Estimating Models with Dispersed Information

We conduct likelihood evaluation of a DSGE model in which firms have imperfect common knowledge. Imperfect common knowledge is found to be more successful than price stickiness à la Calvo to account for the highly persistent effects of nominal shocks on output and inflation. Our likelihood analysis suggests that firms pay little attention to aggregate nominal conditions. This paper shows that such allocation of attention is plausible because it is optimal for firms with a reasonably small size of information frictions and a size of idiosyncratic uncertainty that is in line with the micro evidence on price changes.