Detailansicht

Jürgen Bierbaumer-Polly, Sandra Bilek-Steindl, Marcus Scheiblecker, Nowcasting Economic Activity Using Large-Scale Dynamic Factor Models with Regional Block-Structure. An Application for the Austrian Economy

Projektberichte (in Arbeit), Jänner 2018
We propose a hierarchical four-level dynamic factor model (DFM) for now- and forecasting the quarterly Austrian GDP. In the model, the business cycle dynamics are defined not only by common, but also by block and subblock specific factors. These factors are related to domestic versus international data origin, as well as to characteristics according to Austria's regional structure (manufacturing vs. services dominated). While the dynamics in the national GDP growth are mainly determined by common shocks, a considerable part of the variation in the regional data can be explained by the subblock specific factors. In a "pseudo" real-time forecasting exercise we show that our proposed model outperforms a standard two-level DFM. To deal with missing observations at the beginning/end of the dataset we apply the expectation maximisation (EM) procedure to obtain a balanced data panel. The outcome of this are, i.a., backcasts of the quarterly Austrian GDP series which seem to fit the real-time estimates reasonably well. In a real-time setting, this backcast can give an early (approximately t+15 days) indication of quarterly GDP growth, with its first official release (Flash estimates) published at t+30.
Report to the "Jubiläumsfonds der Oesterreichischen Nationalbank", Grant No 15999 • Project Co-ordinator: Peter Mayerhofer
 
Auftraggeber: Gefördert vom Jubiläumsfonds der Oesterreichischen Nationalbank
Studie von: Österreichisches Institut für Wirtschaftsforschung
Forschungsbereich:Makroökonomie und europäische Wirtschaftspolitik
Sprache:Englisch

Verwandte Einträge

Jürgen Bierbaumer-Polly (Projektleitung), Nowcasting Economic Activity Using Large-Scale Dynamic Factor Models with Regional Block-Structure: An Application for the Austrian Economy

Aktuelle Forschungsprojekte (in Arbeit)
Publication delays of macroeconomic data, like Quarterly National Accounts figures, make it difficult to assess the state of the economy in real time. Nowcasts of, for example, quarterly GDP can mitigate this issue. We propose a hierarchical three-stage dynamic factor model with regional blocks, which will be used for nowcasting both Austrian quarterly national and regional economic activity. In the model the business cycle dynamics are defined not only by common, but also by block-specific factors. These factors are in the first stage related to a certain region, and in the second stage defined with respect to the region's main sectoral predominance. Using this block-structure approach we are able to exploit the full range of the comprehensive dataset, including not only national and international data, but also a high number of regional indicators. Having obtained the model we evaluate its forecasting performance for national GDP by looking at its pseudo out-of-sample forecasts.
 
Auftraggeber: Jubiläumsfonds der Oesterreichischen Nationalbank
Studie von: Österreichisches Institut für Wirtschaftsforschung
Login-Formular