by Adam Tooze
During Greenspan’s refinancing boom of 2000–2003, it wasn’t just the GSEs that were busy. The huge surge in issuance meant that there was plenty of unconventional, “nonconforming” business to go around too. But the decisive thing was what happened in early 2004 when interest rates had reached rock bottom, the refinancing boom had run its course and the GSEs were stopped in their tracks. With the pipeline ready and waiting, it was at this point that the private mortgage industry took over. Leaving behind the GSE-centered model of the 1990s, they deprioritized conforming mortgages in favor of private label “unconventional” lending—subprime, slightly better Alt-A and oversized jumbo loans.
What the private issuers discovered was that if scrutinizing conventional mortgages was profitable, subprime was even more so.40 The financial engineering was more elaborate and one could charge more money for the services. The techniques of the fixed-income investment bankers were now brought fully into play. A surprisingly large share even of nonconforming private label MBS could still attract an AAA rating once combined in structured products. To manage the risks, the production of credit default swaps (CDS), once the preserve of bespoke investment banks, was industrialized. Mainline insurers like AIG offered CDS insurance on exotic securitized products. Given the quality of the underlying mortgages, not all the tranches were good. But that stimulated the investment banks to expand the collateralized debt obligation (CDO) business. CDOs were derivatives based on repackaged middle-ranking “mezzanine” tiers of other securitized mortgage deals. By combining them together and tranching, you could make a large pool of BBB assets yield further tranches of AAA securities. Once you had done that you could then go one step further. You could take the low-rated mezzanine slices of the CDO and pool and tranche them once more to create CDO-squared. And once again by the logic of independent risks and the good graces of the ratings agencies, a portion even of those securities would warrant an AAA rating.
III
By the early 2000s, the private mortgage industry was waiting for the starter’s gun. It had its new raw material—securitized mortgages. It had its mechanics and its engineers at the ready. The end of the refinancing boom of 2003 in conventional mortgages triggered the push into unconventional lending. To unleash the final phase of the boom, it needed one last ingredient. Someone had to be interested in buying the hundreds of billions of dollars in securities that were being produced. If there had been no demand to meet the supply, the price of MBS would have fallen, yields would have surged leading borrowing rates to rise, choking off the mortgage boom. Not only did this not happen, but long-term interest rates remained flat and the spread—the premium that nonconforming borrowers had to pay—declined. This points to the third historic transformation that made possible the 2000s boom, a change not on the supply but on the demand side: the surging demand for safe assets and the mobilization of institutional cash pools for mortgage finance.41 It is at this point that the technical mechanics of mortgage banking reconnects with the grand theme of the rise of China, the emerging markets and the mounting inequality and wealth polarization in the Western world.
To understand this connection, the best place to start is to go back to the most scandalous thing about MBS, their credit rating. The AAA label was important because it placed them in a class of assets like Treasurys that attracted investors looking for safe assets.42 AAA was a badge of quality, and, like any certificate of this type, it signaled that if what you were looking for was safety, you had to look no further. Such assets constitute as close as the unstable capitalist economy can offer to a neutral safe position. They are desired and in some cases legally required by all investors with a particular aversion to risk and little capacity for independent research or evaluation—pension funds, cash funds, insurance funds and so on. As one of the key economists in the field has remarked: “[A]lmost all human history can be written as the search for and the production of different forms of safe assets.”43 This may be true, but it begs the question of what was happening in the late 1990s and early 2000s to drive a huge surge in the demand for safe assets.
The first part of the answer is the development of the emerging market economies from the 1990s. As a result of their trade surpluses and their desire to self-insure against the risk of a repeat of the 1994–1998 crises, they wanted reserve assets that they could liquidate in an emergency. And the assets that best fit that description were US Treasurys, long- and short-dated. In the early 2000s, China and other emerging market sovereigns bought all the Treasurys that even the gaping budget deficits of the first Bush administration could provide. Macroeconomists worried about the current account imbalance that resulted and the possibility of a catastrophic sudden stop unwinding. What they did not pay attention to, because they did not dirty their hands with technicalities like MBS, was the effect the influx of emerging market funds might have in financial markets. Emerging market investors bought first Treasurys and then GSE-issued agency debt. This left other institutional investors looking for alternatives. What filled the gap was financial engineering. If pension funds, life insurers and the managers of the gigantic cash pools accumulated by profitable corporations and the ultrawealthy needed safe assets, AAA-rated securities were a product America’s mortgage machine knew how to synthesize.
