Working Paper nº 1426 August, PDF

Description
Instituto Complutense de Análisis Económico Volatility Spillovers from Australia's major trading partners across the GFC David E. Allen School of Mathematics and Statistics, the University of Sydney, and

Please download to get full document.

View again

of 26
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information
Category:

Recruiting & HR

Publish on:

Views: 18 | Pages: 26

Extension: PDF | Download: 0

Share
Transcript
Instituto Complutense de Análisis Económico Volatility Spillovers from Australia's major trading partners across the GFC David E. Allen School of Mathematics and Statistics, the University of Sydney, and Center for Applied Financial Studies, University of South Australia Michael McAleer Econometric Institute, Erasmus School of Economics, Erasmus University Rotterdam, Tinbergen Institute, The Netherlands, Department of Quantitative Finance, College of Technology Management, National Tsing HuaUniversity Hsinchu, Taiwan, Distinguished Chair Professor, College of Management, National Chung Hsing University, Taichung, Taiwan, Department of Quantitative Economics, Complutense University of Madrid, Spain, Abhay K. Singh School of Accounting, Finance and Economics, Edith Cowan University, Australia Abstract This paper features an analysis of volatility spillover effects from Australia's major trading partners, namely, China, Japan, Korea and the United States, for a period running from 12th September 2002 to 9th September This captures the impact of the Global Financial Crisis (GFC). These markets are represented by the following major indices: The Shanghai composite and the Hangseng. (In the case of China, as both China and Hong Kong appear in Australian trade statistics), the S&P500 index, the Nikkei225 and the Kospi index. We apply the Diebold and Yilmaz (2009) Spillover Index, constructed in a VAR framework, to assess spillovers across these markets in returns and in volatilities. The analysis conrms that the US and Hong Kong markets have the greatest inuence on the Australian one. We then move to a GARCH framework to apply further analysis and apply a tri-variate Cholesky-GARCH model to explore the eects from the US and Chinese market, as represented by the Hang Seng Index. Keywords: Volatility Spillover Index, VAR analysis, Variance Decomposition, Cholesky-GARCH JL Classification G11, C02. Working Paper nº 1426 August, 2014 UNIVERSIDAD COMPLUTENSE MADRID ISSN: WEB DE LA COLECCIÓN: Copyright 2013, 2014 by ICAE. Working papers are in draft form and are distributed for discussion. It may not be reproduced without permission of the author/s. Volatility Spillovers from Australia's major trading partners across the GFC David E. Allen a,, Michael McAleer b, Robert J. Powell c, and Abhay K. Singh c a School of Mathematics and Statistics, the University of Sydney, and Center for Applied Financial Studies, University of South Australia b Econometric Institute, Erasmus School of Economics, Erasmus University Rotterdam, Tinbergen Institute, The Netherlands, Department of Quantitative Finance, College of Technology Management, National Tsing Hua University Hsinchu, Taiwan, Distinguished Chair Professor, College of Management, National Chung Hsing University, Taichung, Taiwan, Department of Quantitative Economics, Complutense University of Madrid, Spain, c School of Accounting, Finance and Economics, Edith Cowan University, Australia Abstract This paper features an analysis of volatility spillover eects from Australia's major trading partners, namely, China, Japan, Korea and the United States, for a period running from 12th September 2002 to 9th September This captures the impact of the Global Financial Crisis (GFC). These markets are represented by the following major indices: The Shanghai composite and the Hangseng. (in the case of China, as both China and Hong Kong appear in Australian trade statistics), the S&P500 index, the Nikkei225 and the Kospi index. We apply the Diebold and Yilmaz (2009) Spillover Index, constructed in a VAR framework, to assess spillovers across these markets in returns and in volatilities. The analysis conrms that the US and Hong Kong markets have the greatest inuence on the Australian one. We then move to a GARCH framework to apply further analysis and apply a tri-variate Cholesky-GARCH model to explore the eects from the US and Chinese market, as represented by the Hang Seng Index. Keywords: Volatility Spillover Index, VAR analysis, Variance Decomposition, Cholesky-GARCH 1. Introduction The Global Financial Crisis (GFC) had a major impact on the world's nancial markets. This paper examines whether there is evidence of spillovers of volatility from Australia's main trading partners, namely, China, Japan, Korea and the United States, for a period running from 12th September 2002 to 9th September 2012, to the Australia stock market. The paper features an application of Diebold and Yilmaz's (2009) Spillover Index model, to assess the impact of the GFC on spillovers to the Australian market, on both returns and volatility series. This is followed