Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Investigating Financial Contagion as a Prelude to its Mitigation: Australia and its Key Trading Partners Sandra Mukulu*, Samanthala Hettihewa*† and Christopher S. Wright*** The incidence and virulence of global and regional financial crises has encouraged analysts to re-examine extant economic policies and tools. An issue of concern is how beneficial financial links can turn malignant and transmit a financial crisis across borders. This study explores financialcontagion processes in financial markets with an emphasis on Australia. Monthly data from Jul/97-Jun/14 was obtained from Yahoo Finance for the composite-stock-market indices of Australia and its key trading partners. Eviews7 software was used to explore the relationship between stock indices of those countries via co-integration and pair-wise Granger Causality (GC) tests. This study found that stock returns in Australia are co-integrated with returns of the other countries. However, when appropriate lagged variables were used, only past values of returns in the Chinese, Singapore and US stock-indices have predictive value in terms of future Australian-stock-index values. This study should be of interest to policy makers, especially in terms of its suggestion of the flow-on effects of sharp discontinuities in the Chinese, Singapore and US stock markets on the Australian stock market. Bilateral trade links appear to be a key linkage in the transmission of financial contagion. However, more research is needed to identify and understand other factors that contribute to the spread of financial crises. Introduction The recent global financial crisis (GFC) enhanced interest among financial researchers on the spread of financial crisis across borders and on finding appropriate policy stances to mitigate such contagion. In particular, there is interest as to how and why some nations are vulnerable to financial contagion, whereas, others appear to be more immune. Kaminsky and Reinhart (2000, p.51) define contagion as the process by which ―…financial difficulties spread from one economy to another in the same region and beyond‖ via trade and financial linkages. However, Caramazza et al. (2004) note that, because countries tend to concurrently establish regional trade agreements and the interbank linkages, it is difficult to determine if financial contagion is mostly due mostly to financial or to trade links. However, empirical studies on linkages may help develop early-warning indicators to facilitate _________________ † Contact Author, Faculty of Business, Federation University, PO Box 663, Ballarat, Victoria 3353, Australia; Email: s.hettihewa@ballarat.edu.au; Tel: +61 3 5327 9158 * Faculty of Business, Federation University, Ballarat, Victoria ** Higher Education Faculty, Holmes Institute, Melbourne, Victoria 1 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 timely intervention to ameliorate future crises. This empirical study seeks to examine the role that inter-country linkages play in the spread of financial crises by focusing on the notion that an understanding of financial linkages can explain why and how financial crises spread across borders. Further, this study hypothesises that a country is more likely to experience financial contagion from a country with which it has extensive trade links. This study investigates Australian economy and its top trading partners. The Department of Foreign Affairs and Trade (DFAT, 2012), asserts that Australia’s top six bi-lateral trading partners in 2011 were (in order of importance) China, Japan, US, Korea, Singapore and UK. Minerals are Australia’s key export. The rest of the paper is organised as follows: A literature review considering the role of policy makers during a financial crisis is given first; an outline of the data selection, research method and analysis and findings are discussed next; and The paper is drawn to a close with limitations of the study and conclusion. Literature Review The Role of Policy Makers during a Financial Crisis In the aftermath of a financial crisis, researchers often seek to understand what cause a crisis, and, in the case of financial contagion 9 the 2007-09 ―Global Financial Crisis‖ (GFC)), the contributing factors of how it spread. A common view is that timely intervention might have avoided or minimized the resulting losses. Information asymmetry Generally, some level of information asymmetry is present in all financial markets and limits the ability of investor to differentiate between profitable and non-profitable securities (Mishkin, 1999). Edgar (2009) argues that a lack of accountability on the part of the regulatory authorities may have created a toxic environment where regulators could make decisions aimed at self-preservation as opposed to the common good of society. Systemic weakness Dabrowski (2010) notes that, during the GFC, even though policy response was delayed and poorly co-ordinated, systemic weaknesses were prevalent in European banks that were over-leveraged. Regulators could address future vulnerability of the