A study of Facebook friend networks has shown that people from low-income households are more likely to grow up to make a higher salary if they had wealthier friends as children
1 August 2022
Children who grow up in low-income households but who make friends that come from higher-income homes are more likely to have higher salaries in adulthood than those who have fewer such friends.
“There’s been a lot of speculation… that the individual’s access to social capital, their social networks and the community they live in might matter a lot for a child’s chance to rise out of poverty,” says Raj Chetty at Harvard University. To find out if that holds up, he and his colleagues analysed anonymised Facebook data belonging to 72.2 million people in the US between the ages of 25 and 44, accounting for 84 per cent of the age group’s US population. It is relatively nationally representative of that age group, he says.
The team used a machine-learning algorithm to determine each person’s socio-economic status (SES), combining data such as the median income of people who live in the same region, the person’s age and sex and the value of their phone model as a proxy for individual income.
The median household income was found to be close to $58,000. The researchers then split the individuals into two groups: those who were below the median SES and those who were above.
If people made friends randomly, you would expect half of each person’s friends to be in each income group. But instead, for people below the median SES, only 38 per cent of their friends were above the median SES. Meanwhile, 70.6 per cent of the friends of people above the median SES were also a part of the same group.
The team then compared these figures to economic mobility figures produced by a research project at Harvard University called the Opportunity Atlas. This project uses census and tax data to determine the average household income at age 35 for someone born in the US between 1978 and 1983. The figures are categorised by race, gender, location and parental income.
Chetty’s team combined the data with the Facebook analysis and found that poorer children were more likely to have higher incomes in adulthood if they were born in areas where poorer people had a higher number of richer friends.
For example, for people below the median SES in Minneapolis, Minnesota, about 49 per cent of their friends are in the above-median SES group, while in Indianapolis, Indiana, the figure is just 32 per cent. And at age 35, the average household income is $34,000 in Minneapolis, compared with $24,700 in Indianapolis.
The researchers then mapped people’s friendships back to where they were formed. They did this by categorising the formation of about 30 per cent of the friendships to a specific location such as a high school, religious group or shared neighbourhood. They then modelled what would happen to adults in below-median-SES groups if they were exposed to the same number of above-median-SES individuals as the average person in that group.
The team found that about half of the economic disparity between the two groups could be explained by a lack of exposure. This could be due to various factors such as living in different neighbourhoods or going to different schools, says Chetty. “You can’t become friends with someone you never meet,” he says.
“This is a very impressive study which has implications for our understanding of education and social mobility,” says Hugh Lauder at the University of Bath in the UK. “The first major point is that in schools which are well-mixed in terms of students’ family incomes, students from low-income families are more likely to make friends with those from higher-income families.”
“The second is that the more segregated students are by geographic location the less likely they will have the opportunity to make friends with students from higher-income families,” he says. “These findings should renew policy-makers’ efforts to provide well-mixed schools in terms of student composition.”
Journal reference: Nature, DOI: 10.1038/s41586-022-04996-4
More on these topics: