First, the funny chart. Mortgage rates and build-to-rent single-family homes. Apparently, ZIRP was the only thing holding this industry back!
Actually, it’s possible that high interest rates are causally related to the rise in build-to-rent.
Build to Rent and Interest Rates
My favorite explanation for the brief tick up in multi-family starts is that the Covid pandemic moved a lot of municipal community engagement online instead of through in-person meetings. It’s a lot harder for NIMBYs to be effective when they are muted on a zoom meeting than when they are filling a room, shouting, interrupting, and displaying anger. Multi-family projects tend to take a couple years to complete, so that tick up in starts in 2021 and 2022 became a tick up in completions in 2023 and 2024. Maybe Covid conditions helped push through an extra 300,000 or 400,000 units that would have otherwise been bogged down with community engagement, and now multi-family starts naturally settled back at near the pre-Covid pace.
Well, it so happens that when multi-family starts turned back down, interest rates were high, and especially short-term interest rates were high. As readers know, I have an unhealthy skepticism of the causal effect of interest rates on housing markets in general, but it is possible that they have had some effects on market share of new home construction in recent years.
Because multi-family construction takes a couple of years, and because of the structure of the industry, short-term interest rates are kind of important. Developers usually depend on securing short term construction loans to fund projects before they are ready to lease up.
Single-family homes can be built much more quickly – in just a few months – and single-family builders tend to be funded mostly with equity capital rather than debt.
Under current conditions, where total production appears to still be supply-constrained on some dimensions, it is possible that the differences in construction schedules and capital costs led to a shift in market share from multi-family to single-family construction.
Maybe, under those conditions, high interest rates were causally related to the persistence of single-family build-to-rent construction. Given a limited number of units that could be completed, maybe high short-term rates marginally shifted that activity to single-family construction.
As I pointed out in the recent construction data update, starts and completions have oddly diverged in recent quarters in single-family build-to-rent. As seen in Figure 2, starts have declined sharply in the last few quarters, but as seen in Figure 1, completions are still elevated. I speculated that builders have been starting homes with the intention of selling them to owner-occupiers, and that investors are coming in as a plan “B”. Maybe those homes flow to “for sale” on the starts survey but to “for rent” on the completions survey?
Or, maybe I am wrong about interest rates and they actually have an important causal effect that works with a 1 year lag on multi-family starts and a 3 year lag on single-family build-to-rent starts. (I’m joking, though I suppose I’ll have to eat that joke if single-family build-to-rent completions fall back to pre-COVID levels.)
The shift in mortgage rates was so quick and significant, it might also have affected household formation patterns. Again, in terms of aggregate housing supply and demand, I don’t think interest rates are that important, and I think home prices, even for owner-occupiers, are mostly explained by rental values. But, the shift from 3% mortgage rates to 7% rates in a year has certainly reduced transaction volume among homeowners, and likely has delayed some capital constrained first-time homebuyers.
The New Housing Cycle
I think our current conditions, where we are 15 or 20 million units short, may create some systematic patterns here. Under normal conditions, where the market for new homes is demand-constrained, cyclical trends might temporarily reduce household formation. It might lead to a reduction in total construction. Under normal conditions, the stock of existing homes is a big stable set of supply, and new construction adds 1% or 3% to that in a given year. In terms of total housing demand, it is very stable, but in terms of new construction, it’s very volatile. A difference of just a couple of percentage points in real income growth, associated with a change of just a couple of percentage points in real demand for housing, could double or triple demand for new homes under that normal context.
Of course, that’s why construction was so cyclical before we had the shortage. Before the 1990s, new home construction 5 years before or after any given date was likely half or double the rate it was at the given date. Since the 1990s, think of it as basically a line up and to the right, straight as a stick, with a giant earthquake in the middle that created a 1.5 million unit fissure. (And, of course, a bit of volatility in permits after 2020 because the builders lacked the capacity to turn more permits into completed units.)
This makes construction much less cyclical. When there are 15 million households in a holding pattern waiting to form, cyclical shifts of a million here or there don’t amount to as much as they did when household formation was generally satisfied and new housing depended on new households.
