Regression Promises And Aggregation Bias Illusions The Application Of Market Delineation To Land Valuation Models

Regression is one of the best tools for consistently deriving market-based adjustments in the appraisal of real estate. There are limitations in regression, however, and the potential for misleading results must be recognized. A principal violation of the validity of a regression model is aggregation bias, which has received limited attention in appraisal literature but is discussed here. This article shows how aggregation bias may creep into a regression model, and how professional appraisers are equipped to avoid it with the tools of market delineation and segmentation. There is a pervasive misunderstanding that a large data sample will minimize the negative impact of inappropriate or incorrect data points (comparables). In truth, the quality of data is as important in large regression modeling data sets as it is in small data sets in the conventional sales comparison approach. This article offers a case study of vacant industrial land to illustrate the misleading results of over aggregation (aggregation bias) and demonstrates how aggregation bias can be avoided through market delineation and segmentation. Only after a data set has been delineated and segmented in accordance with the market can issues related to model specification be effectively addressed.