Housing Tax Policy, A Series: Part 12 - Low income mortgages couldn't have caused the recession.

I have been outlining data from the Survey of Consumer Finances to double check the story that the bank delinquency data seems to tell, that the widely believed story about the housing boom and the recession is based on evidence that is much more flimsy than we might have thought.  In the previous post, I introduced data from the SCF that showed how the growth in homeownership came from high income households and that households didn't increase their debt payment/income ratios or their relative consumption of housing during the boom.  The evidence against the standard narrative is even more stark when we look at dollar levels, because, despite frequent implications to the contrary, low income households don't tend to take on nearly as much debt as high income households.

(In fact, the normal relationship is overwhelming.  Higher debt is associated with higher incomes, whether comparing over time, comparing cross-sections of a population, or comparing across nations.  When we hear a story about desperate American households accumulating debt just to keep their heads above water, our BS sensors should go up.  We should be thinking, "That contradicts practically everything we know."  I think this is from a confusion between thinking of states of disequilibrium and states of equilibrium.  We think of our own lives, and the times that are memorable to us are times where something went wrong, and we took drastic measures that might be unsustainable without some good fortune.  So, we remember that time when we lost our job, and we ended up with $15,000 on our credit card, and wondered how we'd ever get our life back in order.  But, we don't think about the fact that, 10 years later, when we were pulling in $100,000 a year, we took out a $200,000 mortgage.  Even within an individual life with periods of upheaval, the correlation between debt and economic success is massive.  The story that says an entire economy has been in a constant state of disequilibrium for decades is highly suspect.  Especially when it coincides with a massive bidding war on the most fundamental middle class asset.)

To the charts!

The first chart shows the mean level of debt per household, by income level.  In 1989, the average household held about $32,000 in mortgages.  By 2007, this had risen to about $82,000.  But this debt is overwhelmingly held by high income households.

The second chart makes this even more clear.  This is a measure of the total level of mortgage debt in the US.  The top 20% have never held less than 50% of total mortgage debt.  The top 40% have held roughly 80% of mortgage debt throughout this period. From 1992 to 2013, the bottom 40% of households held 7-10% of mortgages outstanding.  In 2007, this amounted to $857 billion out of a mortgage market worth nearly $11 trillion.

Keep in mind, the homeownership rate has been over 64% since the 1960's.  The homeownership rate of the bottom 20% of households has floated around 40%, and the 20-40% quintile has a homeownership rate over 50%.  So, this bottom 40% represents more than a quarter of all homeowners, and has for decades.  They just don't have much mortgage debt.

There are widely read authors - many, in fact - who claim that (1) the gains in the economy have been going to the "top 1%", and who also claim that (2) the reason the economy failed is because low income households have been living on massive levels of debt, just to get by.  They claim that the economy finally toppled due to the unsustainability of that process.

How could they be so wrong?  Let's forgive for a moment that those frightening graphs of rising household debt are showing debt that is 80% held by the top 40% of households.  Maybe low income households don't have high debt levels in absolute value, but maybe they are high compared to their own incomes.

Yesterday's post included this graph of debt payment / income ratios.  For 24 years, debt payments look pretty manageable and stable.

Here is a graph of total debt / income.  Aggregation is always dangerous, and here, for instance, we know that the aggregation within income quintiles hides the difference in leverage between renting and owning households.  But, that being said, mostly what is striking to me here is how uniform debt/income ratios are throughout the income ranges.  Debt levels did rise somewhat over the first 12 years, then they did rise quite sharply in 2004, concurrent with the sharp rise in home prices.  Then, they remained high after the crash.  But, the pattern is very similar among the entire bottom 90% (except for the bottom 20% after the crash, which is probably an artifact of formerly high income families moving down the income scale).  Are we to believe that 80-90% percentile households, with incomes over $100,000, are exhibiting the same distressed debt accumulation as the bottom 20%?  And, if this debt is a product of distress, how in the world does that result in a tripling of home prices over the course of a decade?

As I have argued before, the anchoring of real estate value on the future value of expected rents means that when long term interest rates decline, nominal home values must rise.  There is no way that households could have held debt levels steady in this environment.  And, the stability of debt payments is a signal of this issue.  The increase in debt is all mortgage-based.  There is some rise in non-mortgage debt in the bottom 40%, but some of the same issues could be at play to a lesser extent regarding other debt secured by durables, such as auto loans.

But, let's not lose sight of the broader problem here, the notion that mortgages which were pushed on low income households created a systemic crisis.

This next graph shows the net additional mortgage debt held by households, disaggregated by household income quintile, with 1992 as a baseline.  For instance, in 1992, the average bottom quintile household held $4,500 in mortgage debt.  By 2007, they held $12,500 in mortgage debt, a net gain of $8,000 per household (in 2013 dollars).  In total, households in the lowest quintile held $185 billion in net extra mortgage debt by 2007.  The second quintile held $324 billion in extra mortgage debt.  These together represents 9% of the new mortgage debt.  77% of the new debt over that time went to the top 40% of households.

So, if every single net new mortgage debt issued to the bottom 40% of the income distribution had defaulted, the total value of the losses, before recovery, would have been $324 billion.  Keep in mind that homeownership rates did not rise among these income groups during the boom, and that this income group had stable home ownership rates for decades before this.

Homeownership rates did rise for the median income quintile, from about 62% to 72%.  If every single net new dollar loaned to them during this period defaulted, that would amount to $833 billion before recovery.

