- Email: [email protected]
Hedging stocks with oil
Accepted Manuscript Hedging stocks with oil
Jonathan A. Batten, Harald Kinateder, Peter G. Szilagyi, Niklas F. Wagner PII: DOI: Reference:
S0140-9883(19)30191-4 https://doi.org/10.1016/j.eneco.2019.06.007 ENEECO 4422
To appear in:
Energy Economics
Received date: Revised date: Accepted date:
2 October 2018 14 April 2019 10 June 2019
Please cite this article as: J.A. Batten, H. Kinateder, P.G. Szilagyi, et al., Hedging stocks with oil, Energy Economics, https://doi.org/10.1016/j.eneco.2019.06.007
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Title page
Hedging Stocks with Oil
IP
T
Jonathan A. Batten1, Harald Kinateder2, Peter G. Szilagyi3 and Niklas F. Wagner4 School of Economics, Finance and Banking, Universiti Utara Malaysia, 06010 Sintok, Kedah Darul Aman, Malaysia. Email: [email protected]
2.
Department of Business and Economics, University of Passau Innstrasse 27, 94030 Passau, Germany. Tel: +49 (0)851 509 3243, Email: [email protected]
3.
Department of Economics and Business, Central European University, Nador utca 9, 1051 Budapest, Hungary. Email: [email protected]
4.
Department of Business and Economics, University of Passau Innstrasse 27, 94030 Passau, Germany. Tel: +49 (0)851 509 3240, Email: [email protected]
AC
CE
Dat e Ap ril 12, 20 19
PT
ED
M
AN
US
CR
1.
Keywords: Beta-Hedge; Brent Oil, Commodities; Economic Policy Uncertainty (EPU); Hedge ratio; International Asset Pricing; Market Risk; Systemic Risk; WTI Oil; VIX.
We study the feasibility of hedging stocks with oil. The Dynamic Conditional Correlation (DCC) approach allows for the calculation of optimal hedge ratios and corresponding hedge portfolio returns. Our results show that there are distinct economic benefits from hedging stocks with oil, although the 1
ACCEPTED MANUSCRIPT
CR
IP
T
effectiveness of hedging is both time-varying and market-state-dependent. The event of the Global Financial Crisis (GFC) is shown to affect the effectiveness of hedging. During the GFC, a positive jump in the hedge ratios occurs and hedge effectiveness increases. Among a set of common financial and macroeconomic drivers, we identify the implied volatility index VIX as being the most important. During times of global financial uncertainty, investors reduce stock positions more than commodity positions, thus VIX shocks negatively affect the portfolio returns of stock-oil hedges. The results also show that an appreciation of the U.S. dollar against the euro is associated with reduced hedge portfolio returns. From the GFC onwards, we document an increased significance of the gold price and the term spread in explaining hedge portfolio returns. (JEL F15, F2, F36, G10, G15)
AC
CE
PT
ED
M
AN
US
*We would like to thank participants at the Applied Financial Modelling Conference (Kampar, Malaysia), the Annual Financial Market Liquidity Conference (Budapest, Hungary), International Symposium on Environment and Energy Finance Issues (Paris, France) as well as five anonymous referees for helpful comments. All omissions and errors remain with the authors.
2
ACCEPTED MANUSCRIPT
Hedging Stocks with Oil 1. Introduction Understanding the co-movements between stock and oil markets is important for at least two
IP
T
main reasons1. First, portfolio managers can use the co-movement information to reduce the risk of unfavourable price fluctuations. For example, there is now an extensive empirical literature
CR
that shows these risks can be offset, i.e. "hedged", in the stock market by holding commodities,
US
which include energy assets2. The use of an energy hedge is motivated by its time-varying correlation, (due to the complex interaction effects of both demand and supply shocks), that
AN
offers both diversification as well as hedging properties. Second, recent developments in energy
M
policy will impact future energy prices and thereby affect stock portfolio outcomes. For example, one key impact arising from the COP21 and COP233 agreements are the expected ongoing
ED
decline in the future demand for fossil fuels such as coal, oil and gas 4 and, at the same time, an
PT
increasing demand for alternate energy sources. Consequently, portfolio investors will need to adjust the weighting and hedge the associated risks of those stocks affected by fossil fuel demand
1
AC
CE
in their domestic and international portfolios.
Note that Bernanke (2016) at the Brookings Institute raised the importance of analyzing the time varying correlation between stocks and oil, which by extension allows for its use of oil as a hedge of the macroeconomic risks that affect stock prices. 2
See for exxample: Chkili, Aloui and Nguyen 2014; Lin, Wesseh and Appiah 2014; Basher and Sadorsky 2016; Batten, Kinateder, Szilagyi and Wagner 2017. 3
COP21 and COP23 refer to the agreement from the 2015 United Nations Climate Change Conference in Paris for emissions targets for 2021 and 2023, respectively. The key result was an agreement to set a goal of limiting global warming to less than 2 degrees Celsius (Β°C) compared to pre-industrial levels. The agreement calls for zero net anthropogenic greenhouse gas emissions to be reached during the second half of the 21st century. 4
See for example. Vandyck, Keramidas, Saveyn, Kitous and Vrontisi 2016; Panagiotis, Paroussos, Capros and Tsani 2017.
3
ACCEPTED MANUSCRIPT In this paper, we address these issues by studying the feasibility of hedging stock indices with oil. For this purpose, we study stock-oil hedging in a comprehensive set of international stock markets. Second, not only hedge effectiveness but also the drivers of hedge portfolio returns are analyzed, which provides a clearer understanding of the benefits of stock-oil hedging
T
for investors. Importantly, while we show that there are economic benefits derived from hedging,
IP
an important feature of this paper is that we also identify those financial and macroeconomic
CR
factors that drive uncertainty in the hedging process. Since stock-energy hedging can be used to address the energy policy challenges caused by climate change, this study is also relates to
US
broader concerns of stock-oil market integration. Thus, this study extends earlier work in this
AN
area, including the earlier stud by Batten, Kinateder, Szilagyi and Wagner (2018) that considers the time-varying degree of integration between stock and oil markets and its implications for
M
COP21 (and now COP23). In this paper the time-varying hedge ratios are more precisely derived
ED
using Engleβs (2002) Dynamic Conditional Correlation (DCC), Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model. Since constant correlations are not supported
PT
empirically (e.g. Lin, Wesseh and Appiah, 2014), this improved approach allows better
CE
estimation and more accurate reporting of the hedge ratios and portfolio returns, thereby allowing better determination of hedge effectiveness.
AC
This paper offers several contributions to the energy finance, as well as the asset pricing literature more generally. First, despite the extensive literature that analyzes the co-movement between stock and oil markets, there are few studies that examine the effectiveness of hedging and the causes of hedge portfolio returns (e.g. Primbs and Yamada 2006; Tankov and Voltchkova 2009). By investigating the impact of a set of key financial explanatory variables on the hedge portfolio returns, we are able to contribute to this literature by providing a more 4
ACCEPTED MANUSCRIPT detailed understanding of the factors driving uncertainty in the stock-oil relation. Our results show that the most important driver of the time-varying hedge portfolio returns are changes in market estimates of expected stock market volatility, as proxied by the Chicago Board Options Exchange implied volatility index (VIX). In times of global uncertainty, investors tend to more
T
easily reduce more liquid stock positions, than commodity positions. Thus, shocks in the VIX
IP
negatively affect the portfolio returns of stock-oil hedges.
CR
The second contribution is the incorporation of key macroeconomic variables when analyzing drivers of hedge portfolio returns. Motivated by the results reported by Batten,
US
Kinateder, Szilagyi and Wagner (2018) that show increased stock-oil market integration with
AN
United States (U.S.) dollar appreciation, we include a USD/EUR spot exchange rate variable. We show that appreciation of the U.S. dollar is associated with reduced hedge portfolio returns.
M
However, the level of significance is market-dependent, especially for North American and
ED
emerging markets where there is no relation. Moreover, during periods of high stock-oil market integration, we document an increased impact of the U.S. dollar/euro (USD/EUR) spot rate in all
PT
markets. In addition, we find that changes in U.S. economic policy uncertainty (EPU) and hedge
CE
portfolio returns are negatively related, especially for Europe. The gold price and term structure also show increased negative effects in the later part of the sample, which includes the
AC
quantitative easing period associated with the GFC. This finding is also consistent with the recent literature highlighting the limited use of gold as an inflation hedge (e.g. Narayan, Narayan and Zheng (2010)). Third, in the estimation, we combine stocks with the worldwide price of oil represented by Brent oil futures to form portfolios with developed and emerging country stock markets. Previous empirical studies on these portfolios have highlighted their time-varying correlation 5
ACCEPTED MANUSCRIPT relationships, often with risk spillovers between specific stock and energy markets. The focus of many of these studies tends to be on the impact of energy prices on developed stock markets that have historically been oil importers, and emerging markets that tend to be oil exporters such as those in the MENA region5 (e.g. Huang, Masulis and Stoll 1996; Arouri and Rault 2012;
T
Demirer, Lee and Lien 2015; Tsai 2015; Gupta 2016; Kyrtsou, Mikropoulou and Papana 2016;
IP
Balcilar, Gupta and Wohar 2017). Despite these many studies, there are mixed results on how oil
CR
prices interact with worldwide stock prices. For example, Basher and Sadorsky (2016) find that
US
among key assets (i.e. gold, oil and bonds), oil is the best asset to hedge emerging market stock prices. The inconclusive nature of these findings provides further motivation to more precisely
AN
assess the hedging properties of stock-oil hedges.
Consequently, in this paper we investigate the broader impacts on region-wide and global
M
portfolios. Over the past two to three decades of globalization, the removal of capital controls
ED
and financial market regulatory convergence, has tended to remove those institutional barriers that once prevented investors from undertaking cross-border investments. Consequently, as the
PT
focus of many international studies is on the importance of including emerging stock markets
CE
(Bekaert and Harvey 1995, 1997, 2000; Gerard, Thanyalakpark and Batten 2003; Jeon, Oh and Yang 2006), we also include emerging stock markets in our study.
AC
In investigating energy hedges, this paper builds upon existing portfolio theory that has been applied to international financial markets (e.g. Solnik 1977; Stulz 1981) and regional stock markets (e.g. Guesmi and Teulon 2014; Guesmi, Teulon and Muzzafar 2014). Importantly, theory shows that by holding uncorrelated financial assets in an international portfolio, from
5
The Middle East and North Africa (MENA) countries include: Algeria, Bahrain, Djibouti, Egypt, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Libya, Malta, Morocco, Oman, Qatar, Saudi Arabia, Syria, Tunisia, United Arab Emirates, West Bank and Gaza, and Yemen. Ethiopia and Sudan are sometimes included.
6
ACCEPTED MANUSCRIPT markets that may be isolated by geography, regulation or function, the risk of one stock or asset market can be used to offset the risk of the other. In the case of stock markets, combining portfolios across industries and other markets with varying degrees of liquidity and market access, allows an investor to eventually form diversified portfolios that minimize risk and
T
transactions costs while maximizing expected return. This approach is critical for investors that
IP
wish to either hedge, or diversify, the risk associated with changing demand for financial assets,
CR
whose value is based on the price of fossil fuels. If investors have insufficient capital to diversify their portfolio, they can hedge this position against unfavorable prices changes by taking an
US
offsetting position. To diversify this portfolio, the investor should invest in multiple assets with a
AN
suitable correlation structure (ideally negative). Simply stated, the goal of portfolio diversification is to eliminate unsystematic risk in the portfolio. This is a different goal to
M
hedging, which involves eliminating, or minimizing, all risk in the portfolio, respectively.
ED
But exactly how can investors hedge their stock positions with energy assets? We begin by providing a clear understanding of the dynamic hedge ratios for a cross market hedge. For this
PT
purpose, we calculate risk-minimizing hedge ratios between a set of international stock market
CE
indices and Brent oil. The effectiveness of the estimated hedge ratios is determined by estimating the proportion of market risk that can be offset by the energy hedge using Brent oil. Finally, we
AC
use a regression approach to establish the key financial and macroeconomic drivers of hedge portfolio returns.
The paper is organized as follows: Next a brief overview of the hedging literature with a focus on the use of stock-oil hedging is provided to highlight the context of the paper. Then, we discuss the theoretical motivation and empirical modelling of the hedge portfolio returns. The data used in the study are described in Section 4. Section 5 outlines the results from DCC7
ACCEPTED MANUSCRIPT GARCH(1,1) estimation and analyzes the estimated hedge ratios and portfolio returns. In addition, drivers of hedge portfolio returns are analyzed. The final section allows for concluding remarks.
IP
T
2. Literature
The hedging literature in finance consists of three main strands. The first considers the positive
CR
effects of hedging on firm value (e.g. Gilje and Taillard 2017) or takes a broader macroeconomic
US
perspective. The second addresses the modelling of hedge ratios and determines hedging effectiveness (e.g. Sadorsky 2014). The third identifies key features of risk management products
AN
used by economic agents as part of a broader discussion on financial market design (e.g.
M
Tsetsekos and Varangis 2000; Carr, Geman and Madan 2001). This last group of papers also includes a rich literature that shows how financial market participants can hedge macroeconomic
ED
news (e.g. Beber and Brandt 2009).
PT
Those papers in the first group that take a macroeconomic perspective, also consider the positive effects of hedging certain types of assets from the viewpoint of economic stability. For
CE
example, Narayan, Narayan and Zheng (2010) investigate whether gold is an effective hedge
AC
against inflation, while the more recent study by Raza, Shahzad and Raza (2018) determines if certain commodities can hedge risk in real estate. The recent stock-oil hedge literature takes this perspective and typically considers the relationships between broad classes of financial and nonfinancial assets. For example, Yang, Zhou and Wang (2009) and Ciner, Gurdgiev and Lucey (2013) examine the complex interaction between assets including stocks, bonds, gold, oil and exchange rates. While there are clear portfolio management benefits from articulating the best strategy to hedge one asset portfolio with another asset, there are also important lessons for 8
ACCEPTED MANUSCRIPT policymakers with respect to maintaining macroeconomic stability. For example, it is important to ensure that there is enough liquidity in financial markets to facilitate trading and arbitrage. The second group of papers typically take more technical perspectives and address statistical concerns associated with the modelling of hedge ratios and hedging effectiveness.
T
When determining how best to hedge a bought, or long, position in one asset with a sold, or
IP
short, position in another asset, the standard approach is to identify a hedge ratio by regressing
CR
historical price data from the physical market against the equivalent price data from the futures market. Early studies that estimate the hedge ratio as the regression beta, estimated from the
US
asset and its hedge, includes Figlewski (1984) and Cecchetti, Cumby and Figlewski (1988). Due
AN
to liquidity and timing considerations, futures products seldom match actual positions held in the cash market. Thus, the estimation of optimal hedge strategies is usually undertaken using
M
mismatched futures contracts, which introduces basis risk.
ED
Myers (1991), amongst others, argues that the fundamental problem with this approach is the assumption that these optimal hedge ratios - and therefore basis risk - is constant over time.
PT
New econometric procedures, discussed in the next section, are better able to tackle the
CE
estimation problems associated with time-variation in the hedge ratio. The preferred approach for modelling time-varying hedge ratios in the literature utilize bivariate GARCH models. There is
AC
an extensive literature in this area that includes an extensive set of papers. For example, see Chang, McAleer and Tansuchat (2011), Arouri, Jouini and Nguyen (2011), Arouri, Jouini and Nguyen (2012), Chkili, Aloui and Nguyen (2014), Lin, Wesseh and Appiah (2014), Basher and Sadorsky (2016), Maghyereh, Awartani and Tziogkidis (2017) and Junttila, Pesonen and Raatikainen (2018).
9
ACCEPTED MANUSCRIPT This literature also includes studies that focus on the specifics of which futures contract maturity improves the hedging effectiveness of one specific asset. For example, Varela (1999) finds that near term gold futures contracts are better predictors of the future cash price. Finally, the third group of papers, which often take a practitioner perspective, identify what financial
T
products are best used to hedge certain types of risks in both developed and developing
IP
economies (e.g. Cheung and Chinnβs (2001) survey of products used by currency traders in the
CR
U.S.).
This paper contributes to both the first and the second strands of the hedging literature.
US
While specific financial contracts that trade on exchanges are not specifically considered, the
AN
benefits accrued to investors from hedging that we show later in the paper highlight the importance of including risk management instruments in the design of financial markets, a point
M
also made by Tsetsekos and Varangis (2000). Consequently, this paper builds upon earlier work
ED
in the stock-oil hedge literature and extends these papers by identifying a set of macroeconomic
PT
and financial market drivers that affect hedge effectiveness.
CE
3. Method
AC
This section contains the econometric methods used in this paper. First, the modelling of the time-varying hedge ratio of a stock-oil hedge and corresponding hedge portfolio returns is explained. Finally, the framework for analyzing the drivers of the hedge portfolio returns is provided.
10
ACCEPTED MANUSCRIPT 3.1 Modelling hedge portfolio returns Hedging between stock and oil markets implies a combined position consisting of stock market and oil returns, whose conditional covariance is time-varying. In our setting, investors hold stocks and wish to hedge stock market risk by holding a commodity futures, in this case crude
T
oil. In the spirit of Kroner and Sultan (1993), we model the hedge portfolio return (or pay-off
CR
IP
from buying stocks and going short in π½π‘ units of π π,π‘ ) as π β,π‘ = π π ,π‘ βπ½π‘ π π,π‘ ,
(1)
US
where π π ,π‘ is the return of stock market π at time π‘, and π π,π‘ is the return of the oil futures price at
AN
time π‘. π½π‘ represents the time-varying hedge ratio. Formally, the hedge portfolio return π β,π‘ represents the return of a two-asset difference portfolio. As a result, the conditional variance of
M
π β,π‘ for an information set consisting of information up to time π‘ β 1 is
ED
πππ(π β,π‘ |β±π‘β1 ) = πππ(π π ,π‘ |β±π‘β1 ) + π½π‘2 πππ(π π,π‘ |β±π‘β1 ) β 2π½π‘ πΆππ£(π π ,π‘ , π π,π‘ |β±π‘β1 ). (2)
PT
Next, we take the first derivative of Eq. (1) with respect to π½π‘ , which yields ππππ(π β,π‘ |β±π‘β1 )
= 2π½π‘ πππ(π π,π‘ |β±π‘β1 ) β 2πΆππ£(π π ,π‘ , π π,π‘ |β±π‘β1 ).
(3)
CE
ππ½π‘
1993):
AC
Rearranging Eq. (3) yields the optimal hedge ratio conditioned at time π‘ (Kroner and Sultan
π½π‘ =
πΆππ£(π π ,π‘ ,π π,π‘ |β±π‘β1 ) πππ(π π,π‘ |β±π‘β1 )
.
(4)
The optimal hedge ratio or risk-minimizing hedge deals with the question of how much a long position of one dollar in the stock market should be hedged by a short position of π½π‘ dollar in the oil market (e.g. Lin, Wesseh and Appiah 2014). As we focus on cross market hedging, it is 11
ACCEPTED MANUSCRIPT basically not optimal to use a hedge ratio of one. Instead, the hedge ratio is optimal that minimizes the risk (i.e. variance) of the hedged position. Therefore, we use a minimum variance hedge in our paper. The time-varying hedge ratio given in Eq. (4) implies that we must estimate the conditional
T
variance and covariance, respectively. For this purpose, we use the Dynamic Conditional
IP
Correlation (DCC) GARCH model proposed by Engle and (2002), which models the bivariate
(5)
US
πΉπ‘ = ππ‘ + π―π‘ 1/2 ππ‘ ,
CR
(2 Γ 1) vector of conditional stock and oil returns πΉπ‘ = (π π ,π‘ , π π,π‘ )β² as
where ππ‘ is a (2 Γ 1) vector of conditional means, ππ‘ is a (2 Γ 1) vector of i.i.d.
AN
innovations with πΈ(ππ‘ ) = 0 and πΈ(ππ‘ ππ‘ β² ) = π°, where π° is the identity matrix, and π―π‘ is the
M
conditional covariance matrix that can be decomposed into (6)
ED
π―π‘ = π«π‘ πͺπ‘ π«π‘ ,
PT
1/2 where π«π‘ is a (2 Γ 2) diagonal matrix that consists of the conditional standard deviations βπ,π‘
with π β {π , π}, and πͺπ‘ is a (2 Γ 2) conditional correlation matrix. For the univariate return
CE
series, we assume that their conditional mean ππ,π‘ = πΈ(π π,π‘ |β±π‘β1 ) is given by an autoregressive
AC
process of order one, i.e.
