This article analyses the formation of urban and suburban employment centralities and their impact on local housing values. Urban sprawl combined with the formation of suburban employment poles have produced more complex spatial patterns characterized by a polycentric city. This paper assumes that these changes in urban forms have influenced residential dynamics and have thus influenced housing prices. The presence of suburban centralities is expected to improve local attractiveness and thus to be capitalized in housing prices. The aim of the article is to disentangle the specific impact of the various factors explaining housing prices differentials. We are specifically interested in the influence of economic centralities among other factors such as accessibility or the presence of diverse amenities. We use semi-parametric hedonic regression modelling which allows us to estimate sharply the impact of urban subcenters on housing prices among other factors, by considering varying distance thresholds and heterogeneous spatial patterns of housing prices. The dataset come from residential transactions recorded by notary offices. The precise geographical coordinates of each transaction is used in addition with economic value and intrinsic characteristics of the buildings, allowing a sharp estimation of the spatial pattern of housing prices. Explanatory variables come from several data sources recorded at a fine spatial scale and concern mainly accessibility, socio-economic attributes, local amenities and employment density. The impact of economic centralities on housing prices is studied in three major French cities: Paris, Lyon and Bordeaux. This allows a comparison of three different urban patterns. The contribution of the paper is to investigate the impact of urban structure on housing prices and to overcome several limitations of hedonic models using semi-parametric regressions. This paper relies on the interactions between urban economics and real estate economics.