When the COVID-19 pandemic began in 2020, daily life changed for billions of people in countless ways. The most important, from the perspective of real estate in superstar cities, was a sudden and massive shift to remote work, a change that was enabled by widespread adoption of videoconferencing and file-sharing technology. In the blink of an eye, millions of employees were working not at their offices downtown but from their bedrooms and couches.
That shift affected their behavior in three main ways. First, of course, it changed where they worked. Second, it changed where they lived; untethered from their daily commutes, many of them moved away from urban cores. Third, it changed where they shopped, as online shopping and stores near home became more attractive than urban establishments. Those three behavioral changes, in turn, affected the three classes of real estate—office, residential, and retail—that we discuss in the next chapter.
The three changes were not all permanent, but they have left cities permanently altered. In this chapter, we examine them in detail. Our research suggests that remote work will continue, that accelerated urban out-migration is slowing but not reversing, and that shopping in cities will remain weaker than it used to be.
Hybrid work is here to stay
After abandoning their offices at the start of the pandemic, employees are now working there more often—though still far less than they used to. Office attendance varies by city; for example, it tends to be lower in cities with expensive housing and a large share of knowledge-economy workers. It also varies by industry and workers’ demographic characteristics. There are several reasons to think that office attendance has stabilized and will continue at current rates.
Office attendance is still 30 percent lower than it was before the pandemic
In early 2020, office attendance plummeted for workers in superstar cities. Lockdowns, office closures, and uncomfortable masks gave them a reason to work remotely, and existing technology gave them the means. Office attendance fell by 90 percent in New York City and San Francisco.1 The drops in Austin, Dallas, Houston, and Los Angeles were less severe but still stark at about 70 percent.
As of fall 2022, workers were going to the office just 3.5 days per week, on average, some 30 percent below prepandemic norms.
Three years later, remote work has given way to hybrid work (a combination of in-office and remote work), and office attendance has rebounded substantially. But as of fall 2022, workers were going to the office an average of just 3.5 days per week, some 30 percent below prepandemic norms, according to our survey. (For more information about the survey, see the technical appendix.)
Of all the survey respondents, 37 percent go to the office every day (Exhibit 1). An additional 56 percent have hybrid work arrangements and thus spend one to four days per week in the office. And 7 percent work fully remotely.
Office attendance is lower in metropolitan areas with expensive housing and a large share of knowledge-economy workers
Office attendance varies among metropolitan areas. Employees in Beijing, for example, go to the office 3.9 days per week, on average, while those in London go just 3.1 days per week. The other metropolitan areas we studied ranked between those extremes and were mostly clustered around 3.3 to 3.6 days in the office. Similarly, the percentage of workers who do not work from home at all varies by country. In Japan, more than half of all office workers spend five days per week in the office. In the United Kingdom, just 28 percent do.
Two urban characteristics seem to matter most in determining office attendance in a metropolitan area: its proportion of knowledge-economy workers and its housing costs (Exhibit 2). According to our regression analysis of US counties, the higher a county’s ratio of knowledge-economy workers to other workers, the lower its average office attendance. (We define the knowledge economy as the professional services, information technology, and finance industries.) Similarly, the higher home prices are in an urban core relative to its suburbs, the lower the metro area’s average office attendance is (see Box E1, “How we define cities,” in the executive summary). Other characteristics of US counties with lower office attendance, according to our analysis, are a larger share of commuters and a smaller share of retail employees (which suggests a more limited retail presence). It seems reasonable to think that those relationships would also exist in superstar cities elsewhere in the world, but we were not able to confirm that.
Culture and government policy also play a role in office attendance. For example, Chinese cities’ high rates of office attendance today can probably be explained by the widespread “996” office culture (for “9 a.m. to 9 p.m. in the office, six days a week”). And Beijing’s and Shanghai’s citywide lockdowns reduced office attendance to near zero when they were in effect.
