- Email: [email protected]
Impact of oil price uncertainty on Middle East and African stock markets
Energy 123 (2017) 189e197
Contents lists available at ScienceDirect
Energy journal homepage: www.elsevier.com/locate/energy
Impact of oil price uncertainty on Middle East and African stock markets Anupam Dutta*, Jussi Nikkinen, Timo Rothovius Department of Accounting & Finance, University of Vaasa, P.O. Box 700, FI-65101 Vaasa, Finland
a r t i c l e i n f o
a b s t r a c t
Article history: Received 14 March 2016 Received in revised form 15 August 2016 Accepted 24 January 2017 Available online 30 January 2017
This paper investigates whether the implied crude oil volatility index (OVX), a forward-looking measure of oil market uncertainty published by Chicago Board Options Exchange (CBOE), impacts the realized volatility of Middle East and African stock markets. Using an extended version of the GARCH model, we show that the oil market uncertainty has substantial effects on the realized volatility of most of the markets under study. Our findings also reveal that, even after controlling for the effect of the implied volatility index of S & P 500 (VIX), the impact of the OVX on the Middle East and African equity markets still holds for almost half of the markets considered. Additionally, the application of the GARCH-jump model shows that stock returns of majority of the sampled markets are sensitive to the fluctuations occurring in the implied oil volatility index and that time-varying jumps do exist in the stock returns. Thus, the market participants' anticipation of the future oil market uncertainty is an important factor explaining the returns and volatilities of the Middle East and African equity markets. © 2017 Elsevier Ltd. All rights reserved.
Keywords: Oil price uncertainty VIX OVX Middle East Africa GARCH-jump model
1. Introduction The recent downturn in crude oil price is substantially affecting the economies all over the world, oil being one of the major production factors in most countries. In oil-depending economies, changes in oil price and volatility can be expected to influence also their equity markets. For example, Ciner [15] argues that oil price shocks can cause changes in expected cash flows, by affecting the overall economy as well as the discount rate used to value equities by changing inflationary expectations. Consequently, large oil price movements tend to increase uncertainty of future equity prices as evidenced, for example, by Bernanke [5] and Pindyck [31]. In this paper, our objective is to examine whether the implied crude oil volatility index (OVX), a forward-looking measure of oil market uncertainty published by Chicago Board Options Exchange (CBOE), impacts the volatility of Middle East and African stock markets. Over the past years, the relation between oil price changes and stock returns has received an increasing attention in the energy and finance literature. Accordingly, numerous empirical studies have investigated the return and volatility transmission mechanism between oil price and stock returns. Notable contributions include
* Corresponding author. E-mail address: adutta@uwasa.fi (A. Dutta). http://dx.doi.org/10.1016/j.energy.2017.01.126 0360-5442/© 2017 Elsevier Ltd. All rights reserved.
e.g. Sadorsky [32], Papapetrou [30], Killian and Park [22], Malik and Ewing [26], Lee and Chiou [24], Arouri, Jouini and Nguyen [1,2] and Bouri [9,10]. Collectively, the above articles show that oil and stock markets are associated with each other. A unifying feature of these articles is that they use crude oil price series to investigate the volatility. Our study contributes to the existing literature in two major ways. First, our paper models realized volatilities of emerging stock markets, taking into account the global anticipation of future oil price uncertainty, measured using the implied crude oil volatility index, OVX. This is an advantageous approach, since the implied oil volatility index, which is derived from option prices, is generally considered to be a good indicator of oil market uncertainty (see, e.g., Liu, Ji and Fan [25]). According to Liu et al. [25], this is because implied volatilities not only contain historical volatility information, but also investors' expectation of future market conditions. r and Westgaard [20] find Additionally, Haugom, Langeland, Molna that the day-ahead and week-ahead volatility forecasts can be significantly improved by including information from the OVX. Given the useful properties of the OVX, our paper provides a novel extension to papers such as Sadorsky [32], Papapetrou [30], Killian and Park [22], Malik and Ewing [26], Lee and Chiou [24], Arouri et al. [1,2] and Bouri [9,10] that have used the crude oil price series to model volatility. Second, many recent studies have used implied volatility
190
A. Dutta et al. / Energy 123 (2017) 189e197
indices, such as VIX, to assess the uncertainty transmission among € m [28]; Wang [33]; the financial markets (Nikkinen and Sahlstro Bollerslev, Osterrieder, Sizova and Tauchen [8]; Liu et al. [25]). Of these studies, Liu et al. [25], for instance, find that the US stock market uncertainty, measured by VIX, affects oil price uncertainty measured by using OVX. Nonetheless, OVX has received very little attention in the literature which is very surprising for the oil market considering its economic importance (Haugom et al. [20]). Thus, our paper contributes also to the uncertainty transmission literature by examining whether global oil market uncertainty affects the realized volatilities of Middle East and African equity markets. In our empirical approach, we apply a modified form of the generalized autoregressive conditional heteroskedasticity (GARCH) model to examine the relationship between oil market uncertainty and the volatility of different equity markets under study.1 Our findings based on the extended model indicate significant links between oil price uncertainty and the realized stock market volatility. We also combine OVX with VIX to observe their joint effects on the selected markets and document that, association between oil price uncertainty and realized stock market volatility mostly hold even after controlling for the effect of VIX. Specifically, after inserting VIX as a regressor in the GARCH process, oil market uncertainty evidently affects the variance of nearly half of the markets considered. We further report that, conditioned on the past information, the volatility in each market is very persistent. This latter outcome suggests that the current volatility of stock returns for a specific market is directly affected by its past volatility. Additionally, we consider the application of the GARCH-Jump mode to assess whether variations in the oil volatility index have any significant influences over the stock returns. The results suggest that such associations hold for majority of the sampled markets and that jumps exist in the returns which are time-varying. The results of the present study are useful for energy policy makers, investors and researchers in several aspects. For example, the findings can be used for framing sound asset pricing models and making global asset allocation decisions. They are also beneficial for the market participants in understanding the interaction of the stock markets of Middle East and African countries in relation to the crude oil market, as well as the global equity market uncertainties. In addition, our research could receive special attention from those investors who make use of new financial tools to hedge oil price volatility risk and are potentially interested in futures and option trading on OVX (Liu et al. [25]). The article proceeds along the following lines. The next section reviews the relevant literature. Data and their properties are presented in the third section. Section 4 outlines the research methodology we consider in our empirical investigation. In section 5, we discuss our findings. The last section contains conclusions. 2. Literature review A growing body of literature has examined the association between oil and stock markets. However, the findings of this literature are somewhat mixed. For example, Kling [23] and Papapetrou [30] find a negative relation between oil price changes and stock returns. Chen, Roll and Ross [12], on the other hand, report that stock returns are not affected by oil price fluctuations. More recently, however, Basher and Sadorsky [4] and Choi and Hammoudeh [14] identify a positive relation between oil price changes and stock returns. While the relationship between oil prices and stock returns has
1 Thus, we follow the approach, for example, by Blair et al. [6] and Kambouroudis and McMillan [21] that use VIX as a regressor in the GARCH variance equation.
been extensively analyzed, a strand of literature also examines the persistence and transmission of volatility from oil markets to stock markets. Malik and Hammoudeh [27] employ BEKKeGARCH(1,1) model to study the volatility and shock transmission mechanism among the U.S. equity market, the global crude oil market, and three Gulf equity markets that include Bahrain, Kuwait, and Saudi Arabia. In their empirical analyses, they document that Gulf equity markets are the receivers of volatility from the oil market. In addition, they report a significant volatility transmission from the stock market to the oil market only in Saudi Arabia. By applying an autoregressive conditional jump intensity model with structure changes, Chiou and Lee [13] report the existence of a negative and statistically significant impact of oil prices on stock returns. They also detect that the asymmetric effect has statistical significance only in a high-fluctuation state for both spot and futures oil price contracts. The authors further find that with changes in oil price dynamics, oil price volatility shocks have asymmetric effects on stock returns. Fowowe [18] also employs the GARCHjump model to investigate the relationship between oil prices and returns on the Nigerian Stock Exchange. The study reports a negative but insignificant effect of oil prices on stock returns in Nigeria. More recently, Ewing and Malik [17] use the univariate and bivariate GARCH models to examine the volatility of oil and US stock market prices incorporating structural breaks using daily data from July 1, 1996 to June 30, 2013. The authors do not report any evidence of volatility spillover between oil prices and US stock market when structural breaks in variance are ignored in the model. However, after accounting for structural breaks in the model, they report strong volatility association between the two markets. Bouri [9] has inspected the return and volatility linkages between oil prices and the Lebanese stock market by applying the VAR-GARCH model to weekly data from 30 January 1998 to 30 May 2014. The author finds some unidirectional return and volatility transmissions from oil prices to the Lebanese stock market. He also adds that the interrelationship between oil prices and Lebanese stocks increase during the crisis and tends to reduce significantly in the post-crisis period. Additionally, few other studies use sector indices to investigate the volatility transmission between oil and stock prices. Malik and Ewing [26], for example, estimate bivariate BEKKeGARCH (1,1) models to examine the volatility transmission between oil prices and five US sector indices. The sectors considered in their study include Financials, Industrials, Consumer Services, Health Care, and Technology. The paper finds that there exists significant transmission of shocks and volatility between oil prices and different stock market sectors. Moreover, Arouri et al. [2] make use of European equity sector indices to assess the volatility spillovers between oil and stock prices. The authors show that the volatility transmission between oil price and sector stock returns is significant. 3. Data In this paper, unlike the ones described in previous chapter, we use implied crude oil volatility index, OVX, instead of conventional oil price index, to estimate the effect of oil price uncertainty on the stock market volatilities. This is advantageous, because implied volatilities not only contain historical volatility information, but also investors' expectation of future market conditions. The CBOE publishes OVX index, from the middle of 2007, as a measure of expected 30-day volatility of crude oil prices. The OVX considers real-time bid/ask quotes of nearby and second nearby options with at least 8 days to expiration, and weights these
A. Dutta et al. / Energy 123 (2017) 189e197
options to derive a constant, a 30-day estimate of the expected volatility (Liu et al. [25]). Our daily data on stock prices consist of five Middle East and seven African stock market indices. The markets of Saudi Arabia, United Arab Emirates (UAE), Kuwait, Qatar and Jordan from Middle East, and Nigeria, Egypt, South Africa, Tunisia, Morocco, Kenya and Mauritius from Africa, presented in the order of their oil production, are included in the sample. We use Morgan Stanley Capital International (MSCI) country indices along with the information on OVX and VIX, which are all extracted from the Data Stream database. Our sample starts in May 10, 2007 and ends December 31, 2014, providing a total of 1995 observations. We choose this period because the OVX data are available from the commencing date of our sample. We use log returns which are measured in US dollars, so that we can conveniently compare them across the countries considered in our analyses.
191
Fig. 1 demonstrates the OVX and VIX indices for the whole sample period. The graph indicates several major spikes in both indices during the period under study. The first one appears in the latter part of 2008, when Lehman Brothers filed a bankruptcy, and both volatility indices increase rapidly. Interestingly, the hikes seem to be a consequence of either economic or political events. For example, the spike occurring in OVX during the beginning of 2011 can be attributed to the Libyan war for which the oil price uncertainty increases markedly (Liu et al. [25]). 4. Methodology 4.1. Modified GARCH (1,1) model The first step in our GARCH methodology is the specification of the mean equation. For each of the return series, the mean equation assumes the following form:
Rit ¼ pi þ wi Rit1 þ εit
3.1. Descriptive statistics Table 1, Panel A reports the descriptive statistics for each of the return series considered. The performance of these indices, measured by the daily mean returns, is highest in the Qatar stock market followed by stock markets in Kenya, Mauritius, Saudi Arabia, South Africa and Tunisia, during the sample period. We find negative mean returns for the remaining countries. The equity indices for South Africa, United Arab Emirates and Egypt show higher volatility than the rest of the indices. Most of the indices exhibit negative skewness, implying that large negative stock returns are more common than large positive returns in those markets, but the indices for Kuwait, Tunisia, Kenya and Mauritius are positively skewed. In terms of kurtosis, all the return series are leptokurtic, which is a common phenomenon for financial time series data, hence the GARCH process, proposed by Bollerslev [7], is appropriate (Malik and Hammoudeh [27]). Moreover, the Jarque-Bera test rejects the null hypothesis that all the return series are normally distributed. Panel B of Table 1 displays the descriptive statistics of the OVX and VIX indices. The standard deviations of these two indices indicate that the fluctuations in OVX are more severe than in VIX. In addition, both these indices are positively skewed. The results further show that the indices have kurtosis higher than 3, implying that each volatility series have leptokurtic distribution with asymmetric tails.