European financial sector to contagion of a financial crisis by improving financialinstitution regulation and via timely policy implementation. In the absence of systemic weaknesses, adaptive policies may minimise the effects of a crisis on an economy (e.g. strict supervisory practices by the Australian Prudential Regulation Authority’s (APRA) is a major reason why Australian banks fared well during the GFC; Pais and Stork, 2011; Edwards (2010)). Also since 2003 Australia enjoyed a mining boom driven by China’s demand (Sykes, 2010). There is, however, speculation that this mining boom has ended and that Australia may be less immune to future financial crises or contagion. Thus, it may be prudent for Australia to consider the potential for future contagion from its key trading partners. 2 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Research Questions This research aims to answer the following research questions: a) Does the risk of financial contagion in equity markets reflect the trade linkages that exist between Australia and its key trading partners? b) Can past values of composite stock indexes of Australia’s key bilateral partners be used to use to predict movements of the Australian composite stock index? Data Selection, Research Methods, Analysis and Findings Data Selection Monthly data (Jul/97-Jan/14) for closing prices of the All Ordinaries (Australia), Hang Seng (China), S&P500 (United States), Kospi (South Korea), Nikkei 225 (Japan), Straits Time (Singapore) and FTSE 100 (United Kingdom) stock indexes were obtained from the Yahoo finance database. All stock time series are transformed to natural logarithms and the summary statistics of the transformed indexes are in Table 1. INSERT TABLE 1 HERE This study also considers a subset of the entire dataset ranging from Jan/00-Jan/14 (see Table 2). The second data set excludes the 1997-1999 Asian crisis that originated in Thailand and subsequently affected neighbouring countries. Statistical analysis is performed using the Eviews7 econometric package. INSERT TABLE 2 HERE Research Methods The main objective of this study is to explore causal relationships between stock indexes of Australia and its key trading partners. This section discusses the Granger causality procedure and briefly highlights the empirical tests required prior to conducting Granger causality tests. Granger causality was first introduced by Granger (1969) who argued that it occurs when past values of one series ( ) can be used to predict the current value of another series ( ). is said to Granger cause if it contains information that predicts series and vice versa. The nature of causality may be unidirectional or bidirectional. Granger causality relationships between economic or financial variables can form the basis for risk management (e.g. a study on causal relationships between world oil and agricultural commodity prices could reveal that causal relationships exist; Nazlioglu and Soytas, 2012). Granger-causality tests check for existence of short-run causal relationships between two series using bivariate vector autoregressive (VAR) models. The structure of bivariate VAR models is formulated based on the properties of the individual series and the co-integrative relationship between the two series of interest. A stationary series has mean reverting tendencies and contains no unit root while a non-stationary series follows a random walk and contains a unit root. This study uses the standard Augmented Dickey Fuller (ADF) tests developed by Dickey and Fuller (1979) to check for the presence of a unit root (Engle and Granger, 1987). 3 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 The second step of Granger causality testing involves checking whether variables are co-integrated. The use of non-stationary variables that are integrated of the same order and yet are not co-integrated could result in spurious causality regressions (Dakurah and Sampath, 2001) Granger and Newbold (1974). We discuss two well accepted co-integration tests—the Engle and Granger (1987) technique and the Johansen framework (Johansen, 1988; Johansen, 1991)—and explain why we opted for the use of the latter Granger (1969) recommends that when two series are level stationary (meaning they are integrated of order zero I(0)), the Granger causality relationship can be tested using the bivariate vector autoregressive (VAR) model in equation (1) and (2). ∑ ∑ ∑ ∑ Eqn (1) Eqn (2) Where and represent logged stock indexes at the level, and are the ith lagged coefficients of stock index and respectively, and are the constant terms and and are the error terms of the estimated VAR models. It was found that that Granger causes if any is not equal to zero and Granger causes if any is not equal to zero. Thus, if all and are equal to zero no causal relationship exists between and . According to Granger et al. (2000), if two series are not stationary and are not cointegrated, the bivariate VAR model for Granger causality test should be specified using the differenced form of the series as shown in equations (3) and (4), where is the