A0510
Title: Matrix GARCH model: Inference and applications
Authors: Cheng Yu - Tsinghua University (China) [presenting]
Ke Zhu - University of Hong Kong (Hong Kong)
Feiyu Jiang - Fudan University (China)
Dong Li - Tsinghua University (China)
Abstract: Matrix-variate time series data are largely available in plenty of applications. However, no attempt has been made to study their conditional heteroskedasticity, which is often observed in economic and financial data. To fill the gap, a new matrix generalized autoregressive conditional heteroskedasticity (GARCH) model is proposed, which can capture the dynamics of conditional row and column covariance matrices of the matrix time series. The key element of the matrix GARCH model is a univariate GARCH-type specification for the trace of conditional row or column covariance matrix since the conditional row, and column covariance matrices can not be identified without this trace specification. Moreover, the quasi-maximum likelihood estimator is proposed for the model, and the portmanteau tests are constructed for model diagnostic checking. To handle the large dimensional matrix time series, a matrix factor GARCH model is further raised. Finally, three applications are given on credit default swap prices, global stock sector indices, and future prices to demonstrate the advantage of the matrix GARCH and matrix factor GARCH models over the existing multivariate GARCH-type models in risk management and portfolio allocation.