(1)
Within this framework, Rit denotes the log return on index i between time t and t-1, pi is a long-term drift coefficient and εit refers to the error term for the return on series i at time t. The GARCH model, proposed by Engle [16] and Bollerslev [7], formulates the conditional variance, ht , as a function of past errors and the lagged conditional variances leaving the unconditional variance constant. In order to examine effect of oil price uncertainty, measured by OVX, on the volatility of different equity markets considered, we propose to estimate the following extended GARCH variance equation: 2 ht ¼ u þ aε2t1 þ bht1 þ gOVXt1
(2)
where u > 0, a 0, b 0 and g 0 to guarantee the positivity of ht and εt ¼ zt ht ; zt i.i.d. (0,1). For a nonzero value of g, we report that the volatility of the corresponding equity market is significantly affected by the crude oil price uncertainty. A number of articles (e.g. Blair, Poon and Taylor [6]; Kambouroudis and McMillan [21] and others) have used implied volatility index as a regressor in the GARCH variance equation. Blair et al. [6], for instance, report that the VIX index provides the most accurate projections of conditional variance for all forecast horizons and performance measures considered in their analyses. Kambouroudis and McMillan [21] also argue that the inclusion of VIX as a regressor in the GARCH variance equation improves the overall performance
Table 1 Descriptive statistics of the stock returns. Countries Panel A: Return Indices Saudi Arabia United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius Panel B: Volatility Indices OVX VIX
Mean
Standard deviation
Skewness
Kurtosis
JarqueeBera
0.000074 0.000026 0.000341 0.000282 0.000442 0.000260 0.000006 0.000032 0.000033 0.000292 0.000250 0.000149
0.0149 0.0194 0.0138 0.0148 0.0119 0.0114 0.0102 0.0177 0.0198 0.0145 0.0129 0.01225
1.13 0.56 0.10 0.72 0.50 0.19 0.15 0.96 0.21 0.31 0.10 0.05
21.58 16.76 15.75 19.30 12.38 6.97 10.16 12.09 7.90 6.83 14.32 16.54
29120.79*** 15847.49*** 13918.86*** 22266.59*** 7396.37*** 1323.13*** 4270.12*** 7189.19*** 2012.04*** 1256.43*** 10659.93*** 15254.48***
35.778140 22.227120
14.50000 10.32000
1.614684 2.115989
8.62 6.32
4120.132*** 1783.254***
Notes: This table reports the main descriptive statistics of the daily return indices used. Panel A contains the results for the return indices, while Panel B shows the outcomes for the volatility indices. Jarque-Bera test rejects the null hypothesis of normality at 1% level of significance.
192
A. Dutta et al. / Energy 123 (2017) 189e197
120 100 80 60 40 20 0 2007
2008
2009
2010 VIX
2011
2012
2013
2014
OVX
Fig. 1. OVX and VIX from 10/05/2007-31/12/2014.
of their employed models. Moreover, with a view to investigating the association between oil price uncertainty and stock market volatility after controlling for the effect of implied volatility index of S & P 500, we consider estimating another extended form of GARCH process as given below: 2 2 ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1
(3)
where u > 0, a 0, b 0, g 0 and q 0 to guarantee that ht is positive. In model (3), a statistically significant value of the parameter q implies that there exists a direct connection between the US equity market uncertainty and the stock market volatility of the countries considered in our empirical inspection. Now, in order to examine the potential OVX-stock market volatility relationship, when extreme movements occur in the implied volatility index, we estimate the following GARCH equation: 2 ht ¼ u þ aε2t1 þ bht1 þ qVIXt1 þ
2 X
2 di Di OVXt1
(4)
i¼1
where u > 0, a 0, b 0, q 0 and di 0 to guarantee that ht is positive. Also,
2 2 ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1 þ ls2t1
where u > 0, a 0, b 0, g 0, q 0 and l > 0 to guarantee that ht is positive. In model (5), the term s2t1 indicates the regional market volatility effect at period t-1. Now to obtain s2t , we first find c ht by estimating a univariate GARCH (1,1) process for all the markets under study. We then have a matrix consisting of twelve columns, where each column contains the time series observations of c ht of a specific market. In the next step, we perform a principal component analysis to obtain a column vector consisting of the values of s2t . Finally, in order to check the robustness of our findings, we also present the results from exponential GARCH (EGARCH) model. The EGARCH (1,1) model, proposed by Nelson [29], takes the following form:
lnðht Þ ¼ c þ
ajεt1 j þ vεt1 pffiffiffiffiffiffiffiffiffiffi þ b lnðht1 Þ ht1
(6)
In model (11), when εti is positive or there is good news, the total effect of εt1 is ð1 þ vÞ jεt1 j. In contrast, when εti is negative or there is bad news, the total impact of εt1 is ð1 vÞ jεt1 j. Now to observe the effects of implied volatility indices, VIX and OVX, equation (6) can be extended as follows:
lnðht Þ ¼ c þ D1 ¼ 1; if OVX < Q1 ; 0; otherwise;
(5)
ajεt1 j þ vεt1 2 2 pffiffiffiffiffiffiffiffiffiffi þ b lnðht1 Þ þ gOVXt1 þ qVIXt1 ht1 (7)
and
D2 ¼ 1; if OVX < Q3 ; 0; otherwise; 4.2. GARCHejump model where Q1 and Q3 indicate the first and third quartiles of the variable OVX. The next step is to examine whether the oil market uncertainty has any potential impacts on different equity markets in the presence of regional market volatility factor, we estimate the following multivariate GARCH model:
We advance our analysis by considering the application of the GARCH-jump model, proposed by Chan and Maheu [11], to examine whether the stock returns of the sampled markets are sensitive to high fluctuations occurring in the implied oil volatility index. Such methodology provides a beneficial approach to further assess the
A. Dutta et al. / Energy 123 (2017) 189e197
robustness of our results, since unlike the traditional GARCH models, it can capture the effects of extreme news or abnormal information arising from earnings surprises, crashes, terrorist attacks and similar other events (Fowowe [18]). Our estimated model assumes the following form2:
Rt ¼ t þ m1 Rt1 þ m2 Rt2 þ jROt1 þ εt
Table 2 Estimates of OVX effects on the volatility of selected stock markets. Countries
u
a
b
g
Saudi Arabia
0.038** (0.05) 0.032*** (0.00) 0.008** (0.01) 0.006*** (0.00) 0.050*** (0.00) 0.135*** (0.00) 0.103*** (0.00) 0.105*** (0.00) 0.049** (0.04) 0.085*** (0.00) 0.095*** (0.00) 0.005 (0.27)
0.061*** (0.00) 0.074*** (0.00) 0.061*** (0.00) 0.062*** (0.00) 0.090*** (0.00) 0.114*** (0.00) 0.188*** (0.00) 0.068*** (0.00) 0.074*** (0.00) 0.197*** (0.00) 0.240*** (0.00) 0.153*** (0.00)
0.925*** (0.00) 0.890*** (0.00) 0.906*** (0.00) 0.923*** (0.00) 0.784*** (0.00) 0.588*** (0.00) 0.618*** (0.00) 0.793*** (0.00) 0.872*** (0.00) 0.707*** (0.00) 0.570*** (0.00) 0.758*** (0.00)
0.000535*** (0.00) 0.000020** (0.03) 0.000010*** (0.00) 0.000001 (0.32) 0.000040*** (0.00) 0.000100*** (0.00) 0.000040*** (0.00) 0.000100*** (0.00) 0.000105** (0.01) 0.000010*** (0.00) 0.000100*** (0.00) 0.000010*** (0.00)
United Arab Emirates
(8) Kuwait
where Rt denotes the log return of stock prices between time t and t-1, ROt1 is the log return on the OVX series at time t-1 and εt refers to the error term at time t which has two components as follows:
Qatar Jordan Morocco
εt ¼ ε1t þ ε2t
(9) Tunisia
The first component ε1t is a mean-zero innovation with normal stochastic process assuming the following form:
ε1t ¼
pffiffiffiffiffi ht zt ;
Egypt South Africa
zt NIDð0; 1Þ
Nigeria 0
ht ¼ u0 þ a0 ε21t1 þ b ht1
(10)
The second component ε2t is a jump innovation which consists of abnormal price movements with Eðε2t jIt1 Þ ¼ 0, where It1 denotes the information set. Now ε2t is defined as the discrepancy between the jump component and the expected total jump size between t-1 and t:
ε2t ¼
nt X
utl kmt
193
Kenya Mauritius
Notes: This table presents the estimates of the oil volatility index (OVX) effects on the current stock market volatility for different countries under study. These results 2 : are obtained by estimating the following model: ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 Values in the parentheses indicate the p-values. ** and *** represent significance at 5% and 1% levels, respectively.
(11)
l¼1
where utl refers to the jump size and is assumed to be normally Pnt distributed with mean k and variance d2 , u is the jump l¼1 tl component and nt designates the number of jumps. It is assumed that nt is distributed as a Poisson variable with an autoregressive conditional jump intensity (ARJI) given by
mt ¼ m0 þ rmt1 þ 4xt1
(12)
where mt is the time-varying conditional jump intensity parameter and mt > 0, m0 > 0, r > 0 and 4 > 0. Now the log-likelihood function can be expressed as:
LðUÞ ¼
T X
log f ðrt jIt1 ; UÞ
t¼1 0
0
0
where U ¼ ðt; m1 ; m2 ; j; u ; a ; b ; k; d; m0 ; r; 4Þ. 5. Empirical results 5.1. Findings of the modified GARCH (1,1) model In this section, we analyze our findings to examine whether the oil market uncertainty affects the volatility of different stock markets considered, and second, whether oil market uncertainty impacts on various equity markets in the presence of regional market volatility factor. To find out the effect of oil market uncertainty on volatility on each stock market, we estimate the extended GARCH model, presented in equation (2), for twelve different countries. The results for each of the variance equations are reported in Table 2. The
2 Fowowe [18] employs a similar mean equation to address the impact of oil price fluctuations on Nigerian stock market returns. Fowowe uses traditional oil price series in the jump model, while we consider the oil volatility index.
hypothesis that expected volatility is transmitted from the oil 2 , market to the stock market is supported if the coefficient of OVXt1 g, is significantly different from zero. Table 2 indicates that the implied volatility in stock returns for each country, except Qatar, is affected by the oil market implied 2 . volatility, as evidenced by the significant coefficients on OVXt1 The estimated coefficient for Saudi Arabia seems to be higher than that for other markets, implying that the global oil market plays a pivotal role in the Saudi Arabian economy. The outcomes of Table 2 further reveal that the coefficients of ε2t1 and ht1 are statistically significant for all of the countries under investigation. This suggests that the volatility of stock return is influenced by its own volatility. Furthermore, the positive values of the coefficient b indicate that higher levels of conditional volatility in the past are associated with higher conditional volatility in the current period. Table 3 documents the empirical results estimated from the variance equation (3). These results enable us to assess the connection between oil and equity markets after controlling for the effect of implied stock market volatility index (VIX). The outcomes reveal that, although VIX plays as a moderator in the adopted 2 , g, are still statistically model, the estimated coefficients of OVXt1 significant for almost half of the markets analyzed. 2 Additionally, we find significant results for the variable VIXt1 for nine out of the twelve stock markets under study, which indicates strong links between the US equity market uncertainty and the implied volatility of those nine equity markets. Moreover, the values of the coefficients for both volatility indices are significant for four markets: Saudi Arabia, Kuwait, Tunisia and Mauritius. For example, in case of Kuwaiti market, the parameter estimates for 2 2 , suggesting that the OVXt1 is twice as high as that for VIXt1 Kuwaiti market volatility is two times more sensitive to the oil market uncertainty than to US stock market uncertainty. The results of Table 3 further reveal that the coefficients of ε2t1 and ht1 , a and b respectively, are still statistically significant (at 1% level) for all the stock markets except for South Africa.
