Impact of news-based equity market volatility on international stock markets

Alqahtani, Abdullah; Wither, Michael J.; Dong, Zhankui; Goodwin, Kimberly R. Article Impact of news-based equity market volatility on international stock markets Journal of Applied Economics Provided in Cooperation with: University of CEMA, Buenos Aires Suggested Citation: Alqahtani, Abdullah; Wither, Michael J.; Dong, Zhankui; Goodwin, Kimberly R. (2020) : Impact of news-based equity market volatility on international stock markets, Journal of Applied Economics, ISSN 1667-6726, Taylor & Francis, Abingdon, Vol. 23, Iss. 1, pp. 224-234, https://doi.org/10.1080/15140326.2020.1729571 This Version is available at: https://hdl.handle.net/10419/314089 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Journal of Applied Economics ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/recs20 Impact of news-based equity market volatility on international stock markets Abdullah Alqahtani, Michael J. Wither, Zhankui Dong & Kimberly R. Goodwin To cite this article: Abdullah Alqahtani, Michael J. Wither, Zhankui Dong & Kimberly R. Goodwin (2020) Impact of news-based equity market volatility on international stock markets, Journal of Applied Economics, 23:1, 224-234, DOI: 10.1080/15140326.2020.1729571 To link to this article: https://doi.org/10.1080/15140326.2020.1729571 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. Published online: 11 Mar 2020. Submit your article to this journal Article views: 2912 View related articles View Crossmark data Citing articles: 14 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=recs20 ARTICLE Impact of news-based equity market volatility on international stock markets Abdullah Alqahtani a , Michael J. Wither b , Zhankui Dong c and Kimberly R. Goodwin d a Suffolk University, Boston, MA, USA; b Powerlytics, USA; c Henan University, Kaifeng, China; d The University of Southern Mississippi, Hattiesburg, MS, USA ABSTRACT This study examines the long run impacts of equity market volatility on index returns of nine major international stock exchanges in the Western and Asian regions. This study employs the text-based Economic Market Volatility (EMV) index to measure the degree of uncertainty in the U.S. stock market. Using monthly data from December 2001 to August 2018, the estimation results derived using the standard and nonlinear ARDL models deliver several key messages. First, rising U. S. stock market volatility exhibits significant and negative impacts on stock market returns, except for the stock markets of China, Hong Kong, and India whose impacts are negative but insignificant. Second, the use of the nonlinear ARDL model does not show any signs of asymmetry in the relationship between stock market returns and changes in the EMV index, suggesting that the change in the EMV index has symmetric effects on the changes in major stock indices. ARTICLE HISTORY Received 30 April 2019 Accepted 10 February 2020 KEYWORDS Equity market volatility index; vector autoregressive; international investment 1. Introduction Many economists consider the global financial crisis during 2007 and 2008 to have been the worst financial crisis since the Great Depression. Consequently, researchers have been keen to analyze the warning signs, causes, and global impact of the global financial crisis. About a decade later, the S&P 500 and NASDAQ hit all-time highs by August 2018, and the Dow Jones Industrial Average was not far behind them. In August 2018, the S&P 500 broke 2872 points, the then all-time high recorded in January of the same year. As of August 2019, the ongoing bull market had lasted for more than 4,000 days since the recovery began, and the S&P 500 reached 2,978 points at the end of August 2019, which was 3.7 times its lowest point in March 2009. While 2017 was evidently a healthy year with low volatility and commendable growth in the equity markets, high volatility returned to major equity indices in 2018. The classical asset pricing models (CAPM) (Lintner, 1965;Merton,1973;Mossin,1966;Sharpe,1964)relatean asset’s return to either the variance of the asset’s returns or the covariance of the asset’sreturn with the overall market return. Likewise, the CAPM predicts a positive relationship between CONTACT Zhankui Dong [email protected] Institute for Management Science and Engineering, Henan University, Kaifeng, China JOURNAL OF APPLIED ECONOMICS 2020, VOL. 