58

The application of spatial interaction models to migration flows in an intra-urban environment should take into account the different types of geographical distance that influence flows within a city. Kilometric distance, topological distance, contiguity, and administrative and social discontinuities form a set of explanatory variables whose specific effects are difficult to dissociate without an adequate methodological framework. Poisson regression offers an interesting solution to this problem. Applied to residential migration in the Warsaw metropolitan area in 1985, it shows that the effect of kilometric distance is compounded by two barrier effects, one between the centre and the periphery of Warsaw, and the other between the western and the eastern bank of the Vistula. We could further enhance this model by combining factors of spatial proximity with measures of social proximity such as socio-economic similarities of the districts.