Price Indexes for Short Horizons, Thin Markets or Smaller Cities

We propose a modification to the repeat sales procedure that incorporates pricing information from houses in close proximity that have sold only once. This nearest neighbor framework allows for both idiosyncratic property-specific effects and for common location effects in constructing sale pairs. The distinguishing feature of our model is that a significant percentage of discarded (non-repeat sale) observations are utilized. Like the repeat sales approach, the availability of housing attributes is not required to implement our model, in contrast to other hybrid models. The only additional calculation required is pairwise distances, which are used in lieu of property attributes in hybrid models. Our resulting model is not only parsimonious but also produces relatively tighter pricing interval estimates. A key advantage of our approach is that it can be used to obtain price estimates in thin markets and smaller cities that have relatively fewer transactions.