sign in sign up

<<

Does Gold Act As a Hedge and a Safe Haven for China's Stock Market?

Does Gold Act As a Hedge and a Safe Haven for China's Stock Market?

A Service of

Leibniz-Informationszentrum econstor Wirtschaft Leibniz Information Centre Make Your Publications Visible. zbw for Economics

Chen, Ke; Wang, Meng

Article Does gold act as a hedge and a safe haven for China's stock market?

International Journal of Financial Studies

Provided in Cooperation with: MDPI – Multidisciplinary Digital Publishing Institute, Basel

Suggested Citation: Chen, Ke; Wang, Meng (2017) : Does gold act as a hedge and a safe haven for China's stock market?, International Journal of Financial Studies, ISSN 2227-7072, MDPI, Basel, Vol. 5, Iss. 3, pp. 1-18, http://dx.doi.org/10.3390/ijfs5030018

This Version is available at: http://hdl.handle.net/10419/195652

Standard-Nutzungsbedingungen: Terms of use:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your Zwecken und zum Privatgebrauch gespeichert und kopiert werden. personal and scholarly purposes.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich purposes, to exhibit the documents publicly, to make them machen, vertreiben oder anderweitig nutzen. publicly available on the internet, or to distribute or otherwise use the documents in public. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, If the documents have been made available under an Open gelten abweichend von diesen Nutzungsbedingungen die in der dort Content Licence (especially Creative Commons Licences), you genannten Lizenz gewährten Nutzungsrechte. may exercise further usage rights as specified in the indicated licence.

https://creativecommons.org/licenses/by/4.0/ www.econstor.eu International Journal of Financial Studies

Article Does Gold Act as a Hedge and a Safe Haven for China’s Stock Market?

Ke Chen and Meng Wang * ID

School of Finance and Statistics, Hunan University, Changsha, Hunan 410079, China; [email protected] * Correspondence: [email protected]; Tel.: +86-157-0077-9936

Academic Editor: Nicholas Apergis Received: 28 June 2017; Accepted: 8 August 2017; Published: 17 August 2017

Abstract: This paper examines the dynamic relationships between gold and stock markets in China. Using daily gold and stock indexes data, we estimated the DCC-GARCH model for the ﬁve bear markets since 31 October 2002, and simultaneously used different segments of China’s stock markets for analysis. Our main objective was to examine the time-varying correlations between gold and stock and to check the effectiveness of gold as a hedge or a safe haven for stocks. Results showed that: (1) the dynamic conditional correlations switched between positive and negative values over the periods under study; (2) due to the increasing investment demand of gold, the hedging effect of gold on China’s stock market has strengthened remarkably. Gold acts as a safe haven for only the latest two of the ﬁve bear markets analyzed (12 June 2015–26 August 2015 and 22 December 2015– 29 February 2016); and (3) for non-bear markets, gold does not offer good risk hedging.

Keywords: gold; bear markets; correlation; safe haven; hedge

JEL Classiﬁcation: G10; G11; G14; G15

1. Introduction Gold is one of the most malleable, ductile, dense, conductive, non-destructive, brilliant, and beautiful of metals. This unique set of qualities has made it a coveted object throughout history by humans in almost every civilization, and there have been active gold markets for over 6000 years (Green 2007). As money, as an investment, as a store of value, gold has long fascinated the financial media, investors and researchers in equal measure. Since the breakdown of the Bretton Woods System, gold is no longer a central cornerstone of the international monetary system, but nevertheless still attracts considerable attention from investors and researchers. Owing to the increasing uncertainty of the global financial markets, diversifying a portfolio through hedging becomes more important (Beckmann et al. 2015) especially, since during the global financial and economic crisis that started in 2007, financial assets (in particular stock prices) exhibited losses while the gold price experienced an intense increase. Figure1 shows the evolution of the Shanghai Stock Exchange (SSE) Composite Index and the price of gold from 31 October 2016. Since the beginning of the financial crisis in October 2007, the SSE Composite Index has fallen 38%, while the gold price has risen 22%1. The performance of gold is most impressive given the losses suffered in other asset classes during the crisis. The paper aimed to investigate the dynamic relationships between gold and stocks and to test whether gold represents a safe heaven or as a hedge against China’s stock market.

1 From October 2007 to November 2008.

Int. J. Financial Stud. 2017, 5, 18; doi:10.3390/ijfs5030018 www.mdpi.com/journal/ijfs Int. J. Financial Stud. 2017, 5, 18 2 of 18 Int. J. Financial Stud. 2017, 5, 18 2 of 18

￥450 7000 ￥ 400 6000 ￥350 5000 ￥300 ￥250 4000 ￥200 3000 ￥150 2000 ￥100 ￥50 1000 ￥0 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

gold the SSE Composite Index

Figure 1. Price for gold and the Shanghai Stock Exchange (SSE) Composite Index. Figure 1. Price for gold and the Shanghai Stock Exchange (SSE) Composite Index.

Among all financial assets, gold is quite unique and virtually sits as its own asset class that differs fromAmong other precious all financial metals assets, including gold issilver, quite platinum, unique and and virtuallypalladium sits (Batten as its et own al. 2010, asset 2014 class). thatOne differsreason fromis that other its usefulness precious metalsas an industrial including metal silver, isplatinum, small and anddeclining palladium when ( Battencompared et al. with 2010 its, 2014investment). One reason and uses is that in its jewelry. usefulness The otheras an industrial precious metals metal is still small have and signif decliningicant useswhen in compared industry: withplatinum its investment is commonly and used uses in in catalysts, jewelry. palladium The other is precious now mixed metals into still many have of significant the alloys usesthat are in industry:replacing gold platinum in dentistry is commonly and silver used can in be catalysts, used in the palladium production is nowof solar mixed panels into (O’Connor many of et the al. alloys2015). thatAccording are replacing to the gold 2016 in Chine dentistryse Gold and Yearbook silver can published be used in by the the production China Gold of solar Association, panels (O’Connorphysical gold et al. transactions 2015). According in the Shanghai to the 2016 gold Chinese exchange Gold have Yearbook been the published highest in by the the world China for Gold nine Association,consecutive physicalyears. Figure gold 2 transactions shows the gold in the production Shanghai goldand consumption exchange have by beenChina the from highest 2009 in–2016. the worldGold production for nine consecutive increased years.annually Figure and2 reachedshows the 453,486 gold productiontons in 2016. and China consumption has been the by world’s China fromlargest 2009–2016. gold producer Gold productionfor 10 consecutive increased years. annually The gold and consumption reached 453,486 of China tons inreached 2016. 975.38 China tons has beenin 2016, the and world’s China largest has been gold the producer world’s top for 10gold consecutive consumer for years. four The consecutive gold consumption years; however, of China gold reachedconsumption 975.38 in tons 2016 in declined 2016, and by China 6.74% has when been compared the world’s to top2015. gold From consumer 2015 to for2016, four the consecutive use of gold years;in jewelry however, in China gold decreased consumption by 18.91%, in 2016 while declined the byuse 6.74% of gold when bars comparedincreased toby 2015.28.19% From and2015 the use to 2016,of gold the coins use of increased gold in jewelry by 10.14%. in China Although decreased the consumption by 18.91%, while of gold the usehas ofsharply gold bars declined, increased gold by as 28.19%an investment and the usein China of gold has coins increased increased dramatically, by 10.14%. with Although a total the growth consumption of almost of gold30%. has Gold sharply has a declined,dual nature gold of ascommodity an investment and finance, in China so has the increased demand dramatically,for gold is generally with a totaldivided growth into oftwo almost parts: 30%.the consumption Gold has a dual demand nature and of the commodity investment and demand. finance, With so the the demand development for gold of isChina’s generally gold divided market, intothere two is a parts: growing the variety consumption of gold demand investment and demands the investment as well demand.as gold investment With the developmentproducts. As the of China’sworld’s gold largest market, gold there consumer is a growing and producer, variety of checki gold investmentng whether demands gold is asa safe well haven as gold or investment hedge for products.China’s stock As the market world’s is largestof great gold significance consumer for and Chinese producer, investors, checking as whether well as goldinvestment is a safe product haven ordesigners. hedge for While China’s the stock research market related is of to great this significancefield are rather for scarce Chinese, one investors, existing as study well on as the investment hedging productpotential designers. of gold traded While on the the researchrelatively related new Shanghai to this field Gold are Exchange rather scarce, was that one of existingHoang et study al. (2015) on . theThey hedging investigated potential the ofrole gold of gold traded quoted on the on relatively the Shanghai new Gold Shanghai Exchange Gold in Exchange the diversification was that ofof HoangChinese et portfolios al.(2015). Theyfrom investigated2004–2014. Against the role this of goldbackground, quoted on this the paper Shanghai estimated Gold the Exchange DCC-GARCH in the diversificationmodel for daily of Chinese gold and portfolios stock data from for 2004–2014. the five bear Against markets this background, in China, examined this paper the estimated dynamic therelationships DCC-GARCH between model the for gold daily and gold stock and markets, stock data and for checked the five the bear effectiveness markets in of China, gold examinedas a hedge theor a dynamic safe haven relationships for stock marke betweents. the gold and stock markets, and checked the effectiveness of gold as a hedge or a safe haven for stock markets.

Int. J. Financial Stud. 2017, 5, 18 3 of 18 Int. J. Financial Stud. 2017, 5, 18 3 of 18

1400 1176.4 1200 985.9 975.38 1000 832.18 886.18 761.05 800 600 531.36 454 403.047 428.163 451.799 450.053 453.486 340.876 360.957 400 313.98 200 0 2009 2010 2011 2012 2013 2014 2015 2016

consumption(ton) gold production(ton)

Figure 2. Gold production and consumption in China (ton) 2009–2016.2009–2016.

This paper is organized as follows. The next section briefly reviews the existing literature in This paper is organized as follows. The next section brieﬂy reviews the existing literature in related areas of research. Section 3 outlines the data and empirical methodologies used, Section 4 related areas of research. Section3 outlines the data and empirical methodologies used, Section4 presents the results and their discussion, and Section 5 concludes. presents the results and their discussion, and Section5 concludes.

