This paper aims to rationalize why the assumptions made by economists when eliciting risk attitudes map onto such a limited external validity of the measures obtained. In particular, we set up an experiment to answer two research questions that are clearly intertwined.
The first one investigates how different layers of ignorance impact the choice between a safe amount and an uncertain alternative. The idea is that risk elicitation methods may fail because they shoot towards the wrong target, i.e. the variance of the outcomes instead of a limited knowledge of the set of outcomes (that we label (deep uncertainty), ambiguity, or the presence of losses.
The second research question focuses on the external validity of the measures of risk preferences elicited in this experiment. In principle, we want to see whether knowing more or less about the situation triggers choices which are closer to external benchmarks of risk attitudes, unrelated to the choices made in this experiment. Toward this goal we exploit stated preferences and a measure of risk taking in real life collected through a Daily Reconstruction Method. Both benchmarks have been collected for the same subjects some months earlier.