23, NO. 1, 224–234 https://doi.org/10.1080/15140326.2020.1729571 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/ licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. the required return and the risk premium of an asset and a negative relationship between risk and the asset price. The theoretical and empirical relationships between stock returns and volatility have gained much attention in the finance literature since the introduction of the CAPM. The theoretical work of Ross (1976), for instance, showed that there was a time-varying relationship between the risk premium and conditional covariance of an asset’s returns. Black (1976) and Christie (1982) suggest a negative relationship between stock prices and market volatility because a drop in stock prices increases leverage, which causes expected volatility to rise. Models of Campbell (1993,1996) showed that uncertainty in expected future market returns would influence investors’decision-making. Chen (2002) extended the models of Campbell (1993,1996)) by incorporating investor’s distress with total future volatility risk. The studies above suggest that increasing expected volatility results in fewer investment opportunities in the stock market. As a result, investors are willing to hedge against uncertainty in market volatility, which consequently leads to lower stock prices. On the other hand, the volatility feedback hypothesis of Campbell and Hentschel (1992) and French, Schwert, and Stambaugh (1987) claimed that high volatility tends to raise required returns on corporate stocks, thereby leading to lower stock prices. The existing empirical works, however, provide mixed findings on the linkage between expected volatility and equity returns. For instance, Fleming, Ostdiek, and Whaley (1995), Connolly, Stivers, and Sun (2005,2007), Banerjee, Doran, and Peterson (2007), Sarwar (2012a) and Smales (2017) document a significant and negative relationship between U.S. stock market returns and volatility. Similar results are also found in regions other than the U.S., including Europe (Hammoudeh, Mensi, Reboredo, & Nguyen, 2014), BRICS markets (Mensi, Hammoudeh, Reboredo, & Nguyen, 2014; Sarwar, 2012b), Latin America (Sarwar & Khan, 2017), and the MENA region (Abuzayed, Al-Fayoumi, & Arabiyat, 2018). In contrast to the findings of a negative association between returns and risk, Giot (2005) and Guo and Whitelaw (2006) found a positive relationship between changes in volatility and future stock market returns. Giot (2005) explained that rising implied volatility indicates an oversold stock market, which is usually followed by a bounce back and increase in future stock prices. This ultimately leads to a positive relationship between volatility changes and future stock market returns. The findings above are consistent with French et al. (1987) and Fleming et al. (1995), who reported positive relationships between future stock market returns and changes in market volatility. Instead of assuming a linear or symmetric relationship between stock market returns and volatility, some studies explore the possibility of asymmetric relationships between the two. Using the VIX index as a proxy of stock market volatility, Fleming et al. (1995) showed that changes in implied volatility tend to associate with negative stock market returns rather than positive stock market returns. Bekaert and Wu (2000) argued that the expected returns of the U.S. equity market are negatively correlated with the conditional returns volatility, and the linkage between risk and return may appear asymmetric across different periods. This argument is consistent with Nelson (1991) and Glosten, Jagannathan, and Runkle (1993), who conjectured that the relationship between expected returns and volatility could be either positive or negative over a finite period. Sarwar (2012a,2012b) and Mensi et al. (2014) also documented the finding of intertemporal relations between stock market risk and returns, as Sarwar (2012a) focused on S&P 100, JOURNAL OF APPLIED ECONOMICS 225 500 and 600 returns, while Sarwar (2012b) and Mensi et al. (2014) examined the stock markets in Brazil, Russia, India, and China. Each of these studies suggests that the VIX index is more of a gauge of investor fear than investor positive sentiment. More recently, Sarwar and Khan (2017) reported that the negative impact of the VIX index on Latin American stock markets was much pronounced