$($c$)$ Prior experimental evidence suggests that, while subjects' beliefs might be

quite close to those calculated using Bayes rule for the symmetric case, sub$-$

jects' find it difficult to accurately incorporate new information $($at least accord$-$

ing to the Bayes rule benchmark$)$ when events are very asymmetric. Explain

how you would design a set of treatments to test this hypothesis, and to ex$-$

plore whether it is important to use a belief elicitation procedure that provides

proper incentives even in the case where subjects are not risk neutral. Make

sure you explain how you could use the data generated by the experiment to

test the hypothesis.

$[10$ marks$]$