194
A. Dutta et al. / Energy 123 (2017) 189e197
Table 3 Estimates of OVX and VIX effects on the volatility of selected stock markets. Countries
u
a
b
g
q
Saudi Arabia
0.065*** (0.00) 0.031*** (0.00) 0.009*** (0.00) 0.006*** (0.00) 0.053*** (0.00) 0.146*** (0.00) 0.105*** (0.00) 0.121*** (0.00) 0.239 (0.10) 0.086*** (0.00) 0.099*** (0.00) 0.010 (0.22)
0.057*** (0.00) 0.076*** (0.00) 0.058*** (0.00) 0.062*** (0.00) 0.092*** (0.00) 0.115*** (0.00) 0.191*** (0.00) 0.071*** (0.00) 0.002 (0.91) 0.198*** (0.00) 0.239*** (0.00) 0.180*** (0.00)
0.929*** (0.00) 0.885*** (0.00) 0.913*** (0.00) 0.923*** (0.00) 0.770*** (0.00) 0.567*** (0.00) 0.609*** (0.00) 0.756*** (0.00) 0.260 (0.36) 0.709*** (0.00) 0.550*** (0.00) 0.650*** (0.00)
0.000496*** (0.00) 0.000006 (0.59) 0.000020*** (0.00) 0.000001 (0.32) 0.000004 (0.70) 0.000091** (0.03) 0.000082*** (0.00) 0.000070* (0.06) 0.000080 (0.61) 0.000044 (0.11) 0.000019 (0.46) 0.000021*** (0.00)
0.000675*** (0.00) 0.000076** (0.03) 0.000010*** (0.00) 0.000028* (0.06) 0.000100*** (0.00) 0.000100* (0.07) 0.000089** (0.02) 0.000273*** (0.00) 0.004219** (0.01) 0.000025 (0.62) 0.000190*** (0.00) 0.000394*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: This table displays the estimates of the oil volatility index (OVX) effects on the current stock market volatility after controlling for the effect of US equity market 2 2 : Values in the parentheses indicate the p-values. *, uncertainty (VIX). We find these outcomes by estimating the following model: ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1 ** and *** represent significance at 10%, 5% and 1% levels, respectively.
To examine the impact of oil market uncertainty on the equity markets in the presence of regional market volatility factor, we estimate equation (5) for all the markets considered and display the results in Table 5. The findings reveal that in the presence of regional market volatility factor, the oil price uncertainty has significant impacts on half of the equity markets considered. Additionally, the VIX factor is found to be significant for most of the countries implying that shocks still spill over from the US equity
Table 4 presents the results on equation (4) for each market. The findings indicate that for Saudi Arabia, United Arab Emirates, Qatar, Jordan and Morocco, the corresponding volatility in returns is significantly affected by the higher values of the oil volatility index. However, when the oil price uncertainty is low, only Kenyan equity market significantly reacts to this change. As an additional analysis, we also perform similar analysis for oil price levels and find consistent results.
Table 4 Estimates of OVX effects with extreme movements. Countries
u
a
b
q
d1
d2
Saudi Arabia
0.014 (0.73) 0.077*** (0.00) 0.010 (0.29) 0.020*** (0.00) 0.168*** (0.00) 0.014 (0.76) 0.092*** (0.00) 0.129*** (0.00) 0.040 (0.87) 0.089* (0.05) 0.049 (0.27) 0.015 (0.39)
0.044*** (0.00) 0.071*** (0.00) 0.058*** (0.00) 0.059*** (0.00) 0.091*** (0.00) 0.106*** (0.00) 0.186*** (0.00) 0.071*** (0.00) 0.006 (0.77) 0.198*** (0.00) 0.235*** (0.00) 0.181*** (0.00)
0.931*** (0.00) 0.885*** (0.00) 0.913*** (0.00) 0.924*** (0.00) 0.708*** (0.00) 0.568*** (0.00) 0.615*** (0.00) 0.754*** (0.00) 0.249 (0.36) 0.702*** (0.00) 0.533*** (0.00) 0.644*** (0.00)
0.000646*** (0.00) 0.000061* (0.09) 0.000024* (0.07) 0.000009 (0.19) 0.000128*** (0.00) 0.000201** (0.02) 0.000078** (0.04) 0.000275*** (0.00) 0.004238*** (0.00) 0.000021 (0.69) 0.000181** (0.01) 0.000404*** (0.00)
0.000033 (0.61) 0.000005 (0.87) 0.000003 (0.84) 0.000013 (0.10) 0.000049 (0.23) 0.000157 (0.10) 0.000023 (0.66) 0.000009 (0.90) 0.000670 (0.14) 0.000019 (0.78) 0.000199** (0.01) 0.000011 (0.69)
0.000318*** (0.00) 0.000061*** (0.00) 0.000001 (0.93) 0.000011*** (0.00) 0.000115*** (0.00) 0.000124*** (0.00) 0.000026 (0.26) 0.000005 (0.88) 0.000028 (0.88) 0.000013 (0.69) 0.000024 (0.43) 0.000002 (0.88)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: This table describes how the selected equity markets respond when extreme movements occur in the oil volatility index. Doing so considers two dummy variables. We define these dummies as follows:D1 ¼ 1, if OVX < Q1 ; 0, otherwise and D2 ¼ 1, if OVX > Q3 ; 0, otherwise, where Q1 and Q3 indicate the first and third quartiles of the variable P 2 2 : Values in the parentheses indicate the p-values. *, ** OVX. The following model has been estimated to obtain these results: ht ¼ u þ aε2t1 þ bht1 þ qVIXt1 þ 2i¼1 di Di OVXt1 and *** represent significance at 10%, 5% and 1% levels, respectively.
A. Dutta et al. / Energy 123 (2017) 189e197
195
Table 5 Estimates of regional market volatility effects. Countries
u
a
b
g
q
l
Saudi Arabia
0.049*** (0.00) 0.065** (0.02) 0.021** (0.01) 0.010 (0.15) 0.083*** (0.00) 0.143** (0.01) 0.135*** (0.00) 0.084* (0.06) 0.770* (0.05) 0.042* (0.08) 0.105** (0.01) 0.072** (0.03)
0.055*** (0.00) 0.074*** (0.00) 0.059*** (0.00) 0.059*** (0.00) 0.092*** (0.00) 0.114*** (0.00) 0.190*** (0.00) 0.073*** (0.00) 0.012 (0.56) 0.201*** (0.00) 0.238*** (0.00) 0.218*** (0.00)
0.930*** (0.00) 0.882*** (0.00) 0.906*** (0.00) 0.924*** (0.00) 0.758*** (0.00) 0.574*** (0.00) 0.598*** (0.00) 0.758*** (0.00) 0.099 (0.75) 0.705*** (0.00) 0.547*** (0.00) 0.459*** (0.00)
0.000203*** (0.00) 0.000014 (0.30) 0.000019*** (0.00) 0.000009*** (0.00) 0.000001 (0.90) 0.000089** (0.03) 0.000081*** (0.00) 0.000076** (0.04) 0.000024 (0.90) 0.000047* (0.08) 0.000017 (0.55) 0.000020 (0.42)
0.000308*** (0.00) 0.000066* (0.06) 0.000029** (0.04) 0.000014** (0.01) 0.000091*** (0.00) 0.000149* (0.07) 0.000120** (0.01) 0.000303*** (0.00) 0.004630*** (0.00) 0.000089 (0.10) 0.000189** (0.01) 0.000560*** (0.00)
0.443950*** (0.00) 0.017581 (0.20) 0.007520 (0.12) 0.002190 (0.60) 0.014887* (0.08) 0.000555 (0.97) 0.014032 (0.22) 0.019492 (0.25) 0.263737 (0.10) 0.025445** (0.01) 0.003852 (0.81) 0.056397*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: The numbers shown in this table are used to examine the effects of oil market uncertainty on considered equity markets in the presence of regional market volatility 2 2 factor (s2t ). To obtain these results, we estimate the model given by: ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1 þ ls2t1 . Values in the parentheses indicate the p-values. *, ** and *** represent significance at 10%, 5% and 1% levels, respectively.
robustness test using the EGARCH model. These findings, given in Table 6, are consistent with those reported in Table 3 with one exception. The GARCH (1,1) specification shows that crude oil volatility index plays a potential role in forecasting the stock market volatility of Mauritius, while we do not report such significant result using the EGARCH (1,1) approach. Since all other outcomes are analogous in Tables 3 and 6, we thus conclude that our results are robust to various changes in the specification of the adopted GARCH model. Furthermore, we divide our sample period into two subsamples
market to the other stock markets. Moreover, there is significant association between second moments of regional market and equity markets of Saudi Arabia, Kenya and Mauritius. Thus, the stock market volatility of most of the countries is not impacted by the regional market volatility. We further document that only for Saudi Arabia, the global oil market, US equity market and the stock markets of other neighboring states play crucial roles, suggesting that the Saudi economy is significantly influenced by all these factors. In order to check the sensitivity of our findings, we conduct a
Table 6 Estimates of EGARCH model. Countries
c
a
b
v
g
q
Saudi Arabia
1.23*** (0.00) 1.11*** (0.00) 1.49*** (0.00) 0.13*** (0.00) 0.21*** (0.00) 0.30*** (0.00) 0.35*** (0.00) 0.06*** (0.00) 0.06*** (0.00) 0.28*** (0.00) 0.39*** (0.00) 0.29*** (0.00)
0.184*** (0.00) 0.168*** (0.00) 0.166*** (0.00) 0.167*** (0.00) 0.205*** (0.00) 0.246*** (0.00) 0.348*** (0.00) 0.130*** (0.00) 0.103*** (0.00) 0.360*** (0.00) 0.392*** (0.00) 0.310*** (0.00)
0.891*** (0.00) 0.966*** (0.00) 0.675*** (0.00) 0.988*** (0.00) 0.893*** (0.00) 0.706*** (0.00) 0.845*** (0.00) 0.885*** (0.00) 0.962*** (0.00) 0.883*** (0.00) 0.814*** (0.00) 0.937*** (0.00)
0.125*** (0.00) 0.061*** (0.00) 0.011 (0.50) 0.024*** (0.00) 0.035*** (0.00) 0.034 (0.18) 0.008 (0.67) 0.114*** (0.00) 0.117*** (0.00) 0.007 (0.73) 0.008 (0.71) 0.011 (0.43)
0.000449*** (0.00) 0.000004 (0.73) 0.000640*** (0.00) 0.000005 (0.23) 0.000001 (0.94) 0.000049** (0.03) 0.000038*** (0.00) 0.000018* (0.05) 0.000003 (0.63) 0.000023 (0.17) 0.000006 (0.71) 0.000001 (0.88)
0.000354*** (0.00) 0.000010** (0.02) 0.000622*** (0.00) 0.000043* (0.08) 0.000074*** (0.00) 0.000078* (0.05) 0.000031** (0.04) 0.000023*** (0.00) 0.000021** (0.03) 0.000021 (0.33) 0.000132*** (0.00) 0.000065*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
jþvεt1 2 2 : Values in the parentheses indicate the ppffiffiffiffiffiffiffi Notes: This table displays the estimates of the EGARCH model given as: lnðht Þ ¼ c þ ajεt1 þ qVIXt1 þ b lnðht1 Þ þ gOVXt1 ht1 values. *, ** and *** represent significance at 10%, 5% and 1% levels, respectively.
