How is the Athens Stock Exchange Affected?

International Research Journal of Finance and Economics ISSN 1450-2887 Issue 7 (2007) © EuroJournals Publishing, Inc. 2007 http://www.eurojournals.com/finance.htm

Interdependence of Major World Stock Exchanges: How is the Athens Stock Exchange Affected? Michalis Glezakos Associate Professor Department of Statistics and Insurance Sciences, University of Piraeus 80 Karaoli & Dimitriou str., 185 34 Piraeus, Greece Email: [email protected] Anna Merika Associate Professor Department of Economics, Deree College 33-35 Marathonos, Panorama Voulas, 166 73, Athens, Greece Email: [email protected] Haralambos Kaligosfiris Department of Statistics and Insurance Sciences, University of Piraeus 80 Karaoli & Dimitriou str., 185 34 Piraues, Greece Abstract In the context of globalization, through a growing process of economic integration among countries and financial markets, the interdependency among major world financial markets is more than evident The paper covers the most recent period 2000-2006 using monthly data and investigates and examines the short and long-run relationships between major world financial markets with particular attention to the Greek stock exchange. The research methodology employed includes testing for stationarity, both with the DickeyFuller and the Phillips-Perron tests, the use of a VAR model for the implementation of the Granger Causality test, and Cointegration tests according to Johansen-Juselious. The results confirm the dominance of the USA financial market and the strong influence of DAX and FTSE on all other markets of the sample. The influence of Germany and the DJ index is especially noticeable on the Athens stock exchange. We conduct cointegration tests with variance decomposition and estimate the impulse response functions. Keywords: Market Interdependency, VAR models, Variance decomposition, Impulse response functions JEL Classification: G15, C32

1. Introduction The objective of this paper is to investigate the short and the long-term financial integration of major European markets as well as those of the U.S and Japan with the Greek stock market. The objective of an international investor is to minimize his/her portfolio risk at a given level of expected return. The modern portfolio theory suggests that low correlations between assets result in lower risk where return on the portfolio is measured by its mean and risk is measured by its standard

25

International Research Journal of Finance and Economics - Issue 7 (2007)

deviation. In this respect less than perfect positive correlation between assets provide opportunities for portfolio diversification. From the perspective of an international investor who is willing to make portfolio investments in the world’s advanced financial markets it is important to know if he can achieve diversification. Accordingly, correlation has been used as the main indicator of diversification within asset classes and on international basis within countries. Developed countries with higher economic and trade linkages amongst them, had higher correlation and as a result assets traded in their capital markets responded to certain common factors. The first part of this paper undertakes an extensive literature review in the field and formulates hypotheses which are to be tested in the empirical part. The second part discusses the available data and defines methodology. The next part of the paper shows estimation results and analyses the empirical findings. Finally conclusions are drawn and future research directions are suggested.

2. Literature Review In the context of globalization, through a growing process of economic integration among countries and financial markets, the interdependency among major world financial markets becomes more and more evident. The interdependency between financial markets has been at the focus of interest since the 1960´s decade. The majority of studies in that early period reach the conclusion that the degree of interdependency between markets is quite low, since the prime factors in the development of financial markets are of domestic nature. Even in those years some studies were published that supported the existence of limited interdependency between markets. Agmon (1972), establishes some degree of interdependency between the markets of the US, UK, Germany and Japan during 1961 until 1966. He accounts for it on the basis of the important economic relations that existed between the countries under investigation. Also, Ripley (1973) finds that there is interdependency but only between those countries that are open to foreign capital investments, in contrast with the isolated markets that do not show any interdependence with the other countries. So the limited nature of interdependency between markets in the 1970s decade was attributed to the legal and technical restrictions on the movements of capital among countries. Later on, when these restrictions were abolished, the degree of interdependency between markets increased. Furthermore, the desire of portfolio managers to reduce the systematic risk related to a specific market, and benefit through the diversification effect, pushed them to invest in more than one country. As a result the independency among countries reduced. During the decades of 1980 and 1990, the interest of the academic community for this and other related issues was prominent. A number of different econometric techniques were employed for the investigation of market interdependency, especially short-term effects among markets (Kugioumtzaki et.al, 2006). The development of cointegration theory from Engle and Granger {1987}provided the theoretical framework for the building up of models in the context of which both short and long term relations among international markets can be investigated. Most of the studies that followed showed that there is a considerable degree of interdependency among markets, which is going to increase through time. That is mainly due to the globalization process, the abolishment of any restrictions in capital movements as well as the improvement of telecommunications {Internet}. Table I, summarizes the main studies of market interdependency.


26

Table 1: Published Work on the Interdependency of World Stock Markets Study 1.Eun and Shim (1989) 2.Taylor and Tanks (1989)

3.Koch and Koch (1991) 4.Yan-Leung Cheung and SuiChoi Mak (1992) 5.Malliaris and Urrutia (1992)

Markets Αustralia, Canada,France, Germany, Hong-Kong, Japan, Switzerland, Britain, USA Britain,Germany, USA,Holland, Japan

Period 1980-1985

Μethodology VAR model Impulse responses

Results Market Interdependency exerts dominant influence

1973-1979 1979-1986

Cointegration and Granger test

Japan, Αustralia, Singapore, Hong- Κοng, Switzerland Germany,Britain, USA USA, Japan, Hong-Kong, Malaysia, Indonesia, Philippines, S.Korea, Taiwan, Τailand USA,Japan, Britain, HongKong, Singapore, Australia

1972,1980,1987

Dynamic System of Simultaneous Equations Cointegration test

Market cointegration between Britain, Germany, Japan and Holland after the abolishment of currency restrictions in Britain 1979 Interdependency between markets within a 24 hour period. Also the geographic proximity influences positively the interdependency The USA markets exert dominant influence in most of the cases under examination.

