The pandemic had major effects on office workers in superstar cities, as the previous chapter described. Above all, the pandemic changed their working habits, encouraging many employees to work from home instead of in the office. As they abandoned offices in urban cores, they also started moving to the suburbs in search of larger homes in greener areas. And they shopped less at stores near the office and more online.
Those three changes could have equally significant effects on demand for office, residential, and retail real estate. To find out how large those effects might be, we built a model that projects demand for office, residential, and retail space in a number of scenarios, including a moderate one and a severe one. The model used information from the large global survey that we conducted. We also considered a wide variety of factors, including long-term population trends; employment trends, such as the ongoing effects of automation on employment; office attendance patterns by industry; employee coordination, defined as the maximum share of workers in the office at a given time that employers need to maintain; workers’ ages and incomes; the share of a city’s population that commutes from elsewhere; housing prices; and shopping trends, such as the ongoing increase in online shopping.
In the scenarios that we modeled, demand for office and retail space is generally lower in 2030 than it was in 2019, but the reductions are smaller than those projected by many other researchers. Residential space is less affected, though the price differences between urban cores and suburbs are narrower than they used to be. (For more about the modeling methods and differences among the scenarios, see the technical appendix.)
There will be less demand for office space
Demand for office space in superstar cities has been falling since the start of the pandemic as a result of lower office attendance. Our model suggests that the problem will only worsen over the next seven years, at least in most of the cities we studied.
Demand for office space has already fallen, and vacancy rates have grown by as much as 13 percent in the cities we studied.
Since the start of the pandemic, lower office attendance has driven demand for office space down and vacancy rates up. In every city we studied, the percentage of office space that was vacant rose from 2019 to 2022 (Exhibit 20). The increase ranged from two to 13 percentage points, depending on the city. Moreover, the size of the increase in the cities we studied correlated with the degree to which office attendance there fell, showing that the two phenomena were related. For example, in San Francisco, which experienced one of the steepest declines in office attendance in our study, office vacancy was more than ten percentage points higher in 2022 than it was in 2019. That change was the largest of the vacancy rate increases we observed and seven percentage points higher than the national average. It was also a stark reversal of prepandemic conditions: San Francisco was the tightest US market in 2019, when just 6 percent of its office space was vacant (four percentage points below the national average).
In most superstar cities, lower office attendance has similarly driven down asking rents in real terms.1 Rents in the US cities we studied fell especially sharply—for example, by 28 percent in San Francisco and by 18 percent in New York City from 2019 to 2022. Rents in the European cities we studied have been more resilient. In Paris, London, and Munich, they fell by 10 percent, 12 percent, and 9 percent, respectively. In only two of the cities we studied, Munich and Tokyo, did rents actually increase; they grew the most in Tokyo, increasing by 4 percent from 2019 to 2022.
Occupancy and rents may fall further still. Because their employees are coming to the office less often, many employers have downsized spaces to reduce costs. More will do so; thanks to the lengthy nature of leases in the office sector, nearly half of all tenants have not made a renewal decision since the onset of the pandemic, and macroeconomic uncertainty could give them further reasons to downsize. Indeed, some tenants have chosen not to wait for their renewal dates and instead have bought their way out of long-term contracts. Some landlords have offered tenants rent abatements and higher tenant improvement allowances (capital for physical improvements or alterations), but those actions, while they may have propped up occupancy, have reduced effective rent. And though office attendance seems to be stabilizing, tenants may remain hesitant to commit to new long-term leases, unsure of how much space they will need in a few years.
There is historical precedent for rents’ falling further than they have so far. During the global financial crisis of 2007 and 2008, and also when the dot-com bubble burst in the early 2000s, asking rents in London, New York City, and San Francisco fell by more than 30 percent. They dropped by 21 percent, on average, from their peak in the fourth quarter of 2019 to the fourth quarter of 2022.
