Working Paper: CEPR ID: DP8755
Authors: Domenico Giannone; Michele Lenza; Giorgio E. Primiceri
Abstract: Vector autoregressions (VARs) are flexible time series models that can capture complex dynamic interrelationships among macroeconomic variables. However, their dense parameterization leads to unstable inference and inaccurate out-of-sample forecasts, particularly for models with many variables. A potential solution to this problem is to use informative priors, in order to shrink the richly parameterized unrestricted model towards a parsimonious naïve benchmark, and thus reduce estimation uncertainty. This paper studies the optimal choice of the informativeness of these priors, which we treat as additional parameters, in the spirit of hierarchical modeling. This approach is theoretically grounded, easy to implement, and greatly reduces the number and importance of subjective choices in the setting of the prior. Moreover, it performs very well both in terms of out-of-sample forecasting, and accuracy in the estimation of impulse response functions.
Keywords: Bayesian methods; forecasting; hierarchical modelling; impulse responses; marginal likelihood
JEL Codes: C11; C32; C52; E37
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Informative priors (D83) | Out-of-sample forecasting accuracy (C53) |
| Hierarchical modeling (C59) | Estimation of impulse response functions (C51) |
| Tighter priors (C11) | Accuracy of forecasts (C53) |
| Looser priors (C11) | Effectiveness in simpler models (C52) |
| Prior selection method (C52) | Quality of impulse response estimates (C22) |