Aggregate US productivity did not clearly rise or fall from remote work spreading — but fully remote arrangements cut measured output per hour 4-19%, while hybrid arrangements raised it 5-13%

Summary

Two separate questions are bundled here, and the literature answers them differently. At the aggregate, economy-wide level, U.S. labor productivity growth did not settle into a clearly higher or lower trend because of the shift to remote work. It swung sharply -- surging in 2020 (+5.3% that year), giving most of it back in 2021-2022 (even falling -1.5% in 2022), then running a bit above its pre-pandemic pace in 2023-2025 (2019-2025 compound annual growth ~2.2%/year vs. ~1.3%/year in 2009-2019). Economists who tested whether this pattern was caused by remote work specifically -- by comparing productivity growth across industries that differ in how teleworkable their jobs are -- found essentially no relationship once pre-pandemic industry trends are controlled for (San Francisco Fed, Fernald/Goode/Li/Meisenbacher 2024). The aggregate swings are attributed mainly to the pandemic's cyclical shock and recovery, not to telework itself.

At the level of individual jobs and firms, however, remote work does move productivity, and by a lot -- but the direction depends on how much remote work, not on whether a job merely can be done remotely. Controlled studies of fully remote arrangements find lower output per hour: a Fortune 500 call center found a 4% productivity decline when previously on-site workers were forced fully remote (Emanuel & Harrington 2023, FRB New York); a large Indian IT-services firm found an 8-19% productivity decline after an abrupt full-remote shift (Gibbs, Mengel & Siemroth, cited in Barrero, Bloom & Davis 2023). Controlled studies of hybrid arrangements (1-4 remote days/week) instead find productivity gains of roughly 5-13%: +13% at a Chinese travel agency (Bloom, Liang, Roberts & Ying 2015, QJE), +5% (rising to +13% cumulative with work-from-anywhere) among U.S. patent examiners (Choudhury et al. 2021), and +10% in an Italian firm's flexible-schedule trial (Angelici & Profeta 2023). Roughly 37% of U.S. jobs, accounting for 46% of wages, are classified as feasible to do entirely from home (Dingel & Neiman 2020) -- but feasibility alone says nothing about whether remote work in practice makes those jobs more or less productive; the intensity and design of the arrangement does.

How much remote work actually spread

Two measures, at different strictness: the Census ACS asks a worker's primary means of travel to work and counts only those who report working entirely from home. By that stricter measure, the U.S. share was flat around 4-5% for a decade (2009-2019), then rose to 13.4% by the 2019-2023 5-year estimate (askamerica census.acs_commuting, national aggregate). The Barrero-Bloom-Davis Survey of Working Arrangements and Attitudes (SWAA), which counts any paid workday spent at home including hybrid days, shows a much larger swing: from about 5% of paid workdays pre-pandemic, to a peak above 60% in 2020, settling to about 28-30% of paid workdays by mid-2023 -- roughly four times the 2019 rate. The two series diverge because ACS only captures workers who never commute (fully remote), while SWAA captures every remote workday including a single day a week.

Dingel and Neiman (2020) established the upper bound on what's structurally possible: 37% of U.S. jobs (46% of wages) could in principle be performed entirely at home, based on O*NET occupational task data -- concentrated in computing/math (100%), education (98%), legal (97%), finance (88%) and management (87%), and near zero in food service, cleaning, construction, farming and production. Actual work-from-home rates before the pandemic were far below this technical ceiling (well under a quarter of full-time workers did any home-based work on an average day per the 2018 American Time Use Survey), which is why the pandemic could push realized remote work up so fast without needing any change in job content.

The aggregate productivity trend, and why it is not a clean remote-work story

Source: BLS Major Sector Productivity and Costs, nonfarm business series (askamerica econ.labor_productivity, series PRS85006093), quarterly index averaged to annual.