6.Kasa (1992)

USA, Britain, Japan, Canada

7.Arshanapalli and Doukas (1995)

Μάιος 1987 Μάρτιος 1988

Granger causality test

No Granger causality among markets before and after the crash of October 1987. The dominant role of USA is not confirmed. The stochastic trend behind the long-run movement of markets is more important in Japan and less important in Canada. Before 1987 crush there was no dependency among the European stock exchanges and that of the New York. After the crush there was. The dominant role of USA is also confirmed. The Japanese market does not relate with other markets. There is no interdependency among the 5 markets and as a result there is no long run relationship among them. There is cointegration during the second period under examination.

1974-1990

Error correction model

Britain,Germany, USA,France, Japan

1980-1990

Cointegration test

8.Bayers and Peel (1993)

USA, Britain, Germany, Japan and Holland

1979-1989

Cointegration test

9.Blackman, Holden and Thomas (1994) 10.Richards (1995)

17 Stock Markets

1970-1979 και 1984-1989

Cointegration test

1970-1994

Cointegration test

There is no interdependency among the markets under investigation

11.Hassan and Naka (1996)

Αustralia, Austria, Canada,France, Germany,Denmark, HongKong, Ιtaly, USA,Japan,Britain,Sweden, Switzerland, Holland, Νorway, Spain USA,Britain, Germany,Japan

1984-1991

Vector error correction model (VECM)

12.Koutmos (1996)

Britain,France, Germany,Italy

1986-1991

Multivariate VAREGARCH model Multivariate VAR-

There is an increasing interdependency among markets in the short as well as in the long run. The dominant role of the USA is established There is interdependency among European markets There is an asymmetry in the transmission mechanism of the error variance Independency of markets despite the trade relations among them

13.Both, Martikainen

and

Denmark, Sweden, Finland

Germany,

1978-1988

USA

Νorway,

1988-1994

27


Ken (1997)

EGARCH model

14.Choudhry (1997)

Αrzentina, Βrazil, Chile, Colombia, Μεxico, Venezuela, USA

1989-1993

Cointegration test

15.Elyasiasi, Perera and Puri (1998) 16.Moschos and Xanthakis (1998)

Sri Lanka, Taiwan, Singapore, Japan, S.Korea, Hong-Kong, Ιndia, USA Britain,USA, Greece

1989-1994

Multivariate VAR model

1990-1992

Autoregressive model

17.Janakiramanan and Lamba (1998)

Αustralia, Hong-Kong, Japan, New Ζeland, Singapore, USA, Indonesia, Malaysia, Τailand

1988-1996

VAR models

18.Huang, Yang and Hu (2000)

USA, Japan, China, South Growth Triangle (HongKong, Taiwan, South China)

1992-1997

Cointegration test, Granger causality test

19.Chen, Firth and Rui (2000)

Βrazil, Mexico, Chile, Αrzentina, Κοlombia, Venezuela

1995-2000

Cointegration test

t 20 .Östermark (2001) 21.Masih and Masih (2001)

Finland and Japan

1990-1993

USA, Britain, Japan, Germany, S.Korea, Singapore, Hong-Kong, Taiwan, Αustralia

1992-1994

Cointegration test Cointegration test

22.In, Kim, Yoon and Viney (2001)

Hong-Kong, Τailand

1997-1998

Κorea,

Multivariate VAREGARCH model

There is an asymmetry in the transmission mechanism of the error variance The markets are cointegrated with or without the presence of the USA which appears to exert dominant influence. The market of Sri Lanka is not influenced by any other market. The changes of S&P 500 of New York contribute to improved predictions in the movement of the Athens Stock Exchange The changes in the Athens Stock Index are attributed mainly on domestic factors. Countries which are geographically close with strong economic ties appear to be financially interdependent and highly integrated. The dominant role of the USA market is confirmed. There is no cointegration among the countries of the SCGT and also no long –r un relationship is found among the countries of the SCGT and Japan or the USA. In the shortrun the USA market leads the rest. There is cointegration among the markets under examination up to 1999. Since then though this longrun relationship breaks down. Cointegrated markets. There is interdependency among the Asian markets and the already developed countries of the OECD. The markets of the USA and Britain have a dominant role both in the short and the long-run. Cointegrated markets. Hong-Kong plays a dominant role

We observe that thirteen out of the twenty-two studies covered in this literature review find substantial evidence of interdependency among world financial markets both in the short and the long run. Seven out of these thirteen shows that the role of the USA market worldwide is dominant and the DJ index exerts a substantial influence to the majority of financial markets (Tserkezos & Mylonakis, 2006). Also, two of the studies establish that interdependency among markets increased since the crash of 1987, Malliaria & Urrutia (1992) and Arshanapalli & Dukas (1993). The present study covers the most recent period 2000-2006 using monthly data and investigates the following hypotheses: • Is there interdependency among financial markets and if there is what is the direction of causality (short-run effects) • Is there a long-run relationship among markets

International Research Journal of Finance and Economics - Issue 7 (2007) • •

28

Is the Athens stock market interdependent with major world financial markets(short and longterm effects) What does the variance decomposition show in the case of market inter dependency and also estimation of the impulse response functions to determine the time period required to restore equilibrium, after an external shock hits the market.

