PhD defence of Richard Kleijn on Thursday 28 January 2016
On Thursday 28 January 2016 Richard Kleijn will defend his PhD thesis entitled 'Essays on Bayesian Model Averaging using Economic Time Series'. Supervisor is Professor Herman van Dijk (Erasmus School of Economics). Other members of the Doctoral Committee are Professor Richard Paap, Professor Casper de Vries (both Erasmus School of Economics) and Dr. Lennart Hoogerheide (VU University Amsterdam).
About Richard Kleijn
Richard Kleijn (1975) obtained his Master's degree in Econometrics from Erasmus School of Economics in 1998. After graduation, he became a researcher at the Tinbergen Institute and the Econometric Institute at the same university. His academic research has been presented at international conferences and published in the Journal of Applied Econometrics and the Journal of Forecasting. Since 2005 he has been working in the pension fund industry in the areas of asset liability management and investment policy advisory.
Abstract of 'Essays on Bayesian Model Averaging using Economic Time Series.'
In many situations econometric inference may give different results for different models both in a Bayesian and a Frequentist approach. A common approach is that of model selection in which one single model is selected amongst a variety of possible models. All subsequent analysis such as forecasting is carried out using the selected model. The consequence is that information from the models that were not selected is ignored. For this reason, in this thesis we not only consider model selection but also model averaging, which is an approach in which information from multiple models is combined. The motivating idea is that one single model will not be the best model under all circumstances.
This thesis contributes to this literature by introducing a novel and very flexible class of models and combining it with model selection and model averaging and by developing extensions of model averaging to time varying model weights. Other topics covered in this thesis include a study into the existence of posterior probabilities and higher moments within a cointegration model and Bayesion simulation-based inference of an unobserved components model applied to the analysis of real exchange rate behaviour.