<p>This paper investigates the macroeconomic drivers influencing excess real estate return indices in the US, Canada, and the UK. We conduct this analysis in a Big Data setting with a large set of predictors. We use tree-based methods that address the high dimensionality problem and detect nonlinearities in the relationship of interest. We focus on both the short-term (1 month) and long-term (2 years) drivers for each country and explore their time-varying dynamics. We document heterogeneity in the macroeconomic drivers across countries. Demand-side, labor market, and interest rate variables are relevant in the US and Canada, both in the short and long run. In the UK, a composite leading indicator, unemployment, and the Economic Policy Uncertainty index are important drivers. We document significant time variation of drivers’ importance. Our out-of-sample evidence shows that predictability is weak in the US and Canada but reasonably good in the UK when compared to a benchmark model. Thus, our results show that the US and Canadian markets are more informational efficient than the UK market. A portfolio evaluation tends to support these results. Finally, we find no effect of geopolitical variables on real estate markets.</p>