3. Data and Methodology 3.1 The Sample Table II shows the general stock indices of the countries which make up the sample of the present study. They were chosen for the following reasons. • The USA by general concession is the strongest financial market worldwide • The Japanese market is one of the strongest worldwide and the leader in the Asia region • England, France and Germany have the strongest European markets • The Italian and the Spanish markets are the most dynamic in the Mediterranean region, part of which is the Greek financial market • Belgium and Holland are two dynamic countries but rather of small size like Greece. Therefore, the sample, based on the above criteria, includes the strongest financial markets as well as those that have common characteristics with the Athens Stock Exchange. Table 2: Stock Exchanges and Stock Indices under Investigation Country USA ENGLAND FRANCE GERMANY ΙΤΑLY SPAIN HOLLAND BELGIUM GREECE JAPAN

Index DJ FTSE CAC DAX MILAN MADRID HOL BEL GEN NIKKEI

Dow Jones FTSE – 100 France Cac 40 Dax 30 Performance Milan Comit General Madrid se General Aex Index Bel 20 Athens General Nikkei 225

The research covers the time period between 18/12/2000 up to 9/3/2006 and in total we have at our disposal 1354 daily prices for each one of the indices presented in Table II. In the cases that a stock exchange remains closed, the price of the relevant index is calculated by simulation for the particular day. The data are provided by the data base of Finance Yahoo. The use of daily prices was chosen because the weekly or monthly intervals might be large enough so that do not allow the underpinning of interrelations that conclude within one or only a few days (Stivaktakis et.al., 2006). The timing of daily operations among the different stock exchanges do not coincide, therefore during the evaluation of our results we must consider the differences among the daily running of different stock exchanges. For example, the Athens Stock Exchange closes before the opening of the NYSE, so the New York Stock Exchange exerts influence on the Athens Stock Exchange next day as well. On the contrary, since the stock exchanges of London and Frankfurt do not differ substantially in their running hours, their interrelations occur within the day. 3.2 Methodology In the context of the methodology employed for the purposes of the current study we undertake: • Testing for stationarity, both with the Dickey- Fuller and the Phillips-Perron tests.

29

International Research Journal of Finance and Economics - Issue 7 (2007) • •

•

Using a VAR model for the implementation of the Granger Causality test, aims to identify interdependency between financial markets in the short-run. Cointegration tests as they were put forward by Johansen-Juselious, for pinpointing the longrun relationships among the financial markets under investigation. The purpose is to grasp the mechanism of influence exertion. In this context we use the Engle-Granger and Johansen tests as well as the error correction model –ECM, which allows disequilibrium correction and the attainment of long-run equilibrium. Variance Decomposition and estimating the Impulse Response Functions.

4. Estimation and Analysis of Empirical Results During the period under examination, most markets showed positive returns (with the exceptions of the French, British and the Dutch), the highest return came from the Spanish market. During the same period DAX showed the highest volatility and Dow Jones the lowest, while the Athens Stock Exchange moves within average levels of risk and return. The values of asymmetry and kurtosis shown on Table III suggest that the sample stock returns are not normally distributed, which is verified with the Jarque-Bera statistic. Table 3: Characteristics of Distributions of the Stock Indices Under Investigation Mean Median Maximum Minimum St. Deviation Skewness Kurtosis Jarque-Bera Probability

Bel 0.000264 0.000415 0.097835 -0.054558 0.011885 0.496099 10.50539 3228.767 0.000000

Cac -1.06E-05 7.49E-05 0.072533 -0.073907 0.014779 0.099050 6.413099 658.4515 0.000000

Dax 6.03E-05 0.000618 0.078452 -0.062931 0.016781 0.101412 5.456822 342.3441 0.000000

Dj 8.30E-05 0.000150 0.063481 -0.048144 0.011027 0.413149 6.680660 801.6249 0.000000

Ftse -1.43E-05 0.000419 0.060815 -0.054355 0.011473 -0.035681 6.763103 798.0199 0.000000

Gen 0.000162 4.51E-06 0.069781 -0.077424 0.011832 -0.102907 7.085884 942.8402 0.000000

Hol -0.000114 0.000311 0.099844 -0.072544 0.016106 0.244534 7.242931 1027.613 0.000000

Madrid 0.000342 0.000882 0.051467 -0.042460 0.011634 0.050700 4.999955 225.9025 0.000000

Milan 3.53E-05 0.000335 0.070705 -0.074224 0.011253 -0.262831 7.879171 1356.655 0.000000

Nikkei 0.000170 0.000334 0.059029 -0.066342 0.013746 -0.112675 4.388327 111.4405 0.000000

Table IV shows the return correlations among the various indices. We deduce the following: • Generally, the correlations among the returns of the countries under investigation are high, this is a first indication for the existence of interdependency among them • The highest of correlations is among Holland and France (over 92%) and the lowest among New York and Tokyo (about 12%).I t must also be pointed out that the Tokyo stock exchange shows low correlations with all other markets (up to 20%). • The European markets are quite highly correlated with the exception of Athens, which is not as tightly connected. • Table V also shows that the European, as well as, the Japanese markets is affected substantially by the developments in the New York market. This influence is one-way; the New York stock exchange is not affected by what is happening in the other markets. As far as the Athens stock exchange is concerned, does not seem to influence any other. On the contrary it is influenced by the markets of Germany and USA. The above points need to be verified by the Granger causality test and the cointegration test that will follow.


30

Table 4: Correlations of Returns of the Stock Indices Under Investigation Bel Cac Dax Dj Ftse Gen Hol Madrid Milan Nikkei

Bel 1 0.830150 0.726982 0.457317 0.759172 0.375032 0.861638 0.734090 0.744976 0.202653

Cac 0.830150 1 0.854066 0.495664 0.858851 0.394791 0.926560 0.861904 0.878620 0.237165

Dax 0.726982 0.854066 1 0.600458 0.752124 0.398430 0.826573 0.779085 0.804706 0.207081

Dj 0.457317 0.495664 0.600458 1 0.451907 0.235376 0.479693 0.464006 0.474455 0.127431

Ftse 0.759172 0.858851 0.752124 0.451907 1 0.383141 0.850954 0.767371 0.780987 0.229307

Gen 0.375032 0.394791 0.398430 0.235376 0.383141 1 0.426616 0.384265 0.370978 0.224747

Hol 0.861638 0.926560 0.826573 0.479693 0.850954 0.426616 1 0.823185 0.829365 0.241106

Madrid 0.734090 0.861904 0.779085 0.464006 0.767371 0.384265 0.823185 1 0.817792 0.231388

Milan 0.744976 0.878620 0.804706 0.474455 0.780987 0.370978 0.829365 0.817792 1 0.211681

Nikkei 0.202653 0.237165 0.207081 0.127431 0.229307 0.224747 0.241106 0.231388 0.211681 1

