Linear and Nonlinear Predictablity of International Securitized Real Estate Returns: A Reality Check

Abstract

This paper examines the short-horizon return predictability ofthe ten largest international securitized real estate markets,paying special attention to possible nonlinearity-in-mean as wellas nonlinearity-in-variance predictability. Although internationalsecuritized real estate returns are generally not predictable basedon commonly-used statistical criteria, there is much evidence forthe predictability based on economic criteria (i.e., direction ofprice changes and trading rule profitability), which is more oftendue to nonlinearity-in-mean. The forecast combinations forvarious models appear to improve the forecasting performance,while the allowance of data-snooping bias using White's realitycheck substantially mitigates spurious out-of-sample forecastingperformance and weakens otherwise overwhelmingly strongpredictability. Overall, there is robust evidence for thepredictability in many international securitized real estatemarkets.