We present a flexible structural vector autoregressive model with identification via heteroskedasticity. It encompasses a range of volatility models and allows for imposing over-identifying restrictions. Consequently, statistical methods can be used for comparing models with alternative sets of restrictions that just-identify a conventional SVAR model. Efficient Bayesian algorithms are derived for estimating larger models that are difficult to handle in a frequentist analysis. We propose a novel marginal data density estimator for structural models that are not globally identified. Comparing three classical models for the U.S., we find that models that induce monetary policy shocks from non-borrowed reserves outperform those that use federal funds rate for that purpose.

Assessing Monetary Policy Models: Bayesian Inference for Heteroskedastic Structural VARs (with Matthieu Droumaguet)

Fri 1 Apr 2016 3:30pm5:00pm


Room 103, Colin Clark Building (#39)