It's never overbuilding, and that won't change soon.
It's never overbuilding, and that won't change soon.
The convergence back to a functional housing supply is a tricky situation. Supply obstructions create rents and prices that include what economists refer to as “rents” - unearned income. Those economic rents are ascribed to land value, which translates to higher costs for builders and developers. So, the inflation of obstruction gets spread out into the marketplace in complex ways. (I have a long-form paper on my to-do list about how even high construction costs are often downstream of obstructed development.)
This poses a problem for convergence, because when the disequilibrium becomes the norm, the return to a functioning equilibrium looks like it is disequilibrium. This has metastasized from a regional problem to a national problem, so that now, rents are elevated in most places, and at the same time, observers and insiders frequently voice concerns about the large number of units under construction, rising inventory of units for sale or rent in various markets, rent deflation, etc.
Some are worried that a glut of supply will create a negative wealth shock or a recession. Others claim that developers will never increase supply enough to lower rents.
A lack of supply explains essentially all of the long-term rise in housing costs and supply explains none of the cyclical changes. Here, I’m going to run through a summary of some data on vacancies, etc., to lay out what I’ve found, to hopefully create some clarity about what you should worry about versus what you should not worry about.
Housing Construction and Vacancies
One thing that is common in housing analysis is to show production in absolute numbers. Normally, this is obviously inappropriate because all measures tend to rise over time with population growth, inflation, etc. Figure 1 is a chart from one of my Mercatus papers where I compare housing vacancies over time to housing starts (annual) over time. Vacancies increased over time as the country grew. Housing starts did not.
I think an additional takeaway from that chart is that if housing production ever had an important cyclical effect on supply and demand dynamics (and it arguably did not) it certainly does not today. The simple arithmetic is clear on this. We cannot build too many homes.
This does create a NIMBY (not-in-my-back-yard) vs. YIMBY (yes-in-my-back-yard) signal. A NIMBY might see this data and react by saying we don’t need any more homes. There are 16 million sitting empty. A YIMBY will see this and realize how low production has become in scale.
There are two types of vacancies - existing homes and new homes that haven’t been sold yet. And, within existing homes, there are various kinds of vacancies - waiting to be sold or rented, under renovations, seasonal use, etc. Most of the vacancies aren’t waiting to be sold or rented. But even units being sold or rented have increased from roughly equal to the number of homes constructed annually in the 1970s to several times more than the number of homes constructed annually today. Changes in demand (both for tenancy and for ownership) are much more important cyclically than changes in supply.
New Home Sales and Homes for Sale
The top panel of Figure 2 shows new single-family homes for sale and single-family homes started (annual rate) over time. This is in the usual presentation that makes it look like the 2000s boom was a massive supply event.
The narrative this appears to tell is that builders become overly optimistic over the course of an expansion until finally construction becomes unsustainable. Sales crash, but units for sale continue to pile up for a bit because of that optimism and the naturally cyclical nature created by construction schedules, etc. And 2005 was the whopper of all cycles.
There is some truth, obviously, to that story. One problem with it is causality. The American business cycle has been driven by nominal (monetary) shocks more than by real shocks (oversupply). There are points where sales and inventory shifted without a recession - the 1960s, the mid-90s, and today. Where the contraction and the overhang of inventory has been related to recessions, it has been due to deep cyclical changes in demand - mostly due to frictions for home buyers.
In the bottom panel of Figure 2, I have recast the measures in per capita terms, and I have added a measure for total housing starts. This accounts for other kinds of new housing, besides new single-family homes built for sale. In the bottom panel, you can see that there has been a long-term decline in new homes per capita. The rise in single-family homes sold and for sale was not particularly high in the 2000s on a per capita basis and to the extent that it was, it was entirely due to a shift of market share from other types of housing (which have become basically illegal in a way that becomes more infuriating the deeper one digs into the topic).
Single-family sales peaked in July 2005 and by August 2008 it had reversed to a record low (on a per capita basis) by August 2008. This was a king-sized example of blaming a demand shock on oversupply. To this day, even new papers that cast doubt on the conventional bubble narrative tend to hold on to the optimistic builder/oversupply narrative. Until we can rid the zeitgeist of that myth, policymakers will continue to be tempted to throw grenades into the business cycle on false pretenses.
Currently, single-family homes for sale amount to about 0.3% of the existing housing stock and homes under construction amount to about 1%. Supply does not have the scale to create cyclical shifts.
Construction and Vacancies Across Regions
In Figure 3, I will walk through vacancy rates (of units for sale or rent) and permitting rates (which represents changing supply over time) in Arizona, Texas, and California over the past 30 years, so show how unimportant supply is to cyclical conditions and how the 2008 debacle is evidence on this point (contrary to the unfortunate conventional wisdom).
The top panel is annual permits issued. The bottom panel is vacancy rate.
The problem starts with the national measure of vacancy. As you can see in the bottom panel, the vacancy rate had risen a bit from 2000 to 2005. Then, it increased even higher by 2009. There is a one-two punch of misdirection here. First, the late rise in vacancies is attributed to oversupply. Second, the rise in vacancies that happened during the building boom seems to suggest that vacancies and housing production were leading indicators. Those over-optimistic builders just kept piling on new supply even when anyone could see we had too much production.
The first problem here is that, as I showed above, there is nothing in the historical data to suggest that vacancy rates are a cyclical indicator. They certainly aren’t a cyclical indicator of over- or under- production. Sharp cyclical changes in vacancies after 2005 were a novel experience.
So, we start with the historically tempting but wrong narrative that cycles are a product of overzealous production, that wrong narrative seems to be confirmed by national vacancy trends in the 2000s, and then the continued rise in vacancies until 2009 seems to magnify the false narrative.
