Summary
- Stanford University researchers quantify Covid-19’s effects on real estate markets within and across US cities.
- They find real estate demand has moved from central business districts towards lower-density suburban zip codes, which they call the ‘Donut Effect’.
- This migration from financial, commercial and business centres has happened within large cities, but less so across cities.
- Despite the Donut Effect, the authors expect real estate markets in more densely populated metros to outperform long term.
Introduction
Until now, digitalisation has coincided with substantial growth in ‘superstar’ cities. However, the pandemic has shown that working from home (WFH) is feasible and, for many, has gone better than expected. And so, will the unexpected transition to remote work lead to a permanent reallocation of employment opportunities across regions?
To answer this question, a new NBER working paper examines migration patterns and housing and rental prices within and across US cities during the pandemic. Specifically, the authors determine whether Covid-19 has moved individuals away from the cities in which they work, thereby making it more likely that the full-time WFH model will persist. In doing so, they look at Zillow and US Postal Service (USPS) data to find:
- Housing and rental prices in central business districts (CBDs) of the 12 most populated US metros have fallen significantly relative to surrounding suburbs. This has coincided with large population and business outflows.
- This hollowing of city centres is primarily a large city phenomenon, with less densely populated cities experiencing only small price and migration flow effects.
- The reallocation of economic activity has predominantly been within metros (centres to suburbs), not between metros.
Data and Methodology
To track changing migration patterns during the pandemic, the authors use the USPS, which provides National Change-Of-Address (NCOA) data. The dataset contains zip code-month level inflow and outflow data for the universe of US zip codes over the last four years.
Zillow’s Observed Rental Index (ZORI) and Home Value Index (ZHVI) are used to measure rental and house prices at zip code and metropolitan statistical area (MSA) levels. The ZHVI covers the full universe of US metro areas, while the ZORI is only provided for the 100 largest.
To measure how residents changed their housing demand in response to the pandemic, the authors require an idea of the share of jobs that can be done from home within each zip code (WFH exposure). They obtain the job industry distribution for residents across US zip codes (US Census Bureau), which they overlay with research on the types of jobs that can be done at home.
Lastly, the authors map zip codes to their corresponding metro area’s central business district (CBD). They define the area of a CBD to be all zip codes with centroids (geometric centres) within two kilometres of the CBD coordinates. Finally, they calculate the population densities for each zip code and CBD.
Results
First, the authors document changes in rental and housing prices among the 12 most populated US metros (Chart 1). Rental prices for all zip codes in these 12 metros were increasing prior to the pandemic, but diverged significantly after February 2020. Running zip-code-level regressions, the authors find that areas with greater population densities, greater WFH shares and less distance to the CBD had larger drop-offs in rent. This indicates that the rise of WFH has made dense areas near city centres less attractive relative to the suburbs.
Migration data follows a similar pattern (Chart 2). The outflows from CBDs are especially striking. Monthly population outflows increased to almost 2% of the pre-Covid population for the CBDs of the 12 largest US metros. Overall, CBDs of the top 12 US cities have seen net population and business outflows cumulating to about 15% of their pre-pandemic levels. Again, the largest outflows are from areas that are closest to CBDs, have the highest population densities and have a high WFH exposure.
Next, the authors look at migration flows and prices across smaller cities. They find house prices in less densely populated metros appreciated at a similar rate to low-density zip codes in the top 12 largest metros, and there was little variation across zip codes. Similarly, the outflows from smaller metros were far less significant compared with the largest 12 MSAs (Chart 3). These smaller MSAs (13-365) typically had lower WHF shares, indicating that most WFH jobs are located in the largest cities (tech agglomeration).
Lastly, the authors examine the relationship between metro-level population density and both migration patterns and home price changes post-Covid. They find two key results. First, the reallocation of economic activity is much stronger within than between metros. This reaffirms earlier results showing individuals were more likely to remain in the same metro, rather than move between metro areas.
Second, a positive relationship exists between population density and home price changes at the metro level. That is, larger metros on average experienced larger price increases than smaller metros, despite the CBDs in these larger metros experiencing price declines during the pandemic. The relationship can be explained by higher property prices in larger cities, meaning the mass migration to the suburbs has increased overall spending by more than in smaller metros.
Real estate markets in large metros should continue to outperform in the long run. This is because little evidence exists of between-metro migration during the pandemic, and there is a positive relationship between population density and home price changes. In essence, the pandemic has made the suburbs of large cities more unaffordable, without doing much to make city centres more affordable.
Bottom Line
‘Zoom Towns’ are smaller regional communities that sprung up in response to shifting work patterns during the pandemic. While such remote work hubs sound enticing, the broader long-term trend in the data shows most people have only relocated to the suburbs of the cities in which they work.
This has two implications. First, individuals have remained relatively close to workplaces, so we will likely see at least a partial return to the office post-pandemic. Second, a lack of substantial between-city migration means pre-Covid real estate trends could remain intact, with more densely populated locations predicted to do well.
Citation
Ramani, A., Bloom, N., (2021), The Donut Effect of COVID-19 on Cities, NBER, Working Paper (28876), https://www.nber.org/system/files/working_papers/w28876/w28876.pdf
Authors of Working Paper
Arjun Ramani is a graduate student in computer science at Stanford University and is currently on leave writing for The Economist newspaper in London. He has written cover stories for The Economist on data-driven policy and the venture capital industry.
Nick Bloom is a Professor in the department of economics and Professor, by courtesy, at the Graduate School of Business. He is also the Co-Director of the Productivity, Innovation and Entrepreneurship program at the National Bureau of Economic Research (NBER), and a fellow of the Centre for Economic Performance, and the Stanford Institute for Economic Policy Research.