Does increased housing supply improve affordability?
By Joe Cortright
Why a recent Fed study tells us very little about supply and affordability
The takeaway: A recent Federal Reserve study which seems to show that building more housing won’t improve affordability has little relevance to supply-constrained cities; among other things, it unrealistically assumes that any increase in housing supply will be exactly offset by an increase in population.
A couple of weeks back, a Federal Reserve Bank research paper got prominent play in the media, mostly because it apparently reached some surprisingly counter-intuitive conclusions about the way housing markets work. An article in Forbes by columnist Erik Sherman trumpeted “Additional building won’t make city housing more affordable, say Fed Study.” (To his credit, Sherman quickly walked back the apparent claim made in his article’s headline. In a revision to the article, he added a series of six caveats to the study.)
The study seems like ready-made evidence for the housing supply skeptics who maintain that building more market rate housing will do little or nothing to solve our affordability problems.
The Federal Reserve research paper, “Can more housing supply solve the affordability crisis? Evidence from a neighborhood choice model,” was written by economists Elliott Anenberg and Edward Kung. In the paper they construct a model of metro housing markets and then simulate the effect of a relatively small increase in housing units in the most expensive neighborhoods on the effect of housing prices in that neighborhood. Their model suggests that the increased number of units has only a small negative effect on prices.
Not surprisingly, we expect that most readers aren’t going to download and wade through the working paper. There’s a lot of highly technical detail that will be opaque to non-economists. But we’ve taken a closer look at City Observatory, and so too, has the Brookings Institution’s resident housing economist, Jenny Schuetz, who shared her thoughts on twitter.
Upon closer inspection, the article actually says a great deal less that either its title or it press reports imply.
Like all models, this one makes a number of simplifying assumptions about the way markets work. From our standpoint there are three aspects of the model that should give us pause in leaning too heavily on this paper for insights about policy.
First, and perhaps most importantly, the model is constructed based on the assumption that increases in supply are exactly balanced (or offset) by increases in population.
. . . because our model assumes a zero vacancy rate, increasing the number of housing units will increase the population in the city . . . (page 13)
That is, if new housing units are built, population increases by exactly the same amount. In effect, this assumes away the question of interest: does increasing the amount of housing, relative to population, have an effect on rents? From a policy standpoint, we’re looking to increase housing relative to population, and explicitly to increase the vacancy rate. That’s the canonical way in which more supply translates into lower rents. But the way the Anenberg/Kung model is constructed, it precludes even considering this channel. As a result, as Jenny Schuetz tweeted:
“The simulated cities don’t look much like supply-constrained cities in real world.’
Second, the paper assumes that housing markets are in equilibrium, both before and after the construction of new housing. But as we’ve frequently discussed at City Observatory, housing markets are often in a dis-equilibrium state: there’s a temporal disconnect between demand and supply; demand can surge ahead of supply (as it has it recent years in many metro markets), and unless and until supply catches up, rents tend to surge. When we observe rising rents in cities like San Francisco, Seattle or Washington, its because there’s a dis-equilibrium. And in fact, it is this surge in rents that typically prompts the additional supply, which ultimately brings rents back down. Assuming that rents are in equilibrium is essentially assuming away the underlying economic processes behind the affordability problem and its solution.
Third, this model doesn’t really look at metropolitan level supply-demand balance. As Jenny Schuetz notes, the paper is really focused on the neighborhood level effects of small increments to the housing stock: does building more housing in one neighborhood have much effect on rents in that neighborhood. The model doesn’t test whether large, metro-wide increases in housing would have an effect on rent levels.
In sum, what the Anenberg/Kung paper tells us is that small changes to the housing supply in high income neighborhoods that are exactly balanced by an increase in population will produce only small reductions in average neighborhood rents; and this finding holds only for neighborhoods where supply and demand were in rough equilibrium before the new housing was added. The paper has little to say about the effect of significant changes in housing supply at the metropolitan level , especially in situations where supply increases relative to population and increases vacancy rates, and in markets where there is an apparent disequilibrium (i.e. where growth in demand has outpaced the growth in supply). For policy makers, there’s no evidence here that should lead you away from steps to aggressively expand housing supply, if you’re looking to deal with affordability problems.
More from Jenny Scheutz, via Twitter
For those who don’t have direct access to Twitter, we’ve reproduced Jenny Schuetz’s tweets on the Anenberg & Kung paper here.
…if you read the whole paper carefully, it’s less clear that the results apply to current policy debates, for two reasons.
First, A&K simulate adding new housing to wealthy n[eighbor]hoods and ask whether rent in those n[eighbor]hoods drops. Answer: not very much. But more policy relevant question is: if wealthy n[eighbor]hoods built more h[ousin}g, would rents rise less in nearby low-mod inc[ome] n[eighbor]hoods?
For example, if DC’s Ward 3 built more houses, would there be less pressure to build luxury condos in CoHi [Colonial Heights], Shaw, Petworth – and thus less gentrification & displacement of lower-income residents in those n[eighbor]hoods?
(technically, A&K are estimating own-neighborhood rent elasticity in high-income areas, not cross-n[eighbor]hood rent elasticity b[etween] high and low-income areas. @Noahpinion [Bloomberg’s Noah Smith] had great thread explaining how this works.)
Second limitation of drawing policy inferences from this paper: it’s a simulation built on a formal model, and makes *lots* of simplifying assumptions on key dimensions to make math feasible.
For instance: pop[ulation] growth exactly equals new housing supply, household formation and size are independent of rents, h[ousin]g markets start in equilibrium, quality of new h[ousing]g relative to older h[ousing]g isn’t discussed.
The simulated cities don’t look much like supply-constrained cities in real world — therefore we should be very cautious applying their predicted outcomes to policy changes that could happen in real life. (that’s true for all theoretical models, not just this paper.)
Bottom line: policy advocates who want to cite academic papers to support their positions should *read* the whole paper first. And academics should be careful over-claiming results & policy implications in title & intro, [because] that’s probably as far as most people will read.