π π,π‘ = ππ + ππ π π,π‘β1 + ππ,π‘ ,
(7)
1/2
where ππ,π‘ = βπ,π‘ π§π‘ denotes the unstandardized innovations, which are drawn from a stationary distribution with mean zero und conditional variance βπ,π‘ . To estimate the univariate conditional variances of stock and oil returns βπ,π‘ = πππ(π π,π‘ |β±π‘β1 ), we use the GJR-GARCH(1,1) model of Glosten, Jagannathan and Runkle (1993) that can be represented as 12
ACCEPTED MANUSCRIPT 2 2 βπ,π‘ = ππ,0 + ππ,1 ππ,π‘β1 + ππ,2 πΌ{ππ,π‘β1 <0} ππ,π‘β1 + ππ,3 βπ,π‘β1 ,
(8)
where ππ,0 > 0, ππ,1 , ππ,2 , ππ,3 β₯ 0, and πΌ{β} is an indicator function that takes the value 1 if the innovation ππ,π‘ at time π‘ β 1 is negative and 0 otherwise.-
T
The conditional correlation matrix πͺπ‘ can be decomposed into (9)
IP
πͺπ‘ = ππππ{πΈπ‘ }β1/2 πΈπ‘ ππππ{πΈπ‘ }β1/2,
CR
where πΈπ‘ is a positive definite (2 Γ 2) covariance matrix. According to Engle (2002), the matrix
US
πΈπ‘ is given by
Μ + πππ‘β1 πβ²π‘β1 + ππΈπ‘β1 , πΈπ‘ = (1 β π β π)πͺ
AN
where ππ‘β1 is a (2 Γ 1) vector of standardized innovations, π’π,π‘ =
(10)
π π,π‘ βππ βππ π π,π‘β1 ββπ,π‘
and the matrix
M
Μ represents the unconditional correlation matrix of ππ‘ . The parameters of Eq. (10) are restricted πͺ
ED
to π + π < 1. In the original setting of Engle (2002), the distribution of ππ‘ is assumed to be multivariate normal, i.e. ππ‘ ~π(0, π»π‘ ). As empirical evidence shows that returns often have
PT
leptokurtic distributions, we relax the original assumption of multivariate normality and use the
CE
multivariate Student-t distribution instead, i.e. ππ‘ ~ππ(0, π»π‘ , π) with π degrees of freedom. To estimate the DCC-GJR-GARCH model with Student-t innovations, we use the technique of
AC
Μ empirically and applies quasiBauwens and Laurent (2005), which models the matrix πͺ maximum likelihood for parameter estimation. After estimating the covariance matrix πΈπ‘ , we can calculate the conditional correlation between the stock and oil return as ππ ,π,π‘ = where
the
conditional
covariance
ππ ,π,π‘ βππ ,π ,π‘ ππ,π,π‘
between
,
the
(11)
stock
and
oil
returns
is
ππ ,π,π‘ 13
ACCEPTED MANUSCRIPT = πΆππ£(π π ,π‘ , π π,π‘ |β±π‘β1 ) and the diagonal elements of πΈπ‘ represent the conditional variances of stock returns ππ ,π ,π‘ = πππ(π π ,π‘ |β±π‘β1 ) and oil returns ππ,π,π‘ = πππ(π π,π‘ |β±π‘β1 ). Based on these estimates, we can compute the time-varying hedge ratio in Eq. (4). 3.2 Determinants of hedge portfolio returns
based on a set of π independent variables:
CR
π β,π‘ = πΌ + πΎπ β,π‘β1 + πβ² ππ‘ + ππ‘ ,
IP
T
To analyze the determinants of the estimated hedge portfolio return, we use a regression model
(12)
US
where the dependent variable π β,π‘ is the hedge portfolio returns between the stock and oil returns
AN
defined in Eq. (1) and ππ‘ are the residuals. We use an AR(1) term to account for potential serial correlation. As we use a dynamic hedge ratio, which is updated monthly, it could be possible that
M
the hedge and, therefore, the resulting portfolio return π β,π‘ is quite constant, i.e. it may not abate
ED
immediately. πβ² is a (π Γ 1) vector that contains the regression coefficients of the (π Γ 1) vector of independent variables ππ‘ which consists of the following variables:
PT
βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. Apart from
CE
changes in macroeconomic fundamentals and implied volatility, EPU represents a different type of uncertainty. Baker, Bloom and Davis (2015) develop a time-varying index of economic policy uncertainty using U.S. newspaper coverage frequency. The
AC
ο·
authors document a positive relation between EPU and stock price volatility, and find that the EPU index spikes around times of economic policy uncertainty (e.g. the collapse of Lehman Brothers or U.S. presidential elections). Given this evidence, we assume that positive jumps in EPU produce negative stock returns. As a result, we expect a negative sign for βπΈπππ‘ . 14
ACCEPTED MANUSCRIPT ο·
π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing, obtained from Deutsche Bundesbank. In crisis periods, gold is a traditional safe haven (e.g. Ciner, Gurdgiev and Lucey (2013). As investors tend to reduce equity positions in stock market crises, positive gold returns may indicate lower
T
stock market returns. There is an extensive literature6 that investigates the relationships
IP
between stock and oil prices that could affect the hedging dynamics. Therefore, our
CR
hypothesis is that positive gold returns should be negatively related to hedge portfolio
ο·
US
returns. We use gold spot prices.
βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban
AN
Consumers, sourced from the Board of Governors of the Federal Reserve System (U.S.). An increase in U.S. inflation can cause the U.S. Federal Reserve to raise its key interest
M
rates. This tends to lower stock market prices, which should make the hedge portfolio
βπππ‘ are changes in the U.S. term structure given by the difference between the 10-year
PT
ο·
ED
return negative.
and 2-year Treasury bond yields, sourced from the Board of Governors of the Federal
CE
Reserve System (U.S.). An increase in the term spread (or term structure slope) is an
AC
indicator for prospective growth, which should affect stock market returns positively. We thus expect a positive sign for this variable. ο·
βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate, sourced from the Board of Governors of the Federal Reserve System (U.S.). We choose this currency pair for two
6
For example, Das et al. (2019) demonstrate bilateral causality in the mean and variance for gold and crude oil returns with respect to financial stress; Husain, Tiwari, Sohag and Shabaz (2019) demonstrate volatility spillovers between key previous metals (palladium, gold, platinum and silver) and crude oil; and Singhal, Choudhary, Biswal (2019) show that gold prices positively affect the stock price of the leading oil producer Mexico, while oil prices negatively affects stock prices.
15
ACCEPTED MANUSCRIPT different reasons. First, it is the most-traded currency pair. Second, Batten, Kinateder, Szilagyi and Wagner (2018) find that if the U.S. dollar appreciates, or depreciates, against the euro, this is associated with increased, or decreased, integration of the oilstock portfolio relation. Given this result, a positive change of βπππ·/πΈππ π‘ is associated
βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index
CR
ο·
IP
result, our expectation for the sign of βπππ·/πΈππ π‘ is negative.
T
with increased oil-stock integration, which induces lower hedge portfolio returns. As a
(VIX), which is constructed of options on the S&P 500 index and contains the marketβs
US
expectations of future volatility. The VIX is often regarded as a global βfear indexβ. Our
AN
hypothesis is that positive jumps in VIX have a greater impact on stock than on oil markets. As a result, π π ,π‘ drops more than π π,π‘ , which yields a negative hedge portfolio
ED
M
return. Therefore, we expect a negative sign for βππΌππ‘ .
PT
4. Data and preliminary analysis
We collect monthly closing prices of ICE-Brent near month futures contracts and stock market
CE
indices from January 1990 to December 2017. There are two key measures of the oil price, the
AC
ICE-Brent and NYMEX West Texas Intermediate (WTI) contract. Note that henceforth, for convenience, these contracts are simply referred to as Brent and WTI oil. Both contracts have an underlying value of 1,000 barrels and are deliverable at maturity or settled against cash. We use Brent oil instead of WTI oil for three reasons. First, WTI contains a local price spread due to the transport logistics associated with the movement of oil at the Cushing, Oklahoma oil storage and transport hub. Second, Table 1a reports that the volatility of WTI oil futures returns (9.30) is slightly higher than the volatility of Brent oil futures returns (9.28). Third, our results in Section 16
ACCEPTED MANUSCRIPT 5 reveal that Brent oil is more appropriate for hedging as the hedge effectiveness is higher than the one of WTI oil. All data are measured in U.S. dollars. The stock portfolios comprise the following indices:7 MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France,
T
Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore);
IP
MSCI North America (NA; Canada, U.S.); and the S&P 500 (U.S. only).
CR
The MSCI EM captures large and mid-cap constituents from 24 emerging markets (Brazil,
US
Chile, China, Colombia, Czech Republic, Egypt, Greece, Hungary, India, Indonesia, Korea, Malaysia, Mexico, Pakistan, Peru, Philippines, Poland, Russia, Qatar, South Africa, Taiwan,
AN
Thailand, Turkey and the United Arab Emirates). The MSCI MXWO, sometimes called MSCI World, captures large and mid-cap stocks across 23 developed markets (Australia, Austria,
M
Belgium, Canada, Denmark, Finland, France, Germany, Hong Kong, Ireland, Israel, Italy, Japan,
ED
the Netherlands, New Zealand, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, United Kingdom and the U.S). The MSCI ACWI represents a combination of emerging and
PT
developed markets, which captures all large and mid-cap representations from the MSCI EM and
CE
MSCI MXWO.
The MSCI Europe index represents the performance of large and mid-cap equities across
AC
15 developed countries in Europe (Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the U.K.). The index has a number of sub-indexes that cover various sub-regions, market segments/sizes and sectors, and covers approximately 85% of the free float-adjusted market capitalization in each 7
Basically, investors have two opportunities to invest in these stock market indices. First, they can buy an exchange traded fund that replicates the index. Second, they can replicate the index on their own by buying the respective stocks of the index.
17
ACCEPTED MANUSCRIPT country.8 The MSCI Far East index captures large and mid-cap representation across three countries (Japan, Singapore, and Hong Kong) and has 392 constituents. The index covers approximately 85% of the free float-adjusted market capitalization in each country. (Insert Table 1a about here)
IP
T
Key descriptive statistics are reported in Table 1a. The statistics are for the full sample period
CR
from January 1990 to December 2017. We report statistics for monthly continuously compounded percentage returns: π π,π‘ = (ln ππ,π‘ β ln ππ,π‘β1 ) β 100, where ππ,π‘ ist the closing price
US
of asset π. All asset returns (stock markets and oil) display negative skewness, which is a common finding especially for stock markets. This means that the right tail of all series is shorter
AN
than the left tail. In addition, we detect substantial kurtosis in all markets, with the highest and
M
lowest values observed for the MSCI Emerging Markets (6.47) and MSCI Far East (4.19)
ED
indices, respectively. Oil returns show higher volatility than stock returns. Nevertheless, oil returns have a positive mean, which is smaller than the mean of the stock indices. However, this
PT
is not the case for the MSCI Far East as this index has a negative mean. Finally, as expected due
CE
to the presence of third and fourth moments, the Jacque-Bera test rejects the null hypothesis of
AC
normality at the 1 percent level for both oil and stock returns. (Insert Table 1b about here)
Table 1b reports Pearson pairwise correlations for all the variables to highlight the difficulty that many investors face when constructing diversified international stock portfolios. First, all unconditional pairwise stock portfolio correlations are positive and significant, with the highest correlations between developed country stock market portfolios (e.g. the correlation 8
Source: https://www.msci.com/market-cap-weighted-indexes
18
ACCEPTED MANUSCRIPT between the MSCI MXWO and the MSCI ACWI is 0.9978), and the lowest correlations between the MSCI Far East and the MSCI NA and S&P 500 indices (e.g. the correlation between the MSCI Far East and the S&P 500 is 0.5189). The correlation between oil and the stock indices is
T
not significant at the 1 percent level for Brent oil and the MSCI NA and S&P 500.
IP
5. Results
CR
This section includes results from the empirical analysis that examines the hedge properties
US
between eight stock market indices and Brent oil. First, we present the results of the DCC-GJRGARCH(1,1). Second, we provide the optimal hedge ratio and examine the effectiveness of the
AN
hedges. Finally, we analyze the determinants of hedge portfolio return.
M
(Insert Table 2 about here)
ED
Table 2 reports estimated coefficients from a bivariate DCC-GJR-GARCH(1,1) model and associated p-values between the eight different monthly stock market returns and Brent oil.
PT
Interestingly, the conditional univariate mean equation has a highly significant AR(1) term at the
CE
1% level only for the MSCI EM, whereas returns of MSCI Fareast, MSCI NA and S&P 500 show mild autocorrelation at the 10% percent level. A possible explanation is that the MSCI EM
AC
reacts to new information in a delayed manner due to illiquidity (see e.g. Bao, Pan and Wang 2011). The lower significance of developed indices underpins this hypothesis, as developed markets basically provide higher levels of market liquidity. Our results for the conditional univariate variance equation show that the GARCH coefficients (π3 ) are highly significant, which suggests there is pronounced autocorrelation in the conditional volatility, a finding consistent with high persistence of volatility. Oil shows the 19
ACCEPTED MANUSCRIPT lowest persistence followed by the MSCI NA and the S&P 500. The highest persistence is documented for the MSCI Far East. The latter also shows the highest significance of the asymmetric volatility parameter (π2 ). Overall, from the p-values we conclude that there is mild evidence of asymmetric volatility, except for the MSCI EM (p-value = 0.3899). The estimates
T
for the first DCC parameter π show high, or almost medium, significance. Since the parameter π
IP
is highly significant for all index-oil combinations, except MSCI Fareast (p-value = 0.1069), the
CR
conditional correlation is time-varying rather than constant. The weaker significance of MSCI
US
Far East's DCC parameters indicates that the respective hedge ratio, on average, is expected to fluctuate not as much as the ones of the other indices. Overall, the BIC indicates the best model
AN
fit for the MSCI NA and the S&P 500, while the lowest one is achieved for the MSCI EM. We also check whether the GARCH approach is specified correctly. For this purpose, we
M
analyse whether standardized, as well as squared standardized, residuals show serial correlation.
ED
The Box-Ljung test is applied for low (i.e. lag 5) and high (i.e. lag 10) orders of serial correlation. Our results clearly show that the null hypothesis of no serial correlation cannot be
(Insert Figure 1 about here)
AC
5.1 Hedge ratio
CE
PT
rejected in any case. Therefore, we conclude that our GARCH setting is calibrated correctly.
The optimal hedge ratios (i.e. time-varying betas) are estimated from Eq. (4) using a DCC-GJRGARCH(1,1) model. Figure 1 shows the time-varying evolution of the hedge ratios. The analysis reveals some interesting findings. First of all, it seems that the GFC has significantly changed the hedge ratios of all indices, except for the MSCI Far East. Before the GFC, the hedge ratios of the MSCI MXWO, MSCI ACWI, MSCI G7, MSCI Europe, MSCI NA and S&P 500 indices mostly commute between -0.2 and 0.1. However, there is a positive jump during the GFC and average 20
ACCEPTED MANUSCRIPT levels of the hedge ratios also remain positive afterwards. A possible reason can be the onset of quantitative easing (QE) programs all over the world (e.g. the U.S. Federal Reserve started QE in November 2008 and the European Central Bank in May 2009). QE appears to have changed the covariance between stocks and oil from negative to positive.
T
This phenomenon is typical for financial crisis periods, where correlations among different
IP
asset classes tend to increase. However, due to QE, positive hedge ratios are not only present
CR
during the GFC, but also afterwards. The lowest (i.e. negative) hedge ratios for the S&P 500 are documented during the Iraq War (1990-1991) and in 2008. Another finding is that the lowest
US
hedge ratio for the MSCI EM was in the period 1993-1998. A possible reason could be that
AN
emerging market economies, after the collapse of the Soviet bloc, did not grow as much as developed markets, and oil exporters such as Russia and Brazil suffered from lower oil prices.
M
The above findings and the fact that the oil price has undergone structural breaks (e.g.
ED
Narayan and Gupta 2015; Narayan and Liu 2015) raise the additional question ofwhether there is a significant structural break in the stock-oil hedge ratios. To address this question, we conduct
PT
structural breakpoint tests for the hedge ratios between returns of various stock market indices
CE
and Brent oil. Table 3 reports the date and the significance of a structural break obtained from a supremum F (supF) test. The results show that September 2008 marks a highly significant
AC
structural break for all indices, except for MSCI Far East. On September 2008, a major GFC event occurred, namely the collapse of Lehman Brothers in September 2008. In addition, the supF test also shows that for six of eight indices September 2008 is the optimal date of the structural break. (Insert Table 3 about here) 5.2 Hedge effectiveness 21
ACCEPTED MANUSCRIPT The above-mentioned hedge ratios provide only a general understanding of the hedge properties. However, they do not tell us how effective hedging is over time. Therefore, we analyze the time-varying hedge effectiveness (HE) with a different measure, which corresponds to the proportion of the variance eliminated by the hedge (see Hull and White 1987; Hull and
πππ(π
|β±
T
Suo 2002): )
π‘β1
(13)
CR
π ,π‘
IP
π»πΈ = π½π‘2 πππ(π π,π‘|β± π‘β1) .
varying evolution of HE is plotted in Figure 2.
US
Thus, a perfect hedge refers to HE = 1 and no hedge effectiveness induces HE = 0. The time-
AN
(Insert Figure 2 about here)
M
Figure 2 shows that HE is clearly time-varying and changes significantly after 2008 for all indices except the MSCI Far East. After 2008, HE increases to levels up to 0.69 (December
ED
2011, MSCI EM), meaning that 69% of the proportion of the total positionβs variance can be
PT
eliminated by the hedge. The S&P offers the best hedging only for a few parts of the sample period, including from the Iraq War and in 2004-2005. In 2017, however, the S&P 500βs average
CE
HE is reduced to just 3.45%, whereas other developed market indices such as the MSCI Europe
AC
and the MSCI MXWO maintain high HE levels9. The MSCI EM outperforms the other indices from 2008 to May 2010 and from August 2015 to 2016. We next examine HE for the MSCI Far East, which reveals a completely different pattern. In contrast to the other indices, this indexβs time evolution does not show large regime shifts
9
In 2017, the average HE of MSCI Europe and MSCI MXWO is 9.50% and 5.31%, respectively.
22
ACCEPTED MANUSCRIPT over time. The average HE of the MSCI Far East mostly outperforms the other indices from January 1990 to August 1990, in 1993, and between 1995 and 2004. We further compare HE across the various indices by providing key summary statistics and Pearson correlations in Tables 3a and 3b. The time-varying HE for the MSCI Far East has quite
T
constant slope and therefore shows a lower standard deviation (0.0547) relative to that for the
IP
other indices, ranging from 0.0938 (MSCI Europe) to 0.1264 (MSCI NA). This implies that
CR
while the other indices can achieve a higher HE, the MSCI Far East offers hedgers a HE that conditional mean does not fundamentally change over time even during crisis periods. HE for the
US
MSCI Far East also reveals the lowest pairwise correlation among all indices. This completely
AN
different behavior may be due to various reasons. Notably, two of the three MSCI Far East
is also low in the index at just 0.89%.10
M
countries (Singapore and Hong Kong) are major financial centres. The share of the energy sector
ED
(Insert Tables 4a and 4b about here)
PT
The correlation matrix in Table 4b shows that the HE of S&P 500 exhibits the lowest correlation with MSCI Far East (0.2734). The correlation between the HE of S&P 500 and the
CE
MSCI NA is 0.9937.All correlations are positive and significant. Interestingly, the correlation of
AC
MSCI Far East with all other indices is quite low. This is likely due to the fact the MSCI Far East has a HE where the conditional mean is fairly constant over time. To check the robustness of the above results, we recalculate hedge ratios and HE with WTI oil. The estimated hedge ratios in Figure A.1 (see Appendix) show a similar pattern to the ones for Brent oil (see Figure 1). Next, we investigate whether WTI oil offers a different HE for 10
See https://www.msci.com/documents/10199/138e1129-188d-48d1-b2c8-8c91d6ce4f1e (retrieved Feb 12, 2018).
23
ACCEPTED MANUSCRIPT investors. Table A.1 (see Appendix) summarizes descriptive statistics of HE. The results show that for all indices, except MSCI EM, WTI oil has on average a lower HE than Brent oil. As a result, we conclude that Brent oil is more appropriate for hedgers than WTI oil. Since Brent is the more valuable oil for stock-oil hedges, we use Brent oil for the following analysis.
IP
T
5.3 Economic significance of hedging strategy
CR
Apart from analysing the time-varying effectiveness of hedging with the HE measure, we analyse whether the hedge portfolio offers utility gains for investors. In doing so, we follow the
US
approach of Kroner and Sultan (1993) and Narayan and Sharma (2016), who assume that
AN
investors optimize their utility with a mean-variance utility function. The expected average annualized percentage utility gain, βπΈ(π), is defined as the difference between the utility of the
M
hedged portfolio (HP) and the unhedged portfolio (UP): (14)
ED
βπΈ(π) = [πΈ(π(π»ππ‘ |β±π‘β1 )) β πΈ(π(πππ‘ |β±π‘β1 ))] β 12,
PT
where the expected utility of the hedged portfolio is (15)
CE
πΈ(π(π»ππ‘ |β±π‘β1 )) = πΈ(π β,π‘ |β±π‘β1) β πΎπππ(π β,π‘ |β±π‘β1 ) β ππΆ,
where πΎ > 0 is the parameter that accounts for different levels of risk aversion and ππΆ β₯ 0
AC
allows us to study whether hedging is still profitable when transaction costs are included. We obtain the return of the hedged portfolio, π»ππ‘ , directly from Eq. (1). The expected utility of the unhedged portfolio is πΈ(π(πππ‘ |β±π‘β1 )) = πΈ(π π ,π‘ |β±π‘β1 ) β πΎπππ(π π ,π‘ |β±π‘β1),
(16)
where πΎ > 0 is the parameter that accounts for different levels of risk aversion.
24
ACCEPTED MANUSCRIPT (Insert Table 5 about here) As the expected utility depends on the level of risk aversion, we consider different types of investors ranging from less risk averse investors (πΎ = 3) to highly risk averse investors (πΎ = 12). Table 5 reports the estimated average annualized percentage utility gain for all stock indices
T
using a stock-oil hedge as described in Eq. (1). We consider two scenarios: One without
IP
transaction costs and one with 0.5% of transaction costs. First, we discuss the findings of Panel
CR
A, where no transaction costs are included. The results show that the utility increases with risk
US
aversion. This means that the reduction of variance in the hedged position is larger than in the unhedged position, which underpins our previous findings of Section 5.2.11 Overall, hedging is
AN
always profitable. However, utility gains are higher for non-U.S. markets. The best hedge
M
performance is detected for the MSCI EM and MSCI Europe. For example, a highly risk averse investor obtains an average annualized percentage utility gain of 7.43% (MSCI EM) and 4.13%
ED
(MSCI Europe), respectively if they use a stock-oil hedge.
PT
We next discuss the results of Panel B, where transaction costs of 0.5% are included. Overall, our results demonstrate that transaction costs reduce the hedge profitability but it
CE
depends on the level of investor risk aversion. Low risk averse investors earn only for MSCI EM,
AC
MSCI ACWI, MSCI Europe and MSCI Far East positive utility gains. When compared to the other scenarios, we find that stock-oil hedge of these indices offer the best economic gains. A possible reason could be that these indices also show the highest correlation with Brent oil (see Table 1b). In contrast to low risk averse investors, medium and high risk averse investors earn
11
Hedge effectiveness was previously analysed with the HE measure. Table 4a reports a significant non-zero mean
of HE for all indices.
25
ACCEPTED MANUSCRIPT positive utility for all indices. 5.4 Hedge portfolio returns Before we analyze the determinants of the time-varying hedge portfolio returns, we provide descriptive statistics in Table 6. While HE measured the proportion of the variance eliminated by
IP
T
the hedge, the hedge portfolio return as defined in Eq. (1) is the absolute difference between the
CR
stock portfolio return and the Brent oil return times the hedge ratio. A positive (negative) value indicates that the hedged portfolio offers gains (losses) for the investor. As the variances of the
US
stock portfolios are not identical, the HE ratios do not apply equally to the hedge portfolio
AN
returns.
(Insert Tables 6about here)
M
Table 6 shows that on average, the hedge portfolio returns are positive for all indices
ED
except the MSCI Far East, which has negative average hedge portfolio returns. The S&P 500 and the MSCI NA have the largest average hedge portfolio returns at 0.58%, as well as the lowest
CE
PT
standard deviation of hedge portfolio returns.
