Supporters and critics of free schools in England have had differing expectations about whether free schools would emerge in socially disadvantaged areas, and whether they would become socially selective. We investigate the outcomes, using information from the first three years since the introduction of the first new schools in 2011, drawn from the National Pupil Database. We find that, as policy proposers had expected, free schools have been situated in neighbourhoods that are somewhat above average in terms of the proportions entitled to free school meals, a well-known indicator of social disadvantage. Nevertheless, we also find that the free schools are socially selective within their neighbourhoods. These two effects balance out so that, overall, compared with the average for all England, there are no great differences between the social composition of secondary free schools and that of the national average. However, at primary school level there is evidence that free schools are enrolling children with above average ability. Moreover, there are very substantive differences between the ethnic composition of free schools and other schools. Despite these differences, we find little evidence that the presence of free schools is having an effect on the social composition of intakes to other schools in their neighbourhoods or on segregation in the local authority as a whole. We suggest this may be because there are still too few free schools, with very small cohort sizes, recruiting from very dispersed areas.