Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/212307 
Year of Publication: 
2014
Series/Report no.: 
Bank of Finland Research Discussion Papers No. 32/2014
Publisher: 
Bank of Finland, Helsinki
Abstract: 
In this paper discrete wavelet filtering techniques are applied to decompose macroeconomic data so that they can be simultaneously analyzed in both the time and frequency domains. The MODWT (Maximal Overlap Discrete Wavelet Transform) is applied to U.S. quarterly GDP data from 1947–2012 to obtain the underlying cyclical structure of the GDP components. A MATLAB program is then used to design optimal fiscal policy within a LQ-tracking model with wavelet decomposition, and the results are compared with an aggregate model with no frequency decomposition. The results show that fiscal policy is more active under the wavelet-based model, and that the consumption and investment trajectories under the aggregate model are misaligned. We also simulate FHEC (Frequency Harmonizing Emphasis Control) strategies that allow policymakers to concentrate the policy thrust on tracking frequencies that are optimally aligned with policy goals under different targeting priorities. These strategies are only available by using time-frequency analysis. This research is the first to construct fiscal policy in an applied optimal control model on the short and cyclical lag structures obtained from wavelet analysis. Our wavelet-based optimal control procedure allows the policymaker to construct a pragmatic tracking policy, avoid suboptimal policies gleaned from an aggregate model, and reduce the potential for destabilization that might otherwise result due to improper thrust and timing.
Subjects: 
LQ tracking
macroeconomics
optimal control
discrete wavelet analysis
fiscal policy
JEL: 
C49
C61
C63
C88
E61
Persistent Identifier of the first edition: 
ISBN: 
978-952-323-015-6
Document Type: 
Working Paper

Files in This Item:
File
Size





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