Robust Inference in Risk Elicitation Tasks

Recent experimental evidence suggests that noisy behavior correlates strongly with cognitive ability. This puts previous studies that found a negative relation between cognitive ability and risk aversion into perspective and in particular raises the question of how to achieve robust inference in this domain. This paper shows that using structural estimation that models heterogeneity of noise in combination with a balanced design allows us to mitigate the bias problem. Our estimations show that cognitive ability is related to noisy behavior rather than risk preferences. We also find age and education to be strongly related to noise, but the personality characteristics obtained using the Big Five inventory, are less related to noise and more robustly correlated to risk preferences.