Using modern portfolio theory, the traditional asset allocation process employs measurements of risk and return delivered by asset classes—for example, stocks, bonds, and real estate—to build efficient portfolios. To build efficient portfolios in practice using this type of analysis requires that the risk and return characteristics of the asset class can be replicated in real portfolios. This may be true of stocks and bonds, but is it true of real estate? Using new analysis coupled with previous UK‐based research based on the uniquely rich MSCI (IPD) dataset for UK direct real estate, this paper compares the risk and return characteristics of real estate investment approaches (direct exposure, balanced and specialist unlisted funds, a multi‐manager approach, and listed securities) relative to a UK market index. Based on a random stochastic simulation of historic performance data from 2003 to 2012, we conclude that the difficulty of diversifying away specific risk in such a lumpy asset class means that it is extremely difficult and/or costly to access or replicate direct property market returns. This suggests that an investor/manager setting out to deliver returns in line with a market index would have to demonstrate significant levels of skill. While listed real estate, which is more readily diversifiable, fails to deliver returns that are correlated with direct real estate in the short term (one to five years), it is clear that multi‐manager strategies were able to deliver returns that more effectively replicated a direct benchmark. However, multi‐manager fees negatively impacted on net returns. Specific risk can be avoided by real estate investors, but at a cost.