Urban infrastructure and its role in shaping Moscow’s residential real estate market

Purpose This paper analyzes the apartment prices in Moscow using geographically weighted regression (GWR) to account for the spatial non-stationarity and local variations in the pricing of Moscow’s heterogeneous real estate market on the sample of 3,504 flats. The results suggest that the impact of infrastructure on property values varies considerably across neighborhoods, which is indicative of underlying socioeconomic trends and urban development patterns which are not captured by traditional global ordinary least squares (OLS) models. By empirically updating and extending the largely city-wide, transaction-era literature on Moscow housing with a post-pandemic, spatially disaggregated data set, and by contrasting global and locally varying estimates, this paper contributes to the urban economics literature on spatial heterogeneity in housing markets and offers policy-relevant, location-specific evidence for stakeholders engaged in urban development and residents’ investment decisions. Design/methodology/approach The empirical strategy involves a comparative analysis. First, an OLS model is established to provide a baseline understanding of the average city-wide effects. Subsequently, the GWR model is implemented to uncover and map the spatial variation of these effects, offering a more nuanced and geographically contextualized interpretation. Findings In particular, the analysis highlighted the significant role of proximity to schools, medical facilities, transportation facilities, cultural institutions and recreational areas in shaping housing prices. Proximity to schools consistently raised property values. However, the impact of medical facilities was mixed, with some districts showing price increases due to convenience, while others experienced price reductions, possibly due to noise or negative associations with hospitals and illnesses. Proximity to railway stations boosted prices in most areas, particularly in the southeast, while the effect of airports was more complex, with prices increasing at an optimal distance of around 5.5 km but decreasing if too close probably due to noise and pollution. Research limitations/implications Several limitations qualify the findings and should guide their interpretation. First, the price variable is constructed from asking prices posted on cian.ru rather than actual transaction prices; asking prices can diverge from final sale prices through negotiation, and this gap is unlikely to be constant across neighborhoods or property types, so estimated infrastructure premiums should be interpreted as premiums in listed rather than realized prices. Second, the cross-sectional design, based on a single month data (May 2024), rules out any causal interpretation of the estimated coefficients: proximity to infrastructure and housing prices are more plausibly jointly determined, since infrastructure is itself sited partly in response to existing or anticipated demand for housing (reverse causality), and neither the OLS nor the GWR/ multiscale GWR (MGWR) specification used here − unlike instrumental-variable or quasi-experimental designs based on the opening of new infrastructure − can separate the capitalization of infrastructure into prices from the endogenous location of infrastructure itself. Third, despite the inclusion of a broad set of structural, locational and infrastructure variables, omitted variable bias cannot be ruled out; unobserved factors such as building-specific renovation quality, precise floor plan or local air and noise pollution are likely correlated with both infrastructure proximity and price. Fourth, as discussed in the literature review, this paper focuses on GWR/MGWR because the research question concerns spatial heterogeneity in coefficients rather than global unbiasedness; it does not report a formal comparison against spatial lag or spatial error specifications, which would instead prioritize correcting spatial autocorrelation in a single global model, and such a comparison is left for future work. Finally, because the sample is restricted to the secondary housing market of a single, atypically large and monocentric post-Soviet city, the magnitude − though plausibly not the qualitative pattern − of the estimated infrastructure premiums may not generalize directly to other Russian cities or to primary-market housing. Practical implications These findings can guide urban planning and infrastructure investment, suggesting that enhancing social infrastructure in underdeveloped areas could reduce price disparities and increase the overall attractiveness of residential areas. The GWR model’s ability to capture these spatial variations offers a valuable tool for stakeholders involved in urban development, enabling more precise and localized decision-making that reflects the diverse dynamics of Moscow’s real estate market. Originality/value This approach effectively captured the spatially diverse relationships between the price per square meter of flats and various social infrastructure objects, offering more nuanced insights into the factors affecting real estate pricing across different areas of Moscow.