Office attendance is lower in large firms in the knowledge economy
Employees at larger firms report significantly lower office attendance than do those at smaller firms (Exhibit 3). For example, at firms with 1,000 to 9,999 employees, workers go to the office 3.3 days per week, on average; at firms with 50 to 99 employees, attendance averages 3.7 days. One potential reason is that larger companies tend to have more resources and technology to support working from home. Another is that the sense of community may be stronger at smaller companies, prompting more employees to visit the office.
Also, client meetings are less of a reason to come to the office for employees at larger companies than for those at smaller ones. At companies with 1,000 to 24,999 employees, 25 percent of survey respondents ranked meeting with clients as one of the top three reasons to come to the office, whereas at companies with fewer than 100 employees, 32 percent did. A similar pattern exists for employees in client-facing roles: among salespeople at companies with 1,000 to 24,999 employees, 42 percent go to the office five days per week, a share that grows to 54 percent at companies with 2 to 99 employees.
And employees in the knowledge economy go to the office 0.2 fewer days per week than do those in other industries. One reason could be that knowledge-economy industries are highly digitized and therefore amenable to remote work. Another could be that employees in those industries therefore see less need for commuting and have a lower tolerance for it. Even though their commutes are no longer than those of other workers, they were much likelier than other workers to call saving commuting time one of their top three reasons for working from home.
Office attendance varies in subtle ways by age, income, and seniority
The average number of days that respondents work in the office does not vary much by age, income, or seniority (Exhibit 4, left side). For example, millennials reported working 3.4 days per week in the office, and members of Generation X and baby boomers reported 3.5. The starkest variation we found was among millennials in senior positions, who go to the office 3.7 days per week, more than any other group we studied.
However, there are some differences in the likelihood of those groups to go to the office a particular number of days per week (Exhibit 4, right side). Two differences in particular stand out. First, baby boomers are much likelier to go to the office five days per week than younger workers are. Second, junior-level workers are much likelier to go to the office five days per week than midlevel and senior employees are; perhaps counterintuitively, they are also more likely to work entirely remotely, never going to the office at all.
Furthermore, cluster analysis of respondents in the United States reveals statistically significant groups among those who mostly go to the office and those who mostly work from home. There are two such groups among people who mostly go to the office: high-earning male C-suite employees and low-earning female junior employees. There are also two such groups among people who mostly work from home: knowledge-economy workers who earn over $200,000 per year and, again, low-earning female junior employees.
Employees work from home to save commuting time, and they work from the office to see their teams
When we asked survey respondents who were able to work from home what their top reason was for doing so, the most popular answer globally, as we expected, was to save commuting time, followed by increasing productivity and saving money (Exhibit 5). The answers varied by country, however. For example, 38 percent of respondents in Japan cited saving commuting time as their top reason, whereas just 11 percent of those in China said the same. Saving money was the top reason for 11 to 15 percent of respondents in Europe but for just 3 percent of those in Japan. In all countries, however, respondents called increasing productivity one of the top reasons.
The most popular reasons given by workers with flexible work arrangements for choosing to come to the office were to be able to work with their teams, to comply with an employer’s policy, and to increase productivity. (The fact that increasing productivity was cited as a top reason for both going to the office and working from home suggests that where employees feel most productive may differ by individual or task.) Those findings were substantially consistent in France, Germany, the United Kingdom, and the United States. In China and Japan, however, few respondents cited an employer’s policy as their top reason for going to the office.
The current rate of office attendance may persist
Three indicators suggest that the current rate of office attendance may have stabilized. First, in ten of the most populous cities in the United States, office attendance has held roughly steady since mid-2022, aside from a normal seasonal dip around the winter holidays (Exhibit 6).2 Second, three numbers—the number of days per week that survey respondents go to the office, the number of days that they expect to go to the office after the pandemic ends, and their preferred number—are not far apart (Exhibit 7).
