Kociecki, Andrzej Rubaszek, Michał Ca' Zorzi, Michele
Year of Publication:
ECB Working Paper 1492
The paper provides a novel Bayesian methodological framework to estimate structural VAR (SVAR) models with recursive identification schemes that allows for the inclusion of over-identifying restrictions. The proposed framework enables the researcher to (i) elicit the prior on the non-zero contemporaneous relations between economic variables and to (ii) derive an analytical expression for the posterior distribution and marginal data density. We illustrate our methodological framework by estimating a backward looking New-Keynesian model taking into account prior beliefs about the contemporaneous coefficients in the Phillips curve and Taylor rule.
Bayesian inference overidentifying restrictions Structural VAR