Table 5: Correlation of Returns between Periods t and t+1 Bel Cac Dax Dj Ftse Gen Hol Madrid Milan Nikkei

Bel(+1) 0,0702 0,1428 0,2821 0,0597 0,0576 0,0752 0,0598 0,0767 0,0098

Cac(+1) 0,0193 0,0770 0,2684 - 0,0048 0,0487 - 0,0104 - 0,0104 - 0,0110 0,0040

Dax(+1) - 0,0188 - 0,0448 0,1325 -0,0389 0,0040 - 0,0450 - 0,0274 - 0,0418 0,0079

Dj(+1) -0,0131 -0,0214 0,0022 -0,0157 -0,0187 -0,0337 -0,0104 -0,0120 -0,0199

Ftse(+1) 0,0301 -0,0009 0,0739 0,2744 -0,0003 -0,0214 -0,0040 -0,0125 0,0004

Gen(+1) 0,0809 0,0645 0,1143 0,2301 0,0359 0,0534 0,0621 0,0865 -0,0203

Hol(+1) 0,0397 0,0137 0,0956 0,2870 -0,0057 0,0099 0,0036 0,0140 0,0018

Madrid(+1) 0,0131 - 0,0114 0,0655 0,2235 - 0,0030 0,0333 - 0,0015 - 0,0026 - 0,0202

Milan(+1 Nikkei(+1) 0,0503 0,1957 0,0259 0,2823 0,0731 0,3124 0,2019 0,3179 0,0255 0,2362 0,0710 0,1203 0,0266 0,2482 0,0248 0,2555 0,2882 0,03 -

It appears from Tables VI and VII, which show the results of the Dickey-Fuller (ADF test), that the time-series are not stationary. On the contrary, the first differences of the logarithmic transformations of the series are stationary. So, we say that the series are integrated of order one I(1) or they contain one unit root. Therefore, all prerequisites are present for implementing cointegration tests. Table 6: ADF Test on the Levels of the Series (lnPt) lnPt levels Belgium (lnbel) France (lncac) Germany (lndax) USA (lndj) Britain (lnftse) Greece (lngen) Holland (lnhol) Spain(lnmadrid) Italy (lnmilano) Japan (lnnikkei)

Lag length p 1 0 0 0 0 1 0 0 0 0

ADF statistic - 1,2227 - 0,3420 - 0,22028 - 2,13262 - 0,28781 - 1,51679 - 2,08331 - 1,36013 - 1,61630 - 1,76237

p-value 0,9047 0,5620 0,6070 0,2319 0,5823 0,8237 0,2516 0,8720 0,7864 0,7225

LM test X2 statistic 0,1169 0,3089 1,43212 2,37079 2,92836 0,00473 0,04483 0,32534 0,17528 0,00516

p-value 0,7323 0,5783 0,2316 0,1238 0,0872 0,8278 0,8323 0,5685 0,6755 0,9427

Deterministic terms Trend & intercept None None Intercept None Trend & intercept Intercept Trend & intercept Trend & intercept Trend & intercept

31


Table 7: ADF Test on the First Differences of the Series (lnPt) lnPt first differences Belgium Dlnbel France Dlncac Germany Dlndax USA Dlndj Britain Dlnftse Greece Dlngen Holland Dlnhol Spain Dlnmadri Italy Dlnmilano Japan Dlnnikkei

Lag length p 0 0 0 0 0 0 0 1 0 0

ADF statistic - 32,6421 - 37.4012 - 38,0509 - 38,4511 - 38,3619 - 33,7612 - 36,9762 - 26,4226 - 63,3735 - 36,7089

p-value 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

LM test X2 statistic 0,14630 1,01777 2,35521 1,27091 0,00351 0,04025 2,45483 0,66545 3,62447 1,07989

p-value 0,7021 0,3132 0,1251 0,2597 0,9527 0,8410 0,1173 0,4147 0,0571 0,2989

Deterministic terms None Trend & intercept Trend & intercept None None Trend & intercept None None Trend & intercept Trend & intercept

The critical values from MacKinnon (1996) for rejection of Η0 1% level - 3,964852 - 3,434980 - 2,566675

Trend & intercept Intercept None

5% level - 3,413140 - 2,863472 - 1,941058

10% level - 3,128583 - 2,567848 - 1,616542

4.1 Phillips-Perron Test The Phillips-Perron test is less restrictive and provides an alternative way for checking the stationarity of a time-series. Tables 8 and 9 that follow present the results for the variables in levels, having taken their logarithmic transformation, and the first differences of their logarithmic transformations. We make the same conclusions that we did with the Dickey-Fuller tests. The series are I(1) so we proceed with coitegration tests. Table 8: Phillips-Perron test on the Levels of the Series (lnPt) lnPt levels Belgium (lnbel) France (lncac) Germany (lndax) USA (lndj) Britain (lnftse) Greece (lngen) Holland (lnhol) Spain(lnmadrid) Italy (lnmilano) Japan (lnnikkei)

Bandwidth 20 23 9 6 13 9 17 1 7 5

PP test statistic - 0,974519 - 0,384666 - 0,227267 - 2,029892 - 0,324389 - 1,640686 - 2,056956 - 1,348277 - 1,644775 - 1,714128

p-value 0,9456 0,5457 0,6044 0,2740 0,5687 0,7765 0,2626 0,8751 0,7748 0,7448

Deterministic terms Trend & intercept None None Intercept None Trend & intercept Intercept Trend & intercept Trend & intercept Trend & intercept

Table 9: Phillips-Perron test on the First Differences of the Series (lnPt) lnPt first differences Belgium Dlnbel France Dlncac Germany Dlndax USA Dlndj Britain Dlnftse Greece Dlngen Holland Dlnhol Spain Dlnmadrid Italy Dlnmilano Japan Dlnnikkei

