The shape of the utility function under risk in the loss domain and the "ruinous losses" hypothesis: some experimental results

This paper reports some preliminary experimental results as regards the shape of the utility function for losses when elicited over a wide interval of consequences. Individual utility functions are elicited using the trade-off method, which, unlike standard elicitation procedures, is robust to probability weighting (and avoids most cognitive biases). Even though most utility functions exhibit the usual convex shape, nearly 25% of them appear to be inverse-S shaped, with convexity over moderate losses changing to concavity as losses grow. Though not conclusive (due mainly to the small size of our subject pool), this result brings some new support to the old idea that ruinous or unacceptable losses may induce some abrupt change in the shape of the utility function. Most importantly, it paves the way for more systematic investigation of the "ruinous losses" hypothesis.