The damage from the mortgage moral panic is not subtle.
Charlotte, Dallas, Denver, and Houston: Case Studies in Self Harm.
I’m continuing here with the theme from the earlier posts, to review how this model picks up various aspects that affect home prices: Yes, there was a cyclical housing boom before 2008, but inadequate supply had more to do with high prices at the time than it has been given credit for. And, the bust came in two steps: (1) The cyclical boom died down from 2005 to the end of 2007 (which is also when construction was declining). (2) Price inflation from inadequate supply was increasing during the first phase (the non-disruptive phase) of the contraction, and then, suddenly, declining prices switched from a shift in the cyclical element to a shift in the supply element.
Now, obviously, we didn’t suddenly start building a bunch of homes in 2008 to bust the supply problem. But, we did massively change lending standards. And, so, I attribute a lot of the change in prices from 2008 onward to that credit shock. In places where there aren’t enough homes, they become too expensive for families with lower incomes, and we “solved” that problem by denying those families mortgages. Voila! Affordability! Figure 2 shows the cyclical and supply factors after adjusting for credit conditions.
Basically, we had a "housing bubble” from 2002 to 2006, and that “bubble” ended by late 2007. The unwinding of the bubble was too uneventful, however, to satisfy our demand that *lessons needed to be learned*, so most of the bad stuff associated with busting the bubble came along with the sharp drop in the supply component, after the “bubble” had actually reversed. You really have to crap all over the economy and the housing market to get the supply component to reverse like that, but we managed to do it.
Today, as an initial investigation into that problem, I will look at Charlotte, Dallas, Denver, and Houston - 4 metro areas that took no part at all in the housing boom that peaked in early 2006 in the rest of the country. None of those cities had unusually high price appreciation during the boom years.
Figure 4 shows their price components without the credit factor.
Just a reminder, there is not really anything theoretical in these measures. I ascribe a theory to the supply component that it is related to inadequate housing in my analysis of these markets. But, these are simply measures of relative price levels in these cities. “Cyclical” is a relative price level that is shared across the metro area, and “Supply” is a relative price level that is income sensitive. (It is positive if homes are relatively more expensive in ZIP codes with lower incomes. I have a series of papers coming out on this over the next several weeks that build the case for ascribing that price differential to supply constraints.)
Neither Dallas nor Houston had price/income levels until recently that were far from neutral. They each were cyclically a little low before 2008, which was countered by a slightly above neutral supply component.
Charlotte, oddly, had a bit of a cyclical boom from 2006 to 2009. You can see that in the Case-Shiller indexes shown in Figure 3. And, the supply component was low, but was just starting to point up a bit at the end of 2007.
Denver is a little different than the others. Denver had a bit of a growth spurt in the 1990s, but by the 2000s, it was pulling back a bit. You can see that in the Case-Shiller index numbers from Figure 3. You can see it in Figure 4, where the cyclical component is a bit low before 2008. The supply component is high, suggesting a pre-existing supply problem, but Denver was bucking the national trend, and the supply component was moderating during the 2000s. Construction activity was moderate in Denver at the time, and vacancies were a bit elevated. Pre-crisis Denver is a good example of a city overcoming a supply problem. Somewhat elevated vacancies will be a part of that process.
The cities I looked at in the previous post are a bit more complicated, with home prices dropping 40% or more. LA and Phoenix did have booms before the collapse. Atlanta, on the other hand, was like the cities here, and had no price bubble. So while its collapse was stronger, it poses the same mystery of why a collapse happened.
These 4 cities all saw 20% declines or more, mostly coming from a decline in the supply component. And, while every city has its own idiosyncrasies, they all share this single, very clear shock in the supply component at the beginning of 2008.
Here is what these 4 cities look like with the Credit component added. The Credit component scales with local income. Figure 5 is the chart of price components for the average ZIP code in each of these cities, and so, by construction, the Credit component affects them all equally. It would have a larger scale in ZIP codes with lower incomes in each city:
Again, as with the other components, my Credit component is just measuring something that happened across the country. I attribute it to tightening credit standards, but if you have a better theory for something that immediately and permanently affected home prices in an income sensitive way, you’re welcome to it. That theory is important for public policy arguments. For my purposes here, of giving you insight into home price expectations, it doesn’t matter so much. Something caused home prices to collapse in a way that was sensitive to incomes across every single metro area, regardless of pre-existing conditions. In other words, there is a collapse, discernable in the data, that was unrelated to the bubble but was fairly uniformly related to local incomes.
