It's never overbuilding. More detail.

In the previous post, I discussed how overbuilding wasn’t the cause of high vacancies after 2008 and really has never been associated with high vacancies.

In this post, I will poke around a little more in the Census Bureau’s data on production and housing inventory data.

Figure 1 uses a 2 year time frame to cut down on statistical noise. The blue line is 2 years of production. Here I am using the sum of completed new site-built homes and shipped manufactured homes. This is a more comprehensive metric than simply using the more commonly cited construction data. Manufactured homes have become such a small portion of new production that it doesn’t matter much in a lot of analysis, but in long-term analysis of the housing stock, it does matter because a relatively large portion of new housing was manufactured housing until the 1970s, before the homebuilders conspired to regulate much of it away.

So, the supposed building bubble of the 2000s looks even more puny when this data is appropriately included in the measure of housing production. It’s a pretty pitiful thing to have a generation-defining moral panic over, but I suppose that’s how moral panics always play out.

The black line is the 2 year change in occupied houses. The red line is the 2 year change in vacant houses. (The largest bump in vacancies in the mid-1980s is partially due to a change in methodology, I think.)

Figure 1

You can see how both occupied and vacant units have risen and fallen with building booms (back when we had such things), and so changes in vacancy rates over the building cycle were relatively small. (The sum of the change in vacant units and occupied units tends to run slightly higher than the number of new units because some units are lost over time. There is a lot of noise and the number of lost units is difficult to track in real time. That is, in part, because there isn’t as much of an organized bureaucracy associated with tearing down units, some units are combined or created from existing units, and the deteriorating condition of some older units is not a binary event as in when a new home is completed.)

In a way, I almost feel foolish for engaging in analysis of the vacancy rate of the 2000s building boom, since there wasn’t much of a building boom. It’s sort of like running tests on a hypochondriac. But, there are things to learn here, I think.

One thing this chart helps to highlight is how weird the late vacancy surge after 2005 was. The rise in vacancies does tend to lag building booms, because the business cycle has tended to be driven by demand shocks. Recessionary frictions cut down purchase activity. Household formation and construction decline while vacancies continue to rise for a bit. (Keep in mind, this is vacancies in the existing stock. Not unsold new inventory. This is related to an inability to utilize the existing stock of homes rather than builder expectations and construction schedules. Though, all of these things affect each other.)

By 2009, those frictions had been so strong and so persistent that for nearly every new home that had been completed in the previous two years, there was an additional vacant home. That was in spite of the fact that construction was down by half.

In other words, the red line nearly touched the blue line. That’s never happened before. Vacancies spiked because household formation cratered when the economy crashed and mortgage access was cut off. Not because builders were optimistically completing new homes.

The collapse in household formation was similar to the collapse in the 1980 recession. In both cases, the rate of household formation over a two-year period declined by more than 2%. However, after 1979 it went from more than 4% to less than 2%. After 2005 it declined from under 3% to nearly zero.

What made the collapse after 2005 different was that household formation, and home building, started the period at much lower rates, so the relative collapse was much larger and the rise in vacancies relative to the slower rate of growth in the housing stock was much larger.

In both cases, by the time household formation bottomed, homebuilding had contracted enough that vacancies weren’t growing any more. In 1983, household formation recovered, and so did homebuilding and the number of vacancies. In 2011, household formation recovered, but homebuilding and the number of vacancies did not.

By 2011, the number of vacant units started to decline. This is also something that has never happened before. It can’t happen sustainably. And yet, it continues to be the case 13 years later.

One way to visualize the utilization of the housing stock is to compare the growth in households to the production of homes. In other words, the relative difference between the black line and the blue line in Figure 1. That is shown in Figure 2.

That maxes out at around 80% when household growth is high, like it was in the 1970s and 1980s. In other words, for every 100 new homes constructed, about 80 were occupied.

If the number of households stopped growing, the ratio would, of course, fall to zero, because the numerator would be zero, as it nearly did in 2010.

