and Forum, CUE2016: Low carbon
cities and urban energy systems.
Keywords: Risk-return relationship; Volatility risk; Downside risk; Jump risk; High-frequency data

1. Introduction

The crude oil market is an indispensable part of the economic system [1]. The crude oil is the
foundation of a nation’s economic development [2][3]. Thus, analyzing the crude oil futures has attracted
considerable attention from academics, governments and investors.
Among the various research topics on the crude oil futures, estimating the risk-return relationship in
crude oil futures market is of special interest for energy researchers. However, the researchers’ empirical
evidence is mixed. Some researchers found that the relation between risk and return in crude
oil futures market was positive. Considine and Larson [4] applied a stochastic model to test the existence

* Corresponding author. Tel.: +86-15210024348; fax: +86-10-82500256.


E-mail address: xiaxh.email@gmail.com

1876-6102 © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the Applied Energy Symposium and Forum, CUE2016: Low carbon cities
and urban energy systems.
doi:10.1016/j.egypro.2016.12.078
Xu Gong et al. / Energy Procedia 104 (2016) 462 – 467 463

of risk premia on crude oil and natural gas. Their empirical results provide rather strong support for the
existence of risk premia on crude oil market and natural gas. Recently, Cifarelli and Paladino [5] used a
univariate GARCH(1,1)-M model to estimate the volatility risk premium. The evidence suggested there
were positive feedback trading and positive volatility risk premium in the oil market. However, a number
of studies support the contention that the risk premia is negative. Trolle and Schwartz [6] studied variance
risk premia in crude oil and natural gas market by using a robust model-independent approach. Their
empirical results indicated the average variance risk premia were significant negative for crude oil market.
Li et al. [7] found a intertemporal negative relation between return on the price of oil futures and volatility
components. Kristoufek [8] also found the correlation between returns and volatility risk of both Brent
and WTI crude oils were negative.
The research results from above studies are inconsistent, and accurately estimating the risk-return
relationship in crude oil futures market becomes a challenging work. In this paper, we will
comprehensively analyze the relationship between contemporaneous/ intertemporal risk and return in the
crude oil futures market. Compared with the existing literature, our study has the following advantages.
Firstly, our research is more comprehensive. We examine both the contemporaneous risk-return
relationship and intertemporal risk-return relationship in the crude oil futures market. Secondly, the
existing studies focus mainly on the correlation between volatility risk and return of the crude oil futures.
However, we not only estimate the volatility risk-return relationship but also investigate the downside/jump
risk-return relationship in the crude oil futures market. Thirdly, we use the high-frequency transaction
data to measure the volatility, downside and jump risks of crude oil futures. The high-frequency
transaction data contains far more information than the low-frequency transaction data, which is able to
more accurately measure the risks. Thus, our empirical results are more reliable.
The remainder of this paper is organized as follows. In the next section, we measure the volatility,
downside and jump risks. Section 3 describes the data. In Section 4, we estimate the relationship between
contemporaneous risk and return in crude oil futures market using high-frequency transaction data.
Section 5 analyzes the relationship between intertemporal risk and return in crude oil futures market
through high-frequency data. Section 6 provides the conclusions.

2. Alternative Risk Measures

2.1 Volatility Risk

The volatility risk in financial market cannot be observed, and it needs to use a method to measure. In
this paper, we choose the realized volatility ([9], [10] and [11]) to measure the volatility risk of the crude
oil futures. The daily realized volatility can be written as
M
RVt cd = ¦ rt 2c,i (1)
i=1
where rt c,i is the ith return (i=1, " , M) in day t c . Pt c,i is the ith closing price in day t c .
In Corsi [12], he used the average realized volatility between day t c and t c  H (where H is the
number of days in a month) to measure the monthly realized volatility. Following [12], the monthly
volatility risk VRt is defined as
RVt cd,1  RVt cd,2  "  RVt cd, H
VRt =RVt m (2)
H

