Modelling and forecasting volatility in Applied Econometrics

Here, I investigate a variety of volatility models with varying parameters. Then compare two new Conditional heteroscedasticity models with varying parameters. In the first proposed model, I allow the parameters, which are attached to the lagged squared error term and lagged conditional variance term in the conditional variance under a GARCH specification, to be time-varying. The second proposed model is to introduce functional parameters into the GARCH process, resulting in a new model called fGARCH, based on the temporal interactions between stock and bond returns. They both demonstrate strong improvement in volatility prediction relative to constant parameter GARCH models.