The price of housing just keeps going up and up and up.
In many urban neighborhoods, one can’t help but notice the rise of luxury apartment buildings. To the casual observer, it must be that they are driving up the cost of housing. But is this a confusion of correlation with causation? Naturally, the buildings they replace are more expensive, but what about their impacts on their respective neighborhoods? Do they cause their prices to rise as well?
Supply versus Demand
To the economist, supply is supply—if you build more housing, it must be that prices will fall. But the typical person will respond, “I don’t believe it. I see life getting more expensive each day and it must be from the rich boxing out the poor.” But as I teach my students in Economics 101, if you observe price movements, you cannot infer anything about what’s driving them since prices reflect the intersection of supply and demand. Changes in either one or both can be the cause.
The problem is that in vibrant cities, housing markets are in a perennial wrestling match, where demand is the 300-pound giant on steroids, while supply is the skinny guy in khakis and a cardigan. We can’t see the supply benefits because they keep getting body-slammed by demand. If we really want to know the impact of new housing, we must figure out a way to identify what economists call the “pure supply effect.” In other words, how does one “surgically” extract supply and remove the demand side to look solely at how new construction affects rents?
In the last few years, several economists have directly attacked this problem in some clever ways. And what they find, collectively, is that yes, more housing—even at the upper end—produces improvements in other people’s lives, all else equal. Greater supply means lower prices, plain and simple. But let me stress the “all else equal” part. The statistical methods isolate the supply effect holding constant or controlling for local demand. In this sense, these economists implicitly measure what would happen to prices in the absence of new construction.
All the research discussed below does, in fact, find that new construction is associated with “gentrification” in that this construction is then followed by an increase in neighborhood amenities like restaurants and cafes. But the evidence also suggests the new construction takes place in neighborhoods that are already on the upswing. Developers strike when they see the writing on the wall. “Gentrifiable” neighborhoods draw new construction, leading to more restaurants and a sprucing up of the housing stock, which then attracts more construction, and so on.
This research also suggests that the benefits of new supply are highly localized. If a building appears, the small reduction in rents tends to be within walking distance. If new buildings are relatively rare, it’s that much harder to discern this because of the larger tidal forces of gentrification and urban growth.
Treatment and Control
This crop of research uses so-called event studies. The basic idea is to identify a “treatment group”—that is, a neighborhood gets a new building—and compare it to a “control group”—a neighborhood that does not—to see if there is a difference in outcomes, in this case, apartment rents.
Vital is that treatment is either randomly administrated or independent of the larger demand for housing. This is important because if developers build where housing prices are appreciating, it will seem both to the naked eye and statistically as if the developments are driving the price increases. If price increases draw more development, we can’t determine how supply affects rents properly. This two-way relationship needs to be “cleaved.”
One way to overcome this measurement hurdle is to look for random events that take old buildings offline and put new, larger ones in their place. Dr. Kate Pennington, in her paper, “Does Building New Housing Cause Displacement?: The Supply and Demand Effects of Construction in San Francisco,” uses a very clever way to identify a the supply effect.
Her method is to find buildings destroyed or highly damaged by accidental fire (excluding ones that burned by suspicious circumstances, such as arson). The fire is thus unrelated to housing prices in the neighborhood. Between 2003 and 2017, she found 158 fires, and 47 were so severe they required the construction of new apartment buildings.
Based on this strategy, when a new project is completed, Pennington found that “monthly rents fall by $22.77 – $43.18 relative to trend, roughly 1.2-2.3% for people living within 500 m [1640 feet] of a new project.” In other words, demand may be causing housing prices to rise overall, but the new building is reducing that rise for people who live close by.
New York City
Permits versus Occupancy
Dr. Xiaodi Li, in her paper, “Do New Housing Units in Your Backyard Raise Your Rents?,” studies the case of New York City. She measures two variables related to new construction. First is the year that the building received its initial construction permit. Next is the year when the certificate of occupancy is issued, and the building has been cleared to receive tenants. The idea is to look at the impact of building openings, conditional on, or holding constant, the time since the permit was issued.
