Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/192924 
Year of Publication: 
2018
Series/Report no.: 
KOF Working Papers No. 438
Publisher: 
ETH Zurich, KOF Swiss Economic Institute, Zurich
Abstract: 
GDP forecasters face tough choices over which leading indicators to follow and which forecasting models to use. To help resolve these issues, we examine a range of monthly indicators to forecast quarterly GDP growth in a major emerging economy, Russia. Numerous useful indicators are identified and forecast pooling of three model classes (bridge models, MIDAS models and unrestricted mixedfrequency models) are shown to outperform simple benchmark models. We further separately examine forecast accuracy of each of the three model classes. Our results show that differences in performance of model classes are generally small, but for the period covering the Great Recession unrestricted mixed]frequency models and MIDAS models clearly outperform bridge models. Notably, the sets of top]performing indicators differ for our two subsample observation periods (2008Q1. 2011Q4 and 2012Q1.2016Q4). The best indicators in the first period are traditional real]sector variables, while those in the second period consist largely of monetary, banking sector and financial market variables. This finding supports the notion that highly volatile periods of recession and subsequent recovery are driven by forces other than those that prevail in more normal times. The results further suggest that the driving forces of the Russian economy have changed since the global financial crisis.
Subjects: 
Forecasting
mixed frequency data
Russia
GDP growth
JEL: 
C53
E27
Persistent Identifier of the first edition: 
Document Type: 
Working Paper

Files in This Item:
File
Size
1.11 MB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.