Precautionary Price Stickiness (original) (raw)

2000, SSRN Electronic Journal

In this paper, we consider the pricing behavior of near-rational …rms subject to idiosyncratic shocks which can adjust their prices at any time, but may make mistakes. We model error-prone behavior by assuming …rms play a dynamic logit equilibrium. Prices are therefore endogenously sticky: when a …rm's current price is su¢ ciently close to the optimum, it prefers to leave well enough alone, avoiding the risk of accidentally choosing a worse price. This way of modeling price stickiness is consistent with several observations from microdata that have been hard to explain in most existing frameworks. First, even though the decision to adjust prices has an (S,s) structure, nonetheless many small price changes coexist with larger ones (Klenow and Malin, 2009, "Fact 7"). Second, error-prone choice implies that the price adjustment hazard exhibits negative duration dependence (Nakamura and Steinsson, 2008, "Fact 5"; Klenow and Malin, 2009, "Fact 10"), since …rms will choose to adjust again quickly if they perceive that they have made a mistake. Third, the size of the price change is largely independent of the time since last adjustment (Klenow and Malin, 2009, "Fact 10"). When we estimate the logit rationality parameter, our model matches both the frequency of adjustment and the size distribution of price changes quite successfully. Since our setup guarantees that …rms making su¢ ciently large errors will choose to adjust, it generates a large "selection e¤ect" in response to a nominal shock which eliminates most of the real e¤ects of money. Thus, a model in which price stickiness is a precautionary phenomenon implies that money shocks are almost neutral, as in Golosov and Lucas (2007), but …ts microdata better than their speci…cation does. In ongoing work, we are also estimating a generalized model in which both the price chosen, and the decision of whether or not to adjust the price, are subject to logit errors. This should allow us to distinguish whether intermittent adjustment is driven primarily by a risk of errors or by stickiness per se.