International linkages of the Japanese stock market

International linkages of the Japanese stock market

Japan and the World Economy 20 (2008) 601–621 www.elsevier.com/locate/econbase International linkages of the Japanese stock market Terence Tai-Leung ...

1MB Sizes 0 Downloads 32 Views

Japan and the World Economy 20 (2008) 601–621 www.elsevier.com/locate/econbase

International linkages of the Japanese stock market Terence Tai-Leung Chong a, Ying-Chiu Wong a, Isabel Kit-Ming Yan b,* b

a Department of Economics, The Chinese University of Hong Kong, Hong Kong, China Department of Economics and Finance, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong, China

Received 27 August 2006; received in revised form 9 April 2007; accepted 20 June 2007

Abstract Using daily open-to-close and close-to-open stock prices, this paper examines whether there are any lead–lag relationships between the Tokyo Stock Exchange and the other G7 stock markets. In particular, this paper analyzes whether the movements of other markets in the preceding trading session can be used to formulate profitable strategies to trade in Nikkei. # 2007 Elsevier B.V. All rights reserved. JEL classification : F30; F36; G15 Keywords: Japanese stock market; International linkages

1. Introduction With growing financial integration,1 co-movements among international equity markets have become increasingly evident in the past few decades. Voluminous empirical studies suggest that there is a high degree of interdependence among international stock markets. The strong interdependence across stock markets suggests that news that surfaces in one stock market on a given date is often transmitted to other markets and triggers correlated reactions. One important implication of the growing interdependence is that price movements in one market contain important information about the price movements in the other correlated markets in the subsequent trading sessions. This information is valuable in formulating trading strategies in different national markets with different dealers’ hours.

* Corresponding author. E-mail address: [email protected] (I.-M. Yan). 1 For example, Lane and Milesi-Ferretti (2001, 2003) provides evidence of financial integration using data on different countries’ accumulated stock of international portfolios of equity and FDI. 0922-1425/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.japwor.2007.06.004

602

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

Fig. 1. Measure of financial integration.

In this study, we first examine the extent of lead–lag relationships between the Japanese stock market and the other G7 countries, and then analyze whether the movements of Nikkei in the preceding trading session can be used to formulate profitable strategies to trade in the other markets. We also examine whether one can derive profitable trading rules for the Nikkei 225 Index based on the preceding rise or decline in the indices of other markets as signals. In the literature, the relationship between the New York and Tokyo Stock Exchanges is widely examined. Barclay et al. (1990) report evidence of positive correlation in the daily close-to-close returns of common stocks dually listed on the New York and Tokyo Stock Exchanges. Becker et al. (1990), using data from 1985 to 1988, find high correlation between the open-to-close return of the U.S. equity market and the Japanese equity market in the subsequent trading day. Chowdhury (1994) reports that both the U.S. and the Japanese markets respond significantly to each other’s shocks, with the Japanese market responds to shocks in the U.S. market with a 1-day lag, while the U.S. market responds contemporaneously to shocks in the Japanese market. Vagnes et al. (1996) find that the patterns of returns in the Japanese ADR (American Depository Receipt) are affected by their trading in the U.S. stock market. Neely and Weller (2000) investigate the predictability of equity returns in the U.S.–Japan, U.S.–U.K. and U.S.–Germany markets via a VAR framework but find weak predictability. Goetzmann et al. (2001) examine the profitability of using the information of international correlations in mutual fund trading. They find profitable trading strategies in the international mutual fund markets based on information in the U.S. fund return. Recent studies like Gagnon and Karolyi (2003) also find that there is significant crossmarket correlation between the close-to-open as well as open-to-close returns of S&P 500 and Nikkei 225 during the pre- and post-crash period. A number of other studies have examined the interdependence between the U.S. market and the European stock markets. With the integration of the European economies, the role of the European stock markets in the international financial market has become increasingly important. The total market capitalization of the stock markets of U.K., Germany, France and Italy amounted to about 20 percent of the world total. Longin and Solnik (1995, 2001) consider seven OECD countries from 1960 to 1990 and report that the average correlations in stock market

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

603

returns between the U.S. and the other countries had risen by about 0.36 over this period. Longin and Solnik (2001) also report significant correlation of ‘‘extreme returns’’ between the U.S. market and the U.K., French, German and Japanese markets. Similarly, Connolly and Wang (2003) show that there are strong return co-movements in the equity markets of the U.S., U.K. and Japan from 1985 to 1996. Using a comprehensive data set on the macroeconomic news announcements, they show that a large portion of the observed co-movements in the returns of the international equity markets is attributable to the trading of private information rather than the public announcement of economic fundamentals. Studies that have examined linkages between stock markets in other regions include Fernandez-Serrano and Sosvilla-Rivero (2001), which study linkages among Asian markets, and Goetzmann et al. (2005) document the correlation structure of world equity markets over a long sample period from 1850 to the present. Unlike many of the existing studies which use the U.S. as the core country in the analysis, this paper examines the lead–lag relationships in the international stock markets with Japan as the core country. In particular, we predict that the lead–lag relationships are more significant in the more recent periods of our sample during which the degree of market integration gets stronger.2 Also, we examine various trading strategies especially from the point of view of a Japanese trader. The rest of this paper proceeds as follows: Section 2 analyzes the synchronization of Nikkei and the other G7 members from 1992 to 2003 and examines the performance of a simulated Japanese trader who buys or sells based on the movements in the other markets on the previous trading session. Section 3 provides the data description. Section 4 reports the empirical findings. Section 5 concludes and summarizes the results. 2. Methodology Following Becker et al. (1990), we study the synchronization of stock by using regression analysis and simulated trade. It is worth mentioning that there is no overlapping trading hours between the Tokyo Stock Exchange and the remaining G7 markets. The opening and closing times for the seven stock markets are provided in Fig. 2. The Tokyo Stock Exchange closes first among the four largest Stock Exchanges (U.S., U.K., Germany and Japan) and thus her closing price is known to the other three markets within the 2

