Estimating distance gradient in Bratislava with different types of distance measurements

Proximity to the central business district (CBD) is considered to be a significant determinant of the real estate price formation as households are willing to pay more for housing that is closer to CBD in order to decrease the commuting costs. There is no consensus on the measure of distance, rather, various measures are used by different authors. Within a city, the opportunity costs of time in transportation represent a major part of the commuting costs. Thus, many authors believe that using travel time represents the accessibility to CBD more accurately than distance variables. An interesting question is how different distance measures influence the overall quality of the hedonic model and the estimated parameters in the model.This article draws on recent development in GIS techniques for computing travel distance and time between a number of points on a map using different travel modes to examine hedonic price functions for the apartment market in Bratislava. If real estate market is in equilibrium, attractiveness of location is fully capitalised into property prices. Our research is based on information from 1334 internet offers of apartments for sale in year 2016, which includes information about basic characteristics of individual apartments and the building in which the apartment is located. Using Google Maps Distance Matrix API service, we computed travel distance and travel time to the city centre for walking, public transportation (transit) and driving travel mode. In addition to distance to the city centre we also include the district in which the apartment is located. As the hedonic models of apartment prices usually suffer from the presence of spatial autocorrelation leading to biased and/or inefficient OLS estimates, we use both spatial and non-spatial models.In general, we have confirmed the negative slope of the distance gradient in Bratislava. District specific parameters generally decline with moving away from CBD and all distance measures showed negative coefficients. In addition, we found that the role of district-specific particularities is over-estimated in both spatial and non-spatial models when using crow fly measurement of distance to the city centre compared with other types of distance measurements.