196
A. Dutta et al. / Energy 123 (2017) 189e197
volatility of stock returns, the jump model now confirms that OVX affects the mean returns as well. It is also evident from Table 7 that the jump parameters are generally significant implying that jumps do exist in the stock returns and they are time-varying. The positive (negative) coefficient of the jump mean indicates that the jump behavior driven by abnormal information has a positive (negative) impact on returns, while the positive (negative) coefficient of the jump variance infers that volatility driven by abnormal information has a positive (negative) effect on volatility of returns (Fowowe [18]). The findings also document that the jump intensity parameters (m0 ; r; 4) are mostly statistically significant suggesting that the jump intensity varies over time (Chiou and Lee [13]). Additionally, since these parameters satisfy the constraints that m0 > 0, r > 0 and 4 > 0, it can be concluded that the GARCH-ARJI model is appropriate for describing the jump behavior in the stock market returns. Furthermore, the positive r and 4 suggest that the current jump intensity (mt ) is affected by the most recent jump intensity (mt1 ) and intensity residuals (xt1 ). We further find that the high values of r and 4 indicate a high degree of persistence in the jump intensity.
with equal number of observations. These results indicate some variation in terms of the statistical significances, in particular, which may be due to the substantially reduced sample sizes.3 Our results therefore collectively suggest that global oil and stock market uncertainties, measured by OVX and VIX, impact the stock market volatility of several Middle East and African countries. The association is clearly strongest in the case of Saudi Arabia, which is the most important oil exporting country. 5.2. Findings of the GARCH-Jump model Table 7 displays the results of the autoregressive conditional jump intensity model. We find that the GARCH parameters are statistically significant at 1% level indicating the existence of ARCH 0 0 and GARCH effects. The sum of a and b also implies high degree of persistence in the return fluctuations. Our findings further reveal that stock returns of majority of the sampled markets are affected when changes occur in the crude oil volatility index. To be specific, we find significant oil-stock links for the following countries: Qatar, Jordan, Nigeria, Egypt, South Africa, Tunisia, and Mauritius. It is unexpected that the Saudi, UAE and Kuwaiti stock markets are not affected by the variations in oil price volatility. However, these findings are consistent with those reported by Arouri et al. [3] and Hammoudeh and Choi [19]. Arouri et al., for example, argue that such an insignificant oil-stock link can be explained by the lower reliance of stock markets on oil companies (Kuwait), and the low annual turnover (Saudi Arabia and UAE). Hammoudeh and Choi also document that Kuwaiti market is highly sensitive to fads and herding which could limit its dependence on the oil market. We further report that all the statistically significant coefficients are found to be negative. Such outcomes are not surprising, since a rise in the oil price uncertainty may cause the stock prices to decrease. Thus in addition to the modified GARCH model, which documents that oil price uncertainty, measured by OVX, impacts the realized
6. Conclusions This paper examines whether the implied crude oil volatility index (OVX), a forward-looking measure of oil market uncertainty published by Chicago Board Options Exchange (CBOE), impacts the realized volatility of Middle East and African stock markets. The data include daily stock index observations from Saudi Arabia, United Arab Emirates, Kuwait, Qatar, Jordan, Morocco, Tunisia, Egypt, South Africa, Nigeria, Kenya and Mauritius. The study contributes to the standing literature mainly in two aspects. First, while earlier studies such as Malik and Ewing [26], Lee and Chiou [24], Arouri et al. [1,2] and Bouri [9,10] utilize the
Table 7 Estimated results of GARCH-Jump models. Countries
t
m1
m2
j
u0
a0
b0
k
d2
m0
r
4
Saudi Arabia
0.00011 (0.24) 0.00903*** (0.00) 0.02095 (0.11) 0.03244** (0.01) 1.1091*** (0.00) 0.02797 (0.18) 0.01645 (0.36) 0.00004 (0.86) 0.06612 (0.07)* 0.01080 (0.64) 0.08186 (0.00)*** 0.00808 (0.57)
0.00019 (0.35) 0.01034*** (0.00) 0.01076 (0.60) 0.06227*** (0.00) 0.11254 (0.24) 0.05958*** (0.00) 0.03745 (0.11) 0.00010*** (0.00) 0.00709 (0.73) 0.31975*** (0.00) 0.26243*** (0.00) 0.03749* (0.07)
0.00009 (0.31) 0.01646*** (0.00) 0.01346 (0.49) 0.03039 (0.16) 0.11571 (0.16) 0.03531 (0.10) 0.00698 (0.76) 0.00008 (0.76) 0.03965* (0.07) 0.02549 (0.26) 0.07296*** (0.00) 0.07174*** (0.00)
0.00002 (0.29) 0.00010 (0.80) 0.00045 (0.90) 0.01295*** (0.00) 0.04291*** (0.00) 0.00658 (0.10) 0.00610* (0.08) 0.00003* (0.09) 0.02129*** (0.00) 0.01235*** (0.00) 0.00295 (0.46) 0.00655** (0.03)
0.030*** (0.00) 0.034*** (0.00) 0.070*** (0.00) 0.017*** (0.00) 1.08*** (0.00) 0.005 (0.12) 0.133*** (0.00) 0.109*** (0.00) 0.007** (0.02) 0.200*** (0.00) 0.144*** (0.00) 0.025*** (0.00)
0.062*** (0.00) 0.063*** (0.00) 0.059*** (0.00) 0.085*** (0.00) 0.140*** (0.00) 0.023*** (0.00) 0.155*** (0.00) 0.057*** (0.00) 0.004* (0.06) 0.167*** (0.00) 0.208*** (0.00) 0.148*** (0.00)
0.921*** (0.00) 0.832*** (0.00) 0.668*** (0.00) 0.740*** (0.00) 0.858*** (0.00) 0.957*** (0.00) 0.529*** (0.00) 0.908*** (0.00) 0.988*** (0.00) 0.340** (0.01) 0.442*** (0.00) 0.759*** (0.00)
0.006 (0.77) 0.029 (0.29) 0.082 (0.18) 0.068 (0.52) 13.68*** (0.00) 0.048 (0.70) 0.239* (0.08) 0.004 (0.99) 0.175* (0.08) 0.032 (0.31) 0.311*** (0.00) 0.186 (0.31)
1.16*** (0.00) 1.45*** (0.00) 1.42*** (0.00) 2.05*** (0.00) 5.31*** (0.00) 1.51*** (0.00) 1.56*** (0.00) 1.32*** (0.00) 1.81*** (0.00) 1.07*** (0.00) 1.34*** (0.00) 2.08*** (0.00)
0.013*** (0.00) 0.159*** (0.00) 0.008*** (0.00) 0.238*** (0.00) 0.022*** (0.00) 0.109 (0.14) 0.001 (0.20) 0.063*** (0.00) 0.010** (0.03) 0.027** (0.04) 0.003 (0.10) 0.086*** (0.00)
0.988*** (0.00) 0.920*** (0.00) 0.992*** (0.00) 0.697*** (0.00) 0.706*** (0.00) 0.394 (0.32) 0.990*** (0.00) 0.964*** (0.00) 0.977*** (0.00) 0.970*** (0.00) 0.996*** (0.00) 0.677*** (0.00)
0.069*** (0.00) 0.229*** (0.00) 0.227*** (0.00) 0.132*** (0.00) 0.881*** (0.00) 0.583** (0.01) 0.076** (0.03) 0.066*** (0.00) 0.667*** (0.00) 0.419*** (0.00) 0.114** (0.01) 0.126*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: Values in the parentheses indicate the p-values. *, ** and *** represent significance at 10%, 5% and 1% levels, respectively.
3
Results are available from the authors.
crude oil price series to model the volatility, we extend their works by assessing the association between realized volatilities of sample
A. Dutta et al. / Energy 123 (2017) 189e197
stock markets and the global oil price uncertainty measured by OVX. This can be considered as an important contribution, since OVX, in contrast to traditional oil price indices, contains both historical volatility information as well as investors' expectation of future market conditions and hence turns out to be a good indicator of oil market uncertainty [25]. Second, our paper also contributes to the existing literature on uncertainty transmission by examining whether global oil market uncertainty impacts the realized volatility of the Middle East and African stock markets. At the empirical stage, we apply a modified form of the generalized autoregressive conditional heteroskedasticity (GARCH) model to examine the relationship between expected oil market volatility and the realized volatility of different equity markets under study. In addition, we combine OVX with VIX to observe their joint effects on the stock markets. The findings of our analysis show significant links between oil price uncertainty and the realized stock market volatility. We also find that such associations seem to hold even after controlling for the effect of global stock market uncertainty. We further report that, conditioned on the past information, the volatility in each market is persistent, which suggests that the current volatility of stock returns for a specific market is directly affected by its past volatility. In addition, we employ the GARCH-Jump mode to investigate whether fluctuations in the implied oil volatility index have any major impacts on the stock returns. The findings indicate that such association is significant for majority of the sampled markets and that time-varying jumps do exist in the stock returns. To sum up, our results suggest that the implied crude oil volatility index, i.e. the market participants' anticipation of the future oil market uncertainty is an important factor explaining the returns and volatilities of the Middle East and African stock markets. Acknowledgements Editor Isabel Soares and two anonymous reviewers are gratefully acknowledged for their valuable comments, as well as the participants of the Multinational Finance Society conference in 2016 for their suggestions. Jussi Nikkinen acknowledges the generous financial support from the OP Group Research Foundation. All the remaining errors are solely on the authors' responsibility. References [1] Arouri M, Jouini J, Nguyen D. Volatility spillovers between oil prices and stock sector returns: implications for portfolio management. J Int Money Finance 2011;30:1387e405. [2] Arouri M, Jouini J, Nguyen D. On the impacts of oil price fluctuations on European equity markets: volatility spillover and hedging effectiveness. Energy Econ 2012;34:611e7. [3] Arouri M, Lahiani A, Nguyen D. Return and volatility transmission between world oil prices and stock markets of the GCC countries. Econ Model 2011;28:
197
1815e25. [4] Basher SA, Sadorsky P. Oil price risk and emerging stock markets. Glob Finance J 2006;17:224e51. [5] Bernanke BS. Irreversibility, uncertainty, and cyclical investment. Q J Econ 1983;98:85e106. [6] Blair BJ, Poon SH, Taylor SJ. Forecasting S&P 100 volatility: the incremental information content of implied volatilities and high-frequency index returns. J Econ 2001;105:5e26. [7] Bollerslev T. Generalised autoregressive conditional heteroskedasticity. J Econ 1986;31:307e27. [8] Bollerslev T, Osterrieder D, Sizova N, Tauchen G. Risk and return: long-run relationships, fractional cointegration, and return predictability. J Financ. Econ 2013;108:409e24. [9] Bouri E. Return and volatility linkages between oil prices and the Lebanese stock market in crisis periods. Energy 2015a;89:365e71. [10] Bouri E. Oil volatility shocks and the stock markets of oil-importing MENA economies: a tale from the financial crisis. Energy Econ 2015b;51:590e8. [11] Chan WH, Maheu JM. Conditional jump dynamics in stock market returns. J Bus Econ Stat. 2002;20:377e89. [12] Chen NF, Roll R, Ross SA. Economic forces and the stock market. J Bus 1986;59: 383e403. [13] Chiou JS, Lee YH. Jump dynamics and volatility: oil and the stock markets. Energy 2009;34:788e96. [14] Choi K, Hammoudeh S. Volatility behavior of oil, industrial commodity and stock markets in a regime switching environment. Energy Policy 2010;38: 4388e99. [15] Ciner C. Oil and stock returns: frequency domain evidence. J Int Financ. Mark Inst. Money 2013;23:1e11. [16] Engle R. Autoregressive conditional heteroscedasticity with estimates of the variance of the U. K. inflation. Econometrica 1982;50:987e1008. [17] Ewing BT, Malik F. Volatility spillovers between oil prices and the stock market under structural breaks. Glob Finance J 2016;29:12e23. [18] Fowowe B. Jump dynamics in the relationship between oil prices and the stock market: evidence from Nigeria. Energy 2013;56:31e8. [19] Hammoudeh S, Choi K. Behavior of GCC stock markets and impacts of US oil and financial markets. Res Int Bus Finance 2006;20:22e44. r P, Westgaard S. Forecasting volatility of the [20] Haugom E, Langeland H, Molna U.S. oil market. J Bank Finance 014; 47: 1e14. [21] Kambouroudis DS, McMillan DG. Does VIX or volume improve GARCH volatility forecasts? Appl Econ 2016;48:1e19. [22] Killian L, Park C. The impact of price shocks on the U.S. stock market. Int Econ Rev 2009;50:1267e87. [23] Kling JL. Oil price shocks and stock market behavior. J Portf. Manag 1985;12: 34e9. [24] Lee YH, Chiou JS. Oil price sensitivity and its asymmetric impact on the stock market. Energy 2011;36:168e74. [25] Liu ML, Ji Q, Fan Y. How does oil market uncertainty interact with other markets: an empirical analysis of implied volatility index? Energy 2013;55: 860e8. [26] Malik F, Ewing BT. Volatility transmission between oil prices and equity sector returns. Int Rev Financ. Anal. 2009;18:95e100. [27] Malik S, Hammoudeh S. Shock and volatility transmission in the oil, US and Gulf equity markets. Int Rev Econ Finance 2007;17:357e68. €m P. International transmission of uncertainty implicit in [28] Nikkinen J, Sahlstro stock index option prices. Glob Finance J 2004;15:1e15. [29] Nelson DB. Conditional heteroskedasticity in asset returns: a new approach. Econometrica 1991;59:347e70. [30] Papapetrou E. Oil price shocks, stock market, economic activity and employment in Greece. Energy Econ 2001;23:511e32. [31] Pindyck R. Irreversibility, uncertainty, and investment. J Econ Lit. 1991;29: 110e48. [32] Sadorsky P. Oil price shocks and stock market activity. Energy Econ 1999;21: 449e69. [33] Wang K. Volatility linkages of the equity, bond and money markets: an implied volatility approach. Account Finance 2009;49:207e19.
