Is sharing the same space enough to bridge class lines?

Sharing a setting is only the first stage

Two people cannot become friends if they never share a setting. Yet sharing a school, neighborhood or organization does not guarantee that they enter the same classes, teams, rooms or conversations. Diversity at the institution level can coexist with separation inside it.

The companion to the economic-connectedness study separates those stages. It calls the share of higher-SES people present in someone’s groups exposure. It calls the shortfall in cross-SES friendship after accounting for that exposure friending bias.

The second term is a network statistic, not a moral diagnosis of an individual. It can reflect group size, academic tracking, architecture, norms and many other structures that shape repeated interaction.

The researchers traced friendships back to institutions

The primary sample included 70.3 million US Facebook users aged 25–44 whose friendships could be allocated to at least one setting where they formed. The analysis compared schools, colleges, neighborhoods and organizations, using the socioeconomic-status estimates developed in the companion paper.

The main setting analysis could assign roughly 30% of friendship links to a group. An expanded correction covered about 44% and produced similar conclusions, but most links still could not be assigned to a setting.

The researchers decomposed the difference between low- and high-SES people’s shares of high-SES friends into unequal exposure and conditional friending. They then used changes in the composition of high-school cohorts and school entry-date cutoffs to test whether extra exposure led to extra cross-class friendship in more causally credible local designs.

Exposure explained only half the divide

Nationally, differences in exposure and friending bias each accounted for about half of the disconnection between lower- and higher-SES people. That is an average decomposition under the paper’s definitions, not a fixed law for every institution.

It means composition alone leaves a large part of the question unanswered. Some places give people from different backgrounds many chances to meet but still produce relatively few friendships across the line. Others convert a similar level of exposure more effectively.

More exposure created more friendship in school cohorts

For students born from 1990 to 2000 who could be linked to parents and high schools, the researchers compared lower-SES students in different cohorts of the same school. A 10-percentage-point increase in higher-parental-SES peers led on average to an 8.9-percentage-point increase in the share of higher-parental-SES friends among lower-parental-SES students.

The result was not uniform. Additional exposure produced fewer cross-SES friendships in schools with higher measured friending bias. A separate regression-discontinuity design around school-entry-date cutoffs supported the same qualitative pattern.

This is the study’s most careful causal evidence, but it is local to cohort changes among lower-parental-SES students in US high schools; parental-SES groups compare the top and bottom quintiles. It is not an experiment in adult workplaces or communities.

Opportunity to meet is necessary—not sufficient

The paper discusses group size, academic tracking, physical space and new domains for repeated interaction as potential institutional levers. Those examples are proposals and case illustrations, not all tested interventions. They should guide hypotheses, not be advertised as proven recipes.

The broader cross-sectional findings also remain observational. More cross-class friendship is associated with higher mobility in the companion paper; this study does not prove that every additional connection creates an economic gain.

Several authors worked for or through Meta, four academic authors had a Meta consulting agreement and two had previously received an unrestricted Facebook gift to NYU Stern. Opportunity Insights receives core Chan Zuckerberg Initiative funding, which the paper says was not used for this research. The authors state that Meta did not influence the conclusions. Raw individual network data remain inaccessible outside the company; public aggregates include privacy noise.

Design what happens after the introduction

The answer to the title is not by itself. More exposure caused more cross-class friendship on average in the high-school cohort analysis, but the conversion was weaker where institutional friending barriers were higher.

The findings suggest that connection may require both meaningful exposure and conditions that lower barriers to interaction. That is a structural responsibility. It should not become another instruction for an individual to perform social mobility or another score attached to their contacts.