The recurrence of global financial crisis attracts academic attention of volatility transmission in international stock markets. This project strives to produce a thorough and in-depth review of extant literatures in this field. This topic is to be investigated from three perspectives, theoretical foundation, econometric models and empirical findings. Firstly, economic factors, information transmission and contagion are explained to pave the theoretical foundation of international transmission. Secondly, the evolution of relevant econometric models is reviewed with specific emphasis on the ARCH family and data related issues. Thirdly, empirical evidence is provided in terms of transmission direction, leverage effect and transmission during crisis. The final conclusions indicate directions for further research.
The objective of this project is to present a review of existing literatures regarding cross-border volatility spillover. Although word limit obstructs an exhaustive review of the abundant literature, three strands of papers are investigated.
The theoretical rationale (first part of body) provides underlying logic of international volatility transmission. The visible actors, economic fundamental factors, include international trade and investment links, exchange rate regime and stochastic policy coordination. Invisible Information transmission also accounts for volatility spillover. In this case, volatility transmission shows an asymmetric property, which is dubbed "leverage effect". Market contagion is used to explain excessive transmission during crises.
The second part of review concentrates on the realm of econometric methods. GARCH is effective in modeling volatility transmission, but it cannot incorporate asymmetric transmission. EGARCH model is thus adapted to account for leverage effect. Multi-variate GARCH models also excel in grasping the asymmetric feature, but they involve too many parameters. Simplified versions are developed. The selection of data frequency is also discussed with academic implication.
Empirical findings (the third part of body) vary across different markets using various data and methods. The spillover direction is under examination. Evidence show that inter-transmission exists among developed countries. However, the transmission between developed and emerging markets is ambiguous both in existence and in direction. The leverage effect is tested with real world data. The pattern of volatility transmission in turbulent episodes indicates contradicting results between 1987 Crash and 1997 Asia Crisis. Empirical findings seem to dovetail with theoretical anticipation that the transmission does show asymmetric response to heterogeneous information.
Finally, the project identifies three trends for future research: (1) establish micro foundation; (2) improve econometric models; (3) investigate the recent crisis.
1. Introduction
Recently the global economy is suffering from an unprecedented financial crisis. The recurrence of world-wide financial disturbance stimulates the appetites of both academics and practitioners for exploring the mechanism for international transmission. As early as the 1987 Crash, scholars started to show substantial interest in investigating the spillover of stock returns and volatility across world markets (Bennett and Kelleher 1988; King and Wadhwani 1989). The financial globalisation leads to an intimate interrelationship across international stock markets, due to rapid technology progress and excessive capital flows (Fernandez-Izquierdo and Lafuente 2004). The co-movements in international stock markets have been documented by numerous scholars (Roll 1989; King, Sentana et al. 1994; Kearney 2000).
The terms of co-movement, co-integration, inter-dependence, international linkage etc. are too vague and inaccurate themselves for academic studies. The more appropriate phrasing is to supplement these terms with the order of moment. Two major strands of literatures exist, focusing respectively on transmission of return (first moment) and of volatility (second moment). Despite the abundant researches regarding the first moment transmission (Hilliard 1979; Eun and Shim 1989; Becker, Finnerty et al. 1990), the focus of this essay remains on the second moment transmission. In fact, some markets manifest more interdependence in volatility than in return (Soriano and Climent 2005).
This essay, by means of current literatures, explains the mechanism and manner of volatility transmission in international stock markets and explores the econometric methods to model such transmission. Studying volatility transmission is immensely valuable in academic and practice. It has profound implications, inter alia for optimal asset location, corporate capital budgeting and international policy coordination. (Billio and Pelizzon 2003; Connolly and Wang 2003)
The remaining parts of this essay endeavor to review extant literatures concerning volatility transmission in international stock markets. Section 2 generalises the theoretical rationales which can explain the transmission mechanism. Section 3 introduces and critically comments on the econometric methods employed in empirical studies. The results of these empirical studies are presented in Section 4. Section 5 concludes.
2. Theoretical Rationale
There is a growing consensus that volatility transmission does exist between different and even geographically distant national stock markets. Section 4 will delve into minutiae the empirical evidences. This section aims to elaborate the latent logic of transmission.
