An Influence of U.S. and U.K.

Stock Market Returns for Two Stock Markets: An Evidence Case by Singapore and Japans Stock Markets
Wann-Jyi Horng / Associate Professor
Department of Hospital and Health Care Administration Chia Nan University of Pharmacy & Science Tainan, Taiwan e-mail: hwj7902@mail.chna.edu.tw

Liu-Hsiang Hsu / Assistant Professor

Department of Business Administration Ling Tung University Taichung, Taiwan e-mail: lucian@mail.ltu.edu.tw

Hui-Hsin Hsu / Lecturer

Department of Business Administration Ling Tung University Taichung, Taiwan e-mail: hueishin@mail.ltu.edu.tw

Abstract- The empirical results show that the dynamic conditional correlation (DCC) and the bivariate IGARCH (1, 1) model is appropriate in evaluating the relationship of the Singapore and the Japans stock markets. The empirical result also indicates that the Singapore and the Japans stock markets is a positive relation. The average estimation value of correlation coefficient equals to 0.495, which implies that the two stock markets is synchronized influence. Besides, the empirical result also shows that the Singapores and Japans stock markets do not have the asymmetrical effect. The return volatility of the Singapore and Japans stock markets receives the influence of the U.S. and the U.K. return volatility rates. The square item of U.S. and U.K. stock market returns affects the variation risk of the Singapores stock market. And the square item of U.S. stock market returns also affects the variation risk of the Japans stock market. Keywords- DCC; bivariate IGARCH model; stock market return.

market, he/she will usually care about the international capital the motion situation, the international politics and the economical situation change, in particular, in the U.S. and the U.K. stock markets change. There is a close relationship for Singapore and Japan based on the trade and the circulation of capital with the U.S. and the U.K., but the U.S. and U.K. are also powerful global economical nations. Therefore, the relationship between the Singapore stock market and the Japanese stock market is worth further discussion. The purpose of the present paper is to examine the relations of the Singapore and the Japanese stock markets. This paper also further discusses the affect of the U.S. and the U.K. stock returns volatility for the Singapore and the Japanese stock market returns. The organization of this paper is as follows: Section II descibes the data characteristics; Section III introduces the asymmetric test of the DCC and the bivariate IGARCH(1, 1); Section IV presents the proposed model and the empirical results, and finally Section V summarizes the conclusions of this study. II. DATA CHARACTERISTICS A. Data Sources The data of this research included the U.S., the U.K., the Singapore and the Japanese stock price collected between January, 2000 and December, 2008. The source of the stock data was the Taiwan economic Journal (TEJ), a database in Taiwan. The U.S. stock price refers to the S&P500 index, the U.K. stock price refers to the FTSE100 index. The Singapore stock price refers to the Singapore Strait times stock price index, the Japanese stock price refers to Tokyo NK-225 index. During the process of data analysis, in case that there was no stock market price available on the side of the Singapore and

I. INTRODUCTION We know that Singapore is one of Asian four dragons, also Singapore economy of growth in 2006 is 7.9%, and the forecast value of the grow rate is 5.5-7.5% in the future. And Singapore is also the Asia main financial center, its foreign exchange market is the fourth big trading market in the world. Apparently, Singapore plays an important role in the global economical financial system. Besides, Singapore and Japan have a close relationship based on the trade and the circulation of capital. We also know that Japan is one of big eight industrial countries in the global economical financial system and also has been very influential in the global economy. For example, in year 2000, the turnover of Japans Tokyo stock market achieves to US$3,160 billion, which is only less than those of the New York and the London stock exchange. When the investor has an investment in the international stock