AC
5.4.1 Determinants of hedge portfolio returns We next run the regression model in Eq. (12) to identify the determinants of the time-varying hedge portfolio returns. We use the set of explanatory variables summarized in Section 3.2. Table 7a reports the estimated coefficients and associated Newey and West (1987) p-values. (Insert Table 7a about here) The results show that the model fit is lower for the MSCI Far East (adj. π 2 = 0.1340) and the MSCI EM (adj. π 2 = 0.3355) indices compared to the developed market indices. 26
ACCEPTED MANUSCRIPT Nevertheless, the F-statistics are highly significant for all indices. Interestingly, the MSCI Far East, MSCI NA and S&P 500are the only indices whose hedge portfolio returns have no significant first-order autocorrelation (i.e. persistence). The positive coefficients of the other indices show that hedge portfolio return shocks do not abate immediately but persist into the
IP
index can be partly explained by a lack of such time series persistence.
T
following month. The modelβs considerably lower explanatory power for the MSCI Far East
CR
The table shows that EPU changes have the expected negative effect on hedge portfolio returns, although it is not significant for the MSCI EM, MSCI NA and S&P 500. Those indices,
US
where EPU is significant, cover regions being top exporters to the U.S., thus increased U.S. EPU
AN
may have a spillover effect on those regions. The effect of gold price returns is not consistent, with moderate negative effects documented for MSCI Europe and the S&P 500. There is no
M
significant relation shown for the other indices, although it is interesting that the coefficients on
ED
gold returns are positive for MSCI EM and MSCI Far East and negative elsewhere. Changes in U.S. inflation as well as changes in the U.S. term spread have a negative coefficient, whereas
PT
there is no pronounced influence on hedge portfolio returns.
CE
Table 7a provides evidence that the most important driver of hedge portfolio returns is changes in the VIX, with each coefficient negative and highly significant. This result confirms
AC
the hypothesis that positive jumps in the VIX can be regarded as a global fear index as investors tend to reduce stock rather than commodity positions, which leads to negative hedge portfolio returns. However, for the MSCI Far East, the influence is half as strong as for the other indices, with the estimated coefficient only -0.2980. The other indices coefficients range from -0.5076 (MSCI Europe) to -0.6628 (S&P 500), which means, for example, that a 1% increase of the VIX leads to a 0.66% reduction of S&P 500βs hedge portfolio return. This quantitative difference may 27
ACCEPTED MANUSCRIPT be explained by the fact that the MSCI Far East is the only index whose hedge effectiveness does not show a regime break during the GFC on September 2008 (see Table 3). Overall, our results on VIX are consistent with the findings of Akay, Senyuz and Yoldas (2013) who show that VIX
T
is an important driver of contagion in hedge fund returns.
IP
The negative coefficients on the USD/EUR spot rate are also in line with theory and are
CR
significant for all indices except MSCI EM, MSCI NA and the S&P 500. Estimated coefficients of all significant indices clearly show that changes in the USD/EUR spot rate have a significantly
US
lower impact on hedge portfolio returns as compared with VIX. We document the largest
AN
coefficient for MSCI Europe (-0.3323), which has 65% of the magnitude of βππΌππ‘ βs coefficient, whereas the lowest significant coefficient of MSCI G7 (-0.1349) only corresponds to 23% of the
M
magnitude of βππΌππ‘ βs coefficient.
ED
5.4.2 Additional robustness test An additional robustness test is undertaken with respect to estimation of the regression Model
PT
(13). In Section 5.1, we have shown that there is a significant structural break in hedge ratios on
CE
September 2008 for all indices, except MSCI Far East. Given this finding, we investigate whether the reported results in Section 5.4.1 are different in the subsamples before and after
AC
September 2008.
(Insert Tables 7b and 7c about here)
Tables 7b and 7c report the results of a subsample analysis, where subsample 1 covers the period from January 1990 to August 2008 and subsample 2 from the period from September 2008 to December 2017. Our previous finding that changes in the VIX are the most important 28
ACCEPTED MANUSCRIPT driver of hedge portfolio returns, both statistically as well as economically, is confirmed by the results from subsample 1 and 2. Moreover, the second most important driver, changes in the USD/EUR spot rate, show the same results as in the full sample. However, in subsample 2, the significance of the USD/EUR spot rate is higher for most markets. The increased impact of the
T
USD/EUR spot rate in subsample 2 may have been caused by the increasing integration between
CR
IP
oil and stock markets.
In this context, Batten, Kinateder, Szilagyi and Wagner (2018) document that from 2008
US
onwards, WTI oil has a near-perfect level of integration with the major stock market indices,
AN
which is fairly constant over time (see Figure 2 in Batten, Kinateder, Szilagyi and Wagner 2018). This finding may also explain why gold returns have a higher level of significance in subsample
M
2 as the increased level of WTI-stock market integration occurs simultaneously with worldwide
ED
central bank quantitative easing operations. These actions also increased correlations between other asset markets. In addition, gold has a negative sign for all indices and its coefficient is
PT
significant for all markets, except for MSCI EM and MSCI Far East. Apart from gold, the term
AC
Far East.
CE
spread is a significant driver which affects indices negatively in subsample 2, except for MSCI
Another difference in subsample 2 compared to the entire sample is that there is no persistence in hedge portfolio returns. Interestingly, inflation matters strongly for MSCI Far East. Overall, from the subsample analysis we conclude that changes in VIX and USD/EUR spot rate are the major drivers of hedge portfolio returns. From September 2008 onwards, gold returns and changes in the term structure also have a negative influence.
29
AC
CE
PT
ED
M
AN
US
CR
IP
T
ACCEPTED MANUSCRIPT
30
ACCEPTED MANUSCRIPT 5.4.3 Managerial recommendations First of all, our results show that for developed markets Brent oil provides better hedge effectiveness, although for emerging markets, WTI oil is more appropriate. Basically, hedge strategies can be applied, not only for a specific index, but also for other market indices, as hedge effectiveness is highly positively correlated across all stock markets. However, for MSCI Far
IP
T
East, we report a lower, but still significant, correlation. Moreover, in the entire sample, the
CR
lowest average hedge portfolio return is achieved for MSCI Far East. To further develop hedge strategies, it is of vital importance to be aware of hedge portfolio determinants. Among various
US
determinants, the most important one is the VIX. Hedge portfolio returns across all markets are highly negatively related to jumps in the VIX, and its economic significance is quite high, i.e.
AN
coefficients are larger than 0.5 (except for MSCI Far East). Therefore, portfolio managers should
M
take these factors into account when adjusting their hedge strategies.
ED
6. Conclusion
PT
Our study is an analysis of the characteristics of stock-oil hedges. We have used a
CE
comprehensive sample of international stock market indices both from developed as well as emerging countries and Brent oil as the key energy asset. Our results indicate that the hedge
AC
ratios of major stock market indices are time-varying and have significantly increased after the GFC. Apart from studying hedge ratios and effectiveness, this study has contributed to the literature by identifying those factors that actually drive hedge portfolio returns. Despite the hedge ratios differing among the various indices, we have identified two common drivers of hedge portfolio returns: the most important driver is the changes in the VIX, significant in all markets; the second major driver is changes in the USD/EUR spot rate. Hedge portfolio returns 31
ACCEPTED MANUSCRIPT are related negatively to both these variables. In addition, we find that since the GFC hedge effectiveness has increased and hedge portfolio returns are additionally influenced by gold returns and changes in the term structure. Our findings are important for portfolio managers, especially during periods of market stress, since they can use this information to further improve
T
their hedging performance.
IP
Moreover, future research can use these outcomes to provide additional insights in other
CR
hedging strategies with other energy assets, such as natural gas. Note that to address COP21 and COP23 concerns, investors not only need a thorough understanding of stock energy market
US
integration (see Batten, Kinateder, Szilagyi and Wagner (2018)) but also of hedging techniques,
AN
which allow them to develop adequate risk management strategies to address the impacts on
AC
CE
PT
ED
M
financial markets of climate change.
32
ACCEPTED MANUSCRIPT References Akay, O., Z. Senyuz, and E. Yoldas. 2013. Hedge fund contagion and risk-adjusted returns: A Markov-switching dynamic factor approach. Journal of Empirical Finance 22:16-29. Arouri, M. E. H, J. Jouini, and D. K. Nguyen. 2011. Volatility spillovers between oil prices and
T
stock sector returns: Implications for portfolio management, Journal of International Money and
CR
IP
Finance 30: 1387-1405.
Arouri, M. E. H., and C. Rault. 2012. Oil prices and stock markets in GCC countries: Empirical
US
evidence from panel analysis. International Journal of Finance and Economics 17:242-53.
AN
Arouri, M. E. H., J. Jouini, and D. K. Nguyen. 2012. On the impacts of oil price fluctuations on European equity markets: Volatility spillover and hedging effectiveness. Energy Economics
M
34:611-17.
ED
Baker, S. R., N. Bloom, and S. J. Davis. 2015. Measuring economic policy uncertainty. NBER
PT
Working Paper No. 21633. National Bureau of Economic Research, Cambridge, MA. Balcilar, M., R. Gupta, and M. E. Wohar. 2017. Common cycles and common trends in the stock
CE
and oil markets: Evidence from more than 150 years of data. Energy Economics 61:72-86.
66:911-46.
AC
Bao, J., J. Pan, and J. Wang. 2011. The illiquidity of corporate bonds. Journal of Finance
Basher, S. A. and P. Sadorsky. 2016. Hedging emerging market stock prices with oil, gold, VIX, and bonds: A comparison between DCC, ADCC and GO-GARCH. Energy Economics 54:23547. Batten, J. A., H. Kinateder, P. G. Szilagyi, and N. Wagner. 2017. Can stock market investors 33
ACCEPTED MANUSCRIPT hedge energy risk? Evidence from Asia. Energy Economics 66:559-70. Batten, J. A., H. Kinateder, P. G. Szilagyi, and N. Wagner. 2018. Addressing COP21 using a stock and oil market integration index. Energy Policy 116:127-36. Bauwens, L. and S. Laurent. 2005. A new class of multivariate skew densities, with application
IP
T
to generalized autoregressive conditional heteroscedasticity models. Journal of Business and
CR
Economic Statistics 23:346-354.
Beber, A. and M. W. Brandt 2009. Resolving macroeconomic uncertainty in stock and bond
US
markets. Review of Finance 13:1-45.
AN
Bekaert, G. and C. R. Harvey. 1995. Time-varying world market integration. Journal of Finance 50:403-44.
M
Bekaert, G. and C. R. Harvey. 1997. Emerging equity market volatility. Journal of Financial
ED
Economics 43:29-77.
of Finance 55:565-613.
PT
Bekaert, G. and C. R. Harvey. 2000. Foreign speculators and emerging equity markets. Journal
CE
Bernanke, B. 2016. The relationship between stocks and oil prices. Brookings Institute. Friday,
AC
February 19, 2016 https://www.brookings.edu/blog/ben-bernanke/2016/02/19/the-relationshipbetween-stocks-and-oil-prices/ Carr, P., H. Geman, and D. B. Madan. 2001. Pricing and hedging in incomplete markets. Journal of Financial Economics 62:131-67. Cecchetti, S. G., R. E. Cumby, and S. Figlewski. 1988. Estimation of the optimal futures hedge. Review of Economics and Statistics 70:623-30. 34
ACCEPTED MANUSCRIPT Chang, C.-L., M. McAleer, and R. Tansuchat. 2011. Crude oil hedging strategies using dynamic multivariate GARCH, Energy Economics 33: 912-923. Cheung, Y. W. and M. D. Chinn. 2001. Currency traders and exchange rate dynamics: A survey of the US market. Journal of International Money and Finance 20:439-71.
IP
T
Chkili, W., C. Aloui, and D. K. Nguyen. 2014. Instabilities in the relationships and hedging
CR
strategies between crude oil and U.S. stock markets: Do long memory and asymmetry matter? Journal of International Financial Markets, Institutions & Money 33:354-66.
US
Ciner C., C. Gurdgiev, and B. M. Lucey. 2013. Hedges and safe havens: An examination of
AN
stocks, bonds, gold, oil and exchange rates. International Review of Financial Analysis 29: 20211.
M
Das, D., Kumar, S.B., Tiwari, A.K., Shahbaz, M., Hasim, H.M. 2018. On the relationship of
ED
gold, crude oil, stocks with financial stress: A causality-in-quantiles approach. Finance Research
PT
Letters, 27: 169-174.
Demirer, R., H. T. Lee, and D. Lien. 2015. Does the stock market drive herd behavior in
CE
commodity futures markets? International Review of Financial Analysis 39:32-44.
AC
Engle, R. 2002. Dynamic conditional correlation - A simple class of multivariate GARCH models. Journal of Business and Economic Statistics 20:339-50. Figlewski, S. 1984. Hedging performance and basis risk in stock index futures. Journal of Finance 39:657-69. Gerard, B., K. Thanyalakpark, and J. A. Batten. 2003. Are the East Asian markets integrated? Evidence from the ICAPM. Journal of Economics and Business 55:585-607. 35
ACCEPTED MANUSCRIPT Gilje, E. P. and J. P. Taillard. 2017. Does hedging affect firm value? Evidence from a natural experiment. Review of Financial Studies 30:4083-4132. Glosten, L. R., R. Jagannathan, and D. E. Runkle. 1993. On the relation between the expected value and the volatility of the nominal excess return on stocks. Journal of Finance 48:1779-
IP
T
1801.
CR
Guesmi, K., F. Teulon, and A. T. Muzaffar. 2014. The evolution of risk premium as a measure for intra-regional equity market integration. International Review of Financial Analysis 35:13-9.
US
Guesmi, K. and F. Teulon. 2014. The determinants of regional stock market integration in
AN
Middle East: A conditional ICAPM approach. International Economics 137:22-31. Gupta, K. 2016. Oil price shocks, competition, and oil & gas stock returns - Global evidence.
M
Energy Economics 57:140-53.
ED
Huang, R. D., R. W. Masulis, and H. R. Stoll. 1996. Energy shocks and financial markets.
PT
Journal of Futures Markets 16:1-27.
Hull, J. and A. White. 1987. Hedging the risks from writing foreign currency options. Journal of
CE
International Money and Finance 6:131-52.
AC
Hull, J. and W. Suo. 2002. A methodology for assessing model risk and its application to the implied volatility function model. Journal of Financial and Quantitative Analysis 37: 297-318. Husain, S., Tiwari, A.K., Sohag, K., Shahbaz, M. (2019). Connectedness among crude oil prices, stock index and metal prices: An application of network approach in the USA. Resources Policy, 62: 57-65. Jeon, J., Y. Oh, and D. Yang. 2006. Financial market integration in East Asia: Regional or 36
ACCEPTED MANUSCRIPT global? Asian Economic Papers 5:73-89. Junttila, J., J. Pesonen, and J. Raatikainen. 2018. Commodity market based hedging against stock market risk in times of financial crisis: The case of crude oil and gold. Journal of International Financial Markets, Institutions and Money 56: 255-280.
IP
T
Kroner, K. F. and J. Sultan. 1993. Time-varying distributions and dynamic hedging with foreign
CR
currency futures. Journal of Financial and Quantitative Analysis 28: 535-51. Kyrtsou, C., C. Mikropoulou, and A. Papana. 2016. Does the S&P500 index lead the crude oil
US
dynamics? A complexity-based approach. Energy Economics 56:239-46.
AN
Lin, B., P. K. Wesseh Jr., and M. O. Appiah. 2014. Oil price fluctuation, volatility spillover and the Ghanaian equity market: Implication for portfolio management and hedging effectiveness.
M
Energy Economics 42:172-82.
ED
Maghyereh, A.I., B. Awartani, and P. Tziogkidis. 2017. Volatility spillovers and cross-hedging
Economics 68: 440-453.
PT
between gold, oil and equities: Evidence from the Gulf Cooperation Council countries, Energy
CE
Mensi, W., Hammoudeh, S., Al-Jarrah, I.M.W., Al-Yahyaee, K.H., Kang, S.H. (2019
AC
forthcoming). Risk spillovers and hedging effectiveness between major commodities, and Islamic and conventional GCC banks Journal of International Financial Markets, Institutions and Money.
Merton, R. C. 1973. An intertemporal capital asset pricing model. Econometrica 41:867-87. Myers, R.J. 1991. Estimating time-varying optimal hedge ratios on futures markets. Journal of Futures Markets 11:39-53. 37
ACCEPTED MANUSCRIPT Narayan, P. K., S. Narayan, and X. Zheng. 2010. Gold and oil futures markets: Are markets efficient? Applied Energy 87:3299-3303. Narayan, P. K. and R. Gupta. 2015. Has oil price predicted stock returns for over a century? Energy Economics 48:18-23.
IP
T
Narayan, P. K. and R. Liu. 2015. A unit root model for trending time-series energy variables.
CR
Energy Economics 50:391-402.
Narayan, P. K. and S. S. Sharma. 2016. Intraday return predictability, portfolio maximisation,
US
and hedging. Emerging Markets Review 28:105-116.
AN
Newey, W. K. and K. D. West. 1987. Hypothesis testing with efficient method of moments estimation. International Economic Review 28:777-87.
M
Panagiotis F., N. Tasios, L. Paroussos, P. Capros, and S. Tsani. 2017. Energy system impacts and
ED
policy implications of the European Intended Nationally Determined Contribution and low-
PT
carbon pathway to 2050. Energy Policy 100:216-26. Primbs, J. A. and Y. Yamada. 2006. A moment computation algorithm for the error in discrete
CE
dynamic hedging. Journal of Banking and Finance 30:519-40.
AC
Raza, N., S. Ali, S. J. H. Shahzad, and S. A. Raza. 2018. Do commodities effectively hedge real estate risk? A multi-scale asymmetric DCC approach. Resources Policy. Advance Access Published February 1, 2018. Sadorsky, P. 2014. Modeling volatility and correlations between emerging market stock prices and the prices of copper, oil and wheat. Energy Economics 43:72-81. Singhal, S., Choudhary, S., Biswal, P.C. 2019. Return and volatility linkages among 38
ACCEPTED MANUSCRIPT International crude oil price, gold price, exchange rate and stock markets: Evidence from Mexico Resources Policy, 60: 255-261. Solnik, B. H. 1977. Testing international asset pricing: Some pessimistic views. Journal of Finance 32:503-12.
IP
T
Stulz, R. M. 1981. A model of international asset pricing. Journal of Financial Economics
CR
9:383-406.
Tankov, P. and E. Voltchkova. 2009. Asymptotic analysis of hedging errors in models with
US
jumps. Stochastic Processes and their Applications 119: 2004-27.
AN
Tsai, C. L. 2015. How do U.S. stock returns respond differently to oil price shocks pre-crisis, within the financial crisis, and post-crisis? Energy Economics 50:47-62.
ED
Research Observer 15: 85-98.
M
Tsetsekos, G. and P. Varangis. 2000. Lessons in structuring derivatives exchanges. World Bank
PT
Vandyck, T., K. Keramidas, B. Saveyn, A. Kitous, and Z. Vrontisi. 2016. A global stocktake of
41:46-63.
CE
the Paris pledges: Implications for energy systems and economy. Global Environmental Change
AC
Varela, O. 1999. Futures and realized cash or settle prices for gold, silver and copper. Review of Financial Economics 8:121-38. Yang, J., Y. Zhou, and Z. Wang. 2009. The stock-bond correlation and macroeconomic conditions: One and a half centuries of evidence. Journal of Banking and Finance 33: 670-80.
39
ACCEPTED MANUSCRIPT Table 1a Descriptive statistics of returns Mean Median Std. dev. Skewness Kurtosis Jarque-Bera p-value Observations
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500 Brent Oil WTI Oil 0.501669 0.389993 0.387431 0.373763 0.383295 -0.017262 0.598390 0.602257 0.369065 0.302224 0.917182 1.037763 0.916619 0.836937 0.860256 0.271430 1.074099 1.043291 0.588023 0.731658 6.670794 4.294740 4.378431 4.999437 4.230733 5.619391 4.168421 4.129745 9.279001 9.298459 -1.056040 -0.861708 -0.897674 -0.816177 -0.805854 -0.135577 -0.835864 -0.801277 -0.182283 -0.162851 6.468644 5.131678 5.403398 5.008401 4.889204 4.187888 5.065367 4.856276 5.433638 4.751331 230.8932 0.000000
105.1990 0.000000
125.9943 0.000000
93.77552 0.000000
86.33376 0.000000
336
336
336
336
336
I R
T P
20.78444 0.000031
98.84587 0.000000
84.19516 0.000000
84.77703 0.000000
44.55758 0.00000
336
336
336
336
336
C S
U N
This table reports descriptive statistics (mean, standard deviation, skewness, kurtosis and Jarque-Beraβs normality test) for returns of various stock market indices and ICE-Brent as well as NYMEX WTI near month futures contract. Returns of asset i are measured as the difference in the natural logarithm of intermonth prices in percentage, i.e. π π,π‘ = (ln ππ,π‘ β ln ππ,π‘β1 ) β 100. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
A
D E
M
T P
E C
C A
40
ACCEPTED MANUSCRIPT Table 1b Pearson correlations of variables MSCI EM p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 1.000000 0.769324 0.0000
1.000000 -
MSCI ACWI p-value
0.804429 0.0000
0.997809 0.0000
1.000000 -
MSCI Europe p-value
0.722559 0.0000
0.921759 0.0000
0.922014 0.0000
1.000000 -
MSCI G7 p-value
0.745116 0.0000
0.996772 0.0000
0.992349 0.0000
0.896877 0.0000
MSCI Far East p-value
0.588430 0.0000
0.755847 0.0000
0.755483 0.0000
0.598100 0.0000
MSCI NA p-value
0.716476 0.0000
0.917140 0.0000
0.914326 0.0000
S&P 500 p-value
0.699329 0.0000
0.912564 0.0000
MSCI NA
S&P 500
Brent Oil p-value
0.204103 0.0002
C S
A
U N
1.000000 -
1.000000 -
0.810193 0.0000
0.921160 0.0000
0.524313 0.0000
1.000000 -
0.907763 0.0000
0.805403 0.0000
0.917456 0.0000
0.518910 0.0000
0.997384 0.0000
1.000000 -
0.165824 0.0023
0.163056 0.0027
0.148326 0.0065
0.157928 0.0037
0.114896 0.0353
0.091852 0.0928
D E
Brent Oil
I R
0.763566 0.0000
T P
E C
MSCI Far East
T P
MSCI MXWO p-value
0.153232 0.0049
MSCI G7
M
1.000000 -
C A
This table reports the pairwise Pearson correlations and associated p-values for monthly variable returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
41
ACCEPTED MANUSCRIPT Table 2 Parameter estimates of bivariate DCC-GJR-GARCH(1,1) MSCI EM MSCI MXWO MSCI ACWI MSCI Europe Mean Equation π p-value
MSCI G7
MSCI Far East MSCI NA
0.492116 0.238689
0.495957** 0.026613
0.500949** 0.030876
0.525236** 0.044339
0.525380** 0.013494
0.143560 0.752264*** 0.625527 0.000007
0.154673*** π p-value 0.005325 Variance Equation 3.334564** π€0 p-value 0.049513
-0.015708 0.781742
-0.005653 0.919682
-0.008022 0.906867
-0.021915 0.704889
0.098521* 0.094447
1.532206 0.161893
1.664649 0.112212
2.164136 0.346865
1.292936 0.154903
0.785842 0.473571
S&P 500
Brent Oil
0.746079*** 0.000010
0.002115 0.683801
-0.101174* 0.081787
-0.104200* 0.084766
0.211395*** 0.000873
0.833691* 0.056992
0.870437* 0.064882
0.000985** 0.035416
0.065160 0.118022** 0.500583 0.042473
0.102502* 0.068234
0.073852 0.117378
0.171770 0.134080
0.192938 0.105204
0.171506 0.118710
I R
T P
C S
π€1 p-value
0.069289 0.398194
0.002912 0.961365
0.000000 1.000000
0.000000 0.999999
0.020125 0.706355
π€2 p-value
0.082864 0.389866
0.222101* 0.055759
0.225322* 0.051161
0.165688 0.238272
0.803628*** 0.000000
0.775475*** 0.000000
0.772163*** 0.000000
0.806286*** 0.000001
U N
0.777880*** 0.000000
0.850578*** 0.742873*** 0.000000 0.000000
0.744195*** 0.000000
0.688586*** 0.000000
0.039244 0.000628***
0.053439 0.025825**
0.048700 0.013370**
0.028480 0.004485***
0.062791 0.038128**
0.113628 0.024849 0.057417* 0.004067***
0.072932 0.005392***
-
0.951988*** π p-value 0.000000 Diagnostics Log-likelihood -2,263.610 Bayesian Info Criterion 13.820 Q(5) 7.626 Q(10) 15.096 QΒ²(5) 0.985 QΒ²(10) 6.647
0.933797*** 0.000000
0.940893*** 0.000000
0.963372*** 0.000000
0.922291*** 0.000000
0.559227 0.917488*** 0.106860 0.000000
0.912883*** 0.000000
-
-2,114.544 12.933 1.705 4.409 0.244 0.838
-2,118.986 12.959 1.775 4.105 0.271 1.038
-2,169.326 13.259 2.389 3.592 2.554 6.843
-2,109.935 12.905 1.863 4.929 0.234 0.628
-2,101.142 12.853 2.006 7.146 0.776 1.626
4.355 11.924 1.082 4.156
π€3 p-value DCC Equation π p-value
C A
E C
A
D E
T P
0.209685** 0.040917
M
0.108119** 0.036111
-2,214.271 13.526 7.712 12.895 1.317 8.017
-2,100.840 12.851 1.909 6.957 0.557 1.520
This table reports estimated coefficients from a bivariate DCC-GJR-GARCH(1,1) model and associated p-values on monthly returns of various stock market indices and ICE-Brent near month futures contract. The first part contains results of the univariate conditional mean (see Eq. (7)) and conditional variance equation (see Eq. (8)). Q(k) and QΒ²(k) represent the Ljung-Box test for serial correlation up to order k applied to standardized residuals and squared standardized residuals, respectively. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from January 1990 to December 2017.
