Historically, cities have been disease sinks, and new contagious disease outbreaks have led to large-scale urban flight. There is a colorful description of this phenomenon in The Cholera Years, which documents the impact of the first cholera pandemic in 1832 on New York City urban flight:
Rumors that cholera was moving west and not south from Canada could not stem the growing panic; mass exodus from the city had already begun. A hyperbolic and sarcastic observer remarked later that Sunday had seen ‘fifty thousand stout hearted’ New Yorkers scampering ‘away in steamboats, stages, carts, and wheelbarrows.’
Ultimately, John Snow’s discovery of the water-borne nature of the disease and the subsequent construction of New York City water works ended that particular scourge. But the mobility, crowding, and international connections of cities have left them vulnerable to new pandemics—while new technologies mean that migration is an even more viable escape valve for individuals faced with these new threats.
Documenting Urban Flight
In a new paper with Joshua Coven and Iris Yao, we document the determinants and causes of urban flight in the context of Covid-19. I think the underlying facts are pretty striking and should be remembered as a distinguishing feature of this pandemic. We plot the fraction of resident New Yorkers that leave the city over the course of Spring 2020:
This is really massive flight—over 50% of the resident population in some of the wealthiest neighborhoods in downtown Manhattan leave. Breaking this out by borough, this exodus was pretty large and immediate beginning in early March. As in other studies, this behavioral response preceded formal lockdown orders. About 15% of the total population of Manhattan was gone by mid-April; a fraction which continues declining out until mid-July or so when we last observe people.
We look across major metropolitan areas and the country as a whole. As you might imagine, populations that flee tend to be richer, whiter, and younger. This migration channel, as a consequence, helps to amplify some of the disparities in Covid exposure we discussed last newsletter—the Zooming class gets to flee cities towards other places, while essential workers people are stuck in place in more contagious environments.
While migration is a possible protective mechanism, it carries an important externality—migrants can help seed cases in the rest of the country. We look at the national exodus shown, here in March:
And find that this has an impact on case growth in destination areas. Below, I’m plotting the cases per capita in regions that had high (red) or low (green) influx from New York City. This difference narrows a bit over time, suggesting that early influx brought forward some cases which may have still otherwise happened later. The timing factor is important, as we have seen improved at treatments and masking over the course of the pandemic, resulting in lower mortality for later cases.
To get a better sense at causality, we use an instrument drawn from Facebook connectivity data. There is a good discussion of this data in the context of Covid here by my colleagues Theresa Kuchler, Dominic Russel, and Johannes Stroebel. I’d really recommend using the Facebook data for those of you who are curious—it turns out to be a very nice summary measure of social connections applicable in many contexts (you can access it here).
With migration—it turns out that these urban exodus moves can be instrumented for by Facebook connections (i.e., New York residents disproportionately leave to areas where they have more friends and family), and using this instrumented relationship causally identifies the role of this pandemic-associated migration on destination area case growth.
One of the big implications of this is I think the value of travel restrictions or testing-on-entry policies. While we obviously have a lot of restrictions on international travel, domestic transit remains relatively unimpeded (quarantine order on entry are generally not enforced). The migration channel is a natural behavioral response of individuals facing local cases, but it has a large externality by pushing cases to areas receiving migrants. These spillovers lead to waves of the pandemic washing all over the country, and could be mitigated by screening migrants (for instance, with the latest generation of rapid tests).
What Does Urban Exodus Imply for the Future of Cities?
A lot of these migratory changes represent transitory sheltering in the face of a contagious disease. Right now, as cases spread in rural areas—we are seeing that low density isn’t fully protective, particularly with many rural areas also lacking hospital capacity. So it’s certainly premature to write off cities as “dead,” and I think they will remain vibrant and interesting places in the years to come.
Still, I think it’s quite likely that cities are still going to be reshaped in important ways as a result of this pandemic—particularly the largest “superstar” cities with high rents. These high rents represent, in spatial equilibrium, a premium for accessing strong local labor markets and high urban amenities. But the labor market access has grown dislocated to an extent, as either partial or full-time work-from-home grows in popularity. And while cities have actually used the pandemic to increase some amenities (pedestrianized streets, outdoor dining)—they are clearly worse off in others (no Broadway shows, indoor bars, etc.).
So I would take seriously survey evidence that suggests a sizable fraction of New Yorkers (44% among those who earn more than $100k annually) are at least thinking about leaving the city. If even a small fraction of these folks do end up heading out, the resulting loss in tax revenue will strain public service provision—making it more difficult to continue to cater to those that remain, and potentially producing the same negative spiral that made cities much less exciting places in the 1970s and 1980s.
Cleaning Up #TrashCity
To head this off, I think cities really need to think more about proactively attracting and retaining talent, rather than just assuming that people have no other choices. This means addressing some of the basic amenity factors, resulting from congestion and density, which can sometimes make cities unpleasant places. Here in New York, high on the list is the fact that we just dump trash on the sidewalks, leading to our nickname #TrashCity:
I wrote a piece about fixing this by putting covered containers on streets, so sidewalks are free of trash. It’s not exactly a revolutionary idea to suggest New York should just do what basically all other cities in the world do and move all of this smelly and unsightly trash into bins.
But fixing cities doesn’t necessarily require new groundbreaking ideas. We have plenty of examples around the world of thriving cities, and can borrow from these examples to ensure that cities remain vibrant and healthy places in the future.
Ben Keys and Philip Mulder have a nice paper on the impact of climate change risk on the valuation of real estate properties, with a good NYT writeup here. One complicated feature is the role of depreciation. One other interesting paper on climate adaptation (Desmet, Kopp, Kulp, Nagy, Oppenheimer, Rossi-Hansberg, Strauss) argues that sea level rise takes longer than standard depreciation anyway, so the direct cost to structures is not so high. Under that view, the loss of some Florida beachfront properties might not be too great (we can just stop maintaining them)—but lost agglomeration economies from Manhattan or Miami might be more severe.
There have some interesting (and slightly conflicting) papers in the impact of giving the homeless more housing. An RCT finds 1) large effects of giving homeless people more housing on their housing stays (pretty intuitive); 2) Reductions in some outcomes such as psych ED visits; but 3) minor and imprecise estimated outcomes on total hospital utilization and jail visits. By contrast, an interesting JMP by Elior Cohen uses a judicial variation-like instrument to document much larger effects of housing assistance on subsequent medical utilization and crime. It could be the power of this second analysis (using much broader sample of cases across LA) leads to a better measurement, but it will be helpful to see some additional research here.
An interesting paper here by Glen Weyl and Anthony Zhang on depreciating licenses. They have a nice intuition: while asset ownership is nice for incentivizing investment, it can hurt reallocation if ownership remains sticky and there is a friction to change the asset to new entrants. One way to address this is to make asset ownership fractional—selling, new equity shares in a (say) a telecom license to new bidders. Higher property taxes, I would argue, are another way to increase the cost of carry of an asset in ways that encourage the asset to be held by the highest and best user.