Overreaction in expectations: uncertainty about the persistence of a crisis
In this paper, Acharaya et al. shed light on how uncertainty about the persistence of a crisis affects agents’ welfare and economic decisions. They use a model that depicts disasters in reduced form as regimes in which the stock of wealth is subject to exogenous destruction. The economy transitions stochastically between these episodes and “normal times.” Agents optimally solve their investment/consumption problem, whose solution depends on both the current state and on current information about the unobserved switching parameters. Their model shows:
- In a disaster, uncertainty about the duration of a crisis leads to an intensified response, and acts as an amplification mechanism. Agents alter their optimal investment and consumption conservatively relative to the full-information benchmark, and the welfare cost of parameter uncertainty can be extreme. The intuition is that as precision about information declines the longer the crisis lasts, the longer it is expected to last.
- On the other hand, in advance of a disaster, uncertainty about the arrival rate can be welfare-increasing as agents are acting as if “best-case scenarios” predominate in their beliefs. As a result, agents may exhibit diminished preparedness against future disasters, i.e., optimally invest less in mitigation than under full-information, which also manifests as them willing to pay less for insurance against the next disaster.
Disasters with Unobservable Duration and Frequency: Intensified Responses and Diminished Preparedness
Authors: Viral V. Acharya, Timothy Johnson, Suresh Sundaresan, Steven Zheng
From: New York University, University of Illinois at Urbana-Champaign, Columbia University, University of California, Berkeley
Overreaction in expectations: the importance of recent observations
In this paper, Afrouzi et al. conduct a large-scale randomized experiment where participants forecast stable random processes. The experiment allows the authors to control the data generating process and the relevant information sets. This is not feasible in survey data, which can give rise to major complications in interpreting results from survey data. They find that forecasts display significant overreaction: they respond too much to recent observations.
Overreaction is particularly pronounced for less persistent processes and longer forecast horizons. We also find that commonly-used models, estimated in our data, do not easily account for the variation in overreaction. Some predict too much overreaction when the process is transitory (e.g., adaptive expectations and simple extrapolation), while others predict too little (e.g., diagnostic expectations and constant gain learning).
Overreaction in Expectations: Evidence and Theory
Authors: Hassan Afrouzi, Spencer Yongwook Kwon, Augustin Landier, Yueran Ma, David Thesmar
From: Columbia University, Harvard University, HEC, University of Chicago, MIT
