Parametric and Non-Parametric Statistical Methods in the Assessment of the Effect of Property Attributes on Prices

Abstract

Abstract One of the basic problems in the comparison-based property valuation process is to determine the influence of property attributes on their price differential. Due to the qualitative character of the majority of property attributes as well as to the distributions of both prices and attributes, their effect on the price differential is increasingly often assessed by means of non-parametric statistical methods. As a tool for determining the effect of attributes on prices, many authors propose parametric methods, in particular multiple regression models. The study presents a comparison of the results of property market attribute weight estimation obtained by means of the Spearman rank correlation coefficient with the ceteris paribus adjustment and the multiple regression model based on a set of transactions with built-up land property. In both of the analyzed methods, qualitative variables were modeled with the use of the Osgood semantic differential scale. The results of the analysis show the equivalence of the applied methods. Property attribute weights calculated using the method based on the rank correlation coefficient with the ceteris paribus adjustment and the multiple regression model, both with the same level of relevance, showed almost identical values. This indicates that both parametric and non-parametric methods can be used to estimate weights.