first difference operator for the logged time series. ∑ ∑ ∑ ∑ Eqn (3) Eqn (4) Furthermore, if two variables are non-stationary at the level yet co-integrated, Engle and Granger (1987) recommend the inclusion of an error correction term (ECT) to equation (3) and (4) to avoid model misspecification. The resultant error correction model (ECM) is as shown in equation (5) and (6), where and are the coefficients for the error correction term for series and respectively. and are the lagged error correction terms for equation (5) and (6) respectively. ∑ ∑ ∑ ∑ Eqn (5) Eqn (6) Results Unit Root Tests Graphical representations of the logged stock indexes in Figure 1 and 2 indicate that the series are non-stationary. INSERT FIGURE 1 HERE 4 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 INSERT FIGURE 2 HERE This study checked each series stationary properties using the standard Augmented Dickey Fuller (ADF) tests developed by Dickey and Fuller (1979). The ADF tests the null hypothesis that a series contains a unit root. The mathematical expressions for testing the null hypothesis as show in equation (7) and (8). Equation (7) has a constant and no trend while equation (8) has a constant and a trend term. Eqn (7) Eqn (8) Where is the first difference of the stock index, α is a constant term, β is the coefficient of the trend term, t is the trend term and is the correlation coefficient of the lagged stock index. is coefficient of the first difference of the first lag of the stock index, is the coefficient of the 1st difference of the pth lag of the stock index and is the error term. Ng and Perron (2001) recommend the used of the Modified Akaike Criterion (MAC) to select the number of lags ( to include in equations (7) and (8). The null hypothesis of a unit root is rejected if , and it will be concluded that a series is stationary; conversely, if , the null hypothesis of existence of a unit root cannot be rejected and a conclusion should be made that the series is non-stationary. Table 3 show the results of ADF unit root tests for the first and second dataset. As expected, the tests confirm that all stock indexes contain a unit root and all series are non-stationary at 5% level of significance. INSERT TABLE 3 HERE In order to determine the order of integration we use the first difference of all series. There is no evidence of a deterministic trend component in the differenced series as shown in Figure 3. Consequently, the ADF test for a unit root is performed using the equation (7), which excludes the trend component. The results in table 4 show that the first difference of the stock indexes (the stock returns) is stationary since the null hypothesis for no unit root is rejected at all levels of significance. Consequently, the stock returns of the indexes are used for the subsequent analysis of Granger Causality. INSERT FIGURE 3 HERE INSERT TABLE 4 HERE Co-integration Tests The results of the unit root tests show that all stock index series are integrated of order one (I(1)) and are stationary after taking the first difference. The Engle and Granger (1987) technique we use involves a two-step-testing process that tests for a co-integrating relationship between two variables by estimating an ordinary least squares regression. Thereafter, the residuals of the regression are examined in order to establish whether the two series are co-integrated (i.e. a linear 5 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 combination of the variables has residuals ( ) that are stationary ( ). As a group of M variables can have up to M-1 co-integrating relationships (Koop, 2006), we can have up to six co-integrating relationships in this study. The Johansen framework was used to resolve this issue (Johansen, 1988; Johansen, 1991). Johansen (1988) developed a framework using a vector-autoregressive (VAR) model. The VAR of order p is expressed in matrix notation as shown in equation (9) (Johansen, 1995; Johansen, 1991). As the Johansen technique is sensitive to the variation of lags, the Akaike Information Criterion (AIC) is used to specify the appropriate lag length (p) for performing a Johansen tests. The AIC recommends the use of VAR model with two lags for both datasets. ∑ Eqn (9) ∑ are n x n matrices of unknown parameters where ∑ . is the first difference of a k-vector of I(1) variables, vector of deterministic variables and is the vector of innovations. and , and is a d- If the coefficient matrix Π has reduced rank r < p, then there exist p × r matrices α and β such that Π = α β΄ and β΄ is stationary, even if is nonstationary (Johansen, 1988: 170). The rank r shows the number of cointegrating relations, α indicates the speeds of adjustment and each column of β contains the long-run cointegrating vectors. Two likelihood-ratio tests are recommended for successive hypothesis tests (i.e. the trace test and maximum-eigenvalue test). The trace test statistic and maximum-eigenvalue test statistic are calculated where is the sample size, is the ith largest eigenvalue of the matrix and is the total number of endogenous variables under consideration). ∑ Eqn (10) Eqn (11) In the null hypothesis, the trace tests is and, in alternative hypothesis, . The maximum eigenvalue tests against where k = 0 through 6. The sequential hypothesis testing starts with the test for zero or no cointegrating relationships versus the hypothesis of one or more co-integrating relationships. Subsequent hypothesis tests are performed by successively increasing the value of k by one unit. See Tables 5 and 7. INSERT TABLE 5 HERE From table 5, the null of no co-integration is rejected by the trace test, at a 5% level of significance, since the trace test statistic of 125.9897 is greater than the critical value of 125.6154. The null hypothesis of the subsequent hypothesis tests is rejected because the calculated trace statistics are less than the corresponding critical values. Consequently, the trace test shows that one co-integrating relationship exists among seven variables. Turning to the maximum eigenvalue test results (Table 5), the maximum eigenvaluetest statistic for the null of no co-integration of 43.8964 is smaller than the critical 6 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 value of 46.2314. Thus, the null of no co-integrating relationships is accepted at a 5% level of significance. Similar results are obtained from subsequent tests where findings fail to reject the null hypotheses because the calculated maximum eigenvalue test statistics are smaller than the critical values. Based on the results of the maximum eigenvalue test, there is no co-integration relationship between the seven stock indexes. The two likelihood-ratio tests for the whole period (Jul/97-Jan/14) yield conflicting results, with the trace test indicating that there is one co-integrating relationship and the maximum-eigenvalue test indicating that no co-integrating relationship exists. However, Enders (1995) prefers the use of the maximum-eigenvalue test due to the ―sharper alternative hypothesis‖ (p.393). Similarly, Kennedy (2008) considers the maximum-eigenvalue test a superior test for co-integration compared to the trace test. The resultant co-integration equation can be expressed as shown in equation (12) and the long-run coefficients of the resultant co-integrated equation are shown in table 6. Eqn (12) INSERT TABLE 6 HERE Table 7 shows the Johansen-co-integration test results for the second dataset (Jan/00-Jan/14). For the null hypothesis of no co-integration, the test statistics are 135.1233 for the trace test and 51.6265 for the maximum-eigenvalue test. As both test statistics exceed the critical values at the 5% significance level, the null hypothesis (no co-integration) is rejected. Since subsequent trace and maximumeigenvalue tests statistics fall below the 5% critical value, there is no statistical evidence to support the existence of more than one co-integrating relationship. The co-integrating relationship can be expressed as shown in equation (12) and the estimated co-integration coefficients are reported in Table 8. INSERT TABLE 7 HERE INSERT TABLE 8 HERE Given that the series are non-stationary and are co-integrated an error correction term must be included in the Granger-causality tests. The error correction model (ECM) applicable in this instance is as specified in equations (5) and (6). We specify the appropriate lag-length for conducting Granger-causality tests based on the recommendations for the Johansen-co-integration test. Consequently, the ECM is estimated using two lags and one co-integrating relationship. The resultant VECM’s for the 2 datasets are shown in Tables 9 and 10. INSERT TABLE 9 HERE INSERT TABLE 10 HERE Pairwise Granger causality tests were performed using the estimated error correction terms for VECMs with two lags. Eviews tests for the null hypothesis of Granger noncausation, whereby a rejection of the null hypothesis means that a causal 7 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 relationship exists. The results of the Granger causality tests for both data sets are contained in Table 11. For the period from Jul/97-Jan/14, the null hypothesis that the Singapore stock index does not Granger cause South Korea’s stock index is rejected, at a 1% significance level. Similarly, the null hypothesis for non-causality is rejected at a 5% significance level revealing six Granger-causality relationships. Specifically, a bidirectional causation exists between Singapore and Japan (Singapore causes Japan and vice versa) and unidirectional-Granger-causation relationships exist such that China causes Australia, Singapore causes Australia, China causes Singapore and South Korea causes Singapore. At a 10% significance level, US Granger causes Australia, Australia Granger causes Singapore and US Granger causes Singapore. For the period from Jan/00-Jan/14, there are fewer causality relationships with Singapore Granger causing South Korea (1% level of significance), Japan Granger causing Singapore (5% significance level) and US granger causing Australia and Japan Granger causing the US (10% significance level). No other Granger causality relationships are identified. INSERT