978-1-4244-5326-9/10/$26.00 2010 IEEE

the Japanese stock market or on the side of the U.S. and the U.K. stock market due to holidays, the identical time stock price data from one side was deleted. After this, the four variables samples are 2,046. B. Return Calculation and Statistics To compute the return of the Singapore stock market adopts the natural logarithm difference, rides 100 again. The return of the Japan stock market also adopts the natural logarithm difference, rides 100 again. The return of the U.S. stock market also adopts the natural logarithm difference, rides 100 again. The return of the U.K. stock market also adopts the natural logarithm difference, rides 100 again. Table I presents the four sequences kurtosis coefficients are all bigger than 3, which this result implies that the normal distribution test of Jarque-Bera [1] is not normal distribution. Therefore, the heavy tails distribution is used in this paper. C. Unit Root and Co-integration Tests This paper further uses the unit root test of KSS [2] to determine the stability of the time series data. The KSS [2] examination results is listed in Table II. It shows that the Singapore stock returns, the Japanese stock returns, the U.S. stock returns, and the U.K. stock returns do not have the unit root characteristic, this is, the four markets are stationary series data, under = 1% significance level. Using Johansens [3] co-integration test as illustrated in Table III at the significance level of 0.05 ( =5%) does not reveal of max statistic. This indicated that the Singapore stock market, the Japanese stock market, the U.S. stock market and the U.K. stock market do not have a co-integration relation. Therefore, we do not need to consider the model of error correction.
TABLE I. BASIC STATISTICS Statistics Mean S-D Kurtosis J-B N (p-value) Sample RSING -0.0140 1.3936 15.2231 13032.86 (0.0000) 2045
***

TABLE III. CO-INTEGRATION TEST (LAG=7)

H0
None At most 1 At most 2 At most 3

max
14.7153 12.0327 4.2029 1.6518

Critical value 30.8151 24.2520 17.1477 3.8415

The critical value is given under the level 5%.

Notes: The lag of VAR is selected by the AIC rule [4].

TABLE IV. ARCH EFFECT TEST SING Statistic (p-value) JAPAN Statistic (p-value) Engle LM test 218.9690
***

Tsay F test 4.1593 *** (0.0000) Tsay F test 18.8242 *** (0.0000)
***

(0.0000) Engle LM test 586.0470 *** (0.0000)

Notes

denotes significance at the level 1%.

TABLE V. ASYMMETRIC TEST OF THE DCC AND THE BIVARIATE-IGARCH(1, 1) JAPAN F statistic (p-value) SING F statistic (p-value)
*

Negative size bias test 0.0267 (0.8703) Negative size bias test 1.5937 (0.2069)
Notes: (1) p-value < denotes significance ( =1%, (2) denotes significance at the level 10%,
**

Joint test 1.9123 (0.1255) Joint test 1.0844 (0.3545)

=5%, =10%).

denotes significance at the level 5%.

RJAPAN -0.0351 1.6673 10.1427 4395.67 (0.0000) 2045

***

RUS -0.0222 1.4090 10.2847 4541.11 (0.0000) 2045

***

RUK -0.0188 1.3939 11.4872 6140.44 *** (0.0000) 2045

***

D. ARCH Effect Test Based on the formula (1) and (2) as below, we uses the methods of LM test [5] and F test [6] to test the conditionally heteroskedasticity phenomenon. In Table IV, the results of the ARCH effect test show that the two markets have the conditionally heteroskedasticity phenomenon exists. This result suggests that we can use the GARCH model to match and analyze it. III. ASYMMETRIC TEST OF THE BIVARIATE-GARCH(1, 1) MODEL The bivariate-IGARCH(1, 1) model with a DCC can be constructed in the next section. The asymmetric test methods are used in the following two methods as: negative size bias test and joint test [7]. By the negative size bias test and the joint test shows that the Singapore stock market does not have the asymmetrical effect and the Japanese stock market does not also have the asymmetrical effect in Table V. IV. PROPOSED MODEL AND EMPIRICAL RESULTS Based on the results of the asymmetric test, we follows the ideas of the papers of Engle [8] and Tse & Tusi [9] to discuss the Singapore and Japans stock markets. In the formula (3)

Notes: (1) J-B N is the normal distribution test of Jarque-Bera. (2) S-D is denoted the standard deviation (3) denotes significance at the level 1%.

TABLE II. UNIT ROOT TEST OF KSS FOR THE RETURN DATA KSS Statistic Critical value Significant level RSING -16.69 *** -2.82 RJAPAN -21.89 *** -2.22 RUS -27.43 *** -1.92 RUK -25.47 ***

=1%

=5%
Notes:
***

=10%
denotes significance at the level 1%.

and (4) as below, we also considers the factors of U.S. and U.K. stock markets. After model process selection, in this paper, we may use the bivariate GARCH model to construct the relationships of the Singapore and the Japans stock market returns, the GARCH(1, 1) model is illustrated as follows:

RSING t = 0 + 1 RSING t 1 + 2 RJAPAN t 1

+ 3 RUS t 1 + 4 RUK t 1 + a1,t ,
RJAPAN t = 0 + 1 RJAPAN t 1 + 2 RSING t 1
(1)

