Some visuals on cyclical home prices
Here’s a way to think about the cyclical ups and downs of the housing market over the past 20 years. I’m going to walk through these ups and downs to end up at a picture of today’s market, and a framework for you to consider what you think normalcy will be.
In these charts, the x-axis is the ZIP code income (the axis origin is the average for each metro, and each vertical gridline is .25 log points above or below. The y-axis is the change in home prices over time. Here, I am using changes in the price/income ratio, which helps to control for unusual price changes versus price changes that are just growing in line with incomes. Again, it is a log scale. (A log scale is the same as a percentage scale, but it is just compounded. So, think of a 0.25 log change as 25 sequential 1% changes, either up or down. Log scales are useful because they are symmetrical. A 0.25 log decline followed by a 0.25 log gain ends up back right where you started. Percentages stated below are all log changes.)
Figure 1 shows relative price changes from January 2002 to December 2003. This is a classic image of a boom, similar to what we have seen recently. There is no correlation between price trends and incomes, and differences in price trends are very metro-specific. If a ZIP code is above or below the trendline, it is very likely that this is because most ZIP codes in that city are above or below the trendline. So, during this time - which largely pre-dates the “subprime” boom, and definitely pre-dates the CDO boom - there was an average increase in prices of about 15% relative to incomes.
The first panel of Figure 2 shows the 2004-2006 bubble period. There really shouldn’t be any relationship between income and price trends. And, this was one reason why credit has been attributed such a strong role in rising prices during that period. I think we can attribute that pattern to local supply constraints more than to credit access. (I swear I have a paper coming out on this any minute.) In the next post, I will go into some local markets, where you can clearly see that the pattern in 2002-2006 is a combination of anomalous localities with various supply constraint problems.
The equation in each panel gives you a good way to think about this. The equation is basically the Erdmann Housing Tracker estimate of the change in relative prices that is income sensitive (either due to supply constraints or credit access) given by the x-coefficient, and the change in relative prices that is cyclical (the rise or fall of prices in high-income ZIP codes that aren’t credit constrained or supply constrained) given by the intercept value. The intercept minus the x-coefficient is roughly the average. (So, the average price/income level increased by about 19% from December 2003 to June 2006 : .039 + .146 ). So, notice that while the average price change in the 2020-2022 (below in Figure 4) period is similar to the average from 2002-2003 and again from 2003-2006, these are very different markets! 2020-2022 looks like 2002-2003, but not 2003-2006.
Over the course of 2006-2007 (panel 2), the Fed engineered a cyclical housing downturn. By December 2007, national housing starts were down more than 50%. By the end of 2007, a deep cyclical correction had taken place. A cyclical rise of about 4% remained (the sum of the intercept values), after reversing the gains of 2002-2006.
Supply issues continued to create higher housing costs at the low end, so the relationship between income and price was still negative. The average home only lost about 3% during the cyclical reversal. So, the average home, on net, from January 2002 to December 2007 was still up by about 31%.
We had a housing supply problem which was misinterpreted as a cyclical bubble, and so we mistakenly implemented a cyclical contraction, which simply gave us another problem. Now we had a supply problem (raising average prices and really raising prices where incomes were low) and a cyclical contraction (lowering all prices). And, you might see how this led to disaster, because since nobody understood this, the universal interpretation of the net change from 2002 to 2007 was that there had been a 31% housing bubble that had hardly even begun to reverse. So, while housing starts were blaring a red signal that this contraction was long in the tooth, the country barreled ahead, locked arm in arm, determined to bust a bubble-that-wasn’t. And, determined that it must be those damned lenders and speculators that were making it such a hard bubble to bust.
Let me take a moment to backfill a bit here. There was a credit boom sometime in that 2002-2007 period, and it likely had some interaction with rising prices. In my books, I have gone into lots of detail as to why the interaction is overstated. But I am not trying to erase that activity or claim that it was unimportant in every way.
I have another paper that may come out as soon as this month where I use this framework to try to estimate the effect of various factors driving 2002-2006 home prices, and in that paper I estimate that credit access was related to an average increase in price/income levels from 2002 to 2006 of about 8%. That is larger than the estimate that I have seen isolated (rather than implied or extrapolated) in the existing literature! But, the effect of credit access is much smaller than the effect related to constrained supply, and its interaction with prices is complicated during the 2002-2006 period.
So, if you want to carve out 8% of that 31% rise in the average home value before 2008 and attribute it to the credit boom, that’s fine. There are complicated reasons why I would caution against giving it that much weight, which I will go into more with the upcoming paper, and which I will come back to below. At the end of the day, the way I would put it is that there was a credit bubble before 2008, and it was, at most, coincidental with or facilitating the supply driven price boom, and at least was orthogonal to it, having little to do with the most extreme increases in prices. And, believe me, I know some of you saw busloads of unqualified investors being shuffled into bad investments in 2006 with your own eyes. Just because you saw a witch at the edge of the woods, it doesn’t mean her spell is the reason for the bad harvest. (Again, I will address some of this in the next post, though it will probably be only for subscribers.)
Anyway, back to the main point: Now we get to the disaster part - Panel 3. And it became a disaster because of that mistake - mistaking the result of inadequate supply for the result of speculative fervor. The Fed continued to push for a slower economy, at least through 2008, and you can see here that the cyclical shift moved down another 8% from December 2007 to December 2011 (the intercept value in panel 3). But, now, there was a massive crackdown on lending, it having been blamed for the bubble, and so the price/income in the average ZIP code declined by 41% during the bust, more than wiping out the gains of the boom period.
This was universally accepted as a reversal of the 2003-2007 period, but it was not. In addition to being of a larger scale than the boom had been, when you look locally, the bust areas just don’t match up well with the boom areas when you look at it this way. I will go into that in the next post. Furthermore, as you can see in Panel 4, the bust was mostly re-reversed from 2011 to 2020, even though lending standards have remained tight, there is no real private subprime or Alt-A mortgage market, and much of the new activity is cash-funded. Note, also, that even in the 2011-2020 recovery period, in which mortgage rates continued to decline to all-time lows, the price/income level of the high-income ZIP code at the intercept continued to decline another 6%.
The reason the bust price collapse reversed isn’t because we created a new credit bubble. It’s because credit was never a necessary element. The same thing that was happening from 2003 to 2007 was also happening from 2012 to 2020, and we threw a credit market neutron bomb into it from 2008 to 2011 that left the homes intact while ruining the families that owned them. (And, there had been a bit of a credit bubble before 2008 that explains much less of all of this than the supply problem does.)
That brings us to the Covid boom, from March 2020 to May 2022, in Panel 5, which looks very much like the 2002-2003 boom in the national numbers, pushing home prices up by about 17% across income levels. The cyclical reversal, in Panel 6, is too young to show much of anything, and so we are left with the question of what to expect in 2023 and 2024.
First, though, let’s take one last look at 2011-2020. My recent paper goes into this. One reason - the main reason, surely - that price trends from 2020 to 2022 look different than price trends from 2011 to 2020 is that rent trends look different in those two periods. (Zillow rent data is only available for 2015 on.) As I have written about extensively, it is typical for a 1% increase in rents to be associated with much more than 1% increase in prices, so rents can explain most, if not all, of the general rise in prices. (As with the pre-2008 period, this doesn’t mean that there aren’t some other issues at play. In fact, some metro areas appear to have been bid up more than others, and there are surely some isolated price deviations that are not fully justified by rent trends.)
So, this leaves us at what to expect from here. I will leave the specific numbers for you to decide, but I will leave you with 3 thoughts:
At recent nominal GDP growth rates, a 17% cyclical impulse could be mostly reversed relatively painlessly with prices that are simply flat for a couple of years. In fact, previous housing busts in the US have not been that much shallower than the 2008 bust, in real terms. But higher inflation rates helped them be much less disruptive because nominal prices didn’t decline so deeply.
If you add up the intercepts from 2002 to 2020, there was a cumulative decline of 11%. So, much of the cyclical boost from 2020 to 2022 may have been a recovery from a less-than-normal starting point.
From 2002 to 2022, the average ZIP code has a 26% cumulative gain in price/income levels. If you’re waiting for a 26% price reversal in the aggregate national housing market, you’re going to be waiting for a really long time. The only way to get that would be to build millions of new homes, or to throw a neutron bomb in the credit market, and we don’t have any more bombs. Sub-760 credit score borrowers are still casualties.
Or there is a 26% price bubble because of low interest rates. Unfortunately, I am not talented enough to put together an analysis similar to that above that can identify it, so you’ll have to find another substack to explain that thesis to you. I’m under the impression that there may be a few.