TABLE 11 HERE Discussion of Findings and Conclusions The objective of this study is to examine whether movements in the stock markets of Australia’s key trading partners can explain movements in Australia’s equity markets. The Johansen co-integration test revealed one long-run equilibrium relationship. When the Jul/97-Jan/14 period was examined, more causality relationships were observed as compared to Jan/00-Jan/14. Thus, in a period in which more crises have occurred, there appears to be more causality relationships (i.e. perceived risk of contagion). Thus, episodes of crisis appear to nurture more perceived risk of contagion than would be observed in less turbulent times. Concerning Australia, in the short-run, past values of the Chinese, Singaporean and American stock index could help explain the current value of the Australia Stock index when considering the first period (1997-2014). However, there is no evidence of reverse causality from Australia to China or Australia to the US stock markets. Despite the presence of extensive trade links and economic partnerships, only three significant short-run relationships were found in the first period. Empirical results indicate stronger Granger causation from China to Australia and Singapore to Australia equity markets. We also find Granger Causation from the US to Australia but the significance is lower at a 10% level compared to the 5% level of the two aforementioned relationships. This means that Singapore has more impact on the Australian equity market than vice versa. Furthermore, movements in the Chinese and the US equity markets can affect the Australian equity markets but not vice versa. This study found that certain trade links appear to play a more of a role in contagion risks to the Australian economy than others. It is interesting that Singapore, has more influence on the movements in the Australia equity market. Accordingly, Australian policy makers should consider the degree of the impact of the Chinese, Singaporean and the US stock market on the Australian stock market. Nonetheless, while impact of trade links are important, more research is needed to 8 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 identify and understand other factors that contribute to financial crises and the complex nature of how those factors interact with each other. References Caramazza F, Ricci L and Salgado R. 2004. International financial contagion in currency crises. Journal of International Money and Finance: 51-70. Dabrowski M. 2010. The global financial crisis: lessons for European integration Economic Systems 34: 38-54. Dakurah AH, Davies, S. P and Sampath RK. 2001. Defense spending and economic growth in developing countries A causality analysis. Journal of Policy Modeling 23: 651-658. DFAT. 2012. Trade at a Glance 2012. Available at: http://www.dfat.gov.au/publications/trade/trade-at-a-glance-2012.pdf. DFAT. 2013. China Country Fact Sheet. Available at: http://www.dfat.gov.au/geo/fs/chin.pdf. DFAT. 2013b. Japan Country Fact Sheet. Available at: http://www.dfat.gov.au/geo/fs/japan.pdf. DFAT. 2013c. United States of America Country Fact Sheet. Available at: http://www.dfat.gov.au/geo/fs/usa.pdf. Dickey DA and Fuller WA. 1979. Distribution of the Estimators for Autoregressive Time Series with Unit Root. Journal of American Statistical Association 74: 427-431. Edgar RJ. 2009. The Future of Financial Regulation: Lessons from the Global Financial Crisis. The Australian Economic Review 42: 470-476. Edwards J. 2010. Australia after the global financial crisis. Australian Journal of International Affairs 64: 359-371. Enders W. 1995. Applied Econometric Time Series New York: John Wiley & Sons. Engle RF and Granger CWJ. 1987. Co-Integration and Error Correction: Representation, Estimation, and Testing. Econometrica 55: 251-276. Granger CWJ. 1969. Investigating Causal Relations by Econometric Models and Cross-Spectral Methods. Econometrica 37: 424-438. Granger CWJ, Huang B-N and Yang C-W. 2000. A bivariate causality between stock prices and exchange rates: evidence from recent Asian flu. The Quarterly Review of Economics and Finance 40: 337-354. Granger CWJ and Newbold P. 1974. Spurious Regressions in Econometrics. Journal of Econometrics 2: 111-120. Johansen S. 1988. Statistical analysis of cointegrating vectors. Journal of Economic Dynamic and Control 12: 231-254. Johansen S. 1991. Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models. Econometrica 59: 1551-1580. Johansen S. 1995. Likelihood-based inference in cointegrated vector autoregressive models, Oxford: Oxford University Press. Kaminsky GL and Reinhart CM. 2000. On crises, contagion, and confusion. Journal of International Economics 51: 145-168. Kennedy P. 2008. A Guide to Econometrics, 6th edition, Cambridge, MA: WileyBlackwell. Koop G. 2006. Analysis of Financial Data, Chichester: John Wiley & Sons, Ltd. 9 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Mishkin FS. 1999. Lessons from the Asian crisis. Journal of International Money and Finance 18 709–723. Nazlioglu S and Soytas U. 2012. Oil price, agricultural commodity prices, and the dollar: A panel cointegration and causality analysis. Energy Economics 34: 1098-1104. Ng S and Perron P. 2001 Lag length selection and the construction of unit root tests with good size and power. Econometrica 69: 1519–1554. Pais A and Stork PA. 2011. Contagion risk in the Australian banking and property sectors. Journal of Banking & Finance 35: 681-697. Sykes T. 2010. Six Months of Panic: How the Global Financial Crisis Hit Australia, Crows Nest, NSW: Allen & Unwin. 