Formação de expectativas de inflação em um ambiente de racionalidade limitada: uma abordagem de escolha discreta

Abstract

We propose a discrete (ternary) choice model to study the frequency distribution of inflation predictors. In every reappraisal period of the foresight strategies, each agent chooses only one among three predictors (naive, adaptive and VAR) to forecast the monthly inflation rate. The predictor selection process is modeled as a discrete choice dynamics based on two attributes, namely, accuracy less the average cost predictor (private attributes) and dispersion in cognitive abilities. The calibrated agent-based computational model shows that heterogeneity in inflation expectations is persistent, that is, less accurate predictors coexist with the most accurate predictor due to the dispersion in cognitive abilities of agents.

Keywords:
Heterogeneous inflation expectations; Dispersion in cognitive abilities; Agent-based computational models