Foreign Reserves and Institutional Investors Compete for US Short-Term Safe Assets (in $ billions)
Outstanding Amounts:
2005
2006
2007
2008
2009
2010
Short-term Treasury securities*
1,146
1,173
1,192
1,909
2,558
2,487
Short-term agency securities**
568
489
560
903
844
618
Total
1,714
1,662
1,752
2,812
3,402
3,105
(−) Foreign Official Holdings:
Short-term Treasury securities
216
193
181
273
562
na
Short-term agency securities
112
110
80
130
34
na
Total
328
303
261
403
596
na
(−) Demand from Institutional Cash Pools:
Institutional cash pools (based on available data)
1,771
2,120
2,216
/> 1,834
2,041
1,911
Institutional cash pools (estimate of total volume)
3,120
3,735
3,852
3,467
3,596
3,432
Average
2,445
2,927
3,034
2,650
2,818
2,672
= Deficit of safe, liquid, short-term products
(1,059)
(1,568)
(1,543)
(241)
(12)
na
*Includes Treasury bills and Treasury securities with a remaining maturity of one year or less.
**Includes agency discount notes.
Source: Zoltan Pozsar, “Institutional Cash Pools and the Triffin Dilemma of the US Banking System,” Financial Markets, Institutions & Instruments 22, no. 5 (2013): 283–318, figure 5.
But once again we have to be careful. The shuffling of global demand for dollar-denominated safe assets helps to explain why the mortgage pipeline did not result in an oversupply of AAA securities. But to the extent that private label asset-backed securities were actually sold off to investors, little more was heard of them. When the market turned bad, they would sit on balance sheets as an illiquid entry. They were no longer counted as safe assets. There would be lawsuits against investment banks that had knowingly repackaged unsafe mortgages. Certainly the losses would have an impact on investment allocation and on the spending decisions of millions of pensioners. But this would not by itself create a financial crisis, with bank failures rippling out across the world. The comparison with the dot-com bubble is instructive. It created a huge surge in wealth followed by a collapse. It triggered a severe recession. But it did not lead to banking crises. The subprime mortgage boom of the early 2000s led to a financial crisis because, contrary to the professed logic of securitization, hundreds of billions of private label MBS were not spread outside the banking system, but were stockpiled on the balance sheets of the mortgage originators and securitizers themselves.44
Why did the securitizers end up holding their own product? In part it was a matter of the production system itself. Securitization produced some attractive tranches and some less so. The less attractive tranches needed to be held off the market. Furthermore, the banks operating the pipeline believed their own business proposition. Holding MBS was very profitable at prevailing funding costs. The banks in the mortgage supply chain were at the source of the profit. So why not get rich too? It was a choice. Not every bank did it. Those that took the biggest risks were large mortgage originators and the most aggressively expansive commercial banks—Citigroup, Bank of America and Washington Mutual—and the two smallest and scrappiest investment banks—Lehman and Bear Stearns. By contrast, J.P. Morgan began throttling back its mortgage pipeline already in 2006 and bought as much protection as it could in the CDS market. Goldman Sachs went beyond hedging to place a large bet on an imminent housing market collapse.45
Their caution was easy to justify given the kind of business that subprime lending involved. But it also reflected a more basic banking consideration. Building a big balance sheet of MBS didn’t just involve risk on the asset side. It also involved expanding the liabilities of the bank on the funding side. And this brings us to the true heart of the 2007–2008 crisis. If the mortgage production line was holding hundreds of billions of private label MBS and ABS on its own balance sheet, how were those holdings funded? Here too it was the new model of investment banking that provided the answer. If an upstart mortgage lender like Countrywide didn’t have depositors, neither did Lehman. Lehman got its funding wholesale by tapping the cash pools and so too would the new mortgage lenders, including Lehman. This was the truly lethal mechanism at the heart of the crisis. Funds from money market cash pools were channeled into financing the holding of large balance sheets of MBS.
