Consistent assumptions for multinational asset models

Document description

Building an international ALM requires assumptions about expected returns on bonds, equities and currencies around the world. How can we be sure these are consistent? We test several methods of deriving assumptions. Using simple methods, we discover that the larger the model gets, the greater the chance that even sensible looking assumptions imply major market inefficiencies. The question is, can we find a better assumption set that passes the efficiency test but where the numbers still make some sort of intuitive sense? Should we bodge the mean returns, or the variance covariance matrix - or do we have do invent some other psychological bias to explain standard assumptions?