We consider a small scale DSGE with trend in.ation, where the price dispersion is a non-observable state variable. Hence, the model lacks of a finite VAR representation and the VAR analysis based on this model may suffer of the truncation bias problem. First, we use the DSGE model as data-generation process to create artificial pseudo-data. Second, using these pseudo-data, we employ a sign restrictions VAR to evaluate the effects of a monetary policy shock. The true generation process implies a strong response of output after a monetary shock, instead the VAR shows monetary neutrality. We conjecture that this discrepancy is due to a truncation bias which a¤ects the statistical representation of our true data generation process.