Essays on Modeling Time-Varying Parameters

Abstract:

This dissertation contains four essays on econometric time series modelling. More specifically, the focus is on theoretical properties as well as multivariate applications of time-varying parameter models. This dissertation is therefore split in two parts: a more theoretical part and a more applied part.

The more theoretical part considers optimality properties of score-driven models. The class of score-driven models has gained considerable popularity in the recent statistical literature. Score-driven models are typically appreciated for their robustness properties since the models flexibly adapt themselves to the distribution of the innovations. Despite being relatively new, a wide range of applications of score-driven models is already available in the literature. This part further extends the theoretical motivations for score-driven models.

The more applied part considers two multivariate applications.

The first application is motivated by structural changes observed in a number of key macroeconomic variables, such as interest rates, GDP growth and inflation. This application contributes to a growing literature on how best to model time variation in macro time series models in a forecasting context.

The second application investigates covariance matrix estimation in vast-dimensional spaces of 1,500 up to 2,000 stocks using fundamental factor models. In particular, it evaluates whether recent linear and non-linear shrinkage methods help to reduce the estimation risk in the asset return covariance matrix.

About the author:

Andries van Vlodrop graduated from the Tinbergen MPhil program in 2014. Upon completion of this program he joined the Finance department at the Vrije Universiteit Amsterdam as a PhD student. Currently he is working as a quantitative risk specialist at UBS.