The matrix exponential spatial specification approach for big datasets: The analysis of Istanbul office market

Abstract

Our aim is to develop a hedonic office rent model considering the spatial dependency in order to determine the factors affecting office rents. Based on a comparison between spatial lag, spatial error, and spatial Durbin models, the spatial lag model was selected as the most appropriate model explaining the relationship according to some criteria. Spatial models were estimated using the data on rent levels and property characteristics of 2348 business offices located in 28 different counties of Istanbul during the first quarter of 2018. According to the estimation results of the spatial lag model, the most effective independent variables are average vacancy rate, building type, and Bosporus view. Since the big dataset might cause some misleading estimations, the matrix exponential spatial specification model was estimated. It was observed that the estimated coefficients of both models are almost identical.