So, I think now when an owner-occupier household doesn’t form, it mostly creates an equal and opposite reaction in renter household formation. Even if there is a 25 year old living with their parents who might have purchased a home when they moved out to form a household. Their lack of household formation allows one of the other 15 million potential households to move up in the queue.
I think we can see this in the data on new home construction and household formation.
In our shortage market, rentals are what’s left over after buyers are sated.
The black line in Figure 4 is the cumulative growth of homeowners since 2016. That number has grown by about 1 million households a year, on average, but it has slowed down a bit in the last couple of years, coinciding with higher mortgage rates. (There were a couple of quarters in 2020 when Census Bureau surveys had unusual noise, and I have manually smoothed those quarter.)
The yellow line is the cumulative number of new single family homes sold to homeowners. The red line is the sum of the total number of new single family homes sold to homeowners plus the decline in the number of vacant homes.
The number of newly built homes plus newly occupied homes weren’t enough to supply the total number of new homeowners. The gap – the difference between the red line and the black line – is the number of formerly rented existing homes that were purchased by homeowners to make up the difference. The difference between the red and the black line is the number of existing rental units that were lost from the rental market to the owner-occupier market.
Now, let’s look at rental housing.
Again, in Figure 5, the black line is the cumulative change in the number of households. Here, renter households.
The yellow line here is that difference from Figure 4 – the number of previously occupied rental homes that transitioned to the owner-occupied market. By the way, if this data is accurate, then it may no longer be true that investors have been net sellers in the existing home market. According to this estimate, investors were net sellers from 2016 to 2023, but they have been net buyers since then.
I will soften my claims on that matter. Though, it could be the case that large investors have been net sellers of scattered homes while they transition to purpose-built rentals and the recent shift is a return to a normal growth of small scale investors into the depreciated older units in the existing stock of homes.
The orange line is the number of multi-unit homes completed added to the change in the number of existing rental homes. The red line is the number of single-family build-to-rent homes added to the number of new multi-unit homes and the change in the number of existing rental homes.
In other words, the black line is the cumulative number of new renter households and the red line is the cumulative change in the number of rental homes.
I think we can tell a story here. Owner-occupiers are driving the market. They pay a premium compared to investors, and so they own the homes they choose, in the aggregate. Under current conditions, renter household formation depends on housing supply, and rental supply depends on new construction activity and owner-occupier purchase activity.
The blue line in Figure 5 is a hypothetical level of demand for new rental homes, based on a stable increase in rental households of about 1.1 million households per year. Over the past decade, that would have allowed about 500,000 new rental households annually to meet natural growth plus 500,000 or 600,000 of catch up growth annually, which would have closed the household gap by roughly half or a bit less.
The red line in Figure 6 is the difference between the red line in Figure 5 (the cumulative supply of rental units) and the blue line (the hypothetical demand for new rental units).
The black line in Figure 6 is cumulative rent inflation, estimated from the difference between Zillow’s US rent estimate (with a small technical adjustment) and excluding shelter.
This relationship would suggest that when there is no growth in the functional stock of rental housing, rent inflation runs about 2% above general inflation. That makes sense. Real incomes tend to grow by about 2% annually. There is a strong tendency for families to spend a set amount of their incomes on housing, in the aggregate. When there aren’t new structures, rents rise at roughly the rate of incomes for the existing units.
And, this relationship suggests that when the functional stock of rental housing increases by about 1.1 million units annually, it has facilitated enough household formation to level out rent inflation.
Odd Thoughts
I think we can’t really think about this market in equilibrium terms. This is more like a flow rate of pent up demand back toward equilibrium. There are around 15 million households that would form, if there were homes. The flow from potential to actual households of about 1.1 million renters annually appears to be associated with moderating rent inflation.
In the before times, household formation would have determined the size of the market for new homes. And that was associated with volatile construction activity. Today, the supply of homes determines household formation, and so it is associated with stable aggregate construction trends and changing rent and price trends.
