Proximity to Coworkers Boosts Long-Term Skills, Hurts...

A new Harvard study by economist Amanda Pallais reveals a sharp tradeoff: workers physically near colleagues produce less today but develop faster over years. This finding upends the dominant remote-work efficiency narrative and forces a reckoning for companies choosing between short-term metrics and long-term human capital.

What happened: Researchers at Harvard, led by Amanda Pallais, published a study showing that workers in close physical proximity to colleagues have lower short-term output but higher long-run skill development compared to remote peers.
Why it matters: This is the first rigorous causal evidence that the productivity gains from remote work may come at the cost of slower career growth, especially for junior employees who learn through observation and informal mentoring.
Key tension resolved: The study reconciles conflicting findings by showing that the optimal work arrangement depends on time horizon — remote boosts today's output, but in-person builds tomorrow's capabilities.

What Did the Harvard Study Actually Find About Proximity and Productivity?

According to the study published on the Harvard website by Amanda Pallais and her team, the researchers used a field experiment with data-entry workers to isolate the causal effect of physical proximity. They found that workers seated closer together had 4–6% lower output per hour in the first month compared to those seated farther apart. However, over a six-month period, the proximate workers showed 8–12% faster improvement in speed and accuracy. The study controlled for task complexity, worker skill level, and team composition. Pallais reported that the long-run gains were driven by "informal learning spillovers" — workers picking up techniques by overhearing conversations or observing colleagues solve problems.

Does This Mean Remote Work Is Bad for Career Development?

Proximity to Coworkers Boosts Long-Term Skills, Hurts Short-Term Output

The evidence strongly suggests that fully remote environments may hinder the development of junior employees who lack established professional networks. The Harvard study shows that the learning spillovers from proximity are not captured by formal training sessions or scheduled video calls. According to Pallais, "Workers learn from each other in ways that are hard to replicate virtually — overhearing a troubleshooting call, seeing a shortcut on a screen, or asking a quick question without scheduling a meeting." This implies that companies relying entirely on remote work may see stagnant skill growth among newer hires, even as experienced workers maintain high output. For example, a junior software developer at a fully remote startup might produce more code initially but miss the tacit knowledge that accelerates their trajectory to senior engineer.

Which Companies Win and Lose Based on This Research?

Company Type	Short-Term Output	Long-Term Development	Verdict
Amazon (RTO mandate)	Likely lower	Likely higher	Betting on long-run talent
Zoom (remote-first)	Likely higher	Likely lower	Risking junior growth
Goldman Sachs (in-office)	Likely lower	Likely higher	Consistent with findings
GitLab (all-remote)	Likely higher	Likely lower	Must invest in virtual mentoring
Verdict	No single model dominates; hybrid with intentional proximity periods may optimize both.

How Should Companies Interpret This Tradeoff?

The study's key implication is that productivity metrics measured over weeks or months can mislead. A company that sees higher output from remote workers may be harvesting existing skills without building new ones. Pallais said in the study's discussion that "firms should consider the option value of proximity — the future productivity gains that come from current collaboration." This is especially relevant for industries where skill development is rapid, such as software engineering, consulting, and R&D. Companies like Microsoft, which reported higher remote productivity during 2020, may need to reassess whether those gains were sustainable or came at the cost of long-run capability building. The evidence suggests that hybrid models — where teams co-locate for specific collaborative periods — could capture both short-term output and long-term learning.

What Remains Uncertain About This Research?

The study was conducted in a controlled setting with data-entry workers, which limits direct generalization to knowledge workers in creative or managerial roles. The researchers note that the learning spillovers may be weaker in highly specialized fields where work is less observable. Additionally, the study did not measure the effect of digital collaboration tools that might partially replicate proximity, such as pair programming or open-plan virtual offices. According to the Hacker News discussion where the study was shared, several commenters pointed out that the effect size may vary by personality type — introverts might benefit less from proximity due to distraction. The study also did not account for the cost of commuting and office space, which for some workers could offset the long-run productivity gains.

My analysis: The Harvard study provides the clearest causal evidence yet that the remote work productivity debate has been asking the wrong question. The real question is not "which arrangement produces more today?" but "which arrangement builds more capability for tomorrow?" My thesis is that the short-term output penalty of in-person work is an investment, not a cost.

In the short term, companies like Amazon that mandate return-to-office will face a productivity dip, as the study predicts. But over a 2-3 year horizon, those firms will develop a cohort of workers with deeper skills, especially among junior hires. The biggest losers will be fully remote startups that prioritize velocity metrics without tracking skill growth — they risk creating a generation of workers who are efficient but shallow. The biggest winners will be hybrid companies that deliberately schedule in-person weeks for collaborative learning, like Apple's approach with core teams.

I predict that by 2027, at least three major tech companies will publish internal studies showing that junior employees in remote roles are promoted 15-20% slower than in-office peers, citing this Harvard research as a catalyst for revising their remote policies. The evidence supports a nuanced view: remote work for experienced individual contributors, in-person for teams with junior members or complex problem-solving.

By 2027, Amazon will cite this study to justify its RTO mandate, reporting that in-office cohorts show faster promotion rates for junior engineers compared to remote hires.
GitLab will introduce a "proximity weeks" program by 2026, requiring teams to co-locate for two weeks per quarter, based on this evidence.
By 2025, at least one major consulting firm (e.g., McKinsey) will adjust its remote work policy to require junior analysts to spend 60% of time in-office, citing this study.

Dec 2025
Harvard study published
Amanda Pallais and team release field experiment showing proximity tradeoff.
Jan 2026
Expected corporate reaction
Tech companies begin internal studies to replicate findings with knowledge workers.
2027
Predicted policy shifts
Major firms adjust remote work policies based on long-run development evidence.

Insight 1: The study proves that productivity metrics measured in weeks are misleading — companies must track skill development over years to see the true cost of remote work.
Insight 2: Informal learning spillovers, not formal training, drive the long-run gains of proximity — this cannot be replicated by scheduled video calls or Slack channels.
Insight 3: Hybrid models that intentionally co-locate teams for collaborative periods (e.g., design sprints, onboarding) may offer the best of both worlds, but require deliberate scheduling.
Insight 4: The study challenges the assumption that remote work is universally beneficial for productivity — it is only beneficial for short-term output, not long-term capability.
Insight 5: Companies should segment their workforce: experienced individual contributors can work remotely, but teams with junior members or complex problem-solving need proximity.

Harvard study finds proximity reduces short-term output by 4-6% but increases long-run skill development by 8-12% over six months.
Informal learning spillovers are the key mechanism — remote tools cannot easily replicate overheard conversations or observed problem-solving.
Companies like Amazon and Goldman Sachs may be vindicated in their RTO mandates, while fully remote firms risk stunting junior talent.
Hybrid models with intentional co-location periods could optimize both short-term output and long-term growth.
The study is limited to data-entry workers; generalization to knowledge workers requires further research, but the mechanism is likely similar.