Economists have spent years trying to explain why an Ivy League degree is worth roughly $101,000 more per year a decade after graduation. The answer they eventually found, after stripping out instruction quality, alumni networks in the conventional sense, and prestige signalling, was peer environment. Who you are in the room with.
Consider what that finding makes of one particular story shared recently in The Atlantic. A Yale graduate secured his first job in financial services not through the application process, but by beating a senior executive at beer pong at a party. You could read it as a colourful exception. Raj Chetty and David Deming at the Harvard Kennedy School, and John Friedman of Brown University, would read it differently.
Their research, conducted as part of Harvard’s Opportunity Insights project, points to peer environment as the dominant mechanism of elite university outcomes. The transformation comes from being in sustained physical proximity to people who are highly ambitious, highly connected, and for whom a particular way of moving through the world is simply normal. “Steel sharpens steel” is the shorthand. The degree is the credential, and the peer network is the learning curve.
When the people around you are preparing for first-round interviews at McKinsey or Goldman Sachs, those destinations shift from aspiration to working assumption. You absorb the vocabulary of elite professional life before you have any experience of it – what a case study requires, what a credit analyst actually does, how a fund is structured. You learn it not from a course but from the person two doors down who interned at Bridgewater last summer, or from a study group in which half the room already has offers from bulge-bracket banks and discusses them the way others discuss term papers.
Referrals, warm introductions, knowledge of which partners to approach and which firms are discretely expanding their analyst intakes is information that circulate freely among people who already have it, and rarely reaches those who do not.
The beer pong game was not a lucky break. That is how it works.
Why the peer effect is also an access problem
Admission to the most selective universities is a filter for transformation, and who gets the nod from the gatekeepers matters more than what happens inside.
Economist Russell Weinstein at the University of Illinois documented what happened to campus recruiting during the 2008 financial crisis. Investment banks did not cut their campus presence evenly. They concentrated it at a small number of elite institutions and withdrew from almost everywhere else. The logic was when hiring budgets shrink, recruit where the networks already exist.
A first-generation student with a first-class degree at a less selective university, applying to financial services in a downturn, wasn’t just facing a prestige penalty. He or she was facing a hiring logic that treats the peer environment theynever had access to as a proxy for competence. The exam they sat was not the one being marked.
Raj Chetty’s broader research on mobility shows that “economic connectedness,” the proportion of high-income people in a person’s social network, is among the strongest predictors of upward mobility in America. The universities producing the best long-term outcomes for low-income students are not necessarily the most academically rigorous. They are the ones where low-income students actually form cross-class friendships. The network is the key. Some universities design for this deliberately, but most do not.
How hybrid working removed the fallback for first-generation graduates
Set the US research alongside what is emerging from UK labour economists, and the importance of being in the room extends further.
London School of Economics researchers, Peter John Lambert and Yannick Schindler, now at the Ellison Institute of Technology Oxford examined why entry-level graduate hiring fell sharply between 2022 and 2025. Beyond generative AI replacing workers, their paper, The Broken Ladder: AI, Remote Work, and Early-Career Hiring describes working from home as a primary driver. Managers found it harder to develop junior staff when teams were working partly remotely. The tacit learning that happens in physical proximity degraded when offices emptied three days a week: the informal correction, the overheard negotiation, the lunch where institutional knowledge passes almost unnoticed.
This is the process Chetty, Deming and Friedman identified at elite universities. Informal, proximity-based learning from people further along is what produces the transformation. The LSE paper is describing the same one breaking down at the next stage of a young person’s working life.
For graduates who never had access to the elite peer environment at university, the workplace was always the fallback. The first job was where the informal education would happen, where the mentor would fill the role the peer network did not. Hybrid working has made that fallback considerably less available. Employers have responded not by redesigning how junior talent is developed in a distributed world, but by hiring less of it.
AI is removing the bottom rung of the career ladder
Hybrid working explains part of the decline in entry-level hiring, but artificial intelligence is making it unnecessary.
Generative AI now handles the tasks that made entry-level roles economically rational for employers: research, drafting, data analysis, document review, basic financial modelling. It does them faster and at a fraction of the cost. Entry-level job postings in the United States fell 35% between January 2023 and June 2025. SignalFire found a 50% decline in new role starts by people with less than one year of post-graduate work experience between 2019 and 2024. The National Association of Colleges and Employers reported a 22% drop in internship offers from Fortune 500 companies over the same period; tech internships fell 34%.
When AI automates enough tasks within an entry-level role, employers do not hire someone to perform the remainder. They redistribute the remaining work upward to senior staff, or invest in tools that allow a smaller team to cover the same volume. The junior position disappears not because the graduate was inadequate, but because the tasks that justified the hire have become cheaper to automate than to train.
This is particularly damaging for graduates without elite networks. The entry-level job was never only income. It was the environment where professional knowledge was absorbed about how institutions operate, who the decision-makers are, and how things actually get done. Friedman’s research found that proximity to experienced, connected people is a considerable advantage. That is equally true of the first years of a professional career as it is of an elite university. AI is not removing the first steps on the ladder where the climb was supposed to begin.
How graduate hiring really works in the nepotism economy
In the UK, 13% of 21 to 24-year-olds were not in education, employment, or training in 2025. A first-generation graduate applying for competitive graduate roles sends out roughly 20% more applications than a peer from a higher-income household and receives fewer responses. The gap is not narrowing.
A significant share of the stepping-stone opportunities that feed into graduate hiring, the insight days, spring weeks, and structured work experience that become lines on a CV and introductions to hiring managers, remains largely unadvertised. Access runs through prior connection. A parent who makes a call is not technically nepotism, but the outcome is the same.
This is where the beer pong game fits into the larger picture. Network-based selection has always been present in graduate hiring. What the combined data suggests is that it is growing as a proportion of how competitive roles are filled, at exactlythe moment when hybrid working and AI have reduced the alternatives available to someone who lacks the relevant network.
The peer environment gap can be closed
Chetty’s work on economic connectedness implies that the peer environments that produce upward mobility can be built, not just inherited.
Georgia State University is a well documented case. Ranked by the Brookings Institution and Harvard’s CLIMB initiative in the top 1% of universities nationally for social mobility, it has closed the achievement gap between low-income and higher-income students almost entirely. The school’s predictive analytics system identifies students at risk early enough to intervene, paired with proactive advising that meets students where they are rather than waiting for them to seek help. It now graduates more African American students than any other university in the country, and more low-income students than any other university in Georgia.
The Posse Foundation targets the peer environment more directly. It selects cohorts of ten high-potential students from diverse urban backgrounds and places them together at elite partner universities, including Vanderbilt and the University of Virginia, on full scholarships. Students arrive with a built-in peer group from similar backgrounds, which allows them to participate in the broader peer environment Friedman describes without being isolated within it. Ninety percent graduate. Since 1989, Posse has placed more than 12,000 students and awarded over $2 billion in scholarships.
Neither model is a complete answer. Both treat the problem as environmental rather than motivational.
Employers face a version of the same choice. If informal proximity-based learning is what makes junior hires valuable, and hybrid working and AI have disrupted the conditions for it, the response is not to reduce junior hiring. Some firms are rebuilding mentoring for distributed teams. Others are opening structured work experience to candidates outside their existing networks.
The research didn’t find that elite universities teach better. The Ivy League generates a particular kind of exposure that builds over a lifetime. Steel sharpens steel, and it is clear about who gets to hold the steel.
Interested in this topic? You might also like…