Contents lists available at ScienceDirect
Energy journal homepage: www.elsevier.com/locate/energy
Impact of oil price uncertainty on Middle East and African stock markets Anupam Dutta*, Jussi Nikkinen, Timo Rothovius Department of Accounting & Finance, University of Vaasa, P.O. Box 700, FI-65101 Vaasa, Finland
a r t i c l e i n f o
a b s t r a c t
Article history: Received 14 March 2016 Received in revised form 15 August 2016 Accepted 24 January 2017 Available online 30 January 2017
This paper investigates whether the implied crude oil volatility index (OVX), a forward-looking measure of oil market uncertainty published by Chicago Board Options Exchange (CBOE), impacts the realized volatility of Middle East and African stock markets. Using an extended version of the GARCH model, we show that the oil market uncertainty has substantial effects on the realized volatility of most of the markets under study. Our findings also reveal that, even after controlling for the effect of the implied volatility index of S & P 500 (VIX), the impact of the OVX on the Middle East and African equity markets still holds for almost half of the markets considered. Additionally, the application of the GARCH-jump model shows that stock returns of majority of the sampled markets are sensitive to the fluctuations occurring in the implied oil volatility index and that time-varying jumps do exist in the stock returns. Thus, the market participants' anticipation of the future oil market uncertainty is an important factor explaining the returns and volatilities of the Middle East and African equity markets. © 2017 Elsevier Ltd. All rights reserved.
Keywords: Oil price uncertainty VIX OVX Middle East Africa GARCH-jump model
1. Introduction The recent downturn in crude oil price is substantially affecting the economies all over the world, oil being one of the major production factors in most countries. In oil-depending economies, changes in oil price and volatility can be expected to influence also their equity markets. For example, Ciner [15] argues that oil price shocks can cause changes in expected cash flows, by affecting the overall economy as well as the discount rate used to value equities by changing inflationary expectations. Consequently, large oil price movements tend to increase uncertainty of future equity prices as evidenced, for example, by Bernanke [5] and Pindyck [31]. In this paper, our objective is to examine whether the implied crude oil volatility index (OVX), a forward-looking measure of oil market uncertainty published by Chicago Board Options Exchange (CBOE), impacts the volatility of Middle East and African stock markets. Over the past years, the relation between oil price changes and stock returns has received an increasing attention in the energy and finance literature. Accordingly, numerous empirical studies have investigated the return and volatility transmission mechanism between oil price and stock returns. Notable contributions include
* Corresponding author. E-mail address: adutta@uwasa.fi (A. Dutta). http://dx.doi.org/10.1016/j.energy.2017.01.126 0360-5442/© 2017 Elsevier Ltd. All rights reserved.
e.g. Sadorsky [32], Papapetrou [30], Killian and Park [22], Malik and Ewing [26], Lee and Chiou [24], Arouri, Jouini and Nguyen [1,2] and Bouri [9,10]. Collectively, the above articles show that oil and stock markets are associated with each other. A unifying feature of these articles is that they use crude oil price series to investigate the volatility. Our study contributes to the existing literature in two major ways. First, our paper models realized volatilities of emerging stock markets, taking into account the global anticipation of future oil price uncertainty, measured using the implied crude oil volatility index, OVX. This is an advantageous approach, since the implied oil volatility index, which is derived from option prices, is generally considered to be a good indicator of oil market uncertainty (see, e.g., Liu, Ji and Fan [25]). According to Liu et al. [25], this is because implied volatilities not only contain historical volatility information, but also investors' expectation of future market conditions. r and Westgaard [20] find Additionally, Haugom, Langeland, Molna that the day-ahead and week-ahead volatility forecasts can be significantly improved by including information from the OVX. Given the useful properties of the OVX, our paper provides a novel extension to papers such as Sadorsky [32], Papapetrou [30], Killian and Park [22], Malik and Ewing [26], Lee and Chiou [24], Arouri et al. [1,2] and Bouri [9,10] that have used the crude oil price series to model volatility. Second, many recent studies have used implied volatility
190
A. Dutta et al. / Energy 123 (2017) 189e197
indices, such as VIX, to assess the uncertainty transmission among € m [28]; Wang [33]; the financial markets (Nikkinen and Sahlstro Bollerslev, Osterrieder, Sizova and Tauchen [8]; Liu et al. [25]). Of these studies, Liu et al. [25], for instance, find that the US stock market uncertainty, measured by VIX, affects oil price uncertainty measured by using OVX. Nonetheless, OVX has received very little attention in the literature which is very surprising for the oil market considering its economic importance (Haugom et al. [20]). Thus, our paper contributes also to the uncertainty transmission literature by examining whether global oil market uncertainty affects the realized volatilities of Middle East and African equity markets. In our empirical approach, we apply a modified form of the generalized autoregressive conditional heteroskedasticity (GARCH) model to examine the relationship between oil market uncertainty and the volatility of different equity markets under study.1 Our findings based on the extended model indicate significant links between oil price uncertainty and the realized stock market volatility. We also combine OVX with VIX to observe their joint effects on the selected markets and document that, association between oil price uncertainty and realized stock market volatility mostly hold even after controlling for the effect of VIX. Specifically, after inserting VIX as a regressor in the GARCH process, oil market uncertainty evidently affects the variance of nearly half of the markets considered. We further report that, conditioned on the past information, the volatility in each market is very persistent. This latter outcome suggests that the current volatility of stock returns for a specific market is directly affected by its past volatility. Additionally, we consider the application of the GARCH-Jump mode to assess whether variations in the oil volatility index have any significant influences over the stock returns. The results suggest that such associations hold for majority of the sampled markets and that jumps exist in the returns which are time-varying. The results of the present study are useful for energy policy makers, investors and researchers in several aspects. For example, the findings can be used for framing sound asset pricing models and making global asset allocation decisions. They are also beneficial for the market participants in understanding the interaction of the stock markets of Middle East and African countries in relation to the crude oil market, as well as the global equity market uncertainties. In addition, our research could receive special attention from those investors who make use of new financial tools to hedge oil price volatility risk and are potentially interested in futures and option trading on OVX (Liu et al. [25]). The article proceeds along the following lines. The next section reviews the relevant literature. Data and their properties are presented in the third section. Section 4 outlines the research methodology we consider in our empirical investigation. In section 5, we discuss our findings. The last section contains conclusions. 2. Literature review A growing body of literature has examined the association between oil and stock markets. However, the findings of this literature are somewhat mixed. For example, Kling [23] and Papapetrou [30] find a negative relation between oil price changes and stock returns. Chen, Roll and Ross [12], on the other hand, report that stock returns are not affected by oil price fluctuations. More recently, however, Basher and Sadorsky [4] and Choi and Hammoudeh [14] identify a positive relation between oil price changes and stock returns. While the relationship between oil prices and stock returns has
1 Thus, we follow the approach, for example, by Blair et al. [6] and Kambouroudis and McMillan [21] that use VIX as a regressor in the GARCH variance equation.
been extensively analyzed, a strand of literature also examines the persistence and transmission of volatility from oil markets to stock markets. Malik and Hammoudeh [27] employ BEKKeGARCH(1,1) model to study the volatility and shock transmission mechanism among the U.S. equity market, the global crude oil market, and three Gulf equity markets that include Bahrain, Kuwait, and Saudi Arabia. In their empirical analyses, they document that Gulf equity markets are the receivers of volatility from the oil market. In addition, they report a significant volatility transmission from the stock market to the oil market only in Saudi Arabia. By applying an autoregressive conditional jump intensity model with structure changes, Chiou and Lee [13] report the existence of a negative and statistically significant impact of oil prices on stock returns. They also detect that the asymmetric effect has statistical significance only in a high-fluctuation state for both spot and futures oil price contracts. The authors further find that with changes in oil price dynamics, oil price volatility shocks have asymmetric effects on stock returns. Fowowe [18] also employs the GARCHjump model to investigate the relationship between oil prices and returns on the Nigerian Stock Exchange. The study reports a negative but insignificant effect of oil prices on stock returns in Nigeria. More recently, Ewing and Malik [17] use the univariate and bivariate GARCH models to examine the volatility of oil and US stock market prices incorporating structural breaks using daily data from July 1, 1996 to June 30, 2013. The authors do not report any evidence of volatility spillover between oil prices and US stock market when structural breaks in variance are ignored in the model. However, after accounting for structural breaks in the model, they report strong volatility association between the two markets. Bouri [9] has inspected the return and volatility linkages between oil prices and the Lebanese stock market by applying the VAR-GARCH model to weekly data from 30 January 1998 to 30 May 2014. The author finds some unidirectional return and volatility transmissions from oil prices to the Lebanese stock market. He also adds that the interrelationship between oil prices and Lebanese stocks increase during the crisis and tends to reduce significantly in the post-crisis period. Additionally, few other studies use sector indices to investigate the volatility transmission between oil and stock prices. Malik and Ewing [26], for example, estimate bivariate BEKKeGARCH (1,1) models to examine the volatility transmission between oil prices and five US sector indices. The sectors considered in their study include Financials, Industrials, Consumer Services, Health Care, and Technology. The paper finds that there exists significant transmission of shocks and volatility between oil prices and different stock market sectors. Moreover, Arouri et al. [2] make use of European equity sector indices to assess the volatility spillovers between oil and stock prices. The authors show that the volatility transmission between oil price and sector stock returns is significant. 3. Data In this paper, unlike the ones described in previous chapter, we use implied crude oil volatility index, OVX, instead of conventional oil price index, to estimate the effect of oil price uncertainty on the stock market volatilities. This is advantageous, because implied volatilities not only contain historical volatility information, but also investors' expectation of future market conditions. The CBOE publishes OVX index, from the middle of 2007, as a measure of expected 30-day volatility of crude oil prices. The OVX considers real-time bid/ask quotes of nearby and second nearby options with at least 8 days to expiration, and weights these
A. Dutta et al. / Energy 123 (2017) 189e197
options to derive a constant, a 30-day estimate of the expected volatility (Liu et al. [25]). Our daily data on stock prices consist of five Middle East and seven African stock market indices. The markets of Saudi Arabia, United Arab Emirates (UAE), Kuwait, Qatar and Jordan from Middle East, and Nigeria, Egypt, South Africa, Tunisia, Morocco, Kenya and Mauritius from Africa, presented in the order of their oil production, are included in the sample. We use Morgan Stanley Capital International (MSCI) country indices along with the information on OVX and VIX, which are all extracted from the Data Stream database. Our sample starts in May 10, 2007 and ends December 31, 2014, providing a total of 1995 observations. We choose this period because the OVX data are available from the commencing date of our sample. We use log returns which are measured in US dollars, so that we can conveniently compare them across the countries considered in our analyses.
191
Fig. 1 demonstrates the OVX and VIX indices for the whole sample period. The graph indicates several major spikes in both indices during the period under study. The first one appears in the latter part of 2008, when Lehman Brothers filed a bankruptcy, and both volatility indices increase rapidly. Interestingly, the hikes seem to be a consequence of either economic or political events. For example, the spike occurring in OVX during the beginning of 2011 can be attributed to the Libyan war for which the oil price uncertainty increases markedly (Liu et al. [25]). 4. Methodology 4.1. Modified GARCH (1,1) model The first step in our GARCH methodology is the specification of the mean equation. For each of the return series, the mean equation assumes the following form:
Rit ¼ pi þ wi Rit1 þ εit
3.1. Descriptive statistics Table 1, Panel A reports the descriptive statistics for each of the return series considered. The performance of these indices, measured by the daily mean returns, is highest in the Qatar stock market followed by stock markets in Kenya, Mauritius, Saudi Arabia, South Africa and Tunisia, during the sample period. We find negative mean returns for the remaining countries. The equity indices for South Africa, United Arab Emirates and Egypt show higher volatility than the rest of the indices. Most of the indices exhibit negative skewness, implying that large negative stock returns are more common than large positive returns in those markets, but the indices for Kuwait, Tunisia, Kenya and Mauritius are positively skewed. In terms of kurtosis, all the return series are leptokurtic, which is a common phenomenon for financial time series data, hence the GARCH process, proposed by Bollerslev [7], is appropriate (Malik and Hammoudeh [27]). Moreover, the Jarque-Bera test rejects the null hypothesis that all the return series are normally distributed. Panel B of Table 1 displays the descriptive statistics of the OVX and VIX indices. The standard deviations of these two indices indicate that the fluctuations in OVX are more severe than in VIX. In addition, both these indices are positively skewed. The results further show that the indices have kurtosis higher than 3, implying that each volatility series have leptokurtic distribution with asymmetric tails.