Fig. 1 provides a measure of financial integration based on Lane and Milesi-Ferretti’s (2001, 2003) data on different countries’ accumulated stock of international portfolios of equity and FDI. The data summarizes the total holdings of equity and FDI by domestic residents in the rest of the world and nonresidents’ holdings in the domestic economy. This stock data arguably constitutes a better indication of integration than the flow data since they are less volatile from year to year and is less confounded by the turnover rate. Following Lane and Milesi-Ferretti, an equity-based measure of financial integration is given by the following equation which accumulates FDI and portfolio stocks as a share of GDP: GEQGDPit ¼

PEQAit þ FDIAit þ PEQLit þ FDILit GDPit

where PEQA(L) and FDIA(L) are the stocks of portfolio equity and FDI assets (liabilities). In other words, GEQGDP is an indicator of the level of equity (portfolio and FDI) cross-holdings. The figure shows that the growth in this ratio has been rapid—it has been tripled over 1983–2001 in the world. This indicates that there is an enormous increase in financial integration in the last two decades.

604

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

Fig. 2. Opening and closing time of the seven stock markets. Source: The Dow Jones Guide to the World Stock Markets (1998). There are noon breaks of 1.5 h during the trading day of Japan.

same day (Hirayama and Tsutsui, 1998). As a result, new information in the same calendar day is first revealed in the Japanese market and the response to a shock in the Japanese market should be evident internationally on the same day. As mentioned in von Furstenberg and Jeon (1989), the stock market’s responses to news on a given date surface at ‘‘brokers’ hours’’ from east to west. In view of this, Japan is chosen as the leader in this study. In doing so, we are not assuming away the impacts of the U.K., U.S. and German markets on the subsequent prices of Japan. We aim at tracing out the effects of the Japanese market on the other markets in this part of the paper. Traders may consider the Nikkei 225 in the preceding trading session as a predictor of market movements in the other markets, or they can examine changes in the other markets in the preceding day as indicators of the performance in the Tokyo Stock Exchange. In our analysis, we define the open-to-close return of the Japanese market ðNIKoc Þ as: t NIKoc t



NIKct ¼ ln NIKot

 (1)

where NIKct and NIKot refer to the closing and opening price of Nikkei 225 Stock Average at time t, respectively. Similar definition is given to INDEXoc , which represents the open-to-close t return of stock market indexes in London, New York, Paris, Frankfurt, Milan and Toronto separately.

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

605

Regressions are used to determine the relationship between the Tokyo Stock Exchange and the next day performance in the other G7 markets. To test the hypothesis that the Japanese market leads the others in the subsequent day, regressions with the open-to-close return of Nikkei (NIKoc ) as the t explanatory variable and the open-to-close returns in the other markets (INDEXoc ) as the t dependent variable are estimated. The specification of the regressions is as Eq. (2) below: ¼ alocation þ blocation NIKoc þ et INDEXoc t t

(2)

where alocation and blocation for location = {London, New York, Paris, Frankfurt, Milan and Toronto} are the intercept and slope coefficients to be estimated. et refers to the conventional stochastic errors. A significant blocation shows that the Japanese market leads the stock market in that location. Apart from the above regression analysis, the correlation between the open-to-close returns of NIKoc and INDEXoc is also computed for cross-validation. t t To test the hypothesis that other markets lead the Japanese market, the lagged INDEXoc t1 returns are used as the explanatory variable in the second regression. We estimate the following regression and check the significance of the OLS estimate of blocation. ¼ alocation þ blocation INDEXoc NIKoc t t1 þ et

(3)

To analyze the effects of the Japanese daily return on the overnight performance of the other markets, we examine the impacts of the Nikkei return on the subsequent close-to-open returns of the other markets. INDEXco ¼ alocation þ blocation NIKoc þ et ; t t

(4)

We define the close-to-open return as follows: INDEXco t



 INDEXot ¼ ln ; INDEXct1

(5)

To measure the influence of the lagged INDEXoc t1 returns on the overnight return in Japan, we run the following regression: ¼ alocation þ blocation INDEXoc NIKco t t1 þ et ;

(6)

where NIKco t



 NIKot ¼ ln : NIKct1

(7)

After analyzing the lead–lag relationships between the Japanese market and the other G7 markets, we examine the performance of a simulated trader in Japan who is assumed to buy or sell at the opening price of Nikkei, depending on the rise or fall in the other markets on the previous day. The detailed analysis is presented in Section 4B. 3. Data Our data set includes the daily open-to-close and close-to-open stock prices of seven major stock markets, namely, Tokyo, London, New York, Paris, Toronto, Milan and Frankfurt over a 12-

606

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

Table 1 Distribution of countries and their relative importance to world market portfolio in 1999 Country

Proportion of world market (%)

USA Japan U.K. Germany France Italy Canada

47.2 13.0 9.3 4.1 3.8 2.1 2.1

Total

81.6

Note: All data from this table were sourced from DataStream.

year period, from January 1, 1992 to December 31, 2003. The daily open and close data of Toronto 300, CAC 40 Instantaneous (Paris), MIB 30 (Milan), DAX (Frankfurt), Nikkei 225 stock average (Tokyo), FTSE 100 (London), Dow Jones Industrials (New York) and S&P 500 (New York) are mainly obtained from DataStream. Table 1 shows a complete list of countries and their relative share in the world portfolio in 1999. The total capitalization of the G7 stock markets was about 80 percent of the world market. Testing for the lead–lag relationships among the international markets presents a problem of data synchronization due to time-zone differences. The technical problem in studying pricing relations across markets is due to the existence of non-synchronous holidays and the bi-weekly Saturday trading on the Tokyo Exchange. Nevertheless, an advantage of the trading strategies used in this paper is that the trading activity in Tokyo does not need to be concurrent to the other six stock markets. Statistical characterizations of the returns in our sample are provided in Table 2. It is observed that the average returns over this long sample period vary across markets. While Dow Jones boosts daily returns above 0.04 percent, Nikkei 225 could have lost a meager 0.06 percent. However, the high returns of the Dow Jones Industrial Average have been accompanied by moderately higher risk, as measured by a daily standard deviation of almost 1.03 percent. On the other end of the spectrum, the Toronto 300 market’s volatility is only 0.74 percent. Most returns are left-skewed, as indicated by the negative skewness in Table 2. In addition, the kurtosis coefficients are well above 3 for most markets. This implies that extreme return realizations occurred more frequently than it could have been expected if returns are normally distributed. 4. Empirical findings 4.1. Regression and correlation analysis In this subsection, we examine the lead–lag relationships between the Japanese market and the other markets. A graphical summary of the lead–lag relationships examined in Tables 3–6 is provided in Fig. 3. Table 3 presents the results of regressions and correlations between the open-to-close returns in Japan and the other markets (NIKoc and INDEXoc ). t t Table 3 shows that the correlation between the open-to-close Nikkei return and open-to-close returns in German DAX, U.K. FTSE 100, U.S. Dow Jones and S&P 500 range from 0.055 to 0.1491, and are all significant at the 1 percent level. This indicates significant positive linear