The permitting year, presumably, captures the fact that developers start a new project after seeing significant local price appreciation. But a building’s opening time after that is essentially random—due to many factors outside the developer’s control, such as waiting for banks to approve loans, weather, lawsuits, supply chains, architectural and engineering issues, and so on. She finds that within 500 feet (152 m) of the new building, for every 10% increase in the housing stock, rents decreased by 1%.
End of Subsidies
While Li’s paper finds evidence for a net supply benefit, a paper by Divya Singh and Luis Baldomero-Quintana does not. In their paper, “New Residential Investment and Gentrification,” the authors use a similar data set as Li and look at new construction in New York City. Their primary strategy is to identify the supply impact that came from New York’s tax subsidy program changes.
In July 2008, New York altered the law that made new-construction tax subsidies (“421a”) less generous in many parts of the city. Public discussions about the changes had started as early as February 2006, and it took about two years to make its way through the legislative channels. During this time, developers, seeing a more expensive future, rushed to get building permits before the expiration date. This flooding of the market, one would expect, should lead to a net supply effect.
To better isolate the supply from demand, they take a two-step procedure. For each rental building in New York, they “drew” a 150-meter (492 feet) radius around it and then determined if there was a vacant parcel inside the enclosing circle. They then use the presence of a vacant lot or not to predict if there was “rushed construction,” near the building. Then they used this new predicted supply measure to see if the “rushed construction” impacted prices. The idea is that the presence of a vacant lot in a neighborhood is likely to be a random event, unrelated to gentrification, and based on the personal history of the landowners.
However, they find that rental buildings that experienced a 1% increase in the rental stock within 150 meters (due to “rushed construction”) were subject to a rise in rents of 1.8%, on average. This suggests that new construction prompted by the changes in the law pushed faster gentrification than what might have happened otherwise.
In a previous blog post, I discussed statistical work that I performed on supply and rents in New York. While my statistical method is not as cutting-edge or sophisticated—and has a broader geographical scope—it does suggest that supply does matter across New York City.
My strategy was also to use a two-step procedure. First, I predicted the changes in the number of units based on the historical restrictiveness of zoning across community districts, given that zoning restrictiveness is independent of current price changes. I then used my new measure of supply to see how it affected prices. I find that a 1% increase in units is associated with a 0.7 to 1.1% reduction in median rents across community districts.
At an even larger scale, research in Germany by Andreas Mense, in his paper, “The Impact of New Housing Supply on the Distribution of Rents,” predicts housing supply “shocks” based on summer rain patterns. Low rain summers mean more housing supply by the end of the year since construction can continue with fewer interruptions. Since rain patterns are essentially random, they can be used to create a random measure of supply. In this case, he finds, “adding one new housing unit to the stock for every 100 rental housing units offered on the market in a given month reduces rents by 0.4–0.7%.”
The above research looks at new construction, but let’s flip the coin on its head. What happens to neighborhood rents when housing is removed from the market? Dr. Hector Blanco explores this in his paper, “Pecuniary Effects of Public Housing Demolitions: Evidence from Chicago.” The federal HOPE VI program incentivized the demolition of high-rise housing projects in the 1990s and 2000s, 20% of which were in Chicago, offering an interesting laboratory to see how this impacted rents.
The treatment group was neighborhoods in which public housing was demolished. Then, based on a statistical algorithm, Blanco created a “synthetic” control group, using neighborhoods within Chicago that were very similar in terms of racial and ethnic characteristics but did not see public housing demolitions. He looked that the difference in price changes between the two groups and finds “that house prices increased by up to 20 percent over a ten-year period in census tracts near the demolitions.”
Another pair of papers use a similar strategy of comparing “near” buildings—those within a small radius of new construction, to “far” buildings—those beyond the radius border. The key assumption (backed up by evidence) is that the effects of new construction are very localized but that general demand impacts, as well as housing characteristics, are similar in a larger neighborhood area. Thus, researchers can compare rents in a small radius to rents just beyond this to see what would have happened if the new building was not completed, particularly when the new building is a “pioneer, i.e., the first building after a long hiatus in construction.