(1)
Within this framework, Rit denotes the log return on index i between time t and t-1, pi is a long-term drift coefficient and εit refers to the error term for the return on series i at time t. The GARCH model, proposed by Engle [16] and Bollerslev [7], formulates the conditional variance, ht , as a function of past errors and the lagged conditional variances leaving the unconditional variance constant. In order to examine effect of oil price uncertainty, measured by OVX, on the volatility of different equity markets considered, we propose to estimate the following extended GARCH variance equation: 2 ht ¼ u þ aε2t1 þ bht1 þ gOVXt1
(2)
where u > 0, a 0, b 0 and g 0 to guarantee the positivity of ht and εt ¼ zt ht ; zt i.i.d. (0,1). For a nonzero value of g, we report that the volatility of the corresponding equity market is significantly affected by the crude oil price uncertainty. A number of articles (e.g. Blair, Poon and Taylor [6]; Kambouroudis and McMillan [21] and others) have used implied volatility index as a regressor in the GARCH variance equation. Blair et al. [6], for instance, report that the VIX index provides the most accurate projections of conditional variance for all forecast horizons and performance measures considered in their analyses. Kambouroudis and McMillan [21] also argue that the inclusion of VIX as a regressor in the GARCH variance equation improves the overall performance
Table 1 Descriptive statistics of the stock returns. Countries Panel A: Return Indices Saudi Arabia United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius Panel B: Volatility Indices OVX VIX
Mean
Standard deviation
Skewness
Kurtosis
JarqueeBera
0.000074 0.000026 0.000341 0.000282 0.000442 0.000260 0.000006 0.000032 0.000033 0.000292 0.000250 0.000149
0.0149 0.0194 0.0138 0.0148 0.0119 0.0114 0.0102 0.0177 0.0198 0.0145 0.0129 0.01225
1.13 0.56 0.10 0.72 0.50 0.19 0.15 0.96 0.21 0.31 0.10 0.05
21.58 16.76 15.75 19.30 12.38 6.97 10.16 12.09 7.90 6.83 14.32 16.54
29120.79*** 15847.49*** 13918.86*** 22266.59*** 7396.37*** 1323.13*** 4270.12*** 7189.19*** 2012.04*** 1256.43*** 10659.93*** 15254.48***
35.778140 22.227120
14.50000 10.32000
1.614684 2.115989
8.62 6.32
4120.132*** 1783.254***
Notes: This table reports the main descriptive statistics of the daily return indices used. Panel A contains the results for the return indices, while Panel B shows the outcomes for the volatility indices. Jarque-Bera test rejects the null hypothesis of normality at 1% level of significance.
192
A. Dutta et al. / Energy 123 (2017) 189e197
120 100 80 60 40 20 0 2007
2008
2009
2010 VIX
2011
2012
2013
2014
OVX
Fig. 1. OVX and VIX from 10/05/2007-31/12/2014.
of their employed models. Moreover, with a view to investigating the association between oil price uncertainty and stock market volatility after controlling for the effect of implied volatility index of S & P 500, we consider estimating another extended form of GARCH process as given below: 2 2 ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1
(3)
where u > 0, a 0, b 0, g 0 and q 0 to guarantee that ht is positive. In model (3), a statistically significant value of the parameter q implies that there exists a direct connection between the US equity market uncertainty and the stock market volatility of the countries considered in our empirical inspection. Now, in order to examine the potential OVX-stock market volatility relationship, when extreme movements occur in the implied volatility index, we estimate the following GARCH equation: 2 ht ¼ u þ aε2t1 þ bht1 þ qVIXt1 þ
2 X
2 di Di OVXt1
(4)
i¼1
where u > 0, a 0, b 0, q 0 and di 0 to guarantee that ht is positive. Also,
2 2 ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1 þ ls2t1
where u > 0, a 0, b 0, g 0, q 0 and l > 0 to guarantee that ht is positive. In model (5), the term s2t1 indicates the regional market volatility effect at period t-1. Now to obtain s2t , we first find c ht by estimating a univariate GARCH (1,1) process for all the markets under study. We then have a matrix consisting of twelve columns, where each column contains the time series observations of c ht of a specific market. In the next step, we perform a principal component analysis to obtain a column vector consisting of the values of s2t . Finally, in order to check the robustness of our findings, we also present the results from exponential GARCH (EGARCH) model. The EGARCH (1,1) model, proposed by Nelson [29], takes the following form:
lnðht Þ ¼ c þ
ajεt1 j þ vεt1 pffiffiffiffiffiffiffiffiffiffi þ b lnðht1 Þ ht1
(6)
In model (11), when εti is positive or there is good news, the total effect of εt1 is ð1 þ vÞ jεt1 j. In contrast, when εti is negative or there is bad news, the total impact of εt1 is ð1 vÞ jεt1 j. Now to observe the effects of implied volatility indices, VIX and OVX, equation (6) can be extended as follows:
lnðht Þ ¼ c þ D1 ¼ 1; if OVX < Q1 ; 0; otherwise;
(5)
ajεt1 j þ vεt1 2 2 pffiffiffiffiffiffiffiffiffiffi þ b lnðht1 Þ þ gOVXt1 þ qVIXt1 ht1 (7)
and
D2 ¼ 1; if OVX < Q3 ; 0; otherwise; 4.2. GARCHejump model where Q1 and Q3 indicate the first and third quartiles of the variable OVX. The next step is to examine whether the oil market uncertainty has any potential impacts on different equity markets in the presence of regional market volatility factor, we estimate the following multivariate GARCH model:
We advance our analysis by considering the application of the GARCH-jump model, proposed by Chan and Maheu [11], to examine whether the stock returns of the sampled markets are sensitive to high fluctuations occurring in the implied oil volatility index. Such methodology provides a beneficial approach to further assess the
A. Dutta et al. / Energy 123 (2017) 189e197
robustness of our results, since unlike the traditional GARCH models, it can capture the effects of extreme news or abnormal information arising from earnings surprises, crashes, terrorist attacks and similar other events (Fowowe [18]). Our estimated model assumes the following form2:
Rt ¼ t þ m1 Rt1 þ m2 Rt2 þ jROt1 þ εt
Table 2 Estimates of OVX effects on the volatility of selected stock markets. Countries
u
a
b
g
Saudi Arabia
0.038** (0.05) 0.032*** (0.00) 0.008** (0.01) 0.006*** (0.00) 0.050*** (0.00) 0.135*** (0.00) 0.103*** (0.00) 0.105*** (0.00) 0.049** (0.04) 0.085*** (0.00) 0.095*** (0.00) 0.005 (0.27)
0.061*** (0.00) 0.074*** (0.00) 0.061*** (0.00) 0.062*** (0.00) 0.090*** (0.00) 0.114*** (0.00) 0.188*** (0.00) 0.068*** (0.00) 0.074*** (0.00) 0.197*** (0.00) 0.240*** (0.00) 0.153*** (0.00)
0.925*** (0.00) 0.890*** (0.00) 0.906*** (0.00) 0.923*** (0.00) 0.784*** (0.00) 0.588*** (0.00) 0.618*** (0.00) 0.793*** (0.00) 0.872*** (0.00) 0.707*** (0.00) 0.570*** (0.00) 0.758*** (0.00)
0.000535*** (0.00) 0.000020** (0.03) 0.000010*** (0.00) 0.000001 (0.32) 0.000040*** (0.00) 0.000100*** (0.00) 0.000040*** (0.00) 0.000100*** (0.00) 0.000105** (0.01) 0.000010*** (0.00) 0.000100*** (0.00) 0.000010*** (0.00)
United Arab Emirates
(8) Kuwait
where Rt denotes the log return of stock prices between time t and t-1, ROt1 is the log return on the OVX series at time t-1 and εt refers to the error term at time t which has two components as follows:
Qatar Jordan Morocco
εt ¼ ε1t þ ε2t
(9) Tunisia
The first component ε1t is a mean-zero innovation with normal stochastic process assuming the following form:
ε1t ¼
pffiffiffiffiffi ht zt ;
Egypt South Africa
zt NIDð0; 1Þ
Nigeria 0
ht ¼ u0 þ a0 ε21t1 þ b ht1
(10)
The second component ε2t is a jump innovation which consists of abnormal price movements with Eðε2t jIt1 Þ ¼ 0, where It1 denotes the information set. Now ε2t is defined as the discrepancy between the jump component and the expected total jump size between t-1 and t:
ε2t ¼
nt X
utl kmt
193
Kenya Mauritius
Notes: This table presents the estimates of the oil volatility index (OVX) effects on the current stock market volatility for different countries under study. These results 2 : are obtained by estimating the following model: ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 Values in the parentheses indicate the p-values. ** and *** represent significance at 5% and 1% levels, respectively.
(11)
l¼1
where utl refers to the jump size and is assumed to be normally Pnt distributed with mean k and variance d2 , u is the jump l¼1 tl component and nt designates the number of jumps. It is assumed that nt is distributed as a Poisson variable with an autoregressive conditional jump intensity (ARJI) given by
mt ¼ m0 þ rmt1 þ 4xt1
(12)
where mt is the time-varying conditional jump intensity parameter and mt > 0, m0 > 0, r > 0 and 4 > 0. Now the log-likelihood function can be expressed as:
LðUÞ ¼
T X
log f ðrt jIt1 ; UÞ
t¼1 0
0
0
where U ¼ ðt; m1 ; m2 ; j; u ; a ; b ; k; d; m0 ; r; 4Þ. 5. Empirical results 5.1. Findings of the modified GARCH (1,1) model In this section, we analyze our findings to examine whether the oil market uncertainty affects the volatility of different stock markets considered, and second, whether oil market uncertainty impacts on various equity markets in the presence of regional market volatility factor. To find out the effect of oil market uncertainty on volatility on each stock market, we estimate the extended GARCH model, presented in equation (2), for twelve different countries. The results for each of the variance equations are reported in Table 2. The
2 Fowowe [18] employs a similar mean equation to address the impact of oil price fluctuations on Nigerian stock market returns. Fowowe uses traditional oil price series in the jump model, while we consider the oil volatility index.
hypothesis that expected volatility is transmitted from the oil 2 , market to the stock market is supported if the coefficient of OVXt1 g, is significantly different from zero. Table 2 indicates that the implied volatility in stock returns for each country, except Qatar, is affected by the oil market implied 2 . volatility, as evidenced by the significant coefficients on OVXt1 The estimated coefficient for Saudi Arabia seems to be higher than that for other markets, implying that the global oil market plays a pivotal role in the Saudi Arabian economy. The outcomes of Table 2 further reveal that the coefficients of ε2t1 and ht1 are statistically significant for all of the countries under investigation. This suggests that the volatility of stock return is influenced by its own volatility. Furthermore, the positive values of the coefficient b indicate that higher levels of conditional volatility in the past are associated with higher conditional volatility in the current period. Table 3 documents the empirical results estimated from the variance equation (3). These results enable us to assess the connection between oil and equity markets after controlling for the effect of implied stock market volatility index (VIX). The outcomes reveal that, although VIX plays as a moderator in the adopted 2 , g, are still statistically model, the estimated coefficients of OVXt1 significant for almost half of the markets analyzed. 2 Additionally, we find significant results for the variable VIXt1 for nine out of the twelve stock markets under study, which indicates strong links between the US equity market uncertainty and the implied volatility of those nine equity markets. Moreover, the values of the coefficients for both volatility indices are significant for four markets: Saudi Arabia, Kuwait, Tunisia and Mauritius. For example, in case of Kuwaiti market, the parameter estimates for 2 2 , suggesting that the OVXt1 is twice as high as that for VIXt1 Kuwaiti market volatility is two times more sensitive to the oil market uncertainty than to US stock market uncertainty. The results of Table 3 further reveal that the coefficients of ε2t1 and ht1 , a and b respectively, are still statistically significant (at 1% level) for all the stock markets except for South Africa.