2.2 Downside Risk


464 Xu Gong et al. / Energy Procedia 104 (2016) 462 – 467

In this paper, we applied the downside realised semivariance ([13]) as a proxy of downside risk in the
crude oil futures market. Referring to [13], the downside realised semivariance can be expressed as
M
RSVt c ¦r
j =1
2
t c, j I (rt c, j d 0) (3)

where I (˜) is the indicator function taking the value 1 if the argument I is true.
In this paper, we use the monthly downside realised semivariance to measure the monthly downside
risk. So we get the expression of monthly downside risk
RSVt c,1( d )  RSVt c,2( d )  "  RSVt c, (Hd )
DRt =RSVt  ( m ) (4)
H

2.3 Jump Risk

The daly jump risk J tdc can be defined by


J tdc I ( Z t c ! ID )( RVt c  RBVt c ) (5)
where I (˜) is a indicator function; Z t c is Z-statistics (Huang and Tauchen [14]); D is chosen 0.99;
RBVt is the realized bipower variation.
In this paper, we use the monthly discontinuous jump variation to measure the monthly jump risk of
crude oil futures market. Similar to Eq. (2) and Eq. (4), the monthly jump risk can be written as
(m)
J tdc,1  J tdc,2  "  J tdc, H
JRt =J t
(6)
H

3. Data description

This paper uses 5-minute high-frequency transaction data from the NYMEX-CME for the front-month
WTI crude oil futures contract. The full sample period is from January 1998 to April 2014, which
contains 196 monthly observations.
The resulting summary statistics reported in Table 1. The table shows the return of crude oil futures is
negative skewness and fat tail, and the volatility, downside and jump risks all have the property of
positive skewness and fat tail. In addition, the Ljung–Box Q–statistics reported in the table show the
volatility, downside and jump risks of crude oil futures indicate significant dependencies. According to
the t–statistics, we find all variables refuse the null hypothesis that it has unit root.
Table 1. Summary statistics for all variables.
Mean Std.Dev. Skewness Kurtosis Q(5) Q(10) Q(15) Q(20) t-statistic
Rt 0.8618 9.6718 -0.5114 4.6432 4.5950 13.845 31.709*** 35.016** 12.724***
VRt 4.9353 4.6107 3.7780 22.271 286.39*** 292.45*** 292.56*** 293.45*** 4.7833***
DRt 2.4966 2.1937 2.8440 13.076 276.00*** 285.57*** 287.19*** 289.39*** 4.7900***
JRt 0.6910 1.2687 5.4728 43.949 52.684*** 54.863*** 58.248*** 66.080*** 9.3837***

4. Contemporaneous relation between risk and return

4.1 Econometric model


Xu Gong et al. / Energy Procedia 104 (2016) 462 – 467 465

We investigate the contemporaneous risk-return relationship in the crude oil futures market. The
econometric model that we analyze the contemporaneous risk-return relationship at the monthly
frequency is written:
Rt =D +E X t  H t (7)
where Rt is the monthly return of crude oil futures. X t represents VRt , DRt or JRt . VRt is the
monthly volatility risk as defined in Eq.(2). DRt is the monthly downside risk as defined in Eq.(4). And
JRt is the monthly jump risk as defined in Eq.(6).

4.2 Empirical results

Table 2 presents the parameter estimates of Eq.(7). For volatility risk, E is statistically significant
negative, which shows the return of crude oil futures is lower as the volatility risk level for the market
increases. For downside risk, we can see E is statistically significant negative. The result indicates that
the contemporaneous downside risk-return relationship is negative in the crude oil futures market.
Similarly, the result of Column 6 implies the contemporaneous relation between jump risk and return of
crude oil futures is negative and statistically significant. In addition, comparing the coefficient and t-
statistic of all E , we can find the contemporaneous negative relation between downside risk and return is
stronger than two others.
Table 2. Estimated results for the contemporaneous risk-return relationship.
Volatility Risk Downside Risk Jump Risk
Coefficient t-Statistic Coefficient t-Statistic Coefficient t-Statistic
D 3.9182*** 4.0365 5.1466*** 5.3171 1.7828** 2.2945
E -0.6193*** -4.3039 -1.7163*** -5.8863 -1.3329** -2.4734
Adj.R2 0.0825 0.1472 0.0256
Note: Asterisks indicate statistical significance at the 10% (*), 5% (**) or 1% (***) level.