This Credit component is a big deal. I’ve just stuck a big giant trend change into all these charts that you simply don’t see anywhere else. I’m not aware of even a natural disaster that created that sharp of a shift in any component in any single city, and yet here is a massive shock that hits every city at exactly the same time.
And, so what I find especially striking here is that when you apply what I call the Credit component to the model, this massive adjustment to price trends that is mechanically the same in every single city makes the Supply trend in these 4 cities from late 2007 to early 2008 as straight as an arrow. Where it was slightly rising, like in Charlotte, it kept rising gently. Where it was lightly declining, in Denver, it continued.
Now eventually, the Supply component follows a typical, though not exactly parallel pattern in different cities, sort of flattening out for a few years before starting to climb around 2014. I’m sure I’ll have more to say about that in future posts. But those moves happened slowly over time, which is what supply constraints will typically do.
This is especially noticeable if I display prices in ZIP codes with lower incomes. Here are representative ZIP codes in the four cities with adjusted gross incomes of $50,000. Here’s what the supply component looks like in the raw data, and what it looks like with the credit adjustment.
It’s amazing that for 14 years, conventional wisdom has treated the breaks in those yellow lines as just the natural course of a housing cycle. People like to blame the purported bubble on complacency. The popular theory is that people became convinced that home prices never go down, and so they dangerously bid prices up too high.
There is a good reason to believe that people think that way! In most places at most times, home prices don’t go down! It’s very unusual! Home prices don’t just suddenly drop by 20%. And, yet, at the beginning of 2008, suddenly, in countless cities across the country where housing had been downright boring, in defiance of all the upheavals and volatility in some cities that was capturing our attention, suddenly, in all of them, at exactly the same time, there was a shock that knocked 20% off the average home value, and knocked much more off of the homes in poorer neighborhoods (systematically, in a similar way, at a similar scale, in city after city).
The conventional story of the boom and bust tells a story that’s basically all about the Cyclical component. There was a housing cycle that busted. Seems so obvious.
That story is wrong. But, the worst part about it being wrong is that it satisfied our thirst for answers, and so the Credit component (or whatever you think suddenly caused working class homeowners in boring markets across the country to loose massive amounts of net worth) is simply missing from the conversation.
I may have missed something somewhere, but all the papers I have read treat price declines explicitly as if they are mechanical reversals of the excesses that came before. It’s one of the 1,000 ways that the oversupply myth pollutes the academy. Where price declines weren’t preceded by unusually high prices, they are attributed to oversupply. It’s a big giant kludge to explain Dallas and Charlotte when you haven’t added that blue line to you model. But, nobody bothered to confirm it. Everyone thought they knew there were too many homes without needing to confirm it.
Oversupply could obviously explain a price decline. Right? In fact, that is central to my model. A massive addition of supply, in my own model, would cause prices to drop proportionately to incomes. But, for it to happen the way it did, a few million new homes would have needed to have been suddenly built in early 2008. To the contrary, new construction had been in steep decline since early 2006.
At least the concept makes sense. As “oversupply” has been used conventionally, it could explain a 10% drop, 20%, 40%. When someone mentions that an oversupply of homes was a cause of the financial crisis, they mention it as a sort of McGuffin. Nobody ever says, “Well, I thought an oversupply could lead to a 10% drop in prices, but when prices dropped 20% or 25%, I started to doubt that oversupply could explain it.” It was like a magic button that tied all the lose ends of all the studies of our demand-side excesses together. Nobody attempted to quantify it. It’s hard to quantify a phantom. So, it was available for filling in any explanation of declining home prices of any scale at any time.
I wonder if the availability of housing supply as a malleable kludge variable was the result of the complete lack of evidence for oversupply. If there had been some evidence that could be applied with some local precision, it could have been quantified, and then there would have been a literature to cite, which would provide guard rails. But, maybe a lot of economists tried to quantify the oversupply, failed to find any reasonable signals from vacancies, construction activity, etc. that could explain the collapse in places like Dallas and Charlotte, and gave up, deciding that they were just personally not up to the task of measuring the thing-that-everyone-knew. And, so it remained, as an unquantified kludge variable that was a much more dangerous rhetorical and quantitative crutch than it could possibly have been if there had been some weak evidence to quantify in some way - even in a sloppy, overstated way.
As a public policy issue, this is tragic. As an asset manager or trader, you can be thankful that the chief economist of your counterparty is armed with a legion of poorly specified literature.
I will post a Part 2 of this for paid subscribers where I will touch on the shape and scale of the Credit component and look at Charlotte through the present to see how things are proceeding more recently.