Third, a sizable and influential group of office workers strongly prefer to continue working remotely. About 10 percent of respondents worldwide said that they were both likely to quit their jobs if required to work at the office every day and willing to trade more than 20 percent of their compensation to work their preferred number of days from home. That group is comparatively senior and well paid; 44 percent of them are in senior roles, and 33 percent of them earn more than $150,000 per year (Exhibit 8). Their seniority and high incomes suggest that they are probably decision makers who can protect remote work at the team or company level.
Nevertheless, it is not certain that office attendance will stay at its current level. For example, employees might gain bargaining power and reduce the number of days they work in the office, or employers might gain bargaining power (in the event of a recession, for example) and increase that number. Improvements in hybrid work technology and processes could make it less necessary to come to the office.
Office attendance could also rise if employers find ways to attract workers back to the office. As pandemic restrictions loosened, employers often offered perks, such as free meals and in-office recreational areas, to encourage employees to return. But that approach appears to have succeeded only on a very limited scale; just 3 percent of respondents who were able to choose where to work called such perks their top reason for going to the office.
In the long run, what chiefly determines the rate of office attendance will be its impact on productivity. If research conclusively indicates either a negative or a positive relationship between hybrid work and productivity, that could push office attendance up or down, respectively. Studies conducted by companies and research institutions have so far shown no negative relationship between hybrid work and productivity. 3 For example, a previous MGI report found that about 20 to 25 percent of the workforces in advanced economies could work from home between three and five days a week without a loss of productivity.4 Employees themselves seem to believe that hybrid work helps productivity; in our survey, 87 percent of survey respondents indicated that they would be more productive if they worked their desired number of days at home than if they had to work at the office five days per week.
Up to 7 percent of the people in urban cores left for good
For more than 20 years, superstar cities were the center of global growth. Just 1 percent of the Earth’s surface was home to more than 50 percent of global GDP growth and population growth from 2000 to 2019.5 The two types of growth reinforced each other: “agglomeration economies” fueled economic and productivity growth in these cities, resulting in high-income jobs that attracted still more people.6 They came even though housing tended to be expensive—especially in superstar cities where the supply of housing was very constrained, such as San Francisco.
Then the pandemic began, prompting a wave of households to leave superstar cities and thus driving down population (Exhibit 9). Munich’s population grew by 1.0 percent per year from 2014 to 2019, for instance, but it grew by only 0.3 percent per year from 2020 to 2022. In Boston, population growth during the first period turned into population decline during the second. Urban cores were particularly affected. For example, New York City’s urban core lost 5 percent of its population from mid-2020 to mid-2022, and San Francisco’s lost 7 percent. One reason for the surge of out-migration was hybrid work. Now that people were not working at the office, they cared less about living nearby and could instead prioritize better housing conditions.
Out-migration has since declined, but it has neither ended nor reversed, and it seems to remain higher than it was before the pandemic began. That is, the people who left the urban cores are not coming back, and many others are still leaving.
During the pandemic, a wave of households in superstar cities moved from the urban cores to the suburbs, a trend that was strongest in the United States
In all but two of the superstar cities we studied, after the pandemic began, the difference between urban and suburban population growth shifted to the suburbs’ advantage (Exhibit 10).7 Exactly how much varied from city to city. The difference between urban and suburban population growth was largest in London, Dallas, Houston, New York City, and San Francisco. In most US cities, suburbanization had already been happening and was simply accelerated by the pandemic. By contrast, all the European cities that we studied except Paris and Munich had been urbanizing before the pandemic, so the suburbanization represented a reversal. The Japanese cities had also been growing more urban before the pandemic, but there, the trend continued even after the pandemic started.