2. Related Literature Investors and financial financial analysts often emphasize that gold acts as an an alternative alternative investment investment asset to counter counter market market risk risk in timesin times of market of market stress; stress; not surprisingly, not surprisingly, the hedge the and hedge safe heavenand safe potentials heaven ofpotentials gold have of gold been have extensively been extensively analyzed analyzed in market in fluctuations. market fluctuations Baur and. Baur Lucey and(2010 Lucey) were (2010) the were first tothe formulate first to formulate empirically empirically testable definitionstestable definitions for a hedge for anda hedge a safe and haven a safe with haven regard with to regard financial to assetsfinancial such assets as stocks. such as Following stocks. Following their definitions, their definitions, a hedge (safe a hedge haven) (safe is an haven) asset that is an is uncorrelatedasset that is (negativelyuncorrelated correlated) (negatively with correlated) another asset with or another portfolio asset on or average portfolio (only on in average times of (only market in timesstress orof turmoil)market stress (Beckmann or turmoil) et al. (2015Beckmann) and reported et al. 2015 that) and gold reported was a safe that haven gold forwas stocks a safe in haven the US, for thestocks UK andin the Germany. US, the GoldUK and was Germany. also found Gold as a hedgewas also for found stocks as in thea hedge US and for the stocks UK. in Their the analysisUS and revealedthe UK. thatTheir gold analysis was revealed not a safe that haven gold for was stocks not a at safe all times,haven butfor stocks only in at extreme all times, bearish but only stock in marketsextreme andbearish that stock the safemarkets haven and property that the wassafe short-lived.haven property Baur was and short Mcdermott-lived. Baur(2010 and) also Mcdermott distinguished (2010) betweenalso distinguished a strong and between a weak a form strong of theand hedge a weak and form the safeof the haven hedge property and the (Beckmann safe haven et property al. 2015). Gürgün(Beckmann and et Ünalmıs al. 2015(2014). Gürgün); Beckmann and Ünalmıs et al.(2015 (2014);); Nguyen Beckmann et al. (et2016 al. );(2015); Iqbal( 2017Nguyen) and et Shahzad al. (2016); etIqbal al.( 2017(2017)) all and found Shahzad that goldet al. could(2017) act all asfound a hedge that andgold safe could haven act inas emerginga hedge and and safe developing haven in countries,emerging European and developing stock markets, countries, five EuropeanEurozone peripheral stock markets, GIPSI countries,five Eurozone and Pakistan peripheral and GIPSI India, respectively.countries, and For Pakistan countries and withIndia, a religionrespectively factor. For such countries as Malaysia with a and religion the GCC factor (Gulf such Cooperation as Malaysia Council:and the GCC Bahrain, (Gulf Kuwait, Cooperation Oman, Qatar,Council: Saudi Bahrain, Arabia Kuwait, and United Oman, Arab Qatar, Emirates), Saudi theArabia domestic and United Islamic goldArab account Emirates), provided the domestic hedge and Islamic safe haven gold to account sharia compliant provided stocks hedge ( Ghazali and safe et al. haven 2015 ; to Mensi sharia et al.compliant 2015, 2016 stocks). Ciner (Ghazali et al.( et2013 al.); 2015; Chen Mensi and Lin et (al.2014 2015,); Choudhry 2016). Ciner et al. et( 2015al. (2013);) and ChenSmiech and and Lin Papiez (2014); (Choudhry2016) also et depicted al. (2015) that and gold Smiech had characteristicsand Papiez (2016) of a also hedge depicted and safe that haven gold for had the characteristics US stock market. of a Accordinghedge and tosafe the haven analysis for above-mentioned,the US stock market. most According of the previous to the researchanalysis has above shown-mentioned, that gold most can actof asthe a previous safe haven research against has extreme shown market that gold movement can act as and a safe as a haven hedge against on average. extreme market movement and asOther a hedge studies on average. have examined the dynamic relationship between gold and stock. Souˇcek(2013) foundOther that duringstudies unstablehave examined periods, the the dynamic correlation relationship between gold between and stock,gold and proxied stock. by Soucek open interest, (2013) tendedfound tha tot beduring weak unstable or negative. periods, Thus, the correlation gold can servebetween as angold investors’ and stock, safe proxied haven. by Barunopen interest,ík et al. (tended2016) investigated to be weak or the negative. dynamic Thus, correlations gold can between serve as goldan investors’ and stocks. safe Theirhaven. analyses Baruník showed et al. (2016) that heterogeneityinvestigated the in correlations dynamic correlations across a number between of investment gold and horizons stocks. Their between analyses gold and showed stock wasthat aheterogeneity dominant feature in correlations during times across of a economicnumber of downturn investment and horizons financial between turbulence. gold and Heterogeneity stock was a prevaileddominant in feature correlations during between times ofgold economic and stocks. downturn After the and 2008 financial crisis, correlations turbulence. among Heterogeneity gold and stockprevailed increased in correlations and became between homogenous. gold and stocks. After the 2008 crisis, correlations among gold and stock increased and became homogenous. A number of studies have focused on the relationship between gold and stock in emerging economies. For example, Jain and Biswal (2016) investigated the dynamic linkages between the prices

Int. J. Financial Stud. 2017, 5, 18 4 of 18

A number of studies have focused on the relationship between gold and stock in emerging economies. For example, Jain and Biswal(2016) investigated the dynamic linkages between the prices of gold and Indian stocks and uncovered a strong relationship between gold and stocks, suggesting the importance of using gold to restrain stock market volatility. However, the study by Basher and Sadorsky(2016) was based on data from 23 emerging economies, which indicated that there was a positive link between gold and stocks in most emerging economies. Bouri et al.(2017) also found that there was a positive nonlinear relationship between gold and the stock market in India. Dee et al. (2013); Arouri et al.(2015) and Huang et al.(2016) all focused their studies on the relationship between gold and stocks in China. From the existing literature, we uncovered several things. For the method, most of the studies used the quantile-GARCH model, and some of them used the DCC-GARCH model, Copula and wavelet analysis. For the results, gold was negatively related to stocks in most countries, so gold could be used as a hedge and a safe haven asset. The main feature of these studies, which partially motivated our research, was to ﬁrst use a quantile regression approach which included a simple GARCH (1,1) model; Baur and Lucey(2010) analyzed gold’s safe-haven role in the stock market, however, their analyses were based on the short-term extreme negative impact of stocks on days, which adapted to the characteristics of the “slow bull and quick bear” in western stock markets. For the characteristics of a “quick bull and slow bear” in China’s stock markets, bear markets are usually measured in months or years. Dividing the time period into bear market periods and nonbear market periods based on the ﬂuctuating situation in China’s stock market was more realistic to study whether gold could play the role of a safe haven and hedge in different periods. Second, the previous literature has only used the SSE Composite Index as the proxy variable for China’s stock market, therefore, this paper selected the SSE Composite Index, the SZSE Component Index, the CSI 300 Index, the SME Index, and GEM Index simultaneously to research the effectiveness of gold as a hedge or a safe haven for different stock indexes during the bear and nonbear markets. The results could describe the overall relationship between gold and China’s stock market.

3. Data, Deﬁnitions, Empirical Methodology

3.1. Data As indicated above, the ﬁve stock indexes used were the Shanghai Stock Exchange Composite Index (SSE)2, the Shenzhen Stock Exchange Component Index (SZSE)3, the Capitalization-weighted Stock Index 300 Index (CSI)4, the Small and Medium Enterprise Board Index (SME)5 and the Growth Enterprise Market Index (GEM)6. The paper used the AU9995 spot price from the Shanghai Gold Exchange (SGE) as the proxy variable for China’s gold price. The start date of AU9995 was 31 October 2002, so the sample period spanned 31 October 2002 to 18 April 2017. However, the start date of the CSI 300 Index, the SME Index and the GEM Index were April 2005, June 2005 and June 2010, respectively. Thus, the starting times of the above-mentioned three kinds of index were different from the SSE Composite Index and the SZSE Component Index, as shown in Table1. Daily data of the ﬁve stock indexes and gold spot prices were obtained from Datastream and SGE. The continuous returns for both

2 The SSE Composite Index is a stock market index of all stocks (A shares and B shares) that are traded at the Shanghai Stock Exchange. 3 The SZSE Component Index is an index of 500 stocks that are traded at the Shenzhen Stock Exchange (SZSE). It is the main stock market index of SZSE. 4 The CSI 300 is a capitalization-weighted stock market index designed to replicate the performance of 300 stocks traded in the Shanghai and Shenzhen stock exchanges. 5 SME Board is a supplement to the main board market, and it is a direct financing platform for small and medium enterprises. 6 GEM Index is a stock market index set up by Stock Exchange of Hong Kong for growth companies that do not fulfill the requirements of profitability or track record. Int. J. Financial Stud. 2017, 5, 18 5 of 18 stock and gold prices were calculated by taking the natural logarithm of the ratio of two consecutive prices, which was then, multiplied it by 100.

Table 1. The sources and intervals of variables.

Name Data Sources Sample Interval Rgold SGE 31 October 2002–18 April 2017 Rsse Datastream 31 October 2002–18 April 2017 Rsz Datastream 31 October 2002–18 April 2017 Rcsi Datastream 4 November 2005–18 April 2017 Rsme Datastream 8 June 2005–18 April 2017 Rgem Datastream 2 June 2010–18 April 2017 Note: Rgold, Rsse, Rsz, Rcsi, Rsme, and Rgem represent the return of gold price, the return of the SSE Composite Index, the return of the SZSE Component Index, the return of CSI 300 index, the return of SME Index and the return of GEM Index, respectively.

Table2 presents the descriptive statistics for the daily return series of the gold and five stock indexes and unit root tests for returns. As shown in panel A, all stock index returns varied more dramatically than the gold returns as indicated by their simple standard deviation. Skewness coefficients showed that the return distribution for all time series were negatively and significantly skewed. Kurtosis coefficients indicated that all return series were far from normally distributed. In addition, the Jarque–Bera test statistics clearly confirmed the rejection of the null hypothesis of normality for all returns series at the 1% significance level for both the stock indexes and gold series.