42
ACCEPTED MANUSCRIPT Table 3 Structural breaks of hedge ratio Date SupF p-value F(September 2008) p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500 September 2007 September 2008 September 2008 September 2008 September 2008 September 2013 September 2008 September 2008 608.46 474.81 534.45 987.99 403.33 19.06 471.06 414.48 0.0000 0.0000 0.0000 0.0000 0.0000 0.0004 0.0000 0.0000 441.24 0.46 0.0000 0.4943
T P
I R
This table reports results of structural breakpoint tests for the hedge ratio (i.e. time-varying beta) between returns of various stock market indices and ICE-Brent near month futures contract (see Eq. (4)). The date and the significance of a structural break are obtained from a supremum F (supF) test. In addition, an additional Chow test is performed for September 2008 of MSCI EM and MSCI Far East. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
C S
U N
A
D E
M
T P
E C
C A
43
ACCEPTED MANUSCRIPT Table 4 Descriptive statistics of hedge effectiveness Mean Median Maximum Minimum Std. dev. Skewness Kurtosis
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.096762 0.080642 0.084693 0.068768 0.039339 0.023668 0.020845 0.013007 0.401539 0.507081 0.504094 0.350585 0.000000 0.000002 0.000000 0.000000 0.114444 0.120469 0.124176 0.093751 1.122816 1.735279 1.605678 1.190511 2.820630 4.814569 4.361887 3.099851
MSCI G7 MSCI Far East 0.082618 0.058725 0.028973 0.047789 0.523758 0.409864 0.000000 0.000036 0.121580 0.054695 1.783704 2.330051 4.977962 11.72842
MSCI NA 0.084449 0.024876 0.544066 0.000003 0.126393 1.835536 5.259867
T P
I R
C S
S&P 500 0.074702 0.023809 0.496231 0.000030 0.111926 1.894844 5.557018
This table reports the descriptive statistics (mean, standard deviation (std. dev.), skewness, and kurtosis) of the hedge effectiveness (see Eq. (13)) between monthly returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only).The sample period is from January 1990 to December 2017.
U N
A
D E
M
T P
E C
C A
44
ACCEPTED MANUSCRIPT Table 4b Pearson correlations of hedge effectiveness MSCI EM p-value
MSCI EM MSCI MXWO 1.000000 -
MSCI ACWI
MSCI Europe
MSCI G7
MSCI Far East
0.923958 0.0000
1.000000 -
MSCI ACWI p-value
0.942035 0.0000
0.996656 0.0000
1.000000 -
MSCI Europe p-value
0.916585 0.0000
0.930474 0.0000
0.952912 0.0000
1.000000 -
C S
MSCI G7 p-value
0.909533 0.0000
0.995157 0.0000
0.985435 0.0000
0.897711 0.0000
1.000000 -
MSCI Far East p-value
0.341264 0.0000
0.315677 0.0000
0.304704 0.0000
MSCI NA p-value
0.845860 0.0000
0.967960 0.0000
D E
S&P 500 p-value
0.799041 0.0000
0.939874 0.0000
E C
S&P 500
T P
MSCI MXWO p-value
T P
MSCI NA
I R
U N
A
M
0.228511 0.0000
0.339429 0.0000
1.000000 -
0.950744 0.0000
0.840021 0.0000
0.978696 0.0000
0.279494 0.0000
1.000000 -
0.916668 0.0000
0.786859 0.0000
0.956904 0.0000
0.273429 0.0000
0.993695 0.0000
1.000000 -
C A
This table reports the pairwise Pearson correlations and associated p-values for the estimated hedge effectiveness (see Eq. (13)) between monthly returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
45
ACCEPTED MANUSCRIPT Table 5 Economic significance of hedging strategy Risk aversion
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe Panel A: No transaction costs
MSCI G7
MSCI Far East
MSCI NA
S&P 500
πΎ = 3 (low risk)
0.872512
0.427729
0.770677
0.332818
0.150124
πΎ
1.957354 1.366790 4.127039 1.628678 1.957354 1.198902 1.455294 0.272512 1.024826 0.699807 1.357354
1.198902
1.455294
1.024826
0.699807
2.741247
2.824528
2.408843
1.799174
πΎ
πΎ πΎ
1.392524 0.540875 0.697479 0.540875 3.374201 1.366790 1.628678 = 6 (medium risk) 0.697479 1.366790 1.366790 1.366790 0.872512 7.337556 3.018620 3.491076 = 12 (high risk) 1.628678 1.628678 1.628678 0.427729 1.957354 1.957354 1.957354 0.770677 Panel1.198902 B: Transaction1.198902 costs (TC = 0.51.198902 %) 0.332818 1.455294 1.455294 1.455294 0.150124 0.792524 -0.059125 0.097479 = 3 (low risk) 1.024826 1.024826 1.024826 0.699807 0.699807 0.699807 2.774201 0.766790 1.028678 = 6 (medium risk)
πΎ = 12 (high risk)
6.737556
2.418620
2.891076
M
I R
C S
-0.172271
0.170677
-0.267182
-0.449876
0.598902
0.855294
0.424826
0.099807
2.141247
2.224528
1.808843
1.199174
U N
A
3.527039
T P
This table reports the estimated average annualized percentage utility gain from Eq. (14) for different levels of risk aversion πΎ β {3,6,12}. Panel A shows results without transaction costs, and Panel B shows results with transaction costs (TC), where TC is set to 0.5%. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). Negative values indicate a situation where hedging is economically not efficient. The sample period is from January 1990 to December 2017.
D E
T P
E C
C A
46
ACCEPTED MANUSCRIPT Table 6 Descriptive statistics of hedge portfolio returns Mean Median Maximum Minimum Std. dev. Skewness Kurtosis
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.452573 0.366240 0.367954 0.356068 0.918969 0.696979 0.682551 0.711725 14.27125 10.77109 10.68602 12.30143 -34.72463 -14.49810 -15.28647 -14.91119 6.244585 4.018779 4.072345 4.688381 -0.920370 -0.604363 -0.608776 -0.582073 5.829862 3.899688 3.957707 3.912106
MSCI G7 MSCI Far East 0.354674 -0.010090 0.625749 0.253597 11.10513 18.99771 -14.19026 -19.36308 3.969503 5.447554 -0.574044 0.003237 3.799931 4.013427
MSCI NA 0.568457 0.926287 9.564921 -15.83233 3.931093 -0.707119 4.251608
T P
I R
C S
S&P 500 0.568961 0.970744 10.31122 -16.11077 3.940545 -0.745142 4.476434
This table reports the descriptive statistics (mean, standard deviation (std. dev.), skewness, and kurtosis) of hedge portfolio returns (see Eq. (1)) between monthly returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
U N
A
D E
M
T P
E C
C A
47
ACCEPTED MANUSCRIPT Table 7a: Determinants of hedge portfolio returns (full sample) Constant p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.6550** 0.4755** 0.4976** 0.4908* 0.0272 0.0200 0.0137 0.0821
MSCI G7 MSCI Far East 0.4607** 0.1818 0.0235 0.5304
MSCI NA 0.6001*** 0.0051
S&P 500 0.5758** 0.0120
0.0560 0.4776
T P
0.0627 0.1653
0.0593 0.2079
-0.1047* 0.0563
-0.0493 0.2193
-0.0525 0.2048
0.0230 0.6444
-0.0891 0.1145
-0.1018* 0.0778
-0.0398 0.4271
-0.0529 0.3251
-0.0130 0.7863
-0.0015 0.9778
-0.0551 0.1961
-0.0393 0.3668
-0.0482 0.2901
-0.0496 0.2748
-0.3323*** 0.0000
-0.1349*** 0.0025
-0.1268*** 0.0054
-0.0095 0.8362
-0.0154 0.7389
AR(1) p-value
0.1752*** 0.0001
0.1019** 0.0299
0.1047** 0.0204
0.1085*** 0.0064
0.0892* 0.0731
βπΈππ p-value
-0.0806 0.1303
-0.1006* 0.0545
-0.0963* 0.0654
-0.1397*** 0.0016
-0.0997* 0.0640
π πΊπππ p-value
0.0663 0.1938
-0.0784 0.1312
-0.0629 0.2120
-0.1073** 0.0236
βπΌπΉ p-value
-0.0840 0.0612
-0.0441 0.3839
-0.0535 0.2899
-0.0443 0.4552
βππ p-value
-0.0056 0.9028
-0.0553 0.1825
-0.0529 0.1970
-0.0586 0.1573
βπππ·/πΈππ p-value
-0.0334 0.4853
-0.1631*** 0.0002
-0.1597*** 0.0002
βππΌπ p-value
-0.5333*** 0.0000
-0.5870*** 0.0000
-0.5927*** 0.0000
-0.5076*** 0.0000
-0.5792*** 0.0000
-0.2980*** 0.0000
-0.6615*** 0.0000
-0.6628*** 0.0000
Adj. π 2 F-statistic p-value
0.3355 25.087*** 0.0000
0.4475 39.642*** 0.0000
0.4498 40.011*** 0.0000
0.4815 45.308*** 0.0000
0.4255 36.335*** 0.0000
0.1340 8.384*** 0.0000
0.4680 42.978*** 0.0000
0.4757 44.295*** 0.0000
E C
T P
D E
-0.0859 0.1125
U N
A
M
C S
I R
This table reports estimated regression coefficients from Eq. (12) in standardized form. The dependent variable is the estimated hedge portfolio return (see Eq. (1) and Table 6) on monthly frequency. Robust p-values are calculated according to Newey and West (1987). βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing. βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban Consumers. βπππ‘ are changes in the U.S. term structure given by the difference between 10-year and 2-year Treasury bond yields. βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate. βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index (VIX). The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from January 1990 to December 2017.
C A
48
ACCEPTED MANUSCRIPT Table 7b Determinants of hedge portfolio returns (subsample 1: January 1990 β August 2008) Constant p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.3002 0.5043* 0.4838 0.5834 0.5213 0.0940 0.1087 0.1059
MSCI G7 MSCI Far East 0.4989 -0.3977 0.1003 0.4113
MSCI NA 0.8042*** 0.0071
S&P 500 0.8254*** 0.0054
0.0430 0.6565
T P
0.0629 0.2070
0.0501 0.3318
-0.1035 0.2412
-0.0388 0.5587
-0.0420 0.5353
0.0795 0.1857
-0.0238 0.6840
-0.0336 0.5629
-0.0482 0.4723
0.0285 0.6882
-0.0537 0.3352
-0.0593 0.2901
-0.0294 0.5242
-0.0672 0.1973
-0.0025 0.9588
-0.0024 0.9611
-0.3494*** 0.0000
-0.1488** 0.0123
-0.1251** 0.0175
-0.0148 0.8161
-0.0209 0.7415
AR(1) p-value
0.2295*** 0.0003
0.1122* 0.0768
0.1208** 0.0487
0.1374** 0.0164
0.0926 0.1597
βπΈππ p-value
-0.1790*** 0.0019
-0.1149 0.1994
-0.1212 0.1684
-0.1648** 0.0187
-0.1084 0.2409
π πΊπππ p-value
0.0498 0.4610
-0.0184 0.7501
-0.0132 0.8190
-0.0675 0.2410
βπΌπΉ p-value
0.0040 0.9468
-0.0466 0.4920
-0.0422 0.5388
-0.0406 0.5609
βππ p-value
0.0896 0.0868*
-0.0262 0.5648
-0.0160 0.7222
-0.0302 0.5912
βπππ·/πΈππ p-value
-0.0300 0.6329
-0.1752*** 0.0026
-0.1689*** 0.0034
βππΌπ p-value
-0.5351*** 0.0000
-0.5720*** 0.0000
-0.5794*** 0.0000
-0.5074*** 0.0000
-0.5593*** 0.0000
-0.2755*** 0.0005
-0.6521*** 0.0000
-0.6540*** 0.0000
Adj. π 2 F-statistic p-value
0.3574 18.637*** 0.0000
0.3647 19.205*** 0.0000
0.3721 19.795*** 0.0000
0.3823 20.628*** 0.0000
0.3449 17.697*** 0.0000
0.086 3.965*** 0.0004
0.431 25.040*** 0.0000
0.440 25.920*** 0.0000
E C
T P
D E
-0.0220 0.7124
U N
A
M
C S
I R
This table reports estimated regression coefficients from Eq. (12) in standardized form. The dependent variable is the estimated hedge portfolio return (see Eq. (1) and Table 6) on monthly frequency. Robust p-values are calculated according to Newey and West (1987). βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing. βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban Consumers. βπππ‘ are changes in the U.S. term structure given by the difference between 10-year and 2-year Treasury bond yields. βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate. βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index (VIX). The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from January 1990 to August 2008.
C A
49
ACCEPTED MANUSCRIPT Table 7c Determinants of hedge portfolio returns (subsample 2: September 2008 β December 2017) Constant p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.4847 0.4007 0.4350* 0.2104 0.1653 0.1062 0.0813 0.5947
MSCI G7 MSCI Far East 0.4036* 0.6915*** 0.0896 0.0091
MSCI NA 0.3853* 0.0912
S&P 500 0.3581 0.1313
0.0968 0.1347
0.0529 0.4710
0.0617 0.4102
-0.1250* 0.0857
-0.0761 0.1940
-0.0490 0.1874
-0.1277 0.1817
-0.2681** 0.0375
-0.1988** 0.0329
0.0289 0.6260
-0.1748*** 0.0008
0.0847 0.1520
0.1378* 0.0724
-0.1686* 0.0792
-0.0266 0.7354
-0.2074* 0.0762
-0.1189* 0.0662
-0.3538*** 0.0000
-0.1724*** 0.0030
-0.1848** 0.0137
-0.0929 0.1102
-0.1827 0.1266
T P
AR(1) p-value
0.0540 0.2434
0.0583 0.2638
0.0616 0.2144
0.0494 0.2864
0.0539 0.3341
βπΈππ p-value
-0.0076 0.9076
-0.0948* 0.0856
-0.0828 0.1270
-0.1221** 0.0320
-0.0993* 0.0784
π πΊπππ p-value
-0.0171 0.8291
-0.2439** 0.0187
-0.2142** 0.0330
-0.2128*** 0.0042
-0.2605** 0.0196
βπΌπΉ p-value
-0.1291* 0.0664
0.0184 0.7635
-0.0087 0.8888
-0.0013 0.9884
βππ p-value
-0.1785** 0.0221
-0.1665* 0.0572
-0.1632* 0.0508
-0.1339** 0.0246
βπππ·/πΈππ p-value
-0.2584*** 0.0000
-0.2097*** 0.0004
-0.2259*** 0.0001
βππΌπ p-value
-0.5143*** 0.0000
-0.6236*** 0.0000
-0.6203*** 0.0000
-0.5069*** 0.0000
-0.6324*** 0.0000
-0.4363*** 0.0000
-0.6686*** 0.0000
-0.6715*** 0.0000
Adj. π 2 F-statistic p-value
0.507 17.320*** 0.0000
0.661 31.967*** 0.0000
0.657 31.418*** 0.0000
0.650 30.474*** 0.0000
0.644 29.719*** 0.0000
0.371 10.361*** 0.0001
0.623 27.204*** 0.0000
0.6304 28.049*** 0.0000
E C
T P
D E
U N
A
M
C S
I R
This table reports estimated regression coefficients from Eq. (12) in standardized form. The dependent variable is the estimated hedge portfolio return (see Eq. (1) and Table 6) on monthly frequency. Robust p-values are calculated according to Newey and West (1987). βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing. βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban Consumers. βπππ‘ are changes in the U.S. term structure given by the difference between 10-year and 2-year Treasury bond yields. βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate. βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index (VIX). The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from September 2008 to December 2017.
C A
50
ACCEPTED MANUSCRIPT Figure 1 Hedge ratio
Hedge Ratio
T IP CR
0.2
1993
1996
1999
2002
2005
2008
2011
2014
2017
ED
M
1990
AN
-0.2
US
0.0
Beta
0.4
0.6
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500
AC
CE
PT
This figure plots the hedge ratio (i.e. time-varying beta) between returns of various stock market indices and ICEBrent near month futures contract (see Eq. (4)), estimated by a DCC-GJR-GARCH(1,1) model. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
51
ACCEPTED MANUSCRIPT Figure 2 Hedge effectiveness
0.6
Hedge Effectivness
T IP CR
0.3
1993
1996
1999
2002
2005
2008
2011
2014
2017
M
1990
AN
0.0
US
0.1
0.2
HE
0.4
0.5
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500
AC
CE
PT
ED
This figure plots the time-varying hedge effectiveness (HE) between returns of various stock market indices and ICE-Brent near month futures contract (see Eq. (13)). HE is defined as the proportion of the variance reduced by the hedge. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
52
ACCEPTED MANUSCRIPT Appendix Table A.1 WTI oil: Descriptive statistics of hedge effectiveness Mean Median Maximum Minimum Std. dev. Skewness Kurtosis
MSCI EM 0.106206 0.058941 0.480002 0.000000 0.124252 1.253202 3.278239
MSCI MXWO 0.074713 0.026829 0.461438 0.000010 0.108192 1.806564 5.121009
MSCI ACWI 0.079762 0.027955 0.474238 0.000000 0.114478 1.730364 4.833269
MSCI Europe 0.064518 0.020395 0.336601 0.000000 0.085059 1.356519 3.738521
MSCI G7 0.076508 0.030881 0.472236 0.000000 0.108770 1.832160 5.229974
MSCI Far East 0.054788 0.047530 0.348687 0.000708 0.041836 2.881463 16.60682
C S
I R
T P
MSCI NA 0.076188 0.023000 0.497951 0.000000 0.115645 1.883080 5.493120
S&P 500 0.067648 0.022678 0.452211 0.000002 0.102751 1.926613 5.768798
This table reports the descriptive statistics (mean, standard deviation (std. dev.), skewness, and kurtosis) of the hedge effectiveness (see Eq. (13)) between monthly returns of various stock market indices and NYMEX WTI near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only).The sample period is from January 1990 to December 2017.
U N
A
D E
M
T P
E C
C A
53
ACCEPTED MANUSCRIPT Figure A.1 WTI oil: Hedge ratio
Hedge Ratio
T IP CR
0.2
1996
1999
2002
M
1993
2005
2008
2011
2014
2017
ED
1990
AN
-0.2
US
0.0
Beta
0.4
0.6
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500
AC
CE
PT
This figure plots the hedge ratio (i.e. time-varying beta) between returns of various stock market indices and NYMEX WTI near month futures contract (see Eq. (4)), estimated by a DCC-GJR-GARCH(1,1) model. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
54
ACCEPTED MANUSCRIPT
Hedging Stocks with Oil Research Highlights
T
IP
CR
US AN M ED PT
ο·
CE
ο· ο· ο· ο·
The feasibility of hedging stocks with oil is determined with a dynamic conditional correlation approach. The effectiveness of energy hedges is time-varying and market-dependent. The global financial crisis changed the effectiveness of historic hedging. The most common driver of hedge portfolio returns is the VIX. The appreciation of the U.S. dollar against the euro is associated with reduced hedge portfolio returns. Brent oil is more appropriate in developed markets for stock-oil hedges than WTI oil.
AC
ο·
55
Figure 1
Figure 2
Jonathan A. Batten, Harald Kinateder, Peter G. Szilagyi, Niklas F. Wagner PII: DOI: Reference:
S0140-9883(19)30191-4 https://doi.org/10.1016/j.eneco.2019.06.007 ENEECO 4422
To appear in:
Energy Economics
Received date: Revised date: Accepted date:
2 October 2018 14 April 2019 10 June 2019
Please cite this article as: J.A. Batten, H. Kinateder, P.G. Szilagyi, et al., Hedging stocks with oil, Energy Economics, https://doi.org/10.1016/j.eneco.2019.06.007
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Title page
Hedging Stocks with Oil
IP
T
Jonathan A. Batten1, Harald Kinateder2, Peter G. Szilagyi3 and Niklas F. Wagner4 School of Economics, Finance and Banking, Universiti Utara Malaysia, 06010 Sintok, Kedah Darul Aman, Malaysia. Email: [email protected]
2.