Population trends in Beijing were very different from those in other cities. For years before the pandemic, the government had relocated Beijing’s residents to suburban areas. During the pandemic, however, it imposed policies restricting movement; as a result, population growth in Beijing’s urban core nearly matched that in its suburbs.8
The suburbanization effect may have been stronger in US superstar cities than in European and Japanese ones because the urban cores of US superstar cities tend to be office-dense and retail-poor, whereas the European and Japanese urban cores tend to follow a mixed-use model in which office, residential, and retail space exist alongside one another. Supporting that contention is our statistical analysis of US counties, which indicates that those experiencing the highest rates of out-migration during the pandemic tended to have four characteristics: expensive homes, high office density, a high share of workers in the knowledge economy, and a low share of workers in retail, which suggests a more limited retail presence (Exhibit 11).9
The same patterns held true at the zip code level during the pandemic. When we conducted an analysis of neighborhoods in San Francisco County, New York County (Manhattan), and Harris County (which is mostly Houston), we found that people moved out of expensive, office-dense zip codes and into cheaper ones with more mixed use of real estate.
Hybrid work encouraged the out-migration
To learn more about why people left urban cores during the pandemic, we identified survey respondents who had moved after March 2020 for pandemic-related reasons. They represented the majority of all respondents who had moved after March 2020 (Exhibit 12). Roughly a third of them said that their moves would not have been possible without remote work.
Pandemic-related out-migration after March 2020 was led by young households, and in particular by young people who had caregiving responsibilities and who valued remote work. Respondents who moved because of the pandemic were younger than those who did not move (averaging 36 years of age as opposed to 43); were more prone to say that they were “likely” or “extremely likely” to quit their jobs if forced to go to the office five days a week (39 percent versus 26 percent); were likelier to be caregivers (59 percent versus 44 percent); and had bigger households, on average (3.1 people versus 2.9).
Many of these moves happened because employees untethered from their daily commutes began to care less about how far they lived from the office. Of all survey respondents who went to the office less than five days per week, 38 percent said that remote work had made them more willing to live farther from the office. In the United States, 55 percent of respondents who moved during the pandemic went to places that were farther from the office than their previous homes had been, while just 33 percent moved closer.
Out-migration has since slowed, but people are still leaving urban cores
Out-migration from superstar cities’ urban cores seems to have slowed of late. Between 2020 and 2021, 0.32 percent of the US population left superstar urban cores; the following year, that share fell to 0.24 percent (Exhibit 13).10 But that loss is still larger than it was before the pandemic. In other words, the people who moved out during the pandemic are not moving back, at least not yet. The main beneficiaries of out-migration seem to be rural and suburban areas near non-superstar cities.
Our survey suggests that the people who left urban cores in other countries are not coming back either. In China, 41 percent of pandemic-related movers said that they might move back, but that share was less than a quarter in France, Germany, Japan, and the United Kingdom. Pandemic-related movers’ strong preference for hybrid work—and their professed willingness to quit their jobs if denied it—also implies that they are unlikely to move back.
Shopping remains depressed, especially in urban cores
Just as the shift to hybrid work has changed people’s residential preferences, it has changed their shopping preferences. During the height of the pandemic, foot traffic near brick-and-mortar stores plummeted, especially in urban cores. The office workers who used to sustain those stores shopped there far less, shifting their purchases to the internet and to stores near their homes in the suburbs. More recently, foot traffic near urban stores has risen again, but it remains lower than it was before the pandemic, spelling problems for retailers.
As people stayed home during the pandemic, they radically changed the way they shopped
In March and April 2020, as the pandemic began, foot traffic near stores plummeted in the superstar cities we studied (Exhibit 14).11 The declines were starkest in London and Paris, where it fell by about 80 percent. In New York City and San Francisco, it fell by 50 to 60 percent during the same period, while in Houston and Tokyo, it declined by only about 30 percent. The different outcomes may have been due to variations in government policies, degrees of tourism, acceptance of remote work, and other cultural norms. For example, the strict lockdowns that London and Paris imposed in March may have contributed to those cities’ precipitous declines.
One trend that helped consumers visit stores less often was a quickening shift to online shopping (Exhibit 15). Online spending as a share of all retail spending had already been growing; in the three years before the pandemic, that share grew by 0.4 percentage point per year in Japan and by 2.2 percentage points per year in China, for example. But then the pandemic began, and the share spiked from the end of 2019 to the end of 2020. The spike occurred in all the countries we studied; it was smallest in Japan (1.3 percentage points) and largest in China (9.9 percentage points). The pandemic effectively accelerated the use of online spending by three to four years.