What if the entire bottom 60% of the income distribution had sub-prime level 10% default rates on these net new mortgages.  That's a loss of $116 billion before recovery.  In dollar terms, that's 1.1% of mortgages outstanding.

The bottom 60% of the income distribution represents about 1/2 of all homeowners.  At the highpoint, subprime mortgages were around 20% of originations, reaching more than $600 billion per year.  That is almost the entire level of mortgages outstanding for households in the bottom 40% of incomes.  If every single mortgage origination going to the bottom 60% of households, by income, was subprime, it would have been a small share of subprime loans.

Between 2001 and 2007, when subprime originations increased from $173 billion to over $600 billion each year, total mortgages to the entire bottom 60% only increased by $908 billion over the entire period.  Mortgages to the top 40% (all with incomes over $100,000 $85,000 per year in 2013 dollars), increased by $3,527.  Mortgages to the top 40% didn't just increase more in absolute terms; they increased at a faster rate than low income mortgages.

There simply isn't any way that more than a small handful of subprime loans were going to low income households.  And, here, again, it's useful to look at the timing of events.  Here is a graph from this St. Louis Fed summary of mortgage originations.

Think about the timing here.  Short term rates were rising by 2004.  The proportion of subprime and Alt-A loans was still very low in 2003.  There weren't that many subprime loans originated when rates were low.  Subprime and Alt-A originations were increasing at high interest rates.  More than $2 trillion worth of these loans were originated at higher rates.

As the St. Louis Fed points out, many of these loans were for investment homes.  Here is a graph of mortgages for non-Primary residential real estate.  These subprime loans were going to the top 10% of households.

Thousands of 2013 $, per household

Next is a graph of non-primary residential mortgages, by age.  (Some of the 75+ group movement could be affected by outliers with small sample sizes.) These loans were funding investments by high income baby boomer professionals who were lining up long-term low-risk income for their retirement.  Even in 2004, homes were a good alternative to bonds.  Even now, with all that has happened, those homes over a 20 or 30 year period will earn decent income for their owners who held on to them.

In the aggregate, this is not a story of poor workers duped into overpriced homes with toxic mortgages.  This is a story, mostly, of very wealthy, very high income individuals making reasonable investments, given their alternatives, and then, after the nominal values of those assets were impaired, making the decision to put them on the banks, which again, given their alternatives, was reasonable.  Along with that, a wave of unemployment caused by the dislocation drove more Americans out of their homes because of income shocks.


Here is an estimate of homeowner equity.  (I show the equivalent measure from the Fed's Flow of Funds report.  It has a similar pattern, but with lower equity.  The debt levels from both sources are similar, since they are easily documented.  The difference appears to come from what survey takers think their own homes are worth versus what a grumpy economist at the Fed thinks it's worth.  The truth is probably closer to the Flow of Funds number, but the patterns should be similar.)  Equity did fall (leverage rose) from 1989 to 1992, then generally leveled off until the crash destroyed home values.  But, note that low income households tend to have very low leverage, because they are mainly divided between renters with no mortgage debt and retirees with years of equity.  Typical leverage levels increase as income increases.  The mortgage interest deduction might have a lot to do with this.  Someday, I'd like to see if this differs from the pre-1986 period.

Here, we see the net new mortgage debt, averaged per household.  In 2007, the median household ($47,000 annual income) had seen their mortgage grow by $36,000 while their home's nominal value had grown by $74,000 over 15 years (in 2013 dollars).

In this last graph, I think we have a window into the distributional effects of housing tax benefits.  Over time, to the extent that I can estimate the portion of home price appreciation that can be explained with tax policies, the distribution of the change in real estate values provides a sort of present value of future tax savings.  And, the distribution of real estate capital gains suggests that these benefits skew extremely toward high income households.  They simply hold much more real estate and much more debt than other households do.  This value would represent taxes saved.  At the other end of the spectrum, taxes on real estate without those benefits are paid by tenants, embedded in higher rents, and paid to the government through their landlords.


How Can We Be So Wrong

If you have an epileptic fit, and there happens to be a witch in the room, it really, really seems like the witch is the culprit.  The American public was convinced that homes and mortgages were a problem before the crisis even began.  The same political zeitgeist that led to the Fed's disastrous policy already had its scapegoats built in.  The tight money policy itself was a product of distrust of the housing and mortgage industry.  Yes, there were shady characters out there.  There were financiers who were too aggressive.  There were bureaucrats pushing dangerous programs.  There were investors using specious logic.  There were traders who thought the traders on the other side of the deal were making a bad trade.  (Can you imagine?!)  But, there are always all of these characters, in every industry, every walk of life, in good times and in bad times.  The reason we have the narrative we have is because bankers are today's witches.  If a banker is in the room, you don't need to confirm the connection.  We all know bankers cause crises.  What is there to prove?  And, once you are there, it is really easy to blame a financial crisis, of all things, on them.  The dots practically connect themselves.  There are plenty of anecdotes and empirical data to build a satisfying, if careless, narrative.  And being careless is the easiest thing in the world to do.  (Added:  Even if the witch is in the corner, doing incantations about seizures, that is still not evidence that the witch caused the seizures.  But, if you believe that witchcraft is a powerful problem, you will have a very hard time accepting this caveat.  This error is central to the human condition and to the greater part of human existence.)