Consistent inconsistencies? Evidence from decision under risk