Willingness to Accept and Willingness to Pay on Residential Properties - A Hedonic Approach.xlsx

Data used in the study was obtained from Property24 August 2021 to January 2022. Property24 is a property portal where property listings for sale and to rent from leading real estate agents are advertised. Property24 helps sellers, home buyers, and renters find apartments, houses, townhouses, vacant land, and farms across South Africa. Residential properties from two leading metros in the Eastern Cape Province, Buffalo City Municipality and Nelson Mandela Bay were considered. The data obtained was limited to the data displayed on each property by Property24 that evaluations obtained The study adopted a parametric and non-parametric techniques to estimate the effects of hedonic housing characteristics on willingness to pay and willingness to accept. Hedonic regression model used multiple ordinary least squares regression to examine how each hedonic characteristic adds to the residential properties' entire worth. Ordinary least squares regression has been widely used in estimating house prices over the decades and it is well document in literature (McCord et al., 2018; My-Linh, 2020; Olamide &amp; Adepoju, 2013; Owusu-Ansah, 2013). The study further adopted the Multilayer perceptron of the Artificial Neural Network which is a predictive model that is made up of multiple techniques that measure each self-contained component independently despite its linearity (Papadopoulos et al., 2021). The Artificial Neural Network has gained momentum recently due to its ability to ability to predict house prices (Ghorbani &amp; Afgheh, 2017; Limsombunchai, 2004; Moreno et al., 2011). These techniques were used to estimate the effect of property attributes in influencing the willingness to accept and the willingness to pay in the Eastern Cape province of South Africa. <strong>Model specification and estimation techniques </strong> When building hedonic models, underlying assumptions of linearity, homoscedasticity, independence, normality, and correct model specification should be met. To meet these assumptions, the correct function form and the number of explanatory variables matters. The number of explanatory variables included in the model were determined after several pseudo model evaluations (Sirmans &amp; Macpherson, 2003).