In this paper a new residential property price index (RPPI) for Austria a socalled Òspatial imputation indexÓ is presented. The basic principle of this index is to evaluate hedonic price models over all census tracts in Austria for representative bundles of characteristics and to compare the results over time. By this it is possible to aggregate on arbitrary spatial scales above census tract levels, taking into account various perspectives on the housing market by applying different weighting schemes for aggregation. Methodologically, we use Generalized Additive Models (GAM), which account for possibly nonlinear functional relationships using regression splines and for unexplained spatial heterogeneity (beyond what can be explained by location covariates) using random effects. The paper gives an overview over the methods and variables used for the hedonic models and provides examples for the regional index on census tract levels all over Austria. The results of the aggregated index are further used by the Austrian central bank.