Table 2. Descriptive statistics (31 October 2002–18 April 2017) and unit root tests for returns.

Rsse Rg(sse) Rsz Rg(sz) Rsci Rg(sci) Rsme Rg(sme) Rgem Rg(gem) Panel A: Descriptive Statistics Mean 0.021 0.035 0.035 0.035 0.042 0.031 0.066 0.032 0.038 0.004 Median 0.068 0.06 0.06 0.06 0.102 0.052 0.192 0.052 0.108 0.009 Max 9.034 9.328 14.25 9.33 8.931 9.328 9.27 9.328 6.914 4.527 Min −9.256 −9.474 −9.75 −9.47 −9.695 −9.474 −9.87 −9.474 −9.332 −9.474 Std. Dev. 1.656 1.102 1.876 1.095 1.834 1.155 1.995 1.161 2.128 1.014 Skewness −0.491 −0.392 −0.324 −0.388 −0.528 −0.378 −0.582 −0.378 −0.554 −0.682 Kurtosis 7.13 10.069 6.757 10.11 6.456 9.769 5.58 9.682 4.873 11.582 JB 2637.04 *** 7402.45 *** 2120.62 *** 7463.47 *** 1589.86 *** 5648.30 *** 963.19 *** 5435.87 *** 329.39 *** 5250.75 *** Obs. 3512 3512 3502 3502 2922 2922 2885 2885 1669 1669 Panel B: Unit Root Test Statistics DF −57.938 *** −62.295 *** −56.115 *** −62.172 *** −52.426 *** −56.584 *** −50.653 *** −56.248 *** −37.577 *** −43.123 *** ADF −57.939 *** −62.323 *** −56.120 *** −62.198 *** −52.441 *** −56.609 *** −50.687 *** −56.278 *** −37.566 *** −43.110 *** DF-GLS −57.603 *** −62.231 *** −56.024 *** −40.748 *** −9.766 *** −56.427 *** −1.939 −55.255 *** −4.553 *** −41.993 *** PP −57.998 *** −62.252 *** −56.229 *** −62.122 *** −52.499 *** −56.582 *** −50.636 *** −56.199 *** −37.501 *** −43.085 *** Notes: JB is the Jarque-Berra test for normality, *** indicates the rejection of null hypotheses at the 1% level. Rgold, Rsse, Rsz, Rcsi, Rsme and Rgem represent the return of gold price, the return of the SSE Composite Index, the return of the SZSE Component Index, the return of CSI 300 index, the return of SME Index and the return of GEM Index, respectively. As the sample intervals of SSE, SZSE, CSI300, SME and GEM are different, the corresponding return of gold price in different sample intervals are also different, which are expressed as Rg(sse), Rg(sz), Rg(sci), Rg(sme) and Rg(gem) respectively.

Panel B presents the results of the four unit root and stationarity tests namely the DF, ADF, DF-GLS and PP tests. The DF, ADF, and PP test statistics were significant at the 1% significance level, rejecting the null hypothesis of unit root for all time return series. Referring to the DF-GLS results, we could not reject the null hypothesis of stationarity only for the SME Index. Consequently, all the return series were stationary and thus suitable for further analysis. Financial time series often exhibit correlation. Under such conditions, GARCH modeling is particularly appropriate. Table3 shows the results of the Ljung-Box test and ARCH test. The Ljung-Box test indicated evidence of autocorrelation in the return series for both the gold and stock indexes. Furthermore, the empirical statistics of the Engle(1982) test for conditional heteroscedasticity were significant for all cases suggesting the presence of ARCH effects in returns. All these features Int. J. Financial Stud. 2017, 5, 18 6 of 18 justified our choice of GARCH type models to examine the dynamic relationship between gold and stock indexes.

Table 3. Ljung-Box Q test and ARCH test statistics.

Q(10) ARCH(1) ARCH(5) ARCH(10) Rsse 37.954 (0.000) *** 106.269 (0.000) *** 295.041 (0.000) *** 352.277 (0.000) *** Rg(sse) 29.23 (0.001) *** 99.536 (0.000) *** 170.455 (0.000) *** 224.957 (0.000) *** Rsz 34.332 (0.000) *** 71.440 (0.000) *** 215.862 (0.000) *** 313.094 (0.000) *** Rg(sz) 33.109 (0.000) *** 90.553 (0.000) *** 142.107 (0.000) *** 178.969 (0.000) *** Rsci 36.5 (0.000) *** 83.465 (0.000) *** 233.190 (0.000) *** 292.751 (0.000) *** Rg(sci) 24.404 (0.007) *** 81.784 (0.000) *** 137.488 (0.000) *** 181.913 (0.000) *** Rsme 25.732 (0.004) *** 78.634 (0.000) *** 236.576 (0.000) *** 278.628 (0.000) *** Rg(sme) 24.422 (0.007) *** 78.974 (0.000) *** 132.345 (0.000) *** 175.468 (0.000) *** Rgem 27.884 (0.002) *** 82.632 (0.000) *** 205.258 (0.000) *** 232.847 (0.000) *** Rg(gem) 9.5233 (0.483) 74.896 (0.000) *** 88.509 (0.000) *** 93.421 (0.000) *** Note: Numbers in parenthesis indicate the prob. of the coefﬁcient. *** indicates the rejection of null hypotheses at the 1% level. Rgold, Rsse, Rsz, Rcsi, Rsme and Rgem represent the return of gold price, the return of the SSE Composite Index, the return of the SZSE Component Index, the return of CSI 300 index, the return of SME Index and the return of GEM Index, respectively. As the sample intervals of SSE, SZSE, CSI300, SME and GEM are different, so the corresponding return of gold price in different sample intervals are also different, which are expressed as Rg(sse), Rg(sz), Rg(sci), Rg(sme) and Rg(gem), respectively.

3.2. Definitions of Bear and Nonbear Stock Markets in China At present, there are two main methods to divide the cycle of the bull market and the bear market: (i) the parameter method where Hamilton(1989) divided the bear market and bull market by estimating the parameters of the Markov-switching model; and (ii) the nonparametric method, which divides the bull and bear markets by looking for peaks and troughs. The parameter method is more suitable for mature markets such as the US and Europe. Considering the particularity of the Chinese capital market, which is immature and vulnerable to national policy, the nonparametric method was adopted in this paper to divide bull and bear markets in China. Referring to Eliot wave theory (Frost and Prechter 2005) and combining the characteristics of policy intervention and institutional change in China’s stock market development (Lu and Xu 2004), this paper found that the SSE composite index contained five bear markets from 31 October 2002 to 18 April 20177, as shown in Table4 and Figure3. In addition to the five bear markets, other periods including bull markets and shock markets were considered as normal or nonbear markets. The “bear market periods” and the “nonbear market periods” of the other four stock indexes were all consistent with the SSE Composite Index.

7 The conventional deﬁnition for a bear market which is usually deﬁned as time periods when the stock market down more than 20% from its most recent highs to its corresponding relative minima (Chen and Lin 2014). Int. J. Financial Stud. 2017, 5, 18 7 of 18

Int. J. Financial Stud. 2017, 5, 18 Table 4. Bear markets in China since October 2002. 7 of 18

Time Period Maximum Minimum Drop Origin 31 October 2002Table– 4. Bear markets in China since October 2002. Bear market I 1507.50 1030.94 32% Reduction of state-owned shares 7 June 2005 Time Period Maximum Minimum Drop Origin 1631 October October 2007 2002–– BearBear market market II I 1507.506092.06 1030.941706.70 32%72% ReductionThe global of state-owned financial crisis shares 4 November7 June 2005 2008 16 October 2007– Bear market II 4 August 2009– 6092.06 1706.70 72%Restarting The globalIPO, tightening ﬁnancial crisismacro policies, Bear market III 4 November 2008 3471.44 1959.77 44% 3 December 2012 the European debt crisis 4 August 2009– Restarting IPO, tightening macro policies, Bear market III 3471.44 1959.77 44% 3 December 2012 De-leveraging,the European checking debt crisis OTC Future 12 June 2015– Bear market IV 5166.35 2927.29 43% De-leveraging,Financing by the checking CSRC OTC(China Future Security 12 June 2015– Bear market IV 26 August 2015 5166.35 2927.29 43% Financing by the CSRC (China Security 26 August 2015 Regulatory committee) Regulatory committee) 2222 December December 2015 2015–– ImplementingImplementing the the fuse fuse mechanism mechanism and and the the BearBear market market V V 3651.773651.77 2687.982687.98 26%26% 29 February 2016 2016 registrationregistration system, system, devaluation devaluation of of RMB RMB

7,000

6,000 -43% 5,000 -72%

4,000 -26% -44% 3,000

2,000 -32%

1,000

0 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 the SSE Composite Index

Figure 3. Five b bearear markets in China since October 2002.

3.3. M Methodologyethodology Following the definitionsdefinitions of Baur and Lucey (2010 (2010)),, a a hedge hedge (safe (safe haven) haven) is is an an asset asset that that is is uncorrelated or negatively correlated with another asset or portfolio on average (in times of market stress or turmoil). We estimated the dynamic conditional corre correlationslations between the return of gold and fivefive stock indexes in times of bear market periods and nonbear market periods. That is, if gold was uncorrelated or negatively correlated with stocksstocks in times of bear market periods (or nonbear market periods), gold cancan be a safe haven (hedge) for stockstock markets.markets. The aim of this paper is to investigate contemporaneous time time-varying-varying correlation between gold and stock index in different periods, the correlations thus obtained will shed light on the dynamic relationships am amongstongst the variables. The The DCC DCC-GARCH-GARCH model model by by EngleEngle ((20020022)) isis usedused toto examineexamine time varying correlations between two or more series. So, using the DCC DCC-GARCH-GARCH framework is more appropriate than other GARCH models. The DCC DCC-GARCH-GARCH model is estimated in two steps: in the firstfirst step, the GARCH parameters are estimated; and in the second step, the conditional correlations are estimated. Ht =DtRtDt H t Dt Rt Dt Ht is a n × n conditional covariance matrix, Rt is the conditional correlation matrix, and Dt is a Ht is a n × n conditional covariance matrix, Rt is the conditional correlation matrix, and Dt is a diagonal matrix with time-varying standard deviations on the diagonal. diagonal matrix with time-varying standard deviations on the diagonal. D = diag h1/2 1/ 2h1/2 1/ 2 t  diag1,t ,... ,...n,t Dt h1,t hn,t  −1/2 −1/2 −1/2 −1/2 Rt = diag(q1,t ,... qn,t )Qt diag(q1,t ,... qn,t )

Int. J. Financial Stud. 2017, 5, 18 8 of 18

The expressions for h are univariate GARCH models (H is a diagonal matrix). For the GARCH (1,1) model, the elements of Ht can be written as:

2 hi,t = ωi + αiεi,t−1 + βihi,t−1 where Qt is a symmetric positive deﬁnite matrix.

