Using transaction prices and characteristics including longitude and latitude for dwellings sold in a city in a given time period we show how geospatial house-price spline surfaces can be constructed that impose minimal prior restrictions on the functional form of the hedonic model. Prices for individual dwelling can then be imputed from the spline surface, and these imputed prices used to compute single or double-imputation hedonic price indexes for regions in a city. Using repeat sales in our data set as a point of reference we find that our double imputation price relatives are more accurate than their single imputation counterparts, thus leading us to prefer the double imputation approach. We then show how the double imputation approach can be extended to include dwellings that did not sell in either of the periods being compared, thus potentially increasing the representativity of the index. Also, we develop an approach for dealing with missing characteristics in the data set which makes maximum use of the available data. We illustrate our methodology using data for Sydney, Australia for the years 2001 to 2009 and highlight some differences between our spline-based indexes and more standard indexes obtained by estimating either a linear or semilog hedonic model.
Keywords: Hedonic price index; Geospatial data; Spline; Missing date
Biography: Robert Hill is Professor of Macroeconomics at University of Graz. He holds a PhD from University of British Columbia, and his research interests include the housing market, price indices, comparisons of living standards across countries, and the design of performance benchmarks for funds. He has published in journals such as American Economic Review, International Economic Review, Review of Economics and Statistics, and Journal of Econometrics and he is a member of the Technical Advisory Group of the International Comparisons Program at the World Bank.