国产第一福利影院草草

Every college campus has a hidden curriculum that never appears in any syllabus. It is the curriculum of who you eat lunch with, who you study with, who you end up talking to at a party you almost did not attend. It is the web of friendships, casual and close, accidental and deliberate, that students form over four years of shared space. And according to a groundbreaking dataset built from the Facebook friendships of millions of college students, this hidden curriculum may matter more for economic mobility than almost anything colleges officially teach.

The Social Capital Atlas, constructed by Raj Chetty and his collaborators using anonymized Facebook data from 21 million Americans, measures something that sociologists have theorized about for decades but never been able to observe at scale: the degree to which people from different economic backgrounds actually become friends. At the college level, this metric, called economic connectedness, captures whether low-income students form genuine friendships with high-income peers or whether they remain isolated within their own socioeconomic stratum even when sharing a campus.

We merged the Social Capital Atlas data for 1,224 four-year colleges with the Mobility Report Cards analyzed in our previous installment. The central finding is striking: economic connectedness at a college is one of the strongest predictors of whether that school鈥檚 low-income students will reach the top of the income distribution. The correlation (r = 0.55) is stronger than the relationship between success and SAT scores, instructional spending, or graduation rates. Cross-class friendships are not just a nice byproduct of college life. They appear to be one of its most powerful engines.

Measuring the Invisible: What Economic Connectedness Captures

Economic connectedness, as defined by Chetty and his team, is a precise metric derived from observed Facebook friendships. For each college, it calculates twice the share of high-socioeconomic-status friends among low-SES individuals. A score of 2.0 would mean low-income students have exactly the same proportion of high-income friends as the campus population would predict: perfect integration. A score of 1.0 means low-income students have half as many high-income friends as random mixing would produce. The national average for four-year colleges is about 1.56.

The researchers decompose economic connectedness into two components that reveal why cross-class friendships form or fail to form. The first is exposure: the degree to which low-SES and high-SES students participate in the same groups, classes, and social settings. The second is friending bias: the tendency of students to befriend peers of their own economic background even when exposed to others. A school can have low economic connectedness because it simply does not bring different classes together (low exposure) or because students self-segregate even when they share the same spaces (high bias).

Figure 1: Left panel: Median economic connectedness and exposure to high-SES peers by institutional tier. Right panel: Friending bias (negative = more cross-class friendships than expected). Source: Social Capital Atlas.

The tier-level data reveals a clear pattern. Ivy Plus schools and other elite institutions score highest on economic connectedness (approximately 1.83), followed by highly selective privates (1.79) and highly selective publics (1.75). Selective publics, the large state universities that serve most American students, score notably lower at 1.51. Nonselective public four-year colleges bring up the rear at 0.99, meaning their low-income students have essentially half the cross-class friendships that random mixing would predict.

But the decomposition tells the more important story. Friending bias is near zero everywhere. At every tier, from the Ivy Plus (bias: +0.004) to selective publics (鈭�0.005) to nonselective publics (+0.013), the bias term is negligible. Students do not appear to resist cross-class friendships when given the opportunity. The entire gap in economic connectedness across tiers is driven by exposure: by whether the campus puts students from different backgrounds in the same rooms in the first place.

Friending bias is near zero at every type of college. Students don鈥檛 resist cross-class friendships. The question is whether colleges create the conditions for them to happen.

The Friendship鈥揗obility Connection

The reason economic connectedness matters is not sentimental. It is economic. When we plot economic connectedness against the success rate, the percentage of a school鈥檚 bottom-quintile students who reach the top quintile, the relationship is immediate and powerful.

Figure 2: Each point is a four-year college. X-axis: economic connectedness (cross-class friendship index). Y-axis: success rate (% of bottom-quintile students reaching top quintile). r = 0.55 (p < 0.001). Source: Social Capital Atlas merged with Mobility Report Cards.

The correlation is 0.55, meaning economic connectedness explains roughly 30 percent of the variation in success rates across four-year colleges. Schools where low-income students form more friendships with high-income peers produce dramatically better outcomes for those students. At the extremes, the difference is cavernous: colleges in the top decile of economic connectedness have a median success rate above 40 percent, while those in the bottom decile average below 12 percent.

