Doctoral Programme Quantitative Economics

Peter Wakker - Erasmus University Rotterdam: "Measuring Ambiguity Attitudes for All (Natural) Events"

Feb 01, 2016 from 04:15 PM to 05:45 PM

Kiel Institute for the World Economy - Lecture hall


Since Keynes (1921) & Knight (1921) we know that uncertainties in economics usually do not come with objective statistical probabilities.  Ellsberg’s (1961) paradox showed that in most cases no subjective probabilities can be assigned in any traditional sense (ambiguity).  Hence we need fundamentally new models.  Only at the end of the 1980s, Gilboa & Schmeidler succeeded in introducing such models, initiating a large rewriting of much in economics and finance to come, and making it the hottest topic in decision theory today.  Tversky & Kahneman (1992) incorporated the Gilboa-Schmeidler ideas into the Nobel-awarded prospect theory, leading to the first empirically realistic model of ambiguity.

        Ambiguity attitudes have so far been measured only for artificial events, where subjective beliefs can be derived from constructed symmetries.  Examples include Ellsberg urns with compositions kep secret and experimenter-specified probability intervals with symmetry about their center.  Ellsberg (2011) and many others emphasized the importance of extensions to natural events as occurring in every-day life, to obtain external validity and to avoid the confound of artificial-experimenter-secret aversion.  However, no revealed-preference extension was known as yet because no method was known to control for subjective beliefs in the absence of symmetry.

        This paper introduces a control for subjective beliefs even when they are unknownbasically by making them drop from the equations.  Thus we can measure ambiguity attitudes for all events, including natural ones.  We introduce indexes of ambiguity aversion and ambiguity perception (or sensitivity) that generalize and unify many existing indexes.  Our indexes are easy to elicit in practice.  Thus they can easily be used as an add-on in regressions.  In an experiment on time pressure we find plausible results, supporting validity.  For applications to natural ambiguities, perception/sensitivity of ambiguity is more important, and aversion is less pronounced, than for artificial Ellsberg-events.

Add to your iCal calendar