Thursday, February 11, 2016

Housing Part 116 - The 2006-2007 Mystery and Synthetic CDOs

As the bubble story goes, synthetic CDOs were an insidious new creation that helped create the subprime bubble and allowed mortgage originators to keep originating mortgages with marginal borrowers.  You might have already guessed that I will disagree.  But, this one is so complicated it almost broke my brain, so I have put it below the fold.

One detail that I think is frequently missed is that the explosion of ABS CDOs (packages of fixed income securities with mortgage backed securities as collateral) came after the peak of homeownership and generally was after the peak in home prices and housing starts.  And, Synthetic CDOs (packages of fixed income securities that generally pair collateral like Treasury bonds with Credit Default Swaps (CDS) to mimic the cash flows of a cash CDO) nearly all were developed after the collapse of housing starts and when subprime originations were sharply declining.

I think the timing of Synthetic CDOs is especially important.  They are frequently described as a mechanism that was used to originate more questionable mortgages that were bound to default because they gave speculators a pure way to benefit from mortgage defaults.  The CDS's that enabled them are also frequently described as a mechanism that allowed banks to make money from pure financial engineering instead of investing in real assets.  But, I will argue here that, as with so many interpretations of the housing market at that time, these interpretations have it completely upside down.  Synthetic CDO's happen to have developed as the number of new originations was collapsing in late 2006 and 2007.  And, they still required the use of capital.  There are countless ways that capital is filtered through American financial markets to fund real investments.  Financial engineering may complicate that picture, but the capital went somewhere.  This capital funded "real" investments just as much by buying treasuries as collateral for these CDO's as it would have if savers had simply bought treasuries in an unengineered transaction.  And, mortgage spreads did not rise until after home prices began to collapse in late 2007.  This suggests that, at the margin, the explosion of Synthetic CDO's didn't significantly affect the flow of capital to mortgages, relative to treasuries.  (After writing this, I realized that this is another case where the popular narrative includes two things that are mutually exclusive, yet both wrong.  Either synthetic CDO's funded home purchases for new marginal households or they created profits for banks without investing in anything real.  They could hardly do both at the same time.  As it happens, the evidence suggests that synthetic CDO's did little to boost actual subprime mortgage growth, but the capital did flow into treasuries, which, on the margin, increased the basket of invested capital just as any other marginal new capital would have.)

The mystery is, why didn't these investors just fund mortgages?  Why resort to these Frankenstein mortgage securities?  The conventional explanation is that these were created for the benefit of the shorts, who enabled the last phase of atrocious underwriting.  I will argue that by late 2006, the number of mortgage borrowers were being reduced by contractionary monetary policy, and that the sudden sharp rise of these Synthetic CDO's was a product of that contraction.  Capital markets could still fund new mortgages, but the borrowers were drying up.  I apologize in advance.  This is all still a bit mysterious to me, and the path this post takes will be long and complicated.  I don't know how to simplify it at this point.

In late 2005, the Federal Reserve pushed short term rates high enough that the yield curve flattened and then inverted.  This is a classic signal of a coming contraction in lending and production and a coming recession.  In this cycle, it was two years before the contagion reached the broader economy, but we can see the effect immediately in the housing market.

I think it is interesting that the pattern we see here is a pattern common in contractions that have come after rising Fed Funds rates led to flattened yield curves.  Housing starts tend to drop, as a leading indicator, and rent inflation tends to rise, going into the recession.  This seems like an important facet in the debate over Fed policy, and the problems caused by inflation targeting.  If a common cause of business cycle volatility is pro-cyclical and overzealous Fed policy, then is this pattern a sign that late in expansionary periods the Fed tends to overtighten, causing mortgage financing and home sales to contract?  Since this policy tendency creates a contraction in housing supply, rents rise in response to the supply shift, and the Fed takes the higher inflation as a signal to tighten even more.  We might be able to explain the drop housing starts as an early sign of inevitable contractions, and we might be able to explain the rise in rent inflation as a late sign of the inflationary pressures of over-expansion.  But for them both to repeatedly appear at the point where the yield curve flattens as a result of Fed policy decisions seems to me like a story of overzealous monetary policy.