The relationship holds within tiers, not just across them. Among selective public universities, the 386 schools that educate the largest share of American college students, schools with higher economic connectedness still show higher success rates than their lower-EC peers. This is not simply a proxy for selectivity or wealth. It appears to capture something about the social environment itself.

Several mechanisms are plausible and likely complementary. High-income friends provide access to professional networks, internship referrals, and job leads that low-income students would not encounter through their own families. They normalize career ambitions and salary expectations. They share tacit knowledge about how industries work, how to negotiate, and which opportunities to pursue. And the research literature on social capital suggests that even weak ties, acquaintances rather than close friends, can be decisive in career advancement, because they bridge otherwise disconnected social worlds.

The Hidden Curriculum: Cliques, Civic Life, and Campus Culture

Economic connectedness captures whether friendships form across class lines. But the Social Capital Atlas also measures two other dimensions of campus social life that prove to be informative: clustering and civic engagement.

Clustering measures how cliquish a campus is: specifically, the fraction of a student鈥檚 friend pairs who are also friends with each other. High clustering means tight-knit, insular social circles. Low clustering means more open, sprawling networks where a student鈥檚 friends come from different groups and do not necessarily know each other.

Volunteering rate captures the share of students who belong to volunteering or activism groups on Facebook, a proxy for civic engagement and the density of community organizations on campus.

Figure 3: Left: Success rate and median earnings by campus clustering quartile (low = open networks, high = cliquish). Right: Same metrics by volunteering rate quartile. Source: Social Capital Atlas merged with Mobility Report Cards.

The clustering data tells a cautionary story. Campuses in the most cliquish quartile have a median success rate of 19.4 percent, compared to 24.4 percent at the least cliquish campuses. Median earnings follow the same pattern: $38,750 at the most cliquish schools versus $42,100 at the most open. Tight social circles may feel comfortable, but they constrain the bridge-building that drives economic mobility. When everyone in your friend group already knows everyone else, the network provides emotional support but fewer novel connections to opportunity.

The volunteering data adds a complementary finding. Schools in the highest civic-engagement quartile have a success rate of 28.2 percent and median earnings of $44,800, substantially above the 20.1 percent success rate and $38,400 earnings at schools in the lowest quartile. Civic engagement appears to be both a marker of institutional culture and a mechanism for building the kinds of weak-tie networks that research has linked to career mobility.

Notably, clustering and volunteering move in opposite directions on mobility. Clustering is negatively correlated with mobility (r = 鈭�0.21), while volunteering is positively correlated with success rates (r = +0.16). The ideal campus environment for social mobility, these data suggest, combines open social networks (low clustering) with high civic participation, a combination where students encounter diverse peers through shared activities rather than retreating into homogeneous cliques.

The most cliquish campuses produce a success rate of 19%. The most open produce 24%. Tight social circles feel comfortable, but open networks build opportunity.

The Exposure Problem: Why It鈥檚 About Who鈥檚 on Campus

Perhaps the most important policy implication of the social capital data is the finding that friending bias is essentially zero. Students at every type of institution are willing to form cross-class friendships at roughly the rate that random mixing would predict. The barrier is not attitudes. It is architecture.

At nonselective four-year publics, the exposure score is just 1.05, meaning low-SES students encounter high-SES peers at about half the rate a balanced campus would produce. At the Ivy Plus, exposure is 1.85. The gap is not surprising: elite schools draw from the top of the income distribution by design, and their low-income students find themselves surrounded by wealthy peers simply because that is who attends. The question is whether the exposure advantage of elite schools can be replicated more broadly.

This is where our earlier social mobility rate analysis connects directly to the social capital story. The CUNY campuses that dominate the mobility ranking achieve moderate economic connectedness scores, around 1.2 to 1.5, but they compensate with sheer access. Their low-income students may have somewhat fewer cross-class friendships than their Ivy League counterparts, but there are seven to eight times more of those students, and the ones who do form upward-bridging friendships benefit just as much.

