Analysis of Impact of Accessibility on Residential Property Values in Gotri Area of Vadodara City, India Using OLS and GWR

Abstract

Valuation of residential property is a complex task involving multiple factors. Also, variation in residential property values are so high and diverse that it becomes difficult to analyse which variables are having greater or lesser impact on residential property valuation. The present study uses statistical regression technique to explain the effects of accessibility on residential property values in the Gotri area of Vadodara city, India. The article also includes a comparison between Ordinary Least Squares (OLS) and Geographically Weighted Regression (GWR) methods. The present study has used shortest distance via a road network to selected facilities/amenities from sample residential properties to assess the accessibility. Data on more than 30 parameters related to structural, locational and amenities were collected for 161 residential properties. One dependent variable i.e. residential property value per sq. ft. and fifteen explanatory (structural and accessibility) variables were considered for the study. Due to high multiple correlation between the variables, three variables were removed and Ordinary Least Square (OLS) regression, a global model applied on one dependent and twelve explanatory variables (R2 - 0.40 for ln values of variables). GWR, local model was also applied on same number of variables but did not execute. This was due to severe global or local multicollinearity i.e. redundancy among model explanatory variables. Finally, GWR was implemented considering one dependent and eight explanatory variables which executed successfully (R2 - 0.67 for real values of variables). GWR outperformed the OLS model which suggests that GWR can be favourably used for such applications.