Friday, July 24, 2015

Housing Tax Policy, A Series: Part 46 - GSE's and the Housing Bubble

One version of the housing bust story says that one source of the problem in the housing boom was the increasing tendency of banks and mortgage brokers to troll for households of lower and lower credit quality and to use the expanding generosity of the GSE's to ultimately stick taxpayers with the credit risk.  Respondents, usually from the political left, argue that the public mortgage agencies didn't play a large role in the boom.  I think the data does bear out the GSE defenders.

I will begin my comments by using this portion of a Wikipedia article as a summary of perspectives on the crisis (see the article for quotation attributions).  First, defenders of the GSE's point out that there was a "bubble of similar magnitude in commercial real estate in America" which would be unrelated to GSE's and public homeownership policies.  Others report, "We find limited evidence that substantial deterioration in CMBS [commercial mortgage-backed securities] loan underwriting occurred prior to the crisis."  And others note, "most of the commercial real estate loans were good loans destroyed by a really bad economy. In other words, the borrowers did not cause the loans to go bad, it was the economy."

The defenders of the GSE's are correct on all of these counts.  The irony is that these are similarities between the commercial mortgage market and the residential market. In fact, shouldn't this defense of the GSE's be a prema facie criticism of the conventional narrative of financial crisis?  I mean, if commercial mortgages had an almost identical rise and fall as residential mortgages, without a subprime component, and with a bad economy triggering the bust, shouldn't we have a strong presumption that both markets shared causal factors?  Logically, the behavior of CMBS tends to argue against both the GSE's as a cause of the bust and subprime loans generally as a cause of the bust.  Isn't it interesting that CMBS are only deployed to defend the GSE's, but not subprimes in general?

Notice that total mortgage growth was not particularly unusual in the 2000s.  The growth rate at the top end of long term ranges reflected the influence of low real long term interest rates.  But, as I have chopped through the weeds of this topic, I have found that the germ of this idea that there was a problem - a "bubble" - came from the rise in individual home prices, which is largely the result of housing constrictions in the large core cities.  We have misinterpreted a supply problem as a demand problem.  As a start, note that even at the end of the boom, there was nothing particularly unusual about the rate of residential or commercial mortgage growth compared to a half century of experience.  And the 1990s had seen the weakest growth for both types of mortgages in the modern era.

Critics of the GSE's, in the Wikipedia article, point out, "[f]rom 2004 to 2006, the two [GSEs] purchased $434 billion in securities backed by subprime loans, creating a market for more such lending."  Further, "In 2003, after the use of subprimes had been greatly expanded, and numerous private lenders had begun issuing subprime loans as a competitive response to Fannie and Freddie, the GSE's still controlled nearly 50% of all subprime lending. From 2003 forward, private lenders increased their share of subprime lending, and later issued many of the riskiest loans. However, attempts to defend Fannie Mae and Freddie Mac for their role in the crisis, by citing their declining market share in subprimes after 2003, ignore the fact that the GSE's had largely created this market, and even worked closely with some of the worst private lending offenders, such as Countrywide. In 2005, one out of every four loans purchased by Fannie Mae came from Countrywide."

It sounds like there was a surge of securitization activity in the 2000s, especially after 2003, as GSE's stretched their standards in order to compete with the recklessly expanding private market.  So, what does the data look like?  I am using the Federal Reserve's Mortgage Debt report for data.

Securitized mortgage financing had increased in importance from the late 1960s until about 1994, after which it leveled off as a proportion of mortgages outstanding.  At about the time that the trend in homeownership rates sharply shifted up, a few years before the price boom began, securitization plateaued as a proportion of mortgages, and only rose slightly as a proportion as a result of the collapsed banking sector after 2006.

In the next graph, we can see the growth rates of mortgages retained by the banks and securitized mortgages.  The growth rate of securitized mortgages outpaced bank mortgages until 1994, after which both types of mortgages grew.  In fact, from 2003 to 2006, mortgages retained by the banks tended to grow faster than securitized mortgages, and grew much faster than mortgages securitized through Federal Agencies and GSE's (the light purple line).  Including Ginnie Mae securities, total mortgages through the public agencies barely grew in 2004 and 2005.  Can you believe that 2004-2005 saw the least expansion in public mortgage securitizations since those programs had begun?  (We've managed to match that record every year since 2010.)

