Anxiety and Pro-cyclical Risk Taking with Bayesian Agents

37 Pages Posted: 21 Feb 2015 Last revised: 23 Jan 2019

See all articles by Thomas M. Eisenbach

Thomas M. Eisenbach

Federal Reserve Bank of New York

Martin C. Schmalz

CEPR; University of Oxford - Finance; CESifo; European Corporate Governance Institute (ECGI)

Date Written: June 28, 2018


We provide a model that can explain empirically relevant variations in confidence and risk taking by combining horizon-dependent risk aversion (“anxiety”) and selective memory in a Bayesian intrapersonal game. In the time series, overconfidence is more prevalent when actual risk levels are high, while underconfidence occurs when risks are low. In the cross section, more anxious agents are more prone to biased confidence and their beliefs fluctuate more. This systematic variation in confidence levels can lead to objectively excessive risk taking by “insiders” with the potential to amplify boom-bust cycles.

Keywords: overconfidence, dynamic inconsistency, biases, deception, risk taking

JEL Classification: C72, D03, D81, D83, G02

Suggested Citation

Eisenbach, Thomas M. and Schmalz, Martin C. and Schmalz, Martin C., Anxiety and Pro-cyclical Risk Taking with Bayesian Agents (June 28, 2018). FRB of New York Staff Report No. 711, Available at SSRN: or

Thomas M. Eisenbach (Contact Author)

Federal Reserve Bank of New York ( email )

33 Liberty Street
New York, NY 10045
United States
212-720-6089 (Phone)


Martin C. Schmalz

CEPR ( email )

United Kingdom

University of Oxford - Finance ( email )

United States

CESifo ( email )

Poschinger Str. 5
Munich, DE-81679

European Corporate Governance Institute (ECGI) ( email )

c/o the Royal Academies of Belgium
Rue Ducale 1 Hertogsstraat
1000 Brussels

Do you have negative results from your research you’d like to share?

Paper statistics

Abstract Views
PlumX Metrics