Like many other governments, the Dutch government has simultaneously pursued the contradictory goals of liberalizing the housing market and countering the concentration of low-income groups. This paper discusses how the tension between promoting market forces and countering segregation has played out, using Amsterdam as a case study. The findings suggest that the policy may have mitigated but did not prevent a deepening division between the city’s increasingly privileged core and its periphery. This is at least in part because social mixing was pursued also in neighbourhoods already prone to gentrification. Like several other Western European governments transitioning from a universal to a residual welfare state, the Dutch government has been faced with a dilemma. On the one hand, it wants to implement market rule in the field of housing, which will, ceteris paribus, result in cities that are segregated along class lines. On the other hand, the government wants to prevent segregation and the formation of ‘no go areas’, which requires state intervention. Through the so-called ‘restructuring policy’ the government has pursued the contradictory goals of simultaneously promoting market forces and countering segregation. By selectively investing in neighbourhoods with a weak position in the urban housing system, the government claimed to promote ‘social mixing’ and prevent ‘divided cities’ (Priemus et al. 1998; Uitermark 2003; Uitermark et al. 2007; Boschman et al. 2013). The Netherlands is far from unique in this respect. As Leeset al.(2012, p. 2) note, encouraging socially mixed ‘communities by bringing middle-income people into low-income neighbourhoods has become . . . a major policy goal in North America and in a number of Western European countries’. It is sometimes suggested that this amounts to a thinly disguised gentrification strategy as public and private actors ‘recapture prime urban real estate despite the resultant displacement of many of those households that the strategy is purported to help’ (Joseph and Chaskin, cited in Lees et al. 2012, p. 7; see also Lees 2008).