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

607

Table 2 Descriptive statistics for international stock market returns Index

No. of observations

Mean

S.D.

Skewness

Kurtosis

Nikkei 225 Toronto 300 CAC 40 MIB 30 DAX FTSE 100 Dow Jones Ind. S&P 500

2944 2887 2876 2463 2885 2900 2944 2755

0.0006 0.0001 0.0001 0.0001 0.0004 0 0.0004 0.0003

0.0135 0.0074 0.0122 0.0117 0.0130 0.0106 0.0103 0.0102

0.2186 0.1401 0.0141 0.0147 0.2462 0.1721 0.3155 0.1555

2.4920 6.2440 3.6182 1.3768 5.1823 3.2079 4.9013 4.9661

Average

2832

0.0001

0.0111

0.1050

3.9986

All the returns in this paper are local currency returns. Note: The table provides the descriptive statistics of the daily intraday-stock market returns over the period 1992 to 2003.

Table 3 OLS regression and correlation results (INDEXoc ¼ alocation þ blocation NIKtoc þ et ) t Location

Index

Intercept (alocation)

NIKtoc (blocation)

F-value

R2

Correlation

Canada France Italy Germany U.K.

Toronto 300 CAC 40 MIB 30 DAX FTSE 100

0.0001 0.0001 0.0001 0.0004 0.0001

0.0064 (0.61) 0.0053 (0.31) 0.0096 (0.54) 0.0815** (4.54) 0.1174** (8.09)

0.37 0.10 0.29 20.61 65.41

0.0001 0.0000 0.0001 0.0071 0.0222

0.0114 0.0058 0.0108 0.0845** 0.1491**

U.S.A.

Dow Jones Ind. S&P 500

0.0697** (4.97) 0.0427** (2.88)

24.67 8.28

0.0084 0.0030

0.0915** 0.0550**

(0.45) (0.51) (0.43) (1.63) (0.59)

0.0004* (2.07) 0.0004 (1.79)

t-Value is in parentheses. * Significant at the 5 percent level for a two-tailed test. ** Significant at the 1 percent level for a two-tailed test.

association between the pairwise open-to-close returns series. Relatively speaking, among these four indices, Nikkei and FTSE 100 have the strongest correlation whereas Nikkei and S&P 500 have the weakest correlation. However, as a finding of correlation does not give any indication of lead–lag direction, we turn to our regression results for the clue. Table 3 shows that the parameter blocation in regression (2) for the case of DAX, FTSE 100, Dow Jones and S&P 500 are significantly different from zero at the 1 percent level, implying that the Japanese stock market performance in the current period does explain the subsequent fluctuations in the DAX, FTSE 100, Dow Jones and S&P 500 open-to-close returns. This finding is similar to that of Hirayama and Tsutsui (1998), which shows that large price movements (those that are more than 0.4 percent) in the Japanese market Granger cause price movements in the U.K., U.S. and German markets. As for the regression and correlation coefficients between the returns of Nikkei and Toronto 300, Nikkei and MIB 30 as well as Nikkei and CAC 40, they are all insignificant at the 10 percent level. This indicates that there is no evidence that the open-to-close performance of Japan stock market leads the Canada, France and Italy stock markets. Table 4 shows the test results for the influence of the lagged returns in the other markets oc (INDEXoc ). t1 ) on the subsequent open-to-close return in Japan (NIKt

608

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

Table 4 oc ¼ alocation þ blocation INDEXt1 þ et ) OLS regression and correlation results (NIKoc t Location

Index

oc INDEXt1 (blocation)

Intercept (alocation) *

**

F-value

R2

Correlation

Canada France Italy Germany U.K.

Toronto 300 CAC 40 MIB 30 DAX FTSE 100

0.0006 (2.24) 0.0006* (2.41) 0.0005 (2.01) 0.0005 (1.92) 0.0006* (2.35)

0.2044 0.1538** 0.1248** 0.1212** 0.1556**

(6.12) (7.47) (5.54) (6.30) (6.63)

37.44 55.86 30.73 39.72 43.98

0.0129 0.0192 0.0123 0.0137 0.0151

0.1136** 0.1386** 0.1111** 0.1170** 0.1227**

U.S.A.

Dow Jones Ind. S&P 500

0.0006** (2.58) 0.0006* (2.34)

0.1693** (7.04) 0.2054** (8.45)

49.60 71.34

0.0167 0.0254

0.1292** 0.1595**

F-value

R2

Correlation

197.48 300.49 212.76 184.89 109.00

0.0645 0.0953 0.0796 0.0607 0.0365

0.2540** 0.3087** 0.2822** 0.2464** 0.1911**

18.20 61.30

0.0062 0.0219

0.0787** 0.1481**

R2

Correlation

t-Value is in parentheses. * Significant at the 5 percent level for a two-tailed test. ** Significant at the 1 percent level for a two-tailed test.

Table 5 ¼ alocation þ blocation NIKtoc þ et ) OLS regression and correlation results (INDEXco t Location

Index

NIKtoc (blocation)

Intercept (alocation) **

**

Canada France Italy Germany U.K.