In the next graph, total mortgage growth is shown, with the portion of total growth broken out by bank-retained, securitized, and other.  I have also included the yield curve slope (10 year treasuries minus the Fed Funds Rate) here (the black line).  There is a persistent pattern here that when the yield curve is steep (10yr-FFR > 1%), bank retained mortgage growth increases, and when the yield curve is flat or negative (10yr-FFR < 1%) bank retained mortgage growth declines.  Well, this was the case until 2008.

We can see in the previous graph of growth rates that securitizations and bank-retained mortgages have tended to grow inversely with one another.  This suggests that the securitization agencies have helped to reduce cyclical fluctuations in mortgage credit.  And, they continued to do that in the 2000s.

 Here are graphs that are replications of the graphs above, except that the agency and mortgage pool holdings are broken out between those facilitated by public agencies and private pools (line 69 in this table).  If we take private securitizations out of the picture, there was a very brief and quite normal boost in mortgage growth from 1999 to 2003, followed by a return to the low growth levels of the 1990s.

The next graph is a graph of the size of mortgage pools backed by each agency plus private pools.  (Be careful with interpretation.  I used a log scale so that steady growth is linear.)  Here, we can see again that total mortgage securitizations outstanding were growing at a new, lower trend that remained fairly linear going back to about 1990.  Agency pools (including Ginnie Mae), in total, declined from trend slightly after 2003, as did both Fannie and Freddie, individually.  Ginnie Mae declined sharply.

In October 2005, the GAO issued a report to Congress, titled "HOUSING FINANCE :Ginnie Mae Is Meeting Its Mission but Faces Challenges in a Changing Marketplace".  It includes this statement (page 16):
Since 2000, Ginnie Mae’s volume of MBS outstanding has fallen from $612 billion to $453 billion in 2004, a drop of approximately 26 percent. The primary factor contributing to this decline has been the increase in borrowers who have refinanced out of FHA and VA loan programs into conventional loans. Falling interest rates and rising home prices have led to a boom in refinancing over the last 10 years, particularly from 1997 to 1999 and 2001 to 2004. At the peak of the refinancing boom in 2003, refinancings represented about 65 percent of mortgage originations. As some borrowers with mortgages insured by FHA and guaranteed by VA have built up equity in their homes, they have been able to refinance out of these programs into conventional loans that may offer more favorable and flexible terms and interest rates.
The rise in Freddie and Fannie balances until 2003 were balanced out by declines in Ginnie Mae balances.  Ginnie Mae is the agency that specifically issues FHA loans which are not and never were conventional.  These were loans that were always issued with low down payments.  Some of the long term decline in the importance of Ginnie Mae has come from the growth of private sources of subprime mortgages and from competition with Fannie and Freddie.  But, the primary shift in securitized mortgages during the boom was out of Ginnie Mae mortgages by refinancing into more favorable terms and with higher levels of equity, because of the significant capital gains, which presumably tended to relieve borrowers of the mortgage insurance fees that come with FHA loans.

Now, there certainly was some equity withdrawal going on there, and probably some households that only avoided default because they were tapping marginal new equity - possibly a small portion of those fed into subprime loans instead of conventional loans.  The fact remains - in total, all the federal agencies and GSE's had no unusual growth, and, in fact, declined from trend after 2003, when balances at Freddie and Fannie leveled out and Ginnie Mae continued to decline.

So, we have two distinct phases.  Through 2003, traditional securitizations grew at their regular pace and marginal new mortgage growth came from mortgages retained by the banks.  After 2003, growth of the GSE and agency portfolios collapsed and new growth came from the new private securitization market.

But, as I have with other issues here, I am going to push back against the idea that this was a "subprime bubble".  As with so many issues here, through the haze of chaos, the bubble narrative hits all the right buttons, and honestly seems to make too much sense to deny.  But, when we look more closely at the details and the timing, there are many pieces of evidence that don't fit the narrative well at all, or that present mysteries.

1) The rise in the homeownership rate pre-dates the price boom in general.  Most of the rise in ownership happened before the 2001 recession and it peaked in 2Q 2004, before the subprime phase kicked in.  There is no connection between the growth of the private securitization market and growth in homeownership.  In fact, homeownership rates were declining as private securitizations shot up.