194
A. Dutta et al. / Energy 123 (2017) 189e197
Table 3 Estimates of OVX and VIX effects on the volatility of selected stock markets. Countries
u
a
b
g
q
Saudi Arabia
0.065*** (0.00) 0.031*** (0.00) 0.009*** (0.00) 0.006*** (0.00) 0.053*** (0.00) 0.146*** (0.00) 0.105*** (0.00) 0.121*** (0.00) 0.239 (0.10) 0.086*** (0.00) 0.099*** (0.00) 0.010 (0.22)
0.057*** (0.00) 0.076*** (0.00) 0.058*** (0.00) 0.062*** (0.00) 0.092*** (0.00) 0.115*** (0.00) 0.191*** (0.00) 0.071*** (0.00) 0.002 (0.91) 0.198*** (0.00) 0.239*** (0.00) 0.180*** (0.00)
0.929*** (0.00) 0.885*** (0.00) 0.913*** (0.00) 0.923*** (0.00) 0.770*** (0.00) 0.567*** (0.00) 0.609*** (0.00) 0.756*** (0.00) 0.260 (0.36) 0.709*** (0.00) 0.550*** (0.00) 0.650*** (0.00)
0.000496*** (0.00) 0.000006 (0.59) 0.000020*** (0.00) 0.000001 (0.32) 0.000004 (0.70) 0.000091** (0.03) 0.000082*** (0.00) 0.000070* (0.06) 0.000080 (0.61) 0.000044 (0.11) 0.000019 (0.46) 0.000021*** (0.00)
0.000675*** (0.00) 0.000076** (0.03) 0.000010*** (0.00) 0.000028* (0.06) 0.000100*** (0.00) 0.000100* (0.07) 0.000089** (0.02) 0.000273*** (0.00) 0.004219** (0.01) 0.000025 (0.62) 0.000190*** (0.00) 0.000394*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: This table displays the estimates of the oil volatility index (OVX) effects on the current stock market volatility after controlling for the effect of US equity market 2 2 : Values in the parentheses indicate the p-values. *, uncertainty (VIX). We find these outcomes by estimating the following model: ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1 ** and *** represent significance at 10%, 5% and 1% levels, respectively.
To examine the impact of oil market uncertainty on the equity markets in the presence of regional market volatility factor, we estimate equation (5) for all the markets considered and display the results in Table 5. The findings reveal that in the presence of regional market volatility factor, the oil price uncertainty has significant impacts on half of the equity markets considered. Additionally, the VIX factor is found to be significant for most of the countries implying that shocks still spill over from the US equity
Table 4 presents the results on equation (4) for each market. The findings indicate that for Saudi Arabia, United Arab Emirates, Qatar, Jordan and Morocco, the corresponding volatility in returns is significantly affected by the higher values of the oil volatility index. However, when the oil price uncertainty is low, only Kenyan equity market significantly reacts to this change. As an additional analysis, we also perform similar analysis for oil price levels and find consistent results.
Table 4 Estimates of OVX effects with extreme movements. Countries
u
a
b
q
d1
d2
Saudi Arabia
0.014 (0.73) 0.077*** (0.00) 0.010 (0.29) 0.020*** (0.00) 0.168*** (0.00) 0.014 (0.76) 0.092*** (0.00) 0.129*** (0.00) 0.040 (0.87) 0.089* (0.05) 0.049 (0.27) 0.015 (0.39)
0.044*** (0.00) 0.071*** (0.00) 0.058*** (0.00) 0.059*** (0.00) 0.091*** (0.00) 0.106*** (0.00) 0.186*** (0.00) 0.071*** (0.00) 0.006 (0.77) 0.198*** (0.00) 0.235*** (0.00) 0.181*** (0.00)
0.931*** (0.00) 0.885*** (0.00) 0.913*** (0.00) 0.924*** (0.00) 0.708*** (0.00) 0.568*** (0.00) 0.615*** (0.00) 0.754*** (0.00) 0.249 (0.36) 0.702*** (0.00) 0.533*** (0.00) 0.644*** (0.00)
0.000646*** (0.00) 0.000061* (0.09) 0.000024* (0.07) 0.000009 (0.19) 0.000128*** (0.00) 0.000201** (0.02) 0.000078** (0.04) 0.000275*** (0.00) 0.004238*** (0.00) 0.000021 (0.69) 0.000181** (0.01) 0.000404*** (0.00)
0.000033 (0.61) 0.000005 (0.87) 0.000003 (0.84) 0.000013 (0.10) 0.000049 (0.23) 0.000157 (0.10) 0.000023 (0.66) 0.000009 (0.90) 0.000670 (0.14) 0.000019 (0.78) 0.000199** (0.01) 0.000011 (0.69)
0.000318*** (0.00) 0.000061*** (0.00) 0.000001 (0.93) 0.000011*** (0.00) 0.000115*** (0.00) 0.000124*** (0.00) 0.000026 (0.26) 0.000005 (0.88) 0.000028 (0.88) 0.000013 (0.69) 0.000024 (0.43) 0.000002 (0.88)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: This table describes how the selected equity markets respond when extreme movements occur in the oil volatility index. Doing so considers two dummy variables. We define these dummies as follows:D1 ¼ 1, if OVX < Q1 ; 0, otherwise and D2 ¼ 1, if OVX > Q3 ; 0, otherwise, where Q1 and Q3 indicate the first and third quartiles of the variable P 2 2 : Values in the parentheses indicate the p-values. *, ** OVX. The following model has been estimated to obtain these results: ht ¼ u þ aε2t1 þ bht1 þ qVIXt1 þ 2i¼1 di Di OVXt1 and *** represent significance at 10%, 5% and 1% levels, respectively.
A. Dutta et al. / Energy 123 (2017) 189e197
195
Table 5 Estimates of regional market volatility effects. Countries
u
a
b
g
q
l
Saudi Arabia
0.049*** (0.00) 0.065** (0.02) 0.021** (0.01) 0.010 (0.15) 0.083*** (0.00) 0.143** (0.01) 0.135*** (0.00) 0.084* (0.06) 0.770* (0.05) 0.042* (0.08) 0.105** (0.01) 0.072** (0.03)
0.055*** (0.00) 0.074*** (0.00) 0.059*** (0.00) 0.059*** (0.00) 0.092*** (0.00) 0.114*** (0.00) 0.190*** (0.00) 0.073*** (0.00) 0.012 (0.56) 0.201*** (0.00) 0.238*** (0.00) 0.218*** (0.00)
0.930*** (0.00) 0.882*** (0.00) 0.906*** (0.00) 0.924*** (0.00) 0.758*** (0.00) 0.574*** (0.00) 0.598*** (0.00) 0.758*** (0.00) 0.099 (0.75) 0.705*** (0.00) 0.547*** (0.00) 0.459*** (0.00)
0.000203*** (0.00) 0.000014 (0.30) 0.000019*** (0.00) 0.000009*** (0.00) 0.000001 (0.90) 0.000089** (0.03) 0.000081*** (0.00) 0.000076** (0.04) 0.000024 (0.90) 0.000047* (0.08) 0.000017 (0.55) 0.000020 (0.42)
0.000308*** (0.00) 0.000066* (0.06) 0.000029** (0.04) 0.000014** (0.01) 0.000091*** (0.00) 0.000149* (0.07) 0.000120** (0.01) 0.000303*** (0.00) 0.004630*** (0.00) 0.000089 (0.10) 0.000189** (0.01) 0.000560*** (0.00)
0.443950*** (0.00) 0.017581 (0.20) 0.007520 (0.12) 0.002190 (0.60) 0.014887* (0.08) 0.000555 (0.97) 0.014032 (0.22) 0.019492 (0.25) 0.263737 (0.10) 0.025445** (0.01) 0.003852 (0.81) 0.056397*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: The numbers shown in this table are used to examine the effects of oil market uncertainty on considered equity markets in the presence of regional market volatility 2 2 factor (s2t ). To obtain these results, we estimate the model given by: ht ¼ u þ aε2t1 þ bht1 þ gOVXt1 þ qVIXt1 þ ls2t1 . Values in the parentheses indicate the p-values. *, ** and *** represent significance at 10%, 5% and 1% levels, respectively.
robustness test using the EGARCH model. These findings, given in Table 6, are consistent with those reported in Table 3 with one exception. The GARCH (1,1) specification shows that crude oil volatility index plays a potential role in forecasting the stock market volatility of Mauritius, while we do not report such significant result using the EGARCH (1,1) approach. Since all other outcomes are analogous in Tables 3 and 6, we thus conclude that our results are robust to various changes in the specification of the adopted GARCH model. Furthermore, we divide our sample period into two subsamples
market to the other stock markets. Moreover, there is significant association between second moments of regional market and equity markets of Saudi Arabia, Kenya and Mauritius. Thus, the stock market volatility of most of the countries is not impacted by the regional market volatility. We further document that only for Saudi Arabia, the global oil market, US equity market and the stock markets of other neighboring states play crucial roles, suggesting that the Saudi economy is significantly influenced by all these factors. In order to check the sensitivity of our findings, we conduct a
Table 6 Estimates of EGARCH model. Countries
c
a
b
v
g
q
Saudi Arabia
1.23*** (0.00) 1.11*** (0.00) 1.49*** (0.00) 0.13*** (0.00) 0.21*** (0.00) 0.30*** (0.00) 0.35*** (0.00) 0.06*** (0.00) 0.06*** (0.00) 0.28*** (0.00) 0.39*** (0.00) 0.29*** (0.00)
0.184*** (0.00) 0.168*** (0.00) 0.166*** (0.00) 0.167*** (0.00) 0.205*** (0.00) 0.246*** (0.00) 0.348*** (0.00) 0.130*** (0.00) 0.103*** (0.00) 0.360*** (0.00) 0.392*** (0.00) 0.310*** (0.00)
0.891*** (0.00) 0.966*** (0.00) 0.675*** (0.00) 0.988*** (0.00) 0.893*** (0.00) 0.706*** (0.00) 0.845*** (0.00) 0.885*** (0.00) 0.962*** (0.00) 0.883*** (0.00) 0.814*** (0.00) 0.937*** (0.00)
0.125*** (0.00) 0.061*** (0.00) 0.011 (0.50) 0.024*** (0.00) 0.035*** (0.00) 0.034 (0.18) 0.008 (0.67) 0.114*** (0.00) 0.117*** (0.00) 0.007 (0.73) 0.008 (0.71) 0.011 (0.43)
0.000449*** (0.00) 0.000004 (0.73) 0.000640*** (0.00) 0.000005 (0.23) 0.000001 (0.94) 0.000049** (0.03) 0.000038*** (0.00) 0.000018* (0.05) 0.000003 (0.63) 0.000023 (0.17) 0.000006 (0.71) 0.000001 (0.88)
0.000354*** (0.00) 0.000010** (0.02) 0.000622*** (0.00) 0.000043* (0.08) 0.000074*** (0.00) 0.000078* (0.05) 0.000031** (0.04) 0.000023*** (0.00) 0.000021** (0.03) 0.000021 (0.33) 0.000132*** (0.00) 0.000065*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
jþvεt1 2 2 : Values in the parentheses indicate the ppffiffiffiffiffiffiffi Notes: This table displays the estimates of the EGARCH model given as: lnðht Þ ¼ c þ ajεt1 þ qVIXt1 þ b lnðht1 Þ þ gOVXt1 ht1 values. *, ** and *** represent significance at 10%, 5% and 1% levels, respectively.
