Student-t distribution based VAR-MGARCH: an application of the DCC model on international portfolio risk management

Abstract

Significant second-moment transmission effects and obvious time-varying patterns of correlation coefficients among major equity and currency markets in the US, Japan and the UK are found to exist. Such observations inspire the time-varying setting of dynamic conditional correlation coefficients in MGARCH models. On the other hand, the multivariate Student-t distribution is suitable for analysing the visible leptokurtosis that is common in financial markets. Both are important for international portfolio risk management. Thus, a comparison on the hedging efficiency of hypothetical portfolios consisting of stock and currency future positions is conducted in order to justify the multivariate Student-t distribution based on the DCC-MGARCH model.

Notes

¹ There are numerous studies devoted to this issue. Please refer to Black (1974), Stulz (1981), Solnik (1984), Cho et al . (1986), Korajczyk and Viallet (1989), Ferson and Harvey (1994), Dumas and Solnik (1995) and Fama (1998), among others.

² For example, please refer to Ajayi and Friedman (1998), Kanas (2000), Nasseh and Strauss (2000) and Kim (2003).

³ Please refer to King and Wadhwani (1990), Booth et al . (1997) and Kanas (2000).

⁴ For instance, Fama (1965) claims that a stable Paretian distribution with a characteristic exponent less than two can be used to replace the normal setting. Kon (1984) and Hull and White (1998) assert that a discrete mixture of normal distributions can be used to explain the observed patterns of significant kurtosis and a positive skewness of daily data.

⁵ The detailed ADF test results for different specifications are not presented here for parsimony's sake. They are available from the authors on request.

⁶ The detailed statistical results are not presented here for the sake of parsimony. They are available from the authors on request.

⁷ For example, please refer to Hsuku et al . (2006).