Back to the previous graph, new home sales began to sharply fall at the end of 2005 and at the same time home prices flattened out.  Soon after, rent inflation begins to rise, which suggests to me that the collapse in new home sales was not the result of over-supply.

But, here's the strange thing.  If this was the result of a flat yield curve curbing bank lending, then shouldn't we see a decline in lending, too?  And, we already know that homeownership rates peaked two years prior and that by the end of 2006 subprime lending was collapsing.

But, household mortgage levels were growing until late 2007.  Annualized mortgage growth in 3Q 2007 was still 7%.  That is lowish, compared to historical ranges, but considering the striking collapse in housing starts, it is strong.  And, looking back at the graph above, homeowner leverage began to skyrocket.  I have previously blamed the run-up in homeowner leverage on falling prices.  But, I was being careless about that.  The run-up in leverage began even while prices were holding steady.

Source: Financial Crisis Inquiry Commission
But, housing starts were collapsing, right? So this seems like it is a strong confirmation that home buyers who had over-extended themselves were grasping at home equity credit in order to keep making payments.  When home prices were rising, they could do that while pretending that they were still accumulating equity, but now that prices had plateaued, the unsustainability of this game became clear.

But, this gives us a clear set of trends that we should expect to see in the late 2006 and 2007 time period.  We should see a rise in subprime originations for refinancing and a decline in originations for purchase.  But, we see the opposite of this.

Subprime originations dropped sharply in 2007.  By the first half of 2007, the level of mortgages in private pools peaked and total non-agency mortgages outstanding began to decline in nominal terms.  With the flat yield curve, bank lending was flattening out, too.  By this time, lending was increasingly happening through the GSE's.

And, as we should expect, refinancing correlates with interest rates.  Mortgage applications for refinancing had been low since the Fed began raising rates in 2004, and remained low, only spiking again when the Fed finally began letting rates fall in 2008.  The same is true for home equity lines of credit at commercial banks, which leveled out after 2005 (graph is below).

But, mortgage applications were strong for purchases.  Purchase applications dropped slightly when housing starts began to collapse, but they remained near the boom levels until 2008.

There is further confirmation of this pattern in measures of equity withdrawals (MEW).  Here is a graph of MEW from Bill McBride at Calculated Risk.  It is true that in 2005, equity withdrawals were at a high level.  But, at the same time that new home sales began to collapse, this figure also began to collapse.  The rate of decline seems a little surprising, given that mortgage levels were still growing while home building was falling sharply, and homeowner leverage was rising sharply.

In the post I grabbed the chart from, Bill McBride generously makes the spreadsheet available that he uses to estimate MEW.  From that spreadsheet, I have charted estimated equity withdrawal and mortgage originations for purchase of new homes.  We can see that while equity withdrawal was falling sharply, purchases remained strong.

And, among those purchase mortgages at the GSE's, an increasing number of them in 2006 and 2007 were for 1st time buyers!

From spreadsheet available here.
So, we have an increasing number of first time buyers, growing levels of mortgage debt, and sharply rising homeowner leverage.  But, at the same time, we see collapsing new home sales, collapsing subprime originations, and falling homeownership levels.

The evidence, itself, doesn't seem to fit a coherent narrative.  It certainly doesn't fit the narrative of increasingly predatory lenders  and over-extended homeowners.  But, if purchases were strong and borrowing was increasingly going to first-time buyers, why were new home sales collapsing so sharply?

We might think that home prices finally got so high that first-time buyers couldn't afford them any more, and only existing owners with their inflated equity levels would be able to buy homes at the peak prices.  This intuition is undermined by the fact that housing starts were collapsing in the Open Access cities.  If affordability was the problem, then we might have seen prices or starts collapse in San Francisco or New York City.  But, prices and starts were collapsing where homes were still affordable.  And, clearly the upsurge of first-time buyers undermines that intuition.

There must have been a decent amount of churn between new homebuyers and existing homebuyers at the time, and I think this starts us on a path of figuring out what was going on here.  The equity extraction we see in the graph above, after 2005, wasn't so much a product of refinancings as it was of this churn between existing owners and new buyers.  Home equity lines had plateaued in late 2005.  Desperation borrowing wasn't happening in 2006 and 2007.  It happened in 2008, when monetary deprivation had become severe.

