This paper analyses the diversification potential of direct and listed real estate markets. In a first step, we consider a parametric and non-parametric correlation analysis among 15 national real estate market annual return indices covering the period 2001-2017. To correct for small sample potential biases, confidence intervals are estimated with a percentile t-bootstrap methodology. Our findings are in line with previously reported results in the literature in that it confirms that real estate markets do not present significant diversification opportunities except for a few countries including Australia, Germany and South Africa. In a second step, we calculate the associated coherences to distinguish between long-run  and short-run diversification benefits. It appears that the prospects for diversification are generally low in the long run as markets tend to move together internationally whereas diversification can be achieved in  the short run in many cases. Finally, we consider a well-known CAPM specification using the S&P global All Property Index. The model is estimated in a panel fashion as in Petersen (2009) and Ando and Bai (2014) with country fixed effects. Three CAPMs are estimated using the direct real estate returns by property type in the 15 markets, including their short and long-run components.