Department of Business and Economics, University of Passau Innstrasse 27, 94030 Passau, Germany. Tel: +49 (0)851 509 3243, Email: [email protected]
3.
Department of Economics and Business, Central European University, Nador utca 9, 1051 Budapest, Hungary. Email: [email protected]
4.
Department of Business and Economics, University of Passau Innstrasse 27, 94030 Passau, Germany. Tel: +49 (0)851 509 3240, Email: [email protected]
AC
CE
Dat e Ap ril 12, 20 19
PT
ED
M
AN
US
CR
1.
Keywords: Beta-Hedge; Brent Oil, Commodities; Economic Policy Uncertainty (EPU); Hedge ratio; International Asset Pricing; Market Risk; Systemic Risk; WTI Oil; VIX.
We study the feasibility of hedging stocks with oil. The Dynamic Conditional Correlation (DCC) approach allows for the calculation of optimal hedge ratios and corresponding hedge portfolio returns. Our results show that there are distinct economic benefits from hedging stocks with oil, although the 1
ACCEPTED MANUSCRIPT
CR
IP
T
effectiveness of hedging is both time-varying and market-state-dependent. The event of the Global Financial Crisis (GFC) is shown to affect the effectiveness of hedging. During the GFC, a positive jump in the hedge ratios occurs and hedge effectiveness increases. Among a set of common financial and macroeconomic drivers, we identify the implied volatility index VIX as being the most important. During times of global financial uncertainty, investors reduce stock positions more than commodity positions, thus VIX shocks negatively affect the portfolio returns of stock-oil hedges. The results also show that an appreciation of the U.S. dollar against the euro is associated with reduced hedge portfolio returns. From the GFC onwards, we document an increased significance of the gold price and the term spread in explaining hedge portfolio returns. (JEL F15, F2, F36, G10, G15)
AC
CE
PT
ED
M
AN
US
*We would like to thank participants at the Applied Financial Modelling Conference (Kampar, Malaysia), the Annual Financial Market Liquidity Conference (Budapest, Hungary), International Symposium on Environment and Energy Finance Issues (Paris, France) as well as five anonymous referees for helpful comments. All omissions and errors remain with the authors.
2
ACCEPTED MANUSCRIPT
Hedging Stocks with Oil 1. Introduction Understanding the co-movements between stock and oil markets is important for at least two
IP
T
main reasons1. First, portfolio managers can use the co-movement information to reduce the risk of unfavourable price fluctuations. For example, there is now an extensive empirical literature
CR
that shows these risks can be offset, i.e. "hedged", in the stock market by holding commodities,
US
which include energy assets2. The use of an energy hedge is motivated by its time-varying correlation, (due to the complex interaction effects of both demand and supply shocks), that
AN
offers both diversification as well as hedging properties. Second, recent developments in energy
M
policy will impact future energy prices and thereby affect stock portfolio outcomes. For example, one key impact arising from the COP21 and COP233 agreements are the expected ongoing
ED
decline in the future demand for fossil fuels such as coal, oil and gas 4 and, at the same time, an
PT
increasing demand for alternate energy sources. Consequently, portfolio investors will need to adjust the weighting and hedge the associated risks of those stocks affected by fossil fuel demand
1
AC
CE
in their domestic and international portfolios.
Note that Bernanke (2016) at the Brookings Institute raised the importance of analyzing the time varying correlation between stocks and oil, which by extension allows for its use of oil as a hedge of the macroeconomic risks that affect stock prices. 2
See for exxample: Chkili, Aloui and Nguyen 2014; Lin, Wesseh and Appiah 2014; Basher and Sadorsky 2016; Batten, Kinateder, Szilagyi and Wagner 2017. 3
COP21 and COP23 refer to the agreement from the 2015 United Nations Climate Change Conference in Paris for emissions targets for 2021 and 2023, respectively. The key result was an agreement to set a goal of limiting global warming to less than 2 degrees Celsius (Β°C) compared to pre-industrial levels. The agreement calls for zero net anthropogenic greenhouse gas emissions to be reached during the second half of the 21st century. 4
See for example. Vandyck, Keramidas, Saveyn, Kitous and Vrontisi 2016; Panagiotis, Paroussos, Capros and Tsani 2017.
3
ACCEPTED MANUSCRIPT In this paper, we address these issues by studying the feasibility of hedging stock indices with oil. For this purpose, we study stock-oil hedging in a comprehensive set of international stock markets. Second, not only hedge effectiveness but also the drivers of hedge portfolio returns are analyzed, which provides a clearer understanding of the benefits of stock-oil hedging
T
for investors. Importantly, while we show that there are economic benefits derived from hedging,
IP
an important feature of this paper is that we also identify those financial and macroeconomic
CR
factors that drive uncertainty in the hedging process. Since stock-energy hedging can be used to address the energy policy challenges caused by climate change, this study is also relates to
US
broader concerns of stock-oil market integration. Thus, this study extends earlier work in this
AN
area, including the earlier stud by Batten, Kinateder, Szilagyi and Wagner (2018) that considers the time-varying degree of integration between stock and oil markets and its implications for
M
COP21 (and now COP23). In this paper the time-varying hedge ratios are more precisely derived
ED
using Engleβs (2002) Dynamic Conditional Correlation (DCC), Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model. Since constant correlations are not supported
PT
empirically (e.g. Lin, Wesseh and Appiah, 2014), this improved approach allows better
CE
estimation and more accurate reporting of the hedge ratios and portfolio returns, thereby allowing better determination of hedge effectiveness.
AC
This paper offers several contributions to the energy finance, as well as the asset pricing literature more generally. First, despite the extensive literature that analyzes the co-movement between stock and oil markets, there are few studies that examine the effectiveness of hedging and the causes of hedge portfolio returns (e.g. Primbs and Yamada 2006; Tankov and Voltchkova 2009). By investigating the impact of a set of key financial explanatory variables on the hedge portfolio returns, we are able to contribute to this literature by providing a more 4
ACCEPTED MANUSCRIPT detailed understanding of the factors driving uncertainty in the stock-oil relation. Our results show that the most important driver of the time-varying hedge portfolio returns are changes in market estimates of expected stock market volatility, as proxied by the Chicago Board Options Exchange implied volatility index (VIX). In times of global uncertainty, investors tend to more
T
easily reduce more liquid stock positions, than commodity positions. Thus, shocks in the VIX
IP
negatively affect the portfolio returns of stock-oil hedges.
CR
The second contribution is the incorporation of key macroeconomic variables when analyzing drivers of hedge portfolio returns. Motivated by the results reported by Batten,
US
Kinateder, Szilagyi and Wagner (2018) that show increased stock-oil market integration with
AN
United States (U.S.) dollar appreciation, we include a USD/EUR spot exchange rate variable. We show that appreciation of the U.S. dollar is associated with reduced hedge portfolio returns.
M
However, the level of significance is market-dependent, especially for North American and
ED
emerging markets where there is no relation. Moreover, during periods of high stock-oil market integration, we document an increased impact of the U.S. dollar/euro (USD/EUR) spot rate in all
PT
markets. In addition, we find that changes in U.S. economic policy uncertainty (EPU) and hedge
CE
portfolio returns are negatively related, especially for Europe. The gold price and term structure also show increased negative effects in the later part of the sample, which includes the
AC
quantitative easing period associated with the GFC. This finding is also consistent with the recent literature highlighting the limited use of gold as an inflation hedge (e.g. Narayan, Narayan and Zheng (2010)). Third, in the estimation, we combine stocks with the worldwide price of oil represented by Brent oil futures to form portfolios with developed and emerging country stock markets. Previous empirical studies on these portfolios have highlighted their time-varying correlation 5
ACCEPTED MANUSCRIPT relationships, often with risk spillovers between specific stock and energy markets. The focus of many of these studies tends to be on the impact of energy prices on developed stock markets that have historically been oil importers, and emerging markets that tend to be oil exporters such as those in the MENA region5 (e.g. Huang, Masulis and Stoll 1996; Arouri and Rault 2012;
T
Demirer, Lee and Lien 2015; Tsai 2015; Gupta 2016; Kyrtsou, Mikropoulou and Papana 2016;
IP
Balcilar, Gupta and Wohar 2017). Despite these many studies, there are mixed results on how oil
CR
prices interact with worldwide stock prices. For example, Basher and Sadorsky (2016) find that
US
among key assets (i.e. gold, oil and bonds), oil is the best asset to hedge emerging market stock prices. The inconclusive nature of these findings provides further motivation to more precisely
AN
assess the hedging properties of stock-oil hedges.
Consequently, in this paper we investigate the broader impacts on region-wide and global
M
portfolios. Over the past two to three decades of globalization, the removal of capital controls
ED
and financial market regulatory convergence, has tended to remove those institutional barriers that once prevented investors from undertaking cross-border investments. Consequently, as the
PT
focus of many international studies is on the importance of including emerging stock markets
CE
(Bekaert and Harvey 1995, 1997, 2000; Gerard, Thanyalakpark and Batten 2003; Jeon, Oh and Yang 2006), we also include emerging stock markets in our study.
AC
In investigating energy hedges, this paper builds upon existing portfolio theory that has been applied to international financial markets (e.g. Solnik 1977; Stulz 1981) and regional stock markets (e.g. Guesmi and Teulon 2014; Guesmi, Teulon and Muzzafar 2014). Importantly, theory shows that by holding uncorrelated financial assets in an international portfolio, from
5
The Middle East and North Africa (MENA) countries include: Algeria, Bahrain, Djibouti, Egypt, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Libya, Malta, Morocco, Oman, Qatar, Saudi Arabia, Syria, Tunisia, United Arab Emirates, West Bank and Gaza, and Yemen. Ethiopia and Sudan are sometimes included.
6
ACCEPTED MANUSCRIPT markets that may be isolated by geography, regulation or function, the risk of one stock or asset market can be used to offset the risk of the other. In the case of stock markets, combining portfolios across industries and other markets with varying degrees of liquidity and market access, allows an investor to eventually form diversified portfolios that minimize risk and
T
transactions costs while maximizing expected return. This approach is critical for investors that
IP
wish to either hedge, or diversify, the risk associated with changing demand for financial assets,
CR
whose value is based on the price of fossil fuels. If investors have insufficient capital to diversify their portfolio, they can hedge this position against unfavorable prices changes by taking an
US
offsetting position. To diversify this portfolio, the investor should invest in multiple assets with a
AN
suitable correlation structure (ideally negative). Simply stated, the goal of portfolio diversification is to eliminate unsystematic risk in the portfolio. This is a different goal to
M
hedging, which involves eliminating, or minimizing, all risk in the portfolio, respectively.
ED
But exactly how can investors hedge their stock positions with energy assets? We begin by providing a clear understanding of the dynamic hedge ratios for a cross market hedge. For this
PT
purpose, we calculate risk-minimizing hedge ratios between a set of international stock market
CE
indices and Brent oil. The effectiveness of the estimated hedge ratios is determined by estimating the proportion of market risk that can be offset by the energy hedge using Brent oil. Finally, we
AC
use a regression approach to establish the key financial and macroeconomic drivers of hedge portfolio returns.
The paper is organized as follows: Next a brief overview of the hedging literature with a focus on the use of stock-oil hedging is provided to highlight the context of the paper. Then, we discuss the theoretical motivation and empirical modelling of the hedge portfolio returns. The data used in the study are described in Section 4. Section 5 outlines the results from DCC7
ACCEPTED MANUSCRIPT GARCH(1,1) estimation and analyzes the estimated hedge ratios and portfolio returns. In addition, drivers of hedge portfolio returns are analyzed. The final section allows for concluding remarks.
IP
T
2. Literature
The hedging literature in finance consists of three main strands. The first considers the positive
CR
effects of hedging on firm value (e.g. Gilje and Taillard 2017) or takes a broader macroeconomic
US
perspective. The second addresses the modelling of hedge ratios and determines hedging effectiveness (e.g. Sadorsky 2014). The third identifies key features of risk management products
AN
used by economic agents as part of a broader discussion on financial market design (e.g.
M
Tsetsekos and Varangis 2000; Carr, Geman and Madan 2001). This last group of papers also includes a rich literature that shows how financial market participants can hedge macroeconomic
ED
news (e.g. Beber and Brandt 2009).
PT
Those papers in the first group that take a macroeconomic perspective, also consider the positive effects of hedging certain types of assets from the viewpoint of economic stability. For
CE
example, Narayan, Narayan and Zheng (2010) investigate whether gold is an effective hedge
AC
against inflation, while the more recent study by Raza, Shahzad and Raza (2018) determines if certain commodities can hedge risk in real estate. The recent stock-oil hedge literature takes this perspective and typically considers the relationships between broad classes of financial and nonfinancial assets. For example, Yang, Zhou and Wang (2009) and Ciner, Gurdgiev and Lucey (2013) examine the complex interaction between assets including stocks, bonds, gold, oil and exchange rates. While there are clear portfolio management benefits from articulating the best strategy to hedge one asset portfolio with another asset, there are also important lessons for 8
ACCEPTED MANUSCRIPT policymakers with respect to maintaining macroeconomic stability. For example, it is important to ensure that there is enough liquidity in financial markets to facilitate trading and arbitrage. The second group of papers typically take more technical perspectives and address statistical concerns associated with the modelling of hedge ratios and hedging effectiveness.
T
When determining how best to hedge a bought, or long, position in one asset with a sold, or
IP
short, position in another asset, the standard approach is to identify a hedge ratio by regressing
CR
historical price data from the physical market against the equivalent price data from the futures market. Early studies that estimate the hedge ratio as the regression beta, estimated from the
US
asset and its hedge, includes Figlewski (1984) and Cecchetti, Cumby and Figlewski (1988). Due
AN
to liquidity and timing considerations, futures products seldom match actual positions held in the cash market. Thus, the estimation of optimal hedge strategies is usually undertaken using
M
mismatched futures contracts, which introduces basis risk.
ED
Myers (1991), amongst others, argues that the fundamental problem with this approach is the assumption that these optimal hedge ratios - and therefore basis risk - is constant over time.
PT
New econometric procedures, discussed in the next section, are better able to tackle the
CE
estimation problems associated with time-variation in the hedge ratio. The preferred approach for modelling time-varying hedge ratios in the literature utilize bivariate GARCH models. There is
AC
an extensive literature in this area that includes an extensive set of papers. For example, see Chang, McAleer and Tansuchat (2011), Arouri, Jouini and Nguyen (2011), Arouri, Jouini and Nguyen (2012), Chkili, Aloui and Nguyen (2014), Lin, Wesseh and Appiah (2014), Basher and Sadorsky (2016), Maghyereh, Awartani and Tziogkidis (2017) and Junttila, Pesonen and Raatikainen (2018).
9
ACCEPTED MANUSCRIPT This literature also includes studies that focus on the specifics of which futures contract maturity improves the hedging effectiveness of one specific asset. For example, Varela (1999) finds that near term gold futures contracts are better predictors of the future cash price. Finally, the third group of papers, which often take a practitioner perspective, identify what financial
T
products are best used to hedge certain types of risks in both developed and developing
IP
economies (e.g. Cheung and Chinnβs (2001) survey of products used by currency traders in the
CR
U.S.).
This paper contributes to both the first and the second strands of the hedging literature.
US
While specific financial contracts that trade on exchanges are not specifically considered, the
AN
benefits accrued to investors from hedging that we show later in the paper highlight the importance of including risk management instruments in the design of financial markets, a point
M
also made by Tsetsekos and Varangis (2000). Consequently, this paper builds upon earlier work
ED
in the stock-oil hedge literature and extends these papers by identifying a set of macroeconomic
PT
and financial market drivers that affect hedge effectiveness.
CE
3. Method
AC
This section contains the econometric methods used in this paper. First, the modelling of the time-varying hedge ratio of a stock-oil hedge and corresponding hedge portfolio returns is explained. Finally, the framework for analyzing the drivers of the hedge portfolio returns is provided.
10
ACCEPTED MANUSCRIPT 3.1 Modelling hedge portfolio returns Hedging between stock and oil markets implies a combined position consisting of stock market and oil returns, whose conditional covariance is time-varying. In our setting, investors hold stocks and wish to hedge stock market risk by holding a commodity futures, in this case crude
T
oil. In the spirit of Kroner and Sultan (1993), we model the hedge portfolio return (or pay-off
CR
IP
from buying stocks and going short in π½π‘ units of π π,π‘ ) as π β,π‘ = π π ,π‘ βπ½π‘ π π,π‘ ,
(1)
US
where π π ,π‘ is the return of stock market π at time π‘, and π π,π‘ is the return of the oil futures price at
AN
time π‘. π½π‘ represents the time-varying hedge ratio. Formally, the hedge portfolio return π β,π‘ represents the return of a two-asset difference portfolio. As a result, the conditional variance of
M
π β,π‘ for an information set consisting of information up to time π‘ β 1 is
ED
πππ(π β,π‘ |β±π‘β1 ) = πππ(π π ,π‘ |β±π‘β1 ) + π½π‘2 πππ(π π,π‘ |β±π‘β1 ) β 2π½π‘ πΆππ£(π π ,π‘ , π π,π‘ |β±π‘β1 ). (2)
PT
Next, we take the first derivative of Eq. (1) with respect to π½π‘ , which yields ππππ(π β,π‘ |β±π‘β1 )
= 2π½π‘ πππ(π π,π‘ |β±π‘β1 ) β 2πΆππ£(π π ,π‘ , π π,π‘ |β±π‘β1 ).
(3)
CE
ππ½π‘
1993):
AC
Rearranging Eq. (3) yields the optimal hedge ratio conditioned at time π‘ (Kroner and Sultan
π½π‘ =
πΆππ£(π π ,π‘ ,π π,π‘ |β±π‘β1 ) πππ(π π,π‘ |β±π‘β1 )
.
(4)
The optimal hedge ratio or risk-minimizing hedge deals with the question of how much a long position of one dollar in the stock market should be hedged by a short position of π½π‘ dollar in the oil market (e.g. Lin, Wesseh and Appiah 2014). As we focus on cross market hedging, it is 11
ACCEPTED MANUSCRIPT basically not optimal to use a hedge ratio of one. Instead, the hedge ratio is optimal that minimizes the risk (i.e. variance) of the hedged position. Therefore, we use a minimum variance hedge in our paper. The time-varying hedge ratio given in Eq. (4) implies that we must estimate the conditional
T
variance and covariance, respectively. For this purpose, we use the Dynamic Conditional
IP
Correlation (DCC) GARCH model proposed by Engle and (2002), which models the bivariate
(5)
US
πΉπ‘ = ππ‘ + π―π‘ 1/2 ππ‘ ,
CR
(2 Γ 1) vector of conditional stock and oil returns πΉπ‘ = (π π ,π‘ , π π,π‘ )β² as
where ππ‘ is a (2 Γ 1) vector of conditional means, ππ‘ is a (2 Γ 1) vector of i.i.d.
AN
innovations with πΈ(ππ‘ ) = 0 and πΈ(ππ‘ ππ‘ β² ) = π°, where π° is the identity matrix, and π―π‘ is the
M
conditional covariance matrix that can be decomposed into (6)
ED
π―π‘ = π«π‘ πͺπ‘ π«π‘ ,
PT
1/2 where π«π‘ is a (2 Γ 2) diagonal matrix that consists of the conditional standard deviations βπ,π‘
with π β {π , π}, and πͺπ‘ is a (2 Γ 2) conditional correlation matrix. For the univariate return
CE
series, we assume that their conditional mean ππ,π‘ = πΈ(π π,π‘ |β±π‘β1 ) is given by an autoregressive
AC
process of order one, i.e.
π π,π‘ = ππ + ππ π π,π‘β1 + ππ,π‘ ,
(7)
1/2
where ππ,π‘ = βπ,π‘ π§π‘ denotes the unstandardized innovations, which are drawn from a stationary distribution with mean zero und conditional variance βπ,π‘ . To estimate the univariate conditional variances of stock and oil returns βπ,π‘ = πππ(π π,π‘ |β±π‘β1 ), we use the GJR-GARCH(1,1) model of Glosten, Jagannathan and Runkle (1993) that can be represented as 12
ACCEPTED MANUSCRIPT 2 2 βπ,π‘ = ππ,0 + ππ,1 ππ,π‘β1 + ππ,2 πΌ{ππ,π‘β1 <0} ππ,π‘β1 + ππ,3 βπ,π‘β1 ,
(8)
where ππ,0 > 0, ππ,1 , ππ,2 , ππ,3 β₯ 0, and πΌ{β} is an indicator function that takes the value 1 if the innovation ππ,π‘ at time π‘ β 1 is negative and 0 otherwise.-
T
The conditional correlation matrix πͺπ‘ can be decomposed into (9)
IP
πͺπ‘ = ππππ{πΈπ‘ }β1/2 πΈπ‘ ππππ{πΈπ‘ }β1/2,
CR
where πΈπ‘ is a positive definite (2 Γ 2) covariance matrix. According to Engle (2002), the matrix
US
πΈπ‘ is given by
Μ + πππ‘β1 πβ²π‘β1 + ππΈπ‘β1 , πΈπ‘ = (1 β π β π)πͺ
AN
where ππ‘β1 is a (2 Γ 1) vector of standardized innovations, π’π,π‘ =
(10)
π π,π‘ βππ βππ π π,π‘β1 ββπ,π‘
and the matrix
M
Μ represents the unconditional correlation matrix of ππ‘ . The parameters of Eq. (10) are restricted πͺ
ED
to π + π < 1. In the original setting of Engle (2002), the distribution of ππ‘ is assumed to be multivariate normal, i.e. ππ‘ ~π(0, π»π‘ ). As empirical evidence shows that returns often have
PT
leptokurtic distributions, we relax the original assumption of multivariate normality and use the
CE
multivariate Student-t distribution instead, i.e. ππ‘ ~ππ(0, π»π‘ , π) with π degrees of freedom. To estimate the DCC-GJR-GARCH model with Student-t innovations, we use the technique of
AC
Μ empirically and applies quasiBauwens and Laurent (2005), which models the matrix πͺ maximum likelihood for parameter estimation. After estimating the covariance matrix πΈπ‘ , we can calculate the conditional correlation between the stock and oil return as ππ ,π,π‘ = where
the
conditional
covariance
ππ ,π,π‘ βππ ,π ,π‘ ππ,π,π‘
between
,
the
(11)
stock
and
oil
returns
is
ππ ,π,π‘ 13
ACCEPTED MANUSCRIPT = πΆππ£(π π ,π‘ , π π,π‘ |β±π‘β1 ) and the diagonal elements of πΈπ‘ represent the conditional variances of stock returns ππ ,π ,π‘ = πππ(π π ,π‘ |β±π‘β1 ) and oil returns ππ,π,π‘ = πππ(π π,π‘ |β±π‘β1 ). Based on these estimates, we can compute the time-varying hedge ratio in Eq. (4). 3.2 Determinants of hedge portfolio returns
based on a set of π independent variables:
CR
π β,π‘ = πΌ + πΎπ β,π‘β1 + πβ² ππ‘ + ππ‘ ,
IP
T
To analyze the determinants of the estimated hedge portfolio return, we use a regression model
(12)
US
where the dependent variable π β,π‘ is the hedge portfolio returns between the stock and oil returns
AN
defined in Eq. (1) and ππ‘ are the residuals. We use an AR(1) term to account for potential serial correlation. As we use a dynamic hedge ratio, which is updated monthly, it could be possible that
M
the hedge and, therefore, the resulting portfolio return π β,π‘ is quite constant, i.e. it may not abate
ED
immediately. πβ² is a (π Γ 1) vector that contains the regression coefficients of the (π Γ 1) vector of independent variables ππ‘ which consists of the following variables:
PT
βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. Apart from
CE
changes in macroeconomic fundamentals and implied volatility, EPU represents a different type of uncertainty. Baker, Bloom and Davis (2015) develop a time-varying index of economic policy uncertainty using U.S. newspaper coverage frequency. The
AC
ο·
authors document a positive relation between EPU and stock price volatility, and find that the EPU index spikes around times of economic policy uncertainty (e.g. the collapse of Lehman Brothers or U.S. presidential elections). Given this evidence, we assume that positive jumps in EPU produce negative stock returns. As a result, we expect a negative sign for βπΈπππ‘ . 14
ACCEPTED MANUSCRIPT ο·
π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing, obtained from Deutsche Bundesbank. In crisis periods, gold is a traditional safe haven (e.g. Ciner, Gurdgiev and Lucey (2013). As investors tend to reduce equity positions in stock market crises, positive gold returns may indicate lower
T
stock market returns. There is an extensive literature6 that investigates the relationships
IP
between stock and oil prices that could affect the hedging dynamics. Therefore, our
CR
hypothesis is that positive gold returns should be negatively related to hedge portfolio
ο·
US
returns. We use gold spot prices.
βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban
AN
Consumers, sourced from the Board of Governors of the Federal Reserve System (U.S.). An increase in U.S. inflation can cause the U.S. Federal Reserve to raise its key interest
M
rates. This tends to lower stock market prices, which should make the hedge portfolio
βπππ‘ are changes in the U.S. term structure given by the difference between the 10-year
PT
ο·
ED
return negative.
and 2-year Treasury bond yields, sourced from the Board of Governors of the Federal
CE
Reserve System (U.S.). An increase in the term spread (or term structure slope) is an
AC
indicator for prospective growth, which should affect stock market returns positively. We thus expect a positive sign for this variable. ο·
βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate, sourced from the Board of Governors of the Federal Reserve System (U.S.). We choose this currency pair for two
6
For example, Das et al. (2019) demonstrate bilateral causality in the mean and variance for gold and crude oil returns with respect to financial stress; Husain, Tiwari, Sohag and Shabaz (2019) demonstrate volatility spillovers between key previous metals (palladium, gold, platinum and silver) and crude oil; and Singhal, Choudhary, Biswal (2019) show that gold prices positively affect the stock price of the leading oil producer Mexico, while oil prices negatively affects stock prices.
15
ACCEPTED MANUSCRIPT different reasons. First, it is the most-traded currency pair. Second, Batten, Kinateder, Szilagyi and Wagner (2018) find that if the U.S. dollar appreciates, or depreciates, against the euro, this is associated with increased, or decreased, integration of the oilstock portfolio relation. Given this result, a positive change of βπππ·/πΈππ π‘ is associated
βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index
CR
ο·
IP
result, our expectation for the sign of βπππ·/πΈππ π‘ is negative.
T
with increased oil-stock integration, which induces lower hedge portfolio returns. As a
(VIX), which is constructed of options on the S&P 500 index and contains the marketβs
US
expectations of future volatility. The VIX is often regarded as a global βfear indexβ. Our
AN
hypothesis is that positive jumps in VIX have a greater impact on stock than on oil markets. As a result, π π ,π‘ drops more than π π,π‘ , which yields a negative hedge portfolio
ED
M
return. Therefore, we expect a negative sign for βππΌππ‘ .
PT
4. Data and preliminary analysis
We collect monthly closing prices of ICE-Brent near month futures contracts and stock market
CE
indices from January 1990 to December 2017. There are two key measures of the oil price, the
AC
ICE-Brent and NYMEX West Texas Intermediate (WTI) contract. Note that henceforth, for convenience, these contracts are simply referred to as Brent and WTI oil. Both contracts have an underlying value of 1,000 barrels and are deliverable at maturity or settled against cash. We use Brent oil instead of WTI oil for three reasons. First, WTI contains a local price spread due to the transport logistics associated with the movement of oil at the Cushing, Oklahoma oil storage and transport hub. Second, Table 1a reports that the volatility of WTI oil futures returns (9.30) is slightly higher than the volatility of Brent oil futures returns (9.28). Third, our results in Section 16
ACCEPTED MANUSCRIPT 5 reveal that Brent oil is more appropriate for hedging as the hedge effectiveness is higher than the one of WTI oil. All data are measured in U.S. dollars. The stock portfolios comprise the following indices:7 MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France,
T
Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore);
IP
MSCI North America (NA; Canada, U.S.); and the S&P 500 (U.S. only).
CR
The MSCI EM captures large and mid-cap constituents from 24 emerging markets (Brazil,
US
Chile, China, Colombia, Czech Republic, Egypt, Greece, Hungary, India, Indonesia, Korea, Malaysia, Mexico, Pakistan, Peru, Philippines, Poland, Russia, Qatar, South Africa, Taiwan,
AN
Thailand, Turkey and the United Arab Emirates). The MSCI MXWO, sometimes called MSCI World, captures large and mid-cap stocks across 23 developed markets (Australia, Austria,
M
Belgium, Canada, Denmark, Finland, France, Germany, Hong Kong, Ireland, Israel, Italy, Japan,
ED
the Netherlands, New Zealand, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, United Kingdom and the U.S). The MSCI ACWI represents a combination of emerging and
PT
developed markets, which captures all large and mid-cap representations from the MSCI EM and
CE
MSCI MXWO.
The MSCI Europe index represents the performance of large and mid-cap equities across
AC
15 developed countries in Europe (Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the U.K.). The index has a number of sub-indexes that cover various sub-regions, market segments/sizes and sectors, and covers approximately 85% of the free float-adjusted market capitalization in each 7
Basically, investors have two opportunities to invest in these stock market indices. First, they can buy an exchange traded fund that replicates the index. Second, they can replicate the index on their own by buying the respective stocks of the index.
17
ACCEPTED MANUSCRIPT country.8 The MSCI Far East index captures large and mid-cap representation across three countries (Japan, Singapore, and Hong Kong) and has 392 constituents. The index covers approximately 85% of the free float-adjusted market capitalization in each country. (Insert Table 1a about here)
IP
T
Key descriptive statistics are reported in Table 1a. The statistics are for the full sample period
CR
from January 1990 to December 2017. We report statistics for monthly continuously compounded percentage returns: π π,π‘ = (ln ππ,π‘ β ln ππ,π‘β1 ) β 100, where ππ,π‘ ist the closing price
US
of asset π. All asset returns (stock markets and oil) display negative skewness, which is a common finding especially for stock markets. This means that the right tail of all series is shorter
AN
than the left tail. In addition, we detect substantial kurtosis in all markets, with the highest and
M
lowest values observed for the MSCI Emerging Markets (6.47) and MSCI Far East (4.19)
ED
indices, respectively. Oil returns show higher volatility than stock returns. Nevertheless, oil returns have a positive mean, which is smaller than the mean of the stock indices. However, this
PT
is not the case for the MSCI Far East as this index has a negative mean. Finally, as expected due
CE
to the presence of third and fourth moments, the Jacque-Bera test rejects the null hypothesis of
AC
normality at the 1 percent level for both oil and stock returns. (Insert Table 1b about here)
Table 1b reports Pearson pairwise correlations for all the variables to highlight the difficulty that many investors face when constructing diversified international stock portfolios. First, all unconditional pairwise stock portfolio correlations are positive and significant, with the highest correlations between developed country stock market portfolios (e.g. the correlation 8
Source: https://www.msci.com/market-cap-weighted-indexes
18
ACCEPTED MANUSCRIPT between the MSCI MXWO and the MSCI ACWI is 0.9978), and the lowest correlations between the MSCI Far East and the MSCI NA and S&P 500 indices (e.g. the correlation between the MSCI Far East and the S&P 500 is 0.5189). The correlation between oil and the stock indices is
T
not significant at the 1 percent level for Brent oil and the MSCI NA and S&P 500.
IP
5. Results
CR
This section includes results from the empirical analysis that examines the hedge properties
US
between eight stock market indices and Brent oil. First, we present the results of the DCC-GJRGARCH(1,1). Second, we provide the optimal hedge ratio and examine the effectiveness of the
AN
hedges. Finally, we analyze the determinants of hedge portfolio return.
M
(Insert Table 2 about here)
ED
Table 2 reports estimated coefficients from a bivariate DCC-GJR-GARCH(1,1) model and associated p-values between the eight different monthly stock market returns and Brent oil.
PT
Interestingly, the conditional univariate mean equation has a highly significant AR(1) term at the
CE
1% level only for the MSCI EM, whereas returns of MSCI Fareast, MSCI NA and S&P 500 show mild autocorrelation at the 10% percent level. A possible explanation is that the MSCI EM
AC
reacts to new information in a delayed manner due to illiquidity (see e.g. Bao, Pan and Wang 2011). The lower significance of developed indices underpins this hypothesis, as developed markets basically provide higher levels of market liquidity. Our results for the conditional univariate variance equation show that the GARCH coefficients (π3 ) are highly significant, which suggests there is pronounced autocorrelation in the conditional volatility, a finding consistent with high persistence of volatility. Oil shows the 19
ACCEPTED MANUSCRIPT lowest persistence followed by the MSCI NA and the S&P 500. The highest persistence is documented for the MSCI Far East. The latter also shows the highest significance of the asymmetric volatility parameter (π2 ). Overall, from the p-values we conclude that there is mild evidence of asymmetric volatility, except for the MSCI EM (p-value = 0.3899). The estimates
T
for the first DCC parameter π show high, or almost medium, significance. Since the parameter π
IP
is highly significant for all index-oil combinations, except MSCI Fareast (p-value = 0.1069), the
CR
conditional correlation is time-varying rather than constant. The weaker significance of MSCI
US
Far East's DCC parameters indicates that the respective hedge ratio, on average, is expected to fluctuate not as much as the ones of the other indices. Overall, the BIC indicates the best model
AN
fit for the MSCI NA and the S&P 500, while the lowest one is achieved for the MSCI EM. We also check whether the GARCH approach is specified correctly. For this purpose, we
M
analyse whether standardized, as well as squared standardized, residuals show serial correlation.
ED
The Box-Ljung test is applied for low (i.e. lag 5) and high (i.e. lag 10) orders of serial correlation. Our results clearly show that the null hypothesis of no serial correlation cannot be
(Insert Figure 1 about here)
AC
5.1 Hedge ratio
CE
PT
rejected in any case. Therefore, we conclude that our GARCH setting is calibrated correctly.
The optimal hedge ratios (i.e. time-varying betas) are estimated from Eq. (4) using a DCC-GJRGARCH(1,1) model. Figure 1 shows the time-varying evolution of the hedge ratios. The analysis reveals some interesting findings. First of all, it seems that the GFC has significantly changed the hedge ratios of all indices, except for the MSCI Far East. Before the GFC, the hedge ratios of the MSCI MXWO, MSCI ACWI, MSCI G7, MSCI Europe, MSCI NA and S&P 500 indices mostly commute between -0.2 and 0.1. However, there is a positive jump during the GFC and average 20
ACCEPTED MANUSCRIPT levels of the hedge ratios also remain positive afterwards. A possible reason can be the onset of quantitative easing (QE) programs all over the world (e.g. the U.S. Federal Reserve started QE in November 2008 and the European Central Bank in May 2009). QE appears to have changed the covariance between stocks and oil from negative to positive.
T
This phenomenon is typical for financial crisis periods, where correlations among different
IP
asset classes tend to increase. However, due to QE, positive hedge ratios are not only present
CR
during the GFC, but also afterwards. The lowest (i.e. negative) hedge ratios for the S&P 500 are documented during the Iraq War (1990-1991) and in 2008. Another finding is that the lowest
US
hedge ratio for the MSCI EM was in the period 1993-1998. A possible reason could be that
AN
emerging market economies, after the collapse of the Soviet bloc, did not grow as much as developed markets, and oil exporters such as Russia and Brazil suffered from lower oil prices.
M
The above findings and the fact that the oil price has undergone structural breaks (e.g.
ED
Narayan and Gupta 2015; Narayan and Liu 2015) raise the additional question ofwhether there is a significant structural break in the stock-oil hedge ratios. To address this question, we conduct
PT
structural breakpoint tests for the hedge ratios between returns of various stock market indices
CE
and Brent oil. Table 3 reports the date and the significance of a structural break obtained from a supremum F (supF) test. The results show that September 2008 marks a highly significant
AC
structural break for all indices, except for MSCI Far East. On September 2008, a major GFC event occurred, namely the collapse of Lehman Brothers in September 2008. In addition, the supF test also shows that for six of eight indices September 2008 is the optimal date of the structural break. (Insert Table 3 about here) 5.2 Hedge effectiveness 21
ACCEPTED MANUSCRIPT The above-mentioned hedge ratios provide only a general understanding of the hedge properties. However, they do not tell us how effective hedging is over time. Therefore, we analyze the time-varying hedge effectiveness (HE) with a different measure, which corresponds to the proportion of the variance eliminated by the hedge (see Hull and White 1987; Hull and
πππ(π
|β±
T
Suo 2002): )
π‘β1
(13)
CR
π ,π‘
IP
π»πΈ = π½π‘2 πππ(π π,π‘|β± π‘β1) .
varying evolution of HE is plotted in Figure 2.
US
Thus, a perfect hedge refers to HE = 1 and no hedge effectiveness induces HE = 0. The time-
AN
(Insert Figure 2 about here)
M
Figure 2 shows that HE is clearly time-varying and changes significantly after 2008 for all indices except the MSCI Far East. After 2008, HE increases to levels up to 0.69 (December
ED
2011, MSCI EM), meaning that 69% of the proportion of the total positionβs variance can be
PT
eliminated by the hedge. The S&P offers the best hedging only for a few parts of the sample period, including from the Iraq War and in 2004-2005. In 2017, however, the S&P 500βs average
CE
HE is reduced to just 3.45%, whereas other developed market indices such as the MSCI Europe
AC
and the MSCI MXWO maintain high HE levels9. The MSCI EM outperforms the other indices from 2008 to May 2010 and from August 2015 to 2016. We next examine HE for the MSCI Far East, which reveals a completely different pattern. In contrast to the other indices, this indexβs time evolution does not show large regime shifts
9
In 2017, the average HE of MSCI Europe and MSCI MXWO is 9.50% and 5.31%, respectively.
22
ACCEPTED MANUSCRIPT over time. The average HE of the MSCI Far East mostly outperforms the other indices from January 1990 to August 1990, in 1993, and between 1995 and 2004. We further compare HE across the various indices by providing key summary statistics and Pearson correlations in Tables 3a and 3b. The time-varying HE for the MSCI Far East has quite
T
constant slope and therefore shows a lower standard deviation (0.0547) relative to that for the
IP
other indices, ranging from 0.0938 (MSCI Europe) to 0.1264 (MSCI NA). This implies that
CR
while the other indices can achieve a higher HE, the MSCI Far East offers hedgers a HE that conditional mean does not fundamentally change over time even during crisis periods. HE for the
US
MSCI Far East also reveals the lowest pairwise correlation among all indices. This completely
AN
different behavior may be due to various reasons. Notably, two of the three MSCI Far East
is also low in the index at just 0.89%.10
M
countries (Singapore and Hong Kong) are major financial centres. The share of the energy sector
ED
(Insert Tables 4a and 4b about here)
PT
The correlation matrix in Table 4b shows that the HE of S&P 500 exhibits the lowest correlation with MSCI Far East (0.2734). The correlation between the HE of S&P 500 and the
CE
MSCI NA is 0.9937.All correlations are positive and significant. Interestingly, the correlation of
AC
MSCI Far East with all other indices is quite low. This is likely due to the fact the MSCI Far East has a HE where the conditional mean is fairly constant over time. To check the robustness of the above results, we recalculate hedge ratios and HE with WTI oil. The estimated hedge ratios in Figure A.1 (see Appendix) show a similar pattern to the ones for Brent oil (see Figure 1). Next, we investigate whether WTI oil offers a different HE for 10
See https://www.msci.com/documents/10199/138e1129-188d-48d1-b2c8-8c91d6ce4f1e (retrieved Feb 12, 2018).
23
ACCEPTED MANUSCRIPT investors. Table A.1 (see Appendix) summarizes descriptive statistics of HE. The results show that for all indices, except MSCI EM, WTI oil has on average a lower HE than Brent oil. As a result, we conclude that Brent oil is more appropriate for hedgers than WTI oil. Since Brent is the more valuable oil for stock-oil hedges, we use Brent oil for the following analysis.
IP
T
5.3 Economic significance of hedging strategy
CR
Apart from analysing the time-varying effectiveness of hedging with the HE measure, we analyse whether the hedge portfolio offers utility gains for investors. In doing so, we follow the
US
approach of Kroner and Sultan (1993) and Narayan and Sharma (2016), who assume that
AN
investors optimize their utility with a mean-variance utility function. The expected average annualized percentage utility gain, βπΈ(π), is defined as the difference between the utility of the
M
hedged portfolio (HP) and the unhedged portfolio (UP): (14)
ED
βπΈ(π) = [πΈ(π(π»ππ‘ |β±π‘β1 )) β πΈ(π(πππ‘ |β±π‘β1 ))] β 12,
PT
where the expected utility of the hedged portfolio is (15)
CE
πΈ(π(π»ππ‘ |β±π‘β1 )) = πΈ(π β,π‘ |β±π‘β1) β πΎπππ(π β,π‘ |β±π‘β1 ) β ππΆ,
where πΎ > 0 is the parameter that accounts for different levels of risk aversion and ππΆ β₯ 0
AC
allows us to study whether hedging is still profitable when transaction costs are included. We obtain the return of the hedged portfolio, π»ππ‘ , directly from Eq. (1). The expected utility of the unhedged portfolio is πΈ(π(πππ‘ |β±π‘β1 )) = πΈ(π π ,π‘ |β±π‘β1 ) β πΎπππ(π π ,π‘ |β±π‘β1),
(16)
where πΎ > 0 is the parameter that accounts for different levels of risk aversion.
24
ACCEPTED MANUSCRIPT (Insert Table 5 about here) As the expected utility depends on the level of risk aversion, we consider different types of investors ranging from less risk averse investors (πΎ = 3) to highly risk averse investors (πΎ = 12). Table 5 reports the estimated average annualized percentage utility gain for all stock indices
T
using a stock-oil hedge as described in Eq. (1). We consider two scenarios: One without
IP
transaction costs and one with 0.5% of transaction costs. First, we discuss the findings of Panel
CR
A, where no transaction costs are included. The results show that the utility increases with risk
US
aversion. This means that the reduction of variance in the hedged position is larger than in the unhedged position, which underpins our previous findings of Section 5.2.11 Overall, hedging is
AN
always profitable. However, utility gains are higher for non-U.S. markets. The best hedge
M
performance is detected for the MSCI EM and MSCI Europe. For example, a highly risk averse investor obtains an average annualized percentage utility gain of 7.43% (MSCI EM) and 4.13%
ED
(MSCI Europe), respectively if they use a stock-oil hedge.
PT
We next discuss the results of Panel B, where transaction costs of 0.5% are included. Overall, our results demonstrate that transaction costs reduce the hedge profitability but it
CE
depends on the level of investor risk aversion. Low risk averse investors earn only for MSCI EM,
AC
MSCI ACWI, MSCI Europe and MSCI Far East positive utility gains. When compared to the other scenarios, we find that stock-oil hedge of these indices offer the best economic gains. A possible reason could be that these indices also show the highest correlation with Brent oil (see Table 1b). In contrast to low risk averse investors, medium and high risk averse investors earn
11
Hedge effectiveness was previously analysed with the HE measure. Table 4a reports a significant non-zero mean
of HE for all indices.
25
ACCEPTED MANUSCRIPT positive utility for all indices. 5.4 Hedge portfolio returns Before we analyze the determinants of the time-varying hedge portfolio returns, we provide descriptive statistics in Table 6. While HE measured the proportion of the variance eliminated by
IP
T
the hedge, the hedge portfolio return as defined in Eq. (1) is the absolute difference between the
CR
stock portfolio return and the Brent oil return times the hedge ratio. A positive (negative) value indicates that the hedged portfolio offers gains (losses) for the investor. As the variances of the
US
stock portfolios are not identical, the HE ratios do not apply equally to the hedge portfolio
AN
returns.
(Insert Tables 6about here)
M
Table 6 shows that on average, the hedge portfolio returns are positive for all indices
ED
except the MSCI Far East, which has negative average hedge portfolio returns. The S&P 500 and the MSCI NA have the largest average hedge portfolio returns at 0.58%, as well as the lowest
CE
PT
standard deviation of hedge portfolio returns.
AC
5.4.1 Determinants of hedge portfolio returns We next run the regression model in Eq. (12) to identify the determinants of the time-varying hedge portfolio returns. We use the set of explanatory variables summarized in Section 3.2. Table 7a reports the estimated coefficients and associated Newey and West (1987) p-values. (Insert Table 7a about here) The results show that the model fit is lower for the MSCI Far East (adj. π 2 = 0.1340) and the MSCI EM (adj. π 2 = 0.3355) indices compared to the developed market indices. 26
ACCEPTED MANUSCRIPT Nevertheless, the F-statistics are highly significant for all indices. Interestingly, the MSCI Far East, MSCI NA and S&P 500are the only indices whose hedge portfolio returns have no significant first-order autocorrelation (i.e. persistence). The positive coefficients of the other indices show that hedge portfolio return shocks do not abate immediately but persist into the
IP
index can be partly explained by a lack of such time series persistence.