Highly selective publics like SUNY Stony Brook, UC Irvine, and Binghamton sit in an intriguing middle ground. They have economic connectedness scores around 1.75, nearly as high as the Ivy Plus, combined with bottom-quintile enrollment three to five times higher. These schools achieve genuine cross-class mixing at meaningful scale, and their mobility rates reflect it. They may represent the most replicable model for social-capital-driven mobility in American higher education.

Caveats and Questions

The social capital data is unprecedented in its scope but carries important limitations. The Facebook friendship measure captures digital connections among users aged 25 to 44 as of 2022, which reflects friendships formed during college years (roughly 2000鈥�2014) but may also include connections made afterward. The researchers use birth-cohort restrictions to minimize this issue, but some post-college friendships inevitably enter the data.

More fundamentally, the correlation between economic connectedness and success rates does not prove causation. Schools with higher EC scores also tend to be more selective, better resourced, and located in stronger labor markets. It is possible that economic connectedness is a marker of institutional quality rather than a cause of mobility. The researchers address this concern with within-tier analyses and controls, and the relationship survives, but a definitive causal claim would require experimental variation that does not yet exist.

Additionally, the economic connectedness metric measures friendships between SES groups defined by individual characteristics (for own-SES measures) or parental characteristics (for parent-SES measures). It does not directly measure the quality, depth, or professional utility of those friendships. A Facebook connection between a low-income engineering student and a wealthy finance major may transmit very different social capital than a connection between two students in the same seminar.

The Bottom Line: The Social Architecture of Opportunity

The social capital data challenges a deeply ingrained assumption about how colleges create economic mobility. We tend to focus on what colleges teach: curricula, degrees, credentials, skills. The Social Capital Atlas suggests that whom colleges bring together may matter just as much. When low-income students form friendships with high-income peers, they gain access to networks, norms, and information that no course can replicate.

The finding that friending bias is near zero is, in many ways, the most hopeful data point in this entire analysis. It means the barrier is structural, not interpersonal. Students are willing to befriend across class lines. Colleges just need to put them in the same room.

This has concrete implications for admissions, housing, extracurricular programming, and financial aid. Mixed-income residential housing, first-generation bridge programs, need-blind admissions, and inclusive student organizations are not just equity initiatives. They are, if this data is to be believed, investments in the single most powerful predictor of whether low-income students will climb the economic ladder. The hidden curriculum of college, the friendships students form without even realizing it, may turn out to be the most important curriculum of all.

Data Notes & Methodology

Data Sources: (1) Chetty, R., Jackson, M., Kuchler, T., et al. (2022). 鈥淪ocial Capital I & II.鈥� Nature 608. Dataset: Social Capital by College (Social Capital Atlas, Opportunity Insights). (2) Chetty, R., Friedman, J., et al. (2017). Mobility Report Cards (Opportunity Insights). (3) College-Level Characteristics (Opportunity Insights).

Sample: 1,224 four-year institutions with both social capital measures and mobility data. Social capital data derived from Facebook friendships among users aged 25鈥�44, 1986鈥�1996 birth cohorts (own-SES) or 1990鈥�2000 cohorts (parental SES). Mobility data based on 1980鈥�1984 birth cohorts.

Key Variables: ec_own_ses_college (economic connectedness: cross-class friendship index), exposure_own_ses_college (mixing rate), bias_own_ses_college (friending bias), clustering_college (network cliquishness), volunteering_rate_college (civic engagement), kq5_cond_parq1 (success rate), mr_kq5_pq1 (mobility rate).

Methodology Notes: Economic connectedness = 2脳 share of high-SES friends among low-SES individuals. EC 鈮� exposure 脳 (1 鈭� bias). Clustering = average fraction of friend pairs who are mutual friends. Volunteering = % of users in predicted volunteering/activism groups. Noise added per differential privacy standards (Chetty et al. 2022). Minimum cell sizes: 100 low-SES and 100 high-SES users per college.

Limitations: Facebook friendships 鈮� all friendships. Digital connections may include post-college ties. Correlation between EC and success does not establish causation. SES is estimated from Facebook signals, not tax data. Mobility cohorts (1980鈥�84) predate social capital measurement period. Privacy noise may affect estimates for smaller institutions.