The source of stock market volatility has always consumed the thoughts of researchers. Domestic economic factors and business cycles are always under thorough securitization (Hamiltion and Lin 1996; Engle, Ghysels et al. 2008). In the era of financial globalisation, especially after several earthshaking financial crises in 1990s and this century, researchers begin to realise that the international transmission is an innegligible source of volatility.
2.1 Economic Fundamental Factors
The international capital asset pricing model (Merton 1973) is among the first theoretical frameworks to link the domestic market with global market. The traditional beta in this case represents the covariance with the global stock market. The construction and restructuring of international portfolios involve significant amount of cross-border capital flow. The role of capital flow is also supported by Ramchand and Sumel (1998).
The linkage between stock market can also be perceived as an extension of economic integration. The stock market is deemed as the barometer for national economy, so that the co-movement in two economies (usually trade partners) will also be exhibited in the stock market behaviour. Lin, Engle, et al. (1994) attribute cross-border spillover of stock return and volatility to trade and investment ties, increasing financial market integration, and market contagion. Real exchange rate, the most important variable in bilateral economies, is thought to be a potential source of stock market volatility (Bekaert and Hodrick 1992).
The concept of stochastic policy coordination is employed by Ito, Engle, et al. (1992) and Leachman and Francis (1996) to account for the international transmission. When central banks in two countries simultaneously choose similar monetary policy, say increase interest rate, the stock markets in two nations may show similar form of volatility.
2.2 Information Transmission
The explanation with economic fundamentals is marginally acceptable in explicating why global stock markets fell simultaneously and with such surprising uniformity as the 1987 Crash (King and Wadhwani 1989). A considerable number of literatures stress the dissemination of information across international stock markets. Ross (1989) illustrates that price volatility is a direct result of information flow. The information affecting traders' behaviour is generated not only domestically but also from foreign stock markets (Koutmos and Booth 1995). A special transmission channel of information is the presence of stocks which are listed in two or more capital markets (Bae, Cha et al. 1999). The introduction of information transmission can well explain the fact that there is volatility transmission between two distant markets, regardless of geographical location, since the transmission of information is almost costless and exceptionally rapid. Wongswan (2003) defines information as important macroeconomic announcement. Information is more academically termed as innovation from asset pricing models.
A distinctive feature of information transmission is leverage effect (Black 1976; Christie 1982), referring to the phenomenon that stock volatility is asymmetrically influenced by good news and bad news. Negative information of stock devaluation increases debt-to-equity ratio along with financial risk, which in term lowers stock price, causing additional volatility. As the saying goes, "bad news travels fast", when bad new is disseminated, the entire international stock market is more volatile than when good news is generated. This is in consistence with the observation that volatility transmission is more intense in turmoil than in tranquil periods (Bennet and Kelleher 1988; Koutmos and Booth 1995; Arouri, Bellalah et al. 2008).
2.3 Market Contagion in Financial Crises
The 1987 Crash, 1997 Asian Financial Crisis and the recent Credit Crunch have witnessed a roller-coaster pattern of movement in many national stock markets. This sudden drastic oscillation across international stock markets is vividly labeled as a "meteor shower" by Engle, Ito et al (1990). The excessive volatility in time of crisis cannot be fully explained either by the fundamentals or by information. This effect is called market contagion by Ito and Lin (1993). It is certainly different from normal transmission or interdependence (Forbes and Rigobon 2002; Soriano and Climent 2005). There are divergent opinions on the accurate definition of market contagion. However, market contagion is empirically measured as the unanticipated transmission of shocks (Beirne, Caporale et al. 2008). Karolyi (2003) compares financial crises contagion with epidemic spread and postulates the "disease" carriers to be the irrational behaviours (i.e. panic, herding, and loss of confidence) of market participants (i.e. commercial banks, global mutual funds and speculators).
3. Econometric Methods
3.1 Econometric Models:
A theory is only a hypothesis before being proved empirically. Econometric models are the key to the proof.
