In this paper, we calculate a transaction based price index for apartments in Paris (France). The heterogeneous character of real estate is taken into account by using a multiple regression model. The functional form is specified using a general Box/Cox-function. The data base covers about 65% of all transactions of the Paris housing market in 1990-1999. Correction for incomplete data leads to a sample of 84†686 transactions. This sample is by far the largest ever used in comparable studies, and is therefore internationally a unique data coverage for a housing market.