Many western European housing policies have tried to increase the residential mix of advantaged and disadvantaged groups. Unfortunately, policymakers have given little consideration to how these groups will interact as neighbours. There are numerous theoretically grounded mechanisms by which the social mix of a neighbourhood may influence socio-economic outcomes of its residents. These mechanisms differ on the basis of which group is generating the social externality in the neighbourhood, whether this externality is positive or negative, whether it affects all residents equally, and whether the marginal externality generated by adding one more member of a particular group is constant, proportional, or is characterized by a threshold effect. This paper demonstrates that a social mix housing policy can be justified only under a circumscribed set of the preceding parameters. Indeed, depending on the mechanism assumed, social efficiency implies that neighbourhoods should be either: equally mixed, have the disadvantaged group dispersed as widely as possible, or rigidly segregated; for other mechanisms, mix becomes irrelevant. Thus, for formulating and justifying a mixed housing policy on either efficiency or equity grounds it is crucial to understand exactly what sort of neighbourhood effect(s) is operating in neighbourhoods.