10 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Appendices Table 1: Descriptive Statistics for Stock Indexes (Jul/97-Jan/14) Australi Singapor China Japan Korea UK a e Mean Median Maximum Minimum Std. Dev. JarqueProbability Bera 8.274 8.307 8.822 7.810 0.254 10.578 0.005 9.658 9.651 10.35 8.892 3 0.321 9.236 0.010 9.420 9.373 9.920 8.932 0.255 6.974 7.013 7.693 5.697 0.519 0.001 13.060 0.002 12.102 7.698 7.702 8.244 6.753 0.318 8.724 0.013 8.597 8.637 8.844 8.180 0.153 0.000 18.151 Table 2: Descriptive Statistics for logged Stock Indexes (Jan/00-Jan/14) Australi China Japan Korea Singapor UK Mean 8.335 9.717 9.371 7.088 7.757 8.586 a e Median Maximum Minimum Std. Dev. JarqueProbability Bera 8.375 8.822 7.930 0.225 8.107 0.017 9.751 10.35 9.064 3 0.300 8.981 0.011 9.314 9.920 8.932 0.242 7.222 7.693 6.173 0.452 0.006 10.217 0.000 15.607 7.798 8.244 7.145 0.289 11.638 0.003 8.625 8.817 8.180 0.159 0.001 13.575 US 7.09 7.11 7 7.52 4 6.60 2 0.17 0 2.32 6 0.31 6 3 US 7.10 7.12 7 7.52 8 6.60 2 0.18 0 2.80 0 0.24 6 6 Table 3: Augmented Dickey Fuller Unit Root Test Results of Stock Indexes Jul/97-Jan/14 Jan/00-Jan/14 Country Variable Constant Constant Trend Constant Constant Trend Australia LN(Aus) -1.514325 -1.929803 -1.452244 -1.772412 China LN(China) -1.645075 -3.183113* -1.468013 -2.353573 UK LN(UK) -1.848818 -1.886797 -1.636531 -2.051673 Singapore LN(Singapore) -1.724273 -3.015349 -1.383193 -2.538849 US LN(US) -1.895666 -2.065813 -1.400415 -1.900108 Japan LN(Japan) -2.111976 -1.69135 -2.490175 -2.186308 Korea LN(Korea) -1.283361 -3.181857* -0.797688 -3.062941 Note: *, **, *** are statistically significant at, respectively, the 10 %, 5 % and 1 % level. Table 4: Augmented Dickey Fuller Unit Root Test Results of Stock Indexes Constant Country Variable Jul/97-Jan/14 Jan/00-Jan/14 Australia China UK Singapore US Japan Korea ∆LN(Aus) ∆LN(China) ∆LN(UK) ∆LN(Singapore) ∆LN(US) ∆LN(Japan) ∆LN(Korea) -5.856639*** -6.055578*** -4.582995*** -4.587288*** -5.399556*** -5.957932*** -6.763743*** -5.119305*** -5.798779*** -3.619911*** -7.606968*** -5.224065*** -6.138804*** -6.213806*** 11 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Table 5: Seven-country Johansen co-integration test results (Jul/97-Jan/14) Trace Rank test Number of co-integrating Eigenvalu Trace Statistic 5% critical relationships e value None * At most 1 At most 2 At most 3 At most 4 At most 5 At most 6 Maximum Eigenvalue Rank test Number of co-integrating relationships None At most 1 At most 2 At most 3 At most 4 At most 5 At most 6 * Hypothesis rejected at the 5% level 0.2007 0.1504 0.1402 0.0318 0.0307 0.0261 0.0149 Eigenvalu e 0.2007 0.1504 0.1402 0.0318 0.0307 0.0261 0.0149 125.9897 82.0933 50.1547 20.5530 14.2209 8.1167 2.9411 Max-Eigen Statistic 43.8964 31.9387 29.6017 6.3321 6.1042 5.1756 2.9411 125.6154 95.7537 69.8189 47.8561 29.7971 15.4947 3.8415 5% critical value 46.2314 40.0776 33.8769 27.5843 21.1316 14.2646 3.8415 Table 6: Co-integrating equation for the seven countries (Dependent variable: Australia, Time period: Jul/97-Jan/14) Independent variables Normalized co-integrating coefficients Standard error China Japan Korea Singapore UK US 0.2264 0.7223 0.4716 0.1879 -0.9485 -0.1010 (0.2906) (0.1415) (0.1745) (0.4195) (0.3001) (0.3606) 12 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Table 7: Seven-country Johansen co-integration test results (Jan/00-Jan/14) Trace Rank test Number of co-integrating Eigenvalu Trace Statistic 5% critical relationships e value None * At most 1 At most 2 At most 3 At most 4 At most 5 At most 6 Maximum Eigenvalue Rank test Number of co-integrating relationships None * At most 1 At most 2 At most 3 At most 4 At most 5 At most 6 * Hypothesis rejected at the 5% level 0.2673 0.1511 0.1271 0.0938 0.0572 0.0242 0.0211 Eigenvalu e 0.2673 0.1511 0.1271 0.0938 0.0572 0.0242 0.0211 135.1233 83.4968 56.2990 33.7404 17.3866 7.6096 3.5349 Max-Eigen Statistic 51.6265 27.1978 22.5586 16.3538 9.7769 4.0748 3.5349 125.6154 95.7537 69.8189 47.8561 29.7971 15.4947 3.8415 5% critical value 46.2314 40.0776 33.8769 27.5843 21.1316 14.2646 3.8415 Table 8: Co-integrating equation for the seven countries (Dependent variable: Australia, Time Period: Jan/00-Jan/14) Independent variables Normalized co-integrating coefficients Standard error China Japan Korea Singapore UK US -0.2392 0.5936 0.0789 1.0439 -0.5854 -0.2578 (0.1485) (0.0696) (0.0891) (0.2299) (0.2082) (0.1709) 13 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Table 9: Vector Error Correction Estimates for Australia and Top Six Bilateral Traders (Jul/97Jan/14) ∆ AUSTRALIA 