This is one of a series of interviews by Bloomberg Opinion columnists on how to solve the world’s most pressing policy challenges. It has been edited for length and clarity.

Justin Fox: The remote work revolution unleashed by the pandemic has brought huge changes in the labor market, with social and economic implications that we’ll be dealing with for generations. You’ve been one of the most important chroniclers and analysts of this phenomenon. But you started looking at it well before March 2020, right?
Nicholas Bloom, William D. Eberle Professor of Economics, Stanford University: I started working on this in 2004. At that point I’d already been doing a lot of work on management practices, and I began to notice that what I call being-nice-to-people practices—maternity leave, paternity leave, job sharing, part-time, work from home—vary tremendously. It felt strange because management practices seemed to be norming towards the view that you have performance evaluation data, provide people good targets, reward them if they meet the targets, get them training if they fail to. But there didn’t seem to be any consensus on, what’s the right length of maternity leave? What’s the right degree of work from home? And I thought, the first step is just to collect data on what actually exists out there.

JF: What did you find? Was there a specific set of policies that worked better?
NB: I found that pro-work-life-balance policies—maternity leave, job sharing, part-time work and work from home—were positively correlated with good management and positively correlated with good firm performance. It’s impossible to tell what causes what, but firms that perform well appear to have more progressive policies toward their employees.

JF: I’m most aware of this paper you published in 2015 that explicitly looked at remote work in terms of productivity. How did that come about?
NB: That was a classic Stanford experience. I’m teaching my Ph.D. second-year labor class, which is applied statistics, using datasets to look at interesting topics. I discover about a quarter of the way through that one of the people sitting at the back of the class is the co-founding CEO and, at this point, chairman, of Ctrip.com [now Trip.com], one of the world’s largest travel agents. I was kind of amazed and surprised, and started talking to him. James Liang, the student, said, “We’re thinking of evaluating work-from-home because office space in Shanghai is very expensive and we have this large call center.” I was like, “That sounds great, I’d love to get involved.” We went on to run a 250-person randomized controlled trial.

Going into the study, the company’s view was that work from home would save space but come at a productivity cost: working from home, shirking from home. The joke back then was that the three great enemies of work from home are the bed, the television and the fridge.

When the experiment was up and running and we started to see the data, we were astounded to see that productivity was up, not down. Maybe in 2022 it doesn’t seem surprising, but back then in 2011, it was pretty amazing. Productivity was up 13% for the people working from home, which is a huge improvement. Of that 13% increase, about two-thirds was due to the fact they were working more minutes because they were late less, and took shorter lunch and toilet breaks. Then one-third was that they were more productive per minute.

At the end of the experiment the company said everyone working in the call centers could work from home. But the uptake was low. By 2020, pre-pandemic, there were only 50, 60 employees left doing it. I asked James Liang why, and he said there was a negative stigma associated with volunteering to work from home. There’s a paper by [Natalia] Emanuel and [Emma] Harrington showing the same thing in U.S. call centers pre-pandemic. People are about 10% more productive when they work from home, but there’s about a minus 10% selection effect. The pandemic obviously changed all of this because everyone was doing it. Now this negative stigma has disappeared and it’s become normalized.

In July 2021, I met James for dinner in London. He said, “We’re running another big experiment, this time on hybrid, would you be interested in getting involved?” I jumped at the chance. They took 1,600 people that are coders, finance and marketing professionals—around 25% managers—and randomized between fully in the office and working from home two days a week.

The results came in in February 2022, and they showed productivity performance is about flat. Maybe a mild positive, but nothing substantive. The big benefit was that quit rates dropped by a third and employees’ job satisfaction, work-life balance and intention to stay in the firm were significantly higher.

JF: You’re at home right now. Have you worked from home all along?
NB: I actually have a very unusual working pattern. I live on the Stanford campus so I have about a five-minute commute. Even before the pandemic, I worked hybrid, as in 50% at home, 50% in the office, but I tended to go in every day. There’s a range of activities I do each week that are definitely better in person: research seminars, teaching, advising students. Then there are other activities that I prefer to do at home: quietly working on data, writing, reading, Zoom calls with collaborators in different locations.

JF: In the U.S. you’ve been involved in another data collection project, WFH Research. Describe that and how it got started.
NB: It was a Stanford student, Jack Blundell. I was his advisor, and he’d been running these online surveys looking at gender in the labor market in the U.K. I thought, I’d love to collect information on what’s happening right now on working from home, and so started to run these surveys. Since May 2020, we’ve been surveying 5,000 Americans aged 20 to 64 every month, and asking relatively straightforward questions on levels of work from home, post-pandemic desires by them, post-pandemic intentions by their employers and various other factors.

This is now maybe the best data in the U.S. on work from home. You’d think there’d be an official statistic where the Bureau of Labor Statistics would provide data. They do, but their question unfortunately was well-designed for the pandemic but is not well designed for 2022. It asks, “How many days a week do you work from home, because of the pandemic?” The “because of the pandemic” is a problem, because in 2022 it excludes people who worked from home pre-pandemic and people who are now working from home because their job offers it, independent of the pandemic. So that BLS series has fallen to below 10%, and it looks like no one’s working from home, which is not the case.

