The two-component Beta- t -QVAR-M-lev: a new forecasting model

Published version
Repository DOI

Change log
Haddad, Michel Ferreira Cardia  ORCID logo
Blazsek, Szabolcs 
Arestis, Philip 
Fuerst, Franz 
Sheng, Hsia Hua 

We introduce a new joint model of expected return and volatility forecasting, namely the two-component Beta-t-QVAR-M-lev (quasi-vector autoregression in-mean with leverage). The maximum likelihood estimator for the two-component Beta-t-QVAR-M-lev is an extension of theoretical results of the one-component Beta-t-QVAR-M. We compare the volatility forecasting performance of the two-component Beta-t-QVAR-M-lev and two-component GARCH-M (generalized autoregressive conditional heteroscedasticity), also considering their one-component frameworks. The results for G20 stock market indices indicate that the forecasting performance of the two-component Beta-t-QVAR-M-lev is superior compared with the two-component GARCH-M and their one-component versions.


Acknowledgements: We thank Alfred Hero, Andrew Harvey, Christian Hafner, Matthew Copley, Peter Hansen, Rutger-Jan Lange, the journal editor (Markus Schmid) and two anonymous referees for their helpful comments. All remaining errors are our own. In addition, Michel F. C. Haddad acknowledges funding from the Coordination for the Improvement of Higher Education Personnel of Brazil (CAPES) and Cambridge Commonwealth, European and International Trust (Grant BEX 2220/15-6), and Szabolcs Blazsek acknowledges funding from Universidad Francisco Marroquín.

Funder: Universidad Francisco Marroquín

C52, C58, Dynamic volatility models, C32, F21, G15, Dynamic conditional score (DCS), Generalized autoregressive score (GAS), G20, Volatility forecasting
Journal Title
Financial Markets and Portfolio Management
Conference Name
Journal ISSN
Volume Title
Springer US
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (BEX 2220/15-6)