Here is a graph of the Fed Funds Rate, 10 Year Treasuries, and 30 Year Mortgage rate, in blue and green, and year-over-year currency growth in red.  Even in 2003-2005, currency growth was low relative to earlier periods.  Even during the peak of the housing boom, the Fed was already tightly controlling currency growth to compensate.  That is one reason why inflation remained low throughout the period, even when credit was expanding.  In this graph, we can see how sharp drops in currency growth tend to signal when the Fed has pushed rates too high.  Usually, the Fed reacts by adjusting their monetary posture, and currency growth begins to recover.

But, currency growth was already low in 2005, and it continued to drop after the Fed Funds Rate was pegged at 5.25% from early 2006 to late 2007.  Even as this transfer of households between new buyers and old home owners was pushing mortgage credit levels higher, the harvesting of equity was not enough to counter the posture of the Fed.

By mid-2006, domestic investment, nominal, and real GDP were all falling, and were at levels associated with previous recessions.

Here is a log-scale graph of credit and currency growth, going back to the early 1970s.  Both currency and bank credit had grown along a trend line of 2% per quarter since then.  There was one downward shift in credit in the early 1990s, which coincided with a previous sharp downward real valuation of homes.  In this graph, we can see how sharp the decline in currency growth was, back to late 2005.  Was the remaining credit expansion in 2006 and 2007 the last gasp of a greedy banking sector running out of dupes, or was it a desperate attempt to create scrip in an economy starved of currency?  By 2009, the breakdown in the mortgage and housing markets meant that both currency and credit levels had shifted significantly below trend.

I think that the persistent pegging of interest rates at a level above the natural rate meant that the intrinsic value of homes as financial securities fell sometime in late 2005.  (By this, I mean, the value of home ownership as a discounted present value of future rent payments.)  But, there are two basic regulators of home prices.  The first is the present value of expected rent payments, which is sensitive to real long term interest rates.  The second is the cost of building the structure.  In Closed Access cities, when constricted supply raises rents, the high intrinsic value accrues to the land.  But, housing has its own zero-lower-bound issue, and I think the Fed hit its first zero lower bound in 2006.  In Open Access cities, prices weren't particularly high.  When artificially high interest rates caused the value of homes to fall below the cost of building the structures, there was no land value to give up, relative to other uses.  The price floor established by building costs was higher than the intrinsic value from future rent payments.  And, it was the Open Access cities where builders were building.  This is why the initial response of the housing market to high interest rates and low currency growth was for housing starts in the low priced areas to collapse.

And, I am going to make a statement here that sounds bolder than it should, because observations about capital markets are imbued with a fat dollop of attribution error.  Households, given access to credit, may not lap it up like drug addicts on a bender.  In the aggregate, households with access to credit, and even households sitting on piles of freshly harvested capital gains, faced with home prices above their intrinsic values, might choose to forego that credit.  Shocking!  Call it decentralized macroprudential regulation.  Households naturally compare the relative merits of renting vs. owning, and, on the margin, maybe, just maybe, they adjust their purchasing decisions to account for relative values.  As early as 2004, mortgage demand was declining, and by 2006, demand for commercial real estate loans was falling.

And, what might we expect to happen if interest rates have been pushed above the neutral rate?  What might we expect if monetary and credit policies were creating asset deflation?  Wouldn't we expect there to be an excess of lenders and a shortage of buyers?  Isn't that the problem synthetic CDO's were solving?  Maybe, synthetic CDO's were the result of the first phases of monetary contraction.  Remember, their explosive growth coincides with the collapse of housing starts, the pegging of the Fed Funds Rate at 5.25%, and the decline in mortgage growth.  And, it appears to also have coincided with a large churn of households out of homeownership.