5. Intertemporal relation between risk and return

5.1 Econometric models

To investigate the intertemporal risk-return relationship, we propose two following econometric


models.
Rt =D +E X t 1  H t (8)
Rt =D +E Et ( X t )  H t (9)
where X t 1 is the lagged monthly volatility risk VRt 1 , lagged monthly downside risk DRt 1 or
lagged monthly jump risk JRt 1 . Et ( X t ) denotes the predicted value of volatility risk, downside risk or
jump risk at time t. The predicted value is gotten by using the AR(1), AR(3) or HAR(3) model.
The AR(1), AR(3) and HAR(3) models are written:
X t =a  bX t 1  H t (10)
X t =a  bX t 1  cX t  2  dX t 3  H t (11)
d d d d
Y
t c  21 =a  bY  H tc
t c 1  cY
t c5  dYt c  22 (12)
d
where Y the average volatility/downside/jump risk between day t c and t c  21 . If t c is the first
t c 21 is
d
day of the t month, Yt c 21 approximately equal to the risk X t as computed in Eq.(2), Eq.(4) or Eq.(6) at
th
d
time t. We extract all Yt c 21 at the first day of all months, and get all predicted values of monthly risks. In
466 Xu Gong et al. / Energy Procedia 104 (2016) 462 – 467

d d d d d
addition, Yt c1 is the daily risk; Yt c5 (Yt c1  Yt c 2  "  Yt c5 ) / 5 is the weekly risk; and
Yt cd 22 (Yt cd1  Yt cd 2  "  Yt cd 22 ) / 22 is the monthly risk.

5.2 Empirical results

Table 3 reports the results for the intertemporal relation between risk and return of the crude oil futures.
The table indicates all E are negative, but they are non-significant. The results suggest the intertemporal
risk-return relationship is weak in the crude oil futures market.
Table 4 lists the results for the intertemporal risk-return relationship as defined in Eq.(9). The
predicted values of risk Et ( X t ) are gotten by using the rolling window prediction method. In this section,
five years are used as the rolling window length. That is to say, the rolling window length of AR(1) and
AR(3) models is 60, and the HAR(3) model is 1296. In Columns 3 and 7 of Table 4, all E are not
significant, which shows both the intertemporal volatility risk-return relationship and the intertemporal
jump risk-return relationship is weak. In Column 5, E in the AR(3) and HAR(3) models are negative and
statistically significant. The results indicate there is negative intertemporal relation between downside risk
and expected return in the crude oil futures market.
Table 3. Estimated results for the intertemporal risk-return relationship as defined in Eq.(8).
Volatility Risk Downside Risk Jump Risk
Coefficient t-Statistic Coefficient t-Statistic Coefficient t-Statistic
D 1.5128 1.4816 1.8531* 1.7578 0.9133 1.1513
E -0.1271 -0.8417 -0.3869 -1.2215 -0.0438 -0.0798
Adj.R2 -0.0015 0.0025 -0.0051

Table 4. Estimated results for the intertemporal risk-return relationship as defined in Eq.(9).
Volatility Risk Downside Risk Jump Risk
Coefficient t-Statistic Coefficient t-Statistic Coefficient t-Statistic
D 1.5422 1.4867 2.1477 1.9066 0.6798 0.8196
AR(1) E -0.1367 -1.0244 -0.5208 -1.6033 0.2174 0.7079
Adj.R2 0.0004 0.0115 -0.0037
D 1.5448 1.5156 2.1696** 1.9837 0.7308 0.8931
AR(3) E -0.1409 -1.0729 -0.5525* -1.7271 0.1508 0.6043
Adj.R2 0.0011 0.0145 -0.0047
D 1.7721* 1.6981 2.3797** 2.0948 0.5457 0.5907
HAR(3) E -0.1864 -1.3424 -0.6401* -1.8549 0.4553 0.6494
Adj.R2 0.0059 0.0178 -0.0043

