An individual’s social network and community — their ‘social capital’ — has been thought to influence outcomes ranging from earnings to health. But measuring social capital is challenging. In two papers in this week’s issue, Raj Chetty and his colleagues use data on 21 billion friendships from Facebook to construct a Social Capital Atlas containing measures of social capital for each ZIP code, high school and college in the United States. The researchers measure three types of social capital: connectedness between different types of people, social cohesion and civic engagement. They find that children who grow up in communities where people of low and high socio-economic status interact more have substantially greater chances of rising out of poverty. The team then examines what might limit social interactions across class lines, finding a roughly equal contribution from lack of exposure — because children in different socio-economic groups go to different schools, for example — and friending bias, the tendency for people to befriend people similar to them.