Inflation, part 2: Shelter in Canada and the United States
Part 1 of this series discussed the role of shelter costs in Canada’s consumer price index, in which I hypothesized that a right-half-plane zero exists in the interest rate → inflation transfer function, via the scaling of shelter consumption with debt costs. The Bank of Canada’s policy interest rate has increased by 275 basis points since that article was written.
I had planned part 2 to be a mathematical exploration of the rest of the inflation system outside of shelter, but this update is due, and the theoretical discussion deserves a separate post.
Finally, let me repeat that I am neither an economist nor a financial analyst. Nor is this an econ blog. I am an engineer, and nothing I write here should be interpreted as financial advice. It isn’t even engineering advice. I’m investigating this subject out of personal curiosity and sharing what I’ve found. And if you have your own views, whether you agree you disagree, I’d be delighted to hear from you!
(Updated 2022-12-07 with some additional numbers reflecting today’s announced +50 bps interest rate increase by the Bank of Canada.)
Update on Canada
(Apologies in advance for all the numbers.)
When I wrote part one in May pointing out that shelter and in particular mortgage interest payments could soon create inflationary pressure in Canada, interest rates had only begun their rise. In fact, at that time the latest data still showed average mortgage interest payments falling 4.4% year-over-year. But, as of last month’s CPI print, mortgage interest costs are now growing by 11.4% year-over-year. (If you prefer annualized month-over-month numbers, it’s now increasing +36% per year).
That is signal, not noise.
It is now November, and Canada’s interest rates are at 3.75%, with another 50 basis points expected to be on the way. The best available mortgage rates are now somewhere around 5% or higher, which is creating some stressful situations for borrowers. The increase in interest costs are, of course, not distributed evenly. It falls mostly on the shoulders of recent buyers, who bought at the highest prices and the lowest rates. These unfortunate individuals are seeing interest cost increases of up to +250%.1
In Canada, we include mortgage interest payments in inflation measurements. So the question I was focusing on was: what does this mean for shelter inflation going forward?
Last time I modelled two scenarios: housing prices remain constant, or drop by half from their peak. It’s hard to decide which scenario makes me seem less alarmist, but for brevity I'll just show the business-as-usual projection:2
Mortgage interest makes up 3.11% of the Canadian CPI; that basket weighting is based on 2021 spending patterns and should gradually grow as interest payments become an increasingly significant expenditure for Canadian homeowners. At this weighting and a
22% 25% growth rate, this sub-category alone should increase overall CPI inflation in 2023 by about +0.7% +0.8%.3
Meanwhile the rental market is heating up and driving up shelter costs even as housing prices fall, now down 10% from their peak.
I do not know how this all gets resolved, but the hoped-for soft landing with a smooth return to 2% inflation by the end of 2024 seems simply not consistent with these initial conditions. We won’t see a return to the status quo because the status is not quo. To balance the equation as it stands, something has to give: either housing prices fall much further, or else shelter pushes inflation above 4% for the rest of the decade, or else some third factor will have to drive strong deflation in other baskets faster than shelter costs can grow.4
Shelter inflation is mostly predictable because it already happened, but it happened in asset prices rather than the CPI. As long as interest rates kept falling, we could maintain an illusion that rising housing prices don’t imply rising shelter consumption.5 The pressure was building up, but it was hidden; now rising interest rates are uncorking the shelter bottle.6
Simply put, this is a badly designed control system, and bad metrics are partially to blame. You don’t want your feedback sensors to be blinded to a growing problem. You definitely don’t want your shutdown switch inadvertently wired to trigger the explosion. But it’s too late for changing the metrics to fix the problem.
Canada vs the United States
I wrote On Shelter Futures from a Canadian perspective, and Canada is still my focus (because I live here, and built these models to inform my own decisions). But I seem to have many American readers who are interested in the extent to which these conclusions apply south of the border.
The short answer is: there are many similarities, and three enormous differences.
