Fig. 1.

a Distribution of the preference stability index (number of pairwise comparisons in which implied parameter intervals overlap) for the experimental data ($n=185$). b Simulation exercise with virtual subjects choosing uniformly and independently from the available choices in each of the four risk preference elicitation methods. c Simulation exercise with virtual subjects choosing independently from the choice distribution of each task observed in the experiment. d Simulation exercise with virtual subjects with stochastic preferences, where a noise term $ϵ\sim N(0,0.3)$ is added directly to subjects’ crra parameter $\phi \sim N(0.6,0.3)$. $n=10,000$ for each simulation