0 Qt = (1 − θ1 − θ2)Q + θ1zt−1zt−1 + θ2Qt−1 where Q is the n × n unconditional correlation matrix of the standardized residuals p zi,t zi,t = εi,t/ hi,t . The parameters θ1 and θ2 are non-negative, and are associated with the exponential smoothing process used to construct the dynamic conditional correlations. The DCC model is mean reverting as long as θ1 + θ2 < 1. Dependence on only parameters θ1 and θ2 is one of the strengths of this model. Irrespective of the number of variables, only these two parameters need to be estimated, making it more likely to reach an optimal solution. When the standardized residuals from two variables rise or fall together, they push the correlation up. This elevated level will gradually decrease back to the average level over the passage of time due to the complete absorption of information. When the residuals move in different directions, they pull the correlation down, which moves up with time. The speed of this process is controlled by the parameters θ1 and θ2. The correlation estimator is qi,j,t ρi,j,t = √ qi,i,tqj,j,t

For the purposes of this study, the focus of interest is ρi,j,t, which represents the conditional correlation between the return of gold and the return of each stock index pair.

4. Empirical Results Time series graphs of returns show how volatility has changed across time (Figures4–9). Each series displays several periods of volatility clustering, which also confirms the conclusions from Table3 that ARCH effects exist in each yield series8. It is reasonable to assume that the GARCH model is appropriate. First, the time-varying variances of the series were estimated using a univariate GARCH specification, and the parameter estimates are presented in Table5. α and β represent the estimated ARCH (1) and GARCH (1) parameters, respectively. All univariate GARCH processes showed a high degree of persistence, that is, the sum of α and β were all close to one, indicating that all the models were of good fit. The low values of α and the high values of β indicated that the correlation process was resistant to shocks and reverted to mean quickly. This indicated that the correlations amongst the variables were stable. The magnitude of the GARCH cofficient (β) for stock indexes were high and varied between 0.914 and 0.945, and the magnitude of the GARCH cofficient (β) for gold as around 0.9, indicating the high persistence of volatility over time.

8 Evidence of conditional heteroscedasticity lends support to the use of ARCH-type models and, in particular, to the use of GARCH (1,1) models to capture the volatility behavior of the data series. Int. J. Financial Stud. 2017, 5, 18 9 of 18 Int. J. Financial Stud. 2017, 5, 18 9 of 18 Int. J. Financial Stud. 2017, 5, 18 9 of 18 Int. J. Financial Stud. 2017, 5, 18 9 of 18 Int. J. Financial Stud.10 2017, 5, 18 9 of 18 10 10 5 105 5 0 50 0 -5

-50

2010/2/2 2012/3/6 2012/9/6 2014/4/8

2004/5/31 2005/6/13 2006/6/27 2007/7/10 2008/1/11 2008/7/18 2009/1/22 2009/7/30 2010/8/11 2011/2/23 2011/8/24 2013/3/20 2013/9/26 2015/4/20 2016/4/28 2016/11/4

-5 2003/5/16

2010/2/2 2012/3/6 2012/9/6 2014/4/8

2003/11/17 2004/11/30 2005/12/13 2006/12/27 2014/10/14 2015/10/26

-10 2002/10/31

2004/5/31 2005/6/13 2006/6/27 2007/7/10 2008/1/11 2008/7/18 2009/1/22 2009/7/30 2010/8/11 2011/2/23 2011/8/24 2013/3/20 2013/9/26 2015/4/20 2016/4/28 2016/11/4

-5 2003/5/16

2003/11/17 2002/10/31 2004/11/30 2005/12/13 2006/12/27 2014/10/14 2015/10/26

2012/3/6 2012/9/6 2014/4/8

-10 2010/2/2

2003/5/16 2004/5/31 2005/6/13 2006/6/27 2007/7/10 2008/1/11 2008/7/18 2009/1/22 2009/7/30 2010/8/11 2011/2/23 2011/8/24 2013/3/20 2013/9/26 2015/4/20 2016/4/28 2016/11/4

2003/11/17 2002/10/31 2004/11/30 2005/12/13 2006/12/27 2014/10/14 2015/10/26

2012/3/6 2012/9/6 2014/4/8

-10 2010/2/2

2004/5/31 2005/6/13 2006/6/27 2007/7/10 2008/1/11 2008/7/18 2009/1/22 2009/7/30 2010/8/11 2011/2/23 2011/8/24 2013/3/20 2013/9/26 2015/4/20 2016/4/28 2016/11/4

-15 2003/5/16

2003/11/17 2004/11/30 2005/12/13 2006/12/27 2014/10/14 2015/10/26 -10-15 2002/10/31 -15 FigureFigure 4. 4.Return-15 Return of of the the Shanghai Shanghai Stock Stock ExchangeExchange Composite Index Index (31 (31 October October 2002 2002–18–18 April April 2017). 2017). Figure 4. Return of the Shanghai Stock Exchange Composite Index (31 October 2002–18 April 2017). Figure 4. Return of the Shanghai Stock Exchange Composite Index (31 October 2002–18 April 2017). 20.00 Figure 4. Return of the Shanghai Stock Exchange Composite Index (31 October 2002–18 April 2017). 20.0015.00 20.0015.0010.00 15.0020.0010.005.00 10.0015.005.000.00 10.00-5.005.000.00 -10.00-5.000.005.00

-10.00-15.00-5.000.00

2009/4/8 2012/7/6 2013/8/5 2016/4/6

-10.00-15.00-5.00 2006/8/4

2013/1/16 2003/5/22 2004/6/16 2005/7/11 2006/1/18 2007/2/14 2007/8/29 2008/3/14 2008/9/19 2010/4/29 2011/6/13 2014/2/20 2014/8/27 2015/3/16 2015/9/21

2006/8/4 2009/4/8 2012/7/6 2013/8/5 2016/4/6

2003/11/27 2004/12/22 2009/10/20 2010/11/29 2011/12/20 2016/10/19

-15.00 2002/10/31

2013/1/16 2004/6/16 2005/7/11 2006/1/18 2007/2/14 2007/8/29 2008/3/14 2008/9/19 2010/4/29 2011/6/13 2014/2/20 2014/8/27 2015/3/16 2015/9/21

-10.00 2003/5/22

2002/10/31 2003/11/27 2004/12/22 2009/10/20 2010/11/29 2011/12/20 2016/10/19

2006/8/4 2009/4/8 2012/7/6 2013/8/5 2016/4/6

2013/1/16 2004/6/16 2005/7/11 2006/1/18 2007/2/14 2007/8/29 2008/3/14 2008/9/19 2010/4/29 2011/6/13 2014/2/20 2014/8/27 2015/3/16 2015/9/21

-15.00 2003/5/22

2002/10/31 2003/11/27 2004/12/22 2009/10/20 2010/11/29 2011/12/20 2016/10/19

2006/8/4 2009/4/8 2012/7/6 2013/8/5 2016/4/6

2013/1/16 2003/5/22 2004/6/16 2005/7/11 2006/1/18 2007/2/14 2007/8/29 2008/3/14 2008/9/19 2010/4/29 2011/6/13 2014/2/20 2014/8/27 2015/3/16 2015/9/21

2003/11/27 2004/12/22 2009/10/20 2010/11/29 2011/12/20 2016/10/19 2002/10/31

Figure 5. Return of the Shenzhen Stock Exchange Component Index (31 October 2002–18 April 2017). Figure 5. Return of the Shenzhen Stock Exchange Component Index (31 October 2002–18 April 2017). FigureFigure 5. 5.Return Return of of the the Shenzhen Shenzhen StockStock ExchangeExchange C Componentomponent Index Index (31 (31 October October 2002 2002–18–18 April April 2017). 2017). Figure 5. Retur10 n of the Shenzhen Stock Exchange Component Index (31 October 2002–18 April 2017). 10 105 105 50 50 -50 -50

-10-5

2008-11-11 2015-02-26 2005-09-16 2006-03-08 2006-08-15 2007-01-25 2007-07-11 2007-12-18 2008-06-02 2009-04-23 2009-09-28 2010-03-16 2010-08-23 2011-02-10 2011-07-19 2011-12-27 2012-06-13 2012-11-21 2013-05-10 2013-10-24 2014-04-03 2014-09-10 2015-08-03 2016-01-13 2016-06-27 2016-12-06

-10-5 2005-04-11

2008-11-11 2015-02-26 2005-04-11 2005-09-16 2006-03-08 2006-08-15 2007-01-25 2007-07-11 2007-12-18 2008-06-02 2009-04-23 2009-09-28 2010-03-16 2010-08-23 2011-02-10 2011-07-19 2011-12-27 2012-06-13 2012-11-21 2013-05-10 2013-10-24 2014-04-03 2014-09-10 2015-08-03 2016-01-13 2016-06-27 2016-12-06

-10

2008-11-11 2015-02-26 2005-09-16 2006-03-08 2006-08-15 2007-01-25 2007-07-11 2007-12-18 2008-06-02 2009-04-23 2009-09-28 2010-03-16 2010-08-23 2011-02-10 2011-07-19 2011-12-27 2012-06-13 2012-11-21 2013-05-10 2013-10-24 2014-04-03 2014-09-10 2015-08-03 2016-01-13 2016-06-27 2016-12-06 -15 2005-04-11