An interesting fact is that the countries seeing the sharpest increases, China and the United Kingdom, were the ones where online spending represented the largest share of all retail spending before the pandemic. Perhaps those countries were the most primed for e-commerce growth—with greater customer awareness, more mature distribution, and more adept retailers—and were therefore best able to meet the surge in demand during the pandemic.
Shopping behavior has started to return to prepandemic trends, yet foot traffic and purchasing in stores remain lower
Although it spiked during the first year of the pandemic, online spending as a share of retail spending is now at the same level that it would have reached by now if the pandemic had not happened. It has done that either by growing more slowly (as in France and Germany) or by shrinking (as in China and the United States), though of course it remains higher everywhere than it was before the pandemic.
Yet foot traffic near stores remains down. As of October 2022, it appeared to have stabilized at a level 10 to 20 percent lower than the prepandemic level in the metropolitan areas we studied.
Spending in stores also remains down, though not as much as foot traffic (Exhibit 16). In most metropolitan areas, it was estimated to be as much as 10 percent lower in 2022 than it had been in 2019 in real terms—though in Beijing, where lockdowns remained in force through most of 2022, it was far lower. Although this is bad news for stores, at least the decline in spending has not been as steep as the decline in foot traffic: shoppers have consolidated trips, making purchases more frequently when they do visit stores and spending more per trip.
And though consumers famously shifted their spending from services to goods during the pandemic, they have been slowly shifting it back to services since mid-2021. In the United States, for instance, spending on goods has stabilized since the end of 2021, but spending on services has continued to grow past prepandemic levels in real terms (Exhibit 17). In France, spending on services grew in 2021 in real terms; in 2022, it grew again, even as spending on goods fell.
Those trends will pose challenges for brick-and-mortar stores’ sales. The upward march of online sales will eat away at the stores’ revenues, even though that march has slowed since 2020. Lost foot traffic will do the same unless stores find ways to improve their marketing or encourage the customers who do visit to make more and larger purchases. (The most successful retailers are rethinking their store formats to support omnichannel experiences—those in which customers’ in-store and online experiences, including the marketing aimed at them, are integrated.) The spending shift from goods back to services may further reduce foot traffic to stores selling goods. And other trends could further complicate stores’ prospects. For example, “revenge spending,” consumers’ tendency to spend heavily after the pandemic to make up for lost time, will dissipate, or economies around the world could enter recessions.
Declines in foot traffic near stores have been particularly persistent in urban cores and office-dense neighborhoods
Retailers in urban cores confront even greater challenges. As of October 2022, foot traffic had recovered noticeably less near those stores than near suburban ones (Exhibit 18). In New York, for example, foot traffic near suburban stores was 16 percent lower than it had been in January 2020, but foot traffic near urban stores was 36 percent lower.
And retail traffic has been still lower in cities’ office-dense neighborhoods. For example, the City of London, the borough in London with the highest ratio of office jobs to residents, experienced a nearly 50 percent decrease in foot traffic near stores from February 2020 to June 2022—by far the most severe and sustained decline of any borough in London. Foot traffic near stores was also hurt by declines in tourism, but more mildly; in London, boroughs that are popular with tourists, such as Kensington and Westminster, fared much better than the City of London.
Part of the reason for office-dense neighborhoods’ declines in foot traffic seems to be that many of the employees who might shop there spend more time working from home than they used to. In our survey, respondents in the United States who worked at the office no more than one day per week reported doing much less of their total retail spending near the office than those who worked in the office for two to five days a week reported (Exhibit 19).
However, the struggles of stores in urban cores may present opportunities for those in the suburbs. The same survey respondents—those who spent less at stores near the office—also spent more at stores near home.