Insight into apartment attributes and location with factors and principal components

Purpose - The purpose of this paper is to better understand the spatial structure of the Lyon urban area focusing on real estate. For this, two aims are formulated. The first aim is to identify and geographically analyse latent structure underlying apartment variables and location. The second aim is to decrease a number of explanatory variables in a hedonic model of real estate prices applying latent constructs. Design/methodology/approach - For the first aim of a parsimonious representation among measured variables, exploratory factor analysis is applied. For the second aim of data reduction, principal component analysis (PCA) is used. The exploited regression methodologies are global and geographically weighted ordinary least squares. Findings - Four factors are extracted, of which two represent apartment attributes and other two – location attributes. Principal components provide better insight into location attributes dividing the service employment centres into two geographical groups. The inclusion of principal components in hedonic price equation instead of initial location variables decreases goodness of fit, but does not gradually change non‐location estimates and other parameters. Originality/value - Differently from previous applications of factor analysis and PCA in the real estate domain, oblique rotation is applied, which allows the extracted factors or components to be correlated. The scores of factors and components are interpolated from points to raster maps creating a continuous geographical distribution. Hedonic models with and without principal components are compared in detail.