by an application of a Cholesky-GARCH trivariate model to directly model the inuence of both the US Corresponding author. Acknowledgements: For nancial support, the rst author acknowledges the Australian Research Council, and the third author is most grateful to the Australian Research Council, National Science Council, Taiwan, and the Japan Society for the Promotion of Science.We are grateful to the anonymous reviewers for helpful comments. address: (David E. Allen) Preprint submitted to Elsevier August 13, 2014 and Chinese markets, as represented by Hong Kong and the Hang Seng Index, not Shanghai and the Shanghai Composite Index, because the Spillover Index analysis, conducted in a VAR and variance decomposition framework, reveals that these two markets are the most inuential on the Australian market, even though they are currently ranked fourth and fth, in importance, as trading partners, (see Figure 2). The recent GFC crisis commenced in 2007 and continued through to the European sovereign debt crisis. Alan Greenspan (2010) took the view that: The bubble started to unravel in the Summer of But unlike the debt-like deation of the earlier dotcom boom, heavy leveraging set o serial defaults, culminating in what is likely to be viewed as the most virulent nancial crisis ever. The major failure of both private risk management and ocial regulation was to signicantly misjudge the size of tail risks that were exposed in the aftermath of the Lehman default. The U.S. subprime mortgage and credit crisis was characterized by turbulence that spread from subprime mortgage markets to credit markets more generally, and then to short-term interbank markets as liquidity evaporated, particularly in structured credit then on to stock markets globally. Gorton (2010) suggested that the GFC was not particularly dierent from previous crises except that, prior to 2007, most investors had never heard of the markets that were involved. Concepts such as subprime mortgages, asset-backed commercial paper conduits, structured investment vehicles, credit derivatives, securitization, or repo markets were not common knowledge. Gorton (2010) suggests that the securitized banking system is a real banking system that is still vulnerable to a panic. He argues that the crisis, beginning in August 2007, can best be understood as a wholesale panic involving institutions, where large nancial rms, ran on other nancial rms, making the system insolvent. Fidrmuc and Korhonen (2010), analyze the transmission of global nancial crisis to business cycles in China and India using GDP data and dynamic correlation analysis. They report a signicant link between trade ties and dynamic correlations of GDP growth rates in emerging Asian countries and OECD countries. Cheng, and Glascock (2005), examine the linkages among three Greater China Economic Area (GCEA) stock markets, including Mainland China, Hong Kong, and Taiwan, and two developed markets, Japan and the United States. They nd that a random walk model is outpredicted by an autoregressive GARCH model, and an ARIMA model, in all three GCEA markets, and that there is no evidence of cointegration between the markets. Chung et al. (2010), examine the informational role of the TED spread as perceived credit risk. They apply a Vector Autoregressive (VAR) model, Granger causality tests, cointegrating Vector Error Correction Model (VECM), to analyse the leadership of the US market with respect to UK, Hong Kong, Japan, Australia, Russia and China markets, during the crisis, and nd evidence of increased interdependence during the crisis. They suggest that the impact of orthogonalized shocks from the US market, on other global markets, increases by at least two times during the crisis, and that of the TED spread, even more so. Didier et al. (2012), examine the determinants of comovement in stock market returns during the crisis. They explore the inuence of the United States (US), via analysis of the factors driving the comovement between US stock market returns and stock market returns in 83 countries. Their analysis distinguishes between the period before and after the collapse of Lehman Brothers, and their ndings indicate that comovement was driven largely by nancial linkages. Dooley and Hutchison (2009) explore the transmission of the crisis to the emerging markets. They suggest that whilst initially these markets were largely shielded from the deleterious eects on world trade ows, they subsequently had strong eects after the Lehmann bankruptcy. Huang et al. (2000), apply similar causality and cointegration relationships among the stock markets of the United States, 2 Japan and the South China Growth Triangle (SCGT) region, and report no evidence of cointegration amongst these markets, save the Chinese ones of Shanghai and Shenzen. Kotkatvuori-Örnberg et al. (2013), explore stock market correlations during the nancial crisis across 50 equity markets. They measure the value of covariance information using an the augmented DCC model, and show that by taking into account the change in the level of variance in high volatility periods, the