-10

2008-11-11 2015-02-26 2005-09-16 2006-03-08 2006-08-15 2007-01-25 2007-07-11 2007-12-18 2008-06-02 2009-04-23 2009-09-28 2010-03-16 2010-08-23 2011-02-10 2011-07-19 2011-12-27 2012-06-13 2012-11-21 2013-05-10 2013-10-24 2014-04-03 2014-09-10 2015-08-03 2016-01-13 2016-06-27 2016-12-06 -15 2005-04-11 -15 Figure-15 6. Return of the Capitalization-weighted Stock Index 300 (11 April 2005–18 April 2017). Figure 6. Return of the Capitalization-weighted Stock Index 300 (11 April 2005–18 April 2017). Figure 6. Return of the Capitalization-weighted Stock Index 300 (11 April 2005–18 April 2017). Figure 6. Return of the Capitalization-weighted Stock Index 300 (11 April 2005–18 April 2017). Figure15 6. Return of the Capitalization-weighted Stock Index 300 (11 April 2005–18 April 2017). 15 1510 1510 105 105 50 50 -50

-50

2010/4/7 2008/7/4 2010/9/9 2011/8/1 2012/1/9 2014/9/2 2016/6/6

-10-5 2005/6/8

2007/3/14 2006/4/24 2006/9/27 2007/8/17 2008/1/24 2009/5/21 2011/2/25 2012/6/20 2013/5/13 2014/3/31 2015/2/10 2015/7/21

2010/4/7 2005/6/8 2008/7/4 2010/9/9 2011/8/1 2012/1/9 2014/9/2 2016/6/6

2008/12/10 2009/10/29 2012/11/26 2013/10/23 2015/12/28 2016/11/15

-10-5 2005/11/11

2007/3/14 2006/4/24 2006/9/27 2007/8/17 2008/1/24 2009/5/21 2011/2/25 2012/6/20 2013/5/13 2014/3/31 2015/2/10 2015/7/21

2005/11/11 2008/12/10 2009/10/29 2012/11/26 2013/10/23 2015/12/28 2016/11/15

2010/4/7 2008/7/4 2010/9/9 2011/8/1 2012/1/9 2014/9/2 2016/6/6

-10 2005/6/8

2007/3/14 2006/9/27 2007/8/17 2008/1/24 2009/5/21 2011/2/25 2012/6/20 2013/5/13 2014/3/31 2015/2/10 2015/7/21

-15 2006/4/24

2005/11/11 2008/12/10 2009/10/29 2012/11/26 2013/10/23 2015/12/28 2016/11/15

2010/4/7 2008/7/4 2010/9/9 2011/8/1 2012/1/9 2014/9/2 2016/6/6

-10-15 2005/6/8

2007/3/14 2006/9/27 2007/8/17 2008/1/24 2009/5/21 2011/2/25 2012/6/20 2013/5/13 2014/3/31 2015/2/10 2015/7/21

2006/4/24

2008/12/10 2009/10/29 2012/11/26 2013/10/23 2015/12/28 2016/11/15 -15 2005/11/11 Figure-15 7. Return of the Small and Medium Enterprise Board Index (8 June 2005–18 April 2017). Figure 7. Return of the Small and Medium Enterprise Board Index (8 June 2005–18 April 2017). Figure 7. Return of the Small and Medium Enterprise Board Index (8 June 2005–18 April 2017). FigureFigure 7. 7.Return Return of of the the Small Small and and MediumMedium Enterprise Board Board I Indexndex (8 (8 June June 200 2005–185–18 April April 2017). 2017).

Int. J. Financial Stud. 2017, 5, 18 10 of 18 Int. J. Financial Stud. 2017, 5, 18 10 of 18

10

5

0

-5

2010/6/2 2010/9/1 2011/9/8 2013/7/3 2014/1/3 2014/4/9 2014/7/9 2015/1/9

2010/6/2 2010/9/1 2011/9/8 2013/7/3 2014/1/3 2014/4/9 2014/7/9 2015/1/9

2010/12/8 2011/3/14 2011/6/14 2012/3/19 2012/6/20 2012/9/17 2013/3/27 2013/10/8 2015/4/15 2015/7/14 2016/1/14 2016/4/19 2016/7/19 2017/1/19

2010/12/8 2011/3/14 2011/6/14 2012/3/19 2012/6/20 2012/9/17 2013/3/27 2013/10/8 2015/4/15 2015/7/14 2016/1/14 2016/4/19 2016/7/19 2017/1/19

2012/12/19 2014/10/13 2015/10/19 2016/10/24

-10 2011/12/13

2012/12/19 2014/10/13 2015/10/19 2016/10/24 -10 2011/12/13

-15

FigureFigure 8. 8.Return Return of of the the Growth Growth EnterpriseEnterprise Market Index (2 (2 June June 2010 2010–18–18 April April 2017). 2017).

15

10

5

0

-5

2006/2/6 2007/3/5 2014/3/7 2015/4/2

2006/2/6 2007/3/5 2014/3/7 2015/4/2

2009/11/6 2003/12/1 2004/6/21 2005/7/18 2006/8/15 2007/9/11 2008/3/28 2009/4/24 2010/5/20 2010/12/6 2011/6/21 2012/7/18 2013/1/29 2013/8/19 2014/9/15 2016/4/27

-10 2003/5/23

2009/11/6 2003/12/1 2004/6/21 2005/7/18 2006/8/15 2007/9/11 2008/3/28 2009/4/24 2010/5/20 2010/12/6 2011/6/21 2012/7/18 2013/1/29 2013/8/19 2014/9/15 2016/4/27

-10 2003/5/23

2016/11/10 2002/10/31 2004/12/28 2008/10/13 2011/12/29 2015/10/16

2016/11/10 2002/10/31 2004/12/28 2008/10/13 2011/12/29 2015/10/16 -15 -15

FigureFigure 9. 9.Return Return ofof goldgold (31(31 October 2002 2002–18–18 April April 2017). 2017).

Table 5. Univariate GARCH (1,1) parameter estimates.

Table 5. UnivariateCoefficient GARCH Std. (1,1) Error parameter z-Statistic estimates. Prob. C 0.014 *** 0.005 2.774 0.006 Coefﬁcient Std. Error z-Statistic Prob. Rsse α 0.058 *** 0.008 7.541 0.000 C 0.014 *** 0.005 2.774 0.006 β 0.939 *** 0.007 128.774 0.000 Rsse α β 0.0580.939 *** *** 0.0080.007 7.541128.774 0.0000.000 β C 0.9390.017 *** *** 0.0070.004 128.7743.794 0.0000.000 Rg(sse)C α 0.0170.068 *** *** 0.0040.009 3.7947.264 0.0000.000 Rg(sse) α β 0.0680.921 *** *** 0.0090.010 7.26492.139 0.0000.000 β 0.921 *** 0.010 92.139 0.000 C 0.024 *** 0.008 2.977 0.003 Rsz Cα 0.0240.057 *** *** 0.0080.008 2.9777.287 0.0030.000 α Rsz β 0.0570.938 *** *** 0.0080.008 7.287119.071 0.0000.000 β β 0.9380.938 *** *** 0.0080.008 119.071119.071 0.0000.000 C 0.016 *** 0.004 3.705 0.000 C 0.016 *** 0.004 3.705 0.000 Rg(sz)Rg(sz) α α 0.0640.064 *** *** 0.0090.009 7.1597.159 0.0000.000 β β 0.9250.925 *** *** 0.0100.010 96.01396.013 0.0000.000 CC 0.0090.009 * * 0.0050.005 1.8871.887 0.0590.059 RsciRsci α α 0.0560.056 *** *** 0.0080.008 7.1157.115 0.0000.000 β 0.945 *** 0.007 135.173 0.000 β 0.945 *** 0.007 135.173 0.000 C 0.022 *** 0.006 3.620 0.000 C 0.022 *** 0.006 3.620 0.000 Rg(sci) α 0.073 *** 0.011 6.614 0.000 Rg(sci) β α 0.9140.073 *** *** 0.0120.011 76.4446.614 0.0000.000 β 0.914 *** 0.012 76.444 0.000 Cβ 0.0410.914 *** *** 0.0130.012 3.10476.444 0.0020.000 Rsme α C 0.0780.041 *** *** 0.0110.013 7.0523.104 0.0000.002 Rsme β α 0.9140.078 *** *** 0.0110.011 85.6937.052 0.0000.000 β 0.914 *** 0.011 85.693 0.000

Int. J. Financial Stud. 2017, 5, 18 11 of 18

Table 5. Cont.

Coefficient Std. Error z-Statistic Prob. C 0.024 *** 0.007 3.665 0.000 Rg(sme) α 0.071 *** 0.011 6.445 0.000 β 0.913 *** 0.012 73.841 0.000 C 0.018 0.012 1.569 0.117 Rgem α 0.054 *** 0.011 5.120 0.000 β 0.943 *** 0.010 94.613 0.000 C 0.038 *** 0.012 3.079 0.002 Rg(gem) α 0.074 *** 0.016 4.541 0.000 β 0.888 *** 0.023 38.610 0.000 Note: * and *** denote significance at the 10% and 1% level of significance, respectively. Rgold, Rsse, Rsz, Rcsi, Rsme and Rgem represent the return of gold price, the return of the SSE Composite Index, the return of the SZSE Component Index, the return of CSI 300 index, the return of SME Index and the return of GEM Index, respectively. As the sample intervals of SSE, SZSE, CSI300, SME and GEM are different, the corresponding return of gold price in different sample intervals are also different, which are expressed as Rg(sse), Rg(sz), Rg(sci), Rg(sme) and Rg(gem), respectively.