during financial crises than normal periods. The continued inconsistency in defining the relationship between stock market volatility and returns across the U.S. and international markets emphasizes the need for additional evidence by incorporating varying time horizons, model specifications, and alternative measures of volatility in the analyses. This study addresses the need for alternative measures of volatility by examining the relationship between stock market returns and volatility in major international stock markets using a text-based measure of equity market volatility. This study, therefore, contributes to the literature in two important ways. First, it examines the hypothetical asymmetric relationship between volatility and stock market returns in international markets by using the nonlinear ARDL model (Shin, Yu, & Greenwood-Nimmo, 2014). Since the volatility index tracks U.S. market volatility, the direct effects apply to U.S. market indices, while the international effects are experienced through an indirect effect by the relationship between U.S. and other international markets. Second, it is among the first studies to incorporate a text-based measure of equity market volatility, namely the Equity Market Volatility (EMV) index, into the analysis of the risk-return relationship. The Equity Market Volatility Index is constructed through a textual analysis of economic and stock market coverage in major U.S. news publications. To the best of our knowledge, none of the existing studies examine the asymmetric risk-return relationship using the EMV index as the proxy for stock market volatility. The study that is closest in spirit is Nikkinen and Peltomäki (2019), which uses the Google Search Volume Index, a web-search frequency indicator, to reflect investors’fear of market crashes. Applying new text-based measures of volatility to an unanswered research question may help shed light on how volatility relates to stock returns. 2. Method and data 2.1. The data This study uses newly developed data, namely the Equity Market Volatility (EMV) index (Baker, Bloom, Davis, & Kost, 2019), to measure U.S. equity market volatility. While the Equity Market Volatility index is created to mirror the Volatility Index (VIX) of the Chicago Board Options Exchange (CBOE), the two indices are developed using distinct methodologies. Specifically, the EMV index begins with real-time frequency counts of selected words or terms in eleven major newspapers. 1 The textual analysis algorithm searches each newspaper for articles that contain one of more terms in each of the following sets: 1 The 11 newspapers include The Miami Herald, the Dallas Morning News, the Houston Chronicle, the San Francisco Chronicle, USA Today, the New York Times, the Los Angeles Times, the Boston Globe, the Chicago Tribune, the Wall Street Journal, and the Washington Post. 226 A. ALQAHTANI ET AL. ●E: {economic, economy, financial} ●M: {“stock market”, equity, equities “Standard and Poors”,“Standard & Poors”, “Standard and Poor”,“Standard and Poor’s”,“Standard & Poor’s”} ●V: {uncertain, uncertainty, risk, risky, volatile, volatility}. Next, the raw frequency counts undergo a set of scaling and standardization processes in order to develop the final monthly index values. The algorithm captures the concerns of the media and investors about stock market volatility in real-time, which is an essential component in understanding investors’behavior toward volatility in depth. The unique methodology of the EMV index also reflects the role of wars, policy risks, and other hardto-quantify sources on stock market volatility (Baker et al., 2019). Figure 1 depicts the EMV tracker from December 2000 to August 2018. The trend in the EMV index clearly reveals the drastic rise of uncertainty during the dot com bubble burst in 2000, the 2008 Global financial crisis, and the U.S. debt crisis in August 2011. In addition, Figure 1 shows the trend of the monthly VIX index from January 2004 to August 2018. The two volatility measures share a correlation of about 0.77 from January 2004 to August 2018. This study uses data on the monthly EMV index spanning from December 2001 to August 2018. In terms of the proxy for stock markets returns, this study uses the monthly indices of nine international major stock exchanges from Bloomberg over the same observation period. We loosely categorize the stock markets into the West and