Bandwidth 22 25 11 7 12 7 18 1 6 3

PP test statistic - 32,41849 - 38,12335 - 38,11547 - 38,50221 - 38,67738 - 33,91756 - 37,05449 - 37,29343 - 36,38112 - 36,71778

p-value 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000 0,000

Deterministic terms None Trend & intercept Trend & intercept None None Trend & intercept None None Trend & intercept Trend & intercept


32

4.2 Granger Causality Test The Granger Causality test is based on Var(p) where p is the no of lags. The objective is to pinpoint the causality direction in the interdependent relationship among financial markets. The estimation of a Var(p) model though assumes determination of p. Given that there is no theoretical solution to this problem, we assumed various specifications with different p which we checked with the criteria Schwarz(SC) and Hannan-Quinn(HQ).The best application proved to be the one with Var(p) where p=1. We made use of the first differences of the logarithms of prices because they are stationary as shown by the Dickey-Fuller and Phillips-Perron tests. Initially we looked at the causality among the financial markets of Athens (GEN) and New York (NYSE), which is the top financial market worldwide. So, we estimated the two VAR (1) models that follow. D( LNGEN ) t = c1 + a1 D( LNGEN ) t −1 + b1 D( LNDJ ) t −1 + u1t D( LNDJ ) t = c 2 + γ 1 D( LNDJ ) t −1 + δ 1 D( LNGEN ) t −1 + u 2t The coefficients α1, b1, γ1 and δ1 are presented on Table X: Table 10: Estimation of the Coefficients Sample(adjusted): 3 1353 Included observations: 1351 after adjusting endpoints Standard errors in () & t-statistics in [] D(LNGEN) 0.037614 D(LNGEN(-1)) (0.02726) [1.37958] 0.238956 D(LNDJ(-1)) (0.02930) [8.15524] 9.76E-05 C (0.00031) [0.31138]

D(LNDJ) -0.006680 (0.02606) [-0.25640] -0.043660 (0.02800) [-1.55924] 2.82E-05 (0.00030) [0.09426]

Testing for Granger causality from the DJ to GEN we writeH0:b1=0, this is rejected so we deduce that there is Granger causality from DJ towards GEN. The repetition of this hypothesis testing for GEN influencing DJ now, leads to non-rejection of the null hypothesis, so we conclude that there is no such influence. On Table XI (Appendix 1) we present the results of Granger causality tests for all possible pairs of the financial stock indices. From the results of Table XI we deduce that: 1. The New York financial market plays the most important role in the world financial scene, since the stock index DJ Granger causes all other stock indices of the financial markets under examination. 2. The London financial market does not cause any of the European financial markets. On the contrary, it is Granger caused by the markets of Germany, New York, France and Belgium. 3. With reference to the Greek capital market, we could say that it is Granger caused by the markets of Germany, Belgium, New York and Italy. This is explained by the size of these markets as well as the operation of electronic media centers like Reuters, Bloomberg etc. 4. The German financial markets have a dominant role in Europe, since it appears to Granger cause all European markets in the sample. Also, it Granger causes the Japanese market, Nikkei 225.

33


4.3 Cointegration Tests It is important to note that, the Granger causality test cannot be used to identify the extent of the influence. Furthermore, the tests developed on the basis of VAR models, are not very reliable due to the absence of a methodology that determines the lag structure of the VAR models used. Therefore, we apply as a supportive or supplementary test the cointegration test through the specification of Error Correction Models. The results from this test allow us to detect more accurately the existence of interrelationships among the stock indices on a long-run basis. Let us note, that cointegration presupposes the implementation of the Engle-Granger procedure on all the stock pairs of the sample. The purpose is to identify cointegration and long-run equilibrium relations. Wherever these exist, we use an error correction model to depict the short-run relation among them and then through this model we perform the Granger causality test which takes into account the short as well as long-term relationships between the markets. Table XII (Appendix 2) presents the results of the analysis and more specifically, the cointegration test among pairs of markets and the Granger Causality test. The results of the Table confirm the dominant role of the New York stock exchange in the world financial arena. This market influences all other, while none of the others can influence it. It appears that the New York market can exert substantial influence and shape the returns of major world financial markets. In the context of the European Union, the German stock exchange seems to play an important role, since it influences all other markets and is not influenced in any way by them. So it assumes dominant role and overpowers the traditional leader the London stock exchange. The London stock exchange is being influenced by the market of France, Holland etc. The Greek stock exchange is being influenced by changes in the stock markets of USA and Germany and to a lesser extent by Spain and Belgium. It is rather impressive that no influence is exerted by the London market on the Athens Stock Exchange. 4.4 Variance Decomposition The variance decomposition of the stock indices under investigation is based on the analysis of responses of the variables to shocks. According to this analysis, when there is a shock through the error term we study the influence of this shock to the other variables of the system. So, we collect information for the percentage of the error variance in predicting the prices of an index which is interpreted by some other variable of the system. Also we collect information, about the time horizon within which the whole influence is completed. This way we deduce information for the importance of the influence of each variable, on all other variables of the VAR model. It must be mentioned that the variance decomposition is affected by the ordering of the variables. In this study the stock indices are ordered by descending order of their respective markets. The results of variance decomposition for all stock indices of the sample can be summarized as follows. • A shock in the New York market accounts for about 11% of the variance of the GEN in the Athens stock exchange. The respective percentage for the Frankfurt stock exchange is 8% and the London stock exchange is about 1%. • The German stock exchange is influenced only by the DJ and to a relatively large extent, since a shock initiated in the New York market accounts for 40% of the total variation in DAX. • The London market dependency on the New York stock exchange is also quite considerable (25% up to 30% of the total variation) in contrast with the Tokyo stock exchange where the dependency rate is 10%-12%. • The German market influences to a great extent the French market (about 40%) which is also substantially affected by the New York market (about 30%-33%). • The Milan stock exchange is shaped by about 60% from the markets of the USA and Germany, about the same percentage with the Madrid market. Influences among the remaining markets, despite the fact that they are numerous are negligible.


