Remote work and productivity: five years of evidence after the great experiment

Between March 2020 and roughly June 2021, somewhere between 30% and 50% of the workforce in advanced economies abruptly began working from home. Most managers had assumed productivity would collapse. Most workers assumed they'd accomplish less without the office. Both groups were partially wrong.

Five years and a substantial research literature later, the picture is clearer than the discourse suggests. The popular framing as a "remote work works / doesn't work" binary obscures what the data actually shows: remote work produces specific gains and specific losses, distributed unevenly across job types, worker demographics, and team configurations.

1. The productivity findings

Nicholas Bloom's lab at Stanford has run the most-cited series of studies on remote-work productivity, beginning with a 2015 randomized trial at Ctrip (a Chinese travel company) and continuing through COVID-era observational work.

The Ctrip experiment: 250 employees were randomly assigned to work from home for nine months. The remote group's productivity rose by 13% relative to the in-office group. The gain came almost equally from fewer breaks and a calmer environment, not from working longer hours (Bloom et al., 2015).

COVID-era data, though less cleanly experimental, is consistent. Several large-firm studies found productivity increased by 5-10% during full remote, then decreased by 5-10% during hybrid arrangements relative to either extreme — suggesting hybrid is the most expensive mode (Barrero et al., 2021).

2. The differentiation

The productivity findings are not uniform. They depend on:

Task type. Routine knowledge work (writing, coding, individual analysis) improves remote. Creative collaboration and brainstorming, on aggregate, do not.

Career stage. Junior employees who have not yet built tacit firm knowledge appear to lose more from remote than mid-career employees. The deficit shows in promotion rates 18-24 months later (Emanuel & Harrington, 2023).

Manager quality. Teams with managers skilled in remote coordination (clear written communication, structured async check-ins) outperform their in-office equivalents. Teams with managers who depend on visual supervision underperform.

3. The mental health story

Self-reported wellbeing data from remote workers is mixed and demographically uneven. Workers with long commutes report large quality-of-life gains. Workers with small homes and young children report stress. Solo workers report increased loneliness; partnered workers with separated workspaces report increased intimacy.

The most consistent finding across large surveys: workers with the option to work remotely some of the time report higher job satisfaction than workers in either pure-remote or pure-office arrangements (Choudhury et al., 2021).

4. What employers got wrong

Many employer concerns about remote work turned out to be empirically unsupported:

• Output decline: did not occur on average
• Surveillance need: tracking software showed no meaningful change in actual work hours
• Collaboration collapse: collaboration patterns changed shape but did not collapse

What did occur, and was less anticipated:

• Onboarding new hires became substantially harder
• Cross-team weak-tie communication declined; strong-tie within-team communication intensified
• Promotion rates for fully-remote workers lagged behind hybrid colleagues, controlling for performance

5. The honest summary five years later

The remote-work debate has rarely been a productivity debate. It is mostly a debate about what kind of organization people want to belong to. The productivity evidence supports remote work for most knowledge tasks. The career-development evidence is mixed and tilts toward in-person for early career. The wellbeing evidence supports flexibility, not any particular configuration.

The data don't settle the cultural question. They do reframe it: the productivity case for fully-in-office work is weaker than executives often claim, and the productivity case for fully-remote is stronger than skeptics admit. Most of the actual disagreement is downstream of values, not facts.