Working Paper: CEPR ID: DP4033
Authors: Fabio Canova; Matteo Ciccarelli
Abstract: This Paper proposes a method to conduct inference in panel VAR models with cross-unit interdependencies and time variations in the coefficients. The set-up used is Bayesian, and Markov chain Monte Carlo (MCMC) methods are used to estimate the posterior distribution of the features of interest. The model is re-parameterized to resemble an observable index model and specification searches are discussed. The approach can be used to construct multi-unit forecasts, leading indicators and to conduct policy analysis in multi-unit set-ups. The methodology is employed to construct leading indicators for inflation and GDP growth in the euro area.
Keywords: Bayesian methods; Leading indicators; Markov chain Monte Carlo methods; Panel VAR
JEL Codes: C30; C50; E50
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Bayesian hierarchical approach (C11) | estimation of time-varying coefficients (C51) |
| estimation of time-varying coefficients (C51) | capture dynamics of economic indicators over time (E32) |
| shocks in one country (F69) | significant effects on others (D91) |
| reparameterization of the model into a factor structure (C38) | reduces complexity of estimating numerous individual coefficients (C51) |
| reparameterization of the model into a factor structure (C38) | more efficient estimation of the overall system dynamics (C51) |
| Bayesian hierarchical approach (C11) | improved forecasting accuracy for leading indicators of inflation and GDP growth (E37) |
| model (C59) | effectively trace the impact of shocks in economic indicators (E32) |