34

4.5 Impulse Response Analysis The impulse response analysis investigates the influence on the markets of the sample that a random shock might have. A change in the market I for example, will not only affect market I, but it will transmit to other markets, even with a time lag. So, the impulse response functions show the effect that a random (unpredicted) shock which occurs through one of the errors (innovations), has on the current and future index prices. The coefficients determining the size of the response of each market to shocks in individual markets (for each period) are derived from the coefficients of the VAR model which we can only estimate. So, these response coefficients are also liable to statistical error. To evaluate the response of each market to an unforeseen change in others, it is useful to estimate confidence intervals. From our analysis, it appears that a shock in the DJ index, equal to one standard deviation in size, influences the prediction error of the Athens stock exchange to a substantial extent and the influence s completed after nine days. Nevertheless, these influences are liable to a large error and are statistically significant therefore only the first day after the shock. Therefore, the influence of a shock which occurs in the New York market, on the Athens General Stock Index, is completed within a day and it is especially important since it evidently upsets the Greek market. Also, domestic market shocks cause reactions of the Athens stock exchange, which last for seven days, but after the first day die gradually out, a fact which confirms the theory of finance, when it states that any event of economic nature is immediately reflected in the stock exchange. Investigating the interdependency of the Greek market with the others (apart from NYSE) we observe that any events occurring in them, cause only small and statistically insignificant reactions in the Greek stock exchange. Finally, concerning the influences occurring among the remaining markets we must note the following: • It appears that the New York market is shaped solely by the influence of domestic factors (USA) given that a shock originating domestically influences DJ substantially but only for the first day after it is announced. The other markets do not appear to influence DJ, since any influences are small and statistically insignificant. • The DAX index responds instantaneously to changes in domestic factors which complete their influence also within a day. From the foreign markets only the USA influences substantially the German market since all the other markets exert negligible and statistically insignificant influences. • For the London market, it appears that it is considerably influenced by the New York stock exchange as well as the DAX. These influences are absorbed within the day by the London stock exchange, so they last for as long as the domestic changes. • The Japanese market reacts sharply to shocks originating in the New York market. The influences last and they are statistically significant for a two day period, in contrast with previous cases where influences lasted for a day at most. • Finally, all the other markets in the sample are heavily influenced by shocks in the DJ, DAX and FTSE indices.

5. Conclusions This study offers a continuation of the research on the issue of financial markets growing interdependency, in the context of globalization. The empirical results reveal that both the long-run cointegrating relationships and the short-run dynamic linkages among major world financial markets are strengthened through time. The US global influence is noticeable on all major world financial markets. It also responds significantly to primarily domestic shocks. Furthermore, our findings suggest that the Athens stock market is strongly affected by the US and the German markets but the influence as the estimation of the impulse response function suggested is completed within a day.

35


If the results of this study are extended and contrasted with previous studies, they clearly suggest that financial market integration is time-varying. Further research might be taken towards this direction.

References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21]

Arshanapalli B, Doukas J, Lang L. 1995. “Pre- and post-October 1987 stock market linkages between U.S. and Asian markets”. Pacific-Basin Finance Journal 3: 57-73 Bekaert, G, Harvey CR. 1995. “Time-varying world market integration”. Journal of Finance 50: 403-444 Bessler DA, Yang, J. 2003. “The structure of interdependence in international stock markets”. Journal of International Money and Finance 22: 261-287 Chen G, Firth M, Rui O. 2002. “Stock market linkages: evidence from Latin America” Journal of Banking and Finance 26: 1113-1141 Cheung YW, Lai KS, Bergman M. 2004. “Dissecting the PPP puzzle: the unconventional roles of nominal exchange rate and price adjustment”. Journal of International Economics 64: 135150 Choudhry T. 1997. “Stochastic trends in stock prices: Evidence from Latin American Markets”. Journal of Macroeconomics 19: 285-304 Eun C, Shim S. 1989. “International transmission of stock market movements”. Journal Financial and Quantitative Analysis 24: 241-256 Froot KA, Scharfstein DS, Stein JC. 1992. “Heard on the Street: Informational inefficiencies in a market with short-term speculation”. Journal of Finance 47: 1461-1484 Gelos R, Sahay R. 2001. “Financial market spillovers in transition economies”. Economics of Transitions 9: 53-86 Gonzalo J. 1994. “Five alternative methods of estimating long-run equilibrium Relationships”. Journal of Econometrics 60: 203-233 Griffin JM, Nardari F, Stulz R. 2004. “Stock market trading and market conditions”. NBER Working Paper 10719, 1-48 Jochum C., Kirchgassner G, Platek M. 1999. “A long-run relationship between Eastern stock markets? Cointegration and the 1997-1998 crisis in emerging markets”. Weltwirtschaftliches Archiv. 135: 454-479 Johansen S. 1991. “Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models”. Econometrica 59: 1551-1580 Kasa K. 1992. “Common stochastic trends in international stock markets”. Journal of Monetary Economics 29: 95-124 King M, Sentana E, Wadhwani S. 1994. “Volatility and links between national stock markets”. Econometrica 62: 901-933 King M, Wadhwani S. 1990. “Transmission of volatility between stock markets”. Review of Financial Studies 3: 5-33 Koch PD, Koch TW. 1991. “Evolution in dynamic linkages across daily national stock indexes”. Journal of International Money and Finance 10: 231-251 Koop G. Pesaran MH, Potter SM. 1996. “Impulse response analysis in nonlinear multivariate models”. Journal of Econometrics 74: 119-147 Kugioumtzaki O., Tserkezos D., Mylonakis J. 2006. “Portfolio Analysis Strategies for Investing in World Stocks Markets”. Business Journal for Entrepreneurs, Volume 2006, Issue 3, pp. 109-120 Lee T, Tse Y. 1996. “Cointegration tests with conditional heteroscedasticity”. Journal of Econometrics 73: 401-410 Lin W, Engle RF, Ito T. 1994. “Do bulls and bears move across boarders? International transmission of stock returns and volatility”. Review of Financial Studies 7: 507-538