T
following month. The modelβs considerably lower explanatory power for the MSCI Far East
CR
The table shows that EPU changes have the expected negative effect on hedge portfolio returns, although it is not significant for the MSCI EM, MSCI NA and S&P 500. Those indices,
US
where EPU is significant, cover regions being top exporters to the U.S., thus increased U.S. EPU
AN
may have a spillover effect on those regions. The effect of gold price returns is not consistent, with moderate negative effects documented for MSCI Europe and the S&P 500. There is no
M
significant relation shown for the other indices, although it is interesting that the coefficients on
ED
gold returns are positive for MSCI EM and MSCI Far East and negative elsewhere. Changes in U.S. inflation as well as changes in the U.S. term spread have a negative coefficient, whereas
PT
there is no pronounced influence on hedge portfolio returns.
CE
Table 7a provides evidence that the most important driver of hedge portfolio returns is changes in the VIX, with each coefficient negative and highly significant. This result confirms
AC
the hypothesis that positive jumps in the VIX can be regarded as a global fear index as investors tend to reduce stock rather than commodity positions, which leads to negative hedge portfolio returns. However, for the MSCI Far East, the influence is half as strong as for the other indices, with the estimated coefficient only -0.2980. The other indices coefficients range from -0.5076 (MSCI Europe) to -0.6628 (S&P 500), which means, for example, that a 1% increase of the VIX leads to a 0.66% reduction of S&P 500βs hedge portfolio return. This quantitative difference may 27
ACCEPTED MANUSCRIPT be explained by the fact that the MSCI Far East is the only index whose hedge effectiveness does not show a regime break during the GFC on September 2008 (see Table 3). Overall, our results on VIX are consistent with the findings of Akay, Senyuz and Yoldas (2013) who show that VIX
T
is an important driver of contagion in hedge fund returns.
IP
The negative coefficients on the USD/EUR spot rate are also in line with theory and are
CR
significant for all indices except MSCI EM, MSCI NA and the S&P 500. Estimated coefficients of all significant indices clearly show that changes in the USD/EUR spot rate have a significantly
US
lower impact on hedge portfolio returns as compared with VIX. We document the largest
AN
coefficient for MSCI Europe (-0.3323), which has 65% of the magnitude of βππΌππ‘ βs coefficient, whereas the lowest significant coefficient of MSCI G7 (-0.1349) only corresponds to 23% of the
M
magnitude of βππΌππ‘ βs coefficient.
ED
5.4.2 Additional robustness test An additional robustness test is undertaken with respect to estimation of the regression Model
PT
(13). In Section 5.1, we have shown that there is a significant structural break in hedge ratios on
CE
September 2008 for all indices, except MSCI Far East. Given this finding, we investigate whether the reported results in Section 5.4.1 are different in the subsamples before and after
AC
September 2008.
(Insert Tables 7b and 7c about here)
Tables 7b and 7c report the results of a subsample analysis, where subsample 1 covers the period from January 1990 to August 2008 and subsample 2 from the period from September 2008 to December 2017. Our previous finding that changes in the VIX are the most important 28
ACCEPTED MANUSCRIPT driver of hedge portfolio returns, both statistically as well as economically, is confirmed by the results from subsample 1 and 2. Moreover, the second most important driver, changes in the USD/EUR spot rate, show the same results as in the full sample. However, in subsample 2, the significance of the USD/EUR spot rate is higher for most markets. The increased impact of the
T
USD/EUR spot rate in subsample 2 may have been caused by the increasing integration between
CR
IP
oil and stock markets.
In this context, Batten, Kinateder, Szilagyi and Wagner (2018) document that from 2008
US
onwards, WTI oil has a near-perfect level of integration with the major stock market indices,
AN
which is fairly constant over time (see Figure 2 in Batten, Kinateder, Szilagyi and Wagner 2018). This finding may also explain why gold returns have a higher level of significance in subsample
M
2 as the increased level of WTI-stock market integration occurs simultaneously with worldwide
ED
central bank quantitative easing operations. These actions also increased correlations between other asset markets. In addition, gold has a negative sign for all indices and its coefficient is
PT
significant for all markets, except for MSCI EM and MSCI Far East. Apart from gold, the term
AC
Far East.
CE
spread is a significant driver which affects indices negatively in subsample 2, except for MSCI
Another difference in subsample 2 compared to the entire sample is that there is no persistence in hedge portfolio returns. Interestingly, inflation matters strongly for MSCI Far East. Overall, from the subsample analysis we conclude that changes in VIX and USD/EUR spot rate are the major drivers of hedge portfolio returns. From September 2008 onwards, gold returns and changes in the term structure also have a negative influence.
29
AC
CE
PT
ED
M
AN
US
CR
IP
T
ACCEPTED MANUSCRIPT
30
ACCEPTED MANUSCRIPT 5.4.3 Managerial recommendations First of all, our results show that for developed markets Brent oil provides better hedge effectiveness, although for emerging markets, WTI oil is more appropriate. Basically, hedge strategies can be applied, not only for a specific index, but also for other market indices, as hedge effectiveness is highly positively correlated across all stock markets. However, for MSCI Far
IP
T
East, we report a lower, but still significant, correlation. Moreover, in the entire sample, the
CR
lowest average hedge portfolio return is achieved for MSCI Far East. To further develop hedge strategies, it is of vital importance to be aware of hedge portfolio determinants. Among various
US
determinants, the most important one is the VIX. Hedge portfolio returns across all markets are highly negatively related to jumps in the VIX, and its economic significance is quite high, i.e.
AN
coefficients are larger than 0.5 (except for MSCI Far East). Therefore, portfolio managers should
M
take these factors into account when adjusting their hedge strategies.
ED
6. Conclusion
PT
Our study is an analysis of the characteristics of stock-oil hedges. We have used a
CE
comprehensive sample of international stock market indices both from developed as well as emerging countries and Brent oil as the key energy asset. Our results indicate that the hedge
AC
ratios of major stock market indices are time-varying and have significantly increased after the GFC. Apart from studying hedge ratios and effectiveness, this study has contributed to the literature by identifying those factors that actually drive hedge portfolio returns. Despite the hedge ratios differing among the various indices, we have identified two common drivers of hedge portfolio returns: the most important driver is the changes in the VIX, significant in all markets; the second major driver is changes in the USD/EUR spot rate. Hedge portfolio returns 31
ACCEPTED MANUSCRIPT are related negatively to both these variables. In addition, we find that since the GFC hedge effectiveness has increased and hedge portfolio returns are additionally influenced by gold returns and changes in the term structure. Our findings are important for portfolio managers, especially during periods of market stress, since they can use this information to further improve
T
their hedging performance.
IP
Moreover, future research can use these outcomes to provide additional insights in other
CR
hedging strategies with other energy assets, such as natural gas. Note that to address COP21 and COP23 concerns, investors not only need a thorough understanding of stock energy market
US
integration (see Batten, Kinateder, Szilagyi and Wagner (2018)) but also of hedging techniques,
AN
which allow them to develop adequate risk management strategies to address the impacts on
AC
CE
PT
ED
M
financial markets of climate change.
32
ACCEPTED MANUSCRIPT References Akay, O., Z. Senyuz, and E. Yoldas. 2013. Hedge fund contagion and risk-adjusted returns: A Markov-switching dynamic factor approach. Journal of Empirical Finance 22:16-29. Arouri, M. E. H, J. Jouini, and D. K. Nguyen. 2011. Volatility spillovers between oil prices and
T
stock sector returns: Implications for portfolio management, Journal of International Money and
CR
IP
Finance 30: 1387-1405.
Arouri, M. E. H., and C. Rault. 2012. Oil prices and stock markets in GCC countries: Empirical
US
evidence from panel analysis. International Journal of Finance and Economics 17:242-53.
AN
Arouri, M. E. H., J. Jouini, and D. K. Nguyen. 2012. On the impacts of oil price fluctuations on European equity markets: Volatility spillover and hedging effectiveness. Energy Economics
M
34:611-17.
ED
Baker, S. R., N. Bloom, and S. J. Davis. 2015. Measuring economic policy uncertainty. NBER
PT
Working Paper No. 21633. National Bureau of Economic Research, Cambridge, MA. Balcilar, M., R. Gupta, and M. E. Wohar. 2017. Common cycles and common trends in the stock
CE
and oil markets: Evidence from more than 150 years of data. Energy Economics 61:72-86.
66:911-46.
AC
Bao, J., J. Pan, and J. Wang. 2011. The illiquidity of corporate bonds. Journal of Finance
Basher, S. A. and P. Sadorsky. 2016. Hedging emerging market stock prices with oil, gold, VIX, and bonds: A comparison between DCC, ADCC and GO-GARCH. Energy Economics 54:23547. Batten, J. A., H. Kinateder, P. G. Szilagyi, and N. Wagner. 2017. Can stock market investors 33
ACCEPTED MANUSCRIPT hedge energy risk? Evidence from Asia. Energy Economics 66:559-70. Batten, J. A., H. Kinateder, P. G. Szilagyi, and N. Wagner. 2018. Addressing COP21 using a stock and oil market integration index. Energy Policy 116:127-36. Bauwens, L. and S. Laurent. 2005. A new class of multivariate skew densities, with application
IP
T
to generalized autoregressive conditional heteroscedasticity models. Journal of Business and
CR
Economic Statistics 23:346-354.
Beber, A. and M. W. Brandt 2009. Resolving macroeconomic uncertainty in stock and bond
US
markets. Review of Finance 13:1-45.
AN
Bekaert, G. and C. R. Harvey. 1995. Time-varying world market integration. Journal of Finance 50:403-44.
M
Bekaert, G. and C. R. Harvey. 1997. Emerging equity market volatility. Journal of Financial
ED
Economics 43:29-77.
of Finance 55:565-613.
PT
Bekaert, G. and C. R. Harvey. 2000. Foreign speculators and emerging equity markets. Journal
CE
Bernanke, B. 2016. The relationship between stocks and oil prices. Brookings Institute. Friday,
AC
February 19, 2016 https://www.brookings.edu/blog/ben-bernanke/2016/02/19/the-relationshipbetween-stocks-and-oil-prices/ Carr, P., H. Geman, and D. B. Madan. 2001. Pricing and hedging in incomplete markets. Journal of Financial Economics 62:131-67. Cecchetti, S. G., R. E. Cumby, and S. Figlewski. 1988. Estimation of the optimal futures hedge. Review of Economics and Statistics 70:623-30. 34
ACCEPTED MANUSCRIPT Chang, C.-L., M. McAleer, and R. Tansuchat. 2011. Crude oil hedging strategies using dynamic multivariate GARCH, Energy Economics 33: 912-923. Cheung, Y. W. and M. D. Chinn. 2001. Currency traders and exchange rate dynamics: A survey of the US market. Journal of International Money and Finance 20:439-71.
IP
T
Chkili, W., C. Aloui, and D. K. Nguyen. 2014. Instabilities in the relationships and hedging
CR
strategies between crude oil and U.S. stock markets: Do long memory and asymmetry matter? Journal of International Financial Markets, Institutions & Money 33:354-66.
US
Ciner C., C. Gurdgiev, and B. M. Lucey. 2013. Hedges and safe havens: An examination of
AN
stocks, bonds, gold, oil and exchange rates. International Review of Financial Analysis 29: 20211.
M
Das, D., Kumar, S.B., Tiwari, A.K., Shahbaz, M., Hasim, H.M. 2018. On the relationship of
ED
gold, crude oil, stocks with financial stress: A causality-in-quantiles approach. Finance Research
PT
Letters, 27: 169-174.
Demirer, R., H. T. Lee, and D. Lien. 2015. Does the stock market drive herd behavior in
CE
commodity futures markets? International Review of Financial Analysis 39:32-44.
AC
Engle, R. 2002. Dynamic conditional correlation - A simple class of multivariate GARCH models. Journal of Business and Economic Statistics 20:339-50. Figlewski, S. 1984. Hedging performance and basis risk in stock index futures. Journal of Finance 39:657-69. Gerard, B., K. Thanyalakpark, and J. A. Batten. 2003. Are the East Asian markets integrated? Evidence from the ICAPM. Journal of Economics and Business 55:585-607. 35
ACCEPTED MANUSCRIPT Gilje, E. P. and J. P. Taillard. 2017. Does hedging affect firm value? Evidence from a natural experiment. Review of Financial Studies 30:4083-4132. Glosten, L. R., R. Jagannathan, and D. E. Runkle. 1993. On the relation between the expected value and the volatility of the nominal excess return on stocks. Journal of Finance 48:1779-
IP
T
1801.
CR
Guesmi, K., F. Teulon, and A. T. Muzaffar. 2014. The evolution of risk premium as a measure for intra-regional equity market integration. International Review of Financial Analysis 35:13-9.
US
Guesmi, K. and F. Teulon. 2014. The determinants of regional stock market integration in
AN
Middle East: A conditional ICAPM approach. International Economics 137:22-31. Gupta, K. 2016. Oil price shocks, competition, and oil & gas stock returns - Global evidence.
M
Energy Economics 57:140-53.
ED
Huang, R. D., R. W. Masulis, and H. R. Stoll. 1996. Energy shocks and financial markets.
PT
Journal of Futures Markets 16:1-27.
Hull, J. and A. White. 1987. Hedging the risks from writing foreign currency options. Journal of
CE
International Money and Finance 6:131-52.
AC
Hull, J. and W. Suo. 2002. A methodology for assessing model risk and its application to the implied volatility function model. Journal of Financial and Quantitative Analysis 37: 297-318. Husain, S., Tiwari, A.K., Sohag, K., Shahbaz, M. (2019). Connectedness among crude oil prices, stock index and metal prices: An application of network approach in the USA. Resources Policy, 62: 57-65. Jeon, J., Y. Oh, and D. Yang. 2006. Financial market integration in East Asia: Regional or 36
ACCEPTED MANUSCRIPT global? Asian Economic Papers 5:73-89. Junttila, J., J. Pesonen, and J. Raatikainen. 2018. Commodity market based hedging against stock market risk in times of financial crisis: The case of crude oil and gold. Journal of International Financial Markets, Institutions and Money 56: 255-280.
IP
T
Kroner, K. F. and J. Sultan. 1993. Time-varying distributions and dynamic hedging with foreign
CR
currency futures. Journal of Financial and Quantitative Analysis 28: 535-51. Kyrtsou, C., C. Mikropoulou, and A. Papana. 2016. Does the S&P500 index lead the crude oil
US
dynamics? A complexity-based approach. Energy Economics 56:239-46.
AN
Lin, B., P. K. Wesseh Jr., and M. O. Appiah. 2014. Oil price fluctuation, volatility spillover and the Ghanaian equity market: Implication for portfolio management and hedging effectiveness.
M
Energy Economics 42:172-82.
ED
Maghyereh, A.I., B. Awartani, and P. Tziogkidis. 2017. Volatility spillovers and cross-hedging
Economics 68: 440-453.
PT
between gold, oil and equities: Evidence from the Gulf Cooperation Council countries, Energy
CE
Mensi, W., Hammoudeh, S., Al-Jarrah, I.M.W., Al-Yahyaee, K.H., Kang, S.H. (2019
AC
forthcoming). Risk spillovers and hedging effectiveness between major commodities, and Islamic and conventional GCC banks Journal of International Financial Markets, Institutions and Money.
Merton, R. C. 1973. An intertemporal capital asset pricing model. Econometrica 41:867-87. Myers, R.J. 1991. Estimating time-varying optimal hedge ratios on futures markets. Journal of Futures Markets 11:39-53. 37
ACCEPTED MANUSCRIPT Narayan, P. K., S. Narayan, and X. Zheng. 2010. Gold and oil futures markets: Are markets efficient? Applied Energy 87:3299-3303. Narayan, P. K. and R. Gupta. 2015. Has oil price predicted stock returns for over a century? Energy Economics 48:18-23.
IP
T
Narayan, P. K. and R. Liu. 2015. A unit root model for trending time-series energy variables.
CR
Energy Economics 50:391-402.
Narayan, P. K. and S. S. Sharma. 2016. Intraday return predictability, portfolio maximisation,
US
and hedging. Emerging Markets Review 28:105-116.
AN
Newey, W. K. and K. D. West. 1987. Hypothesis testing with efficient method of moments estimation. International Economic Review 28:777-87.
M
Panagiotis F., N. Tasios, L. Paroussos, P. Capros, and S. Tsani. 2017. Energy system impacts and
ED
policy implications of the European Intended Nationally Determined Contribution and low-
PT
carbon pathway to 2050. Energy Policy 100:216-26. Primbs, J. A. and Y. Yamada. 2006. A moment computation algorithm for the error in discrete
CE
dynamic hedging. Journal of Banking and Finance 30:519-40.
AC
Raza, N., S. Ali, S. J. H. Shahzad, and S. A. Raza. 2018. Do commodities effectively hedge real estate risk? A multi-scale asymmetric DCC approach. Resources Policy. Advance Access Published February 1, 2018. Sadorsky, P. 2014. Modeling volatility and correlations between emerging market stock prices and the prices of copper, oil and wheat. Energy Economics 43:72-81. Singhal, S., Choudhary, S., Biswal, P.C. 2019. Return and volatility linkages among 38
ACCEPTED MANUSCRIPT International crude oil price, gold price, exchange rate and stock markets: Evidence from Mexico Resources Policy, 60: 255-261. Solnik, B. H. 1977. Testing international asset pricing: Some pessimistic views. Journal of Finance 32:503-12.
IP
T
Stulz, R. M. 1981. A model of international asset pricing. Journal of Financial Economics
CR
9:383-406.
Tankov, P. and E. Voltchkova. 2009. Asymptotic analysis of hedging errors in models with
US
jumps. Stochastic Processes and their Applications 119: 2004-27.
AN
Tsai, C. L. 2015. How do U.S. stock returns respond differently to oil price shocks pre-crisis, within the financial crisis, and post-crisis? Energy Economics 50:47-62.
ED
Research Observer 15: 85-98.
M
Tsetsekos, G. and P. Varangis. 2000. Lessons in structuring derivatives exchanges. World Bank
PT
Vandyck, T., K. Keramidas, B. Saveyn, A. Kitous, and Z. Vrontisi. 2016. A global stocktake of
41:46-63.
CE
the Paris pledges: Implications for energy systems and economy. Global Environmental Change
AC
Varela, O. 1999. Futures and realized cash or settle prices for gold, silver and copper. Review of Financial Economics 8:121-38. Yang, J., Y. Zhou, and Z. Wang. 2009. The stock-bond correlation and macroeconomic conditions: One and a half centuries of evidence. Journal of Banking and Finance 33: 670-80.
39
ACCEPTED MANUSCRIPT Table 1a Descriptive statistics of returns Mean Median Std. dev. Skewness Kurtosis Jarque-Bera p-value Observations
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500 Brent Oil WTI Oil 0.501669 0.389993 0.387431 0.373763 0.383295 -0.017262 0.598390 0.602257 0.369065 0.302224 0.917182 1.037763 0.916619 0.836937 0.860256 0.271430 1.074099 1.043291 0.588023 0.731658 6.670794 4.294740 4.378431 4.999437 4.230733 5.619391 4.168421 4.129745 9.279001 9.298459 -1.056040 -0.861708 -0.897674 -0.816177 -0.805854 -0.135577 -0.835864 -0.801277 -0.182283 -0.162851 6.468644 5.131678 5.403398 5.008401 4.889204 4.187888 5.065367 4.856276 5.433638 4.751331 230.8932 0.000000
105.1990 0.000000
125.9943 0.000000
93.77552 0.000000
86.33376 0.000000
336
336
336
336
336
I R
T P
20.78444 0.000031
98.84587 0.000000
84.19516 0.000000
84.77703 0.000000
44.55758 0.00000
336
336
336
336
336
C S
U N
This table reports descriptive statistics (mean, standard deviation, skewness, kurtosis and Jarque-Beraβs normality test) for returns of various stock market indices and ICE-Brent as well as NYMEX WTI near month futures contract. Returns of asset i are measured as the difference in the natural logarithm of intermonth prices in percentage, i.e. π π,π‘ = (ln ππ,π‘ β ln ππ,π‘β1 ) β 100. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
A
D E
M
T P
E C
C A
40
ACCEPTED MANUSCRIPT Table 1b Pearson correlations of variables MSCI EM p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 1.000000 0.769324 0.0000
1.000000 -
MSCI ACWI p-value
0.804429 0.0000
0.997809 0.0000
1.000000 -
MSCI Europe p-value
0.722559 0.0000
0.921759 0.0000
0.922014 0.0000
1.000000 -
MSCI G7 p-value
0.745116 0.0000
0.996772 0.0000
0.992349 0.0000
0.896877 0.0000
MSCI Far East p-value
0.588430 0.0000
0.755847 0.0000
0.755483 0.0000
0.598100 0.0000
MSCI NA p-value
0.716476 0.0000
0.917140 0.0000
0.914326 0.0000
S&P 500 p-value
0.699329 0.0000
0.912564 0.0000
MSCI NA
S&P 500
Brent Oil p-value
0.204103 0.0002
C S
A
U N
1.000000 -
1.000000 -
0.810193 0.0000
0.921160 0.0000
0.524313 0.0000
1.000000 -
0.907763 0.0000
0.805403 0.0000
0.917456 0.0000
0.518910 0.0000
0.997384 0.0000
1.000000 -
0.165824 0.0023
0.163056 0.0027
0.148326 0.0065
0.157928 0.0037
0.114896 0.0353
0.091852 0.0928
D E
Brent Oil
I R
0.763566 0.0000
T P
E C
MSCI Far East
T P
MSCI MXWO p-value
0.153232 0.0049
MSCI G7
M
1.000000 -
C A
This table reports the pairwise Pearson correlations and associated p-values for monthly variable returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
41
ACCEPTED MANUSCRIPT Table 2 Parameter estimates of bivariate DCC-GJR-GARCH(1,1) MSCI EM MSCI MXWO MSCI ACWI MSCI Europe Mean Equation π p-value
MSCI G7
MSCI Far East MSCI NA
0.492116 0.238689
0.495957** 0.026613
0.500949** 0.030876
0.525236** 0.044339
0.525380** 0.013494
0.143560 0.752264*** 0.625527 0.000007
0.154673*** π p-value 0.005325 Variance Equation 3.334564** π€0 p-value 0.049513
-0.015708 0.781742
-0.005653 0.919682
-0.008022 0.906867
-0.021915 0.704889
0.098521* 0.094447
1.532206 0.161893
1.664649 0.112212
2.164136 0.346865
1.292936 0.154903
0.785842 0.473571
S&P 500
Brent Oil
0.746079*** 0.000010
0.002115 0.683801
-0.101174* 0.081787
-0.104200* 0.084766
0.211395*** 0.000873
0.833691* 0.056992
0.870437* 0.064882
0.000985** 0.035416
0.065160 0.118022** 0.500583 0.042473
0.102502* 0.068234
0.073852 0.117378
0.171770 0.134080
0.192938 0.105204
0.171506 0.118710
I R
T P
C S
π€1 p-value
0.069289 0.398194
0.002912 0.961365
0.000000 1.000000
0.000000 0.999999
0.020125 0.706355
π€2 p-value
0.082864 0.389866
0.222101* 0.055759
0.225322* 0.051161
0.165688 0.238272
0.803628*** 0.000000
0.775475*** 0.000000
0.772163*** 0.000000
0.806286*** 0.000001
U N
0.777880*** 0.000000
0.850578*** 0.742873*** 0.000000 0.000000
0.744195*** 0.000000
0.688586*** 0.000000
0.039244 0.000628***
0.053439 0.025825**
0.048700 0.013370**
0.028480 0.004485***
0.062791 0.038128**
0.113628 0.024849 0.057417* 0.004067***
0.072932 0.005392***
-
0.951988*** π p-value 0.000000 Diagnostics Log-likelihood -2,263.610 Bayesian Info Criterion 13.820 Q(5) 7.626 Q(10) 15.096 QΒ²(5) 0.985 QΒ²(10) 6.647
0.933797*** 0.000000
0.940893*** 0.000000
0.963372*** 0.000000
0.922291*** 0.000000
0.559227 0.917488*** 0.106860 0.000000
0.912883*** 0.000000
-
-2,114.544 12.933 1.705 4.409 0.244 0.838
-2,118.986 12.959 1.775 4.105 0.271 1.038
-2,169.326 13.259 2.389 3.592 2.554 6.843
-2,109.935 12.905 1.863 4.929 0.234 0.628
-2,101.142 12.853 2.006 7.146 0.776 1.626
4.355 11.924 1.082 4.156
π€3 p-value DCC Equation π p-value
C A
E C
A
D E
T P
0.209685** 0.040917
M
0.108119** 0.036111
-2,214.271 13.526 7.712 12.895 1.317 8.017
-2,100.840 12.851 1.909 6.957 0.557 1.520
This table reports estimated coefficients from a bivariate DCC-GJR-GARCH(1,1) model and associated p-values on monthly returns of various stock market indices and ICE-Brent near month futures contract. The first part contains results of the univariate conditional mean (see Eq. (7)) and conditional variance equation (see Eq. (8)). Q(k) and QΒ²(k) represent the Ljung-Box test for serial correlation up to order k applied to standardized residuals and squared standardized residuals, respectively. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from January 1990 to December 2017.