Vector Auto-Regression model provides byproducts of variance decomposition and impulse response function, which is helpful in studying spillover. Diebold and Yilmaz (2008) utilises this framework to formulate a statistical measure of Spillover Index. However, more empirical studies follow ARCH-type models, first invented by Engle (1982). Bollerslev (1986) develops it into Generalised ARCH model with wider applications.
Hamao, Masulis et al (1990) construct a two stages uni-variate GARCH model. First GARCH (1, 1) is applied to each stock index individually to generate a series of estimated residual. In the second stage the squared residuals (estimation of volatility) are used for linear regression. Leachman and Francis (1996) makes a little modification of the second step. They use VAR instead of linear regression. The same strategy is also employed by Christofi and Pericli (1999)
Despite the excellent econometric properties and explanation power, the traditional GARCH fails to model one the most important features in the financial market, the leverage effect as mentioned above. Nelson (1991) therefore constructs an EGARCH model in an effort to simulate the asymmetric transmission. Another advantage of EGARCH is that it does not require its coefficients to be non-negative.
Multi-variate GARCH models are excellent in exploring the dynamics of volatility in the international stock markets. Bi-variate GARCH can be used to test transmission pair-wise countries, e.g. Choudhry (2004). For a system of more than three markets, multi-variate GARCH is appropriate. The only problem with this model is the long queue of coefficients to be estimated. Efforts have been made towards simplification while retaining asymmetry, including BEKK-GARCH model (Engle and Kroner 1995) and later Dynamic Conditional Correlation GARCH model (Engle 2002).
The burgeoning literatures in econometric methods have equipped researchers with modern tools for quantitative analysis. The new econometric weapons lend support to the understanding of volatility transmission as is shown in the cases of Flexible Least Squares (He 2001), Marcov Regime Switching Models (Gallo and Otranto 2005), Stochastic Volatility Models (Tanizaki and Hamori 2009), Wavelet Analysis (Lee 2002).
3.2 Data Related Issues
Selection of data frequency is an exquisite issue in empirical research. Both high frequency (monthly) and low frequency (daily or even intra-daily) have been used by authors with various research objectives. Glosten, Jagannathan et al (1993) suggest that the statistical properties of monthly data and daily data are unequivocally divergent.
Leachman and Francis (1996) advocate that monthly data can evade the possible bias produced by bid-ask effect and non-synchronous trading. More importantly, monthly data help elaborate the transmission of underlying fundamentals across international stock markets. He (2001) also adopts the monthly data, because he invites into his model some macro variables which are not available on a daily frequency.
Eun and Shim (1989) prefers daily data to monthly or even weekly data, because the latter may fail to capture interactions that only last for days. Susmel and Engle (1994) investigates the hourly data of New York and London. Volatility clustering is more significant when using high frequency data than low frequency (Bollerslev 1986). More studies resort to daily data for empirical investigation to capture the prompt market response towards news (Chinzara 2008).
Another tricky problem arises from the high frequency data is that the stock markets are probably located in diverse time zones. They have different opening and closing time and in particular cases the opening hours do not overlap. The most common tri-market model includes New York, London and Tokyo (Koutmos and Booth 1995). There is no time overlap between Tokyo and London, and Tokyo and New York, but there is approximately two hours' overlap between London and New York. There will be a bias, not to take into consideration the excess volatility transmission during overlap. Jeong (1999) uses 5-minute return data to explore the transmission pattern among US, UK and Canada in the overlapping hours and finds persistent bi-directional transmission.
4. Empirical Evidence
A remarkable amount of empirical evidence is unequivocally supportive of the idea that there is volatility transmission in international stock markets. The issues of more academic concern and practical implication are the pattern of the transmission.
4.1 Direction of Transmission
Volatility spillover is found to be uni-directional and bi-directional, depending on the sample and method. On the whole, transmission among developed countries tends to be reciprocal. Lin, Engle et al (1994) studies the interaction of the US and the Japanese stock markets and identifies bi-directional transmission of both return and volatility. Hamao, Masulis et al (1990) inspects a triple market scenario with uni-variate GARCH models and concludes that the volatility disseminates from New York to London and Japan and from London to Japan. Another triangular relation is found among UK, US and Canada (Theodossiou and Lee 1993), although US market is more significant 'exporter' of volatility. A more complex empirical study is carried out by Leachman and Francis (1996) who tests the inter-transmission among G-7 countries. The empirical findings indicate a cob-web relation among them.
