Which social connection best predicts upward mobility?

Social capital hides several different forces

A tightly connected community may have strong mutual support but few bridges to different economic worlds. Another place may contain many organizations yet little friendship across class lines. Calling both high social capital hides the distinction.

A 2022 Nature paper measured three families separately: cross-group connectedness, network cohesiveness and civic engagement. It then asked which one was most closely associated with upward income mobility across US places.

The map covered 72.2 million people

The researchers analyzed privacy-protected data from 72.2 million active US Facebook users aged 25–44 who had at least 100 US-based Facebook friends and a usable ZIP code. The network contained about 21 billion friendship links. The sample represented 84% of the US population in that age range by the paper’s benchmark, but it was not everyone.

The team predicted each user’s socioeconomic-status rank from 22 variables; it did not observe individual income directly. Facebook links were treated as a proxy for real-world friendship, not as a measure of time spent online.

Public releases contain aggregate, differentially private statistics by place. The raw individual network did not leave Meta.

One form of connection stood apart

The paper defines economic connectedness as the share of above-median-SES friends among below-median-SES people, divided by 50%. A value of one would mean that high-SES friends were represented in line with an even national split; lower values indicate under-representation.

Across counties, economic connectedness had a population-weighted correlation of 0.65, with a standard error of 0.04, with upward income mobility for children whose parents were at the 25th income percentile. An EC increase of 0.5 units was associated with an 8.2-percentile increase in those children’s adult income ranks across places.

Other measures—network clustering, supported friendships, volunteering and civic organizations—were much weaker predictors of this particular outcome. Economic connectedness remained prominent after including major observable place-level predictors.

Prediction at place level is not individual causation

The headline relationship is observational and ecological. Counties with more cross-class friendships can also differ in schools, segregation, labor markets, family structure and public institutions. The paper’s controls make the association harder to explain away with measured factors; they do not randomly assign friendship.

The study therefore cannot say that adding one higher-SES contact will raise one person’s income. Its counterfactual models are not intervention results, and an area average cannot forecast an individual’s future.

Economic connectedness is also not the universally best form of social capital. The authors explicitly note that cohesiveness may predict some health outcomes more strongly. The outcome decides what kind of network feature looks most relevant.

The map inherits the platform’s boundaries

Several authors were Meta employees or contractors. Two researchers had received an unrestricted Facebook gift to NYU Stern, and Opportunity Insights receives core funding from the Chan Zuckerberg Initiative, which the paper says did not fund this research. The authors state that Meta did not dispute or influence the findings. The research was also facilitated by a consulting agreement between four academic authors and Meta.

Those disclosures do not invalidate the work. They matter because independent researchers cannot access the raw network and must reproduce only the analyses supported by public aggregates and code.

Opportunity is shaped by places, not contacts alone

The paper shifts the question from who has a powerful network? to in which settings do bridges form across economic lines? That is a more useful and less judgmental frame. It recognizes that opportunity can be relational and institutional rather than an innate trait.

But even a diverse setting may remain socially divided. The next question is whether simply exposing people to different economic groups is enough for friendships to form.