Simple belief elicitation: An experimental evaluation

We present a method for eliciting beliefs about probabilities when multiple realisations of an outcome are available, the "frequency" method. The method is applicable for any reasonable utility function. Unlike existing techniques that account for deviations from risk-neutrality, this method is highly transparent to subjects and easy to implement. Rather than identifying point beliefs these methods identify bounds on beliefs, thus trading off precision for generality and simplicity. An experimental comparison of this method and a popular alternative, the Karni method, shows that subjects indeed find the frequency method easier to understand. Significantly, we show that confusion due to the complexity of the Karni method leads to less cognitively able subjects erroneously stating a belief of 50%, a bias not present in the frequency method.