International Research Journal of Finance and Economics - Issue 7 (2007) [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

36

Longin F, Solnik B. 1995. “Is the correlation in international equity returns constant: 19601990?” Journal of International Money and Finance 14: 3-26 Longin F, Solnik B. 2001. “Extreme correlation of international equity markets”. Journal of Finance 56: 649-676 Masih AMM, Masih R. 1997. “Dynamic linkages and the propagation mechanism driving major international stock markets”. Quarterly Review of Economics and Finance 37:859-885 Manning N. 2002. “Common trends and convergence? South East Asian equity markets 19881999”. Journal of International Money and Finance 21: 183-202 Pesaran MH, Shin Y. 1996. “Cointegration and speed of convergence to equilibrium”. Journal of Econometrics 71: 117-43 Pesaran MH, Shin Y, Smith RJ. 2000. “Structural analysis of vector error correction models with exogenous I(1) variables”. Journal of Econometrics 97: 293-343 Pesaran MH, ShinY. 1998. “Generalized impulse response analysis in linear multivariate models”. Economics Letters 58: 17-29 Rockinger M, Urga G. 2000. “The evolution of stock markets in transition economies”. Journal of Comparative Economics 28: 456-472 Stivaktakis I., Tserkezos D., Mylonakis J. 2006. “Aggregation Over Time and the Hsiao´s Optimizing Bivariate Procedure: Some Empirical and Monte Carlo Results. European Journal of Scientific Research, Vol. 13, No1, pp.113-118 Tserkezos D., Mylonakis J. 2006. “Temporal Aggregation Effects in Determining the Election Cycle in Stock Returns - Some Monte Carlo Results”. International Research Journal of Finance and Economics, Issue 5, September, pp.172-177 Tuluca SA, Zwick B. 2001. “The effects of the Asian crisis on global equity markets”. Financial Review 36: 125-142 Van Rijckeghem C, Weder B. 2001. “Sources of contagion: Is it finance or trade?” Journal of International Economics 54: 293-308

37


Appendix 1 Table 11: Pair wise Granger causality tests Sample: 1 1354 Lags: 1 Null Hypothesis: CAC does not Granger Cause BEL BEL does not Granger Cause CAC DAX does not Granger Cause BEL BEL does not Granger Cause DAX GEN does not Granger Cause BEL BEL does not Granger Cause GEN FTSE does not Granger Cause BEL BEL does not Granger Cause FTSE DJ does not Granger Cause BEL BEL does not Granger Cause DJ HOL does not Granger Cause BEL BEL does not Granger Cause HOL MADRID does not Granger Cause BEL BEL does not Granger Cause MADRID MILANO does not Granger Cause BEL BEL does not Granger Cause MILANO NIKKEI does not Granger Cause BEL BEL does not Granger Cause NIKKEI DAX does not Granger Cause CAC CAC does not Granger Cause DAX GEN does not Granger Cause CAC CAC does not Granger Cause GEN FTSE does not Granger Cause CAC CAC does not Granger Cause FTSE DJ does not Granger Cause CAC CAC does not Granger Cause DJ HOL does not Granger Cause CAC CAC does not Granger Cause HOL MADRID does not Granger Cause CAC CAC does not Granger Cause MADRID MILANO does not Granger Cause CAC CAC does not Granger Cause MILANO NIKKEI does not Granger Cause CAC CAC does not Granger Cause NIKKEI GEN does not Granger Cause DAX DAX does not Granger Cause GEN FTSE does not Granger Cause DAX DAX does not Granger Cause FTSE DJ does not Granger Cause DAX DAX does not Granger Cause DJ HOL does not Granger Cause DAX DAX does not Granger Cause HOL MADRID does not Granger Cause DAX DAX does not Granger Cause MADRID MILANO does not Granger Cause DAX DAX does not Granger Cause MILANO NIKKEI does not Granger Cause DAX DAX does not Granger Cause NIKKEI FTSE does not Granger Cause GEN GEN does not Granger Cause FTSE DJ does not Granger Cause GEN GEN does not Granger Cause DJ HOL does not Granger Cause GEN

Obs 1351 4.15486 1351 0.11294 1351 4.01387 1351 14.0545 1351 0.13512 1351 10.8354 1351 1.64601 1351 4.84165 1351 56.7537 1351 1.20594 1351 1.62996 1351 8.34226 1351 0.00664 1351 4.08523 1351 0.01162 1351 1.07869 1351 125.031 1351 10.4506 1351 39.2382 1351 1.97741 1351 46.1553 1351 21.0496 1351 15.0038 1351 153.076 1351 0.55994 1351 0.06574 1351

F-Statistic 3.14729 0.04171* 10.0885 0.73687 0.36801 0.04533* 2.71331 0.00019* 98.0642 0.71324 3.35500 0.00102* 1.84922 0.19972 0.32554 0.02795* 0.24312 9.0E-14* 42.2049 0.27233 5.01910 0.20193 0.35421 0.00394* 153.141 0.93505 0.08618 0.04346* 0.02909 0.91417 0.03582 0.29918 0.11011 0.00000* 0.52558 0.00126* 0.53870 5.0E-10* 52.2420 0.15989 1.19236 1.6E-11* 0.00118 4.9E-06* 0.74689 0.00011* 0.32684 0.00000* 0.01086 0.45441 66.5079 0.79768 0.47762

Probability 0.07628 0.00153* 0.54419 0.09975 0.00000* 0.06722 0.17410 0.56839 0.62204 1.2E-10* 0.02523* 0.55184 0.00000* 0.76913 0.86460 0.84992 0.74007 0.46860 0.46310 8.2E-13* 0.27505 0.97262 0.38762 0.56762 0.91702 7.9E-16* 0.48962

