We propose a flexible hedonic methodology for computing house price indexes that uses multiple imputation (MI) to account for missing data (a huge problem in housing data sets). Ours is the first study to use MI in this context. We also allow for spatial correlation, include interaction terms between characteristics, between regions and periods, and between regions and characteristics, and break the regressions up into overlapping blocks of five consecutive periods (quarters in our case). These features ensure that the shadow prices are flexible both across regions and time. This flexible structure makes the derivation of price indexes from the estimated regression equations far from straightforward. We develop innovative methods for resolving this problem and for splicing the overlapping blocks together to generate the overall panel results. We then use our methodology to construct temporal and spatial price indexes for 15 regions in Sydney, Australia on a quarterly basis from 2001 to 2006 and combine them to obtain an overall price index for Sydney. Our hedonic indexes differ quite significantly from the official index for Sydney published by the Australian Bureau of Statistics. We also find clear evidence of convergence in prices across regions from 2001-3 (while prices were rising), and divergence thereafter. We conclude by exploring some of the implications of these empirical findings.