0.010556 (0.02074) [ 0.50909] ∆ AUSTRALIA (-1) -0.049056 (0.12784) [-0.38374] ∆ AUSTRALIA(-2) 0.166846 (0.12824) [ 1.30102] ∆ CHINA(-1) 0.142287 (0.06665) [ 2.13481] ∆ CHINA (-2) -0.066014 (0.06618) [-0.99752] ∆ JAPAN (-1) 0.113894 (0.06431) [ 1.77089] ∆ JAPAN(-2) -0.053445 (0.06393) [-0.83600] ∆KOREA (-1) -0.036421 (0.04156) [-0.87632] ∆KOREA (-2) 0.013035 (0.04417) [ 0.29509] ∆ SINGAPORE(1) -0.095219 (0.07114) [-1.33848] ∆ SINGAPORE(2) 0.157205 (0.06729) [ 2.33618] ∆ US(-1) 0.196730 (0.12496) [ 1.57429] ∆ US(-2) -0.158183 (0.12282) [-1.28795] ∆ UK(-1) -0.135641 (0.13241) [-1.02436] ∆ UK(-2) -0.046302 (0.13302) [-0.34809] Constant 0.002710 (0.00277) [ 0.97749] Error Correction Term ∆ CHINA 0.005225 (0.04127) [ 0.12658] 0.007185 (0.25446) [ 0.02823] 0.418073 (0.25527) [ 1.63775] 0.104586 (0.13267) [ 0.78830] 0.034475 (0.13173) [ 0.26171] 0.208632 (0.12802) [ 1.62967] -0.082835 (0.12726) [-0.65093] 0.158585 (0.08273) [ 1.91693] 0.070965 (0.08793) [ 0.80706] ∆ JAPAN 0.026595 (0.03200) [ 0.83096] -0.110370 (0.19732) [-0.55935] 0.347417 (0.19794) [ 1.75514] 0.150406 (0.10288) [ 1.46201] -0.092329 (0.10215) [-0.90388] 0.129494 (0.09927) [ 1.30446] -0.030752 (0.09868) [-0.31165] -0.059715 (0.06415) [-0.93087] 0.062607 (0.06818) [ 0.91823] ∆KOREA ∆ SINGAPORE 0.193251 0.028438 (0.04553) (0.03718) [ 4.24402] [ 0.76496] -0.020419 -0.299923 (0.28073) (0.22920) [-0.07274] [-1.30859] -0.302565 0.413405 (0.28162) (0.22992) [-1.07437] [ 1.79801] 0.265244 0.280558 (0.14637) (0.11950) [ 1.81219] [ 2.34781] 0.038032 0.116788 (0.14533) (0.11865) [ 0.26170] [ 0.98430] 0.076651 0.243952 (0.14124) (0.11531) [ 0.54272] [ 2.11565] 0.170783 -0.175243 (0.14039) (0.11462) [ 1.21648] [-1.52892] 0.069198 0.184020 (0.09127) (0.07451) [ 0.75819] [ 2.46961] -0.068616 0.111169 (0.09701) (0.07920) [-0.70734] [ 1.40368] ∆ US -0.023069 (0.02521) [-0.91513] 0.105079 (0.15542) [ 0.67611] 0.301573 (0.15591) [ 1.93429] 0.096440 (0.08103) [ 1.19018] -0.102959 (0.08046) [-1.27969] 0.135860 (0.07819) [ 1.73757] -0.094352 (0.07772) [-1.21396] -0.075159 (0.05053) [-1.48750] 0.040236 (0.05370) [ 0.74923] ∆ UK -0.022342 (0.02346) [-0.95233] 0.155500 (0.14464) [ 1.07510] 0.219978 (0.14510) [ 1.51607] 0.028965 (0.07541) [ 0.38410] -0.106618 (0.07488) [-1.42391] 0.089854 (0.07277) [ 1.23482] -0.121554 (0.07233) [-1.68049] -0.035226 (0.04702) [-0.74912] -0.001159 (0.04998) [-0.02320] -0.132444 -0.128588 -0.145281 (0.14161) (0.10980) (0.15622) [-0.93530] [-1.17107] [-0.92996] -0.247293 (0.12755) [-1.93887] -0.063938 (0.08649) [-0.73928] 0.009161 (0.08049) [ 0.11382] -0.059988 (0.13395) [-0.44785] 0.067097 (0.24875) [ 0.26974] -0.317365 (0.24447) [-1.29815] -0.412735 (0.26358) [-1.56589] -0.081088 (0.26478) [-0.30625] 0.001103 (0.00552) [ 0.19976] 0.052129 (0.12065) [ 0.43208] 0.306318 (0.22405) [ 1.36720] -0.388598 (0.22020) [-1.76476] -0.375263 (0.23740) [-1.58069] -0.135832 (0.23849) [-0.56956] 0.000910 (0.00497) [ 0.18299] 0.132869 (0.08181) [ 1.62414] 0.036968 (0.15192) [ 0.24333] -0.236421 (0.14931) [-1.58338] -0.043719 (0.16098) [-0.27158] -0.020815 (0.16171) [-0.12872] 0.002509 (0.00337) [ 0.74437] 0.110115 (0.07614) [ 1.44629] 0.186943 (0.14139) [ 1.32219] -0.085248 (0.13896) [-0.61347] -0.306797 (0.14982) [-2.04779] -0.066942 (0.15050) [-0.44479] -0.00007 (0.00314) [-0.02375] 0.224815 (0.10386) [ 2.16449] 0.066356 (0.19288) [ 0.34402] -0.266551 (0.18957) [-1.40608] 0.160189 (0.20438) [ 0.78377] -0.106576 (0.20531) [-0.51909] -0.001548 (0.00428) [-0.36167] 0.451984 (0.14777) [ 3.05864] -0.035590 (0.27442) [-0.12969] -0.306883 (0.26971) [-1.13783] 0.048388 (0.29078) [ 0.16640] -0.322892 (0.29211) [-1.10538] 0.007077 (0.00609) [ 1.16236] Note: Standard errors are shown in ( ) and t-statistics in [ ] 14 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Table 10: Vector Error Correction Estimates for Australia and Top Six Bilateral Traders (Jan/00Jan/14) Error Correction Term ∆ AUSTRALIA (-1) ∆ AUSTRALIA(-2) ∆ CHINA(-1) ∆ CHINA (-2) ∆ JAPAN (-1) ∆ JAPAN(-2) ∆KOREA (-1) ∆KOREA (-2) ∆ SINGAPORE(-1) ∆ SINGAPORE(-2) ∆ UK(-1) ∆ UK(-2) ∆ US(-1) ∆ US(-2) Constant ∆ ∆ CHINA AUSTRALIA -0.089661 -0.056403 (0.04046) (0.07175) [-2.21591] [-0.78613] 0.041350 0.206665 (0.13789) (0.24450) [ 0.29988] [ 0.84525] 0.163209 0.264817 (0.13641) (0.24188) [ 1.19645] [ 1.09482] 0.118460 0.093179 (0.08135) (0.14425) [ 1.45616] [ 0.64595] -0.033651 -0.133118 (0.07897) (0.14004) [-0.42610] [-0.95060] 0.083056 0.054679 (0.06874) (0.12189) [ 1.20827] [ 0.44860] -0.118840 -0.136173 (0.07085) (0.12562) [-1.67743] [-1.08397] -0.065950 -0.015450 (0.06359) (0.11276) [-1.03706] [-0.13701] 0.114634 0.186229 (0.06287) (0.11148) [ 1.82333] [ 1.67049] -0.078151 -0.019571 (0.09411) (0.16688) [-0.83041] [-0.11727] 0.117561 0.046193 (0.08881) (0.15748) [ 1.32374] [ 0.29333] -0.043982 -0.055108 (0.15671) (0.27787) [-0.28067] [-0.19832] 0.013387 0.074799 (0.15510) (0.27502) [ 0.08632] [ 0.27198] 0.147900 -0.037612 (0.15002) (0.26602) [ 0.98585] [-0.14139] -0.288646 -0.306704 (0.14642) (0.25963) [-1.97133] [-1.18130] 0.002159 -0.000338 (0.00293) (0.00519) [ 0.73705] [-0.06514] ∆ JAPAN ∆KOREA 0.111687 (0.06470) [ 