I have taken a stab here and there at elasticities and estimating rent inflation at different quantities. But, I’m not sure it’s possible to quantify that into anything more than rough guidelines. I think this is more of a flows thing than a levels thing – the healing of a disequilibrium more than a change in the equilibrium quantity and value of homes.
Sometimes, working with simple models, I might use a constant demand elasticity. But my hunch is that, in some markets, a substantial increase in construction of new rental units won’t necessarily create a mirror image of rent inflation that accumulated when construction was too low.
The demand elasticities (basically, willingness to pay) of newcomers to a growing city, newly formed households within a city, households leaving a city for broader economic reasons, and households leaving a city because of high housing costs, can be wildly different, and can operate on different margins. There is a lot of thinking that could be done on those details that I haven’t tackled yet.
While constant elasticity coefficients can be useful, it is worth at least keeping in mind what we don’t know on this topic. Elasticities on the way up and on the way down might be fundamentally different. Households mostly make real changes when they move, but they make nominal compromises in order to stay put.
That issue is the cause of a lot of the confusion that leads economists to mistake stagnant cities for “superstars”. They mistake the inelastic demand for housing that families express when they are being economically displaced as newcomers’ high willingness to pay.
What the past decade has demonstrated is that new households are willing to pay more (or consume less to counter higher costs) in every city, if they have to. And, they have to pay more when the formation of their household requires the displacement of another household in some way. The family fighting displacement becomes the price setter. Displacement makes a city excessively expensive, not value.
An added facet here is the difference in demand for renters versus owners. That margin is different than it used to be. Broadly speaking, in the before times, if it was marginally advantageous to be an owner, households became homeowners. Imputed rents of owned homes could be modeled from the market rents of similar rented homes.
Since the 2008 mortgage crackdown, market rents on rental homes are much higher, but the cost of buying those homes is not higher for those who can qualify for mortgage financing or pay cash. So, there is now a discontinuity in the experienced rental value of a home, depending on if it is rented or owned.
(Confusingly, because of other moving parts in home valuations, that doesn’t necessarily mean that the price/rent ratios of those hypothetical homes has declined, but that’s a trip more deep into the weeds than I want to get here.)
It’s hard to maintain an intuition about these disequilibria, so I have been sort of thinking out loud on some of these posts. There may be some errors here. Or, I might later realize that there is an incongruity in some of the logic here, but this is how I’m thinking about it.
It might be hard to forecast exact numbers going forward, but I think the basic narrative outlined by Figures 4, 5, and 6 is a useful way to think about housing supply and demand as we move toward normalization.
Moving forward, the interaction between home buyers and the existing stock of homes might ebb and flow. At best, the transition of some older homes into the rental market might help to meet the annual marginal growth of renter households. Under current conditions, the growth of the rental stock appears to be keeping up with marginal new demand.
But, under current legal norms, the gap between current renter households and pent up demand for rentals (the black and blue lines in Figure 5) will mostly have to be met with single-family build-to-rent homes. And, the scale of the demand gap that will be filled is much higher than the current size of that market.
If demand from owner-occupiers picks up again, it will be split between new and existing homes. It will pull the yellow line in Figure 5 down, and so it will pull the orange and red lines down too. So, ironically, more demand from owner-occupiers will probably increase the market share claimed by build-to-rent on new single-family construction. In part, that will depend on whether multi-family units are constrained more by financing or by local land use obstructions. I suspect local obstructions are still the more important constraint, but only time will tell.
Finally, think of the blue line in Figure 5 as the push back toward household formation equilibrium. At some point, the households that have been delayed by the housing shortage will have formed, and the slope of the blue line will flatten a bit. If Trump’s anti-growth policies lead to substantially lower population growth, the blue line will level out at a lower level than current trends would lead to.
Until the gap between households and potential households is closed, households will generally occupy the new homes that can be constructed. If the gap closes faster, then rents might moderate faster as households transition into the new supply. But, I think construction activity will continue at the pace capacity allows, for the most part, as that happens.
With Trump in office, of course, black swan outcomes are possible, however.