196
A. Dutta et al. / Energy 123 (2017) 189e197
volatility of stock returns, the jump model now confirms that OVX affects the mean returns as well. It is also evident from Table 7 that the jump parameters are generally significant implying that jumps do exist in the stock returns and they are time-varying. The positive (negative) coefficient of the jump mean indicates that the jump behavior driven by abnormal information has a positive (negative) impact on returns, while the positive (negative) coefficient of the jump variance infers that volatility driven by abnormal information has a positive (negative) effect on volatility of returns (Fowowe [18]). The findings also document that the jump intensity parameters (m0 ; r; 4) are mostly statistically significant suggesting that the jump intensity varies over time (Chiou and Lee [13]). Additionally, since these parameters satisfy the constraints that m0 > 0, r > 0 and 4 > 0, it can be concluded that the GARCH-ARJI model is appropriate for describing the jump behavior in the stock market returns. Furthermore, the positive r and 4 suggest that the current jump intensity (mt ) is affected by the most recent jump intensity (mt1 ) and intensity residuals (xt1 ). We further find that the high values of r and 4 indicate a high degree of persistence in the jump intensity.
with equal number of observations. These results indicate some variation in terms of the statistical significances, in particular, which may be due to the substantially reduced sample sizes.3 Our results therefore collectively suggest that global oil and stock market uncertainties, measured by OVX and VIX, impact the stock market volatility of several Middle East and African countries. The association is clearly strongest in the case of Saudi Arabia, which is the most important oil exporting country. 5.2. Findings of the GARCH-Jump model Table 7 displays the results of the autoregressive conditional jump intensity model. We find that the GARCH parameters are statistically significant at 1% level indicating the existence of ARCH 0 0 and GARCH effects. The sum of a and b also implies high degree of persistence in the return fluctuations. Our findings further reveal that stock returns of majority of the sampled markets are affected when changes occur in the crude oil volatility index. To be specific, we find significant oil-stock links for the following countries: Qatar, Jordan, Nigeria, Egypt, South Africa, Tunisia, and Mauritius. It is unexpected that the Saudi, UAE and Kuwaiti stock markets are not affected by the variations in oil price volatility. However, these findings are consistent with those reported by Arouri et al. [3] and Hammoudeh and Choi [19]. Arouri et al., for example, argue that such an insignificant oil-stock link can be explained by the lower reliance of stock markets on oil companies (Kuwait), and the low annual turnover (Saudi Arabia and UAE). Hammoudeh and Choi also document that Kuwaiti market is highly sensitive to fads and herding which could limit its dependence on the oil market. We further report that all the statistically significant coefficients are found to be negative. Such outcomes are not surprising, since a rise in the oil price uncertainty may cause the stock prices to decrease. Thus in addition to the modified GARCH model, which documents that oil price uncertainty, measured by OVX, impacts the realized
6. Conclusions This paper examines whether the implied crude oil volatility index (OVX), a forward-looking measure of oil market uncertainty published by Chicago Board Options Exchange (CBOE), impacts the realized volatility of Middle East and African stock markets. The data include daily stock index observations from Saudi Arabia, United Arab Emirates, Kuwait, Qatar, Jordan, Morocco, Tunisia, Egypt, South Africa, Nigeria, Kenya and Mauritius. The study contributes to the standing literature mainly in two aspects. First, while earlier studies such as Malik and Ewing [26], Lee and Chiou [24], Arouri et al. [1,2] and Bouri [9,10] utilize the
Table 7 Estimated results of GARCH-Jump models. Countries
t
m1
m2
j
u0
a0
b0
k
d2
m0
r
4
Saudi Arabia
0.00011 (0.24) 0.00903*** (0.00) 0.02095 (0.11) 0.03244** (0.01) 1.1091*** (0.00) 0.02797 (0.18) 0.01645 (0.36) 0.00004 (0.86) 0.06612 (0.07)* 0.01080 (0.64) 0.08186 (0.00)*** 0.00808 (0.57)
0.00019 (0.35) 0.01034*** (0.00) 0.01076 (0.60) 0.06227*** (0.00) 0.11254 (0.24) 0.05958*** (0.00) 0.03745 (0.11) 0.00010*** (0.00) 0.00709 (0.73) 0.31975*** (0.00) 0.26243*** (0.00) 0.03749* (0.07)
0.00009 (0.31) 0.01646*** (0.00) 0.01346 (0.49) 0.03039 (0.16) 0.11571 (0.16) 0.03531 (0.10) 0.00698 (0.76) 0.00008 (0.76) 0.03965* (0.07) 0.02549 (0.26) 0.07296*** (0.00) 0.07174*** (0.00)
0.00002 (0.29) 0.00010 (0.80) 0.00045 (0.90) 0.01295*** (0.00) 0.04291*** (0.00) 0.00658 (0.10) 0.00610* (0.08) 0.00003* (0.09) 0.02129*** (0.00) 0.01235*** (0.00) 0.00295 (0.46) 0.00655** (0.03)
0.030*** (0.00) 0.034*** (0.00) 0.070*** (0.00) 0.017*** (0.00) 1.08*** (0.00) 0.005 (0.12) 0.133*** (0.00) 0.109*** (0.00) 0.007** (0.02) 0.200*** (0.00) 0.144*** (0.00) 0.025*** (0.00)
0.062*** (0.00) 0.063*** (0.00) 0.059*** (0.00) 0.085*** (0.00) 0.140*** (0.00) 0.023*** (0.00) 0.155*** (0.00) 0.057*** (0.00) 0.004* (0.06) 0.167*** (0.00) 0.208*** (0.00) 0.148*** (0.00)
0.921*** (0.00) 0.832*** (0.00) 0.668*** (0.00) 0.740*** (0.00) 0.858*** (0.00) 0.957*** (0.00) 0.529*** (0.00) 0.908*** (0.00) 0.988*** (0.00) 0.340** (0.01) 0.442*** (0.00) 0.759*** (0.00)
0.006 (0.77) 0.029 (0.29) 0.082 (0.18) 0.068 (0.52) 13.68*** (0.00) 0.048 (0.70) 0.239* (0.08) 0.004 (0.99) 0.175* (0.08) 0.032 (0.31) 0.311*** (0.00) 0.186 (0.31)
1.16*** (0.00) 1.45*** (0.00) 1.42*** (0.00) 2.05*** (0.00) 5.31*** (0.00) 1.51*** (0.00) 1.56*** (0.00) 1.32*** (0.00) 1.81*** (0.00) 1.07*** (0.00) 1.34*** (0.00) 2.08*** (0.00)
0.013*** (0.00) 0.159*** (0.00) 0.008*** (0.00) 0.238*** (0.00) 0.022*** (0.00) 0.109 (0.14) 0.001 (0.20) 0.063*** (0.00) 0.010** (0.03) 0.027** (0.04) 0.003 (0.10) 0.086*** (0.00)
0.988*** (0.00) 0.920*** (0.00) 0.992*** (0.00) 0.697*** (0.00) 0.706*** (0.00) 0.394 (0.32) 0.990*** (0.00) 0.964*** (0.00) 0.977*** (0.00) 0.970*** (0.00) 0.996*** (0.00) 0.677*** (0.00)
0.069*** (0.00) 0.229*** (0.00) 0.227*** (0.00) 0.132*** (0.00) 0.881*** (0.00) 0.583** (0.01) 0.076** (0.03) 0.066*** (0.00) 0.667*** (0.00) 0.419*** (0.00) 0.114** (0.01) 0.126*** (0.00)
United Arab Emirates Kuwait Qatar Jordan Morocco Tunisia Egypt South Africa Nigeria Kenya Mauritius
Notes: Values in the parentheses indicate the p-values. *, ** and *** represent significance at 10%, 5% and 1% levels, respectively.
3
Results are available from the authors.
crude oil price series to model the volatility, we extend their works by assessing the association between realized volatilities of sample
A. Dutta et al. / Energy 123 (2017) 189e197
stock markets and the global oil price uncertainty measured by OVX. This can be considered as an important contribution, since OVX, in contrast to traditional oil price indices, contains both historical volatility information as well as investors' expectation of future market conditions and hence turns out to be a good indicator of oil market uncertainty [25]. Second, our paper also contributes to the existing literature on uncertainty transmission by examining whether global oil market uncertainty impacts the realized volatility of the Middle East and African stock markets. At the empirical stage, we apply a modified form of the generalized autoregressive conditional heteroskedasticity (GARCH) model to examine the relationship between expected oil market volatility and the realized volatility of different equity markets under study. In addition, we combine OVX with VIX to observe their joint effects on the stock markets. The findings of our analysis show significant links between oil price uncertainty and the realized stock market volatility. We also find that such associations seem to hold even after controlling for the effect of global stock market uncertainty. We further report that, conditioned on the past information, the volatility in each market is persistent, which suggests that the current volatility of stock returns for a specific market is directly affected by its past volatility. In addition, we employ the GARCH-Jump mode to investigate whether fluctuations in the implied oil volatility index have any major impacts on the stock returns. The findings indicate that such association is significant for majority of the sampled markets and that time-varying jumps do exist in the stock returns. To sum up, our results suggest that the implied crude oil volatility index, i.e. the market participants' anticipation of the future oil market uncertainty is an important factor explaining the returns and volatilities of the Middle East and African stock markets. Acknowledgements Editor Isabel Soares and two anonymous reviewers are gratefully acknowledged for their valuable comments, as well as the participants of the Multinational Finance Society conference in 2016 for their suggestions. Jussi Nikkinen acknowledges the generous financial support from the OP Group Research Foundation. All the remaining errors are solely on the authors' responsibility. References [1] Arouri M, Jouini J, Nguyen D. Volatility spillovers between oil prices and stock sector returns: implications for portfolio management. J Int Money Finance 2011;30:1387e405. [2] Arouri M, Jouini J, Nguyen D. On the impacts of oil price fluctuations on European equity markets: volatility spillover and hedging effectiveness. Energy Econ 2012;34:611e7. [3] Arouri M, Lahiani A, Nguyen D. Return and volatility transmission between world oil prices and stock markets of the GCC countries. Econ Model 2011;28:
197
1815e25. [4] Basher SA, Sadorsky P. Oil price risk and emerging stock markets. Glob Finance J 2006;17:224e51. [5] Bernanke BS. Irreversibility, uncertainty, and cyclical investment. Q J Econ 1983;98:85e106. [6] Blair BJ, Poon SH, Taylor SJ. Forecasting S&P 100 volatility: the incremental information content of implied volatilities and high-frequency index returns. J Econ 2001;105:5e26. [7] Bollerslev T. Generalised autoregressive conditional heteroskedasticity. J Econ 1986;31:307e27. [8] Bollerslev T, Osterrieder D, Sizova N, Tauchen G. Risk and return: long-run relationships, fractional cointegration, and return predictability. J Financ. Econ 2013;108:409e24. [9] Bouri E. Return and volatility linkages between oil prices and the Lebanese stock market in crisis periods. Energy 2015a;89:365e71. [10] Bouri E. Oil volatility shocks and the stock markets of oil-importing MENA economies: a tale from the financial crisis. Energy Econ 2015b;51:590e8. [11] Chan WH, Maheu JM. Conditional jump dynamics in stock market returns. J Bus Econ Stat. 2002;20:377e89. [12] Chen NF, Roll R, Ross SA. Economic forces and the stock market. J Bus 1986;59: 383e403. [13] Chiou JS, Lee YH. Jump dynamics and volatility: oil and the stock markets. Energy 2009;34:788e96. [14] Choi K, Hammoudeh S. Volatility behavior of oil, industrial commodity and stock markets in a regime switching environment. Energy Policy 2010;38: 4388e99. [15] Ciner C. Oil and stock returns: frequency domain evidence. J Int Financ. Mark Inst. Money 2013;23:1e11. [16] Engle R. Autoregressive conditional heteroscedasticity with estimates of the variance of the U. K. inflation. Econometrica 1982;50:987e1008. [17] Ewing BT, Malik F. Volatility spillovers between oil prices and the stock market under structural breaks. Glob Finance J 2016;29:12e23. [18] Fowowe B. Jump dynamics in the relationship between oil prices and the stock market: evidence from Nigeria. Energy 2013;56:31e8. [19] Hammoudeh S, Choi K. Behavior of GCC stock markets and impacts of US oil and financial markets. Res Int Bus Finance 2006;20:22e44. r P, Westgaard S. Forecasting volatility of the [20] Haugom E, Langeland H, Molna U.S. oil market. J Bank Finance 014; 47: 1e14. [21] Kambouroudis DS, McMillan DG. Does VIX or volume improve GARCH volatility forecasts? Appl Econ 2016;48:1e19. [22] Killian L, Park C. The impact of price shocks on the U.S. stock market. Int Econ Rev 2009;50:1267e87. [23] Kling JL. Oil price shocks and stock market behavior. J Portf. Manag 1985;12: 34e9. [24] Lee YH, Chiou JS. Oil price sensitivity and its asymmetric impact on the stock market. Energy 2011;36:168e74. [25] Liu ML, Ji Q, Fan Y. How does oil market uncertainty interact with other markets: an empirical analysis of implied volatility index? Energy 2013;55: 860e8. [26] Malik F, Ewing BT. Volatility transmission between oil prices and equity sector returns. Int Rev Financ. Anal. 2009;18:95e100. [27] Malik S, Hammoudeh S. Shock and volatility transmission in the oil, US and Gulf equity markets. Int Rev Econ Finance 2007;17:357e68. €m P. International transmission of uncertainty implicit in [28] Nikkinen J, Sahlstro stock index option prices. Glob Finance J 2004;15:1e15. [29] Nelson DB. Conditional heteroskedasticity in asset returns: a new approach. Econometrica 1991;59:347e70. [30] Papapetrou E. Oil price shocks, stock market, economic activity and employment in Greece. Energy Econ 2001;23:511e32. [31] Pindyck R. Irreversibility, uncertainty, and investment. J Econ Lit. 1991;29: 110e48. [32] Sadorsky P. Oil price shocks and stock market activity. Energy Econ 1999;21: 449e69. [33] Wang K. Volatility linkages of the equity, bond and money markets: an implied volatility approach. Account Finance 2009;49:207e19.