Skip to main content

What 21 billion Facebook friendships say about the economic ladder in the US

What 21 billion Facebook friendships say about the economic ladder in the US

/

Meta made the dataset publicly available

Share this story

A dataset shows how many people have friends outside of their income bracket.
A dataset shows how many people have friends outside of their income bracket.
Image: Social Capital Atlas

Meta publicly released information on 21 billion Facebook friendships as part of a research project looking at economic inequality in the United States, the company announced today. Along with new insights into the intersection of money and friendships in America, the partnership between Meta and the researchers gives us another look at who Facebook is willing to share data with — and why.

The research team wanted to understand why people in some places in the US were more likely to move between economic brackets than in others. Using the information from Meta, along with other data, a research team built a dataset for a pair of studies on economic mobility, published Monday in the journal Nature. One study found that people who grow up in areas where there are more friendships between high- and low-income people are more likely to move out of poverty and up the economic ladder.

“Growing up in a community connected across class lines improves kids’ outcomes and gives them a better shot at rising out of poverty,” Raj Chetty, a Harvard economist and lead researcher on the study, told The New York Times.

Many places, though, don’t allow for much interaction between high- and low-income people, the second of the two studies found. And even when a neighborhood does allow for that kind of interaction, people are still more likely to befriend people in similar economic brackets.

Chetty and his collaborators first got access to Facebook’s data in 2018 as part of an effort to understand economic inequality and the vast income disparities in the US. Meta researchers partnered with Chetty on the project, and other members of the research team are affiliated with groups that have contracts with Meta.

The researchers pulled data on Facebook users in the United States between 25 and 44 years old who used the platform in the past 30 days, had at least 100 friends on the platform, and had connected their account to their zip code — a sample that accounted for 72 million people, which is over 80 percent of the US population in that age bracket. The team measured the socioeconomic status of users based on things like location, education levels, relationship status, and language. The analysis then compared individuals’ socioeconomic status with the status of their Facebook friends.

Now, the full dataset, which covers 21 billion Facebook friendships, is available through Facebook’s Data for Good program. People can search the public-facing website and see the economic connectedness of various communities, including their own. Researchers can download the data for additional studies.

Meta and Facebook tout the Data for Good program as a signal of the company’s willingness to partner with outside research teams and share the troves of data from its platform. And that data is often incredibly useful for researchers; it helped track how people’s movement changed over the pandemic, for example.

But that openness has primarily been focused on research into how people live their lives outside of Facebook, not how Facebook or other Meta platforms, like Instagram, influence people’s lives or behavior. The company has been much less willing to partner with external researchers or share raw data on things like the mental health of people who use Meta’s products.

The new studies offer valuable insight into economic mobility in the US, and the data could help researchers understand how people in the US build relationships. Tech giants like Meta are some of the only ways researchers can find the data they might need to run these sorts of analyses. Reliance on Meta and Facebook for that information, though, means the companies have more control over the research questions that get asked in the first place.