The peak in vacancies was a full year after housing starts per capita had fallen below any previous recorded level. The novel rise in vacancies was the result of a demand shock so massive that it created unprecedented changes in a measure which had never had any significant cyclical behavior before.
The story is even more clear when we look across regions.
Of these three states, the state that had the highest vacancy rate in 2005 was Texas. There was nothing unusual about Texas construction rates, and Texas was an outlier for having moderate home prices at the time. There is no reason to attribute Texas’ high vacancy rate to an over-zealous building boom. And, vacancy rates in Texas actually declined after 2005.
In fact, Texas in the 2000s is a good example of what we lost. Moderately high construction activity with modest demand was slowly bringing rents down. Then, when construction collapsed and vacancies declined, rents started rising again.
The vacancy rates before 2006 were good vacancy rates - the sign of a market creating ample supply from an undersupplied starting point.
When vacancy rates in Texas declined after 2005, they increased in Arizona and California. In California, construction rates are persistently low, and so are vacancy rates. In Arizona, construction and vacancy rates tended to be above normal, but when construction spiked in 2005, vacancy rates hit a decadal low in Arizona. That is because the building in Arizona was largely in response to the surge of migrants escaping the low construction and vacancy rates of California.
The spike in vacancy rates in the boom and bust markets, like California and Arizona, was entirely a lagging result of the deep, years-long demand shock in housing after 2005. I have attributed this to the Fed, but have increasingly come to realize was mostly due to the sharp and novel tightening in mortgage markets that began in late 2007.
The peak in vacancies was 4 years after housing construction started to steeply decline. A lot of variable lags are used to patch the holes of some creaky economic models, but this lag has been one heck of a doozy patching one heck of a hole.
What About 2023?
Today, both rates of construction and vacancy rates are low compared to historical norms. Recency bias puts 2008 in the rear view mirror, but (1) the concern about overbuilding in 2008 was always wrong, and (2) it is much more wrong now than it was then. We need both more vacant homes and more occupied homes.
If we can accomplish that, then, yes, it will bring rents down. But that will take years in the best of circumstances. There is no justification for any policy that would slow down construction on any housing in any urban area today.
I have documented how low construction and low vacancy rates systematically raise housing costs on the poorest residents. Before 2008, American housing refugees largely chose their destinations. And the destinations they chose (like Arizona) built homes for them. Since 2008, a lack of construction everywhere means that the choice of where to migrate to is now based more on affordability. Few places build adequately, and so the relative excess housing expense of different cities directs the migration flow on the margin.
There are always a range of investors engaged in activity that has uncertainty. Some are better than others. Some are luckier than others. Some are safer than others. When a demand shock happens in housing, those uncertainties can become consequences. Some builders are left with properties that operate at a loss when their poor location or poor operations or leverage are exposed by a difficult environment.
Cross-sectionally, those differences are important. In many instances, the developers that take losses made objectively dangerous decisions. But, that doesn’t mean that they were culpable for the difficult environment.
There is a saying attributed to Warren Buffett that “Only when the tide goes out do you learn who has been swimming naked.” That’s all well and good. But, the naked swimmers didn’t cause the tides. And, furthermore, there are times when the tide is strong enough to rip off some trunks.
The psychological pressure of investing at risk creates a large contingent of permabear laymen and professionals who always think the market would be better if something could stop everyone else from overpaying for assets. They religiously think there is always too much leverage, too much optimism, and too much money. And, unfortunately, this sentiment can reach into policymaking circles.
They chant a recessionary refrain. “Rip the bastards’ trunks off!” After 2005, the trunks were ripped off to the tune of 1.5 million new housing units per year and a 40 point rise in the credit score of approved mortgages. That is a riptide that is mathematically impossible to repeat. We aren’t building more than 1.5 million units annually and we don’t have 40 more points to give on credit standards.
It doesn’t matter if we’ve learned the wrong lessons about demand versus supply shocks. We can’t create the demand shock required to make vacancies cyclical. We don’t have the gunpowder.
From a practitioner’s standpoint, that also means that nothing is likely to come along and rip your trunks off.
I have a hobby of checking on U-Haul rates between LA and Phoenix because one of the early indicators of the demand shock that culminated in 2008 was the collapse in housing refugee migration. And, related to Covid, there were migration surges this time around, too.
When the migration stopped, that led to a sharp decline in population growth among cities that had long histories of growing. Arizona was one of those, and that is one reason its lagging vacancy rates spiked.
But, population trends in the Closed Access cities with low vacancy rates were countercyclical. Their trends turned up when housing broke down. In fact, I argue that this has become a systematic trend. Housing deprived cities have population growth that is negatively correlated with housing production as crazy as that sounds.
As shown in Figure 3, the whole country now has the vacancy rates and housing production rates of 2005 California.
I’m not sure that the sorts of migration trend shifts that happened after 2005 - even if they were mathematically possible - would matter again, because every single major metropolitan area now has pent up demand from the intra-metropolitan compromises and migrations that happen when there isn’t enough housing. There are locals who are desperate for more housing. In every city.
It is reasonable for a developer to fear that they are locating a project poorly or mismanaging it. That fear has never been less important than it is today.
Whatever upheavals are still to come in housing markets, the most leveraged, least competent, worst located operators will fail first. But the tides have been altered. The moon has been lassoed. Swimming trunks will remain firmly in place for the foreseeable future.
And, as far as policy goes, there will be many willing swimmers as long as we don’t close the beaches. The housing market doesn’t need help, it just needs permission. Rents will not collapse anywhere because ample supply has never caused rents to collapse. It causes them to moderate. And, they will moderate again, some millions of units in the future.