The Predictive Accuracy of Residential Vacancy Chain Models

Abstract

The focus of this report is on the predictive accuracy of residential vacancy chain models. There are two objectives. The first is to define the mathematical structure of a pair of multisectoral residential vacancy chain models. These models simulate the intersectoral transfer of vacant housing opportunities in response to either the creation of new opportunities or the absorption of existing ones. Residential mobility is given as an analytical by-product. The second objective is to calibrate these models and gauge the accuracy of their projections. Models with six housing sectors each are calibrated with 1975-80 data from three different Swedish municipalities (Gävle, Jönköping and Vasteras). Calibrations are used for projections into the proximate 5-year period. Projections of intersectoral residential mobility for 1980-85 are compared statistically with observed moves for the period. Coefficients of determination range between 0.93 and 0.97. Log-linear analysis shows projection error ranging from 3 to 12 per cent. The projective accuracy of these models exceeds that of any known alternative. The empirical success of these models points to their value in understanding processes affecting the formation and change of urban residential structure.