Probability Weighting Functions

In this paper we begin by stressing the empirical importance of non-linear weighting of probabilities, which expected utility theory (EU) is unable to accommodate. We then go on to outline three stylized facts on non-linear weighting that any alternative theory of risk must address. These are that people: overweight small probabilities and underweight large ones (S1); do not choose stochastically dominated options when such dominance is obvious (S2); ignore very small probabilities and code extremely large probabilities as one (S3). We then show that the concept of a probability weighting function (PWF) is crucial in addressing S1-S3. A PWF is not, however, a theory of risk. PWF's need to be embedded within some theory of risk in order to have significant predictive content. We ouline the two main alternative theories that are relevant in this regard: rank dependent utility (RDU) and cumulative prospect theory (CP). RDU and CP explain S1,S2 but not S3. We conclude by outlining the recent proposal for composite prospect theory (CPP) that uses the composite Prelec probability weighting function (CPF). CPF is axiomatically founded, and is flexible and parsimonious. CPP can explain all three stylized facts S1,S2,S3.