A repeat sales index robust to small datasets

Abstract

PurposeThis paper aims to test the robustness of the trend and volatility estimations for two indices: the classical Weighted Repeat Sales and a PCA factorial index. The estimations are computed from a dataset of Paris commercial properties.Design/methodology/approachFirst, two methodologies are presented, and then the dataset. Finally, the impact of the number of transactions per period are tested on the trend and volatility estimates for each index, and an interpretation of the results are given.FindingsThe trend and volatility estimates are biased for the WRS index and not for the PCA factorial index when the periodicity increases. Consequently, the level of the index at the end of the computing period is significantly different for various periodicities in the case of the WRS index. Globally, the PCA factorial seems to be more robust to the number of transactions.Originality/valueAs suggested by D. Geltner, commercial properties indices have to be built using repeat sales instead of hedonic indices. The repeat sales method is a means of constructing real estate price indices based on a repeated observation of property transactions. These indices may be used as benchmarks for real estate portfolio managers. But the investors, in general, are also interested in introducing real estate performance in their portfolio to enhance the efficient frontier. Thus, expected return and volatility are the two key parameters. To create and to improve contracts on real estate indices, trend and volatility of these indices must be robust regarding to the periodicity of the index and the volume of transactions.