Max Holleran’s book, Richard Schragger’s law review article, and randos on Twitter all find pessimistic views on housing supply from a paper by Xiaodi Li. But the paper is asking a narrow question and yielding an optimistic answer. This post tries to provide some context.
EDITED 3/3: I’ve edited the post to take into account pushback from the authors I’ve criticized. Edits are in boldface.
The touchstone of YIMBYism is the sensible idea that housing markets follow the normal patterns of supply and demand.
It’s true. But it’s a deuce to measure, because housing markets don’t have sharp boundaries – they bleed across distance, tenure, and unit type. Suppose 200 new one-bedroom apartments open up in Bushwick. Do those mostly steal business from similar buildings in their immediate neighborhood? Or do they compete with all types of housing throughout the tri-state area? Or something in between?
Further complicating the story, new investment in a neighborhood can have “amenity” effects (positive or negative). These don’t work like supply and demand, but in any specific case it’s hard to distinguish the amenity effect from the supply/demand effect.
Neighborhood effects
A few years ago, most economists and urbanists (myself included) believed
- At the metro-area level, a broad increase in supply will lower rents.
- At the neighborhood level, amenity effects dominate.
That is, we thought the [edited 3/20] West Palm Bushwick apartments had a mostly-regional supply effect but an entirely local amenity effect. The evidence for this included many gentrification anecdotes: new “luxury” apartment buildings were accompanied by rising rents.
Enter Xiaodi (and friends)
Around 2017, a few economists started testing these beliefs. Could there be local supply effects after all? And do new buildings accelerate or decelerate gentrification? To my surprise, Li and others found that yes, there do appear to be hyperlocal effects from supply and demand.
In Li’s case, identification is based on timing: tall New York buildings take several years to build, and the unpredictability of the end date allows us to treat the year when new condos become available as essentially random (whereas the application date is not random). Li finds that new housing lowers rent within a 500-foot radius, but doesn’t have a statistically detectable in the 500 to 1,000 foot “donut” beyond that.
What Li’s paper doesn’t ask or attempt to answer is a different, broader question, what is the citywide effect of new supply on rent?
Cards on the table, I remain skeptical that Li is interpreting her core finding correctly (she can’t fully rule out disamenities). But Holleran and Schragger aren’t expressing skepticism – they’re misinterpreting her answer to one question as the answer to a very different question.
Clearing up a little misunderstanding
I expect that randos on Twitter (don’t harass them) are going to misapply Li’s work, assuming that the hyperlocal effect – a 10 percent increase in housing supply within 500′ decrease rent by 1 percent – represents the entire supply effect. But when housing scholars are taking it out of context, we need a reminder.
Law professor Richard Schragger cites Li’s paper in footnote 177 here, using it as evidence that new housing supply has “very limited” effects on “overall rents” in “specific locations”:
Some studies indicate a decrease in overall rents from increased market-rate housing, [176] but there are others that indicate the opposite or very limited effects overall. [177]
Schragger, The Perils of Land Use Deregulation, U. Pa. L. Rev. (p. 164)
Sociologist Max Holleran misunderstands it in a slightly different way:
One New York City study showed that every 10 percent increase in market-rate housing in a given neighborhood would result in a 1-percent reduction in rental prices: a supply effect but notone that gives much optimism to public policy officials tasked with solving the affordability crisis.
Holleran, Yes to the City (p. 13)
Both authors appear to clearly think that Li’s estimate is evidence about the overall effectiveness of housing supply in lowering rent. But that’s simply not the question she’s asking; her paper offers no evidence one way or the other on what the market-wide rent effect of a market-wide 10% increase in housing supply would be. Both authors identify Li’s study as a localized one, but then interpret her findings pessimistically. That’s the reverse: this study (along with those of Mast, Pennington, and others) made economists more optimistic about supply effects.
Li’s study doesn’t tell us what would happen if new buildings were simultaneously completed every 1,000 ft through New York City. It instead asks what happens when one – literally one – building is completed.
When Schragger returns to the question Li is actually answering
a few paragraphs later (“Places with a relatively low cost of living may lose that attribute once enough wealthier people move in”), he doesn’t cite her work.
The paper you’re looking for
What papers should they have cited? My go-to estimate is
Albouy, Ehrlich, and Liu‘s: a 3 percent increase in housing stock lowers rent by 2 percent. Older estimates were often in the 1:1 range, which is more optimistic about supply. I dug into this in a recent blog post here, highlighting that broad affordability and unit-specific affordability usually come from totally different channels:
I confronted the hard truth that supply changes need to be very large to make a real dent in prices. If, as Albouy, Ehrlich, and Liu estimate, it takes a 3 percent increase in the housing stock to bring prices down 2 percent, then a major metropolitan area needs a massive increase in housing to make a real dent in rent.
Can we get there faster with composition effects? Let’s do a quick back of the envelope. First, assume population and housing stock would grow 10% each at baseline, with no resulting change in price.
–> Supply only approach: we add 40% to the housing stock without changing the mix of housing types. Result: 20% affordability gains.–> Composition-only approach: we add 10% to the housing stock, but with the average price just 50% as high as the norm. Result: 5% affordability gains from lowering average price.
–> Mixed approach: we add 25% to the housing stock, with the average prices 75% as high as the norm. Result: 15% affordability gain (10% from supply, 5% from lowering average price).
Salim Furth, Is affordability just, “You get what you pay for”? Market Urbanism
How realistic are any of these scenarios? I’m not sure. But my takeaway is that supply remains the primary avenue for broad-based affordability gains. But the “you get what you pay for” and “only pay for what you want” channels are far more important for the affordability of a particular new housing unit.
There are questions worth asking about the impact of broad-based housing supply, and unresolved questions in the hyperlocal housing supply literature, but they’re different questions with presumably different answers.
marfis75 on flickr (CC-BY-SA)
