Modelling Neighbourhood Effects in Three Dutch Cities Controlling for Selection

Abstract The non-random selection of people into neighbourhoods complicates the estimation causal neighbourhood effects on individual outcomes. Measured could be result characteristics context, but they also from selecting based their preferences, income, and availability alternative housing. This paper examines how effect income is altered when geographic correction terms are added as controls, these results vary across three Dutch urban regions. We use a two-step approach in which we first model selection, then include choice components estimating income. Using longitudinal register datasets for major cities: Amsterdam, Utrecht Rotterdam, multilevel models, analysed individuals who moved during 5-year period. show that all cities, average becomes much smaller after controlling explicitly modelled selection. suggests studies do not control most likely overestimate size effects. For strongest followed by Amsterdam Utrecht.