International Research Journal of Finance and Economics - Issue 7 (2007) GEN does not Granger Cause HOL MADRID does not Granger Cause GEN GEN does not Granger Cause MADRID MILANO does not Granger Cause GEN GEN does not Granger Cause MILANO NIKKEI does not Granger Cause GEN GEN does not Granger Cause NIKKEI DJ does not Granger Cause FTSE FTSE does not Granger Cause DJ HOL does not Granger Cause FTSE FTSE does not Granger Cause HOL MADRID does not Granger Cause FTSE FTSE does not Granger Cause MADRID MILANO does not Granger Cause FTSE FTSE does not Granger Cause MILANO NIKKEI does not Granger Cause FTSE FTSE does not Granger Cause NIKKEI HOL does not Granger Cause DJ DJ does not Granger Cause HOL MADRID does not Granger Cause DJ DJ does not Granger Cause MADRID MILANO does not Granger Cause DJ DJ does not Granger Cause MILANO NIKKEI does not Granger Cause DJ DJ does not Granger Cause NIKKEI MADRID does not Granger Cause HOL HOL does not Granger Cause MADRID MILANO does not Granger Cause HOL HOL does not Granger Cause MILANO NIKKEI does not Granger Cause HOL HOL does not Granger Cause NIKKEI MILANO does not Granger Cause MADRID MADRID does not Granger Cause MILANO NIKKEI does not Granger Cause MADRID MADRID does not Granger Cause NIKKEI NIKKEI does not Granger Cause MILANO MILANO does not Granger Cause NIKKEI (*)Rejection of the null hypothesis at5% and therefore there is Granger causality

0.30134 1351 2.50544 1351 7.13413 1351 20.6314 1351 0.06252 1351 0.00203 1351 0.24429 1351 0.75620 1351 84.9956 1351 167.383 1351 98.6404 1351 71.1619 1351 154.617 1351 0.50949 1351 0.93721 1351 94.7170 1351 0.76623 1351 99.5606 1351 129.875

38 0.58314 1.30918 0.11369 5.01514 0.00765* 2.27953 6.1E-06* 168.195 0.80260 1.71152 0.96407 3.54478 0.62121 2.17332 0.38467 0.18452 0.00000* 0.19592 0.00000* 0.21714 0.00000* 0.20097 0.00000* 0.26771 0.00000* 0.44939 0.47548 1.85549 0.33317 0.01764 0.00000* 0.40541 0.38154 0.36650 0.00000* 1.09787 0.00000*

0.25275 0.02529* 0.13133 0.00000* 0.19101 0.05995 0.14066 0.66758 0.65811 0.64130 0.65401 0.60496 0.50274 0.17337 0.89437 0.52442 0.54502 0.29492

39


Appendix 2 Table 12: Cointegration Tests Markets GEN, DJ GEN, FTSE GEN, DAX GEN, CAC GEN, MADRID GEN, MILANO GEN, HOL GEN, BEL DJ, FTSE DJ, DAX DJ, NIKKEI DJ, MADRID DJ, MILANO DJ, HOL DJ, BEL DJ, CAC DAX, FTSE DAX, CAC DAX, MADRID DAX, BEL DAX, MILNO DAX. HOL FTSE, NIKKEI FTSE, BEL FTSE, MILANO FTSE, HOL MADRID, BEL NIKKEI, HOL MADRID, NIKKEI MADRID, HOL MADRID, CAC BEL, MILANO BEL, CAC NIKKEI, BEL MILANO, HOL MILANO, CAC MILANO, NIKKEI CAC, HOL CAC, NIKKEI

Cointegration Test DJ is cointegrated with GEN FTSE is not cointegrated with GEN DAX is cointegrated with GEN CAC is not cointegrated with GEN MADRID is cointegrated with GEN MILANO is not cointegrated with GEN Are not cointegrated BEL is cointegrated with GEN DJ is cointegrated with FTSE DJ is cointegrated with DAX DJ is cointegrated with NIKKEI Are not cointegrated DJ is cointegrated with MILANO DJ is cointegrated with HOL DJ is cointegrated with BEL DJ is cointegrated with CAC DAX is cointegrated with FTSE DAX is cointegrated with CAC DAX is cointegrated with MADRID DAX is cointegrated with BEL Are not cointegrated DAX is cointegrated with HOL FTSE is cointegrated with NIKKEI FTSE is not cointegrated with BEL FTSE is not cointegrated with MILANO FTSE is cointegrated with HOL MADRID is cointegrated with BEL Are not cointegrated MADRID is cointegrated with ΝΙΚΚΕΙ MADRID is cointegrated with HOL MADRID is cointegrated with CAC BEL is cointegrated with MILANO Are not cointegrated BEL is cointegrated with NIKKEI Are not cointegrated MILANO is cointegrated with CAC MILANO is cointegrated with NIKKEI Are not cointegrated CAC is cointegrated with NIKKEI

GEN is not cointegrated with DJ GEN is cointegrated with FTSE GEN is cointegrated with DAX GEN is cointegrated with CAC GEN is not cointegrated with MADRID GEN is cointegrated with MILANO GEN is not cointegrated with BEL FTSE is not cointegrated with DJ DAX is not cointegrated with DJ NIKKEI is not cointegrated DJ MILANO is not cointegrated with DJ HOL is not cointegrated with DJ BEL is not cointegrated with DJ CAC is not cointegrated with DJ FTSE is not cointegrated with DAX CAC is not cointegrated with DAX MADRID is not cointegrated with DAX BEL is cointegrated with DAX HOL is not cointegrated with DAX NIKKEI is cointegrated with FTSE BEL is cointegrated with FTSE MILANO is cointegrated with FTSE HOL is not cointegrated with FTSE BEL is not cointegrated with MADRID ΝΙΚΚΕΙ is not cointegrated with MADRID HOL is cointegrated with MADRID CAC is not cointegrated with MADRID MILANO is not cointegrated with BEL NIKKEI is cointegrated with BEL CAC is not cointegrated with MILANO NIKKEI is not cointegrated with MILANO NIKΚEI is not cointegrated with CAC