Asian regions. The West region includes the U.S. stock market (S&P 500 and Dow Jones Industrial Average), the European Union (STOXX 600), U.K. (FTSE 100), and German (DAX). The Asian region includes Japan (Nikkei 225), China (Shanghai Composite), Hong Kong (Hang Seng), and India (NIFTY 50). All data are natural-log transformed prior to any descriptive or inferential analyses. Table 1 presents the descriptions and selected summary statistics for the variables included in the analysis. Table 2 above shows the pairwise correlations between each of the variables. One can observe that the first movements of all major stock exchanges are positively correlated with each other. The correlations within the West region stock markets range from 0.80 to 0.96, indicating a strong correlation between U.S. and European stock markets. The Figure 1. Monthly EMV index (2000 to 2018) and VIX index (2004 to 2018). JOURNAL OF APPLIED ECONOMICS 227 matrix also highlights the strong correlation between the Japanese market and the rest of the international markets, with the strongest correlation being between Japan and the European Union as a whole. Interestingly, the Chinese stock market has a positive yet moderate correlation with each stock market of the West region. In addition, the correlation matrix shows that the EMV index is negatively correlated with all stock market indices, except for the Shanghai Stock Exchange Composite, implying a possible negative relationship between stock market returns and volatility at first glance. 2.2. The method This study adopts the nonlinear autoregressive distributed lag model (NARDL) (Shin et al., 2014) to explore the asymmetric relationships between stock market returns and volatility. As an extension to the traditional ARDL model (Pesaran, Shin, & Smith, 2001), the NARDL model is designed to capture both short run and long run asymmetries in a variable of interest, while reserving all merits of the standard ARDL approach. To begin with, consider the following long run model equation at the t-th month: Rt¼c0þc1EMVtþet(1) where Rtis the stock market index of each international stock exchange, EMVtis the EMV index, and ε t is the white-noise error term in month t. Equation (1) can be extended to an asymmetric long run equation as: Rt¼α0þα1EMVþ tþα2EMV tþεt(2) Table 1. Description and summary statistics of data. Variable Description Mean Std. Dev. Obs. Stock Market Index SPX S&P 500 Index 7.254 0.316 212 DJIA Dow Jones Industrial Average 9.452 0.298 212 SXXP STOXX Europe 600 Index 5.678 0.210 212 FTSE FTSE 100 Index 8.642 0.181 212 GDAXI DAX Performance-Index 8.796 0.400 212 N225 Nikkei 225 9.470 0.297 212 SSEC Shanghai Stock Exchange Composite 7.736 0.373 212 HIS Hang Seng Index 9.821 0.324 212 NSEI Nifty 50 Index 8.278 0.731 212 Stock Market Volatility EMV Equity Market Volatility Index 2.937 0.323 212 Table 2. Correlations matrix. SPX DJI SXXP FTSE GDAXI N225 SSEC HSI NSEI EMV SPX 1.000 DJI 0.993 1.000 SXXP 0.848 0.808 1.000 FTSE 0.911 0.900 0.918 1.000 GDAXI 0.936 0.936 0.846 0.953 1.000 N225 0.860 0.822 0.912 0.810 0.780 1.000 SSEC 0.540 0.565 0.539 0.611 0.699 0.458 1.000 HSI 0.768 0.793 0.674 0.849 0.883 0.604 0.781 1.000 NSEI 0.787 0.815 0.591 0.794 0.881 0.577 0.712 0.950 1.000 EMV −0.455 −0.438 −0.378 −0.422 −0.356 −0.420 0.025 −0.326 −0.342 1.000 228 A. ALQAHTANI ET AL. where α 0 ,α 1 , and α 2 are long run parameters to be estimated and ε t is the white-noise error term. The constant term α 0 captures all exogenous factors. Following Shin et al. (2014), EMVtin equation (1) can be partially decomposed into EMVþ t¼Xt j¼1ΔEMVþ j¼Xt j¼1max ΔEMVj;0  (3) and EMV t¼Xt j¼1ΔEMV j¼Xt j¼1abs min ΔEMVj;0  (4) as shown in Equation (2). EMVþ tis the partial sum of positive changes in the monthly values of the EMV, while EMV tthe is partial sum of the absolute value of the negative changes in the monthly values of the EMV. Likewise, EMVþ tcaptures the increases in the EMV, thus the value of EMVþ tincreases when the market volatility rises in each period. In contrast, EMV tonly captures the decreases in the EMV and hence the value of EMV t decreases whenever the market volatility falls. The hypothesis of asymmetric or nonlinear relationships between stock market returns and volatility can be tested by evaluating α 1 and α 2 in Equation (2) with the null of α 1 =α 2 , which implies that there is no asymmetric relationship. Rejection of the null