According to Engle(2002), the DCC (1,1) model is the most suitable for ﬁtting a ﬁnancial time series. The multivariate GARCH model of dynamic conditional correlations was estimated using the maximum likelihood estimation. To account for non-normality in the distribution of returns, the DCC-GARCH was estimated with a multivariate student t distribution. The DCC (1,1)–GARCH (1,1) parameter estimates are presented in Table6. The estimated DCC parameters, θ1 and θ2, implied a persistent correlation. The sum of θ1 and θ2. was closer to one, so the dynamic correlation was more obvious. Thus, the dynamic correlation between the SSE Composite Index and gold was the strongest, while the dynamic correlation between the CSI 300 index and gold was the weakest.

Table 6. DCC (1,1)–GARCH (1,1) parameter estimates.

Coef. S.E Z Prob θ 0.017 *** 0.006 2.944 0.003 SSE-Gold 1 θ2 0.957 *** 0.016 60.510 0.000 θ 0.016 *** 0.006 2.623 0.009 SZ-Gold 1 θ2 0.953 *** 0.019 50.333 0.000 θ 0.045 0.028 1.599 0.110 CSI-Gold 1 θ2 0.770 *** 0.218 3.533 0.000 θ 0.021 ** 0.009 2.252 0.024 SME-Gold 1 θ2 0.947 *** 0.030 31.468 0.000 θ 0.012 0.007 1.558 0.119 GEM-Gold 1 θ2 0.960 *** 0.025 38.843 0.000 Note: ** and *** denote signiﬁcance at the 5% and 1% level of signiﬁcance, respectively.

The magnitude of mean dynamic conditional correlation coefficients varied between −1 and 1. If the coefficient was closer to −1, the negative correlation between gold and stock index was stronger. In contrast, when the coefficient was closer to 1, the positive correlation between gold and stock index was stronger. If the coefficient was equal to 0, gold had no relation with the stock indexes. The mean dynamic conditional correlations between stock indexes and gold are presented in Table7. The mean dynamic conditional correlations between SSE-Gold and SZ-Gold were positive in the period of Bear Market I, and both of their values were close to 0.1. This indicates that gold could not act as a safe haven during Bear Market I. The relationship of gold and the CSI 300 Index, the SME Index and the GEM Index could not be analyzed since these three indexes did not exist at this time. Int. J. Financial Stud. 2017, 5, 18 12 of 18 act as a safe haven during Bear Market I. The relationship of gold and the CSI 300 Index, the SME Index and the GEM Index could not be analyzed since these three indexes did not exist at this time. Int. J. Financial Stud. 2017, 5, 18 12 of 18 The mean dynamic conditional correlations between SSE-Gold, SZ-Gold, CSI-Gold, and SME- Gold were all positive in the period of Bear Market II. Similarly, the mean dynamic conditional correlationsThe mean between dynamic SSE conditional-Gold, SZ- correlationsGold, CSI-Gold, between SME SSE-Gold,-Gold and SZ-Gold, GEM-Gold CSI-Gold, were andall positive SME-Gold in Bearwere Market all positive III. This in the finding period indicated of Bear Market that gold II. Similarly,did not act the as meana safe dynamic haven for conditional stock indexes correlations in Bear Marketsbetween II SSE-Gold, and III. SZ-Gold, CSI-Gold, SME-Gold and GEM-Gold were all positive in Bear Market III. ThisWith finding the indicated exception that that gold the did mean not dynamic act as a safe conditional haven for correlation stock indexes coefficient in Bear Marketsof CSI-Gold II and was III. greaterWith than the 0, exception the mean thatdynamic the mean condi dynamictional correlation conditional coefficients correlation of coefficient SSE-Gold, of SZ CSI-Gold-Gold, SME was- Goldgreater and than GEM 0, the-Gold mean were dynamic all less conditional than 0 correlationin Bear Market coefficients IV. These of SSE-Gold, four dynamic SZ-Gold, conditional SME-Gold correlationand GEM-Gold coefficients were allvaried less thanbetween 0 in −0.02 Bear and Market −0.05, IV. indicating These four that dynamic gold was conditional negatively correlation related to mostcoefficients stock indexes, varied between so gold −was0.02 a andsafe −haven0.05, indicatingasset for stocks that gold in Bear was Market negatively IV. Moreover, related tomost the mean stock dynamicindexes, soconditional gold was correlation a safe haven coefficients asset for stocksof SSE- inGold, Bear SZ Market-Gold, IV.CSI Moreover,-Gold, SME the-Gold mean and dynamic GEM- Goldconditional were all correlation less than 0 coefficients in Bear Market of SSE-Gold, V, which SZ-Gold, varied between CSI-Gold, 0 and SME-Gold −0.01. This and showed GEM-Gold that gold were couldall less also than act 0 inas Bear a safe Market haven V, in which Bear Market varied between V. This can 0 and be − explained0.01. This that showed with that the increase gold could in investmentalso act as ademand safe haven for ingold, Bear more Market investors V. This tended can be to explained include gold that in with their the investment increase in portfolio investment to diversifydemand forrisk. gold, more investors tended to include gold in their investment portfolio to diversify risk. TThehe mean mean dy dynamicnamic conditional conditional correlation correlation coefficients coefficients of SSE of-Gold, SSE-Gold, SZ-Gold, SZ-Gold, CSI-Gold, CSI-Gold, SME- Gold,SME-Gold, and GEM and-Gold GEM-Gold were greater were greaterthan 0 in than the 0n inon- thebear non-bear markets. markets. Among them, Among the them, mean thedynamic mean conditionaldynamic conditional correlation correlation coefficient coefficient of CSI-Gold of CSI-Gold was the was largest the at largest 0.0669, at 0.0669,and the and minimum the minimum mean dynamicmean dynamic conditional conditional correlation correlation coefficient coefficient was GEM was GEM-Gold,-Gold, which which was 0.0344. was 0.0344. This Thisshowed showed that thatnot onlynot only was wasgold gold not notused used as an as aninvestment investment hedge, hedge, but but was was instead instead invested invested just just like like a astock stock during during periodsperiods of of non non-bear-bear markets. markets.

TableTable 7. 7. TheThe dynamic dynamic conditional conditional correlations correlations between between stock indexes and gold.

BearBear Markets Markets Non-Bear Market I II III IV V Non-Bear Market I II III IV V SSE-Gold 0.0808 0.0667 0.1151 −0.0410 −0.0770 0.0639 SSE-Gold 0.0808 0.0667 0.1151 −0.0410 −0.0770 0.0639 SZSZ-Gold-Gold 0.0772 0.0772 0.0561 0.0561 0.0926 0.0926 −0.0345−0.0345 −0.0137−0.0137 0.0612 CSICSI-Gold-Gold \\ 0.06230.0623 0.0870 0.0870 0.0182 0.0182 −0.0101−0.0101 0.0669 SMESME-Gold-Gold \\ 0.04570.0457 0.0920 0.0920 −0.0424−0.0424 −0.0337−0.0337 0.0519 GEMGEM-Gold-Gold \\\ \ 0.04050.0405 −0.0285−0.0285 −0.0278−0.0278 0.0344 Note:Note: For convenientconvenient observations, observations, the dynamicthe dynamic conditional conditional correlation correlation coefﬁcients coefficients (ρ) in the bear(ρ) in markets the bear and non-bear market periods are shown by means of arithmetic average. markets and non-bear market periods are shown by means of arithmetic average.

FiguresFigures 10–1144 show the evolutionevolution overover timetime ofof the the dynamic dynamic conditional conditional correlation correlation coefficient coefficientρ . The . The sign sign of of all all correlation correlation coefficients coefficients waswas consistently negative negative over over Bear Bear Markets Markets IV IVand and V (12 V June(12 June 2015 2015–26 –26August August 2015 2015 and and 22 22December December 2015 2015–29–29 February February 2016); 2016); that that is is,, there there was was a a negative negative relationshiprelationship between between gold gold and and the the five five stock stock indexes, indexes, so so gold gold was was a a safe safe haven haven asset asset for for stocks. stocks. However,However, the the correlation correlation coefficients coefficients sign sign was was consistently consistently positive positive in nonbear in nonbear market market periods periods (see non(see-shadow non-shadow section section in Figures in Figures 8–12),8– 12where), where gold gold did didnot act not as act a ashedge a hedge for stocks for stocks in China. in China.

0.4

0.3

0.2

0.1

0.0

-0.1

-0.2

-0.3 2002 2004 2006 2008 2010 2012 2014 2016

FigureFigure 10. 10. DynamicDynamic conditional conditional correlations correlations between between gold gold and and SSE. SSE.

Int. J. Financial Stud. 2017, 5, 18 13 of 18 Int. J. Financial Stud. 2017, 5, 18 13 of 18 Int.Int. J.J. FinancialFinancial Stud.Stud. 20172017,,, 55,,, 11188 1313 ofof 1818

0.3 0.30.3

0.2 0.20.2

0.1 0.10.1

0.0 0.00.0

-0.1 -0.1-0.1

-0.2 -0.2-0.2

-0.3 -0.3-0.3 -0.32002 2004 2006 2008 2010 2012 2014 2016 20022002 20042004 20062006 20082008 20102010 20122012 20142014 20162016

FigureFigure 1 11.11.1. DynamicDynamic conditional conditional correlations correlations between between goldgold andand SZ.SZ.

0.5 0.50.5 0.4 0.40.4 0.3 0.30.3 0.2 0.20.2 0.1 0.10.1 0.0 0.00.0 -0.1 -0.1-0.1 -0.2 -0.2-0.2 -0.3 -0.3-0.3 -0.3 2006 2008 2010 2012 2014 2016 20062006 20082008 20102010 20122012 20142014 20162016 Figure 12. Dynamic conditional correlations between gold and CSI. FigureFigure 1 12.12.2. DynamicDynamic conditional conditional correlations correlations between between goldgold andand CSI.CSI.

0.4 0.40.4

0.3 0.30.3

0.2 0.20.2

0.1 0.10.1

0.0 0.00.0

-0.1 -0.1-0.1

-0.2 -0.2-0.2 -0.22005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 20052005 20062006 20072007 20082008 20092009 20102010 20112011 20122012 20132013 20142014 20152015 20162016

FigureFigure 13. 13. DynamicDynamic conditional conditional correlations correlations between between goldgold andand SME.SME.