The decline in demand has allowed tenants—wary about current macroeconomic conditions, uncertain about how much their workers will come to the office, and therefore uncertain about how much space they will need—to negotiate shorter leases from owners. Shorter leases, in turn, may make it more difficult for owners to obtain financing or may cause banks to adjust valuation models, which rely in part on the duration of existing leases.
Also, because of uncertainty about demand, transaction volume—the dollar value of all sales of office buildings—has fallen dramatically. The difference between the prices requested by owners and the bids placed by potential buyers remains wide, financing costs are higher than they were during a frenzied trading environment in 2021, and debt availability remains depressed.2 In 2022, global volume fell in each consecutive quarter, and the total for the year was nearly 28 percent lower than it was in 2019. At the city level, transaction volumes have broadly correlated with sale prices. For example, transaction volume in New York City was 36 percent lower in 2022 than in 2019, and the average sale price per square foot fell 13 percent over the same period. Conversely, transaction volume in Atlanta was 31 percent higher in 2022 than in 2019, and price per square foot grew by 37 percent over the same period.3
Current owners of office buildings hope for a return to prepandemic levels of net operating income. But potential buyers may not share that hope, according to a common indicator of the market outlook for assets. Capitalization rates, which equal properties’ net operating income divided by their current value, have been rising, and that is typically a sign of lower future valuations. In the fourth quarter of 2022, total returns for the office sector in one index were –4.80 percent, distinctly lower than the –3.50 percent returns for all asset classes.4
Demand for office space in the median city we studied is projected to be 13 percent lower in 2030 than in 2019
According to our model, demand for office space in a moderate scenario will be 13 percent lower in 2030 than it was in 2019 for the median superstar city in our study. Remote work and flexible working arrangements drive that projected decline by reducing office attendance. Also, the amount of space allocated to each office worker is expected to shrink as pandemic-related concerns wane, further reducing demand. However, population growth may partly or fully offset the expected demand decline in some cities. (Our model does not consider price elasticity; that is, it does not account for the fact that when demand decreases, prices fall, pushing demand partway back up.)
In the scenarios we modeled, demand declines most precipitously in San Francisco (Exhibit 21). One reason is that before the pandemic, the share of employees who were office-based in San Francisco was among the highest in the United States, so office attendance there fell unusually sharply. Furthermore, San Francisco had an unusually large share of workers in technology and professional services, who were among the most able to work from home at the start of the pandemic and have subsequently been among those who go to the office least frequently. On their own, those pandemic-driven changes would push down demand for office space in 2030 by 23 percent in the moderate scenario. But population and office employment growth will push demand up and slightly offset that effect so that the net decline in demand is 20 percent.
Houston, by contrast, shows the strongest future demand for office space among cities in our study. In the moderate scenario, such demand is projected to grow by 2 percent between 2019 and 2030. That growth is partly owing to Houston’s high office attendance; pandemic-driven changes in people’s working behavior are thus projected to reduce demand for office space in Houston by just 10 percent, a smaller drop than in some other cities. But it is also owing to strong population growth. Houston has long benefited from a broader trend of migration to US Sunbelt cities, and that trend, by pushing up total office employment, should boost demand by a powerful 13 percent, more than offsetting the pandemic-driven changes.
Our model also estimates a severe scenario for office-space demand, one in which office attendance remains at current levels indefinitely and out-migration from urban cores is higher. In that scenario too, San Francisco fares worst among the cities we studied and Houston best. In San Francisco, demand for office space would be 38 percent lower in 2030 than it was in 2019. In Houston, demand would be 10 percent lower.
The demand shortfall that we project, along with growing supply resulting from the completion of new office buildings, will boost vacancy rates. In 2030, in the superstar cities we studied, there will be excess supply of office space of 7 to 21 percent, according to our model’s moderate scenario (Exhibit 22). We define excess supply as the percentage of office space that is vacant beyond structural vacancy (the average from 2014 to 2019). In essence, it represents the part of projected vacancies attributable to the pandemic. It will be at least 15 percent in more than half the cities we studied and be particularly high in San Francisco, Shanghai, Beijing, London, and Munich.