Toronto 300 CAC 40 MIB 30 DAX FTSE 100

0.0003 (2.70) 0.0005** (3.09) 0.0001* (2.51) 0.0008** (6.22) 0.0002* (2.40)

0.1131 0.1987** 0.1999** 0.1327** 0.0584**

(14.05) (17.33) (14.59) (13.60) (10.44)

U.S.A.

Dow Jones Ind. S&P 500

0.0001 (1.16) 0.0001 (0.67)

0.0171** (4.27) 0.0451** (7.83)

t-Value is in parentheses. * Significant at the 5 percent level for a two-tailed test. ** Significant at the 1 percent level for a two-tailed test.

Table 6 oc ¼ alocation þ blocation INDEXt1 þ et ) OLS regression and correlation results (NIKco t Location

Index

Intercept (alocation) **

(2.71) (4.36) (2.99) (3.92) (3.72)

Canada France Italy Germany U.K.

Toronto 300 CAC 40 MIB 30 DAX FTSE 100

0.0002 0.0004** 0.0003** 0.0003** 0.0003**

U.S.A.

Dow Jones Ind. S&P 500

0.0002** (3.36) 0.0003** (3.27)

oc INDEXt1 (blocation) **

F-value

(17.41) (21.20) (15.24) (24.56) (19.88)

303.20 449.32 232.33 603.39 395.41

0.0958 0.1361 0.0864 0.1742 0.1209

0.3095** 0.3690** 0.2939** 0.4174** 0.3477**

0.2148** (32.97) 0.2170** (27.18)

1086.70 738.88

0.2712 0.2129

0.5208** 0.4615**

0.2105 0.1401** 0.1214** 0.1504** 0.1433**

t-Value is in parentheses. ** Significant at the 1 percent level for a two-tailed test.

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

609

Fig. 3. Relationships between the OLS regression tables (Tables 3–6). Legend: o, open; c, close; NIKtco , overnight closeto-open return of the Japanese market in the preceding night of time t; NIKoc , open-to-close return of the Japanese t oc , open-to-close return of the other G7 markets in time t; INDEXt1 , open-to-close return of the market in time t; INDEXoc t other G7 markets in time t  1; INDEXco , overnight close-to-open return of the other G7 markets in the preceding night t of time t.

It is observed from Table 4 that the correlations between the current open-to-close Nikkei return and the 1-day lagged open-to-close returns of all the other G7 markets are significant at the 1 percent level. Further examination of blocation of regression (3) indicates that the open-toclose performance of the G7 markets on the previous day has a significant impact on the current Japanese open-to-close return. Table 4 shows that the changes in the lagged open-to-close returns of Toronto 300, MIB 30 and CAC 40 have significant impacts on the open-to-close return of the Japanese market, though the reverse impact is less pronounce, as indicated in Table 3. Table 5 presents the correlation coefficients and the results of the regression (4) of the open-toclose return in the Japanese market (NIKoc ) on the overnight close-to-open returns in the others t G7 markets (INDEXco ). t Table 5 strongly suggests that the Japanese open-to-close performance has a significant impact on the overnight close-to-open returns of the remaining G7 countries. Combining this finding Table 7 The number of constituent stocks of different indices that are cross-listed in the other markets Toronto 300 Toronto 300 CAC 40 MIB 30 DAX FTSE 100 DJI S&P 500 Nikkei 225 Total number of constituent stocks

CAC 40

MIB 30

DAX

FTSE 100

DJI

S&P 500

1 0 0 1 0 0 0

0 0 0 0 0 0

0 0 0 0 0

0 0 0 0

0 1 0

29 0

0

223

39

30

30

102

29

500

Nikkei 225

225

610

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

Table 8 Performance of Nikkei 225 in Day t with various indices in Day t  1 as triggers Triggers

Triggert1 up by 0.5% Number of times Mean NIKt return % Up in NIKt Triggert1 down by 0.5% Number of times Mean NIKt return % Down in NIKt Annualized return of the trading strategy (%) Triggert1 up by 1% Number of times Mean NIKt return % Up in NIKt Triggert1 down by 1% Number of times Mean NIKt return % Down in NIKt Annualized return of the trading strategy (%) Triggert1 up by 1.5% Number of times Mean NIKt return % Up in NIKt Triggert1 down by 1.5% Number of times Mean NIKt return % Down in NIKt Annualized return of the trading strategy (%) Triggert1 up by 2% Number of times Mean NIKt return % Up in NIKt Triggert1 down by 2% Number of times Mean NIKt return % Down in NIKt Annualized return of the trading strategy (%)

(a) Toronto 300

(b) CAC 40

(c) MIB 30

(d) DAX

(e) FTSE 100

(f) Dow Jones

(g) S&P 500

544 0.0004 49

905 0.0002 50

744 0.0005 48

708 0.0011 53

793 0.0015 55

836 0.0011 53

771 0.0005 53

531 0.0014 56 3.58

870 0.0011 54 5.39

774 0.0009 52 1.86

737 0.0024 58 22.70

781 0.0032 60 34.54

722 0.0024 59 23.29

666 0.0024 59 16.91

178 0.0027 45

472 0.0003 50

420 0.0013 47

377 0.0009 52

375 0.0033 61

394 0.0009 53

359 0.0003 51

193 0.0014 58 2.10

487 0.0015 55 4.28

428 0.0016 54 0.42

441 0.0031 60 14.51

366 0.0038 62 23.36

348 0.0034 63 13.06

333 0.0020 57 6.03

61 0.0076 34

231 0.0004 50

222 0.0009 49

217 0.0015 53

184 0.0043 63

171 0.0003 50

158 0.0000 47

84 0.0000 48 3.91

267 0.0019 57 2.93

219 0.0012 53 0.09

287 0.0027 60 9.03

192 0.0058 67 16.72

172 0.0028 60 4.20

177 0.0033 61 4.71

30 0.0116 30

117 0.0009 50

98 0.0034 37

134 0.0010 53

87 0.0030 55

72 0.0006 53

69 0.0033 39

35 0.0025 57 2.24

135 0.0020 59 1.09

94 0.0022 59 1.27

177 0.0029 63 5.23

92 0.0076 70 8.01

85 0.0031 58 2.38

76 0.0029 54 0.20

‘‘Number of times’’ refers to the number of times the triggers send out signals. ‘‘Mean NIKt return’’ refers to the mean return of the Nikkei Index given the signals. ‘‘% Up in NIKt’’ refers to the percentage of times the Nikkei Index goes up given the signals. ‘‘% Down in NIKt’’ refers to the percentage of times the Nikkei Index goes down given the signals. ‘‘Annualized return of the trading strategy’’ refers to the annualized return of the trading strategy specified in Eqs. (8) and (9).

