In an earlier article (Borst 2008), the Fourier expansion method was compared to two other methods of accounting for time in a hedonic model, namely, the quarterly dummy variable and the quarterly spline variables. In these three methods, the functional form of the time dependency on sale price must be specified in advance. In this follow-on study, there is no explicit specification of the variation of value over time. Rather, the model itself is allowed to vary. As before, the means of judging the efficacy of a particular method is to examine the trends it produces for reasonability and to compute the improvement in predictive accuracy, compared to a baseline model with no mechanism for accounting for the effect of time on value. This article first describes the previous methods. It then introduces the time-varying parameters model. Study data, time trends, and accuracy comparisons are given for all four methods. A simple cross-check on the validity of the developed time trends provides an independent verification of their efficacy. Finally, conclusions about the time-varying parameters method are presented.
Housing market - models
This article continues the exploration of the treatment of time trends in hedonic models begun with the discussion of the Fourier expansion in Volume 5, Issue 4.
Borst, R. A. (2009). Time-varying model parameters : Obtaining time trends in a hedonic model without specifying their functional form. Journal of Property Tax Assessment & Administration, 6(4), 29-36. Retrieved from https://researchexchange.iaao.org/jptaa/vol6/iss4/2