In this paper we address the issue of building a factor repeat sales index based on factors. This is an extension of a companion paper, Baroni, BarthÈlÈmy and Mokrane (2001, BBM) where we built such an index but as a selected linear function of existing economics and financial variables. Here we construct other factors based on a PCA. The estimator is more robust. 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 properties in the Paris area covering the period 1983-2001 period. We estimate the PCA BBM (2004) index for the Paris and close surrounding area. We study its properties according to the estimation period, the choice of the observations, periodicity and reversibility. Then we compare it to the traditional repeat sales index developed by Case & Shiller (the WRS index in 1987). Finally we show our methodology can be used to forecast apartment prices in opposite to WRS index.