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

611

with those from Table 3, it can be said that both the open-to-close and close-to-open returns of DAX, FTSE 100, Dow Jones and S&P 500 are influenced by the open-to-close return on Nikkei of Japan. At the same time, we find evidence that the open-to-close Nikkei return leads the close-to-open (Table 5) but not the open-to-close (Table 3) returns of Toronto 300, MIB 30 and CAC 40. Table 6 shows the impacts of the open-to-close returns in the rest of the G7 markets co (INDEXoc ). t1 ) on the overnight close-to-open return in the Japanese market (NIKt It can be seen from the correlation coefficients in Table 6 that the Nikkei close-to-open return is correlated to the lagged open-to-close returns of the other G7 indices at the 1 percent level. Moreover, from the results of regression (6), we know that the close-to-open return of the Japanese stock market is affected by all the other G7’s open-to-close returns. Taking this finding together with those suggested by Table 4, we conclude that the lagged open-to-close returns of the other G7 stock markets have significant positive impacts on both the open-to-close (Table 6) as well as close-to-open (Table 4) performance of the Japanese market. To investigate the effect of cross-listing in our analysis, we have reported in Table 7 the information about the number of constituent stocks of different indices that are cross-listed in other markets. Only one company listed in Toronto 300 was cross-listed in CAC 40 and FTSE 100, and one company that was listed in FTSE 100 was cross-listed in S&P 500. Overall, the percentage of cross-listed companies in our sample is very low (less than 1 percent). This indicates that the effect of cross-listing is not likely to be a strong driving force behind our results. 4.2. Simulation trade In this subsection, we use the method discussed in Eq. (8) and (9) to assess the simulated trader performance in our sample. Table 8 presents the performance of the simulated trading using today’s open-to-close return of different indices as triggers to predict the up and down movements of tomorrow’s open-to-close return of Japan’s Nikkei 225 Index. The index triggers include the Toronto 300, CAC 40, MIB 30, DAX, FTSE 100, Dow Jones and the S&P 500 Index. This analysis is based on the framework of Becker et al. (1990). The trading rule is defined as follows: Buy : INDEXoc > x% t

(8)

<  x% Sell : INDEXoc t

(9)

where x = {0.5, 1, 1.5, 2} are the filter rules that are considered in this study. The profits for these positions are accordingly defined as profit ¼ NIKoc t profit ¼ NIKoc t

for a long position for a short position

(10) (11)

The positions are closed at the end of the day. The number of profitable trading days along with the mean returns generated by these trading strategies is counted. Column (a) of Table 8 presents the result when the Toronto 300 Index is used as the predictor. The up triggers from the trading strategy predict profitable trading days 30 percent of the time for the case of a 2 percent filter rule, and 49 percent of the time for the case of a 0.5 percent filter rule. As a result, slight losses are incurred by following this trading rule, as shown by the negative

612

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

mean returns in all cases. However, the down triggers foretell profitable returns the next day with a slightly higher precision of 48 percent of the time for the case of a 1.5 percent filter rule, and 58 percent of the time for the case of a 1 percent filter rule. The results of the trading strategies using today’s CAC 40 open-to-close return as the predictor of the next day’s movement of Nikkei is shown in column (b) of Table 8. We observe that there is only 50 percent precision for up triggers in all cases. Again, a slight loss exists. However, the down triggers foretell negative returns the next day with a slightly higher precision of 54–59 percent. It is found that the higher the filter standard used in the down trigger rule (say, increasing from 0.5 to 2 percent filter), the better is the trading strategy in predicting the future movement of the Nikkei Index. Column (c) of Table 8 presents the results of trading strategies using today’s open-to-close returns of the MIB 30 Index as an indicator to predict the up and down movements of the next day’s open-to-close return of Nikkei 225. The up triggers predict profitable trading days 37–49 percent of the times. Nonetheless, the mean returns of Nikkei are negative in all cases. This implies that the up triggers trading rule of MIB 30 Index is not a good predictor after all. On the other hand, the down triggers forecast negative-return trading days 52–59 percent of the time. Column (d) of Table 8 presents the results of trading strategies using the returns of the DAX Index as a predictor of Nikkei 225. The up triggers successfully predict profitable trading days 52–53 percent of the times, whereas the down triggers successfully forecast negative returns 58– 63 percent of the times. In general, down triggers can predict profitable Japanese trading days better than up triggers. The results of filter rules based on the trading signals from the FTSE 100 Index are shown in column (e) of Table 8. The results demonstrate that the market performance in the London market today is a reasonably good predictor of the next day market performance in Japan. This holds at all trigger levels. The table shows that 55–63 percent of the up triggers are captured by this filter rule. Again, the down triggers foretell negative returns the next day with slightly higher precision of 60–70 percent. The results of trading strategies based on the filter signals from the Dow Jones Industrial Average are shown in column (f) of Table 8. The up triggers predict profitable trading days 50–53 percent of the time, making profitable returns ranging from 0.03 to 0.11 percent. Again, the down triggers forecast negative returns the next day with slightly higher precision of 58–63 percent of the times. All returns generated from the down triggers are negative, ranging from 0.24 to 0.34 percent in magnitude. Column (g) of Table 8 presents the results of trading strategies using the returns of the S&P 500 Index as an indicator for the up and down movements in Nikkei. The result shows that the up triggers predict profitable trading days 39–53 percent of the times. Moreover, the down triggers forecast negative-return trading days 54–61 percent of the time. To gauge the performance of the trading rule based on the buy signal and sell signal stated in Eq. (8) and (9), we compute the annualized returns generated by this trading strategy for various triggers. The annualized return is defined as K hY

ð1 þ r i Þ

i1=12

1

i¼1

where ri refers to the return from trading after receiving the ith signal from the trading rule. K is the total number of signals received over the 12-year sample period. The annualized returns based on different triggers are reported in Table 8.

