Application of multivariate GARCH in the minimum variance optimization of a multi-currency sovereign bond portfolio

• Main Author: Hernandez, Anne Catherine Guevara
• Format: text
• Language: English
• Published: Animo Repository 2012
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/4280; https://animorepository.dlsu.edu.ph/context/etd_masteral/article/11118/viewcontent/CDTG005231_P.pdf
• Institution: De La Salle University

Fixed income portfolio managers are often challenged on how to maximize return and mitigate risk, especially during periods of increased volatility and occurrence of extreme events. Given the dynamic nature of risk, it is prudent to find a methodology where interest rate volatility can be forecasted to determine how to allocate the portfolio. The traditional mean-variance (M-V) optimization, which uses equally weighted time series data, considers the variance (volatility) of financial asset returns as constant or unconditional through time. In reality however, financial asset returns experience periods of relatively low volatilites followed by unusally high or persistent volatilities. To address this, the study models the conditional or time-varying nature of volatility through the use of a Multivariate Generalized Autoregressive Heteroscedasticity (MGARCH) Constant Conditional Correlation (CCC) (1,1) Model. Taking the point of view of Philippine banks managing a foreign currency deposit unit (FCDU), the study is a means to diversify in global markets by expanding the U.S. dollar portfolio to sovereign bond markets of Australia, Germany, and United Kingdom. Using MGARCH predicted volatilities in a monthly rebalancing period, the minimum variance and optimal portfolio allocations are determined using Markowitz mean-variance framework. For practical application, the Philippine central banks regulatory FCDU asset cover constraint is considered in the optimization process. Both unconstrained and constrained portfolios are presented. The concept of currency hedging is examined through the use of one-month foreign currency forward contracts, and assessing the results of hedged and unhedged portfolios based on overall risk and return. APPLICATION OF MULTIVARIATE GARCH IN THE MINIMUM VARIANCE OPTIMIZATION OF A MULTI-CURRENCY SOVEREIGN BOND PORTFOLIO 1-3 The impact of including a risk-free asset, represented by the one-month U.S. Treasury bill (T-Bill), in the multi-currency portfolio is also analyzed. Performance evaluation is conducted in two perspectives, the in-sample and out-of-sample analysis, where the in-sample covers the period of August 2001 until February 2009 and the out-of-sample covers the three year period starting March 2009 until February 2012. Results are analyzed using the average returns, standard deviations, and Sharpe ratios of the generated portfolios. Most of the MGARCH predicted variances had better forecasting performance than M-V historical variances. By using these variances, MGARCH portfolios were found superior to M-V portfolios. Results showed enhanced risk-return profiles through the use of conditional volatilities and through diversifying away from a pure U.S. dollar portfolio. Hedging was found beneficial for unconstrained portfolios. The cost of hedging played a significant role, as costly forward contracts were given a lower allocation. For constrained portfolios, it is interesting to note that currency hedging can be disregarded as the FCDU asset cover already protects the portfolio from currency risk exposure. The study examined different strategies to choose from, such as using traditional mean-variance estimates or using MGARCH predicted volatilites, using the minimum variance allocation or using the optimal allocation strategy, hedging currency risk or leaving foreign currency exposure unhedged, being limited to regulatory constraints or having no constraints, and deciding between a portfolio of risky assets or a portfolio of risky assets and a risk-free asset. The best performing portfolios are discussed and presented with recommended portfolio weights for practical use.