The two-envelope paradox

Abstract

It has been suggested that the two- envelope paradox will not arise for a Bayesian who attaches a prior probability distribution to the sums of money contained in the two envelopes. This paper shows this is incorrect; it describes some prior distributions that generate the paradox. It shows a connection between the two- envelope paradox and the St Petersburg paradox, and argues that the two-envelope paradox has not yet been satisfactorily solved.