Return and Volatility Spillovers and Cojump Behavior Between the U.S. and Korean Stock Markets

ABSTRACT

In this study, we examine return spillover, volatility transmission, and cojump behavior between the U.S. and Korean stock markets. In particular, we focus on cojump behavior between the two markets in order to explain the transmission of unexpected shocks. We find that the U.S. stock market causes return spillover effects in the Korean stock market, and there is significant volatility transmission between the two markets. Importantly, we find a stronger association in size, as compared with intensity, of cojumps between the U.S. and Korean stock markets, particularly during the recent financial crisis.

KEY WORDS:

• BEKK
• cojump
• contagion
• return spillover
• volatility spillover

Acknowledgments

The authors are grateful for helpful comments and suggestions from anonymous referees, Ali M. Kutan (editor), Bo Soo Kang, Dae Sup Kim, Heidi Raubenheimer, Nahe Song, and Robert I. Webb.

Notes

¹. This phenomenon is called the “contagion effect.” There is ample research on the contagion effect (e.g., Arshanapalli and Doukas 1993; Asgharian and Nossman 2011; Berben and Jansen 2005; Calvo and Mendoza 2000; Eun and Shim 1989; Hamao et al. 1990; Hilliard 1979; Kim and Ryu 2015a; Longin and Solnik 1995, 2001; Morana and Beltratti 2008).

2. In 2013, the value of exports from Korea to the United States was $62.1 billion, and the value of imports to Korea from the United States was $41.5 billion. These amounts represent 11.1 percent and 8.0 percent of Korea’s total exports and imports, respectively. Source: Korean Statistical Information Service (http://kosis.kr).

³. For more details regarding the VKOSPI, refer to Han et al. (2012), Kim and Ryu (2014, 2015b), Lee and Ryu (2013), and Ryu (2012).

4. In the equations, contemporaneous returns are not incorporated as independent variables because of the discrepancy in closing times between the U.S. and Korean markets. A description of the relationship between closing times for both markets is provided in Appendix A.

5. Similarly, Aggarwal et al. (1999) and Ramchand and Susmel (1998) employ weekly data with a sufficient number of observations to obtain reliable results for the GARCH model.

6. Submitted orders are transacted by the uniform pricing rule during the last ten minutes (from 2:50 p.m. to 3:00 p.m.). We exclude this interval when using continuous trading session data. On the first trading day of each calendar year and on Korea’s scholastic aptitude test (SAT) date, trading in the KRX begins one hour later (at 10:00 a.m.). For these two days, the KOSPI 200 index has fifty-seven intraday observations. For the detail intraday trading mechanism of the Korean financial market, refer to Ahn et al. (2008, 2010) and Ryu (2011, 2013a, 2013b, 2015), among others.

7. Analyses of volatility spillover using implied volatility indexes (e.g., VIX and VKOSPI) are not appropriate for our empirical framework, which includes BEKK GARCH, VAR/VECM, and jump detection tests exploiting high-frequency data. In the classical BEKK GARCH and VAR/VECM, the general assumption is that the observations (e.g., asset returns) are normally distributed. In the CPR jump detection test, the log change in the observation follows the process that includes the Wiener process and the Poisson process. However, assumptions that the processes of the implied volatilities are normally distributed and are described by the Wiener process and/or Poisson process are not permitted, according to previous literature. We estimate the multivariate GARCH model including the implied volatility series, but the estimated results do not provide any significant implications.

8. The contrarian effect means that price changes are negatively serially correlated, and the contrarian strategy is to sell past “winners” and buy past “losers” (e.g., Lo and MacKinlay 1990).

9. CPR and BNS jump detection tests are developed using high-frequency data. Daily return data are not appropriate for the CPR and BNS jump detection tests.

Additional information

Funding

This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2013S1A5A2A03045406).