Synthetic CDO's clearly had nothing to do with the run-up of the housing boom.  I am flabbergasted by the number of facts that are pressed, pushed, squeezed, and shimmied into the bubble story.  How many pages have been printed about how synthetic CDO's were part of the bubble?  Looking at this carefully, how can they be interpreted as anything but a reaction to a lack of demand from homebuyers - a signal that even in 2006 there were incongruities in credit markets preventing savers and investors from matching?  And, at least in hindsight, shouldn't we notice that the sharp decline in housing starts in every single major city in the country that was happening at the time synthetic CDO's were expanding is a significant piece of evidence that synthetic CDO's were just one more sign of monetary contraction, not accommodation?

By the way, this is why prices in the Closed Access cities fell more steeply, but prices and, especially, housing starts there have recovered more since the bottom of the crisis, because marginal alternative uses are less of a source of price regulation there.  In the Closed Access cities, political intransigence is the constraint to housing.  Very valuable little cubes of air hanging over Manhattan aren't sitting empty because it is more valuable to grow hay in them.  They are sitting empty because political forced prevent them from being utilized for anything.  If political permission can be attained for new housing, even now, in the Closed Access cities, it is worth it to take advantage of it.  Prices in Texas need to recover more in order to induce more homebuilding there.

I'd like to make one other point, which is, admittedly, speculative.  There were, famously, several large players taking short positions in 2006 and 2007 on these housing securities.  Synthetic CDOs were a bridge between housing returns, which were rising (because rents were rising and prices were falling) and treasuries.  I have noted that there should be some relationship between long term real treasury returns and home equity returns, and that to the extent that we have data on both, they appear to have followed similar trends until late 2007.  I think the main effect of synthetic CDO's may not have been to fund mortgages, but instead was to boost the returns to treasuries in 2007.  Those shorts were taking tremendous losses until the Credit Default Swaps started paying off.  They were paying a premium to treasury holders in order to take their short positions.

Their stated positions, and the conventional wisdom about their positions, is that they were positioned to gain from the inevitable collapse of unsustainable housing prices.  But, what their exposure really was to was the persistence of devastating monetary policy.  Synthetic CDO's began to rev up as the yield curve inverted in 2006.  The breakdown in the relationship between real treasuries and the implied housing yield happened in late 2007, when home prices really started collapsing and the shorts on the synthetic CDO's received their huge payouts.  One of the functions financial markets serves is to arbitrage across time.  What these shorts were doing was arbitraging the future disequilibrium of housing returns and real treasury returns (the gap we see now).

The shorts thought they were positioned to gain from borrowers defaulting in an overbuilt housing market.  But, if that had been the case, rents would have been declining in 2006 and 2007, not rising, and implied yields to homeowners would have fallen slightly along with long term real treasury rates.

Interest rates in 2006 and 2007 should probably have been 4% or less. And, the Fed should have been signaling a commitment to stability.  If those policies had, indeed, led to stability, the synthetic CDO shorts would have taken tremendous losses, as well they should have if the country hadn't been chanting, "What do we want? The worst housing crisis since the Great Depression! When do we want it? Now!"  I would say that the insight that made the synthetic CDO shorts billionaires was not an insight into housing markets; it was an insight into political inertia.  My speculation is that the demand for treasuries that was created by synthetic CDO's is one reason that the yield curve didn't invert more than it did, and that, after 2006, there were two possible outcomes - the one we experienced, where home yields shot up and treasury yields collapsed, and the synthetic CDO shorts pocketed some gains, or the alternative where the Fed introduced liquidity in 2006 or early 2007, the housing market healed, home prices recovered, real treasury yields and housing yields continued to move together, in the 2% range, and the synthetic CDO shorts would have eaten large losses.

And, keep in mind, I am not talking about housing markets in San Francisco healing.  Prices there were a supply issue.  I'm talking about increasing home prices of just a few percentage points in places like Texas, that would have triggered a housing start recovery in the Open Access cities.  It is important to keep separate the supply and demand issues.  Compared to the outrageous prices that supply constraints create in the Closed Access cities, the effects of demand shifts where capital flows are liberalized and houses can be built are extremely small - imperceptible by comparison.

Notice, way back at the first complicated graph, there are two rent inflation measures.  The dark red measure is for owner-occupiers and the light red measure is for tenants.  They both rose together in 2006, but as housing starts continued to collapse in 2007, note that tenant rent remained persistently high, and remained high until 2009.  Households withdrawing from homeownership were bidding up the rents of tenant housing, because, on the margin, with artificially high interest rates, renting had become preferable.

