A PCA Factor Repeat Sales Index (1973-2001) To Forecast Apartment Prices in Paris (France)

In this paper we address the issue of building a repeat sales index, based on factors. This is an extension of a companion paper, Baroni, Barthelemy and Mokrane (2001, BBM) in which we had built a factorial index as a selected linear function of existing economics and financial variables. Here we offer a more general and robust model based on a Principal Components Analysis (PCA). We apply this methodology to the Paris residential market. We use the CD-BIEN database that contains more than 220 000 repeat sales transactions for residential apartments in the Paris area covering the period 1973-2001 period. Our PCA index for the Paris and close surrounding area is estimated and its characteristics and robustness are analysed depending on: estimation period, choice of observations, periodicity and reversibility. We then compare it to the traditional WRS repeat sales index developed by Case & Shiller (1987). Finally we show that contrary to the WRS index, our index can be used to forecast apartment prices.