TRANSLATING PRIOR INFORMATION ACROSS SPECIFICATIONS TO IMPROVE PREDICTIVE ACCURACY

Unrestricted nonlinear models typically outperform their simple linear counterparts in the hedonic pricing and mass assessment fields. Economic theory, however, suggests prior information that most naturally applies to the simple linear model. This article examines the consequences of translating this prior information across specifications. The results show that the addition of the prior information improved the ex-sample prediction accuracy over all sample sizes examined. The prior information effectively augments the sample size, thus extending the domain of these models.