In the case of Minneapolis, Anthony Damiano and Chris Frenier, in their paper, “Build Baby Build?: Housing Submarkets and the Effects of New Construction on Existing Rents,” compare rents of units within 300 meters (984 feet) of new construction to a comparison group 300 to 800 meters (985-2624 feet) away. On average, they find no effects on rents. But they then look at neighborhoods of different income levels. They find that in high-income neighborhoods rents dropped by 3.2%. In the medium-income neighborhood, there was no impact. But in the low-income neighborhoods, they found that rents rose by 6.6%, on average. This suggests that in the high-income neighborhoods, there was only a pure supply effect since there would be no gentrification. In the low-income neighborhoods in Minneapolis, it seems that new construction promoted more upgrades to the housing stock that drove rents upward.
Across the U.S.
Brian J. Asquith, Evan Mast, and Davin Reed, in their paper, “Local effects of Large New Apartment Buildings in Low-income Areas,” investigate the impacts of new construction across eleven major cities—Atlanta, Austin, Chicago, Denver, Los Angeles, New York, Philadelphia, Portland, San Francisco, Seattle, and Washington DC—and specifically in low-income neighborhoods (below the median for the city) with pioneer buildings that are presumably gentrifying.
Like Damiano and Frenier, they use the near-far approach for treatment and control, respectively, where their “near” buildings are within 250 meters (820 feet) of a newly constructed apartment building, and the control is 251 to 600 meters (821- 1969 feet). They found that new buildings lower rents in nearby buildings by 5 to 7% relative to the trend.
A few of the above papers also investigate the issue of displacement. For San Francisco, Pennington tracks the out-movers from a neighborhood with a new building to see where they went. She obtained the address histories of 1.24 million people who lived in San Francisco between 2002 and 2017. She defines displacement as the process where the out-movers leave their current neighborhoods for one with lower income. The idea is that higher-housing costs due to gentrification pushes them to look for housing in a cheaper location.
She finds that when a new building is randomly built because of a fire, “On average, an additional project reduces displacement risk by 17.14% for people living within 500m.” In short, she found that new construction makes displacement less likely.
Asquith et al. investigate in-migration by tracking individuals’ addresses over time. They find that new construction decreased the average neighborhood of origin income of in-migrants by about 2%. Additionally, the new construction also increased the share of in-migrants who are from very low-income neighborhoods by about three percentage points, suggesting that new buildings allow low-income people to “filter up.”
In short, while higher-income people move into newly-constructed buildings, they do not find that is also generates a wave of high-income displacers. This is consistent with other research, which finds limited displacement in gentrifying neighborhoods.
The Big Picture
Social science research, however, can be a messy affair. Each project uses varying data sets, with different locations, sources of rent information, new construction measures, neighborhood boundaries, and different times. Each also uses different statistical steps or techniques, which depend not only on the nature of the data but also on researchers’ choices. This is why not all research comes to the same conclusion.
For this reason, we must look at the body of scholarship to see general patterns. To the best of my abilities, I have collected a series of studies that well-represent the latest crop of work being done by economists. And when you look at this corpus, the majority concludes that new supply reduces prices compared to what it would have been otherwise.
Pennington and Li found this in San Francisco and New York City, respectively. Blanco saw that the removal of supply raised prices in Chicago. Asquith et al. estimated that supply increases reduce rents in dense low-income neighborhoods in eleven U.S. cities. Barr and Mense find similar results, though at much larger geographic scales.
On the other hand, Damiano and Frenier and Singh and Baldomero-Quintana see a reduction in prices from new housing in higher-income neighborhoods but a positive effect in lower-income ones. In these cases, the speed of gentrification outpaced the supply effect, on average.
Policies for an Affordable Future
It’s important to reiterate that all these studies focus on gentrifying urban neighborhoods. For good or for bad, non-gentrifying, low-income communities tend not to get new construction. But one thing is clear: You can’t help low-income people by restricting supply. Neighborhoods tend to gentrify because they have good transit and are already close to other exciting places and employment centers. Not building will not stop gentrification—though it might slow it down. So how do you allow for improvements in both neighborhood quality and affordability at the same time?
