Consistent modeling of risk averse behavior with spectral risk measures: Wächter/Mazzoni revisited

This paper introduces a new method for modeling risk averse behavior with spectral risk measures. It is shown that recent approaches, using phenomenological correspondences or results from robust statistics, generally do not generate consistent results. Our method is based on the dual theory of choice. We show that it is possible to encode preference relations in distorted probability measures, which themselves induce admissible spectral risk measures. This way, risk averse behavior can be mapped onto the risk spectrum defining a spectral risk measure and can be quantified using a local Pratt-Arrow-like coefficient.