Real Estate Prices: A Paris Repeat Sales Residential Index

Abstract

Abstract This paper offers an alternative methodology to estimate an index for Paris residential housing prices: the now standard repeat sales method (Case and Shiller, 1987). The same dataset as the official Notaires/INSEE Index is employed, and after discussing the main assumptions and their implications for estimation robustness, the estimated index for the 1982-2002 period is compared with the Notaires/INSEE Index. The findings indicate that the estimation is quite robust whatever the estimation period, and that this index is significantly different from the official residential index for Paris. Introduction Every real estate investor faces an objective difficulty concerning the measurement of real estate investment performance and risk. The reasons explaining this difficulty are numerous: an absence of centralized trading, or even price lists; a low degree of buildings or apartment turnover in investor portfolios; a lack of transparency in transactions; the heterogeneity and indivisibility of real estate properties; and a tradition of confidentiality in the industry. The official price index for the Paris residential market (Notaires/INSEE Index) is a hedonic one based on transaction prices. This index can be used to have an estimation of the price growth by comparing the index value at two different dates. For instance, the price return in capital between December 1985 and December 1991 is 249%, according to this index. This paper offers an alternative methodology to estimate an index for Paris residential housing prices: the now standard repeat sales method. The same dataset as the official index is used, and after discussing the main assumptions and their implications for estimation robustness, the estimated index for the 1982-2002 period is compared with the Notaires/INSEE Index. The next section contains a review of the repeat sales literature. The case and Shiller repeat sales framework (thereafter referred to as WRS) is developed next. The data available to estimate the Paris WRS sales index follows, along with the estimation results and robustness analysis. This sales index is then compared to the French one in an index perspective, and a comparison is then made on a return and volatility point of view. The paper wraps up with some concluding remarks. Repeat Sales Methodology: Literature Review The repeat sales method is a means of constructing real estate price indices based on a repeated observation of property transactions. The method begins by stating that the price of say good z at date t is a function of four terms: the good's quality at date t, the value of the underlying global real estate index at date t, a random walk variable linked to good i at date t and an error term, here again linked to good i at date t (modeled as a white noise). case and Shiller (1987) generalize the work of Bailey, Muth and Nourse (1963) and thus provide the first approach of repeat measures methods for construction of real estate indices. The main merit of this model based on repeat sales is that it does not presuppose any mechanical form for the behavior of the underlying real estate index. Since 1987, the model has attracted a lot of attention and has given rise to a number of improvements or critics. The rest of this section aims at presenting these improvements suggested in the literature along four issues: (1) the constant quality assumption; (2) selection bias; (3) revision; and (4) heteroscedasticity. The Constant Quality Assumption The standard repeat sales approach is based on the assumption that house quality stays unchanged between two sales. This assumption enables one to disregard the idiosyncratic error term of the house-its individual or specific characteristics. The constant quality assumption has an impact on the regression model used to estimate the index. Since it implies that the index changes are solely driven by time effects (i. …