DSGE Model Estimation via VAR IRF Matching: The Short and Long of It

Inflation persistence in dynamic stochastic general equilibrium (DSGE) models tends to imply a high level of price stickiness, while much lower rigidity in prices is found in micro data. Altig et al. (2011), working on two versions of the Christiano et al. (2005) model, show that this mismatch can be resolved by allowing for firm-specific capital in the macroeconomic framework. Crucially, this mechanism is shown to be supported by US time series data, as the theoretical model is estimated with a vector autoregression (VAR) impulse response functions (IRF) minimum distance estimator, relying on a parsimonious VAR dynamic structure. However, recent literature, De Graeve and Westermark (2025) and Baumeister (2025), shows that VAR dynamic misspecification can substantially distort impulse responses and, therefore, lead to incorrect estimation of DSGE frameworks. Revisiting the results in Altig et al. (2011), I show that, when allowing for long-lag VARs, the evidence in favour of low price stickiness in models with firm-specific capital vanishes, and the micro–macro price friction mismatch reemerges.