Excess supply is caused by two factors: decreasing demand for office space (discussed above) and continued supply growth (that is, new office buildings entering the market). For example, in Houston, excess supply is projected to be low (7 percent) because, although supply growth will be about as high as in several other cities, demand is actually expected to grow, thanks to employment growth and in-migration. By contrast, excess supply will be 20 percent in Shanghai, driven by a relatively severe decline in demand and relatively high supply growth.
The total vacancy rate in the cities we studied will be considerably higher than the excess supply. It will be more than 20 percent in all cities except Tokyo and Munich in 2030. Tokyo’s projected vacancy rate, 13 percent, is the lowest in our study, which may be due to Tokyo’s transit-oriented design and general resiliency against the pandemic’s effects.
Excess supply can be reduced in only two ways: boosting demand (say, by attracting companies and jobs to fill vacant spaces) or shrinking supply (by demolishing office buildings or converting them to other uses). We discuss those possibilities in chapter 4.
In the cities we studied, $800 billion in office-space value is at stake
In the nine cities we studied, a total of $800 billion in office-space value (in real terms) is at stake by 2030 in the moderate scenario. On average, the total value of office space declines by 26 percent in the moderate scenario and by 42 percent in the severe one from 2019 to 2030. The impact on value could be even greater if rising interest rates compound it. Similarly, the impact could increase if troubled financial institutions decide to more quickly reduce the price of property they finance or own.
Although other factors could drive a decline in value, our model uses only the impact of changing demand. We assume in both scenarios that market capitalization rates will remain constant from 2022 onward. In the moderate scenario, we assume that rents will remain constant in real terms from 2022 through 2030. In the severe scenario, by contrast, we assume that rents will be 30 percent lower in 2025 than they were in 2019 and then remain constant in real terms until 2030.5
The decline in value is already occurring: capitalization rates for office space in the United States have risen from 5.8 percent to 8.0 percent over the past three years, implying value erosion of more than 35 percent before accounting for net operating income decline.6 In the worst case, falling property values could lead to a “doom loop” for some buildings. A doom loop begins when a building’s value declines and its owner consequently holds less equity in relation to debt. That shifting capital structure makes it harder for the owner to secure financing. As a result, aging and low-quality properties may remain unrenovated or not repurposed, further reducing property values. That phenomenon could be exacerbated by office tenants’ increased preference for higher-quality spaces (see sidebar, “The ‘flight to quality’ in office space”).
The decline in demand for office space has been accompanied by a shift in tenants’ preferences toward high-quality space. From 2020 to 2022, rents and demand for Class A space grew more quickly than those for Class B space in many major US cities (Exhibit 23).
While many tenants are downsizing to save money, others are paying more per square foot to provide attractive space to their employees. Some are doing both: a large professional services firm recently consolidated its footprint in New York City, closing two older spaces covering 800,000 square feet and setting up a new space in 450,000 square feet of top-quality space.
There are a number of reasons for this “flight to quality.” One is that older office space is often not suited to hybrid work; for example, it may have less sophisticated audiovisual equipment. Also, now that hybrid work has reduced the total amount of space that employers need, they can spend their budgets on smaller amounts of higher-quality space rather than larger amounts of lower-quality space. The main reason, though, is that many employers see high-quality space as a way to encourage office attendance among their employees. (Among other things, Class A buildings tend to be near transit and thus can reduce commuting time.)
The survey described in chapter 1 provides some evidence that the strategy could be successful. Among the respondents who said that they would be willing to trade more than 20 percent of their income to work their desired number of days in the office rather than go every day, one of the top reasons given for sometimes going to the office was to work in a more pleasant space. That is, even the people least inclined to work at the office seem to be attracted to quality space.
Demand growth for residences will be muted, especially in urban cores
Before the pandemic, housing prices generally rose more quickly in superstar urban cores than in their suburbs. (For the definition of urban cores, see sidebar, “How we define cities,” in the executive summary.) The pandemic reversed that trend, making prices rise more quickly in the suburbs. Our model projects that housing demand in urban cores will nevertheless be greater in 2030 than it was in 2019.
