Southwest Institute 2005 Program and Proceedings
Based on an assumption of one-way learning, Granato and Wong (2004) consider a framework with two groups of agents, Group L and Group H, where Group L is less "attentive" and uses the expectations of the more or highly "attentive" Group H to update their forecasts. The paper shows the "boomerang effect," which is defined as a situation where the inaccurate forecasts of a less attentive group confound a more attentive group's forecasts. This extended paper relaxes the one-way learning assumption and investigates the case that both groups are learning from each other, i.e., dual learning. Simulations suggest that a boomerang effect still exists. Surprisingly, although the highly attentive group has a full set of information to make forecasts, they still learn from Group L. The reason is that Group H adjusts their forecasts because there is available information in Group L's forecast measurement error.