In my textbook, I have discussed why extreme bounds analysis is NOT a good idea. Basically EBA takes a class of priors (instead of one) and computes posteriors for all of them. Then it looks at the extreme bounds produced by this analysis. This was proposed by Leamer as a way of overcoming subjectivity of priors.
Empirical Bayes tells us that not all priors are equal. Some, which are very far from data, are very unlikey. Others, which are compatible with data are more likely. If we use this information in a sensible way, it should be possible to improve on extreme bounds analysis. I dont know what the paper below does, but it could be used as a jump-off point to introduce new ideas and new type of analysis.
THE DETERMINANTS OF U.S. STATE ECONOMIC GROWTH: A LESS EXTREME BOUNDS ANALYSIS
W. ROBERT REED
Published Online: 8 Apr 2008