A Way of Estimating Time-varying Covariance Matrix of FX Returns with Heavy-tailed Distribution
Estimating time-varying covariance matrix of FX returns is an important component of current practice in FX risk management.Because t distribution can capture the heavy-tailed quality of series of FX returns,the paper introduces the EM algorithm to estimate covariance matrix under multivariate t distribution.And we compare the model with others models using daily data of CAD-USD,GBP-USD,and CHF-USD exchange rate.The result indicates the model using EM algorithm has explicitly iterative formulae and the multivariate GARCH (1,1) model can provide superior estimation of timevarying covariance matrix of FX returns having fat-tailed distributions.However,the model using EM algorithm is simpler and more stable than the multivariate GARCH (1,1) model.
Time-varying Covariance Matrix Heavy-tailed Quality EM algorithm Multivariate t Distribution Multivariate GARCH (1,1) Model
Rongda Chen Jianbo Liu Yi Lv
School of Finance,Zhejiang University of Finance and Economics,Hangzhou
国际会议
长沙
英文
763-770
2008-10-28（万方平台首次上网日期，不代表论文的发表时间）