Vacancy and Quantity
Any housing affordability policy must come from a citywide program, given that all neighborhoods are part of an interconnected urban system. But a review of the above literature suggests some ideas.
The first must come based on neighborhood housing vacancy rates. If the new construction reduces vacancies, this is good. Residents at nearly all income brackets will have more choices and with less pressure on landlords to raise rents.
If new construction, however, lowers the vacancy rates, this is an important indicator that supply cannot keep up with demand. Reducing zoning restrictions and providing more construction subsidies can incentivize more construction. All neighborhoods should have average vacancy rates—in good times—between 7 and 10%. Lower than that can mean rising prices and above that can potentially lead to disinvestment or foreclosures.
Another takeaway from these papers is that low-income household affordability generally comes not from new construction for low-income residents, but rather from the filtering process. As a higher-end building comes online, it “frees up” units for those in the lower income brackets. The best way to help the poor is to flood the market with middle-income housing and let the rents in the lowest-income areas fall as a result. The research consensus here suggests that higher-income housing can help the middle and lower classes through this process. We should not fear new housing when it helps to increase neighborhood vacancy rates.
Housing vouchers—subsidies given directly to renters—can be a helpful way to reduce housing expenses for the lowest-income residents. With vouchers, the government pays the landlord the difference between the rent and 30% of a household’s income. But vouchers can only work well if there is sufficient vacancy. In a tight market, if a landlord has many applicants for a vacant unit, she is likely to go with a non-voucher applicant.
Rhetoric versus Reality
The rhetoric on new buildings in gentrifying neighborhoods tends to be spun as us-versus-them–in some form of “They are destroying our neighborhood.” Residents blame “greedy” landlords for jacking up rents. They blame newcomers for displacement and changing the character of the community.
If you took this argument—however species—back in time, great cities like New York would never have become great. They rose to global metropolis status because the real estate market was flexible enough to accommodate all comers. This must remain so.
Yes, we live in different times, where there’s much greater global wealth and mobility, but the fact of the matter remains: new supply helps keep prices in check. The body of evidence presented here shows that it does. We should not fear new construction but encourage it even more.
Asquith, Brian J., Evan Mast, and Davin Reed. (2021). “Local effects of large new apartment buildings in low-income areas.” The Review of Economics and Statistics, 1-46.
Barr, Jason, (2021). “Housing Gotham (Part II): Supply and Prices in the 21st Century,” Skynomics Blog, December 13.
Blanco, Hector. (2022). “Pecuniary Effects of Public Housing Demolitions: Evidence from Chicago.” Regional Science and Urban Economics, 103847.
Damiano, Anthony, and Chris Frenier. (2020) “Build Baby Build? Housing Submarkets and the Effects of New Construction on Existing Rents.” Center for Urban and Regional Affairs Working Paper, University of Minnesota.
Li, Xiaodi. (2022) “Do New Housing Units in Your Backyard Raise your Rents?.” Journal of Economic Geography, 22(6), 1309-1352.
Mense, Andreas. (2020). “The Impact of New Housing Supply on the Distribution of Rents.” Working Paper.
Pennington, Kate. (2021) “Does Building New Housing Cause Displacement?: The Supply and Demand Effects of Construction in San Francisco.” Working Paper.
Sing, Divya, and Luis Baldomero-Quintana. (2022) “New Residential Investment and Gentrification.” Working Paper.
 My summary of each paper is, by necessity, very brief. All the researchers do a lot of additional statistical work to check the robustness of their results and discern possible mechanisms for price changes. My summaries are just the “big ticket” conclusions. Readers of this post are urged to read the papers for more details.
 Note that my literature review here does distinguish between those papers that have already been published and went through the peer review process versus those that did not. All papers were written by Ph.D. economists or as part of their Ph.D. dissertations and are of high-enough quality to merit inclusion in this post.