The issue of diversification in direct real estate investment portfolios has been one of the most widely studied topics in academic and practitioner literature. Most work, however, has been done using mean returns and risks for broad market segments as inputs asset allocation models, or in a few cases using data from small sets of individual properties. This paper reports results from a comprehensive testing of asset allocation modelling drawing on records of 10,000+ UK properties tracked by Investment Property Databank. It provides for the first time robust estimates of the diversification gains attainable given return, risk and cross-correlations across individual properties actually available to fund managers. The discussion of results covers implications for the number of assets and amount of money needed to construct ìbalancedî portfolios by direct investment, or via indirect specialist vehicles.