Modèle factoriel dynamique contraint à régimes markoviens pour l'évaluation en temps réel du cycle économique
|Abstract:||This thesis is composed of three essays on real-time forecasting dynamic factor models. The main objective is to provide frameworks for high-frequency business cycle analysis in the presence of data revisions. This is relevant for three reasons. First, business cycle forecasting is a central question in macroeconometrics. Secondly, policy-makers would benefit from having access to timely, high-frequency information about business conditions to inform their decisions. Finally, decisions must frequently be made based on data that are subject to revision, and this data uncertainty should be incorporated into the decision-making process. After a review of the empirical business cycle literature and of models of business cycle turning points, we propose a rigorous framework for estimating monthly real US Gross Domestic Product (GDP). A recurring problem in this class of models is that estimates for monthly GDP are generally not consistent with quarterly estimates in the same way that quarterly estimates are not consistent with annual data. Our approach solves this problem. In the first essay (chapter 2), we develop and estimate a dynamic factor model treating the monthly Gross Domestic Product (GDP) as an unobservable latent variable. In contrast with existing approaches, the quarterly averages of our monthly estimates are exactly equal to the Bureau of Economic Analysis quarterly estimates. By construction, our monthly estimates have the advantage of being both timely and easy to interpret. The second essay (chapter 3) extends this framework by adding a Markov-switching model of business cycle regimes to the dynamic factor model. The model is now one with three levels, two of which have latent dependent variables. We pay particular attention to the sensibility of the usual indicators at turning points. The industrial production index, manufacturing and trade sales transmit more information about business cycle shocks to the common component (monthly GDP) than does employment. Finally, we integrate data revisions into our Markov- switching dynamic factor model in order to evaluate the effects of the revisions process on monthly estimates. It appears that data revisions have a significant impact on the co-movement of variables and on turning points without compromising the asymmetric nature of the business cycle. Keywords : Dynamic Factor Model (DFM), High-frequency, Real-time, Markov-switching, unobservable components, Revisions, co-movement, Turning points, Asymmetric, Business cycle.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||19 April 2018|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.