Real effects of monetary policy shocks in menu cost models depend crucially on the distribution of the size of price changes. With a realistically high kurtosis, Midrigan (2011) overturned the standard Golosov and Lucas (2007) result of small and temporary real effects and found them nearly as large as in time-dependent pricing models (Calvo, 1983). In this paper we show, however, that this result hangs on assuming both small aggregate shocks and no trend inflation. For larger shock sizes (of magnitude from 1-3\%) or positive trend inflation, the influence of the distribution turns around and the monetary policy effectiveness in Midrigan's model gets even smaller, or, equivalently, the inflation pass-through of aggregate shocks gets even higher than in the Golosov-Lucas model. The reason is that with leptokurtic shocks, for large aggregate shocks or trend inflation, the extensive margin starts playing a quantitatively important role; a channel that is missing from standard time-dependent pricing models, and is quantitatively weak in the Golosov-Lucas model. We also show that the interaction of leptokurtic shocks with trend inflation (Ball and Mankiw, 1994) implies a quantitatively important asymmetry between the inflation pass-through (and the real effects) of positive and negative monetary policy shocks. We present evidence for very high positive and small negative inflation pass-throughs in a natural experiment of large value-added tax shocks in Hungary, that are in line with the quantitative predictions of Midrigan's calibrated menu cost model.
