Jeremy Horpedahl has an interesting post on spending on housing. I think digging into the details of it might be a useful exercise.
Horpedahl usually does quality analysis, but I think he misses the boat here. That’s forgivable, because this is a particularly tricky subject. His post also alerted me to this particular data set. It is tricky to interpret, but I think it does give up its secrets with a tickle or two.
I will likely do more than one post on this, because I think that with just a few basic inferences, there could be some interesting things to learn here.
Here is a typical chart from the post. Horpedahl looks at average spending, and he breaks it out by age, urban vs. rural, and by region. He doesn’t break it out by tenure (owners vs renters), and that is where this data is tricky. This data just tracks actual cash expenditures. So, for renters that is mainly rent, and it is a pretty straightforward description of housing expenses. For owners, it tracks maintenance, property taxes, mortgage interest, etc. So, frankly, because owners dominate the trends, the top line of this data is really pretty useless.
The Finance 101 way to estimate owner expenses would be to estimate the rental value of their homes. Or, you could take the other side of the equation, and add up all the costs. That would include the costs listed above. But, some of the mortgage interest is just an inflation premium. In a no-inflation environment, mortgage interest rates might be 2%. In that scenario, then interest expense would be straightforward. But, if inflation is running at 2% and the mortgage interest rate is 4%, in stasis, the homeowner is really still only spending 2% on interest expenses, in real terms. The other 2% is basically an equity buydown, and the homeowner is paid back through the 2% growth in home equity that comes from inflation.
In this data, the entire 4% is treated as an expense. So, in that way, owner expenses are overstated.
Also, there is the homeowner’s equity. An owner without a mortgage pays no interest expense. So, in this data, the reported housing costs of free and clear homeowners are very low. Of course, the home isn’t really free in that case. The cost here comes in the form of opportunity costs. What would that capital be earning if the household was renting and invested that capital in some other asset? Now, we’re really getting into some complications.
The owner part of this data ends up having basically nothing to do with true housing expenses at all, at least in any way that wouldn’t get you fired from a corporate financial analyst position. The costs that are reported in this data for owners are affected by inflation, real returns on similar assets, the leverage of each owner, etc.
Now, if you properly accounted for all those things, the total expense should basically equal the estimated rental value. That’s why in many data sets, like the CPI, imputed rents are used to track the cost of living.
Owner Expenditures vs. Renter Expenditures
The data here is broken out between owners and renters since 1984. And, it has further been broken out between mortgaged owners and free and clear owners since 2003. Here’s what it looks like when we compare renters and owners.
Renters have seen an increase of about 5 percentage points of their incomes taken by rent. That is sizeable. That’s a 25% rise in relative spending in this category over a span of about 40 years.
Spending by all households has been flat, as Horpedahl found. And, since, by value, most spending is by homeowners, that is largely because spending by owners has slightly declined. Like I said, though, the owner spending numbers are a conceptual mess. A proper accounting of all opportunity costs, etc., should basically add up to a trend similar to the trend for renters. A combination of low inflation and low leverage, which are biases in owner costs that have little to do with true housing costs, has created the downtrend for owners.
(I think we should expect the trend of rental values for owners to be a little less positive than the trend for renters for reasons that are evident in the Erdmann Housing Tracker data. The housing shortage causes costs to go up the most for families with lower incomes, who are more often renters. But, the trend in imputed rental value as a percentage of income is still likely positive for owners.)
Cost Inflation vs. Real Compromises
The data for renters comports with the data I have reviewed previously from PCE numbers. Total housing as a percentage of all expenditures has increased by 10% to 20% since the 1970s.
Figure 3 introduces another complication however. Total expenditures mostly tells us that housing poses budget constraints. At some margin, households pull back on housing expenditures pretty strongly to try to maintain a steady stream of spending. If Americans are spending more on housing, that will inevitably be paired with attempts to cut spending by reducing real consumption - smaller units, less convenient units, migration to less expensive cities. And, of course we are seeing all of that with American households.
So, spending is not flat. Spending on housing is up 20%. And, according to PCE data, that net increase of 20% is a combination of a 50% rise in relative costs and a cutback of 20% to 30% in real consumption.
It would be hard to spell this out if spending had been flat. Because, it might still be the case that Americans were cutting back on housing in exactly the same proportion that costs were rising. That could come from low productivity in housing production, and while higher productivity is better, some sectors are bound to be less productive than others. So, in that case, the argument could be made that American renters natural make tradeoffs to maintain housing expenditures at roughly 20% of income, and the details hardly amount to a crisis.
But it hasn’t been flat. It has been far from flat. And since it has been far from flat, that means that the problem is much, much worse than nominal expenditures make it seem. If Americans are spending 20% more on housing, they are also engaging in all sorts of other costly coping behaviors, and that 50% PCE inflation figure is probably a better gauge of how bad it is.
Mortgaged Owners vs. Non-Mortgaged Owners
There is also some information to pull from the owner side of expenditures. Previously, I have discussed mortgage affordability. Over time, homes have become more expensive in the housing deprived cities, and you can see that in the mortgage affordability numbers. But in general, especially since the Great Recession, borrowers have been getting a huge discount. Prices were driven low by mortgage suppression, so those who could get a mortgage got a smokin’ deal. The recent rise in mortgage rates have only really just closed the gap back to normal.
The mortgaged and non-mortgaged numbers only date from 2003. The measure for all owners is an average of the two (weighted more toward mortgaged owners because there are more of them and their homes tend to be worth more).
The trend of costs for non-mortgaged owners is at least as steeply rising as the trend for renters. The trend for mortgaged owners is lower. The aggregate signal of flat costs is the average of actual costs, which are rising steeply, and a combination of low rates and low leverage among mortgaged owners, which has roughly counteracted the measure of truly rising costs.
Figure 6 shows the components of shelter expenses for mortgaged owners since 2003. Total expenditures have declined by 2.5 percentage points. Costs of ownership have risen, but it has been more than counteracted by a 4 percentage point decline in mortgage interest expenses. Some of this is lower rates. Some of it is a low inflation premium. Some of it is lower average leverage, which means that opportunity costs, which this data doesn’t attempt to capture, are an increasingly important part of housing expenditures.
There is this idea that interest rates have an important effect on housing expenditures. Families go out and buy the nicest house they can get a mortgage for, so when interest rates rise, buyers have to downsize, and housing expenditures fall.
Yet, as Figure 6 makes clear, the most volatile part of housing expenditures is mortgage interest expenses. Interest expenses, in the aggregate, absolutely do not serve as a valuation anchor or spending anchor in housing. This sloppy notion is so ubiquitous, you can earn trading gains simply by ignoring it while everyone else trades on it. The vast, vast majority of homes are not bought or owned based on a maximum interest expense bound. The anecdata doesn’t add up to anything particularly important in macro analysis of American housing.
I’ll leave it at that, for now. I suspect I’ll come back to this more than once, because I think that with just a few basic inferences, there could be some interesting things to learn here. There is a lot to unpack in those owner cost trends. But, for now, the Consumer Expenditure Survey, unfortunately, confirms the obvious. Americans are spending a lot more on housing than we used to, and we’re getting a lot less for it.
A point about "why buy a house in America?": it has become a lottery ticket for an old-age pension, since pensions have been mostly destroyed.
Thanks for this detailed critique. There's a large cohort of people in America who take the position of "Gosh, there's nothing wrong!" and back it up with bad interpretations of data from folks like Horpedahl. It's why I like your income-to-housing expenses graphs for metro areas. They demonstrate how bad things actually are and also indicate how to make them better.