New Data on Bankers' Risk Aversion

-->

Probably the most controversial claim of the book will prove to be our suggestion that the behavior of bankers before the crisis was actually risk averse, at least in the aggregate. (A key point of Chapter 4, however, is that there was considerable heterogeneity underneath these aggregates.)

The risk-aversion claim rests on two facts: the much-higher-than-legally required capital ratios of commercial banks and savings and loans before the crisis; and the fact that commercial banks and S&Ls overwhelmingly bought the least lucrative and supposedly “safest” mortgage-backed securities: those implicitly guaranteed by the federal government through Fannie and Freddie’s congressional charters; and those rated AAA vs. those rated AA or lower.

Lower-paying bonds pay higher yields than higher-rated bonds, which in turn pay higher yields than “agency” (Fannie/Freddie) bonds. Reckless, greedy bankers should not have bought agencies or AAA bonds; they should have bought BB or BBB bonds.

In Table 2.2 of the book, we claim that banks bought more than twice the quantity of agency bonds as AAA mortgage bonds, and four times the quantity of AAA mortgage bonds as AAA CDOs. CDOs usually tranched mezzanine (sub-AAA) tranches of “regular” mortgage bonds, i.e., private-label mortgage-backed securities (PLMBS), so AAA CDOs were objectively riskier than AAA PLMBS. Thus, one way to view the collective risk profile of U.S. banks and savings and loans is to recognize that the following list shows the least risky and least lucrative of their mortgage bond holdings first, with the riskiest and most lucrative last:

$852 billion in agency bonds

$383 billion in AAA private-label mortgage-backed securities

$90 billion in AAA CDOs

Those are the figures on which row 1 of our Table 2.2 are based. However, none of our tables show bank holdings of actual AA, A, BBB, or BB tranches of PLMBS or CDOs. This is because our data on the distribution of the various types of mortgage bond in banks’ portfolios came from a famous Lehman Brothers study dated April 11, 2008: “Residential Credit Losses—Going into Extra Innings?” (We discuss this study in end note 1 to the conclusion of the book.) The Lehman figures assign PLMBS and CDO holdings by various types of investor (commercial banks and S&Ls, investment banks, hedge funds, etc.) by billions of dollars. But the 2008 Lehman table says that commercial banks held no PLMBS or CDOs rated lower than AAA. That seems implausible, unless the authors of the report could not find holdings totaling more than $500 million, so the figure got rounded down to zero billions of dollars. (We attempted to contact the authors of the report, but they all now work for hedge funds and are barred from communicating with members of the public.)

The Lehman figures have been widely cited by others (e.g., Arvind Krishnamurthy, “The Financial Meltdown: Data and Diagnoses”--download PDF--Table 1 [November 2008]; Viral Acharya and Matthew Richardson, Restoring Financial Stability: How to Repair a Failed System [2008], Table 5). They are more comprehensive and newer than the only comparable data source, a 2007 Lehman Brothers report entitled “Who Owns Residential Credit Risk” (cited by the Financial Crisis Inquiry Commission, “Preliminary Staff Report: Securitization and the Mortgage Crisis” [April 2010], Table 1). However, we wish there were a better source. Unfortunately, as of November 2010, when our book was completed, we knew of no other data. Now (ht Viral Acharya) we do—although this source, too, has limitations.

In August 2011, Isil Erel, Taylor D. Nadauld, and René M. Stulz released an NBER working paper, “Why Did U.S. Banks Invest in Highly Rated Securitization Tranches?” Erel et. al found a way to use regulatory filings from bank holding companies to “back out” their holdings of asset-backed securities, including mortgage bonds.

The key data source, Schedule HC-R of form FR Y-9C, groups assets by risk bucket: 0%, 20%, 50%, 100%. This means that AAA and AA PLMBS, which were both risk weighted at 20%, are grouped together. And the 100% category contains so many different types of securities, such as corporate bonds and equities, that we cannot back out mortgage bonds in that category, namely those rated BBB or BB. Using Erel et al.’s method, however, we are able to provide the following results for the four biggest bank holding companies as of the end of 2006:

Citigroup:

$70 billion in agency bonds

$28 billion in AAA/AA ABS

$10 billion in A-rated ABS

Bank of America:

$157 billion in agency bonds

$4 billion in AAA/AA ABS

$2 billion in A-rated ABS

JPMorganChase:

$75 billion in agency bonds

$1 billion in AAA/AA ABS

$0 billion ($351 million) in A-rated ABS

Wells Fargo:

$27 billion in agency bonds

$27 billion in AAA/AA ABS

$1 billion in A-rated ABS

Aggregate:

$329 billion in agency bonds

$60 billion in AAA/AA ABS

$13 billion in A-rated ABS

In short, the Big Four invested in the safest two classes of asset-backed securities securities, agency mortgage bonds and AAA/AA-rated ABS, at a ratio of 28:1 compared to the riskiest, most lucrative category for which we have figures: A-rated ABS. One should keep in mind that, for example, in mid-2006, agencies paid a 9-bps spread over Treasuries, AAA mortgage bonds paid 18 bps, AA mortgage bonds paid 32 bps, and A mortgage bonds paid 54 bps (Ashcraft and Schuermann, "Understanding the Securitization of Subprime Mortgage Credit," p. 28, Table 16). Reckless, greedy bankers should have bought As--or lower-rated mortgage bonds (BBBs paid 154 bps, and BBB- paid 267 bps) every time--never agencies, AAAs, or AAs.

However, do the new data mean that the Lehman figures on which we relied are wrong, for showing zero (or less than $500 million) in sub-AAA mortgage-bond holdings among commercial banks and savings and loans? Not necessarily. The figures above are from the consolidated regulatory filings of bank holding companies, which include investment-banking arms. The Lehman figures do show $24 billion in sub-AAA bonds among investment banks. Moreover, the data above cover all ABS, including credit-card ABS, auto-loan ABS, student-loan ABS, and more exotic categories, not just mortgage ABS.

As for the data, then, we conclude that the full story remains to be told, but that our basic point stands: the bankers—even at Citigroup, which had the riskiest portfolio—were not behaving in patterns we would expect from reckless greedheads.