It has generally been above 100% since 2011. And that is one way to think about housing production trends going forward. Annual housing production is currently at about 1.5 million units. At current rates of household formation, annual housing production would need to rise to about 2.5 million units just to get the utilization rate down to the sustainable 80% range.

Figure 2

I have included the growth rate of household formation in Figure 2 to highlight how the decline in new housing utilization before 2011 was correlated with declining household formation. (It doesn’t quite work using two linear axes because at very high rates of household formation, like in the 1970s, utilization asymptotes at around 80%, but I think the chart is good enough for the post-1980 period to make the point.)

It is true that household formation has been high. Population growth averaged about 1% annually, right up to the Great Recession. It looks like 0.5% might be the post-Covid norm, though a loosening of immigration policies might move it a bit higher again. But, household formation has been running at more like 1.5% annually.

This isn’t particularly unusual. In the 1970s, population growth averaged just over 1% but household formation was routinely over 2%. That largely played out in declining household size.

As Figure 3 shows, US household size leveled off in the 1990s (when housing constraints became especially strong). Household size in Germany, France, and Japan, where housing constraints aren’t as bad, has continued to decline. How much further would household size decline in the US without constrained housing supply? It’s hard to say. The approximately 3 or 4 million young adults living with their parents, compared to 30 years ago before housing became so constrained suggests that it is substantial.

Figure 3

Today’s rate of homebuilding would be a sustainable level to maintain going forward if population growth remains at 0.5% together with additional production equal to about 0.5%.

Figure 4, from one of my Mercatus papers, compares housing production and population growth over time. The way to think about this chart is that the dashed green lines represent the amount of housing required to accommodate population growth. The solid green line represents changes in the housing stock that are unrelated to population growth.

For most of the last 60 years, population growth has been around 1% annually, so about 4 homes per thousand residents were required for population growth (about 1.3 million units today). During the 2000s bust, cyclical housing demand went negative (the solid green line pushed housing production to the left). But, for most of the last 60 years, cyclical demand ranged from zero to 4 units per thousand residents, averaging around 2 (about 600,000 units today).

Figure 4

Those dots dribbling down to the 0.5% population growth range are the pre-Covid years. Since then, population growth has been between zero and 0.5% and housing production has remained around 4 or 5 units per thousand residents (about 1.5 million). That’s been a bit more than 2 units moving right on the solid green line for excess household formation, and 2 units moving up a now shorter dashed green line.

This leaves us with a baseline housing production level going forward of about 650,000 units annually, to accommodate population growth. Housing bears (Lord help them) will start with that, suggest that population growth will decline from here (which may be true) and will add some precise metrics about age demographics and how it will affect household formation.

I think the precision and certainty of those estimates is seducing, but it is simply swamped by the things we can’t really know. That is a baseline. There are few things that could push us below that baseline. But, there are countless uncertain positive real shocks that may push that number up. Just to name a few:

• The millions of households worth of pent up demand, such as those young adults living with parents.

• About 5 million vacant units just to get back to a normal vacancy rate. (We need to build empty homes! We’re not spherical cows! The real world needs slack!)

• New units needed each year just to stop the bleeding on the vacancy rate.

• A return toward historical immigration rates.

• Rising units for seasonal use, short term use, use by foreigners, etc.

That last category would probably rankle the feathers of some NIMBYs. In a state of deprivation, every source of demand by “outsiders” is a threat. But, in a normal economy, those are natural sources of demand. Of course as we become richer, we will want some additional housing. Of course as the world becomes richer, they will want to have some extra homes to visit and use in the US. That should be no more threatening or outrageous of an expectation than the fact that we get our clothes from trendy boutiques instead of washing our burlap sacks once a week. The world gets better. We have been enticed into a habit of thinking that housing is exempt from that expectation. Or worse, that it should be exempt from it.

That solid green line will continue to be to the right of zero. In most feasible scenarios, the solid green line will continue to be a large source of the demand for new housing. Exactly which of these factors will push it to the right? And how far? I don’t pretend to know. We may lack precision about the future, but we can be confident about its direction. Let those with precision be surprised by the unknown.