0.2 0.20.2

0.2 0.20.2

0.1 0.10.1

0.0 0.00.0

0.0 0.00.0

-0.1 -0.1-0.1

-0.1 -0.1-0.1

-0.2 -0.2-0.2 -0.2 2010 2011 2012 2013 2014 2015 2016 2017 20102010 20112011 20122012 20132013 20142014 20152015 20162016 20172017 Figure 14. Dynamic conditional correlations between gold and GEM. FigureFigure 14. 14. DynamicDynamic conditional conditional correlations correlations between between goldgold andand GEM.GEM.

5.5. SubsampleSubsample TestTest SSubsamplesubsamples ofof bearbear andand nonbearnonbear marketmarket periodsperiods werewere alsoalso estimatedestimated inin TablesTables 88––12.12. TheThe resultsresults werewere almostalmost consistentconsistent withwith TableTable 7,7, wherewhere goldgold actedacted asas aa safesafe havenhaven forfor onlyonly thethe latestlatest twotwo ofof thethe fivefive bearbear marketmarket periodsperiods (12(12 JuneJune 20152015––2626 AugustAugust 20152015 andand 2222 DecemberDecember 20152015––2929 FebruaryFebruary 2016)2016);;;

Int. J. Financial Stud. 2017, 5, 18 14 of 18

5. Subsample Test Subsamples of bear and nonbear market periods were also estimated in Tables8–12. The results were almost consistent with Table7, where gold acted as a safe haven for only the latest two of the ﬁve bear market periods (12 June 2015–26 August 2015 and 22 December 2015–29 February 2016); however, gold did not offer good risk hedging for nonbear market periods. Nevertheless, due to the limited number of observations, many estimates were not stable (see the blue font, and “/” indicates it could not be estimated).

Table 8. DCC (1,1) Model-2 Step Estimation of SSE-GOLD.

Step One Step Two The Mean of Dynamic

c α β θ1 θ2 Conditional Correlations Subsample 1: 31 October 2002–7 June 2005 SSE 0.115 (0.074) * 0.065 (0.003) *** 0.863 (0.000) *** −0.026 (0.246) 0.859 (0.000) *** 0.0941 GOLD 0.015 (0.001) *** 0.056 (0.000) *** 0.921 (0.000) *** Subsample 2: 16 October 2007–4 November 2008 SSE 0.079 (0.133) −0.041 (0.011) ** 1.035 (0.000) *** 0.023 (0.318) 0.958 (0.000) *** 0.0074 GOLD 0.068 (0.193) 0.065 (0.051) * 0.921 (0.000) *** Subsample 3: 4 August 2009–3 December 2012 SSE 0.017 (0.012) ** 0.014 (0.003) *** 0.974 (0.000) *** 0.102 (0.107) 0.694 (0.002) *** 0.1714 GOLD 0.020 (0.000) *** 0.064 (0.000) *** 0.921 (0.000) *** Subsample 4: 12 June 2015–26 August 2015 SSE 4.222 (0.849) −0.039 (0.838) 0.729 (0.641) 0.151408 (0.388) −0.213287 (0.804) −0.0755 GOLD 0.082 (0.214) −0.18245 (0.010) *** 1.148 (0.000) *** Subsample 5: 22 December 2015–29 February 2016 SSE 7.035 (0.399) −0.182 (0.127) 0.120 (0.919) −0.065 (0.628) 0.762 (0.105) −0.2423 GOLD 0.259 (0.465) −0.070 (0.239) 0.785 (0.014) ** Subsample 6: non-bear market SSE 0.007 (0.026) ** 0.059 (0.000) *** 0.940 (0.000) *** 0.030 (0.206) 0.653 (0.007) *** 0.0701 GOLD 0.038 (0.000) *** 0.085 (0.000) *** 0.884 (0.000) *** Note: *, ** and *** denote signiﬁcance at the 10%, 5%, and 1% level of signiﬁcance, respectively.

Table 9. DCC (1,1) Model-2 Step Estimation of SZ-GOLD.

Step One Step Two The Mean of Dynamic

c α β θ1 θ2 Conditional Correlations Subsample 1: 31 October 2002–7 June 2005 SZ 0.207 (0.103) 0.069 (0.019) ** 0.814 (0.000) *** 0.005 (0.697) 0.961 (0.000) *** 0.1065 GOLD 0.015 (0.001) *** 0.056 (0.000) *** 0.921 (0.000) *** Subsample 2: 16 October 2007–4 Novemver 2008 SZ 0.115 (0.065) * −0.046 (0.011) ** 1.037 (0.000) *** 0.057 (0.236) 0.849 (0.000) *** −0.0111 GOLD 0.050 (0.237) 0.046 (0.052) * 0.944 (0.000) *** Subsample 3: 4 August 2009–3 December 2012 SZ 0.043 (0.043) ** 0.011 (0.006) *** 0.972 (0.000) *** 0.119 (0.033) ** 0.591 (0.009) *** 0.1461 GOLD 0.020 (0.001) *** 0.063 (0.000) *** 0.922 (0.000) *** Subsample 4: 12 June 2015–26 August 2015 SZ 4.555 (0.869) −0.049 (0.847) 0.736 (0.685) −0.192 (/) 0.976 (/) −0.0874 GOLD 0.096 (0.003) *** −0.214 (0.003) *** 1.163 (0.000) *** Subsample 5: 22 December 2015–29 February 2016 SZ 10.267 (0.401) −0.174 (0.199) 0.096 (0.936) −0.210 (/) 1.051 (/) −0.4322 GOLD 0.283 (0.469) −0.068 (0.257) 0.757 (0.030) ** Subsample 6: non-bear market SZ 0.022 (0.000) *** 0.053 (0.000) *** 0.940 (0.000) *** 0.002 (0.912) 0.716 (0.417) 0.0682 GOLD 0.038 (0.000) *** 0.085 (0.000) *** 0.884 (0.000) *** Note: *, ** and *** denote signiﬁcance at the 10%, 5%, and 1% level of signiﬁcance, respectively. Int. J. Financial Stud. 2017, 5, 18 15 of 18

Table 10. DCC (1,1) Model-2 Step Estimation of CSI-GOLD.

Step One Step Two The Mean of Dynamic

c α β θ1 θ2 Conditional Correlations Subsample 1: 16 October 2007–4 November 2008 CSI −0.004 (0.979) −0.035 (0.113) 1.036 (0.000) *** 0.049 (0.324) 0.822 (0.000) *** −0.0021 GOLD 0.068 (0.192) 0.065 (0.051) * 0.921 (0.000) *** Subsample 2: 4 August 2009–3 December 2012 CSI 0.028 (0.035) ** 0.017 (0.004) *** 0.967 (0.000) *** 0.136 (0.021) 0.499 (0.039) 0.1669 GOLD 0.020 (0.000) *** 0.064 (0.000) *** 0.921 (0.000) *** Subsample 3: 12 June 2015–26 August 2015 CSI 4.111 (0.863) 0.029 (0.889) 0.692 (0.687) 0.171 (0.355) 0.209 (0.826) −0.0973 GOLD 0.096 (0.003) *** -0.214 (0.003) *** 1.163 (0.000) *** Subsample 4: 22 December 2015–29 February 2016 CSI 5.669 (0.275) −0.168 (0.127) 0.206 (0.804) 0.509 (0.170) −0.157 (0.283) −0.2234 GOLD 0.283 (0.469) -0.068 (0.257) 0.757 (0.030) ** Subsample 5: non-bear market CSI 0.005 (0.107) 0.055 (0.000) *** 0.945 (0.000) *** 0.019 (0.403) 0.624 (0.039) ** 0.0663 GOLD 0.030 (0.000) *** 0.079 (0.000) *** 0.895 (0.000) *** Note: *, ** and *** denote signiﬁcance at the 10%, 5%, and 1% level of signiﬁcance, respectively.

Table 11. DCC (1,1) Model-2 Step Estimation of SME-GOLD.

Step One Step Two The Mean of Dynamic

c α β θ1 θ2 Conditional Correlations Subsample 1: 16 October 2007–4 November 2008 SME −0.014 (0.928) −0.030 (0.085) * 1.032 (0.000) *** 0.048 (0.322) 0.722 (0.006) *** 0.0043 GOLD 0.068 (0.193) 0.065 (0.051) * 0.921 (0.000) *** Subsample 2: 4 August 2009–3 December 2012 SME 0.092 (0.047) ** 0.053 (0.001) *** 0.910 (0.000) *** 0.027 (0.110) 0.941 (0.000) *** 0.1451 GOLD 0.020 (0.001) *** 0.064 (0.000) *** 0.921 (0.000) *** Subsample 3: 12 June 2015–26 August 2015 SME 4.263 (0.874) −0.048 (0.864) 0.760 (0.656) −0.033 (/) 1.037 (/) −0.1767 GOLD 0.096 (0.003) *** -0.214 (0.003) *** 1.163 (0.000) *** Subsample 4: 22 December 2015–29 February 2016 SME 9.928 (0.375) −0.169 (0.205) 0.098 (0.931) −0.216 (/) 1.047 (/) −0.4166 GOLD 0.259 (0.465) −0.070 (0.239) 0.785 (0.014) ** Subsample 5: non-bear market SME 0.042 (0.000) *** 0.065 (0.000) *** 0.922 (0.000) *** 0.014 (0.423) 0.845 (0.000) *** 0.0585 GOLD 0.038 (0.000) *** 0.085 (0.000) *** 0.884 (0.000) *** Note: *, ** and *** denote signiﬁcance at the 10%, 5%, and 1% level of signiﬁcance, respectively.

Table 12. DCC (1,1) Model-2 Step Estimation of GEM-GOLD.