The largest mechanism for funding mortgage holdings was asset-backed commercial paper (ABCP).46 The three biggest American issuers of ABCP were Bank of America, Citigroup and J.P. Morgan. The vehicles for managing this operation were so-called structured investment vehicles (SIV), legal entities provided with a minimum layer of capital by their “sponsors,” but otherwise separate from the balance sheets of their parent banks. Onto these SIVs the parent bank would offload a large portfolio of mortgage bonds, securitized car loans, credit card debt or student debt. The SIV would pay the parent bank for the securities with funds raised by issuing ABCP. These were three-month notes backed by the assets in the SIV and the good name of the parent bank. Though the SIV had no track record, it could issue the commercial paper at competitive rates because of the value of the securities it held and because it was assumed that it enjoyed the backing of the sponsoring bank. Remarkably, under the bank regulations prevailing until the early 2000s, assets parked off balance sheet in the SIV could be backed by a fraction of the capital that would be required if they were on balance sheet. Inflating the balance sheet was risky but it raised rates of return on capital. Further profits were to be made by trading on the spread between long-term returns and short-term funding costs. Typically, an ABCP vehicle would hold a portfolio of securities with maturities of three to five years and would fund those securities by selling commercial paper repayable between three months and as little as a few days. For the managers of cash pools, the commercial paper was more attractive than the underlying securities, because it was very short term and backed by a top-rated commercial bank. For the parent banks, the spread between the return from the high-risk cocktail of assets held in the SIV and the low rate paid on the highly rated ABCP was handsome.
If the SIV-ABCP model involved a degree of maturity mismatch, the investment banks pushed this to extremes. The entire business model of investment banks was based on wholesale funding. The most elastic vehicles for this were so-called repurchase agreements, or repo. In a repo transaction a bank would buy a security and pay for the purchase by immediately reselling it for a period of as little as one night or as long as three months, with a promise to repurchase at a certain price. It was in effect a collateralized short-term funding agreement. The investment bank would buy $100 million in securities and repo them with a mutual fund or another investment bank, with the party repoing the paper paying a small interest charge to the investor it was repoing with. It also accepted a haircut. In exchange for $100 million in Treasurys, it did not receive full value, but only $98 million in cash. It would also repurchase them for $98 million. In the meantime, the haircut determined how much of its own money the investment bank would have to put into holding the securities, and thus the leverage in the deal.47 A 2 percent haircut meant that to fund the purchase of $100 million of securities and to receive the interest paid on those bonds, a bank would need $2 million of its own money. The rest it could get out of the repo transaction. Using this mechanism, a small amount of capital could support a far bigger balance sheet, provided, of course, that the repo could be repeatedly “rolled” and that the haircut did not suddenly increase.
By the 2000s the collateral posted in repo markets in New York ran to several trillions of dollars a day. It was split into two markets—bilateral and triparty repo. Both were over-the-counter professional markets, which were only loosely monitored by the central banks or regulators. The best data we have is for trilateral repo where the trade was managed by a third party—either JPMorgan Chase or
Bank of New York Mellon—which held the collateral for the duration of the repo.48 In triparty repo the collateral used was of top quality—almost exclusively Treasurys or agency MBS. Given the additional layer of protection, triparty repo was where institutional cash pools like MMFs did their repo. Triparty was not used to fund private label MBS. It was in the bilateral repo market that they could be funded. The best available data suggest that the bilateral repo market was three times larger than the triparty segment.49 Because the players in the bilateral market tended to be investment banks and hedge funds, the types of assets acceptable as collateral were more wide ranging. It is here, along with ABCP and various types of interbank and unsecured borrowing, that the investment banks financed their holding of private label MBS and CDO portfolios. Given the wide range of collateral, haircuts in the bilateral repo market ranged in the spring of 2007 from 0.25 percent on US Treasurys to 10 percent or more for asset-backed loans of inferior quality.
As in commercial paper, repo was exposed to serious funding risk. You might not be rolled over. Specifically, the risk was that if an investment bank like Lehman or Bear was thought to have suffered major losses on some big part of its portfolio—whether that was funded by commercial paper, bilateral repo or other types of interbank borrowing—it would suffer a general loss of confidence. It would then be considered ineligible as a counterparty in the triparty market and would find itself shut out from critical funding. The scale of the potential risk was huge. At Lehman at the end of fiscal year 2007, of its balance sheet of $691 billion, 50 percent was funded through repo. At Goldman Sachs, Merrill Lynch and Morgan Stanley, the share was 40 percent.50 If any one of these investment banks was to lose access to the repo markets, at a stroke its business model would collapse, taking its entire balance sheet—not just its MBS business, but its derivatives book, currency and interest swaps—down with it.