This thesis proposes to investigate the macroeconomic risk factors and specific investment characteristics having an influence on real estate investments performance, comparing across direct, listed and non-listed exposures to the market. Linkages between the three exposures are also examined. Then, the methods devoted to treat the smoothing bias, affecting appraisal-based real estate indices, are compared and improvements relying on robust statistic techniques are proposed. In addition, the ability to properly identify risk factors depending on whether one uses transaction-based, appraisal-based or desmoothed series is explored. Furthermore, the way leverage alters the response to risk factor changes is analyzed. Finally, this thesis improves the understanding of the use that should be made of real estate investments, for all exposures, while building a mixed-asset portfolio for various investment horizons. How the allocation to real estate is affected by the introduction of other alternative asset classes in the portfolio is investigated too.