Demand and prices for residential real estate have grown more slowly in superstar urban cores than in suburbs and other cities
In every superstar urban core that we studied, the percentage of residential real estate that was vacant grew between 2019 and 2022. The increase varied from 0.8 percentage point in Tokyo to 9.9 percentage points in London (Exhibit 24).
Furthermore, in every city we studied, the vacancy rate for residences rose more in the urban core than in the suburbs. A major reason was increased suburbanization. As we discussed in chapter 1, in most of the cities we studied, population growth in the suburbs relative to that in the urban cores sped up during the pandemic. In London, New York City, and San Francisco, where that phenomenon was especially severe, the difference in vacancy-rate increases between urban cores and suburbs was correspondingly large. In Munich, Tokyo, and Paris, where suburbanization was less dramatic, the vacancy rate increased at roughly the same rate in urban cores and suburbs, driven by pandemic-related deaths and lower immigration from other countries.
Home prices followed the same pattern, rising eight percentage points more slowly in US superstar urban cores than in their suburbs from the end of 2019 to 2022 (Exhibit 25). Prices also rose 13 percentage points more slowly in the US superstar urban cores than in non-superstar urban cores. It was not an exclusively American phenomenon; an OECD report found that the pandemic caused a global acceleration of home price growth outside major city centers.7
Similarly, prices grew more quickly in zip codes farther away from city centers, according to our analysis (Exhibit 26). In fact, many cities experienced what has been called the “doughnut effect”: rising prices in the suburbs and falling prices in city centers.8 In San Francisco, nominal prices in the most strongly affected neighborhoods fell by 12 percent from the end of 2019 to 2022. Residences in San Francisco’s urban core are now worth $750 billion less than they would have been if prices there had risen at the national average rate.
By 2030, projected demand for residential real estate in superstar urban cores is greater than it was in 2019, despite the pandemic
In most superstar urban cores, even though demand for residential real estate has been weak, it will be greater in 2030 than it was in 2019 in a moderate scenario, according to our model (Exhibit 27). Houston, Munich, and Tokyo experience the strongest projected demand growth. The two exceptions are Paris and San Francisco, where population in the urban core is shrinking; demand in those two cities is expected to be lower in 2030 than it was in 2019. Although the effects of the pandemic figure into all our demand projections, they are responsible for significantly lower demand by 2030 in only three cities—London, New York City, and San Francisco.
Although demand is projected to grow in most urban cores, it is projected to grow still more in the suburbs. As a result, excess supply in our model is far greater in the cores than in the suburbs. (Recall that we define excess supply as the percentage of space that is vacant beyond the average from 2014 to 2019—in essence, the part of projected vacancy that is attributable to the pandemic in our model.) The mismatch would lead to strong price corrections, ultimately reducing the difference in vacancy rates between the urban core and the suburbs.
To better understand how our model estimates demand, consider the moderate scenario for London (Exhibit 28). In both the urban core and the suburbs, the population and the size of the average home are expected to grow, adding to total demand. Furthermore, the number of people in an average household is expected to decline, also adding to total demand. But the effects of migration differ: migration out of the urban core is projected to drive down demand there, whereas migration into the suburbs is projected to drive up demand there.
In a severe scenario, pandemic-driven domestic out-migration is higher than in the moderate one. As a result, demand in 2030 is lower in that scenario than in the moderate one for most cities (Exhibit 29). But even in the severe scenario, net demand increases in most cities.
Both scenarios rest on the assumptions that out-migration will continue to be higher than it was from 2015 to 2019 and that the wave of residents who left cities in the past three years will not return. Those assumptions are supported by our survey results, as we discussed in chapter 1. But if they prove incorrect, housing demand in superstar urban cores could be greater than we have estimated.
Also, recall that our model does not consider price elasticity. That is, the projections are for a situation in which prices have not yet adjusted. But ample research suggests that price elasticity in superstar cities is high, so any available floor space will probably be taken up quickly.9 In other words, lower demand is likely to push down prices and rents, and those lower prices and rents would quickly attract new residents and encourage existing ones to buy or rent more space, preventing vacancy from growing.
