Trend, Cycle and Expectation Formation

Abstract

<p>Using the survey of professional forecasters, we document two facts that contradict the standard assumption of stationary state variables in the expectation formation literature: (i) changes in trend beliefs are negatively correlated with changes in cyclical beliefs; (ii) forecast dispersion increases as the forecast horizon lengthens. To organize these facts, we propose an otherwise standard forecasting model with stochastic trends. In this model, forecasters have imperfect information and cannot perfectly differentiate between trends and cycles. This information friction leads to persistent forecasting errors or rational confusion about both components. Furthermore, we show that this information friction may underlie changes in forecasting behaviors following the implementation of explicit inflation targeting in 2012. In an extension of this framework, we examine how be- havioral biases can amplify the impact of such confusion.<br></p>