6. Conclusion

We comprehensively investigate the contemporaneous risk-return relationship and intertemporal risk-


return relationship in the crude oil futures market. In this paper, we use the realized volatility as a
measure of volatility risk, the downside realized semivariance as a measure of downside risk, and the
discontinuous jump variation as a measure of jump risk. Then, we use the simple linear regression model
to examine the relation between volatility/downside/jump risk and return by applying high-frequency
transaction data from the NYMEX-CME for the front-month WTI crude oil futures contract.
We find the contemporaneous relation between risk (volatility risk, downside risk or jump risk) and
return in the crude oil futures market is statistically significant negative. And the contemporaneous
negative relation between downside risk and return is stronger than two other. However, the intertemporal
volatility/jump risk-return relationship is insignificant, and there is a little negative correlation between
Xu Gong et al. / Energy Procedia 104 (2016) 462 – 467 467

downside risk and excepted return in the crude oil futures market. Our empirical results also suggest the
failure of both contemporaneous and intertemporal risk-return tradeoff in the crude oil futures market.
In future work we plan to investigate the risk-return relationship in other energy markets through the
high-frequency transaction data. It will also be interesting to analyze the dynamic relationship between
risk and return and the economic underpinnings of the negative risk premium in the crude oil futures
market.

Acknowledgements

The financial support of the National Natural Science Foundation of China (Nos. 71371195, 71431008,
71203224).

References:

[1] Ji Q, Fan Y. Dynamic integration of world oil prices: A reinvestigation of globalisation vs.
regionalisation. Appl Energ. 2015;155:171-180.
[2] Wang L, An H, Liu X, Huang X. Selecting dynamic moving average trading rules in the crude oil
futures market using a genetic approach. Appl Energ. 2016;162:1608-1618.
[3] Zhang Y, Zhang L. Interpreting the crude oil price movements: Evidence from the Markov regime
switching model. Appl Energ. 2015;143:96-109.
[4] Considine TJ, Larson DF. Risk premiums on inventory assets: The case of crude oil and natural gas.
Journal of Futures Markets. 2001;21:109-126.
[5] Cifarelli G, Paladino G. Oil price dynamics and speculation. Energy Economics. 2010;32:363-372.
[6] Trolle AB, Schwartz ES. Variance risk premia in energy commodities. The Journal of Derivatives.
2010;17:15-32.
[7] Li Z, Sun J, Wang S. An information diffusion-based model of oil futures price. Energy Economics.
2013;36:518-525.
[8] Kristoufek L. Leverage effect in energy futures. Energy Economics. 2014;45:1-9.
[9] Andersen TG, Bollerslev T. Answering the Skeptics: Yes, ARCH models do provide good volatility
forecasts. International Economic Review. 1998;4:885-905.
[10] Huang C, Gong X, Chen X, Wen F. Measuring and forecasting volatility in chinese stock market
using HAR-CJ-M model. Abstr Appl Anal. 2013;2013:1-13.
[11] Gong X, He Z, Li P, Zhu N. Forecasting return volatility of the CSI 300 index using the stochastic
volatility model with continuous volatility and jumps. Discrete Dyn Nat Soc. 2014;2014:1-10.
[12] Corsi F. A simple approximate long-memory model of realized volatility. Journal of Financial
Econometrics. 2009;7:174-196.
[13] Shephard N, Kinnebrock S, Barndorff-Neilsen O E. Measuring downside risk-realised semivariance.
Department of Economics, University of Oxford, 2008.
[14] Huang X, Tauchen G. The relative contribution of jumps to total price variance. Journal of Financial
Econometrics. 2005;3:456-499.