The proposed paper is about the relationship between new urban developments and the residential land markets in Belgium. In this country, residential land market processes are strongly influenced by the predominance of self-provided constructions, with the fact that a building site is often exchanged twice before being constructed: a first transaction is happening between an initial owner and a land developer (lotisseur-amÈnageur) wishing to service the site; a second transaction is afterwards happening between the land developer and a household wishing to acquire a plot of land in order to built its own residence. The analysis is about the ìsecondaryî markets, where households represent the demand component. The key objective has been to analyse how those markets are spatially structured. Our empirical analysis is based on the theoretical idea that two key factors explain the spatial structure of residential land market: * data availability on prices (relationship with price comparison made by market participants when they prepare negotiations); * substitutability (relationship with the territory prospected by the demand wishing to acquire a plot of land). // The research is based on spatial econometrics, with the use of auto-regressive multiple regressions (estimator of the maximum likelihood). Models that distinguish and compare Flanders and Wallonia have been produced in order to statistically explain the communal mean land prices (n= 242 in Flanders; n = 259 in Wallonia). Three explanation components have been considered. The first is related to demand; the mean income of the population is here a key variable. Concerning the supply side ñ the second explanation component ñ a key data is the quantity of land available for new houses in planning documents. Demand and supply data have been tested at two spatial levels ñ municipality and job catchment area ñ in order to apprehend the substitutability factor. The information is the third explanation component. It is related with the factor of data availability and has been treated through a spatial autoregressive variable. For each commune, the value of this variable is equal to the mean of the dependent variable observed in the neighbouring municipalities. Property researches are unfortunately weakly developed in Belgium and our work represent a first attempt to model land prices in this country. Although some academic publications are dealing with the issue, the spatial structure of market is not yet a well developed property research theme. Moreover, our analysis has been elaborated for the whole of Belgium while the literature on this subject is based on hedonic models elaborated for smaller areas. Our statistical treatments ñ with the fact that the autoregressive variable has a very high explanation power ñ reveal that the comparison mechanism is a key explanation of land prices (relation with the spatial structure factor of data availability on prices). In other words, the market value of a plot of land is, ceteris paribus, very strongly dependent of the market value recorded in the neighbouring plots. From a methodological point of view, this result demonstrates that, in the analysed context, an autoregressive equation form is necessary to achieve robust statistical results. Concerning the substitutability (second spatial structure factor taken into account), the analysis has put into the light that the territory prospected by the demand is highly dependent on the quantity of land available for new constructions. Where the planning authorities have ìgenerouslyî dedicated important superficies for new constructions, the scale of the municipality is appropriate to describe the spatial behaviour of the households wishing to acquire a plot of land. On the contrary, where land available for new construction is scarce, the households are obliged to prospect larger territories. In this case, the scale of the job catchment area is then more appropriate that the municipal scale to consider market mechanisms.