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

613

The table indicates that this trading rule yielded the highest return with the U.K. Index (FTSE 100) as trigger, followed by the U.S. Index (Dow Jones) and the German Index (DAX). In general, trading rules with lower filter values in the triggers give rise to more trading and tend to give higher returns. 4.3. Comparison with Becker et al.’s findings Comparing our findings with those of Becker et al. (1990) on the Japan–U.S. relationship, the main difference is that they find no relationship between the close-to-open return of the Japanese market and that of the U.S. market, while we do find clear evidence of such a relationship.3 To understand what gives rise to the difference, it is important to understand why, and under what circumstances, we can expect transmissions from the Japanese market to the U.S. market. Our conclusions are summarized as follows: (1) While Becker et al. (1990) focuses mainly on the period 1983–1988, our paper employs data from a more recent time period, from 1992 to 2003, during which financial markets are more integrated. Based on the Coordinated Portfolio Investment Survey (CPIS) of the IMF, the equity securities investment between Japan and the U.S. has risen rapidly in the last decade. The amount of equity securities investment from Japan to the U.S. has increased by 64 percent from US$ 87,122 million in 1997 to US$ 142,788 million in 2003.4 During the same period, the amount of equity securities investment from the U.S. to Japan has increased by 89 percent from US$ 135,278 to 255,496 million. With the growing interdependence between the two markets, traders tend to pay more attention to the information revealed in the other market. This explains why studies which use relatively more recent data (e.g., Chowdhury (1994) and Hirayama and Tsutsui (1998)) tend to find more evidence on the transmission from Japan to the U.S. (2) The last decade has witnessed significant changes in the ‘‘price discovery mechanism’’ around the globe due to increasing international cross-listing activities. Studies of price discovery on cross-listed stocks, including Eun and Jang (1997), Eun and Sabherwal (2003), Grammig et al. (2005), Pascual et al. (in press) and Su and Chong (2007), have reported that the bulk of price discovery occurs in the home market rather that in the U.S. market. This is because it is usually the case that smaller countries cross-list their stocks in larger foreign markets, it is not surprising to find more recent evidence that a smaller stock market (like Japan) leads a bigger one (like U.S.) in the short run. 4.4. Robustness checks 4.4.1. Robustness of results during the long-term global up-trend and down-trend We carry out a number of experiments to check the robustness of our results. Firstly, we check the robustness of our results during the global up- and the down-trend periods. To identify the upand down-trend of the global stock market, we apply the procedure of Pagan and Sossounov 3

The empirical findings in the literature are not unanimous. Studies that obtain similar results as Becker et al. (1990) include Eun and Shim (1989), Hamao et al. (1990) and Arshanapalli and Doukas (1993). Studies that arrive at similar conclusions as ours include Chowdhury (1994) and Hirayama and Tsutsui (1998), both of which employ more recent data in their analysis. 4 CPIS is a database that contains global data on the cross-border bilateral holdings of securities and derived portfolio investment liabilities. It was first introduced in 1997.

614

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

Fig. 4. MSCI World Index.

(2003) on the Morgan Stanley Capital International (MSCI) World Market Index.5 This procedure allows us to identify the peaks and troughs of the MSCI series within windows of 8 months and determine turning points of the series between up-phases and down-phases.6 A plot of the MSCI World Market Index is provided in Fig. 4. There is one turning point identified by the procedure, which separates the series into an up-phase and a down-phase. The up-phase is from January 1, 1992 to March 27, 2000, and the down-phase is from March 28, 2000 to December 31, 2003. Overall, we find that the predictive power and the correlations between the Japanese market and the other markets are qualitatively the same during both the up-trend and the down-trend.7 4.4.2. Robustness of results during the short-term global up-trend and down-trend To pursue an investigation of the lead–lag relationships during short-term global up and down phases, we divide the sample into two sub-samples for the short-term up- and down-trend periods based on the return of the MSCI World Market Index. We define the short-term weekly up-trend as the period when the average return in the preceding 5 days (the preceding week8) exceeds the average return in the preceding 20 days (the preceding month). The idea is that, when the global market is going up, the average return over the recent past (say, over the preceding 5 days) should be above that over a longer horizon in the past (say, over the preceding 20 days). We define the short-term down-trend as the time when the average return in the preceding 5 days is below the average return in the preceding 20 days. We reproduce Tables 3–6 for both the up-phase and the down-phase. 5

The Morgan Stanley Capital International is a market capitalization-weighted index that measures the performance of stock markets in 22 countries, including Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Hong Kong, Ireland, Italy, Japan, Malaysia, Netherlands, New Zealand, Norway, Singapore, Spain, Sweden, Switzerland, United Kingdom and United States. 6 See Appendix B of Pagan and Sossounov (2003) for more details. 7 Due to the limited space, the full estimation results for the robustness checks are not included in this paper. They are available upon requests from the authors. 8 The preceding 5 days include period t  4 up to period t.