Japanese stock market return receives before 1 periods impact of the Japanese stock return ( 1 =-0.1078). But the Japanese stock return does not receive before 1 periods impact of the Singapores stock return. The Japanese stock return also receives before 1 periods impact of the U.S. stock return volatility rates ( 3 =0.4231). The Japanese stock return also receives before 1 periods impact of the U.K. stock return volatility rates ( 4 =0.1581). The stock return volatility rates of the U.S. and the U.K. are also truly influent the return of the Japanese stock market. On the other hand, the correlation coefficient average t =0.495) of the Singapore and the Japanese estimation value ( stock return volatility is significant. This result also shows the Singapore stock returns volatility is mutually synchronized influence. In additional, estimated value of the degree of freedom for the Students t distribution is 7.13, and is significant under the significance level of 0.01 ( = 1% ). This also demonstrates that this research data has the heavy tailed distribution. From the Table VI, the estimated coefficients of the conditional variance equation will produce the different variation risks in Singapore and Japanese stock markets. The square item (the lag is equals 2) of the U.S. and U.K. stock market returns also affects the variation risk of the Singapores stock market ( 11 =0.0452 and 12 =0.0242, respectively). The square item (the lag is equals 2) of the U.S. stock market returns also affects the variation risk of the Japans stock market ( 21 =0.0274). The square item (the lag is equals 2) of the U.K. stock market returns does not affect the variation risk of the Japans stock market. We have also the result of 11 + 11 + 11 + 12 = 1 in the Singapores stock market, and 21 + 21 + 21 + 22 = 1 in the Japans stock market. This result conforms the conditionally supposition of the IGARCH model. The empirical results show that the Singapore stock market return is the IGARCH model, the Japan stock market return is the IGARCH model. This result also demonstrates the DCC and the bivariate IGARCH (1, 1) model may catch the Singapore and the Japanese stock return volatilities process. The empirical result shows that the Singapore stock market has a fixed variation risk. The Japanese stock market has also a fixed variation risk. In Table VI, the Singapore and the Japans stock market returns have the different conditional variation risks. The variation risk of the Japan stock market is larger than the variation risk of Singapores stock market. Based on the paper of Engle [8], the explanatory ability of the DCC and the bivariate IGARCH(1, 1) model is better than the traditional model of the bivariate GARCH (1, 1). To test the inappropriateness of the DCC and the bivariate IGARCH(1, 1) model, the test method of Ljung & Box [11] is used to examine autocorrelation of the standard residual error. This model does not show an autocorrelation of the standard residual error. Therefore, the DCC and the bivariate IGARCH(1, 1) model are more appropriate.

+ 3 RUS t 1 + 4 RUK t 1 + a 2,t ,

(2)

h11,t = 10 + 11 a12,t 1 + 11 h11,t 1 + 11 RUS t2 2

+ 12 RUK t2 2 ,
2 2 h22,t = 20 + 21 a 2 ,t 1 + 21 h22, t 1 + 21 RUS t 2

(3)

+ 22 RUK t2 2 ,

(4) (5) (6) (7)

h12,t = t h11,t h22,t ,

t = exp(qt ) /(exp(q t ) + 1) ,
q t = 0 + 1 t 1 + 2 a1, t 1 a 2 ,t 1 / h11, t 1 h22 ,t 1 ,
with the white noise of

at = (a1, t , a2, t ) is obey the bivariate

Students t distribution, this is,

(8) a t ~ Tv (0, (v 2) H t / v ) , Among 0 = (0,0) and H t is the covariance matrix of at = ( a1,t , a2,t ) , and t is the DCC coefficient of a1,t and

a2,t . The maximum likelihood algorithm method of BHHH

[10] is used to estimate the models unknown parameters. The programs of RATS and EVIEWS are used in this paper. From the empirical results, we know that the Singapore and the Japanese stock return volatility may be constructed on the DCC and the bivariate IGARCH (1, 1) model. Its estimate result is stated in Table VI. The empirical results show that the Japanese stock return ( 2 =-0.0475) also affects the return of the Singapore stock market. The Singapores stock return also receives before 1 periods impact of the U.S. stock return volatility rates ( 3 =0.3523). The Singapores stock return does not receive before 1 periods impact of the U.K. stock return volatility rates. The Singapore stock return does not also receive before 1 periods impact of the Singapores stock return. The stock return volatility rates of the U.S. are truly influent the return of the Singapores stock market.

