This paper provides a quantitative framework for choosing the composition of reserve currencies. Assuming that the central bank's performance objectives are defined in terms of ex post returns in different currency numeraires, the currency allocation problem is formulated as a multi-objective optimisation problem. The advantage of the proposed methodology is that it does not require any explicit assumptions about the risk preferences of the central bank or knowledge of the currency numeraire. Using some proxy values for the possible range of ex post returns measured in different currency numeraires, the study shows how the currency allocation problem can be solved. In particular, the proposed method borrows the concept of the degree of satisfaction from fuzzy decision theory and maximises such a function defined on the least favourable return outcome. In this sense, the proposed method differs from standard utility-based approaches which look for solutions that are best on average. The results of the study indicate that central banks on average are dollar-based investors on the basis of current allocations. Further, the study also indicates that if central banks consider an ex post return profile that safeguards the purchasing power of the reserves, then the currency distribution of reserves should more closely resemble the SDR basket.