estimates of the conditional covariance are more ecient in capturing the dynamics of the stock market's variance. Min and Hwang (2012), also use dynamic correlation analysis of US nancial crisis to explore contagion from the US across four OECD countries. They nd a process of increasing correlations (contagion), in the rst phase of the US nancial crisis, and an additional increase of correlations (herding), during the second phase of the US nancial crisis, for the UK, Australia and Switzerland. Mun and Brooks (2012) explore the roles of news and volatility in stock market correlations during the global nancial crisis. Their results show that the majority of the correlations are more strongly explained by volatility than news. Yeh and Lee (2000), analyse the interaction and volatility asymmetry of unexpected returns in the greater China stock markets. They suggest that results, of a near vector autoregressions (VAR) model, reveal that the Hong Kong stock market plays an inuential role as a regional force amongst the Taiwan, Shanghai, and Shenzhen B-share stock markets. There are a number of common themes in this literature. The global studies that include the US suggest that it has a major inuence in the transmission of shocks to both developed and undeveloped markets. The econometric methods, used in these studies, range across a number of time series econometrics techniques including cointegration, VAR models, and applications of models nested in the GARCH framework, including multivariate models such as DCC. However, also germane to the approach adopted in the current paper, is a recent study by Diebold and Yilmaz (2009), who formulate and examine precise and separate measures of return spillovers and volatility spillovers. They base their measurement of return and volatility spillovers on vector autoregressive (VAR) models, in the broad tradition of Engle et al. (1990). They focus on variance decompositions, which they argue are well understood and widely calculated. They use them to aggregate spillover eects across markets, which permits the distillation of a wealth of information into a single spillover measure. We adopt their approach, plus a tri-variate Cholesky- GARCH model which permits an analysis of the relationships with those markets with the greatest inuence on the Australian market, as revealed by the application of their Spillover Index. Our analysis is therefore focussed on the relationship with Australia's main trading partners, China, Japan, the United States, Hong Kong and Korea, given that both trade ows and attached information ows are likely to have an economic impact, and be reected in the behaviour of equity indices (see for example, Evans and Hnatkovska (2014)). In this paper we focus on how the GFC impacted on volatility spillovers across the world to the Australian equity market. Even though the Australian nancial markets were spared the major eects of the GFC, in terms of distress to major nancial institutions, the Australian nancial market was still impacted by these major global events. The degree to which the Australian market is inuenced by extreme events in the US, has implications for portfolio optimization by Australian investors and fund managers alike, and eects the degree to which it is possible to hedge risk during times of nancial turbulence. We examine how both return and volatility spillovers and correlations changed, between the Australian market and the US during the nancial crisis. We focus on the impact of the Chinese market in our analyses, given that its is Australia's main trading partner, together with the inuence of the US market, given that this it has consistently been shown to have the greatest impact on other global markets, in the prior studies mentioned, even though it is only 3 4 Australia's fourth most signicant trading partner. 2. Research Method 2.1. Data set and econometric models We wanted to make sure that we captured the relationships with Australia's main trading partners. Figure 1 below depicts Australia's top ten partners in trades in goods and services in 2013 taken from the DFAT top ten list. Figure 1: Top 10 partners in trade in goods and services 2013 Source: Figure 2 depicts the trend in trade in goods and services from 2001 to 2013, with exports in the top half of the diagram and imports in the bottom. It can be seen that trade with China has become ever more important overtaking trade with Japan as the major trading partner in Trade with Korea, particularly in imports, has increased in importance since 2004 when it overtook trade with the US in relative importance. The fth most important trading partner is Hong Kong whose relative importance has not changed over this period. For the purposes of our analysis we have concentrated on the top six trading partners, in terms of exports, as this reects our major trading partners and has implications for export income. 