Eventually, the collapse of new homebuilding had absorbed as much of the negative shock as it could, and prices of existing homes began to decline sharply also.

Here is a graph of population changes, by city.  The housing boom was facilitating a shift of households out of the Closed Access cities.  When credit was flowing, households were using it to move to Texas and Arizona.  The relief in pressure to utilize the existing housing stock meant lower utilization of the existing stock and population expansion where we could build new homes.  When the flow of credit was crimped and new home building was cut, population trends in the different cities converged.  Now, there is less of a housing outlet in Texas, and households across the country have to be stingy about their real housing consumption (although this is paired with high rent inflation, because they would prefer to consume more housing).  So, population in the Closed Access cities is growing at near the national average, even though this is unsustainable given the anemic growth of their housing stock.  But, for now, our broken mortgage markets prevent us from creating a functionally expanding housing stock anywhere.

Ironically, the high residential investment before 2006 was helping American households to reduce their housing expenses, and when we put a stop to that expansion, we trapped them in cities that have imposed unsustainable costs on them.


  1. Okay, so the sequence here is:
    1. closed-access cities limit building
    2. severe rent inflation in closed-access cities
    3. highish rent inflation nationwide (it's an average after all)
    4. higher CPI inflation than the Fed desires (rent inflation is a major component of CPI inflation)
    5. tighter money / higher interest rates, to fight inflation
    6. reduction in house value as security in open-access cities that lack rent inflation
    7. reduction in housing starts in open-access cities
    8. more rent inflation, now in both closed- and open-access cities, causing home prices to rise (higher expected future rent payments)
    9. tighter money, causing home prices to fall (higher discount rate)
    10. fears of recession increase demand for safe assets
    11. synthetic CDO's create supply of safe assets with investment bankers taking the short side betting on a house price collapse, basically betting that the Fed will kill the economy. (funds invested in synthetic CDO's actually go into treasuries, not housing)
    12. Fed sticks with its guns, compounding policy errors, until the short bet pays off
    13. and the rest is history

    Did I get that mostly right?

    I remember reading Scott's stuff back in like 2011, about how NGDPLT is better than IT because it reacts better to negative supply shocks, when tightening money in the face of some rising prices is unhelpful --- forcing all prices down to prevent one price from rising is just not good policy given nominal rigidities, especially wage rigidities --- IT dooms you to unemployment when there is a negative supply shock, whereas NGDPLT does not (you get extra inflation instead). At the time, this benefit struck me as kind of theoretical. But now it looks key.

    The standard narrative looks so wrong, and yet so hard to defeat at the same time (check out the wikipedia page for Collateralized_debt_obligation...) The consensus appears so firmly established.


    1. This is a work in progress, so maybe I will eventually tweak it based on new evidence, but I think you've basically summarized it. There are some details to fill in regarding subprime/private pool mortgage lending. As I work through the timeline, I think we can go all the way back to the end of 2003 when Fannie and Freddie had accounting scandals. The sharp rise in private mortgage pools in 2004 is almost entirely a replacement for the sharp drop in GSE activity.

      The trick will be to convince readers to set aside their doubts for what must seem like an unbelievable description of events. But, I think this might lend itself to a better framing of the argument. The discussion can get a little tribal when it is framed in terms of "Did the housing bubble cause the recession or did the Fed cause it?" I think instead of being a supply vs. demand argument, or blame the Fed or not argument, what I can do is present the extensive data about the supply problem leading up to the bust, and the extensive data suggesting that there wasn't a surge of low income borrowing, etc. Then, given that background, just walk through a description of a series of negative credit shocks, beginning with the 2004 drop in Fannie & Freddie activity, moving through negative shocks that culminated in the run on shadow banking in late 2007, and finally ending up at the Fed decisions in late 2008.

      Several of these things that have been assumed to be a part of credit excess are more reasonably described as reactions to monetary deprivation, once we remove the presumption that a housing bust was inevitable. My hope is that the evidence against that inevitability is strong enough to fight the inevitable incredulity that such a conceptual regime shift will trigger.