The transmission mode between developed and emerging markets has also been examined. It is commonly believed that in this case developed markets act as a locomotive, driving the volatility in the latter. For example Ng (2000) analyzes the volatility spillover in the Pacific rim and detects evident volatility transmission from US (as a world factor) and Japan (as a regional factor) to six Asian-Pacific markets. However, Bala and Premaratne (2003) spot contrasting phenomenon, that small market (Singapore) does export volatility to large market (US). Egert and Kocenda (2007) probes the volatility transmission among two groups of European countries, i.e. developed countries (UK, France and Germany) and emerging markets (Hungary, Poland and Czech Republic) with DCC-GARCH model using 5-minute data. The transmission within the former group is evident, whereas the transmission neither between the two groups nor within the latter group is significantly observed.
4.2 Abnormal Transmission during Crisis
Transmission in time of crises has received growing attention. Hamao, Masulis et al (1990) test the volatility transmission during 1987 Crash by probing the open-close and close-open data. The pre-crash sample indicates that volatility spillovers from London and New York into Tokyo but not vice versa. However, during the Crash, the volatilities exhibit all-directional transmission. However, several studies (Sola, Spagnolo et al. 2002; Caporale, Pittis et al. 2006) regarding 1997 Crisis comes to a contradicting conclusion that volatility spillover is bi-directional in normal periods but uni-direction during crisis. This reverse directional change is worth further study with recent data.
4.3 Asymmetric Leverage Effect
Bae and Karolyi (1994) examines the asymmetric leverage effect on short-run volatility transmission between US and Japan and reveals that bad news in one country exerts a more considerable impact on the next-day volatility in the other country than the same amount of good news. Koutmos and Booth (1995)'s findings also dovetail with the anticipation. Brooks and Henry (2000) employs multi-variate GARCH model to US, Japan and Australia and obtains asymmetric variance-covariance matrix of returns, indicating that not only the size but also the sign of market innovation is influential to volatility transmission.
5. Summary and Conclusion
(close)The study of volatility transmission across international stock market is of enormous interest both theoretically and empirically. Particularly in view of current financial crisis, this topic has again aroused thick interest and there is a foreseeable increase of research in this field. This essay presents a timely review of existing literature.
The selected papers, ranging from 1976 to 2009, encompass most aspects of the study including theoretical, methodological and empirical. The empirical studies selected include samples from Asia, Africa, Europe, Latin America and North America. The review includes papers published in the best recognised financial journals, e.g. Journal of Finance, Journal of Financial Economics, Review of Financial Studies and Journal of Financial and Quantitative Analysis. Quite a few eminent authors are cited, including Nobel laureate Prof. Robert Engle. Therefore the review can be believed to be thorough and profound.
This review covers a substantial bulk of literatures, which can represent the main stream research in this field. The theoretical rationale for international spillover of volatility is concluded as economic fundamental factors, information transmissions and market contagion. The econometric models are dominantly ARCH type, with a few exceptions. The ARCH-type models are introduced in an evolutionary process. Empirical literatures show the direction of transmission between developed markets and between developed and emerging markets, and even under circumstance of financial crisis. (weakness)However the results are not completely identical. Leverage effective is unanimously confirmed by several studies.
(extent study and future trend)In accordance to the structure of this essay, the research of volatility transmission of international stock markets can be extended in three dimensions. (1) Explain the transmission mechanism in a micro level. Current theories mainly focus on the macro level, lacking in-depth discussion of the behaviour of investors and firms in volatility transmission. (2) Econometric methods proliferate at an amazing rate. It is possible to introduce the state-of-art econometric methods to study volatility transmission in a more accurate way (e.g. spectral analysis). (3) The current credit crunch provides a precious opportunity to empirically study the mode of volatility transmission in financial crisis. A comparison study with 1987 Crash and 1997 Asian Crisis is of immeasurable value for academics, investors and governments. The paucity of such work in these specifications will spur future research.
