A Cholesky-MIDAS model for predicting stock portfolio volatility
Thursday 10 December 2009, 16:00
Lecture Theatre 11, LUMS
We are pleased to welcome a special guest speaker to our seminar series : Dr. Ralf Becker, The University of Manchester
Forecasting multivariate variance-covariance matrices (VCM) has gained significant prominence in recent years. In particular variants of the Dynamic Conditional Correlation models have enabled researchers to produce valid VCM forecasts for small to medium dimensional models.
This paper follows a line of research that attempts to harness the power of intra-day price information and the resulting availability of VCM proxies based on these, in the context of forecasting. This line of research has proven promising in the univariate framework but has not yet been firmly established in a multivariate framework. As demonstrated by Chiric and Voev (2010), it is beneficial to generate forecasts of the Cholesky Decomposition of VCM. This paper explores whether a simple modelling strategy, based on using mixed-frequency data (as in MIDAS) models, can provide a simple and feasible modelling strategy. Initial results are promising.