The USA and Canada have deeply connected economies, and have seen similar trends for inflation and interest rates since the 20th century. You’ll see the same headlines no matter whose newspapers you read. Common features are visible in the macroeconomic data: the ‘70s oil crisis, the Volcker shock, a long period of stable inflation and declining interest rates, the dot-com bust, the great financial crisis, the pandemic deflation and post-pandemic inflation, and so on. Both have raised interest rates almost exactly in sync throughout this year, with the US currently at 4.0% and Canada just behind at 3.75%.
Without labels, the two countries would be hard to tell apart in the graph below:
The shelter story has similarities too. In the US, the consumer price indices (CPI) are tracked by the Bureau of Labour Statistics (BLS). As of 2021, shelter made up 33% of the US CPI basket, which is almost the same as Canada’s 30%. Both the US and Canada have similar ratios of renters to homeowners, and perhaps surprisingly, nearly identical supply of homes per capita:7
That covers the similarities. Now for the differences.
Difference 1: Just like StatsCan, the BLS does not include housing prices in the CPI directly, instead estimating an equivalent cost of shelter-as-a-consumer-expense.8 Unlike StatsCan, the BLS does not directly include mortgage interest costs in the CPI, so rising interest rates are only relevant to shelter inflation to the extent that they influence the rental market. This should reduce the degree to which rising interest rates transfer into shelter inflation in the US.
Difference 2: The US has 30-year fixed-rate mortgages, and they are widespread, especially after the financial crisis of 2008. In Canada, such products are not even available to homebuyers. Canadian “fixed-rate” mortgages would be called adjustable-rate mortgages in the US, as they are typically 3-year or 5-year contracts. So while Canadian fixed-rate mortages bought at low pandemic rates and high pandemic prices may be facing rising interest costs starting in 2025, many Americans who bought a home in 2020 can enjoy low rates until their last payment in 2050.
In the United States, it’s theoretically possible that hardly anybody takes out a mortgage until either rates or housing prices come back down to Earth, in which case combination of high rates at high prices would only land on the most recent buyers and the small fraction Americans that have taken out adjustable-rate mortgages.
The volume of new mortgages is thus a critical input variable for the US analysis, and with home sales dropping, mortgage originations are falling too.
That’s not the case in Canada, where 35% of outstanding mortgage debt that is held at variable rates, and all of the rest is adjustable-rate, so virtually anyone with more than 5 years left on their mortgage will be affected. That means Canada will see little decrease in mortgage activity in the medium term, as outstanding loans will continue being renewed even if new mortgage originations dry up.
Difference 3: In both the US and Canada, housing prices tracked inflation from 1980 - 2000.9 Around the turn of the millennium, house prices began similarly steep climbs in both countries. In the US, home prices underwent a significant correction following the financial crisis of 2008, returning to their inflation-adjusted norm before regaining steam in 2012. Adjusted for inflation, US housing prices have now overtaken their pre-financial-crisis high.
Canada saw no such correction, and even though interest rates have been slightly higher and true fixed mortgages are not available, home prices have continued climbing at about the same uninterrupted pace for 20 years.
Many commenters, including the Dallas Fed, have observed that the US may be experiencing another housing bubble and anticipate another correction. This seems entirely plausible, and I don’t mean to deny or understate the severity of the situation facing American readers, where inflation remains high and housing affordability has fallen to a new low.
I’ve been watching the US shelter situation closely for some time, as Canadians do. One year ago, economist Larry Summers published an article in the Washington Post criticizing the Federal Reserve for suggesting that inflation was “transitory”. At the time, the most recently published US shelter inflation data was only at 3.5%, and was far from the spotlight in most other media coverage.