TABLE VI. PARAMETER ESTIMATION OF THE DCC AND THE BIVARIATE IGARCH(1, 1) MODEL Parameters Coefficient (p-value) Parameters Coefficient (p-value) Parameters Coefficient (p-value) Parameters Coefficient (p-value) Parameters

0
0.0534 *** (0.0057)

1
-0.0252 (0.3339)

2
-0.0475 *** (0.0068)

3
0.3523 *** (0.0000)

4
0.0210 (0.4106)

0
0.0192 (0.4285)

1
-0.1078 *** (0.0000)

2
0.0163 (0.5542)

3
0.4231 *** (0.0000)

4
0.1581 *** (0.0000)

threshold of U.S. and U.K. stock return volatility rates. From the empirical result also obtains that the dynamic conditional t =0.495) of the correlation coefficient estimation value ( Singapore and the Japanese stock return volatility is positive. The U.S. and the U.K. stock return volatility rates affects the stock returns of the Singapores and Japans stock markets. The variation risk of the Singapores stock markets returns is truly received the impact of the U.S. and U.K. stock return volatility rates. The variation risk of the Japans stock markets returns is truly received the impact of the U.S. stock return volatility rates. Based on the viewpoint of DCC, the explanation ability of the bivariate IGARCH (1, 1) is better than the traditional bivariate GARCH (1, 1) model. REFERENCES
[1] C. M. Jarque and A. K. Bera, A test of normality of observation and regression residuals, International Statistical Review, 55, 1987, pp.163-172. [2] G. Kapetanios, Y. Shin, and A. Snell, Testing for a unit root in the nonlinear STAR framework, Journal of Econometrics, 112(2), 2003, pp. 359-379. [3] S. Johansen, Estimation and hypothesis testing of cointegration vector in Gaussian vector autoregressive models, Econometrica, 52, 1991, pp.389-402. [4] H. Akaike, Information theory and an extension of the maximum likelihood principle, In 2nd. International Symposium on Information Theory, edited by B. N. Petrov and F. C. Budapest: Akademiai Kiado, 1973, pp. 267-281. [5] R. F. Engle, Autoregressive conditional heteroskedasticity with estimates of the variance of United Kingdom Inflation, Econometrica, 50, 1982, pp. 987-1007 [6] R. S. Tsay, Analysis of Financial Time Series. New York: John Wiley & Sons, Inc., 2004. [7] R. F. Engle, and V. K. Ng, Measuring and testing the impact of news on volatility, Journal of Finance, 48(5), 1993, pp. 1749-1777. [8] R. F. Engle, Dynamic conditional correlation- a simple class of multivariate GARCH models, Journal of Business and Economic Statistics, 20, 2002, pp. 339-350. [9] Y. K. Tse, and Albert K.C. Tsui, A multivariate GARCH model with time-varying correlations, Journal of Business & Economic Statistics, 20, 2002, pp. 351-362. [10] E. K. Berndt, B. H. Hall, R. E. Hall, and J. A. Hausman,.Estimation and inference in nonlinear structural models, Annals of Economic and Social Measurement, 4, 1974, pp. 653-665. [11] G.. M. Ljung, and G. E. P. Box, On a measure of lack of fit in time series models, Biometrika, 65(2), 1978, pp. 297-303.

10
0.0207 *** (0.0011)

11
0.1062 *** (0.0000)

11
0.8244 *** (0.0000)

11
0.0452 *** (0.0023)

12
0.0242
*

(0.0693)

20
0.0332
***

21
0.0888
***

21
0.8714
***

21
0.0274
*

22
0.0124 (0.3335)

(0.0002)

(0.0000)

(0.0000)

(0.0586)

0
(0.0000) min t 0.3402

1
4.0108 *** (0.0000) max t 0.8322

2
0.0282 *** (0.0005)

t
0.4953 *** (0.0000)

v
7.1332 *** (0.0000)

Coefficient -2.0183 *** (p-value) Parameters Coefficient

Notes(1) p-value < denote significance ( =1%, 10%,

**

=5%). (2) * denotes significance at the level

***

denotes significance at the level 5%,

denotes significance at the level 1%.

V. CONCLUSIONS The empirical results show that the Singapore stock market returns volatility does not have an asymmetric effect and the Japanese stock market returns volatility does not have the asymmetric effect. The Singapore and the Japanese stock market return volatility may construct in the DCC and the bivariate IGARCH (1, 1) model with a positive and negative