Unfortunately, the downward pressure on prices and rents is unlikely to make residences in superstar cities—many of which suffer from expensive housing and homelessness—much more affordable. From December 2019 to December 2022, home prices in the United States rose by 40 percent, more than twice as fast as inflation. In US superstar cities’ urban cores, prices grew more slowly, by 25 percent—but still faster than inflation. Housing there will probably become less expensive than it would have been without the pandemic, but home prices will continue to increase and will remain out of reach for many.
Retail space will continue to be challenged
A great deal of retail space in urban cores became vacant during the pandemic, and there is reason to think that vacancy rates will keep rising. Our model projects that demand for retail space in urban cores will be lower in 2030 than it was in 2019 and that some cities will be hit much harder than others.
Retail vacancy has increased and rents have declined, particularly in office-dense locations
Retail real estate has struggled since the start of the pandemic. From 2019 to 2022, vacancy rates in the urban cores we studied increased by an average of 3.3 percentage points (Exhibit 30). The increase was largest in London at 6.2 percentage points, albeit from a relatively low starting point. During the same period, asking rents decreased 5.4 percent, on average, in real terms. The rents that were actually paid may have fallen even more.
Those changes were less acute in the suburbs. From 2019 to 2022, vacancy rates in London’s suburbs increased by just 0.2 percentage point. Similarly, vacancy rates increased by 3.0 percentage points in Manhattan but by only 0.7 percentage point in the surrounding suburbs. Lower foot traffic near stores in urban cores may help explain the variation.
Asking rents for retail space grew the least in office-dense neighborhoods. In Manhattan, for example, the neighborhoods where office space represented the greatest share of all real estate, such as the Financial District and the area around Grand Central Terminal, had some of the smallest increases in asking retail rent from 2019 to 2022 (Exhibit 31).
There are several reasons to think that vacancy rates could keep rising. First, many retailers’ leases have yet to reach their renewal date, so those retailers have not yet had an opportunity to downsize. Second, landlords have made concessions designed to keep stores from closing permanently, such as deferring rent payments and allowing retailers to pay rents calculated as a percentage of in-store sales. Those concessions have probably encouraged retailers to delay closing, despite drops in the profitability of their stores, but it is unlikely that they will delay forever. Third, though spending in brick-and-mortar stores rebounded in 2022, economic uncertainty and waning consumer confidence may slow such spending in the future.
Demand for retail space in the median city we studied is projected to be 9 percent lower in 2030 than in 2019
According to our model, in the urban cores we studied, demand for retail space in 2030 will be as much as 22 percent lower than it was in 2019 in a moderate scenario (Exhibit 32). In a severe scenario in which various pandemic-driven effects are stronger, demand will be up to 31 percent lower. Once again, our model does not consider price elasticity.
Differences in projected demand among cities can be attributed to the relative contribution of the growth drivers that our model used, including population growth, per capita retail spending, and online spending as a share of all retail spending. For example, in London’s urban core, population outflows from 2019 to 2022 were among the largest that we studied, so population growth is not projected to boost demand much (Exhibit 33). And retail spending in London is projected to decrease by 2.2 percent per year through 2030, more slowly than in any other city in our sample. As a result, such spending does not contribute much to demand for retail space in London. What does have a major effect on demand is online spending as a share of all retail spending, which is higher in the United Kingdom than in any other country we studied except China.
Demand for retail space may also vary between urban cores and their suburbs. For example, in Paris, projected demand decreases by 7.6 million square feet in the urban core but increases by 5.6 million square feet in the suburbs (Exhibit 34). Net out-migration from urban cores to suburbs is a strong driver of this variation; the differences in foot traffic and spending discussed in chapter 1 are others. In some superstar cities, as office attendance declined, workers shopped less near the office. However, in Paris, that behavioral shift was muted and did not drive significant differences in demand between the urban core and the suburbs.
