Assessing the structural VAR approach to exchange rate pass-through

A common approach to evaluate dynamic stochastic general equilibrium (DSGE) models is to compare the impulse responses functions from the DSGE model to impulse responses obtained from identified vector autoregressions (VARs). This paper uses Monte Carlo techniques to address the question: Are impulse responses of prices to a UIP shock a useful tool to evaluate DSGE models with incomplete exchange rate pass-through? The data generating process is a small open economy DSGE model. The results suggest that (i) the estimates obtained from a VAR estimated in first differences exhibit a systematic downward bias, even when the VAR is specified with a large number of lags; (ii) by contrast, estimates derived from a low order vector equilibrium correction model are fairly accurate; but (iii) standard cointegration tests have low power to detect the cointegration relations implied by the DSGE model.