Step One Step Two The Mean of Dynamic

c α β θ1 θ2 Conditional Correlations Subsample 1: 4 August 2009–3 December 2012 GEM 0.091 (0.139) 0.038 (0.008) *** 0.933 (0.000) *** 0.052 (0.369) −0.155 (0.769) 0.0686 GOLD 0.016 (0.009) *** 0.068 (0.000) *** 0.920 (0.000) *** Subsample 2: 12 June 2015–26 August 2015 GEM 2.950 (0.128) −0.318 (0.108) 1.181 (0.000) *** −0.131 (0.360) 0.930 (0.000) *** −0.1414 GOLD 0.096 (0.003) *** −0.214 (0.003) *** 1.163 (0.000) *** Subsample 3: 22 December 2015–29 February 2016 GEM 8.104 (0.017) ** −0.210 (0.056) * 0.466 (0.165) −0.117 (/) 1.014 (/) −0.1418 GOLD 0.259 (0.465) −0.070 (0.239) 0.785 (0.014) ** Subsample 4: non-bear market GEM 0.018 (0.007) *** 0.038 (0.000) *** 0.956 (0.000) *** −0.013 (/) −0.001 (/) 0.0594 GOLD 0.108 (0.000) *** 0.122 (0.000) *** 0.760 (0.000) *** Note: *, ** and *** denote signiﬁcance at the 10%, 5%, and 1% level of signiﬁcance, respectively. Int. J. Financial Stud. 2017, 5, 18 16 of 18

6. Conclusions Against a background of intensified macroeconomic uncertainty, whether gold can play its traditional role of stored value effectively and become a hedge and safe haven asset for the stock market is a question worth discussing. This paper examined the dynamic relationships between the returns of the Shanghai gold spot price and the returns of the SSE Composite Index, the SZSE Component Index, the CSI 300 Index, the SME Index, and the GEM Index, using daily data from 31 October 2002 to 18 April 2017, which contained five bear markets. Our analysis yielded several noteworthy findings. First, gold was a safe haven for the SSE Composite Index, the SZSE Component Index, the CSI 300 Index, the SME Index, and the GEM Index only in Bear Market V (22 December 2015–29 February 2016); with the exception of the CSI 300 Index, gold as a safe haven for the SSE Composite Index, the SZSE Component Index, the SME Index and the GEM Index in Bear Market IV (12 June 2015–26 August 2015), while in the other three bear markets, gold did not act as a safe haven asset for China’s stock market. Second, for nonbear markets, gold also did not offer good risk hedging. The possible reasons may be that, on one hand, China’s capital market is imperfect and investors are irrational and are suffering losses in emerging market stocks rather than seeking an alternative safe haven asset to readjust their portfolios. On the other hand, China’s gold market is still in its infancy with few investment products related to gold. While our findings suggest that with the development and improvement to China’s gold market as well as China’s stock market, the hedging effect of gold on China’s stock market has strengthened remarkably. More investors are willing to include gold as a safe haven and a hedge asset against the volatility of the stock market in their investment portfolios. Nowadays, the reform of China’s capital market has been deepening, and the Chinese stock market is becoming more standard. China’s gold market is also in a period of progress and rapid development: gold future contracts have been listed and traded; Shanghai gold has been officially listed, and China is gradually gaining a place in international gold pricing. The investment use of gold has increased dramatically in China. With the acceptance of gold as a risk management tool, investors can use gold in hedging the volatility risk to Chinese stock markets and in equity-commodity portfolio management. Investment product designers in China should develop more gold related investment products for investors to choose from.

Acknowledgments: Project supported by the Social Science Foundation of Hunan Province, China. “A Research on the Onshore and Offshore RMB Exchange Rates Co-movement and Spreads” Grant No. 16JD11. Author Contributions: All authors, Ke Chen and Meng Wang, have contributed significantly and have approved the content of the manuscript. Conflicts of Interest: The authors declare no conflict of interest.

References

Arouri, Mohamed El Hedi, Amine Lahiani, and Duc Khuong Nguyen. 2015. World Gold Prices and Stock Returns in China: Insights for Hedging and Diversiﬁcation Strategies. Economic Modelling 44: 273–82. [CrossRef] Baruník, Jozef, Evžen Koˇcenda,and Lukáš Vácha. 2016. Gold, oil, and stocks: Dynamic correlations. International Review of Economics & Finance 42: 186–201. Basher, Syed Abul, and Perry Sadorsky. 2016. Hedging emerging market stock prices with oil, gold, VIX, and bonds: A comparison between DCC, ADCC and GO-GARCH. Energy Economics 54: 235–47. [CrossRef] Batten, Jonathan A., Cetin Ciner, and Brian M. Lucey. 2010. The macroeconomic determinants of volatility in precious metals markets. Resources Policy 35: 65–71. [CrossRef] Batten, Jonathan A., Cetin Ciner, and Brian M. Lucey. 2014. On the economic determinants of the gold-inﬂation relation. Resources Policy 41: 101–08. [CrossRef] Baur, Dirk G., and Brian M. Lucey. 2010. Is gold a hedge or a safe haven? An analysis of stocks, bonds and gold. Financial Review 45: 217–29. [CrossRef] Baur, Dirk G., and Thomas K. Mcdermott. 2010. Is gold a safe haven? International evidence. Journal of Banking & Finance 34: 1886–98. Int. J. Financial Stud. 2017, 5, 18 17 of 18

Beckmann, Joscha, Theo Berger, and Robert Czudaj. 2015. Does gold act as a hedge or a safe haven for stocks? A smooth transition approach. Economic Model 48: 16–24. [CrossRef] Bouri, Elie, Anshul Jain, P. C. Biswal, and David Roubaud. 2017. Cointegration and nonlinear causality amongst gold, oil, and the Indian stock market: Evidence from implied volatility indices. Resources Policy 52: 201–06. [CrossRef] Chen, An-Sing, and James Wuh Lin. 2014. The relation between gold and stocks: An analysis of severe bear markets. Applied Economics Letters 21: 158–70. [CrossRef] Choudhry, Taufiq, Syed S. Hassan, and Sarosh Shabi. 2015. Relationship between Gold and Stock Markets during the Global Financial Crisis: Evidence from Nonlinear Causality Tests. International Review of Financial Analysis 41: 247–56. [CrossRef] Ciner, Cetin, Constantin Gurdgiev, and Brian M. Lucey. 2013. Hedges and safe havens: An examination of stocks, bonds, gold, oil and exchange rates. International Review of Financial Analysis 29: 202–11. [CrossRef] Dee, Junpeng, Liuling Li, and Zhonghua Zheng. 2013. Is gold a hedge or a safe haven? Evidence from inflation and stock market. International Journal of Development and Sustainability 2: 1–16. Engle, Robert F. 1982. Autoregressive conditional heteroscedasticity with estimates of the variance of UK inflation. Econometrica 50: 987–1008. [CrossRef] Engle, Robert F. 2002. Dynamic conditional correlation: A simple class of multivariate GARCH models. Journal of Business and Economic Statistics 20: 339–50. [CrossRef] Frost, Alfred J., and Robert R. Prechter Jr. 2005. Elliott Wave Principle, 10th ed. Gainesville: New Classics Library, pp. 31, 78–85. Ghazali, Mohd Fahmi, Hooi Hooi Lean, and Zakaria Bahari. 2015. Sharia compliant gold investment in Malaysia: Hedge or safe haven? Pacific-Basin Finance Journal 34: 192–204. [CrossRef] Green, Timothy. 2007. The Ages of Gold. London: Gold Fields Minerals Services Ltd. Gürgün, Gözde, and Ibrahim Ünalmıs. 2014. Is gold a safe haven against equity market investment in emerging and developing countries? Finance Research Letters 11: 341–48. [CrossRef] Hamilton, James D. 1989. A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica 57: 357–84. [CrossRef] Huang, Shupei, Haizhong An, Xiangyun Gao, and Xuan Huang. 2016. Time–frequency featured co-movement between the stock and prices of crude oil and gold. Physica A: Statistical Mechanics and its Applications 444: 985–95. [CrossRef] Hoang, Thi-Hong-Van, Wing-Keung Wong, and Zhen Zhen Zhu. 2015. Is gold different for risk-averse and risk-seeking investors? An empirical analysis of the Shanghai Gold Exchange. Economic Modelling 50: 200–11. [CrossRef] Iqbal, Jav. 2017. Does gold hedge stock market, inflation and exchange rate risks? An econometric investigation. International Review of Economics & Finance 48: 1–17. Jain, Anshul, and P. C. Biswal. 2016. Dynamic linkages among oil price, gold price, exchange rate, and stock market in India. Resources Policy 49: 179–85. [CrossRef] Lu, Rong, and Longbing Xu. 2004. The asymmetry information effect on bull and bear stock markets. Economic Research 03: 65–72. Mensi, Walid, Shawkat Hammoudeh, Juan C. Reboredo, and Duc Kuong Nguyen. 2015. Are Sharia stocks, gold and US Treasury hedges and/or safe havens for the oil-based GCC markets? Emerging Markets Review 24: 101–21. [CrossRef] Mensi, Walid, Shawkat Hammoudeh, and Aviral Kumar Tiwari. 2016. New evidence on hedges and safe havens for Gulf stock markets using the wavelet-based quantile. Emerging Markets Review 28: 155–83. [CrossRef] Nguyen, Cuong, M. Ishaq Bhatti, Magda Komorníková, and Jozef Komorník. 2016. Gold price and stock markets nexus under mixed-copulas. Economic Modelling 58: 283–92. [CrossRef] O’Connor, Fergal A., Brian M. Lucey, Jonathan A. Batten, and Dirk G. Baur. 2015. The financial economics of gold-A survey. International Review of Financial Analysis 41: 186–205. [CrossRef] Int. J. Financial Stud. 2017, 5, 18 18 of 18

Smiech, Slawomir, and Monika Papiez. 2016. In search of hedges and safe havens: Revisiting the relations between gold and oil in the rolling regression framework. Finance Research Letters 20: 238–44. [CrossRef] Souˇcek,Michael. 2013. Crude oil, equity and gold futures open interest co-movements. Energy Economics 40: 306–15. [CrossRef] Shahzad, Syed Jawad Hussain, Naveed Raza, Muhammad Shahbaza, and Azwadi Ali. 2017. Dependence of stock markets with gold and bonds under bullish and bearish market states. Resources Policy 52: 308–19. [CrossRef]

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).