A Model of Reference-Dependent Belief Updating

62 Pages Posted: 4 Dec 2016  

Johannes Karl-Georg Maier

Ludwig Maximilian University of Munich

Clemens König

NERA Economic Consulting (UK)

Date Written: November 03, 2016

Abstract

We propose a model of instrumental belief choice under loss aversion. When new information arrives, an agent is prompted to abandon her prior. However, potential posteriors may induce her to take actions that generate a lower utility in some states than actions induced by her prior. These losses loom larger than gains in other states in which potential posteriors lead to larger utility than the prior. In choosing her belief, the agent optimally trades off these psychological gains and losses against the belief's objective performance. Consistent with empirical evidence, belief updating in this model is conservative and can be asymmetric, so that bad news is updated more conservatively than good news. These updating biases generate prior-dependent information preferences, such that agents may avoid information when being unconfident but are information seeking otherwise. Because belief updating depends on the decision problem in which new information is going to be used, the model allows us to explore how these predictions change with either individual preferences or the choice context.

Keywords: belief choice, non-Bayesian updating, reference dependence, loss aversion, regret, conservatism, signal valence, information preferences, overconfidence

JEL Classification: D030, D810, D830, D840, G020

Suggested Citation

Maier, Johannes Karl-Georg and König, Clemens, A Model of Reference-Dependent Belief Updating (November 03, 2016). CESifo Working Paper Series No. 6156. Available at SSRN: https://ssrn.com/abstract=2878665

Johannes Karl-Georg Maier (Contact Author)

Ludwig Maximilian University of Munich ( email )

Geschwister-Scholl-Platz 1
Munich, Bavaria 80539
Germany

Clemens König

NERA Economic Consulting (UK) ( email )

15 Stratford Place
London, W1C 1BE
United Kingdom

Paper statistics

Downloads
34
Abstract Views
139