Housing appraisal under model uncertainty: Bayesian model averaging method

Abstract

This study aims to examine which variables have a higher impact on the determination of market values of houses than the others through the Bayesian Model Average method. Therefore, variables, compiled from appraisal reports prepared with 742 housings from different districts in the city center, were used in the study. The 12 different independent variables at different measurement levels as continuous and categorical, thought to affect the housing prices were used in the study. The study results were presented by comparing the traditional statistical application and Bayesian Model Average methods. Both methods concluded that the variables such as the floor area (m2), number of rooms, and year of construction have a very strong effect on housing prices. The results of the remaining nine explanatory variables revealed that there were qualitative differences between the two methods. Setting the factors that determine the values of the houses and developing reliable statistical models are critical in real estate appraisal studies.