fbpx
What Happens to Housing Prices If People Leave a City? What Happens to Housing Prices If People Leave a City?
Four years ago I moved to Boston for my first job, fresh-faced and ready to experience the city. I quickly followed the millennial trajectory... What Happens to Housing Prices If People Leave a City?

Four years ago I moved to Boston for my first job, fresh-faced and ready to experience the city. I quickly followed the millennial trajectory of paying too much for a too-small apartment in the heart of the city. I traded off space and conveniences for access.

I could walk 10 minutes to work, 15 minutes to Fenway Park, and 20 minutes to the Boston Public Garden. I was surrounded by energy. Across the US, cities grew because people, entertainment, and jobs coalesced in these areas to create that energy.

Then, COVID happened. Many of the things that made city life special for me, and others, evaporated overnight. What remains is the price and (lack of) space without many of the benefits that a city can provide. Everything that made a city burst with energy also now makes it unsafe.

This got me thinking — What happens to housing if people leave the city? What happens if, according to a Harris Poll survey, the 40% of Urbanites that are considering a move to the suburbs actually leave? What happens if a fraction of them do?


One simplistic way to think about housing prices is supply and demand. If there are more people for fewer houses, the prices should be higher and vice versa.

Source: US Census Bureau

Further, people can move to a city much faster than places for them to live can be built; this is a phenomenon that Boston has been experiencing.

Between 2010 and 2019, the rate of population change in Suffolk County, which is predominantly Boston and its immediate suburbs, was nearly double the growth in housing units. Consequently, in 2017, the rental vacancy rate in Boston was ~3.4% compared to a nationwide value of ~7.3%.

Between 2010 and 2019, the number of people in Boston grew twice as fast as the number of housing units.

We can study this supply-demand principle with data by analyzing the relationship between a city’s population to housing ratio (the number of people per housing unit; a proxy for supply-demand imbalance) and the price per square foot of homes for sale. Intuitively, there should be an increase in value as the population:housing ratio increases (understanding that many other factors impact housing prices).

As an example, look at the point marked as Boston near the top of the following scatter plot. The City of Boston has a population of ~692,000, an estimated 288,716 housing units, and a median listing price per square foot of $1,009; the population to housing units ratio (population/housing units) is ~2.399.

Each dot on the scatter plot below is the population:housing (p:h) ratio on the x-axis and the median listing price per square foot on the y-axis for cities with more than 150,000 residents in the US (see bottom for data sources).

Supply-Demand plotted on the x-axis and Price on the y-axis.

As expected, there’s a moderately positive relationship but with significant variability. To further refine and study the relationship, we’ll separate the values that fall above their predicted value in blue and below the predicted value in red.

This above-below segmentation is simply a clustering of like values; we assume that the cities that fall above the line all have something unique in addition to the population:housing ratio that drives up prices. For example, some are coastal towns like Miami or San Diego where people value the beach access. Others are cities like Seattle or New York that attract higher prices by nature of being a major metro city.

Comparatively, below the line are places that have scaled down “city-like” environments (Providence, RI and Portland, OR) or population numbers that may be inflated due to residents spending time elsewhere for part of the year (Phoenix, AZ and Tampa, FL).

Considering only those above the line, we see a tighter distribution and relationship between the p:h ratio and price per square foot. The linear regression line predicts that for every change of 1 unit in the p:h ratio (e.g., from 2 people per housing unit to 3), the median price per square foot increases by ~$165.

This establishes that a moderately strong relationship appears to exist between the p:h ratio and value per square foot among this subset of expensive cities.


So, what does this mean for somewhere like Boston? With the number of housing units assumed to be stable, and population decreasing, Boston’s p:h ratio would decrease. We can measure the effect of a declining p:h ratio on Boston’s housing prices by comparing it to its closest alternatives today and its potential future peer group.

If we look at Boston’s current “peer group” in blue (example cities: New York NY, Huntington Beach CA), the average price per square foot (ppsf) is $813; Boston has a 24% premium over their value today at ~$1,009. If Boston were to drop to its next “peer group” in Orange (example cities: Jersey City NJ, Austin TX, Arlington VA), the average ppsf is $712.

Boston’s current (blue) and potential (orange) peer groups along with their avg. ppsf.

If we assume that Boston commands the same 24% premium, that would equate to a price per square foot of ~$883, a drop of ~14.4%. To reach this peer group would require a drop in population of ~11%. In other words, prices would be expected to drop faster than the population decreases.


To answer the question posed at the beginning of the article, there’s a plausible situation where population decline in expensive cities resets the supply-demand equation and puts more power into renters/buyers hands.

Conversely, the deterioration of city prices coincides with the increase in suburban prices. As traffic worsened and cities became safer, suburban living lost some of its appeal. Now, we’re already seeing the reversal of that trend as density is scorned and separation is valued.

The implications of a decline in prices are most prominent in the risks to banks, investors, and recent home buyers. Rental investors and mortgage operators were making calculations based on increasing home values and rents; this may be unlikely in the near-term.

The short-term outlook for cities may be excess units hitting the market as people react to the uncertainty of the world, which would temporarily stagnate prices. Following this could be a period of less housing mobility and units changing hands as the first wave of uncertainty passes and people are comfortable with their new locations. Finally, a new long-term equilibrium will be reached, but we’re too far away to predict with any confidence what that may look like.


What do you think will happen to prices? To the populations of major cities? What will the long-term impacts be? Feel free to reach out to me by email at jordan@jordanbean.com or to connect with me on LinkedIn.


Sources and Data Notes:

  • Housing price per square foot data was sourced from Zillow at the city level.
  • Housing units data was estimated using 2018 American Community Survey (ACS) data sourced through the get_acs R package. Data was acquired at the Zip Code level (with a margin of error) and rolled up to the city level based on a Zip-to-City lookup table. City naming conventions sometimes differed from Zillow naming conventions. Best efforts were made to match the data between the two sources to the extent possible for cities with sufficient population.
  • Population data was sourced from the US Census Bureau.

Originally posted here.

Jordan Bean

Jordan Bean

Jordan is an analytics professional with an interest in data storytelling, visualization, and simplifying complex topics into interpretable messages. He's currently pursuing a Masters degree in Business Analytics at Wake Forest University while working in analytics for Liberty Mutual Insurance. Prior to that, he worked in consulting for Private Equity firms on strategy and buyouts. Feel free to connect at https://www.linkedin.com/in/jordanbean/ to talk through any thoughts on articles or breaking into the data field.

1