There are so many ways in which we were tricked by our very own eyes and our very strong biases toward coming to pre-ordained conclusions about the housing boom. Generally, these biases led us to blame lenders and speculators when prices were rising sharply. One way our predispositions seemed to be confirmed was the appearance of a hot mortgage market in the midst of the boom.
Now, I like to make the point that any context with rising prices, for any reason, is bound to have a hot lending market. You can always put obstacles in the lending market, which is what we have done since 2007, to keep prices lower, but this is different than pushing prices up through lending. It is easy to see how a household who is obstructed from credit will not create demand for homeownership (although, they will still create demand for housing). But, if access to credit creates unsustainable or irrational demand that pushes prices too high, we should expect this to lead to some mitigating factors - the natural drop of demand among buyers who see the market as too expensive, the natural decision of current owners to exit the market, etc. There might be some positive feedback mechanisms from loose lending that cause demand to shift up, so that there is some level of positive feedback in hot markets, but it seems to me that the data says these mechanisms have been overstated, and they certainly weren't causal. In other words, the case for tight lending causing prices to retreat below a sort of non-arbitrage price level is easier to make than the case for loose lending causing prices to rise above a reasonable estimate of a non-arbitrage price level.
One reason that credit seemed important was that mortgage debt was rising during the period. But, according to the Survey of Consumer Finances, the typical mortgaged homeowner had leverage that was declining during the boom, and was only slightly higher than the pre-bubble level even after the bust. (This is probably due to owners overestimating their home values during the bust, in part.) It appears that leverage was mostly rising during the boom because there were fewer households who were unencumbered.
This doesn't quite match the conventional wisdom, because we imagine the housing ATM to mostly apply to households who had Loan To Value (LTV) levels of, say, 80%, getting home equity loans and moving their LTV to 90% or 100%. It doesn't really make sense that the housing ATM would mainly lead to some households taking on mortgages who didn't have any to begin with.
What I have found is that first time homeownership was probably slightly elevated from the mid-nineties on (though the financial character of buyers did not change much throughout the bubble, and if anything was strengthening). This led to rising homeownership, which peaked in 2004. From then until 2006 or 2007, the trend reversal in ownership was coming from a rise in permanent sellers - owners exiting ownership. Then, in 2006 and 2007, the collapse of the mortgage market and public policy pressures against marginal lending caused first time buyers to fall. Selling pressure remained strong, and homeownership rates really began to collapse as we now had weak entry and strong exit rates into homeownership.
So, in that 2004-2007 period - the subprime boom period - I think the primary influence on leverage was not the housing ATM factor, but this shift in ownership - a slightly strong first time buyer market (though no stronger than it had been for a decade) and a strong sellers market. The sellers tended to be long term, unencumbered owners. So, there was a tradeoff of owners without a mortgage selling to first time homebuyers who were leveraged, just as first time buyers have generally always been leveraged in the modern US housing market. The stability in leverage levels, which then began to shoot up when prices stabilized and began to decline, was from a shift in owner composition.
We can see this in the SCF measures of homeownership, both with and without a mortgage.
Looking across all families, mortgaged ownership increased by about 8% from 1995 to 2007, but total ownership only increased by about 4%. (Note that the collapse of the credit market lagged the outflow of owners. From 2004 to 2007, mortgaged ownership increased by just under 1% while unmortgaged ownership fell by more than 1%, leading to a net decline in ownership during the private securitization boom.)
Notice how there is a rise in mortgaged ownership across incomes, but this only led to a rise in total ownership in the top 60%. That 4% fall in unmortgaged ownership seems to cross income levels. It is matched by a 4% increase in mortgaged ownership in the bottom quintiles, that leaves ownership flat, and in the middle and higher incomes, mortgaged ownership increased by more like 10%, leading to a net increase in total ownership.
In the lower quintiles, the rise in mortgaged ownership was entirely offset by a fall in unmortgaged ownership. This suggests the possibility that at least part of the story was a rise in low income lending which is hidden in the aggregate numbers by an outflow of former low income owners.
But, if we look at ownership by age, I don't think this bears out in a significant way. Rising mortgaged ownership is highly correlated with age. And, ownership in the lower two income quintiles are also highly weighted to older age groups. The rising mortgaged ownership in the lower quintiles must be coming mostly from the use of leverage among lightly encumbered older, long term owners. (By the way, the length of tenure for owners in the bottom quintiles, especially in Closed Access cities is very high - typically over 20 years.) And young owners are largely in the high income quintiles, so the rise in mortgages to younger age groups must have been mostly to buyers with high incomes.
So, that rise in mortgaged ownership among the low income quintiles is a bit of a mystery. What was causing so many older owners to retain mortgages? Was it just a cultural shift with baby boomers? Products like reverse mortgages?
As you might suspect, I think migration in and out of Closed Access is an important factor here. ACS data which has details and statistical power regarding these issues only goes back to 2005. The migration patterns in 2005 were very strong, but while their scale was unusual, the direction of the trends should be similar to what had been happening for a decade. We might assume that ownership patterns in Closed Access cities in the mid-1990s were much closer to the ownership patterns we see in the rest of the country, and that the sharp differences that have developed since then are largely the product of the cost-based segregation that has developed. (Although, this isn't borne out by state-level ownership rates, which seem to have been below the national average in California and Massachusetts for some time, so maybe my intuition fails me here.)
