No. 181. Inference in Vector Autoregressive Models with an Informative Prior on the Steady State
by Mattias Villani
Abstract: Vector autoregressions have steadily gained in popularity since their introduction in econometrics 25 years ago. A drawback of the otherwise fairly well developed methodology is the inability to incorporate prior beliefs regarding the system's steady state in a satisfactory way. Such prior information are typically readily available and may be crucial for forecasts at long horizons. This paper develops easily implemented numerical simulation algorithms for analyzing stationary and cointegrated VARs in a parametrization where prior beliefs on the steady state may be adequately incorporated. The analysis is illustrated on macroeconomic data for the Euro area.
Keywords: Cointegration, Bayesian inference, Forecasting, Unconditional mean, VARs.
JEL Classification Numbers: C11, C32, C53, E50.
