EEG - Room 0.31 and Online
Online via E-mail
In this paper, we propose a multiplicative nonstationary volatility model allowing nonlinear behaviour driven by exogenous information. The new model extends the time-varying GARCH model of Amado and Teräsvirta (2013, 2017) by including an additional stochastic variable to allow the conditional variance to change smoothly between regimes. Modelling strategies for the proposed model are developed and they rely on statistical inference. The estimation of the model is simplified by employing maximisation by parts and the asymptotic properties of the proposed estimators are also studied. Finite-sample properties of these procedures and statistical hypothesis tests are examined by simulation. An empirical application illustrates the functioning of the model in practice.