Home page

Arnaud Dufays - ph. D. in Economics and Management Science
Assistant professor at Université Laval (ULaval)
2216 Pavillon J.-A.-DeSève, G1V 0A6, Québec, Canada
Office +1 (418) 6562131 - Poste 7749

Research Interests  : Econometrics, Finance and Bayesian inference

Ph.D. thesis : Modeling structural changes in volatility  (Thesis advisor : Luc Bauwens)

It is well known by economists that forecasting time series is difficult for many reasons, among which the changing (institutional, technological…) environment and behaviour of agents of the economic system. Changes may be slow and progressive or quick and abrupt, and generate the non-stationarity of many time series in the long run, such as changes in their trend or volatility. Ignoring these breaks by assuming constant parameters in econometric models, typically leads to forecasts that are far from realizations.  The thesis focuses on detecting and estimating structural breaks in volatility of financial time series.

Master thesis : Forecasting Time Series Subject To Multiple Structural Breaks (Promotor : Luc Bauwens)

The thesis theoretically and pratically (in matlab) develops  an autoregressive model that deals with structural breaks in parameters (mean and variance). By embedding a hierarchical structure, the model uses information on the magnitude and the frequency of structural breaks exhibited by the time series to forecast structural breaks. It documents the paper of Pesaran, Pettenuzo and Timmerman's (2006) and deepens their results by adding some discussions on the Monte carlo standard error. It also investigates the possibility of having structural break in the frequency of structural break itself. A self-contained theory for understanding the main concept of Bayesian inferences by Markov-Chain Monte carlo (MCMC) lies in appendix.

Recent papers :