The 2020 surge fueled early media speculation that forced remote work and digital adoption were boosting productivity. That reading did not hold up. Fernald and Li (2022, Federal Reserve Bank of Kansas City / San Francisco Fed) showed the 2020 surge and subsequent 2021-2022 giveback were largely a predictable cyclical artifact of the pandemic recession and recovery -- low-productivity jobs and hours were disproportionately cut in the 2020 downturn, mechanically raising measured average output per hour, then reversing as those jobs came back. The SF Fed's follow-up study (Fernald, Goode, Li & Meisenbacher, 2024) tested the remote-work hypothesis directly: across 43 industries, they regressed pandemic-era productivity growth on each industry's teleworkability (using Dingel-Neiman occupation-level scores weighted by industry employment mix). Before controlling for pre-pandemic trends, more-teleworkable industries look modestly faster-growing; but industries that were teleworkable also tended to already be growing faster before the pandemic. Once each industry's pre-pandemic trend is netted out, the relationship between teleworkability and excess pandemic productivity growth is nearly flat and not statistically significant. Their conclusion: the shift to remote and hybrid work "most likely has neither substantially held back nor boosted [aggregate] productivity growth."

The job-level effect: hybrid gains, fully-remote losses, and why they differ

The controlled micro-evidence is much less ambiguous than the aggregate data, and splits cleanly by degree of remoteness:

• Hybrid arrangements (1-4 remote days/week) consistently show productivity gains. Bloom, Liang, Roberts & Ying's randomized cTrip.com call-center trial (2015, QJE) found average daily output rose 13% for workers switched to 4-days-home/1-day-office -- 9 percentage points from extra worktime (fewer breaks and less sick leave) and 4 points from genuinely more calls handled per hour. Choudhury et al.'s study of U.S. patent examiners (staggered, quasi-experimental adoption starting 2006) found patent actions rose 5% after a shift to a hybrid schedule (up to 4 home days/week), with no increase in rework/quality problems, and a further 8% gain when examiners later moved to fully-flexible work-from-anywhere. Angelici & Profeta's nine-month field experiment at an Italian firm found a 10% productivity gain (by self- and supervisor-assessment) for employees given schedule/location flexibility one day a week, with no productivity spillover, positive or negative, onto co-workers.
• Fully remote arrangements consistently show productivity losses relative to on-site or hybrid work. Emanuel & Harrington (2023, FRB New York Staff Report 1061) studied a Fortune 500 firm's call centers that employed both remote and on-site workers in the same jobs pre-pandemic: remote workers already answered 12% fewer calls/hour than on-site workers before COVID (worker-selection effect), and when the pandemic forced previously on-site workers fully remote, their productivity fell an additional 4% relative to already-remote colleagues (the causal treatment effect of forced full-remote work) -- with call quality and promotion rates also declining, especially for less experienced staff. Gibbs, Mengel & Siemroth's study of a large Indian IT-services firm forced to full-remote in March 2020 (cited and detailed in Barrero, Bloom & Davis 2023) found labor productivity (output per hour) fell 8-19%, driven mainly by a rise in time spent on meetings/coordination crowding out focused work time, alongside a narrowing of employees' internal and external collaboration networks.

The pattern across all five studies is the same: some remote time (a day or two a week) tends to raise measured output per hour, mainly via fewer sick days/breaks, quieter home environments, and modestly higher effort from happier employees; all remote time tends to lower it, mainly via higher coordination/meeting overhead, narrower professional networks, and reduced on-the-job mentoring, especially for junior staff. Barrero, Bloom & Davis (2023, NBER Working Paper 31686) also flag a conceptual complication: workers count commute-time savings as a productivity gain (roughly 68 minutes/week, or about 2.8% of a 40-hour week, saved as of mid-2023) that standard employer/statistical-agency output-per-hour measures do not credit at all -- which is part of why workers report more favorable views of remote-work productivity than managers do (43% of ACS-eligible remote workers say they're more productive at home vs. 14% who say less, per SWAA data).

Bottom line

How much did aggregate labor productivity change as remote work spread? Not in a way attributable to remote work itself -- the visible 2020-2025 swings (a 2020 surge, a 2022 dip, a modest 2023-2025 acceleration to roughly +2.2%/year vs. +1.3%/year pre-pandemic) track the pandemic business cycle, not teleworkability, once industry-level pre-pandemic trends are controlled for. How much does it differ between remote-capable and non-remote work? Feasibility alone (the ~37% of jobs Dingel-Neiman classify as remote-capable) is the wrong axis for productivity differences -- the studies that actually measure output find hybrid arrangements raising productivity roughly 5-13% and fully remote arrangements lowering it roughly 4-19%, both relative to on-site work in the same jobs. The controlling factor is the intensity of remote work, not whether the job can technically be done remotely at all.