Back to all papers

Paper #1953

Title:
Quantifying dynamic misspecification biases in VARs
Author:
Oriol Gonzalez-Casasus
Date:
September 2026
Abstract:
Applied macroeconomists favor parsimonious VARs, which selection criteria and specification pre-tests routinely endorse. I show that weak, long-lived omitted dynamics escape those diagnostics and cannot be consistently estimated, yet the bias they induce in impulse responses and forecasts admits valid confidence intervals. Researchers can then keep the parsimonious VAR and state the direction and magnitude of the distortion it carries. In US data, the bias in the price response to a monetary contraction is significantly positive at several horizons, accounting for part of the price puzzle without expanding the cross-sectional information set.
Keywords:
vector autoregression, dynamic mis-specification, robust inference, price puzzle
JEL codes:
C12, C32, C52, E52.
Area of Research:
Statistics, Econometrics and Quantitative Methods

Download the paper in PDF format