42
ACCEPTED MANUSCRIPT Table 3 Structural breaks of hedge ratio Date SupF p-value F(September 2008) p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500 September 2007 September 2008 September 2008 September 2008 September 2008 September 2013 September 2008 September 2008 608.46 474.81 534.45 987.99 403.33 19.06 471.06 414.48 0.0000 0.0000 0.0000 0.0000 0.0000 0.0004 0.0000 0.0000 441.24 0.46 0.0000 0.4943
T P
I R
This table reports results of structural breakpoint tests for the hedge ratio (i.e. time-varying beta) between returns of various stock market indices and ICE-Brent near month futures contract (see Eq. (4)). The date and the significance of a structural break are obtained from a supremum F (supF) test. In addition, an additional Chow test is performed for September 2008 of MSCI EM and MSCI Far East. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
C S
U N
A
D E
M
T P
E C
C A
43
ACCEPTED MANUSCRIPT Table 4 Descriptive statistics of hedge effectiveness Mean Median Maximum Minimum Std. dev. Skewness Kurtosis
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.096762 0.080642 0.084693 0.068768 0.039339 0.023668 0.020845 0.013007 0.401539 0.507081 0.504094 0.350585 0.000000 0.000002 0.000000 0.000000 0.114444 0.120469 0.124176 0.093751 1.122816 1.735279 1.605678 1.190511 2.820630 4.814569 4.361887 3.099851
MSCI G7 MSCI Far East 0.082618 0.058725 0.028973 0.047789 0.523758 0.409864 0.000000 0.000036 0.121580 0.054695 1.783704 2.330051 4.977962 11.72842
MSCI NA 0.084449 0.024876 0.544066 0.000003 0.126393 1.835536 5.259867
T P
I R
C S
S&P 500 0.074702 0.023809 0.496231 0.000030 0.111926 1.894844 5.557018
This table reports the descriptive statistics (mean, standard deviation (std. dev.), skewness, and kurtosis) of the hedge effectiveness (see Eq. (13)) between monthly returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only).The sample period is from January 1990 to December 2017.
U N
A
D E
M
T P
E C
C A
44
ACCEPTED MANUSCRIPT Table 4b Pearson correlations of hedge effectiveness MSCI EM p-value
MSCI EM MSCI MXWO 1.000000 -
MSCI ACWI
MSCI Europe
MSCI G7
MSCI Far East
0.923958 0.0000
1.000000 -
MSCI ACWI p-value
0.942035 0.0000
0.996656 0.0000
1.000000 -
MSCI Europe p-value
0.916585 0.0000
0.930474 0.0000
0.952912 0.0000
1.000000 -
C S
MSCI G7 p-value
0.909533 0.0000
0.995157 0.0000
0.985435 0.0000
0.897711 0.0000
1.000000 -
MSCI Far East p-value
0.341264 0.0000
0.315677 0.0000
0.304704 0.0000
MSCI NA p-value
0.845860 0.0000
0.967960 0.0000
D E
S&P 500 p-value
0.799041 0.0000
0.939874 0.0000
E C
S&P 500
T P
MSCI MXWO p-value
T P
MSCI NA
I R
U N
A
M
0.228511 0.0000
0.339429 0.0000
1.000000 -
0.950744 0.0000
0.840021 0.0000
0.978696 0.0000
0.279494 0.0000
1.000000 -
0.916668 0.0000
0.786859 0.0000
0.956904 0.0000
0.273429 0.0000
0.993695 0.0000
1.000000 -
C A
This table reports the pairwise Pearson correlations and associated p-values for the estimated hedge effectiveness (see Eq. (13)) between monthly returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
45
ACCEPTED MANUSCRIPT Table 5 Economic significance of hedging strategy Risk aversion
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe Panel A: No transaction costs
MSCI G7
MSCI Far East
MSCI NA
S&P 500
πΎ = 3 (low risk)
0.872512
0.427729
0.770677
0.332818
0.150124
πΎ
1.957354 1.366790 4.127039 1.628678 1.957354 1.198902 1.455294 0.272512 1.024826 0.699807 1.357354
1.198902
1.455294
1.024826
0.699807
2.741247
2.824528
2.408843
1.799174
πΎ
πΎ πΎ
1.392524 0.540875 0.697479 0.540875 3.374201 1.366790 1.628678 = 6 (medium risk) 0.697479 1.366790 1.366790 1.366790 0.872512 7.337556 3.018620 3.491076 = 12 (high risk) 1.628678 1.628678 1.628678 0.427729 1.957354 1.957354 1.957354 0.770677 Panel1.198902 B: Transaction1.198902 costs (TC = 0.51.198902 %) 0.332818 1.455294 1.455294 1.455294 0.150124 0.792524 -0.059125 0.097479 = 3 (low risk) 1.024826 1.024826 1.024826 0.699807 0.699807 0.699807 2.774201 0.766790 1.028678 = 6 (medium risk)
πΎ = 12 (high risk)
6.737556
2.418620
2.891076
M
I R
C S
-0.172271
0.170677
-0.267182
-0.449876
0.598902
0.855294
0.424826
0.099807
2.141247
2.224528
1.808843
1.199174
U N
A
3.527039
T P
This table reports the estimated average annualized percentage utility gain from Eq. (14) for different levels of risk aversion πΎ β {3,6,12}. Panel A shows results without transaction costs, and Panel B shows results with transaction costs (TC), where TC is set to 0.5%. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). Negative values indicate a situation where hedging is economically not efficient. The sample period is from January 1990 to December 2017.
D E
T P
E C
C A
46
ACCEPTED MANUSCRIPT Table 6 Descriptive statistics of hedge portfolio returns Mean Median Maximum Minimum Std. dev. Skewness Kurtosis
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.452573 0.366240 0.367954 0.356068 0.918969 0.696979 0.682551 0.711725 14.27125 10.77109 10.68602 12.30143 -34.72463 -14.49810 -15.28647 -14.91119 6.244585 4.018779 4.072345 4.688381 -0.920370 -0.604363 -0.608776 -0.582073 5.829862 3.899688 3.957707 3.912106
MSCI G7 MSCI Far East 0.354674 -0.010090 0.625749 0.253597 11.10513 18.99771 -14.19026 -19.36308 3.969503 5.447554 -0.574044 0.003237 3.799931 4.013427
MSCI NA 0.568457 0.926287 9.564921 -15.83233 3.931093 -0.707119 4.251608
T P
I R
C S
S&P 500 0.568961 0.970744 10.31122 -16.11077 3.940545 -0.745142 4.476434
This table reports the descriptive statistics (mean, standard deviation (std. dev.), skewness, and kurtosis) of hedge portfolio returns (see Eq. (1)) between monthly returns of various stock market indices and ICE-Brent near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
U N
A
D E
M
T P
E C
C A
47
ACCEPTED MANUSCRIPT Table 7a: Determinants of hedge portfolio returns (full sample) Constant p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.6550** 0.4755** 0.4976** 0.4908* 0.0272 0.0200 0.0137 0.0821
MSCI G7 MSCI Far East 0.4607** 0.1818 0.0235 0.5304
MSCI NA 0.6001*** 0.0051
S&P 500 0.5758** 0.0120
0.0560 0.4776
T P
0.0627 0.1653
0.0593 0.2079
-0.1047* 0.0563
-0.0493 0.2193
-0.0525 0.2048
0.0230 0.6444
-0.0891 0.1145
-0.1018* 0.0778
-0.0398 0.4271
-0.0529 0.3251
-0.0130 0.7863
-0.0015 0.9778
-0.0551 0.1961
-0.0393 0.3668
-0.0482 0.2901
-0.0496 0.2748
-0.3323*** 0.0000
-0.1349*** 0.0025
-0.1268*** 0.0054
-0.0095 0.8362
-0.0154 0.7389
AR(1) p-value
0.1752*** 0.0001
0.1019** 0.0299
0.1047** 0.0204
0.1085*** 0.0064
0.0892* 0.0731
βπΈππ p-value
-0.0806 0.1303
-0.1006* 0.0545
-0.0963* 0.0654
-0.1397*** 0.0016
-0.0997* 0.0640
π πΊπππ p-value
0.0663 0.1938
-0.0784 0.1312
-0.0629 0.2120
-0.1073** 0.0236
βπΌπΉ p-value
-0.0840 0.0612
-0.0441 0.3839
-0.0535 0.2899
-0.0443 0.4552
βππ p-value
-0.0056 0.9028
-0.0553 0.1825
-0.0529 0.1970
-0.0586 0.1573
βπππ·/πΈππ p-value
-0.0334 0.4853
-0.1631*** 0.0002
-0.1597*** 0.0002
βππΌπ p-value
-0.5333*** 0.0000
-0.5870*** 0.0000
-0.5927*** 0.0000
-0.5076*** 0.0000
-0.5792*** 0.0000
-0.2980*** 0.0000
-0.6615*** 0.0000
-0.6628*** 0.0000
Adj. π 2 F-statistic p-value
0.3355 25.087*** 0.0000
0.4475 39.642*** 0.0000
0.4498 40.011*** 0.0000
0.4815 45.308*** 0.0000
0.4255 36.335*** 0.0000
0.1340 8.384*** 0.0000
0.4680 42.978*** 0.0000
0.4757 44.295*** 0.0000
E C
T P
D E
-0.0859 0.1125
U N
A
M
C S
I R
This table reports estimated regression coefficients from Eq. (12) in standardized form. The dependent variable is the estimated hedge portfolio return (see Eq. (1) and Table 6) on monthly frequency. Robust p-values are calculated according to Newey and West (1987). βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing. βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban Consumers. βπππ‘ are changes in the U.S. term structure given by the difference between 10-year and 2-year Treasury bond yields. βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate. βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index (VIX). The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from January 1990 to December 2017.
C A
48
ACCEPTED MANUSCRIPT Table 7b Determinants of hedge portfolio returns (subsample 1: January 1990 β August 2008) Constant p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.3002 0.5043* 0.4838 0.5834 0.5213 0.0940 0.1087 0.1059
MSCI G7 MSCI Far East 0.4989 -0.3977 0.1003 0.4113
MSCI NA 0.8042*** 0.0071
S&P 500 0.8254*** 0.0054
0.0430 0.6565
T P
0.0629 0.2070
0.0501 0.3318
-0.1035 0.2412
-0.0388 0.5587
-0.0420 0.5353
0.0795 0.1857
-0.0238 0.6840
-0.0336 0.5629
-0.0482 0.4723
0.0285 0.6882
-0.0537 0.3352
-0.0593 0.2901
-0.0294 0.5242
-0.0672 0.1973
-0.0025 0.9588
-0.0024 0.9611
-0.3494*** 0.0000
-0.1488** 0.0123
-0.1251** 0.0175
-0.0148 0.8161
-0.0209 0.7415
AR(1) p-value
0.2295*** 0.0003
0.1122* 0.0768
0.1208** 0.0487
0.1374** 0.0164
0.0926 0.1597
βπΈππ p-value
-0.1790*** 0.0019
-0.1149 0.1994
-0.1212 0.1684
-0.1648** 0.0187
-0.1084 0.2409
π πΊπππ p-value
0.0498 0.4610
-0.0184 0.7501
-0.0132 0.8190
-0.0675 0.2410
βπΌπΉ p-value
0.0040 0.9468
-0.0466 0.4920
-0.0422 0.5388
-0.0406 0.5609
βππ p-value
0.0896 0.0868*
-0.0262 0.5648
-0.0160 0.7222
-0.0302 0.5912
βπππ·/πΈππ p-value
-0.0300 0.6329
-0.1752*** 0.0026
-0.1689*** 0.0034
βππΌπ p-value
-0.5351*** 0.0000
-0.5720*** 0.0000
-0.5794*** 0.0000
-0.5074*** 0.0000
-0.5593*** 0.0000
-0.2755*** 0.0005
-0.6521*** 0.0000
-0.6540*** 0.0000
Adj. π 2 F-statistic p-value
0.3574 18.637*** 0.0000
0.3647 19.205*** 0.0000
0.3721 19.795*** 0.0000
0.3823 20.628*** 0.0000
0.3449 17.697*** 0.0000
0.086 3.965*** 0.0004
0.431 25.040*** 0.0000
0.440 25.920*** 0.0000
E C
T P
D E
-0.0220 0.7124
U N
A
M
C S
I R
This table reports estimated regression coefficients from Eq. (12) in standardized form. The dependent variable is the estimated hedge portfolio return (see Eq. (1) and Table 6) on monthly frequency. Robust p-values are calculated according to Newey and West (1987). βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing. βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban Consumers. βπππ‘ are changes in the U.S. term structure given by the difference between 10-year and 2-year Treasury bond yields. βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate. βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index (VIX). The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from January 1990 to August 2008.
C A
49
ACCEPTED MANUSCRIPT Table 7c Determinants of hedge portfolio returns (subsample 2: September 2008 β December 2017) Constant p-value
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe 0.4847 0.4007 0.4350* 0.2104 0.1653 0.1062 0.0813 0.5947
MSCI G7 MSCI Far East 0.4036* 0.6915*** 0.0896 0.0091
MSCI NA 0.3853* 0.0912
S&P 500 0.3581 0.1313
0.0968 0.1347
0.0529 0.4710
0.0617 0.4102
-0.1250* 0.0857
-0.0761 0.1940
-0.0490 0.1874
-0.1277 0.1817
-0.2681** 0.0375
-0.1988** 0.0329
0.0289 0.6260
-0.1748*** 0.0008
0.0847 0.1520
0.1378* 0.0724
-0.1686* 0.0792
-0.0266 0.7354
-0.2074* 0.0762
-0.1189* 0.0662
-0.3538*** 0.0000
-0.1724*** 0.0030
-0.1848** 0.0137
-0.0929 0.1102
-0.1827 0.1266
T P
AR(1) p-value
0.0540 0.2434
0.0583 0.2638
0.0616 0.2144
0.0494 0.2864
0.0539 0.3341
βπΈππ p-value
-0.0076 0.9076
-0.0948* 0.0856
-0.0828 0.1270
-0.1221** 0.0320
-0.0993* 0.0784
π πΊπππ p-value
-0.0171 0.8291
-0.2439** 0.0187
-0.2142** 0.0330
-0.2128*** 0.0042
-0.2605** 0.0196
βπΌπΉ p-value
-0.1291* 0.0664
0.0184 0.7635
-0.0087 0.8888
-0.0013 0.9884
βππ p-value
-0.1785** 0.0221
-0.1665* 0.0572
-0.1632* 0.0508
-0.1339** 0.0246
βπππ·/πΈππ p-value
-0.2584*** 0.0000
-0.2097*** 0.0004
-0.2259*** 0.0001
βππΌπ p-value
-0.5143*** 0.0000
-0.6236*** 0.0000
-0.6203*** 0.0000
-0.5069*** 0.0000
-0.6324*** 0.0000
-0.4363*** 0.0000
-0.6686*** 0.0000
-0.6715*** 0.0000
Adj. π 2 F-statistic p-value
0.507 17.320*** 0.0000
0.661 31.967*** 0.0000
0.657 31.418*** 0.0000
0.650 30.474*** 0.0000
0.644 29.719*** 0.0000
0.371 10.361*** 0.0001
0.623 27.204*** 0.0000
0.6304 28.049*** 0.0000
E C
T P
D E
U N
A
M
C S
I R
This table reports estimated regression coefficients from Eq. (12) in standardized form. The dependent variable is the estimated hedge portfolio return (see Eq. (1) and Table 6) on monthly frequency. Robust p-values are calculated according to Newey and West (1987). βπΈπππ‘ are changes in the U.S. economic policy uncertainty (EPU) index. π πΊπππ,π‘ are continuously compounded returns of gold spot prices from the London afternoon fixing. βπΌπΉπ‘ are changes in U.S. inflation measured by the Consumer Price Index for All Urban Consumers. βπππ‘ are changes in the U.S. term structure given by the difference between 10-year and 2-year Treasury bond yields. βπππ·/πΈππ π‘ are changes in the USD/EUR spot rate. βππΌππ‘ are changes in the Chicago Board Options Exchange implied volatility index (VIX). The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The 1, 5 and 10 percent significance levels are denoted by ***, ** and *, respectively. The sample period is from September 2008 to December 2017.
C A
50
ACCEPTED MANUSCRIPT Figure 1 Hedge ratio
Hedge Ratio
T IP CR
0.2
1993
1996
1999
2002
2005
2008
2011
2014
2017
ED
M
1990
AN
-0.2
US
0.0
Beta
0.4
0.6
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500
AC
CE
PT
This figure plots the hedge ratio (i.e. time-varying beta) between returns of various stock market indices and ICEBrent near month futures contract (see Eq. (4)), estimated by a DCC-GJR-GARCH(1,1) model. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
51
ACCEPTED MANUSCRIPT Figure 2 Hedge effectiveness
0.6
Hedge Effectivness
T IP CR
0.3
1993
1996
1999
2002
2005
2008
2011
2014
2017
M
1990
AN
0.0
US
0.1
0.2
HE
0.4
0.5
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500
AC
CE
PT
ED
This figure plots the time-varying hedge effectiveness (HE) between returns of various stock market indices and ICE-Brent near month futures contract (see Eq. (13)). HE is defined as the proportion of the variance reduced by the hedge. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
52
ACCEPTED MANUSCRIPT Appendix Table A.1 WTI oil: Descriptive statistics of hedge effectiveness Mean Median Maximum Minimum Std. dev. Skewness Kurtosis
MSCI EM 0.106206 0.058941 0.480002 0.000000 0.124252 1.253202 3.278239
MSCI MXWO 0.074713 0.026829 0.461438 0.000010 0.108192 1.806564 5.121009
MSCI ACWI 0.079762 0.027955 0.474238 0.000000 0.114478 1.730364 4.833269
MSCI Europe 0.064518 0.020395 0.336601 0.000000 0.085059 1.356519 3.738521
MSCI G7 0.076508 0.030881 0.472236 0.000000 0.108770 1.832160 5.229974
MSCI Far East 0.054788 0.047530 0.348687 0.000708 0.041836 2.881463 16.60682
C S
I R
T P
MSCI NA 0.076188 0.023000 0.497951 0.000000 0.115645 1.883080 5.493120
S&P 500 0.067648 0.022678 0.452211 0.000002 0.102751 1.926613 5.768798
This table reports the descriptive statistics (mean, standard deviation (std. dev.), skewness, and kurtosis) of the hedge effectiveness (see Eq. (13)) between monthly returns of various stock market indices and NYMEX WTI near month futures contract. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only).The sample period is from January 1990 to December 2017.
U N
A
D E
M
T P
E C
C A
53
ACCEPTED MANUSCRIPT Figure A.1 WTI oil: Hedge ratio
Hedge Ratio
T IP CR
0.2
1996
1999
2002
M
1993
2005
2008
2011
2014
2017
ED
1990
AN
-0.2
US
0.0
Beta
0.4
0.6
MSCI EM MSCI MXWO MSCI ACWI MSCI Europe MSCI G7 MSCI Far East MSCI NA S&P 500
AC
CE
PT
This figure plots the hedge ratio (i.e. time-varying beta) between returns of various stock market indices and NYMEX WTI near month futures contract (see Eq. (4)), estimated by a DCC-GJR-GARCH(1,1) model. The stock market indices are: MSCI Emerging Markets (EM); MSCI MXWO (Developed Markets); MSCI ACWI (Emerging and Developed Markets); MSCI Europe; MSCI G7 (Canada, France, Germany, Italy, Japan, United Kingdom, U.S.); MSCI Far East (Japan, Hong Kong, Singapore); MSCI North America (NA; Canada, U.S.); S&P 500 (U.S. only). The sample period is from January 1990 to December 2017.
54
ACCEPTED MANUSCRIPT
Hedging Stocks with Oil Research Highlights
T
IP
CR
US AN M ED PT
ο·
CE
ο· ο· ο· ο·
The feasibility of hedging stocks with oil is determined with a dynamic conditional correlation approach. The effectiveness of energy hedges is time-varying and market-dependent. The global financial crisis changed the effectiveness of historic hedging. The most common driver of hedge portfolio returns is the VIX. The appreciation of the U.S. dollar against the euro is associated with reduced hedge portfolio returns. Brent oil is more appropriate in developed markets for stock-oil hedges than WTI oil.
AC
ο·
55
Figure 1
Figure 2