A multilevel analysis of housing submarkets defined by the municipal boundaries and by the street connections in the metropolitan area: Istanbul

Abstract

This paper focuses on the performance of the multilevel model within the metropolitan area and explores the submarkets of housing prices at the metropolitan level. In addition to the district and neighborhood segmentation within municipal boundaries, areas using the same street for transportation are segmented into submarkets to reveal the effect of the street on price. A fully-random multilevel model is used for finding out a correlation between the prices of housing units and their location. This study compares four submarket modeling strategies based on the data on housing sale transactions in Istanbul. The first model is a simple hedonic model using ordinary least squares and taken as a city-wide 'benchmark' model. Three different types of multilevel models are used for models 2,3 and 4. Model 2 and model 3 use district and neighborhood boundaries, which are respectively predefined as spatial divisions. Model 4 represents the geographically smaller areas that use the same street for transportation. The results reveal that multilevel models with smaller spatial units of geography are capable of improving predictive performance in determining dwelling prices in the metropolitan area and that the sub-regional spatial units that use the same street for transportation have an effect on the spatial dimension of submarket boundaries.