The Optimized Monetary Policy ZLB Mandate in NK Behavioural and RE Models Compared

In this paper, we extend the basic ‘workhorse’ New Keynesian model by relaxing the rationality assumption in favor of bounded rationality. We broadly follow (Gabaix, 2020) whereby agents are unable to anticipate macroeconomic developments perfectly and assumed to be both partially myopic and inattentive. In my set-up, we assume that agents form beliefs over the future infinite time horizon of aggregate states and prices which are exogenous to their decisions. we then directly apply agents’ inattentiveness on these exogenous market factors into their infinite-time-horizon decision rules. This approach produces identical decision rules as the “sparse agent” approach of Gabaix. Most importantly, with my method we can derive a fully non-linear form of my model, which is necessary to provide a correct welfare rankings in my mandate study. Results by the standard linear Bayesian estimation technique in this paper show that the boundedly rational expectation (BR) model outperforms the fully rational expectation model (RE). In the BR model, introducing the zero-lower-bound (ZLB) episode results in a higher welfare cost compared to the RE model. As a result, the optimal steady state inflation level of the BR model is higher given a probability of the nominal interest hitting the ZLB. Optimized interest rate rules for the RE and BR models closely mimics a price-level targeting rule. However, under the RE model, the nominal interest rate reacts more aggressively to the price inflation.