2) In the face of this massive increase in non-conventional mortgage securitization, the evidence points to surprisingly stable buyer characteristics - whether incomes, loan to value, FICO scores and whether the data is for national first-time or repeat borrowers of prime loans, for subprime borrowers, or for national surveys, borrower characteristics are surprisingly stable for all groups given the tremendous changes in originations during the period.

3) There is no apparent relationship between housing starts and subprime mortgage levels.  Housing starts rose steadily through 2005, then collapsed sharply.  Subprime loan balances continued to grow until the middle of 2007 with no interruption at all.

4) Rent inflation has been high for at least 20 years.  As housing starts peaked in 2004 and 2005, rent inflation finally moved down to around the Fed's 2% target level, suggesting that during this period, homebuilding was only just beginning to match demand for housing.  When housing starts collapsed in 2006, rent inflation immediately shot back up to over 4%.  Oddly, the high level of private mortgage lending didn't help to maintain housing starts even in the face of this clear signal of a supply shock.  Note that owner-occupied rent inflation subsided along with subprime mortgage balances and home prices in 2007, as the collapse of credit markets caused demand to drop along with supply.  But rent inflation for renters remained high even into 2009.

5) Home prices peaked at the end of 2005 along with housing starts, also with no correlation to the rise in private securitizations.

6) Among prime mortgages, at least, repeat buyers declined after 2003 and first time buyers increased.  First time buyers increased especially sharply in 2006-2007.  This sits oddly with the fact that ownership peaked in 2004.  There had to be tremendous churn of households exiting homeownership during this period.

7) The private securitization market grew by about $1.1 trillion between the end of 2003 and the end of 2005.  This was the steepest period of home price increases.  During that same period, household real estate market values increased by about $6.7 trillion.  An increasing share of this was due to investors.  That is implied by the declining homeownership rate and strong housing starts, and confirmed by Survey of Consumer Finances survey data.  At the end of the boom, coincidental with the rise of subprime and private securitizations, home values were increasingly equity based and investor-driven.

8) While equity levels had slightly declined until 2003, they slightly recovered during the subprime phase, only falling when home values began to collapse.  This is especially surprising given the large amount of churn that was taking place in home ownership.  First time buyers, which appear to be a large part of the buyer's market in 2004-2007 are typically at their highest leverage level, and home sellers would tend to have more equity, especially in a market where prices had been increasing by double digits for several years.  Yet, despite these pressures, equity levels were healthy until home prices collapsed.

9) The private securitization phase didn't begin until after ARM rates began to rise.  Much of the additional private securitization balances occurred when ARM rates were already at their maximum levels.  Private securitizations and subprime mortgages were still a small portion of the market when ARM rates were low.  Private securitizations continued to grow as a proportion of the mortgage market until the middle of 2007, which was a year after ARM rates peaked.

One facet of this to think about is that this tremendous churn in owners belies the notion that homebuyers max out available credit, and that loose monetary policy or credit markets will lead households to mindlessly bid up assets.  Even though home prices continued to rise until 2005, there had to be extensive downward pressure from the sellers that were exiting home ownership.  And, after 2005, that pressure became so great that housing starts collapsed and home prices began to fall, even though mortgage funding continued to grow until mid 2007.  Obviously, access to credit is an obstacle to demand for home ownership, so that fewer constraints will be associated with some buying pressure.  But, why should we believe that, even given excessive access to credit, households will bid up the prices of houses with no regard for value.  Even if there are some marginal buyers willing to do that, there will be many sellers who will counter that demand.  There appear to have been many of those sellers in this case.  Maybe we can take the healthy level of existing home sales in 2004 and 2005 and the large amount of owner churn as a sign that many owners considered home prices to be above intrinsic value.  Existing home sales began to collapse in late 2005, suggesting that the extreme price declines in 2007 were from the collapse of demand, not a surfeit of sellers.  I am not going to fully work it out in this blog post, but it appears to me that, if the strong appearance of sellers in 2004-2005 was a reaction to excessive market values that this natural market reaction was happening when aggregate intrinsic values and market values were still within a few percentage points of one another.


Considering the relatively broad base of defaults, in terms of credit quality, it seems likely that even the evidence that conventional securitizations experienced lower defaults than private securitizations includes a bit of statistical confusion.  Private pools held an especially large portion of new originations when the bottom fell out of the housing market and pulled so many households underwater.  Young mortgages are especially vulnerable to a nominal shock, even with responsible underwriting.  Defaults in the private pools would have been worse because defaults in 2006 and 2007 vintage mortgages were worse for all types of mortgages.

