Working Paper: CEPR ID: DP2685
Authors: Ventzislav Ivanov; Lutz Kilian
Abstract: An important preliminary step in impulse response analysis is to select the vector autoregressive (VAR) lag order from the data, yet little is known about the implications of alternative lag order selection criteria for the accuracy of the impulse response estimates. In this Paper, we compare the criteria most commonly used in applied work in terms of the mean-squared error of the implied impulse response estimates. We conclude that for monthly VAR models, the Akaike Information Criterion (AIC) produces the most accurate structural and semi-structural impulse response estimates for realistic sample sizes. For quarterly VAR models, the Hannan-Quinn Criterion (HQC) appears to be the most accurate criterion with the exception of sample sizes smaller than 120, for which the Schwarz Information Criterion (SIC) is more accurate. For persistence profiles based on quarterly vector error correction (VEC) models, the SIC is the most accurate criterion for all realistic sample sizes. Sequential Lagrange-multiplier and likelihood ratio tests cannot be recommended.
Keywords: Impulse Responses; Lag Orders; Model Selection; VAR; VEC
JEL Codes: C32; C51; E37; E47
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Akaike Information Criterion (AIC) (C52) | accuracy of impulse response estimates (C51) |
| Hannan-Quinn Criterion (HQC) (C51) | accuracy of impulse response estimates (C51) |
| Schwarz Information Criterion (SIC) (C52) | accuracy of impulse response estimates (C51) |
| sample size (C83) | effectiveness of lag order selection criteria (C52) |
| sequential Lagrange multiplier and likelihood ratio tests (C51) | effectiveness in capturing causal relationships (C90) |