Why Are Inflation Forecasts Sticky? Theory and Application to France and Germany

This paper proposes to adapt the model of pricing decisions developed by Alvarez, Lippi, and Paciello (2011) to the decision process of forecasters. The model features both a fixed cost of announcing a revised forecast and a fixed cost of updating the information set and adapting the forecast accordingly. Basically, the former fixed communication costs determine state dependence, which implies that the forecaster changes its forecast only when it is far enough from the optimal forecast, i.e., beyond a fixed threshold; the latter fixed information costs determine time dependence, which implies that the forecaster updates its information set only every other T periods, where T is optimally chosen. We show that survey data of inflation forecast updates as well as the last known monthly inflation rates can be used to estimate the threshold implied by the theoretical model. This threshold estimate is then crucial to uncover the existence of both types of costs as well as an upper bound of the optimal time between two information observations. French and German data suggest that the maximum optimal time to next observation is six months, while the observation cost is at most twice as large as the communication cost.