“Housing prices and rents are both up 15 to 20 percent in the past year. These movements are far from fully reflected in the shelter component of the consumer price index, which represents one-third of the CPI, implying substantial pressures to come.” — Larry Summers, 15 Nov. 2021
On this point Summers would be proven correct, and shelter inflation progressed upwards in a remarkably (if predictably) straight line over the course of the year. (Note that linear inflation growth represents superexponential price growth). Shelter is now leading core inflation in the US, and for the Fed as for Summers this should not have been hard to see coming one year ago.
Marginal market rent indices have now peaked in the US, but by my estimation, the CPI still has some distance left to go to catch up with where the market has climbed to. This perhaps puts me in disagreement with Nobel laureate Paul Krugman, who believes that peaking rents now indicate that “as a driver of inflation, at least, [shelter] is fading away”.
Zillow’s data represents posted rents on the market today, meaning the average prices you’d see if you were looking for a rental somewhere in the US. That’s not quite the same as what the CPI tracks, which is the average rent paid by all renters in the country, including long-term tenants whose rent hasn’t changed recently. It makes sense that the CPI is playing catchup.
Just to add some very ballpark math, looking at the same data as Krugman, we can see that market rents increased by about 26% over two years, from $1619 in October 2020 to $2040 in October 2022. Yet, over the same period, the US CPI rent index has increased by only about 10%. That means that at its current inflation rate, the CPI will still take another two years just to catch up to where market rents are today. If market rents fall from here, that will happen quicker, but Krugman is not pointing to falling rents; he’s pointing to slowing rent growth.
Despite all this, it seems to me that the US housing market is much healthier than Canada’s. And that means that at least on the shelter front, America probably has a lot less inflationary pressure still trapped in the bottle.
This post isn’t intended as professional engineering advice. If you are looking for professional engineering advice, please contact me with your requirements.
When I model scenarios showing housing prices or interest rates flatlining, it isn’t because I expect that they will flatline. It’s because I want to illustrate the long-term mathematical consequences of today’s prices or rates being at their current levels.
(Updated 2022-12-07). The numbers in strikethrough assume average market mortgage rate throughout 2023 remains at 5%, which was the market rate when this article was posted. The non-strikethrough numbers assume 5.5%, reflecting the Bank of Canada rate increase that was announced on December 7th.
It’s hard to use a model make narrow predictions about the future, because the space of real possibilities is always much wider than any model can sanely capture. Many things could happen, especially in a system with multiple degrees of freedom. So I’m not going to predict how this situation will resolve itself, because there are too many possible futures. But that doesn’t mean that any outcome is equally likely.
I’m curious about whether the same could be said for other asset classes, eg. rising asset valuations for farmland or factories converting into food or consumer product price inflation when interest rates increase, but I haven’t done the math on it yet. If you’re aware of any papers about this, please let me know in the comments!
As a reminder, this was the thesis of part 1: that rising interest rates are moving inflation from an unmeasured quantity (asset prices) to a measured quantity (consumer prices). Unmeasured, at least, as far as the CPI — and the policy feedback intended to regulate it — are concerned.
Credit to Doug Porter at BMO for bringing attention to this, and to Mark Mitchell for bringing attention to Doug Porter.
Until 1980, the BLS calculated homeownership costs including the home purchase price in the cost of ownership, along with mortgage interest and property taxes. But the 1980s saw some upheaval in how shelter costs were calculated, and the basket weights were heavily revised, introducing the new category of “owner’s equivalent rent” making up the majority of homeownership costs. This makes pre- and post-1980s comparisons a little bit difficult. Owner’s equivalent rent is based on how much homeowners think their home would rent for, if they were renting it. It tends to track rent very closely, and it dominates the shelter calculation.
Home price indices don’t always agree with each other. In particular, the CES study seems to suggest that home prices in the US matched inflation not only from 1980 onward but even as far back as 1948. My guess is that the differences between the CES data and the others might be related to different approaches to adjusting for inflation. As mentioned in footnote 7, before 1980 the BLS included housing prices in the CPI, which makes “housing prices trended with inflation” almost a truism.