B1617
Title: Structural periodic vector autoregressions under general linear restrictions
Authors: Daniel Dzikowski - TU Dortmund University (Germany) [presenting]
Carsten Jentsch - TU Dortmund University (Germany)
Abstract: While seasonality inherent to raw macroeconomic data is commonly removed by certain seasonal adjustment techniques before it is used for structural inference, this approach might distort the information contained in the data. As an alternative method to commonly used structural vector autoregressions (SVAR) for seasonally adjusted macroeconomic data, the purpose is to offer an approach in which the seasonality of non-seasonally adjusted raw data is modelled directly by periodic structural vector autoregressions (SPVAR). In comparison to a VAR, the periodic VAR allows for periodically time-varying intercepts and periodic autoregressive parameters and innovations' variances, respectively. Thus, the SPVAR allows the capture of seasonal effects and enables a direct and more refined analysis of seasonal patterns in macroeconomic data. Moreover, based on such SPVARs, a general concept is proposed for structural impulse response analyses that take seasonal patterns directly into account. Asymptotic theory is provided for estimators of periodically reduced form parameters and structural impulse responses under flexible linear restrictions. Further, residual-based (seasonal) bootstraps for constructing confidence intervals are introduced. A real data application to a three-dimensional system of macro variables is provided, showing that the most common seasonal adjustment methods generally work very well but that useful insights about the data structure may be lost.