2.1 Data set and econometric models 5 Figure 2: Australia's major trading partners Source: Department of Foreign Aairs and Trade. We took a series of major stock market indices representing these six countries; namely the Shanghai Stock Exchange (SSE) composite index, the Hang Seng Index, the Australian All Ordinaries Index, the Nikkei 225 Index, the S&P500 Index and the Kospi Index. The data set includes daily data for each index from 1st January 2004, until 30th June The indexes are total market indexes, based on market capitalizations, and are taken from Datastream standardised in US dollar terms. Daily returns are calculated as follows: y it = ln(p it ) ln(p it 1 ) (1) The data sets used are shown in Table 1. (We lagged the US S&P 500 index returns by one day to make them more comparable in time with the Australian and other Asian series). Country Index USA S&P500 AUSTRALIA All Ordinaries CHINA Hang Seng CHINA Shanghai Stock Exchange Composite JAPAN Nikkei 225 KOREA Kospi Table 1: List of countries and indices There are a variety of models that could be used to test for the existence of time-varying volatility, and for spillover eects in returns and volatility across markets. 2.2 The Diebold and Yilmaz (2009) Spillover Index 6 One approach is to use a time series Vector Autoregressive (VAR) framework. This was recently formalised by Diebold and Yilmaz (2009) and Diebold and Yilmaz (2012), who modelled spillovers using VAR models and variance decompositions. They constructed a spillover index based both on return spillovers and volatility spillovers. Their approach is initially attractive for our purposes, because it enables us to see which of the six markets considered makes the largest contribution to the spillover of returns and volatilities into the Australian market. We commence with an application of their model which enables us to determine which of Australia's trading partners contributes most to equity shocks. Once we have used their model as an initial lter we then proceed to apply models within a GARCH framework. The returns spillover index uses our basic index data. For the volatility based estimates we departed from the procedure used by Diebold and Yilmaz (2009), which used weekly ranged-based estimates to assess volatility. We preferred to use realised volatility metrics. We employ the the Oxford-Man Institute of Quantitative Finance's realised library which contains daily nonparametric measures of how the volatility of nancial assets or indexes were in the past. Each day's volatility measure depends solely on nancial data from that day. We choose the series constructed by sampling at 10 minute intervals within the day. Data is available for download on the Oxford- Man website (http://realized.oxford-man.ox.ac.uk/), and we took daily estimates for all our markets which were available for the full sample period, apart from the Shanghai Stock Exchange composite index which is not covered. The raw high frequency data is taken from Reuters DataScope Tick History database. These RV estimates were then utilised for the spillover index analysis of volatility The Diebold and Yilmaz (2009) Spillover Index Diebold and Yilmaz (2009) suggest that the advantage of the adoption of a VAR framework and the use of variance decompositions is that they permit the aggregation of spillover eects across markets, distilling a wealth of information into a single spillover measure. This suits our current purposes and permits an examination of the relative contributions to spillovers made by the six markets in our sample. They proceed to develop their measure by taking each asset i, and adding the shares of its forecast error variance coming from shocks to asset j, for all j i, all in the context of an n variable VAR. They then sum these error variances across all i = 1,..., N. If we take the case of a covariance stationary, rst-order, two variable VAR, we have; x t = Φx t 1 + ε t where x t = (x 1t, x 2t ) and Φ is a 2 2 parameter matrix. In the empirical analysis which follows, x will be either a vector of index returns or a vector of index volatilities. The moving average representation of the VAR can be written, given the existence of covariance stationarity, as; x t = Θ(L)ε t where Θ(L) = (1 ΦL) 1. The moving average representation can be conveniently written as; x t = A(L)u t where A(L) = Θ(L)Q 1 t, u t = Q t ε t,e(u t u t) = I, and Q 1 t is the unique lower-triangular Choleski factor of the covariance matrix of ε t. Diebold and Yilmaz (2009) then proceed to consider the optimal 1 step ahead forecast, given by; 2.3 GARCH models 7 x t+1,t = Φx t, with the corresponding one-step ahead error vector [ a0,11 a e t+1,t = x t+1 x t+1,t = A 0 u t+1 = 0,12 a 0,21 a 0,22 which has the covariance matrix; ( ) E e t+1,t e t+1,t = A 0 A 0 ] [ ] u1,t+1 u 2,t+1 This suggests that the variance of a one-step ahead error in forecasting x 1,t is a 2 0,11 + a 2 0,12 and the variance of the one-step ahead error in forecasting x 2,t is a 2 0,21 + a 2 0,22. Diebold and Yalmiz (2009) demonstrate that it is possible to to split the forecast error variances of each variable into components attributable to the various system shocks. In this two variable system it is possible to distinguish between shocks to t
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x