Achieving parsimony in Bayesian vector autoregressions with the horseshoe prior

Abstract In the context of a vector autoregression (VAR) model, or any multivariate regression model, the number of relevant predictors may be small relative to the information set that is available. It is well known that forecasts based on (un-penalized) least squares estimates can overfit the data and lead to poor predictions. Since the Minnesota prior was proposed, there have been many methods developed aiming at improving prediction performance. The horseshoe prior is proposed in the context of a Bayesian VAR. The horseshoe prior is a unique shrinkage prior scheme in that it shrinks irrelevant signals rigorously to 0 while allowing large signals to remain large and practically unshrunk. In an empirical study, it is shown that the horseshoe prior competes favorably with shrinkage schemes commonly used in Bayesian VAR models as well as with a prior that imposes true sparsity in the coefficient vector. Additionally, the use of particle Gibbs with backwards simulation is proposed for the estimation of the time-varying volatility parameters. A detailed description of relevant MCMC methods is provided in the supplementary material.

[1]  Dimitris Korobilis,et al.  VAR Forecasting Using Bayesian Variable Selection , 2009 .

[2]  J. Griffin,et al.  Inference with normal-gamma prior distributions in regression problems , 2010 .

[3]  Jean Boivin,et al.  Measuring the Effects of Monetary Policy: A Factor-Augmented Vector Autoregressive (FAVAR) Approach , 2003 .

[4]  James G. Scott,et al.  The horseshoe estimator for sparse signals , 2010 .

[5]  Todd E. Clark,et al.  Common Drifting Volatility in Large Bayesian VARs , 2012 .

[6]  Michele Lenza,et al.  Prior Selection for Vector Autoregressions , 2012, Review of Economics and Statistics.

[7]  G. Koop,et al.  Bayesian Multivariate Time Series Methods for Empirical Macroeconomics , 2009 .

[8]  Fabio Canova,et al.  Forecasting and Turning-Point Predictions in a Bayesian Panel VAR Model , 2004 .

[9]  James G. Scott,et al.  Shrink Globally, Act Locally: Sparse Bayesian Regularization and Prediction , 2022 .

[10]  Robert B. Litterman,et al.  Forecasting and Conditional Projection Using Realistic Prior Distributions , 1983 .

[11]  Todd E. Clark,et al.  Macroeconomic Forecasting Performance under Alternative Specifications of Time-Varying Volatility , 2015 .

[12]  A. Doucet,et al.  Particle Markov chain Monte Carlo methods , 2010 .

[13]  Dorota Kurowicka,et al.  Generating random correlation matrices based on vines and extended onion method , 2009, J. Multivar. Anal..

[14]  J. Geweke,et al.  Comparing and Evaluating Bayesian Predictive Distributions of Asset Returns , 2008 .

[15]  Dongchu Sun,et al.  Bayesian stochastic search for VAR model restrictions , 2008 .

[16]  Florian Huber,et al.  Adaptive Shrinkage in Bayesian Vector Autoregressive Models , 2019 .

[17]  R. Snee,et al.  Ridge Regression in Practice , 1975 .

[18]  Dimitris Korobilis,et al.  Adaptive hierarchical priors for high-dimensional vector autoregressions , 2019 .

[19]  Catherine Doz,et al.  A Two-Step Estimator for Large Approximate Dynamic Factor Models Based on Kalman Filtering , 2007 .

[20]  Sune Karlsson.  Forecasting with Bayesian Vector Autoregression , 2013 .

[21]  David H. Small,et al.  Nowcasting: the real time informational content of macroeconomic data releases , 2008 .

[22]  F. Diebold,et al.  Comparing Predictive Accuracy , 1994, Business Cycles.

[23]  Fredrik Lindsten,et al.  On the use of backward simulation in the particle Gibbs sampler , 2012, 2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[24]  D. Giannone,et al.  Large Bayesian vector auto regressions , 2010 .

[25]  A. Gelman.  Prior distributions for variance parameters in hierarchical models (comment on article by Browne and Draper) , 2004 .

[26]  R. Tibshirani.  Regression Shrinkage and Selection via the Lasso , 1996 .

[27]  C. De Mol,et al.  Forecasting Using a Large Number of Predictors: Is Bayesian Regression a Valid Alternative to Principal Components? , 2006, SSRN Electronic Journal.

[28]  J. Bai,et al.  Determining the Number of Factors in Approximate Factor Models , 2000 .

[29]  N. Shephard,et al.  Stochastic Volatility: Likelihood Inference And Comparison With Arch Models , 1996 .

[30]  Dimitris Korobilis,et al.  Hierarchical Shrinkage Priors for Dynamic Regressions with Many Predictors , 2011 .

[31]  Michael I. Jordan,et al.  Dimensionality Reduction for Supervised Learning with Reproducing Kernel Hilbert Spaces , 2004, J. Mach. Learn. Res..