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

615

We find that the predictive power of the open-to-close returns in Japan (NIKoc ) on those of t the other markets (INDEXoc ) during the up-trend is very similar to the results estimated in the t original benchmark regression using the whole sample as reported in Table 3. The open-to-close returns of Japan are found to be significant predictors of the returns in Germany, the U.K. and the U.S. During the down-trend, the open-to-close returns of Japan remain a significant predictor of the returns in the U.K., but they become less informative as predictors of the returns in Germany and the U.S. Robustness check also shows that the open-to-close returns of all other G-7 indices are significant predictors of the open-to-close returns of Japan in the subsequent trading session during both the short-term up-phase and the short-term down-phase, with the exception of Italy. The results resemble what we find based on the regressions with the whole sample in Table 4. Similarly, robustness check also indicates that the estimation results in Tables 5 and 6 are robust during both the short-term up-trend and down-trend periods. This suggests that the opento-close returns in Japan have significant predictive power on the subsequent close-to-open returns in all the other countries during both the up-phase and the down-phase. Robustness check on Table 6 also indicates that the predictive power of the open-to-close returns in the other markets on the subsequent close-to-open returns of the Japanese market is significant during both phases, with the coefficients being relatively larger during the up-phase. 4.4.3. Robustness of results during the low and high volatility periods As the second robustness check, we examine the robustness of our results during the low and high volatility periods. We employ two approaches to identify the high volatility periods. The first approach defines the high volatility periods as the turmoil periods in our sample. They include the time during the Mexican crisis (December 1994), the East Asian crisis (October and November 1997) and the technology bubble in the U.S. (January 2000). The second approach defines the high volatility periods as the periods when the absolute returns of the independent variable in the regression exceed 1.5 standard deviations from the mean. This approach is similar to the method used by Fabozzi and Francis (1977) in the identification of substantial up and down stock market movements. Based on the first approach (the crisis dates approach), we find that the predictive powers of the open-to-close Nikkei returns on the open-to-close returns of the other countries during the low volatility periods are very similar to those of the whole sample which are reported in Table 3. For the high volatility periods, the regression coefficients and the correlation between the open-to-close returns of Japan and those of the other countries are generally higher. Similarly, the coefficients in the regression of the open-to-close returns of the other markets on those of the Japanese market (that is, under the specification of Table 4) are found to be higher during the high volatility periods. Such pattern is also observed under the specification of Table 5 (which uses the open-to-close returns in the Japanese market to predict the subsequent close-toopen returns in the other markets during the low and high volatility periods). Nevertheless, the model specified in Table 6 (which uses the open-to-close returns in the other markets to predict the subsequent close-to-open returns in the Japanese market) do not exhibit stronger coefficients during the high volatility periods. Based on the second approach (the standard deviation approach) in the identification of the high volatility periods, we find that the open-to-close returns of Japan can be used to predict more countries during the low volatility periods than the high volatility periods. During the high

616

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

volatility periods, the returns in Japan are significant predictors of the returns in Germany, U.K. and the U.S., as in the case for the full sample given in Table 3. During the low volatility periods, the returns in Japan are found to be significant predictors of the returns in France as well. Using this approach, we also find that the open-to-close returns of the indices in the six countries are all significant predictors of the open-to-close returns in Japan (under the setting of Table 4) both in the high volatility and the low volatility periods. The patterns found in Tables 5 and 6 are similar to those found based on the first approach. 4.4.4. Robustness of results for different years For each of the regressions in Tables 3–6, we run the regression once for each year and plot the estimated slope coefficient (the coefficient that controls the lead–lag relationship) for different

Fig. 5. Plots of the estimated slope coefficient b of the regressions in Table 3 for different years: INDEXtoc ¼ alocation þ blocation NIKtoc þ et . means significant at 10 percent level; means significant at 5 percent level.

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

617

Fig. 6. Plots of the estimated slope coefficient b of the regression in Table 4 for different years: NIKtoc ¼ alocation þ blocation INDEXoc means significant at 10 percent level; means significant at 5 percent level. t1 þ et .

years. Figs. 5–8 plot the estimated slope coefficients of the regressions presented in Tables 3–6, respectively. The coefficients denoted by (the slant-stripe bar) and (the horizontal-stripe bar) represent coefficients which are significant at the 5 and 10 percent level, respectively. Broadly speaking, the magnitude and the significance of the coefficients in these year-by-year regressions are consistent with what we find with the pooled sample period. For instance, in Fig. 5 (which plots the estimated coefficients of the regressions presented in Table 3 of the paper), we find that Germany, the U.K. and the U.S. are the countries with the most significant coefficients over different periods. Fig. 6 (which plots the estimated coefficients of the regressions presented in Table 4 of the paper) indicates that the estimated coefficients are significant in more than half of the periods for all countries, with the exception of Italy only. Similarly, Figs. 7 and 8 (which plot the estimated coefficients of the regressions presented in Tables 5 and 6, respectively) show that the estimated coefficients are significant in at least

618

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

Fig. 7. Plots of the estimated slope coefficient b of the regression in Table 5 for different years: INDEXtco ¼ alocation þ blocation NIKtoc þ et . means significant at 10 percent level; means significant at 5 percent level.

three-fourth of the sample periods. In particular, the coefficients for Canada, France, Italy and Germany are unanimously significant in all periods after 1998. One point that is worth mentioning is that, when we compare the magnitude and significance of the coefficients in the first half of the sample period (1992–1997) and the second half of the sample period (1998–2003), we find that on average more coefficients are significant in the second half of the sample. This pattern is particularly evident in Figs. 7 and 8. For example, for four of the six countries (Canada, France, Italy and Germany) in Fig. 7, the coefficients are all significant in the second half of the sample period but not in the first half. For the U.S., the coefficients are significant in over half of the years in the second sample period but not in the first half. This contrast is even larger in Fig. 8. The coefficients for all periods in the second half of the sample are significant for all countries, but this is not the case in the first half of the sample for Canada, Italy, Germany and the U.K. This stylized fact

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

619

Fig. 8. Plot of the estimated slope coefficient b of the regression in Table 6: NIKco ¼ alocation þ blocation INDEXoc t t1 þ et . means significant at 10 percent level; means significant at 5 percent level.

