Inter-day return and volatility dynamics between Japanese ADRs and their underlying securities

Abstract

In this study, we apply a more refined statistical procedure to test the dependencies and direction of inter-day spillover effects between the ADRs and their underlying shares on two nonsynchronous international markets. The empirical results provide evidence of contemporaneous return and volatility spillovers from Tokyo to New York, and vice versa. In the lagged spillover test, the evidence also suggests that the dominant market (home market) adjusts to the information from the satellite market (foreign market) in an efficient manner. In contrast, the satellite market reacts to the information from the dominant market with a delay.

Acknowledgements

I would like to thank Mark Taylor (the editor) and an anonymous referee for constructive comments and suggestions on an earlier draft of this article. I retain all responsibility for content.

Notes

¹ Recent studies on the emerging stock markets show that returns in emerging markets are more predictable than those in industrial countries and are influenced by the developed financial markets (e.g. Bekaert, 1995; Harvey, 1995; Wei et al ., 1995; Pagan and Soydemir, 2000).

² For example, Lau and Diltz (1994), Bae et al . (1999) and Kim et al . (2000) use the SUR method and close-to-close returns of ADRs and their underlying stocks in their studies. However, the SUR method does not incorporate the time-varying distribution of the stock returns, and it does not allow for the investigation of simultaneous spillover effects on the mean and the conditional variances. Moreover, to reduce the effect of non-synchronous trading between US and Japan stock markets, it would be ideal to use both the open and close prices in this study. The two-stage GJR-GARCH model applied in this article allows for time-varying conditional variances and enables us to analyse the data (open and close prices) not only on the return spillover effects but also on volatility spillover effects.

³ Since 4 of the 16 multinationals are listed in US stock market after 2 September 1993, the sample starting dates for them are determined by their actual listing dates.

⁴ It should be noted that the TSE has a one-and-half hour lunch break, so the actual trading hour is from 9:00–11:00 and 12:30–15:00, Tokyo time.

⁵ These 16 Japanese multinationals are mostly listed in the NYSE. It is noted, however, that five of them are listed in the NASDAQ.

⁶ The result is consistent with the findings of Hill et al . (1990), Tse et al . (1996) and Wang et al . (2002). For example, Hill et al . (1990) study the magnitudes of differences between trading and non-trading variances in futures markets; Tse et al . (1996) find that the index prices of the Eurodollar futures are more volatile at the Chicago Mercantile Exchange (CME) and at the London International Financial Futures Exchange (LIFFE) when the markets are trading, while they are more volatile at the Singapore International Monetary Exchange (SIMEX) when the market is closed. Wang et al . (2002) investigate the dually traded stocks in the Stock Exchanges of Hong Kong (SEHK) and London (LSE).

⁷ The standard GARCH model is symmetric in its response to past innovations. However, there are theoretical arguments that suggest a differential response in conditional variance to previous positive and negative innovations. Several alternative GARCH model specifications have been proposed in an attempt to capture the asymmetric nature of the volatility responses. Engle and Ng (1993), in a test of volatility models on Japanese stock return data, find strong support for the GJR-GARCH model which explicitly incorporates the potential for asymmetry in the conditional variance equation. Kim and Kon (1994) find that the GJR-GARCH model is the most descriptive for individual stocks.

⁸ French (1980) documents that the mean return for Mondays is more negative than that for any other day of the week on the NYSE and AMEX. He concludes that the persistently negative returns for Mondays are caused by some weekend effect. The Day of the week effect is further documented for several international equity markets. For example, Jaffe et al . (1989) and Lee et al . (1990) find Japanese indexes show negative returns on Tuesdays.

⁹ It is noted that the New York markets were closed during the week of attack and reopened on the next Monday, 17 September 2001.

¹⁰ The meteor-shower hypothesis suggests inter-day volatility spillovers from one market to the next while the heat-wave hypothesis suggests that the volatility has only country-specific autocorrelations.