1.72617] -0.054856 (0.22049) [-0.24879] 0.221216 (0.21813) [ 1.01416] 0.062660 (0.13008) [ 0.48169] -0.124222 (0.12628) [-0.98367] 0.135365 (0.10992) [ 1.23151] -0.040229 (0.11329) [-0.35511] -0.112247 (0.10169) [-1.10382] 0.149972 (0.10053) [ 1.49176] 0.079797 (0.15049) [ 0.53024] 0.280785 (0.14201) [ 1.97719] 0.184727 (0.25058) [ 0.73719] 0.003493 (0.24801) [ 0.01408] -0.026630 (0.23989) [-0.11101] -0.384441 (0.23414) [-1.64195] -0.002425 (0.00468) [-0.51773] 0.169021 (0.07477) [ 2.26044] 0.343576 (0.25481) [ 1.34835] -0.089269 (0.25208) [-0.35413] -0.043817 (0.15033) [-0.29147] -0.203081 (0.14594) [-1.39152] 0.043284 (0.12703) [ 0.34074] -0.025407 (0.13092) [-0.19407] -0.127809 (0.11752) [-1.08757] 0.116820 (0.11618) [ 1.00548] 0.108275 (0.17392) [ 0.62257] 0.512227 (0.16412) [ 3.12109] 0.101048 (0.28959) [ 0.34894] -0.015037 (0.28661) [-0.05246] -0.040864 (0.27724) [-0.14740] -0.544854 (0.27058) [-2.01364] 0.004373 (0.00541) [ 0.80780] ∆ SINGAPORE ∆ UK ∆ US 0.041327 -0.060584 -0.035163 (0.06229) (0.04572) (0.04921) [ 0.66346] [-1.32521] [-0.71457] 0.086999 0.158200 0.164885 (0.21227) (0.15579) (0.16769) [ 0.40985] [ 1.01545] [ 0.98325] 0.231662 0.167470 0.239477 (0.20999) (0.15412) (0.16590) [ 1.10318] [ 1.08660] [ 1.44354] 0.220996 -0.006256 0.077536 (0.12523) (0.09191) (0.09894) [ 1.76467] [-0.06806] [ 0.78371] 0.007788 -0.112747 -0.137879 (0.12158) (0.08923) (0.09604) [ 0.06406] [-1.26358] [-1.43557] 0.228977 0.097834 0.163256 (0.10582) (0.07766) (0.08360) [ 2.16385] [ 1.25970] [ 1.95289] -0.189422 -0.139073 -0.088056 (0.10906) (0.08005) (0.08616) [-1.73682] [-1.73742] [-1.02200] -0.042528 -0.092164 -0.130860 (0.09790) (0.07185) (0.07734) [-0.43442] [-1.28271] [-1.69202] 0.129042 0.079947 0.076033 (0.09678) (0.07103) (0.07646) [ 1.33329] [ 1.12547] [ 0.99441] -0.175216 0.079463 0.006075 (0.14488) (0.10633) (0.11445) [-1.20939] [ 0.74731] [ 0.05307] 0.193892 0.123473 0.108515 (0.13672) (0.10034) (0.10801) [ 1.41820] [ 1.23052] [ 1.00471] -0.260303 -0.325737 -0.093477 (0.24124) (0.17705) (0.19058) [-1.07903] [-1.83976] [-0.49049] -0.231480 0.051256 -0.034868 (0.23876) (0.17524) (0.18862) [-0.96950] [ 0.29250] [-0.18486] 0.230654 0.216012 0.058875 (0.23095) (0.16950) (0.18245) [ 0.99873] [ 1.27439] [ 0.32269] -0.149733 -0.248211 -0.189749 (0.22541) (0.16543) (0.17807) [-0.66428] [-1.50037] [-1.06558] 0.000334 -0.001048 0.000377 (0.00451) (0.00331) (0.00356) [ 0.07411] [-0.31657] [ 0.10573] Note: Standard errors are shown in ( ) and t-statistics in [ ] 15 Proceedings of Global Business and Finance Research Conference 5-6 May, 2014, Marriott Hotel, Melbourne, Australia, ISBN: 978-1-922069-50-4 Table 11: Granger Causality tests for seven countries Null hypothesis ∆China /→∆Australia ∆Japan /→∆Australia ∆Korea /→∆Australia ∆Singapore /→∆Australia ∆US /→∆Australia ∆UK /→∆Australia ∆Australia /→∆China ∆Japan /→∆China ∆Korea /→∆China ∆Singapore /→∆China ∆US /→∆China ∆UK /→∆China ∆Australia /→∆Japan ∆China /→∆Japan ∆Korea /→∆Japan ∆Singapore /→∆Japan ∆US /→∆Japan ∆UK /→∆Japan ∆Australia /→∆Korea ∆China /→∆Korea ∆Japan /→∆Korea ∆Singapore /→∆Korea ∆US /→∆Korea ∆UK /→∆Korea ∆Australia /→∆Singapore ∆China /→∆Singapore ∆Japan /→∆Singapore ∆Korea /→∆Singapore ∆US /→∆Singapore ∆UK /→∆Singapore ∆Australia /→∆US ∆China /→∆US ∆Japan /→∆US ∆Korea /→∆US ∆Singapore /→∆US ∆UK /→∆US ∆Australia /→∆UK ∆China /→∆UK ∆Japan /→∆UK ∆Korea /→∆UK ∆Singapore /→∆UK ∆US /→∆UK Jul/97 to Jan/14 Chi-squared stat. 6.148606** 3.846562 0.845459 7.897418** 4.810567* 1.069674 2.703734 3.087865 4.386563 1.006291 1.895752 2.452175 3.643162 3.317314 1.677681 6.583385** 2.278909 1.094597 1.156003 3.286126 1.769241 10.84112*** 1.295698 1.381592 5.507432* 6.015007** 6.83891** 8.204443** 5.786174* 2.562723 3.965457 3.46726 4.5094 2.732502 3.436728 0.079384 3.143417 2.339001 4.36508 0.562589 2.092181 2.403572 Probability 0.0462 0.1461 0.6553 0.0193 0.0902 0.5858 0.2588 0.2135 0.1116 0.6046 0.3876 0.2934 0.1618 0.1904 0.4322 0.0372 0.3200 0.5785 0.5610 0.1934 0.4129 0.0044 0.5232 0.5012 0.0637 0.0494 0.0327 0.0165 0.0554 0.2777 0.1377 0.1766 0.1049 0.2551 0.1794 0.9611 0.2077 0.3105 0.1128 0.7548 0.3513 0.3007 Jan/00 to Jan/14 Chi-squared stat. 2.398278 4.259484 4.60438 3.091037 5.31076* 0.10512 1.789301 1.372824 2.83995 0.121352 1.39589 0.150036 1.132524 1.270545 3.620714 3.917147 2.699275 0.574953 2.022524 1.974963 0.153304 9.747623*** 4.058175 0.142817 1.32686 3.117297 7.672407** 2.03058 1.590946 1.68379 2.861353 2.842628 4.844266* 4.03347 1.040747 0.244582 2.063913 1.596974 4.590128 3.066883 1.747912 4.311002 Probability 0.3015 0.1189 0.1000 0.2132 0.0703 0.9488 0.4088 0.5034 0.2417 0.9411 0.4976 0.9277 0.5676 0.5298 0.1636 0.1411 0.2593 0.7502 0.3638 0.3725 0.9262 0.0076 0.1315 0.9311 0.5151 0.2104 0.0216 0.3623 0.4514 0.4309 0.2391 0.2414 0.0887 0.1331 0.5943 0.8849 0.3563 0.4500 0.1008 0.2158 0.4173 0.1158 Notes: The null hypothesis for Granger causality tests for non-causation and is stated in the form , where ― /→‖ stands for ―does not Granger cause‖. *** statistically significant at 1% , ** statistically significant at 5%, *statistically significant at 10% 16