This same effect would apply to high default rates among ARM loans vs. fixed rate loans.  ARM loans were especially popular at the end of the boom, so much of the excess defaults for ARM loans vs. fixed rate loans could be a product of loan vintage.  The idea that ARM mortgages would have higher defaults because they were taken out by more reckless borrowers seems to make so much sense.  But, remember, the high default vintages were taken out when short term rates were at their peak.  If anything, ARMs in those vintages saw actual rates below what the borrowers had expected to see.
D & VH, Figure 3
D & VH, Figure 4
This is what Demyanyk and Van Hemert found.  Regarding these graphs from their paper on subprime mortgages, they say:

"It is important, though, to realize that this result (KE: the higher defaults of adjustable rate loans after 2005 compared to fixed rate) is driven by an aging effect of the FRM pool, caused by a decrease in the popularity of FRMs from 2001 to 2006. In other words, FRMs originated in 2006 in fact performed unusually poorly (Figure 3, upper-right panel), but if one plots the delinquency rate of outstanding FRMs over time (Figure 4, left panel), the weaker performance of vintage 2006 loans is masked by the aging of the overall FRM pool."

The pundits on the left are correct.  Fannie and Freddie aren't to blame.  They are also correct that commercial real estate went bust as a result of the economic bust, not the other way around.  Pundits on the right are also correct - that predatory banks were not systematically pressing marginal households into oversized homes.  Everyone is correct about the whole mess except for believing that the other side is wrong and that a bust was necessary or inevitable.

Can we all just have a group hug and start letting middle class families buy houses again?


  1. This is your best takedown of the bubble story I have seen. The points about commercial real estate are compelling. (Of course, I already agreed with you, so maybe just more confirmation bias on my part :-)


  2. For a long time I have mentioned the commercial real estate bust---thanks for bringing it up.

    It is remarkable to think that the American right has become anti-housing and regards any stock market increase as a bubble and forever wants ultra-tight money.

    The narrative that desirable neighborhoods do not want condo towers and that is a major reason for housing inflation is not one that will generate any traction on the American right.

    So this is the new standard: the stock market should be depressed, and homeownership rates should fall.

    Not very appealing...

    1. Yes, unfortunately. And, I probably don't give enough attention in my posts to your point that some of this is just as bad in the conservative suburbs as it is in the progressive core cities. I think partly it's just because the urban progressives are so vocal and outrageous about their prejudices whereas the suburban limits happen quietly in small committee meetings. They both, essentially say, "We don't want those people to move into our neighborhoods.", but the progressives are shameless enough to put it on a poster and march around outside city hall pounding their chests about it.

  3. You cannot put up a 60-story condo tower in any coastal city in CA, whether GOP or D controlled.
    As a result, the middle class has been pushed into the Inland Empire, where it is a 110 in summer.
    I keep hoping a pro-business pro-growth political party emergrs...

  4. Very well argued case that the GSEs didn't cause the "bubble".
    I still think we should be winding down Fannie and Freddie. The argument for their existence was always that they increased the amount of credit available for housing and that housing was an important goal for us to subsidize. I didn't believe that home ownership should be subsidized, and now we know that Fannie and Freddie aren't really increasing the amount of credit available, but are displacing private credit. I found the availability of commercial credit, the role of CMBS and the same pricing dynamics in CRE to be telling.

    1. You know, I agree with you about all of that, and I thought we should get rid of them too. I also think we should have some version of free banking and eliminate the government monopoly on currency management. But, I think I've changed my mind about the GSEs. If we are going to have the Fed exercising discretionary power over the currency, then the banks are basically institutions that are naturally short the dollar. Their main vulnerability is to a value that the government has monopoly power over. Given that, then I think there should be a mechanism that allows the banks to put that risk back on the taxpayer. I realize that this would never be an acceptable political position, but morally and practically, it seems like the only reasonable framework to me. Possibly this would be unnecessary under a NGDPLT policy.

    2. Goal #1: NGDPLT. Head and shoulders importance over winding down the GSEs.
      Goal #2: Much higher equity requirements for the banks.
      #3, maybe: FDIC coverage of 98% instead of 100% for deposits over $10,000.
      And I'd still like to wind down the GSEs. Especially the whole semi-private mishagas we had prior to Sept 2008.