A dualistic manner in spatial analysis : location and sector-based investigations for housing price determinants : ersa : regional integration: Europe, the Mediterranean and the world economy : 53rd ERSA Congress, 27-31 August 2013

Investigation of the patterns in housing preferences has always been a strong emphasis on spatial analysis literature, mostly to the extent that housing price varies within urban macro form. The purpose of this paper is to examine the variances of housing prices in Istanbul, by using separate determinant sets in separate models, while employing location based investigation referring to distance and/or proximity effect and sector based investigation, referring to land price effect as a most prominent indicator in residential market. GIS-based spatial analyses methods are utilised in these dual models, in order to describe the variances. Distance effect as a part of locational domain in hedonic price model which represent the demand side of housing price. In this model, distance among the bundle of alternative characteristics is assumed to affect any property price, as a retail product. Although this paper does not accomplish a full hedonic price model, it calculates the effect of distance and/or proximity determinants in housing prices in Istanbul. This first model is conducted with a multiple regression analysis with the contribution of independent variables as population density and distance to the CBD, sub-centre, main transportation arteries, and coast. As assumed in the hypothesis, the result indicates that 66.3% of the variance in housing price can be determined by these variables. In the second model, land price effect in the variance of housing price is investigated in order to illustrate the emergence of residential market. Within this investigation residential market is defined depending on land values whether there are areas where land prices are high but housing prices are relatively lower, or vice versa. In other words, if land and housing prices are not in consistency, there may be possibility of changing residential market features, to be interpreted with the inclusion of changing determinants. In this model, geographic weighted regression is conducted and then relative predictive performance is tested by standard residuals which refer to inconsistency of the variances between land and housing price levels. The paper finalizes with a comparison of the results obtained from location and sector based models. Concluding remarks indicate that spatial variations in housing price, predicted within these two models can also estimate the segmented structure of the city, Istanbul and residential market itself. Such a dualistic manner which corresponds to combination of statistical and spatial analysis techniques can be useful in order to comprehend the residential market in a spatial pattern.