Minnesota-type adaptive hierarchical priors for large Bayesian VARs

Large Bayesian VARs with stochastic volatility are increasingly used in empirical macroeconomics. The key to making these highly parameterized useful is the use of shrinkage priors. We develop a family priors that captures best features two prominent classes priors: adaptive hierarchical and Minnesota Like priors, new ensure only ‘small’ coefficients strongly shrunk zero, while ‘large’ remain intact. At same time, can also incorporate many such as cross-variable shrinking on higher lags more aggressively. introduce fast posterior sampler estimate BVARs this priors—for BVAR 25 variables 4 lags, obtaining 10,000 draws takes about 3 min standard desktop computer. In forecasting exercise, we show outperform both