Nonlinearity of housing price structure

Abstract

Purpose – The purpose of this article, starting from linear regression, was to estimate a switching regression model, nonparametric model and generalized additive model as a semi-parametric model, perform function estimation with multiple nonlinear estimation methods and conduct comparative analysis of their predictive accuracy. The theoretical importance of estimating hedonic functions using a nonlinear function form has been pointed out in ample previous research (e.g. Heckman et al. (2010). Design/methodology/approach – The distinctive features of this study include not only our estimation of multiple nonlinear model function forms but also the method of verifying predictive accuracy. Using out-of-sample testing, we predicted and verified predictive accuracy by performing random sampling 500 times without replacement for 9,682 data items (the same number used in model estimation), based on data for the years before and after the year used for model estimation. Findings – As a result of estimating multiple models, we believe that when it comes to hedonic function estimation, nonlinear models are superior based on the strength of predictive accuracy viewed in statistical terms and on graphic comparisons. However, when we examined predictive accuracy using out-of-sample testing, we found that the predictive accuracy was inferior to linear models for all nonlinear models. Research limitations/implications – In terms of the reason why the predictive accuracy was inferior, it is possible that there was an overfitting in the function estimation. Because this research was conducted for a specific period of time, it needs to be developed by expanding it to multiple periods over which the market fluctuates dynamically and conducting further analysis. Practical implications – Many studies compare predictive accuracy by separating the estimation model and verification model using data at the same point in time. However, when attempting practical application for auto-appraisal systems and the like, it is necessary to estimate a model using past data and make predictions with respect to current transactions. It is possible to apply this study to auto-appraisal systems. Social implications – It is recognized that housing price fluctuations caused by the subprime crisis had a massive impact on the financial system. The findings of this study are expected to serve as a tool for measuring housing price fluctuation risks in the financial system. Originality/value – While the importance of nonlinear estimation when estimating hedonic functions has been pointed out in theoretical terms, there is a noticeable lag when it comes to testing based on actual data. Given this, we believe that our verification of nonlinear estimation’s validity using multiple nonlinear models is significant not just from an academic perspective – it may also have practical applications.

A Pricing Model for Urban Rental Housing Based on Convolutional Neural Networks and Spatial Density: A Case Study of Wuhan, China

“…Therefore, it has received considerable attention and has effectively applied in the field of economy and real estate [14,35,36]. However, GWR also assumed that the relationships between independent and explanatory variables are linear, which has a clear limitation in housing price modeling, because the patterns in the housing and rental price are nonlinear and complicated [13,37]. Up-to-date studies also pointed out the disadvantages of GWR in complex spatial prediction tasks [8,38] and criticized for its reliability and restrictions [39].…”

“…These diverse locational and neighborhood characteristics contain very complex relationships, and the urban facilities relevant to housing contain a massive quantity of spatial density characteristics. First, complicated relationships exist among the structural, locational and neighborhood variables of housing, and these relationships cannot be easily characterized in a simple way [13,14]. If these variables are treated as a one-dimensional vector to be modeled, as in ordinary least squares (OLS), geographically weighted regression (GWR), or some one-dimensional deep learning models [15,16], the accuracy of price forecasting would be limited.…”

Predictive Statistical Cost Estimation Model for Existing Single Family Home Elevation Projects

“…The GAM is used to identify the relationship between input and output variables in nonlinear models. It relaxes the strictly linear relationship between the response and the regressors, allowing regressors to have a general and flexible relationship to the response, but maintains additive or non-interactive structure (Moore et al, 2011;Shimizu et al, 2014;Larsen, 2015;Taghinezhad et al, 2020b). Although we do not consider it here, GAMs can additionally accommodate non-normal responses with added flexibility through a nonlinear link function (Xiang, 2001;Han et al, 2009;Calabrese and Osmetti, 2015).…”

“…Lowe et al (2006) used multiple linear regression, Jrade and Alkass (2007) developed a set of linear regression models in a computer-based cost estimation program, and Sonmez (2008) used a combination of linear regression and bootstrap techniques for construction cost modeling. Additionally, Shimizu et al (2014) used switching regression model and generalized additive model (GAM) to predict the housing price, and Liu et al (2018) used random forest and GAM to predict construction productivity using environmental factors. Specific to natural hazard mitigation, Jafarzadeh et al (2015) applied multiple linear regression to establish construction cost models for seismic retrofit of confined masonry buildings.…”