Title: Forecasting and policy analysis with trend-cycle bayesian VARs
Authors: Jan Bruha - Czech National Bank (Czech Republic) [presenting]
Michal Andrle - IMF (United States)
Abstract: Trend-Cycle Bayesian VARs (TC-BVARs) are introduced for use in macroeconomic forecasting and policy analysis. Economic theory supports the view that trends and cycles are dominated by different shocks and transmission channels. Each variable is decomposed into trend and cyclical components. The flexibility of TC-BVARs comes from the fact that the model specifies flexible processes for low-frequency movements (trends) of variables and flexible VAR process for the cyclical frequencies. There is a clear distinction of cycles, trends, or exogenous time-varying policy targets. TC-BVARs benefit from the flexibility of VARs and from careful anchoring of the models long-run behavior. The state-space form of the model helps to work with missing data, mixed frequencies, and various forms of expert judgment and conditional forecasting. Structural TC-BVARs benefitfrom less biased reduced-form model specification.