This paper examines whether the predictability of securitized real estate returns differs from that of stock returns. It also provides a cross-country comparison of securitized real estate return predictability. In contrast to most of the literature on this issue, the analysis is not based on a multifactor asset pricing framework as such analyses may bias the results. We use a time series approach and thus create a level playing field to compare the predictability of the two asset classes. Forecasts are performed with ARMA and ARMA-EGARCH models and evaluated by comparing the entire empirical distributions of prediction errors, as well as with a trading strategy. The results, based on daily data for the 1990-2007 period, show that securitized real estate returns are generally more predictable than stock returns in countries with mature and well established REIT regimes. ARMA-EGARCH models are found to have portfolio outperformance potential even in the presence of transaction costs, with generally better results for securitized real estate than for stocks.