On Sept. 9, Microsoft canceled its plans for a back-to-the-office target date of Oct. 4 and said it would no longer forecast when it would fully reopen its U.S. work sites. Other companies, notably Apple and Amazon, have delayed their planned reopenings to early 2022 because of the surge in coronavirus infections and the uncertainty surrounding the delta variant. Meanwhile, Kastle Systems’ index of office usage in 10 major metropolitan areas shows that office-going rose steadily from January to July 2021 and then abruptly reversed course.

As long as the coronavirus rages, a sizable share of the American workforce is going to be staying on Zoom. What will that mean for economic productivity? Some commentators are hopeful: After all, there are signs that productivity — the inflation-corrected value of goods and services produced per hour of work — has actually risen in the working-from-home era. Between the second quarter of 2020 and the second quarter of 2021, labor productivity increased by 1.8 percent compared with an average annual increase of 1.4 percent from 2005 to 2019.

It’s hard to interpret that data, though. Some of the largest productivity gains have occurred in durable-goods manufacturing (the making of items like cars and appliances) — a highly capital-intensive sector largely unaffected by the working-from-home trend. Meanwhile, many of the service-sector businesses, such as bookstores and shoe stores, that had the biggest booms in productivity from 2019 to 2020 also saw significant drops in employment, and it is perhaps unsurprising that a skeleton staff of the most able workers is going to be more productive, per person. When job losses disproportionately occur in low-productivity sectors, such as service work — as has happened during the pandemic — that can also boost overall productivity figures. Nonetheless, some economists have calculated that a substantial shift toward a working-from-home economy could result in a permanent increase in productivity, perhaps because of reduced commuting and fewer of the distractions that can come with office life.

But we don’t think that’s likely. Research suggests that a switch to permanent remote work would make us all less productive.

People who shift to working from home can temporarily increase the amount of work they get done in a given day. But over the medium to long term, long-distance employment can’t deliver key benefits — including learning and new friendships — that come from face-to-face contact. In-person work fosters innovation, the effects of which on productivity almost certainly exceed the gains from working harder at home for possibly unsustainable stretches. An even slightly higher growth rate once people return to offices will quickly outpace the one-time gain from saved commuting time.

None of this is to say that all white-collar workers should return to the office five days a week. The pandemic will surely hasten a move toward a hybrid model that involves some days at home and some in the office, giving workers welcome new flexibility. Still, it’s important to understand the downsides of the working-from-home economy.

Research has long linked inequality with reduced economic growth, and working from home will make society more unequal. In May 2020, when more than two-thirds of Americans with advanced degrees were telecommuting, less than 15 percent of Americans with only a high school degree or less were doing so. But there are divisions even among white-collar workers: The remote world may seem heavenly for middle-aged professionals with extensive networks of colleagues and comfortable home offices, for example, but it is decidedly less appealing for 20-somethings who are trying to find their way in a new company and to get work done in dark, cramped apartments. As time passes, those workers may find it harder to land promotions and to thrive more generally.

Studies of remote work both before and during the pandemic support that intuition. In separate analyses, economists examined what happened when a Chinese travel agency and a major American retailer split their call centers into groups that worked from home and groups that came into the office. (The Chinese company randomized its workers into the two groups. The American retailer divided its employees less formally.) In both cases, workers at home handled at least as many calls as those who were in the office. In the case study of the Chinese company, published in 2015, productivity went up 13 percent among those who worked at home. In the U.S. case study, employees who went remote — this, too, happened before the pandemic — saw their calls per hour increase by 7.5 percent, a healthy boost to short-run productivity.

But in both studies, the probability of being promoted for the remote workers roughly halved relative to people who worked in person — suggesting serious long-term consequences of remote work. In the U.S. study, for example, 23 percent of on-site workers were promoted within their first 12 months of work, while the remote promotion rate was 10 percent. Promotion meant being assigned to deal with more difficult calls, often with more irate customers. How would you learn to handle those calls if you couldn’t take cues from the workers around you? How would your boss learn that you had the right touch for that particularly angry buyer from Toledo?

A similar dynamic appears to exist in the tech world. Examining hard data on its programmers’ progress, such as specific software contributions, bug fixes and new features, Microsoft reported in January that “developer productivity was stable or increasing at Microsoft and elsewhere post-COVID.” (Still, 30 percent of the company’s workers who were surveyed reported lower productivity during the pandemic.) More recently, however, Microsoft researchers concluded that “firm-wide remote work caused the collaboration network of workers to become more static and siloed, with fewer bridges between disparate parts.” They also found a shift from synchronous communication (like meetings, phone calls and video sessions) to asynchronous communication (emails and text messages, for example). This reduction in the number of new connections and real-time conversations led the researchers to fear that it will be “harder for employees to acquire and share new information across the network.”

The difficulties with learning from a distance appear to make companies particularly wary of hiring new workers into remote positions. When the pandemic began, employment and postings fell for all jobs as demand cratered across sectors. By October 2020, however, non-remote job openings were back to pre-covid levels. In contrast, postings for remote jobs were down 35 percent from February to December 2020. In industries where people worked remotely, the online world seemed to function better for those who were already insiders than for outsiders who were hoping for a better future.

The information-rich nature of face-to-face contact led the great English economist Alfred Marshall to write more than 130 years ago that when workers gather in dense clusters, “the mysteries of the trade become no mystery but are, as it were, in the air.” This remains true. Economic studies find that migrants to metropolitan areas see their wages increase month by month, year by year, which is most compatible with the view that people become more productive by becoming enmeshed in a dense, vibrant, face-to-face setting.

Many other studies affirm this insight, approaching the question from multiple angles. Consider data on emergency call centers — the equivalent of 911 in England. In some cases, the person who takes the call is in the same room as the person who dispatches the officer, and in some cases, they’re in different rooms. Data shows that the time to dispatch an officer is lower when the two are in the same room, and shorter still when their desks are closer together. This is despite the fact that the initial information is filled out and transmitted electronically. In the academic world, research papers written by collaborators in the same building are cited more frequently later on than papers from collaborators at different universities (or even co-authors in different buildings at the same university).

And while some people might be inclined to separate debates over online schooling from discussions of remote work, both involve the creation and dissemination of new ideas. There is copious research showing that the online classroom experience has been disastrous for learning: One study in the Netherlands found that even under ideal conditions, including high rates of broadband access and ample funding, “students made little or no progress while learning from home” and that “learning loss was most pronounced among students from disadvantaged homes.”

The allure of the work-from-home economy is not new. Forty years ago, the futurist Alvin Toffler argued that the new communications technologies of the time (fax machines, early computers) would lead people to abandon offices and leave cities empty. Toffler was wrong then — metropolitan economies rose from their postindustrial ashes — and remains so.

Some amount of teleworking will continue forever; it is just too convenient to go away entirely. But 21st-century companies compete by deploying knowledge and creativity, and these things are sparked more readily when people are in the same room. Video can help us muddle through for a while, but to soar again, we’ll need to get back into the office.