Among mortgaged homeowners, in the massive outflows that were happening in 2005 and 2006, what we see is a sharp net outflow of households that were young and households that had low incomes. For the general population of Closed Access MSAs, the bottom 60% of households were moving away, on net, at a rate of about 2% a year, at the height of the bubble. But, for homeowners, it was more like 3%. Keep in mind, middle class and poor households don't tend to own homes in Closed Access cities, so in absolute numbers, this is a small portion of the total population of migrants. But, it gives us a window into these shifting ownership patterns.
Let's take a look at ownership patterns at the top income quintile and the 2nd quintile. Here, the orange bars are Contagion cities, the red bars are Closed Access cities, and the green bars are everywhere else. The dark bars are 2005, the light bars are 2014, the solid bars are mortgaged and the striped bars are unmortgaged owners.
Notice, in the top income quintile, ownership in Closed Access cities is similar to other cities. At the top end, Closed Access residents have to make some compromises, regarding the size of the unit, etc. But, in terms of budget and ownership status, their spending and ownership patterns are maintained. As we move down the age groups, Closed Access ownership declines, because the older owners were more likely to buy their homes before the bubble, and younger households have been priced out for 20 years. Also, note how sharply mortgaged ownership has collapsed for younger households in the highest income quintile.*
Now, look at the 2nd income quintile. This is a different story. Ownership among older households is still fairly similar across cities. These are generally households that have owned homes for a long time. But for working age households, now there is a sharp difference between Closed Access and other cities. Among younger households with moderately low incomes, ownership in Closed Access cities is negligible. Across working age groups, it runs 20% or more less than in other cities.
There would be a natural reaction from existing owners to rising prices in constrained cities to capture those capital gains, partly through selling and moving, and partly through taking out mortgage debt. The migration and ownership data suggest that selling is a significant factor. Where capital gains would be caused by supply issues, of course they would be captured in a number of ways, some of which include using the home for collateral. As we move down the age and the income distribution, Closed Access ownership, both mortgaged and not mortgaged, falls farther below the levels we see in other places. This suggests that, on net, supply pressures have a sharp effect on the quantity demanded of homeownership. To the extent that there is a notion about the bubble that expanding credit creates its own demand, pushing prices into an ever-climbing spiral, that appears to be a broad overstatement. In terms of quantity, it seems clear that among existing owners, a city with rising prices will have lower demand for housing units than a city with stable prices, even as some of those households capture their capital gains through home equity borrowing.
My shorthand for the housing bubble in the Closed Access cities is that non-conventional loans weren't putting middle income and poor households in homes; they were facilitating the sale of homes from wealthy old homeowners to high income young buyers. But, looking at SCF national data that goes back to the 1990s or AHS data on cities that goes back to about 2000, there isn't an obvious shift in ownership patterns.** There certainly isn't any shift that is a counterpoint to the sharp downward shift we have seen in young, high income homeownership in Closed Access cities since the bust.
In the end, I don't think there was a shift. There was a slight increase in ownership among the young, nationwide, but there was nothing peculiar about what was happening in the Closed Access cities - the "bubble cities". The non-conventional loans were simply facilitating the same sorts of purchases from the same types of demographic groups that had always bought homes. To the extent that they were critical in facilitating rising prices, they were allowing normal transactions at price points that had reached levels too high for conventional loans to handle. They were simply accelerating the normal process of in and out migration from Closed Access cities. The slight drop in Closed Access population during the boom appears to be about half-explained by the difference in family size between small families that move in and larger families that move out. Thus, the migration acceleration led to population decline.
* Consider what this tells us regarding causality. If creative financing caused the bubble, then when those mortgage markets collapsed, we should have seen prices falling and ownership remaining fairly level for high income groups. Instead, Closed Access ownership among high income young households has collapsed. The reason it has collapsed is because high prices are caused by supply constraints which push rents up. Housing expenses in those cities take a large chunk of any household's budget - even a household in the top income quintile. Supply constraints caused local housing costs to rise above the norms we use for conventional mortgages. Creative mortgage terms were allowing high income buyers to purchase homes outside the norms of conventional mortgages, which is the only way to purchase a home in Closed Access cities. Young, high income homeownership has collapsed in Closed Access cities, in spite of some retraction in prices, especially in lower tiers of the market, because creative mortgages weren't creating the price problem; they were just facilitating ownership in cities that had high costs. Now we have imposed constraints on creative financing, so young households with six-figure incomes are trapped in the renter's market and prices are too low to entice the remaining long-tenured owners to sell.
** Upon review, I think I made an error here. The income quintiles I use are based on national quintiles. The bottom two quintiles in the Closed Access cities only amount to about 16% of Closed Access population each and the top quintile amounts to about 30% of Closed Access population. If both renters are owners are leaving, the relative number of low quintile owners could be declining even if the homeownership rate for those quintiles isn't falling that much. So, the homeownership rate in the Closed Access cities could have risen during the boom because of changing composition to more high income households, even if the rate within each quintile, as I have defined them, remains fairly level or declines slightly. So, my shorthand would be true, even with the homeownership rate data that we see. In fact, this should have been obvious to me as a product of thinking about these things through the lens of national income segregation.