coincides with our argument that the financial market integration is getting stronger towards the more recent periods. 5. Conclusion This paper investigates the lead–lag relationships between the stock index in Japan and those in the other G7 countries. We investigate to what extent the stock exchange in one market may become a price discovery mechanism for the stock exchanges in the other countries. We use the daily open-to-close and close-to-open data from 1992 to 2003 to identify the patterns of linkage between these markets. By examining the synchronization of stock price movements and performing simulated trade, we find various lead–lag effects from the open-to-close returns of stocks in Toronto, Paris, Frankfurt, London, Milan and New York Stock Exchange in the previous trading day to the Japanese equity market return in the current period. The lead–lag relationships

620

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

are found to be slightly stronger generally in the short-term global up-trend than in the downtrend, and are similar during the high and low volatility periods. The simulated trading strategies reveal that the filter rules perform reasonably well in forecasting the up and down price movements in the Japanese stock market. In particular, the stock market index of Japan is well predicted by the movement of the FTSE 100 Index. Regardless of the trigger level, the results reveal that the next day market performance in Japan can be predicted by trading strategies using the signals from most of the G7 markets. The results in this paper are of valuable reference to global investors (Levy and Sarnat, 1970; Solnik, 1974; Adler and Dumas, 1983) and have important implications to the G7 policy makers (Roll, 1989). Acknowledgement We would like to thank Andy Kwan and Sunny Kwong for helpful comments. References Adler, M., Dumas, B., 1983. International portfolio choice and corporation finance: a synthesis. Journal of Finance 38, 925–984. Arshanapalli, B., Doukas, J., 1993. International stock market linkages: evidence from the pre-and post-October 1987 period. Journal of Banking and Finance 17, 193–208. Barclay, M.J., Litzenberger, R.H., Warner, J.B., 1990. Private information, trading volume, and stock-return variances. Review of Financial Studies 3, 233–253. Becker, K.G., Finnerty, J.E., Gupta, M., 1990. The intertemporal relation between the U.S. and Japanese stock markets. Journal of Finance 45, 1297–1306. Chowdhury, A.R., 1994. Stock market interdependencies: evidence from the Asian NIEs. Journal of Macroeconomics 16, 629–651. Connolly, R., Wang, F., 2003. International equity market comovements: economic fundamentals or contagion. PacificBasin Finance Journal 11, 23–43. Eun, C., Sabherwal, S., 2003. Cross-border listings and price discovery, evidence from U.S.-listed Canadian stocks. Journal of Finance 58 (2), 549–575. Eun, C., Jang, H., 1997. Price interaction in a sequential global market: evidence from the cross-listed stocks. European Financial Management 3 (2), 209–235. Eun, C., Shim, S., 1989. International transmission of stock market movements. Journal of Financial and Quantitative Analysis 24, 241–256. Fabozzi, F.J., Francis, J.C., 1977. Stability tests for alphas and betas over bull and bear market conditions. Journal of Finance 32 (4), 1093–1099. Fernandez-Serrano, J.L., Sosvilla-Rivero, S., 2001. Modelling evolving long-run relationships: the linkages between stock markets in Asia. Japan and the World Economy 13, 145–160. Gagnon, L., Karolyi, A., 2003. Information, trading volume and international stock market comovements. in: Choi, J., Hiraki, T. (Eds.), International Finance Review, 4. Goetzmann, W.N., Zoran, I., Rouwenhorst, K.G., 2001. Day trading international mutual funds: evidence and policy solutions. Journal of Financial and Quantitative Analysis 36 (3), 287–309. Goetzmann, W.N., Li, L., Rouwenhorst, K.G., 2005. Long-term global market correlations. Journal of Business 78 (1), 1– 38. Grammig, J., Melvin, M., Schlag, C., 2005. International cross-listed stock prices during overlapping trading hours: price discovery and exchange rate effects. Journal of Empirical Finance 12, 139–164. Hamao, Y., Masulis, R.W., Ng, V., 1990. Correlations in price changes and volatility across international stock markets. Review of Financial Studies 3 (2), 281–307. Hirayama, K., Tsutsui, Y., 1998. Threshold effect in international linkage of stock prices. Japan and the World Economy 10, 441–453. Lane, P.R., Milesi-Ferretti, M.G., 2001. The external wealth of nations: measures of foreign assets and liabilities for industrial and developing countries. Journal of International Economics 55 (2), 263–294. Lane, P.R., Milesi-Ferretti, M.G., 2003. International financial integration. IMF Staff Papers 50 (special issue).

T.T.-L. Chong et al. / Japan and the World Economy 20 (2008) 601–621

621

Levy, H., Sarnat, M., 1970. International diversification of investment portfolios. American Economic Review 668–675. Longin, F., Solnik, B., 1995. Is the correlation in international equity returns constant: 1960–1990? Journal of International Money and Finance 14, 3–26. Longin, F., Solnik, B., 2001. Extreme correlation of international equity markets. Journal of Finance 61 (2), 649–675. Neely, C., Weller, P., 2000. Predictability in international asset returns: a reexamination. Journal of Financial and Quantitative Analysis 35 (4), 601–620. Pagan, A.R., Sossounov, K.A., 2003. A simple framework for analyzing bull and bear markets. Journal of Applied Econometrics 18, 23–46. Pascual, R., Pascual-Fuster, B., Climent, F. Cross-listing, price discovery and the informativeness of the trading process. Journal of Financial Markets, in press. Roll, R., 1989. Price volatility, international market links, and their implications for regulatory policies. Journal of Financial Services Research 3, 211–246. Solnik, B., 1974. Why not diversify internationally rather than domestically? Financial Analysts Journal 30, 48–54. Su, Q., Chong, T.T.L., 2007. Determining the contribution to price discovery for Chinese cross-listed stocks. Pacific-Basin Finance Journal 15, 140–153. The Dow Jones Guide to the World Stock Markets (1998). Prentice Hall, Englewood, NJ. Vagnes, E., Fatemi, A.M., Park, J., 1996. Seasonal patterns in Japanese ADR returns and the US stock market influence. Japan and the World Economy 8, 65–79. von Furstenberg, G.M., Jeon, B.N., 1989. International stock price movements: links and messages. Brookings Papers on Economic Activity 125–179.