JF: In your time series, the percentage of days worked from home is stuck at a pretty high level. What’s changed is what people say about their own and their employer’s intentions going forward.
NB: This is like a play in three parts. In May 2020, Mark Zuckerberg said Facebook has decided it may stick with working from home post-pandemic. It was huge news. Up until that point, no firm had committed to sticking with it. It was seen as a short-run pandemic measure.

Roll it forward a year to May 2021, and for professionals and managers, most firms are saying you’re going to work from home post-pandemic, maybe one, two days a week. Three in the office, two at home, is the most popular plan.

Now we’ve had another year of heavy work from home, of tight labor markets, of productivity growth. Work from home has become entrenched and has actually grown. A good example of this is Apple. Apple announced its “three-two” plan in June 2021. There were complaints from employees, but our survey data suggested that was roughly in the middle of what people were wanting and what employers were promising a year ago. Now Apple’s plan looks kind of mean because the rest of tech has become increasingly generous.

It’s hard to know where we’ll settle down. I think most professionals, managers, most people reading this should expect to see something like two to three days a week in the office. One of my neighbors is a doctor. She now goes into the office four days a week and does telemedicine one day a week, which is new. But she says that’s roughly the mix that her patients want.

JF: One thing that’s striking in your survey data is when you ask people their preferences, the hybrid option that everybody’s talking about, two or three days in the office, is the least popular.
NB: This is why means and medians can be misleading. Just to explain in words, roughly one quarter of people don’t want to work from home at all. They want to go into the office five days a week. If you look at their demographics, they tend to be younger singles or older empty-nesters. Then one third are the opposite extreme. They want to work from home five days a week. They tend to be married with younger kids and have a long commute. The rest is spread out in between. You have a distribution that I call dumbbell-shaped. There are two masses at the edge and then a long, skinny bit in the middle.

It looks unlike most distributions you see in economics. For example, if I survey people about what temperature they want the air conditioning in the office, that’s typically a normal distribution: most of them in the middle, a few extremes. If you put something in the middle, most people are close to happy. With work from home, it’s really hard. The mean of that distribution is about 2.5, but as you point out, not many people actually live in that two or three days a week.

What I think is going to happen in the long run could be a bifurcation of the labor market. You’ll have some firms that say our thing is in-person, we are going to be a firm where people come in five days a week. One quarter of the labor market will love that and they’ll flock to that firm and the other three quarters will slowly drift away. Then there’d be other firms that are going to say our thing is fully remote, and if you’re a fully remote type person, go work for that company. I think there’s going to be a great reshuffle as companies start to differentiate themselves on work-from-home policies.

JF: A lot of your not-work-from-home work has been about management practices and which ones lead to more productivity, and on an economy-wide scale more growth. Is there potential for big productivity gains coming out of this re-sorting of work?
NB: There are two angles for gain from this. One is just productivity of people that can work from home. Our estimates are that it might increase their productivity by 3%, 4%, 5%. They’re maybe half the labor force, but about two-thirds of earnings, which is what’s relevant for GDP. So you can think about that as a 3%, 4% increase in GDP, stretched out over the next few years.

The other benefit that possibly is even larger in the long run is the positive impact on labor supply. There are a number of groups that are more able to work because of working from home. A lot of people—think folks with young kids, people that are disabled, people that are close to retirement age, students—may be happy to work three days a week for six hours a day without the commute, but wouldn’t be prepared to do that for five days a week, including commuting. I think that could push up labor supply by 2%, 3%, 4%, which would be a very large impact on growth.

JF: What about the negatives? One would seem to be the impact on cities. New York seems pretty lively, although I guess we’re going to have to figure out how to pay for our subway if usage stays 30% below what it used to be. And there’s the half of society that can’t work remotely.
NB: A lot of individuals, professionals, managers that used to live in city centers of New York, San Francisco, L.A., Chicago, etc., have decided, look, if I’m only having to come into work two or three days a week, I can live out in the suburbs and commute in. So you see evidence of 5% to 10% of the population leaving the city center. That’s pushed down city center rents and property prices. I see that as good because at the margin, some essential service workers that need to get to work five days a week are going to find it’s relatively more affordable to live in city centers. It’s still extremely expensive, but it’s kind of pushed back a little bit on the affordability crisis, which was a big 2019 issue.

The main worry I have is the one you highlighted, which is the strain on city budgets and particularly mass transit. The issue with mass transit is that almost all the costs are fixed, but the revenues are variable. So if you have a 30, 40% drop in ridership on the subway or BART or the London Underground, which is what’s forecasted, then your costs fall by a few percent and your revenues fall by 30%. Who pays for it? When you say the city should pay, the city is also facing lower property taxes, lower retail taxes, lower hotel taxes. Suburbs of big cities are doing very well, and I think we need to think about redistribution from suburbs to core city centers. Because if you let cities go bankrupt or just shut down the subway, that’s Carmageddon—and everyone suffers.

Justin Fox is a Bloomberg Opinion columnist covering business. A former editorial director of Harvard Business Review, he has written for Time, Fortune and American Banker. He is author of The Myth of the Rational Market.