hypothesis will prove the existence of an asymmetric relationship and vice versa. Equation (2) can be framed into a standard ARDL bound test setting as follow: ΔRt¼β0þβ1Rt1þβ2EMVþ t1þβ3EMV t1þX n1 p¼1 θ1ΔRtpþX n2 p¼0 θ2ΔEMVþ tp þX n3 p¼0 θ3ΔEMV tpþμt(5) where all symbols carry similar denotations as in Equation (2). Estimation of the bound testing model (5) undergoes several procedures. First, the optimal lag lengths n=(n1, n2, n3) are determined using the Akaike information criterion (AIC). Next, the model must be exempt from serial correlation and parameter instability (Pesaran et al., 2001). Finally, the bound test approach (Pesaran et al., 2001) can be applied to Equation (5) to detect the presence of cointegration with the null hypothesis of joint insignificance: β 2 =β 3 =0. Rejection of the null hypothesis appears when the resulting Wald F-statistics is greater than the I(1) upper bounds value as stated in Pesaran et al. (2001). Rejection of the null hypothesis concludes a cointegrating relationship between stock market returns and volatility. For the sake of comparison, this study will estimate both linear and nonlinear fashions of the stock market return model using the ARDL framework. 3. Results and discussions As a preliminary exercise, this study performs the Phillips-Perron unit root test on each variable to detect whether there are any variables integrated at second order or above. This exercise is important as the NARDL model is not appropriate if there are any I(2) variables present in the estimation. Table 3 reports the result of Phillips-Perron unit root JOURNAL OF APPLIED ECONOMICS 229 test of each variable at first-difference. The first column reports the results assuming only a constant term present in the test model. The second column allows for a constant term and a linear trend in the test model. As shown in Table 3, all variables achieve stationary after taking the first-differences, indicating that there are no I(2) variables present. Next, Table 4 presents the results of ARDL bound tests for both standard and nonlinear ARDL specifications of the model. Based on Table 4, the reported Wald-Fstatistics from both the linear and nonlinear models support a similar conclusion. The bound test result suggests that the stock market returns and volatility are cointegrated in all Western countries and in Japan at the 1% significance level. On the contrary, the Wald-Fstatistics of the China, Hong Kong, and India stock markets are lower than the upper I(1) bound value, indicating that there is no significant cointegration between stock market returns and volatility in the Asian region, except for the Japanese market. The following section reports the estimated long run relationships of the linear and nonlinear ARDL models. Table 5 presents the estimated long run relations using the linear specification of the ARDL model. Panel A reports the estimated long run coefficients. The long run coefficients on EMV are negative for all major stock markets, indicating a negative relationship between stock market returns and volatility. However, significant relationships are only Table 3. Phillips-Perron unit root tests. Variable 1st Difference, Intercept 1st Difference, Intercept and Trend SPX −12.588*** −12.738*** DJI −13.138*** −13.238*** SXXP −12.166*** −12.207*** FTSE −14.431*** −14.475*** GDAXI −13.268*** −13.330*** N225 −12.168*** −12.222*** SSEC −13.447*** −13.420*** HSI −12.779*** −12.766*** NSEI −13.810*** −13.778*** EMV −57.366*** −56.741*** EMV + −15.079*** −15.076*** EMV − −18.396*** −18.486*** *, **, *** denote rejection of null hypothesis at 10%, 5%, and 1%, respectively. Figures above are the tau statistics. Table 4. Bound tests. Dependent Variable Equation 1 Standard ARDL model EMV (K=1) Equation 2 Nonlinear ARDL model EMV (K=2) Equation 3 Standard ARDL model VIX (K=1) Equation 4 Nonlinear ARDL model VIX (K=2) SPX 12.612*** 8.025*** 11.614*** 04.421*** DJI 08.917*** 5.471*** 07.437*** 07.222*** SXXP 19.465*** 6.939*** 11.614*** 04.912*** FTSE 15.029*** 7.677*** 07.370*** 04.862*** GDAXI 11.770*** 5.791*** 05.454*** 04.227*** N225 15.108*** 7.877*** 10.181*** 06.287*** SSEC 04.493*** 2.783*** 03.478*** 03.510*** HSI 04.594*** 1.807*** 03.188*** 01.767*** NSEI 03.335*** 1.295*** 02.807*** 01.335*** *, **, *** denote rejection of null hypothesis at 10%, 5%, and 1%, respectively. Figures above are the Wald-Fstatistics. 230 A. ALQAHTANI ET AL.