Supply skepticism lite

A recent “supply skeptic” paper by various academics has gotten a lot of attention in housing-related social media. The somewhat sensationalistic title is: “Inequality, not regulation, drives America’s housing affordability crisis.” But unlike most random rants from “not in my back yard” activists, the authors do not deny that the law of supply and demand matters: instead, they argue that demand is important and supply is not.

The paper’s arguments seem to fall into four general categories:

1. “Regulation isn’t related to housing supply”– As a matter of common sense, it seems obvious that if laws that allow almost no new housing will be amended to allow some housing, more housing will be built than if the law was unreformed. But the authors write that “evidence suggests that upzoning is not meaningfully related to improvements in regional affordability.”  (page 7) They cite a 2023 study by Yonah Freemark.  But Freemark’s study does not support this.  Freemark’s introduction to his article states  “Downzoning policies are largely associated with reduced construction and less affordability. … Early data suggest that upzonings generate positive effects on regional construction and affordability, but more research is needed. (emphasis mine)”  This might be a more ambiguous than YIMBYs would like, but it certainly is not the flat rejection of upzoning that the authors seem to say it is.

Second, they claim that  because “between 2000 and 2020, no major urban area experienced growth in household formation that exceeded its growth to supply, suggesting that there were not significant differences in supply growth between regulated and unregulated regions” (emphasis mine) (pages 7-8). Census data show that about 34,000 of San Francisco’s housing units were built after 2000, and that about 150,000 of Austin’s were. So if the authors really believe that there is no difference between high growth and low growth cities, they must have been very careless indeed.

I think the first half of the sentence (about “growth in household formation”) shows what the authors really mean: the number of households didn’t grow as much as San Francisco as in Austin, therefore supply must have been consistent with demand rather than lagging behind demand. This would make sense if demand for San Francisco was limited to people who stayed there. But as a matter of common sense, demand for San Francisco should include people priced out of San Francisco. Thus, the number of households in San Francisco is not an appropriate measure of demand.

2. “Housing costs will never increase fast enough for rent to become affordable”- Through complicated calculations, they claim that in San Francisco, it would take as many as 124 years to meet their definition of affordability.  (page 13)

But this is based on two questionable assumptions: first, they assume that there can at m ost be a 1.5% annual growth rate in housing.  But they admit that Austin grew by twice this much- despite the fact that Austin still has conventional zoning.  If Austin had no zoning, perhaps its housing supply would grow by even more. 

More important, the authors’ 1.5% assumption is based on growth rates between 2000 and 2020.  But because of the 2008 recession, this was not exactly a high-growth period. About 34 million housing units were built between 2000 and 2020, about the same as between 1960 and 1980- despite the fact that the U.S. population increased by over 50 percent (from 179 million to 281 million) between 1960 and 2000.   To put it another way, in the 2000-2020 period, Americans built one housing unit for every eight people, while in the 1960-1980 period, Americans built one for every five people.  So if the authors really wanted to imagine a deregulation-fueled building boom, they should have used data from the most growth-friendly metros in the 1960s.  (Nationwide, there were 58 million housing units in 1960, and so if 34 million units were built in 1960-1980, this means that the nationwide housing supply increased by almost 60 percent, or about 3 percent per year.  Presumably, housing supply in some places increased more rapidly).

Second, their idea of an “affordable rent” is very low indeed.  They define “affordability” as making the median one-bedroom apartment affordable to someone earning between $30,000-40,000 a year (well below median U.S. salaries) – which means bringing the average rent in San Francisco down to $971 (p. 13),, cutting rents by more than half. What’s wrong with this? It seems to me that as a matter of principle, any rent decline is a good one: just as reducing cancer or car crashes by 5 percent is worth pursuing, so is reducing rents.

By contrast, when describing their preferred policies, the authors show no interest in quantitative data or weighing trade-offs. The authors vaguely conclude that cities need to “directly increase access to affordable housing among low-to-moderate income households.” (p. 30). But do they bother to ask how many years of housing construction it would take to make half of San Francisco’s apartments this affordable, or how many billions of tax increases this would take? Of course not.

3. “Our arguments above were actually too optimistic”  The authors make several arguments as to why they think the scenario discussed above may be too optimistic rather than too pessimistic.  First, they argue that “Upzoning and deregulation likely have positive impact on land values and increase wealth inequality between landowners and renters (p. 14)” But Freemark writes: “Keeping land-use policies as they are today—or, more problematically, implementing downzonings—could reinforce inequalities. Research on downzoning shows that it, too, can increase housing costs by limiting construction in the most desirable neighborhoods.” So it does not seem to me that the authors have read Freemark correctly.

Moreover, if retaining the regulatory status quo kept land costs down, the regions with lots of new housing would have exploding land costs while those with very little new housing (such as Long Island) would have stable land costs.  In fact, land costs seem to go up even in regions that basically shut off new housing. (I wrote about this issue here).

Second, they write that new supply “ought to significantly increase the short-term cost of construction labor … influencing the cost of new units.” (p. 14).  Do the authors really think that construction labor is cheaper in New York or San Francisco than in fast-growing, pro-construction Sunbelt cities like Houston? If so, they are wrong. (Having said that, I don’t know whether the pay gap between these cities has grown or shrunk in recent decades).

Third, they argue that “developers often delay construction in anticipation of higher future returns” (p. 15) and that “there is no reason to believe that developers would ever knowingly build enough housing to drive investment returns below market rates.” (id.) In other words, developers somehow have enough foresight to never build too much housing.  If this was true, development at an Austin-like level can never happen, because developers are so wise that they know not to build when rents are coming down.   But of course, such development IS happening in Sunbelt cities and happened even more rapidly in the mid-20^th c.  So I am not quite sure I understand the authors’ argument.

4. “Its all about demand.”  The authors heavily rely on one statistic: growth rates in mean income.  They point out that San Francisco’s rents have risen at about the same rate as mean income, and therefore conclude that rising incomes cause rising rents (pages 25-26).  (They also look at Houston and Cleveland, but not any other cities).   

Leaving aside this article’s miniscule sample size, a few things seem questionable about this part of the article.  Their definition of mean income includes “non-wage and salary employment income (e.g. employer contributions to health insurance or 401ks).”  But since no one is going to use their health insurance premiums or retirement accounts to pay rent, this means that their definition doesn’t seem all that relevant to rent growth.  And because health insurance premiums were lower in 1980 than they are today, this calculation method seems designed to exaggerate income growth.   Furthermore, if the authors wanted to compare income and rent among renters, they should have looked at mean income for renters rather than mean income for all human beings. (In fairness, I do not know if data exists on this point).

Furthermore, their definition of median rent seems unrelated to real-world rent growth: they define rent as Census-calculated median rent, but Census calculations seem unrelated to market data.  For example, the Census says that the median gross rent for one-bedroom apartments is a little above $2400- but out of the 1900 apartments available on Zillow, 1570 rent for more than this.  (Having said that, if you are trying to compare rent growth over time, maybe the Census data is the best you have).

In sum, there are many things in this article that seem questionable or even downright wrong. Having said that, there is a grain of truth to it: demand goes up and down a lot more rapidly than supply (as I learned in the early 2020s, when rents in Manhattan bounced up and down rapidly). On the other hand, supply could be a lot more flexible than it is today.

*This suggestion seems to me to be simply weird. Lower marriage/birth rates and longer lifespans mean that MORE people will be living alone or with just one or two other people- in other words, households should be smaller. And in turn, smaller households mean demand for MORE housing units, not fewer.